Crizotinib [Xalkori; PF02341066; (R)-3-[1-(2,6-dichloro-3-fluoro-phenyl)-ethoxy]-5-(1-piperidin-4-yl-1H-pyrazol-4-yl)-pyridin-2-ylamine] is an orally available dual inhibitor of anaplastic lymphoma kinase (ALK) and hepatocyte growth factor receptor. The objectives of the present studies were to characterize: 1) the pharmacokinetic/pharmacodynamic relationship of crizotinib plasma concentrations to the inhibition of ALK phosphorylation in tumors, and 2) the relationship of ALK inhibition to antitumor efficacy in human tumor xenograft models. Crizotinib was orally administered to athymic nu/nu mice implanted with H3122 non–small-cell lung carcinomas or severe combined immunodeficient/beige mice implanted with Karpas299 anaplastic large-cell lymphomas. Plasma concentration-time courses of crizotinib were adequately described by a one-compartment pharmacokinetic model. A pharmacodynamic link model reasonably fit the time courses of ALK inhibition in both H3122 and Karpas299 models with EC50 values of 233 and 666 ng/ml, respectively. A tumor growth inhibition model also reasonably fit the time course of individual tumor growth curves with EC50 values of 255 and 875 ng/ml, respectively. Thus, the EC50 for ALK inhibition approximately corresponded to the EC50 for tumor growth inhibition in both xenograft models, suggesting that >50% ALK inhibition would be required for significant antitumor efficacy (>50%). Furthermore, based on the observed clinical pharmacokinetic data coupled with the pharmacodynamic parameters obtained from the present nonclinical xenograft mouse model, >70% ALK inhibition was projected in patients with non–small-cell lung cancer who were administered the clinically recommended dosage of crizotinib, twice-daily doses of 250 mg (500 mg/day). The result suggests that crizotinib could sufficiently inhibit ALK phosphorylation for significant antitumor efficacy in patients.
Recent advances in a number of molecular profiling technologies, the so-called “-omics” sciences, such as genomics, transcriptomics, proteomics, and metabolomics, have significantly enhanced the potential and actual development of personalized medicine, which targets individualized treatment and care based on genetic and individual variations. Some examples of successful approaches to personalized medicine have been reported in cancer therapy, e.g., small-molecule epidermal growth factor receptor (EGFR) inhibitors, gefitinib and erlotinib, for patients with non–small-cell lung cancer (NSCLC) with activating EGFR mutations (Pao et al., 2004) and an anti-Her2 monoclonal antibody, trastuzumab, for patients with breast cancer with HER2 gene overexpression (Spector, 2008). Such accomplishments illustrate the ever-increasing evidence that a personalized medicine approach can yield promising clinical responses in a subset of patients with cancer. On the other hand, rapidly acquired drug resistance still remains an important limitation to long-term successful cancer therapy (Engelman and Settleman, 2008). For example, in patients with NSCLC who demonstrated a primary response to EGFR inhibitors, acquired resistance against EGFR inhibitor monotherapy typically develops after 6 to 12 months (Sharma et al., 2007). One of the mechanisms for this acquired resistance seems to be genomic amplification of the hepatocyte growth factor receptor MET (also named cMet or HGFR), because a combination of MET and EGFR inhibitors can potentially improve efficacy (Engelman et al., 2007; McDermott et al., 2010). Thus, a clinical outcome improvement in some patients with cancer is likely to be achieved by identification of the molecular events that underlie their specific pathogenesis.
The identification of activating mutations or translocations of the anaplastic lymphoma kinase (ALK) gene has been reported in various types of cancer such as anaplastic large-cell lymphoma (ALCL) (Kutok and Aster, 2002), inflammatory myofibroblastic tumor (Pulford et al., 2004), neuroblastoma (George et al., 2008; Mossé et al., 2008), and NSCLC (Soda et al., 2007; Mano, 2008). Nucleophosmin is the most common fusion partner of ALK (80% of translocations) in ALCL. Echinoderm microtubule-associated protein-like 4 (EML4)-ALK fusion protein has been identified in ∼7% of patients with NSCLC, whereas other rare fusion partners for ALK have also been detected in patients with NSCLC (Takeuchi et al., 2009). The genetic ALK rearrangements rarely coexist with Kirsten rat sarcoma viral oncogene homolog or EGFR mutations, and patients with ALK rearrangement tend to be younger than those without the rearrangements and most of them have had little or no exposure to tobacco (Soda et al., 2007; Mano, 2008; Perner et al., 2008; Wong et al., 2009; Kwak et al., 2010). These findings suggest that the ALK rearrangement is a promising therapeutic target as well as a diagnostic molecular marker in patients with NSCLC.
