Badee, Justine Marine, Huth, Felix, Poller, Birk, Schiller, Hilmar, Chenal, Gaelle, Birlinger, Bertrand-Luc, Streckfuss, Judith, Gu, Helen, Einolf, Heidi and Deshmukh, Sujal (2025) Evaluation of various experimental conditions and mechanistic static vs. dynamic models to predict time-dependent CYP3A4/5 inhibition potential of drugs. Drug Metabolism Disposition.

Abstract

The use of mechanistic static and dynamic physiologically-based pharmacokinetic (PBPK) models by incorporating cytochrome P450 (CYP)3A4/5-mediated time-dependent inhibition (TDI) parameters from human liver microsomes (HLM) can potentially give rise to significant overprediction of drug-drug interactions (DDI) caused by TDI, which may result in conducting unnecessary clinical DDI trials. This work aimed to evaluate the predictive performance of mechanistic static and dynamic PBPK models employed to predict the likelihood and the magnitude of clinical DDI caused by drugs with in vitro CYP3A4/5 TDI parameters measured in HLM and human hepatocytes (HHEP). We examined the effect of differences in in vitro CYP3A4/5 TDI parameters such as the inhibition constant (KI) (total or unbound) in experimental conditions (supplementation of glutathione in HLM incubations or plasma in HHEP incubations) on the magnitude of predicted DDI risk in comparison to clinical results. In mechanistic static models, the average unbound organ exit concentrations and the maximum organ entry concentrations were compared for projecting DDI risks. Model performance was assessed using false-negative rates and negative predictive errors for a cutoff value of either 1.25- or 2-fold change in midazolam exposure. DDI caused by CYP3A4/5-mediated TDI was reliably predicted using mechanistic static model with average unbound organ exit concentrations or dynamic PBPK modeling, yielding less marked overpredictions of DDI. Models using in vitro KI corrected for incubation unbound fraction generated in either HLM or HHEP buffer incubations showed best statistical performance while maintaining high prediction accuracy and precision.

Item Type:	Article
Date Deposited:	10 Feb 2026 00:45
Last Modified:	10 Feb 2026 00:45
URI:	https://oak.novartis.com/id/eprint/56854