According to the FDA’s Guidance for Industry on Drug-drug interactions (DDIs), assessment of a new drug’s DDI liability has three major objectives:
- determining whether any interactions necessitate dosing adjustment,
- informing the extent of therapeutic monitoring that may be required and
- identifying any potential contraindications to concomitant use when lesser measures cannot mitigate risk
Physiologically-based pharmacokinetic (PBPK) models have been used to support the clinical pharmacology reviews of new drug applications. They can predict changes in PK due to DDIs and support appropriate dose adjustments. These models are often “fit for purpose.” Although they may account for a drug’s ability to mediate DDIs via enzyme inhibition and enzyme induction, they may not account for the effect the drug has on its own metabolism (auto-inhibition or auto-induction). In our recently published paper, we used the example of efavirenz to illustrate best practices for developing a unified mechanistic PBPK model.
Efavirenz—a drug with complex DDI properties
The antiviral drug, efavirenz, is a known perpetrator of DDIs via CYP2B6 and CYP3A4 induction. Upon multiple dosing, efavirenz also is a DDI victim due to inducing its own clearance via CYP2B6. Patients who are CYP2B6 poor metabolizers risk experiencing efavirenz toxicity and thus usually require dose adjustments. Because of these properties, regulatory agencies expect sponsors to quantify the DDI potential of novel CYP3A4-substrate drugs due to efavirenz co-administration.
A conceptual framework for a unified mechanistic model of efavirenz PK
To develop a model that would account for both its victim and perpetrator properties, we used a step-wise approach to verify the different model components. The Simcyp population-based PBPK Simulator v14.1 was utilized to perform all simulations in virtual healthy volunteers. A model was deemed successful in predicting PK parameters and DDIs if the predicted/observed exposure ratio fell within a 1.5-fold range. Drug exposure was defined as maximum plasma concentration (Cmax) or the area under the concentration-time curve (AUC).
Model development and verification strategy
The efavirenz base model used physiochemical data, in vitro, and clinical data to simulate the time-concentration profile of efavirenz following multiple dosing. Next, in vitro CYP3A4 induction data were incorporated into the base model to predict the DDI with different CYP3A4 substrates. Studies with intravenous and oral alfentanil (a CYP3A4 victim drug) were used to refine the model. Then, the updated model was verified by predicting the DDI for different CYP3A4 drugs, maraviroc, atazanavir and clarithromycin.
Now that the model was able to predict DDIs mediated by CYP3A4 induction, it was expanded to predict DDIs due to CYP2B6 induction. In vitro CYP2B6 data were incorporated into the model which was then used to predict the DDI for the CYP2B6 substrate, bupropion. Finally, the model was refined using in vitro data describing the fractional contribution of metabolizing enzymes. The final model was then used to simulate the exposure resulting from single and multiple dosing of efavirenz in healthy volunteers.
Developing complex models for predicting DDIs
Our model is the first to predict DDIs caused by efavirenz inducing CYP3A4 and CYP2B6 as well as auto-induction. In addition, our model has the novel ability to account for organ-specific induction of CYP3A4 in the liver and gut. This verified mechanistic model can help assess the impact of co-administration of moderate CYP3A4 inducers and thus streamline PBPK packages in regulatory submissions.
Ke A, Barter Z, Rowland-Yeo K, Almond L. Towards a best practice approach in PBPK modeling: Case example of developing a unified efavirenz model accounting for induction of CYPs 3A4 and 2B6. CPT Pharmacometrics Syst Pharmacol. 2016;5(7):367-376.
To learn more about how sponsors are using PBPK to evaluate DDIs, please watch this webinar by Dr. Marylore Chenel, director of Pharmacometrics at Servier.