Prenatal levels of organophosphorus pesticide (OPs) biomarkers have been associated with hallmark
symptoms of Attention Deficit Hyperactivity Disorder (ADHD), including deficits in working memory, social
responsiveness, and ADHD indices. Cross-sectional studies have also linked concurrent pyrethroid pesticide
metabolites with ADHD and ADHD behaviors in children. Toxicology studies report mixture effects for these
pesticides on health outcomes. Three major gaps exist in this literature: 1) No studies have evaluated
prenatal and childhood exposures to these pesticides and clinical ADHD diagnoses. 2) Prenatal
epidemiological studies of OPs/pyrethroids typically rely on urinary DAP biomarkers, which reflect ingestion of
both non-toxic metabolites and toxic parent pesticides. Urinary biomarkers from spot urine samples also do
not characterize pregnancy-wide exposures, since OPs and pyrethroids are rapidly metabolized. 3)
Studies of pesticide mixtures in epidemiology are scarce, and use of biomarkers for such mixtures analysis is
problematic. For instance, administration of chlorpyrifos (an OP) results in increased tissue concentrations of
cypermethrin while reducing urinary excretion of the pyrethroid metabolite 3-phenoxbenzoic acid. Thus, the
mixture itself may affect biomarker levels and increase exposure misclassification. A geospatial framework for
pesticide exposure can address some of the limitations of urinary biomarkers: exposures from agricultural
pesticide applications can be estimated for an entire pregnancy rather than a few days; estimates reflect the
toxic parent pesticide rather than non-toxic OP metabolites; and estimates reflect actual exposures rather than
a post-metabolism level. However, geospatial (GIS) methods of pesticide exposure assessment for
epidemiology in the US have only been done in California, and usually rely on distance-to-field measures.
GIS exposure may be enhanced with drift models that incorporate heat, humidity, inversions, atmospheric
stability, and wind, while external validity may be increased by studying a population outside of California.
We propose to assess the relationship between OPs, pyrethroids, and ADHD in an Arizona population. To
identify a study population, we will apply a validated phenotyping algorithm with exceptional diagnostics to
Arizona Medicaid records to identify 4,000 childhood ADHD cases and 16,000 controls. In the mentored phase,
the Candidate will develop geospatial, phenotyping, exposure assessment, mixture modeling (Bayesian Kernel
Machine regression [BKMR]), and machine learning skills while constructing the case-control study. In the R00
phase, the Candidate will compare the drift model against traditional distance-to-field measures in a frequentist
framework (Aim 2), and model associations between prenatal OP and pyrethroid pesticide mixtures and ADHD
with BKMR (Aim 3). These results will expand GIS studies beyond California, contribute to sparse but critical
literature on pesticide mixtures and neurodevelopment, and be among the first to report associations between
GIS estimates of prenatal pesticide exposures and ADHD case status.