Organophosphate pesticides are established neurotoxicants, and the question of whether ordinary dietary-level prenatal exposure affects neurodevelopment has produced inconsistent findings for two decades. This pooled prospective cohort analysis is the largest attempt yet to settle the autism part of it.
It harmonised data from mother-child pairs across 20 sites in the US Environmental influences on Child Health Outcomes cohort, recruited between 1997 and 2019. Exposure was measured biologically, using maternal urinary concentrations of trichloropyridinol and the sum of six dialkyl phosphate metabolites. Autism-related traits were assessed by parent-reported Social Responsiveness Scale in 3,339 children at a mean age of 6.4 years.
Crude models suggested an inverse association — that is, apparently protective — which is itself a warning sign. After full adjustment for maternal sociodemographic and behavioural factors, both associations disappeared: per doubling of trichloropyridinol, the coefficient was -0.13 (95% CI -0.57 to 0.31), and per doubling of dialkyl phosphates, -0.42 (-1.03 to 0.19). Results held in sex-stratified analyses, in quantile and logistic models, and after adjusting for diet. There was no interaction with maternal diet quality.
The finding should be reported to worried parents as what it is: at exposure levels typical of a Western diet, no association with autism traits was found in over 3,000 children.
The caveat matters more in India than in the US. This cohort's exposure came mainly through food residues. Occupational and agricultural exposure — a pregnant woman working in a sprayed field, or living beside one — is a different and higher exposure that this study does not address, and organophosphate poisoning remains a common acute presentation in Indian practice.
- Reassure parents that dietary-level prenatal organophosphate exposure was not linked to autism traits in over 3,000 children
- Do not extend that reassurance to occupational or agricultural exposure, which was not studied here
- Note the exposure was measured biologically in maternal urine, which is stronger than questionnaire-based estimates
- Continue standard advice on washing produce, which addresses several exposures rather than this one
- Autism traits were parent-reported on a rating scale, not clinically diagnosed, which limits the outcome's precision
The statistics, in plain English
The confidence intervals here both cross zero and are narrow enough to be informative: the trichloropyridinol interval, -0.57 to 0.31, excludes anything larger than a third of a point on the Social Responsiveness Scale in either direction, which is far below clinical relevance. The more instructive detail is what happened between the crude and adjusted models. An apparently protective crude association that vanishes on adjustment usually means the exposure was tracking something else — here, plausibly, socioeconomic factors that correlate with both diet and reported child behaviour. When an implausible direction appears in unadjusted data, confounding is almost always the explanation.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for paediatrics, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free