|
(21-20) Prediction of mixture toxicities: How to deal with Uncertainty ? Scholze, Martin*,1, Backhaus, Thomas1, Faust, Michael1, Junghans, Marion1, Meyer, Wiebke1, Grimme, LH1, 1 Dept. of Biology and Chemistry, University of Bremen, Bremen, Germany ABSTRACT- Two different concepts are used to calculate predictions of mixture toxicities: Concentration Addition and Independent Action. For that purpose the concentration-response relationships of all mixture components have to be known. They are usually estimated on the basis of experimental data. Thus, both the single substance toxicity values and the predicted mixture toxicities are subject to stochastic uncertainty. This always incorporates a random and systematic part, i.e. precision and accuracy. Random errors can be estimated and, if applicable, reduced by adding more data. This is usually impossible for the systematic part which is unknown. However, with increasing numbers of components the chance increases that one or more toxicity estimates are subject to bias, and hence the overall prediction of the mixture toxicity may also be affected by a bias (error propagation). Potentially this may result in a strong under- or overestimation of mixture toxicity. In order to avoid the introduction of major systematic errors, it is essential to identify relevant processes. We demonstrate strategies to avoid inaccurate predictions by performing a sensitivity and uncertainty analysis for different sets of experimental single substance toxicity data (algae, daphnids, fish). Biometrical demands for a predictive assessment of mixture toxicities in terms of data quality and quantity will be discussed. Key words: Mixture toxicity, Uncertainty, Concentration Addition, Independent Action |
