Large-scale study challenges link between acetaminophen use during pregnancy and autism risk in children

A team of Taiwanese researchers have used a nationwide, population-based cohort to examine whether taking acetaminophen during pregnancy might be linked to a higher likelihood of attention-deficit/hyperactivity disorder (ADHD) or autism spectrum disorders (ASDs) in the offspring.

The researchers analyzed over 2 million singleton births—pregnancy with only one child—recorded between 2004 and 2015. Initial data suggested a link between medication use and neurodevelopmental conditions, with the risk increasing with an increase in the frequency of use or higher doses of the medication.

The team then introduced a sibling-matched analysis—a method that compares siblings to account for shared genetics and environment, helping researchers isolate the effect of specific exposures. When the data were viewed through this lens, the previously observed association disappeared.

When the team probed the sibling data further, the story became cloudier. While one analysis suggested an increased risk, another pointed in the opposite direction, leaving the researchers unable to draw a clear conclusion.

The findings are published in JAMA Pediatrics.

To worry or not to worry

Acetaminophen is a widely used over-the-counter drug that is colloquially known as paracetamol or Tylenol, depending on which part of the world one lives in. The drug is an antipyretic and an analgesic, meaning it helps relieve pain and reduce fever. It is often prescribed to manage pregnancy-related discomfort or pain.

Concerns about the possible effects of acetaminophen on children’s brain development began to grow after several cohort studies reported dose-response–like patterns. These studies suggested that prolonged use of acetaminophen during pregnancy was associated with a higher risk of ADHD, autism spectrum disorders (ASDs), and poorer behavioral outcomes in children.

To investigate this further, researchers in Sweden used a sibling-matched design, comparing siblings within the same family to account for shared familial factors. In this analysis, the earlier associations could not be confirmed. A Japanese study using a large national database initially reported positive associations even after statistical adjustments. However, when the same data were analyzed using a sibling-matched design, the results reversed.

One possible way to move the debate forward is through large, population-based studies with robust designs that can more clearly examine these links.

Sifting through millions of records

The researchers took on this challenge by designing an observational study using a national insurance database to identify mothers who received at least two acetaminophen prescriptions during pregnancy. They then analyzed data to see whether children exposed to the drugs in the womb were more likely to be diagnosed with ADHD or ASD.

Initially, they observed that maternal acetaminophen prescriptions during pregnancy were associated with a 12% higher risk of developing ADHD and a 6% higher risk of autism spectrum disorders.

The team carried out sibling-matched analysis, where they compared siblings from the same family. The earlier link disappeared, suggesting that family-related factors, rather than the medication itself, might be driving the association.

To avoid potential bias, the team also performed a bidirectional analysis, examining whether results changed depending on whether the older or younger sibling was exposed to the drug.

They found that the outcome of exposure flipped based on the birth order. When only the older sibling was exposed, the risk of being diagnosed with ADHD and ASD was higher, whereas when the exposure was limited to the younger sibling, the likelihood of getting diagnosed with the conditions was lower.

The very different results produced by the bidirectional analysis revealed that the sibling-matched approach may still be affected by hidden sources of bias.

The researchers conclude that no firm link can be established using this specific sibling-comparison method. They suggest that future studies use sibling designs with investigator-led, age-standardized assessments to help reduce biases caused by changes in diagnostic patterns over time and across different regions.

© 2026 Science X Network