Development of a nutrient database and distributions for use in a probabilistic risk–benefit analysis of human seafood consumption

Human consumption of seafood can be promoted because of its positive health effects. Conversely, it is a source of chemical contaminants. Due to this dilemma, a probabilistic intake assessment of nutrients and contaminants via seafood is of interest to provide more detailed information. A key component of such an assessment is the selection of the most appropriate input distributions to describe the consumption and concentration data. This paper describes the construction of a nutrient database, pooling vitamin D and omega-3 fatty acid concentrations in seafood from different publications and the encountered problems related to a lack or inconsistency of information given in these publications: food description, number of samples, sampling plan, sources of the values, limit of quantification (LOQ), etc. Different solutions have been proposed and the study resulted in a huge database allowing the description of distributions of vitamin D, eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) concentration and their variability in 34 seafood species relevant for Belgian consumption. The distribution fitting and selection procedure resulted in different models for the different species and nutrients. The normal and lognormal distributions are most frequently used, followed by the uniform, beta and loglogistic distribution.