Drug–Adverse Event Causality Assessment

Metoclopramide ↔ Tardive dyskinesia

This is the single largest PRR among the pairs on this page, and the underlying mechanism is textbook: metoclopramide is a dopamine D2-receptor antagonist, the same pharmacological class responsible for antipsychotic-induced tardive dyskinesia. FDA's 2009 boxed-warning decision followed continued spontaneous reports concentrated in patients who had taken the drug for more than 3 months, and published analyses since have identified metoclopramide as the single most common cause of drug-induced movement disorders overall — not just among gastric-motility agents, but across all drug classes. Risk rises sharply with treatment duration and cumulative dose, which is why the boxed warning specifically limits use beyond 12 weeks except in rare cases. This is one of the best mechanistically-understood adverse drug reactions in this dataset — a predictable, class-wide, dose/duration-dependent effect, not merely a statistical coincidence.

Regulatory basis: FDA boxed warning (2009), metoclopramide prescribing information.

This table is this page's own qualitative read of the literature cited above, not an automated or citation-gated assessment — see the note at the bottom of this page.

Clozapine ↔ Agranulocytosis / severe neutropenia

Clozapine-induced agranulocytosis is arguably the most consequential single adverse-drug-reaction discovery in modern psychiatry: it is the reason clozapine, despite being one of the most effective antipsychotics for treatment-resistant schizophrenia, can only be dispensed in the U.S. through a mandatory REMS program requiring regular absolute neutrophil count (ANC) monitoring for the life of the prescription — weekly for the first 6 months, biweekly for the next 6, then monthly if counts stay in range. Reported incidence is 1–2% of patients, overwhelmingly concentrated in the first 6 months of therapy, which is itself a form of temporal and dose-independent specificity (the event is a sudden, often zero-count crash, not a gradual dose-related decline). The mechanism remains formally unresolved — proposed as an idiosyncratic, likely immune-mediated toxic effect on bone marrow myeloid precursors or their progenitors — but the monitoring program's mandatory, indefinite existence is itself regulators' institutional verdict that the causal link is settled, not merely a statistical association.

Regulatory basis: Clozapine REMS program (mandatory ANC monitoring); clozapine prescribing information.

This page's own qualitative read, not an automated/citation-gated assessment.

Ciprofloxacin (fluoroquinolones) ↔ Tendon rupture / tendinitis

Fluoroquinolone-associated tendinopathy, most notoriously Achilles tendon rupture, carries an FDA boxed warning since 2008 that applies class-wide, not just to ciprofloxacin. This is one of the more mechanistically dense entries on this page: at least five distinct pathways have been proposed and partly confirmed in vitro, not just one. Ciprofloxacin upregulates matrix metalloproteinase-3 in tendon-derived cells while suppressing tenocyte proliferation and migration (via reduced FAK phosphorylation), directly degrading and under-replacing collagen. Separately, fluoroquinolones chelate magnesium and iron, cofactors tendon cells need to hydroxylate proline residues during collagen maturation — chelation leaves collagen fibers under-cross-linked and mechanically weaker, the same chelation chemistry responsible for the well-known interaction with magnesium-containing antacids. A third pathway implicates fluoroquinolone-driven reactive oxygen species production in tenocyte mitochondria. Clinical risk factors are well replicated across independent studies: age over 60, concurrent corticosteroid use, and solid-organ transplant history all multiply risk. This is a case where a drug class effect was identified, mechanistically investigated from multiple independent angles, and confirmed beyond spontaneous reports alone.

Regulatory basis: FDA boxed warning for tendinitis/tendon rupture (2008); fluoroquinolone class labeling, with further systemic safety-label restrictions added in 2016. Mechanism review: PMC4080593.

This page's own qualitative read, not an automated/citation-gated assessment.

Ranitidine (Zantac) ↔ Cancer (breast, prostate, bladder, colorectal, renal)

This is the clearest "mechanism, not mystery" entry on this page. Ranitidine itself is not believed to cause cancer; the problem was contamination. In 2019–2020, independent and FDA laboratory testing found that ranitidine can degrade — especially under heat or over time in storage — into NDMA (N-nitrosodimethylamine), a compound classified as a probable human carcinogen based on animal studies. The elevated cancer signals across multiple unrelated organ systems in this dataset (breast, prostate, colorectal, bladder, renal all flagged independently) is exactly the pattern a systemic carcinogen contaminant would produce, rather than the organ-specific pattern a drug's own pharmacology usually leaves. FDA requested market withdrawal of all ranitidine products in the U.S. in April 2020. This is a textbook example of a real, serious, and ultimately regulatory-confirmed signal — but the causal agent was a manufacturing/stability impurity, not ranitidine's intended pharmacological action.

