Do vaccines cause autism?

Anchor. Madsen et al., N Engl J Med 2002;347:1477-82, read in full (third-party PDF copy): all 537,303 children born in Denmark Jan 1991 to Dec 1998; adjusted relative risk of autistic disorder in vaccinated versus unvaccinated children 0.92 (95% CI 0.68 to 1.24) and of other autistic-spectrum disorders 0.83 (0.65 to 1.07); no association with age at vaccination (P=0.23), interval since vaccination (P=0.42) or calendar period (P=0.06) [Table 2 and text]. Hviid et al., Ann Intern Med 2019;170:513-20, read in full (third-party PDF copy): 657,461 children born in Denmark 1999 to 2010, 5,025,754 person-years, 6,517 autism diagnoses (129.7 per 100,000 person-years), fully adjusted HR 0.93 (0.85 to 1.02); "no increased risk for autism after MMR vaccination was consistently observed" in subgroups by sibling history, autism risk score or other vaccines, or in specified periods after vaccination; the authors describe their cohort as nonoverlapping with Madsen's and state the limitation that the date of first diagnosis is probably later than symptom onset, which can bias toward an effect; funded by the Novo Nordisk Foundation and the Danish Ministry of Health; protocol, code and data "not available" [read]. The two PDF copies were not compared with the publishers' own files (nejm.org and acpjournals.org refused this environment); their numbers match the Europe PMC abstracts.

Would change if. a register-based cohort of comparable size finds a hazard ratio clearly above 1 with a confidence interval that excludes 1; or the publisher's own files of Madsen 2002 or Hviid 2019 differ from the third-party copies read; or an independent re-analysis of the Danish registers fails to reproduce them

Anchor. Walker-Smith v GMC [2012] EWHC 503 (Admin), para 153: the paper's ethics statement "was untrue and should not have been included in the paper" [read]. The same judgment quashed the GMC's findings against Walker-Smith on other grounds (paras 186-187) and did not disturb this one. The GMC panel's own findings of fact (28 Jan 2010) were read in a third-party copy: charge 37 against Wakefield, "The Lancet paper stated that the investigations reported in it were approved by the Ethical Practices Committee of the Royal Free Hospital NHS Trust", found proved; "In fact, you did not have ethical approval for the investigations", found proved; the statement was "irresponsible" and "contrary to your duty to ensure that the information ... was accurate", found proved; but "dishonest" was found NOT proved, because "the Panel is not certain that you intended to deceive by inserting that statement in the Lancet Paper" [read]. The paper itself, read in a third-party copy of the Lancet PDF (round 7), carries under "Ethical approval and consent" the sentence "Investigations were approved by the Ethical Practices Committee of the Royal Free Hospital NHS Trust, and parents gave informed consent" [read].

Would change if. the Royal Free ethics committee records show the investigations were approved as the paper stated

Anchor. Gut biopsies from 25 children with autism and bowel symptoms and 13 controls with bowel symptoms only were tested for measles RNA in three laboratories blinded to diagnosis, one of them the laboratory that made the original finding. Measles RNA was found in 1 case (4%) and 1 control (8%). MMR before bowel-symptom onset: 12 of 25 cases vs 3 of 13 controls (P=0.13). The earlier report had found measles RNA in the ileum of 75% of children with autism vs 6% of controls [read, as cited in the paper].

Would change if. a blinded multi-laboratory study finds measles vaccine RNA in gut biopsies of children with autism and bowel symptoms at clearly higher rates than in controls

Anchor. AAP and US Public Health Service, MMWR 1999;48(26): "there are no data or evidence of any harm caused by the level of exposure that some children may have encountered"; nevertheless, "because any potential risk is of concern", thimerosal-containing vaccines "should be removed as soon as possible". The same statement says some children "could be exposed to a cumulative level of mercury over the first 6 months of life that exceeds one of the federal guidelines on methyl mercury" [read].

Would change if. a 1999 government document shows officials had evidence of harm when they made the recommendation

Anchor. Table 2: 23 autism cases among 16,511 vaccinated children and 1 among 1,957 unvaccinated; IRR 1.16 (0.16 to 8.62). Table 3: adjusted HR 0.62 (0.10 to 3.69), p=0.60. The authors write that no significant association was found between vaccine exposure and autism, adding that the number of cases was small [read].

Would change if. a later version of the study with more cases reports an autism association

Anchor. Transcript p. 44: "This is the result for autism, in which we don't see much of a trend except for a slight, but not significant, increase for the highest exposure. The overall test for trend is statistically not significant." The preliminary screen's signals concerned speech delay, attention deficit disorder and the neurodevelopmental group [read; full text searched, all eight autism passages read]. Later, limiting to children seen at least twice, the autism relative risk "is no different than one" and was replicated across case definitions [read]. Dr. Bernier, answering a question about copies, asks participants to treat the data as "embargoed" and "very highly protected" until a public release at the ACIP meeting on 21-22 June 2000, weeks later: a request for confidentiality ahead of a planned public release, which is what the transcript shows, not a decision to withhold [read]. The study was published in 2003 with both phases: no consistent associations, and no significant increase for autism in any analysis [abstract].

