← Evidence MatrixResearch hub →

Open Source Medicine Foundation

Flagged VAERS Signal Pairs: Causality Assessment of the Clearest Signals

The statistics for every vaccine–event pair, with age, co-administration and reporting-era checks, are in the VAERS signal browser.

A rank-aggregated disproportionality scan across decades of VAERS flags thousands of vaccine–event pairs. Most are administrative noise. This page grades the handful that are genuinely clinical, highest-profile, or best-documented in the outside literature. Grades now come in two layers: hand-curated brand-level cards for the highest-profile named products, then a second, independently computed vaccine-class PRR scan that confirms the textbook anchors below.

Causality: well established

Rotavirus vaccine (RotaShield) ↔ Intestinal intussusception / GI haemorrhage

RotaShield ↔ GI haemorrhage: n=94 · PRR 343.7 · ROR 361.0 · EBGM 112.6 · agg. rank 2 of 200
Current productsnPRRROREBGMIC_LB (BCPNN)
RotaTeq ↔ Intussusception86722.222.77.94.02
Rotarix ↔ Intussusception16417.117.47.43.68

This is the textbook case of a VAERS disproportionality signal that was real and actionable. RotaShield, the first licensed rotavirus vaccine, was withdrawn from the U.S. market in 1999 within a year of launch after passive-surveillance reports of intussusception (a bowel obstruction) triggered a CDC/FDA investigation; subsequent case-control studies confirmed a clear excess risk, concentrated in the days immediately after the first dose. The vaccine was redesigned (RotaTeq, Rotarix), and both current products still carry a small, labeled, and actively monitored excess intussusception risk — smaller than RotaShield's, real, and disclosed on the package insert rather than merely a VAERS artifact. This is the rare case where a top-ranked VAERS signal was independently confirmed by controlled epidemiological study, not just a flagged association.

Not a current dossier topic on this site under the C1–C6 framework; see the Rotavirus vaccine page, Section 4 for this site's own adverse-event summary.

Causality: well established

Smallpox vaccine (ACAM2000) ↔ Myocarditis / Pericarditis

Pericardial disease: n=64 · PRR 555.5 · agg. rank 5 Myocarditis: n=113 · PRR 95.5 · agg. rank 27 Pericarditis: n=85 · PRR 82.2 · agg. rank 49 Myopericarditis: n=27 · PRR 161.1 · agg. rank 44

Myo/pericarditis after live vaccinia (smallpox) vaccination is one of the best-documented vaccine-cardiac signals in the literature, predating COVID-19 mRNA vaccine myocarditis by two decades. It was identified clearly during the U.S. military's 2002–2003 smallpox vaccination program, when systematic cardiac monitoring (the same monitoring protocol that also generates the many “normal troponin/ECG” entries elsewhere in this dataset) caught a signal CDC and ACIP subsequently acknowledged as a genuine, mechanistically plausible adverse effect of live vaccinia replication. ACIP guidance has since carried cardiac-risk-factor screening and contraindication language for ACAM2000 specifically because of this signal, which is a much stronger form of evidence than a VAERS rank alone — official recommendation language changed because of it.

Likely surveillance artifact, not a true AE signal

HPV vaccine (Gardasil) ↔ Abnormal Pap smear / HPV test positive / Cervical dysplasia

Smear cervix abnormal: n=388 · PRR 319.9 · agg. rank 24 Papilloma viral infection: n=244 · PRR 309.5 · agg. rank 34 HPV test positive: n=228 · PRR 327.0 · agg. rank 35 Cervical dysplasia: n=127 · PRR 349.1 · agg. rank 47

These are among the highest PRR values on this entire page — and almost certainly a confounding-by-indication / surveillance-bias artifact, not an injury signal. HPV vaccination and cervical cancer screening (Pap smears, HPV DNA testing) happen in the same population, often at the same clinical visit, and HPV vaccination does not prevent infection with strains a person was already carrying before the shot. A positive HPV test or abnormal smear reported in temporal proximity to vaccination mostly reflects pre-existing infection being detected by routine screening that coincides with the vaccination visit, not a new condition the vaccine caused. This is a deliberately included counter-example: a very high disproportionality score that illustrates why rank alone is not evidence of harm.

