Physiologically Based Pharmacokinetic Modeling to Assess Perpetrator and Victim Cytochrome P450 2C Induction Risk

Abstract

Background: Accurate assessment of CYP2C induction-mediated drug-drug interactions (DDIs) remains a challenge, despite the importance of CYP2C enzymes in drug metabolism. Limitations in available models and scarce clinical induction data have hampered quantitative preclinical DDI risk evaluation. Methods: In this study, the authors utilized an all-human hepatocyte triculture system to capture CYP2C induction using the perpetrators rifampicin, efavirenz, carbamazepine, and apalutamide. In vitro induction parameters were quantified by measuring changes in both mRNA and enzyme activities for CYP2C8, CYP2C9, and CYP2C19. These induction parameters, along with CYP-specific intrinsic clearance (CL_int) for the victim compounds, were incorporated into a physiologically based pharmacokinetic (PBPK) model, and pharmacokinetics (PK) of known CYP2C substrates were predicted with and without co-administration of perpetrator compounds using clinical dosing regimens. The results were quantitatively compared with the currently utilized mechanistic static modeling (MSM) approach and the reported clinical DDI outcomes. Results: By incorporating the measured f_m of CYP2C substrates into PBPK modeling, we observed a lower propensity to over- or underpredict the exposure of these substrates as victims of CYP2C induction-based DDIs when co-administered with known perpetrators, which resulted in an excellent correlation to observed clinical outcomes. The MSM approach predicted the CYP3A4 induction-based DDI risk accurately but could not capture CYP2C induction with similar precision. Conclusions: Overall, this is the first study that demonstrates the utility of PBPK modeling as a complementary approach to MSM for CYP2C induction-based DDI risk assessment.

Keywords: CYP2C induction; DDI; PBPK; hepatocytes; pharmacokinetics.

Figure 1

PBPK modeling workflow. The PBPK modeling workflow depicted in the figure involves use of two validation steps: (1) modifying compound files for victim drugs and perpetrator rifampicin with the induction data obtained from the HTC model and validating the model with observed C_max, AUC, and clearance differences; and (2) re-validating the finalized model by overlaying the observed clinical data and simulated data to finalize the model using the rifampicin–midazolam DDI as a positive control.

Figure 2

Representative concentration response graphs from donor 1 for HTC model ((A)—mRNA, (B)—CYP activity). Carbamazepine in closed black circles, rifampicin in closed red triangles, efavirenz in open black circles, apalutamide in closed green circles. The fold change in mRNA level compared to the vehicle control value is displayed on the y-axis, with nominal added concentrations displayed on the x-axis. Values represent mean ± SD of n = 4 values.

Figure 3

Prediction of CYP induction with HTC data using PBPK vs. static modeling after treatment with (A) rifampicin, (B) efavirenz, (C) carbamazepine and (D) apalutamide. AUC ratios of victim drugs with and without rifampicin are shown as green circles denoting PBPK simulated data AUC ratios, red circles denoting MSM predicted AUC ratios, and blue circles denoting clinically observed AUC ratios for the victim drugs. Victim drug f_m values for each CYP involved in their metabolism are listed below the graph. The dashed red lines indicate the acceptable average bioequivalence range (0.8–1.25), and the dotted blue lines indicate the non-significant DDI range (0.5–2.0). AUC ratio values are represented on the y-axis as mean ± SD (n.s. = Not significant; * = p-value < 0.01; ** = p-value < 0.005).

Figure 4

PBPK−predicted PK profiles of victim drugs (A) omeprazole, (B) tolbutamide, (C) glyburide, and (D) repaglinide, administered with/without rifampicin, overlayed with the clinically observed PK profiles obtained from DIDB^TM.