Regulatory basis: FDA market withdrawal request for all ranitidine products (April 2020), citing NDMA contamination.

This page's own qualitative read, not an automated/citation-gated assessment.

Oxycodone ↔ Drug dependence / withdrawal syndrome

The largest report volume of any pair on this page, and the least scientifically contested: oxycodone is a full mu-opioid receptor agonist, and physical dependence and withdrawal on cessation are expected, dose- and duration-related pharmacological effects of that mechanism, not an idiosyncratic reaction. This pair is included less because it is surprising and more because its sheer scale in this dataset (over 100,000 co-reports) is itself a public-health data point, central to the opioid-crisis regulatory response, including REMS requirements for opioid analgesics and prescribing-guideline changes from CDC.

Regulatory basis: opioid analgesic REMS program; CDC clinical practice guideline for prescribing opioids.

This page's own qualitative read, not an automated/citation-gated assessment.

Risperidone ↔ Gynaecomastia

One of the largest PRR values on this page that is not circular or a coding artifact. Risperidone raises prolactin more than most other second-generation antipsychotics because, unlike many of its peers, it does not readily cross back out of the pituitary once it blocks dopamine D2 receptors there. Elevated prolactin drives breast tissue growth in males (gynecomastia) and galactorrhea in both sexes. This dose-related, mechanistically direct, and well-replicated across independent case series and the prescribing information itself.

Regulatory basis: risperidone prescribing information (hyperprolactinemia warning).

This page's own qualitative read, not an automated/citation-gated assessment.

TNF-inhibitors (adalimumab) ↔ Drug-induced lupus

TNF-alpha inhibitors (adalimumab, infliximab, etanercept) are well documented in the rheumatology literature to induce a lupus-like syndrome, with clinically evident cases in roughly 0.2% of treated patients across the class — strikingly close to this dataset's own trial-derived rate for adalimumab specifically (see the trial box below). Published reviews put infliximab and etanercept ahead of adalimumab in how often they trigger antinuclear/anti-dsDNA antibody formation, so adalimumab is, if anything, a conservative example of a class-wide effect rather than its worst case. Onset is typically around 11 months into therapy (joint and skin manifestations predominate: malar/annular rash, arthralgia, sometimes myositis), and symptoms resolve over three weeks to six months after the drug is stopped — a real dechallenge pattern, not instant, but documented. This is a recognized, labeled class effect, not specific to adalimumab.

Regulatory basis: TNF-inhibitor class labeling (autoimmunity/lupus-like reactions warning). Anti-TNF-α-induced lupus, review: PMC6852950.

This page's own qualitative read, not an automated/citation-gated assessment.

Erythropoiesis-stimulating agents (darbepoetin alfa) ↔ Mortality / cardiovascular risk

Unlike most entries on this page, the darbepoetin/death signal is backed by dedicated randomized controlled trials, not just spontaneous reports. The CHOIR trial (hemoglobin targets in chronic kidney disease, NEJM 2006) and the TREAT trial (darbepoetin alfa specifically, NEJM 2009) both found that targeting higher hemoglobin levels with an ESA increased cardiovascular events (including stroke, in TREAT) without a clear survival benefit, and ESAs also carry a labeled warning about shortened survival and faster tumor progression when used in certain cancer patients. FDA added a boxed warning and, for a period, required participation in a dedicated REMS program (ESA APPRISE) before this evolved into label-based risk communication. The raw "Death" PT in FAERS is non-specific, but the underlying causal story here is unusually well controlled compared to most of this page.

Regulatory basis: FDA boxed warning for ESAs; Pfeffer MA, et al. (TREAT). N Engl J Med. 2009;361(21):2019–2032; Singh AK, et al. (CHOIR). N Engl J Med. 2006;355(20):2085–2098.

