Captan exposure and evaluation of a pesticide exposure algorithm among orchard pesticide applicators in the Agricultural Health Study

Abstract

Pesticide exposure assessment in the Agricultural Health Study (AHS) has relied upon two exposure metrics: lifetime exposure days and intensity-weighted lifetime exposure days, the latter incorporating an intensity score computed from a questionnaire-based algorithm. We evaluated this algorithm using actual fungicide exposure measurements from AHS private orchard applicators. Captan was selected as a marker of fungicide exposure. Seventy-four applicators from North Carolina and Iowa growing apples and/or peaches were sampled on 2 days they applied captan in 2002 and 2003. Personal air, hand rinse, 10 dermal patches, a pre-application first-morning urine and a subsequent 24-h urine sample were collected from each applicator per day. Environmental samples were analyzed for captan, and urine samples were analyzed for cis-1,2,3,6-tetrahydrophthalimide (THPI). Task and personal protective equipment information needed to compute an individual's algorithm score was also collected. Differences in analyte detection frequency were tested in a repeated logistic regression model. Mixed-effects models using maximum-likelihood estimation were employed to estimate geometric mean exposures and to evaluate the measured exposure data against the algorithm. In general, captan and THPI were detected significantly more frequently in environmental and urine samples collected from applicators who used air blast sprayers as compared to those who hand sprayed. The AHS pesticide exposure intensity algorithm, while significantly or marginally predictive of thigh and forearm captan exposure, respectively, did not predict air, hand rinse or urinary THPI exposures. The algorithm's lack of fit with some exposure measures among orchard fungicide applicators may be due in part to the assignment of equal exposure weights to air blast and hand spray application methods in the current algorithm. Some modification of the algorithm is suggested by these results.