Does prenatal or early-childhood acetaminophen exposure cause autism, and what do the best-controlled studies (sibling controls, negative controls) show?

The short version: the best-controlled studies—sibling comparisons in large populations—find that the association between prenatal acetaminophen and autism essentially disappears once you account for the things that make mothers different from each other.

The evidence on this question has a clear shape. The early cohort studies, which compared one child to another across different families, reported modest associations. A six-cohort European meta-analysis of 73,881 children found about a 19% increase in autism-spectrum symptoms with prenatal exposure [3]. A smaller biomarker study using cord blood found a stronger signal: children in the highest exposure group had about 3.6 times the odds of an autism diagnosis [1, 2]. These are the numbers that get cited.

But these studies share a fundamental weakness. Mothers who take acetaminophen during pregnancy are different from mothers who do not. They have fevers, infections, pain, and inflammation—conditions that themselves affect fetal brain development. The conventional cohort design cannot fully separate the drug from the reason it was taken. That is where the sibling-control studies matter.

The largest and most rigorous sibling analysis comes from Sweden, covering nearly 2.5 million births. In the conventional analysis, prenatal acetaminophen use was associated with a 5% increase in autism risk. But when the same analysis compared siblings within the same family—children who shared the same mother, the same household, the same genetic background, but differed in whether acetaminophen was used during that pregnancy—the association vanished. The hazard ratio dropped to 0.98, with a confidence interval that crossed 1.00 [2]. A Norwegian sibling study of nearly 10,000 pairs found the same pattern: associations in conventional models, null results in within-family comparisons [1, 2].

This is what the best-controlled evidence shows. The apparent link between prenatal acetaminophen and autism is almost certainly driven by confounding by indication—the underlying reason the mother took the drug, not the drug itself. The sibling design is not perfect; it cannot control for factors that change between pregnancies. But it is the strongest observational tool we have, and it consistently fails to support a causal role for acetaminophen.

The mechanistic evidence is mixed. Acetaminophen generates a toxic metabolite, NAPQI, and can cause oxidative stress and mitochondrial dysfunction [1]. Animal studies show altered brain development markers at high doses [1]. But a 2025 study using human cortical organoids—three-dimensional brain tissue grown from stem cells—found that acetaminophen at physiologically relevant concentrations produced only modest transcriptional changes and no effect on organoid growth, neuronal differentiation, or electrical activity [4]. The human tissue model did not replicate the damage seen in animal studies.

My call: the best-controlled human studies do not support a causal link between prenatal acetaminophen exposure and autism. The sibling analyses consistently null out the association. The question is not settled—no long-term randomized trial exists, and the mechanistic picture is incomplete—but the evidence as it stands points to confounding, not causation. Confidence: moderate.