Analysis of multivariate extreme intakes of food chemicals

A recently published multivariate Extreme Value Theory (EVT) model [Heffernan, J.E., Tawn, J.A., 2004. A conditional approach for multivariate extreme values (with discussion). Journal of the Royal Statistical Society Series B 66 (3), 497¿546] is applied to the estimation of population risks associated with dietary intake of pesticides. The objective is to quantify the acute risk of pesticide intake above a threshold and relate it to the consumption of specific primary food products. As an example daily intakes of a pesticide from three foods are considered. The method models and extrapolates simultaneous intakes of pesticide, and estimates probability of exceeding unobserved large intakes. Multivariate analysis was helpful in identifying whether the avoidance of certain food combinations would reduce the likelihood of exceeding a threshold. We argue that the presented method can be an important contribution to exposure assessment studies.