Separation of True Uncertainty and Interindividual Variability in Human Exposure Modelling

Abstract

Listen

Translate article

P-628 Abstract: Input parameters for exposure modeling can be uncertain due to a lack of knowledge (true uncertainty) or due to variability (e.g. interindividual variability). Ordinary Monte-Carlo simulation combines both types of uncertainty in the variance of one output distribution, which impedes an unambiguous interpretation of the model output. Separation of true uncertainty and interindividual variability can be realized by means of nested Monte-Carlo simulation. In this study, nested Monte-Carlo simulation was applied to the human exposure model NORMTOX. NORMTOX is an integrated human exposure model, which predicts lifetime averaged daily intake levels of contaminants from soil, air, surface and drinking water and food products. The output indicates which fraction of the population is at risk and details the probability of this risk, e.g. “there is a probability of 10% that 5% of the population will exceed a daily intake of 5 mg”. This provides risk managers with the opportunity to evaluate population fractions at risk as well as the reliability of the risk prediction. Here, NORMTOX is applied to test the coherence between environmental quality objectives (EQOs) and the acceptable daily intake (ADI). Because EQOs for different media are often derived independently, they may be incoherent, i.e. simultaneous exposure to multiple media that are all polluted up to their EQO may result in an intake exceeding the ADI. Coherence is expressed using a coherence indicator (CI): the ratio between the predicted intake and the ADI. The coherence of Dutch EQOs and ADIs was calculated for 54 substances; mainly heavy metals and pesticides. Most substances show a CI value clearly below 1.0 for the entire population, indicating that the ADI will not be exceeded when all media are polluted up to their EQO. For four substances, there is a probability of 99% that the entire population exceeds a CI value of 1.0. Dutch EQOs should be lowered for these substances in order to guarantee that the ADI is not being exceeded. Furthermore, there are eleven substances for which the CI is around 1.0. This means that part of the population exceeds the ADI or that the ADI will be exceeded with a particular probability. Interindividual variability is the major source of variance in the predicted daily intake. True uncertainty is relevant only for for lead and cadmium. The presentation will focus on the methods used in NORMTOX and on the interpretation of the generated results in terms of uncertainty and interindividual variability.