Abstract

Early detection of drug-drug interactions (DDIs) can facilitate timely drug development decisions, prevent unnecessary restrictions on patient enrollment, resulting in clinical study populations that are not representative of the indicated study population, and allow for appropriate dose adjustments to ensure safety in clinical trials. All of these factors contribute to a streamlined drug approval process and enhanced patient safety. Here we describe a new approach for early prediction of the magnitude of change in exposure for cytochrome P450 (P450) CYP3A4-related DDIs of small-molecule anticancer drugs based on the model-based extrapolation of human-CYP3A4-transgenic mice pharmacokinetics to humans. Victim drugs brigatinib and lorlatinib were evaluated with the new approach in combination with the perpetrator drugs itraconazole and rifampicin. Predictions of the magnitude of change in exposure deviated at most 0.99- to 1.31-fold from clinical trial results for inhibition with itraconazole, whereas exposure predictions for the induction with rifampicin were less accurate, with deviations of 0.22- to 0.48-fold. Results for the early prediction of DDIs and their clinical impact appear promising for CYP3A4 inhibition, but validation with more victim and perpetrator drugs is essential to evaluate the performance of the new method.

SIGNIFICANCE STATEMENT The described method offers an alternative for the early detection and assessment of potential clinical impact of CYP3A4-related drug-drug interactions. The model was able to adequately describe the inhibition of CYP3A4 metabolism and the subsequent magnitude of change in exposure. However, it was unable to accurately predict the magnitude of change in exposure of victim drugs in combination with an inducer.

Introduction

Discovery and mapping of drug-drug interactions (DDIs) as early as practically possible to assure safety during clinical trials can pay off, as DDIs can have a profound negative impact on the benefit-risk balance of a drug. The investigation of DDIs of a novel compound with cytochrome P450 (P450) enzymes can be carried out in an in vitro setting, enabling early determination during drug development (https://www.fda.gov/media/134582/download). However, it is challenging to extrapolate these findings to in vivo conditions, especially in a quantitative manner, which is crucial for predicting the clinical impact of such interactions.

Clinical DDI studies are conducted before drug registration during phases II and III of clinical development. According to European Medicines Agency (EMA) and US Food and Drug Administration (FDA) guidelines, DDI studies have to be performed using strong inhibitors and inducers to follow a worst-case approach (https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-investigation-drug-interactions-revision-1_en.pdf; https://www.fda.gov/media/134581/download). Subsequently, physiologically based pharmacokinetic (PBPK) predictions are used to predict in silico the effects of moderate and weak inhibitors or inducers (perpetrator drugs) (https://www.fda.gov/media/101469/download). The performance of these clinical trials and subsequent PBPK model development and verification is time consuming. As a result, there is often a lack of clinical data on weak and moderate perpetrator drugs, which are more frequently used in the clinic than strong perpetrator drugs. PBPK simulations can help overcome this problem, but data may not always be available immediately after drug approval (Molenaar-Kuijsten et al., 2021).

Previous research described the development of an extrapolated population pharmacokinetic (PK) model for small-molecule anticancer drugs in human-CYP3A4-transgenic mice (a transgenic mouse model with its endogenous Cyp3a gene knocked out and replaced with the human CYP3A4 gene) that adequately predicted the human PK for brigatinib and lorlatinib (Damoiseaux et al., 2022). Predictability of the model for human PK can be validated, and misspecifications in interspecies extrapolation can be resolved using phase I trial results. Our hypothesis is that a human-CYP3A4-transgenic mouse can represent the situation of normal CYP3A4 enzyme activity and that a Cyp3a knockout mouse represents a situation in which CYP3A4 is fully inhibited. Subsequently, the inhibition or induction of CYP3A4 can be predicted as a fold difference between both mouse species. The fraction of CYP3A4 inhibition can be defined by the initial velocity equation appropriate for the mechanism of inhibition (competitive, uncompetitive, noncompetitive, or mixed). For induction, the rise in CYP3A4 activity from baseline by the perpetrator drug can be modeled using an enzyme turnover model (Yamashita et al., 2013). Our aim was to develop a method for early prediction of the magnitude of change in exposure for weak, moderate, and strong CYP3A4-related DDIs of small-molecule anticancer drugs as victim drugs using human-CYP3A4-transgenic mice.