Would change if. the unread transcript pages show an autism finding presented as significant, or the published study is shown to have dropped such a finding

Anchor. Madsen et al., N Engl J Med 2002;347:1477-82, read in full (third-party PDF copy): all 537,303 children born in Denmark Jan 1991 to Dec 1998; adjusted relative risk of autistic disorder in vaccinated versus unvaccinated children 0.92 (95% CI 0.68 to 1.24) and of other autistic-spectrum disorders 0.83 (0.65 to 1.07); no association with age at vaccination (P=0.23), interval since vaccination (P=0.42) or calendar period (P=0.06) [Table 2 and text]. Hviid et al., Ann Intern Med 2019;170:513-20, read in full (third-party PDF copy): 657,461 children born in Denmark 1999 to 2010, 5,025,754 person-years, 6,517 autism diagnoses (129.7 per 100,000 person-years), fully adjusted HR 0.93 (0.85 to 1.02); "no increased risk for autism after MMR vaccination was consistently observed" in subgroups by sibling history, autism risk score or other vaccines, or in specified periods after vaccination; the authors describe their cohort as nonoverlapping with Madsen's and state the limitation that the date of first diagnosis is probably later than symptom onset, which can bias toward an effect; funded by the Novo Nordisk Foundation and the Danish Ministry of Health; protocol, code and data "not available" [read]. The two PDF copies were not compared with the publishers' own files (nejm.org and acpjournals.org refused this environment); their numbers match the Europe PMC abstracts.

Would change if. a register-based cohort of comparable size finds a hazard ratio clearly above 1 with a confidence interval that excludes 1; or the publisher's own files of Madsen 2002 or Hviid 2019 differ from the third-party copies read; or an independent re-analysis of the Danish registers fails to reproduce them

Anchor. 2004: "no causal link was established between MMR vaccine and autism as the data were insufficient", and The Lancet's admission that the authors had failed to disclose financial interests, as described in Indian J Psychiatry 2011 [read; the 2004 notices themselves were not read]. Signatories: 10 of the 13 authors, not Wakefield, Linnell or Harvey [author list read]. 2010: the Lancet retracted the whole paper, announced on 2 Feb 2010 (notice in the 6 Feb 2010 issue, Lancet 2010;375:445), after the GMC's findings of fact of 28 Jan 2010 ([2012] EWHC 503, para 1, read); the editor, Richard Horton, told the Guardian "the statements in the paper were utterly false" and "I feel I was deceived" [Guardian, 2 Feb 2010, read directly in round 3]. The notice itself was NOT read, so its wording is not quoted here; the earlier pass's quotation of it (via BMJ c696, "contrary to the findings of an earlier investigation", naming the referral and ethics-approval claims) could not be re-checked and the gloss is not used. For context, the Guardian reports that in 2004 the Lancet asked the Royal Free to investigate and the Royal Free's Prof. Hodgson wrote that it "had found no problems": the 2004 inquiry, not the GMC hearing, is the "earlier investigation" in that account. 2011: the BMJ records comparison and the editors' fraud judgment [read via fetch tool; see va-wakefield-records-mismatch].

Would change if. the Lancet notices themselves give different reasons from how they are quoted

Anchor. [2012] EWHC 503 (Admin), Mitting J, 7 March 2012: "The panel's determination cannot stand. I therefore quash it" (para 187); Wakefield "has subsequently abandoned his appeal" (para 1); the GMC hearing began 16 July 2007 and decided on 24 May 2010 (para 1) [all read].

Would change if. a later court or GMC decision reverses either outcome

Anchor. NHS Immunisation Statistics, England 2004-05, Table 2: 91.8% (1995-96), 88.3% (1998-99), 84.1% (2001-02), 81.8% (2002-03), 79.9% (2003-04), 80.9% (2004-05) [read]. Eurosurveillance, 3 July 2008, reporting the HPA: measles endemic again 14 years after local transmission had stopped; 461 confirmed cases in England and Wales to the end of May 2008 [read]. Notifications in England and Wales: 3,670 (2007), 5,088 (2008) [read, UKHSA table].

Would change if. the published coverage figures are revised

Anchor. Denmark, 467,450 children: autism RR 0.85 (0.60 to 1.20), and 0.98 (0.90 to 1.06) per 25 micrograms of ethylmercury [abstract]. US case-control with clinically confirmed cases (256 cases, 752 controls): no increased risk at any exposure window [abstract]. California caseload kept rising through 2007 after removal [abstract]. Danish incidence rose after removal in 1992 [abstract]. Independent populations: Denmark (Hviid 2003; Madsen 2003) and the US (Price 2010; Schechter 2008).

Would change if. a cohort with individual exposure records finds autism rising with thimerosal dose, or diagnoses fall in birth cohorts after removal once reporting changes are accounted for

Anchor. Andersson et al., Ann Intern Med 2025: 1,224,176 children born 1997-2018; any neurodevelopmental disorder HR 0.93 (0.90 to 0.97) per mg [abstract]; autism spectrum disorder HR 0.93 (0.89 to 0.97) [secondary: ACP press release; not in the abstract]. A 2026 BMJ systematic review found no association [abstract].

