{"method":"Drafted by web-searching research agents (Claude), one per drug x condition pair, three at a time; every claim is backed by the listed sources, which the agent visited while drafting. candidability is the agent's own 0-10 judgement of how strong an adverse-drug-reaction candidate the association is. Machine-drafted and pending expert review -- not a curated clinical statement.","notes":{"amitriptyline|G470":{"verdict":"co-prescription population","headline":"Insomnia is mostly why amitriptyline is given; the labelled ADR is rare and the X-linked hit is intergenic","assessment":"Insomnia is not a licensed indication for amitriptyline (the SmPC lists major depression, neuropathic pain, chronic tension-type headache and migraine prophylaxis, plus paediatric enuresis), but low-dose amitriptyline 10-25 mg at night has been prescribed off-label for insomnia disorder for decades, so coded G47.0 in this drug's users is largely a prescribing reason rather than an outcome. The direction of drug effect is the decisive point: in the placebo-controlled DREAMING trial amitriptyline reduced Insomnia Severity Index scores at 6 weeks (mean difference -3.4, 95% CI -6.3 to -0.4) and in a head-to-head trial it was non-inferior to CBT-I at 12 weeks, i.e. the drug moves this condition toward improvement on average, and its very common labelled effect is somnolence, not insomnia. A drug-caused mechanism is nonetheless describable and label-supported at low frequency: insomnia and nightmare are listed as 'uncommon' undesirable effects, amitriptyline suppresses REM sleep and can produce vivid dreaming and disturbed sleep, occasional paradoxical insomnia is reported (a 25 mg case with onset within 3 weeks and resolution one week after withdrawal), and rebound insomnia on tapering is common (68% reported worse sleep after taper), which in an EHR would generate a post-exposure insomnia code without the drug 'causing' chronic insomnia. The atlas numbers do not support an ADR reading either: enrichment is 0.33, i.e. coded insomnia is three-fold depleted in amitriptyline users versus users of other drugs (the comparator presumably includes hypnotic and psychotropic users), and temporality is only 49.2% over 120 dated participants in a cohort that inflates this quantity, so roughly half of these insomnia diagnoses pre-date first exposure. The FAERS 'STRONG' flag with PRR 2.25 is the weakest link in the chain, since insomnia is a near-universal co-reported term for psychotropics and is confounded by indication, withdrawal reporting and the underlying depression. Overall this reads as a sedative given to poor sleepers, with a genuine but uncommon paradoxical/withdrawal insomnia sitting far behind it.","gene_comment":"rs186140708 maps to chrX:4,781,444 (GRCh38) and is annotated intergenic; the nearest features within 500 kb either side are unnamed lncRNAs and mitochondrial processed pseudogenes (MTND6P12, MTCYBP12), so this is an Xp22.2 gene desert with no credible gene assignment and no sleep or TCA-pharmacology candidate. Being a rare X-chromosome variant, it is also exposed to hemizygous-dosage coding artifacts, and nothing here should be presented as mechanism.","score_check":"beta_adr 4.155 (OR ~64) for a 0.84% outcome at only log10p 6.16 - short of genome-wide significance - is the classic sparse-data signature of a rare variant with a handful of carrier cases, and the disease arm is flat noise (beta 0.344 +/- 0.315, |beta| < 1.96*se), so z_diff 4.26 is driven entirely by the unstable treated-arm estimate. The epidemiological fields are internally coherent but point away from an ADR: depleted enrichment and ~50% temporality both argue the condition largely precedes the drug.","candidability":2,"candidability_reason":"Condition is chiefly an off-label prescribing reason and the drug improves it on average; the one variant is a rare intergenic X-desert SNV with an implausible effect size below genome-wide significance.","sources":[{"title":"Amitriptyline 10 mg Film-Coated Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/10849/smpc"},{"title":"DREAMING: low-dose amitriptyline and mirtazapine for insomnia disorder in general practice, randomised placebo-controlled trial (BJGP/PMC)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12199994/"},{"title":"Low-dose amitriptyline versus CBT-I for insomnia with medical comorbidity: randomised non-inferiority trial (PMC)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12696364/"},{"title":"Amitriptyline-Induced Insomnia in a Young Lady Diagnosed With Cyclic Vomiting Syndrome: A Case Report (PMC)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10491943/"},{"title":"Ensembl REST variation record for rs186140708 (chrX:4,781,444, intergenic)","url":"https://rest.ensembl.org/variation/human/rs186140708?content-type=application/json"}]},"amitriptyline|G500":{"verdict":"licensed indication","headline":"Amitriptyline treats trigeminal/facial neuropathic pain; the NPAS3 hit is a rare-variant artifact","assessment":"Trigeminal neuralgia (G50.0) is not an adverse effect of amitriptyline but a reason to prescribe it: the harmonised EU/UK SmPC lists 'treatment of neuropathic pain in adults' as a licensed indication, and amitriptyline is a standard, in places first-line, drug for atypical trigeminal neuralgia and persistent idiopathic facial pain, with a reported NNT around 2.8 in chronic facial pain. It is second-line/adjunctive rather than first-line for classical TN, where carbamazepine and oxcarbazepine dominate, so amitriptyline users coded G50.0 are typically the harder or atypical facial-pain cases rather than newly injured trigeminal nerves. The drug therefore moves this condition in the opposite direction to an ADR reading, which is decisive: no neuralgia or facial pain appears in section 4.8 of the SmPC (paraesthesia is common and polyneuropathy very rare, but neither is trigeminal neuralgia), and on_bnf=0 confirms no label-level ADR listing. The FAERS 'STRONG' signal with PRR 6.2 is exactly the pattern disproportionality analyses generate through confounding by indication and protopathic bias, where the treated disease is reported as the event; it adds no causal weight here. Temporality of 65.5% over 84 dated participants with a 3.46-year median gap is unremarkable given that amitriptyline is often started for depression, migraine or undifferentiated pain years before a facial-pain diagnosis is formally coded, and the atlas itself warns that a high value is weak evidence. Enrichment is null, so the one honest test of whether this drug's users are unusually likely to carry the condition could not be run, leaving nothing to offset the indication reading.","gene_comment":"rs117645447 is an intronic variant in NPAS3 with a global minor allele frequency near 0.26%; NPAS3 is a neurodevelopmental bHLH-PAS transcription factor tied to schizophrenia and related psychiatric risk, with no described role in trigeminal nociception, and the established amitriptyline pharmacogenes are CYP2D6 and CYP2C19, not NPAS3. An intronic rare variant in a neurodevelopmental transcription factor is a locus label, not a mechanism.","score_check":"beta_adr of 6.844 (odds ratio in the hundreds) for a 0.26%-frequency variant in a within-users analysis of at most a few hundred trigeminal-neuralgia cases is the classic sparse-data separation artifact, and the disease arm is flatly null (beta 0.576, se 0.404, so |beta| < 1.96*se; log10p 0.81). The z_diff of 5.17 is therefore driven entirely by the inflated treated-arm estimate rather than by a real difference in effect under drug.","candidability":1,"candidability_reason":"Confounding by indication on a licensed neuropathic-pain use, with an implausibly large effect at a rare intronic variant in a gene unrelated to pain or amitriptyline disposition.","sources":[{"title":"Amitriptyline 10 mg Film-Coated Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/10849/smpc"},{"title":"Trigeminal neuralgia: a practical guide (Pract Neurol, PMC8461413)","url":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8461413/"},{"title":"Management of trigeminal neuralgia and its atypical variant (Medicine Today)","url":"https://medicinetoday.com.au/mt/2009/november/feature-article/management-trigeminal-neuralgia-and-its-atypical-variant"},{"title":"Amitriptyline-perphenazine therapy for persistent idiopathic facial pain (J Anesth Analg Crit Care, PMC11656548)","url":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11656548/"},{"title":"CPIC Guideline for CYP2D6 and CYP2C19 Genotypes and Dosing of Tricyclic Antidepressants: 2016 Update","url":"https://files.cpicpgx.org/data/guideline/publication/TCA/2016/TCA_2016.pdf"},{"title":"dbSNP entry for rs117645447 (NPAS3, intron variant)","url":"https://www.ncbi.nlm.nih.gov/snp/rs117645447"},{"title":"Facing Our FAERS: Disproportionality Analyses, Signal Detection, and Interpretation (J Invest Dermatol)","url":"https://www.jidonline.org/article/S0022-202X(25)02237-7/fulltext"}]},"amitriptyline|H931":{"verdict":"contested","headline":"Tinnitus is a labelled amitriptyline ADR and an off-label target for it; the atlas genetics are not credible","assessment":"Tinnitus is not a licensed indication for amitriptyline but it is an explicitly labelled adverse reaction: the US prescribing information lists tinnitus among CNS/neuromuscular adverse reactions, and case reports describe prolonged tinnitus after even low-dose amitriptyline. The direction is genuinely contested, however, because tricyclics are also used off-label to treat tinnitus distress: the Cochrane review of antidepressants for tinnitus covered amitriptyline, nortriptyline and trimipramine and found slight apparent improvement that was probably methodological bias, i.e. insufficient evidence either way, and Tinnitus UK lists amitriptyline as a tinnitus treatment with a possible risk of significant harm. A recent FAERS disproportionality analysis of antidepressants and inner-ear events does find tinnitus reported above background for the class, which is consistent with this atlas's FAERS PRR of 3.39, although the drugs driving that analysis were mostly SSRIs and SNRIs rather than tricyclics. Within the cohort, tinnitus is not enriched in amitriptyline users (enrichment 0.79, cotx 0.62%), so this is not simple confounding by indication, and the variants show no effect on tinnitus in the drug-free population (beta_disease 0.016 +/- 0.122, far inside noise; log10p_disease 0.05), which is the shape a true drug-conditional effect would take. Against that, the within-user effect size is not believable: beta_adr 4.397 is an odds ratio near 80, reported on four SNVs, two of them with rare-variant rsIDs, in a case set whose datable denominator is only 88 - the classic signature of quasi-separation in sparse data rather than a real large effect. Temporality of 80.7% with a median 3.0 years to onset is the uninformative direction given that prescription records reach further back than diagnosis records, so it adds nothing. The honest reading is a real class-level ototoxicity concern that this particular locus does not support.","gene_comment":"None of the four assignments is credible as mechanism: ENSG00000304974 is an unannotated novel model, F11-AS1 and LINC01449 are lncRNAs, and OGT is an X-linked housekeeping glycosyltransferase whose only auditory link is hearing impairment within the severe OGT-CDG intellectual-disability syndrome. No locus sits near RCOR1 or any of the eleven suggestive loci from the UK Biobank tinnitus GWAS.","score_check":"The disease-arm and prescription-arm numbers are clean and internally consistent (both null, z_diff 5.7 driven entirely by the ADR arm), but beta_adr 4.397 on 88 dated cases with rare alleles is far more likely sparse-data separation than a real OR of ~80. Treat log10p_adr 8.16 as unreliable until the effect is re-estimated with Firth or exact methods.","candidability":5,"candidability_reason":"Tinnitus is a genuine labelled amitriptyline adverse reaction with a class-level FAERS signal, but the genetics offered here - implausibly large effect, rare variants, lncRNA/housekeeping gene assignments - are unconvincing.","sources":[{"title":"DailyMed - AMITRIPTYLINE HYDROCHLORIDE tablet (adverse reactions include tinnitus)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=29ec1221-04d8-4b21-822e-2779fcaae62c"},{"title":"An unusual case of prolonged tinnitus following low-dose amitriptyline (PubMed 18308819)","url":"https://pubmed.ncbi.nlm.nih.gov/18308819/"},{"title":"Antidepressants for patients with tinnitus - Cochrane review (Baldo et al., CD003853)","url":"https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD003853.pub3/abstract"},{"title":"Amitriptyline - Tinnitus UK","url":"https://tinnitus.org.uk/tinnitus-treatments/amitriptyline/"},{"title":"Inner ear signs and symptoms induced by antidepressants: a disproportionality analysis based on FAERS","url":"https://link.springer.com/article/10.1007/s00210-025-04677-9"},{"title":"GWAS suggests variation at the RCOR1 locus is associated with tinnitus in UK Biobank","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7979698/"},{"title":"An intellectual disability syndrome with single-nucleotide variants in O-GlcNAc transferase (OGT)","url":"https://www.nature.com/articles/s41431-020-0589-9"}]},"amitriptyline|J189":{"verdict":"statistical artifact","headline":"Depleted in users and driven by an OR~115 rare intronic EYS variant: artifact, not an amitriptyline ADR","assessment":"Pneumonia is in no sense a licensed indication for amitriptyline, whose SmPC indications are major depression, neuropathic pain, chronic tension-type headache and migraine prophylaxis, and nocturnal enuresis in children; section 4.8 lists no pneumonia, aspiration or dysphagia, only 'congested nose' as very common and alveolitis/Loeffler's syndrome as very rare. A drug-caused mechanism is at least conceivable at class level, because amitriptyline is the prototypical strongly anticholinergic and sedating TCA, and anticholinergic burden in the elderly is associated with pneumonia (aOR 1.33, 95% CI 1.18-1.49 in a 64,430-patient Taiwanese case-control study) through xerostomia, impaired oropharyngeal swallowing and reduced salivary antimicrobial defence. The best drug-specific evidence, however, runs in the opposite direction: in a population-based case-control study of elderly Parkinson's patients, amitriptyline users had a lower risk of incident pneumonia (current aOR 0.87, recent 0.86, past 0.90), with the whole TCA class at aOR 0.86/0.83, attributed by the authors to suppression of Th1 cytokines. The atlas agrees with that direction rather than with an ADR reading: enrichment is 0.68, i.e. pneumonia is depleted among amitriptyline users relative to users of other drugs, and cotx_pct is only 0.83%. The 98.9% temporality over 177 dated participants and median 4.1 years to onset carry essentially no information here, since pneumonia is an acute incident event that will nearly always postdate a chronic prescription in a cohort whose prescription records reach further back than its diagnoses; the FAERS signal is weak (1/3 terms, PRR 1.92), the size expected from a frail, elderly, polypharmacy user base. What remains is a single rare intronic variant with an impossible effect size, so this should be read as a sparse-data artifact sitting on top of a drug-condition pair whose real-world direction is, if anything, protective.","gene_comment":"rs186348321 is an intronic variant (GRCh38 chr6:64,665,130) at 0.5% global and 0.7% non-Finnish-European frequency inside EYS, a ~2 Mb, 44-exon, retina-specific photoreceptor gene whose only established phenotype is autosomal recessive retinitis pigmentosa. There is no pulmonary, immune, aspiration or amitriptyline-pharmacokinetic route from EYS to pneumonia, and a gene this large collects intronic hits by chance; the assignment is positional, not mechanistic.","score_check":"beta_adr 4.748 implies an odds ratio near 115 for a 0.5%-frequency variant in a subgroup with only ~177 dated cases, which is the classic signature of quasi-complete separation rather than a real effect, and the disease arm is flatly null (beta 0.033 +/- 0.147, |beta| well under 1.96*se; log10p 0.08) as is the prescription arm (0.51). The z_diff of 5.74 is therefore inherited entirely from the unstable within-users beta and does not constitute independent evidence of a drug-specific genetic effect.","candidability":1,"candidability_reason":"Backwards at the population level (enrichment 0.68, literature aOR ~0.86 for amitriptyline) and genetically an implausible sparse-data hit in a retina-specific gene.","sources":[{"title":"Amitriptyline 25mg Film-Coated Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/14335/smpc"},{"title":"Antidepressants Usage and Risk of Pneumonia Among Elderly Patients With Parkinson's Disease: A Population-Based Case-Control Study (Front Med 2022, PMC8896435)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8896435/"},{"title":"The Prevalence of Anticholinergic Drugs and Correlation with Pneumonia in Elderly Patients: A Population-Based Study in Taiwan (PMC7503732)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7503732/"},{"title":"Ensembl variation record for rs186348321 (intronic, EYS, gnomAD frequencies)","url":"https://rest.ensembl.org/variation/human/rs186348321?content-type=application/json;pops=1"},{"title":"A hypomorphic variant in EYS detected by genome-wide association study contributes toward retinitis pigmentosa (Commun Biol 2021, PMC7846782)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7846782/"},{"title":"Association between anticholinergic medication uses and the risk of pneumonia in elderly adults: a meta-analysis and systematic review (Ann Med 2023)","url":"https://doi.org/10.1080/07853890.2023.2209736"}]},"amitriptyline|M069":{"verdict":"co-prescription population","headline":"Amitriptyline is an adjunct analgesic in RA, not a cause of it; rare intergenic 7q21 hit is unsupported","assessment":"Rheumatoid arthritis is not a licensed indication for amitriptyline (the EU SmPC lists major depressive disorder, neuropathic pain, chronic tension-type headache and migraine prophylaxis, plus paediatric nocturnal enuresis), but low-dose amitriptyline is a long-standing off-label adjunct for RA-related chronic pain and for the depression that accompanies it, so RA patients are a natural amitriptyline-prescribing population. The SmPC section 4.8 lists no musculoskeletal or rheumatological adverse reactions at all - no arthralgia, no arthritis, no lupus-like syndrome - and the Cochrane review of antidepressants in inflammatory arthritis found only tolerability issues (somnolence, dry mouth, dizziness), never new joint disease. If anything the drug moves the endpoint the other way: the one trial that beat placebo did so by reducing painful/tender joint counts, and amitriptyline is repeatedly shown in vitro and in vivo to lower TNF-alpha, IL-1beta and IL-6 and to inhibit NF-kB, which is the opposite direction from an inflammatory arthritis. The atlas numbers fit the co-prescription reading rather than an ADR: enrichment is only 1.12, the indication score is 8.5 with predisposition 0, and the temporality of 91.5% (median 5.39 years) is exactly the pattern the cohort's prescription-record depth manufactures, so it carries almost no weight here. The FAERS 'STRONG' flag with PRR 2.24 is the classic confounding-by-indication artifact of spontaneous reporting, where RA is listed as a concomitant condition on reports about a drug given for its pain. Nothing in the label, guideline or mechanistic literature supports amitriptyline as a cause of rheumatoid arthritis.","gene_comment":"rs79676622 (chr7:82,561,437, GRCh38) has no gene assignment and Ensembl classifies it as intergenic - it sits roughly 117 kb past the 3' end of CACNA2D1 and about 190 kb upstream of PCLO, too far to name either as mechanism, and it appears nowhere in the RA genetics literature, which is dominated by HLA-DRB1 and PTPN22. With a minor allele frequency near 0.5% in a case set of only a few hundred, this is a rare-variant locus with no biological anchor.","score_check":"The within-user signal is internally consistent (beta 2.465 with an implied SE near 0.5, log10p 6.46) and the disease arm is honestly null (beta 0.219 +/- 0.136, |beta| < 1.96*se), so z_diff 4.47 is arithmetically real; but an OR near 12 for a 0.5%-frequency SNV in ~180 dated cases is the shape of sparse-data bias, and would need replication before being believed.","candidability":2,"candidability_reason":"No labelled or literature-described joint ADR, a drug that is anti-inflammatory and used to treat RA pain, modest enrichment, and a rare intergenic variant with no gene - this reads as confounding by indication.","sources":[{"title":"Amitriptyline 10 mg Film-Coated Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/10849/smpc"},{"title":"The Efficacy and Safety of Antidepressants in Inflammatory Arthritis: A Cochrane Systematic Review (J Rheumatol)","url":"https://www.jrheum.org/content/90/21"},{"title":"Antidepressants for pain management in rheumatoid arthritis (Cochrane, PubMed 22071859)","url":"https://pubmed.ncbi.nlm.nih.gov/22071859/"},{"title":"Antidepressant analgesia in rheumatoid arthritis (PubMed 3236298)","url":"https://pubmed.ncbi.nlm.nih.gov/3236298/"},{"title":"The Anti-Inflammatory Potential of Tricyclic Antidepressants (TCAs) (PubMed 40006011)","url":"https://pubmed.ncbi.nlm.nih.gov/40006011/"},{"title":"Amitriptyline and nortriptyline inhibit interleukin-1 release by rat mixed glial and microglial cell cultures (PubMed 15963243)","url":"https://pubmed.ncbi.nlm.nih.gov/15963243/"},{"title":"Ensembl REST: variation rs79676622 and genes at 7:82.3-82.9 Mb","url":"https://rest.ensembl.org/variation/human/rs79676622?content-type=application/json"}]},"amitriptyline|M797":{"verdict":"licensed indication","headline":"Amitriptyline is a guideline-recommended treatment for fibromyalgia, not a cause of it","assessment":"Fibromyalgia is one of the best-established uses of low-dose amitriptyline: it is off-label against the US label (MDD only) but is an EMA-approved/EU-recognised analgesic indication for neuropathic pain and carries a weak positive recommendation in the 2016/2017 EULAR revised recommendations, with StatPearls listing chronic pain including fibromyalgia among its standard off-label uses. The Cochrane and Rheumatology systematic reviews find amitriptyline 25 mg/day reduces fibromyalgia pain (~26% mean pain reduction) and improves sleep and fatigue in a minority of patients, i.e. the drug moves this condition in the OPPOSITE direction to an ADR, which is decisive against a drug-caused reading. No mechanism by which amitriptyline induces widespread pain is described; its recognised harms are anticholinergic, sedative, cardiac (QTc, orthostatic hypotension) and weight gain. The atlas numbers are exactly what confounding by indication looks like: cotx_pct 1.38% with enrichment 1.46 (fibromyalgia over-represented among amitriptyline users), a null prescription-propensity signal (log10p 0.30, so the variant does not predict who gets the drug), and a FAERS PRR of 9.71 that in this direction almost certainly reflects fibromyalgia being entered as the reported indication rather than as a reaction. Temporality of 84.2% (median 7.0 years after first exposure) is the uninformative direction here and is further inflated by the well-known lag between starting amitriptyline for pain, insomnia or migraine and receiving a formal fibromyalgia label. The one genuinely interesting element is genetic rather than pharmacological: GRIA1 is a replicated gene-based hit in the large multi-ancestry fibromyalgia GWAS (85,139 cases), which supports the predisposition column (39.1) rather than an ADR.","gene_comment":"GRIA1 (GluA1 AMPA subunit) is a credible fibromyalgia susceptibility gene, significant in the EUR gene-based analysis of the 2025/2026 fibromyalgia GWAS, and GPC4 is the astrocytic proteoglycan that recruits AMPA receptors to synapses, so the pair is biologically coherent for pain susceptibility; neither has any described role in tricyclic pharmacology, and the three low-frequency SNVs (rs114846642, rs143889705, rs73239386) are not themselves reported variants at either locus.","score_check":"The disease-arm effect is believable (beta_disease 0.611, se 0.182, z=3.4, log10p 3.11), but beta_adr 2.548 (OR ~13) on a within-user analysis with only a few hundred datable cases is a sparse-data / winner's-curse inflated estimate, so the z_diff of 4.13 should not be read as a true drug-by-genotype interaction. Enrichment 1.46 with a null prescription-propensity arm is consistent with the condition driving the prescription rather than the reverse.","candidability":1,"candidability_reason":"Fibromyalgia is the reason amitriptyline is prescribed and the drug improves it, so this is confounding by indication with a plausible disease-susceptibility gene attached, not an ADR candidate.","sources":[{"title":"Amitriptyline - StatPearls, NCBI Bookshelf","url":"https://www.ncbi.nlm.nih.gov/books/NBK537225/"},{"title":"EULAR revised recommendations for the management of fibromyalgia (Ann Rheum Dis)","url":"https://ard.eular.org/article/S0003-4967(24)02857-7/abstract"},{"title":"Amitriptyline for fibromyalgia in adults - Cochrane Database of Systematic Reviews (Moore et al., 2015)","url":"https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD011824/full/de"},{"title":"Amitriptyline in the treatment of fibromyalgia: a systematic review of its efficacy (Rheumatology)","url":"https://academic.oup.com/rheumatology/article-pdf/47/12/1741/5046568/ken317.pdf"},{"title":"The genetics of fibromyalgia and its relationships to psychiatric and medical traits (PMC13415558)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13415558/"},{"title":"Astrocyte-Secreted Glypican 4 Regulates Release of Neuronal Pentraxin 1 to Induce Functional Synapse Formation (PMC5663462)","url":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5663462/"}]},"amitriptyline|R103":{"verdict":"statistical artifact","headline":"Lower-abdominal pain is what amitriptyline is prescribed to relieve; the variant is a rare-allele artifact","assessment":"R10.3 (pain localised to other parts of the lower abdomen) is not a licensed indication for amitriptyline, whose only FDA/EU approval is major depression, but it is one of the commonest off-label reasons the drug is used: low-dose amitriptyline is a gut-brain neuromodulator recommended by NICE, ACG and AGA for irritable bowel syndrome and other painful disorders of gut-brain interaction, and it is also used for chronic pelvic and bladder pain syndrome. The direction of the drug effect is therefore the opposite of an ADR: the ATLANTIS RCT (n=463) showed titrated 10-30 mg amitriptyline beat placebo on IBS-SSS (mean difference 27 points) and on a >=30% reduction in abdominal pain, and TCAs rank first for abdominal pain relief in IBS network meta-analyses. The only drug-caused route to a lower-abdominal pain code is indirect - anticholinergic constipation (1-10%) and rarely paralytic ileus - which would be coded as constipation far more often than as R10.3, and is not what a pain-relieving indication signal looks like. The atlas numbers fit an indication/comorbidity reading rather than an ADR: enrichment is 0.86 (if anything depleted versus users of other drugs, so not even a strong co-prescription enrichment), indication score 17.2, temporality 94.8% over 251 dated participants is the uninformative direction given the cohort's prescription-record lead-time, and FAERS support is weak (1/3 signals, PRR 1.66). The genetic claim does not survive inspection either: log10p_adr 6.2 is sub-genome-wide, the same variant is flatly null for the condition off-drug (beta 0.013 +/- 0.129), and the z_diff of 4.88 is generated entirely by an implausibly large within-users beta on a very rare allele.","gene_comment":"rs181365610 (chr2:65,832,689 A>G, GRCh38) is a rare SNV, gnomAD genomes G frequency ~0.36% and TOPMED ~0.32%, sitting inside LINC02934 - a ~765 kb lncRNA annotation (chr2:65.44-66.20 Mb) that is effectively a gene desert label. The nearest protein-coding genes are SPRED2 ~400 kb away and MEIS1 ~600 kb away, and neither this lncRNA nor those neighbours has any described role in visceral nociception or TCA pharmacology, so there is no mechanism here to dress up.","score_check":"beta_adr 3.587 (OR ~36) on an allele carried by roughly 0.3% of people is the classic sparse-data explosion from a handful of carrier cases, and the null, tight disease-arm estimate (0.013 +/- 0.129) shows the locus does nothing for abdominal pain generally. The prescription arm (log10p 1.35) is also unremarkable, so nothing independently corroborates the within-users hit.","candidability":1,"candidability_reason":"Backwards pharmacology - the drug is given to reduce this pain - plus a sub-genome-wide, rare-allele effect in a lncRNA desert with a null disease arm.","sources":[{"title":"Low-dose titrated amitriptyline as second-line treatment for adults with IBS in primary care: the ATLANTIS RCT (NIHR Journals Library, Discussion)","url":"https://www.ncbi.nlm.nih.gov/books/NBK608080/"},{"title":"Low-Dose Tricyclic Antidepressants for IBS: Definitive Evidence of Benefit from ATLANTIS (ACG, Evidence-Based GI)","url":"https://gi.org/journals-publications/ebgi/schoenfeld_dec2023/"},{"title":"Amitriptyline - StatPearls, NCBI Bookshelf (indications, off-label uses, anticholinergic and GI adverse effects)","url":"https://www.ncbi.nlm.nih.gov/books/NBK537225/"},{"title":"dbSNP RefSNP report rs181365610 (chr2:65,832,689 A>G; gnomAD/TOPMED frequencies)","url":"https://www.ncbi.nlm.nih.gov/snp/rs181365610"},{"title":"Ensembl gene record LINC02934 (ENSG00000204929), lncRNA, chr2:65,435,852-66,200,373","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/LINC02934"},{"title":"Effect of amitriptyline on symptoms in treatment-naive patients with interstitial cystitis/painful bladder syndrome (PubMed 20303115)","url":"https://pubmed.ncbi.nlm.nih.gov/20303115/"}]},"amitriptyline|R21":{"verdict":"plausible ADR","headline":"Rash is a labelled amitriptyline hypersensitivity reaction, but a chr18 lncRNA hit at 3-year lag is not it","assessment":"Rash is not an indication for amitriptyline (the US label lists only relief of symptoms of depression) but it is an explicitly labelled adverse reaction: the Allergic section of the FDA SPL reads 'Skin rash; urticaria; photosensitization; edema of face and tongue', and the case literature supports it, including a probable (WHO-UMC and Naranjo) generalised maculopapular rash with urticarial pruritus and angio-oedema on 10 mg/day, amitriptyline-induced hypersensitivity syndrome with erythroderma, and rare Stevens-Johnson syndrome; a large inpatient pharmacovigilance series put clinically significant cutaneous ADRs at ~0.1% of psychotropic exposures with older tricyclics carrying more risk than newer agents. So a drug-caused mechanism exists and is described - immune-mediated (type I/type IV) hypersensitivity plus photosensitivity - but it is a rare idiosyncratic event, and against it stands the fact that amitriptyline is a potent H1 (and H2) antagonist of the same chemical class as doxepin, which raises the histamine itch threshold and is used therapeutically for chronic idiopathic urticaria; at population level the drug can plausibly suppress, not raise, urticarial rash. The atlas numbers point the same way: rash is depleted rather than enriched in amitriptyline users (enrichment 0.56, cotx_pct 0.47%), and FAERS shows no disproportionality (PRR 1.3, weak 1/3). Temporality is 100% post-exposure but on only 48 dated participants with 57.5% undated, and the median 3.06 years from first exposure to first rash record is wrong by orders of magnitude for drug hypersensitivity, which classically appears within hours to 8 weeks - that lag looks like the record-ordering compression the atlas itself warns about, not causation. The within-user signal is genuinely stratum-specific (log10p_adr 6.98 against a null disease arm, beta_disease 0.114 +/- 0.134, and a null prescription-propensity arm, z_diff 5.09), which is the right shape for a pharmacogenomic hit, but it rests on a handful of cases and an implausible locus.","gene_comment":"rs75178920 is annotated to NETO1-DT, a divergent-transcript lncRNA at 18q22.3 sitting antisense to a neuronal NMDA-receptor auxiliary subunit gene, with no known skin, immune or drug-metabolism role; the established genetics of drug-induced maculopapular exanthema is HLA class I/II (e.g. HLA-A*31:01, HLA-B*38:02, HLA-DRB1*04:06 for antiepileptics), and nothing of that kind appears here. This is a weak gene assignment and must not be presented as mechanism.","score_check":"beta_adr = 4.057 on the log-odds scale for a single SNV is far larger than any credible common-variant effect and, with roughly 48 dated cases, is the signature of a rare-allele sparse-data inflation rather than a real effect size. The comparison arms are internally consistent (disease and prescription arms both null, |beta_disease| well under 1.96*se_disease), so the stratum specificity is believable even though the magnitude is not.","candidability":4,"candidability_reason":"Rash is a genuine labelled amitriptyline ADR, but this instance is depleted rather than enriched, lags exposure by a median of three years, carries no FAERS support, and rests on an inflated effect at a neuronal lncRNA with no skin or immune plausibility.","sources":[{"title":"Amitriptyline Hydrochloride Tablets, USP - FDA label (SPL)","url":"https://www.accessdata.fda.gov/spl/data/0f12f50f-7087-46e7-a2e6-356b4c566c9f/0f12f50f-7087-46e7-a2e6-356b4c566c9f.xml"},{"title":"Acute Hypersensitivity Reaction With Maculopapular Exanthem and Angio-Oedema Following Low Dose Amitriptyline for Migraine Prophylaxis: A Case Report","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12491646/"},{"title":"Hypersensitivity syndrome caused by amitriptyline administration (Postgrad Med J)","url":"https://pubmed.ncbi.nlm.nih.gov/10824052/"},{"title":"Severe skin complications in patients treated with antidepressants: a literature review","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC4112250/"},{"title":"Inhibition of histamine-induced pruritus by topical tricyclic antidepressants (doxepin, amitriptyline)","url":"https://pubmed.ncbi.nlm.nih.gov/7298924/"},{"title":"HLA Risk Alleles in Aromatic Antiepileptic Drug-Induced Maculopapular Exanthema","url":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8187898/"},{"title":"NETO1-DT Gene - GeneCards (lncRNA, 18q22.3)","url":"https://www.genecards.org/cgi-bin/carddisp.pl?gene=NETO1-DT"}]},"amlodipine|B349":{"verdict":"statistical artifact","headline":"Nonspecific viral-infection code, absent from label and FAERS; amlodipine if anything trends antiviral","assessment":"B34.9 'Viral infection, unspecified' is neither an indication for amlodipine nor a recognised adverse effect: the drug is licensed for hypertension, chronic stable angina, vasospastic angina and documented CAD, and its adverse-effect profile is peripheral oedema, flushing, headache, dizziness, gingival hyperplasia and rare hepatotoxicity, with no infectious component. A 2025 FAERS disproportionality analysis of amlodipine found 27 consistent signals dominated by cardiac, respiratory, gastrointestinal, skin and gingival events and reported no infectious-disease pattern at all. The atlas numbers agree that this is not confounding by indication - cotx_pct is only 0.35% and enrichment 0.90, i.e. viral-infection codes are marginally less common in amlodipine users than in users of other drugs - but they equally give no positive support for an ADR. Temporality of 97% is uninformative here: an acute, self-limiting coded viral illness will almost always be recorded after the start of a lifelong antihypertensive, so this is exactly the upward compression the cohort is known to produce, and the median 3.6-year gap argues for background incidence rather than a drug reaction. The within-user effect (beta 5.671, OR of order 10^2) is biologically implausible for a common acute outcome and is the classic shape of a rare variant with a handful of carrier cases; the disease arm is frankly null (beta 0.27, se 0.178, |beta| well under 1.96*se, log10p 0.89) and the prescription arm is null too, so z_diff 5.01 is carried entirely by the unstable treated-arm estimate. If anything the literature points the other way: dihydropyridine CCBs including amlodipine inhibit post-entry replication of SARS-CoV-2 and several other RNA viruses in vitro, amlodipine use was associated with reduced COVID-19 case fatality in hypertensive patients, and CCB users show lower, not higher, infection incidence than users of other antihypertensive classes - a direction of effect that is decisive evidence against reading this as a drug-caused infection risk.","gene_comment":"ENSG00000295272 is a novel unnamed lncRNA on chr5p15.33 (chr5:2,558,708-2,559,874), sitting in a gene desert near the IRX cluster with no immune, antiviral or drug-disposition annotation, and rs191054889 is a rare variant with no published association literature. This is a coordinate with a transcript label attached, not a mechanism, and it cannot be dressed up as one.","score_check":"The numbers are internally consistent only in the sense that everything except the treated-arm beta is null; a beta of 5.67 alongside a null disease effect, null prescription effect, enrichment below 1 and log10p 7.02 that does not clear genome-wide significance is the signature of sparse-data bias in a rare-variant test, not of a real effect. Nothing here survives contact with the label, FAERS or the CCB infection literature.","candidability":1,"candidability_reason":"Nonspecific garbage outcome code, no label/pharmacovigilance support, null disease and prescription arms, an implausibly large rare-variant beta, and literature pointing in the protective direction.","sources":[{"title":"Amlodipine - StatPearls, NCBI Bookshelf","url":"https://www.ncbi.nlm.nih.gov/books/NBK519508/"},{"title":"Adverse events associated with amlodipine: a pharmacovigilance study using the FDA Adverse Event Reporting System","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12089059/"},{"title":"Calcium channel blocker amlodipine besylate therapy is associated with reduced case fatality rate of COVID-19 patients with hypertension (Cell Discovery)","url":"https://www.nature.com/articles/s41421-020-00235-0"},{"title":"Calcium Ion Channels: Roles in Infection and Sepsis; Mechanisms of Calcium Channel Blocker Benefits in Immunocompromised Patients at Risk for Infection","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC6164603/"},{"title":"Association between antihypertensive medication and the risk of urinary tract infection of outpatients: a retrospective cohort study","url":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10042971/"},{"title":"Ensembl REST lookup, ENSG00000295272 (novel lncRNA, chr5:2,558,708-2,559,874)","url":"https://rest.ensembl.org/lookup/id/ENSG00000295272?content-type=application/json"},{"title":"2026 ICD-10-CM Diagnosis Code B34.9: Viral infection, unspecified","url":"https://www.icd10data.com/ICD10CM/Codes/A00-B99/B25-B34/B34-/B34.9"}]},"amlodipine|E669":{"verdict":"co-prescription population","headline":"Obesity is a driver of the hypertension amlodipine treats, not an amlodipine adverse effect","assessment":"Amlodipine is licensed only for hypertension and for chronic stable / vasospastic angina; obesity (E66.9) is neither an indication nor a labelled adverse reaction. The US label lists 'weight gain' and 'weight decrease' together among postmarketing General events at <1-2%, i.e. non-specific and bidirectional, and the mechanistically real weight signal for a dihydropyridine is dose-dependent peripheral oedema (10.8% at 10 mg vs 0.6% placebo), which raises measured body weight through fluid, not adiposity, and would not be coded as obesity. The causal arrow in the literature runs the other way: 60-70% of adult hypertension is attributable to adiposity and obesity prevalence among US hypertensives rose to 55.4% by 2023, so obese patients are systematically funnelled into antihypertensive prescribing. The atlas numbers are consistent with that reading rather than with an ADR: enrichment is only 1.12 and cotx_pct 6.34% (E66.9 is heavily under-coded), the indication flag is 49.6, and temporality of 94.9% is exactly the uninformative direction the cohort compresses toward, since obesity is a lifelong trait opportunistically coded years into chronic care (median 4.16 y). Metabolically amlodipine is regarded as neutral, with ALLHAT showing no adverse metabolic profile for amlodipine relative to comparators, so there is no described mechanism by which it would create obesity. The within-user genetic signal (log10p 7.64) sits on a null disease arm (beta 0.117 +/- 0.341) and a null prescription arm, but the effect size makes it a sparse-data artifact rather than an interaction worth chasing.","gene_comment":"LRCOL1 (leucine-rich colipase-like 1, a CLPS paralog) is an essentially uncharacterised human gene whose only annotations - digestion, lipid catabolism - are computational predictions by paralogy, with no ClinGen curation and no established obesity or BMI GWAS locus. A LOF burden hit in such a gene with no independent support carries no mechanistic weight.","score_check":"beta_adr of 7.209 for a LOF burden test corresponds to an odds ratio in the hundreds-to-thousands, which is not a believable effect for common obesity and is the signature of near-separation in a handful of LOF carriers. The comparison arms are all null (log10p_disease 0.14, log10p_prescribed 0.08) and FAERS support is negligible (weak 1/3, PRR 1.45), so the large z_diff reflects the unstable within-user estimate rather than a real treated-vs-untreated difference.","candidability":1,"candidability_reason":"Obesity is a cause of the treated condition, not an amlodipine reaction, and the genetics rest on an implausibly large LOF burden effect in an uncharacterised gene.","sources":[{"title":"DailyMed - AMLODIPINE BESYLATE tablet (US prescribing information)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a516b336-8f60-4b86-8921-c7d6c07629f2"},{"title":"Trends in Obesity Prevalence Among Adults With Hypertension in the United States, 2001 to 2023 (Hypertension, AHA)","url":"https://www.ahajournals.org/doi/10.1161/HYPERTENSIONAHA.124.24123"},{"title":"Obesity-Related Hypertension: Epidemiology, Pathophysiology, and Clinical Management (Am J Hypertens)","url":"https://academic.oup.com/ajh/article/23/11/1170/197863"},{"title":"Does amlodipine cause weight gain? - Drugs.com medical answers","url":"https://www.drugs.com/medical-answers/amlodipine-weight-gain-3567986/"},{"title":"Metabolic and Clinical Outcomes in Nondiabetic Individuals With the Metabolic Syndrome Assigned to Chlorthalidone, Amlodipine, or Lisinopril (ALLHAT, Diabetes Care)","url":"https://diabetesjournals.org/care/article/31/2/353/25246/Metabolic-and-Clinical-Outcomes-in-Nondiabetic"},{"title":"LRCOL1 Gene - GeneCards","url":"https://www.genecards.org/cgi-bin/carddisp.pl?gene=LRCOL1"}]},"amlodipine|E86":{"verdict":"contested","headline":"Amlodipine's documented fluid effect is oedema, not depletion; E86 rides a rare intergenic variant","assessment":"ICD-10 E86 (dehydration, hypovolaemia, unspecified volume depletion) is not an amlodipine indication - the licensed uses are hypertension, chronic stable angina, vasospastic angina and documented CAD - and volume depletion is not a labelled reaction; the canonical amlodipine fluid effect runs in the opposite direction, arteriolar-selective vasodilatation driving interstitial fluid accumulation and dose-related peripheral oedema in roughly 10.8% of patients at 10 mg, and the standard teaching is explicitly that this oedema must not be diuresed. Two indirect drug-caused routes to an E86 code do exist and are described: dihydropyridines are acutely natriuretic and lower aldosterone, and amlodipine's own hypotension, dizziness and syncope are coded in the same clinical episodes; more importantly, there is a well-documented prescribing cascade in which CCB oedema is misread and a loop diuretic added (about 3.5% of older CCB starters within a year, 9.5% for any diuretic), with overdiuresis, AKI and falls as the named downstream harms - but that pathway is mediated by the added diuretic, not by amlodipine pharmacodynamics or by a patient's genotype. The atlas itself argues against a drug-specific burden: cotx_pct is only 0.74% with enrichment 0.98, i.e. volume depletion is recorded no more often in amlodipine users than in users of any other drug, so this is background comorbidity of an older, polypharmacy population rather than an amlodipine excess. The comparison arms are, however, clean - the variant is null for the condition in the drug-free population (beta 0.142 +/- 0.112, well inside noise) and null for being prescribed amlodipine (log10p 0.93) - so this is not straightforward confounding by indication or a general predisposition allele leaking through. The FAERS support is weak in substance despite the STRONG label: PRR 2.1 is modest, dehydration-type terms co-report heavily with diuretics and antihypertensives generally, and the published FAERS analysis of amlodipine surfaces gingival, cardiac and fluid-retention terms, not dehydration or hypovolaemia. Temporality of 100% on 176 dated participants is the least informative possible value given the cohort's known upward compression, and a median 4.66 years to onset fits ageing and accumulating comorbidity at least as well as it fits a drug effect.","gene_comment":"rs77625269 (chr16:28,093,273, GRCh38) is intergenic in dbSNP - it sits about 4.7 kb upstream of the XPO6 transcription start and about 30 kb past the end of GSG1L, so 'XPO6' here is a nearest-gene label, not an implicated gene. XPO6 is the profilin-actin nuclear export receptor with no described role in sodium, water or volume homeostasis, and the variant is rare (MAF ~0.7%) with no ClinVar entry, so it supplies no mechanism.","score_check":"Internally the numbers are consistent - log10p_adr 6.4 with beta 2.652 implies SE around 0.53, and z_diff 4.69 against a null disease effect is coherent - but an OR near 14 on a ~0.7% allele means very few carriers, which is exactly the regime where sparse-data bias inflates effects, and no independent replication supports it. The flat enrichment (0.98) and the absent BNF listing are the more telling numbers and they point away from a drug-specific event.","candidability":3,"candidability_reason":"No excess of volume depletion in amlodipine users, the drug's documented fluid effect is oedema in the opposite direction, and the only credible drug-caused route is a diuretic-mediated prescribing cascade that a rare intergenic variant near XPO6 does nothing to explain.","sources":[{"title":"Amlodipine - StatPearls, NCBI Bookshelf (indications, adverse effects, contraindications)","url":"https://www.ncbi.nlm.nih.gov/books/NBK519508/"},{"title":"Adverse events associated with amlodipine: a pharmacovigilance study using the FDA Adverse Event Reporting System","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12089059/"},{"title":"Potentially harmful 'prescribing cascade' common in older adults with hypertension (CCB oedema to loop diuretic)","url":"https://www.healio.com/news/primary-care/20200224/potentially-harmful-prescribing-cascade-common-in-older-adults-with-hypertension"},{"title":"dbSNP rs77625269 (chr16:28093273, intergenic, MAF ~0.7%)","url":"https://www.ncbi.nlm.nih.gov/snp/rs77625269"},{"title":"Ensembl REST overlap, human chr16:28,040,000-28,150,000 (XPO6 start 28,097,976; GSG1L end 28,063,754)","url":"https://rest.ensembl.org/overlap/region/human/16:28040000-28150000?feature=gene;content-type=application/json"}]},"amlodipine|I64":{"verdict":"co-prescription population","headline":"Amlodipine cuts stroke risk in trials; this rare intronic lncRNA hit in ~100 cases points the wrong way","assessment":"Stroke (I64) is not a licensed indication for amlodipine and is not a labelled adverse reaction: the FDA Norvasc label lists hypertension and coronary artery disease/angina as indications, edema, palpitations, dizziness and flushing as the characteristic adverse effects, and explicitly frames blood-pressure lowering as reducing the risk of stroke and myocardial infarction. The relationship in the data is therefore comorbidity of the treated population - amlodipine users are hypertensives, and hypertension is the dominant modifiable stroke risk factor - not drug causation. Crucially the drug moves this outcome in the opposite direction: a meta-analysis of 31 RCTs and 273,543 participants gives calcium-channel blockers an OR of 0.68 (95% CI 0.61-0.75) for stroke versus placebo and 0.79 versus beta-blockers, and amlodipine-based therapy in ASCOT-BPLA reduced stroke by 23% versus atenolol, which is decisive evidence against an ADR reading. The atlas is consistent with this rather than with an ADR: enrichment is 0.77, i.e. stroke is if anything less frequent among amlodipine users than among users of other drugs, the variant does not predict being prescribed amlodipine (log10p 0.21), and the FAERS signal is weak (1/3, PRR 1.44) with no BNF listing. The high temporality (90.9%) is uninformative here - only 44 of roughly 100 affected participants had a usable date (57.7% undated), and diagnoses after first prescription are exactly what a lifelong antihypertensive predicts. No mechanism by which amlodipine would cause unspecified stroke is described in the literature I found; the one theoretical route, excessive blood-pressure lowering with hypoperfusion, is a dose/frailty phenomenon and is not what a single rare intronic SNV would tag.","gene_comment":"ENSG00000249998 is an unnamed havana lncRNA ('novel transcript') at 4p15.32, and rs13113983 (MAF ~2%, C/T) is an intron variant inside it with no entries in the GWAS Catalog; the only other features in the interval are a second unnamed lncRNA and a mitochondrial pseudogene (MTND5P4), with LDB2 the nearest protein-coding gene some way upstream. This is not the well-known 4q25/PITX2 cardioembolic-stroke locus - that is the other chromosome arm - so there is no mechanistic story to attach to this assignment.","score_check":"beta_adr 2.228 (OR ~9) on a 2%-MAF variant with only ~100 cases among amlodipine users, at log10p 6.26 (below genome-wide significance), is the classic sparse-data pattern of an inflated effect rather than a real one, and the z_diff of 4.34 is driven almost entirely by that inflated within-drug beta. The disease-arm effect is essentially null (beta 0.228, se 0.118, |beta| = 1.9*se), so the variant is not even a convincing stroke predisposition allele in the drug-free population.","candidability":1,"candidability_reason":"Backwards signal: amlodipine reduces stroke in randomised trials, stroke is depleted rather than enriched in its users here, and the genetic hit is a sparse-data effect in an anonymous lncRNA.","sources":[{"title":"The Effects of Calcium Channel Blockers in the Prevention of Stroke in Adults with Hypertension: A Meta-Analysis of 273,543 Participants in 31 RCTs (PLOS ONE)","url":"https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0057854"},{"title":"NORVASC (amlodipine besylate) tablets - FDA-approved label","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2017/019787s062lbl.pdf"},{"title":"Amlodipine in the current management of hypertension (PMC10497034)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10497034/"},{"title":"Ensembl REST: gene ENSG00000249998 (lncRNA, novel transcript, chr4:16,973,275-17,073,903)","url":"https://rest.ensembl.org/lookup/id/ENSG00000249998?content-type=application/json"},{"title":"Ensembl REST: variant rs13113983 (chr4:17,056,292, intron_variant, MAF 0.021)","url":"https://rest.ensembl.org/variation/human/rs13113983?content-type=application/json"},{"title":"GWAS Catalog REST: associations for rs13113983 (none reported)","url":"https://www.ebi.ac.uk/gwas/rest/api/singleNucleotidePolymorphisms/rs13113983/associations"}]},"amlodipine|I959":{"verdict":"plausible ADR","headline":"Hypotension is amlodipine's on-target exaggerated effect; the TWIST1 rare variant is not credible","assessment":"Hypotension is not an indication for amlodipine but the extension of its pharmacology: the Norvasc label states amlodipine 'is a peripheral arterial vasodilator that acts directly on vascular smooth muscle to cause a reduction in peripheral vascular resistance and reduction in blood pressure', warns that 'symptomatic hypotension is possible, particularly in patients with severe aortic stenosis', and describes overdose as 'excessive peripheral vasodilation with marked hypotension'. So an ADR reading is mechanistically certain rather than merely plausible, and there is no opposite-direction problem - the drug moves BP in exactly the direction of the coded outcome. Independent support is consistent: ALLHAT's older participants had increased falls in the first year after amlodipine initiation versus chlorthalidone and lisinopril (though no long-term excess of measured orthostatic hypotension), amlodipine is routinely counted among orthostatic-hypotension-inducing drugs in fall-risk prescribing work, and the atlas's own FAERS arm is STRONG with PRR 3.49. The atlas's enrichment of 0.69 says hypotension is actually depleted among amlodipine users relative to users of other drugs, which is what one expects when the treated population is hypertensive - it argues against confounding by indication rather than for it, and log10p_prescribed = 0.0 rules out a prescription-propensity artifact. Temporality of 99% over 201 dated participants with a median 5.19 years to first hypotension code is directionally right but, as the cohort compresses this measure upward, carries little independent weight, and a five-year median fits age and cumulative polypharmacy at least as well as a drug-initiation effect. The weak link is entirely genetic: this is a single ultra-rare SNV carrying an implausible within-user effect on a condition for which the same variant is flatly null in the drug-free population (beta_disease 0.056 +/- 0.144).","gene_comment":"rs537752533 (chr7:19,099,338, GRCh38) is a G>A SNV at ~0.3% global and ~0.5% non-Finnish-European frequency whose only Ensembl VEP consequence is intronic in a nonsense-mediated-decay isoform of TWIST1 - it is not in the canonical two-exon coding gene and is not one of the known 7p21.1 regulatory variants (rs2107595, rs8084351) that make TWIST1 a credible causal gene for atherosclerotic vascular disease. TWIST1 is a craniofacial/mesenchymal bHLH transcription factor with no described role in vascular tone, calcium-channel biology or dihydropyridine pharmacokinetics, so the gene label here is positional, not mechanistic.","score_check":"beta_adr 4.416 is an odds ratio near 83 for a variant carried by roughly one in a hundred people, in a stratum where only ~201 participants have a dated hypotension record - that is the signature of a handful of carrier cases, not a real effect of that magnitude, and the log10p of 9.37 and z_diff of 6.04 inherit the same sparse-count instability. The non-genetic numbers are internally coherent (null disease arm, null prescription arm, enrichment below 1, strong FAERS), so the phenotype-level claim survives even though the variant-level claim does not.","candidability":5,"candidability_reason":"A genuine, on-label, mechanistically inevitable drug effect whose genetics here rest on one ultra-rare intronic variant with an implausible effect size and no pharmacological link to TWIST1.","sources":[{"title":"DailyMed - NORVASC (amlodipine besylate) tablet, full prescribing information","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=abd6a2ca-40c2-485c-bc53-db1c652505ed"},{"title":"FDA-approved label, Norvasc (amlodipine besylate) tablets, NDA 019787","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2011/019787s047lbl.pdf"},{"title":"Juraschek et al., Effects of Antihypertensive Class on Falls, Syncope, and Orthostatic Hypotension in Older Adults (ALLHAT), Hypertension","url":"https://www.ahajournals.org/doi/10.1161/hypertensionaha.119.13445"},{"title":"Amlodipine and frusemide: pharmacological factors contributing to increased fall risk in concurrently treated patients (PMC12336181)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12336181/"},{"title":"Ensembl REST: rs537752533 population frequencies and VEP consequences (GRCh38 chr7:19,099,338)","url":"https://rest.ensembl.org/vep/human/id/rs537752533?content-type=application/json"},{"title":"Twist1 promotes endothelial phenotypic transition and unstable plaque phenotype during atherosclerosis (TWIST1 as the 7p21.1 causal gene)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13544331/"}]},"amlodipine|J310":{"verdict":"plausible ADR","headline":"Rhinitis is a labelled uncommon amlodipine ADR, but the signal rests on an implausible CNV in a gene desert","assessment":"Chronic rhinitis is not an indication for amlodipine: the EMC SmPC and the Norvasc US label licence it only for hypertension, chronic stable angina and vasospastic (Prinzmetal) angina. It is, however, an explicitly labelled adverse reaction - the EU SmPC lists 'Cough, rhinitis' as uncommon (>=1/1,000 to <1/100) under respiratory disorders, while the US label lists epistaxis rather than rhinitis at 0.1-1%. A mechanism is described and points the right way: dihydropyridine calcium-channel blockers relax vascular smooth muscle, and vasodilation of the nasal mucosal sinusoids with interstitial oedema produces congestion and rhinorrhoea; StatPearls names calcium channel blockers among the antihypertensives causing drug-induced rhinitis, and there is no sense in which amlodipine decongests, so no opposite-direction evidence exists. The atlas is consistent with 'not confounding by indication' but also with 'no cohort-level excess': cotx_pct is only 0.77% with enrichment exactly 1.0 (chronic rhinitis is no commoner in amlodipine users than in users of other drugs) and the prescription-propensity arm is null (log10p 0.98). Temporality of 84.3% on 108 dated participants, median 3.6 years after first exposure, is a high value and therefore weak evidence in this cohort, where prescription records reach further back than diagnosis records. The FAERS disproportionality flag (PRR 2.88, STRONG) supports a drug-level rhinitis signal, though the published FAERS-wide amlodipine study found gingival hypertrophy and dyspnoea, not nasal events, among its leading signals. The genetic layer adds nothing: the whole association is carried by one CNV deletion call in a region containing no protein-coding gene, and the same call is nominally associated with rhinitis in the drug-free population in the same direction.","gene_comment":"The single 'gene' is a CNV-deletion window at chr6:76.60-76.70 Mb, which by Ensembl contains only a processed pseudogene, an unnamed lncRNA and a few enhancer/CTCF annotations; it sits in a ~1.4 Mb desert between IMPG1 (ends 76.07 Mb, retina-specific) and HTR1B (starts 77.46 Mb), neither of which has any nasal-mucosa relevance. This is an intergenic CNV call, not a mechanism, and should not be presented as one.","score_check":"beta_adr of 7.44 on a log-odds scale (OR ~1700) for a rare CNV is a sparse-data / quasi-separation artifact rather than a real effect, so z_diff 3.48 largely measures that artifact; the only interpretable estimate is the drug-free arm's 2.23 +/- 0.737 (z ~3.0), which says the deletion call tracks rhinitis with or without the drug. Enrichment of exactly 1.0 and a null prescription arm rule out co-prescription confounding but equally show no population-level excess of rhinitis under amlodipine.","candidability":5,"candidability_reason":"Rhinitis is a genuine, labelled, mechanistically coherent uncommon amlodipine ADR, but the genetics offered here - one implausibly large CNV effect in a gene desert that also predisposes without the drug - is unconvincing as a pharmacogenomic finding.","sources":[{"title":"Amlodipine 10mg tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/4718/smpc"},{"title":"NORVASC (amlodipine besylate) label - DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=abd6a2ca-40c2-485c-bc53-db1c652505ed"},{"title":"Rhinitis Medicamentosa - StatPearls, NCBI Bookshelf","url":"https://www.ncbi.nlm.nih.gov/books/NBK538318/"},{"title":"Adverse events associated with amlodipine: a pharmacovigilance study using the FDA Adverse Event Reporting System","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12089059/"},{"title":"Ensembl REST overlap, human chr6:76,600,000-76,700,000 (genes and regulatory features)","url":"https://rest.ensembl.org/overlap/region/human/6:76600000-76700000?feature=gene;feature=regulatory;content-type=application/json"}]},"amlodipine|K589":{"verdict":"statistical artifact","headline":"Amlodipine slows gut transit, but IBS-without-diarrhoea is the wrong code and AGBL4 the wrong gene","assessment":"Irritable bowel syndrome is nowhere in the amlodipine label: the licensed indications are hypertension and coronary artery disease, and the only bowel entries are abdominal pain (1.6% vs 0.3% placebo), nausea (2.9% vs 1.9%) and constipation, dyspepsia and diarrhoea each at <1% but >0.1%. A class-level mechanism for slowed transit is real and described - L-type calcium channel blockade reduces intestinal smooth-muscle contraction - and a prospective cohort of 556 medical inpatients found dihydropyridine CCBs an independent risk factor for constipation (OR 1.8, 95% CI 1.2-2.8), with amlodipine cited at roughly 2% versus 4-9% for verapamil. But that effect is constipation (K59.0), not IBS: a clinician coding a new antihypertensive's bowel side effect does not write K58.9, and IBS is typically a long-standing functional diagnosis in younger, more often female patients, which fits the enrichment of 0.73 (IBS-without-diarrhoea is depleted, not enriched, among amlodipine users) and the null prescription-propensity arm (log10p 0.89). The direction is also not one-way: gut-selective L-type calcium antagonists (pinaverium, otilonium) are Grade A/B first-line IBS antispasmodics with NNT around 4-4.5, improving pain, bloating and stool consistency, so the same pharmacology that hardens stool relieves the spasm and pain that define IBS. Temporality of 81.7% over 180 dated participants with a 3.84-year median lag is exactly the value this cohort's record-depth asymmetry produces by construction and adds nothing. FAERS shows no signal (PRR 0.88) and the association is not on the BNF list.","gene_comment":"AGBL4 (CCP6) is a tubulin deglutamylase at 1p33 whose gene body spans about 1.5 Mb and whose expression is brain- and testis-dominant, with no enteric, smooth-muscle or calcium-handling role; a single intronic rare SNV in a target that large is a high-prior incidental annotation. AGBL4 is not among the six replicated IBS susceptibility loci (NCAM1, CADM2, PHF2/FAM120A, DOCK9, CKAP2/TPTE2P3, BAG6).","score_check":"beta_adr 4.025 implies an odds ratio near 55 for a common functional bowel diagnosis, which is the classic sparse-data inflation signature of a low-frequency variant with few carriers rather than a real effect size, and z_diff 5.06 is essentially the ADR z-score alone because the disease arm is flatly null (beta -0.069 +/- 0.115, |beta| well under 1.96*se). Nothing else in the sheet corroborates it: no disease-arm effect, no prescription-propensity effect, no FAERS signal, and enrichment below 1.","candidability":2,"candidability_reason":"Real class effect exists but it is constipation, not IBS; the coded outcome is depleted in users and the genetic hit is an implausibly large rare-variant effect in a gut-irrelevant gene.","sources":[{"title":"DailyMed - AMLODIPINE BESYLATE tablet (US label: indications and adverse reactions)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=767c1e67-4cfb-4358-a963-3d58122c8e83"},{"title":"Prevalence, Recognition, and Risk Factors of Constipation among Medically Hospitalized Patients: A Cohort Prospective Study","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10385149/"},{"title":"Expert Opinion on the Efficacy and Safety of Antispasmodics with a Focus on Irritable Bowel Syndrome","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13007261/"},{"title":"Genome-wide analysis of 53,400 people with irritable bowel syndrome highlights shared genetic pathways with mood and anxiety disorders (Nat Genet 2021)","url":"https://pubmed.ncbi.nlm.nih.gov/34741163/"},{"title":"AGBL4 - gene overview (locus 1p33, function, tissue expression)","url":"https://en.wikipedia.org/wiki/AGBL4"}]},"amlodipine|K922":{"verdict":"contested","headline":"GI bleeding is a contested CCB class effect, not an amlodipine label event; the GRM7 variant is not credible","assessment":"Amlodipine is licensed only for hypertension, chronic stable angina and vasospastic angina, and gastrointestinal haemorrhage appears nowhere in the SmPC's undesirable effects, which list only abdominal pain, nausea, dyspepsia and altered bowel habit as common and pancreatitis, gastritis and gingival hyperplasia as very rare. The class-level question is genuinely contested: Pahor's 1996 Lancet cohort of hypertensive people over 67 reported a relative risk of 1.68 (1.03-2.74) for severe GI bleeding on calcium antagonists versus beta-blockers, and a 2015 meta-analysis found a summary RR of 1.17 (1.01-1.36) that became non-significant once prior bleeding history or anti-ulcer co-medication was adjusted for and was 0.93 in randomised trials. The decisive evidence is randomised: in ALLHAT, amlodipine did not increase hospitalised GI haemorrhage versus chlorthalidone (HR 1.09, 0.92-1.28) and had fewer bleeds than lisinopril, so the observational excess looks like confounding rather than drug effect. The atlas is internally consistent with that reading rather than against it: enrichment is 0.81, i.e. K92.2 is slightly depleted among amlodipine users compared with users of other drugs, and predisposition is 0.0, so this is not obviously confounding by indication either, but neither is it an enriched ADR signal. The 99.6% temporality with a 5.05-year median gap is the expected artefact of prescription records reaching further back than diagnosis records in an elderly hypertensive population and carries almost no weight. The FAERS disproportionality (PRR 2.01) is consistent with the old cohort literature but is subject to the same confounding, since amlodipine users are old and frequently co-treated with antiplatelets, anticoagulants and NSAIDs.","gene_comment":"rs141129125 is a rare intronic T>A variant in GRM7 (gnomAD MAF ~0.4%), absent from ClinVar and with no reported bleeding or vascular phenotype; GRM7 itself is a neuronal metabotropic glutamate receptor whose established disease role is recessive epileptic encephalopathy and intellectual disability. There is no mechanistic route from an intronic GRM7 variant to mucosal or platelet haemostasis, so this must not be presented as mechanism.","score_check":"beta_adr of 3.75 (OR about 42) on a variant carried by well under 1% of the cohort is the signature of a sparse-data artefact, however small the p-value looks, and the drug-free arm is flatly null (beta 0.221 +/- 0.144, well inside noise), so z_diff of 5.63 is driven entirely by the unstable treated-arm estimate. The population-level numbers (cotx 1.75%, enrichment 0.81) are believable and unremarkable; the variant effect size is not.","candidability":3,"candidability_reason":"Class-level GI bleeding signal exists but is refuted by randomised ALLHAT data, GI haemorrhage is not on the amlodipine label, enrichment is below 1, and the driving variant is a rare intronic GRM7 SNV with an implausibly large effect.","sources":[{"title":"Amlodipine 5mg Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/3478/smpc"},{"title":"Risk of Hospitalized Gastrointestinal Bleeding in Persons Randomized to Diuretic, ACE-Inhibitor, or Calcium-Channel Blocker in ALLHAT","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3844932/"},{"title":"Risk of gastrointestinal haemorrhage with calcium antagonists in hypertensive persons over 67 years old (Pahor et al., Lancet 1996)","url":"https://pubmed.ncbi.nlm.nih.gov/8602055/"},{"title":"Systematic review with meta-analysis: the association between the use of calcium channel blockers and gastrointestinal bleeding (Aliment Pharmacol Ther 2015)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=DOI:%2210.1111/apt.13211%22&resultType=core&format=json"},{"title":"rs141129125 RefSNP Report - dbSNP, NCBI","url":"https://www.ncbi.nlm.nih.gov/snp/rs141129125"},{"title":"Biallelic GRM7 variants cause epilepsy, microcephaly, and cerebral atrophy","url":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7261753/"}]},"amlodipine|M751":{"verdict":"co-prescription population","headline":"Rotator cuff syndrome tracks hypertension, not amlodipine; the PTPN14 3'UTR effect size is implausible","assessment":"Rotator cuff syndrome (M75.1) is not a licensed indication for amlodipine, whose label covers hypertension, chronic stable angina and vasospastic angina, and shoulder/rotator cuff disease appears nowhere in the SmPC adverse-reaction table; the only musculoskeletal terms listed are ankle oedema, muscle cramps, arthralgia, myalgia and back pain, and even those are uncommon-to-rare. The condition is, however, strongly tied to the population amlodipine treats: a systematic review and meta-analysis of cardiometabolic risk factors reports increased odds of rotator cuff disease with hypertension (OR 1.40, 95% CI 1.19-1.65), and a clinical series found hypertensive patients roughly twice as likely to have large and four times as likely to have massive tears, with tear size rising with duration of antihypertensive therapy - i.e. the indication itself, plus age, is a sufficient explanation. The only drug-specific human evidence is a single 1999 Journal of Human Hypertension case report of bilateral Achilles tendonitis that resolved on withdrawal and recurred on rechallenge, attributed to amlodipine's characteristic pre-capillary vasodilatory oedema rather than to any tendon-matrix mechanism; drug-induced tendinopathy reviews list fluoroquinolones, statins, glucocorticoids and aromatase inhibitors, and mention amlodipine only through that one case, without a proposed pathway. Where a direction of effect has actually been measured, it points the other way: in a rat Achilles injury model amlodipine improved healing, tensile strength, fibroblast activity and neovascularisation via eNOS/NO-driven perfusion and reduced ROS - if anything a protective vascular effect on a tendinopathy whose leading hypothesis is vascular insufficiency. The atlas numbers are consistent with a population artefact rather than an ADR: enrichment is exactly 1.0 (no excess over users of other drugs), the propensity-to-prescribe arm is null (log10p 0.12), the variant does nothing to the condition in the drug-free population (log10p 0.35, beta 0.152 +/- 0.199, well inside noise), and the 94.4% temporality is the uninformative direction given the cohort's record-depth asymmetry, especially with a median 3.58 years to a degenerative condition of ageing. What is left is a single within-users association with no mechanism and no background support.","gene_comment":"rs191380573 is a real 3'UTR variant of PTPN14 (chr1:214,349,697, GRCh38; PTPN14 spans 214,348,700-214,552,449 on the minus strand), so the gene call is annotationally correct rather than an intergenic guess - but PTPN14 is a Hippo-pathway YAP phosphatase known for lymphatic development and choanal-atresia/lymphoedema syndromes, with no tendon or rotator cuff biology, and it is absent from the known rotator cuff loci (GLCCI1, THSD7A, ZNF804A, TNC, SASH1/SAP30BP, COL23A1).","score_check":"beta_adr 4.418 on a variant with ~0.4% minor allele frequency implies an odds ratio near 80 for a common degenerative shoulder condition, which is the classic signature of a sparse-genotype fit rather than a real effect. The supporting arms all read null (log10p_disease 0.35, log10p_prescribed 0.12, enrichment 1.0), so z_diff 5.01 is driven entirely by that one implausible within-users estimate.","candidability":2,"candidability_reason":"No labelled or mechanistic link, the treated population's hypertension already explains the comorbidity, experimental tendon data point the opposite way, and the variant effect is an implausibly large rare-allele estimate.","sources":[{"title":"Risk factors for rotator cuff disease: A systematic review and meta-analysis of diabetes, hypertension, and hyperlipidemia (PubMed 35257948)","url":"https://pubmed.ncbi.nlm.nih.gov/35257948/"},{"title":"The association between arterial hypertension and rotator cuff tear: the influence on rotator cuff tear sizes (PubMed 22748932)","url":"https://pubmed.ncbi.nlm.nih.gov/22748932/"},{"title":"Cephalexin-associated Achilles Tendonitis: Case Report and Review of Drug-induced Tendinopathy (PMC6433089)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC6433089/"},{"title":"Achilles tendonitis: an unusual complication of amlodipine therapy - Journal of Human Hypertension","url":"https://www.nature.com/articles/1000872"},{"title":"The effect of amlodipine, a calcium channel blocker, on tendon healing: an experimental rat Achilles model (PMC12618398)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12618398/"},{"title":"Genome-Wide Association Study Identifies Genetic Variants Associated with Rotator Cuff Tear - A Pilot Study (PMC9601242)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9601242/"},{"title":"Ensembl REST: rs191380573 (3' UTR variant, MAF 0.0041) and PTPN14 gene coordinates","url":"https://rest.ensembl.org/variation/human/rs191380573?content-type=application/json"},{"title":"PTPN14 is required for the density-dependent control of YAP1 (PubMed 22948661)","url":"https://pubmed.ncbi.nlm.nih.gov/22948661/"}]},"amlodipine|N309":{"verdict":"plausible ADR","headline":"Cystitis is not on the amlodipine label, but CCB-induced impaired bladder emptying is a real pathway","assessment":"Cystitis is in no sense an indication for amlodipine: the US label covers hypertension, chronic stable angina, vasospastic angina and angiographically documented CAD only. Nor is cystitis or urinary tract infection a labelled adverse reaction; the Urinary System entry in section 6.1 reads exactly 'micturition frequency, micturition disorder, nocturia', all at under 1% in placebo-controlled trials, and a 2025 FAERS disproportionality study of amlodipine found 27 recurrent signals (gingival hypertrophy, hypotension, bradycardia, shock) with no urinary signal at all. A drug-caused route to cystitis is nonetheless mechanistically coherent and second-order rather than direct: L-type calcium channel blockade reduces Ca2+ influx into detrusor smooth muscle and impairs contractility, and a cross-sectional study of calcium-antagonist users found markedly worse IPSS scores (15.2 vs 9.3, p<0.001) with severe LUTS odds ratios of 7.3 in men and 10.5 in women on amlodipine/nifedipine, while a 2024 FAERS plus Mendelian-randomisation analysis in Urology listed amlodipine among 78 drugs with a urinary-retention signal. Incomplete emptying and raised post-void residual are the standard precursor of bacteriuria and cystitis, so the drug is not moving this condition in the opposite direction - it plausibly moves it the wrong way, just via retention rather than by any direct urothelial effect. The atlas numbers are consistent with an untargeted comorbidity rather than a treated-population artefact: enrichment is exactly 1.0, so cystitis is no commoner in amlodipine users than in users of other drugs, cotx_pct is only 0.88%, and log10p_prescribed of 0.17 shows the variant does not predict being put on amlodipine. Temporality of 83.8% over 185 dated participants with a median 3.9 years to onset points the right way but is the uninformative direction in this cohort, and the FAERS support is weak (1 of 3, PRR 1.97). The weight of the evidence is therefore a thin class-level plausibility with essentially no genetic content behind it.","gene_comment":"PTPN12 (PTP-PEST) is a genuine protein-coding phosphatase acting on p130Cas, PTK2B, paxillin and PSTPIP1, so this is not an lncRNA or intergenic assignment, but nothing in its literature touches detrusor function, urothelial defence, dihydropyridine handling or UTI susceptibility - its disease links are colorectal cancer and PAPA syndrome. As a single MPC burden mask it carries no variant-level resolution and offers no mechanism for the CCB-retention pathway proposed above.","score_check":"beta_adr of 83.34 log-odds implies an SE near 13 given log10p 10.56 - the signature of quasi-complete separation in a rare-variant burden test on a 0.88% outcome, not a real effect size, so the p-value should not be read at face value. The disease arm is only borderline nominal (beta 4.695, se 2.374, z=1.98, log10p 1.32), which means z_diff of 6.18 is driven almost entirely by the unstable treated-arm estimate.","candidability":4,"candidability_reason":"Real class-level bladder-emptying pathway to cystitis, but not a labelled reaction, no enrichment, weak FAERS, and a separation-artefact burden hit in a gene with no bladder relevance.","sources":[{"title":"DailyMed - AMLODIPINE BESYLATE tablet (full prescribing information, sections 1 and 6.1)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=6f20efca-943b-4bc2-a157-0931bd4d4585"},{"title":"NORVASC (amlodipine besylate) tablets - FDA approved label","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2011/019787s047lbl.pdf"},{"title":"Calcium Antagonists Use and Its Association with Lower Urinary Tract Symptoms: A Cross-Sectional Study","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3689686/"},{"title":"Drugs Associated with Urinary Retention Adverse Reactions: A Joint Analysis of FAERS and Mendelian Randomization (Urology, 2024)","url":"https://pubmed.ncbi.nlm.nih.gov/39222669/"},{"title":"Adverse events associated with amlodipine: a pharmacovigilance study using the FDA Adverse Event Reporting System","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12089059/"},{"title":"PTPN12 Gene - GeneCards","url":"https://www.genecards.org/cgi-bin/carddisp.pl?gene=PTPN12"}]},"amlodipine|N390":{"verdict":"statistical artifact","headline":"UTI is not an amlodipine label event; DPP7 burden OR ~540 is a sparse-data artifact","assessment":"Urinary tract infection is neither a licensed indication for amlodipine (hypertension, chronic stable and vasospastic angina, angiographically documented CAD) nor a labelled adverse reaction: the FDA/DailyMed adverse-reaction table lists only micturition frequency, micturition disorder and nocturia under Urinary System, at <1-2%, and no infection term appears anywhere in the label. A weak class-level mechanism is at least arguable, since L-type calcium channel blockade reduces detrusor contractility and CCBs appear on standard lists of drugs causing urinary retention, and chronic high-risk retention is associated with recurrent culture-proven UTI; but amlodipine is a vascular-selective dihydropyridine, this chain is indirect, and no study has shown CCB use raising UTI incidence. The atlas numbers argue against a drug effect rather than for one: enrichment is 1.04, i.e. UTI is no more common in amlodipine users (3.48%) than in users of other drugs, which is exactly what the geriatric-UTI epidemiology predicts for an elderly hypertensive population where over 10% of women aged 65+ have a UTI each year. Temporality of 87.5% (n=730, median 3.89 years after first exposure) is the uninformative direction given the cohort's known upward compression, and a 3.9-year median gap sits comfortably inside ordinary ageing rather than a drug reaction. The FAERS support is explicitly weak (1/3 sources, PRR 1.87), a level of disproportionality that UTI reaches routinely as an intercurrent event in elderly cardiovascular reporting. The decisive problem is the effect size: beta_adr 6.297 on a LOF burden test corresponds to an odds ratio near 540 for a condition affecting 3.5% of the exposed, which is not a biological effect but near-complete case-carrier separation on a handful of carriers, and the same burden in the drug-free population is flat (beta 0.032, se 0.237, log10p 0.05), so the z_diff of 5.06 inherits the artifact rather than testing a real interaction.","gene_comment":"DPP7/DPP2 is a genuine protein-coding gene with real immune relevance (lysosomal serine dipeptidyl peptidase maintaining resting-lymphocyte quiescence, annotated to neutrophil degranulation), so this is not a lncRNA or intergenic placeholder. But it appears in none of the UTI genetics literature - the 36-locus cross-biobank recurrent-UTI GWAS prioritises PSCA, UMOD, FUT2 and other urothelial-barrier genes and never mentions DPP7 - and no interaction between dipeptidyl peptidases and dihydropyridine CCBs is described.","score_check":"The within-user beta of 6.3 (OR ~540) with only log10p 6.65 is the signature of a burden test carried by a few carriers, not a measurable effect, and the parallel disease-arm estimate is indistinguishable from zero (|0.032| << 1.96 x 0.237). Enrichment of 1.04 and a FAERS PRR of 1.87 add nothing beyond background comorbidity.","candidability":2,"candidability_reason":"No label or literature support for amlodipine-caused UTI, no exposure enrichment, and an implausibly large LOF burden effect in a gene absent from all UTI genetics.","sources":[{"title":"Amlodipine besylate tablets - full prescribing information (DailyMed/FDA)","url":"https://dailymed.nlm.nih.gov/dailymed/fda/fdaDrugXsl.cfm?setid=213502a2-f667-4da6-a4b9-462eb303c401&type=display"},{"title":"Norvasc (amlodipine besylate) tablets label, FDA Drugs@FDA","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2011/019787s047lbl.pdf"},{"title":"Urinary Retention in Adults: Evaluation and Initial Management (AFP 2018)","url":"https://www.aafp.org/pubs/afp/issues/2018/1015/p496.html"},{"title":"Urinary tract infection in older adults (PMC3878051)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3878051/"},{"title":"Inherited Susceptibility to Urinary Tract Infections from Kidney Papilla to Bladder (cross-biobank rUTI GWAS, medRxiv)","url":"https://www.medrxiv.org/content/10.64898/2026.07.07.26357417v2.full"},{"title":"Dipeptidyl Peptidase 2 apoptosis assay determines the B-cell activation stage in CLL (PMC2991601)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC2991601/"}]},"amlodipine|N393":{"verdict":"co-prescription population","headline":"Stress incontinence is not an amlodipine effect; CCBs move the bladder toward retention, not sphincter leak","assessment":"Amlodipine is licensed only for hypertension and angina, so stress incontinence (N39.3) is neither an indication nor a labelled adverse reaction: the Norvasc urinary-system list contains micturition frequency, micturition disorder and nocturia at <1% (1-2% across multiple-dose studies), and the word incontinence does not appear. Where calcium-channel blockade is implicated in the lower urinary tract the direction is impaired detrusor contraction and urinary retention - a 2024 FAERS plus Mendelian-randomisation analysis named amlodipine specifically among 78 retention-associated drugs - which is the opposite pole from a sphincteric/pelvic-floor leak, and any incontinence arising that way would be overflow, not stress. The one clinical report that links amlodipine to stress incontinence at all runs backwards: in the 1998 Obstet Gynecol case, doxazosin (urethral tone) and then enalapril (cough-provoked leakage) caused the incontinence and switching to amlodipine resolved it. A dedicated cohort of 390 older incontinence-clinic patients found CCBs the single most-used class (21.8%) yet detected no association between any medication class and incontinence type or severity. The atlas figures fit confounding rather than causation: only 0.73% of amlodipine users carry the code, enrichment is null so there is no background comparison, and stress incontinence shares its whole risk profile - age, female sex, obesity, metabolic syndrome (SUI prevalence 38-48% in NHANES-type female samples) - with the population that gets an antihypertensive. Temporality of 97.1% on 102 dated participants and a median 3.8 years is exactly the pattern the cohort's prescription-before-diagnosis compression manufactures, and FAERS is weak (1/3, PRR 1.62); the most one can say for a drug contribution is that dose-dependent ankle oedema (10.8% at 10 mg vs 0.6% placebo) redistributing overnight can unmask nocturia, which is a different symptom from the coded one.","gene_comment":"rs147810021 is an intergenic X-chromosome variant (chrX:4,192,674, GRCh38) with no gene assignment in dbSNP and a global MAF of only ~0.3%, sitting in the Xp22.3 gene desert between PRKX and NLGN4X - there is no candidate to build a mechanism on. A rare X-linked marker tested for a near-exclusively female phenotype is also the situation most sensitive to sex coding and hemizygosity handling, so the locus needs technical exclusion before biological interpretation.","score_check":"beta_adr 5.676 at log10p 6.07 implies se ~1.15, i.e. an odds ratio near 300 from a 0.3%-frequency variant in the ~0.7% of users carrying the code - internally significant but the classic sparse-cell rare-variant inflation profile, not a credible effect size. The disease arm is consistent with near-noise (beta 0.399, se 0.20, |beta| ~2.0 se), so the z_diff of 4.51 is driven almost entirely by the unstable within-users estimate.","candidability":1,"candidability_reason":"Not a labelled or literature-supported ADR, the drug's known bladder effect points the other way (retention), and the signal rests on a rare unassigned X-linked variant with an implausible effect size in a comorbidity-matched population.","sources":[{"title":"DailyMed - NORVASC (amlodipine besylate) tablet, full prescribing information","url":"https://dailymed.nlm.nih.gov/dailymed/fda/fdaDrugXsl.cfm?setid=abd6a2ca-40c2-485c-bc53-db1c652505ed"},{"title":"Stress urinary incontinence due to prescription medications: alpha-blockers and angiotensin converting enzyme inhibitors (Obstet Gynecol 1998; PMID 9572189)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:9572189&format=json&resultType=core"},{"title":"Drugs Associated with Urinary Retention Adverse Reactions: A Joint Analysis of FAERS and Mendelian Randomization (Urology 2024; PMID 39222669)","url":"https://pubmed.ncbi.nlm.nih.gov/39222669/"},{"title":"Prevalence of commonly prescribed medications potentially contributing to urinary symptoms in a cohort of older patients seeking care for incontinence (PMC3684540)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3684540/"},{"title":"Synergistic interaction between hyperlipidemia and obesity as a risk factor for stress urinary incontinence (PMC10973469)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10973469/"},{"title":"dbSNP RefSNP rs147810021 (NCBI Variation Services)","url":"https://www.ncbi.nlm.nih.gov/snp/rs147810021"}]},"amlodipine|R060":{"verdict":"plausible ADR","headline":"Dyspnoea is a labelled common amlodipine effect, but this rare intronic GRM8 variant looks sparse-data driven","assessment":"Dyspnoea is not an indication for amlodipine (licensed for hypertension, chronic stable angina and vasospastic angina) and it is explicitly listed in section 4.8 of the UK SmPC as a Common (>=1/100 to <1/10) undesirable effect, alongside very common oedema and common ankle swelling. A drug-caused mechanism is described and coherent: dihydropyridines preferentially dilate precapillary arterioles, raising intracapillary hydrostatic pressure and driving fluid into the interstitium, which produces the classic peripheral oedema and, at the extreme, non-cardiogenic pulmonary oedema; in PRAISE-1, an RCT in severe heart failure, both peripheral oedema and pulmonary oedema were significantly more frequent on amlodipine than placebo (pulmonary oedema P=0.01) even though mortality was neutral. There is no evidence that amlodipine moves dyspnoea in the opposite direction, so an ADR reading is not excluded by pharmacology, though the severe respiratory presentations in the literature are overwhelmingly overdose cases, with only a single reported case at a therapeutic 5 mg dose. Against this, dyspnoea is one of the least specific EHR codes there is and amlodipine users are old, hypertensive and frequently have heart failure or COPD, so confounding by indication and comorbidity is the dominant competing explanation; the atlas offers partial reassurance here in that the code is not enriched in amlodipine users (enrichment 0.91, cotx 1.5%) and the variant is unrelated to being prescribed the drug (log10p_prescribed 0.5). The temporality of 96.9% over 353 dated participants, median 3.67 years after first exposure, is the expected high value in this cohort and is weak evidence on its own, and the 3.67-year median lag fits accumulating cardiorespiratory comorbidity at least as well as a drug effect. FAERS support is weak (PRR 1.92, 1/3), and a recent FAERS disproportionality study of amlodipine recovered dyspnoea only as a label-consistent, not a standout, signal.","gene_comment":"rs117700515 is an intronic variant in GRM8 (chr7:127,192,312, GRCh38) with a minor allele frequency of only ~0.5-0.7%; GRM8 encodes a brain-expressed group III metabotropic glutamate receptor with no established role in pulmonary vascular tone, capillary permeability or dyspnoea. The assignment is positional rather than mechanistic and should not be presented as an explanation of the phenotype.","score_check":"The within-user effect of beta 3.126 (OR ~23) for a variant carried by well under 1% of the cohort, in a phenotype set of a few hundred dated cases, is the signature of a sparse-data artifact rather than a real effect of that size, and the clean null in the drug-free arm (beta 0.033 +/- 0.125, log10p 0.1) is what an unstable rare-variant hit would also look like. The z_diff of 5.47 therefore rests almost entirely on the fragile treated-arm estimate and needs carrier counts and a Firth-corrected or exact re-test before it is believable.","candidability":5,"candidability_reason":"Real, labelled, mechanistically coherent amlodipine ADR, but the specific locus is a rare intronic variant in an implausible gene with an effect size typical of sparse-data bias.","sources":[{"title":"Amlodipine 5mg Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/3478/smpc"},{"title":"Effect of Amlodipine on Morbidity and Mortality in Severe Chronic Heart Failure (PRAISE-1), NEJM 1996","url":"https://pubmed.ncbi.nlm.nih.gov/8813041/"},{"title":"Prospective Randomized Amlodipine Survival Evaluation-1 - trial summary, American College of Cardiology","url":"https://www.acc.org/latest-in-cardiology/clinical-trials/2010/02/23/19/13/praise1"},{"title":"Severe non-cardiogenic pulmonary edema following low-dose amlodipine ingestion: a case report (PMC)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12577906/"},{"title":"Adverse events associated with amlodipine: a pharmacovigilance study using the FDA adverse event reporting system","url":"https://www.frontiersin.org/journals/cardiovascular-medicine/articles/10.3389/fcvm.2025.1504671/full"},{"title":"Ensembl REST variation record for rs117700515 (GRM8 intron, MAF ~0.005)","url":"https://rest.ensembl.org/variation/human/rs117700515?content-type=application/json;pops=1"}]},"amlodipine|R103":{"verdict":"plausible ADR","headline":"Abdominal pain is a labelled amlodipine effect, but this rare lncRNA variant is not credible mechanism","assessment":"R103 (pain localised to other parts of lower abdomen) is not an indication for amlodipine and is not a condition amlodipine is given for; the atlas agrees, with cotx_pct 1.2% and enrichment 0.99, i.e. no excess over users of other drugs, and log10p_prescribed 0.14, so the variant does not predict who gets the drug. Abdominal pain is, however, an explicitly labelled adverse reaction: the NORVASC US label reports abdominal pain in 1.6% of amlodipine-treated patients versus 0.3% on placebo, alongside nausea 2.9% vs 1.9% and constipation, dyspepsia, flatulence and (rarely) pancreatitis at <1%. A mechanism is describable rather than merely assertable: L-type calcium channel blockade relaxes gut smooth muscle and slows transit, and constipation-mediated cramping is the usual route from a dihydropyridine to abdominal pain; amlodipine overdose can present as frank ileus/intestinal obstruction. The same mechanism cuts the other way for colicky visceral pain, since calcium channel inhibition is used deliberately as an antispasmodic (pinaverium, otilonium; topical nifedipine/diltiazem for anal fissure), so a drug-caused increase in lower-abdominal pain is expected via constipation, not via spasm. The ICD code itself is a nonspecific symptom label of the kind that accumulates in an older, polypharmacy, mostly hypertensive population, and R103 in particular carries a large urological/gynaecological load unrelated to the drug. FAERS support is weak (1/3, PRR 1.23), and temporality of 94.7% over 284 dated participants is the uninformative direction given the cohort's compression, so it adds nothing. The drug-condition pair is therefore real at class level, but nothing here is specific to this genotype.","gene_comment":"rs141241625 is a rare (MAF ~0.3%) intronic SNV in an unnamed novel lncRNA at chr10:104.62 Mb (GRCh38), just upstream of SORCS3, a neuronal sortilin receptor known from psychiatric/neuroticism GWAS and with no plausible connection to gut motility or visceral nociception; the gene assignment is an annotation of convenience, not a mechanism.","score_check":"beta_adr 3.267 (OR ~26) for a 0.3%-frequency allele against a 1.2% outcome is a sparse-cell effect size, not a believable biological one, even at log10p 7.23; by contrast beta_disease 0.283 +/- 0.106 is only 2.7 SE from zero and is the more honest estimate, so the z_diff of 4.88 is driven almost entirely by the inflated treated-arm beta rather than by real effect modification.","candidability":4,"candidability_reason":"Abdominal pain is a genuine labelled amlodipine effect with a motility mechanism, but this specific rare lncRNA variant has an implausible effect size and no route to visceral pain.","sources":[{"title":"DailyMed - NORVASC (amlodipine besylate) tablet label, section 6 Adverse Reactions","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=abd6a2ca-40c2-485c-bc53-db1c652505ed"},{"title":"Ensembl REST - gene lookup ENSG00000237761 (lncRNA, chr10:104,566,931-104,624,795, GRCh38)","url":"https://rest.ensembl.org/lookup/id/ENSG00000237761?content-type=application/json"},{"title":"Ensembl REST - variation rs141241625 (A/G, MAF 0.0027, intron variant, chr10:104,617,964)","url":"https://rest.ensembl.org/variation/human/rs141241625?content-type=application/json"},{"title":"Ensembl REST - genes overlapping chr10:104,400,000-104,800,000 (SORCS3, SORCS3-AS1, LINC02620, CFAP58)","url":"https://rest.ensembl.org/overlap/region/human/10:104400000-104800000?feature=gene;content-type=application/json"},{"title":"Antispasmodics for Chronic Abdominal Pain: Analysis of North American Treatment Options (PMC8315189)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8315189/"},{"title":"Acute severe intestinal obstruction secondary to amlodipine toxicity, QJM 105:467","url":"https://academic.oup.com/qjmed/article/105/5/467/1564727"}]},"amlodipine|R13":{"verdict":"plausible ADR","headline":"Dysphagia is a labelled amlodipine adverse reaction with a real class mechanism, but this variant is not it","assessment":"Dysphagia is not an indication for amlodipine, which is licensed only for hypertension, chronic stable angina, vasospastic angina and angiographically documented CAD; the US Norvasc label lists dysphagia explicitly among gastrointestinal adverse reactions occurring in <1% but >0.1% of patients, in the class where causality is described as uncertain. A mechanism is described at class level: L-type calcium channel blockade in oesophageal smooth muscle reduces the amplitude and duration of peristalsis and lowers LES tone, and CCBs are recommended with caution in patients with fragile oesophageal function such as systemic sclerosis, where nifedipine has precipitated aspiration. The direction is not one-way, however - the same LES-relaxing effect is used therapeutically in achalasia, where nifedipine and verapamil lower LES pressure and relieve dysphagia, and a recent Mendelian randomisation study reports genetically proxied amlodipine as protective against GERD, so amlodipine (a vascular-selective dihydropyridine with the weakest gut smooth-muscle effect of the class) is a much weaker candidate than nifedipine or verapamil. The atlas epidemiology is neutral rather than supportive: dysphagia is present in 0.85% of amlodipine users with an enrichment of 0.96, i.e. no excess over users of other drugs, and dysphagia is strongly age- and polypharmacy-driven in exactly this population. The prescription-propensity arm is null (log10p 0.60) and the disease arm is noise (beta 0.269 +/- 0.142, |beta| < 1.96*se), so the entire z_diff of 4.99 rests on the within-users arm alone; temporality of 100% is