MethodsData

A selection of two victim drugs was made from previously developed human-CYP3A4-transgenic mouse models that were extrapolated to humans, lorlatinib and brigatinib, for which the human PK was adequately predicted and open-label crossover DDI studies were available (Damoiseaux et al., 2022). The model structures of the previously developed mouse models were a two-compartment model with dual first-order absorption for lorlatinib and a two-compartment model for brigatinib. These extrapolated mouse models represented normal human CYP3A4 activity without a DDI. The dataset that was used for the development of the mouse population PK models also contained PK data from Cyp3a knockout mice. The difference between Cyp3a knockout mouse PK and human-CYP3A4-transgenic mouse PK was described by covariates on clearance and bioavailability. The PK parameters derived from the Cyp3a knockout mouse were assumed to represent a fully inhibited CYP3A4 enzyme activity. Furthermore, population PK models from literature were used to simulate the PK of itraconazole, rifampicin, and the rifampicin induction dynamics of CYP3A4 (Hennig et al., 2007; Wilkins et al., 2008; Yamashita et al., 2013). Clinical trials in healthy participants investigating DDI of lorlatinib and brigatinib were used to evaluate the concordance of preclinical predictions and clinical observations of the fold change in exposure due to the DDI with the perpetrator drugs itraconazole and rifampicin (Chen et al., 2020; Patel et al., 2020; Tugnait et al., 2020).

Modeling of Inhibition

The DDIs of lorlatinib and brigatinib with itraconazole were modeled as a competitive inhibition, and the workflow for the predictions is summarized in Fig. 1 (model code can be found in Supplemental Model code DDI inhibition). The difference in clearance and bioavailability between the extrapolated Cyp3a knockout and human-CYP3A4-transgenic mouse was considered CYP3A4-related metabolism of the drug from 0% to 100%. This difference can be interpreted as the isolated CYP3A4-related effect on clearance and bioavailability. The isolated effect was then scaled according to the degree and rate of inhibition and reintegrated in the model.

Fig. 1.

Flowchart of the method for early-phase prediction of the magnitude of change in exposure for weak, moderate, and strong CYP3A4-related DDIs of small-molecule anticancer drugs. The extrapolation of mice population PK models to humans (within the dashed area) was described previously (Damoiseaux et al., 2022). Ki, dissociation constant.

To model the effect of CYP3A4 inhibition on clearance, the elimination rate constant for CYP3A4-related clearance (ke,CYP3A4) was calculated as follows: where ke,human-CYP3A4-transgenic represents the elimination rate constant of the extrapolated human-CYP3A4-transgenic mouse and ke,Cyp3a knockout represents the elimination rate constant of the extrapolated Cyp3a knockout mouse. The previously described extrapolations of both mouse models to human was performed by means of allometric scaling of clearance and volume of distribution with an exponent of 0.75 and 1, respectively (Damoiseaux et al., 2022).

The relative enzyme activity (REA) over time was calculated using the initial velocity equation for competitive inhibition: where Vmax is set to 1 to represent the relative maximal velocity of uninhibited CYP3A4-related clearance, Km is the Michaelis-Menten constant of the substrate [S] for CYP3A4, and Ki the dissociation constant of the inhibitor [I]. The Ki of itraconazole is 0.0157 μM (Isoherranen et al., 2004), and the Km values of lorlatinib and brigatinib are 1.52 μM and not available, respectively (https://www.accessdata.fda.gov/drugsatfda_docs/nda/2018/210868Orig1s000MultidisciplineR.pdf). We evaluated Km values of 0.1, 1, and 10 μM for brigatinib because brigatinib has a similar molecular structure as lorlatinib and therefore the Km is probably in the same order of magnitude.

