Advertisement
Research ArticleArticle

Prediction of Drug-Drug Interactions with Crizotinib as the CYP3A Substrate Using a Physiologically Based Pharmacokinetic Model

Shinji Yamazaki, Theodore R. Johnson and Bill J. Smith
Drug Metabolism and Disposition October 2015, 43 (10) 1417-1429; DOI: https://doi.org/10.1124/dmd.115.064618

Abstract

An orally available multiple tyrosine kinase inhibitor, crizotinib (Xalkori), is a CYP3A substrate, moderate time-dependent inhibitor, and weak inducer. The main objectives of the present study were to: 1) develop and refine a physiologically based pharmacokinetic (PBPK) model of crizotinib on the basis of clinical single- and multiple-dose results, 2) verify the crizotinib PBPK model from crizotinib single-dose drug-drug interaction (DDI) results with multiple-dose coadministration of ketoconazole or rifampin, and 3) apply the crizotinib PBPK model to predict crizotinib multiple-dose DDI outcomes. We also focused on gaining insights into the underlying mechanisms mediating crizotinib DDIs using a dynamic PBPK model, the Simcyp population-based simulator. First, PBPK model–predicted crizotinib exposures adequately matched clinically observed results in the single- and multiple-dose studies. Second, the model-predicted crizotinib exposures sufficiently matched clinically observed results in the crizotinib single-dose DDI studies with ketoconazole or rifampin, resulting in the reasonably predicted fold-increases in crizotinib exposures. Finally, the predicted fold-increases in crizotinib exposures in the multiple-dose DDI studies were roughly comparable to those in the single-dose DDI studies, suggesting that the effects of crizotinib CYP3A time-dependent inhibition (net inhibition) on the multiple-dose DDI outcomes would be negligible. Therefore, crizotinib dose-adjustment in the multiple-dose DDI studies could be made on the basis of currently available single-dose results. Overall, we believe that the crizotinib PBPK model developed, refined, and verified in the present study would adequately predict crizotinib oral exposures in other clinical studies, such as DDIs with weak/moderate CYP3A inhibitors/inducers and drug-disease interactions in patients with hepatic or renal impairment.

Introduction

A clinically relevant drug-drug interaction (DDI) is generally considered a modification of pharmacological and/or toxicological responses of one substrate drug by another interacting drug. An evaluation of potential DDIs for new molecular entities (NMEs) are recognized as an important consideration in the drug discovery and development setting as well as the regulatory review process (Zhang et al., 2009; Rowland et al., 2011; Prueksaritanont et al., 2013). The US Food and Drug Administration (FDA) and the European Medicines Agency have recently issued DDI guidances (CDER, 2012; CHMP, 2012), that emphasize the use of an integrated mechanistic approach, such as a physiologically based pharmacokinetic (PBPK) model, to quantitatively predict the magnitude of DDIs in the clinic. The dynamic modeling approach is being employed increasingly in all phases of drug discovery and development to evaluate potential DDI risks for NMEs (Boulenc and Barberan, 2011; Zhao et al., 2011; Huang and Rowland, 2012; Peters et al., 2012; Huang et al., 2013). Additionally, regulatory agencies express keen interest in the use of mechanistic dynamic models to provide a deeper understanding of complex DDIs, including simultaneous effects of two or more interacting drugs (e.g., inhibitors and inducers) on exposures of substrate drugs, as well as drug-disease interactions in patients with hepatic or renal impairment (Zhao et al., 2011; Huang and Rowland, 2012; Huang et al., 2013; Varma et al., 2015). Accordingly, the mechanistic dynamic modeling approach can provide an alternative to evaluate complex DDIs, as the FDA encourages study sponsors to use modeling and simulation to determine the best dosing strategy (Huang and Lesko, 2009; Rowland et al., 2011; Huang et al., 2013).

Crizotinib (PF02341066; Xalkori) is an orally available small-molecule inhibitor of multiple tyrosine kinases, including anaplastic lymphoma kinase and mesenchymal-epithelial transition factor (CDER, 2011). Crizotinib has been reported as a CYP3A substrate, and a time-dependent inhibitor and inducer (Mao et al., 2013; Johnson et al., 2015). Accordingly, a clinical net effect of crizotinib (as the interacting drug) on oral exposures (e.g., area under plasma concentration-time curve, AUC) of the CYP3A probe substrate midazolam was evaluated in cancer patients who received a single oral dose of midazolam (2 mg) before and after 28-day multiple oral administration of crizotinib 250 mg twice daily (Tan et al., 2010). The fold-increase in AUC (AUCR) for midazolam was 3.7, suggesting crizotinib was a moderate CYP3A inhibitor. We previously reported a reasonable prediction of midazolam AUCR by a PBPK modeling approach. (Mao et al., 2013). In addition, crizotinib single-dose DDI studies (as the substrate drug) with a strong CYP3A inhibitor or inducer, i.e., ketoconazole (200 mg twice daily) or rifampin (600 mg once daily), were conducted in healthy volunteers (CDER, 2011; Xu et al., 2011a, b). In these DDI studies, crizotinib AUCR was 3.2 with ketoconazole, whereas it was 0.18 with rifampin. As mentioned above, crizotinib is mainly metabolized by CYP3A (as its own primary clearance), which is moderately inhibited and weakly induced by crizotinib itself, resulting in nonstationary pharmacokinetics during multiple-dose administration (CDER, 2011). Thus, a post-marketing requirement by FDA has been issued to conduct crizotinib “multiple-dose” DDI studies with strong CYP3A inhibitors and inducers to inform potential dose adjustment (CDER, 2011). Whereas this particular example may not be a rare complex DDI case, the number of factors to be considered in order to predict the outcome requires sophisticated mechanistic models as similar to complex DDIs. Furthermore, it is challenging to recruit patients in sufficient numbers for multiple-dose DDI studies of anticancer drugs. There are also ethical concerns about the possibility of sub- or supra-therapeutic exposures to crizotinib in patients with concomitant multiple-dose administration of a strong CYP3A inhibitor or inducer.

Given these challenges to conducting multiple-dose DDI studies, it would be highly beneficial to develop, refine, and verify a PBPK model of crizotinib on the basis of currently available data to quantitatively predict crizotinib exposures in multiple-dose DDI studies. The main objectives of the present study were: 1) to develop and refine crizotinib PBPK model on the basis of the clinical single- and 28-day multiple-dose pharmacokinetic results, 2) to verify crizotinib PBPK model on the basis of the results of crizotinib single-dose DDI studies with ketoconazole or rifampin, and 3) to apply the crizotinib PBPK model to predict crizotinib multiple-dose DDI outcomes. A commercially available Simcyp (Sheffield, UK) population-based dynamic simulator was used in the present study (Jamei et al., 2009). Furthermore, we investigated effects of crizotinib CYP3A net-inhibition on its oral exposures by comparing the prediction outcomes between the crizotinib PBPK models with and without the DDI parameters such as CYP3A time-dependent inhibition and induction. Consequently, we focused on gaining insights into the underlying mechanisms mediating crizotinib DDIs using the mechanistic dynamic modeling approach.

Materials and Methods

Clinical Pharmacokinetic Data.

