Advertisement
Research ArticleArticle

Prediction of Crizotinib-Midazolam Interaction Using the Simcyp Population-Based Simulator: Comparison of CYP3A Time-Dependent Inhibition between Human Liver Microsomes versus Hepatocytes

Jialin Mao, Theodore R. Johnson, Zhongzhou Shen and Shinji Yamazaki
Drug Metabolism and Disposition February 2013, 41 (2) 343-352; DOI: https://doi.org/10.1124/dmd.112.049114

Abstract

Crizotinib (Xalkori) is an orally available potent inhibitor of multiple tyrosine kinases, including anaplastic lymphoma kinase and mesenchymal-epithelial transition factor. Objectives of the present study were as follows: 1) to characterize crizotinib time-dependent inhibition (TDI) potency for CYP3A in human liver microsomes (HLM) and cryopreserved human hepatocytes suspended in human plasma (HSP); 2) to characterize crizotinib enzyme induction potency on CYP3A4 in cryopreserved human hepatocytes; 3) to predict crizotinib steady-state plasma concentrations in patients (e.g., autoinhibition and autoinduction) using the mechanistic dynamic model, Simcyp population-based simulator; and 4) to predict a clinical crizotinib-midazolam interaction using the dynamic model as well as the static mathematical model. Crizotinib inactivation constant (KI) and maximum inactivation rate constant (kinact) for TDI were estimated as, respectively, 0.37 µM and 6.9 h−1 in HLM and 0.89 µM and 0.78 h−1 in HSP. Thus, crizotinib inactivation efficiency (kinact/KI) was ∼20-fold lower in HSP relative to HLM. Crizotinib Emax and EC50 for CYP3A4 induction (measured as mRNA expression) were estimated as 6.4- to 29-fold and 0.47 to 3.1 µM, respectively. Based on these in vitro parameters, the predicted crizotinib steady-state area under plasma concentration-time curve (AUC) with HLM-TDI was 2.1-fold higher than the observed AUC, whereas that with HSP-TDI was consistent with the observed result (≤1.1-fold). The increase in midazolam AUC with coadministration of crizotinib (21-fold) was significantly overpredicted using HLM-TDI, whereas the prediction using HSP-TDI (3.6-fold) was consistent with the observed result (3.7-fold). Collectively, the present study demonstrated the value of HSP to predict in vivo CYP3A-mediated drug-drug interaction.

Introduction

A clinically relevant drug-drug interaction (DDI) is generally considered a modification of pharmacological and/or toxicological effects of one drug (object) by another drug (precipitant). In many cases, DDI can be attributed to modulation of drug-metabolizing enzymes, particularly CYP3A, as these are the most abundantly expressed in the liver and intestines (Shimada et al., 1994; Slaughter and Edwards, 1995). The enzyme inhibition mechanism can be categorized as reversible (competitive, uncompetitive, or noncompetitive) inhibition (RI) or time-dependent (mechanism- or metabolism-based) inhibition (TDI). For in vitro TDI evaluation, an inactivation constant (KI) and a maximum inactivation rate constant (kinact) are typically determined in human liver microsomes (HLM) (Grimm et al., 2009; Zimmerlin et al., 2011), and these kinetic parameters are used to predict potential in vivo DDI (Kanamitsu et al., 2000; Mayhew et al., 2000; Obach et al., 2007). In addition to HLM, TDI evaluation in human hepatocytes (HEP) has recently been reported for the known CYP3A time-dependent inhibitors such as diltiazem, erythromycin, and ritonavir (Zhao et al., 2005; Xu et al., 2009; Chen et al., 2011; Kirby et al., 2011). HEP are also extensively used for the evaluation of enzyme induction (EI) for CYPs (Chu et al., 2009). Compared with HLM, HEP are intact cellular systems containing a full complement of phase I/II–metabolizing enzymes and transporter proteins (Di et al., 2012). The presence of cell membranes in HEP is important in maintaining the effects of active uptake/efflux transporters and passive diffusion on intracellular drug concentration; therefore, these factors may substantially affect the determination of enzyme kinetic parameters such as KI and kinact. Generally, the known CYP3A inhibitors showed a less potent TDI in HEP relative to HLM, which have yielded an impact on the in vivo DDI prediction (Xu et al., 2009; Chen et al., 2011; Kirby et al., 2011). Furthermore, it has been reported that cryopreserved human hepatocytes suspended in human plasma (HSP), compared with those in protein-free media, showed better DDI predictions for some of cytochrome P450 inhibitors (Mao et al., 2011, 2012).

Recently, there has been a growing interest in physiologically based pharmacokinetic (PBPK) models, in which each physiologic compartment (i.e., organ or tissue defined by a tissue volume) is connected with blood flow. The PBPK model provides disposition profiles to be predicted from physico- and biochemical properties of compounds (Jones et al., 2006; Lavé et al., 2007; Nestorov, 2007). This dynamic approach is increasingly being employed in drug discovery and development setting to predict pharmacokinetics (PK) and DDI potential in the clinic. Traditionally, DDI predictions have been performed with static mathematical models using various inhibitor concentrations such as hepatic inlet or outlet concentrations in total (protein bound plus unbound) or unbound form (Mayhew et al., 2000; Obach et al., 2007; Fahmi et al., 2009; Boulenc and Barberan, 2011; Mao et al., 2012). In these reports, hepatic DDI magnitudes for RI were reasonably predicted using the projected unbound portal vein (or hepatic inlet) concentration (Cinlet,u), whereas the use of unbound systemic (or hepatic outlet) concentration (Csys,u) yielded better prediction for TDI and EI compared with Cinlet,u. Based on advancements in the field, the Food and Drug Administration (FDA) has recently issued a revised draft DDI guidance for industry (http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ucm292362.pdf), which emphasizes the use of an integrated approach such as static mathematical model or mechanistic dynamic PBPK models to simultaneously utilize these DDI mechanisms (e.g., TDI, EI, and RI) to predict potential clinical DDI of new chemical entities. In many cases, an inhibitor concentration used for DDI prediction with general PBPK models is considered to be comparable to Csys,u at equilibrium since the tissue compartments are assumed to be perfusion limited without any diffusion barrier.

Crizotinib (Xalkori; PF02341066) was identified as an orally available potent inhibitor of multiple tyrosine kinases, including anaplastic lymphoma kinase and mesenchymal-epithelial transition factor (Fig. 1). Crizotinib was recently approved by FDA for the treatment of patients with locally advanced or metastatic non-small cell lung cancer that is anaplastic lymphoma kinase positive as detected by a FDA-approved test. It has been reported that crizotinib is a substrate of CYP3A (Johnson et al., 2011a). Objectives of the present study were as follows: 1) to characterize crizotinib TDI potency on CYP3A in HLM and HSP; 2) to characterize crizotinib EI potency in cryopreserved HEP; 3) to predict crizotinib steady-state plasma concentrations in patients using the mechanistic dynamic model; and 4) to predict a clinical crizotinib-midazolam interaction, utilizing the mechanistic dynamic model as well as the static mathematical model. The dynamic model used was a commercially available Simcyp population-based dynamic simulator (Jamei et al., 2009).

Fig. 1.
Fig. 1.

Chemical structure of crizotinib.

Materials and Methods

Materials.