Subsequently, the elimination rate constant of lorlatinib and brigatinib in the presence of an inhibitor could be calculated as follows:

To describe the inhibitory effect on bioavailability, the reduction of bioavailability attributed to CYP3A4 metabolism was (FCYP3A4) was determined first: where FCyp3a knockout is the bioavailability of the extrapolated Cyp3a knockout mouse and Fhuman-CYP3A4-transgenic represents the bioavailability of the extrapolated human-CYP3A4-transgenic mouse.

The effect of CYP3A4 inhibition on bioavailability (F) was also described using the initial velocity equation for competitive inhibition (eq. 2):

Modeling of Induction

The effect of the inducer rifampicin on the PK of lorlatinib and brigatinib was modeled using a rifampicin-induced CYP3A4 turnover model (model code can be found in Supplemental Model code DDI induction) (Yamashita et al., 2013). The model was developed for the simulation of DDI once the blood concentration of the inducer reaches a steady state after repeated dosing. The prediction of the induction by this model, fold change induction (CYP3A4fold induction), was subsequently used as REA for clearance.

As for the modeling of inhibition, the difference in clearance between the extrapolated Cyp3a knockout and human-CYP3A4-transgenic mouse was considered the isolated CYP3A4-related metabolism of the drug from 0% to 100%, for which the workflow is summarized in Fig. 1. To model the effect of CYP3A4 induction, the murine CYP3A4-related hepatic clearance (CLH,CYP3A4,mouse) was calculated: with the CLH,CYP3A4,mouse and the hepatic blood flow in mice (QH,mouse) and humans (QH,human), the intrinsic clearance in humans (CLint,human), the human CYP3A4 extraction ratio (ERCYP3A4,human), and the human CYP3A4-related hepatic clearance (CLH,CYP3A4,human) could be calculated following the well-stirred liver model equations (eq. 7–10). Hepatic blood flow in mice (QH,mouse) and humans (QH,human) are approximately 0.12 and 90 l/h, respectively (Lautt, 2009; Xie et al., 2014):

First, the extraction ratio in mice (ERmouse) for CYP3A4-related clearance was calculated using eq. 7. With the ERmouse, CLH,CYP3A4,human and allometric scaling the intrinsic clearance in humans was calculated (eq. 8), assuming a mouse and human body weight (BWT) of 0.03 and 70 kg, respectively. The fraction unbound (fu) was assumed to be similar in mouse and human. Lastly, the elimination rate constant of lorlatinib and brigatinib in the presence of an inducer could be calculated: where ke,Cyp3a knockout represents the elimination rate constant of the Cyp3a knockout mouse extrapolated to humans and Vhuman represents the human volume of distribution.

As for the modeling of inhibition, the difference in bioavailability between the extrapolated Cyp3a knockout and human-CYP3A4-transgenic mouse was considered the isolated CYP3A4-related metabolism of the drug from 0% to 100%. Therefore, FCYP3A4 from eq. 4 applies here as well. Intestinal CYP3A4 metabolism was considered to have a significant contribution to bioavailability of brigatinib and lorlatinib. Therefore, the CYP3A4-mediated effect on bioavailability was split into two parts, metabolism due to intestinal CYP3A4 (FCYP3A4,intestines) and the first-pass effect of liver CYP3A4. The first-pass effect of the liver was described using a one-time extraction of the ERCYP3A4,human from the total amount of drug that reaches the portal vein. The fraction of the bioavailability originating from intestinal CYP3A4 in a noninduced state (FCYP3A4,intestines  base) was calculated with ERmouse (eq. 7), Fhuman-CYP3A4-transgenic and FCyp3a knockout (eq. 4):

