Defining food sampling strategies for chemical risk assessment

Collection of accurate and reliable data is a prerequisite for informed risk assessment and risk management. For chemical contaminants in food, contamination assessments enable consumer protection and exposure assessments. And yet, the accuracy of a contamination assessment depends on both chemical analysis and sampling plan performance. A sampling plan is always used when the contamination level of a food lot is evaluated, due to the fact that the whole lot can not be analysed, but only samples, which are drawn from the lot. An efficient sampling plan enables to take samples from a food lot, following a given protocol, with a relatively low risk of misestimating the true mean concentration of the food lot after analysis of the food samples. Sampling plan performance testing is achieved thanks to mathematical validation methods. The best fit sampling plan is the one that gives the best compromise between the lowering of the risk of misestimating the true lot concentration and the practical feasibility (not too much time consuming nor money consuming). This chapter presents two sampling plan validation strategies: a parametric method developed by Whitaker and co-workers from 1972 and a non parametric method set up by Schatzki et al. (Schatzki, 1995; Campbell et al. 2003). To our knowledge, these are the only two methods sufficiently evolved for having been applied to real situation cases for food sampling validation. These statistical methods are first explained from a theoretical point of view. Then, each one is illustrated by a practical application to a sampling plan validation for a specific chemical risk in a food commodity, thanks to workable contamination data gathered in the literature. According to us, in its general mathematical principle, the non parametric method is more appropriate t