Crizotinib [(R)-3-[1-(2,6-dichloro-3-fluoro-phenyl)-ethoxy]-5-(1-piperidin-4-yl-1H-pyrazol-4-yl)-pyridin-2-ylamine; chemical purity >99%] was synthesized by Pfizer Worldwide Research and Development (San Diego, CA) (Cui et al., 2011). Pooled HLM from 50 individual donors (30 males and 20 females) were prepared and characterized at XenoTech LLC (Lenexa, KS). Cryopreserved HEP suspension (three males and two females) and InVitro GRO HT medium were obtained from Celsis In Vitro Technologies (Baltimore, MD). Human plasma (heparin as anticoagulant) was obtained from Lampire Biologic Laboratories (Pipersville, PA). HEP maintenance medium was obtained from Lonza (Walkersville, MD). Midazolam, 1′-hydroxymidazolam (1′-OH midazolam), and [13C3]-1′-hydroxymidazolam (internal standard) were obtained from BD Gentest (Woburn, MA). Glucose-6-phosphate and glucose-6-phosphate dehydrogenase were from Sigma-Aldrich (St. Louis, MO). All other commercially available reagents and solvents were of either analytical or high performance liquid chromatography grade.

Crizotinib TDI Assay in HLM.

A two-step incubation scheme was used to analyze the inhibition of midazolam hydroxylation by crizotinib in HLM. Crizotinib at final concentrations of 0.3–10 µM (0.1% of final DMSO concentration) was preincubated with HLM (0.5 mg/mL) and NADPH (1 mM) at 37°C for 0, 5, 10, and 20 minutes. The incubation mixture was diluted (1:10) with buffer containing NADPH (1 mM), and midazolam (10 μM) was added for an additional 3-minute incubation period to quantify 1′-hydroxymidazolam as the remaining CYP3A activity. The incubation was then terminated by the addition of 100 µl acetonitrile/methanol (3:1 v/v) containing 3% formic acid and the internal standard (0.1 μM). The incubations were performed in duplicate. Samples were centrifuged at 2000 g for 20 minutes, and an aliquot of the supernatant was analyzed by liquid-chromatography tandem mass spectrometry (LC-MS/MS).

Crizotinib TDI Assay in HSP.

Crizotinib TDI potency was evaluated in HSP following a similar study design, as reported previously (Mao et al., 2011). Briefly, the final crizotinib concentrations used in the incubation mixture were 0.13–100 µM (0.5% of final methanol concentration). Incubations were performed in duplicate. A total of 25 μl of stock cryopreserved HEP (prepared in human plasma as 2 × 106 cells/ml) was added to 50 μl of human plasma containing crizotinib and incubated for 0, 10, and 20 minutes (37°C, 5% CO2) before the addition of midazolam. Midazolam in human plasma (25 µl) at final concentration of 30 µM was added to the incubation mixture, followed by a 35-minute incubation to quantify 1′-hydroxymidazolam as the remaining CYP3A activity. The final concentration of HEP was 0.5 × 106 cells/ml. Reactions were terminated by adding 200 µl of acetonitrile/methanol (3:1 v/v) containing the internal standard (0.15 μM). Samples were centrifuged at 2000 g for 20 minutes, and an aliquot of the supernatant was analyzed by LC-MS/MS.

Crizotinib EI Assay in Cryopreserved HEP.

Crizotinib CYP3A4 EI potency was evaluated in cryopreserved HEP from three donors (lot Hu4026, Hu8020, and HIE) using a procedure described previously (Fahmi et al., 2008). Briefly, cryopreserved HEP (3.5 × 105 viable cells per 0.5 ml of plating medium in each well of collagen I-precoated 24-well plates) were incubated for 24 hours (37°C, 95% relative humidity, 5% CO2) before the addition of crizotinib (0.1% of final DMSO concentration). The cells were treated with crizotinib at final concentrations of 0.25–7 μM daily for 3 consecutive days. Quantification of CYP3A4 mRNA was performed using the TaqMan two-step reverse-transcription polymerase chain reaction method, and the relative quantity of the target CYP3A4 gene compared with the endogenous control was determined by the ΔΔCT method. An effect of crizotinib on cell viability was assessed using the WST-1 cell proliferation reagent.

LC-MS/MS Analysis for Midazolam Metabolite.

Concentrations of the CYP3A-mediated metabolite of midazolam, 1′-hydroxymidazolam, in the in vitro TDI assays were determined by a LC-MS/MS method after the protein precipitation. The LC-MS/MS system consisted of Shimadzu LC-10ADvp pumps (Shimadzu, Columbia, MD), a CTC PAL autosampler (Leap Technologies, Carrboro, NC), and a Sciex API 4000 mass spectrometer (Sciex, Foster City, CA) equipped with a turbo ion spray source. Chromatographic separation was achieved by a reverse-phase column (Agilent Zorbax SB-Phenyl, 5 µm, 50 × 2.1-mm column; Agilent, Santa Clara, CA) with a mobile phase consisting of water with 0.1% formic acid (mobile phase A) and acetonitrile with 0.1% formic acid (mobile phase B) at a flow rate of 0.8 ml/min. The injection volume was 5 µL. Analytes were eluted using a step gradient from 1 to 40% mobile phase B over 3 minutes, and then increased to 90% in the next 0.01 minutes. Mobile phase B was held at 90% from 3.01 to 3.50 minutes, and the column was re-equilibrated to 1% over another 0.5 minutes. The mass spectrometry was operated in the positive ionization mode using multiple reaction monitoring at specific parent ion→product ion transitions of m/z 342→203 for 1′-hydroxymidazolam and 345→171 for [13C3]-1′-hydroxymidazolam. The dynamic range of the assay ranged from 7.5 to 2540 nM. The back-calculated calibration standard concentrations were within ±15% of their nominal concentrations with coefficients of variation of less than 15%. No significant carryover and matrix effects were observed in the study.

Estimation of Crizotinib TDI Kinetic Parameters.

To determine TDI kinetic parameters (i.e., KI and kinact) in the HLM assay, the apparent inactivation rate constant (kobs) for each crizotinib concentration ([I]) was first estimated from the slope of initial linear decline of CYP3A remaining activity on a natural logarithmic scale over the preincubation time. Apparent KI (KI,app) and kinact values were then determined by solving the following nonlinear equation (Mayhew et al., 2000):

Embedded Image(1)

Subsequently, for prediction purpose, an estimate for KI was calculated from KI,app following the correction for in vitro nonspecific binding in microsomes (fu,mic), which was measured by the equilibrium dialysis method (Yamazaki et al., 2011).

In the HSP assay, the previously reported mathematical equations incorporating reversible (Eq. 2), irreversible (Eq. 3), or both reversible and irreversible inhibitions (Eq. 4) were employed to estimate the apparent reversible inhibition constant (Ki,app) and/or TDI parameters (KI,app and kinact) by simultaneously fitting three IC50 curves obtained from different incubation times using a weighted nonlinear regression analysis (Mao et al., 2011):Embedded Image(2)Embedded Image(3)Embedded Image(4)where t is a total incubation time of a preincubation (0, 10, or 20 minutes) with an inhibitor, followed by a 35-minute incubation with midazolam, Vit represents the 1′-OH midazolam formation rate at a given inhibitor concentration during total incubation time, Vct represents the 1′-OH midazolam formation rate of vehicle control (no inhibitor) during total incubation time, Km,MDZ is the Michaelis-Menten constant for 1′-OH midazolam formation in HSP (55 µM total corresponding to 2 µM free), Vmax,t represents maximum 1′-OH midazolam formation rate, [I] is the nominal inhibitor concentration, and [S] is the final incubation concentration of midazolam (30 µM).

The ratio of Vit and Vct normalized the baseline for the enzyme activity at the designated incubation time. Model selection was based on a number of criteria such as Akaike information criterion, estimates for each parameters, and standard errors. Finally, an estimate for KI,app was converted to KI (as the input for Simcyp) following the correction for the unbound fraction in plasma (fu,plasma), which was measured by the equilibrium dialysis method (Yamazaki et al., 2011).

Prediction of Crizotinib and Midazolam Pharmacokinetics with Simcyp.