uninformative given the atlas's own upward compression, though the 3.6-year median lag is at least not incompatible with chronic exposure. External pharmacovigilance is weak (FAERS 1/3, PRR 1.71) and a dedicated FAERS dysphagia analysis surfaced levodopa, not calcium channel blockers.","gene_comment":"rs183450691 is a rare variant (gnomAD exomes minor-allele frequency ~0.5%, global MAF ~0.2%) at chr5:115,581,344 in TICAM2/TRAM, the TIR-domain adaptor that bridges TLR4 to TRIF for type-I interferon induction, and it is intronic in the canonical transcript (missense p.Ile314Val only in some readthrough TMED7-TICAM2 isoforms). Innate-immune TLR4 signalling has no described connection to oesophageal smooth-muscle contraction, calcium handling or swallowing, so this gene assignment supplies no mechanism for the labelled class effect.","score_check":"beta_adr of 3.43 is an odds ratio near 31 for a common symptom code carried by a ~0.2-0.5% frequency variant, which is the classic signature of sparse-cell inflation rather than a real effect of that size, and the disease arm on the same variant is indistinguishable from zero. The p-value (log10p 7.56) is therefore not trustworthy as an effect estimate even if the drug-condition pairing is clinically real.","candidability":5,"candidability_reason":"Genuine labelled ADR with a described class mechanism, but no enrichment signal, ambiguous direction for amlodipine specifically, and a rare irrelevant-gene variant with an implausibly large effect.","sources":[{"title":"NORVASC (amlodipine besylate) tablets - full prescribing information, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/fda/fdaDrugXsl.cfm?setid=abd6a2ca-40c2-485c-bc53-db1c652505ed"},{"title":"Calcium Channel Blockers and Esophageal Sclerosis: Should We Expect Exacerbation of Interstitial Lung Disease? (PMC3304075)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3304075/"},{"title":"Genetic evidence for amlodipine's protective role in gastroesophageal reflux disease: a focus on CACNB2, PLOS ONE","url":"https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0309805"},{"title":"On the potential of drug repurposing in dysphagia treatment: new insights from a real-world pharmacovigilance study and a systematic review (PMC10022593)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10022593/"},{"title":"Ensembl REST variation record for rs183450691 (TICAM2, chr5:115,581,344)","url":"https://rest.ensembl.org/variation/human/rs183450691?content-type=application/json;pops=1"}]},"amlodipine|R410":{"verdict":"plausible ADR","headline":"Confusion is a labelled rare amlodipine ADR, but a 5-year lag and a rare lncRNA variant carry it here","assessment":"Amlodipine is licensed only for hypertension, chronic stable angina and vasospastic angina, so R41.0 disorientation is not an indication; the EU SmPC does list 'confusion' as a Rare psychiatric undesirable effect, while the US Norvasc label omits it and records only dose-related dizziness and somnolence plus symptomatic hypotension. A mechanism is easy to state and hard to prove: dihydropyridine vasodilation causing orthostatic or postural hypotension and cerebral hypoperfusion, which in older polypharmacy patients presents as confusion and difficulty concentrating; that mechanism predicts an event within days to weeks of initiation or dose increase, not the median 5.08 years seen here. At class level the cognitive literature runs the other way or is neutral - meta-analytic and cohort data associate calcium channel blockers with equal or lower dementia risk, and amlodipine specifically 'did not seem to be protective' because of poor brain availability, i.e. no described class harm to cognition. The atlas internals argue against confounding by indication (enrichment 0.81, disorientation is if anything slightly less common in amlodipine users than in users of other drugs; prescription-propensity arm null at log10p 0.05) but equally against an acute drug effect, since only 0.57% of users carry the code and the timing fits incident age-related cognitive decline in a treated hypertensive cohort. FAERS disproportionality is 'STRONG' only at PRR 2.04, and confusional state is among the most reported events for any drug taken by elderly patients, so it adds little independent weight; a dedicated 2025 FAERS analysis of amlodipine surfaced gingival, fluid and cardiorespiratory signals, not confusion or delirium. The honest reading is a genuine but rare labelled reaction whose atlas instance is dominated by age and comorbidity and whose genetic anchor does not support it.","gene_comment":"rs142170134 is a rare intronic variant (global MAF ~0.0008, gnomAD ~0.004) at chr11:44,992,783 inside ENSG00000294396, an unnamed Havana 'novel transcript' lncRNA at 11p11.2 abutting TP53I11 and PRDM11 - no neuronal, calcium-channel or CYP3A4/5 connection, so the gene assignment is decorative rather than mechanistic. A real amlodipine pharmacogenomic signal for a hypoperfusion ADR would be expected at CYP3A4/5 or a vascular calcium-channel locus, not here.","score_check":"beta_adr 5.116 (odds ratio ~165) from a variant with MAF under 1% against ~128 dated cases is the signature of sparse-data separation, not a real effect size, and log10p 6.31 is not genome-wide significant for an atlas-wide scan. The disease arm is flat noise (beta 0.127 +/- 0.203, |beta| < 1.96*se) so z_diff 4.81 is inherited wholesale from the inflated ADR estimate.","candidability":4,"candidability_reason":"Labelled rare ADR, but the 5-year lag, depleted enrichment and an implausibly large effect at a rare intronic lncRNA variant point to age and comorbidity rather than a pharmacogenomic effect.","sources":[{"title":"Amlodipine 5mg tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/4717/smpc"},{"title":"DailyMed label: NORVASC (amlodipine besylate) tablet","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=abd6a2ca-40c2-485c-bc53-db1c652505ed"},{"title":"Adverse events associated with amlodipine: a pharmacovigilance study using the FDA adverse event reporting system","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12089059/"},{"title":"Nimmrich & Eckert, Calcium channel blockers and dementia (Br J Pharmacol)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3831702/"},{"title":"Ensembl REST lookup: ENSG00000294396 (novel lncRNA transcript, chr11:44,978,512-45,056,055)","url":"https://rest.ensembl.org/lookup/id/ENSG00000294396?content-type=application/json"},{"title":"Ensembl REST variation: rs142170134 (intron variant, chr11:44,992,783, MAF 0.0008)","url":"https://rest.ensembl.org/variation/human/rs142170134?content-type=application/json;pops=1"}]},"amlodipine|R529":{"verdict":"statistical artifact","headline":"Amlodipine can cause nonspecific pain, but this signal is two rare intronic SNVs with an impossible OR","assessment":"R52.9 (pain, unspecified) is not an amlodipine indication - the licensed uses are hypertension, chronic stable and vasospastic angina, and documented CAD - but a pain relationship does exist at label level: the UK SmPC lists 'pain' itself as an uncommon general disorder, muscle cramps, abdominal pain and headache as common, and myalgia, arthralgia and back pain as uncommon, while DailyMed records dose-related peripheral oedema (10.8% at 10 mg vs 0.6% placebo) that plausibly presents as limb heaviness and discomfort. A drug-caused mechanism is therefore plausible and is described: dihydropyridine arteriolar vasodilation produces oedema, vasodilatory headache and cramps, and there is no credible opposite-direction argument, since a 2025 review of calcium channel blockers in pain concludes that broad-spectrum L-type dihydropyridines have not shown analgesic efficacy and the field has moved to alpha-2-delta and N/T-type agents. The population arm is reassuring rather than incriminating: only 0.63% of amlodipine users carry the code and enrichment is 0.97, so these patients are no more likely to carry unspecified pain than users of other drugs, and the prescription-propensity arm is null (log10p 0.08). The problem is entirely the genetics: beta_adr 6.79 implies an odds ratio near 900 for a common, nonspecific symptom code, which is not a real effect size but quasi-separation of two ~0.3%-frequency alleles against 88 dated events. The drug-free arm is pure noise (beta 0.081 +/- 0.25, log10p 0.13), so the z_diff of 5.24 measures only the size of the artefact, and FAERS gives no disproportionality (PRR 1.2, weak 1/3). Temporality of 94.3% at a median 3.67 years is exactly what cohort compression produces for a symptom code accumulated over years of follow-up and adds nothing.","gene_comment":"rs143340101 (chr14:33,568,506, gnomAD MAF 0.30%) is intronic in NPAS3, a neurodevelopmental bHLH-PAS transcription factor tied to schizophrenia and mood disorder risk with no link to nociception or dihydropyridine pharmacology; rs115731173 (chr5:3,525,655, MAF 0.28%) is intronic in LINC01019, an uncharacterised lncRNA in a gene-poor 5p15 region, which is effectively a positional label rather than a gene assignment.","score_check":"The numbers are internally consistent but not believable as biology: an OR of roughly 900 from two rare variants over 88 events is the signature of sparse-data separation, and the null disease and prescription arms mean nothing independent corroborates it.","candidability":2,"candidability_reason":"The underlying ADR is real but trivial and nonspecific; the genetic signal is a rare-allele separation artefact on a garbage ICD code with unusable genes.","sources":[{"title":"DailyMed - AMLODIPINE BESYLATE tablet (US label: indications, adverse reactions, peripheral oedema by dose)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=6f20efca-943b-4bc2-a157-0931bd4d4585"},{"title":"Amlodipine 10 mg tablets - Summary of Product Characteristics (emc), section 4.8","url":"https://www.medicines.org.uk/emc/product/4718/smpc"},{"title":"Therapeutic Potential of Calcium Channel Blockers in Neuropsychiatric, Endocrine and Pain Disorders (PMC12293676)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12293676/"},{"title":"Ensembl variation rs143340101 (NPAS3 intron, gnomAD MAF 0.30%)","url":"https://rest.ensembl.org/variation/human/rs143340101?content-type=application/json;pops=1"},{"title":"Ensembl variation rs115731173 (chr5:3,525,655 intron, gnomAD MAF 0.28%)","url":"https://rest.ensembl.org/variation/human/rs115731173?content-type=application/json;pops=1"},{"title":"ICD-10-CM Code R52 - Pain, unspecified (AAPC Codify)","url":"https://www.aapc.com/codes/icd-10-codes/R52"}]},"amlodipine|R80":{"verdict":"plausible ADR","headline":"Amlodipine plausibly raises albuminuria via afferent dilation, but the rare lncRNA variant is uninterpretable","assessment":"Isolated proteinuria is not a licensed indication for amlodipine, whose FDA label covers only hypertension and chronic stable/vasospastic angina, and proteinuria appears nowhere in its adverse-reactions table (the label instead states that in hypertensive patients with normal renal function amlodipine lowered renal vascular resistance and raised GFR without change in filtration fraction or proteinuria). The class-level nephrology literature nonetheless supports a drug-caused increase: dihydropyridine CCBs block L-type channels that are expressed mainly on the afferent arteriole, so they dilate the afferent without the efferent, abolish autoregulation and transmit systemic pressure into the glomerulus, producing glomerular hypertension, hyperfiltration and albuminuria. The amlodipine arm of AASK was stopped early in 2000 for increased proteinuria relative to ramipril and metoprolol; a 2024 JGIM cohort found DCCBs associated with severe albuminuria versus thiazides (HR 1.29) and with kidney failure (HR 1.66) only in patients not co-treated with an ACEi/ARB; a paediatric CKD study found 18.8% higher urine protein/creatinine in dhCCB users without RAAS blockade, with the excess abolished by concomitant ACEi/ARB. So the direction of the atlas signal agrees with the literature rather than opposing it, which is the key point in its favour. Against it, amlodipine users are by definition hypertensive and frequently have hypertensive or diabetic CKD, so proteinuria is expected in this population independent of the drug, and the atlas enrichment of 1.03 says the condition is no more common in amlodipine users than in users of other drugs. Temporality reads 100% but rests on only 25 datable participants out of a set that is 79.5% undated, and a dedicated FAERS proteinuria disproportionality analysis covering 2004-2024 returned 21 signal drugs dominated by oncology, immunosuppressant and antiviral agents with no calcium channel blocker among them, which tempers the fact sheet's PRR of 3.25.","gene_comment":"rs190050073 is an intronic variant in LINC02545, an uncharacterised long intergenic non-coding RNA at chr11:13.85 Mb with MAF around 0.3%; there is no published function for the transcript and nothing linking it to glomerular biology, calcium channels or amlodipine handling, so it carries no mechanistic weight. A beta of 5.82 (odds ratio in the hundreds) for a variant this rare is the signature of a sparse-data estimate, not of a large true effect.","score_check":"The within-user p-value is real but the effect size is not usable: no standard error is reported, the drug-free disease arm is missing entirely (log10p_disease and beta_disease NaN, z_diff null), so the single most discriminating test - does this variant simply predispose to proteinuria without the drug - was never run. The prescription-propensity arm is flat (log10p 0.33), which at least argues the variant does not merely mark who gets amlodipine.","candidability":5,"candidability_reason":"The drug-condition direction is genuinely supported by AASK and cohort data, but the specific rare lncRNA variant is uninterpretable and the drug-free comparison is missing.","sources":[{"title":"Dihydropyridine Calcium Channel Blockers and Kidney Outcomes (J Gen Intern Med 2024, PMC11282043)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11282043/"},{"title":"Amlodipine besylate tablets, FDA label NDA 19-787/S-042","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2007/019787s042lbl.pdf"},{"title":"L-type calcium channel blocker use and proteinuria among children with chronic kidney diseases (PMC8985842)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8985842/"},{"title":"Drug-related proteinuria: a vigilance analysis based on the FAERS database (PMC12990312)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12990312/"},{"title":"Ensembl REST variation record for rs190050073","url":"https://rest.ensembl.org/variation/human/rs190050073?content-type=application/json"},{"title":"The African American Study of Kidney Disease and Hypertension (AASK): new findings","url":"https://pubmed.ncbi.nlm.nih.gov/11498655/"}]},"amlodipine|T810":{"verdict":"statistical artifact","headline":"CCB-bleeding link is real but weak and contested; this PARP6 burden signal is a separation artifact","assessment":"T81.0 (haemorrhage/haematoma complicating a procedure) is not an amlodipine indication - the SmPC licenses it only for hypertension, chronic stable angina and vasospastic angina - and it is not a labelled adverse reaction; the closest label entries are purpura (uncommon) and thrombocytopenia/leucocytopenia (very rare). A class-level bleeding mechanism is at least arguable, since dihydropyridines inhibit calcium-dependent platelet aggregation in vitro, Zuccala's BMJ 1997 cohort found a 2.05-fold adjusted perioperative transfusion risk in hip-fracture patients on calcium antagonists, and a 2015 meta-analysis of 17 studies gave a summary RR of 1.17 (1.01-1.36) for gastrointestinal bleeding, driven by lower-GI bleeds. Against that, the randomised ALLHAT data are flatly null for this drug: amlodipine HR 1.09 (0.92-1.28) versus chlorthalidone for hospitalised GI bleeding, with the lowest absolute incidence of the three arms, and the authors judged the causal evidence weak and confounded. The atlas is consistent with that null: enrichment is 1.02, i.e. procedural haemorrhage is no more common in amlodipine users (1.57%) than in users of other drugs, so there is neither confounding by indication nor an excess to explain. The temporality of 99.5% is uninformative here because T81.0 is an incident procedural event that can only be coded after a hospital contact, and the cohort compresses this figure upward anyway. What is left is a single-variant MPC burden test in PARP6 with beta 37.93, an effect size that corresponds to an odds ratio of order 10^16 and cannot be a real parameter estimate at any plausible standard error - it is the signature of quasi-complete separation with a handful of carriers, and it is what generates both the log10p of 7.56 and the z_diff of 5.31.","gene_comment":"PARP6 is a neuronal mono-ADP-ribosyltransferase whose knockout is perinatally lethal in mice and whose human loss-of-function causes developmental delay, epilepsy and microcephaly; it is expressed almost exclusively in neurons and has no described role in platelets, coagulation or vascular integrity, nor any link to amlodipine metabolism (which is CYP3A4). Nothing about this gene assignment supports a haemostatic mechanism, and it should not be presented as one.","score_check":"The numbers do not hang together: beta_adr of 37.93 is not a credible effect size next to a disease-arm estimate of 1.017 +/- 1.324, and the disease arm itself is indistinguishable from noise (|beta| < 1.96*se). Enrichment of 1.02, log10p_prescribed of 0.12 and a FAERS PRR of 1.31 all say there is no population-level or pharmacovigilance excess for the atlas signal to be detecting.","candidability":2,"candidability_reason":"Directionally arguable at class level but null in randomised data, no enrichment, and the genetic signal is an implausibly large burden estimate in a neuron-specific gene with no haemostasis biology.","sources":[{"title":"Amlodipine 5mg Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/3478/smpc"},{"title":"Risk of Hospitalized Gastrointestinal Bleeding in Persons Randomized to Diuretic, ACE-Inhibitor, or Calcium-Channel Blocker in ALLHAT","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3844932/"},{"title":"Use of calcium antagonists and need for perioperative transfusion in older patients with hip fracture: observational study (BMJ 1997)","url":"https://pubmed.ncbi.nlm.nih.gov/9066477/"},{"title":"Systematic review with meta-analysis: the association between the use of calcium channel blockers and gastrointestinal bleeding","url":"https://onlinelibrary.wiley.com/doi/abs/10.1111/apt.13211"},{"title":"Characterization of PARP6 Function in Knockout Mice and Patients with Developmental Delay","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8224619/"}]},"atenolol|H609":{"verdict":"statistical artifact","headline":"Otitis externa is not an atenolol indication or labelled ADR; the ANOS1 hit is a sparse-data artifact","assessment":"Atenolol is licensed only for hypertension, angina pectoris and haemodynamically stable acute myocardial infarction, and neither the FDA Tenormin label nor its postmarketing list contains any ear disorder, otitis or infection term; the only dermatological entries are rash, dry eyes and psoriasiform rash or exacerbation of psoriasis. Otitis externa is therefore neither an indication nor a recognised adverse reaction, and the modest enrichment here (cotx 0.84%, enrichment 1.24) is the size expected from a common ear-canal infection accumulating in an older, comorbid cardiovascular population rather than from a drug effect. The only mechanistic threads are indirect: beta-blockers, atenolol in particular, raise new-onset diabetes risk by roughly 22-44% in meta-analysis, and diabetes with advancing age is the dominant risk factor for otitis externa and especially necrotising otitis externa; separately, beta-blocker-induced psoriasiform or eczematous eruption can involve the external canal. Neither pathway is described as an atenolol-otitis externa causal link anywhere in the literature I could find, and both would predict a diffuse population-level excess, not a single-variant pharmacogenomic effect. The temporality of 78.8% after exposure with a median of 7.01 years rests on only 80 datable participants and, per the atlas's own caveat, a high value is weak evidence; a 7-year lag fits incident age-related infection rather than a drug reaction. The FAERS 'STRONG' label with PRR 4.8 is unpersuasive for this pair, since atenolol is a background comedication in enormous numbers of reports and otitis externa is a frequent intercurrent event, and I found no published disproportionality analysis linking beta-blockers to otitis externa.","gene_comment":"rs73199009 sits at chrX:8,578,630 (GRCh38), intronic within ANOS1, whose product anosmin-1 guides GnRH and olfactory neuron migration and causes Kallmann syndrome when disrupted - a developmental gene with no known role in skin barrier, cerumen, or ear-canal infection, and the variant has no GWAS Catalog associations. An intronic X-chromosomal variant assessed in a sex-imbalanced treated cohort is also exactly the setting where X-handling and sex-composition artifacts arise.","score_check":"beta_adr of 5.763 corresponds to an odds ratio of roughly 300 for a common infection, which is the classic signature of sparse-data separation rather than a real effect, and log10p 7.53 is barely genome-wide even before the drug x condition multiplicity of an atlas. The comparison arms are internally consistent - beta_disease 0.123 +/- 0.189 is flatly null and prescription propensity is null - so the numbers rule out predisposition but do not make the within-users effect believable.","candidability":2,"candidability_reason":"No label or literature support for the outcome, only an indirect diabetes/age pathway shared with confounding, an implausibly large effect on ~80 dated cases, and a developmental X-linked gene with no relevance to the ear canal.","sources":[{"title":"TENORMIN (atenolol) Tablets - FDA label, 2023","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2023/018240s047lbl.pdf"},{"title":"Atenolol (Tenormin) - indications and postmarketing adverse reactions, RxList","url":"https://www.rxlist.com/atenolol-drug.htm"},{"title":"Mechanisms of Beta-Blocker Induced Psoriasis, and Psoriasis De Novo at the Cellular Level (PMC7398737)","url":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7398737/"},{"title":"A meta-analysis of 94,492 patients with hypertension treated with beta blockers to determine the risk of new-onset diabetes mellitus","url":"https://pubmed.ncbi.nlm.nih.gov/17920367/"},{"title":"The Association Between Malignant Otitis Externa and Diabetes Mellitus: A Systematic Review (PMC10645783)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10645783/"},{"title":"OMIM *300836 - ANOSMIN 1; ANOS1","url":"https://omim.org/entry/300836"},{"title":"Ensembl REST variation record for rs73199009 (chrX:8,578,630, intron variant)","url":"https://rest.ensembl.org/variation/human/rs73199009?content-type=application/json"}]},"atenolol|H931":{"verdict":"statistical artifact","headline":"Tinnitus is not on the atenolol label; the signal rests on one rare intergenic X-chromosome variant","assessment":"Atenolol is licensed only for hypertension, angina, arrhythmias and post-myocardial-infarction care, so tinnitus (H93.1) is neither an indication nor a treatment target. Neither the US TENORMIN label nor the UK SmPC lists tinnitus, hearing loss or any ear-and-labyrinth disorder; the closest labelled neurosensory events are dizziness (13% vs 6% placebo on total reports) and vertigo (2% vs 0.5%), which are central/haemodynamic rather than cochlear. Population data give a modest hypertension-tinnitus link (meta-analytic OR 1.37, 19 studies), but the Fasa adult cohort found the risk tracked blood-pressure grade and control, with no significant association for beta-blockers specifically despite their being 60.8% of treated participants; a 2025 FAERS disproportionality analysis found the propranolol tinnitus signal weakened once the migraine indication was removed, and the authors attributed much of it to the underlying condition. The atlas numbers match that null: cotx_pct is only 0.76% with enrichment 0.98, i.e. atenolol users are no more likely to carry a tinnitus code than users of other drugs, and FAERS support here is weak (1 of 3, PRR 1.94). Temporality of 91.7% over 72 dated participants with a 7.4-year median lag is uninformative given the cohort's known upward compression, and a seven-year median onset is not a drug-reaction time course. What remains is a single-variant hit, and its effect size is the problem rather than the evidence.","gene_comment":"rs73236614 has no gene assignment for good reason: Ensembl places it at chrX:140,839,067 (GRCh38) as an intergenic variant in the Xq27.1 gene desert, with the nearest protein-coding genes SOX3 and LDOC1 roughly 270-340 kb away and only lncRNAs (LINC00632/CDR1 region) nearer. It is also rare (T allele ~0.6% in non-Finnish Europeans, 0.37% globally) and X-linked, so there is no auditory, adrenergic or ototoxic mechanism to attach to it.","score_check":"A beta of 5.533 for a ~0.6% X-chromosome allele in a stratum with only 72 dated cases is the signature of near-separation sparse-data bias, not a real effect of that magnitude, and log10p_adr 6.2 does not survive that reading. The comparison arms are consistent with nothing being there: the disease arm is a flat null (beta 0.013 +/- 0.16, log10p 0.03), prescription propensity is null (0.58), and the z_diff of 4.92 simply reflects the inflated treated-arm estimate.","candidability":1,"candidability_reason":"No label or cohort support for a beta-blocker-caused tinnitus, no enrichment in atenolol users, and the entire signal is one rare intergenic X variant with a sparse-data-sized effect.","sources":[{"title":"TENORMIN (atenolol) tablet - DailyMed label","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=6e850b1e-28a4-4ff8-8ed7-bb0b29faa28a"},{"title":"Atenolol 50 mg Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/14294/smpc"},{"title":"Association between hypertension status and severity and tinnitus: Fasa adult cohort study","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13250232/"},{"title":"A Systematic Review and Meta-Analysis on the Association between Hypertension and Tinnitus","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC4735998/"},{"title":"Evaluating the association between migraine treatments and tinnitus: insights from FAERS (PLOS One 2025)","url":"https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0330493"},{"title":"Ensembl/dbSNP variation record for rs73236614","url":"https://rest.ensembl.org/variation/human/rs73236614?content-type=application/json"}]},"atenolol|I251":{"verdict":"licensed indication","headline":"9p21/rs1333040 CAD locus resurfacing in atenolol users; atherosclerotic heart disease is the indication","assessment":"I25.1 (atherosclerotic heart disease) is not an adverse effect of atenolol but the disease atenolol is prescribed for: the UK SmPC lists management of angina pectoris and of myocardial infarction (acute-phase and post-recovery prophylaxis) among its therapeutic indications, and the 2024 ESC chronic coronary syndrome guidelines keep beta-blockers among the first agents used for symptom and heart-rate control in exactly these patients. The atlas numbers say the same thing: 10.3% of atenolol users carry the code, an enrichment of 2.09 over users of other drugs, which is confounding by indication in its purest form. The 99% temporality with a median 1.05 years is the uninformative direction of that metric, and it is also what you expect when a beta-blocker is started for hypertension or an early anginal presentation and the chronic-CAD code is entered at the first ischaemic event or first cardiology visit thereafter. A drug-caused mechanism in the harmful direction is not merely undescribed but contradicted: beta-blockade reduces myocardial oxygen demand and post-infarct mortality, and BCAPS showed a beta-blocker (metoprolol CR/XL) slowing carotid intima-media thickness progression and increasing plaque echogenicity, with reviews arguing for rather than against an anti-atherosclerotic class effect. The FAERS PRR of 6.31 is the standard signature of indication-driven reporting, since the underlying coronary disease is recorded as an event term in reports about patients taking a cardiac drug. Nothing here separates a drug effect from the population being treated.","gene_comment":"9-22083405-T-C is rs1333040, an intronic variant in CDKN2B-AS1/ANRIL at 9p21.3, the most replicated common coronary-artery-disease locus, with the T allele reported as the risk allele, consistent with the protective beta of -0.167 for C here. The gene assignment is therefore strong but it is a disease-susceptibility locus, not a drug-response locus, and it explains the within-user hit rather than supporting an ADR.","score_check":"The numbers are internally consistent and believable: beta_disease -0.167 +/- 0.016 is a solid ~10-sigma effect matching the known 9p21 CAD signal, log10p_disease 24.74 dwarfs the within-user 6.51, z_diff 1.75 is not significant, and log10p_prescribed 0.56 is null. They hang together as one disease locus seen twice, with no evidence of drug modification.","candidability":1,"candidability_reason":"Atherosclerotic heart disease is the licensed indication for atenolol and the variant is the canonical 9p21 CAD susceptibility locus, with no drug-specific effect (z_diff 1.75) and literature pointing the opposite way.","sources":[{"title":"Atenolol 50 mg film-coated tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/14162/smpc"},{"title":"2024 ESC Guidelines for the management of chronic coronary syndromes: Key Points (ACC)","url":"https://www.acc.org/Latest-in-Cardiology/ten-points-to-remember/2024/09/01/15/01/2024-esc-guidelines-for-ccs-esc-2024"},{"title":"Low-Dose Metoprolol CR/XL and Fluvastatin Slow Progression of Carotid Intima-Media Thickness (BCAPS), Circulation","url":"https://www.ahajournals.org/doi/10.1161/01.CIR.103.13.1721"},{"title":"The association between the chromosome 9p21 CDKN2B-AS1 gene variants and lipid metabolism: a pre-diagnostic biomarker for coronary artery disease (PMC)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC6382903/"},{"title":"Association between the SNP rs1333040 in region 9p21 and risk of coronary heart disease in a Chinese population (PubMed)","url":"https://pubmed.ncbi.nlm.nih.gov/39145594/"},{"title":"Ensembl REST: variants overlapping chr9:22083405 (GRCh38) - rs1333040","url":"https://rest.ensembl.org/overlap/region/human/9:22083405-22083405?feature=variation;content-type=application/json"}]},"atenolol|I500":{"verdict":"statistical artifact","headline":"HF is a labelled atenolol contraindication, not an indication - but this rare intergenic hit is noise","assessment":"Congestive heart failure is not a licensed indication for atenolol: the US label lists hypertension, angina pectoris and acute myocardial infarction only, and states plainly that atenolol 'is contraindicated in sinus bradycardia, heart block greater than first degree, cardiogenic shock, and overt cardiac failure'. A drug-caused mechanism is explicitly described on that same label - beta blockade 'carries the potential hazard of further depressing myocardial contractility and precipitating more severe failure', and 'continued depression of the myocardium with beta-blocking agents over a period of time can, in some cases, lead to cardiac failure' - and a SPRINT secondary analysis reported that beta-blocker exposure in hypertensives without baseline HF was associated with increased incident HF. The class effect is nevertheless directionally contested: bisoprolol, carvedilol, metoprolol succinate and nebivolol are guideline therapy that reduces mortality in HFrEF, so beta blockade moves chronic HF in the opposite direction, and atenolol is simply not one of the agents with HF outcome evidence. The atlas numbers are consistent with prescribers avoiding it here: HF is depleted among atenolol users (enrichment 0.53, cotx 0.75%), which argues against crude channelling of HF patients onto this drug, and the prescription-propensity arm is flat (log10p 0.05). Against that, the strongest competing explanation is background risk rather than drug toxicity - hypertension is the dominant driver of incident HF, and a median 8.2 years from first atenolol script to first HF code in only 94 dated participants is exactly what treated hypertension alone would produce. The genetic layer adds nothing: a beta of 5.85 (odds ratio in the hundreds) for a variant with ~0.2% minor allele frequency is a sparse-data separation artifact, not an effect size, and the same variant's disease-arm effect (0.329 +/- 0.163) is only marginally distinguishable from noise. FAERS support is weak (1/3, PRR 1.75) and the pairing is not on the BNF.","gene_comment":"rs116844059 has no gene assignment because there is none to make: it maps to chr18:64,650,136 (GRCh38) as an intergenic variant, and the 400 kb window around it contains only a processed H3F3B pseudogene and two novel lncRNAs. It has no relation to the genes with real beta-blocker pharmacogenetic evidence (CYP2D6, ADRB1, ADRB2, ADRA2C, GRK4, GRK5 in the CPIC guideline), so it must not be presented as mechanism.","score_check":"log10p_adr 9.45 with beta 5.85 on a ~0.2% MAF variant is not believable as an effect estimate - it is the signature of near-complete separation in a handful of carriers, and z_diff 5.83 is inherited wholesale from that inflated beta. The population-level numbers (enrichment 0.53, flat prescription arm, 94 dated cases) are internally coherent and are the only part of this record worth reading.","candidability":2,"candidability_reason":"A labelled class hazard exists, but here it is carried by a rare intergenic variant with an artifactual effect size and is otherwise explained by hypertension's own progression to heart failure.","sources":[{"title":"Atenolol tablet label (Mylan Pharmaceuticals) - Indications, Contraindications, Cardiac Failure warning, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a7048c81-6769-436b-8397-fbaa7489fcdf"},{"title":"Ensembl variation record for rs116844059 (chr18:64,650,136, intergenic, MAF 0.002)","url":"https://rest.ensembl.org/variation/human/rs116844059?content-type=application/json"},{"title":"Ensembl gene overlap for chr18:64,450,136-64,850,136 (pseudogene and lncRNAs only)","url":"https://rest.ensembl.org/overlap/region/human/18:64450136-64850136?feature=gene;content-type=application/json"},{"title":"Europe PMC search: atenolol and incident heart failure - incl. 'Beta-Blocker Use in Hypertension and Heart Failure (A Secondary Analysis of SPRINT)', Am J Cardiol 2022, PMID 34906366","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=atenolol%20AND%20%22incident%20heart%20failure%22&format=json&pageSize=15&resultType=core"},{"title":"Europe PMC search: beta-blocker pharmacogenomics - CPIC Guideline for CYP2D6, ADRB1, ADRB2, ADRA2C, GRK4, GRK5 and Beta-Blocker Therapy, PMID 38951961 / PMC11502236","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22beta-blocker%22%20AND%20pharmacogenomics%20AND%20ADRB1%20AND%20%22heart%20failure%22&format=json&pageSize=10&resultType=core"}]},"atenolol|K760":{"verdict":"statistical artifact","headline":"Fatty liver is not an atenolol effect on this evidence; two rare variants with OR ~77 are sparse-data noise","assessment":"K76.0 (fatty liver, not elsewhere classified) is not a licensed indication for atenolol, whose US label lists only hypertension, angina pectoris and hemodynamically stable acute myocardial infarction. Hepatic steatosis is also not a labelled adverse reaction: the label mentions only postmarketing elevated liver enzymes and/or bilirubin, and LiverTox scores atenolol as likelihood D, a possible rare cause of clinically apparent idiosyncratic injury, noting that atenolol undergoes little hepatic metabolism. There is a genuine indirect route: atenolol carries well documented adverse metabolic effects (weight gain of roughly 1-3 kg, an approximately 19% rise in triglycerides, and excess new-onset diabetes versus losartan in LIFE), all of which are upstream drivers of steatosis, and beta-adrenoceptor blockade worsened liver injury in a murine NASH model, though that work used non-selective propranolol acting on hepatocyte beta-2 receptors and does not transfer cleanly to beta-1-selective atenolol. Against a drug-caused reading, a dedicated FAERS analysis of beta-blocker metabolic disturbances found 70 hepatic-steatosis reports but no disproportionate reporting signal for any agent, which sits awkwardly with the fact sheet's STRONG FAERS call and PRR of 4.81. The atlas epidemiology is likewise unsupportive: fatty liver appears in only 1.08% of atenolol users with an enrichment of 0.89, i.e. slightly less common than among users of other drugs, so there is not even a confounding-by-indication excess to explain, and the 100% temporality over 102 dated participants with a median 9.1 years to onset is exactly the pattern the cohort inflates and is more consistent with slow metabolic ageing than with a drug event. What remains is a within-users genetic signal whose effect size is not credible.","gene_comment":"rs187564951 is an intronic variant in CDKL5, an X-linked brain-expressed kinase whose only established disease role is developmental and epileptic encephalopathy 2, with no liver or lipid biology; rs77074962 is annotated by Ensembl as intergenic on 8q21 and is merely nearest to CNBD1, a cyclic-nucleotide-binding-domain protein with no curated UniProt function or tissue annotation. Neither assignment supports a steatosis mechanism and both should be treated as positional labels, not genes.","score_check":"Both variants are rare (roughly 0.5-0.9% in non-Finnish Europeans), so a beta_adr of 4.343 (odds ratio near 77) for a 1% outcome inside a drug subgroup is the classic sparse-cell inflation, and the same variants in the drug-free population give only beta 0.359 with se 0.182, barely two standard errors. The z_diff of 4.93 is therefore driven by the inflated treated-arm estimate rather than by a real treatment interaction, and the flat prescription arm (log10p 0.12) tells us nothing either way.","candidability":3,"candidability_reason":"Class-level metabolic mechanism is real but indirect and unlabelled, the condition is not enriched in atenolol users, and the two rare variants give an implausible effect size on genes with no hepatic biology.","sources":[{"title":"Atenolol - LiverTox, NCBI Bookshelf","url":"https://www.ncbi.nlm.nih.gov/books/n/livertox/Atenolol/"},{"title":"Atenolol tablet label (Mylan) - DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a7048c81-6769-436b-8397-fbaa7489fcdf"},{"title":"Propranolol, a beta-adrenoceptor antagonist, worsens liver injury in a model of non-alcoholic steatohepatitis (PMC5226920)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC5226920/"},{"title":"Metabolic Disturbances Associated with Beta-Blocker Therapy: New Insights from a Large-Scale Pharmacovigilance Study (PMC13516744)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13516744/"},{"title":"UniProt O76039 - CDKL5, cyclin-dependent kinase-like 5","url":"https://www.uniprot.org/uniprotkb/O76039/entry"},{"title":"Ensembl variant rs77074962 (intergenic, chromosome 8) - Ensembl REST","url":"https://rest.ensembl.org/variation/human/rs77074962?content-type=application/json;pops=1"}]},"atenolol|L409":{"verdict":"plausible ADR","headline":"Psoriasis is a labelled atenolol ADR, but this rare intronic MICAL3 variant's OR~750 is a sparse-data artifact","assessment":"Psoriasis is not an atenolol indication - the label covers hypertension, angina and haemodynamically stable acute MI - but 'psoriasiform rash or exacerbation of psoriasis' appears verbatim in the atenolol ADVERSE REACTIONS section as a postmarketing report, so the condition is a recognised, labelled adverse effect of the drug itself rather than a comorbidity of the treated group. The wider dermatology literature ranks beta-blockers among the drugs most often blamed for inducing or aggravating psoriasis, with a characteristically long latency of months to years, and a prospective cohort found regular beta-blocker use for >=6 years associated with incident psoriasis (HR 1.39, 95% CI 1.11-1.73). That evidence base is nonetheless dominated by case reports and retrospective series, and a 2025 triangulation study (NHANES with propensity-score matching, FAERS disproportionality, and two-sample Mendelian randomisation) found no association, no robust FAERS signal among 300 psoriasis reports, and an MR estimate pointing towards reduced psoriasis risk - concluding that the classic association is confounding by cardiovascular indication. The atlas is consistent with that null on the drug side: enrichment is 1.08, i.e. psoriasis is no commoner in atenolol users than in users of other drugs, so there is no confounding by indication here but equally no population-level excess, and the atlas's own FAERS field is no_signal with PRR 1.02. Temporality of 83.7% over 86 dated participants with a median 6.34 years to first record matches the long-latency picture but, as the cohort inflates this metric, it is supportive rather than probative. The genetic layer does not carry the claim: beta_adr 6.62 implies an odds ratio near 750, which is not a credible biological effect for a complex inflammatory skin disease and is the signature of separation on a rare allele, and log10p_adr 7.05 (p ~9e-8) does not even clear genome-wide significance for a single-variant scan.","gene_comment":"rs566280265 sits at chr22:17,908,089 (GRCh38), intronic in MICAL3, a ubiquitously expressed actin-depolymerising monooxygenase and Rab effector involved in vesicle trafficking and cytokinesis, with no established keratinocyte, IL-23/Th17 or beta-adrenergic biology and no place among the recognised psoriasis loci (HLA-C/PSORS1, IL12B, IL23R, TNIP1). At gnomAD MAF 0.08% globally and ~0.6% in non-Finnish Europeans the allele is rare enough that a handful of carriers can drive the entire signal, so the gene assignment is positionally correct but mechanistically empty.","score_check":"beta_disease 0.156 +/- 0.21 is indistinguishable from zero, so the variant is not simply a psoriasis predisposition allele, but the z_diff of 5.15 is generated wholly by an implausible treated-arm beta of 6.62 on a MAF-0.08% intronic SNV, which is far more consistent with sparse-data separation than with a real effect. The epidemiological fields (enrichment 1.08, FAERS PRR 1.02, log10p_prescribed 0.2) are internally coherent and simply show no drug-level excess of psoriasis.","candidability":4,"candidability_reason":"Real labelled class ADR, but the atlas shows no excess of psoriasis in atenolol users, FAERS is null, recent triangulated evidence points to no effect or the opposite direction, and the MICAL3 variant is a sparse-data artifact with no plausible mechanism.","sources":[{"title":"Atenolol tablet label (DailyMed) - Indications and Adverse Reactions incl. 'psoriasiform rash or exacerbation of psoriasis'","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a32eccc4-16d1-47a5-9e61-261f5d38638c"},{"title":"Drug-induced psoriasis: clinical perspectives (Psoriasis: Targets and Therapy, PMC5774610)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC5774610/"},{"title":"Revisiting the Association Between Beta-blockers and Psoriasis: Evidence from Real-World Data (Endocr Metab Immune Disord Drug Targets, 2025)","url":"https://pubmed.ncbi.nlm.nih.gov/40873291/"},{"title":"Drug-induced psoriasis - DermNet NZ","url":"https://dermnetnz.org/topics/drug-induced-psoriasis"},{"title":"UniProt Q7RTP6 - MICAL3, [F-actin]-monooxygenase MICAL3","url":"https://rest.uniprot.org/uniprotkb/Q7RTP6.txt"},{"title":"Ensembl variation record for rs566280265 (chr22:17908089, intron_variant, gnomAD MAF 0.0008)","url":"https://rest.ensembl.org/variation/human/rs566280265?content-type=application/json"}]},"atenolol|M549":{"verdict":"co-prescription population","headline":"Unspecified back pain is background morbidity in hypertensive atenolol users, not a labelled atenolol effect","assessment":"Atenolol is licensed only for hypertension, angina, cardiac arrhythmias and myocardial infarction, so dorsalgia is not an indication; neither the US label nor the UK Tenormin SmPC lists back pain, myalgia or cramps among undesirable effects (the label's only musculoskeletal terms are leg pain, up to 3% vs 0.5-1% placebo, and lupus-like syndrome). Non-specific back pain is instead strongly co-morbid with the cardiovascular population atenolol treats: in NHIS 2016-2018 hypertension raised the odds of spinal pain by 40% and cardiovascular conditions by 58%, so any hypertensive cohort carries excess M54.9 without a drug effect. The atlas is consistent with that reading rather than with an ADR: enrichment is 0.95, i.e. back pain is no commoner in atenolol users than in users of other drugs, and cotx_pct is only 0.67%. A drug-caused mechanism is not described; the closest class evidence runs the other way, since propranolol modestly reduced chronic TMD pain in a randomised crossover trial and gave ~4% analgesia to experimental heat pain, an effect judged central because the peripherally acting comparator did nothing - and atenolol, being hydrophilic and poorly CNS-penetrant, would not be expected to reproduce even that. Temporality of 94.5% over 109 dated participants is uninformative given the cohort's known upward compression, and a median 6.7 years between first prescription and first back-pain code is far too long a lag for a pharmacological reaction. The FAERS 'STRONG' flag with PRR 2.22 carries little weight here, because back pain is a high-background, non-specific reported term that tracks the comorbidity of the reporting population.","gene_comment":"Both hits are weak assignments: rs138843581 is intronic/upstream in LINC02934, a lncRNA on chr2, and rs72912279 is intronic in an uncharacterised chr18 locus (LOC105372093, reported here as ENSG00000288545) - neither is a protein-coding gene with any nociceptive, adrenergic or musculoskeletal annotation. Nothing in these loci supports a mechanism, and they should not be presented as one.","score_check":"The numbers are internally coherent but fragile: log10p_adr 6.17 is below genome-wide significance, and beta_adr 3.354 (OR ~28) on variants with TOPMED MAF 0.3-0.7% in ~109 dated cases is the classic shape of a sparse-data effect rather than a real large one. The clean nulls in the disease arm (beta 0.011 +/- 0.099) and the prescribing arm are reassuring against predisposition and channelling, but they cannot rescue an unreplicated rare-variant signal on a non-specific pain code.","candidability":2,"candidability_reason":"Back pain is unlabelled, unenriched (0.95) and explained by hypertensive comorbidity, with a 6.7-year lag and a sub-threshold rare-variant signal in two unannotated lncRNA loci.","sources":[{"title":"Atenolol tablet label (Mylan) - Indications and Adverse Reactions, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a7048c81-6769-436b-8397-fbaa7489fcdf"},{"title":"Tenormin 100 mg Tablets SmPC, sections 4.1 and 4.8 (electronic medicines compendium)","url":"https://www.medicines.org.uk/emc/product/12854/smpc"},{"title":"Spinal Pain, Chronic Health Conditions and Health Behaviors: NHIS 2016-2018 (PMID 37047983)","url":"https://pubmed.ncbi.nlm.nih.gov/37047983/"},{"title":"Tchivileva et al., Effect of COMT polymorphism on response to propranolol therapy in chronic musculoskeletal pain (PMID 20216107)","url":"https://pubmed.ncbi.nlm.nih.gov/20216107/"},{"title":"Schweinhardt et al., Effects of intravenous propranolol on heat pain sensitivity in healthy men (PMID 23070986)","url":"https://pubmed.ncbi.nlm.nih.gov/23070986/"},{"title":"dbSNP rs138843581 (LINC02934, intron/upstream, MAF ~0.004)","url":"https://www.ncbi.nlm.nih.gov/snp/rs138843581"},{"title":"dbSNP rs72912279 (chr18 LOC105372093, intron variant, MAF ~0.007)","url":"https://www.ncbi.nlm.nih.gov/snp/rs72912279"}]},"atenolol|N920":{"verdict":"statistical artifact","headline":"Menorrhagia is not an atenolol effect; signal rests on one rare intergenic variant in a gene desert","assessment":"Atenolol is licensed only for hypertension, angina pectoris and haemodynamically stable acute myocardial infarction; excessive menstruation (N92.0) is neither an indication nor a listed adverse reaction on the US label, whose adverse-reaction and postmarketing sections contain no menstrual, uterine or vaginal bleeding term, and the StatPearls review likewise lists no gynaecological effect. The only bleeding-adjacent label entries are the postmarketing reports of thrombocytopenia and purpura, which are rare, non-specific to atenolol among beta-blockers, and would present as bruising or mucosal bleeding rather than as isolated menorrhagia. No mechanistic or clinical literature links beta-1 blockade to heavy menstrual bleeding: a PubMed search for atenolol and menstrual endpoints returns three papers, none about bleeding, and the beta-adrenergic literature on the uterus concerns beta-2-mediated relaxation and tocolysis, not menstrual blood loss. The atlas' own comparison arms argue against an adverse-reaction reading rather than for it: co-treatment is only 1.06% with enrichment 0.87, i.e. menorrhagia is slightly LESS common in atenolol users than in users of other drugs, which is what one expects when a mostly older hypertensive population is compared with drug users of all ages, and FAERS gives a PRR of 0.48 with no disproportionality signal. Temporality of 94% after 100 dated participants with a median 2.17 years is exactly the pattern the cohort's prescription-versus-diagnosis record depth manufactures, so it carries almost no weight here, and confidence 21.7 with predisposition 0 reflects that. The within-user association therefore looks like a within-drug-stratum multiple-testing hit on a very rare variant rather than a drug-caused phenomenon.","gene_comment":"rs112437286 has no gene assignment for good reason: it is an intergenic A/T variant at chr4:29,761,337 with a global minor allele frequency of about 0.2%, sitting in a 4p15.1 gene desert whose only nearby features are the processed pseudogene EEF1A1P21 and two lncRNAs, with the nearest protein-coding gene PCDH7 roughly 1 Mb away. There is no candidate mechanism to attach to this locus, and nothing in it should be presented as biology.","score_check":"beta_adr of 6.668 (odds ratio in the hundreds) for a 0.2%-frequency allele against a case count small enough that only 100 participants had datable records is the classic sparse-data signature, and the same variant is pure noise for the condition itself (beta_disease 0.105 +/- 0.128, log10p 0.38), so the large z_diff of 5.67 is driven entirely by the unstable treated-arm estimate. The negative background arms - enrichment 0.87, FAERS PRR 0.48, not on the BNF interaction list - are consistent with each other and with the literature, and they all point away from a real effect.","candidability":1,"candidability_reason":"No label or literature support, depleted rather than enriched in atenolol users, no FAERS signal, and a single ultra-rare intergenic variant in a gene desert with an implausibly large effect.","sources":[{"title":"Atenolol tablet - DailyMed label (Caraco Pharmaceutical Laboratories), Indications and Adverse Reactions","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=f36d4ed3-dcbb-4465-9fa6-1da811f555e6"},{"title":"Atenolol - StatPearls, NCBI Bookshelf (NBK539844)","url":"https://www.ncbi.nlm.nih.gov/books/NBK539844/"},{"title":"Ensembl REST: variation rs112437286 (chr4:29,761,337, intergenic, MAF 0.00216)","url":"https://rest.ensembl.org/variation/human/rs112437286?content-type=application/json"},{"title":"Ensembl REST: genes overlapping chr4:28,761,337-30,761,337 (nearest coding gene PCDH7 ~1 Mb away)","url":"https://rest.ensembl.org/overlap/region/human/4:28761337-30761337?feature=gene;content-type=application/json"},{"title":"PubMed search: atenolol AND menstrual (3 results, none on menstrual bleeding)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=atenolol+menstrual"}]},"atenolol|R002":{"verdict":"licensed indication","headline":"Atenolol treats palpitations; the signal is a rare-variant artifact on an unnamed lncRNA","assessment":"Palpitations (R00.2) is a symptom for which atenolol is given, not one it causes: the UK Tenormin SmPC lists 'management of cardiac arrhythmias' among its therapeutic indications alongside hypertension, angina and myocardial infarction, and StatPearls names beta-blockers first-line therapy for symptomatic or frequent ectopy. The US atenolol label (Mylan) restricts indications to hypertension, angina and acute MI, and its cardiac adverse reactions are bradycardia, heart block and worsening heart failure - i.e. atenolol moves heart rate and rhythm awareness in the OPPOSITE direction to palpitations, which is decisive evidence against an ADR reading. The one genuine drug-caused route is beta-blocker rebound: both the SmPC ('should not be withdrawn abruptly... over 7-14 days') and a documented rebound case report describe tachycardia and palpitations days after abrupt cessation, and this is the most likely driver of the FAERS palpitations signal (PRR 2.59) rather than on-treatment harm. The atlas numbers fit the indication reading: enrichment 1.25 shows palpitations are modestly more common in atenolol users than in users of other drugs, exactly as expected when the symptom prompts the prescription, and BNF does not list palpitations as an atenolol side effect (on_bnf 0). Temporality of 97.5% with a 6.2-year median lag is the weak, cohort-compressed direction and cannot rescue this, since patients started on atenolol for hypertension accumulate palpitation codes over years of follow-up for unrelated reasons. Nothing here identifies a pharmacogenomic modifier of an atenolol effect; it identifies patients whose symptom brought them to the drug.","gene_comment":"ENSG00000303889 is an unnamed novel lncRNA at 9q34.3 (chr9:136.60-136.61 Mb) and rs142183691 is an intronic SNV within it with global MAF ~0.0012; the nearest protein-coding genes, EGFL7 and AGPAT2, sit ~50 kb away and have no cardiac rhythm role, and the variant has no publication record. This is a positional label, not a mechanism, and it must not be read as one.","score_check":"The numbers do not hang together: a beta_adr of 4.97 (OR ~140) on a variant with MAF 0.0012 in roughly 160 dated cases implies well under one expected carrier by chance, so the log10p of 7.03 rests on a handful of genotypes - the signature of sparse-data bias or an imputation/genotyping error. The drug-free disease arm is the only internally coherent estimate (beta 0.572, se 0.209, z=2.7, nominal only), and the propensity arm is flat (log10p 0.11), so z_diff 4.61 is inherited from the unstable within-users beta.","candidability":1,"candidability_reason":"Backwards pharmacology plus confounding by indication, on a rare intronic lncRNA variant whose effect size is an artifact of its own sparsity.","sources":[{"title":"Tenormin 100 mg Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/12854/smpc"},{"title":"Atenolol tablet - US prescribing information (DailyMed, Mylan)","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=a7048c81-6769-436b-8397-fbaa7489fcdf"},{"title":"Palpitations - StatPearls, NCBI Bookshelf","url":"https://www.ncbi.nlm.nih.gov/books/NBK436016/"},{"title":"Beta-blocker Rebound Phenomenon in an Adolescent with Graves' Disease (PMC9724061)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9724061/"},{"title":"Ensembl variant record for rs142183691 (chr9:136606925, intronic, MAF 0.0012)","url":"https://rest.ensembl.org/variation/human/rs142183691?content-type=application/json"},{"title":"Ensembl gene record ENSG00000303889 (novel lncRNA, 9q34.3)","url":"https://rest.ensembl.org/lookup/id/ENSG00000303889?content-type=application/json"}]},"atenolol|R030":{"verdict":"licensed indication","headline":"R03.0 is atenolol's indication, not its harm; beta=7 on ultra-rare variants is a sparse-data artifact","assessment":"ICD-10 R03.0 records an episode of elevated blood pressure in a patient with no formal hypertension diagnosis, or an isolated incidental reading - i.e. exactly the finding that leads to a beta-blocker prescription, not a consequence of one. The US label lists hypertension as atenolol's first indication and describes it as a beta1-selective blocker giving 24-hour blood pressure reduction with once-daily dosing; the HDPAL randomised trial confirms the direction, with 44-h ambulatory BP falling on atenolol-based therapy and home BP consistently lower than on the lisinopril arm. The drug therefore moves this condition in the opposite direction to the one an ADR reading requires, which is decisive against causation. The only labelled withdrawal hazard is exacerbation of angina, MI and ventricular arrhythmia after abrupt discontinuation, and the rebound-hypertension literature is dominated by clonidine, methyldopa and minoxidil rather than atenolol, so even a withdrawal mechanism is not the described one here. The atlas numbers are consistent with indication rather than harm: enrichment is 1.02 (R03.0 is no commoner in atenolol users than in users of other drugs, it is simply a generic primary-care code), the indication score is 9.9 and predisposition 0. Temporality of 84% after first exposure with a 4.85-year median is uninformative here, because atenolol users have their blood pressure measured indefinitely, so a post-exposure 'high reading' code reflects surveillance; the fact sheet's own guidance is that only a low temporality carries weight. The FAERS PRR of 3.56 is the expected signature of confounding by indication in spontaneous reports, the very effect that disproportionality studies now handle by excluding reports carrying the drug's own labelled indication.","gene_comment":"rs113236683 (chr2:199,394,154, MAF 0.08%) is intronic in SATB2, a homeobox transcription factor whose established human phenotype is SATB2-associated (Glass) syndrome - craniofacial and neurodevelopmental, with no established blood-pressure role; the other two are worse, rs185398807 being intergenic at chr6:28.86 Mb in the extended-MHC/olfactory-receptor desert and only nominally assigned to the ribosomal-protein pseudogene RPL13P, and rs557049009 intergenic in a 21q21 gene desert. All three are ultra-rare (MAF 0.06-0.16%), so none of these assignments supports a mechanism.","score_check":"beta_adr of 7.04 implies an odds ratio of roughly 1,100 for alleles carried by under 0.2% of people in a stratum with 119 datable cases - the classic separation/sparse-data artifact, not an effect size. The drug-free arm is frankly null (beta 0.246 +/- 0.191, |beta| < 1.96*se, log10p 0.7), so the z_diff of 5.28 is produced entirely by the inflated treated-arm estimate rather than by any real treated-versus-untreated difference.","candidability":0,"candidability_reason":"Definitionally not an adverse reaction - R03.0 is the licensed indication and atenolol lowers BP - and the rare-variant effect size is implausible for its error.","sources":[{"title":"Atenolol tablet - DailyMed label (indications, mechanism, adverse reactions, cessation of therapy warning)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a32eccc4-16d1-47a5-9e61-261f5d38638c"},{"title":"ICD-10-CM R03.0: Elevated blood-pressure reading, without diagnosis of hypertension","url":"https://www.aapc.com/codes/icd-10-codes/R03.0"},{"title":"Hypertension in hemodialysis patients treated with atenolol or lisinopril: a randomized controlled trial (HDPAL)","url":"https://europepmc.org/article/MED/24398888"},{"title":"Atorvastatin-associated neurological adverse events: disproportionality analysis with adjustment for confounding by indication (FAERS)","url":"https://europepmc.org/article/MED/42482918"},{"title":"Ensembl variant record rs113236683 (intronic, chr2:199,394,154, MAF 0.00078)","url":"https://rest.ensembl.org/variation/human/rs113236683?content-type=application/json"},{"title":"Ensembl gene record SATB2 (ENSG00000119042, chr2:199,268,900-199,471,271)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/SATB2?content-type=application/json"}]},"atenolol|R252":{"verdict":"statistical artifact","headline":"Cramp is common in hypertensives but not enriched under atenolol; LYL1 burden beta of 66 is not real","assessment":"Atenolol is licensed only for hypertension, angina and acute MI; cramp and spasm (R25.2) is neither an indication nor a labelled adverse reaction. The US TENORMIN label does record neighbouring lower-limb complaints - leg pain 3% vs 1% placebo on elicited reporting, cold extremities 12% vs 5%, tiredness 26% vs 13% - so a peripheral-vasoconstriction route to leg symptoms under beta-blockade is at least conceivable, but the label never names cramp, spasm or myalgia. The one controlled study of the question (Imai et al., Eur J Clin Pharmacol 1995, 78 hypertensives) found nocturnal calf cramps and raised CPK with pindolol and carteolol only, and explicitly none with propranolol, arotinolol or metoprolol - i.e. the class signal does not extend to the beta-1-selective, hydrophilic agents that atenolol belongs to. A 396-patient cardiovascular cohort found musculoskeletal pain and cramp in over half of patients and no significant association with beta-blocker use, which matches the atlas's own enrichment of 0.93: cramp is no commoner in atenolol users than in users of other drugs, so this is background morbidity of an elderly hypertensive population rather than either confounding by indication or a drug effect. Temporality of 94.8% (n=96, median 6.7 years after first exposure) is the uninformative direction in this cohort and is equally consistent with simple ageing over follow-up. FAERS support is weak (1 of 3, PRR 1.84) and cramp is not a BNF-listed effect for this drug.","gene_comment":"LYL1 is a haematopoietic basic helix-loop-helix transcription factor, expressed mainly in bone marrow and lymphoid tissue and known chiefly for its t(7;19) translocation in T-ALL; it has no described role in skeletal muscle, motor neurons or electrolyte handling, so an MPC burden hit here carries no mechanistic reading for cramp. With a single variant unit and only ~96 dated cases the burden test is also running on very few carriers.","score_check":"beta_adr of 66.6 is not a believable effect size on any coefficient scale for a burden test and is the signature of quasi-complete separation in a sparse cell, and the disease-arm beta of 4.57 +/- 1.73 is likewise implausibly large even though it clears 1.96*se; z_diff of 5.7 therefore compares two unstable numbers. The p-values are real arithmetic but do not survive as evidence of a drug-specific genetic effect.","candidability":2,"candidability_reason":"No enrichment over background, cramp absent from the atenolol label and from the beta-1-selective arm of the only controlled study, and a haematopoietic transcription factor carrying an impossible effect size.","sources":[{"title":"Atenolol - StatPearls, NCBI Bookshelf (NBK539844)","url":"https://www.ncbi.nlm.nih.gov/books/NBK539844/"},{"title":"TENORMIN (atenolol) tablets - DailyMed label, Almatica Pharma","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=746db603-a6e1-4dc3-c2d8-92314419098c"},{"title":"Imai Y et al. Muscle cramps and elevated serum creatine phosphokinase levels induced by beta-adrenoceptor blockers. Eur J Clin Pharmacol 1995 (PMID 7621844)","url":"https://pubmed.ncbi.nlm.nih.gov/7621844/"},{"title":"Su HC et al. Factors affecting the intensity of chronic musculoskeletal pain in patients with cardiovascular disease... muscle cramps. Medicine (Baltimore) 2023, PMC10615485","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10615485/"},{"title":"UniProt P12980 (LYL1_HUMAN) - protein function, expression and disease","url":"https://rest.uniprot.org/uniprotkb/P12980.txt"}]},"atenolol|R53":{"verdict":"plausible ADR","headline":"Fatigue is a labelled atenolol effect, but a rare intergenic GPR37 SNV with OR~140 is a sparse-data artifact","assessment":"Malaise and fatigue (R53) is not an indication for atenolol - the licensed indications are hypertension, angina and haemodynamically stable acute MI - and it is an explicitly labelled adverse reaction: in the Tenormin US label's controlled hypertension studies, elicited tiredness was reported by 26% of atenolol patients versus 13% on placebo, with lethargy 3% vs 0.7% and dizziness 13% vs 6%. The Ko et al. JAMA 2002 quantitative review of 15 placebo-controlled trials in >35,000 patients puts the real magnitude much lower but still positive and significant: an absolute annual excess of 18 reports of fatigue per 1000 treated patients (95% CI 5-30, NNH 57), and the excess was significantly larger for early-generation beta-blockers, the class atenolol belongs to. So the drug-condition pairing itself is genuine and directionally correct - beta1 blockade lowers resting and exercise heart rate and blunts chronotropic and cardiac-output reserve, which is the standard explanation for exertional tiredness, and nothing in the literature moves fatigue in the opposite direction. The atlas design arms behave the way a true drug-specific effect should: enrichment is exactly 1.0 (atenolol users are no more likely to carry R53 than users of other drugs, so there is no channelling of tired patients onto this drug), the prescription-propensity arm is null (log10p 0.76), and the same variant is flat for R53 in the drug-free population (beta 0.146 +/- 0.139, well inside noise), giving z_diff 5.72. The weak link is the genetic signal itself, not the pharmacology: beta_adr 4.92 is an odds ratio around 140 for a variant with ~0.5-0.7% minor allele frequency over only 73 datable cases, which is the signature of quasi-separation in a sparse 2x2 table rather than a credible biological effect size. Temporality of 100% is uninformative here given 39% undated records, a median lag of 6 years and a symptom code that accumulates with age, and the external pharmacovigilance support is thin (FAERS weak, PRR 1.55). R53 is also a residual, non-specific symptom code, so even a true within-users association would be hard to map onto a specific mechanism.","gene_comment":"rs139716193 is a rare non-coding SNV at chr7:124,766,405 (GRCh38), about 300 bp upstream of the GPR37 transcription start site and just outside an annotated promoter/CTCF region, and Ensembl classifies it as intergenic; the GPR37 label is a nearest-gene call with no eQTL or functional support and no GWAS Catalog association for this rsID. GPR37 (PAEL-R) is a parkin substrate and prosaposin receptor with roles in oligodendrocytes and Parkinson disease, with no established connection to beta-adrenergic signalling, exercise capacity or fatigue, so it should not be presented as mechanism.","score_check":"The comparison arms are internally consistent (null disease effect, null prescribing propensity, enrichment 1.0), but beta_adr 4.92 for a ~0.5% allele over 73 cases implies an implausible OR of roughly 140 and is almost certainly a sparse-data inflation of a much smaller or absent effect. The clinical half of the association is solid; the variant-level effect size is not believable as reported.","candidability":5,"candidability_reason":"Fatigue on atenolol is a real, label- and trial-documented class effect with an unconfounded atlas signal, but the genetics is a rare intergenic nearest-gene call with a sparse-data effect size on a non-specific symptom code.","sources":[{"title":"TENORMIN (atenolol) US prescribing information - indications and adverse reactions (openFDA label record)","url":"https://api.fda.gov/drug/label.json?search=openfda.brand_name:%22TENORMIN%22&limit=1"},{"title":"Ko DT et al. Beta-blocker therapy and symptoms of depression, fatigue, and sexual dysfunction. JAMA 2002;288(3):351-7 (PMID 12117400, Europe PMC record)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:12117400&resultType=core&format=json"},{"title":"Ensembl Variation: rs139716193 - consequence, mapping and gnomAD population frequencies","url":"https://rest.ensembl.org/variation/human/rs139716193?content-type=application/json;pops=1"},{"title":"Ensembl: GPR37 gene coordinates (chr7:124,743,885-124,766,099, minus strand) and overlapping regulatory features at the variant position","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/GPR37?content-type=application/json"},{"title":"NCBI Gene summary for GPR37 (G protein-coupled receptor 37, 7q31.33, parkin substrate)","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=gene&id=2861&retmode=json"},{"title":"GWAS Catalog REST: associations for rs139716193 (none reported)","url":"https://www.ebi.ac.uk/gwas/rest/api/singleNucleotidePolymorphisms/rs139716193/associations"}]},"atenolol|R634":{"verdict":"statistical artifact","headline":"Beta-blockers cause weight gain, not loss; rare intronic PPP1R12B hit has an implausible effect size","assessment":"Abnormal weight loss (R63.4) is not an indication for atenolol, whose label covers hypertension, angina and post-myocardial-infarction mortality reduction, and the US label's adverse-reactions section lists no weight, appetite or nutritional term at all. The class effect runs the other way: a systematic analysis of eight randomised trials of at least six months found body weight higher on beta-blockers than on control, median difference +1.2 kg (range -0.4 to +3.5), with the gain concentrated in the first months of treatment, which is decisive evidence against reading weight loss as a beta-blockade effect. The atlas numbers agree with that direction rather than contradicting it: enrichment 0.84 means abnormal weight loss is slightly less common among atenolol users than among users of other drugs, and the prescription-propensity arm is null (log10p 0.87), so there is no confounding-by-indication signal to salvage either. Temporality of 100% over 153 dated participants with a median 7.3 years to onset is exactly the pattern expected when a long-standing chronic prescription precedes the intercurrent illnesses that actually cause late-life weight loss - malignancy, heart failure and cachexia, frailty - and in this cohort a high temporality is weak evidence on its own. The FAERS label of STRONG rests on a PRR of only 2.06 for a symptom term ('weight decreased') that is reported freely in elderly cardiac patients; spontaneous-report disproportionality has no exposed-patient denominator and is explicitly hypothesis-generating rather than evidence of incidence or causality, and it is not adjusted for indication. The genetic signal is the weakest part: log10p_adr 6.56 does not reach the usual genome-wide threshold, the same variant does precisely nothing on the trait in the drug-free population (beta_disease -0.013 +/- 0.136, log10p 0.04), so the entire z_diff of 5.1 is carried by a within-user estimate of beta 4.66 - an odds ratio near 100 - from a variant with a European minor allele frequency of about 0.6% in a case set of 153, which is the signature of sparse-data separation rather than a real effect.","gene_comment":"rs143365951 is a rare intronic SNP (chr1:202,410,290; gnomAD NFE MAF ~0.6%) inside PPP1R12B/MYPT2, the striated-muscle myosin phosphatase targeting subunit studied for cardiac contractility and fibrosis, with no reported link to appetite, body weight, cachexia or beta-blocker response. The assignment is purely positional - the only other feature at the site is the U6 snRNA pseudogene RNU6-89P - so it supplies a gene name, not a mechanism.","score_check":"The within-user beta of 4.66 for a 0.5% allele with 153 dated cases implies an effect far too large to be credible and is almost certainly sparse-data bias, especially against a flat, well-estimated null in the disease arm. The population-level numbers are internally consistent and point away from an ADR: the condition is depleted, not enriched, in atenolol users.","candidability":1,"candidability_reason":"Drug moves weight in the opposite direction, condition is depleted in users, gene is positional only, and the effect size is an implausible rare-variant artifact.","sources":[{"title":"Atenolol tablet - US prescribing information (DailyMed, Watson Laboratories)","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=a32eccc4-16d1-47a5-9e61-261f5d38638c"},{"title":"Sharma AM et al. Hypothesis: beta-adrenergic receptor blockers and weight gain: a systematic analysis. Hypertension 2001 (PMID 11230280)","url":"https://europepmc.org/article/MED/11230280"},{"title":"Ensembl variation record for rs143365951 (intronic, population frequencies)","url":"https://rest.ensembl.org/variation/human/rs143365951?content-type=application/json;pops=1"},{"title":"Ensembl genes overlapping chr1:202,400,000-202,420,000 (PPP1R12B, RNU6-89P)","url":"https://rest.ensembl.org/overlap/region/human/1:202400000-202420000?feature=gene;content-type=application/json"},{"title":"Lee et al. The MYPT2-regulated striated muscle-specific myosin light chain phosphatase limits cardiac myosin phosphorylation in vivo. J Biol Chem 2024 (PMID 38224947)","url":"https://europepmc.org/article/MED/38224947"},{"title":"Europe PMC search: limitations of FAERS disproportionality analysis (no denominators, confounding by indication, hypothesis-generating only)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22disproportionality%20analysis%22%20AND%20%22FAERS%22%20AND%20confounding&format=json"}]},"atenolol|R739":{"verdict":"plausible ADR","headline":"Atenolol really does worsen glycaemia, but this rare non-coding SLC38A8 hit is not credible genetics","assessment":"Hyperglycaemia (R73.9) is not an indication for atenolol - the US label lists only hypertension, angina and haemodynamically stable acute MI - and it is a recognised metabolic liability of non-vasodilating beta-blockers, with atenolol and metoprolol named explicitly as the agents in which 'adverse metabolic effects have been observed', in contrast to carvedilol and nebivolol. Described mechanisms are concrete: beta-2-mediated suppression of first-phase insulin secretion, unopposed alpha-adrenergic activity increasing hepatic glucose output, and reduced insulin-stimulated glucose uptake through vasoconstriction in skeletal muscle. The direction is not entirely one-way: the labelled glycaemic hazard of atenolol is the opposite one - masking the adrenergic warning signs of hypoglycaemia and prolonging severe hypoglycaemia in diabetic, fasting or vomiting patients - so a coded R73.9 in an atenolol user is more likely drug-induced dysglycaemia than drug-caused hypoglycaemia, but the drug moves glucose in both directions depending on context. The atlas support for a drug effect is weak on its own terms: enrichment is 0.79, i.e. hyperglycaemia is coded slightly LESS often in atenolol users than in users of other drugs (unsurprising, since the comparator arm is full of statins, thiazides and other antihypertensives that also raise glucose), and although 97.7% of the 131 dated cases post-date first exposure with a median 9 years, the cohort's record structure inflates that figure and a high value carries little weight. FAERS is supportive but non-specific (PRR 2.0, ROR+PRR+IC concordant), and confounding by indication is severe here because hypertension, obesity and metabolic syndrome are exactly the reasons atenolol is prescribed. The genetic layer, which is what would make this a pharmacogenomic finding rather than a known class effect, is the part that does not hold up.","gene_comment":"rs144889482 is a rare (global MAF ~0.6%) C/T variant at chr16:84,041,941 whose most severe consequence is 'non_coding_transcript_exon_variant', assigned to SLC38A8 - a sodium-dependent glutamine/alanine transporter expressed in brain, spinal cord and retina whose only established disease role is recessive foveal hypoplasia 2 (FVH2). There is no described glucose, insulin or adrenergic biology for SLC38A8, so this is a positional label on a non-coding variant, not a mechanism.","score_check":"beta_adr = 6.687 implies an odds ratio of roughly 800 for a 0.6%-frequency non-coding allele, which is not a believable biological effect and is the classic signature of quasi-separation / sparse-data bias in rare-variant logistic regression, even with log10p 8.36. The comparison arms behave sensibly (beta_disease 0.326 +/- 0.265 is pure noise, log10p_prescribed 0.47 is null, so z_diff 5.44 is driven entirely by the inflated within-user estimate rather than by a genuine treatment interaction).","candidability":5,"candidability_reason":"The drug-condition relationship is a genuine, mechanistically described class effect, but the specific locus is a rare non-coding variant in a retina/brain amino-acid transporter with an implausibly large effect size and no enrichment support.","sources":[{"title":"Atenolol - StatPearls, NCBI Bookshelf (NBK539844)","url":"https://www.ncbi.nlm.nih.gov/books/NBK539844/"},{"title":"Atenolol tablets - US prescribing information (DailyMed, Mylan)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a7048c81-6769-436b-8397-fbaa7489fcdf"},{"title":"Beta-blockers and metabolic modulation: unraveling the complex interplay with glucose metabolism, inflammation and oxidative stress (Front Pharmacol, PMC11695285)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11695285/"},{"title":"UniProtKB - SLC38A8 (Solute carrier family 38 member 8), human","url":"https://rest.uniprot.org/uniprotkb/search?query=gene:SLC38A8%20AND%20organism_id:9606%20AND%20reviewed:true&format=txt"},{"title":"Ensembl REST - variant rs144889482","url":"https://rest.ensembl.org/variation/human/rs144889482?content-type=application/json"},{"title":"Europe PMC search: beta-blockers, incident diabetes and hypertension","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22beta-blocker%22%20AND%20%22incident%20diabetes%22%20AND%20hypertension&format=json&pageSize=10&resultType=core"}]},"atorvastatin|E162":{"verdict":"statistical artifact","headline":"Statins raise glucose, not lower it; hypoglycaemia code tracks treated diabetes, TRIM11 beta implausible","assessment":"Hypoglycaemia (E16.2) is not an atorvastatin indication - the US label lists only LDL/triglyceride lowering and cardiovascular risk reduction - and it is not a labelled adverse reaction; the drug's documented metabolic effect runs the other way, with Warnings 5.4 stating that 'increases in HbA1c and fasting serum glucose levels have been reported with statins, including atorvastatin' and SPARCL reporting diabetes in 6.1% on atorvastatin versus 3.8% on placebo. A statin-caused hypoglycaemia is therefore directionally opposite to the one glycaemic effect that is actually established for the class, and the only substantial clinical study of the plausible interaction pathway (statin added to a sulfonylurea, 592,872 new users) found hazard ratios for severe hypoglycaemia 'consistent with no association', whereas fibrates did show excess risk. What an unspecified-hypoglycaemia code overwhelmingly marks in an EHR cohort is insulin- or sulfonylurea-treated diabetes, and diabetes is precisely the population in which atorvastatin is near-universally prescribed, so the co-occurrence is expected without any drug effect. The atlas's own honest test does not even support strong co-prescription: cotx_pct is 0.96% with enrichment 1.06, essentially flat against other drugs' users, and log10p_prescribed 0.34 shows the variant does not predict being prescribed the drug. Temporality of 93.8% over only 145 dated participants with a 5.69-year median gap is the uninformative high direction the atlas warns about, since prescription records predate diagnosis records by construction. The FAERS 'STRONG' flag with PRR 2.52 is the expected shape of co-medication confounding in spontaneous reports, where atorvastatin is co-reported with insulins and sulfonylureas, not independent support.","gene_comment":"TRIM11 is a brain-enriched RING-type E3 ubiquitin ligase involved in degradation of insoluble ubiquitinated proteins, cortical neurogenesis and AIM2 inflammasome control; UniProt lists no metabolic or insulin-secretion role and the GWAS Catalog gene page shows no glucose or diabetes trait, so the assignment supplies no mechanism for hypoglycaemia. It is a single MPC burden unit rather than a replicated variant, which makes the gene label a bookkeeping entry, not evidence.","score_check":"beta_adr of 65.3 from one MPC burden mask over ~145 dated cases is a sparse-data separation artifact, not an effect size, and the z_diff of 8.14 is driven entirely by that implausible number. The comparison arm is also weak: beta_disease 5.738 with se 2.33 is only ~2.5 standard errors and log10p_disease 1.86 is nominal at best.","candidability":1,"candidability_reason":"Backwards from the drug's documented hyperglycaemic effect, explained by diabetes co-treatment, and carried by an implausible burden-test beta in a gene with no glucose biology.","sources":[{"title":"Atorvastatin Calcium tablets, film coated - full prescribing information (DailyMed, Rising Pharma)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=73cefb0b-1747-4fe0-b252-377f4ed94c6f"},{"title":"Severe hypoglycemia in users of sulfonylurea antidiabetic agents and antihyperlipidemics (Leonard et al., PMID 26566262)","url":"https://pubmed.ncbi.nlm.nih.gov/26566262/"},{"title":"Risk of new-onset diabetes across individual statins in secondary prevention: Korean NHIS cohort (PMC13008627)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13008627/"},{"title":"UniProtKB: E3 ubiquitin-protein ligase TRIM11 (human) - function, tissue specificity, disease","url":"https://rest.uniprot.org/uniprotkb/search?query=gene:TRIM11+AND+organism_id:9606+AND+reviewed:true&fields=accession,protein_name,cc_function,cc_disease,cc_tissue_specificity&format=txt"},{"title":"GWAS Catalog - reported associations mapped to TRIM11 (none for glucose or diabetes traits)","url":"https://www.ebi.ac.uk/gwas/genes/TRIM11"},{"title":"PubMed search: statin and hypoglycemia - no direct risk studies beyond the sulfonylurea interaction cohort","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+hypoglycemia"}]},"atorvastatin|F419":{"verdict":"statistical artifact","headline":"Rare intergenic 9q21.11 SNV with an implausible OR; statins do not raise anxiety risk in the literature","assessment":"Anxiety disorder is neither an indication for atorvastatin nor a labelled adverse reaction: the US SPL lists insomnia in trials and depression, nightmare, dizziness and reversible cognitive impairment post-marketing, but anxiety appears nowhere in the adverse-reactions section. The best population evidence points the other way: a 2026 systematic review and meta-analysis of five studies and ~1.9 million participants found statin use was not associated with incident anxiety (HR 0.75, 95% CI 0.55-1.04), and a primary-care cohort of statin users with depression found no excess anxiety (aOR 1.05-1.07 across 2-12 months) and no excess self-harm, sleep disturbance or suicidality. The atlas's own background arm agrees: enrichment 0.77 means anxiety is depleted, not enriched, among atorvastatin users, and cotx_pct of 0.98% is an order of magnitude below true anxiety prevalence, so the phenotype is sparsely captured. Temporality of 99.6% over 255 dated participants is exactly the uninformative high value the atlas warns about, and the pharmacovigilance side is null (FAERS weak 1/3, PRR 1.09, not on the BNF list). The variant contributes nothing without the drug (log10p_disease 0.08; beta 0.028 +/- 0.139, indistinguishable from zero), so z_diff 5.33 is generated entirely by an inflated within-user beta of 4.14 - an odds ratio near 60 for anxiety, which no real pharmacogenomic effect on a common psychiatric phenotype produces. The literature, the enrichment direction and the effect size all point the same way: this is a sparse-data signal from a rare allele, not a drug-caused finding.","gene_comment":"rs193041764 is a rare intergenic SNV at chr9:68,662,929 (9q21.11) with no gene assignment in dbSNP; the nearest protein-coding gene, PIP5K1B, is ~42 kb away and the intervening annotation is lncRNA (TMEM252-DT, LINC01506), so there is no gene-level mechanism to invoke. It sits in the pericentromeric segmental-duplication belt of chromosome 9 and is rare (G allele ~0.4-0.7% in Europeans, absent in East Asians), a combination that makes genotyping/imputation error and sparse-data bias the leading explanations.","score_check":"beta_adr 4.142 at log10p 7.32 implies a standard error near 0.78 and an odds ratio around 60 for a common psychiatric diagnosis carried by only ~1% of users - a textbook sparse-data artifact for a <1% frequency allele. The disease arm (beta 0.028 +/- 0.139) is a clean null, so the whole z_diff rests on the unstable treated-arm estimate.","candidability":1,"candidability_reason":"Null-to-protective drug-anxiety literature, depleted comorbidity, no label signal, and an implausible odds ratio from a rare intergenic variant in a repeat-rich region.","sources":[{"title":"Atorvastatin calcium tablet, film coated - DailyMed label (Indications; Adverse Reactions incl. postmarketing)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=f2d8b5e9-6ff2-4efa-ab01-e13df29f7371"},{"title":"The association of statin use with the risk of anxiety: a systematic review and meta-analysis (Front Psychiatry 2026)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12957783/"},{"title":"Mortality and adverse events associated with statin use in primary care patients with depression (BMJ Ment Health 2024)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11110566/"},{"title":"The effects of atorvastatin on emotional processing, reward learning and verbal memory: a randomised study (J Psychopharmacol 2021)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8652357/"},{"title":"dbSNP reference SNP report: rs193041764","url":"https://www.ncbi.nlm.nih.gov/snp/rs193041764"},{"title":"Ensembl REST overlap, GRCh38 chr9:68.16-69.16 Mb gene annotation","url":"https://rest.ensembl.org/overlap/region/human/9:68162929-69162929?feature=gene;content-type=application/json"}]},"atorvastatin|G473":{"verdict":"statistical artifact","headline":"Sleep apnoea is a statin-treated comorbidity, not a statin effect; signal rests on a rare lncRNA variant","assessment":"Sleep apnoea (G47.3) is not an indication for atorvastatin, whose label covers LDL lowering and cardiovascular risk reduction, and it appears nowhere in the label's adverse-reaction or postmarketing sections; only insomnia and non-specific sleep disturbance are listed post-marketing, and a meta-analysis of 19 blinded randomised trials (123,940 participants) concluded that blinded trial data do not support a causal relationship between statins and most labelled conditions, sleep disturbance included. The traffic runs the other way: OSA travels with obesity, intermittent hypoxia and dyslipidaemia, so OSA patients are heavily statin-exposed and atorvastatin has itself been trialled as an adjunct in severe OSA (40 mg for 12 weeks improved lipids and systolic BP but not endothelial function; a 2026 pilot found no BP benefit), with no trial reporting worsening of sleep-disordered breathing. The atlas is consistent with that reading in one respect and not another: the prescription-propensity arm is flat (log10p 0.33), and enrichment of 0.93 says atorvastatin users are, if anything, marginally less likely than users of other drugs to carry an OSA code, so this is not a floridly indication-confounded pair either. What is left is a within-users effect of beta 3.36 (OR ~29) at a variant with ~0.4% minor allele frequency across roughly 295 datable cases, which is the classic shape of sparse-data inflation rather than a real interaction; the same variant's effect in the drug-free population is a modest and much more credible beta 0.457 +/- 0.159, i.e. a weak general OSA predisposition allele, and the z_diff of 5.23 simply inherits the inflated numerator. A 98% temporality on that denominator carries little weight given the cohort's known upward compression, and the FAERS PRR of 2.14 for a drug taken by most of the OSA population is expected reporting bias, not corroboration. There is no described mechanism by which HMG-CoA reductase inhibition would cause upper-airway collapse, and none is required to explain these numbers.","gene_comment":"rs114063754 (chr2:78,427,034, MAF ~0.38% in gnomAD) is intronic to ENSG00000298364, an Ensembl 'novel transcript' lncRNA also annotated as uncharacterized LOC124906027, sitting in a chr2p12 gene desert with no known link to respiratory control, craniofacial anatomy or adiposity. This is a positional label, not a mechanism, and it should not be presented as one.","score_check":"The disease-arm estimate (0.457 +/- 0.159, z = 2.87) is believable as a weak predisposition effect, but an OR near 29 in the treated arm for a 0.4%-frequency allele on a few hundred coded cases is far more consistent with sparse-count instability than with a sevenfold drug-specific amplification. The 1.08% co-occurrence rate also implies heavy under-coding of OSA, which further destabilises rare-variant case counts.","candidability":2,"candidability_reason":"No mechanism, no label or trial support for statin-induced sleep apnoea, and the entire within-users signal is an implausibly large odds ratio at a rare intronic variant in an uncharacterized lncRNA.","sources":[{"title":"LIPITOR (atorvastatin calcium) US prescribing information, FDA accessdata","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2022/020702s079lbl.pdf"},{"title":"PubMed results: statins and obstructive sleep apnea (trials of statins given to OSA patients)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+obstructive+sleep+apnea"},{"title":"Joyeux-Faure et al., Response to statin therapy in obstructive sleep apnea syndrome: a multicenter randomized controlled trial (PMID 25221387)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:25221387&format=json&resultType=core"},{"title":"Europe PMC: statins and sleep disturbance as adverse effects, incl. Lancet blinded-trial meta-analysis (PMID 41655587) and polysomnography review (PMID 41742773)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=statin%20AND%20%22sleep%20disturbance%22%20AND%20adverse&format=json&resultType=core&pageSize=6"},{"title":"Europe PMC: obstructive sleep apnoea, intermittent hypoxia and dyslipidaemia mechanisms","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22sleep%20apnea%22%20AND%20dyslipidemia%20AND%20mechanism&format=json&resultType=core&pageSize=6"},{"title":"dbSNP refSNP rs114063754 (NCBI Variation Services)","url":"https://api.ncbi.nlm.nih.gov/variation/v0/refsnp/114063754"},{"title":"Ensembl REST lookup for ENSG00000298364 (lncRNA, novel transcript, chr2)","url":"https://rest.ensembl.org/lookup/id/ENSG00000298364?content-type=application/json"}]},"atorvastatin|H409":{"verdict":"contested","headline":"Statin-glaucoma link is genuinely contested; the atlas genetics are two rare chr4 gene-desert variants","assessment":"Glaucoma is not an indication for atorvastatin and is not listed anywhere in the US label, whose only ocular entry under Special Senses is \"vision blurred\"; the label's indications are lipid lowering and cardiovascular risk reduction. The direction of the drug effect is genuinely contested rather than settled: a 2025 nationwide Taiwanese nested case-control study (32,640 ocular hypertension and 20,390 POAG cases) reported statin users at modestly higher risk of ocular hypertension (aOR 1.12, 95% CI 1.09-1.15) and POAG (aOR 1.07, 95% CI 1.04-1.11), explicitly singling out atorvastatin and rosuvastatin with a duration-response trend, and a 2025 network meta-analysis found increased glaucoma onset for rosuvastatin, simvastatin and pravastatin - but not atorvastatin. Against that, the 2017 JAMA Ophthalmology cohort found continuous statin use protective (adjusted HR 0.79, 95% CI 0.66-0.96), a Japanese claims nested case-control was flatly null (aOR 0.97-0.98), the most-cited protective JAMA Ophthalmology paper was retracted in 2020, and in a TGF-beta2 ocular hypertension mouse model oral atorvastatin lowered IOP from 32.3 to 15.4 mmHg while suppressing trabecular meshwork ECM - i.e. the best mechanistic data point the other way. The atlas's own epidemiology is unremarkable: glaucoma is present in only 0.71% of atorvastatin users and is very slightly depleted relative to users of other drugs (enrichment 0.94), so there is no confounding-by-indication signature, but equally no population-level excess; FAERS is weak (1/3 arms, PRR 1.17) and the condition is absent from the BNF ADR list. Temporality of 89.1% over 221 dated participants with a 2.47-year median lag is exactly what age plus prescription-record-depth would produce in a statin-taking cohort in their sixties, and cannot separate a drug effect from the detection bias that also plagues the Taiwanese claims result, where statin users have far more routine physician contact. The clinical question is live enough to be worth watching, but this particular variant signal does not advance it.","gene_comment":"The fact sheet assigns no gene, and correctly so: rs190862969 (chr4:130,925,968) and rs368321957 (chr4:131,232,615) sit 307 kb apart in a 4q28.3 gene desert whose only annotated content is unnamed lncRNAs, processed pseudogenes and an RNU6 pseudogene, with the nearest named features LINC02479 and LINC02377 hundreds of kilobases away. Both are rare (gnomAD MAF ~0.3%) and at almost identical frequency, so they most likely tag one rare haplotype rather than two independent hits, and no mechanism can be read off them.","score_check":"A beta of 3.967 (OR ~50) for a common late-onset polygenic disease at ~0.3% allele frequency is the classic sparse-data signature the atlas warns about, and the same variants are pure noise in the drug-free population (beta -0.054 +/- 0.184, log10p 0.11), so the z_diff of 5.47 is carried entirely by a handful of carriers on the treated side. The prescription-propensity arm is null (log10p 0.27) and enrichment is 0.94, which rules out indication bias but leaves the effect size itself not believable at face value.","candidability":4,"candidability_reason":"Atorvastatin-specific glaucoma risk has recent but contested and detection-bias-prone support at aOR ~1.1, while this atlas hit is an unreplicated rare intergenic gene-desert variant with an implausibly large effect and no signal in the disease arm.","sources":[{"title":"Association Between Statin Use and Glaucoma Risk: A Population-Based Study (Invest Ophthalmol Vis Sci 2025, PMID 41328992)","url":"https://pubmed.ncbi.nlm.nih.gov/41328992/"},{"title":"Associations of different types of statins with the risk of open-angle glaucoma: systematic review and network meta-analysis (Graefes Arch Clin Exp Ophthalmol 2025, PMID 39212799)","url":"https://pubmed.ncbi.nlm.nih.gov/39212799/"},{"title":"Association of Daily Dosage and Type of Statin Agent With Risk of Open-Angle Glaucoma (JAMA Ophthalmol 2017, PMID 28114645)","url":"https://pubmed.ncbi.nlm.nih.gov/28114645/"},{"title":"Association between statin use and open-angle glaucoma: nested case-control study, Japanese claims database (Sci Rep 2023, PMID 37468563)","url":"https://pubmed.ncbi.nlm.nih.gov/37468563/"},{"title":"Atorvastatin reduces IOP in ocular hypertension in vivo and suppresses ECM in trabecular meshwork (Int J Mol Med 2022, PMID 35417030)","url":"https://pubmed.ncbi.nlm.nih.gov/35417030/"},{"title":"Atorvastatin calcium tablets - US prescribing information (DailyMed SPL)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=634ec8e3-6d83-4cb2-90a3-fc9c973b06bf"},{"title":"dbSNP records for rs190862969 and rs368321957 (NCBI E-utilities esummary)","url":"https://www.ncbi.nlm.nih.gov/snp/rs190862969"},{"title":"Ensembl REST overlap, human chr4:130,400,000-131,800,000 gene annotation","url":"https://rest.ensembl.org/overlap/region/human/4:130400000-131800000?feature=gene;content-type=application/json"}]},"atorvastatin|H919":{"verdict":"contested","headline":"Statin-related hearing loss is real but disputed in direction; the A2ML1 variant here is rare and oversized","assessment":"Hearing loss is not an atorvastatin indication and does not appear in the US label's adverse-reaction tables; the only auditory term in the label is postmarketing 'tinnitus' under Special senses, so H91.9 is at most an off-label safety signal. The literature genuinely disagrees on direction: an All of Us analysis of 90,271 hyperlipidaemia patients reported statin use with hearing loss OR 1.60 and atorvastatin specifically OR 1.27, an FAERS disproportionality study flagged hearing impairment for atorvastatin/ezetimibe (ROR 10.76), and a 2026 case report describes bilateral SNHL three weeks after atorvastatin 40 mg that resolved on withdrawal; against that, statins were protective for sudden SNHL in type 2 diabetes and associated with better mid-to-high-frequency hearing at ages 55-74. A drug-caused mechanism is therefore plausible but not established, and the competing cochlear-microvascular argument predicts benefit, not harm. The atlas's population signals are weak rather than supportive: enrichment 1.25 is the modest excess expected in an older, cardiometabolic