Abstract
More than 20 previously reported lines of independent evidence from clinical observations, studies in laboratory animal models, pharmacokinetic considerations, and numerous temporal and spatial associations indicate that numerous genetic and environmental factors leading to inflammation and oxidative stress confer vulnerability to the aberrant metabolism of acetaminophen during early development, leading to autism spectrum disorder (ASD). Contrary to this conclusion, multivariate analyses of cohort data adjusting for inflammation-associated factors have tended to show little to no risk of acetaminophen use for neurodevelopment. To resolve this discrepancy, here we use in silico methods to create an ideal (virtual) population of 120,000 individuals in which 50% of all cases of virtual ASD are induced by oxidative stress-associated cofactors and acetaminophen use. We demonstrate that Cox regression analysis of this ideal dataset shows little to no risk of acetaminophen use if the cofactors that create aberrant metabolism of acetaminophen are adjusted for in the analysis. Further, under-reporting of acetaminophen use is shown to be a considerable problem for this analysis, leading to large and erroneously low calculated risks of acetaminophen use. In addition, we argue that factors that impart susceptibility to acetaminophen-induced injury, and propensity for acetaminophen use itself, can be shared between the prepartum, peripartum, and postpartum periods, creating additional difficulty in the analysis of existing datasets to determine risks of acetaminophen exposure for neurodevelopment during a specific time frame. It is concluded that risks of acetaminophen use for neurodevelopment obtained from multivariate analysis of cohort data depend on underlying assumptions in the analyses, and that other evidence, both abundant and robust, demonstrate the critical role of acetaminophen in the etiology of ASD.