Investigated, not confirmed on follow-up

Lyme disease vaccine (LYMErix, discontinued) ↔ Arthritis / Polyarthritis

Arthritis: n=273 · PRR 50.1 · agg. rank 40 Polyarthritis: n=29 · PRR 101.4 · agg. rank 42

LYMErix (GSK/SmithKline Beecham) was a first-generation Lyme disease vaccine withdrawn from the U.S. market in 2002, three years after approval. Its withdrawal followed a reported signal of autoimmune arthritis, supported at the time by a published molecular-mimicry hypothesis (structural similarity between the vaccine's OspA antigen and human LFA-1, proposed as a trigger in genetically susceptible, HLA-DR4-positive individuals). The manufacturer faced litigation and discontinued the product amid falling sales and the controversy. However, the FDA's and independent epidemiological follow-up studies conducted after withdrawal did not find a confirmed population-level excess of autoimmune arthritis in vaccinated people. LYMErix is a genuinely contested historical case: a real regulatory and commercial withdrawal driven by a safety signal and a biologically plausible mechanism, where the best subsequent epidemiological evidence did not confirm the causal claim. It belongs on this page as an example of a signal that was taken seriously and specifically investigated, with an outcome more ambiguous than either the rotavirus or smallpox examples above.

Biologically plausible, narrower evidence base

Rubella-containing vaccine (Meruvax II) ↔ Osteoarthritis / arthropathy

Osteoarthritis: n=61 · PRR 74.0 · ROR 75.9 · agg. rank 32

Transient post-vaccination arthralgia and arthritis after rubella-containing vaccines (rubella component of MMR, or monovalent rubella vaccine) is a long-recognized reactogenicity pattern, described in the literature since the 1970s–80s, more common in post-pubertal women than in children. The acute, self-limited form has reasonably strong supporting evidence; chronic or recurrent joint symptoms are less consistently demonstrated across studies, and most reviews describe the association as biologically plausible (immune-mediated, similar to natural rubella infection's own joint-symptom pattern) but not as strongly or independently confirmed outside spontaneous-report data as the rotavirus or smallpox examples above.

See also the MMR vaccine page for VAERS category breakdowns that include arthralgia.

Expected reactogenicity, not a safety signal

DTaP+IPV+Hib (Pentacel) ↔ Post-tussive vomiting

n=109 · PRR 196.1 · agg. rank 26

Included as a contrast case: a very high disproportionality score attached to a mild, well-characterized, and expected reactogenicity event (vomiting following a coughing fit, a known feature of pertussis-containing vaccine reactogenicity profiles) rather than a serious or contested adverse event. High rank does not imply high severity.

Vaccine-class PRR confirmations of the 22 strongest textbook anchors

The pairs above are brand-level: single named products hand-picked from the rank-aggregated rankv_signals.json table. The cards below are a deliberately different, independently computed class-level scan: products were rolled up to their antigen class (influenza, MMR, varicella, HPV, SARS-CoV-2, rotavirus), each report reduced to its distinct MedDRA preferred terms, and PRR (with 95% CI), ROR and Yates’-corrected chi-square computed for every class ↔ event pair against all other classes as the comparator, on the full 1,038,044-report VAERS set. Only the 22 anchors that survive textbook/regulatory validation are shown; the raw scan flags 4,866 pairs and 490 clinically-strong ones that are retained as JSON audit data, not graded here. The rotavirus ↔ intussusception anchor (class PRR 684, n=164) is already the flagship card at the top of this page and is not repeated. Each chip below is one preferred term: n = distinct reports, PRR with 95% CI, ROR, χ2. These remain spontaneous-report associations on an “all other vaccines” baseline — not incidence rates, and not “vaccinated vs. unvaccinated.”

Causality: well established

Influenza vaccine (all classes) ↔ Guillain-Barré syndrome

Guillain-Barré syndrome: n=1,020 · PRR 9.06 (CI 8.35–9.82) · ROR 9.16 · χ2 4,122

GBS is the canonical influenza-vaccine signal. The 2012–13 NASEM/IOM monograph concluded evidence is “in favor of” a causal association between inactivated influenza vaccine and GBS, with an excess risk on the order of roughly one to two additional cases per million doses — small in absolute terms but real, time-locked to the weeks after vaccination, and mechanistically supported by molecular mimicry between vaccine components and peripheral-nerve gangliosides. The class-level PRR of about nine here is an association in spontaneous reports, not a rate: VAERS has no reliable denominator and is inflated by seasonal reporting surges, so read it as confirming that this textbook pair is among the most disproportionate in the dataset, not as an incidence estimate. For context, influenza infection is itself a far stronger GBS precipitant, so the net effect of vaccination is generally considered protective at the population level by preventing the infection that more often triggers GBS.

