Evaluating the performance of sampling plans for phycotoxins in shellfish: Improvement of an existing method

For sampling plan validation purposes, a classical mathematical method was improved for the prediction of variance as function of the mean contamination level thanks to prior knowledge of the theoretical distribution fitting the observed OA levels among individual mussels. Indeed, knowing that, thanks to a regression analysis of literature data, for the lognormal distribution the scale parameter was observed to be directly proportional to the location parameter, the regression bias could be lowered. Literature data from Norway and Sweden showed different levels of variability between contamination events and depuration. However, the highest variability level was chosen to propose a best fit sampling plan in order to have a better approach of reality. It consisted of taking two samples of 50 mussels (Mytilus sp.) for this geographic location (Norway and Sweden).