Physicochemical and in vitro PK parameters of crizotinib used for DDI prediction are summarized in Table 1. Based on these parameters, two compound files of crizotinib were created in Simcyp (version 11.1), as follows: one with the TDI parameters from HLM (henceforth referred to as CRZ-HLM), and the other with the TDI parameters from HSP (henceforth referred to as CRZ-HSP). The only difference in these two crizotinib Simcyp files was the TDI kinetic parameter values (KI and kinact). Crizotinib in vitro inhibitory effect on CYP3A (i.e., RI) using midazolam as a probe substrate was negligible with IC50 value of >30 μM (http://www.accessdata.fda.gov/drugsatfda_docs/nda/2011/202570Orig1s000ClinPharmR.pdf). Crizotinib EI parameters (Emax and EC50) estimated from three individual cryopreserved HEP were normalized to 2.4-fold and 0.84 μM, respectively, with the rifampin data (mean Emax of 90-fold and EC50 of 0.57 μM) by the induction calibrator of prediction tool box implemented in Simcyp. These EI parameters were also incorporated into both CRZ-HLM and CRZ-HSP files.

View this table:
TABLE 1

Physicochemical and pharmacokinetic parameters of crizotinib used for DDI prediction

Simcyp simulations of crizotinib plasma concentration-time profiles were performed by a full PBPK model with a first-order absorption rate constant (ka) based upon the clinically observed plasma concentrations in healthy volunteers (n = 14) after a single 50-mg intravenous 2-hour infusion or a single 250-mg oral administration. These single-dose clinical studies of crizotinib were conducted in a crossover design with a washout period of at least 14 days (http://www.accessdata.fda.gov/drugsatfda_docs/nda/2011/202570Orig1s000ClinPharmR.pdf). In the Simcyp absorption model, ka values were set to predict clinically observed time to reach maximum plasma concentration (tmax), and an unbound fraction in the gut (fu,gut) was assumed to be equal to an unbound fraction in blood (fu,blood) calculated from fu,plasma and blood-to-plasma ratio (Rbp). The blood flow term (Qgut) of crizotinib in the Qgut model was estimated to be 4.0 L/h from crizotinib physicochemical properties. Using a retrograde model implemented in Simcyp, hepatic intrinsic clearance (CLint) was back-calculated from the in vivo intravenous plasma clearance (47 L/h) determined in the single intravenous infusion study, as mentioned above. Crizotinib renal clearance was negligible in humans, and the fraction metabolized by CYP3A4 (fm,CYP3A4) was suggested to be 0.8 based on the in vitro CYP phenotyping and the human mass-balance study with [14C]crizotinib (Johnson et al., 2011a, 2011b); therefore, 80% of the back-calculated CLint was assigned to CYP3A4-mediated CLint (194 μl/min/mg protein) with the remaining 20% as additional CLint (49 μl/min/mg protein). The steady-state volume of distribution (Vss) of crizotinib was predicted using the mathematical model 2 (Rodgers et al., 2005) implemented in Simcyp. To improve goodness-of-fit between predicted and observed plasma concentration-time profiles, crizotinib pKa values were adjusted from 5.4 and 8.9 (diprotic base) to 7.6 (monoprotic base) to predict the clinically observed Vss of 25 L/kg from the single intravenous infusion study (http://www.accessdata.fda.gov/drugsatfda_docs/nda/2011/202570Orig1s000ClinPharmR.pdf). Using the unadjusted pKa values of 5.4 and 8.9, the predicted crizotinib Vss was 7.0 L/kg, which resulted in poor fits to the clinically observed single-dose intravenous infusion and oral plasma concentration profiles. Although a minimal PBPK model-implemented Simcyp reasonably predicted the areas under the plasma concentration-time curve (AUC), the maximum plasma concentration (Cmax) was significantly underpredicted in both the single-dose intravenous infusion and oral plasma concentration profiles. Simulations for midazolam plasma concentration-time profiles were also performed by a full PBPK model using a modified Simcyp-default midazolam file with the adjusted ka of 2 h−1 and pKa values of 7.4 (ampholyte). Mathematical model 2 implemented in Simcyp was used to predict midazolam Vss. These adjustments were required to simulate comparable midazolam plasma concentration-time profiles to the clinically observed results in patients (n = 14) after a single oral administration of midazolam without coadministration of crizotinib (see above). In addition to these adjustments, the predicted liver to plasma partition coefficients (kp) were set to unity from 22 and 1.3 for crizotinib and midazolam, respectively. These changes resulted in minimal effects on the predicted Vss values for both crizotinib (from 25.3 to 24.8 L/kg) and midazolam (from 1.0 to 1.1 L/kg).

Clinical trial simulation in Simcyp was performed with a virtual population of healthy volunteers in 8 trials of 10 subjects, each aged 18–65 years with a female/male ratio of 0.34, whose CYP3A4 degradation rate constant (kdeg) was 0.019 h−1 in the liver and 0.030 h−1 in the gastrointestinal (GI) tract. The output sampling interval in Simcyp simulation tool box was set to 0.2 h in all simulations. Trial designs used were as follows:

  • Trial 1: A single oral dose of 250 mg crizotinib was administered on day 1; crizotinib plasma concentration was simulated for 7 days.

  • Trial 2: Multiple doses of crizotinib, 250 mg twice daily with an interval of 12 hours, were orally administered for 28 days; crizotinib plasma concentration was simulated during crizotinib repeated administration.

  • Trial 3: A single oral dose of 2 mg midazolam was coadministered on day 28 with repeated oral doses of crizotinib (250 mg twice daily with an interval of 12 hours) for 28 days; plasma concentrations of midazolam and crizotinib were simulated during crizotinib repeated administration.

In these simulations, Cmax, tmax, and AUC0-τ were obtained from Simcyp output, whereas AUC0-∞ was calculated from simulated plasma concentrations using the linear trapezoidal rule:Embedded Image(5)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.

Geometric means of PK parameters in eight simulation trials were compared with the clinically observed geometric mean to assess Simcyp prediction accuracy. In addition to the Simcyp outputs of 5th and 95th percentiles of all 80 subjects, the geometric mean values of 5th and 95th percentiles among eight simulation trials were calculated by Microsoft Excel 2007 (Microsoft, Redmond, WA) to compare with the clinically observed 5th and 95th percentiles in the crizotinib-midazolam interaction study.

Sensitivity Analysis for Crizotinib Hepatic kp and CYP3A4 kdeg with Simcyp.

The sensitivity analyses for crizotinib hepatic kp value ranging from 1 to 24 and hepatic CYP3A4 kdeg from 0.693 to 0.00693 h−1 (i.e., t1/2 of 1–100 h) were separately performed in the prediction of crizotinib-midazolam interaction with Simcyp using CRZ-HSP. Clinical trial simulation in Simcyp was the same as trial 3, as described above, and the PK parameters (e.g., Cmax, tmax, AUC0-τ, and AUC0-∞) and geometric mean values of 5th and 95th percentiles among eight simulation trials were either obtained or calculated from Simcyp outputs, as described above.

Prediction of Crizotinib-Midazolam Interaction with a Static Mathematical Model.

A static mathematical model (Fahmi et al., 2008) was used to predict the fold increase in midazolam AUC0-∞ (AUCR) following coadministration with crizotinib:Embedded Image(6)Embedded Image(7)Embedded Image(8)Embedded Image(9)where A, B, and C represent TDI, EI, and RI, respectively, and the subscripts h and g denote liver and GI tract, respectively.

Crizotinib input parameters and CYP3A4 kdeg values in the liver and GI tract used for the static model were exactly the same as Simcyp input parameters to compare in vivo DDI predictions between the dynamic and static models. Therefore, the proposed empirical scaling factor for induction (i.e., d) in the static model was set to unity since the EI parameters were already normalized with rifampin data for the dynamic model.

