Prediction of human metabolic clearance from in vitro systems: retrospective analysis and prospective view

To provide a definitive assessment of prediction of in vivo CL (int) from human liver in vitro systems for assessment of typical underprediction.

A database of published predictions of clearance from human hepatocytes and liver microsomes was compiled, including only intravenous CL (b). The influence of liver model (well-stirred (WS) or parallel tube (PT)), plasma protein binding and clearance level on the relationship between in vitro and in vivo CL (int) was examined.

Average prediction bias was about 5- and 4-fold for microsomes and hepatocytes, respectively. Reduced bias using the PT model, in preference to the popular WS model, was only marginal across a wide range of clearance with a consequential minor impact on prediction. Increasing underprediction with decreasing fu (b), or increasing CL (int), was found only for hepatocytes, suggesting fundamental in vitro artefacts rather than failure to model potentially unequilibrated binding during rapid extraction.

In contrast to microsomes, hepatocytes give a disproportionate prediction with increasing clearance suggesting limitations either at the active site, such as cofactor exhaustion, or with intracellular concentration equilibrium, such as rate-limiting cell permeability. A simple log linear empirical relationship can be used to correct hepatocyte predictions.