Causality: well established

MMR-containing vaccine ↔ Febrile convulsion (febrile seizure)

Febrile convulsion: n=90 · PRR 15.28 (CI 12.13–19.26) · ROR 15.36 · χ2 942

Fever is a predictable part of the measles and rubella immune response, and febrile seizures are a listed reaction: ACIP/CDC document roughly one additional febrile seizure per 3,000–4,000 MMRV doses (about double the rate of giving MMR and varicella as separate injections) in the 5–12 day window after the first dose. Febrile seizures are frightening to witness but, in an otherwise healthy child, are benign and are not associated with epilepsy or long-term neurodevelopmental harm. The strong class-level PRR (about 15) reflects exactly this well-characterized, time-locked reactogenicity pattern. Because varicella-containing combinations straddle both the MMR(MMRV) and varicella buckets, this signal is grouped with the measles/rubella response rather than counted twice.

Causality: well established

SARS-CoV-2 vaccine (predominantly mRNA) ↔ Myocarditis / pericarditis spectrum

Myocarditis: n=1,782 · PRR 7.63 (CI 6.30–9.25) · ROR 7.65 · χ2 602 Pericarditis: n=1,487 · PRR 5.09 (CI 4.27–6.05) · ROR 5.09 · χ2 417 Pericardial effusion: n=640 · PRR 5.53 (CI 4.20–7.29) · ROR 5.54 · χ2 187 Pericardial drainage: n=83 · PRR 9.87 (CI 3.62–26.91) · ROR 9.87 · χ2 29 Pericardial excision: n=35 · PRR 16.64 (CI 2.28–121.47) · ROR 16.64 · χ2 13 Viral myocarditis: n=26 · PRR 6.18 (CI 1.47–26.04) · ROR 6.18 · χ2 7

Myocarditis and pericarditis after SARS-CoV-2 mRNA vaccination — second doses, mostly males aged 12–29 — is now a regulatory-established association: it appears in FDA product labeling, was added as a listed adverse event after ACIP/CDC review of v-safe and VAERS data, and is corroborated by large published cohort studies. Most reported cases are mild and self-limited with recovery; a minority require hospitalization. This is the clearest example on the page of a genuine signal with a known absolute risk (on the order of tens to roughly one hundred-plus per million second doses in young males) that still yields only a modest class-level PRR (about 5–8), because the events are individually rare and because SARS-CoV-2 infection itself carries a substantially higher myocarditis risk than vaccination — and the comparator here is all other vaccine classes, never an unvaccinated baseline. The mRNA and adenoviral-vector platforms are not separable at this class level.

Causality: well established

SARS-CoV-2 vaccine (adenoviral vector) ↔ Cerebral / splanchnic thrombosis with thrombocytopenia (VITT)

Cerebral thrombosis: n=201 · PRR 47.78 (CI 11.87–192.38) · ROR 47.80 · χ2 89 Cerebral venous sinus thrombosis: n=148 · PRR 35.18 (CI 8.72–142.00) · ROR 35.19 · χ2 64 Mesenteric vein thrombosis: n=47 · PRR 22.35 (CI 3.08–161.96) · ROR 22.35 · χ2 19 Cerebral venous thrombosis: n=30 · PRR 7.13 (CI 1.70–29.84) · ROR 7.13 · χ2 9 Thrombotic thrombocytopenic purpura: n=42 · PRR 3.33 (CI 1.41–7.83) · ROR 3.33 · χ2 8

Vaccine-induced immune thrombotic thrombocytopenia (VITT, also called thrombosis with thrombocytopenia syndrome) — unusual-site thrombosis (cerebral venous sinus, splanchnic/mesenteric, arterial) with low platelets and anti-PF4 antibodies, days after an adenoviral-vector SARS-CoV-2 dose (AstraZeneca ChAdOx1, J&J Ad26.COV2.S) — is a mechanistically confirmed, label-recognized entity, and the reason several countries restricted these vectors to older age groups. It is a class-level “established” anchor even though the event terms themselves are rare. Note how the class-level PRR here is unusually high (20–48) relative to the myocarditis card: VITT concentrated in a small set of reports produces strong disproportionality. The wide 95% confidence intervals reflect the low absolute counts, and a residual limitation stands: this scan cannot separate adenoviral-vector from mRNA products within the single “covid” class, so the signal is diluted by the far more numerous mRNA reports in the dataset — which, if anything, makes the surviving thrombosis disproportionality more notable, not less.