This page's own qualitative read, not an automated/citation-gated assessment.

Nirmatrelvir/ritonavir (Paxlovid) ↔ COVID-19 recurrence ("Paxlovid rebound")

"Paxlovid rebound" — recurrence of COVID-19 symptoms or a positive test 2–8 days after finishing a 5-day nirmatrelvir/ritonavir course, sometimes after testing negative in between — became widely enough recognized that CDC issued a Health Alert Network advisory about it in May 2022. The leading mechanistic explanation is that the drug's antiviral course may be short relative to the time needed to fully clear virus in some patients, allowing a viral (and symptomatic) rebound once the drug is stopped, rather than true reinfection. It is included here as a genuinely well-publicized, real, and mechanistically plausible phenomenon, though the precise frequency and whether it differs meaningfully from the "rebound" some untreated COVID-19 cases also show is still an active area of study.

Regulatory basis: CDC Health Alert Network Advisory 467 (May 24, 2022): rebound reported 2–8 days after recovery, median duration 3 days, not attributed to reinfection or resistance.

This page's own qualitative read, not an automated/citation-gated assessment.

Varenicline (Chantix) ↔ Suicidal ideation / depression

This is the most instructive case on this page precisely because the story has two acts. In 2009, after a wave of post-marketing reports of neuropsychiatric events, FDA added a boxed warning to varenicline for suicidal thoughts and behavior — the kind of signal disproportionality data like this would have supported at the time. But FDA also required the manufacturer to run a large, purpose-built randomized controlled trial (EAGLES, published in The Lancet, 2016) directly comparing varenicline, bupropion, nicotine patch, and placebo for neuropsychiatric safety. EAGLES found no statistically significant increase in moderate-to-severe neuropsychiatric events for varenicline versus placebo, and FDA removed the boxed warning in December 2016. The spontaneous-report signal was real and worth investigating; the controlled trial it motivated did not confirm a causal drug effect at the magnitude first feared. This is the clearest example in this dataset of why a disproportionality signal is a prompt for better-controlled study, not a verdict on its own.

Regulatory basis: FDA boxed warning added 2009, removed December 2016, following the EAGLES trial (Anthenelli RM, et al. Lancet. 2016;387(10037):2507–2520).

Almost every viewpoint looked "met" from spontaneous reports alone in 2009; the actual experiment is what changed the picture. This page's own qualitative read, not an automated/citation-gated assessment.

Isotretinoin (Accutane) ↔ Depression / suicidal ideation

Isotretinoin carries a prominent psychiatric warning on its U.S. label, and individual case reports of depression and suicidal ideation following treatment are well documented and plausible (isotretinoin affects retinoid signaling pathways implicated in mood regulation). But the two largest cohort studies on this question directly disagree with each other, and that disagreement is the real story. Jick et al. (2000) followed 7,195 Saskatchewan isotretinoin users and 13,700 acne patients on oral antibiotics (plus a smaller UK cohort), comparing depression/psychosis diagnoses before vs. after exposure, and found relative risks around 1.0 — no detectable excess. Sundström et al., a Swedish registry cohort of 5,756 isotretinoin patients, found the opposite framing of the same question: the standardized incidence ratio for attempted suicide was already elevated in the year before treatment started (SIR 1.57 for all attempts), not just during or after it — the single clearest documented instance of the confounding-by-indication pattern this grade is built on (people with the most psychologically distressing acne are the ones referred for isotretinoin in the first place). Note also that completed suicide specifically is not a flagged signal in this dataset (PRR 1.3), which argues against the most severe form of the proposed association, even while the milder mood-symptom signals remain flagged. This page grades it as genuinely contested rather than established in either direction.

Jick SS, Kremers HM, Vasilakis-Scaramozza C. Isotretinoin use and risk of depression, psychotic symptoms, suicide, and attempted suicide. Arch Dermatol. 2000;136(10):1231–1236. PMID 11030769. Sundström A, et al. Association of suicide attempts with acne and treatment with isotretinoin. BMJ. 2010;341:c5812.

This page's own qualitative read, not an automated/citation-gated assessment.

