The liver poses particular problems in constructing physiologically-based pharmacokinetic models since this organ is not only a distribution site for drugs/chemicals but frequently the major site of metabolism. The impact of hepatic drug metabolism in modelling is substantial and it is crucial to the success of the model that in vitro data on biotransformation be incorporated in a judicious manner. The value of in vitro/in vivo extrapolation is clearly demonstrated by considering kinetic data from incubations with freshly isolated hepatocytes. The determination of easily measurable in vitro parameters, such as V(max) and K(m), from initial rate studies and scaling to the in vivo situation by accounting for hepatocellularity provides intrinsic clearance estimates. A scaling factor of 1200 x 10(6) cells per 250 g rat has proved to be a robust parameter independent of laboratory technique and insensitive to animal pretreatment. Similar procedures can also be adopted for other in vitro systems such as hepatic microsomes and liver slices. An appropriate scaling factor for microsomal studies is the microsomal recovery index which allows for the incomplete recovery of cytochrome P-450 with standard differential centrifugation of liver homogenates. The hepatocellularity of a liver slice has been unsatisfactory in scaling kinetic parameters from liver slices. The level of success varies from drug to drug and substrate diffusion is a competing process to metabolism within the slice incubation system; hence, low clearance drugs are better predicted than high clearance drugs. The use of three liver models (venous-equilibration, undistributed sinusoidal and dispersion models) have been compared to predict hepatic clearance from in vitro intrinsic clearance values. As no consistent advantage of one model over the others could be demonstrated, the simplest, the venous-equilibration model, is adequate for the currently available data in hepatocytes. While these successes are encouraging as they establish the fidelity of in vitro systems for in vivo prediction, the level of success varies from drug to drug. It is important to address the reasons for failure of prediction by each in vitro system and it is noteworthy that the current approach simplifies several key issues.