Biologically plausible, narrower evidence base

Varicella (zoster) vaccine ↔ Herpes zoster reactivation spectrum

Herpes zoster: n=3,282 · PRR 8.54 (CI 8.19–8.91) · ROR 8.94 · χ2 14,040 Ophthalmic herpes zoster / HZO: n=76 & 30 · PRR 15.84 (CI 11.48–21.86) & 112.55 (CI 39.65–319.48) · χ2 506 / 376 Herpes zoster disseminated: n=21 · PRR 105.05 (CI 31.33–352.18) · ROR 105.08 · χ2 257 Herpes zoster oticus (Ramsay Hunt): n=21 · PRR 4.92 (CI 3.01–8.06) · ROR 4.93 · χ2 46

This is the single most disproportionate clinical cluster in the class scan by volume (tens of thousands of chi-square units), but the causality reading is more subtle than the raw numbers suggest, so it is graded plausible, not established. For the live attenuated Oka-strain varicella vaccine, a genuine post-vaccination herpes zoster can occur and is sometimes (not always) confirmed by genotyping to be vaccine-strain virus — a recognized, usually mild event. The far more-used recombinant subunit vaccine (Shingrix) contains no live virus and cannot cause zoster, so much of this volume is almost certainly background/shingles-in-the-recipient noise, age-related reactivation, and — in the live-vaccine buckets — wild-type strain reactivation, not vaccine-strain disease. The eye (ophthalmic/HZO) and disseminated and oticus (Ramsay Hunt) subtypes show extreme PRRs with wide CIs on small counts; these are real, mechanistically coherent phenotypes of zoster where they occur, but the scan cannot tell vaccine-strain from wild-type reactivation from MedDRA codes alone. Treat the cluster as a well-understood, largely benign zoster-reporting pattern concentrated in the varicella class — plausible and expected to appear here, not evidence of a dangerous new signal.

Expected reactogenicity, not a safety signal

HPV vaccine ↔ Vasovagal syncope / presyncope

Syncope: n=1,967 · PRR 4.63 (CI 4.43–4.84) · ROR 4.98 · χ2 5,182 Syncope vasovagal: n=52 · PRR 73.38 (CI 47.25–113.95) · ROR 73.55 · χ2 1,386 Presyncope: n=261 · PRR 3.15 (CI 2.78–3.57) · ROR 3.17 · χ2 357

Included to show that a very high class-level PRR can be entirely benign and expected. Syncope in the minutes after injection — a vasovagal faint, common in adolescents undergoing any needling procedure — is disproportionately reported with HPV vaccines because they are given precisely to that age group and because fainting after them was recognized early, prompting the standing ACIP/CDC recommendation to observe vaccinated adolescents for 15 minutes afterward and to vaccinate them seated or lying down. The mechanism is the common needle-faint reflex, not a toxic or immune-injury effect of the vaccine; the “syncope vasovagal” term reaching a PRR near 73 simply confirms the reports are coded to the reflex itself. The real clinical risk is the fall-injury from an unsupervised faint, which the observation period addresses — so this anchor is a success story for a mitigation that works, and a caution that disproportionality magnitude must never be read as harm. Note this is a distinct point from the separate HPV ↔ abnormal-Pap card above (a surveillance artifact); both show that high HPV PRRs reflect the vaccination context, not vaccine toxicity.

Part of the Open Source Medicine Foundation Network

Open Source Medicine Foundation

Home of the OSMF network — open-source pharmacovigilance tools and evidence-based medicine resources.

Research Tracker

Track peer-reviewed literature on vaccine safety, pharmacovigilance, and post-market surveillance.

Evidence Desk

Soft launch — provisional claim grades; not medical advice. Does not contain vaccine lot or VAERS data.