Thalidomide ↔ Abnormal/increased blood hCG

These are, numerically, among the single highest PRR values in this entire dataset — and they are a near-certain reporting artifact, not a drug effect, included deliberately as a counter-example. Thalidomide is the drug most responsible for the existence of modern pharmacovigilance, after causing severe birth defects (notably phocomelia) in thousands of children in the late 1950s–60s when given to pregnant women for morning sickness. Precisely because of that history, its current U.S. uses (multiple myeloma, erythema nodosum leprosum) are restricted under the mandatory THALOMID REMS program, which requires routine pregnancy testing (serum/urine hCG) for every patient of childbearing potential before and during treatment. A hCG result flagged as "abnormal" or "increased" in that monitoring context is a required-testing artifact — it reflects the surveillance protocol built around the drug, not a new finding the drug caused. Tellingly, classic teratogenicity terms (phocomelia, congenital anomaly) do not even appear as flagged signals in this dataset, precisely because the REMS program has been so effective at preventing pregnancy exposure in the first place. The historical causal link for birth defects is not in question; it simply does not show up as a modern FAERS disproportionality signal, for the same reason the HPV/Pap-smear pattern does not reflect vaccine harm on the companion vaccine signal-pairs page.

Regulatory basis: THALOMID REMS program (mandatory pregnancy testing); historical teratogenicity established independently of FAERS, pre-dating the modern FAERS system.

This entry fails most Hill viewpoints despite a sky-high PRR — a clean illustration of why strength alone proves nothing. This page's own qualitative read, not an automated/citation-gated assessment.

Alendronate (bisphosphonates) ↔ Atypical femoral fracture

Alendronate is the archetypal oral bisphosphonate and first-line therapy for postmenopausal osteoporosis, and it genuinely reduces ordinary osteoporotic fractures. But prolonged use — characteristically more than three to five years of continuous exposure — is associated with a rare, distinctive low-energy subtrochanteric and diaphyseal femoral fracture pattern ("atypical femoral fractures"), mechanically different from the ordinary hip fracture the drug prevents. The FDA added a warning to the bisphosphonate class in October 2010, and the American Society for Bone and Mineral Research convened a task force the same year to standardize the case definition, precisely because the pattern is easy to miss. The FAERS term shown here ("femur fracture") is broader than the atypical pattern, so the raw count conflates the two — a genuine reporting limitation — but the specific atypical-fracture association is well supported in longitudinal cohort data and the label.

Regulatory basis: FDA Drug Safety Communication on atypical femur fractures with bisphosphonates (October 2010); ASBMR task-force case definition (2010); alendronate prescribing information (warnings on atypical subtrochanteric/diaphyseal fractures).

This page's own qualitative read, not an automated/citation-gated assessment.

Denosumab ↔ Osteonecrosis of the jaw

Denosumab is a RANKL inhibitor given as 60 mg every six months for osteoporosis and at 120 mg monthly in cancer/bone-metastasis settings. Osteonecrosis of the jaw (ONJ) — exposed, non-healing bone in the mouth — is a labeled risk for denosumab, just as it is for the bisphosphonates; it is best understood as a class effect of drugs that suppress bone resorption. Risk is strongly dose- and duration-dependent and is driven mainly by the oncology dosing, so ONJ is comparatively rare at osteoporosis dosing. This is a mechanistically coherent, clinically managed adverse effect, not a statistical artifact.

Regulatory basis: denosumab prescribing information (boxed warning for osteonecrosis of the jaw in oncology dosing); class effect shared with the bisphosphonates.

This page's own qualitative read, not an automated/citation-gated assessment.

Pioglitazone ↔ Bladder cancer

Pioglitazone, a thiazolidinedione for type 2 diabetes, was flagged for a possible bladder-cancer risk by a French national cohort in 2011, which led France to suspend the drug and prompted the FDA and EMA to add warnings and restrict its use (FDA label warning, 2011). Yet the picture has not resolved cleanly: several large subsequent observational studies and pooled analyses found little to no increase in bladder-cancer risk, and some of the FAERS over-reporting almost certainly reflects intensified bladder/prostate surveillance in this population plus the regulatory scare itself driving reporting. The honest grade is "contested," not "established" — a labeled caution resting on genuinely disagreeing evidence.

Regulatory basis: FDA drug safety communication and label change (2011); ANSM (France) suspension (2011); EMA review. The epidemiological evidence is mixed, with later large studies finding little to no elevated risk.