population, cotx_pct 1.57% is far below true age-specific hearing-loss prevalence so the phenotype is coding-driven, and temporality 89.1% with a 3.19-year median lag is the uninformative high value the cohort's record structure inflates. The prescription-propensity arm is flat (log10p 0.16), which at least rules out a channelling artifact, and the same variant is only borderline in the drug-free population (beta 0.28, se 0.142, z 1.97). The z_diff of 4.76 rests entirely on a beta_adr of 3.086 in the treated arm, which is the shape of sparse-data bias, not a discovered interaction.","gene_comment":"rs116952594 is an intronic variant at chr12:8,831,374 inside A2ML1 (ENSG00000166535, protein-coding), so the gene call is positionally correct, and A2ML1 has a real if narrow otological literature - spontaneous otitis media in A2ml1-knockout mice (OR 11) and A2ML1 variants associated with 4 kHz hearing loss in an indigenous Filipino cohort. But that is conductive/middle-ear susceptibility with no described link to statin pharmacology, HMG-CoA reductase, or cochlear drug handling, and the allele is rare (gnomAD ~0.27% global, ~0.5% European), so it should not be presented as a mechanism for a statin ADR.","score_check":"A beta of 3.086 (OR ~22) on a ~0.5%-frequency intronic allele in a within-drug case set whose dated denominator is only 377 is almost certainly sparse-data inflation rather than a true effect of that size. The disease-arm estimate (beta 0.28, se 0.142) is barely distinguishable from noise, so the large z_diff is driven by the unstable treated-arm term, not by a credible difference between arms.","candidability":5,"candidability_reason":"Statin-associated hearing loss has label, FAERS and case-report support but contested direction, and the specific rare A2ML1 variant here looks like sparse-data inflation on an age-confounded phenotype.","sources":[{"title":"openFDA drug label API - atorvastatin prescribing information (indications; postmarketing 'Special senses: vision blurred, tinnitus')","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22atorvastatin%22&limit=1"},{"title":"Europe PMC search: atorvastatin AND sudden sensorineural hearing loss / ototoxicity / tinnitus (Homer 2026 All of Us; Li 2024 FAERS; Dari 2026 case report)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=atorvastatin%20AND%20(%22sudden%20sensorineural%20hearing%20loss%22%20OR%20ototoxicity%20OR%20tinnitus)&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC search: statin AND hearing loss (Maldonado 2026 Drugs & Aging; Li 2025 Otolaryngol Head Neck Surg; Zeng 2026 Expert Opin Drug Saf)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=statin%20AND%20%22hearing%20loss%22&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC search: A2ML1 AND otitis media/hearing (A2ml1-knockout mouse OM model, PMID 38759260; A2ML1/FUT2 audiologic measures in an indigenous community, PMID 36719978)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=A2ML1%20AND%20(%22otitis%20media%22%20OR%20hearing)&format=json&pageSize=25&resultType=core"},{"title":"Ensembl REST - rs116952594 (chr12:8,831,374, intron variant, gnomAD MAF ~0.0027)","url":"https://rest.ensembl.org/variation/human/rs116952594?content-type=application/json&pops=1"},{"title":"Ensembl REST - genes overlapping chr12:8,831,374 (A2ML1, ENSG00000166535)","url":"https://rest.ensembl.org/overlap/region/human/12:8831374-8831374?feature=gene;content-type=application/json"},{"title":"Europe PMC search: hyperlipidaemia and hearing loss risk (Peng 2026 bidirectional MR - no causal link for hyperlipidaemia and SSNHL)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22hyperlipidemia%22%20AND%20%22hearing%20loss%22%20AND%20risk&format=json&pageSize=20&resultType=core"}]},"atorvastatin|H931":{"verdict":"contested","headline":"Tinnitus is a labelled atorvastatin reaction, but this rare intronic RABGAP1L hit is sparse-data noise","assessment":"Tinnitus is not an indication for atorvastatin - the label covers hyperlipidaemia and cardiovascular risk reduction - but it is an explicitly labelled adverse reaction: the US atorvastatin calcium prescribing information lists 'Special Senses: vision blurred, tinnitus' among adverse reactions reported in placebo-controlled trials, and postmarketing experience adds dizziness. A published case describes progressive tinnitus from six months and irreversible 'cookie-bite' bilateral hearing loss at eighteen months on atorvastatin 20 mg, scored 'possible' on the Naranjo scale, with the manufacturer acknowledging three unpublished deafness reports. A 2026 All of Us analysis of over 90,000 hyperlipidaemic patients, adjusted for age, race, sex, hypertension and diabetes, found statin use associated with modestly increased odds of sensorineural hearing loss and tinnitus (atorvastatin around OR 1.2-1.3). The direction is genuinely contested, however: statins have been trialled as otoprotective agents, a small series used atorvastatin to treat tinnitus in hyperlipidaemic patients, and a diabetic cohort reported roughly 25% lower hearing-loss/tinnitus risk on statins - so a class effect that pushes tinnitus the other way cannot be excluded. Mechanistically an inner-ear cholesterol/strial microvascular effect is speculative rather than described; no ototoxic pathway for statins is established. The atlas numbers are compatible with a weak real effect at the population level (enrichment 1.11, FAERS PRR 1.92, onset a median 3.5 years after first exposure) but the specific variant signal is not credible as pharmacogenomics.","gene_comment":"rs138349984 is an intronic variant in RABGAP1L (1q25.1, a Rab GTPase-activating protein of endosomal/Golgi trafficking) with non-Finnish European allele frequency around 0.5%; the gene has no known cochlear, otologic or lipid-pathway role and the intronic position gives no mechanistic handle, so this is a locus label rather than a candidate mechanism.","score_check":"beta_adr = 6.27 on a log-odds scale is an odds ratio of several hundred for a 0.5%-frequency intronic variant against a 0.86% outcome - a near-separation, sparse-data artifact rather than a real effect size, and no ADR standard error is reported to check it. The comparison arms behave sensibly (beta_disease 0.39 +/- 0.214 is within noise, prescription arm null), so z_diff 5.24 is driven entirely by the inflated treated-arm estimate; temporality 89.6% on 134 dated participants is the weak direction the cohort inflates anyway.","candidability":5,"candidability_reason":"Tinnitus is a genuinely labelled atorvastatin adverse reaction with supporting cohort and case evidence, but the direction is contested in the literature and this particular rare intronic RABGAP1L signal is a sparse-data artifact with no plausible mechanism.","sources":[{"title":"Atorvastatin calcium tablets, film coated - full prescribing information (DailyMed, Accord Healthcare)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=4e06b20d-cc5d-46eb-991b-5972248de3af"},{"title":"PubMed search results: statin tinnitus (studies on statins, hearing loss and tinnitus)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+tinnitus"},{"title":"Statins and Their Effect on Hearing: An All of Us Database Study (Ann Otol Rhinol Laryngol, 2026; PMID 40728037)","url":"https://europepmc.org/abstract/MED/40728037"},{"title":"Irreversible atorvastatin-associated hearing loss (Pharmacotherapy, 2012; PMID 22392429)","url":"https://europepmc.org/abstract/MED/22392429"},{"title":"Ensembl REST variation record for rs138349984 (intronic RABGAP1L, population frequencies)","url":"https://rest.ensembl.org/variation/human/rs138349984?content-type=application/json;pops=1"},{"title":"MyGene.info record for RABGAP1L (RAB GTPase activating protein 1 like, 1q25.1)","url":"https://mygene.info/v3/query?q=RABGAP1L&species=human&fields=name,summary,genomic_pos,map_location,type_of_gene"}]},"atorvastatin|K30":{"verdict":"statistical artifact","headline":"Dyspepsia is label-listed but at placebo rates; the signal rests on a rare intronic SEL1L3 variant","assessment":"Dyspepsia is not an indication for atorvastatin and is not a co-prescription marker either: the atlas gives enrichment exactly 1.00, so K30 is no more common among atorvastatin users than among users of other drugs, which for a symptom this common in primary care is the expected null. It is a label-listed adverse reaction, but the US prescribing information puts dyspepsia at 4.7% on atorvastatin versus 4.3% on placebo (diarrhoea 6.8 vs 6.3%, nausea 4.0 vs 3.5%), i.e. a background-rate event carried in the table rather than a demonstrated drug effect, and it does not appear in the BNF side-effect list here (on_bnf=0). A drug-caused mechanism is not described and the upper-GI literature points, if anywhere, the other way: a Taiwanese cohort of 48,562 hyperlipidaemic patients found significantly lower peptic ulcer incidence in statin users with a dose-response, a systematic review pooled OR 0.89 (95% CI 0.67-1.18) for peptic ulcer in statin users, and an endoscopic case-control study found no association with ulcer (OR 1.2, 0.7-2.1) or reflux oesophagitis. The FAERS support is explicitly weak (1 of 3, PRR 1.7), which is what an unremarkable, ubiquitous symptom term looks like in spontaneous reports. The temporality of 100% after first exposure across 364 dated cases is uninformative given the cohort's known upward compression, and a median 2.89 years to onset is far too long for a drug-symptom link that would normally declare itself in weeks. What is left is a single rare non-coding variant with beta 2.19 (OR about 9) on a common symptom in roughly 364 cases, a null disease arm (beta 0.089 +/- 0.096) and a null prescribing arm - a shape far more consistent with a sparse-data extreme than with a real pharmacogenomic effect.","gene_comment":"rs548799122 is a rare C/T variant at chr4:25,747,317 (GRCh38, gnomAD MAF ~0.9% overall and only ~0.2% in 1000 Genomes, enriched to ~2.7% in Finns) that falls inside the SEL1L3 intron span and is annotated by Ensembl as non-coding/intergenic in consequence, so the gene call is positional rather than functional. SEL1L3 is a poorly characterised single-pass membrane Sel1-repeat protein with lymphoid-enhanced expression, no established disease link and fewer than 30 PubMed records, and no connection to gastric acid, motility or statin pharmacology - it should not be presented as mechanism.","score_check":"The numbers do not hang together: with ~364 cases and a ~0.5-1% minor allele frequency only a handful of carrier cases can exist, so an OR near 9 at log10p 6.25 is the classic sparse-data inflation pattern, and z_diff 4.69 is driven by that inflated within-user beta rather than by any real divergence from the flat disease-arm effect. The one internally consistent number is enrichment 1.0, which correctly says dyspepsia is simply as common in statin users as anywhere else.","candidability":2,"candidability_reason":"Label-listed at placebo-equal frequency, no enrichment, mechanism if anything protective, and the genetic signal is a rare intronic variant with a sparse-data effect size.","sources":[{"title":"Atorvastatin calcium tablets - full prescribing information (DailyMed), Adverse Reactions Table 1","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=73cefb0b-1747-4fe0-b252-377f4ed94c6f"},{"title":"Wijarnpreecha et al., Statins and risk of peptic ulcer disease: a systematic review and meta-analysis, Arab J Gastroenterol 2020 (PMID 32830090)","url":"https://pubmed.ncbi.nlm.nih.gov/32830090/"},{"title":"Feng et al., The effect of statins on the occurrence of peptic ulcer, Eur J Intern Med 2015 (PMID 26226858)","url":"https://pubmed.ncbi.nlm.nih.gov/26226858/"},{"title":"Fujii et al., Statin use and risk of gastroduodenal ulcer and reflux esophagitis, Hepatogastroenterology 2009 (PMID 19621671)","url":"https://pubmed.ncbi.nlm.nih.gov/19621671/"},{"title":"Ensembl REST variation record for rs548799122 (chr4:25,747,317, consequence and population frequencies)","url":"https://rest.ensembl.org/variation/human/rs548799122?content-type=application/json;pops=1"},{"title":"Ensembl gene record for SEL1L3 (chr4:25,713,965-25,863,760) and overlapping-gene listing for the locus","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/SEL1L3?content-type=application/json"},{"title":"UniProtKB Q68CR1 - Protein sel-1 homolog 3 (SEL1L3), human","url":"https://rest.uniprot.org/uniprotkb/search?query=gene:SEL1L3+AND+organism_id:9606&format=txt"}]},"atorvastatin|K589":{"verdict":"statistical artifact","headline":"IBS is not depleted-not-enriched in atorvastatin users; the signal is one rare lncRNA intron variant","assessment":"IBS (K58.9) is not an atorvastatin indication and is not a labelled adverse reaction: the US label lists only non-specific GI events at essentially placebo rates (diarrhoea 6.8% vs 6.3%, dyspepsia 4.7% vs 4.3%, nausea 4.0% vs 3.5%) plus postmarketing pancreatitis, and no functional bowel disorder. The atlas itself shows no excess: 1.11% of atorvastatin users carry the code with enrichment 0.94, i.e. IBS is marginally less frequent than among users of other drugs, so there is neither confounding by indication nor a population excess to explain. What literature exists points the other way - in 438,805 UK Biobank participants regular statin use was associated with lower incident IBS in men (HR 0.77, 95% CI 0.61-0.97) and null in women, lovastatin has been formulated deliberately to treat IBS-C via intestinal methanogens, and atorvastatin 40-80 mg is currently in a randomised crossover trial (BASTA, NCT07042165) as a treatment for bile acid diarrhoea, a condition overlapping IBS-D. A drug-caused mechanism for statins inducing IBS is therefore neither described nor directionally supported; FAERS agrees (no signal, PRR 0.8) and the drug is not BNF-flagged for this. Temporality of 82.9% over 257 dated participants with a 2.82-year median is exactly what the cohort's record-depth asymmetry produces for a common ageing-population diagnosis, and it carries no weight against a null enrichment. The whole association rests on a single variant whose effect estimate is not credible, so the reading is an artifact rather than a real pharmacogenomic effect.","gene_comment":"rs111407300 is a rare intronic variant (Ensembl MAF ~0.0018, C/T at chr3:194,777,159) inside LINC01968, a long intergenic non-coding RNA at 3q29 with no annotated function, no statin-pharmacology role and no relation to established IBS loci. This is a weak gene assignment that cannot be presented as mechanism.","score_check":"beta_adr 3.62 (OR ~37) on a variant with ~0.2% minor allele frequency and only a few hundred cases means a handful of carriers are driving log10p 8.77 - the classic sparse-data separation pattern, and z_diff 5.7 inherits the same inflated estimate. The comparison arms are consistently null (beta_disease 0.147 +/- 0.103 is well inside noise, log10p_disease 0.82, log10p_prescribed 0.12), so nothing independent corroborates the within-user hit.","candidability":1,"candidability_reason":"No enrichment, no label or FAERS signal, literature pointing the opposite way, and a rare lncRNA intron variant with an implausibly large effect.","sources":[{"title":"Atorvastatin calcium tablets - DailyMed label (indications and adverse reactions)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=73cefb0b-1747-4fe0-b252-377f4ed94c6f"},{"title":"Gender-specific association between the regular use of statins and the risk of irritable bowel syndrome: a population-based prospective cohort study (Front Pharmacol; PMID 36686671)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:36686671&resultType=core&format=json"},{"title":"BASTA trial: atorvastatin for bile acid diarrhoea, randomised placebo-controlled crossover protocol (BMJ Open 2026; PMID 42521312)","url":"https://doi.org/10.1136/bmjopen-2026-119827"},{"title":"PubMed: statins and irritable bowel syndrome (incl. lovastatin lactone for IBS-C, PMIDs 27347377, 28989013)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+irritable+bowel+syndrome"},{"title":"Ensembl REST: rs111407300 variant annotation (intron variant, MAF 0.0018)","url":"https://rest.ensembl.org/variation/human/rs111407300?content-type=application/json"},{"title":"Ensembl REST: LINC01968 gene record (lncRNA, chromosome 3)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/LINC01968?content-type=application/json"}]},"atorvastatin|K760":{"verdict":"co-prescription population","headline":"Fatty liver tracks the dyslipidaemia statins treat; statins improve it, and the variant effect is absurd","assessment":"K76.0 is not a licensed indication for atorvastatin: the DailyMed label covers cardiovascular risk reduction and LDL-C lowering only, and its hepatic content is limited to transaminase elevations (>3x ULN in ~0.7%) plus rare postmarketing hepatic failure, with hepatic steatosis, fatty liver and NAFLD absent from the label entirely. LiverTox likewise describes a hepatocellular/cholestatic idiosyncratic injury at roughly one serious case per million person-years and states that statins can be safely used in patients with nonalcoholic liver disease, never listing steatosis as a statin lesion. The direction of drug effect is the decisive point: a 2023 meta-analysis of 14 RCTs in NAFLD (Medicine, PMID 37390233) found statins lowered ALT (MD -5.53), AST (-3.43) and GGT (-4.05) alongside triglycerides and LDL-C, so the drug moves this condition the right way, not the wrong way. What the atlas is seeing is the shared metabolic-syndrome axis behind MASLD and the atherogenic dyslipidaemia that earns an atorvastatin prescription, amplified by detection bias, because starting a statin triggers LFT checks and abdominal imaging that generate a K76.0 code. The atlas numbers fit that reading rather than an ADR one: enrichment is only 1.17, temporality of 96.8% is exactly what surveillance-after-prescription produces and is the uninformative direction, and the FAERS PRR of 2.54 reflects the same comorbid population and statin liver-reporting notoriety. Nothing here supports a drug-caused steatosis; the co-occurrence is confounding by indication with an ascertainment overlay.","gene_comment":"rs142961571 is a rare 5'UTR variant at chr1:15,757,809 (global MAF ~0.2%, up to ~2% in non-Finnish Europeans) that does fall inside FBLIM1, so the positional call is fair, but FBLIM1/migfilin is a filamin-binding focal-adhesion protein with no hepatic or lipid biology in UniProt and no NAFLD-GWAS pedigree. None of the established steatosis loci (PNPLA3, TM6SF2, HSD17B13, MBOAT7) appears, so there is no mechanism to attach to this hit.","score_check":"beta_adr of 6.69 is an odds ratio near 800 for a common metabolic diagnosis on a variant carried by ~1% of people, which is a sparse-data/quasi-separation signature rather than a real effect, however small the p-value looks. The disease arm is flatly null (beta 0.054 +/- 0.223, log10p 0.09), so z_diff of 6.18 is carried entirely by the implausible treated-arm estimate.","candidability":1,"candidability_reason":"Confounding by indication and surveillance bias for a condition statins are shown to improve, on a rare variant whose effect size is a sparse-data artifact.","sources":[{"title":"Atorvastatin Calcium tablet label (DailyMed, Rising Pharma Holdings)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=73cefb0b-1747-4fe0-b252-377f4ed94c6f"},{"title":"Statins - LiverTox: Clinical and Research Information on Drug-Induced Liver Injury (NCBI Bookshelf NBK548067)","url":"https://www.ncbi.nlm.nih.gov/books/NBK548067/"},{"title":"Statins on nonalcoholic fatty liver disease: a systematic review and meta-analysis of 14 RCTs (Medicine, 2023; PMID 37390233)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22Statins%20on%20nonalcoholic%20fatty%20liver%20disease%22%20meta-analysis&resultType=core&format=json&pageSize=2"},{"title":"Ensembl REST variation record for rs142961571 (position, consequence, population frequencies)","url":"https://rest.ensembl.org/variation/human/rs142961571?content-type=application/json;pops=1"},{"title":"UniProt Q8WUP2 - Filamin-binding LIM protein 1 (FBLIM1 / migfilin)","url":"https://rest.uniprot.org/uniprotkb/Q8WUP2.json?fields=protein_name,gene_names,cc_function,cc_tissue_specificity,cc_disease"},{"title":"EASL-EASD-EASO Clinical Practice Guidelines on the management of MASLD, Executive Summary (Diabetologia 2024; PMID 38869512)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22EASL-EASD-EASO%20Clinical%20Practice%20Guidelines%20on%20the%20management%20of%20metabolic%20dysfunction-associated%20steatotic%20liver%20disease%22&resultType=core&format=json&pageSize=2"}]},"atorvastatin|L918":{"verdict":"statistical artifact","headline":"Statins are antifibrotic in skin; this sub-threshold X/chr9 lncRNA hit for L91.8 runs backwards","assessment":"L91.8 (other hypertrophic disorders of skin, the ICD-10 L91 keloid/hypertrophic-scar family) is neither an indication for atorvastatin nor a labelled adverse reaction: the Lipitor label lists only urticaria in trials and angioneurotic oedema plus bullous rashes (erythema multiforme, SJS, TEN) post-marketing, and the recognised statin skin reactions in the literature are eczematous, photosensitive, lupus-like and bullous, not fibrotic or hypertrophic. The pharmacology in fact points the other way: simvastatin inhibits TGF-beta1-driven type I collagen expression and keloid fibroblast proliferation (Mun 2014, PMID 24471776; Chen 2016, PMID 30024693), pravastatin reverses established radiation-induced cutaneous fibrosis (PMID 30776452), and atorvastatin itself alleviates fibrosis in experimental systemic sclerosis (PMID 41443333) - so statins are being explored as anti-scarring agents, which is decisive evidence against a drug-caused hypertrophic-scar reading. The atlas signal is also internally weak: log10p_adr 6.24 is p~5.7e-7, below genome-wide significance, resting on 117 datable participants and 0.34% co-occurrence with no background arm to compute enrichment, so the honest confounding test was never run. The disease arm is a clean null (beta 0.076 +/- 0.089, well inside noise) and the prescription arm is null too, so z_diff 4.7 is driven entirely by the treated-only estimate rather than by any contrast with a real disease effect. Temporality of 89.7% is the uninformative direction given the cohort's known compression of prescription records earlier than diagnosis records, and FAERS is weak (PRR 1.75, 1 of 3). Nothing here separates a rare-variant sparse-data fluctuation in a small subgroup from a drug effect.","gene_comment":"ENSG00000288098 is an unnamed 'novel transcript' lncRNA spanning ~800 kb of Xq27 across the MAGEC1/MAGEC2 cancer-testis antigen and pseudogene field - no dermal or fibrosis biology, and X-linked hemizygosity in a male-skewed statin cohort inflates rare-variant effects. Worse, the two variants are not even in the same locus: rs190148947 is at X:142,016,244 (NFE MAF ~1%) inside that lncRNA, while rs76043071 is at 9:98,724,864, intergenic between GABBR2 and ANKS6, so the single gene label cannot describe both.","score_check":"beta_adr 2.053 (OR ~7.8) on ~117 dated cases and a ~1-3% minor allele is the classic shape of a sparse-cell overestimate, and the accompanying disease-arm beta of 0.076 +/- 0.089 is indistinguishable from zero. The numbers are arithmetically consistent but the effect size is not believable as a true magnitude.","candidability":1,"candidability_reason":"Sub-genome-wide, sparse, mechanistically unsupported hit for a condition statins are documented to improve rather than cause.","sources":[{"title":"LIPITOR (atorvastatin calcium) tablets - full prescribing information, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a60cc18b-0631-4cf0-b021-9f52224ece65"},{"title":"PubMed: simvastatin keloid (Mun 2014 PMID 24471776; Chen 2016 PMID 30024693; Zhuang 2025 PMID 39313951)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=simvastatin+keloid"},{"title":"PubMed: statins skin fibrosis antifibrotic (atorvastatin in systemic sclerosis PMID 41443333; pravastatin reverses radiation fibrosis PMID 30776452)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statins+skin+fibrosis+antifibrotic"},{"title":"PubMed: statin cutaneous adverse reactions review","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+cutaneous+adverse+reactions+review"},{"title":"Ensembl REST lookup: ENSG00000288098 (lncRNA, novel transcript, X:141,934,496-142,740,581)","url":"https://rest.ensembl.org/lookup/id/ENSG00000288098?content-type=application/json"},{"title":"Ensembl REST variation: rs190148947 (X:142,016,244) and rs76043071 (9:98,724,864) with gnomAD frequencies","url":"https://rest.ensembl.org/variation/human/rs190148947?content-type=application/json;pops=1"}]},"atorvastatin|M720":{"verdict":"co-prescription population","headline":"Dupuytren shares diabetes/age/male-sex risk with statin users; statins are antifibrotic, not profibrotic","assessment":"Palmar fascial fibromatosis (Dupuytren disease) is not an indication for atorvastatin, whose licensed uses are cardiovascular risk reduction and lipid lowering, and it appears nowhere in the US prescribing information: the connective-tissue reactions listed are myopathy, myositis and rhabdomyolysis, not fibromatosis. The two populations overlap heavily for reasons that have nothing to do with the drug: in a UK Biobank case-control study of 4,148 cases, Dupuytren was driven by male sex (OR 3.23), age (OR 1.08/year) and diabetes with complications (OR 2.59, with HbA1c dose-response) plus smoking and alcohol, and atorvastatin is prescribed almost universally to older diabetic men. Direction of effect argues against an ADR reading: statins inhibit fibroblast-to-myofibroblast transition through mevalonate-pathway/GTPase-prenylation blockade and reduced TGF-beta1, atorvastatin specifically attenuates human cardiac fibroblast activation and alpha-SMA expression, and simvastatin has been proposed in the literature as a *therapy* for Dupuytren contracture rather than a cause. Pharmacovigilance does support a statin signal for tendon and ligament disorders in FAERS, but no fibromatosis or Dupuytren signal is reported, and this atlas row itself has faers = n_too_low. The atlas gives no enrichment (no background arm), so the honest test of confounding by indication was never run here; temporality of 89.1% on 175 dated participants with a 3.09-year median lag is exactly the high value the cohort's prescription-before-diagnosis compression produces and is weak evidence on its own. The prescription arm is flat (log10p 0.22), which rules out the variant tagging propensity to be treated but does nothing to rule out the clinical comorbidity route.","gene_comment":"rs560768493 is a rare (MAF ~0.1-0.4%) intergenic T/C variant at chr12:61,170,760 in a 12q14.1 gene desert whose only nearby annotation is a processed pseudogene, and the fact sheet assigns no gene at all. It sits nowhere near the established Dupuytren loci (WNT7B, EPDR1, and the other WNT-pathway signals from the published GWAS), so there is no mechanistic story to attach to it.","score_check":"beta_adr 4.215 (OR ~68) for a variant this rare in a subgroup with only a few hundred coded cases is a sparse-data artifact, not a real effect size, even at log10p 6.67; the drug-free arm is the more believable number and it is only nominal (0.485/0.177, z = 2.74, log10p 2.21). z_diff 4.48 is therefore driven almost entirely by the inflated within-users beta rather than by a genuine treated-versus-untreated divergence.","candidability":2,"candidability_reason":"Shared-risk-factor comorbidity with an unmeasured background arm, a gene-desert rare variant with a sparse-data beta, and a mechanism that runs the opposite way.","sources":[{"title":"Atorvastatin calcium tablets - prescribing information (DailyMed)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=73cefb0b-1747-4fe0-b252-377f4ed94c6f"},{"title":"Modifiable Risk Factors for Prevention in Dupuytren Disease: A UK Biobank Case-Control Study (PMID 37257135)","url":"https://www.ncbi.nlm.nih.gov/pubmed/37257135"},{"title":"Simvastatin may be a useful therapy in Dupuytren contracture. Med Hypotheses 2006 (PMID 16343796)","url":"https://pubmed.ncbi.nlm.nih.gov/16343796/"},{"title":"Genome-wide association study of Dupuytren disease: WNT7B and EPDR1 (PMID 39570219)","url":"https://www.ncbi.nlm.nih.gov/pubmed/39570219"},{"title":"Atorvastatin attenuates human cardiac fibroblast activation (alpha-SMA/myofibroblast transition) (PMID 42123723)","url":"https://www.ncbi.nlm.nih.gov/pubmed/42123723"},{"title":"Tendon and ligament disorders with lipid-modifying agents: FDA AERS disproportionality analysis (PMID 40400906)","url":"https://www.ncbi.nlm.nih.gov/pubmed/40400906"},{"title":"dbSNP record rs560768493 (chr12:61,170,760, intergenic)","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=snp&id=560768493&retmode=json"}]},"atorvastatin|R21":{"verdict":"plausible ADR","headline":"Rash is a labelled uncommon atorvastatin ADR, but this rare 22q11 segdup variant is not credible","assessment":"Rash is not an indication for atorvastatin (licensed for hypercholesterolaemia and cardiovascular prevention) but is an explicitly labelled adverse reaction: the UK SmPC lists urticaria, skin rash, pruritus and alopecia as uncommon (>=1/1000 to <1/100) and lichenoid drug reaction, bullous dermatitis, erythema multiforme, SJS and TEN as rare, and the US label carries urticaria plus post-marketing bullous rashes including SJS/TEN. The atlas numbers are consistent with that framing rather than with confounding by indication: cotx_pct 0.71% is close to the labelled 'uncommon' band, and enrichment 0.89 means R21 is if anything slightly less frequent among atorvastatin users than among users of other drugs, so patients are not being selected for rash. Against this, FAERS gives essentially no disproportionality (PRR 1.06, weak 1/3), which is expected for a high-background non-specific code like R21 but means the pharmacovigilance arm adds nothing. The temporality figure is uninformative: 100% post-exposure rests on only 65 datable participants with 66.8% undated, and a median of 4.84 years from first prescription to first rash record fits chronic background dermatology far better than a drug eruption, which typically appears within weeks to a few months of initiation. The comparison arms are clean in one respect - the variant is flat for rash in the drug-free population (log10p 0.14, beta 0.045 +/- 0.132, i.e. indistinguishable from zero) and flat for being prescribed the drug (log10p 0.91) - but that means the z_diff of 5.33 is carried entirely by the unstable within-user estimate. beta_adr 3.182 implies an odds ratio near 24 for a variant with ~0.5-1% minor allele frequency, an effect size seen for HLA-restricted severe cutaneous reactions and essentially nowhere else, and certainly not for a non-coding neuronal-receptor locus. So the drug-condition pair is real and the class effect is documented, but this particular variant does not support it.","gene_comment":"rs113110884 (chr22:20,273,295, C>T, MAF ~0.5% gnomAD / ~1-2% in northern European panels) falls inside the extended RTN4R transcript at 22q11.21; RTN4R encodes the Nogo-66 receptor mediating CNS axonal growth inhibition, with no skin, keratinocyte or immune role, and dbSNP assigns the variant no gene or functional class. Worse for the assignment, the position lies inside a segmental duplication (chr22:20,262,035-20,290,267, 95.8% identity to chr22:18.95 Mb), exactly the kind of region where rare-variant calls and imputation go wrong; the only replicated genetics for statin cutaneous reactions points at HLA alleles, not here.","score_check":"The disease and prescription arms are properly null and the enrichment argues against confounding by indication, so the design is not obviously broken, but an OR of ~24 for a ~0.5% allele resting on a handful of carriers in a segmental-duplication region is the classic signature of sparse-data inflation or miscalling rather than a real effect. The temporality denominator (n_dated 65, 66.8% undated) is too thin to add support.","candidability":4,"candidability_reason":"Genuine labelled ADR and not confounded by indication, but the specific locus is a rare non-coding variant in a segdup-prone 22q11 region with an implausible effect size, no FAERS signal and a 5-year median latency.","sources":[{"title":"Atorvastatin 10 mg film-coated tablets - SmPC (electronic medicines compendium)","url":"https://www.medicines.org.uk/emc/product/13672/smpc"},{"title":"Atorvastatin calcium tablet, film coated - DailyMed label (setid 25c8f92f-9cb6-482b-82e1-16ad7c94e7b3)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=25c8f92f-9cb6-482b-82e1-16ad7c94e7b3"},{"title":"Forouzan P et al. Atorvastatin-induced Lichenoid Drug Eruption: A Case Report and Review of Statin-associated Cutaneous Adverse Events. Cureus 2020 (PMID 32257699)","url":"https://pubmed.ncbi.nlm.nih.gov/32257699/"},{"title":"Lv M et al. Toxic epidermal necrolysis in a patient on atorvastatin therapy expressing human leukocyte antigen alleles. Medicine (Baltimore) 2021 (PMID 33546081)","url":"https://pubmed.ncbi.nlm.nih.gov/33546081/"},{"title":"dbSNP rs113110884 (allele frequencies, position, no gene/function assignment)","url":"https://www.ncbi.nlm.nih.gov/snp/rs113110884"},{"title":"NCBI Gene: RTN4R reticulon 4 receptor (Nogo-66 receptor), 22q11.21","url":"https://www.ncbi.nlm.nih.gov/gene/65078"},{"title":"UCSC genomicSuperDups track, chr22:20,263,000-20,283,000 (hg38) - variant lies in a 95.8%-identity segmental duplication","url":"https://api.genome.ucsc.edu/getData/track?genome=hg38;track=genomicSuperDups;chrom=chr22;start=20263000;end=20283000"}]},"atorvastatin|R529":{"verdict":"statistical artifact","headline":"Statin muscle pain is real, but R52.9 plus a rare RBMS1 intron variant is not that finding","assessment":"R52.9 'Pain, unspecified' is not a licensed indication for atorvastatin, whose approved uses are dyslipidaemia and cardiovascular risk reduction; pain-type events are, however, genuinely on the label, with arthralgia and pain in extremity listed as common and myopathy, rhabdomyolysis and immune-mediated necrotising myopathy as recognised serious reactions. A drug-caused mechanism therefore exists in principle - statin-associated musculoskeletal symptoms (SAMS) are the subject of a dedicated CPIC guideline built on impaired hepatic uptake via SLCO1B1 c.521T>C and raised systemic statin exposure - but the endpoint tested here is an unspecific symptom code, not myalgia (M79.1) or myopathy, so it cannot distinguish SAMS from back pain, headache or any other complaint. The atlas numbers argue against a drug effect rather than for one: enrichment is exactly 1.00 (coded pain is no commoner in atorvastatin users than in users of other drugs), the prescription-propensity arm is null (log10p 0.45), FAERS gives essentially no disproportionality (PRR 1.08), and the median 3.7-year gap to first pain record does not match SAMS, whose randomised excess is confined to year 1. The blinded evidence sets the ceiling on what could be detected: in the CTT individual-participant meta-analysis of 19 trials (n=123,940), muscle pain was reported by 27.1% on statin vs 26.6% on placebo, rate ratio 1.03, with only about 1 in 15 first-year reports attributable to the drug - so a diffuse pain code in an EHR is overwhelmingly non-drug noise. Meanwhile the same variant is already nominally associated with pain in the drug-free population (beta 0.819, se 0.265, z about 3.1), pointing to background predisposition rather than a drug-by-genotype interaction, even though z_diff is 4.5. Taken together this reads as a sparse-data within-users signal on a diluted outcome, not a pharmacogenomic ADR.","gene_comment":"rs145098727 is a rare intronic variant (chr2:160,292,652, gnomAD MAF about 0.4%) inside RBMS1, an RNA-binding protein gene whose GWAS record is metabolic and psychiatric (type 2 diabetes, beta-cell chromatin, IBD/depression overlap, glaucoma) with no link to statin disposition, muscle or nociception. It has no GWAS Catalog entry of its own, and the established statin-myopathy pharmacogenes are SLCO1B1, ABCG2 and CYP2C9 - so this assignment carries no mechanism and should not be presented as one.","score_check":"beta_adr of 6.48 on a variant with about 0.4% minor allele frequency and only 101 dated cases implies an effect size far beyond anything credible for a symptom code, the classic shape of quasi-separation in a sparse cell. The disease-arm effect (0.819 +/- 0.265) is the only estimate here with a believable magnitude, and it points to background predisposition rather than a drug interaction.","candidability":2,"candidability_reason":"Non-specific pain code with zero enrichment, no FAERS signal, 3.7-year latency, an implausibly large beta on a rare intronic variant in a gene with no statin or muscle mechanism.","sources":[{"title":"Atorvastatin - StatPearls (NCBI Bookshelf): indications and adverse effects","url":"https://www.ncbi.nlm.nih.gov/books/NBK430779/"},{"title":"CPIC Guideline for SLCO1B1, ABCG2 and CYP2C9 and Statin-Associated Musculoskeletal Symptoms (Clin Pharmacol Ther 2022)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9035072/"},{"title":"Effect of statin therapy on muscle symptoms: individual participant data meta-analysis of randomised double-blind trials (Lancet 2022, PMID 36049498) via Europe PMC","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22Effect%20of%20statin%20therapy%20on%20muscle%20symptoms%22&format=json&resultType=core&pageSize=5"},{"title":"Ensembl variation record for rs145098727 (location, consequence, allele frequencies)","url":"https://rest.ensembl.org/variation/human/rs145098727?content-type=application/json;pops=1"},{"title":"Europe PMC literature search: RBMS1 genome-wide association findings","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22RBMS1%22%20AND%20%22genome-wide%20association%22&format=json&resultType=core&pageSize=10"},{"title":"GWAS Catalog REST lookup for rs145098727 (returns 404 - no catalogued association)","url":"https://www.ebi.ac.uk/gwas/rest/api/singleNucleotidePolymorphisms/rs145098727"}]},"atorvastatin|Z955":{"verdict":"co-prescription population","headline":"Post-PCI status code, not a reaction: atorvastatin is licensed to reduce revascularization, not cause it","assessment":"Z95.5 is a WHO ICD-10 Chapter XXI status code, 'Presence of coronary angioplasty implant and graft' - it records that a patient has had PCI, not a disease and not a reaction to anything. The atorvastatin (Lipitor) US label states the drug is indicated 'To reduce the risk of: Myocardial infarction (MI), stroke, revascularization procedures, and angina' both in primary prevention with multiple CHD risk factors and in patients with clinically evident CHD, so the licensed effect of the drug on this exact endpoint is in the opposite direction to an ADR reading; contemporary practice reinforces this, with guideline-directed high-intensity statin after coronary intervention associated with fewer recurrent ASCVD events. The atlas numbers are consistent with a treated-population marker rather than harm: 3.91% of atorvastatin users carry Z95.5 with enrichment 0.69, i.e. no enrichment at all relative to users of other drugs (unsurprising, since the drug-taking comparator is itself full of cardiac patients while atorvastatin is also used broadly in primary prevention). The temporality of 99.6% after first exposure is structurally guaranteed and carries no information here: a stent-presence code can only be entered at or after the procedure, statins are typically started years earlier for dyslipidaemia, and the atlas already warns that prescription records reach further back than diagnoses. The only reading under which the signal is not simply indication is 'incident revascularization despite statin therapy', i.e. residual/progressive atherosclerotic risk or statin non-response - a legitimate pharmacogenomic question, but the opposite of a drug-caused adverse event, and the variant offered does not support it. Confidence 95.7 with indication 40.2 and predisposition 34.3 already points the classifier toward the treated population rather than toward toxicity.","gene_comment":"rs116987503 is a rare intronic variant (MAF ~0.6%, but ~5.7% in Ashkenazi Jewish samples) sitting inside PCDH15, a calcium-dependent adhesion protein of inner-ear and retinal hair-cell mechanotransduction whose established phenotypes are Usher syndrome 1F and DFNB23; there is no cardiovascular or lipid biology attached to it. The assignment is positional within a very large gene, and the strong ancestry gradient of the allele makes population stratification a more parsimonious explanation than coronary mechanism.","score_check":"The within-user beta of 1.181 at log10p 7.48 for a 0.6% allele in ~1100 dated cases is the shape sparse-data and stratification artifacts take, and the background disease effect (beta 0.302, se 0.094, z=3.2) is nominal at best, so z_diff 3.76 rests on a fragile numerator. The null prescription-propensity signal (log10p 0.13) is reassuring but does not rescue the finding.","candidability":1,"candidability_reason":"A procedure-status code the drug is licensed to reduce, with no enrichment, structural temporality and a hearing-loss gene as its only genetic support.","sources":[{"title":"LIPITOR (atorvastatin calcium) tablets - FDA prescribing information, Indications and Usage","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2022/020702s079lbl.pdf"},{"title":"WHO ICD-10 (2019) Z95 - Presence of cardiac and vascular implants and grafts, incl. Z95.5","url":"https://icd.who.int/browse10/2019/en/JsonGetChildrenConcepts?ConceptId=Z95&useHtml=true"},{"title":"NLM Clinical Table Search Service - ICD-10-CM Z95.5 'Presence of coronary angioplasty implant and graft'","url":"https://clinicaltables.nlm.nih.gov/api/icd10cm/v3/search?sf=code,name&terms=Z95.5"},{"title":"UniProt Q96QU1 (PCDH15_HUMAN) - Protocadherin-15: function, tissue specificity, Usher syndrome 1F / DFNB23","url":"https://rest.uniprot.org/uniprotkb/Q96QU1.txt"},{"title":"Ensembl REST - variant rs116987503 (chr10:55,628,926, intronic, MAF 0.0063)","url":"https://rest.ensembl.org/variation/human/rs116987503?content-type=application/json;pops=1"},{"title":"Europe PMC search: statin intensity and outcomes after percutaneous coronary intervention (incl. JAHA 2026, PMID 41804923)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22high-intensity%20statin%22%20AND%20%22percutaneous%20coronary%20intervention%22%20AND%20secondary%20prevention%20guideline&format=json&pageSize=8&resultType=core"},{"title":"Europe PMC search: PCDH15 and cardiovascular/coronary/lipid phenotypes (no relevant associations)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=PCDH15%20AND%20(cardiovascular%20OR%20coronary%20OR%20lipid)&format=json&pageSize=10&resultType=core"}]},"bisoprolol|A099":{"verdict":"contested","headline":"Bisoprolol labels common GI upset, but A09.9 gastroenteritis rests here on a rare intergenic variant","assessment":"Gastroenteritis is not an indication for bisoprolol; the SmPC licenses it for hypertension, angina and stable chronic heart failure, and the US label for hypertension only. Non-specific gastrointestinal complaints are, however, a labelled ADR: the EU SmPC lists nausea, vomiting, diarrhoea and constipation as 'common', and the US label quantifies diarrhoea at 3.5% and nausea at 2.2%, with ischaemic colitis and mesenteric arterial thrombosis carried as rare class-level mentions for beta-blockers. That gives a partly plausible route to an A09.9 code, because A09.9 is a catch-all for gastroenteritis/colitis of unspecified origin and absorbs undifferentiated diarrhoea episodes as well as presumed infectious ones - so the atlas outcome is probably a mixture of drug-related loose stools, incidental infectious gastroenteritis, and occasional ischaemic colitis miscoded. A nationwide 57-million-patient study did report beta-blocker exposure as a risk factor for ischaemic colitis (OR 9.6, unadjusted and heavily confounded by the cardiovascular indication itself), which is directionally consistent but not evidence of causation. The atlas numbers are weak on every axis that would distinguish these: cotx_pct 2.62% with enrichment only 1.4 is what an older, comorbid cardiovascular population produces anyway; temporality of 100% over n=329 with a median 3.46-year lag is uninformative for an acute, recurrent event in chronically treated patients, since almost any acute episode falls after a long-standing prescription; and log10p_prescribed 0.43 at least rules out the variant tracking prescription propensity. The genetic layer is the part that does not hold up, and it is the part the finding depends on.","gene_comment":"rs55854447 has no gene assignment because it is intergenic at chr14:97,288,492 in a 14q32.2 gene desert - the nearest protein-coding gene, VRK1, ends ~330 kb away and everything closer is unnamed lncRNA, so there is no mechanistic candidate for gut injury here. It is also very rare (Ensembl MAF ~0.0018), which is exactly the configuration in which a beta of 2.311 should be read as sparse-data inflation rather than a ten-fold effect.","score_check":"log10p_adr 8.82 on a ~0.2%-frequency variant against a 2.6% outcome means the signal rests on a handful of carriers, and beta_disease 0.138 +/- 0.077 is under 1.96 SE, so the drug-free arm gives no independent support - the z_diff of 5.57 is driven almost entirely by the inflated within-users beta. The FAERS 'STRONG' flag with PRR 3.08 is consistent with the labelled GI complaints but gastroenteritis is a notoriously non-specific spontaneous-report term in an elderly polypharmacy population.","candidability":3,"candidability_reason":"Labelled GI upset makes the drug-condition pair non-absurd, but the modest enrichment fits the treated population and the rare intergenic variant with an unsupported disease arm is most likely sparse-data noise.","sources":[{"title":"Bisoprolol Fumarate Tablets, USP - US prescribing information (DailyMed, Aurobindo Pharma)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=2564646d-183c-4759-8726-155a4c22aa2b"},{"title":"Bisoprolol Fumarate 5 mg Film-Coated Tablets - SmPC (electronic medicines compendium, Sandoz)","url":"https://www.medicines.org.uk/emc/product/4622/smpc"},{"title":"Quantifying risk factors for ischemic colitis: a nationwide, retrospective cohort study (Indian J Gastroenterol, 2020; PMID 32797386)","url":"https://pubmed.ncbi.nlm.nih.gov/32797386/"},{"title":"Ensembl REST - variant record for rs55854447 (chr14:97288492, intergenic, MAF ~0.0018)","url":"https://rest.ensembl.org/variation/human/rs55854447?content-type=application/json"},{"title":"Ensembl REST - genes overlapping chr14:96,788,492-97,788,492 (VRK1 and lncRNAs only)","url":"https://rest.ensembl.org/overlap/region/human/14:96788492-97788492?feature=gene;content-type=application/json"}]},"bisoprolol|E872":{"verdict":"co-prescription population","headline":"Acidosis is not a bisoprolol label event; mostly HF/CKD comorbidity, with a rare real BRASH pathway","assessment":"Acidosis (E87.2) is neither a licensed indication for bisoprolol nor a labelled adverse reaction: the US bisoprolol fumarate label lists hypertension as the indication (EU labels add chronic heart failure and angina) and contains no mention of acidosis, lactic acidosis or ketoacidosis, consistent with on_bnf = 0. The population that receives bisoprolol - heart failure, ischaemic heart disease, hypertension with CKD and diabetes - is exactly the population in which metabolic acidosis is coded (CKD acidosis, cardiogenic hypoperfusion, diabetic ketoacidosis), and the atlas enrichment of 1.51 is the size of effect one expects from that comorbidity alone rather than from a drug effect. A genuine drug-caused route does exist but is narrow: BRASH syndrome (bradycardia, renal failure, AV-nodal blockade, shock, hyperkalaemia), in which beta-blockers are the implicated agent in about three-quarters of published cases and bisoprolol is named specifically, presents with severe metabolic acidosis (reported pH 6.86-7.19), as does frank beta-blocker overdose with cardiogenic shock. Against a general lactic-acidosis reading, beta-adrenergic blockade moves lactate the other way - catecholamines drive aerobic glycolysis and adrenergic antagonists reduce lactic acidosis in shock models - so any bisoprolol-attributable acidosis arises through bradycardia, hypoperfusion and renal failure, not through a metabolic push on lactate. The atlas temporality of 99% over 103 dated participants sits at the ceiling this cohort compresses everything toward and therefore adds nothing, and median 3 years to onset fits accumulating renal and cardiac disease rather than an acute drug event. The FAERS PRR of 4.37 is expected from exactly the overdose, shock and BRASH reports plus co-reported metformin, and does not independently support a common pharmacogenomic effect.","gene_comment":"MYSM1 (chr1 histone H2A deubiquitinase, bone marrow failure syndrome 4) and PAPPA2 (chr1 IGFBP-5 protease, IGF bioavailability and short stature) are real protein-coding genes, but neither touches renal acid excretion, lactate handling, potassium shift or beta-adrenergic signalling. With two low-frequency SNVs and no acid-base biology behind either, the gene assignment carries no mechanistic weight and should not be presented as one.","score_check":"beta_adr = 4.585 (odds ratio near 100) for a common-ish outcome in ~100 dated cases is the signature of sparse-data bias or quasi-separation at a rare variant, not a real effect size, and the same variants are flat in the drug-free arm (beta 0.317 +/- 0.214, |beta| < 1.96*se, log10p 0.86). The z_diff of 5.14 therefore reflects the implausible treated-arm estimate rather than a credible drug-by-genotype interaction.","candidability":3,"candidability_reason":"Real but rare drug-caused pathway (BRASH/overdose) sits behind an otherwise comorbidity-driven signal, and the genetics are an unbelievable sparse-data effect at genes with no acid-base biology.","sources":[{"title":"Bisoprolol fumarate tablet, film coated - DailyMed label (indications, warnings; no acidosis listed)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=200ab93b-6ad3-46cb-8f57-db056a14de16"},{"title":"PubMed: beta-blocker AND lactic acidosis (incl. Adrenergic antagonists reduce lactic acidosis in hemorrhagic shock, J Trauma 1999, PMID 10338406; Beta-blocker intoxication, Presse Med 2000, PMID 10862260)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=beta-blocker+lactic+acidosis"},{"title":"PubMed: bisoprolol AND acidosis (BRASH syndrome case reports, PMID 41431490 and PMID 40391372)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=bisoprolol+acidosis"},{"title":"Europe PMC: BRASH syndrome and acidosis - case series and systematic review (beta-blockers in 76.5% of cases, WJEM 2025 PMID 41246542; metabolic acidosis pH 6.86-7.19)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=BRASH%20syndrome%20acidosis&format=json&pageSize=10&resultType=core"},{"title":"MyGene.info records for PAPPA2 (pappalysin 2, IGFBP-5 protease) and MYSM1 (histone H2A deubiquitinase)","url":"https://mygene.info/v3/query?q=symbol:PAPPA2%20OR%20symbol:MYSM1&fields=symbol,name,summary,genomic_pos,type_of_gene&species=human"}]},"bisoprolol|G439":{"verdict":"statistical artifact","headline":"Bisoprolol prevents migraine in RCTs; a rare intronic ATP8A2 variant at OR~120 is a sparse-data artifact","assessment":"Migraine is not a licensed indication for bisoprolol - the SmPC lists only hypertension, angina pectoris and stable chronic heart failure - but bisoprolol is an evidence-supported off-label migraine preventive: a double-blind placebo-controlled trial in 226 migraineurs (Cephalalgia 1997) found bisoprolol 5 mg cut attack frequency by 39% versus 22% on placebo, and a 1992 double-blind study compared it directly with metoprolol, the beta-blocker class being long-standing guideline prophylaxis. That makes the direction of the real drug effect the opposite of an adverse reaction, which is decisive evidence against reading G43.9 here as bisoprolol-caused. The label does list headache and dizziness as common undesirable effects, but that is non-specific drug-onset headache, not incident migraine as a coded disorder, and migraine appears nowhere in section 4.8. The atlas numbers agree that these patients are not migraineurs by selection - cotx_pct 0.43% with enrichment 0.78 means migraine is actually depleted among bisoprolol users relative to users of other drugs, as expected for an older cardiac population - so this is not classic confounding by indication either; it is a signal with no clinical direction to sit in. The within-user hit is a single rare intronic SNV (rs181219039, gnomAD/Ensembl MAF ~0.12%) carrying beta 4.82, an odds ratio near 120, in a stratum with roughly 83 datable migraine cases: that is the signature of near-complete separation on a handful of carriers, not a real effect size, and log10p 6.53 does not even clear a genome-wide threshold. The same variant is flat for migraine in the drug-free population (beta 0.232, se 0.151, |beta| well under 1.96*se), FAERS support is weak (1/3, PRR 1.42), and the 75.9% temporality is the uninformative direction on a cohort that compresses this measure upward.","gene_comment":"rs181219039 is an intronic variant in ATP8A2 (chr13:25,972,430), a brain/retina P4-ATPase phosphatidylserine flippase whose biallelic loss causes recessive CAMRQ4 ataxia-intellectual disability; it has no published link to migraine, no relation to beta-1 adrenergic signalling, and an intronic rare allele here carries no functional interpretation. The gene assignment is positional only and must not be presented as mechanism.","score_check":"The within-drug p-value is sub-genome-wide and the effect size is implausible for a MAF-0.1% allele in ~83 cases, pointing to sparse-data bias rather than a large true effect. The null disease-arm beta and the enrichment below 1 are internally consistent and are the trustworthy numbers in this sheet.","candidability":1,"candidability_reason":"Bisoprolol reduces migraine in placebo-controlled trials, so the ADR direction is backwards, and the genetic hit is an implausibly large effect on a rare intronic variant in an unrelated gene.","sources":[{"title":"Prophylactic treatment of migraine with bisoprolol: a placebo-controlled study (Cephalalgia 1997, PMID 9251876)","url":"https://europepmc.org/article/MED/9251876"},{"title":"PubMed search: bisoprolol migraine prophylaxis (incl. PMID 1351025 bisoprolol vs metoprolol, 1992)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=bisoprolol+migraine+prophylaxis"},{"title":"Bisoprolol Fumarate 5 mg Film-coated Tablets SmPC (Sandoz) - sections 4.1 and 4.8","url":"https://www.medicines.org.uk/emc/product/4622/smpc"},{"title":"Ensembl variation record for rs181219039 (chr13:25,972,430, intron variant, MAF 0.0012)","url":"https://rest.ensembl.org/variation/human/rs181219039?content-type=application/json;pops=1"},{"title":"UniProt Q9NTI2 - ATP8A2 phospholipid-transporting ATPase IB (function, tissue specificity, CAMRQ4)","url":"https://www.uniprot.org/uniprotkb/Q9NTI2/entry"},{"title":"Europe PMC search: ATP8A2 AND migraine (no reported association)","url":"https://europepmc.org/search?query=ATP8A2%20AND%20migraine"}]},"bisoprolol|L989":{"verdict":"statistical artifact","headline":"Beta-blocker skin reactions are real as a class, but this hit rests on two rare noise variants","assessment":"L98.9 'Disorder of skin, unspecified' is a residual ICD wastebasket code, not a defined dermatologic phenotype, so it cannot by itself support or refute any specific cutaneous ADR. It is certainly not an indication: the bisoprolol fumarate label lists hypertension (with angina and heart failure elsewhere in the class), while rash, acne, eczema, psoriasis, pruritus, alopecia, dermatitis, exfoliative dermatitis and cutaneous vasculitis appear only in the unquantified post-marketing adverse-reaction list. A drug-caused mechanism at class level is described and immunologically framed: beta-blockers are linked to psoriasiform and lichen-planus-like eruptions, contact dermatitis, alopecia and vitiligo, and a prospective cohort found HR 1.39 for psoriasis after long-term use, with latencies up to a year, so the atlas median of 3.01 years post-exposure is not incompatible with a slow class effect. Against that, the atlas's own supporting numbers point the wrong way: cotx_pct 0.86% with enrichment 0.86 means unspecified skin disorder is if anything slightly less common in bisoprolol users than in users of other drugs, FAERS shows no signal (PRR 1.06), and the 87% temporality is uninformative because prescription records in this cohort predate diagnosis records. The genetic signal is the weakest part - beta_adr 6.0 (OR ~400) from two variants with gnomAD MAF 0.3-0.8% is the signature of sparse-data separation in a handful of carriers, not a real effect, and the two variants sit on different chromosomes so they are not even one locus. The literature supports a class-level dermatologic ADR for beta-blockers; it does not support this particular variant-phenotype pair, which the atlas numbers themselves undercut.","gene_comment":"ENSG00000302198 is an unnamed Havana 'novel transcript' lncRNA at chr20:57.99 Mb, and the other variant, rs114012253 (chr2:157.16 Mb), is intergenic with no mapped gene at all. There is no mechanistic gene here, and the lncRNA label should not be presented as one.","score_check":"The numbers do not hang together: beta_disease -0.038 +/- 0.147 is indistinguishable from zero as expected, but beta_adr of 6.0 on rare (MAF <1%) variants with only 92 datable cases is an implausible effect for its error and almost certainly sparse-data separation, which also manufactures the z_diff of 5.1. Enrichment below 1 and a null FAERS PRR give no external corroboration.","candidability":2,"candidability_reason":"Non-specific wastebasket phenotype, no enrichment, null FAERS, and an OR ~400 from two rare intergenic/novel-lncRNA variants - a sparse-data artifact riding on a real but unrelated class effect.","sources":[{"title":"DailyMed - BISOPROLOL FUMARATE tablet, film coated (label: indications and adverse reactions)","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=200ab93b-6ad3-46cb-8f57-db056a14de16"},{"title":"Immunologic adverse reactions of beta-blockers and the skin (Exp Ther Med 2019, PMID 31384329)","url":"https://europepmc.org/article/MED/31384329"},{"title":"Drug-induced psoriasis: clinical perspectives (Psoriasis Auckl 2017, PMC5774610)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC5774610/"},{"title":"Ensembl REST lookup - ENSG00000302198 (novel transcript, lncRNA, chr20:57,985,640-57,987,728)","url":"https://rest.ensembl.org/lookup/id/ENSG00000302198?content-type=application/json"},{"title":"Ensembl REST variation - rs570854059 (intron of chr20 lncRNA, gnomAD MAF 0.30%)","url":"https://rest.ensembl.org/variation/human/rs570854059?content-type=application/json;pops=1"},{"title":"Ensembl REST variation - rs114012253 (intergenic, chr2:157,159,365, gnomAD MAF 0.78%)","url":"https://rest.ensembl.org/variation/human/rs114012253?content-type=application/json;pops=1"}]},"bisoprolol|R252":{"verdict":"plausible ADR","headline":"Cramp is a labelled uncommon bisoprolol effect, but the INTS7 burden hit is a sparse-data artifact","assessment":"Cramp and spasm (R25.2) is not an indication for bisoprolol, whose licensed uses are hypertension, stable chronic angina and chronic heart failure with reduced LVEF; it is instead an explicitly labelled adverse effect, listed in the UK SmPC section 4.8 as 'muscular weakness and cramps' at uncommon frequency and appearing as 'muscle cramps' in the US bisoprolol fumarate label's adverse-experience list, so the atlas's on_bnf=0 flag is a matcher miss rather than an absence of label support. A drug-caused mechanism is describable but not well established: beta-1 blockade reduces cardiac output and peripheral perfusion (the same label reports coldness or numbness of the extremities as common, and end-stage peripheral arterial disease and Raynaud's are contraindications), and heart-failure patients on bisoprolol are near-universally co-treated with loop diuretics, yet an evidence-based review of nocturnal leg cramps implicates iron sucrose, oestrogens, raloxifene, naproxen and teriparatide while noting that even diuretics 'have not been implicated in evidence-based reviews' and not mentioning beta-blockers at all. Nothing points the other way - bisoprolol is not used to treat cramp - so this is not a backwards signal, merely a weakly supported one. The atlas phenotype numbers are flat rather than corroborating: cotx_pct 1.13% with enrichment 1.01 means cramp is no commoner in bisoprolol users than in users of other drugs, and FAERS is weak at PRR 1.22, which argues against confounding by indication but equally against a large drug effect. Temporality of 82.2% over 90 dated participants with a median 3.01 years to first cramp record is consistent with an incident post-exposure event but, as the cohort compresses this measure upward, carries little weight on its own. The genetic layer is where the association breaks down: the within-user hit is a single-variant MPC burden in INTS7 with beta 42.1, an effect size on the log-odds scale that no real biology produces and that is the signature of quasi-complete separation in a handful of carriers, while the same variant is null for cramp in the drug-free population (beta 1.537, se 1.184) and null for being prescribed bisoprolol (log10p 0.55), so the z_diff of 5.95 is driven entirely by the unstable treated-arm estimate.","gene_comment":"INTS7 (Integrator complex subunit 7, chromosome 1) is a ubiquitously expressed housekeeping subunit that terminates promoter-proximal RNA Pol II transcription and participates in S-phase DNA-damage signalling; it has no neuromuscular expression bias and no link to cramp, excitability or channel biology, unlike the SCN4A/CLCN1/CACNA1S class one would expect. With a single qualifying variant in the MPC burden it cannot be dressed up as mechanism.","score_check":"The disease-arm effect is frankly null by its own error bar (1.537 +/- 1.184) and the within-user beta of 42.1 is far outside any believable range for a binary outcome, pointing to sparse-carrier separation rather than a real effect. Enrichment of 1.01 and FAERS PRR 1.22 are both consistent with the literature that cramp on beta-blockers is real but small, so the phenotype side is credible while the genotype side is not.","candidability":4,"candidability_reason":"Genuine labelled but minor ADR; the specific INTS7 signal is an implausible-magnitude burden hit in a housekeeping gene with no neuromuscular rationale and no population-level enrichment behind it.","sources":[{"title":"Bisoprolol 10mg film-coated tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/14906/smpc"},{"title":"BISOPROLOL FUMARATE tablet, film coated - US prescribing information (DailyMed)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=200ab93b-6ad3-46cb-8f57-db056a14de16"},{"title":"Nocturnal Leg Cramps - American Family Physician 2012;86(4):350-355","url":"https://www.aafp.org/pubs/afp/issues/2012/0815/p350.html"},{"title":"UniProtKB Q9NVH2 - Integrator complex subunit 7 (INTS7)","url":"https://www.uniprot.org/uniprotkb/Q9NVH2"},{"title":"Factors affecting the intensity of chronic musculoskeletal pain in patients with cardiovascular disease... muscle cramps (Medicine 2023, PMID 37904395)","url":"https://europepmc.org/article/MED/37904395"}]},"bisoprolol|R410":{"verdict":"plausible ADR","headline":"Disorientation is a labelled beta-blocker class effect, but the SPRTN burden signal is a sparse-data artifact","assessment":"R41.0 Disorientation is not an indication for bisoprolol, which is licensed for hypertension (US label) and for stable chronic heart failure with reduced LVEF (UK SmPC). It is, however, a recognised class adverse effect: the US bisoprolol fumarate label lists 'decreased concentration/memory', vivid dreams, insomnia and somnolence among CNS effects, and explicitly cites for beta-blockers as a class 'an acute reversible syndrome characterized by disorientation to time and place, emotional lability, slightly clouded sensorium' plus hallucinations, while the UK SmPC lists hallucinations and nightmares as rare and sleep disorders and depression as uncommon. An active-comparator FAERS disproportionality analysis (J Psychopharmacol 2025, PMID 40704576) found disorientation, hallucinations, nightmares and somnolence reported more often for beta-blockers than for lisinopril or losartan, which is consistent with the atlas FAERS flag (STRONG, PRR 3.32) - but the effect was driven by lipophilic propranolol, and bisoprolol is a poorly CNS-penetrant, beta1-selective agent, so the class mechanism (central beta-adrenergic blockade after blood-brain-barrier crossing) applies to it only weakly. Confounding by indication is substantial rather than decisive: cognitive impairment affects roughly 40-50% of heart failure patients and about 40% of stroke-free permanent AF patients, and the observed enrichment of only 1.5 with cotx_pct 1.08% is exactly what an old cardiac population coded for a nonspecific symptom would produce. The temporality of 99.3% over 135 dated participants is uninformative under this cohort's known upward compression, and a median of 2.33 years after first exposure is equally compatible with accumulating age-related cognitive decline. The genetic layer does not support a drug-specific reading: the same SPRTN burden is nominally associated with disorientation in the drug-free population (beta 7.49, se 2.69, log10p 2.27), pointing to disease predisposition rather than a bisoprolol-specific interaction, and the atlas's own scores agree (predisposition 30.9 vs indication 15.1).","gene_comment":"SPRTN is a DNA-dependent metalloprotease that resolves DNA-protein crosslinks; biallelic loss causes Ruijs-Aalfs progeroid syndrome with early hepatocellular carcinoma, and it has no described role in beta-adrenergic signalling, brain function or delirium, nor in bisoprolol disposition (which is ~50% renal, ~50% CYP3A4, not CYP2D6). A rare-missense burden hit in this gene is biologically unmotivated here and should not be presented as mechanism.","score_check":"beta_adr of 90.2 on the log-odds scale is not a real effect size but quasi-complete separation among a handful of rare missense carriers, which also makes z_diff = 5.39 untrustworthy since it contrasts that inflated estimate with the drug-free one. The drug-free effect (7.49 +/- 2.69) does exceed 1.96*se and log10p_prescribed of 0.22 shows no prescribing-propensity signal, so the variant is probably a genuine but non-drug-specific predisposition signal rather than pure noise.","candidability":4,"candidability_reason":"Real labelled class ADR at the drug level, but the specific agent is poorly CNS-penetrant, the treated population is heavily confounded by HF/AF cognitive impairment, and the SPRTN burden is a sparse-data artifact that also fires without the drug.","sources":[{"title":"Bisoprolol Fumarate tablet - US prescribing information (DailyMed, Micro Labs)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=c3ee566c-74c4-45f4-a947-526a7313e326"},{"title":"Bisoprolol fumarate - Summary of Product Characteristics (emc, product 465)","url":"https://www.medicines.org.uk/emc/product/465/smpc"},{"title":"Beta-blockers and risk of neuropsychiatric adverse events: an active-comparator restricted disproportionality analysis of FAERS (J Psychopharmacol 2025, PMID 40704576)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:40704576&format=json&resultType=core"},{"title":"PubMed: beta-blocker delirium/confusion literature (incl. metoprolol-associated CNS complications PMID 32582495; delirium after vascular surgery with preoperative beta-blockade PMID 19711147)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=beta-blocker+delirium+confusion+elderly"},{"title":"UniProt Q9H040 (SPRTN_HUMAN): DNA-protein crosslink protease, Ruijs-Aalfs syndrome","url":"https://rest.uniprot.org/uniprotkb/Q9H040.txt"},{"title":"Europe PMC: cognitive impairment prevalence in heart failure (~40-50%) and permanent atrial fibrillation (~40%)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=(atrial%20fibrillation%20OR%20%22heart%20failure%22)%20AND%20(%22cognitive%20impairment%22)%20AND%20prevalence&format=json&pageSize=8&resultType=core"}]},"bisoprolol|R634":{"verdict":"statistical artifact","headline":"Beta-blockers cause weight gain, not loss; the hit is an unnamed lncRNA in a 3q13 gene desert","assessment":"Abnormal weight loss (R63.4) is not a licensed indication for bisoprolol, whose UK SmPC lists only hypertension, stable chronic angina and adjunctive therapy of chronic heart failure with reduced systolic LV function, and it is not a recognised adverse effect: the US FDA label names weight gain under General adverse experiences and mentions weight loss, anorexia, decreased appetite and cachexia nowhere, while the EMC SmPC section 4.8 carries no weight term at all. The pharmacology runs the wrong way for an ADR reading, and this is decisive rather than merely unhelpful: beta-blockade lowers resting energy expenditure and beta-adrenergic lipolysis, Sharma's systematic analysis of eight RCTs of at least six months found body weight higher on beta-blocker than control with a median difference of +1.2 kg, and in COPERNICUS carvedilol made patients with severe heart failure 33% less likely to lose more than 6% of body weight and 37% more likely to gain at least 5%, without fluid retention. Weight loss is nevertheless genuinely frequent in this treated population, as cardiac cachexia in advanced heart failure and as the frailty, malignancy and comorbidity burden of an elderly cardiovascular cohort, so a coded association is expected without the drug causing anything. The atlas numbers are consistent with that non-causal reading rather than with an ADR: enrichment is 0.99, so bisoprolol users carry R63.4 at exactly the rate seen in users of other drugs, and log10p_prescribed of 0.45 shows the variant does not predict receiving the drug. The drug-free disease arm is a flat null (beta_disease -0.003 with se 0.046, well inside noise, log10p 0.02), so the entire z_diff of 5.31 is carried by the within-users arm alone, an unreplicated single-variant hit resting on 156 dated cases. Temporality of 98.7% adds nothing here, since the prompt's own caveat is that this cohort compresses temporality upward and only low values are informative. The FAERS support is negligible on inspection: openFDA returns 2,523 'Weight decreased' events among 135,645 bisoprolol reports (1.86%) against roughly 1.4% background, a PRR near 1.3-1.5 for a term ubiquitous in elderly polypharmacy reporting, and it is accompanied by 1,491 'Weight increased' reports in the same drug.","gene_comment":"ENSG00000242880 is not a named gene but an unnamed 418 kb havana_tagene lncRNA annotated only as 'novel transcript' on 3q13.31 (chr3:115,147,451-115,565,595), and rs4277658 (chr3:115,339,675 G/A, MAF 6.3%, intron_variant, no GWAS Catalog association) falls in a gene desert roughly 190 kb telomeric of the ZBTB20 3' end and 280 kb from GAP43. There is no protein-coding gene near enough, and no adipose or adrenergic biology attached to this transcript, to support any mechanistic story, so the locus should be reported as an anonymous intergenic signal and not dressed up as a gene.","score_check":"The statistics are internally coherent rather than a sparse-data explosion, since beta_adr 1.604 at log10p 7.08 implies a standard error near 0.30 and a z of about 5.4, so the number is not implausible for its error. What is not believable is the interpretation: a single unreplicated variant in an anonymous lncRNA, an outcome with enrichment 1.0 and a completely null disease arm, and a non-specific symptom code that in this cohort mostly indexes cardiac cachexia, frailty and occult malignancy.","candidability":1,"candidability_reason":"Backwards signal: the drug class demonstrably prevents weight loss and promotes weight gain, and the variant is an anonymous intergenic lncRNA hit with no disease-arm or literature support.","sources":[{"title":"Bisoprolol fumarate tablets, US FDA prescribing information (openFDA drug label API) - indications and adverse reactions, listing weight gain and no weight loss","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22bisoprolol+fumarate%22&limit=1"},{"title":"Bisoprolol fumarate 5 mg tablets - Summary of Product Characteristics (emc), sections 4.1 and 4.8","url":"https://www.medicines.org.uk/emc/product/3090/smpc"},{"title":"Sharma AM et al. Hypothesis: Beta-adrenergic receptor blockers and weight gain: A systematic analysis. Hypertension 2001;37:250-4","url":"https://pubmed.ncbi.nlm.nih.gov/11230280/"},{"title":"Clark AL et al. Effect of beta-adrenergic blockade with carvedilol on cachexia in severe chronic heart failure: results from the COPERNICUS trial. J Cachexia Sarcopenia Muscle 2017;8:549-56","url":"https://pubmed.ncbi.nlm.nih.gov/28244261/"},{"title":"Cabassi A et al. Sympathetic modulation by carvedilol and losartan reduces angiotensin II-mediated lipolysis in subcutaneous and visceral fat. J Clin Endocrinol Metab 2005;90:2888-97","url":"https://pubmed.ncbi.nlm.nih.gov/15741261/"},{"title":"Hweidi IM et al. Cardiac cachexia among patients with chronic heart failure: A systematic review. Nurs Forum 2021;56:916-24","url":"https://pubmed.ncbi.nlm.nih.gov/34091923/"},{"title":"Ensembl REST lookup for ENSG00000242880 (lncRNA, 'novel transcript', chr3:115,147,451-115,565,595) and variant rs4277658","url":"https://rest.ensembl.org/lookup/id/ENSG00000242880?content-type=application/json"},{"title":"openFDA FAERS event counts for bisoprolol with 'Weight decreased' versus 'Weight increased'","url":"https://api.fda.gov/drug/event.json?search=patient.drug.openfda.generic_name:%22bisoprolol%22+AND+patient.reaction.reactionmeddrapt:%22Weight+decreased%22&limit=1"}]},"citalopram|A099":{"verdict":"statistical artifact","headline":"SSRIs really do cause GI upset, but this A09 hit is a rare intronic brain-gene variant with an impossible OR","assessment":"Gastroenteritis is not a licensed indication for citalopram; the US label lists only major depressive disorder, and the only GI content is adverse-reaction data (nausea 21% vs 14% placebo, diarrhoea 8% vs 5%, vomiting 4% vs 3%). So a class-level GI effect is genuinely documented, and SSRIs are additionally listed among drug risk factors for microscopic colitis, with antidepressant use associated with Clostridioides difficile infection in a two-design US study - both routes could in principle land a patient under an A09 'gastroenteritis and colitis of unspecified origin' code. The atlas numbers, however, do not fit that reading: enrichment is 0.87, i.e. citalopram users carry this code slightly LESS often than users of other drugs, so there is no excess to explain, and the median first record is 4.39 years after first exposure, whereas SSRI nausea and loose stools are an early-weeks, dose-onset phenomenon that habituates. The temporality of 100% over n_dated=183 is exactly what an acute, self-limiting, high-background-incidence infectious code produces under years of follow-up, and carries no weight here. The within-user genetic signal (log10p 6.5) does not reach genome-wide significance, and its beta of 4.19 implies an odds ratio near 66 for a variant with ~0.5% European frequency - a sparse-data artifact, not an effect. The FAERS 'STRONG' flag at PRR 2.33 is consistent with patients reporting SSRI-induced nausea and diarrhoea that map onto gastroenteritis-like preferred terms, which is a coding overlap rather than independent confirmation of this variant.","gene_comment":"rs145737307 is an intronic SNV (chr7:88134075 A>G, gnomAD NFE MAF ~0.55%) in ADAM22, a non-catalytic LGI1 receptor whose expression is essentially restricted to brain and whose only established disease link is autosomal-recessive developmental and epileptic encephalopathy 61. There is no gut, mucosal or immune biology to connect it to infectious gastroenteritis, and an intronic rare variant carrying an OR of ~66 is a hallmark of low-count instability.","score_check":"The comparison arms behave sensibly - the variant is null for the condition off-drug (beta 0.151 +/- 0.133) and null for prescription propensity (log10p 0.02) - but the ADR arm itself is not believable, since beta 4.19 at only p~3e-7 on a ~0.5% allele means a handful of carrier cases are driving everything. z_diff 4.86 inherits that instability rather than validating it.","candidability":2,"candidability_reason":"Real SSRI GI toxicity exists, but here the code is depleted rather than enriched, onset is 4.4 years out, and the variant is an implausibly large-effect rare intron in a brain-only gene.","sources":[{"title":"Citalopram tablet - US prescribing information (Amneal), DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=6a402640-1c9d-4fca-8b5b-e1ac1dd4ca5a"},{"title":"PubMed: SSRIs and microscopic colitis (Townsend 2019 Frontline Gastroenterol; Menon & Ng 2015; Park 2015 WJG)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=selective+serotonin+reuptake+inhibitor+microscopic+colitis"},{"title":"Rogers MAM et al. Depression, antidepressant medications, and risk of Clostridium difficile infection. BMC Medicine 2013","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=DOI:%2210.1186/1741-7015-11-121%22&resultType=core&format=json"},{"title":"UniProt Q9P0K1 (ADAM22_HUMAN): brain-restricted LGI1 receptor, DEE61","url":"https://rest.uniprot.org/uniprotkb/Q9P0K1.txt"},{"title":"Ensembl / gnomAD v4 record for rs145737307 (7-88134075-A-G, intronic, NFE MAF 0.0055)","url":"https://rest.ensembl.org/variation/human/rs145737307?content-type=application/json;pops=1"},{"title":"Comparative Real-World Safety Profiles of Six SSRIs: A Global Pharmacovigilance Analysis. Cureus 2025","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12775886/"}]},"citalopram|B351":{"verdict":"statistical artifact","headline":"Onychomycosis is not a citalopram effect; a rare intergenic variant with OR~2800 is sparse-data separation","assessment":"Citalopram is licensed only for major depressive disorder, and the FDA label's dermatological adverse reactions run from rash and pruritus through acne, eczema, alopecia and psoriasis to rare cellulitis, with no fungal infection, tinea or nail disorder anywhere in the list, so B35.1 is neither an indication nor a recognised adverse effect. Onychomycosis is instead an extremely common age- and comorbidity-driven condition (point prevalence up to 13.8% in North American adults, driven by age over 60, diabetes, obesity, peripheral vascular disease, smoking and immunosuppression), none of which citalopram causes. The atlas agrees that this is background: cotx_pct is only 1.22% (far below true population prevalence, so nail infections are heavily under-coded in the EHR) and enrichment is 0.93, i.e. citalopram users carry onychomycosis slightly LESS often than users of other drugs, which is the opposite of what confounding-by-indication or a real ADR would produce. If anything the pharmacology points the wrong way for an ADR reading: SSRIs have direct in-vitro antifungal activity against Trichophyton rubrum, the dermatophyte responsible for most onychomycosis, with sertraline shown to disrupt ergosterol biosynthesis and cell-wall/membrane stability genes and being explored for antifungal repurposing. The only genuine clinical link between this drug and this condition runs in the reverse direction and is a drug-drug interaction, not an ADR: onychomycosis reviews specifically warn to monitor patients on terbinafine who also take SSRIs, TCAs, atypical antipsychotics or beta-blockers. The within-user signal rests on one rare variant with beta_adr 7.93, an odds ratio near 2,800, in a subgroup where the outcome is coded in ~1.2% of people, which is the signature of a near-separated contingency table rather than a real effect; the disease arm is flatly null and negatively signed (beta -0.215, se 0.162, |beta| well under 1.96*se, log10p 0.73), the prescription arm is null (log10p 0.23), and the 80% temporality on 90 dated participants is exactly the uninformative high value the cohort's record-depth asymmetry manufactures.","gene_comment":"There is no gene: rs189578781 is an intergenic variant at chr1:116,222,102 (GRCh38) with no mapped gene in Ensembl and a most-severe consequence of intergenic_variant, and at ~0.7-1.5% minor allele frequency in Europeans (higher in Finns) it is rare enough to make a huge beta unstable. Nothing about the locus connects to SSRI pharmacokinetics, cutaneous immunity or dermatophyte susceptibility, so no mechanistic story can be built on it.","score_check":"The numbers do not hang together: an OR of ~2,800 for a 1%-frequency variant on a common, under-ascertained skin condition is implausible for any real biology and is the classic sparse-data/quasi-separation artifact, while the log10p of 8.45 is asymptotic p-value inflation from that same near-empty cell. The supporting fields all point away from a drug effect (enrichment 0.93, null and oppositely signed disease effect, null prescription arm), leaving z_diff 6.02 as an artefact of the unstable ADR-arm estimate rather than evidence of drug modification.","candidability":1,"candidability_reason":"Not a labelled or literature-supported ADR, no enrichment in treated patients, no gene, and an effect size implausible for its error; SSRI pharmacology if anything points the opposite way.","sources":[{"title":"Citalopram tablet - FDA prescribing information (DailyMed, Amneal Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=6a402640-1c9d-4fca-8b5b-e1ac1dd4ca5a"},{"title":"Ensembl REST variation record for rs189578781 (intergenic, chr1:116,222,102, MAF ~0.002)","url":"https://rest.ensembl.org/variation/human/rs189578781?content-type=application/json;pops=1"},{"title":"Onychomycosis: Rapid Evidence Review (AFP 2021) - prevalence, risk factors, terbinafine-SSRI interaction","url":"https://www.aafp.org/pubs/afp/issues/2021/1000/p359.html"},{"title":"PubMed: onychomycosis epidemiology and risk factors reviews (incl. Gupta 2017, J Cutan Med Surg)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=onychomycosis+epidemiology+risk+factors+review"},{"title":"The Antidepressant Sertraline Affects Cell Signaling and Metabolism in Trichophyton rubrum (J Fungi 2023, PMID 36836389)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:36836389&resultType=core&format=json"}]},"citalopram|H612":{"verdict":"statistical artifact","headline":"Earwax impaction is not a citalopram effect; the hit is one rare intergenic allele on 61 dated cases","assessment":"Impacted cerumen (H61.2) is neither a licensed indication for citalopram - the Celexa label lists only major depressive disorder - nor a recognised adverse reaction: no ear, cerumen or hearing term appears in its adverse-reaction or postmarketing sections, and tinnitus is mentioned only as a discontinuation-syndrome symptom. It is instead an extremely common incidental finding of a clinical encounter: NHANES puts otoscopically confirmed cerumen impaction at 18.6% of people aged 12+ and 32.4% of those aged 70+, driven by age, male sex and insurance/access rather than by any drug, and the 2017 AAO-HNS guideline frames it as an anatomical/self-cleaning problem with no pharmacological cause. The atlas cotx_pct of 0.83% is more than an order of magnitude below that population prevalence, which confirms the ICD code here is a marker of who had their ears examined and coded, not of who has earwax. No drug mechanism is described or plausible: citalopram's relevant peripheral effects are mild anticholinergic dryness (dry mouth 20% vs 14% placebo) and increased sweating (11% vs 9%), neither of which has any documented action on ceruminous glands, and the direction of either would be pure speculation. Temporality of 72.1% after first exposure rests on only 61 datable participants with a median of 3.09 years, exactly what accrual of routine primary-care encounters during follow-up produces, and the prescription arm (log10p 0.23) and disease arm (log10p 0.48) are both null. Nothing in the literature supports a citalopram-cerumen link, and nothing in the atlas numbers requires one.","gene_comment":"rs80144718 has no gene: it is an intergenic variant at chr11:75,024,965 (GRCh38) with a minor allele frequency of about 0.2%, roughly 4 kb beyond the end of NEU3 and around 75 kb from SLCO2B1/OR2AT4, so any gene label attached to it would be proximity, not mechanism. There is no biological story connecting this locus to cerumen production or to citalopram handling.","score_check":"beta_adr of 6.015 (odds ratio in the hundreds) for a 0.2%-frequency allele against roughly 61 events is the classic sparse-data separation artifact, and the z_diff of 5.06 is generated entirely by that inflated estimate since the drug-free disease effect is flat noise (beta -0.182, se 0.188). The FAERS PRR of 2.19 for such a nuisance term is reporting background, not corroboration.","candidability":0,"candidability_reason":"Not an adverse reaction by definition, no mechanism, no label or literature support, and the signal is a rare-allele sparse-data artifact on a detection-driven code.","sources":[{"title":"CELEXA (citalopram) US prescribing information, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=4259d9b1-de34-43a4-85a8-41dd214e9177"},{"title":"Clinical Practice Guideline (Update): Earwax (Cerumen Impaction), Otolaryngol Head Neck Surg 2017 (PMID 28045591)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:28045591&resultType=core&format=json"},{"title":"Tolan et al., Cerumen impaction: prevalence and associated factors in the United States population, Laryngoscope Investig Otolaryngol 2024","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC10958939/"},{"title":"Ensembl REST variation record for rs80144718 (intergenic, chr11:75,024,965, MAF ~0.002)","url":"https://rest.ensembl.org/variation/human/rs80144718?content-type=application/json"},{"title":"Ensembl REST gene overlap for chr11:74,800,000-75,300,000 (NEU3, SLCO2B1, ARRB1 neighbourhood)","url":"https://rest.ensembl.org/overlap/region/human/11:74800000-75300000?feature=gene;content-type=application/json"}]},"citalopram|H919":{"verdict":"statistical artifact","headline":"Hearing loss is not a labelled citalopram effect; signal rests on one ultra-rare intronic HK1 variant","assessment":"Hearing loss is not a licensed indication for citalopram (US labels list major depressive disorder only) and unspecified hearing loss is not a labelled adverse reaction; the label's only auditory entry is tinnitus, under 'Special Senses' as an infrequent (1/1000-1/100) premarketing event. Class-level inner-ear signals do exist - a 2004-2023 FAERS disproportionality analysis found excess reporting of tinnitus, vertigo, labyrinthitis and vestibular ataxia for eight antidepressants including citalopram - but these are vestibular/tinnitus terms, not sensorineural hearing loss, and the atlas itself reports FAERS counts too low to test here. The one randomised trial of the exact exposure (citalopram 20 mg/d vs placebo for 6 months in patients over 60 with auditory processing disorder) found no effect on any central auditory test, which argues against a drug-caused deterioration of hearing on this timescale. The far more likely route runs backwards: audiometric hearing loss independently raises the odds of antidepressant use (OR 1.04 per 10 dB worsening in a 29,772-patient EHR study), so citalopram users are enriched for deaf and hard-of-hearing patients by indication and by age, and the atlas's own enrichment of 0.95 says hearing loss is no commoner in citalopram users than in users of other drugs. The 91.1% temporality (n_dated 135, median 3.53 years after first exposure) is exactly the value the cohort's record-length asymmetry inflates, so it adds little. Nothing in the numbers requires a drug effect, and the specific variant driving the hit is not credible as one.","gene_comment":"rs558138996 is an ultra-rare (MAF ~0.0008) non-coding SNP at chr10:69,276,684, which falls inside the HK1 gene span but is annotated intergenic/non-coding by Ensembl and has no GWAS Catalog entry; HK1's established phenotypes are non-spherocytic haemolytic anaemia, HMSN-Russe neuropathy, retinitis pigmentosa 79 and NEDVIBA, with no hearing or cochlear disease among them. Calling this 'HK1' is a positional label, not a mechanism.","score_check":"beta_adr 5.584 (OR ~260) for a variant carried by roughly 1 in 1300 people is the signature of a sparse-cell / near-separation artifact, not a real effect size, and the same variant is flat in the drug-free population (beta 0.178 +/- 0.113, |beta| well under 1.96*se, log10p 0.94), so z_diff 5.5 is inherited entirely from the implausible treated-arm estimate. The epidemiological columns are internally consistent and unremarkable (cotx 1.37%, enrichment 0.95, prescribed log10p 0.21).","candidability":2,"candidability_reason":"Unlabelled outcome with an RCT showing no auditory effect, no enrichment, and a single ultra-rare non-coding variant with an impossible effect size; only weak class-level inner-ear reporting keeps it off zero.","sources":[{"title":"CITALOPRAM (citalopram hydrobromide) tablet, film coated - DailyMed label (indication; tinnitus as infrequent Special Senses event)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=2632b547-2e13-447f-ac85-c774e437d6a8"},{"title":"Polanski et al., The Effect of Citalopram Versus a Placebo on Central Auditory Processing in the Elderly, Otol Neurotol 2017 (PMID 28796088)","url":"https://europepmc.org/article/MED/28796088"},{"title":"Inner ear signs and symptoms induced by antidepressants: a FAERS disproportionality analysis, Naunyn-Schmiedeberg's Arch Pharmacol 2026 (PMID 41171397)","url":"https://europepmc.org/article/MED/41171397"},{"title":"The Association Between Hearing Loss and Depression in a Large Electronic Health Record System, Otolaryngol Head Neck Surg 2025 (PMID 39865418)","url":"https://europepmc.org/article/MED/39865418"},{"title":"UniProt P19367 (HK1_HUMAN) disease annotations: CNSHA5, HMSNR, RP79, NEDVIBA","url":"https://www.uniprot.org/uniprotkb/P19367/entry"},{"title":"Ensembl REST variation record for rs558138996 (MAF 0.00078, intergenic/non-coding, chr10:69,276,684)","url":"https://rest.ensembl.org/variation/human/rs558138996?content-type=application/json"},{"title":"NHGRI-EBI GWAS Catalog search for rs558138996 - no associations found","url":"https://www.ebi.ac.uk/gwas/search?query=rs558138996"}]},"citalopram|K317":{"verdict":"co-prescription population","headline":"Gastric/duodenal polyps are a PPI- and endoscopy-driven finding in SSRI users, not a citalopram effect","assessment":"Citalopram's label carries only major depressive disorder as an indication and lists nausea, diarrhoea, dyspepsia, vomiting and abdominal pain plus post-marketing gastrointestinal haemorrhage; gastric polyps, gastric neoplasia and hypergastrinaemia appear nowhere in it, and a PubMed search for antidepressants or SSRIs with gastric or fundic gland polyps returns no relevant study at all, so K31.7 is neither a licensed indication nor a recognised adverse effect. The one well-established iatrogenic cause of gastric polyps is chronic acid suppression: a meta-analysis of 20 studies and 40,218 subjects found fundic gland polyps in PPI users at OR 2.46 (95% CI 1.42-4.27), rising to OR 4.71 beyond six months and 5.32 beyond twelve, and SSRI users are precisely the group given long-term PPI cover because serotonin-reuptake inhibition raises upper-GI bleeding risk, as the citalopram label's abnormal-bleeding warning states. A second, equally mundane route is ascertainment: gastric polyps are silent and are only ever coded when someone is gastroscoped, and citalopram is used in patients with dyspepsia, gastroparesis and other disorders of gut-brain interaction, who are scoped far more than average. The atlas numbers fit those explanations rather than an ADR: 100% temporality over 102 dated participants with a median 4.3 years to the polyp code is exactly what cumulative PPI exposure and eventual endoscopy look like, and with enrichment null there is no test of whether these patients were already an endoscopy-heavy population. Nothing here suggests citalopram moves gastric polyps in the opposite direction; the point is simply that no mechanism from serotonin reuptake to fundic gland or hyperplastic polyp formation has been described. The FAERS PRR of 5.26 sits on a very small numerator for a condition almost never spontaneously reported except incidentally, and co-reported PPI use is the obvious competing suspect.","gene_comment":"rs142651685 is intronic in NFIB, a brain-development transcription factor, and rs151112840 intronic in N4BP3, a predicted neuronal NEDD4-binding protein; neither has any described role in gastric epithelium, acid secretion or polyp biology, and both are rare (global MAF 0.16% and 0.31%), so the gene labels are proximity assignments rather than mechanism.","score_check":"beta_adr 4.27 with log10p 6.71 implies a standard error near 0.8 and an odds ratio around 70 with an order-of-magnitude confidence interval - classic sparse-data inflation from two rare intronic variants against roughly 100 cases. The drug-free arm is flatly null (beta 0.06 +/- 0.117), so z_diff 5.08 is carried entirely by that uninterpretable treated-arm estimate.","candidability":2,"candidability_reason":"No label or literature link, no mechanism, an obvious PPI co-medication and endoscopic-detection confounder, and rare intronic variants in neurodevelopmental genes with a sparse-data effect size.","sources":[{"title":"Citalopram tablets - DailyMed label (Aurobindo)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=2632b547-2e13-447f-ac85-c774e437d6a8"},{"title":"Martin FC et al. Systematic review with meta-analysis: fundic gland polyps and proton pump inhibitors. Aliment Pharmacol Ther 2016 (PMID 27634363)","url":"https://doi.org/10.1111/apt.13800"},{"title":"PubMed search: (antidepressant OR SSRI OR citalopram) AND (gastric polyp OR fundic gland polyp) - no relevant results","url":"https://pubmed.ncbi.nlm.nih.gov/?term=%28antidepressant+OR+SSRI+OR+citalopram%29+AND+%28gastric+polyp+OR+fundic+gland+polyp%29"},{"title":"PubMed search: fundic gland polyps and proton pump inhibitors","url":"https://pubmed.ncbi.nlm.nih.gov/?term=fundic+gland+polyps+proton+pump+inhibitor"},{"title":"Ensembl variation record rs142651685 (intronic, NFIB, MAF 0.0016)","url":"https://rest.ensembl.org/variation/human/rs142651685?content-type=application/json"},{"title":"Ensembl variation record rs151112840 (intronic, N4BP3, MAF 0.0031)","url":"https://rest.ensembl.org/variation/human/rs151112840?content-type=application/json"},{"title":"MyGene.info records for NFIB and N4BP3","url":"https://mygene.info/v3/query?q=symbol:NFIB&species=human&fields=symbol,name,summary"}]},"citalopram|K635":{"verdict":"statistical artifact","headline":"Colon polyp is a colonoscopy-detected finding, not a citalopram effect; lone rare lncRNA hit","assessment":"Citalopram is licensed only for major depressive disorder (and, in EU labelling, panic disorder); the US label lists nausea, diarrhoea, dyspepsia, vomiting and abdominal pain among GI reactions and gastrointestinal haemorrhage in postmarketing experience, but no colonic neoplasm, adenoma or polyp anywhere. Polyp of colon (K63.5) is not an indication, not a recognised adverse reaction, and the atlas itself finds no disproportionality in FAERS (PRR 0.89, no_signal) and no BNF listing. The neoplasia literature runs neutral to the opposite direction: a Saskatchewan nested case-control study found a decreased colorectal cancer risk with high daily SSRI dose (IRR 0.70, 95% CI 0.50-0.96), and an 11-study dose-response meta-analysis of 109,506 participants found no association for antidepressants overall (RR 0.97, 0.94-1.01) or for SSRIs specifically (RR 0.99, 0.96-1.03). A drug-caused mechanism is not described; the only route that would plausibly raise recorded polyp counts in SSRI users is ascertainment, since SSRIs do increase GI bleeding and occult bleeding or GI symptoms lead to colonoscopy, at which age-prevalent polyps are found incidentally. The atlas numbers fit that reading: cotx_pct is only 1.22%, no background enrichment could be computed, and 97.6% temporality over 210 dated participants with a 3.3-year median lag is exactly what a screening/endoscopy-detected diagnosis looks like in a cohort whose prescription records predate its diagnosis records. The genetic evidence is a single low-frequency SNV with an implausibly large within-user effect (beta 3.13, OR ~23) that is flatly null in the drug-free population (beta 0.071 +/- 0.088, log10p 0.38), so the entire z_diff of 5.08 comes from one sparse, unreplicated arm.","gene_comment":"LINC02015 is a long intergenic non-coding RNA on chromosome 3 (~chr3:177.82-177.94 Mb, Ensembl) with no established function or colorectal biology, and rs145844567 does not appear in the GWAS Catalog for any trait. The assignment is a positional label for an intergenic locus, not a mechanism, and should not be