Paracetamol use in pregnancy not linked to autism or ADHD, major review finds

A large and methodologically rigorous review, published in The Lancet, has reported no evidence that paracetamol use during pregnancy increases the likelihood of autism spectrum disorder, attention-deficit hyperactivity disorder, or intellectual disability in children, reinforcing existing clinical guidance on its safety

A comprehensive systematic review and meta-analysis published in The Lancet has reported no evidence that paracetamol use during pregnancy increases the likelihood of autism spectrum disorder (ASD), attention-deficit hyperactivity disorder (ADHD), or intellectual disability in children. The findings address public and scientific concern surrounding the neurodevelopmental safety of the drug and support current recommendations that paracetamol remains an appropriate first-line treatment for pain and fever during pregnancy, when used as directed.

Paracetamol – also known as acetaminophen and widely as the brand name Tylenol – is the most widely used analgesic and antipyretic in pregnancy worldwide. It has long been preferred to non-steroidal anti-inflammatory drugs and opioids because of its comparatively favourable maternal and foetal safety profile. Nevertheless, a series of observational studies published over the past decade has reported associations between prenatal paracetamol exposure and neurodevelopmental outcomes, particularly ASD and ADHD. These reports have contributed to public anxiety and, in some countries, notably the USA, have fuelled highly politicised debate about medication use in pregnancy.

The authors of the present review aimed to determine whether these reported associations persisted when analyses were restricted to higher-quality evidence capable of accounting more effectively for familial, genetic and environmental confounding.

To achieve this, they undertook a systematic search of MEDLINE, Embase, the Cochrane Library, and ClinicalTrials.gov from the database inception of each to 30 September 2025. Eligible studies were cohort designs that compared pregnancies with and without paracetamol exposure, defined neurodevelopmental outcomes using validated questionnaires or medical records and reported adjusted effect estimates that accounted for maternal comorbidities and treatments. Studies reporting only unadjusted estimates were excluded, as were case reports and small case series.

Study quality was assessed using the Quality In Prognosis Studies tool, which evaluates risk of bias across key domains including study participation, exposure measurement, confounding, outcome assessment and statistical analysis.

In total, 43 studies met the criteria for inclusion in the systematic review, with 17 providing data suitable for quantitative synthesis. Of these, 11 were judged to be at low risk of bias, 23 at moderate risk, and nine at high risk.

The primary analytical focus was placed on sibling-comparison studies. This design compares outcomes between siblings who differ in prenatal exposure, thereby controlling for shared genetic background, socioeconomic circumstances and many aspects of the home environment. Such designs are regarded as particularly valuable in neurodevelopmental research, where familial factors exert a strong influence on both exposure patterns and outcomes. Secondary analyses examined studies assessed as having low risk of bias, all studies reporting adjusted estimates and subsets with follow-up periods longer than five years.

Across all primary analyses, the findings were consistently null. In sibling-comparison studies, prenatal paracetamol exposure was not associated with ASD, with a pooled odds ratio of 0.98 and confidence intervals that excluded clinically meaningful increases in risk. No association was identified for ADHD or intellectual disability, with pooled estimates close to unity. Statistical heterogeneity was low for ASD and modest for the other outcomes, indicating good consistency across the most methodologically rigorous studies.

These results were reinforced by multiple sensitivity analyses. When the meta-analysis was restricted to studies at low risk of bias, no association emerged between paracetamol exposure and any of the three neurodevelopmental outcomes. Analyses that included all adjusted studies, as well as those limited to cohorts with more than five years of follow-up, similarly failed to demonstrate increased risk. Subgroup analyses by trimester of exposure, duration of use, or offspring sex could not be undertaken reliably within the most stringent evidence subsets because too few studies reported stratified data. But no subgroup showed a consistent signal of harm.

The authors contrasted these findings with earlier meta-analyses that reported small increases in the likelihood of ASD or ADHD following prenatal paracetamol exposure. Those earlier syntheses were characterised by substantial heterogeneity and relied heavily on conventional observational designs that are vulnerable to residual confounding. Maternal fever, pain, infection, genetic liability and socioeconomic factors are all associated with both paracetamol use and child neurodevelopment, making confounding by indication a persistent challenge. The present review demonstrated that when these factors are more effectively controlled, particularly through sibling-comparison designs, the apparent associations disappear.

Large population-based cohort studies provided important contextual support. A study from Sweden that included 2.48 million births reported no association between prenatal paracetamol exposure and ASD, ADHD or intellectual disability when sibling comparisons were used. A Japanese cohort of more than 217,000 children identified modest associations in conventional analyses but these were attenuated or reversed in family-based models and further weakened by bias modelling and negative-control analyses.

The current meta-analysis integrated these findings into a coherent framework, suggesting that earlier positive associations are most plausibly explained by unmeasured familial and clinical confounders rather than a causal drug effect.

The review also considered proposed biological mechanisms that have been cited to support concern. Hypotheses have included endocrine disruption, oxidative stress mediated by the toxic metabolite N-acetyl-p-benzoquinone imine, altered prostaglandin signalling and epigenetic modification.

Paracetamol is known to cross the placenta, and the foetal liver has limited detoxification capacity which would make such mechanisms theoretically plausible. However, the authors noted that most supporting evidence has come from animal or in-vitro studies, often using exposure levels that do not reflect typical human use. No human data have established a causal pathway linking standard paracetamol use in pregnancy to adverse neurodevelopmental outcomes.

The clinical implications of the findings are substantial given that paracetamol is often the only safe and accessible option for treating pain, and in particular fever, during pregnancy and is included on the World Health Organization’s ‘List of Essential Medicines’.

Untreated maternal fever is a recognised risk factor for miscarriage, congenital anomalies, preterm birth and adverse neurodevelopment. Discouraging paracetamol use on the basis of weak or biased evidence could therefore expose pregnant individuals and their foetuses to these well-established (and significant) harms.

The conclusions of this review align closely with guidance from major professional and regulatory bodies, including the American College of Obstetricians and Gynecologists, the UK’s Royal College of Obstetricians and Gynaecologists, and the European Union’s medicines regulator, the European Medicines Agency. All continue to recommend paracetamol as the first-line analgesic and antipyretic in pregnancy when used appropriately. The authors argued that the politicisation of scientific uncertainty has contributed to public confusion and that their findings provide a clearer evidential basis for clinical advice.

The study did acknowledge that uncertainty has not been eliminated entirely. Limitations included heterogeneity in exposure measurement, with many studies relying on maternal self-report and variability in outcome definitions reflecting changes in diagnostic criteria over time.

Although sibling-comparison designs reduce confounding from shared familial factors, they also reduce statistical power and cannot fully address confounders that vary between pregnancies. Despite these constraints, the convergence of evidence across multiple high-quality analyses strongly argued against a clinically important causal effect.

The authors concluded that maternal paracetamol use during pregnancy does not increase the likelihood of ASD, ADHD, or intellectual disability in offspring. By prioritising family-based designs and low-bias evidence, the review clarified that previously reported associations are most plausibly explained by residual confounding factors. The findings support existing clinical guidance and reinforce the position that paracetamol remains a safe and appropriate option for pain and fever management during pregnancy, when used as directed.

For further reading please visit: 10.1016/S3050-5038(25)00211-0