Detailed information about crizotinib clinical studies such as a single-dose oral bioavailability study, single-dose DDI studies with ketoconazole or rifampin, and Phase I multiple-dose escalation studies were previously reported by Pfizer Oncology Business Unit (Pfizer Inc., San Diego, CA) (Tan et al., 2010; Xu et al., 2011a, b, 2015). Additional information about crizotinib pharmacokinetics is also available in the FDA website (CDER, 2011). Briefly, in the single-dose oral bioavailability study, a single dose of crizotinib was administered to 14 healthy male adult volunteers either intravenously (50 mg for 2 hours) or orally (250 mg) in a two-way crossover design with a ≥4-day washout period. Plasma concentrations of crizotinib in subjects were determined up to 7 days postdose. The multiple-dose phase I study of crizotinib at the twice daily dose of 250 mg was conducted in 9 cancer patients. Plasma concentrations of crizotinib were determined on day 7 (single dose, n = 9), day 15 (n = 5), and day 28 (n = 5). The crizotinib single-dose DDI studies with ketoconazole or rifampin were conducted in 15 healthy male adult volunteers in a fasted state in a two-way crossover design with a ≥14-day washout period. In the DDI study with ketoconazole, a single oral dose of crizotinib (150 mg) was administered to healthy volunteers either as crizotinib alone or crizotinib on day 4 with coadministration of ketoconazole (200 mg twice daily with 12-hour intervals) from days 1 to 16. The lower dose of crizotinib (150 mg) relative to the recommended dose (250 mg) was selected in this study because an increase in crizotinib oral exposures was expected with coadministration of ketoconazole. In the DDI study with rifampin, a single oral dose of crizotinib (250 mg) was administered to healthy volunteers either as crizotinib alone or crizotinib on day 9 with coadministration of rifampin (600 mg once daily) from days 1 to 14. Plasma concentrations of crizotinib in all subjects were determined up to 14 days postdose in the DDI study with ketoconazole and up to 7 days postdose in the DDI study with rifampin.

Crizotinib Input Parameters into Simcyp.

Physicochemical and pharmacokinetic parameters of crizotinib used for the present PBPK model are summarized in Table 1. Crizotinib hepatic microsomal intrinsic clearance (CLint,hep) was back-calculated from the clinically observed plasma clearance (CL) using a retrograde model implemented in Simcyp version 13.1 (Jamei et al., 2009). Crizotinib intravenous and oral plasma CL estimates (geometric mean) were 46.8 and 108 l/h, respectively, in the single-dose oral bioavailability study (Xu et al., 2015). In the crizotinib single-dose DDI studies with ketoconazole and rifampin, crizotinib oral plasma CL estimates in the control groups (crizotinib alone) were 123 and 119 l/h, respectively (Xu et al., 2011a, 2011b). Crizotinib renal CL was estimated as 2.25 l/h on the basis of urinary excretion (2.3% of the administered dose as parent drug) in a single oral-dose human mass-balance study with [14C]crizotinib (Johnson et al., 2015). The fraction of dose absorbed from gastrointestinal tract (Fa) was estimated at approximately 0.5 since crizotinib excretion into feces was 53% of the administered oral dose (250 mg) in the human mass-balance study (Johnson et al., 2015). By taking into account renal CL, Fa, and Fg (the fraction of dose that escapes intestinal first-pass metabolism), the back-calculated crizotinib CLint,hep values were 222 μl/min per milligram microsomal protein from the intravenous CL (46.8 l/h) and 129 μl/min per milligram microsomal protein from the oral CL (108 l/h) in the bioavailability study, and 146 μl/min per milligram microsomal protein from the mean oral Cl estimate (121 l/h) in the control groups of the DDI studies. A fraction metabolized by CYP3A4 (fm,CYP3A4) was estimated to be 0.8 on the basis of the in vitro cytochrome P450 phenotyping and the human mass-balance study (Johnson et al., 2011, 2015); therefore, 80% of the back-calculated total CLint,hep was assigned to CYP3A4-mediated CLint, whereas the remaining CLint,hep value was assigned to additional microsomal CLint,hep. Clinically observed crizotinib steady-state volume of distribution (Vss) was 1772 liters (25 l/kg) after a single intravenous 2-hour infusion; therefore, the predicted Vss of 7.5 l/kg with the tissue-composition–based model implemented in Simcyp (as the mathematical model 2) was adjusted to 25 l/kg using a Simcyp Kp scalar of 3.4 (Rodgers et al., 2005). Crizotinib time-dependent inactivation constant (KI) and maximum inactivation rate constant (kinact) for CYP3A4 were estimated as 0.89 μM and 0.78 hours−1, respectively, in cryopreserved human hepatocytes suspended in human plasma (Mao et al., 2013). Crizotinib in vitro inhibitory effect on CYP3A as a reversible inhibitor was negligible (IC50 > 30 μM) in human liver microsomes. Crizotinib CYP3A4 induction parameters, Emax and EC50, were estimated from three individual cryopreserved hepatocytes, and then normalized to 2.4 and 0.84 μM, respectively, by the induction calibrator implemented in Simcyp, with rifampin as positive control (mean Emax of 90-fold and EC50 of 0.57 mM).

View this table:
TABLE 1

Physicochemical and pharmacokinetic parameters of crizotinib used for PBPK model

Simcyp Simulation.

Simcyp simulation of crizotinib plasma concentrations was performed by a full PBPK model with a first-order absorption rate constant (ka) set to predict clinically observed tmax (time to reach maximal plasma concentration, Cmax). In the Simcyp Qgut model (Yang et al., 2007), crizotinib blood flow term (Qgut) was predicted to be 4.0 l/h on the basis of crizotinib physicochemical properties, and its unbound fraction in the gut (fu,gut) was assumed to be equal to an unbound fraction in blood (fu,blood). From the Simcyp compound library of the interacting drugs, ketoconazole (sim-ketoconazole 200 mg bid), rifampin (sim-rifampicin), diltiazem (sim-diltiazem), erythromycin (sim-erythromycin), fluconazole (sv-fluconazole), and fluvoxamine (sv-fluvoxamine) were used for the predictions of crizotinib DDI studies. Their Simcyp default DDI parameters on CYP3A4 were as follows: ketoconazole competitive inhibition constant (Ki) = 0.015 μM (fu,mic = 0.97); rifampin induction Emax = 8, EC50 = 0.32 μM, and competitive Ki = 10.5 μM (fu,mic = 1); diltiazem competitive Ki = 36.1 μM, mechanism-based inhibition kinact = 0.702 hours−1, and KI = 4.75 μM (fu,mic = 1); erythromycin competitive Ki = 82 μM (fu,mic = 0.909), mechanism-based inhibition kinact = 2.25 hours−1, and KI = 23.2 μM (fu,mic = 1); fluconazole competitive Ki = 10.7 μM (fu,mic = 1); fluvoxamine Ki = 17.89 μM (fu,mic = 0.441). Rifampin Emax was set at 16 in the present study because it has been reported that a better correlation between Simcyp-predicted and observed midazolam AUCR was observed in several studies using rifampin Emax of 16 rather than 8 (Almond et al., 2012).

All clinical trial simulations in Simcyp were performed with a virtual population of healthy volunteers (fasted state) in six trials of 15 subjects (total 90 subjects), each aged 20–50 years with a female/male ratio of 0.5, whose CYP3A4 degradation rate constant was 0.0193 hours−1 in the liver and 0.030 hours−1 in the intestine. The output sampling interval in Simcyp simulation tool box was set to 0.2 hours in all simulations. Trial designs in the clinical studies and the corresponding simulations were as follows:

  • Trial #1 (Single-dose intravenous pharmacokinetic study): A single intravenous dose of crizotinib (50 mg) was administered for 2 hours; plasma concentrations of crizotinib were simulated up to 7 days.

  • Trial #2 (Single-dose oral pharmacokinetic study): A single oral dose of crizotinib (250 mg) was administered; plasma concentrations of crizotinib were simulated up to 7 days.

  • Trial #3 (Single-dose DDI study with ketoconazole): A single oral dose of crizotinib (150 mg) was administered on day 4 with and without multiple oral coadministration of ketoconazole 200 mg twice daily for 16 days; plasma concentrations of crizotinib and ketoconazole were simulated during the drug treatment period.

  • Trial #4 (Single-dose DDI study with rifampin): A single oral dose of crizotinib (250 mg) was administered on day 9 with multiple oral coadministration of rifampin 600 mg once daily for 14 days; plasma concentrations of crizotinib and rifampin were simulated during the drug treatment period.

  • Trial #5 (Multiple-dose pharmacokinetic study): An oral dose of crizotinib (250 mg) was administered twice daily for 28 days; plasma concentrations of crizotinib were simulated during the drug treatment period.

  • Trial #6 (Multiple-dose DDI study with ketoconazole): An oral dose of crizotinib (150 mg) was administered twice daily for 28 days with and without multiple oral coadministration of ketoconazole 200 mg twice daily for 28 days; plasma concentrations of crizotinib and ketoconazole were simulated during the drug treatment period.