presented as one.","score_check":"log10p_adr 6.83 with beta 3.13 and a completely null disease arm (0.071 +/- 0.088) is the signature of a rare-variant, sparse-cell association rather than a real effect of that magnitude. The internal confidence score (44.4) and the low indication score (17.4) are consistent with a weak, unsupported hit.","candidability":2,"candidability_reason":"Not an indication and not a labelled ADR; literature is null-to-protective for colorectal neoplasia, the phenotype is endoscopy-detection driven, and the genetics rest on one unreplicated rare variant at an intergenic lncRNA.","sources":[{"title":"CITALOPRAM tablet - US prescribing information (DailyMed, Amneal Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=6a402640-1c9d-4fca-8b5b-e1ac1dd4ca5a"},{"title":"Xu W et al. Use of antidepressants and risk of colorectal cancer: a nested case-control study. Lancet Oncol 2006 (PMID 16574545)","url":"https://europepmc.org/article/MED/16574545"},{"title":"Antidepressant use and colorectal cancer morbidity and mortality: a dose-response meta-analysis. Medicine (Baltimore) 2020 (PMID 32481383)","url":"https://europepmc.org/article/MED/32481383"},{"title":"PubMed: antidepressants AND colorectal cancer risk (search results incl. WHI cohort, PMID 29789327)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=antidepressants+AND+colorectal+cancer+risk"},{"title":"Ensembl REST lookup: LINC02015 (lncRNA, chromosome 3)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/LINC02015?content-type=application/json"},{"title":"GWAS Catalog search for rs145844567 (no reported associations)","url":"https://www.ebi.ac.uk/gwas/search?query=rs145844567"}]},"citalopram|M819":{"verdict":"plausible ADR","headline":"Osteoporosis is a recognised SSRI class effect, but this rare intronic lncRNA variant is not credible","assessment":"Osteoporosis is not a licensed indication for citalopram: the SmPC lists only depressive illness and panic disorder with or without agoraphobia. Nor is it a co-prescription artefact here, since osteoporosis affects 1.0% of citalopram users with an enrichment of 0.96 against users of other drugs, and the prescription-propensity arm is flat (log10p 0.04), so the condition is neither the reason for the drug nor over-represented in its population. Reduced bone mineral density and fracture are, however, a recognised SSRI class effect: the citalopram SmPC section 4.8 carries a class-effects statement that epidemiological studies, mainly in patients aged 50 and over, show an increased risk of bone fractures with SSRIs and TCAs, and reviews report roughly a 70% increase in fracture risk among SSRI users. The direction of drug effect is unambiguously toward worse bone, not better: escitalopram impaired fracture callus formation in rats, with reduced trabecular thickness (p=0.031) and reduced callus bone volume (p=0.002), consistent with serotonin-transporter blockade acting on osteoblast proliferation, differentiation and mineralisation. The counterweight is that causal magnitude is contested, with Mendelian-randomisation estimates for antidepressant exposure on osteoporosis of only about OR 1.003, and depression itself lowers BMD, so most of the observational association may be indication and lifestyle confounding rather than drug pharmacology. The atlas therefore agrees with the literature at the drug-condition level, and the FAERS PRR of 2.09 with concordant ROR/IC fits the known class signal, but the pharmacogenomic layer does not: a beta of 4.93 (odds ratio near 140) for a ~0.5%-frequency variant over 132 dated cases is a sparse-data magnitude, and the same allele is already nominally associated with osteoporosis in drug-free participants (beta 0.325, se 0.118, z=2.75), so part of what z_diff attributes to citalopram is baseline predisposition. Temporality of 86.4% with a median 3.74 years to onset is exactly what ageing into osteoporosis on maintenance antidepressants would produce, so it adds little on its own.","gene_comment":"DLEU1 is a long non-coding RNA at 13q14, known as the region deleted in chronic lymphocytic leukaemia, with no established role in bone or skeletal biology. rs138921674 is a rare intronic variant there (T allele ~0.49% in gnomAD, no ClinVar entry), so the gene label supplies a name rather than a mechanism.","score_check":"The condition-level numbers hang together and are believable: no enrichment, no prescription-propensity signal, a modest and properly powered disease-arm effect, and an independent FAERS signal. The variant-level beta of 4.93 for a 0.5%-frequency SNV is the signature of sparse-data bias and should not be read as a real effect size.","candidability":5,"candidability_reason":"The drug-condition link is a genuine, labelled SSRI class effect not explained by indication, but the specific variant is a rare intronic lncRNA hit with an implausibly large effect and partial explanation by baseline predisposition.","sources":[{"title":"Citalopram 20mg Tablets (Aurobindo) SmPC, section 4.1 indications - electronic Medicines Compendium","url":"https://www.medicines.org.uk/emc/product/550/smpc"},{"title":"Citalopram 20mg Tablets (Sandoz) SmPC, section 4.8 class effects on bone fracture - electronic Medicines Compendium","url":"https://www.medicines.org.uk/emc/product/4955/smpc"},{"title":"PubMed: SSRI, bone mineral density and fracture risk (incl. Warden & Fuchs, Curr Osteoporos Rep 2016, PMID 27495351)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=SSRI+bone+mineral+density+fracture+risk+osteoporosis"},{"title":"The SSRI escitalopram oxalate negatively impacts fracture healing in healthy adults and osteoporotic rats (Sci Rep 2025, PMC12572269)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12572269/"},{"title":"Europe PMC: serotonin transporter, osteoblast and SSRI effects on bone (incl. brain-bone axis review PMC12988505)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22serotonin%20transporter%22%20AND%20osteoblast%20AND%20SSRI%20AND%20bone&format=json&resultType=core&pageSize=5"},{"title":"Europe PMC: antidepressants and osteoporosis/fracture risk, incl. Mendelian randomisation (PMC13180680) and drug-induced osteoporosis (PMC12898846)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%28antidepressant%20AND%20osteoporosis%20AND%20meta-analysis%20AND%20%22fracture%20risk%22%29&format=json&resultType=core&pageSize=4"},{"title":"dbSNP rs138921674 - chr13:50464150, intronic in DLEU1, gnomAD MAF 0.49%","url":"https://www.ncbi.nlm.nih.gov/snp/rs138921674"},{"title":"Ensembl REST lookup: DLEU1, lncRNA biotype, chromosome 13","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/DLEU1?content-type=application/json;expand=0"}]},"citalopram|N390":{"verdict":"statistical artifact","headline":"UTI is not a citalopram indication or labelled reaction; the signal is one rare intergenic SNV","assessment":"Citalopram is licensed only for major depressive disorder, and urinary tract infection appears nowhere in its adverse-reaction tables; what the label does list are micturition disorders - polyuria (frequent) and urinary retention, urinary incontinence, micturition frequency and dysuria (all infrequent). A mechanistic route from SSRI to UTI therefore exists in principle, via serotonergically mediated voiding dysfunction and incomplete bladder emptying, and case reports of escitalopram-associated acute urinary retention resolving on withdrawal support it, but the systematic review of antidepressant-associated retention puts the SSRI rate at 0.025% of patients - far too rare to generate a population-level UTI signal. The one purpose-built epidemiological test is against it: in a CPRD nested case-control study of 2664 UTI cases in people with multiple sclerosis, SSRI use versus no antidepressant gave an adjusted OR of 1.15 (0.96-1.37), i.e. null, while TCA/SNRI use did raise risk (1.43, 1.21-1.69) - SSRIs were the comparator that looked clean. The atlas is consistent with that null on the confounding axis rather than the causal one: enrichment 0.82 means UTI is actually less common among citalopram users than among users of other drugs, and log10p_prescribed 0.47 shows the variant has nothing to do with who gets prescribed the drug, so this is not classical confounding by indication either. Temporality of 87.2% over 313 dated cases with a median 2.58 years to onset is uninformative given the cohort's known upward compression, and FAERS support is weak (1/3, PRR 1.91) for a term that is co-reported in almost any elderly polypharmacy series. What is left is a single low-frequency variant with an implausibly large effect and no biology behind it, which is the signature of a sparse-data false positive rather than a pharmacogenomic finding.","gene_comment":"rs143913044 (chr3:192,014,844, GRCh38) is annotated intergenic with no assigned gene: its nearest features are unnamed lncRNAs and the processed pseudogene PERPP1, and the closest protein-coding gene, FGF12, lies ~124 kb away and has no urinary-tract or host-defence biology. At ~1.5% minor allele frequency in non-Finnish Europeans this is a low-frequency variant with no mechanistic story to tell.","score_check":"beta_adr 2.146 (OR about 8.6) for a ~1.5% MAF variant in roughly 313 cases means the estimate rests on a handful of carriers, and log10p_adr 6.92 does not even clear genome-wide significance, while the disease arm is flat noise (beta 0.025 +/- 0.058) so the whole z_diff of 5.18 is carried by the inflated treated-arm beta. The numbers are internally consistent only as a small-count artifact, not as a real effect of that magnitude.","candidability":1,"candidability_reason":"Not a labelled reaction, depleted rather than enriched in citalopram users, contradicted by a dedicated SSRI-UTI study, and driven by one sub-genome-wide intergenic variant with a sparse-data effect size.","sources":[{"title":"Citalopram tablet - DailyMed label (Amneal): indications and adverse reactions","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=6a402640-1c9d-4fca-8b5b-e1ac1dd4ca5a"},{"title":"Faure Walker et al., Linking the evidence between urinary retention and antipsychotic or antidepressant drugs: a systematic review, Neurourol Urodyn 2016 (PMID 26288221)","url":"https://pubmed.ncbi.nlm.nih.gov/26288221/"},{"title":"Leung et al., TCA/SNRI versus SSRI and risk of urinary tract infection in multiple sclerosis: nested case-control study, Pharmacoepidemiol Drug Saf 2025 (PMID 41195597)","url":"https://pubmed.ncbi.nlm.nih.gov/41195597/"},{"title":"Trombetta et al., Escitalopram-associated acute urinary retention, Consult Pharm 2013 (PMID 24129221)","url":"https://pubmed.ncbi.nlm.nih.gov/24129221/"},{"title":"Ensembl REST variation record for rs143913044 (intergenic, chr3:192,014,844, MAF ~0.0055)","url":"https://rest.ensembl.org/variation/human/rs143913044?content-type=application/json"}]},"citalopram|R05":{"verdict":"statistical artifact","headline":"Cough is a label-listed nonspecific event, but this rare lncRNA variant looks like sparse-data noise","assessment":"Cough is not an indication for citalopram, whose only US licensed indication is major depressive disorder, but it is not absent from the label either: the Celexa label lists \"coughing\" as a Frequent event among other adverse reactions observed during premarketing evaluation, with rhinitis 5% vs 3% and upper respiratory tract infection 5% vs 4% on placebo, i.e. essentially no controlled excess. A literature search returned no paper linking citalopram or escitalopram specifically to cough, and the only established SSRI respiratory harm is rare antidepressant-associated eosinophilic/organizing pneumonia, almost entirely sertraline case reports, which presents as dyspnoea and infiltrates rather than as an isolated R05 symptom code. The direction of the serotonergic literature runs the other way: duloxetine reduced cough frequency from 84 to 33 coughs/hour in refractory chronic cough, amitriptyline and tramadol are used as cough neuromodulators, and reviews of cough-depression comorbidity attribute part of the link to shared serotonergic pathways, so central serotonin augmentation is more often studied as a cough suppressant than a cough cause. The atlas numbers agree with that null reading rather than with an ADR: enrichment is exactly 1.0, so cough at 0.69% of citalopram users is no commoner than among users of other drugs, FAERS disproportionality is weak (PRR 1.52, 1/3), and the drug is not on the BNF cough list. Temporality of 100% over 67 dated participants with a 3.4-year median lag is uninformative in a cohort that compresses this measure upward, and a 3-year gap to a self-limited symptom code carries no causal weight. What is left is a single rare intronic SNV whose effect size is far out of proportion to anything the clinical literature would predict.","gene_comment":"ENSG00000287292 is an unnamed novel HAVANA lncRNA spanning chr4:148.44-149.02 Mb, and rs116595899 (chr4:148,774,352, C/T, global MAF 0.29%) sits in its intron inside a gene desert whose nearest protein-coding neighbour, NR3C2, ends ~330 kb away. There is no mechanistic story here connecting the locus to the cough reflex, airway sensory nerves or citalopram pharmacokinetics, and the gene assignment should not be presented as mechanism.","score_check":"beta_adr 5.495 (OR ~240) at a 0.3% MAF variant with roughly 87 cough cases implies a handful of carriers and is the classic shape of sparse-data separation bias, even though log10p 6.84 implies z about 5.4. The comparison arms are consistent with noise rather than biology: the disease arm is flatly null (beta 0.131 +/- 0.147, |beta| well under 1.96*se), prescription propensity is null (log10p 0.91), and z_diff 5.08 is driven entirely by the unstable treated-arm estimate.","candidability":2,"candidability_reason":"No enrichment, null disease and prescription arms, weak FAERS, no citalopram-cough literature, and an implausibly large effect at a rare intronic variant in an unannotated lncRNA.","sources":[{"title":"CELEXA (citalopram) tablet label - DailyMed (Allergan)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=4259d9b1-de34-43a4-85a8-41dd214e9177"},{"title":"Europe PMC search: chronic cough and antidepressants/neuromodulators (duloxetine RCT PMID 41530764; amitriptyline PMID 40716800; cough-depression mechanisms PMID 41936621)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22chronic%20cough%22%20AND%20%28antidepressant%20OR%20amitriptyline%20OR%20paroxetine%20OR%20serotonin%29%20AND%20treatment&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC search: SSRI-associated interstitial lung disease and eosinophilic pneumonia (PMIDs 40909069, 39851477, 33581985, 33948187)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%28%22selective%20serotonin%20reuptake%20inhibitor%22%20OR%20citalopram%20OR%20sertraline%29%20AND%20%28%22interstitial%20lung%20disease%22%20OR%20%22eosinophilic%20pneumonia%22%20OR%20%22drug-induced%20lung%20disease%22%29&format=json&pageSize=20&resultType=core"},{"title":"Europe PMC search: citalopram/escitalopram and cough reflex or chronic cough (no drug-specific hits)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%28citalopram%20OR%20escitalopram%29%20AND%20%28%22cough%20reflex%22%20OR%20%22cough%20hypersensitivity%22%20OR%20%22chronic%20cough%22%29&format=json&pageSize=25&resultType=core"},{"title":"Ensembl REST lookup: ENSG00000287292 (novel lncRNA, chr4:148,442,685-149,024,624)","url":"https://rest.ensembl.org/lookup/id/ENSG00000287292?content-type=application/json;expand=0"},{"title":"Ensembl REST variation: rs116595899 (chr4:148,774,352, intron variant, MAF 0.0029)","url":"https://rest.ensembl.org/variation/human/rs116595899?content-type=application/json"}]},"citalopram|R074":{"verdict":"contested","headline":"Citalopram both treats and can provoke chest pain; the atlas variant sits in a 2q14.3 gene desert","assessment":"Chest pain (R07.4) is not a licensed indication for citalopram, whose label indication is major depressive disorder, but it is not simply an adverse effect either. Citalopram is an established off-label neuromodulator for functional/non-cardiac chest pain: intravenous citalopram raised mechanical and chemical oesophageal perception thresholds in hypersensitive subjects (Broekaert 2006), and a Rome IV functional-chest-pain cohort reported symptom improvement in 47% on citalopram 20 mg versus 11% untreated. That is a documented effect in the OPPOSITE direction and weighs heavily against a straightforward ADR reading. There is nevertheless a real cardiac liability in the other direction: the label carries a dose-dependent QTc prolongation warning with torsade de pointes and sudden death, dose caps of 20 mg over age 60 and in CYP2C19 poor metabolisers, frequent tachycardia, and infrequent angina pectoris and myocardial ischaemia, and MHRA contraindicates it in congenital long QT. Nonspecific chest pain is also the classic somatic presentation of the anxiety and panic comorbidity that gets patients prescribed an SSRI in the first place, so an R07.4 code in this population is diagnostically ambiguous by construction. The atlas numbers do not settle the direction: cotx_pct is only 2.01% with enrichment 0.62, i.e. chest pain is less common among citalopram users than among users of other drugs, and FAERS support is weak (PRR 1.63, 1 of 3).","gene_comment":"rs139074265 (chr2:122,983,571, MAF ~0.0026) is annotated intergenic and lies in a 2q14.3 gene desert; ENSG00000295630 is an unnamed novel lncRNA whose 5' end is ~4 kb away, and the nearest protein-coding genes (TSN, CNTNAP5) are over 1 Mb distant. This assignment carries no mechanistic content and should not be presented as one.","score_check":"The drug-specific pattern is internally consistent (disease arm flatly null at beta 0.057 +/- 0.086, prescribing arm null, z_diff 4.83), but beta_adr 2.841 for a variant at MAF 0.26% is the shape of a sparse-data estimate and needs a standard error and carrier count before it is believable. Temporality of 99.3% over 275 dated cases is uninformative here given the cohort's known upward compression.","candidability":3,"candidability_reason":"Nonspecific symptom code in an anxiety-rich population, a drug that is used to treat this very symptom, and a rare intergenic variant in a gene desert with no mechanism.","sources":[{"title":"Citalopram tablet label (DailyMed, Amneal Pharmaceuticals) - indications, QT warning, cardiovascular adverse reactions","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=6a402640-1c9d-4fca-8b5b-e1ac1dd4ca5a"},{"title":"MHRA Drug Safety Update: Citalopram and escitalopram - QT interval prolongation","url":"https://www.gov.uk/drug-safety-update/citalopram-and-escitalopram-qt-interval-prolongation"},{"title":"Broekaert et al., Influence of citalopram on oesophageal hypersensitivity: a double-blind, placebo-controlled study (Aliment Pharmacol Ther 2006)","url":"https://pubmed.ncbi.nlm.nih.gov/16422995/"},{"title":"Efficacy of citalopram or amitriptyline versus no treatment in patients with functional chest pain (Ann Gastroenterol 2023, PMC9756034)","url":"https://europepmc.org/article/MED/36593804"},{"title":"Psychological comorbidity in patients presenting to the emergency department with low-risk chest pain and anxiety (PMC12828106)","url":"https://europepmc.org/article/MED/40781746"},{"title":"Ensembl REST annotation for rs139074265 (intergenic, 2:122983571, MAF 0.0026) and ENSG00000295630 (novel lncRNA)","url":"https://rest.ensembl.org/variation/human/rs139074265?content-type=application/json"}]},"citalopram|Z966":{"verdict":"statistical artifact","headline":"Z96.6 is a device-status code, not an ADR; the TP63 hit is a rare-variant sparse-data artifact","assessment":"Z96.6 codes the presence of an orthopaedic joint implant - a device-status flag recorded at or after arthroplasty, not a disease and not an adverse reaction, so a within-user GWAS of it tests who has already had a hip or knee replaced. Citalopram's label carries a single licensed indication (major depressive disorder) and no fracture or bone-density warning, though osteoporosis, arthralgia and arthritis appear as low-frequency post-marketing terms and hyponatraemia-related unsteadiness is flagged as a fall mechanism. There is a genuine SSRI bone signal in the literature (any-fracture HR 1.77, 95% CI 1.15-2.74 in women initiating SSRIs; HR 2.39 for fracture with fluoxetine in a post-stroke IPD meta-analysis) and a genuine perioperative one (OR 1.78 for transfusion after lower-limb arthroplasty), but neither makes 'has a prosthesis' an ADR endpoint, and the arthroplasty-specific literature points the other way: fluoxetine and paroxetine were associated with LOWER risk of THA/TKA in osteoarthritis (RR 0.68 and 0.79), and SSRI initiators had lower TJA risk than centrally acting analgesic initiators (HR 1.81 favouring SSRI). The atlas agrees with that direction rather than an ADR reading - enrichment 0.52 means implants are about half as common in citalopram users as in users of other drugs. Temporality of 100% carries no information here because a Z96.6 status code cannot logically be entered before the implant exists, and the ~1.32% prevalence gives only 157 datable cases. The remaining claim is therefore purely genetic, and it does not survive inspection: beta_adr 4.606 (OR ~100) with a p-value implying z about 5.5, i.e. SE near 0.85, is the signature of a handful of carriers, while the same variant is flatly null for the condition off-drug (beta -0.031 +/- 0.112) and null for being prescribed citalopram at all.","gene_comment":"rs186308256 is an intronic variant at chr3:189,833,228 inside TP63 (3q28) with a minor allele frequency of only ~0.1-0.2%; TP63 is an ectodermal/limb-development transcription factor whose human phenotypes are split-hand/foot malformation and EEC/Rapp-Hodgkin syndromes, with no established link to osteoarthritis, bone mineral density, prosthesis survival or serotonergic pharmacology. An intronic ultra-rare allele in that gene is a positional label, not a mechanism.","score_check":"The numbers do not hang together: an OR near 100 for a ~0.2%-frequency allele on a 1.3%-prevalence status code, with a null disease arm and a null prescription arm, is a sparse-data separation artifact rather than a real effect. The z_diff of 5.44 simply restates the same unstable within-user estimate against a disease effect that is indistinguishable from zero.","candidability":1,"candidability_reason":"The outcome is definitionally a device-status code rather than a reaction, the drug is if anything associated with fewer joint replacements, and the single ultra-rare TP63 variant shows a sparse-data effect size no follow-up would replicate.","sources":[{"title":"Citalopram tablets, film coated - US prescribing information (DailyMed, Aurobindo)","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=2632b547-2e13-447f-ac85-c774e437d6a8"},{"title":"The risk of fracture among women starting selective serotonin reuptake inhibitors (J Bone Miner Res, 2025; PMID 40757575)","url":"https://europepmc.org/article/MED/40757575"},{"title":"Fluoxetine and Paroxetine Exhibit a Protective Effect Against Total Joint Arthroplasty in Patients With Osteoarthritis (Cureus, 2025; PMID 41116942)","url":"https://europepmc.org/article/MED/41116942"},{"title":"Associations of centrally acting analgesic use with total joint arthroplasty compared to SSRI among osteoarthritis patients (BMC Med, 2026; PMID 42129763)","url":"https://europepmc.org/article/MED/42129763"},{"title":"Use of serotonergic antidepressants and perioperative complications in lower limb arthroplasty: systematic review and meta-analysis (J Orthop, 2025; PMID 40741070)","url":"https://europepmc.org/article/MED/40741070"},{"title":"Ensembl REST: variant record for rs186308256 (chr3:189833228, intron_variant, MAF ~0.001)","url":"https://rest.ensembl.org/variation/human/rs186308256?content-type=application/json;pops=1"}]},"clopidogrel|I639":{"verdict":"licensed indication","headline":"Cerebral infarction is a labelled clopidogrel indication; drug lowers it. Hit is an intergenic X pseudogene.","assessment":"Cerebral infarction (I63.9) is an explicit licensed indication for clopidogrel: the US label lists patients with a history of recent stroke, recent MI or established peripheral arterial disease, and CAPRIE randomised 19,185 such patients, showing an 8.7% relative risk reduction in the ischaemic stroke/MI/vascular death cluster versus aspirin. The drug therefore moves this condition in the opposite direction to an adverse reaction, which is decisive evidence against an ADR reading of the raw co-occurrence; the atlas's own indication score of 55.8 and enrichment of 1.37 are consistent with prescribing rather than causation, and the FAERS PRR of 4.58 is exactly what confounding by indication plus 'drug ineffective' reporting produces in spontaneous data. There is nevertheless a legitimate pharmacogenomic phenotype hidden inside this cell: ischaemic stroke occurring while on clopidogrel is treatment failure, and CPIC's neurovascular recommendation plus the CHANCE genetic substudy (clopidogrel-aspirin cut new stroke only in CYP2C19 loss-of-function non-carriers, HR 0.51 vs 0.93, P=0.02 for interaction) establish that host genotype modifies it. That is the comparison this atlas cell would need to make, and the locus it recovered is not CYP2C19 or anything downstream of it. The temporality of 84.8% after first exposure with a median of 1.7 years is uninformative here for two reasons: the fact sheet warns the cohort compresses it upward, and both recurrent stroke on treatment and first stroke in a patient anticoagulated for ACS would land in the same bucket, from only 99 dated events. The variant is null in the drug-free population (log10p 0.14, beta 0.063 +/- 0.18), so the z_diff of 5.41 is driven entirely by the within-users arm and rests on the same 99 events.","gene_comment":"rs145901271 is an intergenic X:40,936,652 variant (Ensembl most_severe_consequence: intergenic_variant, gnomAD non-Finnish European MAF ~0.6%) lying ~800 bp from RPS2P55, a processed ribosomal-protein pseudogene, inside a ~150 kb Xp11.4 gap whose nearest protein-coding neighbour, USP9X, is ~149 kb away. There is no gene here in any meaningful sense and no mechanistic link to clopidogrel bioactivation, which runs through CYP2C19 on chromosome 10.","score_check":"A beta of 4.514 (OR ~90) for a 0.6% allele on the X chromosome derived from ~99 dated events is the signature of sparse-data separation, plausibly aggravated by hemizygous male coding in a heavily male cardiovascular population, and log10p 7.73 is not convincing for a multi-phenotype within-drug scan. The comparison arms are internally coherent (flat disease and prescription arms), but coherence across arms does not rescue an effect size that large from that few carriers.","candidability":2,"candidability_reason":"The condition is the drug's own labelled indication and clopidogrel demonstrably reduces it, and the locus is an intergenic pseudogene region with an effect size consistent with sparse-data artifact.","sources":[{"title":"Clopidogrel bisulfate tablet label (DailyMed, Dr. Reddy's) - Indications and Usage, boxed warning on CYP2C19 poor metabolisers","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=45caaa95-8470-b459-5f82-8156072d1878"},{"title":"CPIC Guideline for Clopidogrel and CYP2C19 (Medical Genetics Summaries / PharmGKB, NBK84114)","url":"https://www.ncbi.nlm.nih.gov/books/NBK84114/"},{"title":"Wang Y et al. Association Between CYP2C19 Loss-of-Function Allele Status and Efficacy of Clopidogrel ... Minor Stroke or TIA (CHANCE). JAMA 2016;316(1):70-8","url":"https://pubmed.ncbi.nlm.nih.gov/27348249/"},{"title":"CAPRIE Steering Committee. A randomised, blinded, trial of clopidogrel versus aspirin in patients at risk of ischaemic events. Lancet 1996;348:1329-39","url":"https://pubmed.ncbi.nlm.nih.gov/8918275/"},{"title":"Ensembl REST variation record for rs145901271 (intergenic_variant, X:40936652, gnomAD frequencies)","url":"https://rest.ensembl.org/variation/human/rs145901271?content-type=application/json"},{"title":"Ensembl gene record for RPS2P55 (processed pseudogene, X:40934982-40935864)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/RPS2P55?content-type=application/json"}]},"clopidogrel|K30":{"verdict":"plausible ADR","headline":"Dyspepsia is a label-listed clopidogrel ADR, but this ultra-rare SLC8A1 intronic hit is sparse-data noise","assessment":"Dyspepsia is not an indication for clopidogrel, whose licence is confined to ACS, recent MI, recent stroke and peripheral arterial disease; it is, however, an explicitly labelled adverse reaction in the EU SmPC, where gastrointestinal haemorrhage, diarrhoea, abdominal pain and dyspepsia are all listed as Common, with gastric/duodenal ulcer and gastritis Uncommon. The US Plavix label is more conservative and lists no dyspepsia at all, only postmarketing colitis, pancreatitis, stomatitis, gastric/duodenal ulcer and diarrhoea, and it records that in CAPRIE the only non-bleeding event more frequent on clopidogrel than aspirin was pruritus. A drug-caused mechanism is therefore plausible but weak and partly backwards: clopidogrel produced less GI haemorrhage than aspirin in CAPRIE (2.0% vs 2.7%) and less nausea, vomiting, dyspepsia and diarrhoea than ticagrelor in the FDA safety reviews, so it is the least gastrotoxic of the common antiplatelets, and most upper-GI symptoms in its users arise from the co-prescribed aspirin of dual antiplatelet therapy rather than from clopidogrel itself. The atlas numbers are consistent with a mild, non-specific effect rather than confounding by indication: enrichment is exactly 1.0, so dyspepsia is no commoner in clopidogrel users than in users of other drugs, indication scores only 18, and 100% of the 120 datable cases were first recorded after first exposure (median 1.32 years), with no undated records. Against that, the FAERS signal is weak (1/3, PRR 1.59), and the same variant is flatly null for dyspepsia in the drug-free population (beta 0.128 +/- 0.143, |beta| well under 1.96*se, log10p 0.43), so the entire z_diff of 4.77 rests on the within-users effect. That within-users effect is not believable as stated, and it is what the finding actually stands or falls on. The clinical story (a real but minor labelled GI ADR, mostly aspirin-driven) survives; the pharmacogenomic story does not.","gene_comment":"rs75077292 is an intronic SNV at chr2:40,187,733 inside SLC8A1 (NCX1, the Na+/Ca2+ exchanger) and its antisense lncRNA SLC8A1-AS1, with no coding consequence and no entry in the GWAS Catalog; NCX1 has smooth-muscle and megakaryocyte biology but no established link to dyspepsia or to clopidogrel handling, whose only actionable pharmacogene is CYP2C19. The gene assignment is positional only and should not be presented as mechanism.","score_check":"The variant is ultra-rare (1000G MAF 0.27%, gnomAD NFE 0.5%) yet carries beta_adr 4.976, i.e. an odds ratio near 145 with an implied standard error around 1.0 - the classic signature of quasi-separation on a handful of carriers, and log10p 6.12 is not genome-wide significant. The cohort-level numbers (cotx 1.26%, enrichment 1.0, temporality 100% of 120 dated) are internally coherent, but the genetic effect size is not.","candidability":4,"candidability_reason":"Real but mild label-listed ADR that is largely attributable to concomitant aspirin, resting on an ultra-rare intronic variant whose implausible effect size makes the genetics uninterpretable.","sources":[{"title":"PLAVIX (clopidogrel bisulfate) US prescribing information - DailyMed (sanofi-aventis)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=de8b0b67-eb25-4684-83b5-7ad785314227"},{"title":"Plavix EPAR Product Information (EU SmPC), section 4.8 - European Medicines Agency","url":"https://www.ema.europa.eu/en/documents/product-information/plavix-epar-product-information_en.pdf"},{"title":"Clopidogrel - StatPearls, NCBI Bookshelf","url":"https://www.ncbi.nlm.nih.gov/books/NBK470539/"},{"title":"Serebruany et al. Gastrointestinal adverse events after dual antiplatelet therapy: clopidogrel is safer than ticagrelor. Cardiology 2013","url":"https://pubmed.ncbi.nlm.nih.gov/23860246/"},{"title":"Ng et al. Management and outcome of peptic ulcers or erosions in patients receiving aspirin plus clopidogrel. J Gastroenterol 2008","url":"https://pubmed.ncbi.nlm.nih.gov/18807129/"},{"title":"Ensembl VEP / variation record for rs75077292 (intronic, SLC8A1)","url":"https://rest.ensembl.org/vep/human/id/rs75077292?content-type=application/json"},{"title":"Open Targets Platform - SLC8A1 (ENSG00000183023) target record","url":"https://platform.opentargets.org/target/ENSG00000183023"}]},"clopidogrel|K922":{"verdict":"plausible ADR","headline":"GI haemorrhage is a labelled clopidogrel ADR, but the OR9Q1 hit is a rare-variant artifact","assessment":"Gastrointestinal haemorrhage is not an indication for clopidogrel but a recognised, labelled adverse effect: the Plavix label states that P2Y12 inhibitors including clopidogrel increase the risk of bleeding, reports GI haemorrhage in 2.0% of CAPRIE patients on clopidogrel (versus 2.7% on aspirin), and in CURE major bleeding of 3.7% versus 2.7% when added to aspirin. Pharmacovigilance agrees independently of this atlas: a 2026 FAERS/JADER disproportionality analysis of drug-induced upper GI bleeding ranked clopidogrel 6th in FAERS (ROR 16.7, 1817 cases) and 2nd in JADER (ROR 10.35), matching the fact sheet's STRONG FAERS flag (PRR 6.04). The mechanism is not mucosal toxicity but irreversible P2Y12 blockade: pre-existing erosions, ulcers and angiodysplasia bleed and fail to seal, which is why clopidogrel is dose-independent in a way that erosive drugs are not, and why the risk concentrates in older, aspirin- and NSAID-co-treated, comorbid patients. There is one comparator subtlety worth stating: against aspirin clopidogrel's GI bleeding rate is slightly LOWER (CAPRIE), so clopidogrel is not the worst antiplatelet for the gut - but against no antiplatelet it clearly raises bleeding, so this is not a backwards signal. The atlas descriptives are consistent with a treatment-emergent event rather than an indication: enrichment 1.03 means clopidogrel users carry no more GI bleeding history than users of other drugs, temporality is 98.3% over 115 dated cases with no undated records and a median 2.12 years after first exposure, and the same variant is completely flat in the drug-free population (log10p_disease 0.11, beta_disease -0.043 +/- 0.144), so this is not a predisposition locus leaking through. What does not hold up is the genetics: the phenotype is a real ADR, but nothing here identifies a pharmacogenomic determinant of it.","gene_comment":"rs181975950 is an intronic SNV in OR9Q1, an olfactory receptor gene at 11q12 with no platelet, coagulation or mucosal role, no GWAS Catalog associations and no relevant literature; it is rare (T allele ~0.3% in gnomAD/TOPMED), so beta_adr 4.616 (OR ~100) across only 115 cases means a handful of carriers and is a sparse-data artifact, not an effect size. The established pharmacogene for clopidogrel is CYP2C19, and its boxed warning concerns loss of antiplatelet efficacy in poor metabolizers, not bleeding.","score_check":"The epidemiological numbers are believable and mutually consistent (no enrichment, late onset, null disease arm, strong external FAERS signal), but the variant statistics are not: an odds ratio near 100 on a ~0.3%-frequency intronic variant with 115 cases is the classic signature of a few-carrier sparse-data blow-up, and z_diff 5.04 inherits that instability. The atlas is right about the condition and almost certainly wrong about the locus.","candidability":5,"candidability_reason":"A genuine, well-documented drug-caused ADR with clean temporality and no confounding by indication, but the genetic hit is an implausible rare intronic olfactory-receptor variant with an artifactual effect size, so there is no pharmacogenomic lead to pursue.","sources":[{"title":"PLAVIX (clopidogrel bisulfate) prescribing information, DailyMed (Sanofi-Aventis)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=de8b0b67-eb25-4684-83b5-7ad785314227"},{"title":"Clopidogrel Therapy and CYP2C19 Genotype - Medical Genetics Summaries (NCBI Bookshelf)","url":"https://www.ncbi.nlm.nih.gov/books/NBK84114/"},{"title":"Drug-induced upper gastrointestinal bleeding: a real-world pharmacovigilance study (FAERS/JADER)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12928464/"},{"title":"Characteristics, Risk Stratification, and Outcomes of Upper Gastrointestinal Bleeding in Patients Receiving Antithrombotic Therapy","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13114178/"},{"title":"dbSNP record for rs181975950 (intronic, OR9Q1, chr11:58060137)","url":"https://www.ncbi.nlm.nih.gov/snp/rs181975950"}]},"clopidogrel|M139":{"verdict":"plausible ADR","headline":"Arthritis is on clopidogrel's label as a postmarketing reaction; the rare intronic PHEX hit is unconvincing","assessment":"Clopidogrel is licensed only to reduce myocardial infarction and stroke in acute coronary syndrome, recent MI or stroke, and established peripheral arterial disease, so M13.9 arthritis is not an indication. It is, however, explicitly listed in the postmarketing section of the US prescribing information under 'Musculoskeletal, connective tissue and bone disorders: Myalgia, arthralgia, arthritis', with the label's own caveat that these are spontaneous reports of unknown frequency for which causality cannot be reliably established. No mechanistic literature describes clopidogrel-induced arthritis; the two routes that would make sense are bleeding into a joint from P2Y12 blockade (bleeding is the drug's dominant toxicity, and acquired haemophilia A is a recognised postmarketing event) and an immune-mediated thienopyridine hypersensitivity of the kind implied by the labelled serum sickness and rash entries, but neither is documented for joint disease specifically, and a targeted EuropePMC search returned no case series or pharmacovigilance study on clopidogrel and arthritis or haemarthrosis. Clopidogrel does not treat arthritis, so there is no opposite-direction argument against an ADR reading. The atlas numbers argue against crude confounding by indication rather than for a mechanism: arthritis is depleted, not enriched, among clopidogrel users (enrichment 0.69, cotx 0.89%), indication scores 11.2 and predisposition 0, and the same variant is flat in the drug-free population (beta 0.081 +/- 0.066, well inside noise), giving z_diff 4.94. Against that, FAERS disproportionality is essentially null (PRR 1.17, weak 1/3), M13.9 is a wastebasket code that in an elderly vascular cohort will mostly capture osteoarthritis and undifferentiated joint pain, and the 98% temporality rests on only 100 dated participants in a cohort the atlas itself says inflates this metric.","gene_comment":"rs141297234 is a rare C/T intronic variant at chrX:22,300,612 inside a very large PHEX intron, in a stretch also covered by the PTCHD1-AS lncRNA; PHEX is the X-linked hypophosphatemia gene, and adult XLH does cause enthesopathy and joint disease, so the locus is bone-adjacent rather than absurd. But it is a single rare non-coding hit with no burden support, and PHEX has no relationship to clopidogrel pharmacology, which runs through CYP2C19, CES1 and P2Y12 - this is not a mechanism.","score_check":"beta 2.326 at log10p 6.64 implies an odds ratio near 10 for a rare X-linked variant on a nonspecific ICD code, the classic shape of a sparse-data, winner's-curse estimate rather than a real effect size. The internal comparisons are consistent (null in the untreated, null for prescription propensity, z_diff 4.94), so the signal is drug-specific if it is real, but the magnitude should not be taken at face value.","candidability":5,"candidability_reason":"Arthritis is a genuine label-listed postmarketing reaction and confounding by indication is unlikely here, but the genetics are a single rare intronic variant in a pharmacologically irrelevant gene with a sparse-data effect size and near-null FAERS support.","sources":[{"title":"Clopidogrel tablet, film coated (Aurobindo Pharma) - DailyMed prescribing information","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=0078fb3d-3595-4ae1-a059-1d5e81c879cf"},{"title":"Ensembl REST: variation rs141297234 (chrX:22,300,612, intron variant)","url":"https://rest.ensembl.org/variation/human/rs141297234?content-type=application/json"},{"title":"Ensembl REST: genes overlapping chrX:22,250,612-22,350,612 (PHEX, PTCHD1-AS, CBLL2)","url":"https://rest.ensembl.org/overlap/region/human/X:22250612-22350612?feature=gene;content-type=application/json"},{"title":"PHEX gene - MedlinePlus Genetics","url":"https://medlineplus.gov/genetics/gene/phex/"},{"title":"Europe PMC search: X-linked hypophosphatemia and enthesopathy/osteoarthritis (incl. PMID 40800666)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22X-linked%20hypophosphatemia%22%20AND%20%28osteoarthritis%20OR%20enthesopathy%20OR%20arthritis%29&format=json&pageSize=15"},{"title":"Europe PMC search: clopidogrel and arthritis/arthralgia/haemarthrosis (no dedicated reports found)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=clopidogrel%20AND%20%28arthritis%20OR%20arthralgia%20OR%20haemarthrosis%20OR%20hemarthrosis%29&format=json&pageSize=25"}]},"clopidogrel|M542":{"verdict":"statistical artifact","headline":"Neck pain is not a clopidogrel effect, is depleted in its users, and the lncRNA hit is sparse-data","assessment":"Cervicalgia is not a licensed indication for clopidogrel: the FDA Plavix label indicates it only for acute coronary syndrome (UA/NSTEMI, STEMI) and for recent MI, recent stroke or established peripheral arterial disease. Nor is it a labelled adverse reaction: the postmarketing musculoskeletal entry lists only 'Myalgia, arthralgia, arthritis', with no neck pain or cervicalgia term, and the clinical-trial safety profile is dominated by bleeding. The atlas itself argues against any excess: only 0.89% of clopidogrel users carry M54.2 and the enrichment is 0.58, i.e. neck pain is roughly 40% less common in clopidogrel users than in users of other drugs, which is the opposite of what either an ADR or confounding by indication would produce. The one antiplatelet mechanism that can genuinely present as acute neck pain is bleeding, specifically spontaneous cervical spinal epidural haematoma, which is documented on clopidogrel but only in isolated case reports and would be coded as haematoma with neurological deficit rather than as nonspecific chronic cervicalgia; conversely, cervical artery dissection presents with neck pain and is then treated with antiplatelets, a route by which the diagnosis precedes the drug. Temporality of 68.7% rests on just 67 dated participants and, as the atlas notes, is compressed upward by prescription records reaching further back than diagnosis records, so it is weak evidence on its own. The comparison arms are flat, the disease arm is null (beta -0.177, se 0.129, so |beta| well under 1.96*se) and the prescription arm is null at log10p 0.93, leaving z_diff 5.52 as an artefact of the inflated within-users beta rather than a real treatment-specific effect; external pharmacovigilance support is weak (FAERS 1/3, PRR 1.3) and the pairing is not on the BNF.","gene_comment":"NR2F1-AS1 is an antisense lncRNA at 5q15 opposite NR2F1, studied almost entirely in cancer biology (with a single osteogenic fracture-healing report) and with no connection to nociception, cervical spine disease or platelet function; rs116458234 is an intronic variant at ~0.5% global and ~0.8-1.3% European frequency with no GWAS Catalog association. The only established clopidogrel pharmacogene is CYP2C19 (CPIC), and nothing at this locus belongs in a mechanistic story.","score_check":"beta_adr of 6.963 implies an odds ratio around a thousand for a ~1% intronic variant in a phenotype with 67 dated cases, which is the signature of quasi-separation on a handful of carriers rather than a real effect, and log10p 7.19 inherits that instability. The null disease and prescription arms are believable; the within-users estimate is not.","candidability":1,"candidability_reason":"Non-labelled, nonspecific musculoskeletal code that is depleted in clopidogrel users, with an implausibly large effect at a rare intronic lncRNA variant.","sources":[{"title":"PLAVIX (clopidogrel bisulfate) tablets - FDA prescribing information (indications; postmarketing adverse reactions; CYP2C19 boxed warning)","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2021/020839s074lbl.pdf"},{"title":"CPIC Guideline for CYP2C19 Genotype and Clopidogrel Therapy: 2022 Update (PMID 35034351) via Europe PMC","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22Clinical%20Pharmacogenetics%20Implementation%20Consortium%22%20AND%20TITLE%3Aclopidogrel&format=json&pageSize=5&resultType=core"},{"title":"Europe PMC search: clopidogrel AND (neck pain OR cervicalgia OR musculoskeletal pain) - no causal ADR literature","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=clopidogrel%20AND%20%28%22neck%20pain%22%20OR%20cervicalgia%20OR%20%22musculoskeletal%20pain%22%29&format=json&pageSize=20"},{"title":"Europe PMC search: spontaneous spinal epidural haematoma and clopidogrel (rare bleeding presentation with acute spinal pain)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22spontaneous%20spinal%20epidural%20hematoma%22%20AND%20clopidogrel&format=json&pageSize=10"},{"title":"Europe PMC search: NR2F1-AS1 literature (cancer-dominated lncRNA)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22NR2F1-AS1%22&format=json&pageSize=15"},{"title":"Ensembl REST variation record for rs116458234 (chr5:93,360,700, intronic, MAF ~0.5%)","url":"https://rest.ensembl.org/variation/human/rs116458234?content-type=application/json;pops=1"},{"title":"Ensembl gene summary ENSG00000237187 (NR2F1-AS1, antisense RNA)","url":"https://www.ensembl.org/Homo_sapiens/Gene/Summary?g=ENSG00000237187"}]},"clopidogrel|R13":{"verdict":"co-prescription population","headline":"Dysphagia is a stroke sequela in the very population clopidogrel treats, not a labelled clopidogrel effect","assessment":"Dysphagia (R13) is not an indication for clopidogrel: the Plavix label covers acute coronary syndrome, recent MI, recent stroke and established peripheral arterial disease, and its adverse-reaction sections list GI haemorrhage, gastroduodenal ulcer, colitis, pancreatitis, stomatitis and diarrhoea, with no dysphagia, oesophagitis or oesophageal ulcer. Recent stroke is itself a licensed indication, and post-stroke dysphagia is one of the commonest stroke sequelae, pooled at 42% across 42 studies (26,366 patients) in a 2022 BMC Geriatrics meta-analysis, so a clopidogrel cohort is structurally enriched for swallowing disorders before any drug effect is considered. The published clopidogrel-dysphagia literature is two case reports only, one of pill oesophagitis after supine dosing during PCI and one hypopharyngeal mass in a patient on aspirin plus clopidogrel, i.e. anecdotal mucosal-injury or bleeding mechanisms rather than a characterised ADR, matching the weak FAERS signal here (1/3, PRR 1.4) and absence from the BNF. The atlas numbers do not rescue an ADR reading: enrichment is 1.06, essentially no excess of dysphagia among clopidogrel users versus users of other drugs, and the 99% temporality with a 2.92-year median lag is exactly what the cohort's record-depth asymmetry produces and carries little weight. The within-user hit, beta 3.718 at log10p 6.33, implies a standard error near 0.74 for a variant with roughly 0.7% minor allele frequency in about 110 dated cases, an odds ratio near 40 resting on a handful of carriers and short of genome-wide significance, which is the classic shape of sparse-data inflation. The clean null in the drug-free population (beta 0.082 +/- 0.091) and the null prescription-propensity p make the signal user-specific, but noise in a small treated arm produces that same pattern, so the drug-specificity is not itself evidence of a drug-caused effect.","gene_comment":"rs140016081 is an intronic variant at chr4:42,078,960 (GRCh38) lying inside SLC30A9, a protein-coding gene, so the assignment is positionally defensible rather than an intergenic guess. SLC30A9 encodes the mitochondrial zinc antiporter ZnT9, whose only established disease link is biallelic Birk-Landau-Perez syndrome; it has no connection to clopidogrel bioactivation (CYP2C19/P2Y12), platelet function, oesophageal mucosa or swallowing physiology, so it supplies no mechanism.","score_check":"The within-user effect size is implausibly large for the implied error and carrier count, and it sits alongside a null background effect and a null prescription arm, which reads as sparse-data inflation in about a hundred cases rather than a stable signal. The epidemiological fields are internally consistent but uninformative: enrichment at 1.06 is a flat line, and the 99% temporality is the compressed value the design is known to generate.","candidability":2,"candidability_reason":"No enrichment, no label or pharmacovigilance support, a condition that is a hallmark sequela of the drug's own indication, and a rare intronic variant in a biologically irrelevant gene with a sparse-data-scale effect.","sources":[{"title":"PLAVIX (clopidogrel) label, DailyMed (Sanofi-Aventis U.S. LLC)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=de8b0b67-eb25-4684-83b5-7ad785314227"},{"title":"PubMed search: clopidogrel dysphagia (clopidogrel-induced pill oesophagitis, PMID 36277663; dysphagia with hypopharyngeal mass on clopidogrel/aspirin, PMID 25611864)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=clopidogrel+dysphagia"},{"title":"Prevalence of dysphagia and risk of pneumonia and mortality in acute stroke patients: a meta-analysis (BMC Geriatrics 2022, PMID 35562660)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:35562660&format=json&resultType=core"},{"title":"UniProt Q6PML9 - SLC30A9 / ZnT9, mitochondrial zinc antiporter, Birk-Landau-Perez syndrome","url":"https://rest.uniprot.org/uniprotkb/Q6PML9.txt"},{"title":"Ensembl variation record for rs140016081 (intron variant, chr4:42,078,960, MAF 0.69%)","url":"https://rest.ensembl.org/variation/human/rs140016081?content-type=application/json"},{"title":"Ensembl region overlap 4:42,060,000-42,100,000 confirming SLC30A9 (ENSG00000014824) spans the variant","url":"https://rest.ensembl.org/overlap/region/human/4:42060000-42100000?feature=gene;content-type=application/json"}]},"clopidogrel|T828":{"verdict":"co-prescription population","headline":"Device complications in stented patients: clopidogrel is given to prevent them, and the gene is a rare intron","assessment":"ICD-10 T82.8 covers complications of cardiac and vascular prosthetic devices, implants and grafts - embolism, fibrosis, haemorrhage, pain, stenosis and thrombosis of a stent, valve or graft - so by construction it can only be coded in people who already carry such a device. Clopidogrel's licensed indications are acute coronary syndrome, percutaneous coronary intervention with stent placement, and established atherosclerotic disease, i.e. exactly that population, so 1.33% of clopidogrel users carrying a T82.8 code is expected background; enrichment is null here, so the one honest test of confounding by indication was not run. For the dominant thrombotic and stenotic subcodes the drug is known to move the condition in the opposite direction - dual antiplatelet therapy after PCI exists to prevent stent thrombosis, and premature clopidogrel discontinuation is a recognised cause of it - which is decisive evidence against an ADR reading. The only mechanistically defensible drug-caused slice is T82.83, haemorrhage due to a device, which fits clopidogrel's principal adverse effect, but it cannot be separated inside an aggregated T82.8 code. Temporality of 98.2% (n=112, median 1.01 years) is uninformative here because the device, the prescription and the complication are placed in that order by clinical routine, not by pharmacology. The genuinely pharmacogenomic version of this association is well established and points elsewhere: CYP2C19 loss-of-function carriers form less active metabolite and have higher rates of stent thrombosis and MACE (CPIC strong recommendation to use prasugrel or ticagrelor in ACS/PCI intermediate and poor metabolisers; FDA boxed warning), and the atlas hit does not recover that signal. FAERS PRR 11.82 reflects the same reporting channel - device complications are reported in stented patients who are on clopidogrel - not causality.","gene_comment":"rs151015017 is an intronic variant at the HPSE2 locus on 10q24 with a gnomAD/TOPMed MAF of about 0.06%; HPSE2 encodes inactive heparanase-2, which lacks catalytic activity and is known clinically only for recessive urofacial (Ochoa) syndrome, with no described role in platelet function, coagulation or stent biology. CPIC states explicitly that no gene other than CYP2C19 is a validated determinant of clopidogrel response, so this assignment should not be presented as mechanism.","score_check":"beta_adr 4.594 (OR ~99) from a variant with 0.06% minor allele frequency is the classic sparse-data pattern, and the drug-free-population effect (beta 0.892 +/- 0.342, z=2.6) is far smaller, so the z_diff of 3.81 is more likely driven by a handful of carriers than by a real interaction. log10p_prescribed of 0.05 rules out a prescription-propensity path, but that does not rescue a rare-variant effect of this size.","candidability":2,"candidability_reason":"The condition is what clopidogrel is prescribed to prevent, the drug moves it in the opposite direction, and the variant is an ultra-rare intron in a gene with no antiplatelet biology.","sources":[{"title":"Clopidogrel Therapy and CYP2C19 Genotype - Medical Genetics Summaries, NCBI Bookshelf","url":"https://www.ncbi.nlm.nih.gov/books/NBK84114/"},{"title":"CPIC Guideline for CYP2C19 Genotype and Clopidogrel Therapy: 2022 Update (PMC9287492)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9287492/"},{"title":"ICD-10 T82.8 - Complications of cardiac and vascular prosthetic devices, implants and grafts (subcodes T82.81-T82.89)","url":"https://www.veroscribe.com/icd-10/codes/T82.8"},{"title":"UniProtKB Q8WWQ2 - Inactive heparanase-2 (HPSE2)","url":"https://rest.uniprot.org/uniprotkb/Q8WWQ2.txt"},{"title":"Ensembl REST - variant rs151015017 (intronic, MAF 0.0006, chr10:99237313 GRCh38)","url":"https://rest.ensembl.org/variation/human/rs151015017?content-type=application/json"},{"title":"Clinician-Patient Discord: Exploring Differences in Perspectives for Discontinuing Clopidogrel (PMC2932847)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC2932847/"}]},"dextropropoxyphene|M796":{"verdict":"licensed indication","headline":"Limb pain is what this analgesic was prescribed for; PLPP3 hit is a huge-effect, off-biology outlier","assessment":"Dextropropoxyphene is a weak mu-opioid analgesic that was authorised across the EU for short- and long-term pain, and M79.6 (pain in limb) is a pain symptom code that sits squarely inside that indication rather than outside it. The EMA 2009 referral that led to withdrawal (Commission decision June 2010) turned on narrow therapeutic index and overdose deaths, and on the finding that the drug was a weak analgesic with limited effectiveness - not on any pain-generating adverse effect; the documented adverse-effect profile is constipation, pruritus, drowsiness, nausea and sodium-channel-mediated cardiotoxicity, with no limb or musculoskeletal pain. Its real-world use was dominated by chronic musculoskeletal and rheumatological pain (co-proxamol in the UK was largely a rheumatology-population drug), so a limb-pain diagnosis in these users is the reason for the prescription in a large share of cases. The atlas numbers are equivocal on direction rather than supportive: temporality 57.2% (n_dated 152) is close to a coin flip and, given the known upward compression of this metric, means roughly half these patients were already coded with limb pain before first exposure, while enrichment 0.44 says limb pain is actually less frequent here than among users of other drugs, which argues the signal is not even a strong co-prescription marker. The only mechanistically interesting alternative - opioid-induced hyperalgesia or withdrawal myalgia on stopping - is a real opioid class phenomenon but is not described for dextropropoxyphene specifically, is not diagnosed as M79.6 in practice, and is not what a within-users GWAS of a symptom code would capture. FAERS shows no signal (PRR 1.19), and the drug has been off the market for over a decade, so there is no prospective route to follow this up.","gene_comment":"rs72666457 maps to PLPP3 (PPAP2B), a lipid phosphate phosphatase acting on phosphatidate/LPA/S1P in vascular endothelium and a well-known coronary-artery-disease GWAS locus with no described role in nociception, muscle, or opioid pharmacokinetics; the variant itself returns no record in the GWAS Catalog, so nothing independent supports it. LPA signalling has neuropathic-pain literature, but invoking that here would be reverse-engineering a mechanism from a single unreplicated SNV.","score_check":"beta_adr 4.64 at log10p 6.57 implies an odds ratio near 100 with a standard error close to 0.9 - a sparse-data magnitude that no common-variant effect on a symptom code should have, and the same variant is flatly null on limb pain in the drug-free population (beta 0.063, se 0.094), so z_diff 5.05 is driven entirely by the unstable treated-arm estimate. The phenotype counts are small (n_dated 152) and prescription propensity is null (log10p 0.39), which at least rules out a pure prescribing-channel artifact but leaves the effect size unbelievable at face value.","candidability":1,"candidability_reason":"Limb pain is the indication for an analgesic, not its adverse effect, and the single supporting variant is an off-biology locus with an implausibly large, unreplicated effect.","sources":[{"title":"Dextropropoxyphene - indications, adverse effects, cardiotoxicity and worldwide withdrawal","url":"https://en.wikipedia.org/wiki/Dextropropoxyphene"},{"title":"EMA referral: dextropropoxyphene-containing medicinal products (CHMP opinion 2009; Commission decision 2010)","url":"https://www.ema.europa.eu/en/medicines/human/referrals/dextropropoxyphene"},{"title":"UniProt O14495 - PLPP3 (PPAP2B), phospholipid phosphatase 3","url":"https://rest.uniprot.org/uniprotkb/O14495.txt"},{"title":"GWAS Catalog REST lookup for rs72666457 (no associations recorded)","url":"https://www.ebi.ac.uk/gwas/rest/api/singleNucleotidePolymorphisms/rs72666457/associations?projection=associationBySnp"},{"title":"Europe PMC: Ottewell & Walker, 'Co-proxamol: where have all the patients gone?', Rheumatology (Oxford) 2008 (PMID 18187524)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:18187524&resultType=core&format=json"},{"title":"Europe PMC: PPAP2B/PLPP3 in coronary and vascular biology (e.g. PMID 36575638)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22PPAP2B%22%20AND%20%22coronary%22&format=json&pageSize=8"}]},"diazepam|D649":{"verdict":"statistical artifact","headline":"Anaemia is not a diazepam effect; the signal is one rare intergenic SNV with an implausible OR","assessment":"Anaemia is not a licensed indication for diazepam - the FDA Valium label lists only anxiety disorders, acute alcohol withdrawal, adjunctive skeletal muscle spasm and adjunctive seizure disorders - and it is not a listed adverse reaction: the ADVERSE REACTIONS section covers CNS, GI, cardiovascular, psychiatric, urogenital and skin effects plus elevated transaminases, with the only haematological statement being that 'because of isolated reports of neutropenia and jaundice, periodic blood counts and liver function tests are advisable during long-term therapy' - neutropenia, not anaemia, and only as isolated reports. A targeted PubMed search for diazepam with haematologic toxicity, neutropenia, aplastic or haemolytic anaemia returns 30 records, none of which is a diazepam-induced blood disorder; the agranulocytosis/psychotropic literature that surfaces concerns phenothiazines and clozapine-class agents, not benzodiazepines. No benzodiazepine mechanism for anaemia is described - benzodiazepines have no marrow, haem-synthesis or haemolytic target - so a drug-caused reading has no pharmacological footing, and the atlas's own arms agree: FAERS shows no signal (PRR 1.07) and the condition is not a BNF-listed effect. The only route by which anaemia and benzodiazepine exposure genuinely travel together is confounding by indication through restless legs syndrome, where iron-deficiency anaemia is a recognised cause (25-30% of iron-deficient patients develop RLS) and benzodiazepines are among the drugs used, but the atlas contradicts even that: enrichment is 0.24, i.e. unspecified anaemia is about four-fold LESS common in diazepam users than in users of other drugs, so this population is if anything haematologically healthier than the comparison arm. Temporality of 98.1% on 107 dated participants is exactly the pattern the cohort inflates by construction (prescriptions predate diagnoses) and carries no weight against a null background. D64.9 'anaemia, unspecified' is in any case an administrative catch-all rather than a phenotype a variant would act on.","gene_comment":"rs117410228 has no gene assignment for good reason: Ensembl places it at chr20:9,059,027 as an intergenic variant with MAF 0.0073, in a gene desert whose nearest neighbours are PLCB4 (~9 kb away) and PLCB1 (~90 kb), phospholipase C beta genes with neuronal and craniofacial roles and no erythropoietic function. The GWAS Catalog holds no trait association for it, so there is no mechanism here to describe, only a locus label.","score_check":"beta_adr 3.287 is an odds ratio near 27 for a variant with 0.7% minor allele frequency in a stratum with 107 dated cases - a handful of carriers at most, the textbook shape of a sparse-data artifact rather than a real effect, and the disease arm confirms the variant does nothing (beta 0.087 +/- 0.059, well inside noise, log10p 0.86). z_diff 5.48 is therefore inherited entirely from the unstable within-users estimate, and the null prescription arm (log10p 0.04) plus enrichment 0.24 leave no coherent story for the numbers to tell.","candidability":1,"candidability_reason":"Non-indication, non-labelled, mechanism-free condition, depleted in treated patients, driven by one rare intergenic SNV with an implausibly large effect and a null disease arm.","sources":[{"title":"VALIUM (diazepam) tablets - FDA prescribing information (Roche, Reference ID 4029651)","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2016/013263s094lbl.pdf"},{"title":"Ensembl REST: variant rs117410228 (chr20:9,059,027, intergenic, MAF 0.0073)","url":"https://rest.ensembl.org/variation/human/rs117410228?content-type=application/json"},{"title":"Ensembl REST: genes overlapping chr20:8,559,027-9,559,027 (PLCB1, PLCB4, LAMP5, PAK5)","url":"https://rest.ensembl.org/overlap/region/human/20:8559027-9559027?feature=gene;content-type=application/json"},{"title":"Restless legs syndrome (Am J Health Syst Pharm 2006) - iron deficiency and RLS, benzodiazepines among treatments","url":"https://pubmed.ncbi.nlm.nih.gov/16914630/"},{"title":"Restless legs syndrome (Mayo Clin Proc 1997) - iron-deficiency anaemia as associated condition","url":"https://pubmed.ncbi.nlm.nih.gov/9070203/"},{"title":"PubMed search: diazepam AND (hematologic toxicity OR neutropenia OR aplastic anemia OR hemolytic anemia) - 30 records, none diazepam-induced","url":"https://pubmed.ncbi.nlm.nih.gov/?term=diazepam+AND+%28%22hematologic+toxicity%22+OR+neutropenia+OR+%22aplastic+anemia%22+OR+%22hemolytic+anemia%22%29"}]},"diazepam|J449":{"verdict":"contested","headline":"Benzodiazepines genuinely worsen COPD outcomes, but incident COPD with beta 5.2 is not that effect","assessment":"COPD is not a licensed indication for diazepam: the label lists anxiety, alcohol withdrawal, muscle spasm and adjunctive seizure use, and explicitly contraindicates severe respiratory insufficiency and sleep apnoea, with a dose reduction advised in chronic respiratory insufficiency because of respiratory depression risk. There is a real class-level respiratory hazard behind this pair - Vozoris et al. (Eur Respir J 2014) found new benzodiazepine users with COPD had more outpatient exacerbations (RR 1.45) and more emergency visits for COPD or pneumonia (RR 1.92) - so the drug does move COPD in the harmful direction, but it acts on exacerbations and hypercapnic decompensation in people who already have COPD, not on the development of airflow obstruction. The atlas phenotype here is J44.9, a chronic smoking-related structural disease that diazepam cannot cause; a first J44.9 record a median 4.1 years after first exposure in an anxious, older, largely smoking population is ascertainment, not incidence, and the 92.9% temporality is the uninformative direction. Enrichment of 0.28 says COPD is strongly depleted among diazepam users relative to other drugs' users, which fits label-driven avoidance of benzodiazepines in respiratory disease and argues against a co-prescription-population reading, but it also means the exposed-and-affected group is small (0.5% co-occurrence, n_dated 99). The FAERS PRR of 2.57 is consistent with real reporting of respiratory decompensation but is equally consistent with notoriety and with benzodiazepines being given for dyspnoea-related anxiety in COPD patients, so it does not discriminate. Taken together the drug-condition safety concern is genuine and the atlas signal is real in the FAERS sense, yet the specific within-user genetic hit does not represent it.","gene_comment":"CXCL13 is a credible COPD gene in its own right - it drives the cigarette-smoke-induced lymphoid follicles of severe COPD (Bracke, AJRCCM 2013) - and PDE1C sits in the cyclic-nucleotide phosphodiesterase family that is druggable in obstructive and pulmonary-vascular lung disease, but neither has any connection to GABA-A pharmacology or benzodiazepine handling, so they offer disease biology, not a drug-caused mechanism. Both are rare SNVs assigned to a nearest gene, so even the disease reading rests on an unverified assignment.","score_check":"beta_adr of 5.211 (OR ~180) for a binary outcome at two rare SNVs in a 0.5% co-occurrence stratum is the classic signature of sparse-data separation rather than a real effect, and the same variants are essentially null in the drug-free population (beta_disease 0.227 +/- 0.11, |beta| barely above 1.96*se, log10p 1.41). z_diff of 5.23 is therefore inflated entirely by the unstable treated-arm estimate, and log10p_prescribed of 0.07 only tells us the variants do not predict getting the prescription.","candidability":3,"candidability_reason":"Real class-level respiratory harm from benzodiazepines in COPD, but the atlas phenotype is incident chronic COPD rather than exacerbation, and the genetics are a sparse-data artifact with no pharmacological link.","sources":[{"title":"Diazepam - StatPearls (NCBI Bookshelf): indications, contraindications including severe respiratory insufficiency and sleep apnoea","url":"https://www.ncbi.nlm.nih.gov/books/NBK537022/"},{"title":"DailyMed - DIAZEPAM oral solution label (Chartwell): Indications, Contraindications, respiratory depression warnings","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=167ee280-be68-4a28-a7d8-de41df14bc48"},{"title":"Vozoris et al., Benzodiazepine drug use and adverse respiratory outcomes among older adults with COPD, Eur Respir J 2014 (PMID 24743966), via Europe PMC","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:24743966&resultType=core&format=json"},{"title":"PubMed search: benzodiazepines, COPD, exacerbation and respiratory outcomes (Vozoris 2014; Ann Am Thorac Soc 2019; Front Pharmacol 2020/2026)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=benzodiazepines+COPD+exacerbation+respiratory+outcomes"},{"title":"PubMed search: CXCL13 in COPD lymphoid follicles (Bracke et al., AJRCCM 2013, PMID 23742729; Litsiou et al. 2013; Faner et al. 2016)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=CXCL13+COPD+lymphoid+follicles"},{"title":"Europe PMC search: PDE1C / phosphodiesterase families in COPD, airway smooth muscle and pulmonary disease","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=PDE1C%20AND%20(COPD%20OR%20airway%20smooth%20muscle%20OR%20pulmonary)&resultType=core&format=json&pageSize=5"}]},"diazepam|M255":{"verdict":"statistical artifact","headline":"Joint pain falls inside diazepam's muscle-spasm indication; rare lncRNA hit has an impossible effect","assessment":"Arthralgia (M25.5) is not itself a licensed indication, but the Valium label indicates diazepam as an adjunct for relief of skeletal muscle spasm due to reflex spasm to local pathology, explicitly including inflammation of the muscles or joints and trauma, and as a spasticity adjunct, so patients with painful joints are exactly one of the groups the label points at; benzodiazepine use is also documented as heavily prescribed and frequently problematic in chronic musculoskeletal pain populations (J Pain 2023), while the trial evidence says it does not work (Friedman 2017: diazepam added to naproxen was no better than placebo for acute low back pain, and the 2023 Cochrane overview finds no place for it in low back pain). Arthralgia is not listed anywhere in the label's adverse-reaction section, which reports drowsiness, fatigue, muscle weakness, ataxia and, post-marketing, falls and fractures; FAERS is flat here (PRR 0.97, no signal) and the pair is not on the BNF list. The only routes to a drug-caused joint pain would be indirect and unsupported by this data: injury from benzodiazepine-related falls, or myalgia and stiffness during withdrawal being coded as joint pain. Argued the other way, the atlas cotx numbers do not actually show a treated-population effect either - joint pain is depleted, not enriched, among diazepam users (enrichment 0.46), so this is not a simple confounding-by-indication story so much as an association with no clinical anchor at all. Temporality is 56.3% over 190 dated participants, and since the cohort's prescription records reach back further than diagnoses this figure is inflated by construction, so a value barely above half is weak-to-unsupportive rather than reassuring. The clean comparison arms (log10p_disease 0.35, log10p_prescribed 0.55) look specific, but for a variant this rare a null in every arm is the expected result of low power, not evidence of drug-specificity.","gene_comment":"rs149121643 is an intronic variant in ENSG00000282828, an unnamed 'novel transcript' lncRNA at 2p16.3, with gnomAD MAF ~0.5% overall (0.7% non-Finnish European, 2% Finnish); the nearest protein-coding gene is NRXN1, ~130 kb away, and neurexin-1 is a synaptic adhesion molecule with no connection to GABA-A receptor pharmacology or to joint pathology. This is an anonymous rare intronic locus, not a mechanism, and it should not be described as a NRXN1 finding.","score_check":"beta_adr of 4.81 implies an odds ratio near 120 for a common, non-specific phenotype carried by a ~0.5%-frequency allele, which is a hallmark of sparse-data bias or quasi-separation rather than a real effect, and log10p 7.41 is only borderline even before correcting for the many drug-condition pairs tested. The disease arm confirms nothing is there (beta 0.097 +/- 0.128, well inside noise), so the apparent z_diff of 5.33 is driven entirely by the untrustworthy treated-arm estimate.","candidability":1,"candidability_reason":"No labelled or literature adverse effect, no plausible mechanism, flat FAERS, and a rare intronic lncRNA variant whose effect size is too large to be real.","sources":[{"title":"VALIUM (diazepam) tablet - full prescribing information, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=260e2041-2bb3-482f-850e-b5d47a7bdbe6"},{"title":"Friedman et al. Diazepam Is No Better Than Placebo When Added to Naproxen for Acute Low Back Pain. Ann Emerg Med 2017 (PMID 28187918)","url":"https://pubmed.ncbi.nlm.nih.gov/28187918/"},{"title":"Benzodiazepine Use and Dependence in Relation to Chronic Pain Intensity and Pain Catastrophizing. J Pain 2023 (PMID 36243316)","url":"https://pubmed.ncbi.nlm.nih.gov/36243316/"},{"title":"Pharmacological treatments for low back pain in adults: an overview of Cochrane Reviews, 2023 (PMID 37014979)","url":"https://pubmed.ncbi.nlm.nih.gov/37014979/"},{"title":"Ensembl REST lookup: ENSG00000282828 (lncRNA, novel transcript, chr2:49,563,344-49,784,585)","url":"https://rest.ensembl.org/lookup/id/ENSG00000282828?content-type=application/json"},{"title":"Ensembl REST variation: rs149121643 (intron variant, gnomAD MAF ~0.005)","url":"https://rest.ensembl.org/variation/human/rs149121643?content-type=application/json;pops=1"}]},"diltiazem|M255":{"verdict":"contested","headline":"Label lists infrequent osteoarticular pain, but an OR~280 intronic SQSTM1 hit over ~80 cases is sparse-data","assessment":"Diltiazem is licensed only for hypertension and chronic stable angina (plus rate control in the IV form), so joint pain (M25.5) is in no sense an indication. The US label does, however, list osteoarticular pain, myalgia, muscle cramps and gout among infrequent (<2%) or postmarketing adverse reactions, so a musculoskeletal complaint is a recognised, if non-specific and low-frequency, label term rather than an invented one, and the atlas FAERS flag (PRR 2.08, concordant ROR and IC) is consistent with that weak disproportionality. The atlas numbers argue against confounding by indication rather than for a strong effect: joint pain is present in only 2.91% of diltiazem users and is depleted relative to users of other drugs (enrichment 0.86), the variant has no effect on joint pain in the drug-free population (beta_disease 0.018 +/- 0.064, log10p 0.11) and no association with being prescribed diltiazem (log10p 0.28), so z_diff 5.27 is essentially just the within-user signal restated, not independent corroboration. What does not hang together is the effect size: beta_adr 5.633 at log10p 6.94 implies a standard error near 1.1 and an odds ratio around 280 for a common, non-specific symptom, which is the signature of quasi-separation on a rare allele in a small case set (n_dated 80), not a real pharmacogenomic effect. Temporality of 76.2% with a median 4.1 years to onset is uninformative here, because the brief notes the cohort compresses this upward and only a low value would carry weight. Mechanistically the most defensible route to musculoskeletal pain in diltiazem users is indirect and pharmacokinetic rather than genetic - diltiazem is a CYP3A4 and P-gp inhibitor that raises simvastatin AUC roughly 5-fold (hence the label's 10 mg simvastatin cap) and lovastatin AUC 3-4 fold, and it causes dose-dependent peripheral oedema in up to 15% at 480-540 mg - neither of which has anything to do with SQSTM1.","gene_comment":"rs75363646 maps to chr5:179,825,265 (GRCh38), an intronic variant inside SQSTM1 (ENSG00000161011); SQSTM1 is a genuine bone gene, but only through rare coding mutations such as P392L in Paget's disease, and common intronic SQSTM1 SNPs are not established risk variants for joint pain or osteoarthritis. The gene name is superficially attractive for a bone/joint phenotype, which makes it more important not to read mechanism into an intronic tag with a null effect in the disease arm.","score_check":"The within-user p-value is real arithmetic but the implied odds ratio of roughly 280 for joint pain, on ~80 dated cases and a rare allele, is not a believable biological effect and reads as sparse-data separation. The comparison arms are believable and are all null, which correctly rules out predisposition and prescribing bias but leaves the ADR effect itself unsupported.","candidability":3,"candidability_reason":"Non-specific label-level musculoskeletal term with weak FAERS support, no enrichment in treated patients, and a genetically implausible effect size on an uninformative intronic SQSTM1 variant.","sources":[{"title":"TIAZAC (diltiazem hydrochloride) extended-release capsules - full prescribing information, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=c567fe7e-887e-4291-a0d1-2dd3f25cbf25"},{"title":"DailyMed search: diltiazem hydrochloride product labels","url":"https://dailymed.nlm.nih.gov/dailymed/search.cfm?labeltype=all&query=diltiazem+hydrochloride"},{"title":"Ensembl REST - variation record for rs75363646 (intron variant, chr5:179,825,265, GRCh38)","url":"https://rest.ensembl.org/variation/human/rs75363646?content-type=application/json"},{"title":"Ensembl REST - gene overlap at chr5:179,825,265 returns SQSTM1 (ENSG00000161011)","url":"https://rest.ensembl.org/overlap/region/human/5:179825265-179825265?feature=gene;content-type=application/json"},{"title":"Europe PMC search - SQSTM1 and Paget's disease of bone (rare coding mutations, no common intronic joint-pain variant)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=SQSTM1%20AND%20%22Paget%20disease%20of%20bone%22&format=json&pageSize=10&resultType=core"},{"title":"Europe PMC search - calcium channel blocker peripheral oedema and amlodipine pharmacovigilance","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22calcium%20channel%20blocker%22%20AND%20%22peripheral%20edema%22%20AND%20mechanism&format=json&pageSize=10&resultType=core"}]},"fluoxetine|G560":{"verdict":"statistical artifact","headline":"CTS is a labelled fluoxetine hypersensitivity finding, but this rare GFOD1 intron hit is sparse-data noise","assessment":"Carpal tunnel syndrome is not an indication for fluoxetine, but it is genuinely named in the US label: section 5.3 (Allergic Reactions and Rash) lists carpal tunnel syndrome among the clinical findings reported alongside fluoxetine-associated rash, with fever, arthralgia, oedema, lymphadenopathy and proteinuria, all resolving on discontinuation. That labelled event is an acute, rare, reversible serum-sickness-like phenomenon, not the chronic idiopathic entrapment neuropathy captured by an ICD-10 G56.0 code a median 4.3 years after first exposure, so the label does not underwrite what the atlas has found. PubMed returns no studies of fluoxetine or SSRIs and carpal tunnel syndrome at all; the only antidepressant case reports are tranylcypromine, tricyclic overdose and lithium acting through induced hypothyroidism, and no SSRI mechanism on tenosynovial or median-nerve biology is described. The atlas cohort numbers argue against a drug effect rather than for one: cotx_pct is 1.27% with enrichment 0.86, i.e. carpal tunnel is if anything slightly less common in fluoxetine users than in users of other drugs, which is what you would not see if the drug were adding cases. Temporality of 98.5% over 130 dated participants is exactly the pattern the cohort's record-depth asymmetry manufactures and carries no weight on its own, and the FAERS support is weak (1/3, PRR 1.48). Established carpal tunnel genetics point at connective-tissue and IGF-1 signalling (the DIRC3-IGFBP5 locus shared with trigger finger) and at metabolic drivers such as type 2 diabetes, obesity and hypothyroidism, none of which touches this locus. What remains is a single rare intronic variant with an uninterpretably large effect, which is a sparse-data signature rather than a pharmacogenomic finding.","gene_comment":"rs11756572 is an intronic SNV in GFOD1 (Gfo/Idh/MocA-like oxidoreductase domain containing 1, chr6:13,455,966, 6p24.1) with gnomAD/TOPMed MAF ~0.5%; it has no GWAS Catalog associations and GFOD1 has no described role in nerve, tenosynovium or connective tissue. The gene assignment is positional only and supplies no mechanism.","score_check":"beta_adr 4.80 (OR ~120) for a 0.5%-frequency variant in roughly 130 cases means the signal rests on a handful of carriers near complete separation, which is the classic sparse-data inflation the fact sheet warns about. The comparison arms are consistently null (beta_disease 0.174 +/- 0.119, i.e. within noise; log10p_disease 0.84; log10p_prescribed 0.42), so the z_diff of 5.62 is driven entirely by the implausible treated-arm estimate rather than by a real contrast.","candidability":2,"candidability_reason":"Carpal tunnel appears in the label only as part of an acute hypersensitivity syndrome, the condition is not enriched in fluoxetine users (0.86), and the genetic signal is an implausibly large effect at a rare intronic variant in a gene with no relevant biology.","sources":[{"title":"Fluoxetine capsules, US prescribing information, DailyMed (section 5.3 Allergic Reactions and Rash lists carpal tunnel syndrome)","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=6f4e4e9f-9ded-452f-b3ef-ab6aa9c72952"},{"title":"PubMed search: fluoxetine AND carpal tunnel syndrome (no results)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=fluoxetine+carpal+tunnel+syndrome"},{"title":"PubMed search: antidepressant AND carpal tunnel syndrome (only tranylcypromine, lithium/hypothyroidism and TCA-overdose case reports)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=antidepressant+carpal+tunnel+syndrome"},{"title":"Patel et al., Shared genetic susceptibility between trigger finger and carpal tunnel syndrome: a genome-wide association study, Lancet Rheumatol 2022 (DIRC3-IGFBP5/IGF-1 axis)","url":"https://pubmed.ncbi.nlm.nih.gov/36043126/"},{"title":"Causal effects of type 2 diabetes, obesity, gout, and hypothyroidism on carpal tunnel syndrome: a Mendelian randomization study","url":"https://pubmed.ncbi.nlm.nih.gov/42059059/"},{"title":"Ensembl REST variation record for rs11756572 (intron variant, chr6:13,455,966, MAF 0.0053)","url":"https://rest.ensembl.org/variation/human/rs11756572?content-type=application/json"},{"title":"GWAS Catalog variant page for rs11756572 (no reported trait associations)","url":"https://www.ebi.ac.uk/gwas/variants/rs11756572"},{"title":"MyGene.info record for GFOD1 (Gfo/Idh/MocA-like oxidoreductase domain containing 1, 6p24.1-p23)","url":"https://mygene.info/v3/query?q=symbol:GFOD1&species=human&fields=symbol,name,genomic_pos,summary,map_location"}]},"fluoxetine|I48":{"verdict":"statistical artifact","headline":"Rare intronic PCLO variant with an implausible effect in ~106 AF cases; AF is depleted in fluoxetine users","assessment":"Atrial fibrillation is not an indication for fluoxetine (MDD, OCD, bulimia, panic disorder, and with olanzapine bipolar depression) and is not an established adverse reaction: the US label lists atrial fibrillation only in postmarketing reports explicitly flagged as having no established causal relationship, while the genuine cardiac signal for fluoxetine is QT prolongation and torsades de pointes, i.e. a ventricular repolarisation effect, not an atrial arrhythmogenic one. The best pharmacoepidemiology points away from a drug effect: in 785,254 Danish antidepressant initiators the AF hazard peaked BEFORE the first prescription (aHR 7.65 at 30-15 days pre-treatment, 4.29 in the last 15 days, falling to 3.18 in month 1 and 1.11 by 6-12 months), the signature of depression, acute illness and diagnostic work-up rather than of pharmacology; and a 116,125-patient new-user cohort found no association with chronic AF (RR 0.98, 95% CI 0.86-1.12) with no gradient by serotonin-reuptake potency, despite the 5-HT4 hypothesis. The atlas is internally consistent with that: AF affects only 0.9% of fluoxetine users and is strongly DEPLETED relative to users of other drugs (enrichment 0.23), so this is not confounding by indication either - it is simply a young, largely non-cardiac outpatient population, and FAERS shows no disproportionality (PRR 0.93). What remains is a single rare intronic variant whose effect is confined to the treated stratum (log10p_adr 6.54, z_diff 5.08) while being exactly null for AF in the drug-free population (beta 0.001 +/- 0.101), which is the pattern a subgroup-only false positive produces as readily as a true pharmacogenomic interaction. The temporality of 100% after exposure carries essentially no weight here, both because the cohort compresses it upward and because with a median 3.58 years to diagnosis it is equally compatible with ageing of a treated cohort.","gene_comment":"rs534970958 is an intronic variant in PCLO (chr7:82,781,217, GRCh38) with a global minor allele frequency around 0.4%; PCLO is a very large presynaptic-cytomatrix gene whose only genetic literature is a weakly replicated depression/psychiatric one, it has no role in atrial electrophysiology and does not appear among established AF GWAS loci (PITX2/4q25, ZFHX3, KCNN3 and similar). An intronic rare variant in a huge brain-expressed gene is a positional annotation, not a mechanism for atrial fibrillation.","score_check":"A beta of 3.78 (OR ~44) for a 0.4%-frequency variant tested in roughly 106 dated AF cases implies the signal rests on a handful of carriers, the classic sparse-data inflation regime, and log10p 6.54 does not even clear a genome-wide threshold. The null disease-arm estimate (0.001 +/- 0.101) and the enrichment of 0.23 are believable and both argue that nothing about AF is unusual in this drug's users.","candidability":1,"candidability_reason":"Not a labelled or literature-supported fluoxetine ADR, AF is depleted in users, no FAERS signal, and the sole variant is a rare intronic PCLO SNV with an implausibly large effect from very few carriers.","sources":[{"title":"PubMed search: antidepressant use and atrial fibrillation risk","url":"https://pubmed.ncbi.nlm.nih.gov/?term=antidepressant+atrial+fibrillation+risk"},{"title":"Fenger-Gron M et al., Depression, antidepressants, and the risk of non-valvular atrial fibrillation, Eur J Prev Cardiol 2019 (PMID 30452291, Europe PMC record)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:30452291&resultType=core&format=json"},{"title":"Lapi F et al., The use of antidepressants and the risk of chronic atrial fibrillation, J Clin Pharmacol 2015 (PMID 25427727, Europe PMC record)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:25427727&resultType=core&format=json"},{"title":"openFDA drug label record for fluoxetine (indications; postmarketing QT prolongation, torsades de pointes, atrial fibrillation)","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22fluoxetine%22+AND+adverse_reactions:%22atrial+fibrillation%22&limit=1"},{"title":"Ensembl REST variation record for rs534970958 (location, intron_variant, MAF)","url":"https://rest.ensembl.org/variation/human/rs534970958?content-type=application/json;pops=1"},{"title":"Europe PMC query: atrial fibrillation genome-wide association loci (PITX2/4q25 and related)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22atrial%20fibrillation%22%20AND%20%22genome-wide%20association%22%20AND%20loci%20AND%20review&resultType=core&format=json&pageSize=5"},{"title":"Europe PMC query: PCLO / piccolo and major depressive disorder genetics","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22PCLO%22%20AND%20%22piccolo%22%20AND%20%22major%20depressive%20disorder%22&resultType=core&format=json&pageSize=5"}]},"fluoxetine|K20":{"verdict":"statistical artifact","headline":"Oesophagitis is depleted in fluoxetine users and off-label; signal rests on one rare anonymous chr9 intron","assessment":"Oesophagitis (K20) is not a licensed indication for fluoxetine - the US label lists only major depressive disorder, OCD, bulimia nervosa and panic disorder - and it is not a labelled adverse reaction either: dyspepsia (7-10%) and nausea (21-29%) appear in the trial tables, and postmarketing reporting covers oesophageal ulcer, gastrointestinal haemorrhage and haematemesis, but oesophagitis, reflux and GERD are absent from the label. A drug-caused mechanism is only weakly supported: SSRIs deplete platelet serotonin and raise bleeding and upper-GI mucosal-injury risk (pooled OR 1.39, 95% CI 1.23-1.58 for major bleeding on antithrombotics), and antidepressant-mediated lower-oesophageal-sphincter relaxation has been proposed, but that mechanism is anticholinergic and belongs to tricyclics rather than to fluoxetine. The direction of effect actively argues against an ADR reading for the symptom end of this spectrum: the AGA clinical practice update on functional heartburn recommends SSRIs as neuromodulators, primary or add-on, and a randomised trial found fluoxetine as effective as diltiazem for symptom relief in distal oesophageal spasm - the drug is used to treat oesophageal complaints, not expected to cause them. The atlas numbers point the same way: enrichment 0.66 means coded oesophagitis is roughly a third less common among fluoxetine users than among users of other drugs, so there is no co-prescription or indication signal to explain away, and none to support. Temporality of 98.9% on only 92 dated participants, with a median of 5.06 years to first record, is exactly the pattern the cohort's prescription-versus-diagnosis record depth manufactures, and a five-year lag fits incidental age-related disease better than drug injury. FAERS is weak (1 of 3, PRR 1.98), which is the level of disproportionality any high-volume antidepressant generates against a common GI code.","gene_comment":"No gene is assigned: rs138483345 is a ~0.6% MAF intronic variant at chr9:27,719,841 inside LOC124902135, an uncharacterised locus in the gene desert between MOB3B and LINGO2, with no oesophageal, mucosal or pharmacokinetic candidate anywhere nearby. This is a positional label, not a mechanism, and it should not be written up as one.","score_check":"The within-user beta of 4.78 (OR ~119) for a variant carried by under 1% of people, in a condition with only ~92 datable cases, means a handful of carriers are driving the whole result - the classic sparse-data inflation signature, even though beta/se is internally consistent with z_diff 5.24. The clean null arms (disease beta -0.077 +/- 0.107, prescription log10p 0.6) rule out indication confounding but leave an uncorroborated, biologically implausible effect size.","candidability":1,"candidability_reason":"Not on the label, depleted rather than enriched in treated patients, opposite in direction to fluoxetine's actual use in oesophageal symptoms, and carried by an implausibly large effect at one anonymous rare intron.","sources":[{"title":"Fluoxetine capsules label (Accord Healthcare) - indications and adverse reactions, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=5d3f883a-ab3d-47e0-b24e-6f3a4e91ed70"},{"title":"AGA Clinical Practice Update on Functional Heartburn: neuromodulators including SSRIs as primary or add-on therapy (Gastroenterology 2020, PMID 32017911)","url":"https://pubmed.ncbi.nlm.nih.gov/32017911/"},{"title":"Comparison of the efficacy of diltiazem versus fluoxetine in the treatment of distal esophageal spasm: a randomized controlled trial (Arab J Gastroenterol 2024, PMID 37718154)","url":"https://pubmed.ncbi.nlm.nih.gov/37718154/"},{"title":"Nochaiwong et al., serotonin reuptake inhibitors and major bleeding risk with antithrombotics, meta-analysis (Ann Med 2022, PMID 34955074)","url":"https://pubmed.ncbi.nlm.nih.gov/34955074/"},{"title":"Brahm & Kelly-Rehm, Antidepressant-mediated gastroesophageal reflux disease (Consult Pharm 2011, PMID 21486738)","url":"https://pubmed.ncbi.nlm.nih.gov/21486738/"},{"title":"dbSNP rs138483345 - chr9:27,719,841, intronic in LOC124902135, gnomAD MAF ~0.6%","url":"https://www.ncbi.nlm.nih.gov/snp/rs138483345"}]},"fluoxetine|M549":{"verdict":"statistical artifact","headline":"Back pain is not a fluoxetine ADR; two rare X-linked variants give an implausible OR ~150","assessment":"Dorsalgia is not a licensed indication for fluoxetine (FDA labelling covers major depression, OCD, panic disorder, bulimia and, with olanzapine, bipolar and treatment-resistant depression) and it does not appear anywhere in the current US prescribing information: a full-text scan of a DailyMed fluoxetine SPL returns zero occurrences of 'back pain', 'musculoskeletal', 'myalgia', 'osteo' or 'fracture', and StatPearls likewise lists no musculoskeletal adverse effects. The realistic relationship is the reverse of an ADR: back pain and depression are strongly comorbid (LBP carried OR 1.61 for depression in NHANES young adults), and antidepressants are widely, if ineffectively, given for non-specific low back pain, so pain patients accumulate SSRI prescriptions rather than the other way round. If anything the drug is expected to move pain slightly downward or not at all, since the Cochrane review of antidepressants for non-specific LBP found no pain relief versus placebo (SMD -0.06, 95% CI -0.28 to 0.16) and the 2021 BMJ meta-analysis found benefit only for SNRIs, with TCAs and other antidepressants showing none. The atlas numbers match a non-signal: enrichment 1.07 means dorsalgia is essentially as common in fluoxetine users as in users of other drugs, FAERS disproportionality is negligible (PRR 1.14, 1/3), the drug is not on the BNF list for this event, and the 87.8% temporality is exactly the uninformative direction the cohort inflates. The one arguably real drug-bone link in the literature - SSRI initiation raising fracture risk (HR 1.77 for any fracture in SWAN mid-life women) - concerns fractures, not unspecified dorsalgia, and nothing in this fact sheet tests it. What is left is a within-users genetic hit with beta 5.03 (OR of order 10^2) on 82 datable cases, which is a sparse-data pattern, not a measured effect.","gene_comment":"Neither assignment supports mechanism: rs184784451 is an intergenic variant at Xq22 annotated only by proximity to GUCY2F, a retinal photoreceptor guanylyl cyclase linked to X-linked retinitis pigmentosa with no role in pain, bone or serotonin biology, and rs749501023 is intronic in LINC03070, an unannotated lncRNA at Xp22. Both are rare (gnomAD non-Finnish European MAF around 0.4%) and hemizygous in males, the classic setting for unstable X-chromosome rare-variant effect estimates.","score_check":"The arithmetic is self-consistent (beta 5.03 at log10p 7.53 implies se near 0.9), but an odds ratio of roughly 150 for so ordinary an outcome as unspecified back pain, from 0.4%-frequency variants in a case set of the order of 100, is the signature of quasi-separation rather than a real effect. The disease arm is flatly null (beta -0.083 +/- 0.132, |beta| well under 1.96*se) and the prescription arm is null too, so z_diff 5.58 is driven entirely by the fragile within-users estimate.","candidability":1,"candidability_reason":"No labelled or literature ADR, no enrichment, no FAERS signal, biologically irrelevant X-linked rare variants and an implausibly large effect on a tiny case set.","sources":[{"title":"Fluoxetine - StatPearls (NCBI Bookshelf)","url":"https://www.ncbi.nlm.nih.gov/books/NBK459223/"},{"title":"Fluoxetine capsules prescribing information (DailyMed SPL, Avet Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/services/v2/spls/6f4e4e9f-9ded-452f-b3ef-ab6aa9c72952.xml"},{"title":"Ferreira et al., Efficacy and safety of antidepressants for the treatment of back pain and osteoarthritis: systematic review and meta-analysis, BMJ 2021","url":"https://europepmc.org/article/MED/33472813"},{"title":"Urquhart et al., Antidepressants for non-specific low back pain, Cochrane Database Syst Rev 2008","url":"https://europepmc.org/article/MED/18253994"},{"title":"The risk of fracture among women starting selective serotonin reuptake inhibitors (SWAN), J Bone Miner Res 2025","url":"https://europepmc.org/article/MED/40757575"},{"title":"Ensembl REST variation record for rs184784451 (gnomAD frequencies, intergenic)","url":"https://rest.ensembl.org/variation/human/rs184784451?pops=1"},{"title":"MyGene.info record for GUCY2F (guanylate cyclase 2F, retinal) and LINC03070","url":"https://mygene.info/v3/query?q=symbol:GUCY2F&fields=symbol,name,summary,type_of_gene,genomic_pos"}]},"fluoxetine|N951":{"verdict":"co-prescription population","headline":"Fluoxetine treats menopausal hot flushes off-label - signal runs backwards, beta 104 is sparse-data noise","assessment":"ICD-10 N95.1 is 'Menopausal and female climacteric states', the code that carries hot flushes and other climacteric complaints. It is not a licensed fluoxetine indication - the US label lists only major depression, OCD, bulimia, panic disorder and PMDD - but fluoxetine is a recognised off-label treatment for exactly this condition: the phase III trial of fluoxetine in women with hot flashes (J Clin Oncol 2002, PMID 11896107) reported a 50% fall in hot-flash score versus 36% on placebo, a low-dose pilot found a 36% reduction in daily hot flushes, and the 2023 nonhormone therapy position statement of The Menopause Society gives SSRIs/SNRIs a Level I recommendation for vasomotor symptoms. A drug that reduces hot flushes cannot plausibly be read as causing the code that records them, so the direction of the association is decisive evidence against an ADR interpretation; on top of that, perimenopausal and postmenopausal women are a high-prevalence antidepressant population, which the modest enrichment of 1.33 fits comfortably. The FAERS 'STRONG' signal with PRR 7.24 for a menopause term is the classic indication-reporting artifact rather than independent safety support, since the state is entered as context or indication on reports for a drug prescribed into it. Temporality of 54% over 87 dated participants sits at the low end once the cohort's