The claim that prenatal acetaminophen use causes autism has become one of the most politically amplified questions in pharmacoepidemiology, despite a thin evidentiary base that has only gotten thinner under scrutiny. Acetaminophen (paracetamol) is the most widely used analgesic in pregnancy globally. When early observational studies reported small associations with neurodevelopmental outcomes, the finding moved quickly from journals into headlines. In September 2025, the FDA initiated a label change warning process and the Trump administration publicly discouraged use during pregnancy, actions that ran ahead of the available evidence rather than followed from it.
On 13 April 2026, a research letter in JAMA Pediatrics by Prahm et al. at Copenhagen University Hospital added one of the most comprehensive analyses to date. Using Danish nationwide registers, they followed 1,506,155 children born from singleton pregnancies between 1997 and 2022. Among 31,098 children (2.1%) exposed to prescribed acetaminophen in utero, autism was diagnosed in 1.8%, compared with 3.0% in the unexposed group. After adjusting for maternal age, income, smoking, other medications, and health conditions, the adjusted hazard ratio was 1.03 (95% CI: 0.95 to 1.12). There was no dose-response pattern and no trimester-specific effect.
The study's most important component is the sibling-matched analysis, comparing children within the same family where the mother used acetaminophen during one pregnancy but not another. This design holds constant shared genetics, household exposures, and socioeconomic conditions, the factors most likely to confound population-level associations. It is among the strongest confounding controls available outside a randomised trial. The sibling-matched hazard ratio was 1.09 (95% CI: 0.91 to 1.27): no significant association.
This pattern is not new. A 2024 Swedish cohort in JAMA examined 2.5 million children and found sibling hazard ratios of 0.98 for both autism and ADHD. A November 2025 BMJ umbrella review of nine systematic reviews covering 40 studies concluded that associations consistently disappeared in sibling-controlled designs and rated the overall evidence quality as low to critically low. A signal that does not survive familial confounding is the characteristic fingerprint of a confounded association, not a causal one. Across Danish, Swedish, and meta-analytic evidence, that is what the data show.
What makes the Danish study particularly instructive is its infrastructure. Denmark's register system, built on the CPR number (a unique personal identifier assigned to every resident since 1968), enables individual-level linkage across healthcare, prescription, and administrative databases with virtually no loss to follow-up. The National Prescription Register captures every filled prescription at community pharmacies. The National Patient Register provides validated diagnostic codes. Full population coverage eliminates selection bias. A study of 1.5 million children with linked prescription and diagnostic data, including a sibling design, was feasible because the infrastructure already existed. In most countries, a cohort of this scale would take years and tens of millions in funding. In Denmark, it was a research letter.
The limitations cut in a specific direction. The National Prescription Register only captures prescribed acetaminophen; over-the-counter purchases, which represent the majority of use, are invisible. This underestimates true exposure and biases results toward the null. Mothers obtaining prescriptions may also have had more severe symptoms, introducing confounding by indication. Outcome misclassification is possible, since autism diagnosis depends on clinical recognition and help-seeking behaviour. None of these rescue the causal hypothesis: OTC misclassification makes the null result conservative, the sibling design absorbs confounding by indication, and the convergence across Danish, Swedish, and meta-analytic evidence argues against an artefact of any single analytic choice.
The Danish registers were not built to answer this question. They were built as public infrastructure, maintained for decades, and made available under governance frameworks that balance privacy with population-level benefit. When a public health question becomes urgent, the answer can come in months rather than years. That is the return on long-term investment in data infrastructure.
The practical conclusion is straightforward, even if the political discourse is not. The best available evidence does not support a meaningful association between prenatal acetaminophen use and autism. Untreated fever and pain in pregnancy carry their own risks. Decisions about medication in pregnancy should follow evidence, not warnings issued ahead of it.
Disclaimer: This post is a summary of published research and is intended for informational purposes only. It does not constitute medical advice. Decisions about medication use during pregnancy should be made in consultation with a healthcare provider.