Would change if. the paper's full text, its published correction or its updated supplement shows an autism association in the main analysis (a 2026 reanalysis says the updated supplement does); or another cohort with recorded doses finds one

Anchor. DeStefano, Price, Weintraub 2013 (256 cases, 752 controls): adjusted OR per 25-unit increase in antigens 0.999 at 3 months, 7 months and 2 years, with narrow intervals [abstract].

Would change if. a second study with recorded doses finds an association with total antigen exposure

Anchor. Ozonoff et al. 2010: 25 infants later diagnosed with autism and 25 typically developing infants, assessed at 6, 12, 18, 24 and 36 months. Gaze to faces, social smiles and vocalizations were highly comparable at 6 months and declined afterwards in most of the autism group (19 of 22). By parent report, 11 showed clear signs before the first birthday and no regression, and 4 lost skills [read].

Would change if. larger prospective studies find signs reliably present from birth or at 6 months

Anchor. Thompson, 27 Aug 2014: "my coauthors and I omitted statistically significant information in our 2004 article" [read, third-party copy of the law firm's page]. The CDC's archived statement agrees that race was reported only for the birth-certificate sample and says the data remain available for reanalysis [read]. The paper's whole sample: 624 cases, 1,824 controls; no significant association for vaccination before 18 or 24 months, including in children with developmental regression; vaccination before 36 months OR 1.49 (1.04 to 2.14), attributed by the authors to vaccination required for early-intervention programs [abstract].

Would change if. the study protocol or the analysis files show a different reporting decision

Anchor. Salon, "Correcting our record", 16 Jan 2011: critics had eroded its faith in the story, and deleting it was the best service to readers. The 2005 corrections included a mercury figure that overstated exposure against the EPA limit and a wrong statement that the rotavirus vaccine contained thimerosal [read].

Would change if. Salon restores the article

Anchor. Bruesewitz v. Wyeth (2011), majority opinion: the Act "establishes a no-fault compensation program" designed to work faster and with greater ease than the tort system; for Table injuries "no showing of causation is necessary; the Secretary bears the burden of disproving causation"; for unlisted injuries "the claimant must prove causation"; claimants need not show the vaccine was defective [read]. The syllabus: compensation is paid out of a fund created by an excise tax on each vaccine dose [read].

Would change if. Congress changes the Act's burden of proof

Anchor. Special master's order of 27 Mar 2015, docket 08-504V (a different family's case; read in full, govinfo.gov), footnotes 6, 11 and 12 and the text (footnote 6 also says that while the case was part of the Omnibus Autism Proceeding, "the theory of the case asserted" was "that thimerosal-containing vaccinations can cause autism"): the respondent's Rule 4(c) report said that "based on an initial review" the facts "met 'the statutory criteria for demonstrating that the vaccination [the child] received on July 19, 2000, significantly aggravated an underlying mitochondrial disorder'"; the petitioners "specifically alleged a Table Injury" of "a presumptive MMR vaccine-related injury of an encephalopathy ... [that] eventually manifested as a chronic encephalopathy with features of autism spectrum disorder and a complex partial seizure disorder as a sequel" (quoting the 2011 fees decision); the case "was settled, not adjudicated"; the Decision Awarding Damages (21 Jul 2010) says only that the respondent "elect[ed] not to challenge petitioner's claim", which the order says "cannot be conflated into either a concession of the theory writ large or a judicial determination"; the respondent says it "has never conceded (in [that case] or otherwise) that any vaccine can cause significant aggravation of a pre-existing mitochondrial disorder" and that its recommendation to compensate "was unrelated to the validity of a particular causal theory"; the case "was originally going to be an OAP test case before it settled" and the Petitioners' Steering Committee "later determined that it should not be so designated". CNN (6 Mar 2008): the program concluded in November 2007 that the child's underlying illness was "significantly aggravated" by vaccines; her case had been designated a potential test case before the concession, made shortly before a deadline for expert testimony (the family lawyer's account); CDC director Julie Gerberding said the government had made "absolutely no statement indicating that vaccines are a cause of autism" [read directly in round 3]. NPR (7 Mar 2008): lawyers for HHS "agreed to pay" and "in its ruling, the government said it was possible that the vaccines [the child] received may have exacerbated that disorder" [read directly in round 3]. The two reports therefore word the concession differently ("concluded ... significantly aggravated" versus "possible ... may have exacerbated"); the statement above gives both. A fuller phrase sometimes attributed to the Rule 4(c) report ("features of autism spectrum disorder") is, in the court order, the petitioners' own allegation as quoted from a 2011 decision, not the government's wording. The Rule 4(c) report and the 2010 decision themselves were not read (va-sq-compensation-decision).

