367 - Tylenol, pregnancy, and autism: What recent studies show and how to interpret the data

Recent studies suggest a very low probability of a causal link between prenatal Tylenol use and autism, as observed associations are weak and often disappear after accounting for confounding factors like genetics and family environment.

• Prenatal Tylenol use has a very low probability of causing autism, despite small statistical associations.

• Large-scale sibling-controlled studies effectively abolish any observed link between Tylenol and autism risk.

• For maternal fever during pregnancy, using Tylenol is crucial due to the significant risks fever poses to the fetus.

Recent headlines have linked acetaminophen (Tylenol) use during pregnancy to autism, generating confusion and concern. This discussion emphasizes the importance of a critical, scientific framework for interpreting complex epidemiological data. While a small statistical association between prenatal acetaminophen exposure and autism may exist, strong evidence from large studies, particularly those controlling for genetic and environmental factors through sibling analyses, suggests this association is likely not causal. The dramatic increase in autism diagnoses is primarily attributed to expanded diagnostic criteria, increased awareness, and advancing parental age, rather than a single environmental factor like Tylenol.

Autism Rates and Causality Challenges

• 00:01:36 Autism rates have risen dramatically, increasing fivefold from 6.7 to 32.2 cases per thousand children between 2000 and 2019, but complex conditions rarely have a single cause. The intense desire to find autism triggers leads to the 'multiple comparisons problem,' where statistically significant but spurious correlations are bound to be found by chance, similar to unrelated events like margarine consumption and divorce rates in Maine appearing correlated.

FDA Pregnancy Drug Categories

• 00:05:34 The FDA classifies pregnancy drug risks using categories A, B, C, D, and X, providing guidance on potential fetal harm based on animal and human studies. Category A (e.g., prenatal vitamins) shows no demonstrated risk in human studies (2-5% of drugs), Category B (e.g., Tylenol, metformin) shows no human risk but possible animal signals (15-25%), and Category C (most drugs, 60-75%) means risk cannot be ruled out. Categories D and X indicate known fetal risk, used only if maternal benefit outweighs harm (D) or completely contraindicated (X, e.g., statins).

Critique of Key Review Article

• 00:22:45 Recent alarm about acetaminophen and autism was triggered by a 2021 systematic review that claimed a consistent positive association, but this claim is misleading. Two of the six included studies showed no significant association, and dose-response relationships were often confounded. The smallest study, the Xi study, showed a strong association but suffered from a small sample size, biased participant inclusion (11% autism rate vs. 3% general population), and unreliable cord blood biomarker measurements of long-term exposure, rendering its findings questionable.

Swedish and Japanese Cohort Studies

• 00:36:30 The large Swedish cohort study of 2.5 million children initially found a small (5%) relative risk increase between prenatal acetaminophen use and autism in unadjusted analyses. However, when controlling for family environment and genetics using a sibling-controlled analysis, this association was entirely abolished, suggesting confounding factors. A subsequent Japanese study of 220,000 children, with higher acetaminophen exposure rates, also found that any trend towards increased risk was abolished in sibling analyses, further weakening the case for causality.

Bradford Hill Criteria Evaluation

• 00:47:39 Applying the Bradford Hill criteria to the link between acetaminophen and autism reveals a very weak effect size (1.05x, compared to 10x for smoking and lung cancer). While consistency across studies is moderate, specificity is weak as many other factors are linked to autism. Temporality is strong (exposure precedes outcome), but dose dependency and biological plausibility are modest or weak. Analogy, referencing aspirin's protective effect as another prostaglandin inhibitor, argues against a causal link, while experimental evidence and coherence remain largely unsupportive or inconsistent.

Balancing Risks and Recommendations

• 01:19:02 The probability of a causal link between prenatal acetaminophen use and autism is very low. The increase in autism rates is best explained by expanded diagnostic criteria, increased awareness, and advancing parental age. When considering medication during pregnancy, the risks of the medication must be balanced against the risks of the untreated maternal condition. For minor aches, avoiding acetaminophen might be best, but for fever, acetaminophen use is strongly favored given the known risks of fever itself (e.g., 25-200% higher risk for birth defects, 25-200% higher autism risk) to the developing fetus.