Neural tube defects and maternal residential proximity to agricultural pesticide applications and crops

Residential proximity to agricultural pesticide applications may be an important source of environmental exposure to pesticides, particularly in rural communities. California maintains a database of pesticide-use reports (PUR) of agricultural applications, which can be mapped to a spatial resolution of approximately one square mile. To assess residential proximity to pesticide applications at distances of up to 1,000 meters, an exposure model was developed which improved the spatial resolution of the PUR by linking these data to land-use survey maps. Using this PUR/land-use model as a gold standard, PUR-only and land-use survey-only exposure models had poorer spatial resolution, thus leading to potential nondifferential misclassification and severe attenuation of true odds ratios (ORs) in case-control studies.;This PUR/land-use exposure metric was employed to assess maternal residential proximity within 1,000 meters of 59 specific pesticides during the first two calendar months of gestation in a case-control study of neural tube defects (NTDs) delivered in California. Hierarchical logistic regression was used to adjust for multiple comparisons and correlations between a large number of specific pesticide exposures. A moderate effect was observed from exposure to benomyl, a known teratogen. Employing a polytomous hierarchical model to estimate effects for the NTD subtypes anencephaly and spina bifida, differences in effects between subtypes were observed for certain methyl carbamate and organophosphorus pesticides. In addition, increases in NTD and subtype prevalence were associated with exposures to multiple pesticides.;In a study population of mothers of NTD cases and healthy-infant controls, land-use survey maps were used to validate self-reported residential proximity within 0.25 mile of agricultural crops. Effect estimates based on self-reported residential proximity to crops, except vineyards, appeared to be positively biased compared with those for map-based proximity. This difference appeared to be due to differential recall between cases and controls. Differences in reporting behavior were also observed between geographic regions of residence and categories of maternal demographic characteristics.