Would change if. the Rule 4(c) report or the damages decision itself says something different from how a court order and the news reports quote it

Anchor. Federal Circuit, Cedillo v. HHS, No. 2010-5004 (decided 27 Aug 2010) [opened; passages read]: counsel for the Petitioners' Steering Committee "proposed a 'test case' approach"; Cedillo, Hazlehurst and Snyder "were also designated as test cases"; in footnote 1, "The Special Masters in Hazlehurst and Snyder denied the respective petitions and the Court of Federal Claims affirmed in both cases. Snyder ... was not appealed to this court"; the court affirmed in Cedillo and, "while this appeal was pending", in Hazlehurst (604 F.3d 1343, 13 May 2010). Federal Circuit, Hazlehurst v. HHS, No. 2009-5128 [opened]: the omnibus proceeding was "an effort ... to identify and adjudicate certain test cases as a means of addressing the principal issues in approximately 5,000 autism claims that have been filed under the Vaccine Act"; the Cedillo and Snyder petitioners presented a theory based on the combined effect of MMR and thimerosal-containing vaccines; the Hazlehursts "initially presented that theory of causation, but in post-hearing briefing they relied on the theory that [the child's] autism was caused by the MMR vaccine alone"; the court affirmed. In Cedillo (10-5004) the special master "concluded that the evidence did not demonstrate that the MMR vaccine alone or in combination with TCVs can cause autism in general, or that the MMR vaccine alone or in combination with TCVs caused" the child's autism [read]. The special masters' decisions, opened directly from uscfc.uscourts.gov (captions, theories and conclusions read; the evidentiary bodies, 120 to 310 pages each, were not): Cedillo (98-916V, 12 Feb 2009, Hastings): "The overall weight of the evidence is overwhelmingly contrary to the petitioners' causation theories"; Hazlehurst (03-654V, 12 Feb 2009, Campbell-Smith): caption "Failure to Prove that the Combination of MMR and Thimerosal-Containing Vaccines Causes Autism"; Snyder (01-162V, 12 Feb 2009, Vowell): compensation "DENIED"; King (03-584V, 12 Mar 2010, Hastings): "the petitioners in this case are not entitled to a Program award"; Mead (03-215V, 12 Mar 2010, Campbell-Smith): "Petitioners' claim is dismissed"; Dwyer (03-1202V, 12 Mar 2010, Vowell): "the petition for compensation is therefore DENIED". The court's background sheet, read in full: the Court of Federal Claims judges affirmed the three first-theory decisions (Cedillo, 89 Fed. Cl. 158; Hazlehurst, 88 Fed. Cl. 473; Snyder, 88 Fed. Cl. 706); Cedillo and Hazlehurst were appealed and Snyder was not; the first-theory records held 939 medical articles, 50 expert reports and 28 testifying experts; the second-theory hearings ran in May and July 2008. Whether the petitioners in King, Mead and Dwyer sought review was not found in what was read. Bruesewitz dissent n.25 records that "Respondent notes that there are some 5,000 petitions" pending: that is the respondent's figure, quoted by the dissent, not a finding. CNN 2008 (the family lawyer): "nearly 5,000 other autism claims are pending". The figures "more than 5,000 filed" and "about 4,800 pending" were found in no source read and are dropped.

Would change if. a later decision reopens the theories on new evidence

Anchor. Bruesewitz v. Wyeth LLC, 562 U.S. 223 (2011), 6-2: the 1986 Act preempts all design-defect claims [read]. "Autism" appears only in the dissent, footnote 25 [read].

Would change if. a later Supreme Court case addresses autism

Anchor. 42 CFR 100.3 (version effective 3 Jan 2022): no entry contains "autism"; the MMR entry lists encephalopathy or encephalitis at 5-15 days [read]. A search of the Federal Register for rules amending the Table since 2023 found none [searched].

Would change if. HHS amends the Table

Anchor. Geier v. Maryland State Board of Physicians, 223 Md. App. 404, 116 A.3d 1026 (Court of Special Appeals, No. 1095, Sept. Term 2014, filed 29 May 2015), the appeal of the licence revocation [read in full]: the Board's Final Decision and Order of 22 Aug 2012 revoked the licence. The Board found that he treated patients with Lupron, "a medication that was not approved by the U.S. Food and Drug Administration ... for use on children in the absence of precocious puberty", without an adequate examination; that he prescribed chelation "without documenting a reason ... and without adequate documented informed consent", which "violated the standard of quality care"; and, in setting the sanction, that he provided "informed consent" forms that were "misleading and in at least one case blatantly false". The circuit court affirmed on 9 Apr 2014. The charges included failing to keep adequate medical records. The 1 Oct 2015 opinion (Nos. 722 and 2256, a later suit by the Geiers against the Board) gives the summary suspension of 27 Apr 2011 and the same dates, and the 2019 opinion (No. 338) describes the Board's conclusion in similar words [read]. The Board's own orders, fetched from the Board's site (scanned PDFs, read by OCR, so wording is taken from the court opinions where it is quoted): its Final Decision and Order of 22 Mar 2012 in cases 2007-0083, 2008-0454 and 2009-0308 upheld the summary suspension after a full hearing (read from OCR of the 83-page scan; two further statements were read on pages 8 and 9 in two OCR passes, 130 and 300 dpi, that agree, so they are reliable, but OCR errors remain possible elsewhere in the order): the Board "agrees with the ALJ's assessment of all of the evidence presented", and a footnote says that Lupron treatment "carries a very high risk of skin abscesses and infections, and it is contraindicated in patients with a history of seizures" and that he prescribed it for a patient with uncontrolled seizures; its Final Decision and Order in case 2012-0223 (after 27 Feb 2013), footnote 1, says: "On August 22, 2012, the Board issued a Final Decision and Order in another case revoking Dr. Geier's license" [read]. The 22 Aug 2012 order itself was not reached (see va-sq-geier-revocation-order).

