A Presidential Tylenol Warning Cut ER Orders 10 Percent. Two Reviews Say the Autism Claim Doesn't Hold.

What is actually known at this point is this: two independent reviews, using different methods on different sets of studies, have now looked at whether acetaminophen use during pregnancy causes autism or ADHD in children, and neither found evidence that it does. The most recent, published September 23, 2026 in Developmental Medicine and Child Neurology, pooled seven meta-analyses covering more than 3 million participants. The association held up in the raw, unrestricted data. It did not hold up once the researchers restricted the analysis to the highest-quality studies.

That finding adds to a November 2025 umbrella review in The BMJ, led by Professor Shakila Thangaratinam at the University of Liverpool, which reached a similar conclusion through a different route: it traced the apparent link to shared family factors rather than to the drug itself.

A warning that already changed behavior

The question is not academic. In September 2025, President Trump told pregnant women at a White House event to avoid the drug, standing alongside Health and Human Services Secretary Robert F. Kennedy Jr., National Institutes of Health Director Jay Bhattacharya, and FDA Commissioner Marty Makary. "Taking Tylenol is not good," Trump said. "I'll say it: it's not good."

A study in The Lancet, led by Dr. Jeremy Faust, an emergency physician at Brigham and Women's Hospital in Boston, tracked what happened next in hospital electronic health records. "It happened overnight," Faust said of the drop. Dr. Caleb Alexander, an epidemiology professor at the Johns Hopkins Bloomberg School of Public Health, said the shift did not surprise him: "Words matter. And when they come from someone with as big an audience as the president of the United States, they can change prescriber and patient behavior." Faust's data show ordering had mostly returned to normal by December.

Why the link looked real in the first place

The Liverpool-led review is the more useful piece for understanding why so many earlier studies found something. It graded nine systematic reviews covering 40 primary studies using the AMSTAR 2 tool, a standard measure of review quality, and rated confidence in the findings as low in two reviews and critically low in seven. Only one of the nine had used sibling-controlled analysis, which compares children in the same family, some exposed to acetaminophen in the womb and some not, to strip out genetics and household environment as explanations.

In that one review, whole-cohort data showed a modestly elevated hazard ratio for autism, 1.05, and for ADHD, ranging from 1.07 to 2.02 depending on the study. Once the analysis controlled for siblings, both effects collapsed toward 1.0, the point of no measurable difference: 0.98 for autism, and 0.98 to 1.06 for ADHD. A hazard ratio at 1.0 means no added risk. The Liverpool authors concluded that shared family factors, including parental mental health and genetic predisposition, likely explain most of what earlier whole-cohort studies picked up.

“This disappearance of association in the sibling analyses, previously observed in the whole cohort, suggests that shared family factors, such as parental mental health, genetic predisposition, and socioenvironmental background, explain much of the observed risk.”

What the newer review adds

The September 2026 review, led by Hoda Tayebi Hillali at the University of Santiago de Compostela in Spain, did not use sibling data. It instead compared risk ratios across meta-analyses of differing quality. Unrestricted, the pooled risk ratio was 1.17 for autism and 1.29 for ADHD, small but real-looking numbers. Restricted to the higher-quality subset, both effects moved toward the null. Tayebi Hillali's team wrote that the observed associations are probably explained by methodological limitations in the underlying studies, not by a causal effect of the drug.

What is still uncertain

Neither review is a randomized trial, and none is likely to happen. Assigning pregnant women to take or avoid a pain reliever for research purposes would raise its own ethical problems, so the field is left working from observational data of uneven quality, which is exactly what both reviews say limits confidence either way. Major regulators, including the UK's Medicines and Healthcare products Regulatory Agency, the European Medicines Agency, and Australia's Therapeutic Goods Administration, responded to the original 2025 announcement by reaffirming their existing safety guidance rather than changing it. Kenvue, the company that now makes Tylenol, said in a statement that it stands with the science: "We continue to believe that there is no credible data that shows a proven link between taking acetaminophen and autism."

What changed measurably is not the biology. It is how many pregnant patients got a fever reducer in an emergency room in the fall of 2025, a shift that traces to one televised announcement more directly than it traces to any of the studies now weighing in after the fact.