The Fight Over Tylenol and Autism Just Got Messier

Last September, during a memorably bizarre press conference, President Trump told pregnant women—repeatedly and emphatically—not to take Tylenol. His health secretary, Robert F. Kennedy Jr., pointed to studies that “suggest a potential association” between acetaminophen use during pregnancy and neurodevelopmental disorders. Kennedy was more adamant on The Joe Rogan Experience recently, insisting that he had read 76 studies on the subject over one weekend and concluded that the “science is really clear” in showing a link between the drug and the conditions.

The science, in fact, is not at all clear. Although some studies have indeed found an association between prenatal acetaminophen use and autism or ADHD, others have not, including a recent systematic review of 43 studies. But a new study, carried out in Taiwan and published today in JAMA Pediatrics, seems poised to inflame the controversy anew.

Kennedy’s confident assertions aside, the FDA has offered a more evenhanded assessment of prior evidence. On the day of the September press conference, the agency issued a notice to doctors acknowledging that “a causal relationship has not been established” between prenatal acetaminophen use and neurological conditions. That’s exactly right. The studies that have found an association tend to be small and unable to determine causality or rule out other potential causes. Two recent large studies—one conducted in Japan, the other in Sweden—examined siblings and found no association between acetaminophen and autism. Such studies, common in epidemiology, compare siblings in the same family, making use of the similarities in their genetics and environment to help rule those out as culprits. The Swedish researchers, for example, found a weak association between acetaminophen and neurological disorders among all 2.4 million children in the study. But when the study was narrowed just to siblings, comparing, for instance, one who was exposed prenatally and one who was not, the association vanished, suggesting that a factor other than acetaminophen use was at play. The Japanese results followed a similar pattern.

At first glance, the new Taiwanese study appears to mirror the conclusions of the Swedish and Japanese studies. The researchers analyzed the health records of more than 2 million children born from 2004 to 2015, along with their mothers. Across the entire study, mothers who were prescribed acetaminophen during pregnancy were more likely to give birth to a child who would later be diagnosed with autism or ADHD. (In Taiwan, most acetaminophen is taken by prescription, the study’s authors suggest.) And, as in the Swedish and Japanese studies, the apparent association between acetaminophen and the neurological disorders disappeared when the analysis was narrowed to look only at siblings.

But the researchers conducting the Taiwan study also found something weird. When only an older sibling was exposed to acetaminophen during pregnancy, that child was more likely than the younger, unexposed sibling to have ADHD or autism. When only a younger sibling was exposed, the risk for that child decreased. In other words, the siblings didn’t appear to be acting as reliable controls for each other in the experiment, as would be expected.

It’s a head-scratching result, and even the authors don’t seem to have an explanation for it. (Zeyan Liew and Pei-Chen Lee, who are listed as the study’s corresponding authors, did not respond to my requests for comment.) In the paper, they argue that their findings cast doubt on the validity of sibling studies more generally, including their own, arguing that the study design might accidentally introduce unexplained biases and that the families included might not be a representative sample of the general population. Their skepticism of sibling studies runs counter to the long-established notion that the more narrow approach is more rigorous, but it does seem to support Kennedy’s dismissal of the many studies that indicate acetaminophen’s safety. Late last month, Kennedy told Rogan that such studies “have huge holes in them.” (He did not specify what those holes might be; Andrew Nixon, a spokesperson for the Department of Health and Human Services, did not clarify when I asked him.)

[Read: The Trump administration’s unintended autism experiment]

To try to make sense of the findings, I spoke with Viktor Ahlqvist, an author of the Swedish sibling study. Ahlqvist thought the new study’s results were broadly consistent with his own and was perplexed by the researchers’ emphasis on what seemed to him like a “local quirk,” perhaps related to some unknown variable that’s specific to Taiwan. (For instance, a mother in Taiwan who’s more likely to use acetaminophen in a first pregnancy might be different from other mothers in other ways as well.) Besides, Ahlqvist said, the oddity in the Taiwanese data doesn’t invalidate the overall results of the Swedish or the Japanese studies—or of the Taiwanese study itself, for that matter. Yusuke Okubo, a co-author of the Japanese sibling study, was more cautious. He told me that although the new study mostly confirmed his research, the discrepancy between the outcomes for older and younger siblings “suggests that unaddressed biases may also remain.” Ruling out any potential harm based solely on siblings studies like his own is therefore premature, he said.

And so the science remains just as inconclusive as it was back in September. Since then, several of the administration’s acetaminophen-related promises seem to have fallen by the wayside. Kennedy said at the time that the FDA would start the process of updating acetaminophen’s label to indicate the possible risks during pregnancy; in an email, a spokesperson for Kenvue, which makes Tylenol, said the FDA has yet to contact the company about that proposed change. (“We believe that there is no credible data that shows a proven link between taking acetaminophen and autism,” the spokesperson wrote.) The secretary also teased “a nationwide public-service campaign to inform families” about the alleged dangers of prenatal acetaminophen use; if such a campaign has begun, it has gone under the radar.

Back in the fall, the FDA did advise doctors to consider minimizing the use of acetaminophen for low-grade fevers in pregnancy, though the notice also correctly stated that the drug is safer for pregnant women than aspirin or ibuprofen. A recent analysis in The Lancet found that, in the weeks following the White House press conference, prescriptions for acetaminophen given to pregnant women who visited emergency rooms dropped as much as 20 percent below what would otherwise have been expected; after 11 weeks, prescription rates were approaching typical levels again. (Nixon didn’t respond to questions about the label change or the nationwide campaign, but did point me to an X post about the Lancet data, in which he wrote, “Delivering a message about a specific neurological risk for babies is another example of our commitment to telling the truth about public health.”)

[Read: Trump tells pregnant women to ‘fight like hell’ not to take Tylenol]

Last spring, Kennedy promised that the health department would figure out, in just a few months, not only what has caused autism rates to rise in the United States, but also how to prevent kids from being exposed to whatever it is. This proposal was unlikely, if not absurd. Autism researchers have been trying for decades to understand the complex disorder, and the consensus among them is that the increase can largely be attributed to broader diagnostic criteria and better surveillance. But pinning the blame on acetaminophen, even if the case has always been weak, allowed the Trump administration to appear to meet its self-imposed deadline. That association, at least, is clear.