Evaluation of models for predicting drug–drug interactions due to induction

The various methods for predicting clinical CYP3A induction from in vitro induction data all have demonstrated utility; it is the authors' opinion that the correlation-based approaches offer as good or better predictivity and have simpler input requirements than more complex approaches. Of the different correlation approaches, the relatively simple unbound C(max)/EC(50) or AUC/EC(50) approaches are the simplest and yet show the best correlation to the observed clinical data. While the approaches discussed herein represent an improvement in our understanding of the predictive value of in vitro induction data, it is important to recognize that there is still room for improvement in quantitative prediction of magnitude of drug interactions due to induction.