Although twice the amount of enzymes would be capable of metabolizing twice the amount of drug within the same time, the metabolic rate in the intestines is also dependent on the drug’s supply to and residence time at the enzyme site. To approximate this process without making the model unnecessarily complex, we assumed a natural logarithmic relationship between the amount of CYP3A4 enzymes in the intestines and FCYP3A4,intestines: where the REA represents the predictions (fold change induction) of the induced CYP3A4 turnover model. The induced CYP3A4 turnover model was developed based on CYP3A4 mRNA expression and enzyme activity in human hepatocytes and therefore might not be representative for CYP3A4 induction in enterocytes (Yamashita et al., 2013). Due to the possibility that the induction in the intestines was not accurately simulated in our model, we chose to simulate both situations with and without induction of the intestines to get a better understanding of the contribution of the intestinal induction in our model.

Subsequently, the total effect of rifampicin-induced CYP3A4 on bioavailability was modeled by combining the intestinal and first-pass liver metabolized fractions from eqs. 9 and 13:

Simulation of DDIs

In total, 100 simulations with 10 individuals were performed for each interaction, with a randomly sampled weight corresponding to the characteristics of the clinical trial in healthy participants and without random effects (Chen et al., 2020; Patel et al., 2020; Tugnait et al., 2020). Doses and dosing schedules similar to those in clinical trials were administered in the simulation for brigatinib, lorlatinib, itraconazole, and rifampicin. Itraconazole 200 mg was administered once daily for 10 days starting 4 days (96 hours) before a single dose of lorlatinib 100 mg or brigatinib 90 mg was administered. Rifampicin 600 mg was administered once daily for 11 days starting 7 (168 hours) or 8 days (192 hours) before a single dose of brigatinib 180 mg or lorlatinib 100 mg was administered, respectively. Additionally, simulations were performed without the DDIs for brigatinib and lorlatinib in each situation.

Comparison of Predictions to Clinical Trials

The area under the plasma concentration-time curve from 0 until infinity (AUC0–inf) was calculated by integration of the individually predicted concentration over time. Median AUC0–inf, maximum concentration (Cmax), and corresponding time (tmax) were calculated from the simulation output. Subsequently, the changes in AUC0–inf, Cmax, and tmax between the simulation of the DDI and monotherapy were calculated. These results were compared with clinical trial results, and the order of magnitude between both was determined. A 0.5- to 2-fold difference in the order of magnitude was considered a reasonable prediction of the DDI. To put into perspective, most predictions of competitive inhibition with PBPK models are within these margins (Kilford et al., 2022). A published framework for qualification of the Simcyp Simulator (PBPK) with respect to competitive and mechanism-based inhibition of multiple P450 enzymes evaluated their model for CYP3A4 with 114 competitive inhibitors (Kilford et al., 2022). Comparison with clinical studies gave an average bias of 0.93 and 0.95 for Cmax and AUC ratios, respectively. The predictions were within 1.5-fold of observed Cmax and AUC ratios for 92% and 81%, respectively. Only four and two of the predictions were outside 2-fold of observed Cmax and AUC ratios, respectively.

Software

Nonlinear mixed-effects modeling was performed using NONMEM (version 7.5; ICON Development Solutions, Ellicott City, MD) and Perl-speaks-NONMEM (PsN, version 5.4.0) (Beal et al., 1988). Pirana (version 2.9.9) was used as the graphical user interface for NONMEM, and R (version 4.2.1) was used for processing the data and graphical and statistical diagnostics (Keizer et al., 2013).

Results

The proportion of clearance and bioavailability attributed to CYP3A4 metabolism is presented in Table 1. In case of brigatinib, CYP3A4 was responsible for 51% of the total clearance and a 29% reduction in bioavailability, whereas CYP3A4 was responsible for a lower proportion of the total clearance and bioavailability for lorlatinib of 25% and 13%, respectively.