upward compression is allowed for, and menopause is an age-driven state that will be coded during any multi-year exposure regardless of drug. Nothing here separates a drug effect from the population fluoxetine is given to, and the within-user genetic signal is not of a magnitude that can be taken at face value.","gene_comment":"TERF2IP (RAP1) is a genuine protein-coding shelterin component that represses homology-directed repair at telomeres and also regulates NF-kB, and DNA-damage-repair biology is the main established genetic mechanism behind age at natural menopause, so the gene is not absurd on its face. But this is a single MPC burden test in a female-only, ~0.8%-prevalence phenotype with no supporting variant-level evidence, and the biology plausibly relates to menopause timing, not to a fluoxetine-caused event.","score_check":"beta_adr of 104.5 is not an interpretable effect size on any plausible scale and points to sparse-data separation in the burden test rather than a real effect, and the comparison arms are empty rather than contrastive (log10p_disease 0.02 with beta 0.124 +/- 2.003 is pure noise, log10p_prescribed 0.06). z_diff 5.07 therefore just re-expresses the inflated within-user estimate against an uninformative baseline and adds no independent support.","candidability":1,"candidability_reason":"The drug is a guideline-recommended treatment for this condition, so the association runs backwards, and the genetics are a single sparse-data burden hit with a nonsensical effect size.","sources":[{"title":"Phase III evaluation of fluoxetine for treatment of hot flashes (J Clin Oncol 2002, PMID 11896107) - Europe PMC record","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22fluoxetine%22%20AND%20TITLE%3A%22hot%20flashes%22&format=json&pageSize=20&resultType=core"},{"title":"The 2023 nonhormone therapy position statement of The North American Menopause Society (Menopause, PMID 37252752) - Europe PMC record","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22nonhormone%20therapy%20position%20statement%22&format=json&pageSize=10&resultType=core"},{"title":"PROZAC (fluoxetine) US prescribing information, FDA Drugs@FDA label 018936s108","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2017/018936s108lbl.pdf"},{"title":"UniProtKB Q9NYB0 - TERF2IP / RAP1, telomeric repeat-binding factor 2-interacting protein 1","url":"https://rest.uniprot.org/uniprotkb/Q9NYB0.txt"},{"title":"Progress in genome-wide association studies of age at natural menopause (Reprod Biomed Online 2023, PMID 36572578) - Europe PMC record","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22age%20at%20natural%20menopause%22%20AND%20ABSTRACT%3A%22DNA%20damage%22&format=json&pageSize=10&resultType=core"},{"title":"NLM Clinical Table Search Service - ICD-10-CM N95.1 'Menopausal and female climacteric states'","url":"https://clinicaltables.nlm.nih.gov/api/icd10cm/v3/search?sf=code,name&terms=N95.1"}]},"fluoxetine|R55":{"verdict":"contested","headline":"Syncope is a rare labelled fluoxetine ADR, yet fluoxetine also treats vasovagal syncope; CDH22 hit weak","assessment":"R55 (syncope and collapse) is not an indication for fluoxetine, whose licensed uses are major depression, OCD, bulimia nervosa and panic disorder; syncope appears in the US label only as a rare postmarketing nervous-system reaction, with hypotension and bradycardia listed as infrequent, alongside warnings for hyponatraemia (sodium below 110 mmol/L reported) and postmarketing QT prolongation/torsades. A drug-caused mechanism is therefore describable: a UK primary-care self-controlled case series in 41,005 adults aged 60+ found SSRIs carried the highest antidepressant risk of incident postural hypotension (IRR 4.22, 95% CI 3.76-4.74), concentrated in the first 28 days, and a 38-study meta-analysis ranked fluoxetine among the antidepressants with the highest hyponatraemia risk, a classic route to collapse in older polypharmacy patients. Against that, fluoxetine moves one major syncope subtype in the opposite direction: in a network meta-analysis of 28 RCTs in vasovagal syncope, fluoxetine reduced spontaneous syncope recurrence (RR 0.36, 95% CI 0.16-0.84), so for reflex syncope the drug is a treatment rather than a cause, and R55 as coded does not distinguish reflex from orthostatic events. The atlas is internally mixed on this: FAERS gives a strong disproportionality signal (PRR 2.18), but enrichment is 0.54, meaning fluoxetine users in this cohort carry R55 at roughly half the rate of users of other drugs, which argues against both confounding by indication and a large excess burden. Temporality is 99.1% over only 115 dated participants and is exactly the direction the cohort inflates, and the median 4.16 years from first exposure to first syncope code sits far outside the first-28-day window in which the postural-hypotension mechanism actually operates. The overall shape is a real but rare, largely acute and dose/age-driven class effect that this within-users GWAS has not convincingly localised.","gene_comment":"rs192620892 is a rare intronic variant in CDH22 (chr20:46262088, gnomAD MAF ~0.005), a cadherin with no established role in autonomic control, blood pressure, cardiac conduction or drug disposition; intronic plus rare plus biologically silent makes this a positional label, not a mechanism.","score_check":"beta_adr 3.607 (OR ~37) for a 0.5% allele on a syncope endpoint is far beyond any plausible pharmacogenomic effect size and has the signature of sparse-data bias, even though the clean null in the drug-free arm (beta 0.031 +/- 0.103) and z_diff 5.08 make the contrast look specific. The 115-participant dated denominator and the absence of a reported standard error for beta_adr mean the p-value of 1e-6.66 cannot be taken at face value.","candidability":4,"candidability_reason":"Syncope is a genuine but rare labelled SSRI effect; here it is depleted in fluoxetine users, times out four years after exposure, and rests on an implausibly large rare intronic CDH22 signal.","sources":[{"title":"Fluoxetine capsules - full prescribing information (DailyMed, Accord Healthcare)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=5d3f883a-ab3d-47e0-b24e-6f3a4e91ed70"},{"title":"PROZAC (fluoxetine) FDA-approved label, 2017","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2017/018936s108lbl.pdf"},{"title":"Europe PMC query: SSRIs and postural hypotension / falls risk (Bhanu et al., Br J Gen Pract 2025, PMID 39824621)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=(SSRI%20OR%20%22selective%20serotonin%20reuptake%20inhibitor%22)%20AND%20(syncope%20OR%20%22orthostatic%20hypotension%22)%20AND%20(risk%20OR%20cohort%20OR%20%22case-control%22)&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC query: SSRIs for vasovagal syncope (Behnoush et al., Heart Rhythm 2023, PMID 36509319)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22vasovagal%20syncope%22%20AND%20(fluoxetine%20OR%20paroxetine)&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC query: fluoxetine/SSRI hyponatraemia and SIADH (Li et al., BMC Pharmacol Toxicol 2025, PMID 40764948)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=fluoxetine%20AND%20hyponatraemia%20AND%20SIADH&format=json&pageSize=15&resultType=core"},{"title":"dbSNP rs192620892 (CDH22 intron variant, allele frequencies)","url":"https://www.ncbi.nlm.nih.gov/snp/rs192620892"},{"title":"PubMed: SSRI, syncope and orthostatic hypotension literature","url":"https://pubmed.ncbi.nlm.nih.gov/?term=selective+serotonin+reuptake+inhibitor+syncope+orthostatic+hypotension"}]},"furosemide|E162":{"verdict":"co-prescription population","headline":"Furosemide raises glucose, not lowers it; hypoglycaemia here tracks the CHF/cirrhosis/renal indication","assessment":"Hypoglycaemia is neither a licensed indication for furosemide nor a labelled adverse effect: the US label indicates furosemide for oedema of congestive heart failure, cirrhosis and renal disease including nephrotic syndrome, and its metabolic adverse reactions run the other way - 'increases in blood glucose and alterations in glucose tolerance tests ... and rarely, precipitation of diabetes mellitus', with hyperglycaemia and glycosuria listed among adverse reactions and hypoglycaemia absent from the label entirely. The pharmacology literature agrees that loop diuretics are hyperglycaemic agents, with a 2025 review of diuretics and glucose homeostasis (PMID 40223924) and a transplant cohort reporting loop-diuretic use associated with new-onset diabetes (HR 5.08, PMID 35092430); a drug that reliably moves glucose upward is poor evidence for a drug-caused hypoglycaemia signal. The three diseases furosemide is licensed for are precisely the three organ failures that generate spontaneous or iatrogenic hypoglycaemia - impaired hepatic gluconeogenesis in cirrhosis, reduced renal gluconeogenesis and insulin/sulfonylurea accumulation in CKD (PMID 37076583, PMID 35898692), and cachexia/malnutrition in advanced heart failure - so the enrichment of 1.85 and the 1.02% co-occurrence are exactly what confounding by indication predicts. Only an indirect route is plausible: furosemide-induced volume depletion and renal impairment can reduce clearance of insulin and sulfonylureas in diabetic patients, but that makes hypoglycaemia a consequence of the treated illness and its other drugs rather than of furosemide's own pharmacodynamics. The FAERS 'STRONG' flag with PRR 4.12 does not rescue this: furosemide is one of the most frequently co-reported concomitant drugs in reports of insulin- and sulfonylurea-related hypoglycaemia, and the disproportionality literature is explicit that such signals are hypothesis-generating and vulnerable to confounding by indication and co-medication. Temporality of 100% after first exposure (n_dated 107, median 5.49 years) is the uninformative direction given the cohort's record-depth compression, and 5.49 years is more consistent with progressive cardiorenal and hepatic decline than with an acute drug reaction.","gene_comment":"LEKR1 is a protein-coding gene at 3q25.31 (chr3:156.8-157.0 Mb) sitting next to CCNL1, a locus known for fetal birth weight rather than glucose handling, and rs147448455 is a rare 3'UTR variant (global MAF ~0.5%). Neither the gene nor the variant has any published connection to hypoglycaemia or insulin secretion - the only LEKR1 literature hits concern placental DNA methylation and adipose fatty-acid composition - so this is a positional assignment, not a mechanism.","score_check":"A beta_adr of 3.06 (OR ~21) for a 0.5%-frequency variant in a case set whose datable subset is only 107 people is the classic shape of a sparse-data inflated estimate, even at log10p 8.18. The drug-free arm is more credible and far smaller (beta 0.446 +/- 0.161, z = 2.8, same direction), and log10p_prescribed of 0.26 at least rules out the variant driving who gets furosemide, so the z_diff of 4.74 most likely reflects effect-size inflation in the treated stratum rather than a genuine 7-fold gene-by-drug interaction.","candidability":2,"candidability_reason":"Drug moves glucose in the opposite direction on its own label, the enrichment is explained by furosemide's licensed indications, and the rare 3'UTR variant in a glucose-agnostic gene carries an implausibly large effect.","sources":[{"title":"Furosemide - StatPearls, NCBI Bookshelf (adverse effects incl. hyperglycemia)","url":"https://www.ncbi.nlm.nih.gov/books/NBK499921/"},{"title":"Furosemide tablet US prescribing information (DailyMed) - indications, glucose warnings","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=dba00759-87b4-4ecf-998e-ba03ed3d2f1a"},{"title":"Europe PMC: loop diuretics and hyperglycaemia / new-onset diabetes (PMID 40223924, 35092430)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22loop%20diuretic%22%20AND%20(hyperglycemia%20OR%20%22glucose%20intolerance%22%20OR%20%22new-onset%20diabetes%22)&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC: hypoglycaemia risk in chronic kidney disease (PMID 37076583, 35898692, 41293049)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22chronic%20kidney%20disease%22%20AND%20hypoglycemia%20AND%20risk&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC: limitations of FAERS disproportionality - confounding by indication and co-medication","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22disproportionality%20analysis%22%20AND%20%22confounding%20by%20indication%22%20AND%20(FAERS%20OR%20%22spontaneous%20reporting%22)&format=json&pageSize=20&resultType=core"},{"title":"Ensembl REST: rs147448455 (chr3:156980109, 3'UTR variant, MAF 0.0049)","url":"https://rest.ensembl.org/variation/human/rs147448455?content-type=application/json;pops=0"},{"title":"Ensembl REST: LEKR1 gene record (protein coding, chr3:156825481-157046129)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/LEKR1?expand=0;content-type=application/json"},{"title":"Europe PMC: LEKR1 literature (placental methylation, adipose fatty acids; no glucose role)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=LEKR1&format=json&pageSize=25&resultType=core"}]},"furosemide|E871":{"verdict":"plausible ADR","headline":"Genuine label-recognised electrolyte ADR, but the rare intronic SMAD3 SNV is a sparse-data artifact","assessment":"Hyponatraemia is not an indication for furosemide: the FDA label licenses it for oedema in congestive heart failure, cirrhosis and renal disease including nephrotic syndrome, and it explicitly warns that 'electrolyte depletion may occur during furosemide therapy', naming hyponatraemia and hypochloraemic alkalosis among the imbalances to monitor with frequent serum electrolytes. Pharmacovigilance agrees the pairing is real: a 2026 FAERS disproportionality analysis found furosemide had the largest absolute case count of any drug reported with hyponatraemia (n=1,560, ROR 24.11), which matches the atlas FAERS flag (PRR 6.02), though indapamide and hydrochlorothiazide showed stronger signals - thiazides, not loop diuretics, are the classic culprits and accounted for 94% of severe diuretic-induced hyponatraemia in Sonnenblick's 129-case review. The mechanism is bidirectional rather than backwards: furosemide blocks NKCC2 in the thick ascending limb and dissipates the medullary gradient, impairing urinary concentration as well as dilution, so it is used adjunctively with salt tablets and fluid restriction in SIADH to raise sodium, yet it still precipitates hyponatraemia through volume depletion and secondary ADH release, especially in elderly patients (Beers criteria) and in the decompensated heart-failure and cirrhotic patients who are exactly its indication population. That indication overlap is the main confounder here: hyponatraemia is intrinsic to advanced heart failure and cirrhosis, and an enrichment of only 1.42 with cotx 1.08% is as consistent with sick-population confounding as with drug causation, while the 96.2% temporality over 106 dated participants is the weak, compressed direction of that metric. The genetic layer does not survive scrutiny at all, so the association is a credible drug-condition pair carried by an incredible variant.","gene_comment":"rs188879376 is a rare intronic SMAD3 variant (gnomAD G allele ~0.6%, ClinVar likely benign) with no described role in renal water handling, aquaporin-2, vasopressin or sodium transport - SMAD3 literature is TGF-beta fibrosis, aortopathy and osteoarthritis, none of which is a hyponatraemia mechanism. Assigning this locus to furosemide-induced hyponatraemia would be dressing an intronic tag SNP up as biology.","score_check":"beta_adr 3.97 (OR ~53) for a ~0.5%-frequency SNV in a condition affecting only ~1% of users is the signature of a sparse-data artifact, and the same variant is essentially null on the condition itself (beta_disease 0.273 +/- 0.138, z=1.98), so the z_diff of 4.84 is driven entirely by the inflated treated-arm estimate. A 188,461-participant GWAS of plasma sodium found 31 loci (NFAT5, SLC4A10, WNK2) but no genome-wide significant hit for diuretic-induced hyponatraemia in 14,327 exposed individuals and no gene-by-diuretic interaction, which makes a single p=1.2e-7 rare-variant hit here very unlikely to replicate.","candidability":5,"candidability_reason":"A real, label-documented ADR heavily overlapped by confounding from the oedema/heart-failure indication, attached to a rare intronic SMAD3 variant with an implausible effect size and no mechanism.","sources":[{"title":"Furosemide - StatPearls, NCBI Bookshelf (indications, NKCC2 mechanism, hyponatraemia/SIADH risk in older adults)","url":"https://www.ncbi.nlm.nih.gov/books/NBK499921/"},{"title":"Furosemide tablets US prescribing information (openFDA drug label API): indications, electrolyte depletion precautions, serum electrolyte monitoring","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22furosemide%22&limit=1"},{"title":"Europe PMC search: furosemide and hyponatremia risk - incl. 'Drug-Induced Hyponatremia Surveillance: a disproportionality analysis based on FAERS' (PMID 42257098) and furosemide in refractory SIAD management","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%28furosemide%20AND%20hyponatremia%20AND%20risk%29&format=json&pageSize=20&resultType=core"},{"title":"PubMed: diuretic-induced hyponatremia reviews - Sonnenblick et al., Chest 1993 (thiazides responsible for 94% of 129 severe cases; PMID 8432162)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=diuretic-induced+hyponatremia+review"},{"title":"Genome-Wide Association Study of Plasma Sodium Concentrations with and without Exposure to Thiazide Diuretics, JASN 2025 (PMC12147972)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12147972/"},{"title":"dbSNP RefSNP record for rs188879376 (SMAD3 intronic, gnomAD ~0.6%, ClinVar likely benign)","url":"https://api.ncbi.nlm.nih.gov/variation/v0/refsnp/188879376"}]},"furosemide|I219":{"verdict":"statistical artifact","headline":"Acute MI is not a furosemide indication nor a labelled ADR; signal is a rare-variant sparse-data effect","assessment":"The FDA furosemide label limits the indication to oedema associated with congestive heart failure, cirrhosis and renal disease, and lists no ischaemic heart disease indication and no myocardial infarction adverse reaction; its cardiovascular warnings are orthostatic hypotension, excessive diuresis with circulatory collapse, electrolyte depletion, and vascular thrombosis and embolism in the elderly. Acute MI (I21.9) is therefore not a licensed indication, and it is best read as a comorbidity of the population that receives loop diuretics, which are the mainstay of established heart failure (57.5% of HF patients in a recent Spanish cohort) and are used in the literature as a marker of advanced disease rather than as a cause of events. The atlas, unusually, does not support even that reading strongly: cotx_pct is only 0.78% and enrichment is 0.93, so MI is no more common in furosemide users than in users of other drugs, and log10p_prescribed of 0.82 shows the variant does not predict who gets the drug. Temporality of 92.4% with a median 4.7 years after first exposure is the direction expected of a genuine post-exposure event, but on only 105 dated participants and with the cohort's known upward compression it carries little weight on its own. The comparison arms are internally consistent - the variant is null on MI in the drug-free population (beta_disease 0.257 +/- 0.203, well inside noise, log10p 0.69), so z_diff 5.17 is driven entirely by the within-users arm - but the within-users effect itself, beta 5.945 (an odds ratio near 400) for a variant with roughly 0.9% European allele frequency in a subgroup with about a hundred events, is the signature of quasi-separation on a handful of carriers, not of a real pharmacogenomic effect. FAERS gives only weak corroboration (1/3, PRR 1.74) and there is no BNF interaction entry, so nothing external anchors the finding.","gene_comment":"PRKG1 (cGMP-dependent protein kinase 1) is a real and vascularly relevant gene - gain-of-function variants cause heritable thoracic aortic dissection with coronary ectasia and it mediates NO/cGMP smooth-muscle relaxation - but rs72797346 is a rare intronic SNV (chr10:51,675,580, C/T, ~0.9% NFE) inside a 1.3 Mb gene body with no GWAS Catalog association, so the gene assignment is positional only. Nothing links PRKG1 to loop-diuretic pharmacology (it is the nitrate/PDE5 axis, not the Na-K-2Cl axis), so the name should not be read as mechanism.","score_check":"The comparison arms behave sensibly - null on disease, null on prescription propensity - but beta_adr of 5.945 for a ~1% allele over ~105 events is not a believable effect size and points to sparse-data bias inflating both the estimate and the p-value. Enrichment of 0.93 also means the usual confounding-by-indication story is not what is producing the number.","candidability":2,"candidability_reason":"No label or literature support for furosemide causing MI, a rare intronic variant with an implausibly large effect, and a null disease arm - almost certainly a sparse-data artifact.","sources":[{"title":"Furosemide tablets - FDA label (openFDA drug label API)","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22furosemide%22&limit=1"},{"title":"Europe PMC search: PRKG1 and aneurysm / coronary / blood pressure","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=PRKG1%20AND%20(aneurysm%20OR%20%22coronary%22%20OR%20%22blood%20pressure%22)&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC search: loop diuretics in ischaemic heart disease and heart failure prescribing","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22loop%20diuretics%22%20AND%20%22ischemic%20heart%20disease%22%20AND%20(prescription%20OR%20prevalence%20OR%20%22heart%20failure%22)&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC search: furosemide and myocardial infarction / mortality outcomes","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=furosemide%20AND%20%22myocardial%20infarction%22%20AND%20(mortality%20OR%20outcome%20OR%20%22confounding%20by%20indication%22)&format=json&pageSize=25&resultType=core"},{"title":"Ensembl REST: rs72797346 variant record (GRCh38, gnomAD frequencies)","url":"https://rest.ensembl.org/variation/human/rs72797346?content-type=application/json;pops=1"},{"title":"Ensembl REST: PRKG1 gene coordinates (GRCh38 chr10:50,990,888-52,298,423)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/PRKG1?content-type=application/json"}]},"furosemide|J441":{"verdict":"contested","headline":"Class-level loop-diuretic respiratory harm is plausible, but this rare gene-desert variant is uninterpretable","assessment":"COPD with acute exacerbation is not a licensed indication for furosemide, whose label covers only oedema in congestive heart failure, cirrhosis and renal disease plus hypertension, and it is not listed as a labelled adverse reaction; the label's only relevant entries are hypochloraemic alkalosis and hypokalaemia. There is, however, real observational literature for respiratory harm: in a propensity-weighted Ontario cohort of 51,612 incident diuretic users aged 66+ with COPD, incident diuretic use was associated with more ED visits (HR 1.35), more hospitalisations for COPD/pneumonia (HR 1.22) and higher COPD/pneumonia mortality (HR 1.41), driven by loop diuretics and persisting in patients without prior heart failure or prior exacerbations. The proposed mechanism is coherent and drug-specific: loop diuretics raise serum bicarbonate and arterial pH, and the resulting metabolic alkalosis blunts central and peripheral chemoreceptor drive and worsens hypercapnia, while hypokalaemia can weaken respiratory muscles. Against that, confounding by indication is severe here: about 12% of COPD patients have heart failure, heart failure itself raises exacerbation risk (aHR 1.45-1.65), and incident use of all five heart-failure drug classes - ACEi, ARB, beta-blocker, loop diuretic and MRA - was associated with increased exacerbation, which points at the cardiac illness rather than at furosemide. The drug also moves dyspnoea in the opposite direction in some settings: nebulised furosemide has shown improvement in pulmonary function and dyspnoea in small obstructive-lung-disease trials, and systemic furosemide is the treatment for the pulmonary oedema that is routinely miscoded as a COPD exacerbation. The atlas cannot adjudicate this because enrichment is null (no background arm), so the one honest test of whether furosemide is simply given to cardiorespiratory patients was not run.","gene_comment":"rs73739788 (chr6:23,533,888, MAF ~1.8%) is an intronic variant in ENSG00000289368, an unnamed novel lncRNA, and the whole 6p22.3 window is a protein-coding desert - the nearest coding genes are HDGFL1 at 22.57 Mb and NRSN1 at 24.13 Mb, roughly 0.6-1 Mb away. There is no mechanistic story here, no reported GWAS association for the variant, and nothing that connects it to chloride handling, chemoreceptor drive or airway biology.","score_check":"beta_adr 3.849 for a 1.8%-frequency allele implies an odds ratio near 47, which is not a believable effect for a common exacerbation phenotype and has the signature of sparse-data bias in a small carrier count; the drug-free arm gives only beta 0.569 +/- 0.246 (|beta| barely above 1.96*se, OR ~1.8), so the z_diff of 4.1 is manufactured by the unstable treated-arm estimate rather than by a real interaction. Temporality of 84.8% on just 99 dated participants is weak evidence in a cohort that compresses this measure upward, and log10p_prescribed of 0.63 at least confirms the variant does not predict who receives furosemide.","candidability":4,"candidability_reason":"The class-level respiratory harm has independent cohort support and a mechanism, but this specific signal is an implausibly large rare-variant effect in an unnamed lncRNA, with no enrichment test and heavy heart-failure confounding.","sources":[{"title":"FUROSEMIDE tablet - US prescribing information (DailyMed)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=79d9aef8-cfb9-4f6e-ac15-f830d7ea2324"},{"title":"Vozoris NT et al. Incident diuretic drug use and adverse respiratory events among older adults with COPD. Br J Clin Pharmacol 2018 (PMC5809361)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC5809361/"},{"title":"Axson EL et al. Relationship between heart failure and the risk of acute exacerbation of COPD. Thorax 2021 (PMC8311079)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8311079/"},{"title":"Veldhoen R, Muscedere J. Nebulised furosemide for the treatment of patients with obstructive lung disease: a systematic review protocol. BMJ Open 2023 (PMID 37996224)","url":"https://pubmed.ncbi.nlm.nih.gov/37996224/"},{"title":"Ensembl REST: ENSG00000289368 (novel lncRNA, chr6:23,338,012-23,584,458) and variant rs73739788","url":"https://rest.ensembl.org/variation/human/rs73739788?content-type=application/json"}]},"furosemide|K30":{"verdict":"statistical artifact","headline":"Furosemide GI irritation is labelled, but a rare UQCC1 height-locus SNV with OR~18 is not credible","assessment":"Dyspepsia (K30) is not an indication for furosemide, whose US label covers only oedema in congestive heart failure, cirrhosis and renal disease, and hypertension. Upper-GI intolerance is nonetheless a recognised labelled adverse effect class: the same label lists anorexia, 'oral and gastric irritation', nausea and vomiting (the latter also flagged as a sign of electrolyte imbalance), plus cramping, diarrhoea, constipation and pancreatitis, so a coded dyspepsia diagnosis is at least a biologically adjacent endpoint rather than an absurd one. A mechanism is only weakly described: furosemide is not an NSAID and has no known effect on acid secretion or mucosal prostaglandins, so the plausible routes are non-specific local gastric irritation from the oral tablet and nausea/anorexia secondary to hypovolaemia and electrolyte depletion; there is no published furosemide-dyspepsia pharmacology, and the FAERS signal here is weak (PRR 1.47, 1/3 hits). The atlas epidemiology actively works against an ADR reading at population level: only 1.2% of furosemide users carry K30 and the enrichment is 0.68, i.e. dyspepsia is less common in furosemide users than in users of other drugs, which rules out confounding by indication but equally shows no excess to explain. Temporality is 100% of 102 dated participants, which the cohort's record structure inflates and which is therefore weak evidence on its own, and a median 4.44 years to first record is a long lag for a direct gastric-irritation effect. The remaining claim is purely genetic, and that is where it fails: a single rare intronic variant carrying an implausible effect size, tested on a soft symptom code, at a locus with no gastrointestinal biology.","gene_comment":"rs117783353 is a rare intronic SNV (global MAF ~0.5%, ~1.9% in non-Finnish Europeans) at 20q11.22 inside UQCC1, a mitochondrial complex III assembly factor with no gastric, acid-secretion or diuretic pharmacology; the only established GWAS trait at this exact variant is body height (p = 1e-10), reflecting the neighbouring GDF5/GDF5-AS1 skeletal locus rather than anything digestive. The gene assignment is technically intragenic but mechanistically empty, and should not be presented as a mechanism.","score_check":"beta_adr 2.871 is an odds ratio near 18 for a ~1-2% frequency variant against a common, non-specific symptom code, with no ADR standard error given and log10p 6.33 falling short of genome-wide significance before any correction for the atlas's many phenotypes - the classic shape of sparse-data bias in a small carrier count. The drug-free arm is a clean null (beta 0.036 +/- 0.076, well inside noise) and z_diff 4.93 therefore rests entirely on the unstable treated-arm estimate rather than on a demonstrated interaction.","candidability":3,"candidability_reason":"Labelled GI irritation makes the drug-condition pair non-absurd, but dyspepsia is depleted in furosemide users, the lag is years, and the variant is a rare height-locus SNV with a sparse-data-sized effect and no relevant biology.","sources":[{"title":"FUROSEMIDE tablet - DailyMed label (indications and adverse reactions)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=dba00759-87b4-4ecf-998e-ba03ed3d2f1a"},{"title":"dbSNP/Ensembl variant record for rs117783353 (consequence, position, population frequencies)","url":"https://rest.ensembl.org/variation/human/rs117783353?content-type=application/json;pops=1"},{"title":"GWAS Catalog: rs117783353 genomic context and mapped genes (UQCC1, GDF5)","url":"https://www.ebi.ac.uk/gwas/rest/api/singleNucleotidePolymorphisms/rs117783353"},{"title":"GWAS Catalog: reported associations for rs117783353 (body height, p = 1e-10)","url":"https://www.ebi.ac.uk/gwas/rest/api/singleNucleotidePolymorphisms/rs117783353/associations?projection=associationBySnp"},{"title":"Europe PMC literature search: UQCC1 function and associated phenotypes (complex III assembly, height, grip strength)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=UQCC1%20AND%20(mitochondrial%20OR%20%22complex%20III%22%20OR%20height)&format=json&pageSize=15&resultType=core"},{"title":"Europe PMC literature search: loop diuretics and peptic ulcer / GI bleeding / dyspepsia","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22loop%20diuretic%22%20AND%20(%22peptic%20ulcer%22%20OR%20%22gastrointestinal%20bleeding%22%20OR%20dyspepsia)&format=json&pageSize=20&resultType=core"}]},"furosemide|K922":{"verdict":"co-prescription population","headline":"GI haemorrhage tracks the furosemide population, not the drug; rare DNM3 intronic hit is unconvincing","assessment":"Gastrointestinal haemorrhage is neither an indication for furosemide nor a labelled adverse reaction: the furosemide product information and the StatPearls review list GI effects (nausea, vomiting, diarrhoea, constipation, pancreatitis, hepatic encephalopathy in cirrhosis) but no bleeding or haemorrhage, and the atlas itself records on_bnf = 0. Targeted Europe PMC searches for furosemide or loop diuretics with gastrointestinal / upper gastrointestinal bleeding returned no study establishing an association in either direction, so there is no described drug-caused mechanism to lean on. What furosemide does mark is a population that bleeds for other reasons: it is the standard diuretic for decompensated heart failure, CKD and cirrhotic ascites, and those patients carry oesophageal varices, anticoagulant and antiplatelet therapy, aortic-stenosis angiodysplasia (Heyde syndrome) and advanced age, all of which are established drivers of GI bleeding in the literature retrieved here. An indirect route (hypovolaemia and splanchnic hypoperfusion from aggressive diuresis precipitating ischaemic colitis) is biologically arguable but I found no epidemiological support for diuretics as a risk factor for ischaemic colitis. The FAERS signal (PRR 2.86) is exactly what confounding by comorbidity and co-medication produces in spontaneous-report data, and the atlas's own enrichment of 1.00 says GI haemorrhage is no commoner in furosemide users than in users of other drugs, which undercuts a drug-specific effect. The 100% temporality with a 4-year median gap is the uninformative direction for this cohort and adds nothing.","gene_comment":"rs553821737 is a rare (gnomAD MAF ~0.4%) intronic variant at chr1:172,136,588 inside the 600 kb DNM3 gene; DNM3 encodes a brain- and testis-enriched dynamin GTPase with no role in haemostasis, mucosal integrity or portal circulation. The nearest functional element is the DNM3OS / miR-214 / miR-199a fibrosis-associated cluster ~1.8 kb away, which is suggestive at best and has been studied in lung, kidney and heart fibrosis, not GI bleeding.","score_check":"beta_adr = 4.745 (OR ~115) at log10p = 10.0 implies an SE near 0.75 for a rare 0.4% variant, the classic sparse-data signature of a handful of carrier cases rather than a real effect of that size; the disease arm is flatly null (beta 0.118 +/- 0.129) and the prescription arm is null too, so z_diff = 6.22 is driven entirely by the inflated treated-arm estimate. The numbers are internally consistent but not credible as an effect size, and would need replication in an independent cohort before being taken seriously.","candidability":2,"candidability_reason":"No labelled or literature-supported furosemide-GI-bleeding link, enrichment exactly 1.0, and a rare intronic variant in a functionally irrelevant gene with a sparse-data effect size.","sources":[{"title":"Furosemide - StatPearls (NCBI Bookshelf, NBK499921)","url":"https://www.ncbi.nlm.nih.gov/books/NBK499921/"},{"title":"UniProtKB Q9UQ16 - Dynamin-3 (DNM3), human","url":"https://rest.uniprot.org/uniprotkb/Q9UQ16.txt"},{"title":"Europe PMC search: furosemide AND \"gastrointestinal bleeding\"","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=furosemide%20AND%20%22gastrointestinal%20bleeding%22&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC search: \"loop diuretic\" AND \"upper gastrointestinal bleeding\"","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22loop%20diuretic%22%20AND%20%22upper%20gastrointestinal%20bleeding%22&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC search: diuretics AND \"ischemic colitis\" AND risk","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=diuretics%20AND%20%22ischemic%20colitis%22%20AND%20risk&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC search: heart failure, GI bleeding, anticoagulants (risk factors)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22heart%20failure%22%20AND%20%22gastrointestinal%20bleeding%22%20AND%20anticoagulant%20AND%20risk&format=json&pageSize=20&resultType=core"},{"title":"Ensembl REST: variant rs553821737 (frequencies, consequence)","url":"https://rest.ensembl.org/variation/human/rs553821737?content-type=application/json;pops=1"},{"title":"Ensembl REST: genes overlapping chr1:172,130,000-172,145,000 (DNM3, DNM3OS, MIR214, MIR199A2)","url":"https://rest.ensembl.org/overlap/region/human/1:172130000-172145000?feature=gene;content-type=application/json"}]},"furosemide|L089":{"verdict":"co-prescription population","headline":"Local skin infection tracks the oedematous population furosemide treats; DMD intronic hit is not mechanism","assessment":"L08.9 (other/unspecified local infection of skin and subcutaneous tissue) is not a licensed indication for furosemide, which is approved for volume overload and oedema in heart failure, hepatic and renal disease including nephrotic syndrome. It is, however, exactly the complication of the population that receives the drug: the main risk factors for erysipelas/cellulitis of the leg are lymphoedema and a portal of entry, with leg oedema, venous insufficiency and overweight also associated (Dupuy), and StatPearls lists venous insufficiency, lymphoedema, diabetes and any skin-barrier breach as the standard risk set. Critically, furosemide moves the dominant risk factor in the protective direction - it removes the oedema - so a straightforward drug-caused route to more skin infection is not plausible for the bulk of these cases. A minor genuine route does exist and is documented: furosemide is a long-recognised trigger of bullous pemphigoid (case reports from Fellner 1976 through Lee & Downham 2006 and a 2025 case, one series reporting chronic leg ulcers secondary to furosemide-induced BP) and the label carries photosensitivity, SJS/TEN, exfoliative dermatitis and AGEP, any of which can become secondarily infected - but these are rare idiosyncratic events, not a 0.9% co-treatment burden. The atlas is internally consistent with the confounding reading rather than the ADR reading: enrichment 0.85 means the condition is if anything slightly less common in furosemide users than in users of other drugs, the disease arm is flatly null (beta 0.003 +/- 0.078), and FAERS support is weak (1/3, PRR 1.79). Temporality of 90% after first exposure on only 90 dated participants, with a median 4.8 years, is the uninformative direction given the cohort's known compression and is equally explained by ageing and progressive venous/lymphatic disease during years of diuretic therapy.","gene_comment":"rs148880129 is an intronic variant in DMD (chrX:31603867, gnomAD MAF ~1.1%), and DMD encodes dystrophin - a muscle and cardiac structural protein with no described role in skin barrier or antibacterial defence; because DMD is the largest human gene, 'in DMD' here is a positional label spanning 2.2 Mb of intron rather than a mechanistic assignment. There is no credible path from a rare intronic X-linked dystrophin variant to local skin infection, and it should not be presented as mechanism.","score_check":"beta_adr 2.405 (OR ~11) at log10p 7.23 for a ~1% intronic X-chromosome variant against a small case set is the classic shape of a sparse-data / hemizygosity-coding artifact, and the entire z_diff of 5.34 rests on that one beta since the drug-free arm is exactly null. The population-level numbers (cotx 0.9%, enrichment 0.85, weak FAERS) are believable and point away from any excess to explain.","candidability":2,"candidability_reason":"Condition belongs to the oedematous population furosemide treats, drug moves the main risk factor protectively, no enrichment, null disease arm, and the variant is an unreplicated intronic DMD hit with no plausible skin biology.","sources":[{"title":"Furosemide - StatPearls (NCBI Bookshelf, NBK499921): indications, mechanism, dermatologic adverse reactions","url":"https://www.ncbi.nlm.nih.gov/books/NBK499921/"},{"title":"Cellulitis - StatPearls (NCBI Bookshelf, NBK549770): risk factors and management","url":"https://www.ncbi.nlm.nih.gov/books/NBK549770/"},{"title":"PubMed: risk factors for erysipelas/cellulitis of the leg (Dupuy et al. case-control and review)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=risk+factors+erysipelas+cellulitis+leg+lymphoedema+oedema+case-control"},{"title":"PubMed: furosemide-induced bullous pemphigoid case reports and disproportionality analysis","url":"https://pubmed.ncbi.nlm.nih.gov/?term=furosemide+bullous+pemphigoid"},{"title":"dbSNP rs148880129 (DMD intron variant, chrX:31603867, gnomAD frequency)","url":"https://www.ncbi.nlm.nih.gov/snp/rs148880129"},{"title":"MedlinePlus Genetics: DMD gene (dystrophin)","url":"https://medlineplus.gov/genetics/gene/dmd/"}]},"furosemide|R040":{"verdict":"co-prescription population","headline":"Nosebleeds in the anticoagulated cardiac population; the rare intronic TBC1D4 hit is a sparse-data artifact","assessment":"Epistaxis is not a licensed indication for furosemide - the label covers oedema in congestive heart failure, cirrhosis and renal disease, plus hypertension - and epistaxis does not appear anywhere in the label's adverse-reaction list, which does record rare haematologic events (thrombocytopenia, agranulocytosis, aplastic and haemolytic anaemia) that could in principle present as mucosal bleeding but are far too rare to generate a cohort-level signal. The population that receives furosemide is precisely the population in which nosebleeds are common: older patients with hypertension, cardiovascular and peripheral vascular disease, on antiplatelets or anticoagulants, which are the established risk factors identified in the epistaxis literature (39% hypertensive, 39% on antiplatelet/anticoagulant therapy in one severity cohort). The atlas numbers agree with that reading rather than with an ADR: co-treatment is only 1.26% with an enrichment of 1.33, i.e. a mild excess entirely compatible with shared comorbidity, and the FAERS STRONG flag (PRR 2.81) is what co-reporting of furosemide alongside warfarin/DOACs in heart-failure reports would produce. Direction of effect argues against causation: hypertension raises epistaxis risk (adjusted HR 1.47, 95% CI 1.30-1.66 in a 71,498-person national cohort), and furosemide lowers blood pressure and intravascular volume, so the drug moves the dominant haemodynamic risk factor the protective way. Temporality of 94.9% over only 99 dated participants is exactly the upward-compressed pattern the atlas warns is weak on its own, and a median 2.69 years to onset in an ageing cardiac cohort adds nothing causal. The comparison arms are consistent with no shared biology: the variant is flat for epistaxis in the drug-free population (log10p 0.28; beta 0.139 +/- 0.219, well inside noise) and flat for propensity to be prescribed furosemide (log10p 0.5), so the entire association rests on one implausibly large within-users effect.","gene_comment":"rs117678777 is a rare (alt allele ~0.25% in gnomAD/TopMed) intronic variant in TBC1D4, not a coding change; TBC1D4 encodes AS160, an insulin-regulated GLUT4-trafficking Rab-GAP known for glucose phenotypes (the Greenlandic p.Arg684Ter diabetes variant), with no described role in coagulation, platelet function, vascular fragility or nasal mucosa. There is no mechanistic route from this locus to epistaxis, and the assignment should not be dressed up as one.","score_check":"beta_adr 4.922 (OR ~140) for a 0.25%-frequency variant, in a condition arm with only 99 dated cases, is the signature of sparse-data separation on a handful of carriers rather than a real effect, and z_diff 4.99 is driven wholly by that unstable treated-arm estimate since the disease arm is a clean null. The population and comparison-arm numbers (enrichment 1.33, log10p_prescribed 0.5) are believable; the variant effect size is not.","candidability":1,"candidability_reason":"Condition is explained by the elderly, hypertensive, anticoagulated population furosemide is given to; the drug moves blood pressure in the protective direction, and the sole variant is a rare intronic TBC1D4 hit with an implausible effect size and no mechanism.","sources":[{"title":"FUROSEMIDE tablet - US prescribing information (DailyMed, Leading Pharma)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=b5c6fcf4-fd37-4fee-8dac-1a3273d95ffe"},{"title":"Association of Hypertension With the Risk and Severity of Epistaxis (JAMA Otolaryngol Head Neck Surg, PMC7489409)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7489409/"},{"title":"PubMed: risk factors for epistaxis - hypertension, anticoagulants, antiplatelets (PMIDs 36128022, 37496443, 28844608)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=epistaxis+hypertension+anticoagulant+risk+factors"},{"title":"dbSNP rs117678777 (intronic TBC1D4, alt allele frequency ~0.0025)","url":"https://www.ncbi.nlm.nih.gov/snp/rs117678777"},{"title":"PMC search: TBC1D4 p.Arg684Ter phenotypes - glucose homeostasis and GLUT4, no bleeding phenotype","url":"https://pmc.ncbi.nlm.nih.gov/search/?term=TBC1D4+p.Arg684Ter+Greenlandic+type+2+diabetes"},{"title":"PubMed search returning no studies linking diuretics/furosemide to epistaxis or nasal mucosal dryness","url":"https://pubmed.ncbi.nlm.nih.gov/?term=diuretics+epistaxis+nasal+mucosa+dryness"}]},"gabapentin|A419":{"verdict":"co-prescription population","headline":"Sepsis tracks gabapentin's frail comorbid users, not the drug; the rare chr18 lncRNA hit is sparse-data","assessment":"Sepsis is neither an indication for gabapentin nor a labelled adverse reaction: the US label lists only postherpetic neuralgia and adjunctive partial-onset seizures, and sepsis appears nowhere in it, the only infection entries being non-specific ('infection' 5% vs 4% placebo in the PHN trials, 'viral infection' 11% vs 3% in the 3-12y epilepsy trials), which are upper-respiratory-type events, not bacteraemic sepsis. A Europe PMC search for gabapentinoids with sepsis or serious infection returns no study making that link, so there is no pharmacovigilance or cohort literature behind an ADR reading. What is real is a respiratory-depression signal - label section 5.7 warns of serious respiratory depression with CNS depressants or underlying respiratory impairment, and Rahman et al. (Ann Intern Med 2024) found a 1.39-fold hazard of severe COPD exacerbation with gabapentinoid use, with perioperative studies reporting more pulmonary complications - so a sedation/aspiration-pneumonia route to sepsis is conceivable, but it is indirect, undescribed, and cannot be read off an unspecified-sepsis code. The far simpler explanation of the 1.5-fold enrichment is who gets gabapentin: diabetic neuropathy (diabetic foot infection), CKD, cancer pain, postherpetic neuralgia in the elderly, and spine/immobility patients - populations with high baseline sepsis risk, which is also why the drug-free disease arm shows nothing (beta_disease 0.136 +/- 0.114, indistinguishable from noise) and the prescription-propensity arm is null. Temporality of 100% over 109 dated participants with a median 1.28 years to sepsis is exactly what any acute incident event looks like in a chronically treated frail cohort, and the fact sheet's own caveat applies: a high value here is weak evidence. The atlas therefore contradicts nothing in the literature, but nothing in the literature supports a drug-caused sepsis reading either.","gene_comment":"ENSG00000287907 is an unnamed lncRNA 'novel transcript' on chr18q22 (chr18:68.24-68.54 Mb), and rs117452116 (chr18:68,414,441, T allele ~0.6% TOPMed / ~1-1.5% European) carries no gene or functional annotation in dbSNP. It has no immune, complement or host-defence biology and no relation to gabapentin's alpha2delta targets CACNA2D1/CACNA2D2, so it supplies a locus label, not a mechanism.","score_check":"beta_adr of 4.531 is an odds ratio near 90 for a ~1% variant against a 1.17% outcome - the signature of sparse-data bias or near-separation in a handful of carriers, not a credible effect size, and the z_diff of 5.53 is driven entirely by that inflated estimate rather than by any signal in the null disease arm. The within-users log10p of 8.09 is real arithmetic but should not be read as a real effect until the carrier counts and a Firth-corrected estimate are inspected.","candidability":1,"candidability_reason":"No drug-sepsis literature or mechanism, enrichment explained by the frail comorbid population gabapentin treats, and the sole variant is a rare unannotated chr18 lncRNA SNV with an implausibly large effect.","sources":[{"title":"NEURONTIN (gabapentin) prescribing information, DailyMed (Parke-Davis/Pfizer)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=ee9ad9ed-6d9f-4ee1-9d7f-cfad438df388"},{"title":"Rahman et al., Gabapentinoids and Risk for Severe Exacerbation in COPD: A Population-Based Cohort Study, Ann Intern Med 2024 (PMID 38224592)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:38224592&resultType=core&format=json"},{"title":"PubMed: gabapentinoid AND (pneumonia OR infection) - 19 records, none on sepsis","url":"https://pubmed.ncbi.nlm.nih.gov/?term=%22gabapentinoid%22+AND+%28pneumonia+OR+infection%29"},{"title":"PubMed: gabapentin respiratory depression cohort studies","url":"https://pubmed.ncbi.nlm.nih.gov/?term=gabapentin+%22respiratory+depression%22+risk+cohort"},{"title":"Europe PMC: gabapentinoid AND (sepsis OR serious infection) - no linking study","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=gabapentinoid%20AND%20(sepsis%20OR%20%22serious%20infection%22)&format=json&pageSize=25"},{"title":"Ensembl REST lookup, ENSG00000287907 (lncRNA, novel transcript, chr18)","url":"https://rest.ensembl.org/lookup/id/ENSG00000287907?content-type=application/json"},{"title":"dbSNP rs117452116","url":"https://www.ncbi.nlm.nih.gov/snp/rs117452116"}]},"gabapentin|E780":{"verdict":"statistical artifact","headline":"Off-label phenotype, depleted in gabapentin users, resting on one rare intergenic variant","assessment":"Pure hypercholesterolaemia (E78.0) is neither a licensed indication for gabapentin nor a labelled adverse reaction: the US label lists only postherpetic neuralgia and adjunctive partial-onset seizures, and neither hypercholesterolaemia nor hyperlipidaemia appears anywhere in it. The label does carry a small metabolic footprint - weight gain 2-3% vs 0-2% placebo, hyperglycaemia 1% vs 0%, peripheral oedema 8% vs 2% - and there is a mechanistic route by which that could drift into dyslipidaemia, since gabapentin's target alpha2delta-1 in the ventromedial hypothalamus controls feeding and glucose/lipid balance in mice, where gabapentin infusion into the VMH increases feeding and body-weight gain. That route is indirect and undocumented in human lipid endpoints; a PubMed search for gabapentin and dyslipidaemia/hypercholesterolaemia returns no study showing the drug raises cholesterol, and one small neurochemical study reports gabapentin therapy *lowering* synaptosomal membrane cholesterol, so nothing in the literature supports a cholesterol-raising direction. The atlas arms argue against the association rather than for it: enrichment 0.41 means gabapentin users carry E78.0 roughly 2.4x less often than users of other drugs, so this is neither confounding by indication nor a population-level excess, and log10p_prescribed 0.05 rules out a prescription-propensity artifact. In the drug-free population the same variant does nothing (beta_disease -0.071 +/- 0.043, |beta| below 1.96*se, and pointing the opposite way), so the entire z_diff of 5.12 rests on the within-users estimate alone. Temporality of 92.8% over 345 dated participants is uninformative here given the cohort's known upward compression, and FAERS is essentially null (1/3, PRR 1.53), as is BNF listing.","gene_comment":"rs118173562 is an intergenic variant at chr12:71,284,188 (GRCh38) with MAF ~0.7%, assigned to TSPAN8 only by falling inside the gene's extended span with no transcript consequence. TSPAN8 is a genuine metabolic locus (type 2 diabetes, HDL, apolipoprotein A1, BMI, HbA1c in the GWAS Catalog), but those signals come from common index SNPs such as rs7961581, not this rare non-coding variant, so the gene name lends the finding an air of metabolic mechanism it has not earned.","score_check":"A beta of 1.752 (OR ~5.8) for a 0.7%-frequency variant tested within a drug-user subgroup is the classic shape of a sparse-data, winner's-curse signal, and it is unsupported by the null, oppositely-signed effect in the much larger drug-free arm. The numbers are internally consistent but the only positive arm is the one most vulnerable to sparse rare-variant inflation.","candidability":2,"candidability_reason":"Off-label phenotype that is depleted rather than enriched in gabapentin users, null in the disease and prescription arms, near-null in FAERS, resting on one rare intergenic variant with a large effect.","sources":[{"title":"Gabapentin capsule/tablet label (DailyMed, Laurus Labs) - Indications and Adverse Reactions","url":"https://www.dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=07aa1bcc-df05-44fa-92d4-63894e0e6cd6"},{"title":"PubMed: gabapentin, lipid and cholesterol studies (incl. PMID 25614181, Neurochem Res 2015)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=gabapentin+lipid+cholesterol"},{"title":"PubMed: gabapentinoid weight gain and metabolic effects (PMID 24403154 J Neurosci 2014; PMID 32337532 Endocrinology 2020)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=gabapentinoid+weight+gain+metabolic"},{"title":"Ensembl REST: variation rs118173562 (intergenic, chr12:71284188, MAF 0.0073)","url":"https://rest.ensembl.org/variation/human/rs118173562?content-type=application/json"},{"title":"Ensembl REST: TSPAN8 gene coordinates (chr12:71,124,454-71,441,898)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/TSPAN8?content-type=application/json"},{"title":"GWAS Catalog: traits associated with TSPAN8 (T2D, HDL, ApoA1, BMI, HbA1c)","url":"https://www.ebi.ac.uk/gwas/genes/TSPAN8"},{"title":"PubMed: gabapentin dyslipidemia / hypercholesterolemia search","url":"https://pubmed.ncbi.nlm.nih.gov/?term=gabapentin+dyslipidemia+OR+hypercholesterolemia"}]},"gabapentin|I251":{"verdict":"contested","headline":"Contested class-level CV signal; CPNE4 hit is sparse-data noise and CAD is depleted in gabapentin users","assessment":"Atherosclerotic heart disease (I25.1) is not a licensed gabapentin indication - the US label covers postherpetic neuralgia and adjunctive partial-onset seizures only - and the Neurontin label lists no cardiovascular event in its adverse-reaction or postmarketing sections. There is nonetheless a real, contested observational literature: a 210,064-patient EHR cohort of diabetic neuropathy reported five-year HRs of 1.25 for myocardial infarction, 1.37 for peripheral vascular disease and 1.14 for heart failure on gabapentin, and a 2025 systematic review of five cohorts (>1 million patients) concluded gabapentinoids raise thrombotic risk within three months and cardiovascular risk beyond a year - all from propensity-matched observational data in populations (diabetic neuropathy, chronic pain, immobility) where residual confounding by indication is the standard alternative explanation, and no specific mechanism linking alpha-2-delta ligand binding to coronary atherosclerosis is established. The atlas numbers do not corroborate the drug-population reading: enrichment is 0.25, so chronic CAD is four-fold depleted among gabapentin users relative to users of other drugs, and only 1.45% carry the code, which is the opposite of what confounding by a cardiometabolic indication would produce here. Temporality of 96.6% on only 117 dated participants is uninformative by the atlas's own convention, and the outcome is a chronic-atherosclerosis code whose first recording reflects surveillance and coronary imaging far more than incident drug-induced disease over a 1.73-year median. The genetic layer is the weakest part: log10p_adr 6.81 does not reach genome-wide significance, the disease arm is flatly null (beta 0.039 +/- 0.042), and the whole z_diff of 5.13 rests on an implausibly large within-user beta. FAERS support is weak (1/3 signals, PRR 1.51), consistent with a background-level disproportionality rather than a recognised reaction.","gene_comment":"rs80201980 is an intronic, MAF ~3.5% variant at chr3:131,654,696 inside CPNE4 (copine-4), a calcium-dependent phospholipid-binding protein with no established disease role and no link to the alpha-2-delta/CACNA2D1 biology gabapentin acts on; the variant has no GWAS Catalog entry, so the gene assignment is positional only and carries no mechanistic weight for coronary atherosclerosis.","score_check":"beta_adr 2.324 (OR ~10) for a 3.5%-frequency SNV against roughly a hundred cases is the classic sparse-data inflation pattern, and it sits alongside a null disease arm and a null prescription arm, so the effect size is not believable even if the p-value is real. The enrichment of 0.25 and the 96.6% temporality are internally consistent but neither supports an ADR reading.","candidability":3,"candidability_reason":"Class-level cardiovascular concern exists but is contested and confounding-prone, the outcome is a chronic surveillance-driven code, and the CPNE4 variant is sub-genome-wide with an artifact-grade effect size in an unrelated gene.","sources":[{"title":"NEURONTIN (gabapentin) US prescribing information, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=ee9ad9ed-6d9f-4ee1-9d7f-cfad438df388"},{"title":"Cardiovascular risk of gabapentin and pregabalin in patients with diabetic neuropathy (Cardiovasc Diabetol 2022; PMID 36050764)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:36050764&resultType=core&format=json"},{"title":"Cardiovascular safety of gabapentinoids gabapentin & pregabalin: a systematic review (Indian J Med Res 2025; PMID 40536375)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:40536375&resultType=core&format=json"},{"title":"PubMed search: gabapentin cardiovascular risk / myocardial infarction","url":"https://pubmed.ncbi.nlm.nih.gov/?term=gabapentin+cardiovascular+risk+myocardial+infarction"},{"title":"Ensembl REST: variant rs80201980 (chr3:131654696, intronic, MAF 0.035)","url":"https://rest.ensembl.org/variation/human/rs80201980?content-type=application/json"},{"title":"UniProtKB: human Copine-4 (CPNE4) function and tissue specificity","url":"https://rest.uniprot.org/uniprotkb/search?query=gene:CPNE4+AND+organism_id:9606&fields=accession,protein_name,gene_names,cc_function,cc_tissue_specificity,cc_disease&format=json"},{"title":"GWAS Catalog REST: rs80201980 (no association records returned)","url":"https://www.ebi.ac.uk/gwas/rest/api/singleNucleotidePolymorphisms/rs80201980"}]},"gabapentin|K625":{"verdict":"statistical artifact","headline":"Rectal bleeding is not a gabapentin label event; the rare CRISPLD2 intronic hit is sparse-data noise","assessment":"Gabapentin is licensed only for postherpetic neuralgia and adjunctive partial-onset seizures (FDA Neurontin label), so anorectal haemorrhage (K62.5) is neither an indication nor a labelled reaction: the label's digestive-system tables list diarrhoea 6% vs 3%, constipation 4% vs 2% and dyspepsia, with no rectal or gastrointestinal haemorrhage term at any frequency, and MedlinePlus likewise lists no bleeding. The only drug-caused route is indirect and weak - gabapentin-induced constipation with straining, and constipation is a documented risk factor for haemorrhoidal disease (OR 3.92, 95% CI 1.03-14.2 in one controlled series) - but gabapentin's excess of diarrhoea is larger than its excess of constipation, and a targeted PubMed search returned no study linking gabapentin or pregabalin to GI or rectal bleeding. The far more likely explanation is co-medication rather than gabapentin itself: gabapentin users are chronic-pain and multimorbid patients heavily co-prescribed opioids (strongly constipating) and NSAIDs (a recognised cause of lower-GI bleeding), and the atlas cannot separate those. The atlas is at least consistent on the confounding-by-indication axis: enrichment 0.62 means rectal bleeding is actually depleted among gabapentin users relative to users of other drugs, so the drug is not preferentially given to people who already bleed. Against that, the within-user genetic signal does not survive scrutiny - a beta of 6.25 (odds ratio in the hundreds) for a variant carried by ~0.3% of people, with no corresponding effect in the drug-free population (beta 0.107 +/- 0.112), is the signature of a handful of concordant carriers, not of a real interaction. FAERS support is weak (1 of 3, PRR 1.82) and the 84.8% temporality on 112 dated participants is exactly the direction the cohort's record structure inflates, so it adds little.","gene_comment":"rs551836887 is a rare intronic variant (gnomAD MAF ~0.3%, 1000G 0.06%) at chr16:84915773, within CRISPLD2 - a genuine protein-coding, broadly expressed secreted ECM/heparin-binding gene at 16q24.1, best known for glucocorticoid response in airway tissue and orofacial clefting. It has no described role in haemostasis, mucosal integrity, colorectal vasculature or gut motility, so the assignment is anatomically real but mechanistically empty.","score_check":"log10p_adr 6.59 paired with beta 6.25 for a 0.3%-frequency intronic variant, alongside a flatly null disease arm (0.107 +/- 0.112) and a null prescription arm, is a sparse-data pattern rather than a large true effect. The population-level numbers (enrichment 0.62, 1.31% co-occurrence) are believable; the variant effect size is not.","candidability":2,"candidability_reason":"No labelled or literature-supported gabapentin-rectal-bleeding link, an implausibly large effect for a 0.3% intronic variant with a null disease arm, and opioid/NSAID co-medication as a simpler explanation.","sources":[{"title":"NEURONTIN (gabapentin) FDA prescribing information - indications and adverse reaction tables","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2017/020235s064_020882s047_021129s046lbl.pdf"},{"title":"Gabapentin - MedlinePlus Drug Information","url":"https://medlineplus.gov/druginfo/meds/a694007.html"},{"title":"PubMed: constipation as a risk factor for haemorrhoids, anal fissure and rectal bleeding","url":"https://pubmed.ncbi.nlm.nih.gov/?term=constipation+risk+factor+hemorrhoids+anal+fissure+rectal+bleeding&size=20"},{"title":"PubMed: gabapentinoid gastrointestinal bleeding risk (no direct studies found)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=gabapentinoid+gastrointestinal+bleeding+risk&size=20"},{"title":"Ensembl REST - variant rs551836887 (location, consequence, population frequencies)","url":"https://rest.ensembl.org/variation/human/rs551836887?content-type=application/json;pops=1"},{"title":"UniProt Q9H0B8 - CRISPLD2 function, expression and disease involvement","url":"https://rest.uniprot.org/uniprotkb/Q9H0B8.txt"}]},"gabapentin|R42":{"verdict":"plausible ADR","headline":"Dizziness is a top-tier labelled gabapentin ADR; the ELOVL5 intronic hit is oversized and unreplicated","assessment":"Dizziness (R42) is not an indication for gabapentin but one of its two defining adverse reactions: the US label reports dizziness in 28% of postherpetic-neuralgia patients versus 8% on placebo and 17% versus 7% in adjunctive epilepsy trials, with ataxia at 3% and 13% respectively, plus an explicit driving-impairment warning. The Cochrane review of 37 trials and 5914 participants gives dizziness 19%, gait disturbance 14% and somnolence 14% on gabapentin, with an NNH of 7.5 for any adverse event, so the effect is dose-dependent, class-typical alpha-2-delta CNS depression rather than a rare idiosyncratic reaction. The atlas numbers are directionally consistent with that reading rather than with confounding by indication: enrichment is 0.69, meaning R42 is actually less frequent among gabapentin users than among users of other drugs, so the drug is not being channelled to already-dizzy patients, and the FAERS disproportionality is flagged strong with PRR 2.03. The variant is null for dizziness in the drug-free population (beta_disease 0.021 with se 0.08, log10p 0.1, predisposition 0.0) and unrelated to being prescribed the drug (log10p_prescribed 0.53), which is the pattern a genuine drug-by-genotype interaction should show, and z_diff is 5.67. Against that, temporality of 98.6% rests on only 73 dated participants with 33% undated and is the direction the cohort inflates anyway, so it adds little. The weak point is the genetics, not the pharmacology: a single rare intronic SNV carrying beta 4.097 (odds ratio around 60) for a reaction that afflicts roughly one in five treated patients is biologically incoherent, since a common dose-dependent CNS effect cannot be gated by a 0.26%-frequency allele of that magnitude, and an effect this large at log10p 8.01 in a sparse within-users stratum is the classic signature of few carrier cases. Gabapentin is if anything used off-label to suppress acquired pendular nystagmus and some ocular-motor causes of oscillopsia, but that is a narrow niche indication and does not offset a labelled 17-28% dizziness rate.","gene_comment":"rs141310150 sits at chr6:53,274,512 (GRCh38), genuinely inside ELOVL5 (chr6:53,260,285-53,349,179) as an intronic variant with gnomAD MAF around 0.26%, so this is not a gene desert or a lncRNA guess, and ELOVL5 is neurologically credible because dominant missense mutations in it cause spinocerebellar ataxia 38, a cerebellar syndrome with gait ataxia and nystagmus. That said, a rare intron variant is not an SCA38 allele, ELOVL5 has no known role in gabapentin pharmacokinetics (renal clearance, no CYP metabolism, target CACNA2D1, uptake via LAT1/SLC7A5) or pharmacodynamics, and the cerebellar resonance should be treated as a coincidence to be tested, not as mechanism.","score_check":"The confounding controls are believable (enrichment below 1, flat disease arm with beta well inside its standard error, flat prescription arm), but beta_adr of 4.097 for a 0.26%-frequency allele acting on a reaction with ~20% baseline incidence is not a credible effect size and most likely reflects sparse-data bias over a handful of carrier cases. The temporality figure is uninformative at n_dated 73 with a third undated.","candidability":6,"candidability_reason":"The ADR itself is real, common and clinically consequential, and the confounding checks pass, but the single rare intronic ELOVL5 variant carries an implausibly large effect with no replication or mechanistic link to gabapentin.","sources":[{"title":"Gabapentin capsules - full prescribing information (DailyMed, Amneal Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=507bf6d0-e7a2-413d-9848-cd2445a1411d"},{"title":"Wiffen PJ et al. Gabapentin for chronic neuropathic pain in adults. Cochrane Database Syst Rev 2017 (PMID 28597471)","url":"https://pubmed.ncbi.nlm.nih.gov/28597471/"},{"title":"Di Gregorio E et al. ELOVL5 mutations cause spinocerebellar ataxia 38. Am J Hum Genet 2014 (PMID 25065913)","url":"https://pubmed.ncbi.nlm.nih.gov/25065913/"},{"title":"Ensembl variation record for rs141310150 (alleles, consequence, gnomAD frequencies, GRCh38 position)","url":"https://rest.ensembl.org/variation/human/rs141310150?content-type=application/json;pops=1"},{"title":"Ensembl gene record for ELOVL5 (GRCh38 coordinates and description)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/ELOVL5?content-type=application/json"},{"title":"Strupp M et al. Current treatment of vestibular, ocular motor disorders and nystagmus. Ther Adv Neurol Disord 2009 (PMID 21179531)","url":"https://pubmed.ncbi.nlm.nih.gov/21179531/"}]},"lisinopril|E871":{"verdict":"plausible ADR","headline":"Hyponatraemia is a label-listed lisinopril ADR, but this rare lncRNA variant is a sparse-data artifact","assessment":"Hyponatraemia (E87.1) is not a licensed indication for lisinopril - the US label lists only hypertension, heart failure adjunct therapy and acute myocardial infarction - and it is not treated with an ACE inhibitor. It is, however, an explicitly recognised adverse reaction: the DailyMed lisinopril label carries 'Hyponatremia' in post-marketing experience and 'inappropriate antidiuretic hormone secretion' under endocrine adverse reactions, and the Swedish case-control study of 11,213 hospitalised hyponatraemia cases found newly initiated ACEIs carried the highest risk of the four antihypertensive classes (aOR 2.24, 95% CI 1.87-2.68), while ongoing ACEI use was not elevated. A mechanism is described - blocked peripheral conversion of angiotensin I lets excess substrate reach the brain, where central conversion to angiotensin II enhances hypothalamic ADH release, with reduced aldosterone adding renal sodium loss - but reviews call ACEI-induced hyponatraemia rare and typically early after initiation, and the 2019 case series found only 14 published reports. The atlas numbers fit that reading only loosely: enrichment is a modest 1.23, entirely compatible with heart failure, older age and above all co-prescribed thiazides (lisinopril/HCTZ fixed combinations are a major uncontrolled confounder here), and the median 5.52 years from first exposure to first sodium diagnosis is the wrong timing for the initiation-linked mechanism the literature describes, so the 100% temporality over n=79 carries essentially no weight given the cohort's known upward compression. The one reassuring column is the comparison arms: the variant is flat in the drug-free population (log10p_disease 0.45, beta 0.169 +/- 0.183, i.e. indistinguishable from noise) and flat on prescribing propensity (log10p 0.38), so this is not a disease-predisposition or channelling signal. What sinks it is the effect size itself: beta_adr 6.758 is an odds ratio near 10^3 for a variant with gnomAD MAF 0.36% across at most ~80 cases, which is the signature of a handful of carriers in a near-separated cell rather than a measurable pharmacogenomic effect, and z_diff 5.21 inherits that instability wholesale.","gene_comment":"rs183950360 is a rare intronic C>T SNV (gnomAD T = 0.36%, 1000G 0.11%) at chr8:97,427,479 inside ENSG00000309049, an unnamed Havana 'novel transcript' lncRNA (LOC101927066) on 8q22 with no characterised function and nothing known about sodium, aquaporin or vasopressin biology. This is a locus label, not a gene assignment, and it should not be presented as mechanism for ACEI-associated SIADH.","score_check":"The comparison arms behave sensibly (null disease effect with an honest standard error, null prescribing arm), but beta_adr 6.758 on a 0.36%-frequency variant in a case set of roughly 80 is not a believable effect magnitude and points to sparse-data separation despite log10p_adr 7.19. FAERS support (PRR 2.14, ROR+PRR+IC) validates the drug-condition pair at class level, not this variant.","candidability":5,"candidability_reason":"Genuine, label-listed ADR pair with a described mechanism, but the genetics are an unannotated lncRNA variant with an artifact-scale effect and the 5.5-year lag and thiazide co-prescription undercut drug causation here.","sources":[{"title":"Lisinopril tablet label (Actavis Pharma) - DailyMed: indications, post-marketing hyponatremia, inappropriate ADH secretion","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=43e2c8d1-3704-4323-bcaf-f582572b81f7"},{"title":"Falhammar H et al., Associations Between Antihypertensive Medications and Severe Hyponatremia: A Swedish Population-Based Case-Control Study, J Clin Endocrinol Metab 2020 (PMC7451505)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7451505/"},{"title":"Nakayama T et al., SIADH associated with angiotensin-converting enzyme inhibitor therapy in the perioperative period, J Renin Angiotensin Aldosterone Syst 2019 (PMC6407162)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC6407162/"},{"title":"Capinha M et al., Drug-Induced Hyponatremia: Insights into Pharmacological Mechanisms and Clinical Practice Management, J Clin Med 2025 (PMC12471027)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12471027/"},{"title":"dbSNP rs183950360 - position, alleles, intronic consequence and gnomAD/ALFA allele frequencies","url":"https://www.ncbi.nlm.nih.gov/snp/rs183950360"},{"title":"Ensembl REST lookup ENSG00000309049 - lncRNA 'novel transcript', chr8:97,419,063-97,446,484 (GRCh38)","url":"https://rest.ensembl.org/lookup/id/ENSG00000309049?content-type=application/json"}]},"lisinopril|J449":{"verdict":"co-prescription population","headline":"COPD is background comorbidity in lisinopril users, not drug-caused; P4HA3 burden effect looks sparse-data","assessment":"COPD is not a licensed indication for lisinopril - the US label lists hypertension, heart failure and acute myocardial infarction only - and chronic obstructive pulmonary disease is not a labelled adverse reaction; the only respiratory event on the label is cough, at a 2.5% excess over placebo in hypertension trials, with 6-14% incidence in the literature. The atlas itself argues against an excess: COPD is present in 2.79% of lisinopril users with an enrichment of 0.99, i.e. exactly the background rate among users of other drugs, so this is the ordinary overlap between a smoking-related airway disease and an antihypertensive/cardiac population of the same age. Temporality of 92% after first exposure with a median of 6.98 years is uninformative here: the cohort compresses temporality upward, and a seven-year lag matches age-accrued comorbidity rather than ACE-inhibitor cough, which appears within weeks to months. There is a genuine airway pharmacology - bradykinin and substance P accumulation drives ACE-inhibitor cough and can worsen bronchial hyperresponsiveness in susceptible patients - and a large cohort comparison found ACE inhibitors worse than ARBs in established COPD for pneumonia (aHR 1.22), severe exacerbations (aHR 1.19) and mortality (aHR 1.33); but that is aggravation of existing disease and a competing-drug comparison, not induction of fixed airflow obstruction. The most likely route from lisinopril to a J44.9 code is diagnostic: persistent ACE-inhibitor cough in an older hypertensive smoker triggers a respiratory work-up and an obstructive-airway label, which also plausibly explains the FAERS COPD signal (PRR 2.62) without any incident disease. So the drug-relevant question is misattribution and exacerbation risk, not a new pharmacogenomic ADR phenotype.","gene_comment":"P4HA3 is a real protein-coding gene, a collagen prolyl 4-hydroxylase alpha subunit expressed in smooth muscle including lung and implicated in TGF-beta1-driven collagen synthesis in fibrosis, so extracellular-matrix remodelling is not an absurd story - but there is no COPD or ACE-inhibitor literature for it, and a single missense burden test with no rsIDs supports nothing beyond post hoc plausibility.","score_check":"The ADR-arm beta of 28.8 for an MPC burden test on ~200 dated cases is the classic signature of quasi-complete separation, not a real effect size, and the disease-arm effect (0.642 +/- 0.829) is indistinguishable from noise, so the z_diff of 4.99 rests entirely on the unstable treated-arm estimate. The null prescribing-propensity signal (log10p 0.02) and enrichment of 0.99 are consistent and believable, and they say the condition is neither indication-driven nor over-represented.","candidability":3,"candidability_reason":"Real airway pharmacology exists but COPD sits at exactly background prevalence in these patients and the single-gene burden hit looks like a sparse-data artifact.","sources":[{"title":"Lisinopril tablets - DailyMed label (indications; cough as adverse reaction)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=8f20acd7-2635-4a9b-b732-2a84ea93dea7"},{"title":"PubMed: ACE inhibitor cough - bradykinin/substance P mechanism and incidence (incl. ACCP guidelines, PMID 16428706)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=%22ACE+inhibitor%22+cough+mechanism+bradykinin+substance+P"},{"title":"Comparative effects of ACE inhibitors and ARBs on pneumonia and severe exacerbations in COPD (Int J COPD 2018, PMC5846309)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC5846309/"},{"title":"PubMed: ACE inhibitors, asthma and airway hyperresponsiveness","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ACE+inhibitor+asthma+bronchial+hyperresponsiveness+airway"},{"title":"Human Protein Atlas: P4HA3 (prolyl 4-hydroxylase subunit alpha 3)","url":"https://www.proteinatlas.org/ENSG00000149380-P4HA3"},{"title":"PubMed: P4HA3 literature (collagen synthesis/fibrosis, TGF-beta1)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=P4HA3"}]},"lisinopril|K20":{"verdict":"co-prescription population","headline":"Oesophagitis is a cardiovascular-comorbidity, not a lisinopril effect; the rare 16p13.3 variant is sparse-data","assessment":"Lisinopril is licensed only for hypertension, heart failure and post-myocardial-infarction mortality reduction, so oesophagitis (K20) is not an indication. The US label's gastrointestinal adverse reactions are pancreatitis, constipation, flatulence, dry mouth and diarrhoea, plus nausea/vomiting/abdominal pain in the context of intestinal angioedema; oesophagitis, reflux and GERD appear nowhere in the clinical-trial or post-marketing sections, and StatPearls likewise lists only cough, angioedema, hyperkalaemia, hypotension and renal effects. The pill-oesophagitis literature implicates tetracyclines, NSAIDs, bisphosphonates and potassium chloride, not ACE inhibitors, and no ACE-inhibitor mechanism for oesophageal mucosal injury or lower-oesophageal-sphincter relaxation is described. The only credible indirect route is diagnostic: ACE-inhibitor cough affects 5-30% of users and reflux is a standard differential in chronic cough, but current guidance explicitly advises against routine endoscopy or pH monitoring for extraoesophageal symptoms alone, so this can inflate coding only modestly. The atlas numbers fit comorbidity rather than causation: 1.79% co-occurrence with an enrichment of only 1.36 is the ordinary excess of reflux disease in an older, heavier, poly-treated cardiovascular population (aspirin, NSAIDs, nitrates, calcium blockers all in the same patients), and the 100% temporality on 166 dated participants with a median 5.79 years to onset is exactly the pattern the cohort's record-length artefact produces, not an acute drug reaction. FAERS support is weak (1 of 3, PRR 1.93), consistent with the same background co-occurrence.","gene_comment":"rs563549874 is a rare (MAF ~0.2%) regulatory-region variant at chr16:804,252 in the gene-dense 16p13.3 telomeric region, tens of kilobases from GNG13, so the gene call is proximity-based rather than functional; GNG13 is a G-protein gamma subunit studied in taste buds, cerebellum and tumour prognosis, with no oesophageal or ACE/bradykinin biology to offer a mechanism.","score_check":"An effect of beta 2.671 (OR ~14) at 0.2% allele frequency with only log10p 6.05 - below genome-wide significance - is the classic sparse-data signature of a handful of carrier cases. The z_diff of 4.7 is driven largely by an uninformative comparison arm (beta_disease 0.083 +/- 0.09, essentially zero power for a variant this rare), so 'specific to treated patients' is not established.","candidability":2,"candidability_reason":"No label or literature basis for a lisinopril-caused oesophagitis, modest enrichment explained by cardiovascular comorbidity, and a sub-genome-wide rare-variant hit next to an implausible gene.","sources":[{"title":"LISINOPRIL (lisinopril) tablet - DailyMed label, Camber Pharmaceuticals","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=8f20acd7-2635-4a9b-b732-2a84ea93dea7"},{"title":"Lisinopril - StatPearls, NCBI Bookshelf (NBK482230)","url":"https://www.ncbi.nlm.nih.gov/books/NBK482230/"},{"title":"PubMed: drug-induced / pill esophagitis reviews (implicated drug classes)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=drug-induced+esophagitis+pill+esophagitis"},{"title":"Chronic Cough: Evaluation and Management - Am Fam Physician 2024","url":"https://www.aafp.org/pubs/afp/issues/2024/0800/chronic-cough.html"},{"title":"Ensembl REST: variant rs563549874 (chr16:804,252, MAF 0.002, regulatory region)","url":"https://rest.ensembl.org/variation/human/rs563549874?content-type=application/json"},{"title":"PubMed: GNG13 literature (taste/cerebellum/tumour prognosis)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=GNG13"}]},"lisinopril|N309":{"verdict":"statistical artifact","headline":"Cystitis is not a lisinopril effect: no label signal, ACEi bladder data point the other way, rare X-linked hit","assessment":"Cystitis is neither a licensed indication for lisinopril (hypertension, systolic heart failure, post-MI mortality reduction) nor a labelled adverse reaction: the US prescribing information lists only 'Urogenital: Impotence' among urogenital events and no urinary tract infection or cystitis term. The one genuine ACE-inhibitor link to the lower urinary tract in the literature is indirect and mechanically different - ACEi cough (5-35% of users, bradykinin/tachykinin mediated) can provoke or worsen stress incontinence, which is a continence problem, not an infection. Where ACE inhibition has been studied against bladder injury the direction is protective, not harmful: ACEi use during prostate radiotherapy was associated with roughly half the rate of late haematuria (adjusted HR 0.51, 95% CI 0.28-0.94), and captopril-type agents are used experimentally to blunt chemically induced urotoxic cystitis, so a drug-caused excess of cystitis has no supporting mechanism and some evidence against it. The atlas exposure numbers are also unremarkable: only 1.17% of lisinopril users carry the code and enrichment is 1.15, i.e. cystitis is barely more common in these patients than in users of other drugs, consistent with an elderly, largely female, comorbid hypertensive population rather than with drug causation. The FAERS 'STRONG' flag at PRR 2.2 is the kind of disproportionality that a very widely prescribed antihypertensive generates against a common ambulatory diagnosis, and it is not corroborated by any label or BNF entry (on_bnf = 0). Temporality of 87.2% on 86 dated participants is exactly the pattern the cohort inflates by design (prescription records predate diagnosis records), so it adds essentially nothing. The genetic signal is a single rare intronic SNV whose effect size is implausible for the outcome, so the whole association rests on one fragile locus rather than on a coherent clinical story.","gene_comment":"rs192999440 is an intronic variant on chromosome X (~8.31 Mb) inside ENSG00000285679, an unnamed lncRNA 'novel transcript' with no known bladder or renin-angiotensin biology; an X-linked, uncharacterised, non-coding assignment is about the weakest gene evidence available. It also raises a specific worry for a female-predominant outcome like cystitis, since chrX association can track sex-related genotype handling and case-mix rather than biology.","score_check":"beta_adr 4.886 with log10p 8.13 implies a standard error near 0.85 and an odds ratio around 130 - a sparse-data / near-separation signature for a rare variant, not a credible effect size for cystitis. The comparison arms are flatly null (log10p_disease 0.16 with beta 0.069 +/- 0.17, log10p_prescribed 0.13), so z_diff 5.59 simply reflects the inflated within-users estimate rather than a real treated-versus-untreated difference.","candidability":2,"candidability_reason":"No label or mechanistic support, ACEi bladder literature runs protective, enrichment near 1, and the signal is one rare X-linked intronic lncRNA variant with an implausibly large effect for its error.","sources":[{"title":"Lisinopril tablets - full prescribing information (DailyMed, Camber Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=8f20acd7-2635-4a9b-b732-2a84ea93dea7"},{"title":"Managing therapeutic competition in patients with heart failure, lower urinary tract symptoms and incontinence (PMC3907694)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3907694/"},{"title":"Use of angiotensin converting enzyme inhibitors is associated with reduced risk of late bladder toxicity following radiotherapy for prostate cancer (PMC8986577)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC8986577/"},{"title":"Pharmacological and molecular evidence for kinin B1 receptor expression in urinary bladder of cyclophosphamide-treated rats (PMC1571608)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC1571608/"},{"title":"Ensembl REST: gene ENSG00000285679 (lncRNA, novel transcript, chrX:7,928,443-8,444,335)","url":"https://rest.ensembl.org/lookup/id/ENSG00000285679?content-type=application/json"},{"title":"Ensembl REST: variant rs192999440 (chrX:8,312,268, intron variant)","url":"https://rest.ensembl.org/variation/human/rs192999440?content-type=application/json"}]},"lisinopril|R21":{"verdict":"plausible ADR","headline":"Rash is a labelled lisinopril reaction, but the MED15 rare-variant hit is a sparse-data artifact","assessment":"Rash is not an indication for lisinopril and is a recognised adverse reaction: the US label lists rash, pruritus, urticaria, photosensitivity, erythema, cutaneous pseudolymphoma, Stevens-Johnson syndrome and toxic epidermal necrolysis, and notes that rash may occur alone or as part of a symptom complex with fever, arthralgia, myalgia and vasculitis. The literature supports a real class effect with delayed onset - a case series of 23 elderly hypertensives found generalised or localised eczematous, spongiotic eruptions appearing 4-30 months after starting an ACE inhibitor or ARB, and an Italian surveillance programme found ACE inhibitors to be the main cause of pityriasis rosea-like drug eruptions, all resolving on withdrawal. The atlas numbers are consistent with rash being a background-rate event rather than a channelling artefact: cotx_pct is 1.33% with enrichment 0.94, so rash is no commoner in lisinopril users than in users of other drugs, and log10p_prescribed 0.45 shows no confounding by propensity to be prescribed. The variant, however, does not survive scrutiny: beta_adr 5.598 is an odds ratio of roughly 270 for a 5'UTR/intronic variant with a European allele frequency near 0.5-0.9%, in a case set that can hold at most ~70 rash events, so a handful of carriers - possibly one or two - drive the entire log10p of 8.88. The drug-free comparison is flatly null (beta_disease -0.102 +/- 0.126, log10p_disease 0.38), which is what makes z_diff 6.12 large, but a z_diff built on a separation-scale treated-arm beta measures instability, not specificity. Temporality is uninformative here: 100% after exposure rests on only 24 dated cases out of 67.1% undated, and a median of 8.02 years to first record is far longer than the weeks-to-months onset described for ACE-inhibitor eruptions.","gene_comment":"rs75946839 is a rare (MAF ~0.5-0.9%) 5'UTR/intron variant in MED15 at 22q11.21, a ubiquitously expressed Mediator-complex transcriptional coactivator with no described role in bradykinin/substance P handling, drug hypersensitivity or any cutaneous phenotype. There is no HLA or immune-region signal here, which is what a genuine drug-eruption pharmacogenetic hit would most likely look like.","score_check":"The class-level clinical numbers (enrichment ~1, no prescription signal, null disease effect) are believable and internally coherent, but beta_adr 5.598 for a ~0.5% allele against at most a few dozen rash cases is a sparse-data separation artefact rather than a real effect size. The 100% temporality on 24 dated records with 67% undated carries essentially no weight.","candidability":5,"candidability_reason":"Rash is a genuine, labelled, mechanistically described ACE-inhibitor adverse effect, but the specific MED15 rare-variant signal here is an implausibly large sparse-data hit in a gene with no dermatologic or renin-angiotensin relevance.","sources":[{"title":"LISINOPRIL tablet - DailyMed label (Camber Pharmaceuticals), ADVERSE REACTIONS","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=8f20acd7-2635-4a9b-b732-2a84ea93dea7"},{"title":"Eczematous reactions due to angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers (Immunopharmacol Immunotoxicol, 2013; PMID 23672527)","url":"https://pubmed.ncbi.nlm.nih.gov/23672527/"},{"title":"Pityriasis rosea-like adverse reaction: review of the literature and experience of an Italian drug-surveillance center (Dermatol Online J, 2006; PMID 16638369)","url":"https://pubmed.ncbi.nlm.nih.gov/16638369/"},{"title":"dbSNP/NCBI Variation Services record for rs75946839 (MED15, chr22:20,508,260, 5'UTR/intron; gnomAD MAF ~0.5-0.9%)","url":"https://api.ncbi.nlm.nih.gov/variation/v0/refsnp/75946839"},{"title":"UniProt Q96RN5 - MED15_HUMAN, Mediator of RNA polymerase II transcription subunit 15","url":"https://rest.uniprot.org/uniprotkb/Q96RN5.txt"}]},"lisinopril|R600":{"verdict":"contested","headline":"Lisinopril has a famous swelling ADR - but it is angioedema (T78.3), not R60.0 localised oedema","assessment":"Localised oedema is not an indication for lisinopril; the label lists hypertension, heart failure and early STEMI, and neither peripheral nor localised oedema appears among its clinical-trial or post-marketing adverse reactions. What the label does carry, in a warning, is angioedema of the face, extremities, lips, tongue, glottis and larynx, including fatal cases and intestinal angioedema, driven by impaired bradykinin degradation - but ICD-10 codes that as T78.3 (angioneurotic oedema), while R60.0 is the generic localised-swelling code that in a hypertension/heart-failure population is dominated by dependent ankle oedema, venous insufficiency, decompensated heart failure and co-prescribed dihydropyridine calcium-channel blockers. For that peripheral phenotype the drug moves the needle the other way: ACE inhibitors and ARBs are used to relieve congestive oedema and to blunt amlodipine-type pedal oedema, so an ADR reading only survives if a meaningful share of these R60.0 codes are really mis-coded facial angioedema. The atlas numbers are consistent with a non-indication phenotype rather than with an ADR: cotx_pct 0.69% with enrichment 0.94 means localised oedema is no commoner in lisinopril users than in users of other drugs, and log10p_prescribed 0.17 rules out prescription-propensity confounding. Temporality is 90% after exposure on only 80 dated participants with a median lag of 3.72 years, which the cohort's known upward compression makes weak evidence, and 3.7 years fits accumulating cardiovascular comorbidity better than a bradykinin-mediated reaction. The strong FAERS signal (PRR 12.81) most plausibly reflects the angioedema/swelling reporting cluster for the class rather than validation of R60.0 specifically.","gene_comment":"Both assignments are empty: ENSG00000232389 is a processed spermine-synthase pseudogene on chr6 (~70.6 Mb) and ENSG00000287526 is an unnamed novel lncRNA on chr19 (~34.0 Mb), neither anywhere near the loci that real ACEi-angioedema genetics point at (BDKRB2 14q32, KCNMA1 10q22, XPNPEP2 Xp22, F5/PROCR 1q24, PRKCQ 10p15, MME 3q25). rs150583919 and rs73031262 therefore contribute no mechanism and should not be presented as bradykinin-pathway support.","score_check":"beta_adr of 5.743 (odds ratio in the hundreds) at only log10p 6.05 is the signature of a rare variant in a handful of cases - a sparse-data effect estimate, not a real effect of that size - and the disease arm is frankly null (beta 0.277 vs se 0.168, |beta| < 1.96*se), so z_diff 4.63 is carried entirely by the fragile treated-arm estimate. The population-level fields (enrichment 0.94, prescribed 0.17) look believable; the variant-level effect size does not.","candidability":4,"candidability_reason":"The class has a genuine, label-warned swelling ADR, but this is the wrong ICD code for it, the drug reduces the peripheral oedema R60.0 mostly captures, enrichment is neutral and the two variants sit on a pseudogene and a novel lncRNA with a sparse-data beta.","sources":[{"title":"Lisinopril - StatPearls (NCBI Bookshelf): indications, mechanism, angioedema","url":"https://www.ncbi.nlm.nih.gov/books/NBK482230/"},{"title":"Lisinopril tablets - US prescribing information (DailyMed, Aurobindo Pharma)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=0e6364a4-6d66-4151-8197-d45e2a762895"},{"title":"Genome-wide association study of angioedema induced by ACE inhibitor and ARB treatment (Pharmacogenomics J 2020; KCNMA1)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7674154/"},{"title":"Europe PMC search: ACE-inhibitor angioedema genome-wide association studies (BDKRB2, F5, PROCR, PRKCQ, RIMS1)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22angiotensin-converting%20enzyme%20inhibitor%22%20AND%20angioedema%20AND%20%22genome-wide%20association%22&format=json&pageSize=15&resultType=core"},{"title":"Europe PMC search: calcium-channel-blocker pedal oedema and renin-angiotensin blockade","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22amlodipine%22%20AND%20%22peripheral%20edema%22%20AND%20(%22ACE%20inhibitor%22%20OR%20%22renin-angiotensin%22)%20AND%20reduction&format=json&pageSize=10&resultType=core"},{"title":"Ensembl REST lookup ENSG00000232389 (SMS processed pseudogene, chr6) and ENSG00000287526 (novel lncRNA, chr19)","url":"https://rest.ensembl.org/lookup/id/ENSG00000232389?content-type=application/json"}]},"lisinopril|R739":{"verdict":"statistical artifact","headline":"ACE inhibitors lower, not raise, diabetes risk; a rare RYR2 intron with beta 6.5 is sparse-data separation","assessment":"Hyperglycaemia (R73.9) is not a licensed indication for lisinopril and the direction of the licensed drug effect runs against an ADR reading: the US label carries no hyperglycaemia warning but does warn that lisinopril combined with insulin or oral antidiabetics 'may cause an increased blood-glucose-lowering effect with risk of hypoglycaemia', with post-marketing hypoglycaemia reports in diabetic patients; the only nod to the opposite direction is a legacy laundry-list 'Endocrine: diabetes mellitus' entry among uncontrolled adverse-experience reports. The randomised evidence is concordant and decisive: in the Lancet 2021 individual-participant meta-analysis of 145,939 patients across 19 trials, ACE inhibitors reduced new-onset type 2 diabetes versus placebo (RR 0.84, 95% CI 0.76-0.93), while beta-blockers and thiazides increased it - and in ALLHAT lisinopril was the metabolically favourable comparator against chlorthalidone. The atlas's own comorbidity arm agrees: enrichment is 0.81, i.e. hyperglycaemia is coded slightly less often in lisinopril users than in users of other drugs, which is what a glucose-neutral-to-protective antihypertensive should look like. The FAERS 'STRONG' flag (PRR 2.23) is the expected confounded pattern for a drug prescribed almost exclusively to hypertensive, obese and diabetic patients, in whom hyperglycaemia is a background event rather than a drug effect. The temporality of 88.5% is uninformative here: n_dated is only 78, and the cohort is known to compress this figure upward, so a high value carries no weight. What remains is a nominal predisposition signal (beta_disease 0.71 +/- 0.217, z = 3.3, OR ~2 in the drug-free population) that would make the variant a weak hyperglycaemia risk allele irrespective of lisinopril, not a drug-gene interaction.","gene_comment":"rs138333572 is an intronic T/C variant at chr1:237,808,852 inside RYR2, MAF ~0.9%, annotated likely benign; RYR2 is a ~790 kb gene that is a well-known magnet for intronic false positives. RYR2 does amplify glucose-stimulated insulin secretion via calcium-induced calcium release in beta cells, but reviews of islet ryanodine receptors judge this a modulatory role with limited relevance to human diabetes pathogenesis, and nothing connects it to ACE inhibition.","score_check":"beta_adr of 6.511 (OR ~670) for a 0.9%-frequency variant tested inside a drug-user subset is the signature of quasi-complete separation on a handful of carriers, not a real effect size, and z_diff 4.6 inherits that artifact wholesale. Only the disease arm is numerically believable (beta 0.71 +/- 0.217), and it describes predisposition in untreated people, while log10p_prescribed of 1.02 at least rules out prescription-channel bias.","candidability":2,"candidability_reason":"Backwards signal: the drug class demonstrably reduces incident diabetes and the label warns of hypoglycaemia, the condition is depleted in lisinopril users, and the within-drug beta is a sparse-data artifact.","sources":[{"title":"DailyMed - LISINOPRIL tablet (US prescribing information, Camber Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=8f20acd7-2635-4a9b-b732-2a84ea93dea7"},{"title":"Nazarzadeh M et al. Blood pressure lowering and risk of new-onset type 2 diabetes: an individual participant data meta-analysis. Lancet 2021 (PMID 34774144)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:34774144&resultType=core&format=json"},{"title":"ALLHAT Collaborative Research Group. Major outcomes in high-risk hypertensive patients randomized to ACE inhibitor or calcium channel blocker vs diuretic. JAMA 2002 (PMID 12479763)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:12479763&resultType=core&format=json"},{"title":"Ensembl REST - variation record for rs138333572 (RYR2 intron, MAF 0.0092, likely benign)","url":"https://rest.ensembl.org/variation/human/rs138333572?content-type=application/json;pops=1"},{"title":"Islam MS. Ryanodine receptors in islet cell function: calcium signaling, hormone secretion, and diabetes. Cells 2025 (Europe PMC search)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=RYR2%20AND%20%28%22insulin%20secretion%22%20OR%20%22beta%20cell%22%29&resultType=core&format=json&pageSize=5"},{"title":"PubMed search: ACE inhibitors and new-onset diabetes meta-analyses (incl. Geng DF et al., Int J Cardiol 2013, PMID 22809536)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ACE+inhibitor+new-onset+diabetes+meta-analysis"}]},"losartan|L089":{"verdict":"statistical