99% detection in 3,339 pregnancies turns an autism-traits test into a dose comparison

Start with the denominator. Ninety-nine percent of mothers had detectable organophosphate metabolites in urine collected during pregnancy ^[3], and one percent of 3,339 pairs is roughly 33 women without ^[15]. A design of that shape cannot set exposed children against unexposed children, because the unexposed group barely exists. It compares more exposure with less exposure inside a population where the chemical is everywhere, and lead author Haleigh Cavalier is explicit that the levels observed were relatively low ^[7].

The exposure measure carries its own limit. The authors caution that these metabolites are short-lived, vary over time, and are shaped by each participant's metabolism ^[12]. A urine sample is therefore a noisy stand-in for months of intake. Where that noise is unrelated to the child's later score, the usual consequence is that estimated associations shrink toward zero rather than inflate, which makes a null built on single-occasion sampling weaker evidence of no effect than a null built on repeated sampling. That is my reading of the measurement, not a claim the study makes.

The outcome is a trait scale rather than a diagnosis. The Social Responsiveness Scale is a 65-item questionnaire that rates social communication difficulty on a spectrum, including in children never formally diagnosed, covering things like avoidance of eye contact and difficulty reading other people's emotions ^[5]. So the thing this does not tell you is whether diagnosed autism is distributed differently across exposure levels. It tells you that continuous trait scores, measured at ages five to seven on average, did not track metabolite concentrations ^[6][4].

What the study does buy is scale. Small studies of prenatal organophosphate exposure have gone both ways ^[10], and a 3,339-pair sample drawn from 20 centres in the NIH ECHO programme ^[2] allowed the team to account for environmental, social and biological factors that smaller cohorts could not ^[14]. That is a genuine advance over the existing literature, and it leaves the older findings on IQ loss, cognitive difficulty and ADHD ^[9] untested here rather than overturned.

One loose end worth watching: Cavalier notes that outside agricultural and occupational settings, most Americans meet these insecticides on fresh produce, and suggests the benefits of a fruit-and-vegetable diet may partly offset any effect on the developing brain ^[11]. That is a hypothesis about two opposing effects arriving in the same food, and nothing in this design separates them.