The availability in the GI tract (Fg) and fm,CYP3A4 for midazolam were set as 0.63 and 0.93, respectively (Ernest et al., 2005; Obach et al., 2006). Crizotinib steady-state unbound average plasma concentrations (Cave,u), which were calculated from AUC0-τ divided by the dosing interval of 12 hours, followed by the correction for fu,plasma, were used as the hepatic inhibitor concentrations ([I]h) in the static model (as consistent with a full PBPK model with hepatic kp of unity). The intestinal inhibitor concentrations ([I]g) were calculated from crizotinib parameters summarized in Table 1 using the following equation (Rostami-Hodjegan and Tucker, 2004):Embedded Image(10)where D, ka, Fa, fu,gut, and Qgut represent the dose amount per dose (µmol), the first-order absorption rate constant (h−1), the fraction of dose absorbed, the unbound fraction in the GI tract, and the enterocyte blood flow (L/h), respectively.

Results

In Vitro Effects of Crizotinib on CYP3A.

Crizotinib TDI kinetic parameters for CYP3A were determined in HLM and HSP using midazolam as a CYP3A probe substrate. Estimated KI,app and kinact in HLM were 3.0 µM and 6.9 h−1, respectively (Fig. 2). Crizotinib IC50 curves in HSP were adequately described by the irreversible model (Eq. 3) with KI,app of 9.6 µM and kinact of 0.78 h−1 (Fig. 3). The estimates for KI in HLM and HSP following the correction for in vitro nonspecific binding (e.g., fu,mic and fu,plasma, respectively) were 0.37 and 0.89 µM, respectively. Thus, the KI value in HSP was approximately 3-fold higher than that in HLM, whereas the kinact from HSP was ∼9-fold lower than that from HLM, resulting in the 20-fold difference in the inactivation efficiency (kinact/KI) between HLM and HSP (19 and 0.87 h−1× µM−1, respectively). These KI and kinact values were used for Simcyp simulation (Table 1).

Fig. 2.
Fig. 2.

Time-dependent inhibition by crizotinib for CYP3A in HLM. The x-axis represents crizotinib concentration in micromolars, and the y-axis represents the observed apparent inactivation rate constant, kobs (⬤), in reciprocal hours. The solid line represents the fitting curve (-). Data are the mean of duplicate determination.

Fig. 3.
Fig. 3.

Time-dependent inhibition by crizotinib for CYP3A in HSP. The x-axis represents crizotinib concentration in micromolars on a logarithmic scale, and the y-axis represents the ratio of remaining CYP3A enzyme activity to the vehicle control at the end of 35-minute incubation with midazolam following 0 (⬤)-, 10 (◆⬥)-, or 20 (⬛)-min preincubation with crizotinib alone. The dashed lines represent the best fitting curves for each incubation period by the irreversible model. Data are the mean of duplicate determination.

In cryopreserved HEP from three donors, crizotinib concentration-dependently induced CYP3A4 mRNA expression with estimated Emax of 6.4- to 29-fold and EC50 of 0.47–3.1 μM. It should be noted that relatively low cell viability (<0.7 ratio to vehicle control sample) was observed at higher crizotinib concentrations (≥5 μM); thus, the accuracy of EI parameter estimation might have been impacted. Following the normalization with rifampin data (positive control), mean crizotinib Emax of 2.4-fold and EC50 of 0.84 μM were used for Simcyp simulation (Table 1).

Predicted Crizotinib Plasma Concentrations with Simcyp.

Clinically observed crizotinib Cmax, tmax, and AUC0-∞ were 87 ng/ml, 4 hours, and 1817 ng⋅h/ml, respectively, in cancer patients (n = 8) after a single oral administration of 250 mg crizotinib (http://www.accessdata.fda.gov/drugsatfda_docs/nda/2011/202570Orig1s000ClinPharmR.pdf). By the Simcyp simulation with CRZ-HLM, the predicted Cmax (89 ng/ml) was within 10% of the observed value, whereas the AUC0-∞ (3901 ng⋅h/ml) was overpredicted by 2.1-fold (Table 2). As shown in Fig. 4A, the predicted crizotinib plasma concentrations were slightly, but consistently higher than the observed concentrations after tmax. In contrast, the predicted AUC0-∞ (1745 ng⋅h/ml) by Simcyp with CRZ-HSP was consistent with the observed value (<10%), whereas the predicted Cmax (42 ng/ml) was underpredicted by ∼2-fold (Table 2). It remained unclear why the Cmax was underpredicted. One of the possible reasons could be the effect of multidrug-resistance transport protein, P-glycoprotein, on the absorption since crizotinib was reported to be a substrate of P-glycoprotein (http://www.accessdata.fda.gov/drugsatfda_docs/nda/2011/202570Orig1s000ClinPharmR.pdf). The predicted crizotinib plasma concentrations with CRZ-HSP were in good agreement with the observed profiles after tmax (Fig. 4B).

View this table:
TABLE 2

Clinically observed and Simcyp-predicted pharmacokinetic parameter estimates of crizotinib in humans after a single or 28-day repeated oral administration of crizotinib

Fig. 4.
Fig. 4.

Observed and Simcyp-predicted plasma concentrations of crizotinib in patients after a single or 28-day repeated oral administration of crizotinib. Crizotinib was orally administered at a single dose of 250 mg (n = 8) or multiple doses of 250 mg twice daily (n = 5) to cancer patients. The x-axis represents the time after a single (A and B) or 28-day repeated (C and D) oral administration of crizotinib in hours, and the y-axis represents the observed (○) and Simcyp-predicted (—) plasma concentrations of crizotinib in nanograms per milliliter on a logarithmic scale using TDI parameters from HLM (A and C) or HSP (B and D). The observed and predicted plasma concentrations are expressed as mean ± SD and mean with 90% CI (dashed line), respectively.

Clinically observed crizotinib steady-state Cmax, tmax, and AUC0-τ were 328 ng/mL, 4 hours, and 3054 ng⋅h/ml, respectively, in cancer patients (n = 5) after 28-day repeated oral administration of 250 mg crizotinib twice daily (see above). The predicted Cmax (543 ng/ml) and AUC0-τ (6494 ng⋅h/ml) by Simcyp with CRZ-HLM were overpredicted by 1.7- and 2.1-fold, respectively (Table 2; Fig. 4C). In contrast, the Simcyp-predicted crizotinib steady-state plasma concentrations with CRZ-HSP were in good agreement with the observed concentrations (Fig. 4D). The predicted Cmax (273 ng/ml) and AUC0-τ (3258 ng⋅h/ml) were within 20% of the observed values (Table 2). Additionally, the effect of crizotinib TDI or EI on crizotinib steady-state plasma concentrations was examined in a virtual population of healthy volunteers by Simcyp using CRZ-HSP with/without TDI and/or EI parameters. The predicted Cmax (288 ng/ml) and AUC0-τ (3432 ng⋅h/ml) with only TDI parameters were comparable to the predicted values (273 ng/ml and 3258 ng⋅h/ml, respectively) with both TDI and EI parameters (within 5%), whereas those with only EI parameters (128 ng/ml and 1519 ng⋅h/ml, respectively) were comparable to the predicted values (133 ng/ml and 1587 ng⋅h/ml, respectively) without TDI and EI parameters (within 5%). The predicted Cmax and AUC0-τ with only TDI parameters were ∼2-fold higher than the predicted values without TDI and EI parameters. Based on these comparisons, crizotinib steady-state plasma concentrations appeared to be accumulated by 2-fold due to TDI with a minimal EI effect.

Predicted Crizotinib-Midazolam Interaction with Simcyp.