TABLE 1

Clearance and bioavailability of Cyp3a knockout and human-CYP3A4-transgenic mouse for brigatinib and lorlatinib extrapolated to a 70-kg human

Concentrations over time for perpetrator drugs itraconazole and rifampicin were simulated using models from literature, and curves are presented in Supplemental Figs. 1 and 2 (Hennig et al., 2007; Wilkins et al., 2008). At steady state, the concentrations for itraconazole and rifampicin were approximately 0.2 and 2 mg/l, respectively. The CYP3A4 fold induction that resulted from the steady-state rifampicin exposure was also simulated using a model from literature and showed similar CYP3A4 induction of approximately 8-fold at steady state (Yamashita et al., 2013) (Supplemental Fig. 2).

Concentration-time curves of brigatinib and lorlatinib monotherapy and DDI predictions with itraconazole or rifampicin are shown in Fig. 2, as well as the results of clinical trials for comparison. Predictions were most accurate for the interaction with itraconazole for both brigatinib and lorlatinib (Fig. 2; Table 2). Predictions for the DDI between brigatinib and itraconazole deviated 1.07- to 1.31-fold from the results found in clinical trials for AUC0–inf, Cmax, and tmax based on the Km values 0.1, 1, and 10 μM. Different Km values within this range had minimal influence on AUC0–inf, Cmax, and tmax, but a Km of 10 μM resulted in the most accurate predictions (Table 2). The predictions of the DDI between lorlatinib and itraconazole deviated only 1.11- and 0.99-fold compared with observations in clinical trials for AUC0–inf and Cmax, respectively.

Fig. 2.

Concentration-time curves of brigatinib and lorlatinib monotherapy and DDI predictions with itraconazole or rifampicin compared with results from clinical trials (Chen et al., 2020; Patel et al., 2020; Tugnait et al., 2020). For the DDI predictions involving rifampin, the senario without intestinal induction was also evaluated (referred to in the figure as liver only). A Km of 1 μM for CYP3A4 was used for brigatinib in the prediction of the DDI with itraconazole since the actual Km was not available.

TABLE 2

Model predictions of AUC0–inf, Cmax, and tmax for brigatinib and lorlatinib monotherapy and DDI with itraconazole or rifampicin compared with results from clinical trials (Chen et al., 2020; Patel et al., 2020; Tugnait et al., 2020)

Induction-related DDIs with rifampicin were less accurate, as both brigatinib and lorlatinib exposure were underpredicted and did not meet the 0.5- to 2-fold criteria (Fig. 2; Table 2). Predictions for the DDI between brigatinib and rifampicin resulted in deviations from clinical trials of 0.22-, 0.38-, and 0.83-fold for AUC0–inf, Cmax, and tmax, respectively (Table 2). Without intestinal induction, the predictions for the DDI between brigatinib and rifampicin resulted in deviations from clinical trials of 0.93-, 1.27-, and 0.83-fold for AUC0–inf, Cmax, and tmax, respectively. For lorlatinib, the predictions of the interactions deviated 0.23- and 0.48-fold from clinical trials for AUC0–inf and Cmax, respectively. Without intestinal induction, the predictions for the DDI between lorlatinib and rifampicin resulted in deviations from clinical trials of 2.22- and 3.14-fold for AUC0–inf and Cmax, respectively.

Discussion

The magnitude of change in exposure resulting from CYP3A4 inhibition by itraconazole was adequately predicted for lorlatinib and brigatinib with an extrapolated population PK model of human-CYP3A4-transgenic mice. In contrast, the change in lorlatinib and brigatinib exposure after induction of CYP3A4 could not be accurately predicted and was underpredicted for both compounds. The reason for better prediction of inhibition is probably because inhibition of enzymes involves a limited number of variables compared with induction, making it easier to predict. Induction involves a larger set of variables, including regulation by the pregnane X receptor (PXR) and the constitutive androstane receptor (CAR) that can interact in more complex ways, potentially leading to greater unpredictability.