This page's own qualitative read, not an automated/citation-gated assessment.

Atorvastatin (statins) ↔ Type 2 diabetes mellitus

Pooled randomized-trial data found a small but statistically significant increase in new-onset diabetes among statin recipients (Sattar et al., Lancet 2010, PMID 20167359), and in 2012 the FDA added a warning about increases in blood glucose and HbA1c. The effect is real but modest, appears dose-related, and is generally judged to be outweighed by statins' cardiovascular benefit; whether it reflects a true diabetogenic action or the unmasking of already-predisposed patients remains debated. The very high FAERS PRR here overstates the clinical magnitude — statin users are heavily monitored, so even a modest true effect inflates further in spontaneous reporting.

Sattar N, et al. Statins and risk of incident diabetes: a collaborative meta-analysis of randomised statin trials. Lancet. 2010;375(9716):735-742 (PMID 20167359). FDA prescribing-information change on glucose/HbA1c (2012).

This page's own qualitative read, not an automated/citation-gated assessment.

Drospirenone/ethinyl estradiol ↔ Pulmonary embolism (venous thromboembolism)

Combined oral contraceptives are a well-recognized cause of venous thromboembolism, and the progestin component modifies that risk: national cohort data found drospirenone-containing combined pills carried a modestly higher VTE risk than levonorgestrel-containing pills (Lidegaard et al., BMJ 2009, PMID 19679613). The FDA revised the drospirenone labeling in 2011 to reflect this. The mechanism — an estrogen/progestin-driven procoagulant shift — is established pharmacology, and VTE is the classic serious adverse effect of combined hormonal contraception.

Lidegaard Ø, et al. Hormonal contraception and risk of venous thromboembolism: national follow-up study. BMJ. 2009;339:b2890 (PMID 19679613). FDA prescribing-information revision for drospirenone-containing contraceptives (2011).

This page's own qualitative read, not an automated/citation-gated assessment.

Omeprazole (proton-pump inhibitors) ↔ Chronic kidney disease

Proton-pump inhibitors are among the most-prescribed drug classes, and a widely-cited cohort analysis found that cumulative PPI exposure was associated with an increased risk of chronic kidney disease (Lazarus et al., JAMA Intern Med 2016). The FAERS pattern here spans the class — omeprazole with chronic kidney disease and acute kidney injury, lansoprazole with end-stage renal disease. The signal remains contested, however: confounding by indication and by baseline health is hard to exclude, effect sizes vary across studies, and some later analyses found little or no association. Acute interstitial nephritis is a recognized, likely-immune PPI reaction and provides a plausible mechanism, but the size of any chronic-kidney-disease risk is genuinely disputed.

Lazarus B, et al. Proton pump inhibitor use and the risk of chronic kidney disease. JAMA Intern Med. 2016;176(2). doi:10.1001/jamainternmed.2015.7193. Evidence contested; confounding by indication limits causal inference.

This page's own qualitative read, not an automated/citation-gated assessment.

Interferon beta-1a ↔ Influenza-like illness

Interferon beta-1a is a first-line injectable therapy for relapsing multiple sclerosis, and a flu-like illness — fever, chills, myalgia, malaise — is its single most characteristic and predictable side effect, typically worst in the hours after each injection and improving with continued use. It is the textbook "expected pharmacological effect": cytokine-mediated, dose-related, and managed with dose titration, evening dosing, and antipyretics. Its causality was never in question; the high report count mainly reflects how common and memorable the symptom is.

Regulatory basis: interferon beta-1a prescribing information (influenza-like symptoms listed among the most common adverse reactions).

This page's own qualitative read, not an automated/citation-gated assessment.

Capecitabine ↔ Palmar-plantar erythrodysaesthesia (hand-foot syndrome)

Capecitabine is an oral fluoropyrimidine used for colorectal and breast cancer, and palmar-plantar erythrodysesthesia — "hand-foot syndrome": painful redness, swelling, and peeling of the palms and soles — is its signature toxicity. It occurs in a substantial minority of treated patients, is clearly dose- and schedule-dependent, and is well characterized and manageable (dose reduction, interruption, emollients, pyridoxine). This is a mechanistically direct, expected effect of the drug's own pharmacology, not an idiosyncratic reaction.