Clinically observed midazolam Cmax, tmax, and AUC0-∞ in cancer patients (n = 14) after a single oral administration of 2 mg midazolam were reported to be 13 ng/ml, 0.5 hours, and 32 ng⋅h/ml, respectively, without coadministration of crizotinib (http://www.accessdata.fda.gov/drugsatfda_docs/nda/2011/202570Orig1s000ClinPharmR.pdf). After 28-day oral administration of twice daily doses of 250 mg crizotinib in cancer patients (n = 8), midazolam Cmax and AUC0-∞ increased to 26 ng/ml and 117 ng⋅h/ml, respectively. Thus, the fold increase in midazolam Cmax (CmaxR) and AUCR with coadministration of crizotinib was 2.0 and 3.7, respectively, suggesting crizotinib was a moderate CYP3A inhibitor (http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ucm292362.pdf). The 90% confidence intervals (CI) were 1.4–2.9 for CmaxR and 2.6–5.1 for AUCR. When the crizotinib-midazolam interaction was predicted by Simcyp with CRZ-HLM, the predicted midazolam plasma concentration-time profiles were much higher than the observed profiles (Fig. 5A). The midazolam CmaxR and AUCR were 3.0 and 21, respectively (Table 3; Fig. 6A), resulting in the 1.5- and 5.6-fold overpredictions of CmaxR and AUCR, respectively. In contrast, the predicted midazolam plasma concentration-time profiles with CRZ-HSP were relatively consistent with the observed profiles (Fig. 5B). The midazolam CmaxR and AUCR (1.9 and 3.6, respectively) were in good agreement with the clinically observed results (Table 3; Fig. 6B). The 90% CI for predicted Cmax (1.4–2.9) was relatively consistent with the observed CI, whereas that for predicted AUC0-∞ (1.6–8.8) was slightly larger than the observed result.

Fig. 5.
Fig. 5.

Observed and Simcyp-predicted plasma concentrations of midazolam in patients before and after 28-day repeated oral administration of crizotinib. Midazolam (2 mg) was orally administered to cancer patients before (n = 14) and after (n = 8) twice daily doses of 250 mg crizotinib for 28 days. The x-axis represents the time after a single oral administration of midazolam in hours, and the y-axis represents the observed plasma concentrations of midazolam before (○) and after (⬤) oral administration of crizotinib in nanograms per milliliter on a logarithmic scale along with the predicted plasma concentrations before (- - -) and after (—) crizotinib administration by Simcyp with TDI parameters from HLM (A) or HSP (B). The observed and predicted plasma concentrations are expressed as mean ± SD and mean, respectively.

View this table:
TABLE 3

Clinically observed and Simcyp-predicted pharmacokinetic parameter estimates of midazolam in humans before and after 28-day repeated oral administration of crizotinib twice daily doses of 250 mg

Fig. 6.
Fig. 6.

Clinically observed and Simcyp-predicted fold increase in midazolam AUC0-∞ with coadministration of crizotinib. Midazolam (2 mg) was orally administered to cancer patients before (n = 14) and after (n = 8) twice daily doses of 250 mg crizotinib for 28 days. The x-axis represents the observed (Obs) and predicted 8 clinical trials (1–8) of 10 individuals by Simcyp with TDI parameters from HLM (A) or HSP (B), and the y-axis represents the fold increase in midazolam AUC0-∞ (AUCR) with coadministration of crizotinib. The observed and predicted AUCR values in each trial are expressed as geometric mean with 90% CI. The solid and dashed lines represent geometric mean and 90% CI, respectively, of all trials.

Sensitivity Analysis for Crizotinib Hepatic kp and CYP3A4 kdeg with Simcyp.

The sensitivity analysis for crizotinib hepatic kp values of 1–24 revealed that the midazolam AUCR with coadministration of crizotinib varied markedly from 3.6 to 13, while the predicted crizotinib AUC0-τ values were within 2-fold (3258–5434 ng⋅h/ml) (Supplemental Fig. 1). Therefore, DDI prediction with a full PBPK model in Simcyp was highly sensitive to crizotinib hepatic kp value even though the PBPK model was basically constructed with perfusion-limited compartments in the absence of any diffusion barrier.

In the sensitivity analysis for hepatic CYP3A4 kdeg values of 0.693–0.00693 h−1 (i.e., t1/2 of 1–100 h), the predicted crizotinib steady-state Cmax and Cave varied from 141 to 385 ng/ml and 140 to 383 ng/ml, respectively, indicating that the CYP3A4 t1/2 of 25–35 hours yielded the best prediction for crizotinib Cave (Supplemental Fig. 2). The midazolam CmaxR and AUCR varied from 1.3 to 2.4 and 1.3 to 7.7, respectively. Thus, the AUCR was more sensitive to the hepatic kdeg values relative to CmaxR. The t1/2 of 30–40 hours yielded the best prediction of the midazolam AUCR. Therefore, the hepatic kdeg value of 0.019 h−1 (t1/2 of 36 h) used in this study was within a range of reasonable predictions on both the crizotinib steady-state plasma concentrations and crizotinib-midazolam interaction.

Predicted Crizotinib-Midazolam Interaction by Static Models.

The predicted midazolam AUCR values with the static mathematical model using either clinically observed or Simcyp-predicted crizotinib Cave,u are summarized in Table 4. Using the clinically observed crizotinib Cave,u, the predicted midazolam AUCR values with HLM-TDI and HSP-TDI were 17 and 3.2, respectively. Thus, the midazolam AUCR with HLM-TDI was overpredicted by ∼5-fold, whereas the prediction with HSP-TDI was consistent with the observed AUCR. In the static model, the midazolam AUCR in the liver (11–12) with HLM-TDI was much higher than that with HSP-TDI (2.5–2.6), whereas the AUCR in the intestines was relatively comparable between HLM-TDI and HSP-TDI (1.6 and 1.3, respectively). This comparable intestinal AUCR between HLM-TDI and HSP-TDI appeared to be largely due to midazoalm Fg of 0.63, which resulted in the maximal effect of 1.6-fold increase. Using the Simcyp-predicted crizotinib Cave,u by HLM-TDI and HSP-TDI, the predicted midazolam AUCR values were 19 and 3.3, respectively. Thus, the prediction results of crizotinib-midazolam interaction with the static model using the Simcyp-predicted crizotinib Cave,u were relatively consistent with those with the Simcyp full PBPK models with kp of unity.

View this table:
TABLE 4

Predicted fold increase in midazolam AUC0-∞ by the static models using crizotinib steady-state average plasma concentration

In addition, the predicted midazolam AUCR values with Simcyp static model (Rss) were 13 and 2.3 with HLM-TDI and HSP-TDI, respectively. These AUCR values were ∼1.5-fold lower than the predicted values with the Simcyp full PBPK models (i.e., 21 and 3.6, respectively). The difference in the predicted ratios between the Rss and full PBPK predictions appeared to be largely due to the crizotinib TDI effects on the accumulation of crizotinib steady-state plasma concentrations, which was not taken into account for the Simcyp-Rss prediction, in which the inhibitor concentration was calculated from the input PK parameters such as D and CL.

Discussion

In the present study, the estimated in vitro inactivation efficiency (kinact/KI) for crizotinib TDI potency was ∼20-fold lower in HSP relative to HLM. Consistent with the present results, the known CYP3A inhibitors (e.g., diltiazem, erythromycin, and verapamil) were also reported to be less potent in HEP compared with HLM, and the difference in TDI potency has yielded an impact on in vivo DDI prediction (Xu et al., 2009; Chen et al., 2011). The advantages in utilization of HEP over HLM are considered to be that HEP are intact cellular systems containing not only a full complement of phase I/II–metabolizing enzymes, but also functional cell membranes. The estimated kinact in HEP is generally considered an apparent hybrid kinetic parameter consisting of the following: 1) the rate of diffusion through the cell membrane; 2) the total metabolic consumption rate (including sequential metabolism); and 3) the intrinsic enzyme inactivation rate. Therefore, these factors could be important for some drugs showing substantially different intra- versus extracellular concentrations and/or multiple metabolic pathways, particularly those associated with non-CYPs (Zhao, 2008). Crizotinib was primarily metabolized by CYP3A (Johnson et al., 2011a), and its passive permeability was relatively low (∼1 × 10−6 cm/s) in the recently reported permeability assay system using low-efflux Madin-Darby canine kidney cells (Di et al., 2011). Crizotinib was also reported to be a substrate of P-glycoprotein, as mentioned above. Thus, these factors could potentially lead to differences in intra- versus extracellular concentrations, and, correspondingly, to differences in the estimated TDI potency between HLM versus HSP.