Would change if. the revocation is overturned

Anchor. Hansen et al. 2015: 677,915 Danish children; 60% of the increase attributable to the two reporting changes [abstract]. US special education 1994-2003: autism rose from 0.6 to 3.1 per 1,000 while intellectual disability and learning disability declined [abstract]. CDC surveillance (8-year-olds): 6.7 per 1,000 in 2000 to 32.2 per 1,000 (one in 31) in 2022, across a series whose definition, sources and sites changed (DSM-5 from 2016; record-based case definition from 2018; 6 to 16 sites) [abstracts]. DSM-5 (2013) merged several diagnoses into one autism spectrum [read].

Would change if. a study using constant criteria and case-finding shows a rise of similar size

Anchor. Sandin et al. 2017: 37,570 twin pairs and over 2.6 million sibling pairs; heritability 0.83 (0.79 to 0.87) [read]. Bai et al. 2019: 2,001,631 children in five countries; median 80.8% (73.2% to 85.5%), from 50.9% (Finland) to 86.8% (Israel) [read].

Would change if. large family studies with different designs give much lower estimates

Anchor. Modabbernia et al. 2017: each 10 years of maternal or paternal age raises risk by 18% and 21%; all studies of prenatal valproate found strong associations, likely causal; vaccination, thimerosal and maternal smoking classed as unrelated [read].

Would change if. newer reviews reverse these associations

Anchor. Chess 1977 and Chess, Fernandez, Korn 1978: 243 children with congenital rubella; 7% had autism at preschool age; at re-examination of 205 at ages 8 to 9 there were two remissions and three new cases [abstracts]. Berger, Navar-Boggan, Omer 2011: a simple mathematical model (7.4% of congenital rubella cases presenting with autism, range 3.0% to 10.0%) estimates 16,600 congenital rubella cases and 1,228 autism cases prevented in the US in 2001-2010 [read].

Would change if. re-examination of congenital rubella cohorts finds autism no more common than expected

Anchor. cdc.gov/vaccine-safety/about/autism.html, dated 22 Jul 2026: the key point quoted above; also that studies supporting a link "have been ignored by health authorities"; the heading "Vaccines do not cause Autism*" is kept, with a footnote that scientists have not identified the root causes of autism [read via fetch tool]. First changed 19-20 Nov 2025; Kennedy said he ordered it [secondary: STAT, 21 Nov 2025]. Revised again about 22 Jul 2026 [secondary].

Would change if. the page is changed again

Anchor. [2012] EWHC 503 (Admin), para 156 (quoting the panel's charges 31-32 against Walker-Smith): the panel found that "consecutively referred" implied a routine referral, that the referrals of four children (1, 5, 9, 10) were not routine, and that the description was "irresponsible" and "misleading" [read]. Para 157: the phrase "means no more than that the children were referred successively, rather than as a single batch ... The words did not imply routine referral", and the paper "made it clear that this was not a routine referral. It was a referral generated by the concerns of parents about a possible link" [read]. Para 158: that four referral letters did not mention intestinal symptoms was "factually accurate ... but of no significance", because Walker-Smith elicited gastrointestinal symptoms at clinic [read]. The 2010 Lancet notice is reported to name the statement as incorrect [secondary: BMJ c696, not re-read; the notice was not read]. The GMC panel's findings of fact, read in a third-party copy: charge 33, the paper stated the children were "consecutively referred ..." and described them as a "self referred" group, admitted and found proved; the panel was "satisfied that a general reader would interpret the wording ... to mean that the children had been referred to the gastroenterology department with gastrointestinal symptoms, and that the investigators had played no active part in the referral process"; charge 34, the referrals of children 1, 9, 5 and 10 "did not constitute routine referrals" and those of children 2, 9, 5 and 12 "included active involvement in the referral process by you" (Wakefield); the description of the referral process was therefore "irresponsible", "misleading" and "contrary to your duty to ensure that the information in the paper was accurate", each found proved (charge 34.c, against Wakefield); and his letter to the Lancet of 2 May 1998 saying the children were referred "through the normal channels" was found "dishonest" (charge 35) [GMC findings, read].

Would change if. the Lancet's 2010 notice is read and says something other than how it is reported, or the GMC's own file differs from the third-party copies read

Disputed by.

• Mr Justice Mitting, High Court of England and Wales: 'Consecutively referred' meant only that the children were referred successively; the panel's contrary reading was rejected for Walker-Smith.
  src-walker-smith-2012
• The Lancet editors (2010 retraction, as reported): The 'consecutively referred' statement was incorrect.
  src-lancet-retraction-2010 (source not read here)

Anchor. Henry Ford Health (26 Sep 2025): the draft was shelved on first internal review for serious data and methods problems, listing unequal groups and much shorter follow-up for the unvaccinated children [read via fetch tool]. Aaron Siri's Senate testimony says the authors feared professional consequences [read via fetch tool]. The authors did not respond to Michigan Public [secondary].