artifact","headline":"Rare intergenic variant, OR in the thousands: sparse-data artifact on a comorbidity backdrop","assessment":"Local skin infection (ICD-10 L08.9) is not an indication for losartan, whose licensed uses are hypertension, stroke reduction in LVH and diabetic nephropathy in type 2 diabetes, and it is not a labelled adverse reaction: the DailyMed US label lists only urticaria, pruritus, rash, photosensitivity, and postmarketing angioedema, erythroderma and vasculitis under skin events, all allergic or inflammatory rather than infectious. No mechanism linking AT1-receptor blockade to bacterial skin infection is described, and the drug is not immunosuppressive; the only infection-flavoured label entry is upper respiratory infection at 8% vs 7% on placebo, i.e. not a real excess. What the atlas is almost certainly seeing on the population side is the treated group itself: losartan is specifically licensed for diabetic nephropathy, and diabetes carries a roughly 1.5- to 4-fold increase in common and severe infections, while diabetes, obesity, chronic kidney disease, venous disease and lymphoedema are the standard documented risk factors for cellulitis and other local skin infections. Even so the atlas measures almost none of that here: enrichment is 1.13, cotx_pct 1.88%, FAERS PRR 1.13 with no signal, and the drug is not BNF-flagged, so losartan users are barely more affected than users of other drugs. Temporality of 87.7% on 81 dated participants with a median 3.4 years is uninformative, because an incident, recurrent, acute infection will nearly always postdate the start of a chronic antihypertensive regardless of causation. The genetic claim is the weakest part: a beta_adr of 8.52 on the log-odds scale implies an odds ratio in the thousands for a condition affecting 1.9% of users, which is not a biological effect size but quasi-separation from a handful of carriers of a variant with gnomAD frequency 0.29%.","gene_comment":"rs573957134 has no gene assignment in the fact sheet and dbSNP confirms it is intergenic at chr20:56216436 (GRCh38); the nearest annotated features are a 5S rRNA pseudogene (RNA5SP487) about 11 kb away and MC3R about 32 kb away, neither with any skin-immunity role, so no mechanistic reading is available.","score_check":"The numbers do not hang together: beta_disease 0.049 +/- 0.149 is flatly null in the drug-free population and log10p_prescribed 0.23 shows no prescribing link, so a log10p_adr of 9.0 with beta 8.52 for a rare intergenic SNV exists only in this one small cell. The z_diff of 6.04 is driven by the same implausible within-user estimate rather than by any real difference in effect.","candidability":1,"candidability_reason":"No label or mechanistic basis, negligible enrichment and FAERS signal, and an implausible odds ratio from a very rare intergenic variant.","sources":[{"title":"Losartan Potassium tablets - DailyMed label (indications, adverse reactions, postmarketing)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a8817f22-8478-49b9-8354-ab27ca222c6b"},{"title":"dbSNP rs573957134 (chr20:56216436, intergenic, gnomAD G allele 0.0029)","url":"https://www.ncbi.nlm.nih.gov/snp/rs573957134"},{"title":"Ensembl REST overlap, chr20:56166436-56266436 - nearest genes MC3R and RNA5SP487","url":"https://rest.ensembl.org/overlap/region/human/20:56166436-56266436?feature=gene;content-type=application/json"},{"title":"DermNet NZ - Cellulitis: risk factors (diabetes, CKD, obesity, venous disease, lymphoedema, immunosuppression)","url":"https://dermnetnz.org/topics/cellulitis"},{"title":"Diabetes Mellitus and Infectious Diseases: Current Evidence and Clinical Implications (Diabetes Metab J, 2025; PMID 40859782)","url":"https://europepmc.org/article/MED/40859782"}]},"losartan|N309":{"verdict":"statistical artifact","headline":"Cystitis in losartan users: AT1 blockade protects the bladder experimentally; rare intronic DMD hit is noise","assessment":"Cystitis is not an indication for losartan, whose US label covers hypertension, stroke reduction in LVH, and diabetic nephropathy; the only urinary term in the Adverse Reactions section is 'urinary tract infection' among adverse events in the diabetic-nephropathy (RENAAL) population, where diabetes itself is the dominant UTI risk factor rather than the drug. The atlas is consistent with that reading and not with an ADR: only 1.67% of losartan users carry an N30.9 code and enrichment over other drugs' users is 1.23, i.e. barely above the background expected for an older, hypertensive, often diabetic and female-skewed treated population. Temporality is 73.6% after first exposure on just 72 dated participants with a median 5.3 years' lag, which in a cohort whose prescription records reach further back than diagnoses is close to the null expectation and carries little weight. Pharmacologically the direction of effect runs against an ADR: AT1-receptor blockade abolished inflammation and oedema in murine autoimmune cystitis, olmesartan attenuated TNF-alpha/IL-6/NF-kB and mucosal damage in cyclophosphamide-induced haemorrhagic cystitis in rats, and telmisartan reduced experimentally induced detrusor overactivity, so ARBs are studied as bladder-protective rather than bladder-irritant agents. The FAERS support is explicitly weak (1 of 3, PRR 1.79) and cystitis is not a BNF-listed effect. Taken together this is a common comorbidity of the treated population attached to an uninterpretable genetic hit, not a drug-caused reaction.","gene_comment":"rs149408540 is a rare (roughly 0.1-1% gnomAD) intronic variant at chrX:32,609,243, assigned to DMD only because DMD spans 2.2 Mb of the X chromosome - the largest mutational target in the genome and a gene with no bladder or urothelial biology. A rare X-linked variant tested against a strongly female-biased phenotype is also exactly the configuration in which hemizygous coding and sex-imbalance produce spurious hits, so this is positional annotation, not mechanism.","score_check":"beta_adr = 7.489 implies an odds ratio near 1800 for a rare intronic variant, which is a sparse-data separation artifact rather than a real effect, and the same variant is flat in the drug-free population (log10p_disease 0.67, beta 0.198 +/- 0.16, well inside noise), so z_diff = 6.04 simply reflects the inflated treated-arm estimate. The prescription arm is null (log10p 0.24), which rules out prescribing bias but does nothing to rescue an effect size this implausible.","candidability":1,"candidability_reason":"Non-indication comorbidity of an elderly hypertensive/diabetic population, with pharmacology pointing the opposite way and a huge implausible beta on a rare intronic X-linked variant in DMD.","sources":[{"title":"Losartan Potassium tablet - DailyMed label (indications and adverse reactions)","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=a8817f22-8478-49b9-8354-ab27ca222c6b"},{"title":"Phull H et al. Angiotensin II plays a role in acute murine experimental autoimmune cystitis. BJU Int 2007","url":"https://pubmed.ncbi.nlm.nih.gov/17550411/"},{"title":"Olmesartan ameliorates cyclophosphamide-induced hemorrhagic cystitis in rats via Nrf2/HO-1 signaling. Tissue Cell 2022","url":"https://pubmed.ncbi.nlm.nih.gov/35930992/"},{"title":"Juszczak K, Maciukiewicz P. The angiotensin II receptor type 1 blockage affects the urinary bladder activity. Adv Clin Exp Med 2017","url":"https://pubmed.ncbi.nlm.nih.gov/29211350/"},{"title":"Ensembl REST: rs149408540 (chrX:32609243, intron variant, rare)","url":"https://rest.ensembl.org/variation/human/rs149408540?content-type=application/json"},{"title":"Ensembl REST: DMD gene record (chrX:31,097,677-33,339,609, dystrophin)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/DMD?content-type=application/json"}]},"propranolol|R31":{"verdict":"statistical artifact","headline":"Ultra-rare lncRNA intron variant with an OR near 400; haematuria is not a propranolol reaction","assessment":"Unspecified haematuria is neither a licensed indication for propranolol nor a labelled adverse reaction: the approved indications are hypertension, angina, atrial fibrillation, post-MI, migraine prophylaxis, essential tremor, hypertrophic subaortic stenosis and pheochromocytoma, and the label's genitourinary adverse-reaction entries are only impotence and Peyronie's disease. The only bleeding-adjacent label entries are agranulocytosis and thrombocytopenic/nonthrombocytopenic purpura, a rare haematologic route that has not been reported as producing haematuria, and a Europe PMC sweep of beta-blocker plus haematuria returns no drug-induced haematuria literature at all. Where propranolol and haematuria do meet in the literature the direction is the opposite: propranolol shrank an infantile bladder haemangioma with resolution of the bleeding (BMJ Case Rep 2019), and in cirrhosis-associated IgA nephropathy propranolol given for portal hypertension was followed by a fall in proteinuria and haematuria. The atlas numbers agree that this is not a treated-population artefact either but do not rescue it as an ADR: cotx_pct 1.91% with enrichment 0.60 means haematuria is actually depleted among propranolol users relative to users of other drugs, consistent with propranolol's skew toward younger migraine, tremor and anxiety patients. Temporality of 97.5% rests on only 80 dated participants and, per the atlas's own calibration, a high value is weak evidence; the 6.16-year median lag to first record fits incidental urological disease of ageing rather than a drug effect. The comparison arms are clean nulls (log10p_disease 0.33, log10p_prescribed 0.69), so z_diff 5.98 is entirely a property of one unstable within-user beta rather than of a demonstrated drug-by-genotype interaction.","gene_comment":"rs750141766 is an ultra-rare intronic SNV at chr4:38,455,849 (C>A, gnomAD MAF ~0.2% overall, ~0.4% in non-Finnish Europeans) inside LINC01258, an uncharacterised long intergenic non-coding RNA at 4p14. The nearest protein-coding genes are KLF3 about 200 kb away and the TLR10/TLR1/TLR6 cluster about 320 kb away, far too distant to support any mechanistic story about bleeding or the urinary tract.","score_check":"beta_adr 5.964 is an odds ratio close to 400 for an allele carried by roughly 1 in 250 Europeans, the signature of quasi-complete separation on a handful of carriers rather than a real effect size, and log10p_adr 8.54 inherits that instability. The disease arm is a well-behaved null (beta -0.083, se 0.114), which makes the large z_diff a statement about the fragile arm only.","candidability":1,"candidability_reason":"No label, guideline or case support for propranolol-induced haematuria, published cases point the opposite way, the condition is depleted in users, and the signal is one ultra-rare intronic lncRNA variant with a separation-scale effect size.","sources":[{"title":"Propranolol Hydrochloride Tablet label (Mylan) - Indications and Adverse Reactions, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=b94dcec7-7633-4a33-9cea-48161adeaf93"},{"title":"Propranolol - StatPearls (NCBI Bookshelf, NBK557801)","url":"https://www.ncbi.nlm.nih.gov/books/NBK557801/"},{"title":"Hematuria - StatPearls (NCBI Bookshelf, NBK534213)","url":"https://www.ncbi.nlm.nih.gov/books/NBK534213/"},{"title":"PubMed search: propranolol hematuria","url":"https://pubmed.ncbi.nlm.nih.gov/?term=propranolol+hematuria"},{"title":"Propranolol as an effective therapy for infantile haemangioma of the urinary bladder (BMJ Case Rep 2019, PMID 30709884)","url":"https://europepmc.org/article/MED/30709884"},{"title":"Ensembl REST: rs750141766 variation record and LINC01258 gene record","url":"https://rest.ensembl.org/variation/human/rs750141766?content-type=application/json;pops=1"}]},"propranolol|R51":{"verdict":"licensed indication","headline":"Headache is propranolol's licensed indication; the SLC9A2 burden beta of 22 is a separation artifact","assessment":"R51 (headache) is not an adverse effect of propranolol but the condition propranolol is licensed to prevent: the US prescribing information for propranolol extended-release states it is 'indicated for the prophylaxis of common migraine headache', and StatPearls lists migraine prophylaxis among the FDA-approved indications alongside hypertension, angina, essential tremor and infantile haemangioma. The drug moves this endpoint in the opposite direction to an ADR: a 2025 randomised trial found low-dose propranolol 80 mg/day gave a greater fall in monthly headache days than amitriptyline, with 60% versus 43% of patients reaching 50% reduction, and both agents well tolerated. Headache is not listed among the label's CNS adverse reactions (which are light-headedness, fatigue, lassitude, insomnia, vivid dreams and depression), consistent with on_bnf = 0, and the FAERS signal here is weak (1/3, PRR 1.94) for a term that is among the highest-background reported events for any drug. The atlas confounding markers are mixed rather than reassuring: enrichment 0.95 says unspecified headache is no commoner in propranolol users than in users of other drugs, but that is expected when the true indication is coded as migraine (G43) and R51 captures the residual symptom coding, and the indication score of 18.7 flags the pairing anyway. Temporality of 98.8% on only 84 dated participants is the uninformative direction by the atlas's own rule, and in a drug given for headache a first R51 code a median 1.79 years after starting simply marks continuing or breakthrough headache under prophylaxis. The one direction in which propranolol can genuinely worsen headache — rebound on abrupt withdrawal, for which the label already warns that dosage be tapered over at least a few weeks — is a discontinuation phenomenon that this design cannot see and that the SLC9A2 result does not speak to.","gene_comment":"SLC9A2 encodes NHE2, an apical plasma-membrane Na+/H+ exchanger expressed mainly in colon, kidney and skeletal muscle with no established neuronal or trigeminovascular role, and a Europe PMC search returns no paper linking SLC9A2/NHE2 to migraine, headache or pain. As a single-variant MPC burden test it carries no mechanistic weight for a headache phenotype.","score_check":"beta_adr = 22.48 is not a real effect size but the signature of quasi-complete separation in a sparse burden test, and the companion disease effect (beta 0.531, se 0.735) is indistinguishable from zero, so the z_diff of 4.89 is inherited from the artifact rather than from a genuine treated-versus-untreated contrast. With log10p_disease 0.33 and log10p_prescribed 0.21 there is no supporting structure anywhere else in the sheet.","candidability":1,"candidability_reason":"The condition is the drug's own licensed indication and propranolol reduces it, while the genetic signal is an implausible sparse-burden beta in a gene with no headache biology.","sources":[{"title":"Propranolol hydrochloride extended-release capsules, US prescribing information (DailyMed)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a4edba76-29b8-41fb-bf0b-d4633abba003"},{"title":"Propranolol - StatPearls, NCBI Bookshelf (NBK557801)","url":"https://www.ncbi.nlm.nih.gov/books/NBK557801/"},{"title":"Roy et al., Low-Dose Propranolol versus Amitriptyline for Episodic Migraine Prophylaxis: A Randomized Controlled Trial, Clin Drug Investig 2025 (PMID 40932597)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:40932597&format=json&resultType=core"},{"title":"UniProtKB Q9UBY0 - Sodium/hydrogen exchanger 2 (SLC9A2/NHE2)","url":"https://rest.uniprot.org/uniprotkb/Q9UBY0.txt"},{"title":"Europe PMC literature search: SLC9A2 AND (migraine OR headache OR pain) - no relevant hits","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=SLC9A2%20AND%20%28migraine%20OR%20headache%20OR%20pain%29&format=json&pageSize=15&resultType=core"},{"title":"Europe PMC literature search: propranolol AND migraine prophylaxis","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=propranolol%20AND%20migraine%20prophylaxis%20AND%20SRC%3AMED&format=json&pageSize=15&resultType=core"}]},"ramipril|A099":{"verdict":"plausible ADR","headline":"Labelled ramipril GI effect and the classic misread of ACEi visceral angioedema; PTPRD hit is not mechanism","assessment":"Gastroenteritis (A09.9) is not a ramipril indication - the label covers hypertension, cardiovascular risk reduction and post-MI heart failure - and the atlas agrees, with enrichment 1.07 (no excess of the code among ramipril users) and log10p_prescribed 0.66 (the variant does not predict being prescribed the drug), so confounding by indication is not the explanation here. The US label lists 'gastroenteritis' verbatim among gastrointestinal adverse reactions and reports AIRE-trial rates above placebo for vomiting (2% vs 0.5%), diarrhoea (1% vs 0.4%) and nausea (2% vs 1%), so a nonspecific GI adverse effect is genuinely labelled. A specific and well-described mechanism also exists in the same direction: ACE (kininase II) inhibition raises bradykinin and causes intestinal/visceral angioedema, which the label itself flags as abdominal pain with or without nausea or vomiting, and which the case literature repeatedly shows being managed first as 'presumed acute gastroenteritis' or IBS, sometimes for years, with a FAERS review finding 303 intestinal-angioedema reports in 2004-2024 and ACE inhibitors as the leading suspects. There is no opposite-direction argument - ramipril is not protective against gastroenteritis - and the observed latency (median 5.34 years) is compatible with visceral angioedema, which is reported anywhere from weeks to eleven years after starting an ACE inhibitor. The weakness is the phenotype and the genetics rather than the pharmacology: A09.9 in an EHR is overwhelmingly ordinary infectious or unspecified GI illness, of which true bradykinin-mediated bowel oedema would be a vanishing fraction, so the coded outcome is mostly noise with respect to the mechanism that makes the pairing interesting. The atlas comparison arms are clean (beta_disease 0.127 +/- 0.077 is inside noise, predisposition 0.0, z_diff 4.9), but that only says the variant acts, if at all, within treated patients - it does not connect the variant to the bradykinin pathway. As an ADR the pairing is real and labelled; as a pharmacogenomic finding it rests on a single low-frequency intronic SNV with no mechanistic anchor.","gene_comment":"rs72698987 is a low-frequency intronic variant (European MAF ~1.6%) at chr9:10.43 Mb inside PTPRD, a ~2.3 Mb receptor tyrosine phosphatase that is one of the largest human genes and a brain-expressed synaptic adhesion molecule, with no known role in bradykinin metabolism, bowel permeability or gut immunity. Genes that large accumulate intronic association hits by sheer target size, so the assignment is positional rather than mechanistic and should not be presented as explanation.","score_check":"log10p_adr 7.18 sits just under genome-wide significance in a scan over many drug-condition pairs, and beta_adr 1.821 (OR ~6) at ~1.6% allele frequency is the classic sparse-data / winner's-curse profile, so the effect size is very likely inflated even if the signal is real. Temporality of exactly 100% over 401 dated cases is uninformative rather than supportive - a chronic antihypertensive started years earlier will nearly always precede a first acute GI code - while the null disease and prescription arms and enrichment near 1.0 are the credible parts of the sheet.","candidability":5,"candidability_reason":"A genuinely labelled ADR with a described bradykinin mechanism that is known to masquerade as gastroenteritis, but the phenotype is dominated by infectious illness and the PTPRD variant is an unanchored intronic hit in a giant neuronal gene.","sources":[{"title":"Ramipril capsules - full prescribing information (DailyMed, Lupin Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a8cdec12-81d1-4ef4-82c0-d05d93c37a07"},{"title":"PubMed: ACE inhibitor visceral / intestinal angioedema case series and reports","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ACE+inhibitor+visceral+angioedema+intestinal"},{"title":"Europe PMC: ACE inhibitor-induced isolated intestinal angioedema mimicking functional bowel disorder; and FAERS disproportionality analysis of drug-induced intestinal angioedema","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=ACE%20inhibitor%20visceral%20angioedema&format=json&resultType=core&pageSize=5"},{"title":"UniProt P23468 - PTPRD, receptor-type tyrosine-protein phosphatase delta","url":"https://rest.uniprot.org/uniprotkb/P23468.txt"},{"title":"Ensembl variation record for rs72698987","url":"https://rest.ensembl.org/variation/human/rs72698987?content-type=application/json;pops=1"}]},"ramipril|B379":{"verdict":"statistical artifact","headline":"Not a ramipril effect: flat enrichment, null disease arm, two rare intronic SNVs with an implausible OR","assessment":"Candidiasis (ICD-10 B37.9) is not an indication for ramipril, whose US label covers hypertension, cardiovascular risk reduction in high-risk patients aged 55+, and heart failure after myocardial infarction; that label lists no fungal or candidal adverse reaction, and the only oral entry relevant at all is 'dry mouth' among gastrointestinal events. A published audit of oral adverse drug reactions in the 100 most-prescribed German drugs found xerostomia and dysgeusia to be the dominant oral ADRs, with the highest oral-side-effect scores in antibiotics, analgesics and antidepressants rather than ACE inhibitors, so even the xerostomia-to-thrush route is a weak and unquantified path for ramipril specifically. Reviews of oral and oesophageal candidiasis attribute it to local and systemic immunosuppression, corticosteroids, antibiotics, dentures and diabetes, none of which ramipril supplies; ACE inhibitors are not immunosuppressive and there is no drug-caused mechanism described in the literature. The atlas numbers agree with that reading rather than contradicting it: enrichment 0.95 means candidiasis is no more common in ramipril users than in users of other drugs (cotx 0.66%), so there is not even confounding by indication to explain, and log10p_prescribed 0.2 shows the variants do not predict being given the drug. Temporality of 87.2% on only 86 dated participants is exactly the upward-compressed pattern the cohort produces for any chronic drug started before most diagnosis records begin, and carries little weight on its own. The within-users signal (beta 5.887, log10p 7.62) sits against a completely null drug-free disease arm (beta -0.007 +/- 0.165), so z_diff 5.52 is arithmetic from one inflated coefficient rather than an independent contrast. The most economical explanation is sparse-data separation on two rare variants in a small case set, not a pharmacogenomic effect.","gene_comment":"rs148811762 is an intronic variant at chrX:10,077,376 inside WWC3, a KIBRA/WWC-family Hippo-pathway scaffold studied in cancer and synaptic biology with no role in mucosal antifungal immunity (no IL-17, CARD9 or CLEC7A connection); rs186243613 at chr7:10,083,001 falls in ENSG00000287409, an unnamed 'novel transcript' lncRNA, which is effectively no gene assignment at all. Both are rare (gnomAD MAF ~0.4% and ~0.12%), so neither locus supplies mechanism and neither should be presented as one.","score_check":"beta_adr 5.887 is a log-odds of roughly 360-fold, which for variants at 0.1-0.7% frequency in a case set of about 86 is near-complete separation rather than a real effect size. The disease arm is a clean, well-powered null, so the whole signal rests on the one unstable within-users coefficient.","candidability":1,"candidability_reason":"No labelled or literature-described drug-fungal mechanism, no enrichment in treated patients, a null disease arm, and two rare intronic variants in a lncRNA and a Hippo-pathway gene producing an implausibly large effect.","sources":[{"title":"Ramipril capsules - US prescribing information (DailyMed, Aurobindo Pharma)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=d6d57158-e8f9-4c91-8317-0374e0c87d33"},{"title":"Oral Side Effects of the Most Commonly Prescribed Drugs in Germany (Dent J, 2026)","url":"https://europepmc.org/article/MED/41744921"},{"title":"Pharmacological Management of Oral and Esophageal Candidiasis: A Clinical Pharmacotherapy Perspective (J Clin Med, 2025)","url":"https://europepmc.org/article/MED/41226935"},{"title":"PubMed search: ACE inhibitor and candidiasis (no drug-caused association reported)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ACE+inhibitor+candidiasis"},{"title":"Ensembl variation record rs148811762 (chrX:10,077,376, intronic WWC3, gnomAD MAF ~0.4%)","url":"https://rest.ensembl.org/variation/human/rs148811762?content-type=application/json;pops=1"},{"title":"Ensembl gene record ENSG00000287409 (chr7 lncRNA, 'novel transcript') and rs186243613 mapping","url":"https://rest.ensembl.org/lookup/id/ENSG00000287409?content-type=application/json"}]},"ramipril|C787":{"verdict":"co-prescription population","headline":"Liver metastases in elderly ramipril users: no enrichment, no mechanism, rare-variant signal only","assessment":"C78.7 is a staging code for hepatic metastasis from a primary tumour elsewhere, so it is not a licensed indication for ramipril and cannot plausibly be a de novo drug reaction; the US label lists hypertension, cardiovascular risk reduction and post-MI heart failure as indications and no neoplasm term anywhere in adverse reactions, with animal carcinogenicity studies negative at ~200x the maximum human dose. The atlas numbers agree with that reading: 0.56% of ramipril users carry the code and the enrichment against users of other drugs is 1.05, i.e. exactly the background cancer burden of an elderly hypertensive population, so there is no channelling and no excess. Temporality of 99.3% with a median of 4.91 years after first exposure is uninformative here, because the cohort compresses this upward and because incident metastatic cancer is expected to accrue in a population first prescribed an antihypertensive in later life. The class-level literature on ACE inhibitors and cancer is contested but does not point this way: an 11-study meta-analysis found a modest lung-cancer excess (OR 1.19, 95% CI 1.05-1.36) with I2 = 98%, while an updated overall-cancer meta-analysis found protective effects by hazard ratio and detrimental effects by relative risk and concluded the evidence does not warrant changing prescribing. For hepatic metastasis specifically the described direction is the opposite one: ACE inhibition with captopril reduces growth of colorectal liver metastases in murine models and in patient-derived organoids, and RAS-inhibitor use has been associated with lower colorectal cancer mortality (RR 0.80), which argues against an ADR reading rather than for it. What the atlas has found is a within-users variant association on a comorbidity that the drug's population carries anyway, with the drug-free arm flat (beta_disease 0.17 +/- 0.167, indistinguishable from noise) so the entire z_diff of 4.51 rests on the treated arm alone.","gene_comment":"TRERNA1 is an ~800 bp lncRNA at 20q13.13 (chr20:50,040,707-50,041,504, GRCh38) and rs116976563 is an intronic variant within it at chr20:50,041,241 with a minor allele frequency of about 0.6%; the lncRNA is genuinely metastasis-linked in the literature (SNAI1 enhancer driving EMT via EZH2/CDH1 silencing, poor prognosis and sorafenib resistance in HCC), which makes it superficially attractive. That biology is about metastatic behaviour in general, not about ACE inhibition, so a single rare intronic variant in a small lncRNA at sub-genome-wide significance is a weak assignment that must not be presented as a pharmacogenomic mechanism.","score_check":"log10p_adr of 6.03 is below the genome-wide threshold, and a beta of 3.256 for a ~0.6% MAF variant against a 0.56%-prevalent outcome with only 137 dated carriers-of-record is the classic shape of a sparse-data effect estimate rather than a real effect of that size. The disease arm is flat and the prescription arm is null, so the numbers are internally consistent with a chance within-users hit rather than with a drug-modified genetic effect.","candidability":1,"candidability_reason":"No enrichment, no label or mechanistic basis, class literature points the opposite way for liver metastases, and the signal is a rare intronic lncRNA variant with a sparse-data effect size below genome-wide significance.","sources":[{"title":"DailyMed - RAMIPRIL capsule (Lupin Pharmaceuticals) US prescribing information","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a8cdec12-81d1-4ef4-82c0-d05d93c37a07"},{"title":"Wu et al., Association between angiotensin-converting enzyme inhibitors and the risk of lung cancer: a systematic review and meta-analysis, Br J Cancer (PMID 36396817)","url":"https://europepmc.org/abstract/MED/36396817"},{"title":"Shin et al., ACE inhibitors or angiotensin receptor blockers and cancer risk: an updated meta-analysis of observational studies, Ther Adv Drug Saf 2022 (PMID 36249084)","url":"https://europepmc.org/abstract/MED/36249084"},{"title":"Wen et al., Captopril reduces growth of colorectal cancer liver metastases (PMID 23792575) and Vallejo Ardila et al., captopril impairs metastatic growth (PMID 32448803)","url":"https://europepmc.org/abstract/MED/23792575"},{"title":"Song et al., LncRNA TRERNA1 functions as an enhancer of SNAI1 and promotes metastasis via EMT, Mol Ther Nucleic Acids 2017 (PMID 28918030)","url":"https://europepmc.org/abstract/MED/28918030"},{"title":"Ensembl REST - TRERNA1 gene record and rs116976563 variant record (GRCh38)","url":"https://rest.ensembl.org/variation/human/rs116976563?content-type=application/json"}]},"ramipril|D509":{"verdict":"contested","headline":"ACE inhibitors do lower haemoglobin, but not by iron deficiency; the chr19 pseudogene variant adds nothing","assessment":"Iron deficiency anaemia is not a licensed indication for ramipril, whose SmPC covers hypertension, cardiovascular prevention, diabetic and non-diabetic nephropathy, symptomatic heart failure and post-MI secondary prevention. A haematological adverse effect is nonetheless label-recognised: the SmPC lists red blood cell count decreased and haemoglobin decreased (rare), plus bone marrow failure, pancytopenia and haemolytic anaemia (very rare), and advises monitoring blood counts. The direction is not ambiguous - ACE inhibitors and ARBs are the mainstay treatment for post-transplant erythrocytosis precisely because they lower haemoglobin and haematocrit, and RAAS inhibition has been shown to exacerbate anaemia in other settings, so the drug pushes this phenotype the way an ADR reading requires. The described mechanism, however, is hypoproliferative: ACE inhibition prevents degradation of AcSDKP, which accumulates and suppresses erythropoiesis, alongside blunted erythropoietin production - none of which produces iron deficiency, which requires blood loss or malabsorption. That matters because D50.9 in a cardiovascular cohort is dominated by the treated population itself: iron deficiency is reported in 27% of primary-care heart failure patients and 57-75% by broader definitions, and these are the patients co-prescribed aspirin and anticoagulants with their attendant gastrointestinal blood loss. The atlas is honest about this - enrichment is exactly 1.00, so ramipril users are no more likely to carry the code than users of other drugs, and the FAERS PRR of 2.75 is itself open to the same confounding-by-indication. Temporality of 98.9% over 351 dated participants with a 4.68-year median lag is consistent with drug-then-anaemia but, as the cohort compresses this measure upward, carries little independent weight.","gene_comment":"rs79143240 is a rare intronic variant (European MAF ~1%) in the long intergenic non-coding RNA LINC02987 at chr19:27.98 Mb, and the assigned ENSG00000266906 is a 300 bp transcribed zinc-finger processed pseudogene in the same pericentromeric gene desert. Neither has any described role in erythropoiesis, iron handling or the renin-angiotensin system, so this is a locus label, not a mechanism.","score_check":"log10p_adr 6.76 is p~1.7e-7, short of genome-wide significance, and a beta of 2.599 (OR ~13) on a 1% allele is the signature of sparse-data inflation rather than a real effect of that size. The disease arm is properly null (beta 0.154 vs se 0.098, well inside noise), which rules out simple disease predisposition but also means the z_diff of 4.82 is driven almost entirely by the inflated within-users estimate.","candidability":4,"candidability_reason":"Class-level ACE-inhibitor anaemia is real and label-listed, but the iron-deficiency phenotype does not match that mechanism, enrichment is flat, and the variant is a sub-genome-wide, sparse-data hit in a pseudogene/lncRNA desert.","sources":[{"title":"Ramipril 10mg Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/8067/smpc"},{"title":"PubMed: ACE inhibitors, anaemia, erythropoietin and haemoglobin","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ACE+inhibitor+anemia+erythropoietin+hemoglobin"},{"title":"PubMed: AcSDKP accumulation and suppressed erythropoiesis under ACE inhibition","url":"https://pubmed.ncbi.nlm.nih.gov/?term=angiotensin+converting+enzyme+inhibitor+anemia+AcSDKP+erythropoiesis"},{"title":"PubMed: ACE inhibitors as first-line treatment for post-transplant erythrocytosis","url":"https://pubmed.ncbi.nlm.nih.gov/?term=%22post-transplant+erythrocytosis%22+ACE+inhibitor+treatment"},{"title":"Europe PMC: prevalence and causes of iron deficiency and anaemia in heart failure","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22iron%20deficiency%22%20AND%20%22heart%20failure%22%20AND%20prevalence%20AND%20anaemia&format=json&pageSize=8&resultType=core"},{"title":"dbSNP rs79143240 (chr19, LINC02987 intron, EUR MAF ~1%)","url":"https://www.ncbi.nlm.nih.gov/snp/rs79143240"},{"title":"Ensembl REST lookup ENSG00000266906 (transcribed processed pseudogene, chr19)","url":"https://rest.ensembl.org/lookup/id/ENSG00000266906?content-type=application/json"}]},"ramipril|G439":{"verdict":"statistical artifact","headline":"ACE inhibitors prevent migraine in RCTs; two rare intronic variants with OR~13 point the wrong way","assessment":"Migraine is not a licensed indication for ramipril, whose approved uses are hypertension, cardiovascular risk reduction in patients 55+, and heart failure after myocardial infarction. Non-specific headache is a genuine and common label adverse reaction (5.4% in placebo-controlled hypertension trials versus dizziness 2.2%), and the FAERS disproportionality here (PRR 4.8) most likely reflects that headache/migraine coding rather than a distinct migraine syndrome. The decisive evidence runs the other way: renin-angiotensin blockade is an established migraine preventive, with lisinopril 20 mg reducing headache hours by 20% and migraine days by 21% versus placebo in a randomised crossover trial (Schrader, BMJ 2001), candesartan equalling propranolol in a triple-blind double-crossover study (Stovner, Cephalalgia 2014), and candesartan 16 mg beating placebo by ~1.2 migraine days in a 2025 Lancet Neurology phase 2 trial. Cross-phenotype genetics agrees that lowering diastolic blood pressure should lower, not raise, migraine risk (Guo, Nat Commun 2020). The atlas data are consistent with this: enrichment 0.66 means migraine is depleted among ramipril users relative to users of other drugs, exactly what an older, male-skewed cardiovascular population plus a possible protective class effect would produce, and cotx_pct is only 0.59%. The 73.2% temporality on 112 dated participants and median 3.12 years to first migraine record is the uninformative direction in this cohort and cannot rescue a drug-caused reading against three positive prophylaxis trials.","gene_comment":"rs72748106 is an intronic variant in DELEC1 (9q33.1), a candidate esophageal tumour-suppressor with no neurovascular role, at MAF ~0.001-0.006, and rs78407439 is intronic/downstream in CCDC192 (5q23.2, alias LINC01183, an uncharacterised coiled-coil ORF) at MAF ~0.011-0.033. Neither is a known migraine or blood-pressure locus and neither offers any mechanism; these are weak positional gene labels, not biology.","score_check":"A beta of 2.539 (OR ~13) at variant frequencies of 0.1-3%, with no disease-arm signal at all (beta_disease -0.051 +/- 0.067, |beta| well under 1.96*se) and no prescription-propensity signal, is the classic shape of a sparse-data rare-variant artifact rather than a real pharmacogenomic effect. The z_diff of 5.4 is driven entirely by the inflated within-user estimate, so it adds no independent support.","candidability":1,"candidability_reason":"Drug class is a proven migraine preventive, migraine is depleted in ramipril users, and the hits are rare intronic variants in unrelated genes with an implausibly large effect.","sources":[{"title":"Ramipril capsule label (DailyMed, Lupin Pharmaceuticals) - indications and adverse reactions","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a8cdec12-81d1-4ef4-82c0-d05d93c37a07"},{"title":"Schrader et al., Prophylactic treatment of migraine with ACE inhibitor (lisinopril): randomised, placebo controlled, crossover study, BMJ 2001 (PMID 11141144)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22angiotensin%20converting%20enzyme%20inhibitor%22%20AND%20TITLE%3Amigraine&format=json&pageSize=15&resultType=core"},{"title":"Oie et al., Candesartan versus placebo for migraine prevention: randomised phase 2 trial, Lancet Neurology 2025 (PMID 40975098); Stovner et al., Candesartan versus propranolol, Cephalalgia 2014 (PMID 24335848)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3Acandesartan%20AND%20TITLE%3Amigraine&format=json&pageSize=15&resultType=core"},{"title":"Guo et al., A genome-wide cross-phenotype meta-analysis of the association of blood pressure with migraine, Nat Commun 2020 (PMID 32632093)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID%3A32632093&format=json&resultType=core"},{"title":"NCBI dbSNP summary for rs72748106 (DELEC1, intronic) and rs78407439 (CCDC192, intronic/downstream) with allele frequencies","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=snp&id=72748106,78407439&retmode=json"},{"title":"NCBI Gene records for DELEC1 (9q33.1, deleted in esophageal cancer 1) and CCDC192 (5q23.2, LINC01183)","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=gene&id=728586&retmode=json"}]},"ramipril|H609":{"verdict":"statistical artifact","headline":"Otitis externa is not a ramipril effect; the signal is a rare-variant separation artifact at a novel lncRNA","assessment":"Ramipril is licensed for hypertension, cardiovascular prevention, diabetic and non-diabetic nephropathy, heart failure and post-MI secondary prevention; otitis externa (H60.9) is neither an indication nor a listed reaction. The only ear and labyrinth entries in the EMC SmPC are 'hearing impaired, tinnitus', and the US label mentions only hearing loss and tinnitus among rare post-marketing neurologic effects - no otitis, ear infection or ear pain appears in either label. Otitis externa is a swimming-, trauma- and dermatitis-driven canal infection with a Pseudomonas/S. aureus aetiology peaking at ages 7-14, and no systemic drug is described as a cause; the only drug-adjacent route would be immunosuppression or diabetes predisposing to necrotising otitis externa, neither of which ramipril confers. The atlas numbers agree that this is not a treated-population artefact either: cotx_pct is 0.57% with enrichment 0.83, i.e. otitis externa is slightly DEPLETED in ramipril users relative to users of other drugs, exactly as expected for an elderly hypertensive population versus a younger general drug-taking one, and FAERS shows no signal (PRR 1.08) with no BNF listing. Temporality of 71.6% on only 74 dated participants is uninformative here, since the cohort compresses this upward and otitis externa is an acute, recurrent, lifetime-common event that will often postdate the start of any chronic antihypertensive by chance (median 3.4 years). The genetics do not rescue it: beta_adr of 6.03 is a log-odds of roughly 400-fold for a variant with 0.3% minor allele frequency in a handful of cases, which is the signature of near-complete case-control separation rather than a real effect, and the same variant shows nothing on the condition itself in the drug-free population (beta_disease 0.068 +/- 0.215, well inside noise, log10p 0.12), so the z_diff of 5.07 is driven entirely by the unstable treated-arm estimate.","gene_comment":"ENSG00000287566 is a 'novel transcript' lncRNA at chr10:5.55 Mb, sitting near CALML3/CALML5/ASB13/NET1 but with no ear, skin-barrier or infection biology of its own; rs147431255 is a rare (MAF 0.0031) non-coding transcript exon variant there. The second variant, rs147786361, is not even at that locus - it maps to chrX:118.27 Mb as a regulatory-region variant - so the single gene label covers at most one of the two hits and cannot be presented as mechanism.","score_check":"The within-user p-value is real arithmetic but the effect size behind it is not believable: a log-odds near 6 for a 0.3%-frequency variant over ~74 datable cases is sparse-data separation, and the flat null effect of the same variant on otitis externa in the drug-free population confirms nothing genuine is being detected. Only the epidemiological fields (enrichment 0.83, FAERS PRR 1.08) hang together with the literature, and they point away from an ADR.","candidability":1,"candidability_reason":"No label or literature basis, no plausible mechanism, condition depleted in ramipril users, and the genetic effect is an implausibly large rare-variant estimate with a null disease-arm counterpart.","sources":[{"title":"Ramipril capsules - full prescribing information (DailyMed, Aurobindo Pharma)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=d6d57158-e8f9-4c91-8317-0374e0c87d33"},{"title":"Ramipril 1.25mg Tablets SmPC (electronic Medicines Compendium, Zentiva)","url":"https://www.medicines.org.uk/emc/product/8064/smpc"},{"title":"Otitis Externa - StatPearls, NCBI Bookshelf","url":"https://www.ncbi.nlm.nih.gov/books/NBK556055/"},{"title":"Ensembl REST lookup: ENSG00000287566 (lncRNA, novel transcript, chr10)","url":"https://rest.ensembl.org/lookup/id/ENSG00000287566?content-type=application/json"},{"title":"Ensembl REST variation: rs147431255 and rs147786361","url":"https://rest.ensembl.org/variation/human/rs147431255?content-type=application/json"}]},"ramipril|H931":{"verdict":"plausible ADR","headline":"Tinnitus is on the ramipril label, but this rare-variant signal looks like a sparse-data artifact","assessment":"Tinnitus is not an indication for ramipril (hypertension, cardiovascular risk reduction, post-MI heart failure) but it is a listed adverse reaction: the US label names tinnitus and hearing loss under 'Other Adverse Reactions' among neurologic/psychiatric events, without a frequency, and vertigo was more common on ramipril than placebo in AIRE (2% vs 0.7%). Class-level support exists but is weak and confounded: Figueiredo et al. found ACE inhibitor use in 16.0% of tinnitus patients vs 5.7% of controls (p=0.006), alongside similar excesses for thiazides, potassium-sparing diuretics and calcium-channel blockers, and the authors themselves note multidrug therapy prevents a causal reading. The competing explanation is strong, because hypertension itself raises tinnitus odds (OR 1.34, 95% CI 1.10-1.62; still OR 1.27 when controlled, and OR 2.08 for grade II), so ACE-inhibitor users are enriched for tinnitus by virtue of being hypertensive. The atlas's own indication arm is consistent with that ambiguity but does not resolve it: enrichment is 0.98, i.e. tinnitus is no more common in ramipril users than in users of other drugs (itself a treated, comorbid comparison group), and log10p_prescribed of 0.1 rules out a prescription-propensity artifact. No drug-caused mechanism is described beyond generic speculation about ototoxicity of antihypertensives; the plausible physiological route is blood-pressure or bradykinin-mediated cochlear/perfusion change, and no direction-reversing evidence exists, though ARBs have been reported as protective for hearing in one vestibular-schwannoma series, which cuts against a simple RAS-blockade-causes-tinnitus story. Temporality of 88.9% over only 99 dated participants is the weak direction of that metric and adds little. The clinical pair is a genuine, if unquantified, labelled adverse reaction; the specific genetic claim here is what does not survive scrutiny.","gene_comment":"rs148133930 is a rare intronic variant in AGTPBP1 (chr9:85,557,777; global MAF 0.22%) with no established auditory role, and rs191504316 (chrX:13,687,533) is intergenic with no reported allele frequency, sitting ~1.6 kb upstream of RAB9A in a gene-dense Xp22.2 window that also contains TCEANC, TRAPPC2 and OFD1, so 'RAB9A' is a nearest-gene label rather than a mechanism. Neither gene has a credible connection to cochlear or auditory-pathway biology, and an X-chromosome intergenic hit in a small case set carries extra technical risk.","score_check":"beta_adr of 10.46 on the log-odds scale implies an odds ratio of roughly 3x10^4 from variants at 0.2% frequency or rarer in a set of about 99 dated cases, which is the signature of quasi-complete separation rather than a real effect, so the log10p of 11.26 and z_diff of 6.87 should not be taken at face value. The remaining numbers are internally coherent and are the informative ones: the disease arm is flat (beta -0.08 +/- 0.228, log10p 0.14), the prescription arm is null, and enrichment of 0.98 shows no excess of tinnitus in ramipril users at all.","candidability":3,"candidability_reason":"Real but unquantified labelled ADR whose atlas genetics are a sparse-data artifact on two rare variants, one intergenic, with no enrichment and a hypertension-confounded literature.","sources":[{"title":"Ramipril capsule label (DailyMed, Aurobindo Pharma) - Indications and Adverse Reactions","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=d6d57158-e8f9-4c91-8317-0374e0c87d33"},{"title":"Figueiredo RR et al., Positive Association between Tinnitus and Arterial Hypertension, Front Neurol 2016","url":"https://www.frontiersin.org/articles/10.3389/fneur.2016.00171/full"},{"title":"Association between hypertension status and severity and tinnitus, BMJ Open (PMC13250232)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13250232/"},{"title":"Ensembl variation record for rs148133930 (AGTPBP1 intron, MAF 0.0022)","url":"https://rest.ensembl.org/variation/human/rs148133930?content-type=application/json"},{"title":"Ensembl variation record for rs191504316 (intergenic, chrX:13687533)","url":"https://rest.ensembl.org/variation/human/rs191504316?content-type=application/json"},{"title":"Ensembl genes overlapping chrX:13,637,533-13,737,533 (TCEANC, RAB9A, TRAPPC2, OFD1)","url":"https://rest.ensembl.org/overlap/region/human/X:13637533-13737533?feature=gene;content-type=application/json"},{"title":"Europe PMC search: tinnitus AND ACE inhibitor","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=tinnitus%20AND%20%22ACE%20inhibitor%22&format=json&pageSize=15&resultType=core"}]},"ramipril|I219":{"verdict":"licensed indication","headline":"Ramipril is licensed to prevent MI and treat post-MI patients; a rare KLF12 intron hit does not undo that","assessment":"Acute myocardial infarction is named twice in ramipril's own SmPC section 4.1: as cardiovascular prevention (reduction of cardiovascular morbidity and mortality in patients with manifest atherothrombotic disease or diabetes with a risk factor) and as secondary prevention after acute myocardial infarction in patients with clinical signs of heart failure started >48 h after the event, so I219 is an indication context and an inclusion criterion for the drug, not an outcome the drug is suspected of producing. The randomised evidence points the opposite way to an ADR reading: in HOPE (NEJM 2000, 9297 high-risk patients) ramipril cut myocardial infarction from 12.3% to 9.9%, RR 0.80, p<0.001, and the AIRE programme established a mortality reduction in post-MI patients with heart failure - a drug that lowers MI incidence in trials cannot be credibly read as causing MI from an observational within-users GWAS. The atlas numbers are exactly what confounding by indication looks like: cotx_pct 1.52% with enrichment only 1.21 (ramipril users are modestly more likely to carry an MI code, as expected for a high-CV-risk treated population), temporality 89% at a median 2.95 years, which is uninformative here because ramipril is typically started for hypertension or vascular risk years before an index MI. The label does carry 'myocardial ischaemia including angina pectoris or myocardial infarction' as a rare adverse reaction, but that is understood as an acute hypotension-mediated event in critically stenosed vascular beds, on a timescale of hours to days after dosing or up-titration, not a signal peaking at three years. The FAERS support is weak (1/3, PRR 1.92), and for an ACE inhibitor prescribed overwhelmingly to patients with or at risk of coronary disease a PRR near 2 is fully explained by channelling. The one thing the atlas gets right is that the variant is not a general MI risk allele and not a prescription-propensity artifact (log10p_prescribed 0.12), which leaves a treated-population interaction claim resting entirely on a single low-frequency SNV.","gene_comment":"rs138042714 is a low-frequency C>A variant (gnomAD genomes MAF ~1%, ~1.6-2% in European reference panels) sitting in an intron / genic-upstream region of KLF12 at chr13:74,048,448, with no entry in the GWAS Catalog and no reported cardiovascular association; Open Targets ranks KLF12's strongest cardiovascular link as atrial fibrillation, not coronary artery disease or myocardial infarction. KLF12 is a transcriptional repressor of AP-2-alpha with no described role in atherothrombosis or in the renin-angiotensin axis, so this is a positional gene label, not a mechanism.","score_check":"beta_adr 2.128 implies an odds ratio near 8.4 for a ~1% intronic variant on myocardial infarction, which is far outside anything the CAD/MI genetic architecture supports and is the classic shape of a sparse-data artifact from few minor-allele carriers among 281 dated cases; log10p_adr 6.93 does not even clear genome-wide significance. The disease arm is honestly null (beta 0.189 +/- 0.124, |beta| < 1.96*se), so the z_diff of 4.61 is driven entirely by the unstable within-users estimate rather than by a real difference between treated and untreated effects.","candidability":1,"candidability_reason":"Ramipril demonstrably lowers MI risk in randomised trials and is licensed for MI prevention and post-MI care, so the direction is backwards; the supporting variant is an implausibly large effect at a rare intronic KLF12 site with no cardiovascular genetic support.","sources":[{"title":"Tritace 10mg Tablets - Summary of Product Characteristics (electronic medicines compendium, Sanofi)","url":"https://www.medicines.org.uk/emc/product/2757/smpc"},{"title":"Yusuf S et al. Effects of an angiotensin-converting-enzyme inhibitor, ramipril, on cardiovascular events in high-risk patients (HOPE). N Engl J Med 2000","url":"https://pubmed.ncbi.nlm.nih.gov/10639539/"},{"title":"Anderson AN et al. A South African pharmaco-economic analysis of the Acute Infarction Ramipril Efficacy (AIRE) Study. Cardiovasc J S Afr 2000","url":"https://pubmed.ncbi.nlm.nih.gov/11447469/"},{"title":"dbSNP rs138042714 (NCBI) - chr13:74,048,448 C>A, KLF12 intron/genic-upstream, gnomAD MAF ~0.01","url":"https://www.ncbi.nlm.nih.gov/snp/rs138042714"},{"title":"Open Targets Platform - KLF12 (ENSG00000118922) disease associations","url":"https://platform.opentargets.org/target/ENSG00000118922"},{"title":"GWAS Catalog REST API - rs138042714 (no catalogued associations, HTTP 404)","url":"https://www.ebi.ac.uk/gwas/rest/api/singleNucleotidePolymorphisms/rs138042714"}]},"ramipril|I64":{"verdict":"licensed indication","headline":"Ramipril is licensed to REDUCE stroke (HOPE RR 0.68); an FDX1 burden beta of 57.9 is a separation artifact","assessment":"Stroke is not an adverse reaction to ramipril: the US label carries an explicit indication in patients 55 years or older at high cardiovascular risk 'to reduce the risk of myocardial infarction, stroke, or death from cardiovascular causes', and the HOPE trial that supports it recorded stroke in 3.4% on ramipril versus 4.9% on placebo (RR 0.68, p<0.001). Recent comparative outcome work continues to report that ACE inhibitors as a class significantly reduce stroke and MACE, so the drug moves this endpoint in the opposite direction to the one an ADR reading requires. The atlas numbers agree with that literature rather than contradicting it: stroke is present in only 0.81% of ramipril users and is actually depleted relative to users of other drugs (enrichment 0.84), and there is no prescribing-propensity signal (log10p_prescribed 0.5). What remains is a within-users FDX1 burden hit whose effect size, beta 57.889, is not a biological effect but complete or near-complete separation in a rare-variant test, and the same burden is indistinguishable from noise in the drug-free population (beta_disease 7.932 with se 4.335, |beta| < 1.96*se), which also makes the z_diff of 4.14 an artefact of the numerator rather than a real treated-versus-untreated contrast. The 95.7% 'after exposure' figure rests on 47 datable participants with 54.4% undated and, given the cohort's known upward compression of temporality, carries essentially no weight; the FAERS support is likewise weak (1/3 sources, PRR 1.67). The only mechanistic route worth naming is non-specific - symptomatic hypotension, postural hypotension and syncope are labelled reactions and profound hypotension can in principle precipitate watershed ischaemia - but nothing in the trial or observational record turns that into a net excess of stroke.","gene_comment":"FDX1 encodes mitochondrial adrenodoxin, a [2Fe-2S] electron donor for CYP11A1 steroidogenesis expressed mainly in adrenal, kidney and testis, with no known role in ACE-inhibitor pharmacology or in monogenic or GWAS-established cerebrovascular disease. Its only stroke-adjacent literature is rodent and single-cell work placing FDX1 in cuproptosis pathways after cerebral ischaemia - an expression correlate downstream of infarction, not a germline risk mechanism, and it should not be presented as one.","score_check":"The numbers are internally inconsistent with an ADR: a log-odds of 57.9 from an MPC burden test is a sparse-data separation artefact, not an effect, and the disease-arm estimate (7.932 +/- 4.335) fails its own significance threshold. Enrichment below 1 and a flat prescribed-propensity signal are the only figures here that are believable, and both point away from a drug-caused stroke excess.","candidability":0,"candidability_reason":"Backwards direction against a licensed stroke-reduction indication, stroke depleted in treated patients, and an impossible burden effect size.","sources":[{"title":"Ramipril capsules - full prescribing information (DailyMed, Lupin Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a8cdec12-81d1-4ef4-82c0-d05d93c37a07"},{"title":"HOPE Study Investigators. Effects of an angiotensin-converting-enzyme inhibitor, ramipril, on cardiovascular events in high-risk patients. N Engl J Med 2000 (PMID 10639539)","url":"https://pubmed.ncbi.nlm.nih.gov/10639539/"},{"title":"Europe PMC search: ramipril AND stroke AND prevention (incl. ACE-I vs ARB cardiovascular outcomes, Eur Heart J Open, PMID 42524061)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=ramipril%20AND%20stroke%20AND%20prevention&format=json&pageSize=10&resultType=core"},{"title":"UniProt P10109 - Adrenodoxin, mitochondrial (FDX1)","url":"https://rest.uniprot.org/uniprotkb/P10109.txt"},{"title":"Europe PMC search: FDX1 AND (stroke OR cerebrovascular OR GWAS)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=FDX1%20AND%20(stroke%20OR%20cerebrovascular%20OR%20GWAS)&format=json&pageSize=10&resultType=core"}]},"ramipril|I802":{"verdict":"statistical artifact","headline":"Not a labelled ADR; ACE inhibition is if anything antithrombotic, and the hit is a rare pseudogene variant","assessment":"I80.2 (phlebitis/thrombophlebitis of deep lower-limb vessels) is not a licensed indication for ramipril, whose SmPC indications are hypertension, cardiovascular prevention in atherothrombotic disease or diabetes, diabetic and non-diabetic nephropathy, symptomatic heart failure and post-MI secondary prevention. It is also not a recognised adverse reaction: the Tritace SmPC section 4.8 vascular-disorder entries are hypotension, orthostatic hypotension, syncope, flushing, vascular stenosis, hypoperfusion, vasculitis and Raynaud's phenomenon, and the US ramipril label lists no phlebitis, thrombophlebitis or thrombosis anywhere. The mechanistic direction runs the wrong way for an ADR reading: angiotensin II raises PAI-1 and platelet adhesion factors, and in an atherosclerotic cohort RAS-inhibitor users had a lower VTE incidence (8.4% vs 12.5%; adjusted HR 0.60, 95% CI 0.40-0.90) - blocking the axis is associated with less venous thrombosis, not more. The atlas numbers fit plain comorbidity rather than causation: only 0.44% of ramipril users carry the code and enrichment is 1.09, i.e. a 9% excess over other drugs' users, which is what an older, hypertensive, obese, less mobile treated population predicts on its own. Temporality of 93.2% over 117 dated participants with a median 3.0 years to onset is expected for any chronic midlife drug preceding an age-accruing venous event, and is the weak direction of this metric by the atlas's own convention. FAERS support is weak (1 of 3, PRR 1.97) and the pair is not on the BNF interaction/ADR list.","gene_comment":"RPL32P14 is a ribosomal protein L32 pseudogene with no publications at all in Europe PMC, so the gene label carries no mechanism; rs549725833 (chr5:19,038,101, T/C, intronic, gnomAD MAF ~0.005, absent from the GWAS Catalog) is a rare variant whose assignment to that pseudogene is a proximity call, not biology.","score_check":"A beta of 4.06 (OR near 60) for a variant at 0.5% frequency in a condition affecting ~117 datable users is the classic sparse-data blow-up, and the disease arm is flatly null (beta 0.069, se 0.17, well inside 1.96*se), so z_diff 5.15 is driven by the inflated treated-arm estimate rather than by a real drug-by-genotype interaction. The prescription arm is also null (log10p 0.30), which removes the only alternative real signal.","candidability":1,"candidability_reason":"Not a labelled or literature-supported ADR, mechanism and observational data point the opposite way, enrichment is trivial, and the genetic hit is an implausibly large rare-variant effect on a pseudogene.","sources":[{"title":"Tritace 10mg Tablets - Summary of Product Characteristics (electronic medicines compendium)","url":"https://www.medicines.org.uk/emc/product/2757/smpc"},{"title":"RAMIPRIL capsule - US prescribing information (DailyMed, Aurobindo Pharma)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=d6d57158-e8f9-4c91-8317-0374e0c87d33"},{"title":"Chae YK et al. Inhibition of renin angiotensin axis may be associated with reduced risk of developing venous thromboembolism in patients with atherosclerotic disease. PLoS One 2014","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC3909246/"},{"title":"PubMed search: ACE inhibitor and venous thromboembolism risk","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ACE+inhibitor+venous+thromboembolism+risk"},{"title":"Ensembl variation record for rs549725833 (location, consequence, population frequencies)","url":"https://rest.ensembl.org/variation/human/rs549725833?content-type=application/json;pops=1"},{"title":"GWAS Catalog variant page for rs549725833 (no reported associations)","url":"https://www.ebi.ac.uk/gwas/variants/rs549725833"}]},"ramipril|J90":{"verdict":"co-prescription population","headline":"Pleural effusion is a marker of the heart failure ramipril treats, not a labelled or described ACEi reaction","assessment":"Pleural effusion (J90) is not a licensed indication for ramipril and is not listed anywhere in the ramipril label's adverse-reaction section; the only respiratory terms there are persistent non-productive cough and rare eosinophilic pneumonitis, with no pleural, pleuritic or serositis term. Ramipril is licensed for hypertension, high cardiovascular risk and post-MI heart failure, and congestive heart failure is the single commonest cause of (transudative) pleural effusion, so effusion in this cohort is overwhelmingly the treated disease declaring itself rather than a drug effect. The direction of the drug's known effect argues against an ADR reading: ACE inhibitors reduce congestion and heart-failure morbidity and mortality, so if anything ramipril should lower the incidence of cardiogenic effusion in exposed patients. A EuropePMC sweep of drug-induced pleural disease returns pioglitazone, dasatinib, nilotinib, bosutinib, brigatinib, sunitinib, etoricoxib, valproate, warfarin and statins but no ACE inhibitor, so no mechanism for ramipril-induced pleural effusion is described in the literature. The atlas numbers are internally consistent with the confounding reading: enrichment 1.04 says these patients are no more likely to carry the code than users of other drugs, temporality 98.9% is the uninformative high end of a metric this cohort compresses upward, and the FAERS PRR of 2.79 is itself vulnerable to the same cardiac indication channel. What remains is a single rare intronic SNV with an enormous within-user beta and a flat disease-arm effect, which is a sparse-data pattern rather than a pharmacogenomic one.","gene_comment":"ENSG00000293906 is an unnamed novel lncRNA on chr13 annotated only as 'novel transcript, antisense to SLC46A3', and rs143383919 is an intronic variant in it with a minor allele frequency of about 0.4%. That is a weak gene assignment carrying no mechanistic content for pleural physiology, and it should not be presented as one.","score_check":"beta_adr 3.348 (OR roughly 28) from a variant at 0.4% frequency, against a completely null disease arm (beta 0.084, se 0.144), is the classic signature of sparse-data separation rather than a real effect size, and the z_diff of 5.06 inherits that inflation. The p-value is not implausible for the sample, but the effect magnitude is not believable at face value.","candidability":2,"candidability_reason":"Condition is a complication of the indication, the drug moves it the other way, no described mechanism, and the variant is a rare-allele sparse-data artifact.","sources":[{"title":"Ramipril capsules label (DailyMed, Aurobindo Pharma) - indications and adverse reactions","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=d6d57158-e8f9-4c91-8317-0374e0c87d33"},{"title":"EuropePMC search: heart failure as the most common cause of pleural effusion (incl. StatPearls, PMID 29630281; transudate cohort PMID 42514262)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22pleural%20effusion%22%20AND%20%22most%20common%20cause%22%20AND%20%22heart%20failure%22&format=json&pageSize=15&resultType=core"},{"title":"EuropePMC search: drug-induced pleural disease / pleural effusion - implicated drugs (no ACE inhibitor)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22drug-induced%20pleural%20disease%22%20OR%20%22drug-induced%20pleural%20effusion%22&format=json&pageSize=20&resultType=core"},{"title":"EuropePMC search: captopril and pleural effusion (no ACEi-induced pleuritis literature returned)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22captopril%22%20AND%20%22pleural%20effusion%22&format=json&pageSize=25&resultType=core"},{"title":"Ensembl REST: gene ENSG00000293906 (lncRNA, novel transcript antisense to SLC46A3, chr13)","url":"https://rest.ensembl.org/lookup/id/ENSG00000293906?content-type=application/json"},{"title":"Ensembl REST: variant rs143383919 (chr13:28735638, intron variant, MAF ~0.004)","url":"https://rest.ensembl.org/variation/human/rs143383919?content-type=application/json"}]},"ramipril|K20":{"verdict":"co-prescription population","headline":"Not a labelled ramipril reaction; hypertension itself raises oesophagitis risk, and the locus is a gene desert","assessment":"Oesophagitis (K20) is not a licensed indication for ramipril, whose approved uses on the FDA label are hypertension and post-MI heart failure, and it is not a labelled adverse reaction either: the label's oesophageal/upper-GI terms are dyspepsia and dysphagia, with abdominal pain arising from intestinal angioedema, none of which amounts to oesophagitis. The classic culprits in the drug-induced-oesophagitis literature are doxycycline, bisphosphonates, iron, potassium and oral contraceptives via direct caustic mucosal contact; ACE inhibitors do not appear in those case series, and ramipril's small once-daily capsule is a poor fit for that mechanism. I found no published data on angiotensin II or bradykinin altering lower-oesophageal-sphincter tone or oesophageal motility, so a pharmacological mechanism here is speculative rather than described, though nothing shows ramipril moves oesophagitis in the opposite direction. The competing explanation is well documented: essential hypertension is an independent risk factor for erosive oesophagitis (adjusted OR 1.21, 95% CI 1.04-1.42) and metabolic syndrome tracks with GERD, which is close to the 1.09 enrichment the atlas actually observes in ramipril users. The FAERS PRR of 3.39 is the one external prop, but spontaneous reporting for a drug this heavily prescribed in an older, reflux-prone, polypharmacy population is exactly where notoriety and comorbidity confounding accumulate. The 99.7% temporality on 306 dated participants is uninformative given the cohort's known upward compression of that statistic.","gene_comment":"rs182355394 is an intergenic X-chromosome variant at X:89,108,942 (Xq21.31) with no mapped gene and no gnomAD frequency in Ensembl; the surrounding megabase holds only processed pseudogenes and novel lncRNAs, with the nearest protein-coding gene, the testis-specific CPXCR1, about 354 kb away. There is no gene assignment worth interpreting mechanistically here.","score_check":"The within-user signal is internally consistent (log10p_adr 8.33 against log10p_disease 1.79, a disease effect of 0.14 +/- 0.058 that is only marginal on its own, z_diff 5.13, and a null prescription-propensity arm at log10p 0.12), so this is not simply the background disease variant resurfacing. But a beta of 1.411 for a rare X-linked variant with no reported allele frequency, over roughly 300 dated cases, is the regime where sparse-data bias inflates effect sizes, and no ADR standard error is given to check it.","candidability":3,"candidability_reason":"Oesophagitis is unlabelled and mechanistically undescribed for ACE inhibitors, the modest 1.09 enrichment is matched by hypertension's own effect on erosive oesophagitis, and the variant sits in an X-chromosome gene desert with nothing to interpret.","sources":[{"title":"openFDA drug label: ramipril (indications and adverse reactions)","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22ramipril%22&limit=1"},{"title":"Ensembl variation record for rs182355394","url":"https://rest.ensembl.org/variation/human/rs182355394?content-type=application/json"},{"title":"Ensembl genes overlapping X:88.6-89.6 Mb (locus context for rs182355394)","url":"https://rest.ensembl.org/overlap/region/human/X:88600000-89600000?feature=gene;content-type=application/json"},{"title":"Europe PMC search: pill-induced / medication-induced esophagitis (culprit drug classes)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22pill-induced%20esophagitis%22%20OR%20%22medication-induced%20esophagitis%22&format=json&pageSize=25&resultType=core"},{"title":"Essential Hypertension as an Independent Risk Factor for Erosive Esophagitis (J Inflamm Res 2025, PMID 40224391)","url":"https://europepmc.org/article/MED/40224391"},{"title":"Metabolic syndrome is associated with gastroesophageal reflux disease in a large Taiwanese population study (PMID 40093807)","url":"https://europepmc.org/article/MED/40093807"}]},"ramipril|K589":{"verdict":"plausible ADR","headline":"Ramipril has labelled GI and small-bowel-angioedema effects, but the lncRNA variant here is uninterpretable","assessment":"IBS is not a licensed indication for ramipril: the Tritace SmPC lists hypertension, cardiovascular prevention, diabetic and non-diabetic nephropathy, symptomatic heart failure and post-MI secondary prevention, and the US label adds nothing gastrointestinal. Ramipril does, however, carry labelled GI reactions - 'gastrointestinal inflammation, digestive disturbances, abdominal discomfort, dyspepsia, diarrhoea, nausea, vomiting' as common, and 'small bowel angioedema', upper abdominal pain and constipation as uncommon - with section 4.4 warning explicitly that intestinal angioedema has been reported on Tritace. That mechanism is bradykinin-mediated and well described: ACE inhibition blocks bradykinin degradation, B2-receptor-driven submucosal oedema follows with a latency running from weeks to years, estimated incidence 0.1-0.7%, and the case literature repeatedly calls it an elusive, missed diagnosis whose non-specific recurrent abdominal pain gets attributed to functional bowel disease - a credible route by which a ramipril user acquires a K58.9 code. Two atlas numbers support a non-indication reading: enrichment 0.81 means IBS is if anything less common in ramipril users than in users of other drugs, so this is not confounding by indication, and log10p_prescribed 0.64 shows no prescription-channel signal. Against it, the specific code is 'IBS without diarrhoea' whereas ramipril's common labelled GI effect and the classic visceral-angioedema presentation are both diarrhoea-predominant, and temporality of 87.1% over 186 dated participants is the uninformative direction in this cohort even though the 3.47-year median sits inside the documented weeks-to-years latency. The clinical class effect is therefore real and labelled; nothing in this particular variant supports it.","gene_comment":"ENSG00000301403 is an unnamed novel lncRNA on chr14q24 (GRCh38 14:74,378,107-74,398,858) with no GWAS Catalog entry and no known gut biology, and rs151129650 is an intronic variant within it at MAF ~0.2-0.35%. This is a nameless non-coding locus, not a mechanism, and should not be presented as one.","score_check":"beta_adr 3.845 (OR ~47) for a common functional bowel diagnosis carried by a variant of ~0.3% frequency is a sparse-data artifact, and z_diff 5.06 inherits that inflation; the drug-free arm is the more honest number, beta_disease 0.328 +/- 0.126 (z ~2.6), a modest borderline predisposition rather than a clean drug-specific effect. The FAERS PRR of 24.82 is on a rarely reported, chronic, non-serious term for an extremely widely prescribed antihypertensive, so it is weaker evidence than its magnitude suggests.","candidability":5,"candidability_reason":"Genuine labelled GI/intestinal-angioedema class effect with no indication confounding (enrichment 0.81), but the genetics is a rare intronic variant in an unnamed lncRNA with an implausible effect size.","sources":[{"title":"Tritace 10mg Tablets - Summary of Product Characteristics (electronic medicines compendium)","url":"https://www.medicines.org.uk/emc/product/2757/smpc"},{"title":"Ramipril capsules - US prescribing information (DailyMed, Lupin Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=a8cdec12-81d1-4ef4-82c0-d05d93c37a07"},{"title":"ACE inhibitor induced visceral angioedema: an elusive diagnosis (BMJ Case Reports 2020, PMC7684650)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC7684650/"},{"title":"PubMed: ACE inhibitor visceral angioedema abdominal - result set including BMJ Case Rep 2026/2019, J Emerg Med 2018, Clin Radiol 2006","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=pubmed&term=angiotensin+converting+enzyme+inhibitor+visceral+angioedema+abdominal&retmax=20&retmode=json"},{"title":"Ensembl REST: ENSG00000301403 gene record (lncRNA, novel transcript, chromosome 14)","url":"https://rest.ensembl.org/lookup/id/ENSG00000301403?content-type=application/json"},{"title":"Ensembl REST: rs151129650 variant record (intron variant, MAF 0.00216)","url":"https://rest.ensembl.org/variation/human/rs151129650?content-type=application/json&pops=1"},{"title":"GWAS Catalog search for ENSG00000301403 (no results)","url":"https://www.ebi.ac.uk/gwas/search?query=ENSG00000301403"},{"title":"Colonic hyperalgesia triggered by proteinase-activated receptor-2 in mice: involvement of endogenous bradykinin (Neurosci Lett 2006, PMID 16644120)","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=pubmed&id=16644120&retmode=json"}]},"ramipril|L409":{"verdict":"plausible ADR","headline":"ACE-inhibitor-induced psoriasis is a real class effect with a ramipril FAERS signal; the GLP1R variant is not","assessment":"Psoriasis is not an indication for ramipril, whose licensed uses are hypertension, cardiovascular risk reduction in patients 55+, and post-MI heart failure, and the US label's dermatologic postmarketing list (urticaria, pruritus, rash, photosensitivity, purpura, onycholysis, pemphigus, pemphigoid, erythema multiforme, TEN, Stevens-Johnson) does not name psoriasis. The epidemiology nonetheless supports induction or exacerbation: a systematic review and meta-analysis of 6.4 million individuals put ACE inhibitors at OR 1.67 (1.31-2.13), the highest comparative risk in the network analysis at OR 2.09 (1.39-3.18), and a FAERS disproportionality analysis found a ramipril-specific signal (ROR 1.63, 1.36-1.96), close to the atlas's own FAERS PRR of 1.91. Causality is supported by dechallenge/rechallenge case material rather than by an established mechanism: a 2024 Cureus report describes widespread plaques within two weeks of starting an ACE inhibitor in previously controlled psoriasis, improving on withdrawal and not recurring on an alternative antihypertensive, while the cited authors state the underlying mechanism still needs elucidating; nothing in the literature suggests ramipril improves psoriasis, so there is no backwards-direction objection. Confounding by indication is the main competing explanation, since psoriasis itself carries excess hypertension and cardiometabolic disease, and the atlas's mild enrichment of 1.21 with 1.02% co-occurrence is exactly what that comorbidity alone would produce. Temporality of 78.9% over 133 dated cases with a 3.78-year median lag is compatible with drug-induced onset but, as the cohort compresses this measure upward, adds little on its own. The comparison arms are clean in the right way: the same variant is null for psoriasis in the drug-free population (beta 0.18 +/- 0.167, well inside noise) and null for being prescribed ramipril (log10p 0.23), so the within-user signal is not a disease-predisposition or prescribing artifact, but that leaves a within-treated interaction resting on a single rare variant.","gene_comment":"rs112794418 is a rare non-coding A>G SNV at chr6:39,093,427 (GRCh38), ~2 kb downstream of the GLP1R 3' end (GLP1R spans 39,048,780-39,091,302), with dbSNP assigning it no gene or functional class and gnomAD/TOPMED frequencies of only ~0.5%; it has no GWAS Catalog entries. GLP1R has no established role in psoriasis - the GLP-1 receptor agonist literature is mixed and mostly about obesity-driven IL-17/IL-23 inflammation, with one 2026 cohort even reporting higher psoriasis risk on GLP-1RA than DPP-4i - so the gene label here is nearest-neighbour annotation, not mechanism.","score_check":"beta_adr 4.301 (OR ~70) for a 0.5%-frequency variant across roughly 133 psoriasis cases means only a handful of carriers drive log10p 8.33, the classic shape of a sparse-data effect estimate, and no standard error is reported to check it. The disease-arm and prescription-arm nulls are believable and appropriately quiet, but the z_diff of 5.47 inherits the same fragile numerator and needs replication before it counts as an interaction.","candidability":6,"candidability_reason":"Genuine, literature-backed ACE-inhibitor class ADR with a ramipril-specific pharmacovigilance signal, but the genetics here are a rare non-coding variant with an implausibly large effect and no GLP1R-psoriasis rationale.","sources":[{"title":"Ramipril capsules - US prescribing information (DailyMed, Lupin)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a8cdec12-81d1-4ef4-82c0-d05d93c37a07"},{"title":"Antihypertensive drug use and psoriasis: a systematic review, meta- and network meta-analysis (Br J Clin Pharmacol 2022, PMID 34611920)","url":"https://pubmed.ncbi.nlm.nih.gov/34611920/"},{"title":"Psoriasis associated with ACE inhibitors: an analysis of the FAERS database (Pharmazie 2020, PMID 33305730)","url":"https://pubmed.ncbi.nlm.nih.gov/33305730/"},{"title":"Angiotensin-Converting Enzyme Inhibitor-Induced Psoriasis: A Case Report on Plaque Exacerbation (Cureus 2024, PMID 39364528)","url":"https://pubmed.ncbi.nlm.nih.gov/39364528/"},{"title":"The Causal Association Between Medication Intake and Increased Risk of Psoriasis (Dermatol Pract Concept 2024, PMID 38364394)","url":"https://pubmed.ncbi.nlm.nih.gov/38364394/"},{"title":"dbSNP record for rs112794418 (chr6:39,093,427, MAF ~0.005)","url":"https://www.ncbi.nlm.nih.gov/snp/rs112794418"},{"title":"NCBI Gene: GLP1R, glucagon like peptide 1 receptor (chr6:39,048,780-39,091,302)","url":"https://www.ncbi.nlm.nih.gov/gene/2740"}]},"ramipril|M109":{"verdict":"contested","headline":"Gout is not a ramipril ADR on the label; ramipril is mildly uricosuric, and the variant is artifact-scale","assessment":"Gout is not a licensed indication for ramipril and is not listed as an adverse reaction in the label, which mentions uric acid only among scattered clinical laboratory findings alongside liver enzymes, bilirubin and glucose. The pharmacology points the opposite way: reviews of cardiovascular drugs and serum urate report that captopril, enalapril and ramipril mildly increase uricosuria by reducing proximal tubular reabsorption and can blunt the diuretic-induced rise in serum uric acid, and an acute-dosing study found ramipril increases the fractional excretion of uric acid. The one substantial epidemiological signal in the other direction is Choi et al. (BMJ 2012), a nested case-control study of 24,768 gout cases in UK hypertensives reporting an adjusted RR of 1.24 for ACE inhibitors (versus 2.36 for diuretics, 1.48 for beta blockers, 0.87 for calcium channel blockers and 0.81 for losartan) - a modest excess in a class routinely co-prescribed with thiazides and given to patients with hypertension, heart failure and CKD, all independent gout risk factors. The atlas itself does not support channelling or an ADR: gout is present in only 1.49% of ramipril users with an enrichment of 1.05, i.e. no more common than among users of other drugs, and the 100% temporality over 190 dated participants is the direction the cohort inflates by construction, so it carries little weight. The strongest reading is a real but small class-level urate/gout association driven by comorbidity and diuretic co-therapy rather than by ramipril itself, with the mechanistic literature actively arguing against a ramipril-caused rise in urate. Note that the FAERS PRR of 4.48 for this pair was not independently verified here and would be subject to the same diuretic co-reporting confounding.","gene_comment":"Neither locus is a urate or gout gene: MATN4 encodes matrilin-4, a cartilage/bone extracellular matrix protein studied in osteosarcoma, osteoporosis and oligodendrocyte biology with no renal urate handling role, and LINC01376 is an uncharacterised lncRNA. Established urate/gout GWAS signals sit at SLC2A9, ABCG2, GCKR and the SLC22A/SLC17A transporters, none of which appear here, so no mechanism can be built from this assignment.","score_check":"beta_adr of 7.438 for a rare SNV (rs189976197) corresponds to an odds ratio in the thousands, which is a sparse-data separation artifact rather than a real effect size, and the same variant's disease effect (beta 0.73, se 0.414) fails |beta| < 1.96*se so it is indistinguishable from noise in the drug-free population. The absence of a prescription-propensity signal (log10p 0.15) is reassuring against channelling, but z_diff 4.88 is driven by the inflated treated-arm estimate, so the numbers do not hang together.","candidability":3,"candidability_reason":"Real but modest class-level epidemiological association that is best explained by comorbidity and diuretic co-therapy, with ramipril pharmacology pointing the opposite way and an artifact-scale rare variant on non-urate genes.","sources":[{"title":"Antihypertensive drugs and risk of incident gout among patients with hypertension: population based case-control study (BMJ 2012; PMID 22240117)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:22240117&resultType=core&format=json"},{"title":"Cardiovascular drugs and serum uric acid (Cardiovasc Drugs Ther 2003; PMID 15107595)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:15107595&resultType=core&format=json"},{"title":"Ramipril capsules US prescribing information (Lupin Pharmaceuticals), DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a8cdec12-81d1-4ef4-82c0-d05d93c37a07"},{"title":"PubMed search: ramipril and uric acid (incl. Labeeuw 1987, PMID 2959232, ramipril increases fractional excretion of uric acid)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ramipril+uric+acid"},{"title":"Europe PMC search: angiotensin blockade and hyperuricemia (Wolff 2015, PMID 26568810; Nieradko-Iwanicka 2018, PMID 29853726)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22hyperuricemia%22%20AND%20TITLE%3A%22angiotensin%22&resultType=core&format=json&pageSize=10"},{"title":"Europe PMC search: MATN4 (matrilin-4) function and disease associations","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22MATN4%22&resultType=core&format=json&pageSize=10"}]},"ramipril|M791":{"verdict":"plausible ADR","headline":"Myalgia is a labelled common ramipril ADR, but this rare lncRNA variant carries an implausible OR","assessment":"Myalgia is not an indication for ramipril (licensed for hypertension, cardiovascular prevention, heart failure post-MI and renal disease) but it is an explicitly labelled adverse reaction: the Tritace SmPC lists 'muscle spasms, myalgia' at common frequency and arthralgia as uncommon, and the US label carries myalgia and arthralgia among post-marketing/controlled-trial events. The atlas FAERS flag (STRONG, PRR 2.01) is therefore consistent with an established, if non-specific, label term rather than a new discovery. The population numbers argue against confounding by indication in the usual direction: only 0.63% of ramipril users carry an M79.1 code and enrichment is 0.72, i.e. myalgia is recorded less often in ramipril users than in users of other drugs, so this is not a drug given to people who already have muscle pain. Temporality is 89% after first exposure on only 82 dated participants with a median 2.99 years, which the cohort's record-depth asymmetry inflates, so it is weak supporting evidence at best. The dominant competing explanation is co-medication rather than the ACE inhibitor: ramipril's cardiovascular-prevention population is near-universally co-prescribed statins, and statin-associated muscle symptoms run at roughly 7-29% in observational cohorts, an order of magnitude above the 0.63% seen here, so any residual myalgia signal in ramipril users is very hard to attribute to ramipril itself. A directional counter-argument also exists: ACE inhibitors have been trialled as a treatment for sarcopenia rather than a cause of myopathy, and the LACE trial found perindopril neutral (not harmful) for muscle strength and mass over 12 months, while PubMed contains essentially no ACE-inhibitor myopathy literature beyond a 1984 captopril case report. A bradykinin-mediated mechanism (ACE is kininase II, the accepted route for cough and angioedema) is the only plausible hook, and it is speculative for myalgia, not described.","gene_comment":"rs150777996 is an intronic variant at chr12:126,690,358 assigned to LINC00944, a long non-coding RNA with no established muscle or renin-angiotensin biology, and at ~1.4% minor allele frequency in non-Finnish Europeans it is rare; this is a positional label, not a mechanism, and it should not be presented as one.","score_check":"beta_adr of 5.602 implies an odds ratio near 270 for a 1.4%-frequency variant against a 0.63% outcome, which is a sparse-data/near-separation artifact rather than a real effect size, so the log10p of 7.68 and z_diff of 5.9 are not trustworthy. The disease-arm estimate is also inconsistent with the ADR arm: beta_disease -0.41 +/- 0.199 is barely distinguishable from noise and points the opposite way, and the prescription arm is flat (log10p 0.3), so nothing here is corroborated by a second arm.","candidability":5,"candidability_reason":"A genuinely labelled ramipril adverse reaction, but the genetics are a single rare intronic lncRNA variant with an implausible effect size and a statin co-prescription explanation left standing.","sources":[{"title":"Tritace 10mg Tablets - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/2757/smpc"},{"title":"Ramipril capsules - US prescribing information (DailyMed, Hikma)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=7e9e84f7-ff47-463d-bf47-7a9745960aae"},{"title":"Statin-associated muscle symptoms: impact on statin therapy - EAS Consensus Panel Statement (Eur Heart J 2015, PMID 25694464) and related SAMS prevalence literature","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+muscle+symptoms+myalgia+prevalence+SAMS"},{"title":"LACE trial: ACE I/D genotype and response to ACE inhibitor therapy in older adults with sarcopenia (PLoS One 2023, via Europe PMC)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22perindopril%22%20AND%20%22sarcopenia%22%20AND%20LACE&format=json&resultType=core&pageSize=3"},{"title":"PubMed search: ACE inhibitor myalgia / myopathy (4 results, no ACEi-specific myotoxicity series)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ACE+inhibitor+myalgia+myopathy"},{"title":"Ensembl REST record for rs150777996 (chr12:126,690,358, intronic, gnomAD NFE 1.39%)","url":"https://rest.ensembl.org/variation/human/rs150777996?content-type=application/json;pops=1"}]},"ramipril|N189":{"verdict":"licensed indication","headline":"Ramipril is licensed for and proven to slow proteinuric CKD; an ADR reading runs backwards","assessment":"Chronic kidney disease is not an adverse effect of ramipril but one of the reasons it is prescribed: the UK SmPC for Tritace lists 'incipient glomerular diabetic nephropathy', 'manifest glomerular diabetic nephropathy' and 'manifest glomerular non-diabetic nephropathy defined by macroproteinuria >= 3 g/day' as therapeutic indications, and CKD is in any case near-universal comorbidity in the hypertensive, diabetic and post-MI populations the US label covers. The direction of drug effect is the decisive point: in the REIN trial ramipril cut the monthly GFR decline (0.53 vs 0.88 mL/min, p=0.03) and halved progression to doubling of creatinine or ESRD in proteinuric non-diabetic nephropathy, and MICRO-HOPE showed ramipril reduced overt nephropathy by 24% in diabetics, so a drug that slows CKD cannot plausibly be read as causing incident N18.9. There is a genuine renal adverse effect, but it is a different phenotype and time course - the labels list acute renal failure, renal impairment and raised urea/creatinine as uncommon, driven by RAAS-dependent perfusion in bilateral renal artery stenosis, volume depletion or NSAID/diuretic co-use, typically within weeks of starting; here the median first CKD record is 5.17 years after first exposure, which fits natural progression of hypertensive/diabetic nephropathy on treatment, not a functional haemodynamic hit. The supporting atlas metrics point the same way: enrichment is exactly 1.0 (no excess of CKD in ramipril users over users of other drugs), FAERS shows no signal with PRR 1.11, the condition is not on the BNF adverse-effect list, and temporality of 98.4% is the uninformative direction given the cohort's known upward compression. Confounded surveillance also matters - patients on ramipril have renal function monitored by protocol, so CKD is more likely to be coded in them at all, independent of any causal effect. The only thing here that looks like a finding is the single variant, and it is the weakest part of the package.","gene_comment":"rs117308443 has no gene assignment because there is none to make: it sits at chr13:62,873,344 (GRCh38) in a 13q21 gene desert whose only annotated neighbours within ~900 kb are unnamed lncRNAs and pseudogenes (LINC00448 ends ~77 kb away), with no protein-coding gene anywhere near. With MAF ~0.45% in Ensembl/gnomAD this is a rare intergenic marker that supports no mechanistic story about renal handling, ACE, bradykinin or afferent arteriolar tone.","score_check":"The disease arm is flat and well behaved (beta 0.068 +/- 0.158, |beta| well under 1.96*se, log10p 0.18), so the variant does not predispose to CKD without the drug; the within-user beta of 2.87 at log10p 6.6 implies an odds ratio near 18 with an implied SE around 0.57, which for a 0.45% MAF variant against a few hundred CKD cases is the classic sparse-cell configuration that inflates rare-variant effect sizes. The z_diff of 4.85 is therefore driven entirely by the fragile arm, not by a contrast between two well-estimated effects.","candidability":1,"candidability_reason":"Drug is licensed for and shown to slow this condition, FAERS and BNF null, enrichment 1.0, and the signal rests on a rare intergenic variant in a gene desert with an implausibly large effect.","sources":[{"title":"Tritace 10mg Tablets SmPC (Sanofi) - sections 4.1, 4.4, 4.8","url":"https://www.medicines.org.uk/emc/product/2757/smpc"},{"title":"Ramipril capsules US label (Lupin), DailyMed - Indications and Warnings","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a8cdec12-81d1-4ef4-82c0-d05d93c37a07"},{"title":"GISEN Group, Ramipril Efficacy In Nephropathy (REIN), Lancet 1997;349:1857-63","url":"https://pubmed.ncbi.nlm.nih.gov/9217756/"},{"title":"HOPE/MICRO-HOPE: effects of ramipril on cardiovascular and microvascular outcomes in diabetes, Lancet 2000;355:253-9","url":"https://pubmed.ncbi.nlm.nih.gov/10675071/"},{"title":"Schmidt et al., Serum creatinine elevation after renin-angiotensin system blockade and long term cardiorenal risks, BMJ 2017;356:j791","url":"https://pubmed.ncbi.nlm.nih.gov/28279964/"},{"title":"Ensembl variant record for rs117308443 (intergenic, MAF 0.0045, 13:62873344)","url":"https://rest.ensembl.org/variation/human/rs117308443?content-type=application/json"},{"title":"dbSNP refSNP record rs117308443","url":"https://www.ncbi.nlm.nih.gov/snp/rs117308443"}]},"ramipril|R101":{"verdict":"plausible ADR","headline":"Upper abdominal pain is a labelled ramipril ADR; the rare pericentromeric hit near a pseudogene is not","assessment":"R10.1 is not a licensed indication for ramipril — the Tritace SmPC lists hypertension, cardiovascular prevention, diabetic and non-diabetic nephropathy, symptomatic heart failure and secondary prevention after myocardial infarction — but it is an explicitly labelled adverse reaction: section 4.8 lists 'abdominal pain upper including gastritis' and 'small bowel angioedema' as uncommon, with 'abdominal discomfort, dyspepsia, nausea' and 'gastrointestinal inflammation' as common, and section 4.4 warns that intestinal angioedema presents with abdominal pain. A drug-caused mechanism is described and specific: bradykinin accumulation from ACE inhibition produces visceral/mesenteric angioedema, and a 2025 FAERS disproportionality analysis of drug-induced intestinal angioedema found strong signals for ACE inhibitors (lisinopril ROR 350.6 on 254 reports, captopril 60.0, losartan 19.5, enalapril 16.6), noting that 57% of such patients had already undergone unnecessary GI surgery or biopsy before the drug was recognised. Published cases show both subacute onset weeks after initiation and presentations after 7-11 years of therapy, so the atlas median of 3.81 years after first exposure is not itself disqualifying, though temporality of 89.3% (n_dated 475) is the compressed direction and carries little weight on its own. Against this, the atlas gives enrichment 1.01 — epigastric pain is no more common among ramipril users (2.05%) than among users of other drugs — and log10p_prescribed 0.57, so there is neither confounding by indication nor a population excess to explain; R10.1 is a very nonspecific code that cannot separate bradykinin-mediated bowel angioedema from ordinary dyspepsia or gastritis. The genetic layer is the weak part: a single rare intergenic variant with an implausibly large effect, no disease-arm signal (beta_disease 0.183 ± 0.102, |beta| < 1.96*se, i.e. indistinguishable from noise), and no connection to the loci that ACE-inhibitor angioedema GWAS actually implicate (PRKCQ, RAD51B, RIMS1, and rs500766 at chr10:6.5 Mb — 36 Mb from this hit). So the phenotype is a genuine labelled ADR, but this variant is not evidence for it.","gene_comment":"rs182945706 is an intergenic SNV at chr10:42,722,542 (GRCh38) with gnomAD MAF ~0.36%, and RSU1P1 is HGNC-classified simply as 'Ras suppressor protein 1 pseudogene 1' at 10q11.21 — a non-transcribed pseudogene, not a mechanism. The position is pericentromeric 10q11, a segmental-duplication-rich region where genotyping and imputation are unreliable, so the gene label here is a nearest-feature artefact rather than a candidate.","score_check":"beta_adr 2.844 (OR ~17) for a variant at 0.36% frequency, reaching only log10p 6.27, is the classic sparse-data profile: a handful of carriers driving the estimate, and z_diff 4.62 is inherited entirely from that inflated ADR beta rather than from any real contrast with the drug-free arm, where the effect is flatly null. The phenotype-level numbers (cotx 2.05%, enrichment 1.01, FAERS PRR 3.98) are believable; the variant-level ones are not.","candidability":5,"candidability_reason":"Real, label-documented ACE-inhibitor adverse reaction with a described bradykinin mechanism, but the genetics are a rare intergenic pericentromeric variant near a pseudogene with a sparse-data effect size and no disease-arm or literature support.","sources":[{"title":"Tritace 10mg Tablets (ramipril) SmPC — sections 4.1 and 4.8, electronic Medicines Compendium","url":"https://www.medicines.org.uk/emc/product/2757/smpc"},{"title":"Drug-Induced Intestinal Angioedema: A Disproportionality Analysis (Medical Sciences, 2025)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12735144/"},{"title":"PubMed search: ACE inhibitor visceral angioedema abdominal pain (case series and reports, 1996-2026)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ACE+inhibitor+visceral+angioedema+abdominal+pain"},{"title":"Europe PMC: genetic predictors of ACE-inhibitor-induced angioedema (GWAS, PRKCQ / RAD51B / RIMS1 / rs500766)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22angiotensin-converting%20enzyme%20inhibitor%22%20angioedema%20genetic%20XPNPEP2%20OR%20BDKRB2&format=json&pageSize=8&resultType=core"},{"title":"Ensembl REST: rs182945706 — chr10:42,722,542, intergenic, gnomAD MAF 0.36%","url":"https://rest.ensembl.org/variation/human/rs182945706?content-type=application/json;pops=1"},{"title":"HGNC record 31114 — RSU1P1, Ras suppressor protein 1 pseudogene 1, locus type pseudogene, 10q11.21","url":"https://rest.genenames.org/fetch/hgnc_id/31114"}]},"ramipril|R35":{"verdict":"statistical artifact","headline":"Polyuria is not a ramipril ADR; RAS blockade should reduce it, and the UBR5 hit is a rare-variant artifact","assessment":"Polyuria is neither a licensed indication for ramipril (hypertension, cardiovascular risk reduction, post-MI heart failure, nephropathy) nor a labelled adverse reaction: the US prescribing information lists raised BUN/creatinine, acute renal failure, proteinuria, hyperkalaemia, hyponatraemia and altered blood glucose, and the words polyuria, polydipsia, urinary frequency and diabetes insipidus appear nowhere in it. The direction of effect argues against an ADR reading: intrarenal renin-angiotensin system activation is what drives salt-induced nocturnal polyuria in animal models, suppressing intrarenal RAS improves it, and an ACE inhibitor plus diuretic has actually been patented as a treatment for nocturnal polyuria, so ramipril should if anything lower urine output. R35 in this population is far more readily explained by uncontrolled diabetes, loop or thiazide co-therapy, CKD concentrating defect and age-related nocturia than by the ACE inhibitor, and the atlas agrees that ramipril users are not selected for it (cotx 0.41%, enrichment 0.95, indication score 0.0). The FAERS 'STRONG' flag reproduces on openFDA (201 of 123,139 ramipril reports mention polyuria vs 8,088 of 20.7M overall, ratio ~4.2), but it dissolves on inspection: those 201 reports carry metformin (46), furosemide (37), insulin (23), empagliflozin (22), dapagliflozin (17), hydrochlorothiazide (10) and lithium (4), i.e. ramipril is a background cardiometabolic co-medication in reports whose polyuria belongs to osmotic diuresis, diuretics