For years, studies found that taking Tylenol during pregnancy raised the risk of autism and ADHD. A study of 708,000 children just showed those studies were measuring the wrong thing entirely
Acetaminophen, sold under the brand name Tylenol and known internationally as paracetamol, has been the default recommendation for pain and fever during pregnancy for decades. It crosses the placental barrier, which means the developing fetus is exposed to it, and that biological fact has made it a target of ongoing scientific scrutiny.
Over the past decade, that scrutiny produced a string of alarming findings. Study after study reported that children born to mothers who used acetaminophen during pregnancy were more likely to be diagnosed with autism spectrum disorder or attention-deficit hyperactivity disorder. The associations were modest but consistent enough to attract serious attention. The FDA began reviewing the evidence. Researchers called for precautionary warnings. In September 2025, President Trump publicly linked acetaminophen use in pregnancy to autism, a claim that intensified anxiety among expecting parents already navigating one of medicine’s genuinely contested questions.
A new study published yesterday in JAMA Internal Medicine, one of the most selective clinical journals in medicine, has examined the same question with a fundamentally different method. The findings suggest that the previous associations, real as they appeared in the data, were not measuring what researchers thought they were measuring.
The study analyzed 708,020 mother-child pairs from electronic health records in Hong Kong spanning two decades of data. Its central design feature was a sibling comparison: instead of comparing children of mothers who used acetaminophen to children of mothers who did not, the researchers compared brothers and sisters within the same families, specifically children whose mother took the medication during one pregnancy but not another.
When they did that, the association between acetaminophen and autism disappeared entirely. So did the association with ADHD. The risk ratio for autism was 1.00. For ADHD it was 1.01. In statistical terms, these numbers mean there was no meaningful difference between the exposed and unexposed siblings.
Why the earlier studies got it wrong
The finding does not mean the earlier research was conducted carelessly. It means it was vulnerable to a specific type of error that is genuinely difficult to avoid in observational studies of pregnancy outcomes.
The error is called confounding. Autism and ADHD are both strongly heritable conditions, with heritability estimates around 80% and 74% respectively. Mothers who take acetaminophen during pregnancy tend to do so because they are managing pain, fever, or illness. The underlying conditions that prompt medication use, including inflammatory conditions, infections, and chronic pain disorders, have genetic components and may themselves be correlated with neurodevelopmental risk in offspring.
When you compare all children of acetaminophen users to all children of non-users, you are not just comparing drug exposure. You are comparing two groups of families that differ in health profile, genetics, and dozens of other unmeasured ways. The drug appears associated with autism not necessarily because it causes autism, but because the kind of mother who takes it during pregnancy is, on average, slightly more likely to have a child with autism for entirely unrelated genetic reasons.
The sibling comparison eliminates this problem by design. When you compare a child exposed in utero to a sibling who was not, you are holding constant the mother’s genetics, her health history, her socioeconomic status, and her family environment. The only meaningful difference between the two children is whether they were exposed to the drug.
“Many earlier observational studies could not fully separate the effect of acetaminophen from the reasons it was taken, such as fever, infection, or pain, or from genetic and family factors that may also influence autism and ADHD,” said corresponding authors Shan Luo and Eric Yuk Fai Wan of the University of Hong Kong. “We therefore examined whether the association remained when children from the same family were compared.”
It did not remain.
How the researchers confirmed what they were seeing
The team used several analytical approaches to test whether their null result was genuine rather than a product of their own method.
First, they ran a conventional population-level analysis on the full dataset, without the sibling matching. This replicated the positive associations found in older studies, a small but detectable elevated risk of autism and ADHD among exposed children. That confirmed the earlier literature was not wrong on its own terms. It was finding a real pattern in the population-level data.
Then they conducted a negative control analysis. They identified mothers who were prescribed acetaminophen in the year before becoming pregnant and in the year after giving birth, periods when the drug could not biologically affect fetal brain development. If acetaminophen itself were the causal agent, exposure outside of pregnancy should show no association with the child’s neurodevelopmental outcomes.
It showed the same small elevated association as the during-pregnancy analysis. This is precisely what you would expect if the association were driven by characteristics of the mothers who use the medication rather than by the drug itself. Mothers who take acetaminophen before pregnancy are the same mothers who take it during pregnancy. Their children show higher rates of autism and ADHD not because of the drug but because of who those mothers are genetically and medically.
“This suggests that the apparent association is more likely to reflect underlying characteristics of the women or families who use the medicine, rather than a direct effect of acetaminophen itself,” the researchers concluded.
What the study means for the public debate
The findings do not arrive in a vacuum. The acetaminophen-autism question became politically charged in late 2025 when the Trump administration made public statements linking the drug to autism, prompting widespread concern and, according to the researchers, a significant increase in anxiety among pregnant patients.
The study was designed in part to address that confusion directly.
“The public importance of this question became especially clear in September 2025, when President Trump publicly linked acetaminophen use in pregnancy with autism,” Luo and Wan noted. “The claim intensified anxiety around a medicine commonly used for pain and fever, while the scientific evidence remained difficult to interpret.”
One of the authors described a personal experience that shaped their motivation. During her own pregnancy, she developed shingles and hesitated to take acetaminophen because of alarming reports she had read. “That experience helped me understand how difficult medication decisions can feel when the available information is uncertain or frightening,” she said. “It strengthened my commitment to research on medication safety during pregnancy.”
The practical message from the study is direct. After accounting for shared family factors through a sibling comparison across 708,000 children, there is no evidence that acetaminophen use during pregnancy increases the risk of autism or ADHD. The drug remains appropriate when clinically indicated, and should be used at the lowest effective dose for the shortest necessary duration, in consultation with a healthcare professional.
What the study does not cover
The researchers are careful to draw the boundary of their conclusions precisely.
The study captured only prescription acetaminophen from Hong Kong’s public healthcare system. Over-the-counter purchases, which account for a significant portion of use, were not included. Prescription use typically reflects more severe symptoms requiring a doctor visit, while over-the-counter use covers milder, self-managed pain and fever. If there is a meaningful difference in risk between high-dose prescription use and low-dose over-the-counter use, the study design may not detect it.
The analysis focused specifically on autism and ADHD. The authors explicitly note that their findings should not be generalized to all possible pregnancy or childhood outcomes. Other questions about acetaminophen in pregnancy remain open.
The sibling design, while more rigorous than standard population comparisons, has its own constraints. It restricts the comparison to families where exposure varied between pregnancies, which may not represent all families equally. And it cannot rule out unmeasured factors that affected one pregnancy but not the other.
What it can say, across one of the largest and most methodologically careful analyses yet conducted on this question, is that the association that alarmed millions of pregnant women and prompted public health warnings appears to have been a statistical artifact of unmeasured family characteristics rather than a causal effect of the drug.
The study “Prenatal Acetaminophen (Paracetamol) Use and the Risk of Autism and/or Attention-Deficit/Hyperactivity Disorder Among Sibling-Matched Cohorts” was authored by Shan Luo, Qiaowa Gong, Yujie Ai, Jiayue Zhang, Linda Chan, William Chi Wai Wong, Patrick Ip, Esther Wai Yin Chan, Peter Tanuseputro, Ian Chi Kei Wong, and Eric Yuk Fai Wan at the University of Hong Kong, and published July 31, 2026 in JAMA Internal Medicine.
Source: University of Hong Kong. DOI: 10.1001/jamainternmed.2026.2215
