A0335
Title: PCBs intake assessment using a general Bayesian network depending on the meat safety monitoring system
Authors: Hassan Hachem - Universite Paris-Saclay, AgroParisTech, INRAE, UMR MIA Paris-Saclay, 91120, Palaiseau, France (France) [presenting]
Isabelle Albert - (France)
Abstract: Polychlorinated biphenyls (PCBs) are targeted contaminants in the current European monitoring system due to their overall relevance in terms of safety risks, particularly in animal-derived food. As part of SENTINEL, an ANR research project whose objective is to strengthen the monitoring of chemical safety in food, the exposure to PCBs in France associated with the consumption of pork meat is evaluated under the current monitoring system and an alternative one proposed in SENTINEL based on sample pooling. To this end, a general Bayesian network, using discrete and continuous random variables, is proposed to assess the PCBs exposure in pork meat, gathering the sparse contamination and consumption data along the food pathway. The Bayesian modular model considers conventional and organic production meat modes, intake levels, quantities, and frequencies of pork meat consumption. Pseudo-contamination data are introduced in the contamination module to integrate different safety monitoring strategies All models also include historical data already collected as prior information. The results of the models are the dietary consumption exposure which can be compared to evaluate the safety impact of new monitoring strategies. MCMC sampling algorithms were implemented through the R package RJAGS, and pseudo data were produced from R functions. In the future, it will be interesting to fully integrate the monitoring system in the Bayesian model instead of using pseudo data.