Would change if. the internal peer-review record or the authors' own account is published

Disputed by.

• Henry Ford Health: Shelved on internal review for data and methods problems.
  src-henry-ford-statement
• Aaron Siri: The authors feared professional consequences.
  src-siri-testimony

Anchor. Rests on the 1998 case series of 12 children, in which parents (or, in the full text, parents or physicians) linked onset to MMR in 8, and on later parent surveys. See va-wakefield-paper-account-accurate and va-vax-unvax-surveys.

Would change if. a large cohort with recorded vaccination and diagnosis finds a clear excess after MMR, or a mechanism is shown in controlled human or experimental data

Anchor. Deer, BMJ 2011;342:c5347, summary box: "Three of nine children reported with regressive autism did not have autism diagnosed at all"; five of the 12 described as previously normal "had documented pre-existing developmental concerns"; some onsets reported as within days of MMR were recorded as months later [read via fetch tool]. Godlee et al., BMJ 2011;342:c7452: none of the 12 cases was free of misrepresentation or undisclosed alteration [read via fetch tool]. What the paper itself says, read in a third-party copy of the Lancet PDF (round 7): its Summary reports "Behavioural disorders included autism (nine), disintegrative psychosis (one), and possible postviral or vaccinal encephalitis (two)" and that onset of behavioural symptoms was associated by the parents with MMR in eight of the 12 children; its Table 2 lists child 8 as "Post-vaccinial encephalitis?" and child 10 as "Post-viral encephalitis?", not as autism [read]. The BMJ's comparison with the hospital records is therefore a comparison with these stated diagnoses; the records themselves are not available (va-sq-bmj-2011-records).

Would change if. the children's records, or a second independent comparison of them, show the published account was accurate

Anchor. Mold et al. 2018 measured aluminum in brain tissue from 5 donors with autism (mean 2.30 to 3.82 micrograms per gram by lobe, with standard deviations larger than the means in three lobes) and imaged tissue from 10 donors. There was no control tissue from people without autism, and the paper did not trace the aluminum to any source [abstract and article page].

Would change if. a controlled study shows higher brain aluminum in autism than in matched controls and traces it to vaccines

Anchor. Parents' reports of onset after vaccination, as in the 1998 paper (8 of 12).

Would change if. self-controlled case-series show onset or regression clustering in the weeks after vaccination beyond what age alone predicts

Anchor. Mawson 2017: anonymous online survey of homeschooling mothers, 666 children; autism 19/405 vaccinated (partly or fully) vs 3/261 unvaccinated, OR 4.2 (1.2 to 14.5); the ADHD row of the same table shows identical figures; diagnoses could not be checked against records [read]. Hooker & Miller 2021: voluntary parent survey, 1,565 children of whom only 136 fully vaccinated; autism OR 5.03 (1.64 to 15.5) for fully vaccinated; in the subset whose diagnoses were confirmed in the participating practices' electronic records, autism 27/403 in children with any vaccine vs 3/737 unvaccinated, OR 13.0 (3.73 to 45.0); no correction for multiple tests; recall and selection bias acknowledged [read].

Would change if. a record-based study with independently confirmed diagnoses, adjusted for healthcare use, reproduces an excess of this size

Anchor. Both trends rose over the same decades; see va-rise-mostly-reporting.

Would change if. diagnoses fall in birth cohorts after a vaccine is withdrawn, once reporting changes are accounted for

Anchor. Rossignol & Frye 2012 (meta-analysis): mitochondrial disease in 5.0% (3.2% to 6.9%) of children with autism, against about 0.01% in the general population; in one series of 28 children with both, 17 regressed, 12 of them with a fever over 101F, including 4 of those 12 after a fever following routine vaccination, and one further child in a case report regressed after a post-vaccination fever; the reviewers state it is not clear whether mitochondrial dysfunction contributed to or caused the regression [read]. Shoffner et al. 2010 (that series): none regressed with vaccination unless a fever was present [abstract]. The review's reference 25 (Weissman et al. 2008, read in full) says that in "one of our 25 patients" the deterioration "appeared to follow vaccination", that "such timing does not prove causation", and that "large, population-based studies will be needed"; its reference 85 (the 2006 case report) was read as an abstract only.

Would change if. a prospective study of children with confirmed mitochondrial disease records fevers from all causes and finds regression clustering after them, or finds none

Next step. Prospective follow-up of children with confirmed mitochondrial disease, recording every fever with its cause (infection or vaccination) and any regression, with blinded assessment; test whether post-vaccination fevers differ from infection fevers.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Searched: Europe PMC, 2 Oct 2026, title or abstract containing "vaccination schedule", "immunization schedule" or "vaccine schedule" and autism*, published 2010 to 2026: 22 hits, screened by title. None is a study of autism diagnoses across children grouped by whole-schedule exposure pattern (the hits are uptake, attitude, mechanism and evidence-review papers). A title screen can miss a study whose title does not say so. The IOM's 2013 statement that few studies had assessed the entire schedule is adoption, in adoption_note.

