Defining a Sampling Strategy for Okadaic Acid (OA) Toxins in Shellfish to Determine the Human Health Status of a Growing Area

In order to ensure food safety in Europe, the level of okadaic acid (OA) equivalents in live bivalve molluscs must not exceed 160 μg/kg of raw meat (Regulation 853/2004/EC). Therefore, sampling plans must be set up to monitor production areas, knowing that mussels be used as an indicator species (Regulation 854/2004/EC). During a contamination event, there is a high variability of contamination between mussels. For this reason, determining an appropriate sampling plan is difficult and must be made on a scientific basis. However, there has never been a probabilistic evaluation of the impact that sample size and the number of samples taken, have on the analytical results and the corresponding decision for the public safety of a specific lot of shellfish. In this paper we achieve this aim by the application of scientific method to contamination data already available in scientific publications. Variance data recorded in scientific publications are used to predict variance as a function of OA concentration. A lognormal distribution fits the observed OA distributions among individual mussels. A computer model is used to predict the probability of acceptance for a given sampling plan. By OC (Operating Characteristic) curves, the performance of several sampling plan designs is evaluated to demonstrate how to manipulate sample size, and number of samples analysed to reduce misclassification of production areas. According to these OC curves, a best fit sampling plan is proposed, which consists of taking two samples of 30–40 mussels each.