  • Trial #7 (Multiple-dose DDI study with rifampin): An oral dose of crizotinib (250 mg) was administered twice daily for 28 days with and without multiple oral coadministration of rifampin 600 mg once daily for 28 days; plasma concentrations of crizotinib and rifampin were simulated during the drug treatment period.

  • Trial #8 (Multiple-dose DDI study with a weak or moderate CYP3A inhibitor): An oral dose of crizotinib (150 mg) was administered twice daily for 28 days with and without multiple oral coadministration of diltiazem 120 mg twice a day, erythromycin 500 mg three times a day, fluconazole 200 mg once a day, or fluvoxamine 50 mg once a day for 28 days; plasma concentrations of crizotinib, diltiazem, erythromycin, and fluconazole were simulated during the drug treatment period.

These trial designs are also summarized in Table 2. Two Simcyp compound files of crizotinib with (on) and without (off) crizotinib DDI parameters such as CYP3A4 time-dependent inhibition (KI and kinact) and induction (Emax and EC50) were used to investigate net effects of crizotinib DDI potential on the simulation outcomes #3–7; that is, the on/off ratios in the simulation results suggest net effects of crizotinib autoinhibition/induction potential on crizotinib pharmacokinetics and DDI outcomes. In the single-dose trials #1–4, the area under the plasma concentration-time curve from time zero to infinity (AUC0–∞) was calculated from simulated plasma concentrations using the linear trapezoidal rule:Embedded Image(1)where AUC0–L, CL, and λ represent the area under the plasma concentration-time curve from time zero to the last time point, the plasma concentration at the last time point, and the elimination rate constant in the terminal phase of log plasma concentration-time curves determined by linear regression, respectively.

View this table:
TABLE 2

Simulation outline of crizotinib single- and multiple-dose pharmacokinetic and drug-drug interaction studies

Apparent terminal half-life (t1/2,z) was calculated from 0.693/λ, whereas Cmax and tmax were obtained from Simcyp outputs. To understand crizotinib nonstationary pharmacokinetics, an index of nonstationary pharmacokinetics (NSI) was calculated from the area under the plasma concentration-time curve over the dosing interval (AUC0–τ) at steady state divided by AUC0–∞ in the corresponding single-dose studies. In the DDI studies, the fold-increases in Cmax (CmaxR) and AUCR (e.g., AUC0–∞ ratios in the single-dose studies and AUC0–τ ratios in the multiple-dose studies) were calculated from the ratios of the simulated values in treatment groups relative to control groups. In addition, fractions metabolized/excreted (fm) by each CL pathway at all simulation time points were individually calculated from simulated CLint value in each pathway divided by total CLint value in a Simcyp virtual population. Crizotinib Fg values at all simulation time points were also individually calculated by the Qgut model using the simulated Qgut, fu,gut, and intestinal CLint in a Simcyp virtual population. Mean and median fm and Fg values were then calculated along with percent coefficient of variation (CV%) and the 5th–95th percentiles per each treatment day. All calculations on Simcyp output files were performed with Microsoft Excel 2007 (Redmond, WA). Geometric means of pharmacokinetic parameters in six simulation trials were compared with the clinically observed geometric mean. A goodness-of-fit between the model-predicted and observed plasma concentrations was assessed by a linear regression analysis (r2 > 0.9) with SigmaPlot 11 (Systat Software, San Jose, CA), together with visual inspection, whereas that on the pharmacokinetic parameters such as Cmax and AUC was evaluated with the predicted-to-observed ratio of ≤ ±50%. Plots of goodness-of-fit for the model-predicted and observed plasma concentrations are summarized in Supplemental Fig. S1.

Results

Model Development and Verification

Prediction of Crizotinib Single-Dose Pharmacokinetics.

Following a single intravenous 2-hour infusion of crizotinib 50 mg to healthy subjects, the observed crizotinib Cmax, AUC0–∞, and t1/2,z (geometric mean) were 155 ng/ml, 1067 ng∙h/ml, and 39 hours, respectively (Table 3). As shown in Fig. 1A, crizotinib plasma concentrations were adequately predicted by the PBPK model (r2 = 0.983). Consequently, the predicted crizotinib Cmax (167 ng/ml), AUC0–∞ (1258 ng∙h/ml), and t1/2,z (45 hours) were within 20% of the observed results. Following a single oral administration of crizotinib 250 mg to healthy subjects, the observed crizotinib estimates for Cmax, AUC0–∞, and t1/2,z were 100 ng/ml, 2321 ng∙h/ml, and 29 hours, respectively (Table 3). Crizotinib oral plasma concentrations were also reasonably predicted by the PBPK model (r2 = 0.977) (Fig. 1B). The predicted crizotinib Cmax (103 ng/ml) and AUC0–∞ (2878 ng∙h/ml) were within 20% of the observed results, whereas t1/2,z (58 hours) was slightly overpredicted by 2-fold.

View this table:
TABLE 3

Clinically observed and PBPK model–predicted pharmacokinetic parameters of crizotinib in healthy subjects following a single intravenous or oral administration of crizotinib

Data are expressed as geometric mean with percent coefficient of variation (CV%) in parentheses (n = 14 for the observed; n = 15 per group × 6 groups for the predicted).

Fig. 1.
Fig. 1.

Observed and PBPK model–predicted plasma concentrations of crizotinib in healthy subjects after a single intravenous (A) or oral administration (B). Crizotinib was administered intravenously (50 mg, 2-hour infusion) or orally (250 mg) to 14 healthy subjects in the single-dose oral bioavailability study. Crizotinib plasma concentrations were predicted in a virtual population of healthy subjects (n = 15 per group × 6 groups, total 90 subjects) with a full-PBPK model implemented in the Simcyp simulator. The x-axis represents the time after dosing in hours and the y-axis represents the observed (○) and PBPK model–predicted (—) plasma concentrations in nanograms per milliliter on a logarithmic scale. The observed and predicted plasma concentrations are expressed as mean ± S.D. and mean (solid line) with 5th and 95th percentiles (dashed line), respectively.

Prediction of Crizotinib Pharmacokinetics in Crizotinib Single-Dose DDI Study with Ketoconazole.

In the crizotinib single-dose DDI study with ketoconazole, an oral dose of crizotinib 150 mg was administered to healthy volunteers either as crizotinib alone (control group) or crizotinib on day 4 with coadministration of ketoconazole 200 mg twice daily from days 1 to 16 (treatment group). Clinically observed crizotinib Cmax and AUC0–∞ (geometric mean) were, respectively, 66 ng/ml and 1260 ng∙h/ml in the control group and 94 ng/ml and 3986 ng∙h/ml in the treatment group, resulting in CmaxR of 1.4 and AUCR of 3.2 (Table 4). PBPK model–predicted crizotinib plasma concentrations reasonably matched the observed results in both the control and treatment groups (r2 = 0.946 and 962, respectively) (Fig. 2, A and B). The predicted crizotinib Cmax and AUC0–∞ in both the groups were within 20% of the observed results, whereas the predicted t1/2,z values were roughly within 2-fold of the observed values in both the groups (Table 4). In addition, the simulation was performed without crizotinib DDI parameters (off) to investigate crizotinib CYP3A net-inhibition effects on the DDI outcomes. The predicted crizotinib Cmax, AUC0–∞, and t1/2 were comparable between the simulations on and off (Table 4). The predicted crizotinib CmaxR (1.6) and AUCR (3.6) in the simulation off were also comparable to those (1.6 and 3.4, respectively) in the simulation on. Time-courses of the predicted crizotinib hepatic and intestinal CLint in the simulation on are graphically presented in Fig. 3, A and B, respectively. The hepatic and intestinal CLint values in the control group slightly decreased on day 4 when crizotinib was orally administered and thereafter returned to the baseline level by the end of study. In contrast, the hepatic and intestinal CLint values in the treatment group were significantly inhibited from the beginning of ketoconazole treatment, resulting in a little change in crizotinib CLint values on day 4. On the basis of the predicted time-courses of crizotinib median fm,CYP3A4 and Fg in a Simcyp virtual population of 90 subjects, ketoconazole-mediated CLint inhibition resulted in a decrease in fm,CYP3A4 of 0.80 to 0.19 with an increase in Fg of 0.94 to 1.0 on day 4 (data not shown). Overall, the Simcyp simulation reasonably predicted the effect of ketoconazole on the crizotinib exposures in the crizotinib single-dose DDI study.