In our model, we used a previously published rifampicin-induced CYP3A4 turnover model that was based on CYP3A4 mRNA expression and enzyme activity in human hepatocytes (Yamashita et al., 2013). The induction of intestinal CYP3A4 was also based on this turnover model because no similar turnover model for intestinal CYP3A4 was available. Although the fundamental concept of enzyme induction is similar in both the liver and intestines, the expression of proteins involved in induction and therefore the level of induction might vary between these locations. To get a better understanding of the contribution of the intestinal induction in our model, we chose to predict both situations with induction of both liver and intestinal CYP3A4 and only liver CYP3A4. Based on the results, the induction of intestinal CYP3A4 seems to have different effects for both drugs. In case of brigatinib, predictions appear to be most accurate without the induction of intestinal CYP3A4. For lorlatinib, the results suggest that some extent of induction of intestinal CYP3A4 is required for adequate predictions. The optimal induction of intestinal CYP3A4 appears to be in between the predictions with and without intestinal CYP3A4. Nevertheless, the predicted induction of intestinal CYP3A4 was overpredicted in our model for both drugs.

Furthermore, our assumption of a natural logarithmic relationship between the amount of CYP3A4 enzymes in the intestines and the gastrointestinal metabolism might not adequately represent the complexities of the physiological processes. This relationship has significant impact on the predictions of the bioavailability, possibly contributing to the overprediction of intestinal CYP3A4 induction. Adequate predictions of this relationship might require a more detailed model of the involved physiological processes. CYP3A4 expression levels vary throughout the intestines (Takayama et al., 2021). Brigatinib and lorlatinib are likely absorbed at different locations in the intestines because of their Biopharmaceutics Classification System classes of 1 and 4, respectively. This might clarify that both drugs undergo varying degrees of exposure to intestinal CYP3A4, leading to differences in the impact of inducing intestinal CYP3A4. Furthermore, the transit time of the drug through the intestinal wall (enterocytes) is probably important since the drug would be available to intestinal CYP3A4 for a longer period of time.

The most common way to evaluate the magnitude of change in exposure due to DDIs is through dedicated clinical trials or PBPK modeling based on clinical trial data or even after drug approval (https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-investigation-drug-interactions-revision-1_en.pdf). PBPK models require substantial quantities of drug-specific data acquired from diverse in vitro and in vivo models. Subsequently, these models are further improved with clinical data and system-specific parameters. Although PBPK modeling probably results in more accurate estimates of the magnitude of change in exposure resulting from a DDI, our approach can be applied in early drug development, and, except for the required mice experiments (likely already partly available from preclinical research), the analysis is relatively easy to perform. Early detection of relevant DDIs can contribute to timely drug development decisions, prevention of unnecessary restrictions on patient enrollment, resulting in clinical study populations that are not representative of the indicated study population, and dose adjustments to ensure safety in clinical trials. Furthermore, PBPK models rely on accurate and comprehensive experimental data to make in vivo DDI predictions. The models are a simplified representation of the complex physiological processes involved in DDIs. The lack of mechanistic detail in the model carries the risk of omitting crucial physiological processes during the model development, making in vivo validation essential. In contrast, the use of extrapolated mouse models takes into account unknown physiological aspects as long as they are shared with the human physiology. Nevertheless, knocking out and humanization of mouse genes are also known to cause compensatory changes in expression of other enzymes (Kumar et al., 2017).

