Paracetamol, Pregnancy and Autism

On 22 September 2025 the US President and HHS Secretary announced that acetaminophen (paracetamol/Tylenol) in pregnancy may cause autism, advised pregnant women to avoid it, and the FDA initiated a label change and physician letter. The FDA's own notice contained the caveat that defines this hub: an association has been described in many studies, but a causal relationship has not been established and contrary studies exist. The announcement leaned on observational associations (Boston Birth Cohort, Nurses' Health Study II) and a 2025 Navigation Guide review whose methodology drew immediate criticism — while the American College of Obstetricians and Gynecologists responded that in over two decades of research not a single reputable study has concluded that paracetamol in any trimester causes neurodevelopmental disorders, the EMA and UK MHRA held their guidance unchanged, and obstetric bodies worldwide reaffirmed paracetamol as the analgesic of choice in pregnancy.

Roughly half of pregnancies worldwide involve paracetamol — taken for fevers, infections, chronic pain and inflammation, each independently associated with neurodevelopmental outcomes, by mothers whose own genetics and health also travel to their children. That is textbook confounding by indication and familial factors, and the study designs that can address it keep dissolving the signal: the 2024 Swedish national study of 2.48 million children found the small association vanished entirely in sibling comparisons (same family, different pregnancies — genetics and environment controlled); a 2025 BMJ umbrella review of the systematic-review literature found the evidence does not support causation; and a January 2026 Lancet-published European reanalysis reached the same verdict — associations fall apart once confounding is properly handled. The pattern — crude association, disappearing under family-level control — is the signature of confounding, not causation.

The precautionary framing — when in doubt, avoid the drug — hides an asymmetry: the alternative to treating significant fever and pain in pregnancy is not safety but untreated fever and pain. Maternal hyperthermia, particularly prolonged and in the first trimester, is itself associated with neural-tube defects and adverse outcomes; severe pain drives blood pressure and stress physiology; and the pharmacological alternatives are worse — NSAIDs are contraindicated in later pregnancy (ductus arteriosus, oligohydramnios) and opioids carry their own fetal costs. This is why every major obstetric body landed on the same formulation the FDA itself ultimately echoed: paracetamol remains appropriate in pregnancy when needed — at the lowest effective dose for the shortest necessary duration — advice that was standard before the controversy and remains standard within it.

The announcement arrived inside a political commitment to find autism's cause, and the epidemiology explains the pressure: recorded prevalence has risen to roughly 1 in 31 US eight-year-olds — driven substantially by broadened criteria, screening and service-linked identification, as the adjacent hub on autism prevalence documents — while twin studies place heritability around 80%, making autism among the most genetic of common conditions. The candidate-cause history is instructive: vaccines (refuted at scale), thimerosal (refuted), and now paracetamol each offered a modifiable, blameable exposure; each dissolved under family-controlled designs. Genuine environmental contributions (parental age, prematurity, specific prenatal exposures like valproate) are real and researched — but the search for a single ubiquitous cause of a largely heritable, diagnostically expanded condition is structurally primed to produce exactly this kind of episode.

Claims like this are not free. Mothers of autistic children are handed retrospective guilt for ordinary fever treatment; pregnant women are pushed toward untreated fever, unproven alternatives, or NSAIDs at the wrong gestation; mass-tort litigation — a US multidistrict action over prenatal Tylenol had already been dismissed for unreliable expert evidence before the announcement revived the ecosystem — shapes the information environment with paid expertise on both sides; and regulatory credibility erodes when label changes precede evidentiary standards, as the contrast between the FDA's move and the EMA/MHRA's unchanged assessments made visible. The episode belongs beside the vaccine-autism history as a study in how the association-to-headline pipeline works — and why sibling designs, umbrella reviews and regulatory pluralism are the reader's protection.

For pregnant women: paracetamol remains the recommended option for fever and significant pain — used when needed, at the lowest effective dose, for the shortest necessary time, exactly as before; significant fever should be treated, not endured; and NSAIDs in later pregnancy remain the thing actually to avoid. For parents of autistic children: the best-controlled evidence says your paracetamol use did not cause your child's autism — the guilt being distributed is not supported by the data. For everyone reading headlines: watch the design, not the announcement — crude associations that vanish in sibling comparisons are confounding wearing a causal costume, and the strongest studies here are unanimous. And for clinicians: the counselling script is unchanged — treat the indication, dose conservatively, document, and inoculate patients against both the fear and the guilt.