View this table:
TABLE 4

Clinically observed and PBPK model–predicted pharmacokinetic parameters of crizotinib in healthy subjects following a single oral administration of crizotinib with and without multiple coadministration of ketoconazole 200 mg twice daily or rifampin 600 mg once daily

Data are expressed as geometric mean with percent coefficient of variation (CV%) in parentheses (n = 15 for the observed; n = 15 per group × 6 groups for the predicted).

Fig. 2.
Fig. 2.

Observed and PBPK model–predicted plasma concentrations of crizotinib in healthy subjects after a single oral administration of crizotinib with and without coadministration of ketoconazole (A and B) or rifampin (C and D). A single oral dose of crizotinib was administered to 15 healthy subjects either as crizotinib 150 mg alone (A) or crizotinib 150 mg on day 4 (B) of 16-day multiple coadministration of ketoconazole 200 mg twice daily or either as crizotinib 250 mg alone (C) or crizotinib 250 mg on day 9 (D) of 14-day multiple coadministration of rifampin 600 mg once daily. Crizotinib plasma concentrations were predicted in a virtual population of healthy subjects (n = 15 per group × 6 groups, total 90 subjects) with a full-PBPK model implemented in the Simcyp simulator. The x-axis represents the time after a single oral administration of crizotinib in hours and the y-axis represents the observed (○) and PBPK model–predicted (—) plasma concentrations of crizotinib in nanograms per milliliter on a logarithmic scale. The observed and predicted plasma concentrations are expressed as mean ± S.D. and mean (solid line) with 5th and 95th percentiles (dashed line), respectively.

Fig. 3.
Fig. 3.

PBPK model–predicted crizotinib intrinsic clearance in liver (A and C) and intestine (B and D) of healthy subjects following a single oral administration of crizotinib with and without 16-day multiple oral administration of ketoconazole (A and B) or following 28-day multiple oral administration of crizotinib with and without ketoconazole (C and D). A full–PBPK model simulation was performed in a Simcyp virtual population of healthy subjects (n = 15 per group × 6 groups, total 90 subjects) following a single oral administration of crizotinib (150 mg) alone or crizotinib (150 mg) on day 4 of 16-day multiple oral coadministration of ketoconazole (200 mg twice daily) and following 28-day multiple oral administration of crizotinib (150 mg twice daily) with and without coadministration of ketoconazole (200 mg twice daily). The x-axis represents the treatment period in days and the y-axis represents PBPK model–predicted crizotinib mean intrinsic clearance in liver (A and C) and intestines (C and D) in liters per hour.

Prediction of Crizotinib Pharmacokinetics in Crizotinib Single-Dose DDI Study with Rifampin.

In the crizotinib single-dose DDI study with rifampin, an oral dose of crizotinib 250 mg was administered to healthy volunteers either as crizotinib alone (control group) or crizotinib on day 9 with coadministration of rifampin 600 mg once daily from days 1 to 14 (treatment group). Clinically observed crizotinib Cmax and AUC0–∞ (geometric mean) were, respectively, 102 ng/ml and 2192 ng∙h/ml in the control group and 32 ng/ml and 397 ng∙h/ml in the treatment group, resulting in CmaxR of 0.31 and AUCR of 0.18 (Table 4). PBPK model–predicted crizotinib plasma concentrations reasonably matched the observed results in both the control and treatment groups (r2 = 0.988 and 0.911, respectively) (Fig. 2, C and D). The predicted crizotinib Cmax and AUC0–∞ in both the groups were within 10% of the observed result, whereas the predicted t1/2,z values were within 2-fold of the observed values in the control and treatment groups (Table 4). When the DDI simulation was performed without crizotinib DDI parameters (off), the predicted crizotinib Cmax, AUC0–∞, and t1/2 were comparable to the predicted values in the simulation on. The predicted crizotinib CmaxR (0.30) and AUCR (0.15) in the simulation off were also comparable to those (0.31 and 0.15, respectively) in the simulation on (Table 4). Time-courses of the predicted crizotinib hepatic and intestinal CLint in the simulation on are presented in Fig. 4, A and B, respectively. The hepatic and intestinal CLint values in the treatment group significantly increased from the beginning of rifampin treatment and then reached steady state around day 6. Thereafter, the CLint values slightly decreased on day 9 when crizotinib was orally administered, and then returned to steady state level by the end of study. The predicted time-courses of crizotinib median fm,CYP3A4 and Fg in a Simcyp virtual population showed an increase in fm,CYP3A4 of 0.80 to 0.96 with a decrease in Fg of 0.94 to 0.73 on day 9 (data not shown). Overall, the Simcyp simulation sufficiently predicted the effect of rifampin on the crizotinib exposures in the crizotinib single-dose DDI study.

Fig. 4.
Fig. 4.

PBPK model–predicted crizotinib intrinsic clearance in liver (A and C) and intestine (B and D) of healthy subjects following a single oral administration of crizotinib with and without 14-day multiple oral administration of rifampin (A and B) or following 28-day multiple oral administration of crizotinib with and without rifampin (C and D). A full–PBPK model simulation was performed in a Simcyp virtual population of healthy subjects (n = 15 per group × 6 groups, total 90 subjects) following a single oral administration of crizotinib (250 mg) alone or crizotinib (250 mg) on day 9 of 14-day multiple oral coadministration of rifampin (600 mg once daily) and following 28-day multiple oral administration of crizotinib (250 mg twice daily) with and without coadministration of rifampin (600 mg once daily). The x-axis represents the treatment period in days and the y-axis represents PBPK model–predicted crizotinib mean intrinsic clearance in liver and intestines in liters per hour.

Model Verification and Refinement

Prediction of Multiple-Dose Crizotinib Pharmacokinetics.

Following 28-day multiple oral administration of crizotinib 250 mg twice daily to cancer patients, the observed crizotinib plasma concentration-time profiles were relatively flat during the dosing interval of 12 hours as presented in Fig. 5. Clinically observed crizotinib Cmax and AUC0–τ (geometric mean) on day 28 were 328 ng/ml and 3054 ng∙h/ml, respectively, with NSI of 1.3 (Table 5). The PBPK-model overpredicted crizotinib plasma concentrations (Fig. 5A), resulting in predicted crizotinib Cmax (515 ng/ml) and AUC0–τ (6165 ng∙h/ml) that were, respectively, 1.6- and 2.0-fold higher than the observed results. The predicted NSI (2.1) was also overpredicted by 1.6-fold. Thus, the predicted crizotinib oral exposures did not meet our criteria of goodness-of-model prediction (≤50%), leading us to refine the model to improve the agreement between the observed versus predicted results. On the basis of a sensitivity analysis of crizotinib pharmacokinetic parameters, the PBPK model–predicted plasma concentrations were in good agreement with the observed results when Fa was assumed to be 0.3 (Fig. 5B). The predicted Cmax (266 ng/ml) and AUC0–τ (3182 ng∙h/ml) were within 20% of the observed values (Table 5). The predicted NSI (1.1) was also consistent with the observed result (1.3).

Fig. 5.
Fig. 5.

Observed and PBPK model–predicted plasma concentrations of crizotinib in cancer patients following 28-day multiple oral administration. Crizotinib (250 mg twice daily dose) was orally administered to five cancer patients for 28 days. Crizotinib plasma concentrations were predicted in a virtual population of healthy subjects (n = 15 per group × 6 groups, total 90 subjects) with a full–PBPK model implemented in the Simcyp simulator assuming crizotinib Fa of 0.5 (A) or 0.3 (B). The x-axis represents the time after dosing in hours and the y-axis represents the observed (○) and PBPK model–predicted (—) plasma concentrations in nanograms per milliliter on a logarithmic scale. The observed and predicted plasma concentrations are expressed as mean ± S.D. and mean (solid line) with 5th and 95th percentiles (dashed line), respectively.