Would change if. a study of whole-schedule exposure patterns is published

Next step. Compare autism diagnoses across children grouped by their recorded schedule pattern (on time, delayed, spaced, partial) in a register or Vaccine Safety Datalink cohort, adjusting for healthcare use, with the analysis plan published before the data are read.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. The MMR cohorts compare MMR with no MMR, not all vaccines with none. Hviid 2019 (read in full) states that "a general criticism of observational vaccine effect studies is that they do not include a completely unvaccinated group of children", that the number completely unvaccinated throughout childhood "will be low in a country such as Denmark", and reports that in children with no DTaP-IPV/Hib vaccination in the first year of life "we found no support for an association" between MMR and autism: a vaccine-naive-in-infancy subgroup, not a fully unvaccinated one. The record-based Henry Ford draft is unpublished and small (1,957 unvaccinated, 1 autism case). The surveys are self-selected. Searched: Europe PMC 2010-2026 for vaccinated/unvaccinated and autism in titles (33 hits reviewed by the MMR source pass).

Would change if. such a study is published

Next step. Use a national register (Denmark, Norway or Finland) to compare children with no recorded vaccinations against vaccinated children, matched on birth year, region and healthcare use, with diagnoses from specialist records; publish the plan first. A randomized trial is ruled out by ethics, so this is the strongest feasible design.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Hooker 2014 reanalyzed the data as a cohort with Pearson chi-squared tests instead of the matched design and reported RR 3.36 (p=0.0019) for African-American boys vaccinated before 36 months. The journal retracted it on 3 Oct 2014 for undeclared competing interests that compromised peer review and for concerns about the validity of the methods and statistics [both read].

Would change if. an independent reanalysis with the original matched design is published

Next step. Independent reanalysis of the CDC restricted dataset by a group with no stake, using the original matched case-control design and a pre-registered plan, including race and sex strata; then check the result against a register cohort with race recorded.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Hansen 2015: 60% explained, interval 33% to 87%; the rest unexplained [abstract].

Would change if. a study accounts for the remainder

Next step. Repeat the Danish decomposition for birth cohorts after 1991 with later criteria changes, and add measured candidate factors (parental age, preterm survival, awareness and screening) to see how much of the remainder they explain.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Heritability is high but explains variation, not individual cases (see va-heritability-high).

Would change if. large genetic and exposure cohorts assign causes to most cases

Next step. Combine genome sequencing with prospectively recorded exposures (including vaccination dates and fevers) in existing birth cohorts such as Norway's MoBa and the US SPARK study, so specific causes can be tested case by case.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Hviid 2019: HR 0.93 (0.85 to 1.02), an upper bound of about a 2% relative increase [abstract]. Named subgroups tested so far (siblings of children with autism; children with other risk factors) showed no increase [abstracts]. That a null result cannot show an effect is exactly zero is a property of the method, not a finding of any study.

Would change if. never fully closed by population data; narrowed by tests of named subgroups (see va-mito-fever-regression)

Anchor. Tried 2 Oct 2026: thelancet.com (HTTP 403), sciencedirect.com (Cloudflare block), Unpaywall (not open access), Elsevier's article API (returns only metadata, confirming the issue date 6 to 12 Feb 2010), Europe PMC and PubMed Central (record only, no text; not in PMC), the Internet Archive (blocked by this environment's egress policy), BMJ news report c696 (bmj.com 403).

Would change if. The notice's wording is only reported second-hand (the Guardian of 2 Feb 2010, a 2011 review). If it names different statements, or none, the retraction-reasons claim and its confidence change.

Next step. A library or the publisher: the Lancet's notice is one page (DOI 10.1016/S0140-6736(10)60175-4, issue of 6 to 12 February 2010). Anyone with institutional access to thelancet.com or ScienceDirect, or a British Library copy, can read it and send the exact text.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Tried 2 Oct 2026: thelancet.com (403), doi.org redirect (publisher block), Europe PMC and PubMed (record and author list only, no text), Unpaywall (not open access). Their content is known only through a 2011 review in the Indian Journal of Psychiatry, read in full.

Would change if. If the 2004 notices say something different from the 2011 review's description, the 2004 part of va-retraction-reasons changes, and its confidence can move.

Next step. A library or the publisher: the 2004 retraction (DOI 10.1016/S0140-6736(04)15715-2) and the editors' accompanying statement (Lancet 2004;363:820) can be read with institutional access.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Tried 2 Oct 2026: bmj.com (403 on c5347, c7452, c696), Europe PMC (records only, no pmcid), PubMed Central (not found), Unpaywall (not open access). The figures (3 of 9 children without a regressive-autism diagnosis; 5 of 12 with earlier developmental concerns) were read through a fetch tool in an earlier pass and are not re-checked.

Would change if. Reading the articles in full, and any independent check of the records, could move va- wakefield-records-mismatch toward established or toward unsupported.

