A 2025 Navigation Guide systematic review concluded there was “strong evidence of a likely relationship” between prenatal acetaminophen exposure and neurodevelopmental disorders (NDDs). That conclusion was based on qualitative synthesis without meta-analytic pooling or formal corrections for bias. The current study reanalysed the same body of evidence using standard quantitative synthesis and multiple bias-correction methods to test whether the claimed associations remain robust under these approaches. The authors emphasise that the report is a preprint and has not been peer reviewed; it should not be used to guide clinical care.
From the 2025 Navigation Guide review (PubMed search through 25 February 2025), the authors selected 24 of 46 studies that reported a ratio measure (hazard ratio, odds ratio, relative risk, or incidence rate ratio). Estimates were collapsed so there was one effect per study, outcome group, and design. This produced 30 study-level estimates addressing three outcome groups: ADHD, ASD, and other neurodevelopmental disorders. Reported study sample sizes ranged from 307 to 2,480,797.
The reanalysis applied a random-effects meta-analysis as the primary pooling method. To evaluate and adjust for potential biases, the authors implemented eight publication-bias-correction methods, performed a credibility-ceiling sensitivity analysis, and analysed five quality-filtered subsets of the evidence. They also harmonised reported odds and hazard ratios onto a common risk-ratio scale and repeated all analyses on the harmonised measures. P-curve testing was used to assess right-skew consistent with selective reporting.
When studies were weighted by a single study-level estimate (rather than by number of sub-analyses contributed), the pooled estimate for autism spectrum disorder (ASD) attenuated from 1.14 (95% CI, 1.01–1.28) to 1.08 (95% CI, 0.97–1.19; P = .17) before any additional bias correction. After harmonising effect measures and applying bias-correction and sensitivity procedures, ASD point estimates moved further toward the null and the confidence intervals included 1.00. The authors report that, overall, the ASD association was not robust to the analytic choices made to reduce multiplicity and address bias.
The pooled association for attention-deficit/hyperactivity disorder (ADHD) remained positive under most bias-correction methods. Reported pooled estimates across methods ranged from 1.02 to 1.31. However, under a 15% credibility ceiling sensitivity analysis — a conservative approach that downweights the contribution of studies subject to systematic error — the 95% confidence interval for the ADHD estimate crossed 1.00. Thus, while a modest association with ADHD persisted in many analyses, it was sensitive to stricter bias-adjustment assumptions.
Analyses of the composite category of other neurodevelopmental disorders (other-NDD) produced less consistent results. In one bias-adjustment method the pooled estimate reversed direction (ratio 0.98; 95% CI, 0.65–1.19). P-curve testing for the other-NDD group showed no right-skew (P = .71), suggesting the pattern of reported p-values did not display the distribution expected if there were a consistent true positive effect and selective reporting was absent. Overall, the other-NDD evidence did not support a clear positive association after bias correction.
This reanalysis did not corroborate the prior characterization of “strong evidence” supporting a prenatal acetaminophen–NDD association. After applying standard meta-analytic pooling, multiple publication-bias-correction techniques, credibility-ceiling sensitivity, and harmonisation of effect measures, the association with ASD moved toward null and was not statistically significant by the reported analyses. A modest association with ADHD persisted under many methods but lost statistical significance under a conservative credibility-ceiling assumption. Results for other neurodevelopmental disorders were inconsistent and, in some analyses, reversed direction.
The authors conclude that the weight of evidence does not support the earlier strong claim and that findings are sensitive to analytic choices and bias-correction approaches. They present these results to inform interpretation of the existing epidemiologic literature and related policy statements, while noting limitations inherent to reanalysing observational studies that vary in design, exposure measurement, outcome ascertainment, and adjustment.
All code and data used in the reanalysis are publicly available at the authors’ GitHub repository: https://github.com/ambika241/acetaminophen-ndd-reanalysis. The preprint states that the authors declared no competing interests and confirmed adherence to ethical and reporting guidelines. The manuscript is a preprint posted September 4, 2026, and the authors explicitly note it has not undergone peer review; therefore, the findings should not be used to change clinical practice without further validation and peer-reviewed confirmation.