A limitation of this study is that it does not account for variability in CYP3A4 genotypes in the human population. The interindividual variability in CYP3A4 activity has been estimated to be attributed to genetic influences ranging from 66% to 88% (Klein and Zanger, 2013). Also, compensatory changes in Cyp3a knockout mouse models are known to upregulate other P450 enzymes such as Cyp3a13 (the only Cyp3a member not deleted) and several enzymes of the mouse Cyp2 family (Kumar et al., 2017). Consequently, Cyp3a knockout mouse models might not optimally represent a fully inhibited CYP3A4 enzyme activity for drugs that are also a substrate for one of these upregulated P450 enzymes. Furthermore, brigatinib is a known autoinductor of CYP3A4, as might be the case for other victim drugs (https://drug-interactions.medicine.iu.edu). Autoinhibition and autoinduction of target proteins such as the PXR and the CAR regulates CYP3A4 expression. These proteins differ significantly between mice and humans, which could result in misspecifications for the prediction of steady-state concentrations with and without DDIs (Hasegawa et al., 2011). PK information used in this study of both clinical and preclinical trials was collected after a single dose of the victim drug, and effects of autoinhibition or autoinduction are therefore unlikely. However, applications of the model in case of multiple doses should be handled with caution. Another essential point is the small number of victim and perpetrator drugs that were used in this study. It remains to be elucidated whether this approach will result in equal performance for different victim and perpetrator drugs. In relation to this, the choice of compounds that were included could be interpreted as “cherry picking,” as only two of the four previously developed mouse models were selected. The misspecification in the clearance of the two not-included extrapolated mouse models was an indication that the DDI predictions with these two models were unlikely to be accurate and would not have contributed to the proof of concept. Ideally, we would have used more sensitive probes (e.g., midazolam) to evaluate our concept, but our available data were limited. Additional evaluation of this concept with such probes will be essential in future research.

An important feature of the described method is that it can also be applied to other victim drugs and, just as important, other perpetrator drugs, either weak, moderate, or strong. Extrapolation to other victim drugs will mainly be dependent on the extent to which the extrapolated human-CYP3A4-transgenic mouse is predictable for the human PK. Direct extrapolation by means of allometric scaling alone might not be sufficient for all drugs (Damoiseaux et al., 2022). However, availability of clinical PK information from phase I trials might facilitate optimization of the extrapolated mouse models to account for physiological differences that contribute to mouse and human PK differences. The extrapolation to other perpetrator drugs with an inhibitory effect on CYP3A4 can be achieved with a population PK model and an inhibitory constant of the perpetrator drug, which are often available for most well known perpetrator drugs. In the methods, we give a description of a competitive inhibition, but this method can be extended to uncompetitive, noncompetitive, or mixed inhibition by using the corresponding initial velocity equation.

To conclude, the described method offers an alternative for the early detection and assessment of potential clinical impact of CYP3A4-related DDIs. The model was able to adequately describe the inhibition of CYP3A4 metabolism and the subsequent magnitude of change in exposure. However, it was unable to accurately predict the magnitude of change in exposure of victim drugs in combination with an inducer. Additionally, the method was solely evaluated for two victim drugs and two perpetrator drugs, leaving the performance for other victim and perpetrator drugs yet to be determined.

Data Availability

The datasets generated during and/or analyzed during the present study are available from the corresponding author on reasonable request.

Authorship Contributions

Participated in research design: Damoiseaux, Beijnen, Huitema, Dorlo.

Conducted experiments: Damoiseaux.

Performed data analysis: Damoiseaux.

Wrote or contributed to the writing of the manuscript: Damoiseaux, Beijnen, Huitema, Dorlo.

FootnotesReceived September 7, 2023.Accepted July 30, 2024.

This work received no external funding.

All authors declare that they have no conflicts of interest with the contents of this article.

dx.doi.org/10.1124/dmd.123.001530.

↵This article has supplemental material available at dmd.aspetjournals.org.

AbbreviationsAUC0–infarea under the plasma concentration-time curve from 0 until infinityCLclearanceDDIdrug-drug interactionERextraction ratioKmMichaelis-Menten constantP450cytochrome P450PBPKphysiologically based pharmacokineticPKpharmacokineticREArelative enzyme activitytmaxtime of maximum plasma concentrationCopyright © 2024 by The Author(s)