View this table:
TABLE 5

Clinically observed and PBPK model–predicted pharmacokinetic parameters of crizotinib in cancer patients following 28-day multiple oral administration of crizotinib 250 mg twice daily

Data are expressed as geometric mean with percent coefficient of variation (CV%) in parentheses (n = 5 for the observed; n = 15 per group × 6 groups for the predicted).

To further investigate crizotinib-mediated CYP3A net-inhibition potential during multiple-dose oral administration, steady-state crizotinib pharmacokinetic parameters were compared between the simulations on and off (Table 5). The predicted on/off ratios for crizotinib AUC0–τ were 2.1–2.5, suggesting the crizotinib CYP3A net-inhibition effect on the accumulation of steady-state oral exposure was 2- to 3-fold (if Fa was consistent during the multiple-dose administration). In addition, the time-courses of crizotinib median fm,CYP3A4 in the liver and Fg in the intestine were predicted in the simulation on. Crizotinib fm,CYP3A4 decreased from 0.80 on day 0 to 0.48 on day 28 during 28-day multiple-dose administration, and Fg increased from 0.94 on day 0 to 0.98 on day 28 (data not shown).

Model Application

Prediction of Crizotinib Pharmacokinetics in Multiple-Dose DDI Study with Ketoconazole.

Following 28-day multiple oral administration of crizotinib 150 mg twice daily with and without coadministration of ketoconazole 200 mg twice daily, the predicted crizotinib Cmax and AUC0–τ were, respectively, 114 ng/ml and 1359 ng∙h/ml in the control group and 250 ng/ml and 2978 ng∙h/ml in the treatment group, resulting in the predicted CmaxR of 2.2 and AUCR of 2.2 (Table 6). Thus, the predicted AUCR in the multiple-dose simulation was 1.5-fold lower than that (3.4) in the single-dose simulation (Table 4). When crizotinib DDI parameters were not incorporated into the simulation (off), the predicted Cmax and AUC0–τ in the treatment group (221 ng/ml and 2619 ng∙h/ml, respectively) were 3.3-fold higher than those in the control group (66 ng/ml and 791 ng∙h/ml, respectively). Thus, the predicted AUCR (3.3) in the simulation off was comparable to that (3.4) in the single-dose simulation. It is worth noting that the on/off ratio was diminished from 1.7 in the control group to 1.1 in the treatment group, leading to the lower AUCR of 2.2 in the simulation on.

View this table:
TABLE 6

PBPK model–predicted pharmacokinetic parameters of crizotinib in healthy subjects following 28-day multiple oral administration of crizotinib with and without coadministration of ketoconazole or rifampin

Data are expressed as geometric mean with percent coefficient of variation (CV%) in parentheses (n = 15 per group × 6 groups). Dose levels were crizotinib 150 mg twice daily and ketoconazole 200 mg twice daily in the crizotinib-ketoconazole interaction and crizotinib 250 mg twice daily and rifampin 600 mg once daily in the crizotinib-rifampin interaction.

Time-courses of the predicted crizotinib hepatic and intestinal CLint in the simulation on are graphically presented in Fig. 3, C and D, respectively. The hepatic and intestinal CLint values in the control group gradually decreased and then reached steady state around day 14, whereas those in the treatment group were significantly inhibited from the beginning of the treatment. At the end of treatment (i.e., day 28), the hepatic and intestinal CLint values were, respectively, 3- and 12-fold lower in the treatment group than in the control group. The hepatic CLint inhibition led to a decrease in crizotinib fm,CYP3A4 of 0.80 (median) to 0.61 in the control group and to 0.07 in the treatment group on day 28 (Fig. 6A). In contrast, crizotinib Fg increased from 0.94 (median) to 0.97 in the control group and to 1.0 in the treatment groups on day 28 (Fig. 6B). The predicted crizotinib fm,CYP3A4 and Fg reached steady state around day 14.

Fig. 6.
Fig. 6.

PBPK model–predicted crizotinib fm,CYP3A4 (A and C) and Fg (B and D) in a Simcyp virtual population of healthy subjects following 28-day multiple oral administration of crizotinib with and without ketoconazole (A and B) or rifampin (C and D). A full–PBPK model simulation was performed in a Simcyp virtual population of healthy subjects (n = 15 per group × 6 groups, total 90 subjects) following 28-day multiple oral administration of crizotinib (150 mg twice daily) with and without coadministration of ketoconazole (200 mg twice daily) or following 28-day multiple oral administration of crizotinib (250 mg twice daily) with and without coadministration of rifampin (600 mg once daily). The x-axis represents the treatment period in days and the y-axis represents PBPK model–predicted median fm,CYP3A4 or Fg with 10th, 25th, 75th, and 90th percentiles in fraction.

Prediction of Crizotinib Pharmacokinetics in Crizotinib Multiple-Dose DDI Study with Rifampin.

Following 28-day multiple oral administration of crizotinib 250 mg twice daily with and without coadministration of rifampin 600 mg once daily, the predicted crizotinib Cmax and AUC0–τ were, respectively, 225 ng/ml and 2694 ng∙h/ml in the control group and 24 ng/ml and 275 ng∙h/ml in the treatment group with a large intersubject variability of ∼120% (Table 6). The predicted CmaxR and AUCR were 0.11 and 0.10, respectively. Thus, the predicted AUCR was roughly comparable between the single- and multiple-dose simulations. When crizotinib DDI parameters were not incorporated into the simulation (off), the predicted Cmax and AUC0–τ were, respectively, 110 ng/ml and 1319 ng∙h/ml in the control group and 15 ng/ml and 176 ng∙h/ml in the treatment group. Thus, the predicted AUCR was comparable between the simulations on and off (0.10 and 0.13, respectively).

Time-courses of the predicted crizotinib hepatic and intestinal CLint in the simulation on are graphically presented in Fig. 4, C and D, respectively. The hepatic and intestinal CLint values in the control group gradually decreased, whereas those in the treatment group significantly increased from the beginning of the treatment to around day 6 and then slightly decreased during the rest of the treatment period. The hepatic and intestinal CLint values were 4- to 6-fold higher in the treatment group than in the control group on day 28. The rifampin-mediated CYP3A induction led to an increase in crizotinib fm,CYP3A4 of 0.80 (median) to 0.96 on day 28 (Fig. 6C), whereas crizotinib Fg slightly decreased from 0.94 to 0.80 on day 28 (Fig. 6D). The predicted crizotinib fm,CYP3A4 and Fg reached steady state around day 14.

Prediction of Crizotinib Pharmacokinetics in Multiple-Dose DDI Study with a Weak or Moderate CYP3A Inhibitor.

Following 28-day multiple oral administration of crizotinib 150 mg twice daily with and without coadministration of diltiazem (120 mg twice a day), erythromycin (500 mg three times a day), fluconazole (200 mg once a day), or fluvoxamine (50 mg once a day), the predicted crizotinib Cmax and AUC0–τ were, respectively, 123 ng/ml and 1476 ng∙h/ml with diltiazem, 179 ng/ml and 2141 ng∙h/ml with erythromycin, 199 ng/ml and 2378 ng∙h/ml with fluconazole, and 114 ng/ml and 1368 ng∙h/ml with fluvoxamine (Table 7). Correspondingly, the predicted AUCR in the treatment groups with diltiazem, erythromycin, and fluconazole were 1.1, 1.6, 1.8, and 1.0, respectively.

View this table:
TABLE 7

PBPK model–predicted pharmacokinetic parameters of crizotinib in healthy subjects following 28-day multiple oral administration of crizotinib with and without coadministration of a weak or moderate CYP3A inhibitor

Data are expressed as geometric mean with percent coefficient of variation (CV%) in parentheses (n = 15 per group × 6 groups). Dose levels were crizotinib 150 mg twice daily with diltiazem 120 mg twice a day, erythromycin 500 mg three times a day, fluconazole 200 mg once a day, or fluvoxamine 50 mg once a day.

