A new umbrella review of research into prenatal acetaminophen use finds little evidence that the common painkiller causes autism or attention‑deficit/hyperactivity disorder in children. While earlier reviews suggested small increases in risk, these associations largely diminished when the analysis concentrated on higher‑quality evidence, with only modest signals remaining in some analyses. The study was published in Developmental Medicine & Child Neurology.
Acetaminophen, also known as paracetamol and sold under brands such as Tylenol, is widely available and used to treat pain and fever. Because it is thought to carry fewer risks for birth defects than many other pain relievers, it has become the standard first‑line treatment for pain and fever during pregnancy. Estimates suggest that more than half of pregnant women globally use the medication at some point during gestation.
Over the past decade, conflicting studies have fuelled a debate about the drug’s safety in pregnancy. A 2019 meta‑analysis pooled early observational research and found associations with higher ADHD rates in children, though regulators and clinicians questioned whether bias or confounding factors could explain the link. A 2021 brother‑sister comparison suggested family background and genetics, rather than acetaminophen itself, might account for the apparent association with ADHD. A 2024 Swedish study using sibling comparisons found no link between prenatal acetaminophen exposure and autism or ADHD.
The debate intensified in September 2025 when, in the United States, former President Donald Trump advised against acetaminophen use during pregnancy, citing a possible autism risk, and the Food and Drug Administration began updating the drug’s label to note a possible association. Regulators in the European Union, the UK and Australia soon said they could not find evidence that the drug causes autism, but the mixed messages left many expectant mothers and healthcare professionals uncertain.
Led by Hoda Tayebi‑Hillali of the University of Santiago de Compostela in Spain, the new study set out to determine whether previously reported risks hold up once methodologically weaker studies are set aside. “Previous systematic reviews had reached discordant conclusions, and many counted the same primary studies several times without acknowledging it,” Tayebi‑Hillali said. “We set out to settle this debate rigorously: we quantified the overlap between reviews, appraised their quality, and reanalyzed the underlying primary studies according to their risk of bias. Our goal was to give clinicians, pregnant women, and health authorities an answer grounded in the best available evidence, not in headlines.”
The researchers conducted an umbrella review, pooling evidence from meta‑analyses. They searched six major scientific databases for studies published up to November 2025 and included seven meta‑analyses encompassing more than three million participants. Three of the meta‑analyses looked at autism, while all seven examined ADHD.
In the initial unrestricted analysis, the researchers found small positive associations: a pooled risk ratio of 1.17 for autism and 1.29 for ADHD, indicating exposed children had about a 17 per cent and 29 per cent higher relative risk, respectively, of a diagnosis compared with unexposed children.
Next, the team assessed the quality of each meta‑analysis using a standard checklist. Most studies were rated as low quality, often because they did not report how much acetaminophen women took or did not list excluded studies. When the weakest reviews were dropped, results changed little: a risk ratio of 1.17 for autism and 1.28 for ADHD remained. Only one meta‑analysis, published in 2025, earned a high‑confidence rating and it reported smaller risk ratios—1.10 for autism (not statistically significant) and 1.16 for ADHD.
To further test the findings, the researchers analysed the primary studies themselves, assessing risk of bias with a standard tool. When limited to studies with low risk of bias, autism risk fell to 1.05 (based on two large studies from Sweden and Japan), and ADHD risk to 1.17 (based on three studies). Both estimates hovered slightly above 1, though the ADHD results showed considerable variation between studies.
“One of the most striking findings was how much the estimates changed once we took the quality of the evidence into account,” Tayebi‑Hillali explained. “Associations that initially appeared consistent were substantially attenuated when we excluded lower‑quality reviews and primary studies at higher risk of bias. It was also particularly telling that some sibling‑comparison studies, which help control for the genetic and environmental factors shared within a family, did not find the associations originally reported. This underscores the importance of distinguishing between a statistical signal and a possible causal relationship.”
Given the backdrop of autism and ADHD prevalence worldwide (about 1 in 117 children for autism and about 1 in 60 for ADHD), any absolute increase in risk would be modest even if the remaining association were genuine. “Even taking the figures at face value, the impact would be very small,” Tayebi‑Hillali noted. “Moreover, in observational studies, a risk ratio this small can readily be explained by bias, without the drug playing any role at all. The size of the effect, therefore, does not justify alarm.”
The new findings align with recent PsyPost reporting on sibling‑matched analyses showing no link between prenatal acetaminophen exposure and autism or ADHD. They also sit in contrast to a 2025 review highlighted by PsyPost that reported stronger associations in more rigorous studies, particularly those measuring the drug in biological samples. The authors of the current umbrella review regarded biomarker measurements taken from cord blood, meconium, or maternal blood as being of uncertain validity.
The analysis also contrasts with earlier studies from 2022 and 2016, which linked prenatal acetaminophen to attention and behaviour problems. Those studies relied on parent‑reported symptoms rather than formal diagnoses, and none met the review’s strict risk‑of‑bias criteria.
As with all research, there are caveats. “The main misinterpretation to avoid is the claim that our study proves paracetamol causes autism or ADHD; equally, it would be wrong to conclude that we have shown any exposure, regardless of dose or duration, to be entirely free of risk,” Tayebi‑Hillali cautioned. “Our work is a synthesis of observational studies and therefore cannot establish causality on its own.”
A major consideration is “confounding by indication.” Pregnant women typically take acetaminophen to treat fevers, infections, pain or inflammation, and these underlying health issues can themselves influence fetal brain development, complicating attempts to separate drug effects from illness effects. Many studies also relied on self‑reported questionnaires, which can mislabel exposure timing or amount.
The most rigorous estimates rest on only a small number of studies: a single high‑quality meta‑analysis and two to three primary studies. The review did not formally assess publication bias, and could not evaluate dose, duration or timing of exposure; some earlier work suggested higher risks with use lasting more than 28 days, but timing information was largely absent in most studies.
Looking ahead, the authors called for improved observational study designs that better capture dose, frequency, duration, and timing of paracetamol use, alongside indications driving treatment. They emphasised designs that control for familial, genetic, and environmental factors—notably, sibling‑comparison approaches—taired with rigorous risk‑of‑bias assessments.
Animal studies have suggested acetaminophen might interfere with hormones involved in development, but the authors caution that such experimental evidence does not by itself prove disruption of human fetal brain development. Given the known harms of untreated fever and pain in pregnancy, they conclude that acetaminophen remains an appropriate first‑line treatment when medically necessary, used at the lowest effective dose and for the shortest possible time.
“Headlines that present an association as if it were proven causation can convey an alarm that the evidence does not warrant. At the same time, it is essential to acknowledge uncertainty and to keep investigating,” Tayebi‑Hillali said. “We should also remember that untreated fever and pain during pregnancy can have consequences, so avoiding a clinically necessary medication without consulting a healthcare professional is not a risk‑free decision either.”
The study is titled “Prenatal exposure to paracetamol and the risk of autism and attention‑deficit/hyperactivity disorder: An umbrella review and meta‑meta‑analysis,” and was authored by Hoda Tayebi‑Hillali, María Blanco‑Suárez, Berta Rivas‑Mundiña, Jacobo Limeres‑Posse, Márcio Diniz‑Freitas, and Pedro Diz‑Dios.