Mathematical models used in the Simcyp-PBPK model are the well-stirred and Qgut models in the liver and GI tract, respectively (Yang et al., 2007; Jamei et al., 2009). The well-stirred model assumes that drug distribution into liver is perfusion limited without any diffusion delay; therefore, unbound drug concentration within liver is in equilibrium with unbound drug concentration in liver outlet (i.e., Csys,u). Accordingly, we intended to use crizotinib Csys,u as the inhibitor concentration for DDI prediction by adjusting hepatic kp value (i.e., unity), as described in Materials and Methods. Consistent with our findings, other investigators (Obach et al., 2007; Fahmi et al., 2009; Boulenc and Barberan, 2011) reported that the use of Csys,u (e.g., Cmax,u or Cave,u) as the inhibitor concentrations yielded the most accurate DDI predictions for TDI and EI by the static models. When the projected/calculated Cinlet,u was used, the DDI predictions for TDI and EI were generally overpredicted, while those for RI were more accurately predicted. These findings may suggest that RI mainly occurs during the first-pass metabolism process, whereas TDI and EI continue to take place beyond the first-pass metabolism. Consistently, the clinically observed crizotinib-midazolam interaction was more significant on midazolam AUCR than CmaxR, and the DDI prediction using crizotinib Csys,u yielded a reasonable prediction for both midazolam AUCR and CmaxR. Thus, crizotinib Csys,u appears to be a more appropriate inhibitor concentration than Cinlet,u for the prediction of crizotinib-midazolam interaction.

The Qgut model in the GI tract is basically constructed by the well-stirred model, in which the blood flow term (i.e., Qgut) is a hybrid parameter consisting of a compound-dependent permeability through enterocyte membrane and a physiologic villous blood flow. Crizotinib Qgut value was estimated to be 4 L/h based on a calculated projected human jejunum permeability (Peff,man) value of 0.58 × 10−4 cm/s derived from its physicochemical properties. This Peff,man value was within a moderate range among 17 compounds reported (Yang et al., 2007). Since crizotinib Fg was estimated to be ∼0.9 by the Qgut model, its moderate Peff,man did not appear to be a limiting factor for the first-pass metabolism in the GI tract. In contrast, the experimentally measured crizotinib in vitro permeability (1 × 10−6 cm/s) seemed to be lower than the value expected from its physicochemical properties, suggesting that the calculated Peff,man value might be overpredicted. However, crizotinib permeability can be taken into account to estimate TDI kinetic parameters in HSP because this system contains intact cell membranes. We therefore believe that a perfusion-limited model is appropriate for PK and/or DDI prediction when the kinetic parameters were determined in an in vitro system containing cell membranes. On the other hand, a diffusion rate through the cell membrane would likely be an important factor for PBPK model of low permeable compounds when their kinetic parameters were determined in assay systems without functional cell membranes (e.g., HLM). In such a case, a diffusion-limited model would be likely required for their PK and/or DDI prediction. Moreover, when compounds are metabolized by multiple enzymes, such as microsomal and cytosolic enzymes, HEP would be more appropriate to determine TDI kinetic parameters by taking into account an overall in vivo metabolic pathway. It has been reported that multiple enzymes (e.g., CYP3A and aldehyde oxidase) were involved in the crizotinib metabolism (Johnson et al., 2011a); therefore, this might be another factor that contributed to the difference in the TDI kinetics between HLM and HSP. These considerations are graphically summarized in Fig. 7. For PBPK model-based DDI prediction, the selection of either a perfusion- or diffusion-limited model should be carefully considered based upon each inhibitor’s property and in vitro assay system used. It should also be important to select/simulate appropriate surrogate concentrations of both object and precipitant drugs as the unbound concentrations at target sites. In vivo DDI predictions are expected to be more accurate based on plasma concentration profiles predicted by the PBPK model compared with a fixed inhibitor concentration (e.g., Cmax,u and Cave,u) used for the static model. In the prediction of crizotinib-midazolam interaction, it would be worth noting that the Simcyp minimal and full PBPK models (with hepatic kp of unity) yielded comparable DDI predictions when the predicted crizotinib concentrations in plasma were comparable between these models (Supplemental Table 1). Moreover, these DDI predictions were roughly comparable to the predictions from the static models such as a Simcyp-static model (i.e., Rss) and a static mathematical model (Supplemental Table 1). Therefore, there might not be significant differences in DDI prediction between dynamic and static models if the factors related to DDI were appropriately incorporated into these models. However, these comparable predictions for crizotinib-midazolam interaction could be largely due to the relatively flat crizotinib plasma concentration-time profiles (i.e., Cmax ≈ Cave) during dosing interval, as shown in Fig. 4.

Fig. 7.
Fig. 7.

Proposed general scheme of PBPK model-based in vivo TDI prediction.

Several other factors related to TDI mechanism should also be considered for the in vivo DDI prediction. Quantitative measurement of in vivo CYP3A4 kdeg in humans remains a challenge as kdeg is often associated with an uncertainty in the DDI prediction (Obach et al., 2007; Rowland Yeo et al., 2011). Not surprisingly, a wide range of hepatic kdeg values has been reported for CYP3A4 as turnover t1/2 of 10–140 h (Yang et al., 2008; Grimm et al., 2009; Xu et al., 2009; Chen et al., 2011). We used the Simcyp default kdeg values of 0.019 h−1 (t1/2 = 36 h) in the liver and 0.030 h−1 (t1/2 = 23 h) in the intestines. Based on the sensitivity analysis for hepatic CYP3A4 kdeg values ranging from 0.693 to 0.00693 h−1, that is, t1/2 of 1–100 hours, the hepatic kdeg value of 0.019 h−1 (t1/2 of 36 hours) used in this study was within a range of reasonable predictions on both the crizotinib steady-state plasma concentrations and crizotinib-midazolam interaction (Supplemental Fig. 2). Another important parameter is the fraction metabolized by a drug-metabolizing enzyme, i.e., fm,CYP3A4. The estimated crizotinib fm,CYP3A4 was ∼0.8 based on the in vitro CYP phenotyping study and the human mass-balance study with [14C]crizotinib (Johnson et al., 2011a, 2011b); the fm,CYP3A4 of 0.8 used in the present study yielded reasonable PK/DDI predictions.

Despite these reasonable DDI predictions, the method used in the present study included potential limitations that should be addressed. First, neither in vitro nonspecific binding nor any decrease in crizotinib concentration during incubation was considered for the correction of in vitro crizotinib EI parameters (e.g., Emax and EC50), as previously reported (Fahmi et al., 2008). In contrast, the obtained TDI parameters in HLM and HSP were corrected for the in vitro nonspecific binding. Secondly, and possibly more importantly, the physiologic CYP3A4 kinetic (or systems) parameters (i.e., ksyn and kdeg) for the TDI prediction were assumed to remain constant in the presence of inhibitor, and an additional enzyme degradation pathway was introduced to reach new steady-state concentration as kobs (Eq. 1). This assumption may not be valid when EI occurs in parallel to TDI. In fact, there appears to be limited understanding of underlying simultaneous in vitro to in vivo extrapolation for TDI and EI (and its possible interplay). Moreover, although TDI and EI of CYP3A are known to occur in both liver and GI tract, their magnitudes may be different in these organs. Despite these limitations, a PBPK modeling approach for DDI prediction affords an opportunity to gain greater insight into the underlying mechanisms mediating these complex interactions as a function of time with the effects of intrinsic factors such as organ dysfunction, age, and genetics.