Regulatory basis: capecitabine prescribing information (hand-foot syndrome listed among the most common adverse reactions); confirmed by a randomized comparator arm (§ trial data below). A class effect of the fluoropyrimidines.

This page's own qualitative read, not an automated/citation-gated assessment.

Alprazolam (benzodiazepines) ↔ Drug dependence / abuse

Alprazolam is a short-acting, high-potency benzodiazepine, and physical dependence with withdrawal on cessation is an expected, dose- and duration-related effect of the benzodiazepine class — the drug's own pharmacology, not an idiosyncratic reaction. In September 2020 the FDA updated the Boxed Warning for the entire class to address risks of abuse, addiction, physical dependence, and withdrawal. The high report count reflects both how widely these drugs are prescribed and how reliably dependence develops with sustained use.

Regulatory basis: FDA Drug Safety Communication, 23 September 2020 — Boxed Warning updated for the benzodiazepine class (abuse, addiction, physical dependence, withdrawal).

This page's own qualitative read, not an automated/citation-gated assessment.

Quetiapine (antipsychotics) ↔ Diabetes mellitus

Quetiapine is a second-generation antipsychotic, and weight gain, dyslipidaemia, and hyperglycaemia/diabetes are its best-known metabolic effects — shared across the class but differing in magnitude by molecule. The FDA required a class warning about hyperglycaemia and diabetes for all atypical antipsychotics in 2003. The mechanism (weight gain, insulin resistance, and possibly direct effects on glucose regulation) is well described, and the effect is dose- and duration-related and largely manageable with baseline and periodic metabolic monitoring.

Regulatory basis: FDA class warning on hyperglycemia and diabetes with atypical antipsychotics (2003); quetiapine prescribing information (metabolic monitoring).

This page's own qualitative read, not an automated/citation-gated assessment.

Risperidone ↔ Weight gain

Risperidone, like quetiapine, is a second-generation antipsychotic whose metabolic effects include substantial weight gain — a labeled, dose- and duration-related consequence of the class, with the risk varying meaningfully between agents. In a randomized head-to-head trial of risperidone versus quetiapine XR, "weight increased" was recorded in roughly 6% versus 5% of patients. The very high FAERS PRR reflects that weight gain is both common and reliably reported, but the effect itself is well established, not a statistical curiosity.

Regulatory basis: risperidone prescribing information (weight gain among the most common adverse reactions); FDA class metabolic warning for atypical antipsychotics.

This page's own qualitative read, not an automated/citation-gated assessment.

Docetaxel (taxanes) ↔ Alopecia

Docetaxel is a taxane chemotherapy, and hair loss is a near-universal, mechanistically direct consequence of its action on rapidly dividing cells — including the hair-follicle matrix. In a randomized trial, alopecia was recorded in roughly a third of patients on docetaxel versus under 2% on pembrolizumab, a striking and entirely expected difference. Alopecia is not merely cosmetic; it is among the most distressing chemotherapy side effects, and its severity is dose- and schedule-dependent.

Regulatory basis: docetaxel prescribing information (alopecia among the most common adverse reactions); confirmed by a randomized comparator arm (trial data below). A class effect of the taxanes and most cytotoxic chemotherapies.

This page's own qualitative read, not an automated/citation-gated assessment.

Tenofovir disoproxil fumarate ↔ Bone loss

Tenofovir disoproxil fumarate (TDF) is an antiretroviral used to treat HIV and hepatitis B, and its renal and bone toxicities are precisely why the newer prodrug TAF was developed. TDF reduces bone mineral density and raises markers of bone turnover, with the steepest declines in the first year of therapy and a persistently lower density thereafter; the FDA label carries explicit bone and renal warnings. The very high FAERS PRR reflects an established, mechanism-linked effect (proximal tubular dysfunction with phosphate wasting), though the term "bone loss" aggregates several coded MedDRA terms, so the raw count overshoots the specificity.

Regulatory basis: tenofovir disoproxil fumarate prescribing information (warnings on decreases in bone mineral density and renal toxicity); TAF developed specifically to mitigate these effects.

This page's own qualitative read, not an automated/citation-gated assessment.