or lithium-induced nephrogenic DI. Temporality of 100% over only 116 dated participants is exactly the compressed, uninformative high value the cohort produces by design and adds nothing. The within-user genetic signal is the weakest part: beta_adr 4.329 (OR ~76) on an intronic variant with MAF ~0.4% in a phenotype affecting ~0.4% of users implies a handful of carrier cases, which is quasi-separation rather than a large effect.","gene_comment":"rs138518996 is a rare intronic SNV (MAF ~0.004, gnomAD/TOPMed) in UBR5, a HECT E3 ubiquitin ligase studied almost entirely in cancer and DNA-damage biology, with no GWAS Catalog associations and no documented role in vasopressin signalling, aquaporin trafficking or renal water handling. The assignment is positional only and supplies no mechanism for polyuria.","score_check":"The disease arm is internally coherent but weak (beta 0.427 +/- 0.177, z = 2.4, log10p 1.8), whereas beta_adr 4.329 at log10p 6.83 for a 0.4%-frequency variant in ~116 cases is a sparse-data effect size, so the z_diff of 4.63 measures the instability of the ADR beta rather than a real treatment interaction. Only the null propensity arm (log10p 0.1) and the flat enrichment can be taken at face value.","candidability":1,"candidability_reason":"Not a labelled or mechanistically plausible ADR, drug expected to move polyuria the other way, FAERS signal explained by co-medication, and the variant effect is quasi-separation on a handful of carriers.","sources":[{"title":"Ramipril capsules - US prescribing information (DailyMed, Lupin)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=a8cdec12-81d1-4ef4-82c0-d05d93c37a07"},{"title":"dbSNP rs138518996 (UBR5, intron variant, MAF ~0.004)","url":"https://www.ncbi.nlm.nih.gov/snp/rs138518996"},{"title":"Cerebral Sympathetic Nervous System Activation Promotes Intrarenal Renin-Angiotensin System Activity and Salt-Induced Nocturnal Polyuria (Int Neurourol J, PMID 41942335)","url":"https://europepmc.org/article/MED/41942335"},{"title":"Decreased nitric oxide production is a novel therapeutic target for salt-induced nocturnal polyuria in aging (PMID 39875406)","url":"https://europepmc.org/article/MED/39875406"},{"title":"Patent: Treatment of nocturnal polyuria using an ACE inhibitor in combination with a diuretic","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=ABSTRACT%3A%28%22nocturnal%20polyuria%22%20AND%20%22ACE%20inhibitor%22%29&format=json"},{"title":"openFDA FAERS drug event API - ramipril / polyuria co-reporting counts","url":"https://api.fda.gov/drug/event.json?search=patient.drug.openfda.generic_name:%22ramipril%22+AND+patient.reaction.reactionmeddrapt:%22polyuria%22&limit=1"},{"title":"GWAS Catalog REST - associations for gene UBR5 (none reported)","url":"https://www.ebi.ac.uk/gwas/rest/api/genes/UBR5/associations"},{"title":"PubMed search: ramipril AND polyuria","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ramipril+polyuria"}]},"ramipril|R410":{"verdict":"plausible ADR","headline":"Confusional state is a labelled ramipril ADR, but the rare intronic GJB6 variant is not a mechanism","assessment":"R41.0 (disorientation, unspecified) is not an indication for ramipril, whose SmPC licenses it only for hypertension, cardiovascular prevention, renal disease, symptomatic heart failure and post-MI secondary prevention. It is, however, a labelled adverse reaction at the symptom level: the Tritace SmPC lists 'confusional state' as uncommon and 'disturbance in attention' as rare under psychiatric disorders, with hypotension, orthostatic blood pressure decrease and syncope as common vascular effects, and the FAERS arm here is concordant (STRONG, PRR 6.47). The atlas's own confounding check is reassuring rather than damning: disorientation is coded in 1.16% of ramipril users with an enrichment of only 1.03 against users of other drugs, so this is not simply a symptom of the population that gets prescribed an ACE inhibitor, and the prescription-propensity arm is flat (log10p 0.59). Temporality is 98.7% post-exposure over 152 dated participants with a median 5.53 years, but the cohort compresses this upward and a 5-year lag to an age-linked symptom code in an elderly cardiovascular population is weak evidence on its own. A drug-caused route is easy to describe and does not need novel biology: cerebral hypoperfusion from first-dose or orthostatic hypotension, ACE-inhibitor-associated hyponatraemia (documented since the captopril case reports of hyponatraemia with irreversible neurologic injury, PMID 3907348, and a recognised contributor to confusion and delirium in older patients, PMID 23510827), and acute kidney injury with uraemia. Importantly the drug is not clearly protective in the opposite direction that would refute an ADR reading: the ACE-inhibitor cognition literature is genuinely mixed, with antihypertensives broadly associated with less cognitive decline but ARBs outperforming ACE inhibitors in several analyses and at least one recent study reporting increased Alzheimer risk with ACE inhibition. What does not hold up is the genetics: the phenotype is credible, the variant is not.","gene_comment":"rs142230271 is a rare (MAF ~0.3%) intronic variant at chr13:20,231,113, inside GJB6 (20,221,962-20,232,450) so the gene call is positionally right but functionally empty; GJB6 encodes connexin 30, a gap-junction protein of inner ear, skin and astrocytes whose established phenotypes are non-syndromic deafness and Clouston syndrome, it carries no GWAS Catalog trait associations, and dbSNP flags this allele likely benign. There is no described link between connexin 30 and ACE-inhibitor pharmacology, blood-pressure response or delirium, so this must be reported as a locus, not a mechanism.","score_check":"beta_adr 4.719 is an odds ratio near 110 for a 0.3%-frequency allele against a 1.16% outcome, which is the classic signature of sparse-data separation on a handful of carriers rather than a real effect of that size, even though log10p 9.98 and z_diff 5.93 look clean. The comparison arms are internally consistent and honest - beta_disease 0.275 +/- 0.17 is under 1.96*se so the variant does nothing without the drug, and prescription propensity is null - but the interaction rests entirely on an implausibly large treated-arm estimate that needs carrier counts and a replication cohort before it means anything.","candidability":5,"candidability_reason":"Real labelled ADR with FAERS support and no confounding by indication, but the rare intronic GJB6 variant and its OR ~110 are not believable genetics.","sources":[{"title":"Tritace 10mg Tablets - Summary of Product Characteristics (electronic medicines compendium)","url":"https://www.medicines.org.uk/emc/product/2757/smpc"},{"title":"PubMed: ACE inhibitors and cognitive impairment / dementia (search results incl. PMID 40832424, 35834254, 18294116)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ACE+inhibitor+cognitive+impairment+dementia"},{"title":"PubMed: ACE inhibitor hyponatraemia and confusion in the elderly (incl. PMID 3907348 captopril-induced hyponatraemia, PMID 23510827)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=ACE+inhibitor+hyponatremia+elderly+confusion"},{"title":"Ensembl REST: variant rs142230271 (intronic, MAF 0.0029, likely benign, chr13:20231113)","url":"https://rest.ensembl.org/variation/human/rs142230271?content-type=application/json"},{"title":"MedlinePlus Genetics: GJB6 gene (connexin 30)","url":"https://medlineplus.gov/genetics/gene/gjb6/"},{"title":"GWAS Catalog: gene GJB6 (no reported trait associations)","url":"https://www.ebi.ac.uk/gwas/genes/GJB6"},{"title":"Ensembl REST: GJB6 gene coordinates (chr13:20,221,962-20,232,450)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/GJB6?content-type=application/json"}]},"simvastatin|E785":{"verdict":"licensed indication","headline":"Hyperlipidaemia is simvastatin's licensed indication; the drug lowers LDL ~38%, so this reads backwards","assessment":"E78.5 (hyperlipidaemia, unspecified) is the core licensed indication for simvastatin, not an adverse effect: the US label indicates it as an adjunct to diet to reduce LDL-C in adults with primary hyperlipidaemia, in heterozygous and homozygous familial hypercholesterolaemia, and in primary dysbetalipoproteinaemia and hypertriglyceridaemia. The label's own efficacy data run in the opposite direction to an ADR reading - in 4S at 20 mg, total cholesterol fell 28%, LDL-C 38% and triglycerides 19%, with HDL-C up 8% - so a drug-caused increase in hyperlipidaemia is not merely unsupported but contradicted. The recognised simvastatin adverse reactions are myopathy/rhabdomyolysis, transaminase elevation, immune-mediated necrotising myopathy and modest rises in glucose/HbA1c; dyslipidaemia is not among them. The atlas signals are exactly what confounding by indication looks like: enrichment 1.47 in simvastatin users, temporality 77% with a 2.85-year median gap (chronic E78.5 codes are typically first entered at a review after the statin was started on a lipid result, so a high value here carries no weight), and a FAERS PRR of 6.7 that is the classic artefact of the indication being reported as the event term. The only non-indication reading worth naming is that residual or worsening hyperlipidaemia on treatment could mark statin non-response or non-adherence, but the atlas cannot distinguish that from ordinary indication coding, and the EXOSC8 burden gives no traction on it. The 3.44% co-treatment rate is far below the true indication prevalence, which reflects Polish coding habits (statin users coded under I10, E78.0 or CVD risk rather than E78.5) rather than any weakening of the indication link.","gene_comment":"EXOSC8 is a non-catalytic core subunit of the RNA exosome (Exo-9); its only established human phenotype is autosomal-recessive pontocerebellar hypoplasia type 1C, and neither UniProt nor the EXOSC8 literature annotates any lipid or cholesterol role, with no lipid-trait hits mapped to it in the GWAS Catalog. A biallelic-LOF neurodegeneration gene is not a plausible mediator of statin pharmacology, and the assignment should not be dressed up as mechanism.","score_check":"The disease-arm effect is real but modest and consistent with predisposition rather than drug action (beta 1.187, se 0.46, z = 2.58, log10p 2.0), whereas the within-users beta of 6.901 is an implausible burden effect size (OR in the hundreds) that signals a sparse-carrier LOF artefact despite log10p 7.34, making z_diff 4.25 untrustworthy. log10p_prescribed of 0.33 shows the variant does not predict who gets the drug, so the signal is carried entirely by the sparse ADR arm.","candidability":0,"candidability_reason":"Definitionally not an adverse reaction - it is the licensed indication, and the drug moves the condition in the opposite direction.","sources":[{"title":"Simvastatin tablet, film coated - full prescribing information (DailyMed, Camber Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=751b79c6-bba8-4d29-b69c-a2ac6453fec8"},{"title":"Zocor (simvastatin) - indications and adverse reactions","url":"https://www.rxlist.com/zocor-drug.htm"},{"title":"UniProtKB Q96B26 (EXOS8_HUMAN) - Exosome complex component RRP43","url":"https://rest.uniprot.org/uniprotkb/Q96B26.txt"},{"title":"GWAS Catalog - associations mapped to EXOSC8","url":"https://www.ebi.ac.uk/gwas/genes/EXOSC8"},{"title":"Europe PMC literature search: EXOSC8","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXOSC8&format=json&pageSize=10&resultType=core"}]},"simvastatin|F03":{"verdict":"statistical artifact","headline":"Statins trend protective for dementia; this rare intergenic variant with OR~56 is a sparse-data artifact","assessment":"Unspecified dementia (F03) is not a licensed indication for simvastatin, whose label covers hyperlipidaemia and cardiovascular risk reduction. Cognition does appear on the label, but only as Postmarketing Experience: 'Rare reports of cognitive impairment (e.g., memory loss, forgetfulness, amnesia, memory impairment, confusion) associated with statin use... generally nonserious, and reversible upon statin discontinuation, with variable times to symptom onset (1 day to years) and symptom resolution (median of 3 weeks)' — a reversible, non-progressive complaint that is a different entity from a coded chronic dementia. The class effect on dementia proper runs the other way: pooled observational meta-analyses report statin use associated with lower dementia risk (HR 0.86, 95% CI 0.82-0.91; AD HR 0.82) and RR 0.85 in an earlier dose-response synthesis, and recent randomized/narrative syntheses conclude statins do not harm neurocognitive function — a drug that moves the condition in the opposite direction is decisive evidence against an ADR reading here, even allowing that the protective observational signal is itself healthy-user confounded. The atlas descriptives fit age and indication rather than causation: only 0.28% of simvastatin users carry F03, no background arm exists so enrichment cannot be tested, and the 100% temporality over just 95 dated participants (median 4.82 years after exposure) is exactly what you expect when a midlife lipid prescription precedes a late-life dementia code in a cohort whose prescription records reach further back than its diagnosis records. The within-user signal rests on a single rare intergenic SNV with beta 4.027 — an odds ratio near 56 — which is not a credible effect size for a common geriatric syndrome and is the signature of sparse-data separation in a handful of carriers. The clean null in the prescribing-propensity arm (log10p 0.24) rules out confounding by prescription, but that only means the artifact is a case-count artifact rather than an indication artifact.","gene_comment":"rs145266573 has no gene assignment in the atlas and none in the GWAS Catalog; Ensembl places it at chr2:198,790,131 as an intergenic variant with MAF ~0.4%, in a gene desert between PLCL1 and SATB2 with no reported trait associations. There is no gene here to build a mechanism on, and none should be inferred.","score_check":"The disease arm is properly null (beta -0.006 +/- 0.246, well inside noise) and the prescription arm is null too, so z_diff 4.71 is driven entirely by the within-user estimate; but a beta of 4.027 for a 0.4%-frequency allele against ~95 dementia cases implies a handful of carriers and near-complete separation, so the log10p of 6.05 is not believable as an effect size. FAERS support is weak (1/3, PRR 1.99) and the drug is not flagged on the BNF for this condition, consistent with the signal being noise rather than a suppressed real association.","candidability":2,"candidability_reason":"Backwards direction versus the dementia literature, no gene, and an implausible rare-variant effect size in a tiny case set — near-certainly artifact.","sources":[{"title":"ZOCOR (simvastatin) prescribing information, DailyMed - Postmarketing Experience, Nervous System Disorders","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=8f55d5de-5a4f-4a39-8c84-c53976dd6af9"},{"title":"PubMed: statins and dementia risk - meta-analyses (Westphal Filho 2025 PMID 40199829; Li 2025 PMID 39822593; Zhang 2018 PMID 30045255)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statins+dementia+risk+meta-analysis"},{"title":"Europe PMC records for PMIDs 40199829 / 39822593 / 30045255 - pooled HR 0.86 for dementia, RR 0.85 dose-response","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:39822593%20OR%20EXT_ID:40199829%20OR%20EXT_ID:30045255&format=json&resultType=core&pageSize=5"},{"title":"Europe PMC: statins and cognition in randomized trials (GeroScience 2026, PMID 42010230; Front Neurol 2026, PMID 41716550)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22statin%22%20AND%20%22cognitive%22%20AND%20%22randomized%22&format=json&pageSize=10&resultType=core"},{"title":"Ensembl REST: rs145266573 - chr2:198,790,131, intergenic, MAF 0.0039","url":"https://rest.ensembl.org/variation/human/rs145266573?content-type=application/json"},{"title":"GWAS Catalog: rs145266573 - no mapped gene, no reported associations","url":"https://www.ebi.ac.uk/gwas/variants/rs145266573"}]},"simvastatin|G309":{"verdict":"co-prescription population","headline":"APOE/APOC1 rs4420638 predicts Alzheimer disease and statin prescribing alike; no drug-specific effect","assessment":"Alzheimer disease is not a licensed indication for simvastatin, whose FDA-approved indications are CHD risk reduction, hyperlipidaemia and heterozygous familial hypercholesterolaemia; the label's only cognitive entry is rare postmarketing reports of memory loss, forgetfulness and confusion that are 'generally nonserious, and reversible upon statin discontinuation' (median resolution 3 weeks), which is explicitly not a neurodegenerative dementia. The direction of the literature is opposite to an ADR reading: an updated meta-analysis of 55 observational studies and over 7 million patients reports statin use associated with lower all-cause dementia (HR 0.86, 95% CI 0.82-0.91) and lower Alzheimer disease (HR 0.82, 95% CI 0.74-0.90). The atlas signal is instead the APOE-APOC1 locus doing what it always does: rs4420638 carries log10p 99.47 with beta 1.226 +/- 0.058 for Alzheimer disease in the drug-free population, and within simvastatin users beta is 1.399 with z_diff of only 0.78, i.e. the treated effect is statistically indistinguishable from the untreated one. Critically, the same variant reaches log10p 35.16 on propensity to be prescribed simvastatin, because this locus also governs lipid traits, so it selects the exposed group and predicts the outcome through two independent paths. Only 0.26% of simvastatin users carry a G309 code and no enrichment background is available, so there is no evidence the drug is preferentially given to Alzheimer patients or vice versa; temporality of 89.3% over just 103 dated participants (median 4.27 years after first exposure) is exactly what age and the cohort's record-length asymmetry would produce for a late-onset dementia in a chronically prescribed drug. There is no described mechanism by which simvastatin would cause Alzheimer disease specifically in APOC1 carriers, and the pharmacoepidemiology points the other way.","gene_comment":"rs4420638 is a real, well-mapped SNV in the APOE/APOC1 cluster on 19q13 and is one of the strongest common Alzheimer signals known (p <= 2e-34 in recent work), but it is also a lipid variant (associated with raised triglycerides and lowered apoA1), which is precisely why it flags statin prescribing. The gene assignment is solid; it just indexes background predisposition and treatment selection, not a drug interaction.","score_check":"The numbers are internally consistent and not artifactual: beta_disease 1.226 with se 0.058 is a genuine 21-sigma effect, and the within-user beta sits within noise of it (z_diff 0.78). What they do not support is any drug-specific term, and log10p_prescribed 35.16 makes the selection path explicit.","candidability":1,"candidability_reason":"Null drug-specific effect (z_diff 0.78) at a locus that predicts both the outcome and the prescription, with meta-analytic evidence that statins lower, not raise, Alzheimer risk.","sources":[{"title":"ZOCOR (simvastatin) US prescribing information, FDA (indications; postmarketing cognitive impairment reports)","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2012/019766s085lbl.pdf"},{"title":"Statin use and dementia risk: a systematic review and updated meta-analysis, Alzheimers Dement (N Y) 2025, PMID 39822593 (Europe PMC record)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:39822593&resultType=core&format=json"},{"title":"PubMed: statin dementia Alzheimer meta-analysis (result set, direction of effect across reviews)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+dementia+Alzheimer+meta-analysis"},{"title":"Europe PMC: rs4420638 / APOC1 association with Alzheimer disease (Gholami et al., Psychogeriatrics 2026, PMID 42535978)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=rs4420638%20AND%20(Alzheimer%20OR%20APOE)&resultType=core&format=json&pageSize=5"},{"title":"Europe PMC: rs4420638 and lipid traits (triglycerides, apoA1), Front Endocrinol 2025, PMID 41356011","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=rs4420638%20AND%20(cholesterol%20OR%20lipid)&resultType=core&format=json&pageSize=5"}]},"simvastatin|G629":{"verdict":"contested","headline":"Peripheral neuropathy is on the simvastatin label, but the epidemiology is null and PLB1 is not credible","assessment":"Polyneuropathy is not an indication for simvastatin, whose licensed use is LDL-C lowering and cardiovascular risk reduction; peripheral neuropathy and paraesthesia are listed verbatim among postmarketing nervous-system adverse reactions in the US label, so a labelled ADR relationship exists on paper. The epidemiology behind that listing is genuinely contested: the original Danish case-control study (Gaist 2002) reported OR 3.7 overall and 14.2 for definite idiopathic polyneuropathy, but a much larger Danish replication (Svendsen 2017, 370 cases / 7400 controls) found OR 1.14 (0.84-1.54), a prospective Dutch case-control study (Warendorf 2019) found statin users at LOWER risk (OR 0.56, 0.34-0.95) with risk falling as exposure lengthened, and the 2022 Muscle & Nerve meta-analysis pooled to 1.24 (0.88-1.76) in non-diabetics and 0.82 (0.56-1.21) in diabetics - i.e. no association. A mechanism has been proposed (ubiquinone/cholesterol depletion in Schwann-cell membranes and axonal energy metabolism, the same lipid-lowering axis behind statin myotoxicity) and FAERS disproportionality for atorvastatin still shows axonal and sensory neuropathy signals after stripping label indications, so the hypothesis is not dead - but the best-designed cohort evidence points at null or slightly protective, which undercuts an ADR reading. The atlas is consistent with that null: enrichment is 0.99, meaning polyneuropathy is no more common in simvastatin users than in users of other drugs, and cotx_pct is only 0.58%. Temporality of 97.1% over 174 dated participants with a 4.66-year median lag is exactly the pattern the cohort's record-depth asymmetry manufactures, so it adds nothing; the internal FAERS flag is weak (1/3, PRR 1.34).","gene_comment":"rs75944756 (chr2:28,457,902, GRCh38) sits ~750 bp inside the extreme 5' edge of the PLB1 span and is annotated by Ensembl as an intergenic variant with MAF ~0.8%, so the gene call is proximity-based rather than functional. PLB1 encodes a phospholipase B of intestinal and lung epithelium with no known role in peripheral nerve biology or in statin disposition - none of the established statin pharmacogenes (SLCO1B1, ABCG2, CYP3A4) is implicated here.","score_check":"beta_adr 3.922 at log10p 6.11 implies SE around 0.8 and an odds ratio near 50 for a rare (MAF ~0.8%) non-coding variant, which is the classic sparse-data signature rather than a real effect of that magnitude. The comparison arms behave correctly - the disease arm is properly null (beta 0.136 +/- 0.265, |beta| well under 1.96*SE) and prescription propensity is flat - so z_diff 4.53 is a clean treated-only pattern, but one rare single variant at one locus with an implausible effect size needs replication before it is believable.","candidability":4,"candidability_reason":"Labelled ADR whose best epidemiology is null-to-inverse, no enrichment in the atlas, and a rare intergenic variant with a sparse-data-sized effect in a gene with no nerve or statin biology.","sources":[{"title":"Simvastatin tablet, film coated - DailyMed label (Adverse Reactions, Postmarketing Experience)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=751b79c6-bba8-4d29-b69c-a2ac6453fec8"},{"title":"Gaist D et al. Statins and risk of polyneuropathy: a case-control study. Neurology 2002 (PMID 12011277)","url":"https://europepmc.org/article/MED/12011277"},{"title":"Svendsen TK et al. Statins and polyneuropathy revisited: case-control study in Denmark 1999-2013. Br J Clin Pharmacol 2017 (PMID 28370351)","url":"https://europepmc.org/article/MED/28370351"},{"title":"Warendorf JK et al. Statins do not increase risk of polyneuropathy: a case-control study and literature review. Neurology 2019 (PMID 30737334)","url":"https://europepmc.org/article/MED/30737334"},{"title":"Wannarong T et al. Statins and the risk of polyneuropathy: a systematic review and two meta-analyses. Muscle Nerve 2022 (PMID 34693541)","url":"https://europepmc.org/article/MED/34693541"},{"title":"Atorvastatin-associated neurological adverse events: a disproportionality analysis with adjustment for confounding by indication. Front Med 2026 (PMC13385032)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13385032/"},{"title":"Ensembl REST: rs75944756 variant record and PLB1 gene coordinates (GRCh38)","url":"https://rest.ensembl.org/variation/human/rs75944756?content-type=application/json"}]},"simvastatin|H000":{"verdict":"statistical artifact","headline":"Stye is not a statin effect; statins reduce eyelid gland disease, and the variant is a rare lncRNA intron hit","assessment":"Hordeolum (H00.0) is an acute purulent staphylococcal inflammation of the Zeis/Moll or meibomian glands of the lid margin, common, usually self-limited within about a week and prone to recurrence; the Cochrane review describes it as a benign obstructive/infectious lid-margin condition, not a systemic drug toxicity. It is not an indication for simvastatin, whose FDA labelling covers cardiovascular risk reduction and LDL-C lowering in hyperlipidaemia and familial hypercholesterolaemia, and it is not a labelled adverse reaction: the only ocular items anywhere in the simvastatin label are rodent Harderian-gland adenomas in the carcinogenicity section and rare post-marketing reports of ocular myasthenia gravis. The nearest well-studied analogues point the other way: in a 335,070-patient Taiwanese NHIRD cohort statin users had a LOWER incidence of blepharitis (3.04% vs 3.72%, adjusted HR 0.746, p<0.001), and a Korean cross-sectional study found statin use independently associated with a lower meiboscore, i.e. better meibomian gland morphology, so if statins move lid-gland disease at all they move it in the protective direction, which is decisive against an ADR reading. The atlas's own comparison arms agree with that: FAERS shows no signal (PRR 0.73), the drug is not on the BNF list for this event, only 0.21% of simvastatin users carry the code, and there is no background arm at all (enrichment null) so the 94.2% temporality figure is exactly the cohort's known upward compression of a common, sporadic, any-time-in-life diagnosis rather than evidence of induction. What remains is a single rare intronic SNV with an implausibly large effect on a trivially common infection, which is the signature of sparse-data separation, not pharmacology.","gene_comment":"rs139975518 (chr4:140,225,948, minor allele G, MAF ~0.37%) is an intronic variant in ENSG00000308003, an unnamed novel havana_tagene lncRNA sitting in the gene desert between MAML3 and SCOC; there is no lipid, sebaceous-gland, skin-barrier or immune gene at the locus, so the assignment carries no mechanistic content whatever.","score_check":"beta_adr 4.462 implies an odds ratio near 86 for a 0.37%-frequency allele acting on a 0.21%-prevalence eyelid infection, which no real common-disease effect looks like and which almost certainly reflects a handful of carrier cases; log10p_adr 6.17 is also below genome-wide significance, and the disease arm is flatly null (beta 0.353, se 0.254, |beta| < 1.96*se), so the z_diff of 4.4 is driven entirely by the unstable treated-arm estimate.","candidability":1,"candidability_reason":"No label or literature support, the best evidence has statins reducing lid-gland disease, and the signal is one rare intronic variant in an unnamed lncRNA with a sparse-data effect size.","sources":[{"title":"Simvastatin tablets USP - FDA prescribing information (openFDA drug label API record)","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22simvastatin%22&limit=1"},{"title":"ZOCOR (simvastatin) tablets, FDA label PDF (Merck)","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2012/019766s085lbl.pdf"},{"title":"Feng et al., Statin Use Is Associated With a Lower Risk of Blepharitis: A Population-Based Study, Front Med 2022 (PMID 35372440)","url":"https://pubmed.ncbi.nlm.nih.gov/35372440/"},{"title":"Park et al., Association of Meibomian Gland Dysfunction with Oral Statin Use, J Clin Med 2022 (PMID 35956248)","url":"https://pubmed.ncbi.nlm.nih.gov/35956248/"},{"title":"Cheng et al., Acupuncture for acute hordeolum, Cochrane Database Syst Rev 2017 - background on hordeolum pathophysiology and natural history (PMID 28181687)","url":"https://pubmed.ncbi.nlm.nih.gov/28181687/"},{"title":"Ensembl REST: gene ENSG00000308003 (novel lncRNA, chr4:140,172,795-140,238,180)","url":"https://rest.ensembl.org/lookup/id/ENSG00000308003?content-type=application/json"},{"title":"Ensembl REST: variant rs139975518 (intron variant, MAF 0.0037)","url":"https://rest.ensembl.org/variation/human/rs139975518?content-type=application/json"}]},"simvastatin|H353":{"verdict":"co-prescription population","headline":"Macular degeneration in statin users is shared-risk overlap; meta-analysis puts statins at OR 0.93","assessment":"The ZOCOR (simvastatin) US label is indicated only for cardiovascular risk reduction and lipid lowering, and lists no ocular adverse reaction of any kind in sections 6.1/6.2 - the only retinal or lens findings are optic-nerve/retinal-ganglion changes and cataracts in dogs and rats at 5-30x human exposure. Macular degeneration is therefore neither an indication nor a labelled adverse effect. The pharmacoepidemiology points the other way from an ADR: a 22-study meta-analysis of 2,063,195 participants found a pooled OR of 0.93 (95% CI 0.83-1.05) for AMD in statin users, a 2022 Retina meta-analysis found no difference in incidence or progression, and several 2025-2026 cohort and target-trial-emulation studies report reduced AMD risk with prolonged or high-intensity lipophilic statin use, with an 80 mg atorvastatin pilot showing drusenoid PED regression in 10 of 23 patients. The atlas's own co-occurrence numbers fit confounding rather than causation: enrichment is only 1.24, which is what shared age, smoking and vascular risk between the dyslipidaemic and the AMD population produces, and the 94.8% temporality with a 4.57-year median lag is the expected artefact of statins being started in the fifties-sixties while AMD is diagnosed in the seventies. The FAERS PRR of 5.44 is the one discordant element, but spontaneous reporting for a drug used by hundreds of millions of elderly people against a disease of the elderly is exactly the setting where disproportionality reflects co-occurrence and reporting channel rather than causation, and it is not corroborated by any of the controlled literature. Nothing here supports simvastatin causing macular degeneration; if anything the class is being studied as protective.","gene_comment":"ENSG00000232153 is an Ensembl 'novel transcript' lncRNA at chr2:34.73-34.82 Mb in a gene-poor stretch of 2p22.3, and rs57314774 is a common (MAF ~0.13) intronic variant within it with no entry in the GWAS Catalog and no relation to any established AMD locus (CFH, ARMS2/HTRA1, C3, CFB, APOE, LIPC). This is a positional label, not a mechanism, and a single variant with no supporting neighbours.","score_check":"The numbers are internally consistent - beta_disease -0.013 +/- 0.036 is a well-powered null, so z_diff 4.96 is driven entirely by the within-user beta of 0.635 (implied SE ~0.13 from log10p 6.37), plausible for ~400 dated cases. But that is one sub-genome-wide-significant variant (p ~ 4e-7) claiming an OR of 1.89 that is flatly absent in the far larger disease arm, which is the classic shape of a small-stratum winner's-curse hit.","candidability":2,"candidability_reason":"Controlled evidence puts the statin effect on AMD at neutral-to-protective, the enrichment and temporality are age-driven, and the genetics rest on one intronic SNV in an unannotated lncRNA.","sources":[{"title":"ZOCOR (simvastatin) US prescribing information, FDA Drugs@FDA label 019766s093","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2015/019766s093lbl.pdf"},{"title":"The Relationship between Statin and Risk of Age-Related Macular Degeneration: A Systematic Review and Meta-Analysis (J Ophthalmol 2022, PMC9110218)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9110218/"},{"title":"Statins and Fibrates in Age-Related Macular Degeneration: A Contemporary Clinical Narrative Review (2020-2025), J Clin Med 2026 (PMID 42074763)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22Statins%20and%20Fibrates%20in%20Age-Related%20Macular%20Degeneration%22&format=json&resultType=core&pageSize=1"},{"title":"Regression of some high-risk features of age-related macular degeneration with high-dose atorvastatin, EBioMedicine 2016 (PMID 27077128)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22Regression%20of%20Some%20High-risk%20Features%20of%20Age-related%20Macular%20Degeneration%22&format=json&resultType=core&pageSize=1"},{"title":"Europe PMC title search: statin AND macular degeneration (Retina 2022 meta-analysis PMID 34983903; Ophthalmol Retina 2026 target trial emulation PMID 42526538; Br J Clin Pharmacol 2025 PMID 40494650)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22statin%22%20AND%20TITLE%3A%22macular%20degeneration%22&format=json&resultType=core&pageSize=8"},{"title":"Ensembl REST lookup for ENSG00000232153 (lncRNA, novel transcript, chr2:34,732,287-34,822,726)","url":"https://rest.ensembl.org/lookup/id/ENSG00000232153?content-type=application/json"},{"title":"Ensembl REST variation record for rs57314774 (chr2:34,757,937 C/T, MAF 0.127, intron variant)","url":"https://rest.ensembl.org/variation/human/rs57314774?content-type=application/json"},{"title":"GWAS Catalog entry for rs57314774 (no reported associations)","url":"https://www.ebi.ac.uk/gwas/variants/rs57314774"}]},"simvastatin|I10":{"verdict":"co-prescription population","headline":"Hypertension marks who gets a statin, not what a statin causes; statins modestly lower blood pressure","assessment":"Essential hypertension is not a licensed indication for simvastatin, whose US label covers CV risk reduction in established coronary heart disease, primary hyperlipidaemia, familial hypercholesterolaemia, dysbetalipoproteinaemia and hypertriglyceridaemia, and it appears nowhere in the label's adverse-reaction or postmarketing lists. It is instead the single most characteristic comorbidity of the statin-treated population: statins are allocated on absolute cardiovascular risk, of which hypertension is a core component, and ASCOT-LLA randomised 10,305 explicitly hypertensive patients to atorvastatin, which is the design logic behind the atlas's enrichment of 2.09 and cotx_pct of 28.7%. A drug-caused mechanism in the direction implied is not merely undescribed but contradicted: pooled RCT evidence puts statins at roughly -1.9 mmHg systolic (Strazzullo 2007) and -1.6/-0.96 mmHg systolic/diastolic across 46 placebo trials and 49,087 participants (Alghamdi 2020), and the 2025 meta-analysis in hypertensive patients found diastolic pressure specifically reduced in the simvastatin subgroup. The 97.9% temporality and 3.53-year median lag are exactly what this cohort's prescription-before-diagnosis record depth produces for a chronic, opportunistically coded diagnosis, and carry almost no weight here. The one internally coherent element is genetic specificity: beta_disease 0.017 +/- 0.024 is indistinguishable from zero, log10p_prescribed is 0.02, and z_diff 4.95 says the variant's effect is confined to treated patients rather than reflecting general hypertension predisposition or prescribing propensity. But at log10p_adr 7.77 this is only just past genome-wide significance on a single common SNV with no supporting biology, and the FAERS signal is weak (1/3 detectors, PRR 1.92), which is far too thin to overturn trial evidence pointing the other way.","gene_comment":"ENSG00000293732 is an unnamed 'novel transcript' lncRNA on 10q21.1 and rs12413893 (chr10:64,775,221, MAF 0.063) is intronic within it; the neighbourhood is effectively a gene desert whose only annotated features are processed pseudogenes (RPL17P35, ANXA2P3, and CYP2C61P, which is a pseudogene unrelated to the functional CYP2C cluster at 10q23.33). This assignment supports no mechanistic story and should not be presented as one.","score_check":"The within-users association is arithmetically self-consistent (beta 0.328 at log10p 7.77 on ~10.7k dated participants is a reasonable effect for its precision, and beta_disease 0.017 +/- 0.024 is honestly reported as null), so this is not a sparse-data artifact. What the numbers cannot do is distinguish an interaction from a variant that tags something about the hypertensive subset of statin users, and 7.77 leaves no margin for a locus with no prior support.","candidability":2,"candidability_reason":"Confounding by indication on the atlas's most predictable comorbidity, with the drug's proven blood-pressure effect running the opposite way and the variant sitting in an unnamed lncRNA at barely genome-wide significance.","sources":[{"title":"Simvastatin tablet, film coated - DailyMed label (Indications and Adverse Reactions)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=b2d6c7ac-304b-4cba-974c-d9a23a069925"},{"title":"Strazzullo P et al. Do statins reduce blood pressure? A meta-analysis of randomized, controlled trials. Hypertension 2007 (PMID 17309949)","url":"https://pubmed.ncbi.nlm.nih.gov/17309949/"},{"title":"Alghamdi J et al. Blood pressure-lowering activity of statins: systematic review and meta-analysis of placebo-randomized controlled trials. Eur J Clin Pharmacol 2020 (PMID 32696233)","url":"https://pubmed.ncbi.nlm.nih.gov/32696233/"},{"title":"Chu Z et al. Effects of statin use on blood pressure and hypertension-related outcomes in hypertensive patients: systematic review and meta-analysis. 2025 (PMID 40174858)","url":"https://pubmed.ncbi.nlm.nih.gov/40174858/"},{"title":"Sever PS et al. ASCOT-LLA: atorvastatin in hypertensive patients with average or lower-than-average cholesterol. Lancet 2003 (PMID 12686036)","url":"https://pubmed.ncbi.nlm.nih.gov/12686036/"},{"title":"Ensembl REST: ENSG00000293732 (novel transcript, lncRNA, chr10:64,739,699-64,911,683) and rs12413893","url":"https://rest.ensembl.org/lookup/id/ENSG00000293732?content-type=application/json"}]},"simvastatin|I251":{"verdict":"licensed indication","headline":"I25.1 is simvastatin's own indication; statins reduce atherosclerotic CHD events, not cause them","assessment":"Atherosclerotic heart disease is written into the simvastatin label as an indication, not an adverse effect: the US prescribing information indicates simvastatin 'to reduce the risk of total mortality by reducing risk of coronary heart disease death, non-fatal myocardial infarction and stroke, and the need for coronary and non-coronary revascularization procedures in adults with established coronary heart disease', the trial basis being 4S and the Heart Protection Study. A drug-caused mechanism is therefore not merely unsupported but backwards: simvastatin lowers LDL-C and reduces the incidence of exactly the endpoint coded here, which is decisive evidence against an ADR reading. The atlas numbers match confounding by indication rather than causation: enrichment 1.49 (I25.1 is 49% commoner in simvastatin users than in users of other drugs), and log10p_prescribed 26.74 says the same variants predict merely being put on the drug. The within-user effect (beta_adr 0.317, log10p 6.21) is weaker than, and statistically indistinguishable from, the effect of the same variants on the condition in the drug-free population (beta_disease 0.385 +/- 0.029, log10p 39.98; z_diff 0.98), so there is no drug-by-genotype interaction at all - the atlas is re-detecting ordinary CAD genetics inside a treated subgroup. Temporality of 98.6% with a median 1.03 years to diagnosis is uninformative here for the reason the atlas itself flags, and is anyway the expected natural history: statins are commonly started for hypercholesterolaemia or primary prevention and the ischaemic diagnosis is coded later as the underlying atherosclerosis progresses. The FAERS 'STRONG' signal with PRR 4.75 reflects reporting of cardiac events in a population selected for cardiac risk, not a causal pharmacovigilance finding. The one genuinely pharmacogenomic reading of this row is not an ADR but residual risk: Wei et al. (Circulation 2018) showed LPA rs10455872 predicts CHD events during statin therapy (OR 1.58, p=2.6e-10) independent of the degree of LDL-C lowering, i.e. incomplete protection rather than drug-induced harm.","gene_comment":"Both loci are among the best-replicated coronary-artery-disease genes in human genetics - rs10455872 tags high Lp(a) at LPA, and rs10757270 is an intronic variant in CDKN2B-AS1/ANRIL at chr9:22,072,720 (GRCh38), inside the 9p21.3 CHD interval first reported by McPherson et al. (Science 2007), though rs10757270 itself is a proxy rather than one of the canonical reported SNPs (rs10757274/rs1333049). They are highly plausible genes for atherosclerosis and wholly implausible as a simvastatin toxicity mechanism; statins do not lower Lp(a) and may raise it in low-molecular-weight apo(a) carriers, so LPA marks a risk statins fail to treat rather than one they create.","score_check":"The numbers are internally coherent and not artifactual - beta_disease is 13 SEs from zero on n large enough for log10p ~40 - but they point at the disease's own genetics, with z_diff 0.98 explicitly showing no excess effect in treated patients. The high log10p_prescribed is the tell: the variants predict exposure, which is the signature of confounding by indication rather than of an ADR.","candidability":1,"candidability_reason":"The condition is the drug's licensed indication and the drug moves it in the protective direction; z_diff ~1 shows no treated-vs-untreated genetic difference.","sources":[{"title":"SIMVASTATIN tablet, film coated - DailyMed prescribing information (Indications and Usage)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=b2d6c7ac-304b-4cba-974c-d9a23a069925"},{"title":"Wei WQ et al. LPA Variants Are Associated With Residual Cardiovascular Risk in Patients Receiving Statins. Circulation 2018;138:1839-1849 (PMC6202211)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC6202211/"},{"title":"McPherson R et al. A common allele on chromosome 9 associated with coronary heart disease. Science 2007 (PMID 17478681)","url":"https://europepmc.org/article/MED/17478681"},{"title":"Yahya R et al. Statin treatment increases lipoprotein(a) levels in subjects with low molecular weight apolipoprotein(a) phenotype. Atherosclerosis 2019 (PMID 31327478)","url":"https://europepmc.org/article/MED/31327478"},{"title":"Donnelly LA et al. Robust association of the LPA locus with LDL-cholesterol lowering response to statin treatment, meta-analysis of 30,467 individuals. Pharmacogenet Genomics 2013 (PMID 23903772)","url":"https://pubmed.ncbi.nlm.nih.gov/23903772/"},{"title":"Ensembl REST: rs10757270 mapping, chr9:22,072,720 (GRCh38), intron_variant","url":"https://rest.ensembl.org/variation/human/rs10757270?content-type=application/json"}]},"simvastatin|I48":{"verdict":"co-prescription population","headline":"Statins move atrial fibrillation neutral-to-protective; AF here is the cardiovascular population, not an ADR","assessment":"Atrial fibrillation is not a licensed indication for simvastatin (the label covers cardiovascular risk reduction and hyperlipidaemias) and it is not in the postmarketing adverse-reaction list; it appears only in Table 1 of the 4S clinical-trial tally, at 5.7% on simvastatin versus 5.1% on placebo (N=2221 vs 2223 over a median 5.4 years), an unadjudicated ~13-event difference that is not statistically significant. Randomised evidence points the other way or nowhere: a 32-trial, 71,005-patient meta-analysis found OR 0.69 (95% CI 0.57-0.83) overall, OR 0.37 for postoperative AF, and a flatly null OR 1.00 (0.86-1.15) for new-onset AF, while an earlier 20-trial analysis reported OR 0.49 with a significant benefit specifically in the simvastatin subgroup, and rosuvastatin in JUPITER cut incident AF by 27% (HR 0.73). A drug that is neutral-to-protective in randomised data cannot be read as causing the condition, so the FAERS 'STRONG' signal (PRR 3.38) is best explained by channelling and co-reporting: AF patients are anticoagulated cardiovascular patients who are also statin users, and AF is entered as a concomitant condition on reports. The atlas's own comparison arms say the same thing quietly: enrichment is 1.02, meaning AF is no more common in simvastatin users than in users of other drugs, and the prescription-propensity arm is null (log10p 0.47), so there is no positive drug-specific excess to explain. The 100% temporality over 1708 dated participants with a 4.22-year median lag is exactly the pattern the cohort's record-depth asymmetry manufactures for a chronic drug started in midlife before an age-accumulating arrhythmia, and carries essentially no causal weight on its own.","gene_comment":"RASGEF1C (chr5:180.1-180.2 Mb, 5q35.3) is a real protein-coding RasGEF-domain gene but is predominantly brain-expressed, has no cardiac or ion-channel role, and carries zero entries in the GWAS Catalog for any trait, let alone AF or any known AF locus (PITX2, ZFHX3, KCNN3, CAND2). A single deletion-burden hit in a subtelomeric region is a locus label, not a mechanism, and should not be presented as one.","score_check":"The within-user beta of 6.758 corresponds to an odds ratio of roughly 860 for a condition present in 7.3% of the cohort, which is arithmetically near-impossible and signals near-complete separation on a handful of CNV carriers rather than a real effect; the implied standard error (~1.15) is itself as large as most genuine effects. The disease arm is reassuringly honest by contrast: beta 0.618 +/- 0.476 is 1.3 standard errors from zero, i.e. noise, so the headline z_diff of 4.96 is driven entirely by the unstable treated-arm estimate.","candidability":1,"candidability_reason":"Randomised data put statins neutral-to-protective for AF, enrichment is 1.02, the gene has no cardiac support, and the effect size is a separation artifact.","sources":[{"title":"Simvastatin tablets, full prescribing information (DailyMed SPL) - indications, adverse reactions, Study 4S Table 1","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=3f6fad0b-0278-433f-86db-761b673a5803"},{"title":"Statin therapy and atrial fibrillation: systematic review and updated meta-analysis of published randomized controlled trials (Curr Opin Cardiol 2013, PMID 23160338)","url":"https://europepmc.org/article/MED/23160338"},{"title":"The role of statin therapy in the prevention of atrial fibrillation: a meta-analysis of randomized controlled trials (Br J Clin Pharmacol 2012, PMID 22376147)","url":"https://europepmc.org/article/MED/22376147"},{"title":"Antiarrhythmic effect of statin therapy and atrial fibrillation: a meta-analysis of randomized controlled trials (JACC 2008, PMID 18294568)","url":"https://europepmc.org/article/MED/18294568"},{"title":"High-sensitivity C-reactive protein, statin therapy, and risks of atrial fibrillation: an exploratory analysis of the JUPITER trial (Eur Heart J 2012, PMID 22187510)","url":"https://europepmc.org/article/MED/22187510"},{"title":"UniProtKB Q8N431 - Ras-GEF domain-containing family member 1C (RASGEF1C), human","url":"https://rest.uniprot.org/uniprotkb/search?query=gene:RASGEF1C+AND+organism_id:9606&format=txt"},{"title":"Ensembl gene record ENSG00000146090 (RASGEF1C) and GWAS Catalog association query (0 associations)","url":"https://www.ebi.ac.uk/gwas/rest/api/genes/RASGEF1C/associations?projection=associationByGene"}]},"simvastatin|I517":{"verdict":"statistical artifact","headline":"Statins shrink LV mass rather than enlarge it; the HOXD10 burden effect is a separation artifact","assessment":"Cardiomegaly (I51.7) is neither a licensed indication for simvastatin nor a labelled adverse reaction: the ZOCOR prescribing information lists only lipid-lowering and cardiovascular risk-reduction indications, and the words cardiomegaly, hypertrophy and heart failure do not appear anywhere in the label, including the adverse reactions section. Cardiomegaly is instead an incidental echo/radiographic finding of the hypertensive and ischaemic heart disease that gets patients started on a statin in the first place, which is the obvious source of the FAERS signal (PRR 3.93) as well. The direction of the drug's real effect is the decisive point against an ADR reading: statins reduce left ventricular mass, both in patients (a case-control study of 304 angina patients found significantly lower LV mass index in pravastatin/simvastatin users) and in animal models (simvastatin regressed hypertrophy and fibrosis in a transgenic rabbit model of hypertrophic cardiomyopathy; rosuvastatin reduced hypertrophy, fibrosis and NADPH-oxidase/Rac1 expression in Ren2 rats), with a described mechanism through inhibition of isoprenylation of small G proteins. The atlas is therefore proposing that simvastatin causes the condition it is documented to attenuate. The internal numbers do not rescue this: enrichment is 0.79, i.e. cardiomegaly is if anything less common in simvastatin users than in users of other drugs, the prescription arm is flat (log10p 0.16) and the disease arm is only marginal (beta 5.82 +/- 2.45, z = 2.4). The 97.1% temporality over 412 dated participants is exactly the pattern the cohort's prescription-record depth manufactures and carries no weight on its own.","gene_comment":"HOXD10 is a real protein-coding Abd-B-family homeobox transcription factor (UniProt P28358) that patterns the anterior-posterior axis and limb buds, is expressed mainly in the urogenital tract, and whose only established human disease link is congenital vertical talus with Charcot-Marie-Tooth; it has no cardiac, myocardial-growth or lipid-handling role. A rare-variant burden mask on a 340-residue developmental transcription factor is not a credible route to statin-associated cardiomegaly.","score_check":"beta_adr = 44.5 is a log-odds of roughly 1e19 on the odds scale, which is not an effect size but quasi-complete separation in a sparse MPC burden test, so the log10p of 8.47 and the z_diff of 4.89 both inherit that instability. The comparison arms point the other way (enrichment 0.79, prescription p flat, disease effect only 2.4 SE from zero), so nothing in the sheet supports a real drug-specific genetic effect.","candidability":1,"candidability_reason":"Not a labelled ADR, the drug is documented to move LV mass in the opposite direction, the condition is depleted rather than enriched in users, and the gene effect is an implausible separation artifact in a gene with no cardiac biology.","sources":[{"title":"ZOCOR (simvastatin) US prescribing information, FDA","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2012/019766s085lbl.pdf"},{"title":"Nishikawa H et al. Statins induce the regression of left ventricular mass in patients with angina. Circ J 2004 (PMID 14745145)","url":"https://pubmed.ncbi.nlm.nih.gov/14745145/"},{"title":"Simvastatin induces regression of cardiac hypertrophy and fibrosis in a transgenic rabbit model of human hypertrophic cardiomyopathy (PMID 11457751)","url":"https://pubmed.ncbi.nlm.nih.gov/11457751/"},{"title":"Habibi J et al. Rosuvastatin decreases cardiac oxidative stress and remodeling in Ren2 transgenic rats. Endocrinology 2007 (PMID 17317778)","url":"https://pubmed.ncbi.nlm.nih.gov/17317778/"},{"title":"UniProtKB P28358 (HXD10_HUMAN) Homeobox protein Hox-D10","url":"https://rest.uniprot.org/uniprotkb/P28358.txt"}]},"simvastatin|I802":{"verdict":"co-prescription population","headline":"Factor V Leiden recovered inside statin users; DVT is background thrombophilia, and statins lower VTE","assessment":"Deep-vein phlebitis/thrombophlebitis (I80.2) is neither a licensed indication for simvastatin nor a labelled adverse reaction: the US simvastatin label lists cardiovascular risk reduction and lipid-lowering as indications, and its clinical-trial and postmarketing adverse-reaction sections (myopathy, rhabdomyolysis, hepatotoxicity, interstitial lung disease, cognitive effects, SJS) contain no venous thrombosis, thromboembolism, phlebitis or thrombophlebitis term. The drug is in fact described as moving this condition in the opposite direction: rosuvastatin reduced first VTE in JUPITER (HR 0.57, 95% CI 0.37-0.86) and reviews of statins in VTE conclude they lower incidence and recurrence rather than raise it, so an ADR reading is contradicted by randomised evidence. The atlas numbers say the same thing internally: DVT is present in only 0.53% of simvastatin users with enrichment 1.06, i.e. no more common than among users of other drugs, and the variant does not predict being prescribed the drug (log10p_prescribed 0.97). The lead variant is rs6025, Factor V Leiden, whose effect in the drug-free population is overwhelming (log10p 76.83, beta 1.904 +/- 0.102, OR ~6.7, matching the classic 3-8x heterozygote risk in GeneReviews) while the within-user effect is essentially identical (beta 1.816, z_diff 0.24) - the treated subgroup is simply re-detecting the germline thrombophilia, not a drug-modified effect. Temporality of 89.2% over 203 dated participants with a 4.4-year median is uninformative here, since the cohort compresses temporality upward and DVT incidence rises with the same age at which statins are started. FAERS support is weak (1/3, PRR 1.48), consistent with background reporting rather than a signal.","gene_comment":"F5 rs6025 is Factor V Leiden, the best-characterised thrombophilia variant, but it is a disease-predisposition allele with no statin-specific pharmacology; the second variant rs144737447 is assigned to NME7, a ciliary nucleoside diphosphate kinase with no coagulation role that simply sits ~144 kb from F5 in the same 1q24.2 region, so that assignment is proximity, not mechanism.","score_check":"The numbers are internally believable and well-powered - beta_disease 1.904 +/- 0.102 is a tight, textbook-sized Factor V Leiden effect, not a sparse-data artifact. What they do not support is a drug effect: z_diff 0.24 and enrichment 1.06 place the whole signal in the background disease genetics.","candidability":1,"candidability_reason":"Pure disease-predisposition locus recovered inside drug users, with no enrichment, no effect modification, and randomised evidence that statins reduce VTE.","sources":[{"title":"Simvastatin tablets, film coated - FDA label (DailyMed, Accord Healthcare)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=871251c0-36a1-4a32-9eab-ff6c1e925ca9"},{"title":"Factor V Leiden Thrombophilia - GeneReviews (NCBI Bookshelf NBK1368)","url":"https://www.ncbi.nlm.nih.gov/books/NBK1368/"},{"title":"The role of statins in mitigating venous thromboembolism incidence and severity (Res Pract Thromb Haemost, 2025; PMID 41169532)","url":"https://europepmc.org/article/MED/41169532"},{"title":"Bempedoic Acid and Venous Thromboembolism Risk Among Statin-Intolerant Patients (JAMA Cardiol; PMID 42201706) - cites JUPITER rosuvastatin VTE HR 0.57","url":"https://europepmc.org/article/MED/42201706"},{"title":"Ensembl REST: F5 and NME7 gene coordinates on chromosome 1q24.2","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/NME7?content-type=application/json"}]},"simvastatin|I831":{"verdict":"statistical artifact","headline":"Varicose veins are no statin reaction; huge beta on rare variants with a null disease arm is noise","assessment":"Simvastatin is licensed only for hyperlipidaemias and for cardiovascular risk reduction, and I83.1 (varicose veins of the lower extremity with inflammation) is neither an indication nor a labelled adverse reaction: the DailyMed prescribing information lists no varicose, phlebitic, thrombophlebitic or venous-disorder term in clinical-trial or post-marketing adverse reactions, whose vascular-adjacent content is dominated by myopathy, rhabdomyolysis and myalgia. The experimental literature runs in the opposite direction to an ADR reading - Eschrich et al. (J Am Heart Assoc 2016) showed HMG-CoA reductase inhibition suppresses varicose remodelling of veins in stretch, perfusion and ligation models, with reduced smooth-muscle proliferation and lower MCP1/MMP2/MMP9, although notably simvastatin itself was the one statin of three that did not reproduce the effect, so the fair reading for this drug is neutral rather than protective. Clinically, a secondary analysis of three randomised trials (Wound Repair Regen 2022) found more venous leg ulcers healed among statin users after adjustment, again giving no signal of statin-driven venous harm. Varicose veins are instead an age-, BMI- and immobility-driven condition that is simply common in the population simvastatin is prescribed to, and the atlas cannot rebut that here because enrichment is null - the whole confounding-by-population test is missing - while temporality of 89.7% on only 126 dated participants is exactly the direction the cohort compresses upward and therefore carries little weight, and the FAERS PRR of 2.82 is the expected shape for a chronic comorbidity co-reported in an elderly polypharmacy population rather than evidence of causation. The genetic arm does not rescue it: beta_adr 4.744 corresponds to an odds ratio near 100 for a common complex phlebological trait, at log10p 6.66 that is below the conventional genome-wide threshold, and the same variants are flatly null outside the drug (log10p_disease 0.07, beta -0.057 +/- 0.287), so the z_diff of 5.0 is carried entirely by an implausibly large within-user estimate on rare alleles in a handful of cases.","gene_comment":"ABHD14B and DOCK3 are neighbours on 3p21 - a poorly characterised hydrolase/lysine deacylase and a neuronal Rac guanine-nucleotide exchange factor - and CHRM2 (7q33) is the M2 muscarinic receptor whose GWAS Catalog record is dominated by heart rate, PR interval and behavioural traits, with no venous phenotype for any of the three. None appears among the 46 replicated varicose-vein loci (PIEZO1, EFEMP1, VEGFA, PROX1, COL27A1, DOCK8 - not DOCK3), so the assignments carry no venous-biology support and should not be narrated as mechanism.","score_check":"The numbers do not hang together: an OR of roughly 100 for varicose veins on rare SNVs, with the identical variants showing no effect at all in the drug-free population and only 126 dated cases, is the classic signature of sparse-data separation rather than a real interaction. The FAERS and temporality flags are both metrics this design is known to inflate, and with enrichment unavailable there is no independent check left standing.","candidability":1,"candidability_reason":"Not a labelled or literature-supported statin reaction, drug effect on venous remodelling is neutral-to-protective, genes are venously irrelevant, and the effect size is a sparse-data artifact.","sources":[{"title":"DailyMed - Ezetimibe and Simvastatin tablets, full prescribing information (indications, adverse reactions)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=5ea3ca31-cfee-4978-8b7a-755986f6b091"},{"title":"Eschrich J et al. Varicose Remodeling of Veins Is Suppressed by 3-Hydroxy-3-Methylglutaryl Coenzyme A Reductase Inhibitors. J Am Heart Assoc 2016 (PMC4802467)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC4802467/"},{"title":"Genome-wide association analysis and replication in 810,625 individuals with varicose veins (PMC9163161)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9163161/"},{"title":"Europe PMC search: statins and venous leg ulcer healing (incl. Wound Repair Regen 2022 secondary analysis of three RCTs)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=statin%20AND%20%22venous%20leg%20ulcer%22&format=json&resultType=core&pageSize=5"},{"title":"GWAS Catalog associations mapped to CHRM2","url":"https://www.ebi.ac.uk/gwas/api/search/downloads?q=ensemblMappedGenes%3A%22CHRM2%22&facet=association&efo=true"},{"title":"GWAS Catalog associations mapped to DOCK3","url":"https://www.ebi.ac.uk/gwas/api/search/downloads?q=ensemblMappedGenes%3A%22DOCK3%22&facet=association&efo=true"},{"title":"PubMed search: statin, chronic venous insufficiency and varicose veins","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+chronic+venous+insufficiency+varicose+veins"}]},"simvastatin|J310":{"verdict":"statistical artifact","headline":"Chronic rhinitis is not a simvastatin effect; statins suppress nasal eosinophilic inflammation if anything","assessment":"Chronic rhinitis (J31.0) is neither an indication for simvastatin nor a labelled adverse reaction: the US prescribing information lists only hyperlipidaemia and cardiovascular risk reduction as indications, and its adverse-reaction section names upper respiratory infection (9%), bronchitis and sinusitis, with sinusitis actually lower on simvastatin than placebo in 4S (1.8% vs 2.3%); post-marketing respiratory entries are interstitial lung disease and dyspnoea, not rhinitis. The mechanistic literature points the opposite way from an ADR: simvastatin inhibits IL-5-induced eosinophil chemotaxis and CCR3 expression and reduced eosinophil infiltration in an allergic-rhinitis model, atorvastatin attenuated allergic nasal inflammation by blocking prostaglandin biosynthesis, and simvastatin is being explored as a topical antimicrobial for chronic rhinosinusitis - so if statins move nasal inflammation at all, they move it downward. The atlas itself gives no support for confounding-by-indication either: enrichment is 1.14 with only 0.61% co-occurrence, and the prescription-propensity arm is flat (log10p 0.08), so this is neither an indication signal nor a channelling signal, just an isolated within-users burden hit. Temporality of 86.4% over 198 dated participants with a 3.07-year median lag is exactly the pattern the cohort's record-depth asymmetry manufactures, and is weak evidence on its own. The FAERS flag (PRR 2.15) is the only external corroboration, and spontaneous rhinitis reports on a drug taken by tens of millions of people are a poor discriminator against reporting artefact, especially with no BNF listing. Taken together, the finding reads as a sparse-data burden hit on a common, age-related ENT diagnosis rather than a drug-caused reaction.","gene_comment":"DENND1C encodes DENN-domain protein connecdenn 3, a guanine-nucleotide exchange factor for RAB8A/RAB13/RAB35 involved in membrane trafficking and studied mainly in angiogenesis, podocyte trafficking and as a lung-tumour immune-infiltration marker; it has no established role in nasal mucosal disease, atopy or statin pharmacology, and no allergic-airway GWAS support. A single-gene MPC burden test on such a gene, with no supporting SNV signal, carries essentially no mechanistic weight here.","score_check":"The within-users beta of 25.2 is not a credible effect size for a missense-burden test on a 0.61%-prevalence outcome - that magnitude is the signature of quasi-complete separation in sparse cells, and no standard error is supplied to check it. The comparison arm is also barely significant (beta_disease 3.38 with se 1.565, |beta| only 2.2 se), so the z_diff of 4.12 is driven by the unstable treated-arm estimate rather than by a real divergence.","candidability":2,"candidability_reason":"Implausible effect size for its data density, an implausible gene, no label or literature support, and a mechanism literature that points in the protective direction.","sources":[{"title":"Simvastatin tablets prescribing information (DailyMed, Camber Pharmaceuticals)","url":"https://www.dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=751b79c6-bba8-4d29-b69c-a2ac6453fec8"},{"title":"PubMed: statin and allergic rhinitis - study set incl. simvastatin/IL-5/CCR3 eosinophil chemotaxis (PMID 27275740) and atorvastatin in allergic rhinitis (PMID 36634416)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+allergic+rhinitis"},{"title":"Europe PMC: statin/simvastatin and rhinitis, rhinosinusitis risk - incl. antimicrobial activity of simvastatin against CRS-related S. aureus (PMID 40657706)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%28statin%20OR%20simvastatin%29%20AND%20rhinitis%20AND%20%28cohort%20OR%20risk%20OR%20randomized%29&format=json&pageSize=25&resultType=core"},{"title":"UniProtKB: human DENND1C (DENN domain-containing protein 1C / connecdenn 3), RAB8A/RAB13/RAB35 GEF","url":"https://rest.uniprot.org/uniprotkb/search?query=gene:DENND1C+AND+organism_id:9606&fields=accession,protein_name,gene_names,cc_function,cc_tissue_specificity&format=tsv"},{"title":"Europe PMC: published literature mentioning DENND1C (trait and disease associations)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=DENND1C&format=json&pageSize=20&resultType=core"}]},"simvastatin|J449":{"verdict":"co-prescription population","headline":"COPD in statin users is shared-smoking comorbidity; trials show statins neutral-to-protective, not causal","assessment":"COPD is not an indication for simvastatin: the ZOCOR label lists only cardiovascular risk reduction and lipid disorders, and the only respiratory events in its post-marketing section are interstitial lung disease and dyspnoea, which are pathologically distinct from fixed smoking-related airflow obstruction. Nor is COPD a recognised adverse reaction; in 4S the only respiratory term reaching the >=2% table was bronchitis, at 6.6% on simvastatin versus 6.3% on placebo, i.e. no separation. The direction of the drug effect actually argues against an ADR reading: STATCOPE randomised 885 patients with moderate-to-severe COPD to simvastatin 40 mg or placebo and found identical exacerbation rates (1.36 vs 1.39 per person-year, P=0.54) and identical serious-adverse-event rates, while a 2026 meta-analysis of three RCTs found statins reduced exacerbation-related hospitalisation (RR 0.87) and severe exacerbations (RR 0.88). The co-occurrence the atlas sees is the well-described COPD-cardiovascular overlap driven by shared smoking, age and socioeconomic status, which is exactly why statins are heavily prescribed in this group; an enrichment of only 1.32 is the modest excess that comorbidity predicts, not a drug effect. Temporality of 93.7% and a median 4.63 years to diagnosis are uninformative here, since statins are started in midlife for primary prevention and COPD is typically coded years later - the ordering is age and ascertainment, not causation. The FAERS PRR of 3.2 is unpersuasive for a chronic background comorbidity in an elderly, poly-medicated reporting population, where the underlying disease is routinely captured on the report form.","gene_comment":"CADM2 is a real protein-coding synaptic cell-adhesion gene at 3p12.1, but it has no pulmonary biology; its GWAS signature is behavioural - risk tolerance, BMI and substance-use initiation (it is the top locus for lifetime cannabis use, rs7609594, p=7.4e-20) - so a CADM2 deletion plausibly tags smoking and risk-taking, the very confounders that produce both COPD and a statin prescription. A single CNV- call in a ~1.1 Mb gene is also the variant class most vulnerable to calling error, and no mechanism links it to airway obstruction.","score_check":"The numbers are internally consistent with confounding rather than pharmacology: beta_disease -0.317 +/- 0.63 is indistinguishable from zero in the drug-free population, predisposition is 0, and the prescription-propensity signal is weak (log10p 1.37), so the within-user hit stands alone. A beta of 5.1 (OR ~160) from one rare CNV deletion at log10p 6.28 is far more likely sparse-data inflation than a real effect of that size, and z_diff 4.53 mostly reflects that inflation.","candidability":1,"candidability_reason":"Backwards signal: randomised evidence shows statins do not worsen COPD, and the association is explained by shared smoking/cardiovascular comorbidity plus a behaviour-linked, sparse CNV call.","sources":[{"title":"Criner et al., Simvastatin for the prevention of exacerbations in moderate-to-severe COPD (STATCOPE), N Engl J Med 2014","url":"https://europepmc.org/article/MED/24836125"},{"title":"Statins for the prevention of exacerbations in COPD: systematic review and meta-analysis of RCTs, Systematic Reviews 2026","url":"https://europepmc.org/article/MED/42157340"},{"title":"ZOCOR (simvastatin) US prescribing information via openFDA drug label API","url":"https://api.fda.gov/drug/label.json?search=openfda.brand_name:%22ZOCOR%22&limit=1"},{"title":"COPD and Cardiovascular Diseases: Biomarker-Guided Stratification and Therapeutic Perspectives (2025)","url":"https://europepmc.org/article/MED/41517298"},{"title":"The genetics of cannabis lifetime use (CADM2 rs7609594 top locus), Neuropsychopharmacology 2026","url":"https://europepmc.org/article/MED/41044382"},{"title":"NCBI Gene / mygene.info record for CADM2 (cell adhesion molecule 2, 3p12.1)","url":"https://mygene.info/v3/query?q=symbol:CADM2&fields=name,summary,genomic_pos,type_of_gene,map_location&species=human"}]},"simvastatin|J90":{"verdict":"co-prescription population","headline":"Pleural effusion tracks the elderly cardiac population statins treat; OARD1 LOF signal is sparse-data noise","assessment":"Pleural effusion is not an indication for simvastatin, whose licensed uses are cardiovascular risk reduction and lipid lowering, and it is not a listed adverse reaction on the label; the closest post-marketing entries are interstitial lung disease, dyspnoea, eosinophilia and a lupus-erythematosus-like syndrome. The published human evidence for a pleural effect is a single 2006 case report of eosinophilic pleural effusion attributed to simvastatin after 13 years of exposure, and statins do not appear in the standard reviews of drug-induced eosinophilic pleural effusion, which list anticonvulsants, SSRIs, dantrolene, benzodiazepines and muscle relaxants. Against that, pleural effusion is overwhelmingly a disease of the same population simvastatin is prescribed to: it is present in roughly 40-47% of patients hospitalised with acute or advanced heart failure, and the other big causes (malignancy, mean age about 71, and pneumonia) are equally age- and comorbidity-driven. The atlas numbers say the same thing quantitatively: 1.34% co-occurrence with an enrichment of 1.06 means pleural effusion is essentially no more common in simvastatin users than in users of other drugs, which is what confounding by an old, cardiac, multimorbid population looks like rather than a drug-specific hazard. Temporality of 99.6% over 495 dated participants with a median 4.7 years to onset is exactly the uninformative direction for this cohort, since prescription records reach further back than diagnosis records, and the near-zero prescription-propensity signal (log10p 0.06) at least means the variant is not simply predicting who gets the drug. The FAERS disproportionality (PRR 2.38) is consistent with reporting of effusions in statin-treated heart-failure and cancer patients and cannot separate that from a causal effect. A hypersensitivity or drug-induced-lupus pleuritis mechanism is biologically conceivable and label-adjacent, but it is case-report rare and nothing in this locus points to it.","gene_comment":"OARD1 (C6orf130/TARG1) is a small, ubiquitously expressed ADP-ribose glycohydrolase whose only established human phenotype is a rare recessive neurodegeneration; it has no lipid, immune, pulmonary or pleural biology and no connection to statin pharmacology. A loss-of-function burden test on so small a gene will rest on very few carriers, so this is a gene name attached to a signal, not a mechanism.","score_check":"The within-user beta of 3.98 (odds ratio around 50) from a single LOF burden unit, next to a drug-free-population effect of -0.369 +/- 0.224 that is both null and opposite in sign, is the signature of sparse carrier counts rather than a large real effect, so the z_diff of 5.52 is carried entirely by the unstable treated-arm estimate. Without carrier and case counts the log10p of 6.87 should not be taken at face value.","candidability":2,"candidability_reason":"Flat enrichment, a null and oppositely-signed disease arm, an implausible gene and a sparse-data effect size leave confounding by the treated cardiac population as the parsimonious reading.","sources":[{"title":"Simvastatin tablet, film coated - prescribing information (Indications; Post-marketing adverse reactions incl. interstitial lung disease, dyspnea, eosinophilia, lupus erythematosus-like syndrome), DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=751b79c6-bba8-4d29-b69c-a2ac6453fec8"},{"title":"Roncato-Saberan M, Hustache-Mathieu L, Hoen B. Eosinophilic pleural effusion caused by simvastatin after 13 years of exposure. Eur J Intern Med 2006;17(6) (PMID 16962959), Europe PMC record","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID%3A16962959&format=json&resultType=core"},{"title":"Europe PMC title search: statin/simvastatin AND pleural/pleurisy/pleuritis (returns the 2006 simvastatin case report and a 2022 Cancers paper on statin benefit in lung-cancer pleural fluid)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%28statin%20OR%20simvastatin%29%20AND%20TITLE%3A%28pleural%20OR%20pleurisy%20OR%20pleuritis%29&format=json&pageSize=50&resultType=core"},{"title":"Europe PMC title search: drug-induced pleural effusion - Drug-induced eosinophilic pleural effusion, Eur Respir Rev 2011 (PMID 22130825) and Pneumologia 2014 (PMID 25241560); statins not among the implicated classes","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22drug-induced%22%20AND%20TITLE%3A%22pleural%20effusion%22&format=json&pageSize=30&resultType=core"},{"title":"Europe PMC title search: pleural effusion AND heart failure - effusion in 47% of advanced HF (Circ Heart Fail 2024, PMID 39105292) and 40.5% of elderly acute HF admissions (Am J Med 2026, PMID 42276439)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22pleural%20effusion%22%20AND%20TITLE%3A%28%22heart%20failure%22%29&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC title search: pleural effusion aetiology/epidemiology - malignancy, infection and cardiac disease as leading causes; malignant effusion mean age 71.2 (Respiration 2025, PMID 39837310)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22pleural%20effusion%22%20AND%20TITLE%3A%28epidemiology%20OR%20causes%20OR%20aetiology%20OR%20etiology%29&format=json&pageSize=25&resultType=core"},{"title":"UniProt Q9Y530 - ADP-ribose glycohydrolase OARD1 (C6orf130/TARG1): function, ubiquitous expression, recessive neurodegeneration association","url":"https://rest.uniprot.org/uniprotkb/search?query=gene%3AOARD1+AND+organism_id%3A9606&format=json&fields=accession%2Cid%2Cprotein_name%2Cgene_names%2Ccc_function%2Ccc_disease%2Ccc_tissue_specificity"}]},"simvastatin|K579":{"verdict":"co-prescription population","headline":"Diverticular disease is age-shared comorbidity in statin users; statins are neutral to protective, not causal","assessment":"Diverticular disease is neither a licensed indication for simvastatin nor a labelled adverse reaction: the US SPL lists only lipid-lowering and cardiovascular risk-reduction indications, and the word 'diverticul' does not appear anywhere in the label, whose GI adverse reactions are limited to abdominal pain, gastritis, dyspepsia and constipation. The epidemiological literature points the other way from an ADR reading: a Danish nested case-control study (8,809 cases) found a nominal OR of 1.19 for ever-use that became 1.01 (0.85-1.20) once restricted to colonoscoped cases and controls, and the authors attributed the crude excess to diagnostic bias; a Swedish nationwide study (13,127 cases) gave a fully adjusted OR of 1.00 (0.94-1.06) for acute diverticular disease and 0.70 (0.55-0.89) for emergency surgery; a UK GPRD case-control found current statin use associated with reduced risk of diverticular perforation (OR 0.44, 0.20-0.95). Statins are therefore either null or mildly protective for the complicated end of this phenotype, which is decisive evidence against reading a within-users signal as drug-caused injury. The atlas is consistent with that: enrichment of 1.11 means diverticular disease is barely more frequent in simvastatin users than in users of other drugs, exactly what shared age, obesity and Western-diet risk predict for two conditions of the sixth-to-eighth decade, and no FAERS disproportionality could be computed. The 98.9% temporality and 4.13-year median lag carry almost no weight here because the cohort compresses this metric upward and statins are started for a chronic indication that runs for decades alongside ageing bowel. No mechanism connects HMG-CoA reductase inhibition to colonic wall herniation, and the established pathophysiology from GWAS is connective-tissue, extracellular-matrix and enteric neuromuscular rather than lipid-metabolic.","gene_comment":"rs142406975 is a rare intronic variant (gnomAD MAF ~0.6% overall, 0.48% in non-Finnish Europeans) in SEC14L1, a ubiquitously expressed SEC14-like lipid-binding protein known mainly as a negative regulator of RIG-I antiviral signalling and of cholinergic transporters, with no reported role in colonic wall structure, motility or statin pharmacology. SEC14L1 is not among the established diverticular disease loci from the UK Biobank GWAS work (ARHGAP15, FAM155A, COLQ, ELN/EFEMP1-type ECM and neuromuscular genes), and rs142406975 itself has no GWAS Catalog entry, so an intronic hit in it is a positional label rather than a mechanism.","score_check":"The numbers are internally coherent but fragile: beta_adr 2.629 with log10p 6.83 implies z about 5.2 and se about 0.5, i.e. an OR near 14 on a 0.5%-frequency variant estimated from few carriers, short of genome-wide significance and exactly the shape a sparse-data rare-variant fluctuation takes. The disease arm is flatly null (beta 0.129 +/- 0.155, well inside 1.96*se), so the z_diff of 4.77 is driven entirely by the unreplicated within-users estimate rather than by any demonstrated modification of a real disease effect.","candidability":2,"candidability_reason":"Enrichment of only 1.11 plus published null-to-protective statin effects on diverticular disease, with an unreplicated sub-genome-wide rare intronic variant in a gene with no bowel biology.","sources":[{"title":"Simvastatin tablets, film coated - US prescribing information (DailyMed SPL)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=3f6fad0b-0278-433f-86db-761b673a5803"},{"title":"Statins and risk of diverticular disease: nested case-control study (Pharmacoepidemiol Drug Saf 2021, PMID 33583126)","url":"https://europepmc.org/article/MED/33583126"},{"title":"A population-based case-control study on statin exposure and risk of acute diverticular disease (Scand J Gastroenterol 2016, PMID 26357870)","url":"https://europepmc.org/article/MED/26357870"},{"title":"Concurrent drug use and the risk of perforated colonic diverticular disease (Gut 2011, PMID 20940283)","url":"https://europepmc.org/article/MED/20940283"},{"title":"Statin use and risk of acute diverticulitis: a population-based case-control study (Medicine 2020, PMID 32443369)","url":"https://europepmc.org/article/MED/32443369"},{"title":"GWAS of diverticular disease points towards neuromuscular, connective tissue and epithelial pathomechanisms (Gut 2019, PMID 30661054)","url":"https://europepmc.org/article/MED/30661054"},{"title":"UniProt Q92503 - SEC14-like protein 1 (human)","url":"https://rest.uniprot.org/uniprotkb/Q92503"},{"title":"Ensembl variation record for rs142406975 (17:77212716, intronic, gnomAD frequencies)","url":"https://rest.ensembl.org/variation/human/rs142406975?content-type=application/json;pops=1"}]},"simvastatin|M513":{"verdict":"statistical artifact","headline":"Not a statin ADR: simvastatin is studied as disc-regenerative, and the burden effect size is implausible","assessment":"M51.3 (intervertebral disc degeneration) is not an indication for simvastatin, whose SmPC lists only hypercholesterolaemia, mixed dyslipidaemia and cardiovascular prevention, and it is not a labelled adverse reaction: the musculoskeletal section of the Zocor SmPC covers myopathy/myositis, myalgia, muscle cramps, rhabdomyolysis, muscle rupture and tendinopathy, with no spinal, disc or back-pain entry. There is a genuine class-level musculoskeletal signal for statins - Mansi et al. (JAMA Intern Med 2013) reported higher odds of musculoskeletal disease (OR 1.19), strain/sprain injury (1.13) and musculoskeletal pain (1.09) in statin users - so a diffuse statin myalgia workup being coded as degenerative disc disease is a realistic misclassification channel, but that is a coding artifact, not a disc pathology. The direct disc literature runs the opposite way: intradiscal simvastatin stimulates a chondrogenic phenotype partly via BMP-2 and retards or reverses degeneration in rat stab-injury and degenerative-disc models, i.e. the drug is being developed as a disc-anabolic therapy rather than a disc toxin (albeit locally injected in rodents, which does not directly speak to oral systemic dosing). Against that background the atlas numbers do not describe a drug effect: disc degeneration is an age- and load-driven structural diagnosis, statin users are exactly the old, metabolically loaded population that accumulates it, and the enrichment of 0.41 says the condition is if anything less common among simvastatin users than among users of other drugs, so there is no excess to explain. The temporality of 99.2% with a median 4.37 years after first exposure is uninformative here, since the cohort's prescription records reach back further than its diagnosis records and a high value is expected by construction. A FAERS disproportionality of PRR 3.19 for a degenerative spine diagnosis under the most notoriety-laden drug class in spontaneous reporting is not usable causal evidence.","gene_comment":"MEF2D is a real, skeletal-muscle-enhanced MADS-box transcription factor with GWAS hits for pain susceptibility, bone mineral density and height, so it is not an absurd musculoskeletal gene, but it is not among the established disc degeneration loci (ASPN, CHST3, CILP, THBS2, and more recently PARK2, LRP1, AKR1C1) and a myogenic transcription factor supplies no nucleus pulposus mechanism. The signal is a single-unit MPC burden test, so 'MEF2D' here is a gene-level rare-variant aggregate, not a characterised variant.","score_check":"beta_adr of 23.55 on a log scale is not an effect size but the signature of near-complete separation in a sparse single-unit burden test, which makes the log10p of 7.1 and the derived z_diff of 4.86 untrustworthy. The disease arm is honestly null (beta 1.763 with se 0.917, |beta| = 1.92*se, just short of the 1.96 threshold; log10p_disease 1.26), and the prescription arm at 0.84 shows no channelling.","candidability":2,"candidability_reason":"Not a labelled ADR, no excess in treated patients (enrichment 0.41), an implausibly large sparse-burden beta, and a drug literature pointing in the opposite (disc-protective) direction.","sources":[{"title":"Zocor 40 mg film-coated tablets - SmPC (electronic Medicines Compendium)","url":"https://www.medicines.org.uk/emc/product/7789/smpc"},{"title":"PubMed: simvastatin and intervertebral disc degeneration (intradiscal simvastatin retards/reverses degeneration; BMP-2 chondrogenic stimulation)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=simvastatin+intervertebral+disc+degeneration"},{"title":"Mansi et al., Statins and Musculoskeletal Conditions, Arthropathies, and Injuries, JAMA Internal Medicine 2013","url":"https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/1691918"},{"title":"PubMed: atrogin-1/MAFbx and statin muscle toxicity (mechanisms of statin myopathy)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=atrogin-1+statin+muscle+toxicity"},{"title":"UniProtKB Q14814 (MEF2D_HUMAN) - function and expression","url":"https://rest.uniprot.org/uniprotkb/Q14814.txt"},{"title":"GWAS Catalog associations mapped to MEF2D","url":"https://www.ebi.ac.uk/gwas/genes/MEF2D"},{"title":"PubMed: genetics of intervertebral disc degeneration (reported loci)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=%22disc+degeneration%22+genetics+GWAS+loci"}]},"simvastatin|M706":{"verdict":"statistical artifact","headline":"Trochanteric bursitis is not a labelled statin effect; the signal rests on two ultra-rare non-coding variants","assessment":"Trochanteric bursitis (M70.6) is neither an indication for simvastatin (the label covers lipid lowering and cardiovascular risk reduction) nor a labelled adverse reaction: the musculoskeletal terms on the DailyMed simvastatin label are myalgia, myopathy/rhabdomyolysis, muscle cramps, arthralgia, immune-mediated necrotizing myopathy, polymyalgia rheumatica and arthritis, with no bursitis or tendon disorder. What is now called trochanteric bursitis is mostly gluteal (abductor) tendinopathy in middle-aged women, and its risk factors -- age, female sex, adiposity, impaired sleep -- overlap almost exactly with the population that gets a statin. A statin-tendinopathy link is genuinely contested: a 2025 systematic review plus case report proposes extracellular-matrix and membrane effects but found only three heterogeneous cohort studies, and a 2025 Mendelian randomization study of Achilles tendinopathy found no causal effect of statin therapy in either direction. The single best-powered epidemiological test here is a 2024 cohort of 10,301 bursitis cases vs 44,608 controls: hyperlipidemia itself raised bursitis risk (OR 1.24) and statin use showed only a marginal OR of 1.12, which did not persist within hyperlipidemic patients -- i.e. the apparent statin association tracked the indication, not the drug. The FAERS PRR of 2.11 is the expected notoriety signal for a musculoskeletal complaint reported against a drug famous for muscle side effects, and the atlas cannot test confounding here because enrichment is null (no background arm). The temporality of 96.3% over only 135 dated cases is exactly the direction the cohort's record-depth artefact pushes, so it adds little.","gene_comment":"No gene is assigned, and correctly so: rs117514239 (chr8:98,358,868, GRCh38, MAF ~0.0022) is a regulatory-region variant inside the very large STK3 intron, and rs77119758 (chr6:156,588,192, MAF ~0.0018) is intergenic in a gene desert whose nearest annotated feature is an unnamed lncRNA ~44 kb away. Neither locus has any known connection to tendon or bursal biology, statin pharmacokinetics or muscle toxicity, and STK3 proximity should not be dressed up as a Hippo-pathway mechanism.","score_check":"beta_adr of 4.241 is an odds ratio near 70 for variants with MAF ~0.002 in a case set of roughly 135, where fewer than one carrier is expected by chance -- the classic sparse-data blow-up rather than a real effect. The comparison arms are all null (beta_disease -0.174 +/- 0.171 is well inside noise, log10p_disease 0.51, log10p_prescribed 0.82), so z_diff 5.26 is driven entirely by the one unstable estimate it is contrasting against.","candidability":2,"candidability_reason":"Not a labelled effect, best epidemiology attributes bursitis to hyperlipidemia rather than statins, MR is null, and the genetics are an implausibly large effect on two ultra-rare non-coding variants.","sources":[{"title":"DailyMed - SIMVASTATIN tablet, film coated (full prescribing information: indications, adverse reactions)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=b2d6c7ac-304b-4cba-974c-d9a23a069925"},{"title":"Risk Factors for the Development of Olecranon Bursitis - A Large-Scale Population-Based Study (J Clin Med 2024; PMID 39768728)","url":"https://doi.org/10.3390/jcm13247801"},{"title":"Shoulder Tendinopathy Induced by Statins: A Case Report and Systematic Review (J Pers Med 2025; PMID 40423069)","url":"https://europepmc.org/articles/PMC12112834"},{"title":"Investigating the Controversy Surrounding Statin Therapy and Achilles Tendinopathy Using Mendelian Randomization (Int J Clin Pharm 2025; PMID 40053301)","url":"https://europepmc.org/articles/PMC12432074"},{"title":"Europe PMC search: greater trochanteric pain syndrome as gluteal tendinopathy and its risk factors (Donati 2026 PMID 40973929; Minetto 2025 PMID 41416075)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22greater%20trochanteric%20pain%20syndrome%22%20AND%20%22gluteal%20tendinopathy%22&format=json&pageSize=10&resultType=core"},{"title":"Ensembl REST variation lookup for rs117514239 and rs77119758 (position, consequence, MAF)","url":"https://rest.ensembl.org/variation/human/rs117514239?content-type=application/json"}]},"simvastatin|N328":{"verdict":"co-prescription population","headline":"Bladder disorders are not a simvastatin ADR; trial data are null and the variant is a rare intronic outlier","assessment":"N32.8 (other specified disorders of bladder) is not a licensed indication for simvastatin, whose label covers hyperlipidaemia, familial hypercholesterolaemia and cardiovascular risk reduction only, and whose adverse-reaction section lists no bladder or discrete urinary event - the only urinary term is myoglobinuria as a consequence of rhabdomyolysis. The class literature runs the other way or nowhere: the REDUCE trial found statin use unassociated with either incident LUTS (HR 1.05, 95% CI 0.78-1.41) or LUTS progression (HR 1.13, 95% CI 0.96-1.33), and most observational cohorts report protective or null associations with BPH/LUTS, so a drug-caused mechanism for bladder dysfunction is not described and is not supported by the strongest design available. The one supporting strand is disproportionality: FAERS data mining across 2.7 million reports found small but consistent signals for voiding (ROR 1.16, 95% CI 1.10-1.23) and storage (ROR 1.25, 95% CI 1.20-1.30) symptoms in statin users, with no proposed mechanism - and the atlas's own FAERS PRR of 1.36 sits in exactly that band, i.e. the same weak, confounding-prone spontaneous-report signal rather than independent confirmation. The atlas enrichment of 1.28 is precisely what shared risk factors predict: simvastatin users are older and cardiometabolic, and diabetes, obesity and prostatic enlargement all drive N32.8 coding, so the excess is explained by who gets the drug rather than by the drug. Temporality of 98% over 557 dated participants with a median 3.75 years after first exposure is uninformative here because the cohort's prescription records systematically pre-date its diagnosis records; only a low value would have carried weight. The comparison arms are the most telling part: the variant does nothing to the condition in the drug-free population (log10p 0.85; beta 0.214 +/- 0.146, well inside noise) and nothing to being prescribed the drug (log10p 0.26), so the entire finding rests on a within-users effect at a variant with no biological anchor.","gene_comment":"RN7SL278P is an SRP-RNA pseudogene at 15q24.3 with no known function, and rs143118039 (chr15:76,974,556, GRCh38) is a rare intronic SNP (MAF ~0.27%) that actually falls in an unnamed lncRNA, ~19 kb from PSTPIP1 and ~20 kb from RCN2 - neither with any urological role. The gene assignment carries no mechanistic content and should not be presented as one.","score_check":"A beta of 2.276 (OR ~10) at a 0.27% variant is the classic shape of sparse-data inflation, and log10p 6.79 does not immunise it: a handful of carriers can produce this. The null disease arm and null prescription arm are internally consistent, but they mean the z_diff of 4.5 is driven entirely by the fragile treated-arm estimate.","candidability":2,"candidability_reason":"Not a labelled ADR, randomised evidence is null-to-protective for statins on urinary symptoms, the enrichment is comorbidity of the cardiometabolic treated population, and the genetic hit is a rare intronic variant assigned to a pseudogene.","sources":[{"title":"Ezetimibe and Simvastatin Tablets - full prescribing information (DailyMed)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=5ea3ca31-cfee-4978-8b7a-755986f6b091"},{"title":"Statin-associated lower urinary tract symptoms: data mining of the public version of FAERS (Int J Clin Pharmacol Ther, 2014)","url":"https://pubmed.ncbi.nlm.nih.gov/24472404/"},{"title":"Statin Use and Lower Urinary Tract Symptoms Incidence and Progression in the REDUCE Trial (2022)","url":"https://pubmed.ncbi.nlm.nih.gov/34544265/"},{"title":"PubMed search: statin and lower urinary tract symptoms","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+lower+urinary+tract+symptoms"},{"title":"HGNC record for RN7SL278P (RNA, 7SL, cytoplasmic 278, pseudogene; 15q24.3)","url":"https://rest.genenames.org/fetch/hgnc_id/46294"},{"title":"Ensembl variation record for rs143118039","url":"https://rest.ensembl.org/variation/human/rs143118039?content-type=application/json"},{"title":"Ensembl genes overlapping chr15:76,944,556-77,004,556 (GRCh38)","url":"https://rest.ensembl.org/overlap/region/human/15:76944556-77004556?feature=gene;content-type=application/json"}]},"simvastatin|N390":{"verdict":"statistical artifact","headline":"UTI is flat in randomised simvastatin data; the DUSP9 burden effect size is a separation artifact","assessment":"Urinary tract infection is not a licensed indication for simvastatin, whose ZOCOR label covers cardiovascular risk reduction and hyperlipidaemia only. UTI does appear in the label's adverse-reaction table, but as a background 4S trial event with essentially identical rates on drug and placebo (3.2% simvastatin vs 3.1% placebo) - a randomised null, and almost exactly the 3.73% co-occurrence the atlas reports. No direct mechanism links HMG-CoA reductase inhibition to urinary infection; the statin renal signal in the label is acute kidney injury secondary to myoglobinuria from rhabdomyolysis, which is not infectious. If anything the infection literature points the other way, with sepsis meta-analyses reporting neutral to lower mortality in statin users (heavily healthy-user confounded), so an ADR reading has no directional support. The only indirect chain is statin-induced new-onset diabetes (a 9-13% relative increase in observational meta-analysis, and acknowledged on the label as raised HbA1c and fasting glucose) feeding the well-documented diabetic predisposition to UTI - a second-order route that the flat 4S UTI comparison already argues against, and that would in any case be confounding by a downstream comorbidity rather than a direct reaction. The modest enrichment of 1.23 is what an older, more diabetic, more catheterised statin population would produce on its own.","gene_comment":"DUSP9 is a real Xq28 MAP-kinase phosphatase and a replicated common-variant type 2 diabetes GWAS locus, so it is metabolically interesting, but nothing connects it to urinary infection, and the signal here is a rare missense (MPC) burden test rather than the known GWAS SNP. An X-linked gene is also a standard source of burden-test artefact when dosage or sex coding is imperfect.","score_check":"beta_adr of 14.676 is not a credible log-odds - it corresponds to an odds ratio in the millions and is the signature of near-complete case separation on a handful of rare carriers, which also inflates z_diff of 4.21. The disease arm is borderline at best (beta 1.728 with se 0.857, p about 0.04), prescription propensity is null (log10p 0.08), and temporality of 90.4% is uninformative given the cohort's known upward compression.","candidability":1,"candidability_reason":"Randomised label data show no UTI excess, no mechanism exists, and the reported effect size is a sparse-data separation artifact.","sources":[{"title":"ZOCOR (simvastatin) prescribing information, DailyMed","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=8f55d5de-5a4f-4a39-8c84-c53976dd6af9"},{"title":"PubMed: statins and urinary tract infection","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statins+urinary+tract+infection"},{"title":"PubMed: statin therapy, infection risk and sepsis meta-analyses","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+infection+risk+sepsis+randomized+meta-analysis"},{"title":"PubMed: statins and new-onset diabetes meta-analyses","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+new-onset+diabetes+meta-analysis+randomized"},{"title":"PubMed: diabetes mellitus as a risk factor for urinary tract infection","url":"https://pubmed.ncbi.nlm.nih.gov/?term=diabetes+mellitus+risk+of+urinary+tract+infection+cohort"},{"title":"PubMed: DUSP9 and type 2 diabetes","url":"https://pubmed.ncbi.nlm.nih.gov/?term=DUSP9+type+2+diabetes"}]},"simvastatin|N40":{"verdict":"co-prescription population","headline":"BPH is no statin adverse effect; statins run neutral-to-protective, SPRED1 burden is a sparse artifact","assessment":"Benign prostatic hyperplasia (N40) is not a licensed indication for simvastatin, whose approved use is lipid lowering and cardiovascular risk reduction, and no prostatic or BPH term appears anywhere in the label's adverse reactions or postmarketing list (the only urogenital entry is erectile dysfunction). The direction of the published literature runs against an ADR reading: a 2019 meta-analysis of 11 studies and 49,128 men found no increase in BPH incidence with statins and a protective effect in men over 60, an atorvastatin trial and several cohorts report benefit, and the REDUCE trial analysis