Acknowledgments

The Authors greatly acknowledge Deepak Dalvie, Sascha Freiwald, Lance Goulet, Caroline A. Lee, Danielle Smith and Evan B. Smith [Pharmacokinetics, Dynamics, and Metabolism (PDM), Pfizer, San Diego, CA] for the bioanalytical and biotransformation assays, and Sherri Boldt, Odette A. Fahmi and Gregory S. Walker (PDM, Pfizer, Groton, CT) for the bioanalytical and induction assays. The Authors also thank Weiwei Tan and Keith D. Wilner [Clinical Pharmacology, Pfizer, San Diego, CA] for valuable discussion about clinical pharmacokinetic data, and Odette A. Fahmi [PDM, Pfizer, Groton, CT], R. Scott Obach [PDM, Pfizer, Groton, CT], Bill J. Smith [PDM, Pfizer, San Diego, CA], and Bhasker Shetty [PDM, Pfizer, San Diego, CA] for excellent inputs for the draft manuscript.

Authorship Contributions

Participated in research design: Mao, Yamazaki.

Conducted experiments: Mao, Shen, Yamazaki.

Performed data analysis: Mao, Yamazaki.

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

Footnotes

Abbreviations

AUC
area under the plasma concentration-time curve
CI
confidence interval
Cinlet,u
unbound hepatic inlet concentration
CLint
hepatic intrinsic clearance
Cmax
maximum plasma concentration
Csys,u
unbound systemic concentration
DDI
drug-drug interaction
EI
enzyme induction
FDA
Food and Drug Administration
fm
fraction metabolized by an enzyme
fu
unbound fraction
GI
gastrointestinal
HEP
human hepatocytes
HLM
human liver microsomes
HSP
human hepatocytes suspended in human plasma
kdeg
degradation rate constant
Ki
inhibition constant
KI
inactivation constant
KI,app
apparent KI
kinact
maximum inactivation rate constant
kobs
apparent inactivation rate constant
kp
tissue to plasma partition coefficient
LC-MS/MS
liquid-chromatography tandem mass spectrometry
PBPK
physiologically based pharmacokinetics
Peff,man
projected human jejunum permeability
PK
pharmacokinetics
Rbp
blood-to-plasma ratio
RI
reversible inhibition
TDI
time-dependent inhibition
tmax
time to reach maximum plasma concentration
Vss
volume of distribution at steady state
  • Received September 18, 2012.
  • Accepted November 5, 2012.

