The present study was designed to explain the differences in isoprene toxicity between mouse and rat based on the liver concentrations of the assumed toxic metabolite isoprene diepoxide. In addition, extrapolation to the human situation was attempted. For this purpose, enzyme kinetic parameters Km and Vmax were determined in vitro in mouse, rat and human liver microsomes/cytosol for the cytochrome P450-mediated formation of isoprene mono- and diepoxides, epoxide hydrolase mediated hydrolysis of isoprene mono- and diepoxides, and the glutathione S-transferases mediated conjugation of isoprene monoepoxides. Subsequently, the kinetic parameters were incorporated into a physiologically-based pharmacokinetic model, and species differences regarding isoprene diepoxide levels were forecasted. Almost similar isoprene diepoxide liver and lung concentrations were predicted in mouse and rat, while predicted levels in humans were about 20-fold lower. However, when interindividual variation in enzyme activity was introduced in the human model, the levels of isoprene diepoxide changed considerably. It was forecasted that in individuals having both an extensive oxidation by cytochrome P450 and a low detoxification by epoxide hydrolase, isoprene diepoxide concentrations in the liver increased to similar concentrations as predicted for the mouse. However, the interpretation of the latter finding for human risk assessment is ambiguous since species differences between mouse and rat regarding isoprene toxicity could not be explained by the predicted isoprene diepoxide concentrations. We assume that other metabolites than isoprene diepoxide or different carcinogenic response might play a key role in determining the extent of isoprene toxicity. In order to confirm this, in vivo experiments are required in which isoprene epoxide concentrations will be established in rats and mice.