Did Infant Thimerosal Exposure Cause Autism? The 2004 UK Cohort Study
- Vaccine Affect
- Nov 20
- 5 min read
Andrews N, Miller E, Grant A, Stowe J, Osborne V, Taylor B.
Thimerosal Exposure in Infants and Developmental Disorders: A Retrospective Cohort Study in the United Kingdom Does Not Support a Causal Association.
Pediatrics 2004;114(3):584–591.
Big question:Does getting more thimerosal-containing DTP/DT vaccines as a young infant increase the risk of autism or other neurodevelopmental disorders?
Data & population
Source: UK General Practice Research Database (GPRD).
Cohort: 109,863 children born 1988–1997 and registered with GP practices contributing to the database. PubMed+1
Follow-up: at least 2 years of continuous registration from birth in the database (to allow time for diagnoses to appear in the records). researchportal.ukhsa.gov.uk
Exposure
Thimerosal came from DTP and DT vaccines routinely given in infancy.
Each DTP/DT dose was recorded; each dose contained 50 µg thimerosal (25 µg ethylmercury). researchportal.ukhsa.gov.uk
Exposure categories:
Number of DTP/DT doses by 3 months, 4 months, and
Cumulative exposure by 6 months.
Outcomes
They looked at GP-recorded diagnoses of: PubMed+1
Autism
“General developmental disorders”
Language/speech delay
Unspecified developmental delay
ADHD/“attention-deficit disorder”
Tics
Behaviour problems
Enuresis/encopresis
Analysis
Hazard ratios per extra DTP/DT dose (or per higher cumulative exposure).
Adjusted for sex, practice, calendar period, and some other factors (limited set).
Main findings
From the abstract/summary:
Autism & most disorders: No evidence that higher thimerosal exposure increased risk.
Tics: One analysis showed increased risk with higher exposure at 4 months– HR 1.50 per dose (95% CI 1.02–2.20).
“Protective” results: Higher exposure was associated with lower risk for:
General developmental disorders (HR 0.87)
Unspecified developmental delay (HR 0.80)
ADHD (HR 0.79).
Authors’ conclusion
They conclude that, apart from a possible signal for tics, they found no evidence that infant thimerosal exposure from DTP/DT vaccines causes neurodevelopmental disorders.
2. Key quotes
Study design
“A retrospective cohort study was performed using 109 863 children born 1988–1997 and registered in United Kingdom general practices.”
Tics result
“For tics… hazard ratio 1.50 per dose at 4 months; 95% confidence interval 1.02–2.20.”
Conclusion
“With the possible exception of tics, there was no evidence that thimerosal exposure via DTP/DT vaccines causes neurodevelopmental disorders.”
3. Critics review of the methods
3.1 Data source & case ascertainment
GPRD limitations: Diagnoses come from routine GP coding. Critics argue that:
Autism and other neurodevelopmental disorders are often diagnosed in specialist clinics, not always fully coded in GP records.
Diagnostic practices changed over the 1988–1997 birth cohorts, and this may not be fully controlled.
The minimum 2-year follow-up means many children were still quite young; late-diagnosed autism or ADHD could be missed or misclassified as “no disorder.”
3.2 Exposure definition
The exposure is only thimerosal in DTP/DT, not:
Other thimerosal-containing vaccines (e.g., some hepatitis B products),
Other environmental or medical mercury sources.
Exposure windows are by 3, 4, and 6 months. Critics might say:
This assumes that doses in that narrow window capture the biologically relevant exposure window for all outcomes.
It doesn’t account for total cumulative thimerosal across all vaccines in early childhood.
3.3 Outcomes & multiple comparisons
Nine different categories of developmental/behavioural outcomes, multiple age windows, and multiple exposure metrics generate many statistical tests.
Critics point out:
You can expect some significant findings purely by chance.
The apparent “protective” associations (lower risk with higher exposure) raise the possibility of:
Healthy-vaccinee bias (children who are healthier or from more engaged families may be more likely to follow the schedule on time),
Residual confounding (socioeconomic status, parental concern, health-seeking behaviour).
The tics association may also be a chance finding or a marker of differential ascertainment rather than a true causal signal.
3.4 Confounding
Adjustments are relatively limited (sex, GP practice, some temporal factors).
Critics question whether they fully captured:
Socioeconomic status and parental education,
Perinatal factors (preterm birth, low birthweight) beyond crude categories,
Health service utilisation (families who see doctors more often may both vaccinate earlier and receive more diagnoses).
3.5 Power and autism specifically
Autism cases in this cohort are relatively few (on the order of ~100 cases in later summaries). ResearchGate
Critics argue that:
The study might have limited power to detect modest increases in autism risk from thimerosal.
Confidence intervals around autism estimates may still allow for small increases or decreases in risk.
3.6 Conflicts of interest and context
Authors are based in the UK Health Protection Agency / immunisation program. researchportal.ukhsa.gov.uk
Pro-safety reviews (e.g., AAP, CHOP Vaccine Education Center) later cite this as evidence that thimerosal is not linked to neurodevelopmental harm. Children's Hospital of Philadelphia
Critics sympathetic to the vaccine–autism hypothesis argue this creates institutional incentive to find no association, and they group this paper with other large, government- or industry-linked epidemiological studies that reach similar conclusions.
Summary from a critic’s angle:A critic would frame Andrews et al. as a large but administrative-database study with limited autism case capture, possible misclassification, a narrow exposure definition, and patterns (like “protective” effects) that clearly point to uncontrolled confounding and selection bias. For that critic, the study doesn’t “prove safety,” it just fails to detect a signal under those constraints.
4. External critiques and alternative takes
Here’s what’s been said about this study from outside the paper.
4.1 Anti-thimerosal critiques
Hooker (2014) – “Methodological Issues and Evidence of Malfeasance in Research Purporting to Show Thimerosal in Vaccines is Safe”
Groups Andrews et al. with several other large epidemiologic studies.
Argues that these studies, including Andrews’, suffer from:
Misclassification of exposure and outcomes,
Inadequate control for confounders,
Systematic design choices that tend to produce null results. PMC
Yazbak and others (as summarized in an annotated bibliography)
Criticise thimerosal safety studies presented to the 2004 IOM committee (including the UK GPRD work) for:
“Vagueness” of epidemiological data,
The fact that essentially all large studies end up rejecting the thimerosal–autism hypothesis. VTechWorks
4.2 Alternative/contrasting findings
Geier & Geier (2004) – VAERS analysis on “Neurodevelopmental disorders following thimerosal-containing childhood vaccines” reports increased reporting of NDs associated with thimerosal-containing vaccines. PubMed
This is sometimes presented by critics as counter-evidence to studies like Andrews et al., though VAERS-based analyses are themselves heavily criticised for reporting bias and lack of denominators.
4.3 Mainstream summarising reviews
Parker et al. 2004/2005 (Pediatrics review) and later narrative reviews on thimerosal and autism cite Andrews et al. as one of several consistent null epidemiologic studies. PubMed
The AAP fact-check and the CHOP Vaccine Education Center list Andrews et al. as key evidence that thimerosal exposure is not associated with autism or other neurodevelopmental harm. Children's Hospital of Philadelphia
So the same study is used very differently:
mainstream reviews: “another large, well-done null study,”
critics: “another biased administrative-data study that was almost guaranteed to be null.”