found simply no association (HR 1.05 for incidence, 1.13 for progression). The atlas enrichment of 1.71 is exactly what confounding by indication predicts: BPH and dyslipidaemia share an ageing male, cardiometabolic population, and metabolic syndrome is repeatedly reported as positively associated with prostate volume and symptom severity, so the men who get simvastatin are the men who get diagnosed with BPH. Temporality of 98.9% over 1,315 dated participants carries almost no weight here because the cohort structurally compresses it upward, and a median of 4.45 years to a slowly accumulating age-related diagnosis is uninformative. The within-user signal (log10p 6.77) sits on top of a completely null drug-free disease test (beta 0.814 +/- 1.304, log10p 0.27), so the z_diff of 4.81 is driven entirely by an implausible treated-arm effect rather than by any modification of a real predisposition. Nothing in FAERS supports it either (counts too low to compute a PRR), and the BNF flag is zero.","gene_comment":"SPRED1 is a real gene - a negative regulator of RAS-MAPK signalling whose loss causes Legius syndrome - but it has low tissue specificity, no published link to prostate growth or BPH, and here it is only an MPC burden hit with no named variants. A burden beta of 22 on the log-odds scale is not a biological effect size; it is the signature of a handful of carriers falling on one side of the outcome.","score_check":"The numbers are internally consistent but self-refuting: a beta of 22 from a rare-variant burden test with a null disease-arm effect (0.814 +/- 1.304, well inside noise) is a sparse-data separation artifact, not a measurable odds ratio. The only believable quantities in the sheet are the descriptive ones - 3.72% co-occurrence and 1.71-fold enrichment - and those describe the treated population, not a drug effect.","candidability":1,"candidability_reason":"Not a labelled or literature-supported statin adverse effect, direction of evidence is neutral-to-protective, enrichment is explained by shared cardiometabolic ageing, and the genetic signal is a sparse burden artifact in a gene with no prostate biology.","sources":[{"title":"PubMed search: statins and benign prostatic hyperplasia / lower urinary tract symptoms","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statins+benign+prostatic+hyperplasia+lower+urinary+tract+symptoms"},{"title":"The effects of statins on benign prostatic hyperplasia and the lower urinary tract symptoms: A Meta-analysis (Medicine 2019, PMID 31045838, PMC6504530)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:31045838&resultType=core&format=json"},{"title":"Statin Use and Lower Urinary Tract Symptoms Incidence and Progression in the REDUCE Trial (J Urol 2022, PMID 34544265)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:34544265&resultType=core&format=json"},{"title":"openFDA simvastatin prescribing information (indications and postmarketing adverse reactions)","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22simvastatin%22&limit=1"},{"title":"UniProt Q7Z699 - SPRED1, Sprouty-related EVH1 domain-containing protein 1","url":"https://rest.uniprot.org/uniprotkb/Q7Z699.txt"},{"title":"Europe PMC search: metabolic syndrome / dyslipidaemia and benign prostatic hyperplasia","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22metabolic%20syndrome%22%20AND%20%22benign%20prostatic%20hyperplasia%22%20AND%20dyslipidemia&resultType=core&format=json&pageSize=5"}]},"simvastatin|N951":{"verdict":"statistical artifact","headline":"Menopause is a life stage, not a statin ADR; the SPRTN burden beta of 69 is a sparse-data artifact","assessment":"ICD-10 N95.1 (\"Menopausal and female climacteric states\") is a code for the physiological climacteric and its vasomotor symptoms - hot flushes, sweats, insomnia - applied in gynaecology and primary care, not a drug injury. The US simvastatin label lists hyperlipidaemia and cardiovascular risk reduction as the only indications, and its adverse-reaction section (myopathy/rhabdomyolysis, hepatic dysfunction, GI and CNS effects, post-marketing erectile dysfunction) contains no mention of menopause, climacteric symptoms, flushing or any oestrogen effect; flushing is a niacin, not a statin, phenomenon. A drug-caused mechanism is not merely undescribed but incoherent: simvastatin cannot induce menopause, and the only reproductive-endocrine signal in the literature runs the other way - statin exposure has been reported to lower gonadotropins in postmenopausal women, i.e. blunting rather than producing climacteric endocrinology. The expected confounding here is co-prescription: oestrogen loss at menopause deteriorates the lipid profile and raises cardiovascular risk, so more women acquire a statin indication around the same years the N95.1 code is applied. The atlas does not even show that - cotx_pct is 0.17% with enrichment 0.35, i.e. this code is three-fold depleted among simvastatin users relative to users of other drugs, consistent with N95.1 being a young-perimenopausal, female-only, gynaecology-clinic code sitting in a mostly older and heavily male statin population. Temporality of 58.8% on only 85 dated participants, with a median 1.9 years after first exposure, is exactly the weak, upward-compressed pattern the cohort produces by construction and adds nothing; the FAERS arm is a non-signal (PRR 1.51).","gene_comment":"SPRTN is a genuine protein-coding gene - a DNA-dependent metalloprotease that clears DNA-protein crosslinks, with biallelic loss causing the Ruijs-Aalfs progeroid/hepatocellular-carcinoma syndrome - so the annotation is not a lncRNA or intergenic guess, but it has no described role in lipid handling, statin pharmacokinetics, sex-steroid synthesis or ovarian ageing. A single MPC burden mask in a stratum this small carries no interpretable mechanistic content.","score_check":"The numbers do not hang together: a burden beta of 69.3 is not an effect size but quasi-complete separation in a handful of carriers, and the disease arm is frankly null (beta 2.781 +/- 2.442, |beta| < 1.96*se, log10p 0.59), so the z_diff of 4.98 is manufactured entirely by the unstable within-users estimate. With ~85 dated cases, no prescription-propensity signal (log10p 0.05) and a depleted co-occurrence, log10p_adr 6.88 should be read as sparse-data instability rather than evidence.","candidability":0,"candidability_reason":"Definitionally not an adverse reaction - a physiological life stage - and the supporting genetics are a separation artifact with a null disease arm.","sources":[{"title":"SIMVASTATIN tablet, film coated - full prescribing information (DailyMed, setid 751b79c6-bba8-4d29-b69c-a2ac6453fec8)","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=751b79c6-bba8-4d29-b69c-a2ac6453fec8"},{"title":"ICD-10-CM N95.1 \"Menopausal and female climacteric states\" (NLM Clinical Table Search Service)","url":"https://clinicaltables.nlm.nih.gov/api/icd10cm/v3/search?sf=code,name&terms=N95.1"},{"title":"Europe PMC: N95.1 climacteric coding and symptom content (incl. Prescription patterns of herbal medicine for menopausal disorders, PMID 33665094)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22N95.1%22%20AND%20%22menopausal%20and%20female%20climacteric%20states%22&format=json&pageSize=5&resultType=core"},{"title":"Europe PMC: menopausal transition, oestrogen deficiency and lipid-profile deterioration (Fasero & Coronado, J Clin Med 2025, PMID 40507425)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22menopause%22%20AND%20%22lipid%20profile%22%20AND%20%22LDL%20cholesterol%22%20AND%20transition&format=json&pageSize=8&resultType=core"},{"title":"Europe PMC: statins and reproductive endocrinology - rosuvastatin potentiates gonadotropin lowering in postmenopausal women (Pharmacology 2023, PMID 36791677)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=statin%20AND%20%22anti-Mullerian%20hormone%22%20OR%20(statin%20AND%20%22ovarian%20reserve%22)&format=json&pageSize=8&resultType=core"},{"title":"Europe PMC: SPRTN as a DNA-protein-crosslink metalloprotease and ageing/genome-stability gene (incl. Nat Commun 2025, PMID 40691134; Science 2026, PMID 41610251)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=SPRTN%20AND%20(function%20OR%20syndrome)&format=json&pageSize=10&resultType=core"}]},"simvastatin|R030":{"verdict":"co-prescription population","headline":"Statins modestly lower BP; R03.0 in simvastatin users is background cardiometabolic coding, not an ADR","assessment":"R03.0 (elevated blood-pressure reading without a diagnosis of hypertension) is neither a licensed indication for simvastatin nor a labelled adverse reaction: the FDA label indicates simvastatin for LDL-C lowering and cardiovascular risk reduction, and hypertension/raised blood pressure appears nowhere in its clinical-trial or post-marketing adverse-reaction sections. A drug-caused mechanism in this direction is not merely undescribed but contradicted: three independent meta-analyses of randomised placebo-controlled trials report that statins lower blood pressure by roughly 1.4-1.9 mmHg systolic and 0.8-1.0 mmHg diastolic (Alghamdi 2020, 46 RCTs, SBP -1.6 mmHg [-2.50, -0.60]; Strazzullo 2007, 20 RCTs, SBP -1.9 mmHg [-3.8, -0.1]; Liu 2023, SBP -1.42 [-2.38, -0.46]), attributed to pleiotropic endothelial effects. That the drug moves the phenotype in the opposite direction is decisive against reading this as an adverse reaction. What remains is a population effect: simvastatin is prescribed to a cardiometabolic cohort in which incidental raised readings are routine, and the atlas itself shows no drug-specific excess (cotx_pct 1.24%, enrichment 1.01, i.e. identical to users of other drugs), while the low log10p_prescribed (0.82) says the variant does not predict being prescribed the drug either. The 83.3% temporality and 2.64-year median lag are exactly what the cohort's record-depth asymmetry manufactures for any common primary-care code and carry no weight here. The FAERS 'STRONG' flag with PRR 2.13 is best explained by hypertension being a near-universal co-reported comorbidity on statin reports rather than a signal of causation.","gene_comment":"ACOT12 is a cytosolic, liver-enriched acetyl-CoA hydrolase (UniProt Q8WYK0) with no described role in vascular tone or blood-pressure regulation; a PubMed search for ACOT12 with blood pressure or hypertension returns only two incidental hits (a sheep genome-selection scan and a mouse PCOS-offspring study), neither implicating it. A rare-LOF burden in a hepatic lipid-metabolism thioesterase is not a credible mechanism for an elevated blood-pressure reading.","score_check":"The ADR effect of beta 6.3 for a single-gene LOF burden is a textbook sparse-data separation artifact - an odds ratio of ~550 is not a real effect size for a common symptom code - and the z_diff of 4.08 is driven entirely by that inflated beta, while the drug-free arm is only nominal (beta 1.042 +/- 0.423, z = 2.46, log10p 1.86). The population numbers, by contrast, are believable and unhelpful to the hypothesis: enrichment 1.01 means no excess over other drugs' users at all.","candidability":1,"candidability_reason":"Randomised evidence has statins lowering blood pressure, the label lists no such reaction, enrichment is 1.01, and the genetic signal is an implausibly large LOF burden effect in a gene with no vascular biology.","sources":[{"title":"openFDA drug label API - simvastatin (indications and adverse reactions)","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22simvastatin%22&limit=1"},{"title":"Alghamdi et al. Blood pressure-lowering activity of statins: systematic review and meta-analysis of placebo-randomized controlled trials (PMID 32696233)","url":"https://pubmed.ncbi.nlm.nih.gov/32696233/"},{"title":"Strazzullo et al. Do statins reduce blood pressure? A meta-analysis of randomized, controlled trials (PMID 17309949)","url":"https://pubmed.ncbi.nlm.nih.gov/17309949/"},{"title":"Liu et al. Statin's role on blood pressure levels: meta-analysis based on randomized controlled trials (PMID 36799888)","url":"https://pubmed.ncbi.nlm.nih.gov/36799888/"},{"title":"UniProtKB Q8WYK0 (ACOT12_HUMAN) - acetyl-CoA thioesterase, function and tissue specificity","url":"https://rest.uniprot.org/uniprotkb/Q8WYK0.txt"},{"title":"PubMed search: ACOT12 AND (blood pressure OR hypertension) - 2 incidental records","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=pubmed&term=ACOT12+AND+(blood+pressure+OR+hypertension)&retmode=json"}]},"simvastatin|R13":{"verdict":"plausible ADR","headline":"Statin myopathy can present as dysphagia, but R13 here looks age/stroke-driven and the variant is ultra-rare","assessment":"Dysphagia (R13) is not an indication for simvastatin and appears nowhere in the US label, whose indications are limited to cardiovascular risk reduction and lipid lowering; it is also not listed among its adverse reactions. A drug-caused mechanism nevertheless exists and is described: statin-associated myopathy, and in particular anti-HMGCR immune-mediated necrotising myopathy (IMNM, which the label does carry as a rare warning), can present with oropharyngeal dysphagia, documented in case reports in Ugeskr Laeger 2010, BMJ Case Rep 2020 and Cureus 2023, with resolution on withdrawal in at least one; a small controlled study (J Voice 2020, 75 statin users vs 85 controls) found higher swallowing- and voice-symptom scores in statin users. The drug is not known to move dysphagia in the opposite direction, so nothing decisively refutes an ADR reading. Against that, R13 is a coarse EHR code whose burden in this population is dominated by age, stroke and neurodegenerative disease - all far commoner in statin users than IMNM, which is a per-100,000 event - and simvastatin is prescribed exactly to older post-stroke and cardiovascular patients, so the 0.99% co-occurrence and 4.23-year median lag are as compatible with accruing age and cerebrovascular risk as with drug injury. The mild enrichment (1.17) argues the code is only slightly commoner in simvastatin users than in users of other drugs, and the 99.6% temporality is uninformative given the cohort's known upward compression. FAERS support is weak (1/3, PRR 1.63) and the association carries no BNF listing.","gene_comment":"rs575681638 is an intronic variant in ABTB3/BTBD11 (chr12:107,507,573, within the gene's 107.32-107.66 Mb span) with a minor allele frequency of about 0.04% in gnomAD/TOPMed; ABTB3 has no known role in skeletal muscle, HMG-CoA reductase signalling, or statin disposition, and none of the established statin-myopathy loci (SLCO1B1, ABCG2, CYP3A4, HLA-DRB1*11:01 for IMNM) is implicated here, so the gene assignment is positionally correct but mechanistically empty.","score_check":"A beta of 2.494 (OR ~12) at MAF 4e-4 in a cohort with only 535 dated dysphagia records means the signal rests on a handful of carriers, which is the classic sparse-data signature rather than a large real effect, and the reported log10p of 7.28 should be treated as unreliable until carrier counts and a Firth/exact test are shown. The disease arm is properly null (beta 0.049 +/- 0.129, |beta| far below 1.96*se) and the prescription arm is flat, so z_diff 5.14 is driven entirely by the fragile within-user estimate.","candidability":4,"candidability_reason":"The drug-condition pair has a real but very rare described mechanism (statin myopathy/IMNM with dysphagia), yet the R13 phenotype in this population is mostly age- and stroke-driven and the driving variant is an ultra-rare intron in a gene with no relevant biology.","sources":[{"title":"Simvastatin tablet, film coated - FDA label (DailyMed, setid b2d6c7ac)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=b2d6c7ac-304b-4cba-974c-d9a23a069925"},{"title":"PubMed records: statin-induced dysphagia and statin-associated necrotising autoimmune myopathy (PMIDs 20156405, 32029513, 37719573) and The Prevalence of Dysphonia and Dysphagia Symptoms in Patients on Statin Therapy, J Voice 2020 (PMID 31375400)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+dysphagia"},{"title":"Ensembl REST: variation rs575681638 (chr12:107,507,573, intron, MAF 0.00039)","url":"https://rest.ensembl.org/variation/human/rs575681638?content-type=application/json"},{"title":"Ensembl REST: ABTB3 (ankyrin repeat and BTB domain containing 3), chr12:107,318,421-107,659,642","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/ABTB3?content-type=application/json"},{"title":"PubMed: prevalence of oropharyngeal dysphagia in older adults and post-stroke (PMIDs 34903688, 38643757)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=oropharyngeal+dysphagia+prevalence+older+adults+stroke"}]},"simvastatin|R31":{"verdict":"co-prescription population","headline":"COL4A4 truncating variants cause haematuria with or without simvastatin; z_diff 1.59 shows no drug interaction","assessment":"Haematuria is not an indication for simvastatin, whose licensed uses on the ZOCOR label are reduction of CHD mortality and cardiovascular events and treatment of hyperlipidaemia. Nor is haematuria a listed adverse reaction: the only renal/urinary entries are acute renal failure secondary to myoglobinuria in rhabdomyolysis and, within the rare hypersensitivity syndrome, vasculitis and purpura. A genuine but uncommon drug pathway does exist - statin-induced rhabdomyolysis produces myoglobinuria, which turns the urine dipstick strongly blood-positive with few or no red cells and is readily coded as unspecified haematuria (R31) - and this is the most credible reading of the FAERS PRR of 4.21; but that presentation is acute and dose-related, not something appearing a median 3.56 years after first exposure. The published statin-haematuria epidemiology is essentially a rosuvastatin story: the JASN 2022 cohort of 947,000 new statin users found rosuvastatin carried a hazard ratio of only 1.08 for haematuria versus atorvastatin, attributed to rosuvastatin dosing in advanced CKD rather than to a class effect that would implicate simvastatin. The atlas numbers point the same way: enrichment of 1.75 says haematuria is simply commoner in statin-treated patients, who are older, more often anticoagulated, more likely to have CKD or prostate disease and far more likely to have a urine dipstick done at all. The 98.6% temporality is uninformative given that prescription records predate diagnosis records in this cohort, and the predisposition flag of 100% is the honest summary of what the variant is doing here.","gene_comment":"COL4A4 is about as strong a haematuria gene as exists - heterozygous truncating variants cause thin basement membrane nephropathy and autosomal dominant Alport syndrome, and UK Biobank exome analyses show truncating COL4A4 variants are robustly associated with haematuria and proteinuria in the general population - so the gene assignment is excellent but points to constitutional glomerular basement membrane disease, not to any statin mechanism.","score_check":"The numbers are internally consistent and not artifactual: beta_disease 3.197 with se 0.279 is a solid effect, and log10p_disease 29.63 dwarfs the within-user log10p_adr 7.35, which is what a drug-independent predisposition locus looks like when it is re-tested in a subset. The decisive figure is z_diff 1.59, well short of significance, so the atlas itself provides no evidence that simvastatin modifies the COL4A4 effect.","candidability":3,"candidability_reason":"Real gene, real condition, but the effect is a well-known drug-independent Mendelian predisposition re-detected inside a treated population that is already 1.75-fold enriched for haematuria, with no interaction signal.","sources":[{"title":"COL4A4 gene - MedlinePlus Genetics","url":"https://medlineplus.gov/genetics/gene/col4a4/"},{"title":"ZOCOR (simvastatin) US prescribing information, FDA Drugs@FDA label","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2012/019766s085lbl.pdf"},{"title":"Shin JI et al., Association of Rosuvastatin Use with Risk of Hematuria and Proteinuria, JASN 2022 (PMID 35853713)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:35853713&resultType=core&format=json"},{"title":"Population-scale genomics reveals divergent pathogenicity of variant classes across paralogous collagen IV genes (UK Biobank, COL4A3/COL4A4 truncating variants and haematuria)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22COL4A4%22%20AND%20%22haematuria%22%20AND%20biobank&resultType=core&format=json&pageSize=5"},{"title":"Autosomal Dominant Alport Syndrome, JASN 2026 (PMID 42172079) - heterozygous COL4A3/COL4A4 variants affect ~1% of the population, two thirds with haematuria","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22COL4A3%22%20OR%20%22COL4A4%22%20AND%20%22exome%22%20AND%20%22hematuria%22&resultType=core&format=json&pageSize=5"},{"title":"Gross and Microscopic Hematuria, StatPearls (PMID 30480952) - dipstick-positive blood without red cells suggests myoglobinuria (pseudohaematuria)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22dipstick%22%20AND%20%22myoglobinuria%22%20AND%20%22hematuria%22&resultType=core&format=json&pageSize=4"}]},"simvastatin|R600":{"verdict":"statistical artifact","headline":"Localised oedema is not a simvastatin effect; the 2q11.2 pseudogene CNV signal is a sparse-data artifact","assessment":"R60.0 localised oedema is a symptom code (leg/limb swelling), not a licensed indication for simvastatin and not a recognised statin adverse effect. In the pivotal 4S trial reproduced in the ZOCOR label, 'edema/swelling' occurred in 2.7% of simvastatin patients versus 2.3% on placebo, i.e. no meaningful excess, and StatPearls' adverse-effect list for simvastatin does not mention oedema at all; the only oedema-type reaction in the label is angioedema, listed under rare post-marketing hypersensitivity syndrome, which is a different phenotype coded elsewhere (T78.3/D84.1) and supported only by isolated case reports across the class. The atlas numbers agree with that: enrichment is 0.90, so localised oedema is if anything slightly less common in simvastatin users than in users of other drugs, and cotx_pct is only 0.93%. The far more likely driver of leg swelling in this exact population is co-medication rather than the statin - dihydropyridine calcium-channel blockers such as amlodipine roughly triple peripheral oedema in meta-analysis and are prescribed to the same cardiovascular patients - together with age and venous insufficiency, which also explains the 87% temporality and 3.85-year median lag without any drug effect, since statins are started in midlife and leg oedema accrues later. The FAERS 'STRONG' flag with PRR 4.0 is weak support here because 'oedema peripheral' is one of the most heavily reported terms in spontaneous data and is massively co-reported with the antihypertensives statin users take. The one honest negative-control result is log10p_prescribed = 0.05, showing no prescription-propensity confounding at these loci - but that does not rescue an effect size that is not believable in the first place.","gene_comment":"SPRTN is a real gene (DNA-protein crosslink metalloprotease; biallelic loss causes Ruijs-Aalfs progeroid syndrome) with no lymphatic, venous or vascular-permeability biology that could produce localised oedema, and an MPC burden test on ~0.9% cases rests on very few carriers. The second hit, chr2:97,000,001-97,100,001, is pericentromeric 2q11.2 and contains no credible gene at all - per Ensembl it holds only novel lncRNAs, TRIM43CP and other unprocessed pseudogenes, orphan IGKV pseudogenes and the duplicated FAHD2B - exactly the segmental-duplication-rich territory where read-depth CNV calls are unreliable.","score_check":"beta_adr = 46.53 is not a plausible effect size for any real variant on a 0.93%-prevalence symptom code and is the signature of quasi-separation in a sparse burden/CNV test, so log10p_adr = 9.46 and z_diff = 5.52 are inherited from that artifact rather than from a measured effect. The disease arm is frankly null (log10p_disease 0.97; beta 3.689 +/- 2.285, well inside 1.96*se), so there is no predisposition signal to contrast against either.","candidability":1,"candidability_reason":"Not a label ADR, no excess over placebo in 4S, depleted rather than enriched in users, and the genetics are an implausible beta over a pericentromeric pseudogene CNV region.","sources":[{"title":"Simvastatin - StatPearls (NCBI Bookshelf, NBK532919): adverse effects list","url":"https://www.ncbi.nlm.nih.gov/books/NBK532919/"},{"title":"ZOCOR (simvastatin) prescribing information, DailyMed SPL 8f55d5de - Adverse Reactions (4S: edema/swelling 2.7% vs 2.3% placebo) and Postmarketing (angioedema under hypersensitivity syndrome)","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=8f55d5de-5a4f-4a39-8c84-c53976dd6af9"},{"title":"PubMed: statin angioedema - case reports (Zgolli 2021 PMID 33676755; Al-Qaaneh 2022 PMID 34779391; Shahbaz 2018 PMID 30210953)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=statin+angioedema"},{"title":"PubMed: amlodipine peripheral edema incidence - meta-analysis of randomised placebo-controlled trials (PMID 31107359) reporting ~3-fold increased peripheral edema","url":"https://pubmed.ncbi.nlm.nih.gov/?term=amlodipine+peripheral+edema+incidence"},{"title":"Ensembl REST overlap, GRCh38 chr2:97,000,001-97,100,001 (2q11.2): lncRNAs, TRIM43CP and other pseudogenes, IGKV orphan pseudogenes, FAHD2B","url":"https://rest.ensembl.org/overlap/region/human/2:97000001-97100001?feature=gene"}]},"simvastatin|R80":{"verdict":"statistical artifact","headline":"Statin tubular proteinuria is real but rosuvastatin-dose-specific; this rare NLGN1 hit is sparse-data noise","assessment":"Isolated proteinuria (R80) is not an indication for simvastatin, and the FDA Zocor label contains no mention of proteinuria at all - its only renal text is acute renal failure secondary to myoglobinuria in rhabdomyolysis. A drug-caused mechanism does exist at class level: HMG-CoA reductase inhibition in proximal tubular cells impairs receptor-mediated endocytosis and hence reabsorption of normally filtered low-molecular-weight proteins, producing transient, dose-dependent, functional (non-nephrotoxic) proteinuria; but this is labelled only for rosuvastatin, where the Crestor label's section 5.4 states dipstick-positive proteinuria was more frequent at 40 mg 'when compared to lower doses of CRESTOR or comparator HMG-CoA reductase inhibitors' - i.e. simvastatin is the comparator in which it was NOT seen. Pushing the other way, meta-analyses in non-end-stage CKD find statins modestly REDUCE urinary albumin excretion and 24-h protein excretion, so the net population-level direction of statin effect on proteinuria is downward, not upward. The atlas numbers give no support for an excess either: cotx_pct is 0.44% with enrichment 0.92, i.e. coded proteinuria is if anything marginally less common in simvastatin users than in users of other drugs, and the prescription-propensity arm is flat (log10p 0.05). The 96% temporality rests on only 25 datable participants with 86.2% undated, so it carries almost no weight, and FAERS is weak (1/3, PRR 1.75). What remains is a single rare variant with an enormous effect estimate and no comparison arm, which is the signature of sparse-cell separation rather than of pharmacology.","gene_comment":"rs139771640 is a rare variant (MAF ~0.3%, GRCh38 chr3:173,819,694) whose most severe consequence in Ensembl is 'non_coding_transcript_exon_variant', falling inside the ~900 kb NLGN1 locus - a synaptic cell-adhesion gene with no renal or tubular biology and no literature linking it to proteinuria or statin response. The assignment is positional only and should not be read as mechanism.","score_check":"beta_adr 4.672 (odds ratio ~100) for a 0.3%-frequency variant against a 0.44% phenotype, reported with no standard error and no drug-free disease arm or z_diff, is the classic profile of a sparse-data / quasi-separation artifact rather than a real effect. Enrichment 0.92 and a null prescribed-propensity arm are internally consistent but simply show no excess of proteinuria in simvastatin users.","candidability":2,"candidability_reason":"Class-level tubular proteinuria is real but is a rosuvastatin high-dose effect not attributed to simvastatin, statins net-reduce proteinuria, the atlas shows no enrichment, and the single rare non-coding NLGN1 variant looks like sparse-data separation.","sources":[{"title":"ZOCOR (simvastatin) FDA prescribing information (no proteinuria listed; renal failure only via myoglobinuria)","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2012/019766s085lbl.pdf"},{"title":"CRESTOR (rosuvastatin) FDA label, section 5.4 Proteinuria and Hematuria","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2010/021366s016lbl.pdf"},{"title":"Goyal K, Soman SS, Bhan A. Transient Proteinuria Induced by High-Dose Rosuvastatin. AACE Endocrinol Diabetes 2025 (PMC12332483)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12332483/"},{"title":"Europe PMC search: statin-induced proteinuria and tubular mechanism (Stouffer 2026 PMC13154092; Schulze 2025 PMC12675985)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=statin%20AND%20proteinuria%20AND%20tubular&format=json&pageSize=25&resultType=core"},{"title":"Europe PMC title search: statin effects on proteinuria (network meta-analysis PMID 31719617; CKD meta-analysis PMID 26776964; 'Statins and proteinuria' PMID 16105477)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%28statins%20AND%20proteinuria%29&format=json&pageSize=30&resultType=core&sort=CITED%20desc"},{"title":"Ensembl REST variation record for rs139771640 (MAF 0.0031, non-coding transcript exon variant, chr3:173819694)","url":"https://rest.ensembl.org/variation/human/rs139771640?content-type=application/json"},{"title":"Ensembl REST gene record for NLGN1 (chr3:173,396,284-174,294,372, neuroligin 1)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/NLGN1?content-type=application/json"}]},"simvastatin|Z501":{"verdict":"statistical artifact","headline":"Z50.1 is a rehabilitation-encounter code, not a disease; signal is a rare intergenic sparse-data artifact","assessment":"Z50.1 is not a clinical condition at all: in WHO ICD-10 it sits in Chapter XXI, block Z40-Z54 'Persons encountering health services for specific procedures and health care', under Z50 'Care involving use of rehabilitation procedures', and its immediate sibling Z50.0 is cardiac rehabilitation. It is therefore neither a licensed indication for simvastatin nor a recognised adverse effect; the ZOCOR label lists only lipid-lowering and cardiovascular risk-reduction indications and its muscle warnings (myopathy, rhabdomyolysis, immune-mediated necrotizing myopathy, the 80 mg dose restriction) never mention physiotherapy or rehabilitation. The only mechanistic thread worth naming is indirect: statin-associated musculoskeletal symptoms are real and pharmacogenetically characterised by CPIC for SLCO1B1, ABCG2 and CYP2C9, so a statin could in principle push some patients toward physiotherapy - but a referral code is a health-service proxy several steps downstream of that, and nothing here tests it. The far more parsimonious reading is confounding by indication and by care pathway: simvastatin users are coronary and cerebrovascular patients who are routinely enrolled in rehabilitation (INTERASPIRE found 87% of CHD patients on lipid-lowering therapy, 50% on high-dose statins), and the atlas cannot rebut this because enrichment is null, meaning no drug-free background for Z50.1 was available. Temporality of exactly 100.0% over only 162 dated participants, with a median 4.04 years after first exposure, is precisely the pattern the cohort's record-depth artifact produces for administrative encounter codes and is weak evidence by the fact sheet's own rule. The comparison arms are all flat - the disease arm gives beta 0.18 +/- 0.111 (|beta| well under 1.96*se, indistinguishable from noise, log10p 0.99) and the prescription-propensity arm is empty at log10p 0.03 - so z_diff 4.71 is generated entirely by an implausible within-user beta rather than by any contrast the data support.","gene_comment":"There is no gene: rs142781610 is an intergenic variant in the 5p14.1 gene desert (chr5:29,231,711, GRCh38) with no assigned transcript, no ClinVar record, no GWAS Catalog entry and no publications, and the atlas itself leaves the gene field empty. Nothing mechanistic can be built on this locus, and in particular it has no relation to the SLCO1B1/ABCG2/CYP2C9 pharmacokinetic genes that actually govern statin muscle toxicity.","score_check":"The numbers do not hang together: beta_adr 3.999 (odds ratio ~55) for a variant with gnomAD alternate frequency ~0.39% (1000G MAF 0.16%) inside a phenotype held by only 0.43% of users implies a handful of carrier cases, the textbook sparse-data setting where effect sizes explode, and log10p_adr 6.19 does not reach genome-wide significance for a rare single SNV. The flat disease and prescription arms confirm there is no reproducible effect for the z_diff to be measuring.","candidability":0,"candidability_reason":"The endpoint is an administrative rehabilitation-encounter code rather than an adverse reaction, and the supporting variant is a rare intergenic locus with an effect size implausible for its standard error.","sources":[{"title":"WHO ICD-10 (2019): Z50 Care involving use of rehabilitation procedures - child concepts including Z50.1 Other physical therapy","url":"https://icd.who.int/browse10/2019/en/JsonGetChildrenConcepts?ConceptId=Z50&useHtml=false&showAdoptedChildren=true"},{"title":"WHO ICD-10 (2019): block Z40-Z54 'Persons encountering health services for specific procedures and health care'","url":"https://icd.who.int/browse10/2019/en/JsonGetChildrenConcepts?ConceptId=Z40-Z54&useHtml=false&showAdoptedChildren=true"},{"title":"ZOCOR (simvastatin) US prescribing information, FDA Drugs@FDA label","url":"https://www.accessdata.fda.gov/drugsatfda_docs/label/2022/019766s102lbl.pdf"},{"title":"Cooper-DeHoff RM et al., CPIC Guideline for SLCO1B1, ABCG2 and CYP2C9 Genotypes and Statin-Associated Musculoskeletal Symptoms, Clin Pharmacol Ther 2022 (PMID 35152405)","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pubmed&id=35152405&rettype=abstract&retmode=text"},{"title":"dbSNP entry for rs142781610 (intergenic, chr5:29,231,711 GRCh38, no ClinVar record)","url":"https://www.ncbi.nlm.nih.gov/snp/rs142781610"},{"title":"Ensembl REST variation record for rs142781610 (most severe consequence: intergenic_variant)","url":"https://rest.ensembl.org/variation/human/rs142781610?content-type=application/json"},{"title":"INTERASPIRE: residual dyslipidaemia and statin use in 4,069 coronary heart disease patients across 13 countries, Atherosclerosis 2025 (PMID 40315644)","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pubmed&id=40315644&rettype=abstract&retmode=text"}]},"simvastatin|Z980":{"verdict":"co-prescription population","headline":"Z98.0 is a post-surgical status code, not an ADR: statin-treated obesity precedes bariatric bypass","assessment":"Z98.0 is not a disease at all: the NLM ICD-10-CM table defines it as 'Intestinal bypass and anastomosis status', a Z-chapter code recording that a patient has had bowel bypass or an anastomosis (post-bariatric anatomy, or an anastomosis after bowel resection), so simvastatin cannot cause it in any literal sense. The US simvastatin label lists only lipid indications (primary hyperlipidemia, HeFH/HoFH, dysbetalipoproteinaemia, hypertriglyceridaemia, CHD mortality reduction) and its adverse reactions are myopathy/rhabdomyolysis, hepatic injury, hyperglycaemia and ordinary GI complaints; it says nothing about intestinal bypass, anastomotic complications or malabsorption. The direction of association is the opposite of an ADR reading: roughly 60% of candidates for metabolic surgery already carry dyslipidaemia and are on statins, and a matched cohort of >16,000 surgical patients found 60.4% of prior statin users stopped within 2 years after surgery versus 35.4% of controls, with far less new initiation (2.6% vs 9.6%) - so bypass status marks a population selected into statin therapy, and if anything surgery reduces statin use. The one genuine pharmacological link runs the other way too: bypassing the duodenum changes statin exposure (simvastatin concentrations rise 33-150% at 3 months and roughly double at 6 months before returning to baseline), i.e. the surgery is an effect modifier of the drug, not an outcome of it. Within the atlas the association is carried by a single MPC burden mask in RNF167 with beta_adr 103.079 - a perfect-separation magnitude, not a real effect size - while the drug-free disease arm (beta 5.475, se 5.93) and the prescription-propensity arm are both null, and enrichment is unavailable, so there is no honest test of whether bypass status is over-represented among simvastatin users. Temporality of 100% over 162 dated participants with a median 5.66 years after first exposure looks striking but is exactly what confounding by indication predicts here, since obese dyslipidaemic patients are typically statinised years before they reach surgery, and the cohort's record structure already biases this metric upward.","gene_comment":"RNF167 (17p13.2) is a genuine protein-coding E3 ubiquitin ligase acting on TSSC5, Rab7, Tollip and RIG-I-like receptors, with no described role in lipid metabolism, statin pharmacokinetics, bowel physiology or surgical anatomy. A single-gene MPC burden hit with no supporting variant-level or disease-arm signal offers no mechanism for a surgical status code.","score_check":"The numbers do not hang together: a burden beta of 103 with log10p 7.18 against ~162 dated cases is the signature of sparse-data separation, and the disease arm's |beta| < 1.96*se makes it indistinguishable from noise, so z_diff 4.89 is driven entirely by the implausible treated-arm estimate. The confidence score of 8.2 is far higher than the evidence supports.","candidability":0,"candidability_reason":"Outcome is a surgical status code, definitionally not an adverse reaction, and the burden effect size is a separation artifact.","sources":[{"title":"NLM Clinical Table Search Service, ICD-10-CM Z98.0 'Intestinal bypass and anastomosis status'","url":"https://clinicaltables.nlm.nih.gov/api/icd10cm/v3/search?sf=code,name&terms=Z98.0&maxList=10"},{"title":"DailyMed: Simvastatin tablet, film coated - full prescribing information (indications, adverse reactions)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=3f6fad0b-0278-433f-86db-761b673a5803"},{"title":"Molecular Mechanisms Affecting Statin Pharmacokinetics after Bariatric Surgery (Int J Mol Sci, 2024; PMC11476770)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC11476770/"},{"title":"Statin use trajectories postbariatric surgery: a matched cohort analysis (Surg Obes Relat Dis; PMID 39379259)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22Statin%20use%20trajectories%20postbariatric%20surgery%22&resultType=core&format=json&pageSize=2"},{"title":"NCBI Gene: RNF167, ring finger protein 167 (Gene ID 26001, 17p13.2)","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=gene&id=26001&retmode=json"},{"title":"Europe PMC search: RNF167 functional literature (Rab7, Tollip, RLR ubiquitination)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=RNF167&format=json&pageSize=15"}]},"temazepam|N390":{"verdict":"statistical artifact","headline":"UTI is not a temazepam ADR; the signal is a rare X-linked testis-lncRNA variant with an absurd effect","assessment":"Temazepam is licensed only for short-term treatment of severe or disabling insomnia and as premedication for minor procedures, so urinary tract infection is neither an indication nor anywhere in the indication space. UTI is also not a listed undesirable effect: the UK SmPC lists urinary retention and incontinence among less common reactions, which offers at most an indirect two-step route (sedation and retention leaving a post-void residual that seeds bacteriuria), but dedicated reviews of drug-induced urinary retention and of lower urinary tract disorders as adverse drug reactions attribute retention to anticholinergics, opioids and anaesthetics rather than to benzodiazepines. The only well-documented benzodiazepine-infection link is pneumonia in older adults, an aspiration and sedation pathway that is respiratory and does not transfer to the urinary tract, and a PubMed and Europe PMC search returns no study reporting benzodiazepine or hypnotic use as a risk factor for UTI. What remains is the ordinary frailty pathway shared by every sedative in an elderly population: immobility, falls, hospitalisation, catheterisation, nocturnal continence loss - a marker of who receives temazepam rather than a pharmacological action of it. The atlas numbers argue against even that reading, because UTI is roughly half as common in temazepam users as in users of other drugs (cotx_pct 1.47, enrichment 0.55), so this cohort is not one in which UTI is being accumulated. The 89.4% temporality is the uninformative direction the atlas itself warns about, and rests on only 104 datable participants. Taken together, the phenotype is not a recognised temazepam reaction, the population is not enriched for it, and the genetic evidence is not of a quality that could rescue either.","gene_comment":"SPANXA2-OT1 is a lncRNA overlapping the testis-specific SPANX cancer/testis antigen cluster on Xq27, and rs782188154 is a rare intronic variant within it (GRCh38 chrX:141,412,320, no reported MAF) - no connection to bladder function, uropathogen defence or benzodiazepine pharmacology. An X-chromosome hit on a strongly sex-skewed phenotype like UTI is an added caution, since residual sex handling on chrX is a classic source of spurious association.","score_check":"beta_adr 5.224 is an odds ratio near 185 for a rare intronic variant, which is the signature of sparse-cell separation rather than a measurable effect, and the disease arm is a clean null (beta 0.018 +/- 0.084) as is the prescription arm, so z_diff 5.2 only restates that the ADR arm is the outlier. The FAERS PRR of 2.57 for UTI on a hypnotic given to frail elderly is exactly where co-reporting and confounding dominate, and it is not corroborated by any label or cohort evidence.","candidability":1,"candidability_reason":"Not a labelled or literature-supported reaction, depleted rather than enriched in treated patients, and driven by an implausibly large effect at a rare intronic variant in a testis-specific X-linked lncRNA.","sources":[{"title":"Temazepam 10mg/5ml Oral Solution - Summary of Product Characteristics (emc)","url":"https://www.medicines.org.uk/emc/product/9479/smpc"},{"title":"PubMed search: benzodiazepine AND urinary tract infection","url":"https://pubmed.ncbi.nlm.nih.gov/?term=benzodiazepine+urinary+tract+infection"},{"title":"PubMed search: drug-induced urinary retention reviews (Verhamme 2008 Drug Saf; Dobrek 2023 Pharmaceuticals)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=drug-induced+urinary+retention+review"},{"title":"PubMed search: sedative/hypnotic use and infection risk in older adults (Taipale 2017 CMAJ; Dublin 2011 JAGS)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=sedative+hypnotic+use+infection+risk+older+adults+pneumonia"},{"title":"Europe PMC: Pharmacovigilance evidence of drug induced urinary incontinence (Sci Rep 2025, PMC12578822)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=TITLE%3A%22Pharmacovigilance%20evidence%20of%20drug%20induced%20urinary%20incontinence%22&format=json&pageSize=5&resultType=core"},{"title":"Ensembl REST: variant rs782188154","url":"https://rest.ensembl.org/variation/human/rs782188154?content-type=application/json"},{"title":"Ensembl REST: gene SPANXA2-OT1 (lncRNA, chromosome X)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/SPANXA2-OT1?content-type=application/json"}]},"tramadol|A099":{"verdict":"statistical artifact","headline":"Opioids constipate rather than cause gastroenteritis; rare intronic CPED1 hit is a sparse-data artifact","assessment":"A09.9 (gastroenteritis/colitis of unspecified origin) is not an indication for tramadol, whose licensed use is management of moderate to moderately severe pain. The gastrointestinal profile of tramadol points the opposite way: the US label reports constipation in 24-46% and nausea in 24-40% of users versus diarrhoea in only 5-10%, and the whole opioid class slows gut transit, so an excess of diarrhoeal/gastroenteritis coding is biologically backwards for an on-drug effect. The single route by which tramadol produces diarrhoea is opioid withdrawal after abrupt discontinuation, which the label lists explicitly, but that is a discontinuation phenomenon, is not normally coded as infectious gastroenteritis, and would not be captured by a within-user GWAS of incident A09.9. The atlas numbers agree with that reading rather than contradicting it: enrichment 0.81 means gastroenteritis is actually less frequent in tramadol users than in users of other drugs, so there is not even confounding by indication to explain, and cotx_pct 1.22% is close to background primary-care incidence of an acute self-limiting illness. Temporality of 100% over 276 dated participants is uninformative here, exactly as the fact sheet warns, and the median 3.82 years from first prescription to first coded episode is far more consistent with an unrelated acute infection arising during chronic analgesic use than with a drug reaction. FAERS support is negligible (weak, 1/3 signals, PRR 1.36) and the variant has no effect on the condition off-drug (beta_disease 0.193 with se 0.121, below the 1.96*se noise threshold), so the entire z_diff of 4.53 rests on the within-user arm alone.","gene_comment":"rs189708494 is a rare intronic variant (gnomAD non-Finnish European MAF ~0.8%, overall ~0.4%) at chr7:121,161,842 in CPED1, a gene known essentially only as a positional passenger in the 7q31.31 CPED1-WNT16 bone-mineral-density locus where WNT16 is the favoured effector; CPED1 has no described role in gut physiology, opioid pharmacology or infectious susceptibility. A beta of 2.85 (OR ~17) for an acute, largely environmental infectious phenotype from a rare intronic variant is the classic signature of sparse-data bias, not mechanism.","score_check":"The within-user p-value is internally consistent with the quoted effect size, but an OR near 17 on a ~0.5% allele for acute gastroenteritis is not a believable biological effect and the off-drug arm is flat noise. The prescription-propensity arm (log10p 1.59) and depleted enrichment leave nothing that a drug-specific mechanism would explain.","candidability":1,"candidability_reason":"Backwards direction versus the drug's known constipating effect, condition depleted in tramadol users, and an implausible rare intronic hit in a bone-density-locus gene.","sources":[{"title":"Tramadol - StatPearls, NCBI Bookshelf (adverse effects, mechanism, CYP2D6)","url":"https://www.ncbi.nlm.nih.gov/books/NBK537060/"},{"title":"Tramadol hydrochloride tablet - US prescribing information, DailyMed (indications, GI adverse reaction incidence, withdrawal symptoms)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=ae7c54b1-b440-4cca-97e8-e5b825413d32"},{"title":"Ensembl REST: rs189708494 - location, alleles, gnomAD/1000 Genomes frequencies, intron_variant consequence","url":"https://rest.ensembl.org/variation/human/rs189708494?content-type=application/json;pops=1"},{"title":"Multiple Mechanisms Explain Genetic Effects at the CPED1-WNT16 Bone Mineral Density Locus (Curr Osteoporos Rep, 2023; PMID 36943599)","url":"https://pubmed.ncbi.nlm.nih.gov/36943599/"},{"title":"PubMed literature search for CPED1 (traits reported for the gene)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=CPED1"}]},"tramadol|E785":{"verdict":"statistical artifact","headline":"Rare intergenic 13q22 variant vs hyperlipidaemia in tramadol users: no label ADR, gene desert, sparse data","assessment":"Hyperlipidaemia (E78.5) is not an indication for tramadol, which is licensed only for moderate-to-severe pain, and no lipid abnormality appears anywhere in the adverse-reaction sections of the US label; the metabolic adverse effect that is labelled for tramadol is hypoglycaemia, not dyslipidaemia. Nor is this a co-prescription population signal in the usual direction: only 0.79% of tramadol users carry the code and the enrichment of 0.3 means hyperlipidaemia is markedly DEPLETED among tramadol users relative to users of other drugs, so the atlas is not simply rediscovering that tramadol is given to cardiometabolic patients. Human literature linking tramadol to blood lipids is essentially absent - a PubMed title search returns seven papers, six of which are about intravenous lipid emulsion as an antidote in tramadol poisoning or about lipid peroxidation, and the only dyslipidaemia paper is a 42-day high-dose Wistar rat toxicity model in which tramadol raised total cholesterol, LDL and triglycerides and lowered HDL, a chronic-abuse paradigm that does not transfer to therapeutic dosing. The FAERS 'STRONG' flag rests on roughly 40 reports and the exemplar report lists hyperlipidaemia alongside coronary artery disease and hypertension, i.e. the term is carried in as a concomitant condition of a polypharmacy patient rather than reported as a reaction, which is the classic way a chronic metabolic diagnosis acquires a disproportionality score. The 93.6% temporality on 157 dated participants and a 3.07-year median lag are exactly what the cohort's prescription-before-diagnosis compression produces for a slowly accumulating chronic diagnosis, so they add nothing. What is left is a single rare variant with a very large within-user effect and a null effect on the same condition in the drug-free population, which is the signature of sparse-data instability, not of a drug-gene interaction.","gene_comment":"rs9529353 is an intergenic variant at chr13:68,503,449 with a minor allele frequency of about 0.8%, sitting in the 13q22 gene desert; the nearest annotated feature is the processed pseudogene RPS3AP52 some 131 kb away and the nearest lncRNAs are 370-720 kb off, with no protein-coding gene in a 2 Mb window. There is no lipid or opioid-pharmacology candidate here at all, and the atlas correctly leaves the gene field empty rather than inventing one.","score_check":"The numbers are internally consistent but fragile: beta_adr 2.195 (OR about 9) at log10p 6.36 implies a standard error near 0.43, which for a 0.8%-frequency allele means the estimate rests on a handful of carrier cases, while beta_disease 0.129 +/- 0.067 is not distinguishable from noise, so the z_diff of 4.7 is driven entirely by the unstable treated-arm estimate. The near-null prescription arm (log10p 0.36) rules out a prescribing-propensity explanation but does nothing to make a rare intergenic OR of 9 credible without replication.","candidability":1,"candidability_reason":"No labelled or human-documented lipid effect, condition depleted rather than enriched in tramadol users, and the signal is one rare intergenic variant in a gene desert with a null disease-arm effect.","sources":[{"title":"Ensembl REST: variation rs9529353 (chr13:68,503,449, intergenic, MAF ~0.8%)","url":"https://rest.ensembl.org/variation/human/rs9529353?content-type=application/json"},{"title":"Ensembl REST: genes overlapping chr13:67.5-69.5 Mb (13q22 gene desert)","url":"https://rest.ensembl.org/overlap/region/human/13:67500000-69500000?feature=gene;content-type=application/json"},{"title":"openFDA drug label: tramadol hydrochloride - indications and adverse reactions (hypoglycaemia listed; no lipid abnormality)","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22tramadol%20hydrochloride%22&limit=1"},{"title":"Cucurbita Pepo L. Seed Oil Modulates Dyslipidemia and Neuronal Dysfunction in Tramadol-Induced Toxicity in Wistar Albino Rats (Dose Response, 2024; PMID 39381131)","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pubmed&id=39381131&rettype=abstract&retmode=text"},{"title":"PubMed: tramadol[Title] AND (lipid OR cholesterol OR dyslipidemia OR hyperlipidemia)[Title] - 7 results, mostly lipid-emulsion antidote studies","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=pubmed&term=tramadol%5BTitle%5D+AND+(lipid%5BTitle%5D+OR+dyslipidemia%5BTitle%5D+OR+hyperlipidemia%5BTitle%5D+OR+cholesterol%5BTitle%5D)&retmax=20&retmode=json"},{"title":"openFDA FAERS: tramadol hydrochloride with reaction term Hyperlipidaemia (~40 reports; exemplar lists it with CAD and hypertension)","url":"https://api.fda.gov/drug/event.json?search=patient.drug.openfda.generic_name.exact:%22TRAMADOL%20HYDROCHLORIDE%22+AND+patient.reaction.reactionmeddrapt.exact:%22Hyperlipidaemia%22&limit=1"}]},"tramadol|F419":{"verdict":"plausible ADR","headline":"Tramadol-induced anxiety is real and label-listed; the rare intergenic chr4 variant is not the reason why","assessment":"Anxiety is not an indication for tramadol - the US label licenses it only for pain severe enough to require an opioid - but anxiety is an explicitly listed adverse reaction, pooled in the label's 'CNS Stimulation' term (nervousness, anxiety, agitation, tremor) at 7% within 7 days rising to 14% by 90 days, and it is also a cardinal feature of tramadol's atypical, partly monoaminergic withdrawal syndrome that appears even on rotation to another opioid. A mechanism is described and is dual: tramadol is a serotonin/noradrenaline reuptake inhibitor as well as a mu-agonist, so it can provoke serotonergic activation, insomnia and agitation on treatment, and inter-dose or discontinuation withdrawal anxiety on its short half-life. Independent cohort evidence supports a drug-caused direction: in a Korean nationwide hip-osteoarthritis cohort (22,651 patients) codeine carried a lower anxiety risk than tramadol (aHR 0.81, 95% CI 0.69-0.95), and long-term tramadol use has been linked to increased incident depression, so the effect is not simply 'pain patients are anxious'. The opposite direction exists but is weaker and acute - opioids are transiently anxiolytic and tramadol has been trialled as an MDD adjunct - which makes this contested in sign but does not overturn the chronic-use and withdrawal literature. The atlas is consistent with this at the drug level: FAERS is STRONG with PRR 2.9, and enrichment of 0.68 shows anxiety is actually less common in tramadol users than in users of other drugs, arguing against confounding by indication rather than for it. The genetic layer, however, does not survive: the signal rests on one rare intergenic variant with an effect size that is not credible for 172 dated cases, so this is a believable drug-level ADR attached to an uninterpretable locus.","gene_comment":"rs6829799 is an intergenic variant at chr4:82,119,215 (GRCh38) with MAF ~0.2%; the nearest annotated feature is the processed pseudogene HNRNPA3P13, with RASGEF1B ending ~75 kb upstream and HNRNPD starting ~230 kb downstream, and none of these has any neuropsychiatric or opioid-pharmacology credential. No gene assignment here can be treated as mechanism.","score_check":"beta_adr 4.49 (odds ratio ~90) for a 0.2%-frequency variant in a within-user analysis with 172 dated cases is the classic sparse-data inflation signature, and the accompanying z_diff 5.8 is driven by that inflated beta rather than by a real treated-vs-untreated contrast. The comparison arms are at least clean - the disease arm is flat noise (beta 0.081, se 0.116) and prescription propensity is null - and temporality of 100% is uninformative given the cohort's known upward compression.","candidability":5,"candidability_reason":"A genuine, label-listed and cohort-supported tramadol adverse effect, but the genetics are a rare intergenic variant with an implausibly large effect, so nothing pharmacogenomic is followable here.","sources":[{"title":"Tramadol hydrochloride tablets, film coated - FDA label (DailyMed SPL, Aurobindo)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=e5005d9e-91cb-4ff8-bcf2-b5ee70cf4f3f"},{"title":"Ensembl REST: variant record for rs6829799 (chr4:82,119,215, intergenic, MAF 0.2%)","url":"https://rest.ensembl.org/variation/human/rs6829799?content-type=application/json"},{"title":"Ensembl REST: genes overlapping chr4:81.9-82.4 Mb (RASGEF1B, HNRNPD, HNRNPA3P13)","url":"https://rest.ensembl.org/overlap/region/human/4:81900000-82400000?feature=gene;content-type=application/json"},{"title":"Kim et al., Comparative risk of psychiatric comorbidities associated with codeine and tramadol in hip osteoarthritis, J Glob Health 2026 (PMID 42024864) via Europe PMC","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:42024864&resultType=core&format=json"},{"title":"Withdrawal Syndrome Following Opioid Rotation: Tramadol and Its Unique Pharmacology, Cureus 2026 (PMID 41769572) via Europe PMC","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=EXT_ID:41769572&resultType=core&format=json"},{"title":"Europe PMC search: tramadol, monoamine reuptake inhibition, antidepressant and depression outcomes (incl. J Affect Disord 2026, PMID 41921874)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=tramadol%20AND%20(antidepressant%20OR%20%22reuptake%20inhibition%22)%20AND%20depression&format=json&pageSize=15"}]},"tramadol|L409":{"verdict":"co-prescription population","headline":"HLA-C*06:02 psoriasis locus rediscovered inside tramadol users; no drug effect (z_diff 1.44)","assessment":"Psoriasis is neither an indication for tramadol nor a labelled adverse reaction: the FDA labels license tramadol only for severe and persistent pain requiring an opioid, and a query across all 108 openFDA tramadol hydrochloride labels returns no occurrence of the word psoriasis anywhere, while the labelled skin events are sweating, dermatitis, pruritus, rash and urticaria. Psoriasis patients do receive tramadol for psoriatic-arthritis and axial pain, so some co-prescription is expected, but the atlas enrichment of 0.78 says psoriasis is if anything slightly less common among tramadol users than among users of other drugs, so even the confounding-by-indication story is weak here. The genetics are the whole of the signal: rs12189871 is the classical PSORS1 tag for HLA-C*06:02 (HLA-Cw6), reported for psoriasis at p=3e-50 in the GWAS Catalog and used as an HLA-Cw6 proxy in psoriasis pharmacogenomic panels, and the atlas itself measures it at log10p 126.4 with beta 1.23 +/- 0.051 in the drug-free population. Against that, the within-user signal (log10p 7.08, beta 1.69) is simply the same locus re-detected in a subset, and z_diff of 1.44 means the effect in treated patients is statistically indistinguishable from the effect without the drug. Pharmacovigilance agrees there is nothing drug-specific: the atlas records no FAERS signal with PRR 0.92, and the 321 FAERS reports mentioning both tramadol and psoriasis sit against 134,467 psoriasis reports overall, i.e. background co-medication. The atlas's own decomposition already reads it correctly, scoring predisposition 96.6 against indication 3.3.","gene_comment":"rs12189871 (chr6:31,284,147, GRCh38) lies about 12 kb upstream of HLA-C in the PSORS1 interval and is a well-validated tag for HLA-C*06:02; the co-listed RPL3P2 is a ribosomal-protein pseudogene 2.6 kb away, a nearest-feature annotation with no mechanistic content. The gene call is therefore strong for HLA-C and meaningless for RPL3P2, but it points at inherited psoriasis susceptibility, not at any tramadol pathway.","score_check":"The numbers are internally coherent and believable: beta_disease 1.23 +/- 0.051 is far from noise and matches the published PSORS1 effect, while the near-null z_diff (1.44) and log10p_prescribed (0.13) show neither drug modification nor genotype-driven prescribing. Temporality of 80.8% rests on only 78 dated participants and is exactly the direction the cohort's record-length asymmetry inflates, so it adds no independent support.","candidability":1,"candidability_reason":"Real psoriasis susceptibility locus re-detected within drug users; no labelled or literature-described opioid mechanism, no FAERS signal, condition depleted rather than enriched, and the treated-vs-untreated effect difference is null.","sources":[{"title":"dbSNP rs12189871 (NCBI)","url":"https://www.ncbi.nlm.nih.gov/snp/rs12189871"},{"title":"GWAS Catalog: rs12189871 associations (psoriasis p=3e-50; mapped genes RPL3P2, HLA-C, PSORS1C3)","url":"https://www.ebi.ac.uk/gwas/rest/api/singleNucleotidePolymorphisms/rs12189871/associations?projection=associationBySnp"},{"title":"openFDA drug label API: tramadol hydrochloride indications and adverse reactions (108 labels, no mention of psoriasis)","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22tramadol+hydrochloride%22&limit=1"},{"title":"Stuart PE et al., Genome-wide Association Analysis of Psoriatic Arthritis and Cutaneous Psoriasis Reveals Differences in Their Genetic Architecture, Am J Hum Genet 2015 (PMID 26626624)","url":"https://pubmed.ncbi.nlm.nih.gov/26626624/"},{"title":"HLA-Cw6 and other HLA-C alleles ... associate with optimal response to anti-IL-17A treatment in psoriasis, Expert Opin Biol Ther 2021 (PMID 33297781; rs12189871 in the HLA-C upstream panel)","url":"https://pubmed.ncbi.nlm.nih.gov/33297781/"},{"title":"openFDA FAERS: tramadol + psoriasis co-reports (321) vs all psoriasis reports (134,467)","url":"https://api.fda.gov/drug/event.json?search=patient.drug.openfda.generic_name:%22tramadol%22+AND+patient.reaction.reactionmeddrapt:%22psoriasis%22&limit=1"}]},"tramadol|N309":{"verdict":"plausible ADR","headline":"Opioid retention plausibly seeds cystitis, but TPRA1 burden effect (beta 33) is a sparse-data artifact","assessment":"Cystitis is not a licensed indication for tramadol: the US label restricts it to pain severe enough to require an opioid, and its urogenital adverse reactions are urinary retention and urinary frequency (incidence >=1%) plus dysuria (<1%), with no mention of cystitis or urinary tract infection. A drug-caused pathway is nonetheless coherent and one step removed: opioids suppress the micturition reflex, blunt the sensation of urgency and detrusor contraction and raise post-void residual volume, and a POUR review states plainly that susceptibility to urinary infection rises both directly from incomplete emptying and indirectly through the catheterisation used to relieve retention; a 2026 JADER disproportionality analysis found a significant urinary-retention signal for tramadol specifically, alongside morphine, oxycodone and codeine. Tramadol does not move cystitis in the opposite direction - if anything its analgesia would mask lower urinary tract symptoms and delay, not create, a coded diagnosis. Confounding by indication is not the obvious explanation here: cotx_pct is only 0.58% and enrichment is 0.76, i.e. cystitis is slightly depleted among tramadol users relative to users of other drugs, and current IC/BPS consensus therapy is built on pentosan polysulfate, amitriptyline, hydroxyzine, silodosin and intravesical or neuromodulatory options rather than opioids, so tramadol-for-bladder-pain is not a large prescribing channel. The residual confounders are the ordinary ones - a chronically painful, older, largely female, less mobile and more catheterised population under closer medical surveillance - and the atlas cannot separate those from a true opioid effect. The internal FAERS support is weak (1/3, PRR 1.8), consistent with the label being silent on cystitis, so the pharmacovigilance signal for this exact term is thin even though the retention signal upstream of it is solid.","gene_comment":"TPRA1 (Q86W33, GPR175/TMEM227, chromosome 3) is an orphan seven-transmembrane protein expressed ubiquitously with higher levels in heart, placenta and kidney, with no described role in bladder physiology, opioid pharmacology or urinary infection; it is a different gene from TRPA1, the ankyrin TRP channel that carries the real bladder-nociception and opioid-hyperalgesia literature, and the similarity of the symbols is a standing trap here. A single-unit MPC burden hit in this gene supplies no mechanism and should not be presented as one.","score_check":"beta_adr of 32.98 is not a credible log-odds effect (it implies an astronomically large odds ratio) and is the classic signature of separation in a sparse rare-variant burden test with only 118 dated cases, so z_diff 4.72 is inflated by that same unstable estimate rather than by a genuine treated-versus-untreated contrast. The comparison arms are appropriately null (disease |beta| 1.169 < 1.96*se 1.101; log10p_prescribed 0.43), and temporality of 81.4% after first exposure with a 2.38-year median is consistent with drug-then-event but is the direction the cohort inflates anyway, so it adds little on its own.","candidability":4,"candidability_reason":"Real class-level opioid retention leading to infection makes the phenotype plausible, but cystitis is an unlabelled second-order consequence and the genetic evidence is an implausibly large burden effect in a urologically irrelevant orphan gene.","sources":[{"title":"Tramadol Hydrochloride Tablet, Film Coated - US prescribing information (DailyMed, Aurobindo)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=e5005d9e-91cb-4ff8-bcf2-b5ee70cf4f3f"},{"title":"Disproportionality analysis of opioid-related urinary retention using the JADER database (Drug Discov Ther, 2026; PMID 42633912)","url":"https://doi.org/10.5582/ddt.2026.01040"},{"title":"Postoperative urinary retention (POUR): A narrative review (PMC11033892)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/PMC11033892/fullTextXML"},{"title":"Global Consensus on Interstitial Cystitis/Bladder Pain Syndrome: An Update on Therapeutic Treatments (Neurourol Urodyn, 2026; PMID 40783827)","url":"https://doi.org/10.1002/nau.70106"},{"title":"UniProtKB Q86W33 (TPRA1_HUMAN) Transmembrane protein adipocyte-associated 1 / GPR175","url":"https://rest.uniprot.org/uniprotkb/Q86W33.txt"}]},"tramadol|N328":{"verdict":"plausible ADR","headline":"Bladder/micturition disorder is a labelled tramadol ADR, but this rare intergenic variant is unconvincing","assessment":"Tramadol is licensed only for moderate to moderately severe pain, so N32.8 (other specified disorders of bladder) is in no sense an indication; it is, however, a recognised adverse effect class, since the US tramadol label lists urinary frequency and urinary retention at 1% to <5% and dysuria at <1%, and the atlas FAERS flag (STRONG, PRR 2.68) agrees. A drug-caused mechanism is described and is not direction-ambiguous: mu-opioid agonism inhibits the sacral micturition reflex (morphine-6-glucuronide alone produces urinary retention in unanaesthetised rats), and tramadol's noradrenaline-reuptake inhibition adds urethral sphincter tone, so both of its actions push toward impaired emptying rather than away from it. The one counter-current is that tramadol is used acutely to relieve catheter-related bladder discomfort, which is bladder spasm rather than the chronic structural/functional bladder pathology N32.8 collects, so it does not overturn the ADR reading. The population arm supports rather than undermines causality: only 0.86% of tramadol users carry the code and enrichment is 0.61, i.e. bladder disorders are less common in tramadol users than in users of other drugs, so this is not confounding by indication, and the prescription-propensity arm is flat (log10p 0.02). The disease arm is also clean in the sense that matters here: beta_disease 0.276 with se 0.159 is inside noise, so the variant does not predispose to bladder disorders without the drug, and z_diff 4.76 is driven by the within-user effect. The weakness is entirely on the genetic side, plus the fact that N32.8 is a residual wastebasket code whose members range from bladder hypertrophy to detrusor problems, so the phenotype being tested is heterogeneous.","gene_comment":"The atlas assigns no gene, and correctly so: rs142163545 is an intergenic SNV at chr10:76,769,795 (GRCh38) with global MAF ~0.24%, sitting in a gene desert whose nearest protein-coding neighbour is KCNMA1 about 100 kb away past its 3' end. KCNMA1 encodes the BK channel that dominates detrusor smooth-muscle excitability and whose loss of function produces detrusor overactivity, so the neighbourhood is thematically attractive, but a rare variant 100 kb downstream with no eQTL or regulatory evidence is a coincidence of position, not a mechanism, and should not be written up as one.","score_check":"The numbers are internally consistent but the effect is not credible at face value: beta_adr 4.643 (OR ~100) for a 0.24%-frequency variant against only 135 dated cases is the signature of sparse-data separation, and log10p 6.56 is below the usual genome-wide threshold. The clean null in the disease and prescription arms is reassuring, but this variant needs Firth/exact re-testing and replication before it means anything.","candidability":5,"candidability_reason":"Real, labelled tramadol bladder ADR with a coherent opioid/noradrenergic mechanism and no confounding by indication, but the supporting genetics is a rare intergenic variant with an implausibly large effect and sub-genome-wide p on a wastebasket ICD code.","sources":[{"title":"Tramadol hydrochloride tablet - DailyMed label (indications; urinary frequency/retention 1-<5%, dysuria <1%)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=ae7c54b1-b440-4cca-97e8-e5b825413d32"},{"title":"Tramadol - StatPearls (NBK537060): mechanism, adverse effects, CYP2D6","url":"https://www.ncbi.nlm.nih.gov/books/NBK537060/"},{"title":"Ensembl REST variant record for rs142163545 (intergenic, chr10:76,769,795, MAF 0.0024)","url":"https://rest.ensembl.org/variation/human/rs142163545?content-type=application/json"},{"title":"Ensembl overlap query, chr10:76.67-76.87 Mb - nearest protein-coding gene KCNMA1","url":"https://rest.ensembl.org/overlap/region/human/10:76669795-76869795?feature=gene;content-type=application/json"},{"title":"Petkov GV, Central role of the BK channel in urinary bladder smooth muscle physiology and pathophysiology (PMC4166757)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC4166757/"},{"title":"PMC search: tramadol and urinary retention (incl. morphine metabolites and micturition; tramadol for catheter-related bladder discomfort)","url":"https://pmc.ncbi.nlm.nih.gov/search/?term=tramadol+urinary+retention"}]},"tramadol|R101":{"verdict":"plausible ADR","headline":"Upper abdominal pain is a labelled tramadol effect, but the rare SCD5-upstream variant is not credible","assessment":"R10.1 upper abdominal pain is not a licensed indication for tramadol in the narrow sense - the label indicates it for 'pain severe enough to require an opioid analgesic' generally - but it is an explicitly labelled adverse reaction: the US tramadol tablet label lists abdominal pain among reactions occurring in 1-5% of patients, alongside dyspepsia (5% at 7 days rising to 13% at 90 days), constipation (24% to 46%) and nausea (24% to 40%), so the GI burden is dose- and duration-dependent and substantial. A drug-caused mechanism is well described and runs in the ADR direction: mu-opioid agonism slows gastric emptying and gut transit, and chronic opioid use can produce narcotic bowel syndrome, an opioid-induced hyperalgesia state whose reported prevalence is about 2.8% among opioid users and in which tramadol is specifically implicated and epigastric cramping pain is a documented presentation. The opposite-direction confound is nonetheless real and must be stated: tramadol is an analgesic that is also prescribed for abdominal pain, so a fraction of these codes will be the reason for the prescription rather than its consequence - the atlas indication score of 20.4 acknowledges this. Against that, the atlas comparison arms behave as an ADR should: cotx is only 1.42% and enrichment is 0.77, meaning upper abdominal pain is if anything less common in tramadol users than in users of other drugs, which is the opposite of what channelling to abdominal-pain patients would produce, and 85.5% of 331 datable cases were first coded after first exposure at a median 2.97 years - though the fact sheet's own warning applies, a high temporality figure is weak evidence because the cohort compresses it upward. The three-arm separation is clean: the variant is flat in the drug-free disease arm (beta 0.084, se 0.09, indistinguishable from zero) and null for prescription propensity (log10p 0.33), with z_diff 5.06, so this is not a disease-predisposition or prescribing-propensity artifact leaking through. FAERS support is weak (1 of 3 sources, PRR 1.87), which is unsurprising for a nonspecific symptom code. The condition-drug pair is therefore genuine and mechanistically supported at class level; what is not supported is that this particular variant explains any of it.","gene_comment":"rs139368799 is a rare non-coding SNV (gnomAD MAF 0.35% global, 0.58% non-Finnish European) sitting 4.4 kb upstream of SCD5 on 4q21.22, annotated by VEP only as an upstream_gene_variant of MODIFIER impact - the gene assignment is proximity, not function. SCD5 is a primate-specific stearoyl-CoA desaturase in fatty-acid metabolism with no established role in opioid pharmacokinetics, mu-receptor signalling, gut motility or visceral nociception; the recognised pharmacogenes for tramadol are CYP2D6 (and secondarily OPRM1/COMT), and CYP2D6 slow metabolism is what has actually been tied to tramadol side-effect discontinuation.","score_check":"The arms are internally consistent and the disease/prescription nulls are properly null, but beta_adr 2.95 (OR ~19) on an allele carried by roughly 0.5% of Europeans is exactly the sparse-data shape the atlas warns about, and log10p_adr 6.89 does not reach genome-wide significance. The clinical association is believable; this variant's effect size is not, and would need replication in an independent cohort before being taken at face value.","candidability":5,"candidability_reason":"A real, labelled, mechanistically explained tramadol adverse effect, but attached here to a rare non-coding variant near a functionally irrelevant gene at sub-genome-wide significance.","sources":[{"title":"Tramadol Hydrochloride tablet, coated - FDA label (DailyMed, Amneal Pharmaceuticals)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=58b802cb-2443-4f5b-9718-7d54c6d50cb4"},{"title":"Tramadol - StatPearls, NCBI Bookshelf","url":"https://www.ncbi.nlm.nih.gov/books/NBK537060/"},{"title":"Narcotic Bowel Syndrome, an Under-recognized Cause of Chronic Abdominal Pain in Adults (J Neurogastroenterol Motil, 2022)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC9577581/"},{"title":"Role of cytochrome phenotyping in patients with chronic noncancer pain with inadequate response to tramadol (Pain Rep)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC13384671/"},{"title":"Ensembl VEP annotation for rs139368799 (upstream_gene_variant, SCD5, gnomAD frequencies)","url":"https://rest.ensembl.org/vep/human/id/rs139368799?content-type=application/json"},{"title":"SCD5 stearoyl-CoA desaturase 5 - NCBI Gene 79966","url":"https://www.ncbi.nlm.nih.gov/gene/79966"}]},"tramadol|R31":{"verdict":"plausible ADR","headline":"Label lists haematuria at <1%; a real but thin ADR, and the rare intronic KCNIP2 hit adds no mechanism","assessment":"Haematuria is not an indication for tramadol, which is licensed only for moderate-to-severe pain, but it is not absent from the label either: the US tramadol labelling lists hematuria together with difficulty in micturition, urinary frequency, dysuria and urinary retention among events reported at 0.5% to <1.0%, with the standard caveat that a causal relationship cannot be reliably established from such reports. Beyond that listing the primary literature is essentially empty - a PubMed search for tramadol and hematuria returns only a handful of papers, none of them causal, and the drug is not flagged for this in the BNF (on_bnf = 0), while FAERS support here is weak (1/3, PRR 1.48). The only mechanism one can argue is indirect: opioid mu-receptor effects on detrusor and sphincter tone cause urinary retention, which invites instrumentation, catheterisation and infection, all of which produce visible or dipstick haematuria; nothing points to direct urothelial or glomerular injury by tramadol. The competing explanation is strong, because tramadol is an established analgesic for acute renal colic in multiple randomised trials, and urolithiasis is a common cause of haematuria, so a stone patient acquires both codes for the same reason. The atlas partly resists that reading: enrichment is 0.68, i.e. haematuria is if anything less common in tramadol users than in users of other drugs, and the median interval to the haematuria record is 2.2 years, too long for the same renal-colic episode - though the 99.2% temporality itself is uninformative given the cohort's upward compression. The genetic layer is the weak part: a single rare intronic variant carrying an implied odds ratio around 20 in a gene with no urinary-tract biology, with the same variant showing nothing at all for haematuria off-drug.","gene_comment":"rs144335349 sits at chr10:101,828,941 (GRCh38), genuinely intronic in KCNIP2 (101,825,969-101,844,005) and overlapping the KCNIP2-AS1 antisense lncRNA, so the positional assignment is correct but the gene is a Kv4 potassium-channel accessory subunit of brain and cardiac Ito with no described renal or urothelial role. At gnomAD frequency of roughly 0.2-0.5% it is a rare variant, and an intronic rare variant in a cardiac/neuronal channel subunit is a locus label, not a mechanism for bleeding into urine.","score_check":"The within-user signal is internally consistent (beta 3.06 with log10p 6.01 implies se around 0.63) but a beta of 3.06 for a ~0.3% allele means a handful of carrier cases driving an OR near 20, which is the classic shape of a sparse-data hit; the disease arm is a clean null (beta 0.10 +/- 0.11, |beta| < 1.96*se) and prescription propensity is null too, so the z_diff of 4.67 is real arithmetic but rests entirely on the fragile treated-arm estimate. Nothing here is impossible, but the effect size is far larger than any plausible pharmacogenomic effect on haematuria and should be treated as unreplicated until carrier counts are shown.","candidability":5,"candidability_reason":"Genuinely uncertain: haematuria is a labelled sub-1% genitourinary event with only an indirect retention-mediated mechanism, and the supporting genetics are a rare intronic variant in a channel gene with no urinary biology.","sources":[{"title":"Tramadol - StatPearls, NCBI Bookshelf (NBK537060): indications and adverse effects","url":"https://www.ncbi.nlm.nih.gov/books/NBK537060/"},{"title":"openFDA drug label API - tramadol hydrochloride labelling (indications; genitourinary adverse reactions incl. hematuria 0.5% to <1.0%)","url":"https://api.fda.gov/drug/label.json?search=openfda.generic_name:%22tramadol+hydrochloride%22&limit=1"},{"title":"PubMed search: tramadol AND hematuria (4 results, none causal)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=tramadol+hematuria"},{"title":"PubMed search: tramadol renal colic randomized trials","url":"https://pubmed.ncbi.nlm.nih.gov/?term=tramadol+renal+colic+randomized"},{"title":"Hematuria - StatPearls, NCBI Bookshelf (NBK534213): causes incl. urolithiasis and drug-induced haematuria","url":"https://www.ncbi.nlm.nih.gov/books/NBK534213/"},{"title":"UniProt Q9NS61 (KCNIP2): function, tissue expression","url":"https://rest.uniprot.org/uniprotkb/Q9NS61.txt"},{"title":"Ensembl REST: rs144335349 consequence and allele frequency; genes overlapping chr10:101,820,000-101,840,000","url":"https://rest.ensembl.org/variation/human/rs144335349?content-type=application/json;pops=1"}]},"tramadol|R33":{"verdict":"plausible ADR","headline":"Urinary retention is a labelled tramadol ADR with a clear opioid mechanism; the TBPL1 burden hit is not","assessment":"Retention of urine (R33) is not an indication for tramadol and is not merely background comorbidity: the US tramadol hydrochloride label lists urinary retention and urinary frequency among adverse reactions occurring in 1% to less than 5% of patients, and the Drug Interactions section warns that concomitant anticholinergics increase the risk of urinary retention. The class effect is well described - Verhamme et al. (Drug Safety 2008) review drug-induced urinary retention with opioids as a recognised cause, de Boer et al. (2017) treat postoperative urinary retention as a core opioid-related side effect, and mu-opioid agonism suppresses detrusor contractility and bladder afferent signalling, so the drug moves the condition in the direction the atlas reports, not the opposite. Tramadol specifically has a case-level and pharmacovigilance trail: five Lareb reports of transient voiding impairment or retention on tramadol (Pharmacoepidemiol Drug Saf 1999), a Prescrire International note on tramadol urinary disorders (2001), and a recent JADER disproportionality analysis reporting significant urinary-retention signals for morphine, oxycodone, tramadol and codeine - consistent with the sheet's FAERS PRR of 3.2. The atlas epidemiology is compatible with an ADR rather than confounding by indication: enrichment is 0.85, i.e. retention codes are if anything slightly less common in tramadol users than in users of other drugs, predisposition scores 0, and all 262 datable cases were first recorded after first exposure. The main non-genetic caveat is that the median 1.39 years to onset is long for opioid-induced retention, which is typically acute or perioperative, and chronic tramadol recipients skew older and male (prostatic obstruction), a confounder that enrichment against other-drug users does not fully remove. So the drug-condition link is real and labelled; what the atlas adds here - the variant - is the part that does not hold up.","gene_comment":"TBPL1 (TBP-like 1 / TRF2, UniProt P62380) is a real protein-coding transcription factor driving ribosomal-protein gene expression, ubiquitously expressed with testis/ovary enrichment, and has no documented role in bladder function, smooth muscle, opioid signalling or drug metabolism; the GWAS Catalog gene page lists no associations for it. It is a single MPC missense-burden hit, and the genes one would expect for a tramadol ADR - CYP2D6 (O-desmethyltramadol formation) or OPRM1 - are absent, so this should not be presented as mechanism.","score_check":"beta_adr of 15.5 from a single-gene MPC burden test is an implausible effect size and points to quasi-complete separation on a handful of rare carriers rather than a real 15-log-odds effect, so log10p 6.95 and z_diff 4.94 should be discounted. The comparison arms are at least clean: the same burden is null for the condition without the drug (log10p 0.62, beta 0.73 +/- 0.624, i.e. indistinguishable from noise) and null for prescription propensity (0.59), so nothing suggests indication bias - only sparse-data instability.","candidability":5,"candidability_reason":"A genuine, labelled, mechanistically explained opioid ADR, but the specific genetic finding is a sparse-data TBPL1 burden hit with no biological plausibility.","sources":[{"title":"DailyMed - TRAMADOL HYDROCHLORIDE tablet, coated (Amneal) full prescribing information","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=58b802cb-2443-4f5b-9718-7d54c6d50cb4"},{"title":"PubMed search: tramadol urinary retention (Lareb bladder dysfunction cases; Prescrire; JADER disproportionality)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=tramadol+urinary+retention"},{"title":"PubMed search: opioid-induced urinary retention (Verhamme 2008 Drug Safety; de Boer 2017)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=opioid+induced+urinary+retention"},{"title":"Europe PMC search: tramadol AND \"urinary retention\" (JADER opioid-related urinary retention analysis)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=tramadol%20AND%20%22urinary%20retention%22&format=json&pageSize=25&resultType=core"},{"title":"UniProt P62380 - TATA box-binding protein-like 1 (TBPL1), Homo sapiens","url":"https://rest.uniprot.org/uniprotkb/P62380.txt"},{"title":"GWAS Catalog - gene page for TBPL1 (no reported associations)","url":"https://www.ebi.ac.uk/gwas/genes/TBPL1"},{"title":"MedlinePlus - Tramadol drug information","url":"https://medlineplus.gov/druginfo/meds/a695011.html"}]},"tramadol|R53":{"verdict":"plausible ADR","headline":"Fatigue/asthenia is a labelled tramadol reaction, but this chr7 lncRNA hit is sparse-data and sub-genome-wide","assessment":"R53 (malaise and fatigue) is not an indication for tramadol - the US label restricts it to pain severe enough to require an opioid analgesic - but asthenia and malaise are explicitly listed adverse reactions, with asthenia reported in 6% of patients at 7 days rising to 12% at 90 days, malaise in 1-5%, and somnolence in 16-25%, so the drug-condition direction is right and the drug is not known to improve fatigue. A mechanism is straightforward and described: tramadol is a mu-opioid agonist whose M1 metabolite carries most of the opioid activity, combined with serotonin and noradrenaline reuptake inhibition, giving dose-related sedation and cognitive slowing; chronic opioid exposure additionally causes hypothalamic-pituitary-adrenal suppression and hypogonadism, both of which present as fatigue and are recognised predictable endocrine adverse effects of long-term opioid therapy. Against a pure ADR reading, R53 is a nonspecific symptom code that is common in the chronic-pain population tramadol is prescribed to, though the atlas enrichment of 0.8 says fatigue is if anything slightly less frequent in tramadol users than in users of other drugs, so gross confounding by indication is not the obvious explanation here. The temporality of 99% after first exposure is uninformative given only 97 dated participants, 19.8% undated, and the known upward compression of this metric in the cohort, and the median 2.37 years to onset fits chronic use rather than an acute sedative effect. The genetic side is where the association falls down: a single SNV, not CYP2D6, OPRM1 or COMT, the loci with actual tramadol pharmacogenomic evidence, and no supporting disease-arm or prescription-arm signal (log10p_disease 0.7, log10p_prescribed 0.17). FAERS support is weak (1/3, PRR 1.45), consistent with fatigue being a real but low-salience, poorly reported opioid effect.","gene_comment":"ENSG00000225718 is an unnamed novel lncRNA at chr7:69,145,185-69,430,923 (GRCh38) sitting in a gene desert whose only protein-coding neighbour, AUTS2, begins ~170 kb downstream; the locus has no known link to opioid pharmacokinetics, mu-receptor signalling or fatigue. rs79486133 should be treated as an anonymous intergenic marker, not as mechanism.","score_check":"beta_adr 3.411 implies an odds ratio near 30 for a nonspecific symptom code on ~97 dated cases, which is the signature of a low-frequency variant with sparse-data inflation rather than a real effect of that size, and log10p_adr 6.36 does not reach genome-wide significance. The disease arm is properly null (|beta_disease| 0.144 < 1.96 x se 0.112), so z_diff 5.2 is carried entirely by the inflated within-users estimate.","candidability":4,"candidability_reason":"The drug-condition link is a genuine labelled adverse reaction with a described mechanism, but the specific genetic signal is an anonymous lncRNA marker with an implausibly large, sub-genome-wide effect on a nonspecific symptom code.","sources":[{"title":"TRAMADOL HYDROCHLORIDE tablet, coated - US prescribing information (DailyMed, Advagen Pharma)","url":"https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=93b12089-3a0f-4b57-abb1-2429cf31995d"},{"title":"Tramadol - StatPearls, NCBI Bookshelf NBK537060","url":"https://www.ncbi.nlm.nih.gov/books/NBK537060/"},{"title":"Ensembl REST lookup: ENSG00000225718 (novel lncRNA, chromosome 7)","url":"https://rest.ensembl.org/lookup/id/ENSG00000225718?content-type=application/json;expand=0"},{"title":"Ensembl REST overlap: genes in chr7:68,900,000-70,000,000 (GRCh38)","url":"https://rest.ensembl.org/overlap/region/human/7:68900000-70000000?feature=gene;content-type=application/json"},{"title":"Europe PMC search: opioid-induced adrenal insufficiency and fatigue (incl. 'Opioid Analgesics: Managing the Predictable', PMID 40441357)","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=opioid%20induced%20adrenal%20insufficiency%20fatigue&format=json&pageSize=10&resultType=core"},{"title":"Europe PMC search: AUTS2 and fatigue/opioid/addiction literature","url":"https://www.ebi.ac.uk/europepmc/webservices/rest/search?query=%22AUTS2%22%20AND%20(fatigue%20OR%20opioid%20OR%20addiction)&format=json&pageSize=10"}]},"zopiclone|K297":{"verdict":"statistical artifact","headline":"Dyspeptic symptoms are on zopiclone's label, but this MAF-0.1% intronic hit is sparse-data inflation","assessment":"Zopiclone is licensed only for short-term treatment of insomnia in adults, so gastritis (K29.7) is not an indication; the Zimovane SmPC lists dry mouth as common, nausea and vomiting as uncommon and dyspepsia at unknown frequency, but gastritis itself is not a listed undesirable effect, so the condition sits at the edge of the labelled dyspeptic spectrum rather than being a recognised ADR. A drug-caused mechanism is describable only by analogy and it points at the wrong organ: GABA-A hypnotics blunt the arousal response to nocturnal acid, and zolpidem cut arousal from 89% to 40% of acid events and lengthened acid clearance from ~38 s to >363 s in GERD patients, which is a reflux-oesophagitis argument, not a gastric-mucosal-inflammation one; no zopiclone-specific gastric mechanism is described, and nothing in the literature suggests zopiclone moves gastric disease in the opposite direction. Confounding by comorbidity is the obvious competitor, since functional dyspepsia and insomnia are repeatedly reported as co-occurring (Lacy 2011; Wuestenberghs 2022), and K29.7 in EHR practice is frequently a symptomatic/endoscopic label for exactly those patients. Against that, the atlas enrichment of 0.46 says gastritis is actually depleted among zopiclone users relative to users of other drugs, which is the honest test and argues against straightforward channelling of upper-GI patients to this hypnotic. The genetic layer is where this falls apart: a single ultra-rare intronic SNV with beta_adr 5.643 corresponds to an odds ratio in the hundreds for a condition affecting 2% of users, which is not a believable effect size and is the classic signature of a handful of carriers among the cases. FAERS being STRONG with PRR 2.64 for gastritis on a hypnotic is worth noting but is itself vulnerable to co-reporting and notoriety, and it supports a drug-level GI signal at best, not this locus.","gene_comment":"rs117631852 (chr11:111,249,271, GRCh38) genuinely falls inside POU2AF2/ENSG00000150750 (111,245,725-111,286,401), so the assignment is positionally correct rather than an intergenic guess, and POU2AF2/OCA-T1 is a real coactivator of POU2F3 in the tuft-cell lineage, a cell type present in gastric epithelium. But the variant is intronic with a minor allele frequency of about 0.001, has no established gastric phenotype (this 11q23.1 region is known instead for the COLCA1/COLCA2 colorectal-cancer signal), and tuft-cell biology should not be dressed up as a mechanism for zopiclone gastritis.","score_check":"The internal pattern is coherent - beta_disease 0.054 +/- 0.104 is indistinguishable from zero, log10p_disease 0.22, and z_diff 5.66 all say the effect is confined to treated patients - but that same pattern is what a rare variant with very few carrying cases produces, and beta_adr 5.643 is too large for its context to be taken at face value. Temporality of 96.1% over only 129 dated participants is the uninformative direction given the cohort's prescription-record lead-in, so it adds essentially nothing.","candidability":3,"candidability_reason":"Gastritis is at most an edge-of-label dyspeptic effect with no gastric mechanism, and the single MAF-0.1% intronic variant carries an implausible effect size, though the depleted enrichment keeps it above pure indication bias.","sources":[{"title":"Zimovane 7.5mg (zopiclone) SmPC - sections 4.1 and 4.8, electronic medicines compendium","url":"https://www.medicines.org.uk/emc/product/2855/smpc"},{"title":"Ensembl variation record for rs117631852 (alleles, MAF, position, consequence)","url":"https://rest.ensembl.org/variation/human/rs117631852?content-type=application/json"},{"title":"Ensembl gene record for POU2AF2 (ENSG00000150750, chr11 coordinates)","url":"https://rest.ensembl.org/lookup/symbol/homo_sapiens/POU2AF2?content-type=application/json"},{"title":"OCA-T1 and OCA-T2 are coactivators of POU2F3 in the tuft cell lineage (Nature 2022, PMID 35576971)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=POU2AF2+C11orf53+tuft+cell"},{"title":"Effect of zolpidem on the sleep arousal response to nocturnal esophageal acid exposure (PMID 19426833, abstract via NCBI E-utilities)","url":"https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pubmed&id=19426833&rettype=abstract&retmode=text"},{"title":"PubMed: insomnia / sleep disturbance and functional dyspepsia association (Lacy 2011 PMID 21334461; Wuestenberghs 2022 PMID 35210990)","url":"https://pubmed.ncbi.nlm.nih.gov/?term=insomnia+functional+dyspepsia+association"}]},"zopiclone|R074":{"verdict":"statistical artifact","headline":"Chest pain is a nonspecific label term for zopiclone; the rare intergenic hit is a sparse-data artifact","assessment":"Zopiclone is licensed only for short-term insomnia, so chest pain (R07.4) is not an indication; it does appear on the eszopiclone (S-zopiclone) US label as 'Body as a Whole: Frequent: chest pain' among premarketing adverse events, but that list is unadjusted trial-emergent terminology, not an attributed causal effect, and the FAERS PRR of 2.92 rests on one of the least specific reporting terms in the database. No mechanism links GABA-A alpha1 agonism to chest pain: the same label reports no respiratory depression in healthy volunteers at 2.5-fold the recommended dose and only advises caution in compromised respiratory function, and in heart-failure patients with insomnia the Z-drugs (zolpidem, zopiclone, eszopiclone) had a lower rate of HF rehospitalisation than benzodiazepines, so the cardiac direction of travel is if anything favourable rather than harmful. The likelier explanation is the treated population: non-cardiac chest pain is strongly enriched for panic disorder, obsessive-compulsive disorder and major depression, and those are precisely the patients who receive hypnotics, so a chest-pain code in a zopiclone user is expected without any drug effect. Against that, the atlas's own enrichment of 0.38 says chest pain is actually depleted among zopiclone users relative to users of other drugs, at a low 1.33% co-occurrence, which undercuts a strong population story too and leaves the association resting entirely on one variant. That variant is a rare intergenic SNV with a within-users beta of 3.87 (OR ~48) and a flat, null effect on chest pain in the drug-free population (beta 0.15, se 0.095), which is the signature of a handful of carriers rather than a pharmacogenomic effect. Nothing here supports a drug-caused, genetically modified adverse reaction.","gene_comment":"No gene is assigned, and correctly so: rs74612175 is an intergenic variant at chr10:91,124,837 (GRCh38) with a global MAF of about 0.14%, sitting in a gene desert whose nearest features are an unnamed lncRNA and the NUDT9P1 processed pseudogene, with PCGF5 the closest protein-coding gene roughly 38 kb away. ANKRD1 (cardiac ankyrin repeat protein) lies ~200 kb off, but at that distance and that allele frequency it is a coincidence of position, not a mechanism.","score_check":"The numbers do not hang together: a beta of 3.87 at a variant with MAF ~0.0014 implies an odds ratio near 50 for an extremely common nonspecific symptom code, which is not a believable effect size, while the same variant is indistinguishable from noise in the disease arm (|0.15| < 1.96 x 0.095), so the z_diff of 4.69 is driven wholly by the unstable within-drug estimate. Temporality of 100% over 146 dated records is the saturated value the cohort's record structure inflates and adds no independent evidence.","candidability":1,"candidability_reason":"Nonspecific symptom code with no mechanism, depleted rather than enriched in users, and a rare intergenic variant whose huge effect is a sparse-data artifact.","sources":[{"title":"Eszopiclone tablets, film coated - full prescribing information (DailyMed SPL)","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=2b5b4fbb-9b86-46c6-9508-37d3d9806dec"},{"title":"Ensembl REST: variation rs74612175 (chr10:91,124,837, intergenic, MAF 0.0014)","url":"https://rest.ensembl.org/variation/human/rs74612175?content-type=application/json"},{"title":"Ensembl REST: genes overlapping chr10:90.6-91.6 Mb (locus context for rs74612175)","url":"https://rest.ensembl.org/overlap/region/human/10:90600000-91600000?feature=gene;content-type=application/json"},{"title":"Sato et al., Associations of Benzodiazepine With Adverse Prognosis in Heart Failure Patients With Insomnia, J Am Heart Assoc 2020","url":"https://pubmed.ncbi.nlm.nih.gov/32200713/"},{"title":"Ho et al., Non-cardiac, non-oesophageal chest pain: the relevance of psychological factors, Gut 1998","url":"https://pubmed.ncbi.nlm.nih.gov/9771413/"}]}}}