Next step. A library or the publisher (bmj.com; the BMJ may supply the articles on request), and for the underlying records the BMJ or the author: the hospital records are confidential, so the records themselves can be reached only through the Royal Free, the GMC hearing exhibits or the BMJ.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Tried 2 Oct 2026: jamanetwork.com (HTTP 403, both the article and the PDF link Unpaywall lists), Europe PMC (record titled 'Incorrect Variable Description', no text), PubMed Central (not found).

Would change if. If the erratum changes the study's figures, the US corroboration line changes; the headline claim rests on the two Danish cohorts and does not.

Next step. The publisher or a library: the erratum is one short item (DOI 10.1001/jama.2015.17754). The corresponding author or JAMA can supply it.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Tried 2 Oct 2026: acpjournals.org (403 for the correction and the article), Europe PMC (the correction's record, no text; the article's abstract gives only 'any neurodevelopmental disorder' HR 0.93, 0.90 to 0.97), Unpaywall (not open access), the press release host (403).

Would change if. A reading that confirms the autism figure could raise the claim; one that shows a different autism result, or that the correction changes it, would change the claim.

Next step. The publisher (Annals of Internal Medicine, acpjournals.org) or a library, and the authors for the supplement: the correction, the full text and the supplementary tables will show whether the autism-only hazard ratio (0.93, 0.89 to 0.97, so far known only from a press release) is right.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Tried 2 Oct 2026: govinfo.gov (no opinions for the Poling docket, 02-1466V), uscfc.uscourts.gov (the autism decisions page lists the six test cases, not this case), Google Scholar-style searches. The case is read only through a 2015 special master's order that quotes the Rule 4(c) report and the 2011 fees decision, and through CNN and NPR (2008).

Would change if. The documents' own wording could change the statement of the concession, its confidence, and whether it rested on the Table's presumption.

Next step. The Court of Federal Claims clerk (Office of Special Masters): the decisions are court records and may be requested from the clerk; the Rule 4(c) report may be sealed or redacted. A records request to the Department of Justice or HHS (the Vaccine Injury Compensation Program) is the other route.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Tried 2 Oct 2026: a web search returned only a search lead for Banks (a 2007 ruling described in secondary summaries as compensating a developmental delay after acute disseminated encephalomyelitis following MMR); no decision was opened.

Would change if. Finding several compensated cases with autism-spectrum features would change how the legal item is described (more than one concession), though not the population evidence.

Next step. Search the Court of Federal Claims decisions database (uscfc.uscourts.gov, the Office of Special Masters decision search, and govinfo.gov) for compensated cases that describe autism or autism-like conditions, and read each decision.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Tried 2 Oct 2026: mbp.state.md.us (the 22 Mar 2012 order and a 2013 order were found and read by OCR; the 22 Aug 2012 order was not located among the file names D2425003.1 to .245 tried), a web search, and the court opinions (which quote it).

Would change if. The order's own wording could refine the grounds stated; it is unlikely to change the date or the outcome.

Next step. The Maryland State Board of Physicians (public records request): the order is a public document. The Board's site hosts orders as scanned PDFs; the August 2012 order was not found under the file names tried.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Tried 2 Oct 2026: PMC2536523 and Europe PMC (abstract only). The abstract describes a 19-month-old girl; a 2015 court order puts the vaccination in the compensation case on 19 Jul 2000, and CNN gives that child's birth as December 1998 (age about 19 months). Weissman et al. 2008 (PLoS ONE 3:e3815, read in full, open access) says that in "one of our 25 patients" the deterioration "appeared to follow vaccination" and cites for it both the 2006 case report and a 2008 NEJM perspective on the compensation case; that points to one child but is not a statement that the report's patient is the child in the compensation case. Whether the 2006 case report concerns the same child is not shown by any source read.

Would change if. If the report concerns the same child, the mitochondrial scale counts one fewer independent case and the 'Yes' position loses a separate item.

Next step. The publisher (SAGE, J Child Neurol) or a library for the full text, which may say whether the patient is the child in the compensation case; the authors can confirm.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Tried 2 Oct 2026: all six decisions were downloaded and their captions, theories, tables of contents, conclusions and orders were read; the evidentiary bodies (about 1,300 pages in all) were not read.

Would change if. The full reasoning could add detail or reveal a qualification to the outcomes; it is unlikely to change them.

Next step. Anyone with time to read the six PDFs, which are public on uscfc.uscourts.gov: an independent reader (a law or epidemiology specialist) could summarise the findings on each causation theory, which would also let the omnibus claim be checked against the full reasoning.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)

Anchor. Tried 2 Oct 2026: all three publishers refused this environment (HTTP 403); the dig read third-party copies (Madsen: web.math.princeton.edu; Hviid: autismsciencefoundation.org; the 1998 paper: briandeer.com, a site run by the journalist who investigated it). The Madsen and Hviid copies' numbers match the Europe PMC abstracts; the 1998 copy's ethics sentence and 'consecutively referred' wording match what the High Court judgment and the GMC findings quote.

Would change if. A difference between a copy and the publisher's file would change what the claims read from it; none was seen.

Next step. A library or the publishers (nejm.org, acpjournals.org, thelancet.com): anyone with access can compare the copies' text, tables and figures with the publishers' files and report any difference.

Start this step → (signs you in with GitHub; trusted contributors can send it to the dig agent)