Discussion

It has become common to use PBPK modeling to contribute to our understanding of DDI mechanisms. We have sought to develop and verify a PBPK model that could achieve this goal for crizotinib. The present study clearly illustrates the challenges associated with developing such a model for a substrate drug that has time-dependent inhibition and induction properties when administered in combination with an interacting drug that is a strong inhibitor or inducer. Despite the complexity of these DDI mechanisms, the crizotinib PBPK model described appears successful in providing plausible predictions of multiple-dose DDI outcomes on the basis of available single-dose DDI data along with single- and multiple-dose pharmacokinetic results. However, some issues identified in the present study remain and warrant further discussion.

In the single-dose oral bioavailability study, crizotinib oral bioavailability was estimated to be 0.43 with an Fa × Fg value of 0.96 in healthy subjects at an oral dose of 250 mg relative to an intravenous dose of 50 mg, assuming linear pharmacokinetics (Xu et al., 2015). Using the model-predicted Fg value of 0.94, Fa was calculated to be ∼0.9 in this study. However, the recovery of crizotinib (as parent drug) in feces of healthy subjects in the single-dose mass-balance study (250 mg) was 53% of the administered dose, and it was improbable that this result was confounded by biliary excretion of the parent drug and/or reversible metabolites (Johnson et al., 2015). Therefore, the fecal recovery of crizotinib in the mass-balance study was considered as the fraction of dose unabsorbed (1 – Fa). The discrepancy between Fa estimates from these studies has likely been attributed to nonlinear pharmacokinetics between the oral dose of 250 mg versus the intravenous dose of 50 mg. On the basis of these findings, crizotinib Fa was set to 0.5 in the single-dose simulations, resulting in model-predicted crizotinib plasma concentrations that reasonably matched the observed results in the single-dose studies, including the DDI studies. In contrast, crizotinib steady-state plasma concentrations in cancer patients were over-predicted by approximately 2-fold (Table 5), which led us to refine the model for the multiple-dose pharmacokinetic simulation. Following a sensitivity analysis of crizotinib pharmacokinetic parameters, the model-predicted results were in good agreement with the observed results when Fa was assumed to be 0.3, suggesting that crizotinib Fa might decrease from 0.5 to 0.3 during multiple-dose administration. Given the relatively large dose (250 mg), a twice-daily dosing regimen, and different plasma concentration-time profiles at steady state (i.e., relatively flat profiles) compared with single-dose administration, it could be possible that the absorption process became saturated following multiple-dose administration. Alternatively, crizotinib Fa may differ between healthy subjects (single-dose) and cancer patients (multiple-dose), owing, e.g., to its pH-dependent solubility and/or relatively low permeability with an efflux potential, i.e., a substrate of P-glycoprotein (CDER, 2011). Furthermore, crizotinib has been suggested as a moderate CYP3A inhibitor given the clinically observed DDI result with midazolam (i.e., AUCR of 3.7) following multiple-dose administration of crizotinib 250 mg twice daily (Tan et al., 2010). On the basis of the comparison of crizotinib steady-state simulations between on and off (without an interacting drug), crizotinib CYP3A net-inhibition potential (i.e., on/off ratio) was predicted as 2- to 3-fold (Table 5). This moderate on/off ratio appeared consistent with the midazolam AUCR when taking into account the difference in fm,CYP3A4 and Fg between crizotinib (0.8 and 0.9, respectively) and midazolam (0.9 and 0.6, respectively) (Wang, 2010). Conversely, the observed crizotinib NSI of 1.3 suggested that crizotinib CYP3A net-inhibition potential was negligible to weak (Table 5). This discrepancy could also be explained by a possible decrease in Fa following multiple-dose administration. That is, the lower than anticipated NSI may be a result of the net effect of the opposing mechanisms of crizotinib-mediated CYP3A net inhibition and a decrease in Fa. Additionally, when crizotinib steady-state plasma concentrations were simulated with only crizotinib CYP3A induction parameters (Emax and EC50) without its inhibition parameters (KI and kinact), the simulated plasma concentrations were within 10% of the result of simulation off (data not shown). Thus, it is improbable that crizotinib-mediated CYP3A induction can offset its CYP3A inhibition following multiple-dose administration.

In the single-dose DDI prediction with ketoconazole, the predicted crizotinib Cmax and AUC0–∞ in both the control and treatment groups were comparable between the simulations on and off (Table 4), suggesting a negligible effect of crizotinib CYP3A net inhibition on the single-dose DDI results. These findings were expected on the basis of the overall crizotinib net-inhibition mechanism (i.e., moderate time-dependent inhibition with weak induction). In the multiple-dose DDI prediction with ketoconazole, crizotinib CYP3A net inhibition could theoretically be considered as additive to ketoconazole-mediated inhibition. The predicted Cmax and AUC0–τ in the control group were 1.7-fold higher in the simulation on than off owing to the effect of crizotinib-mediated CYP3A net inhibition (Table 6). In contrast, the predicted crizotinib Cmax and AUC0–τ in the treatment groups were comparable between the simulations on and off, which was most probably attributable to strong (or near-complete) CYP3A inhibition by ketoconazole. Correspondingly, the diminished on/off ratios from the control group (1.7) to the treatment group (1.1) resulted in the 1.5-fold lower predicted CmaxR and AUCR in the simulation on (2.2) than off (3.3). As shown in Fig. 3, the predicted crizotinib hepatic and intestinal CLint values were significantly inhibited by ketoconazole (treatment group) relative to crizotinib (control group). As a result, the predicted crizotinib fm,CYP3A4 in the treatment group decreased from 0.80 to 0.07, whereas the predicted Fg increased from 0.94 to 1.0, suggesting a near-maximal CYP3A inhibition following multiple-dose administration of ketoconazole 200 mg twice daily, which appears consistent with a previous report (Zhao et al., 2009). To confirm this hypothesis, crizotinib DDI prediction was performed with a higher ketoconazole dose of 400 mg “twice daily” (Supplemental Table S1). As expected, the predicted crizotinib Cmax and AUC0–τ in the treatment group were roughly comparable to those with ketoconazole at the dose of 200 mg twice daily. Thus, the inhibition of CYP3A-mediated crizotinib clearance following multiple-dose administration of ketoconazole 200 mg twice daily appears to be nearly complete. Taken together, these findings suggest that, when coadministered with ketoconazole, the effect of crizotinib CYP3A net inhibition on its oral exposures would be negligible following either single- or multiple-dose administration of crizotinib.

In the single-dose DDI prediction with rifampin, the predicted crizotinib Cmax and AUC0–∞ in both the control and treatment groups were comparable between the simulations on and off (Table 4), suggesting a negligible effect of crizotinib CYP3A net inhibition on the single-dose DDI outcome, as anticipated from on crizotinib net-inhibition mechanism. Unexpectedly, the observed apparent t1/2,z in the treatment group (48 hours) was slightly longer than that in the control group (33 hours). In contrast, the predicted t1/2,z (33 hours) in the treatment group was shorter than that (55 hours) in the control group as would be expected from rifampin-mediated CYP3A induction (Table 4). These findings might suggest that the interaction of crizotinib with rifampin could result from not only CYP3A induction but also some other mechanism(s), such as transporter-mediated distribution and/or excretion. Crizotinib has not been identified as a substrate of any uptake transporters; however, crizotinib is a substrate of P-glycoprotein (CDER, 2011), which is also induced by rifampin (Paine et al., 2002; Kim et al., 2008). Therefore, P-glycoprotein may, in part, play a role in the crizotinib-rifampin interaction (e.g., increase in entero-hepatic circulation). In the multiple-dose DDI predictions with rifampin, crizotinib-mediated CYP3A net inhibition could theoretically diminish rifampin-mediated CYP3A induction to some extent. However, the predicted Cmax and AUC0–τ values in both the control and treatment groups were approximately 2-fold higher in the simulation on than off, resulting in the comparable predicted CmaxR and AUCR between the simulations on and off (Table 6). Consistently, there appeared negligible-to-minimal effects of crizotinib on the predicted hepatic and intestinal CLint when the CLint was markedly induced following multiple-dose coadministration of rifampin (Fig. 4). Additionally, the multiple-dose DDI simulation was performed with a lower rifampin Emax of 8. As expected, the difference in the predicted CmaxR and AUCR between the Emax of 8 (0.26–0.28) and 16 (0.10–0.13) was 2- to 3-fold owing to the difference in the Emax values (Supplemental Table S1). However, the predicted CmaxR and AUCR were still comparable between the simulation on and off with Emax of 8, suggesting that the effects of crizotinib on the multiple-dose DDI outcomes could be negligible even though rifampin Emax was decreased to its half. Collectively, these simulations suggest that the effect of crizotinib-mediated CYP3A net inhibition on its oral exposure would be negligible on the multiple-dose DDI outcomes when coadministered with rifampin.