Rituximab ↔ Serious infection

Rituximab is an anti-CD20 monoclonal antibody that depletes B cells, and an increased risk of infection — including serious bacterial, fungal, and viral infections, and notably reactivation of hepatitis B and progressive multifocal leukoencephalopathy (PML) — is a labeled, mechanistic consequence of that depletion. The FDA label carries boxed warnings for PML and HBV reactivation. The population treated (lymphoma, autoimmune disease, transplant) also raises baseline infection risk, so the FAERS count mixes drug-driven and disease-driven infection, but the causal contribution of B-cell depletion itself is well established.

Regulatory basis: rituximab prescribing information (boxed warnings for progressive multifocal leukoencephalopathy and hepatitis B virus reactivation; increased risk of serious infections).

This page's own qualitative read, not an automated/citation-gated assessment.

Morphine (opioids) ↔ Overdose / respiratory depression

Morphine is a full mu-opioid receptor agonist, and the dose-related triad of respiratory depression, sedation, and death in overdose is the defining, mechanistically direct hazard of the opioid class. The opioid-overdose epidemic is a public-health emergency addressed by the CDC clinical practice guideline, opioid REMS, and naloxone access. The FAERS term "overdose" here spans intentional and unintentional excess and reflects a well-established, dose-dependent risk rather than a statistical artifact — though it also aggregates suicides, misuse, and medication errors, so it is not a pure "adverse reaction."

Regulatory basis: opioid analgesic REMS; CDC Clinical Practice Guideline for Prescribing Opioids (2016, updated 2022); morphine prescribing information (boxed warning for respiratory depression and addiction).

This page's own qualitative read, not an automated/citation-gated assessment.

Medroxyprogesterone acetate ↔ Meningioma

Medroxyprogesterone acetate is a progestin used for contraception, abnormal uterine bleeding and, at high doses, as hormonal cancer therapy. The progestogen class has been linked to intracranial meningioma: a French national cohort found a dose-dependent association with high-dose cyproterone acetate (Weill et al., BMJ 2021, PMID 33536184), which prompted European reviews that extended the meningioma warning across several progestogens. Meningiomas frequently express progesterone receptors, giving a plausible mechanism. Whether the effect size for medroxyprogesterone specifically matches cyproterone's is not yet settled, so this is graded "emerging/contested" rather than established — and the extreme FAERS PRR (1,443) substantially overstates the true risk once awareness and increased imaging are accounted for.

Weill A, et al. Use of high dose cyproterone acetate and risk of intracranial meningioma in women: cohort study. BMJ. 2021;372:n37 (PMID 33536184). European regulatory reviews of progestogens and meningioma (from 2020 onward). Signal emerging; the effect size for medroxyprogesterone is not established.

This page's own qualitative read, not an automated/citation-gated assessment.

Estrogen + medroxyprogesterone acetate ↔ Breast cancer

Conjugated equine estrogen plus medroxyprogesterone acetate was the combined hormone-therapy regimen tested in the Women's Health Initiative, the large randomized trial whose 2002 report found an increased risk of invasive breast cancer (as well as cardiovascular events), leading to early termination of that arm and a sharp, worldwide drop in hormone-therapy use (Rossouw et al., JAMA 2002, PMID 12117397). The breast-cancer association is the best-established harm of combined postmenopausal hormone therapy, is duration-dependent, and prompted a boxed warning. Note that this is estrogen plus progestin therapy, not medroxyprogesterone used alone as a contraceptive — the two contexts differ.

Rossouw JE, et al. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results from the Women's Health Initiative randomized controlled trial. JAMA. 2002;288(3):321-333 (PMID 12117397). FDA boxed warning for estrogen+progestin postmenopausal therapy.

This page's own qualitative read, not an automated/citation-gated assessment.

Insulin glargine ↔ Hypoglycaemia (blood glucose decreased)

Insulin glargine is a long-acting basal insulin, and hypoglycaemia is its defining and most important adverse effect — a direct, dose-related consequence of the drug's own mechanism, not an idiosyncratic reaction. Every insulin product carries this risk, and it is the principal safety concern in insulin therapy, requiring careful titration and glucose monitoring. It is the textbook expected pharmacological effect: the drug's intended action (lowering glucose) can overshoot.

Regulatory basis: insulin glargine prescribing information (hypoglycemia listed as the most common adverse reaction and among the most important risks).

