INTRODUCTION
Combined drug therapies have become increasingly popular in recent years due to their ability to simultaneously treat multiple conditions with a personalized combination of drugs (Gertz & Dispenzieri, 2020; Lu et al., 2017). However, the potential for drug-drug interactions (DDIs) to cause adverse effects and serious safety issues is a concern. In particular, when a perpetrator drug inhibits the metabolism of victim drugs via the CYP3A enzyme, it can lead to an increase in exposure of the victim drugs. To minimize the risk of negative effects from DDIs, the potential for interaction between perpetrator and victim drugs can be assessed during the drug discovery and development process and after the drug has been approved (Yang, Pfuma Fletcher, Huang, Zineh & Madabushi, 2021). A mechanistic static model can be used to evaluate the ratio of the concentration-time area under the curve (AUC) for a test article (i.e. a drug) after co-administration with a perpetrator of CYP3A to the AUC for the test article after dosing alone (AUCR, AUCi/AUC) (Gomez-Mantilla, Huang & Peters, 2023). If this ratio is found to be greater than the traditional threshold of 1.25, it is considered an indicator of a potential DDI, and further studies are performed to evaluate the DDI risk. However, static models have limitations in that they use a single concentration of a drug and can over-predict the extent of DDI. Dynamic models that take into account the time-varying concentrations of drugs and metabolites may provide a more accurate assessment of DDIs.
Physiologically based pharmacokinetic (PBPK) modeling is widely adopted by the pharmaceutical industry for evaluating DDIs due to its ability to consider the time-course of drug concentrations, resulting in more accurate predictions (Lin, Chen, Unadkat, Zhang, Wu & Heimbach, 2022). If the PBPK model can predict the observed DDI accurately, and sensitivity analysis indicates minimal impact, a clinical DDI study may not be necessary (Shebley et al., 2018). However, the decision to conduct such a study ultimately depends on various factors, such as regulatory requirements, the potential consequences of the DDI, the stage of drug development, and the predictive performance (accuracy and reliability) of the PBPK-DDI model. To evaluate the predictive performance of the PBPK model for DDIs, we used the PBPK model to predict the DDIs of 35 substrates after co-administration with an inhibitor, and then compare them with the observed results (Ren, Sai, Chen, Zhang, Tang & Yang, 2021). The comparison revealed that 75% of the predicted AUCR values by the model were within a 2-fold range of the observed AUCR values, with the assumption of 100% of the fraction metabolized (fm ). This preliminary analysis suggests that incorporating reported ”fm ” into the prediction model for certain drugs (crizotinib, macitentan, panobinostate, and ruxolitinib) could result in improved accuracy. However, further testing on fm and an expanded study with a larger sample size are necessary to accurately evaluate the improvement in predictive performance.
A commonly used method for quantifying fminvolves conducting an in vitro assay using human liver microsomes (HLM). This assay requires incubating a compound with HLM in the presence and absence of an inhibitor of interest and determining the clearance (Clint) of the compound under both conditions (Murayama et al., 2018). HLM are a prevalent in vitro model for determining drug metabolism or fm , as they are affordable and cost-effective. In addition to traditional methods such as in vitro or in vivo testing, in silico methods can be employed to predict the fate of molecules in the body through computer simulations (Watanabe et al., 2023). These methods are generally more economical. In this paper, we will describe the use of both in vitro and in silico methods to determine fmvalues. We will then use these values to update previous PBPK models of the substrates and further predict DDIs. Finally, we will evaluate the accuracy of the different methods used.