Predictive Performance of Physiologically-Based Pharmacokinetic Models in Predicting Drug-Drug Interactions Involving Enzyme Modulation

BackgroundPhysiologically-based pharmacokinetic (PBPK) modeling in predicting metabolic drug-drug interactions (mDDIs) is routinely used in drug development. Currently, the US FDA endorses the use of PBPK to potentially support dosing recommendations for investigational drugs as enzyme substrates of mDDIs, and to inform a lack of mDDIs for investigational drugs as enzyme modulators.MethodsWe systematically evaluated the performance of PBPK modeling in predicting mDDIs published in the literature. Models developed to assess both investigational drugs as enzyme substrates (Groups 1 and 2, as being inhibited and induced, respectively) or enzyme modulators (Groups 3 and 4, as inhibitors and inducers, respectively) were evaluated. Predicted ratios of the area under the curve (AUCRs) and/or maximum plasma concentration (C(max)Rs) with and without comedication were compared with the observed ratios.ResultsFor Groups 1, 2, 3, and 4, 62, 50, 44, and 43% of model-predicted AUCRs, respectively, were within a predefined threshold of 1.25-fold of observed values (0.8-1.25x). When the threshold was widened to twofold, the values increased to 100, 80, 81, and 86% (0.5-2.0x). For Groups 3 and 4, prediction for mDDI liability (the existence or lack of mDDIs) using PBPK appears to be satisfactory.ConclusionOur analysis supports the FDA's current recommendations on the use of PBPK to predict mDDIs.