Bayesiansk beräkning av dos-responssamband kopplat till försöksdesign för att förbättra riskbedömning av kemikalier
Toxicological dose-response (DR) and dose-effect (ED) data are central parts of the hazard identification in all chemical regulatory domains (plant protection products, biocides, pharmaceuticals, food additives, environmental toxins, chemical products, etc.). It is therefore crucial to understand and develop tools that can be used to accurately and precisely describe DR and DE relationships. Extensive criticism of the traditional NOAEL methodology has resulted in increased use of Benchmark Dose (BMD) modeling. In the proposed project we investigate how the theoretical conclusions regarding BMD-guided experimental design works in practice. This is tested using a standardized test protocol with additional dose groups. The results will then be used to simulate experiments with variable number of doses and animals per dose group. In addition, we will refine the BMD model by implementing Bayesian statistics (instead of the current frequentist statistics). Doing so, we include prior information about models and model parameters (so-called priors or prior distributions). The Bayesian approach is expected to provide more accurate and toxicologically more relevant information regarding the likelihood of various DR/DE models and their parameter values (posteriors) and critical effect sizes. The prior information will be extracted from large amounts of original data from Swetox, NTP or from industrial collaborators.