References

  1. ↵
    1. Boulenc X,
    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
  2. ↵
    1. Chen Y,
    2. Liu L,
    3. Monshouwer M,
    4. Fretland AJ
    (2011) Determination of time-dependent inactivation of CYP3A4 in cryopreserved human hepatocytes and assessment of human drug-drug interactions. Drug Metab Dispos 39:2085–2092.
    OpenUrlAbstract/FREE Full Text
  3. ↵
    1. Chu V,
    2. Einolf HJ,
    3. Evers R,
    4. Kumar G,
    5. Moore D,
    6. Ripp S,
    7. Silva J,
    8. Sinha V,
    9. Sinz M,
    10. Skerjanec A
    (2009) In vitro and in vivo induction of cytochrome p450: a survey of the current practices and recommendations: a pharmaceutical research and manufacturers of America perspective. Drug Metab Dispos 37:1339–1354.
    OpenUrlAbstract/FREE Full Text
  4. ↵
    1. Cui JJ,
    2. Tran-Dubé M,
    3. Shen H,
    4. Nambu M,
    5. Kung PP,
    6. Pairish M,
    7. Jia L,
    8. Meng J,
    9. Funk L,
    10. Botrous I,
    11. et al.
    (2011) Structure based drug design of crizotinib (PF-02341066), a potent and selective dual inhibitor of mesenchymal-epithelial transition factor (c-MET) kinase and anaplastic lymphoma kinase (ALK). J Med Chem 54:6342–6363.
    OpenUrlCrossRefPubMed
  5. ↵
    1. Di L,
    2. Keefer C,
    3. Scott DO,
    4. Strelevitz TJ,
    5. Chang G,
    6. Bi YA,
    7. Lai Y,
    8. Duckworth J,
    9. Fenner K,
    10. Troutman MD,
    11. Obach RS
    (2012) Mechanistic insights from comparing intrinsic clearance values between human liver microsomes and hepatocytes to guide drug design. Eur J Med Chem. [published ahead of print].
  6. ↵
    1. Di L,
    2. Whitney-Pickett C,
    3. Umland JP, Zhang H, Zhang X, Gebhard DF, Lai Y, Federico 3rd JJ, Davidson RE, Smith R,
    4. et al.
    (2011) Development of a new permeability assay using low-efflux MDCKII cells. J Pharm Sci 100:4974–4985.
    OpenUrlCrossRefPubMed
  7. ↵
    1. Ernest CS 2nd.,
    2. Hall SD,
    3. Jones DR
    (2005) Mechanism-based inactivation of CYP3A by HIV protease inhibitors. J Pharmacol Exp Ther 312:583–591.
    OpenUrlAbstract/FREE Full Text
  8. ↵
    1. Fahmi OA,
    2. Hurst S,
    3. Plowchalk D,
    4. Cook J,
    5. Guo F,
    6. Youdim K,
    7. Dickins M,
    8. Phipps A,
    9. Darekar A,
    10. Hyland R,
    11. Obach RS
    (2009) Comparison of different algorithms for predicting clinical drug-drug interactions, based on the use of CYP3A4 in vitro data: predictions of compounds as precipitants of interaction. Drug Metab Dispos 37:1658–1666.
    OpenUrlAbstract/FREE Full Text
  9. ↵
    1. Fahmi OA,
    2. Maurer TS,
    3. Kish M,
    4. Cardenas E,
    5. Boldt S,
    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
  10. ↵
    1. Grimm SW,
    2. Einolf HJ,
    3. Hall SD,,
    4. He K,
    5. Lim HK,
    6. Ling KH,
    7. Lu C,
    8. Nomeir AA,
    9. Seibert E,
    10. Skordos KW,
    11. et al.
    (2009) The conduct of in vitro studies to address time-dependent inhibition of drug-metabolizing enzymes: a perspective of the pharmaceutical research and manufacturers of America. Drug Metab Dispos 37:1355–1370.
    OpenUrlAbstract/FREE Full Text
  11. ↵
    1. Jamei M,
    2. Marciniak S,
    3. Feng K,
    4. Barnett A,
    5. Tucker G,
    6. Rostami-Hodjegan A
    (2009) The Simcyp population-based ADME simulator. Expert Opin Drug Metab Toxicol 5:211–223.
    OpenUrlCrossRefPubMed
  12. ↵
    Johnson TR, Zhou S, Lee C et al. (2011a) Identification of the human enzymes involved in the oxidative metabolism of crizotinib, in 17th North American Regional ISSX Meeting; 2011 Oct 16–20; Atlanta, GA. International Society for the Study of Xenobiotics, Washington, DC.
  13. ↵
    Johnson TR, Goulet L, Smith EB et al. (2011b) Metabolism, excretion and pharmacokinetics of [14C]crizotinib following oral administration to healthy subjects, in 17th North American Regional ISSX Meeting; 2011 Oct 16–20; Atlanta, GA. International Society for the Study of Xenobiotics, Washington, DC.
  14. ↵
    1. Jones HM,
    2. Parrott N,
    3. Jorga K,
    4. Lavé T
    (2006) A novel strategy for physiologically based predictions of human pharmacokinetics. Clin Pharmacokinet 45:511–542.
    OpenUrlCrossRefPubMed
  15. ↵
    1. Kanamitsu S,
    2. Ito K,
    3. Green CE,
    4. Tyson CA,
    5. Shimada N,
    6. Sugiyama Y
    (2000) Prediction of in vivo interaction between triazolam and erythromycin based on in vitro studies using human liver microsomes and recombinant human CYP3A4. Pharm Res 17:419–426.
    OpenUrlCrossRefPubMed
  16. ↵
    1. Kirby BJ,
    2. Collier AC,
    3. Kharasch ED,
    4. Whittington D,
    5. Thummel KE,
    6. Unadkat JD
    (2011) Complex drug interactions of HIV protease inhibitors 1: inactivation, induction, and inhibition of cytochrome P450 3A by ritonavir or nelfinavir. Drug Metab Dispos 39:1070–1078.
    OpenUrlAbstract/FREE Full Text
  17. ↵
    1. Lavé T,
    2. Parrott N,
    3. Grimm HP,
    4. Fleury A,
    5. Reddy M
    (2007) Challenges and opportunities with modelling and simulation in drug discovery and drug development. Xenobiotica 37:1295–1310.
    OpenUrlCrossRefPubMed
  18. ↵
    1. Mao J,
    2. Mohutsky MA,
    3. Harrelson JP,
    4. Wrighton SA,
    5. Hall SD
    (2011) Prediction of CYP3A-mediated drug-drug interactions using human hepatocytes suspended in human plasma. Drug Metab Dispos 39:591–602.
    OpenUrlAbstract/FREE Full Text
  19. ↵
    1. Mao J,
    2. Mohutsky MA,
    3. Harrelson JP,
    4. Wrighton SA,
    5. Hall SD
    (2012) Predictions of cytochrome P450-mediated drug-drug interactions using cryopreserved human hepatocytes: comparison of plasma and protein-free media incubation conditions. Drug Metab Dispos 40:706–716.
    OpenUrlAbstract/FREE Full Text
  20. ↵
    1. Mayhew BS,
    2. Jones DR,
    3. Hall SD
    (2000) An in vitro model for predicting in vivo inhibition of cytochrome P450 3A4 by metabolic intermediate complex formation. Drug Metab Dispos 28:1031–1037.
    OpenUrlAbstract/FREE Full Text
  21. ↵
    1. Nestorov I
    (2007) Whole-body physiologically based pharmacokinetic models. Expert Opin Drug Metab Toxicol 3:235–249.
    OpenUrlCrossRefPubMed
  22. ↵
    1. Obach RS,
    2. Walsky RL,
    3. Venkatakrishnan K,
    4. Gaman EA,
    5. Houston JB,
    6. Tremaine LM
    (2006) The utility of in vitro cytochrome P450 inhibition data in the prediction of drug-drug interactions. J Pharmacol Exp Ther 316:336–348.
    OpenUrlAbstract/FREE Full Text
  23. ↵
    1. Obach RS,
    2. Walsky RL,
    3. Venkatakrishnan K
    (2007) Mechanism-based inactivation of human cytochrome p450 enzymes and the prediction of drug-drug interactions. Drug Metab Dispos 35:246–255.
    OpenUrlAbstract/FREE Full Text
  24. ↵
    1. Rodgers T,
    2. Leahy D,
    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
  25. ↵
    1. Rostami-Hodjegan A,
    2. Tucker G
    (2004) ‘In silico’ simulations to assess the ‘in vivo’ consequences of ‘in vitro’ metabolic drug–drug interactions. Drug Discov Today 1:441–448.
    OpenUrl
  26. ↵
    1. Rowland Yeo K,
    2. Walsky RL,
    3. Jamei M,
    4. Rostami-Hodjegan A,
    5. Tucker GT
    (2011) Prediction of time-dependent CYP3A4 drug-drug interactions by physiologically based pharmacokinetic modelling: impact of inactivation parameters and enzyme turnover. Eur J Pharm Sci 43:160–173.
    OpenUrlCrossRefPubMed
  27. ↵
    1. Shimada T,
    2. Yamazaki H,
    3. Mimura M,
    4. Inui Y,
    5. Guengerich FP
    (1994) Interindividual variations in human liver cytochrome P-450 enzymes involved in the oxidation of drugs, carcinogens and toxic chemicals: studies with liver microsomes of 30 Japanese and 30 Caucasians. J Pharmacol Exp Ther 270:414–423.
    OpenUrlAbstract/FREE Full Text
  28. ↵
    1. Slaughter RL,
    2. Edwards DJ
    (1995) Recent advances: the cytochrome P450 enzymes. Ann Pharmacother 29:619–624.
    OpenUrlPubMed
  29. ↵
    1. Xu L,
    2. Chen Y,
    3. Pan Y,
    4. Skiles GL,
    5. Shou M
    (2009) Prediction of human drug-drug interactions from time-dependent inactivation of CYP3A4 in primary hepatocytes using a population-based simulator. Drug Metab Dispos 37:2330–2339.
    OpenUrlAbstract/FREE Full Text
  30. ↵
    1. Yamazaki S,
    2. Skaptason J,
    3. Romero D,
    4. Vekich S,
    5. Jones HM,
    6. Tan W,
    7. Wilner KD,
    8. Koudriakova T
    (2011) Prediction of oral pharmacokinetics of cMet kinase inhibitors in humans: physiologically based pharmacokinetic model versus traditional one-compartment model. Drug Metab Dispos 39:383–393.
    OpenUrlAbstract/FREE Full Text
  31. ↵
    1. Yang J,
    2. Jamei M,
    3. Yeo KR,
    4. Tucker GT,
    5. Rostami-Hodjegan A
    (2007) Prediction of intestinal first-pass drug metabolism. Curr Drug Metab 8:676–684.
    OpenUrlCrossRefPubMed
  32. ↵
    1. Yang J,
    2. Liao M,
    3. Shou M,
    4. Jamei M,
    5. Yeo KR,
    6. Tucker GT,
    7. Rostami-Hodjegan A
    (2008) Cytochrome p450 turnover: regulation of synthesis and degradation, methods for determining rates, and implications for the prediction of drug interactions. Curr Drug Metab 9:384–394.
    OpenUrlCrossRefPubMed
  33. ↵
    1. Zhao P
    (2008) The use of hepatocytes in evaluating time-dependent inactivation of P450 in vivo. Expert Opin Drug Metab Toxicol 4:151–164.
    OpenUrlCrossRefPubMed
  34. ↵
    1. Zhao P,
    2. Kunze KL,
    3. Lee CA
    (2005) Evaluation of time-dependent inactivation of CYP3A in cryopreserved human hepatocytes. Drug Metab Dispos 33:853–861.
    OpenUrlAbstract/FREE Full Text
  35. ↵
    1. Zimmerlin A,
    2. Trunzer M,
    3. Faller B
    (2011) CYP3A time-dependent inhibition risk assessment validated with 400 reference drugs. Drug Metab Dispos 39:1039–1046.
    OpenUrlAbstract/FREE Full Text
Back to top

In this issue

Download PDF
Article Alerts
Email Article
Citation Tools

Research ArticleArticle

Prediction of Crizotinib-Midazolam Interaction

Jialin Mao, Theodore R. Johnson, Zhongzhou Shen and Shinji Yamazaki
Drug Metabolism and Disposition February 1, 2013, 41 (2) 343-352; DOI: https://doi.org/10.1124/dmd.112.049114
Reddit logo Twitter logo Facebook logo Mendeley logo

Jump to section

Advertisement