This page's own qualitative read, not an automated/citation-gated assessment.

Copper intrauterine device ↔ Foreign body in reproductive tract

This is the single highest PRR in the entire dataset (89,124), and it is a pure coding artifact — included deliberately as the extreme counter-example. A copper intrauterine device is a physical object placed in the uterus; when a report mentions "foreign body in reproductive tract," that is a literal description of the device itself, not an adverse reaction the device caused. The PRR is astronomically high precisely because the reaction term is essentially unique to this product class, which guarantees disproportionality by construction. It is the clearest possible reminder that PRR measures the distinctiveness of what gets reported, not harm — the highest number on the page is also one of the least meaningful.

Methodological note, not a regulatory finding: the "foreign body" term describes the intrauterine device itself (a physical implant), a coding artifact of the reporting scheme rather than a biological effect.

This page's own qualitative read, not an automated/citation-gated assessment.

Warfarin ↔ INR increased

Warfarin's effect is measured by the INR, and the INR is the test used to dose it; a report of "international normalised ratio increased" is therefore partly a description of the drug working and partly of the monitoring built around it, not a discrete adverse event in the usual sense. The over-representation of INR terms for warfarin is expected: the test is nearly exclusive to vitamin-K-antagonist therapy, which guarantees disproportionality by construction. Genuine supratherapeutic INR with bleeding is a real and important warfarin hazard, but the raw "INR increased" signal is a monitoring/coding artifact rather than a novel safety finding. Tellingly, INR terms also appear in the comparator arm of a modern anticoagulant trial (below).

Methodological note, not a regulatory finding: INR is the monitoring parameter used to dose vitamin-K antagonists; its over-representation is a coding/monitoring artifact. Supratherapeutic INR with bleeding remains a real, labeled warfarin risk (warfarin prescribing information).

This page's own qualitative read, not an automated/citation-gated assessment.

Seven large, high-PRR pairs that are really "what the drug treats"

This is the single most common pattern the widened scan surfaced, and it is a data-quality lesson rather than a safety signal: in each pair above, the "reaction" term is the disease the drug is prescribed to treat. Etanercept and dupilumab treat the exact conditions ("rheumatoid arthritis," "dermatitis atopic") listed as their top disproportionality signals; adapalene is a topical acne medication whose top "reaction" is acne; furosemide treats congestive heart failure; levetiracetam treats seizures. FAERS case-report forms often record the condition under treatment in the same reaction-term field used for genuine adverse events — sometimes because the report concerns lack of effect, disease progression despite treatment, or simply because the submitter filled in the indication by habit. A disproportionality algorithm cannot distinguish "the drug caused this" from "the patient had this, which is why they were prescribed the drug" from the term alone. This pattern alone accounts for a large share of the roughly 136,000 pairs the widened scan flagged, and is exactly why this page's selection criterion is the outside literature, not the PRR rank.

Methodological note, not a regulatory finding: a known FAERS data-quality pattern (indication/disease-under-treatment recorded in the reaction field), distinct from a true adverse-event signal.

No per-pair Hill table here — one would be identical for all seven and isn't informative. In Hill terms, every pair in this group fails the same two viewpoints for the same reason: specificity (the "reaction" is definitionally the condition the drug is used for, not a distinguishable outcome) and, in most cases, temporality (the condition usually precedes the prescription, not the reverse). That double failure is what the pattern is — it isn't a borderline call on any one viewpoint.

Extended scan: 24 more pairs, briefer read

The 33 cards above were selected for having an independent regulatory or trial basis. Widening further down the same volume-sorted candidate list surfaces mostly three patterns: genuinely well-established, mechanistically expected drug effects; indication/disease-under-treatment recorded as the "reaction"; and a few real but more nuanced or lower-confidence cases. Rather than pad this page with full essays for pairs that are mostly re-confirming the same two or three lessons above, here is the rest of the top-volume scan in one line each, graded honestly — including where this page's confidence is genuinely lower.

These 24 were drawn from the same volume-sorted, loosened-threshold scan (n ≥ 150, PRR ≥ 4, signal-flagged, deduplicated by reaction term and capped at two per drug) as the 33 full cards above. Several raw top-40 entries were left out as exact duplicates of a pair already covered above (e.g. a second oxycodone-combination dependence entry).