Identification of pesticide mixtures to which French agricultural workers and farm-owners are exposed: Results from the Agriculture and Cancer (AGRICAN) cohort study

Abstract

Farmers, particularly in Europe, are exposed to multiple pesticides during their working life. Such exposures can cause adverse health outcomes. We aimed to identify the main pesticide mixtures to which French agricultural workers are exposed and to classify farmers into clusters based on their mixture exposure profile. The AGRICAN cohort includes farm-owners and farm workers enrolled from 2005 to 2007, with information on exact years of beginning and end of pesticide use on 11 crops and five livestock. We estimated duration of exposure to 390 pesticides identified with the PESTIMAT crop-exposure matrix for 16,905 male pesticide users from 1950 to 2009. We used a Sparse Non-negative Matrix Under-approximation to identify the main pesticide mixtures based on exposure duration, and then applied hierarchical agglomerative clustering to classify farmers sharing similar profiles of co-exposure to the mixtures. SNMU suggested 6 optimal numbers of mixtures (4, 7, 11, 15, 27, 38) explaining from 29 to 91 % of total variance. We selected 27 mixtures. Mixtures contained between four to 22 pesticides and mostly concerned the use of pesticides on wheat/barley, vineyards, corn, fruit and vegetables or on multiple crops together. We selected 11 clusters composed of 395 to 4521 farmers. Some had a higher proportion of individuals working on specific crops (as vineyard or corn), while others were characterized by the diversity of crops (cluster 8:"Permanent crops, potatoes and tobacco"). This is the first study to identify pesticide mixtures in farmers and to classify them into clusters based on their mixture exposure profiles. The next step will be to study the associations between pesticide mixtures and health outcomes such as prostate cancer in AGRICAN.

Keywords: Clustering; Cocktail; Crop-exposure matrix PESTIMAT; Farmers; HAC; Herbicide; SNMU.