Validation and sensitivity analysis of a probabilistic model for dietary exposure assessment to pesticide residues with a Basque Country duplicate diet study

Abstract

The ‘fitness for pulyose’ of a probabilistic model designed to assess dietary exposure to pesticides was validated. The model had to meet two prerequisites. First, it should provide more realistic estimates of intake than conservative methods. Second, it should not underestimate'true' intakes. True intakes were estimated using a duplicate diet study. Three approaches were used to check the prerequisites: visual comparison, a statistical test of a high percentile, and a comparison for each infant of the duplicate diet, conservative and model intake values. Compliance with the prerequisites was met for the six pesticides selected, in the three approaches. Model outcome distributions reduced the uncertainty, considered as the difference between conservative and duplicate diet intakes, by 75–98% for high percentiles, depending on the pesticides. A sensitivity analysis of the model based on analysis of variance for selected factors was conducted for three pesticides. The,factors included concentration and food consumption input data presentations, values assigned to pesticide–food commodities without analysis, values assigned to samples with results below the limit of reporting, unit-to-unit variability and processing factors. Their signficance and relevance were studied. Assigning values to pesticide–food commodities without analysis and processing factors, when available, were the most relevant factors in this study.

Keywords:

• probabilistic modelling
• pesticide residues
• model validation