In the DDI prediction for substrate drugs, hepatic fm,CYP3A and intestinal Fg are considered the most important parameters (Fahmi et al., 2008). The present simulation results suggested that 1) crizotinib hepatic fm,CYP3A decreased from 0.80 (median) to 0.48 following multiple-dose administration of crizotinib 250 mg twice daily, 2) coadministration of crizotinib (150 mg twice daily) with ketoconazole (200 mg twice daily) resulted in a decrease in the fm,CYP3A to 0.07, and 3) coadministration of crizotinib (250 mg twice daily) with rifampin (600 mg once daily) resulted in the increase in the fm,CYP3A to 0.96 (Fig. 7). The changes in hepatic fm,CYP3A would lead to the significant fm changes in other pathways such as non-CYP3A metabolic clearance and renal excretion. In contrast, the gut contribution to crizotinib systemic DDIs was considered minimal because of the predicted high Fg (0.94). The predicted crizotinib Fg increased to 0.98 following multiple-dose administration of crizotinib alone and to 1.0 by coadministration with ketoconazole, whereas it decreased to 0.80 by coadministration with rifampin. The mechanistic dynamic modeling approach affords an opportunity to gain such an insight into the underlying mechanisms mediating DDIs as a function of time.

Fig. 7.
Fig. 7.

Summary of PBPK model–predicted crizotinib fm changes in a Simcyp virtual population of healthy subjects following 28-day multiple oral administration of crizotinib with and without ketoconazole or rifampin. Crizotinib fractions (fm) metabolized/eliminated by each pathway were predicted in a Simcyp virtual population of healthy subjects (n = 15 per group × 6 groups, total 90 subjects) following 28-day multiple oral administration of crizotinib 250 mg twice daily (crizotinib), 28-day multiple coadministration of crizotinib 150 mg twice daily with ketoconazole 200 mg twice daily (crizotinib with ketoconazole), and 28-day multiple coadministration of crizotinib 250 mg twice daily with rifampin 600 mg once daily (crizotinib with rifampin). Data are expressed as median (n = 90) in each clearance pathway.

Overall, the present study demonstrated that the crizotinib PBPK model reasonably predicted plasma concentrations of crizotinib in humans after a single intravenous infusion and single- and 28-day multiple oral administration. Furthermore, the PBPK model adequately predicted the fold-increases in crizotinib oral exposures in humans after a single oral administration of crizotinib with multiple-dose coadministration of ketoconazole or rifampin. These results suggest that the crizotinib PBPK model described has been sufficiently developed, refined, and verified on the basis of the clinically observed results and can be applied to predict crizotinib multiple-dose DDI outcomes. The multiple-dose DDI predictions suggest that crizotinib net inhibition on CYP3A may have a negligible contribution to DDI outcomes when crizotinib is coadministered with strong CYP3A inhibitors or inducers such as ketoconazole and rifampin. Therefore, recommendations for crizotinib dose adjustment in multiple-dose DDI scenarios could be possible on the basis of the currently available single-dose DDI results. Overall, we believe that the present crizotinib PBPK model can be useful for predicting crizotinib exposures in other clinical studies, such as DDIs with weak/moderate CYP3A inhibitors/inducers and drug-disease interactions in patients with hepatic or renal impairment.

Acknowledgments

The authors thank Akintunde Bello, Weiwei Tan, and Huiping Xu, (Clinical Pharmacology, Pfizer, San Diego, CA) for valuable discussion about clinical pharmacokinetics of crizotinib and Bhasker Shetty and Paolo Vicini (Pharmacokinetics, Dynamics and Metabolism, Pfizer, San Diego, CA) for contributions to the draft manuscript.

Authorship Contributions

Participated in research design: Yamazaki.

Performed data analysis: Yamazaki.

Wrote or contributed to the writing of the manuscript: Johnson, Smith,Yamazaki.

Footnotes

Abbreviations

AUC
area under the plasma concentration-time curve
AUCR
fold-increase in AUC by an interacting drug
CmaxR
fold-increase in Cmax by an interacting drug
CL
plasma clearance
CLint
intrinsic clearance
DDI
drug-drug interaction
Fa
fraction of dose absorbed from gastrointestinal tract
Fg
fraction of dose that escapes intestinal first-pass metabolism
fm
fraction metabolized or excreted by each CL pathway
fm,CYP3A4
fraction metabolized by CYP3A4
fu
unbound fraction
Ki
inhibition constant
KI
inactivation constant
kinact
maximum inactivation rate constant
NSI
nonstationary pharmacokinetic index
PBPK
physiologically based pharmacokinetics
t1/2,z
apparent terminal half-life
tmax
time to reach maximum plasma concentration
Vss
volume of distribution at steady state

