Dose response models calculation
Dose response models are uploaded as data or retrieved from the PROASTweb through linked remote repositories. A second possibility is to compute dose response models using an integrated version of PROAST: for each response in a dose response experiment a dose response model is fitted. Depending on the type of data (e.g., response type, covariates y/n, single or multiple substances) a PROAST run is configured and executed. If effect representations are provided, then benchmark responses specified by the effect representations data are used, otherwise only the model fits will be computed without benchmark doses.
PROAST (copyright RIVM National Institute for Public Health and the Environment) is a software package for the statistical analysis of dose-response data. Its main purpose is dose-response modelling of toxicological data, and the derivation of a Benchmark dose (BMD or BMDL) in human risk assessment (or an ECx in ecotoxicological risk assessment). More generally, it can be used for (nonlinear) regression (with covariates) (see Slob (2002), Slob and Setzer (2013))
Figure 98 Dose response model: HepaRG Adipo72 72h.
When PROAST is run without uncertainty, the BMD is estimated and the BMDL and BMDU are based on the profile likelihood method where the Fisher information matrix indicates the estimate’s precision. When uncertainty is specified,the setting Calculate a parametric benchmark dose confidence interval becomes relevant.
checked: a benchmark dose interval (90% confidence) based on the profile likelihood method is estimated,
unchecked: the BMDL (p5) and BMDU (p95) are based on the 90% confidence interval of the bootstrapped BMD values.
In module hazard characterisations specify with setting Use lower limit of BMD whether a BMD or BMDL is used in the risk assessment.