Evaluation of gestational nonpersistent pesticide exposure with newborn size and gestational length in rural Ghana using a novel time-varying extension of multiple informant models.

Background: Gestational pesticide exposure may negatively affect newborn outcomes. Prior results evaluating nonpersistent pesticides are inconsistent.

Objective: To examine associations between gestational exposure to nonpersistent pesticides and newborn outcomes and identify critical windows of susceptibility.

Study Design: In a Ghanaian pregnancy cohort, we measured select biomarkers of organophosphate, pyrethroid, and herbicide pesticides in repeated urine samples (1-5/participant). We developed a new model for assessing critical windows of vulnerability from irregularly-timed measurements of nonpersistent pesticides, leveraging strengths from multiple informant and distributed lag models. We estimated associations of biomarker concentrations with newborn anthropometrics and gestational length, adjusting for confounders and exploring effect modification by infant sex and placental malaria.

Results: 1,211 pregnant women contributed 3,786 gestational urinary samples. In models assuming constant associations with exposures across pregnancy, in a given week a doubling of 3-phenoxybenzoic acid (pyrethroid biomarker) was associated with a -15.8 g difference in birth weight (95 % CI:-28.1,-3.6), and a doubling of the 2,4-dichlorophenoxyacetic acid (2,4-D, herbicide biomarker) was associated with an 11.1 g increase in birth weight (95 % CI:1.0,21.1). In time-varying models, significant associations were identified for pyrethroid exposure measured between weeks 16-27, and for 2,4-D exposure measured during weeks 25-33. Organophosphates were not associated with birth weight. No associations were found for birth length or head circumference for any pesticide. In constant association models, a doubling of weekly 2,4-D was associated with a 0.05 week increase in gestational length (95 %CI:0.01,0.09); no associations were found with other biomarkers.

Conclusions: We identified associations between gestational exposure to nonpersistent pesticides and both birth size and gestational length. Extending multiple informant models to account for the complex data structure allowed us to discern effects in opposing directions by distinct pesticide classes. While estimated effects for a given week were modest, prolonged or repeated exposures could result in larger cumulative impacts.