References

  1. ↵
    Almond LM, Okialda K, Mukadam S, Rowland-Yeo K, Gardner I, Rostami-Hodjegan A, and Kenny JR (2012) Assessment of the induction of systemic clearance versus first-pass effect of midazolam (MDZ) using a physiologically-based dynamic model, at the 113th ASCPT Meeting, 2012 Mar 14–17; National Harbor, MD; American Society for Clinical Pharmacology and Therapeutics, Alexandria, VA.
  2. ↵
    1. Boulenc X and
    2. Barberan O
    (2011) Metabolic-based drug-drug interactions prediction, recent approaches for risk assessment along drug development. Drug Metabol Drug Interact 26:147–168.
    OpenUrlPubMed
  3. ↵
    CDER (2011) Application Number: 202570Orig1s000, Clinical Pharmacology and Biopharmaceutics Review(s). http://www.accessdata.fda.gov/drugsatfda_docs/nda/2011/202570Orig1s000ClinPharmR.pdf. US Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER).
  4. ↵
    CDER (2012) Drug Interaction Studies — Study Design, Data Analysis, Implications for Dosing, and Labeling Recommendations Draft Guidance. http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ucm292362.pdf. US Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER).
  5. ↵
    CHMP (2012) Guideline on the Investigation of Drug Interactions. http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2012/07/WC500129606.pdf. European Medicines Agency, Committee for Human Medicinal Products (CHMP).
  6. ↵
    1. Fahmi OA,
    2. Maurer TS,
    3. Kish M,
    4. Cardenas E,
    5. Boldt S, and
    6. Nettleton D
    (2008) A combined model for predicting CYP3A4 clinical net drug-drug interaction based on CYP3A4 inhibition, inactivation, and induction determined in vitro. Drug Metab Dispos 36:1698–1708.
    OpenUrlAbstract/FREE Full Text
  7. ↵
    1. Huang SM and
    2. Lesko LJ
    (2009) Authors’ response. J Clin Pharmacol 43:370.
    OpenUrl
  8. ↵
    1. Huang SM and
    2. Rowland M
    (2012) The role of physiologically based pharmacokinetic modeling in regulatory review. Clin Pharmacol Ther 91:542–549.
    OpenUrlCrossRefPubMed
  9. ↵
    1. Huang SM,
    2. Abernethy DR,
    3. Wang Y,
    4. Zhao P, and
    5. Zineh I
    (2013) The utility of modeling and simulation in drug development and regulatory review. J Pharm Sci 102:2912–2923.
    OpenUrlCrossRefPubMed
  10. ↵
    1. Jamei M,
    2. Marciniak S,
    3. Feng K,
    4. Barnett A,
    5. Tucker G, and
    6. Rostami-Hodjegan A
    (2009) The Simcyp population-based ADME simulator. Expert Opin Drug Metab Toxicol 5:211–223.
    OpenUrlCrossRefPubMed
  11. ↵
    Johnson TR, Zhou S, Lee C, Clouser-Roche A, Freiwald S, Youdim KA, Aherne K, Batugo MR, Skaptason J, Yamazaki S et al .(2011) Identification of the human enzymes invovlved in the oxidative metabolism of crizotinib, at the 17th North American Regional ISSX Meeting; 2011 Oct 16–20; Atlanta, GA International Society for the Study of Xenobiotics, Washington, DC.
  12. ↵
    Johnson TR, Tan W, Goulet L, Smith EB, Yamazaki S, Walker GS, O’Gorman MT, Bedarida G, Zou HY, Christensen JG, et al. (2015) Metabolism, excretion and pharmacokinetics of [14C]crizotinib following oral administration to healthy subjects. Xenobiotica 45:45–59.
  13. ↵
    1. Kim KA,
    2. Park PW,
    3. Liu KH,
    4. Kim KB,
    5. Lee HJ,
    6. Shin JG, and
    7. Park JY
    (2008) Effect of rifampin, an inducer of CYP3A and P-glycoprotein, on the pharmacokinetics of risperidone. J Clin Pharmacol 48:66–72.
    OpenUrlCrossRefPubMed
  14. ↵
    1. Mao J,
    2. Johnson TR,
    3. Shen Z, and
    4. Yamazaki S
    (2013) Prediction of crizotinib-midazolam interaction using the Simcyp population-based simulator: comparison of CYP3A time-dependent inhibition between human liver microsomes versus hepatocytes. Drug Metab Dispos 41:343–352.
    OpenUrlAbstract/FREE Full Text
  15. ↵
    1. Paine MF,
    2. Wagner DA,
    3. Hoffmaster KA, and
    4. Watkins PB
    (2002) Cytochrome P450 3A4 and P-glycoprotein mediate the interaction between an oral erythromycin breath test and rifampin. Clin Pharmacol Ther 72:524–535.
    OpenUrlCrossRefPubMed
  16. ↵
    1. Peters SA,
    2. Schroeder PE,
    3. Giri N, and
    4. Dolgos H
    (2012) Evaluation of the use of static and dynamic models to predict drug-drug interaction and its associated variability: impact on drug discovery and early development. Drug Metab Dispos 40:1495–1507.
    OpenUrlAbstract/FREE Full Text
  17. ↵
    1. Prueksaritanont T,
    2. Chu X,
    3. Gibson C,
    4. Cui D,
    5. Yee KL,
    6. Ballard J,
    7. Cabalu T, and
    8. Hochman J
    (2013) Drug-drug interaction studies: regulatory guidance and an industry perspective. AAPS J 15:629–645.
    OpenUrlCrossRefPubMed
  18. ↵
    1. Rodgers T,
    2. Leahy D, and
    3. Rowland M
    (2005) Physiologically based pharmacokinetic modeling 1: predicting the tissue distribution of moderate-to-strong bases. J Pharm Sci 94:1259–1276.
    OpenUrlCrossRefPubMed
  19. ↵
    1. Rowland M,
    2. Peck C, and
    3. Tucker G
    (2011) Physiologically-based pharmacokinetics in drug development and regulatory science. Annu Rev Pharmacol Toxicol 51:45–73.
    OpenUrlCrossRefPubMed
  20. ↵
    1. Tan W,
    2. Wilner KD,
    3. Bang Y,
    4. Kwak EL,
    5. Maki RG,
    6. Camidge DR,
    7. Solomon BJ,
    8. Ou SI,
    9. Salgia R, and
    10. Clark JW
    (2010) Pharmacokinetics (PK) of PF-02341066, a dual ALK/MET inhibitor after multiple oral doses to advanced cancer patients. J Clin Oncol 28:15s.
    OpenUrlCrossRefPubMed
  21. ↵
    1. Varma MV,
    2. Pang KS,
    3. Isoherranen N, and
    4. Zhao P
    (2015) Dealing with the complex drug-drug interactions: towards mechanistic models. Biopharm Drug Dispos 36:71–92.
    OpenUrlCrossRefPubMed
  22. ↵
    1. Wang YH
    (2010) Confidence assessment of the Simcyp time-based approach and a static mathematical model in predicting clinical drug-drug interactions for mechanism-based CYP3A inhibitors. Drug Metab Dispos 38:1094–1104.
    OpenUrlAbstract/FREE Full Text
  23. ↵
    Xu H, O’Gorman M, Tan W, Leister C, Monajati M, Brega N, Ni G, Phillips S, Pardo P, and Bello A (2011a) The effect of a potent CYP3A inhibitor ketoconazole on the pharmacokinetics of crizotinib, a dual ALK/MET inhibitor, in healthy subjects. in AAPS Annual Meeting and Exposition; 2011 October 23–27; Washington, DC American Association of Pharmaceutical Scientists, Arlington, VA.
  24. ↵
    Xu H, O’Gorman M, Tan W, Leister C, Monajati M, Brega N, Ni G, Phillips S, Pardo P, and Bello A (2011b) The effect of a strong CYP3A inducer rifampin on the pharmacokinetics of crizotinib, a dual ALK/MET inhibitor, in healthy subjects, in AAPS Annual Meeting and Exposition; 2011 October 23–27; Washington, DC American Association of Pharmaceutical Scientists, Arlington, VA.
  25. ↵
    1. Xu H,
    2. O’Gorman M,
    3. Boutros T,
    4. Brega N,
    5. Kantaridis C,
    6. Tan W, and
    7. Bello A
    (2015) Evaluation of crizotinib absolute bioavailability, the bioequivalence of three oral formulations, and the effect of food on crizotinib pharmacokinetics in healthy subjects. J Clin Pharmacol 55:104–113.
    OpenUrlCrossRefPubMed
  26. ↵
    1. Yang J,
    2. Jamei M,
    3. Yeo KR,
    4. Tucker GT, and
    5. Rostami-Hodjegan A
    (2007) Prediction of intestinal first-pass drug metabolism. Curr Drug Metab 8:676–684.
    OpenUrlCrossRefPubMed
  27. ↵
    1. Zhang L,
    2. Zhang YD,
    3. Zhao P, and
    4. Huang SM
    (2009) Predicting drug-drug interactions: an FDA perspective. AAPS J 11:300–306.
    OpenUrlCrossRefPubMed
  28. ↵
    1. Zhao P,
    2. Ragueneau-Majlessi I,
    3. Zhang L,
    4. Strong JM,
    5. Reynolds KS,
    6. Levy RH,
    7. Thummel KE, and
    8. Huang SM
    (2009) Quantitative evaluation of pharmacokinetic inhibition of CYP3A substrates by ketoconazole: a simulation study. J Clin Pharmacol 49:351–359.
    OpenUrlCrossRefPubMed
  29. ↵
    1. Zhao P,
    2. Zhang L,
    3. Grillo JA,
    4. Liu Q,
    5. Bullock JM,
    6. Moon YJ,
    7. Song P,
    8. Brar SS,
    9. Madabushi R,
    10. Wu TC,
    11. et al.
    (2011) Applications of physiologically based pharmacokinetic (PBPK) modeling and simulation during regulatory review. Clin Pharmacol Ther 89:259–267.
    OpenUrlCrossRefPubMed
Back to top

In this issue

Download PDF
Article Alerts
Email Article
Citation Tools

Research ArticleArticle

Prediction of Crizotinib Drug-Drug Interactions

Shinji Yamazaki, Theodore R. Johnson and Bill J. Smith
Drug Metabolism and Disposition October 1, 2015, 43 (10) 1417-1429; DOI: https://doi.org/10.1124/dmd.115.064618
del.icio.us logo Digg logo Reddit logo Twitter logo Facebook logo Google logo Mendeley logo

Jump to section

Advertisement