Crizotinib (Xalkori, PF02341066, (R)-3-[1-(2, 6-dichloro-3-fluoro-phenyl)-ethoxy]-5-(1-piperidin-4-yl-1H-pyrazol- 4-yl)-pyridin-2-ylamine) was identified as an orally available, ATP-competitive dual inhibitor for ALK and MET. Crizotinib has recently been approved by the Food and Drug Administration for the treatment of locally advanced or metastatic NSCLC in patients who are positive for ALK rearrangement as detected by a Food and Drug Administration-approved test. We previously reported the pharmacokinetic/pharmacodynamic (PK/PD) modeling of crizotinib for the inhibition of MET phosphorylation and antitumor efficacy in athymic mice implanted with GTL16 gastric carcinomas or U87MG glioblastomas (Yamazaki et al., 2008). PK/PD modeling is a useful mathematical approach linking drug exposure to pharmacologic responses as a function of time, providing a quantitative assessment of in vivo drug potency with mechanistic insight of drug action (Derendorf et al., 2000; Chien et al., 2005). The previous PK/PD modeling results of crizotinib suggested that >90% inhibition of MET phosphorylation would be required for significant antitumor efficacy (>50%).
The objectives of the present study were to characterize: 1) the PK/PD relationship of crizotinib plasma concentrations to the inhibition of ALK phosphorylation in tumors (biomarker), and 2) the relationship of ALK inhibition to antitumor efficacy (pharmacological response) in athymic mice implanted with H3122 NSCLC cells harboring the EML4-ALK fusion protein or Karpas299 ALCL cells carrying the nucleophosmin-ALK fusion protein. Subsequently, we compared the present PK/PD relationships in the ALK-driven xenograft mouse models with the previous PK/PD results in the MET-driven xenograft models. Furthermore, based on these nonclinical PK/PD modeling results, the biomarker inhibitions of ALK and MET phosphorylation were projected by PK/PD simulation by using the observed crizotinib clinical PK results in patients with NSCLC. The present PK/PD results will be helpful in understanding PK/PD relationships of crizotinib and guiding dose escalation or de-escalation to achieve or maintain efficacious exposure in patients with ALK or MET-positive tumors.
Materials and Methods
Crizotinib (chemical purity >99%) and a structurally related in-house compound (internal standard for analysis) were synthesized at Pfizer Worldwide Research and Development (San Diego, CA) (Cui et al., 2011). All other reagents and solvents were commercially available and of either analytical or high-performance liquid chromatography grade.
In Vivo PK/PD Study.
The experimental designs and methods of the in vitro and in vivo PK/PD studies were previously reported in part by Zou et al. (2011). In brief, three separate repeated oral-dose PK/PD studies were conducted with crizotinib in athymic nu/nu mice implanted with H3122 NSCLC xenografts (studies 1 and 2) or in severe combined immunodeficient/beige mice implanted with Karpas299 ALCL xenografts (study 3). Mice were treated with crizotinib at doses of 25, 50, 100, and 200 mg/kg once daily in studies 1 and 2 (14 and 18 days of administration, respectively) or doses of 25, 50, and 100 mg/kg once daily in study 3 (13 days of administration). A subset of mice was humanely euthanized at 1, 4, 7, and 24 h after the last dose. Blood samples (n = 3/time point) were collected by exsanguination via cardiac puncture to determine plasma concentrations of crizotinib. Resected tumors (n = 3/time point) were snap-frozen and pulverized by using a liquid nitrogen-cooled cryomortar. Protein lysates were generated, and then the level of total phosphorylated ALK protein (ALK phosphorylation) was determined by using a capture enzyme-linked immunosorbent assay method (studies 1 and 3). Tumor volume was measured during the treatment period by electronic Vernier calipers and calculated as the product of its length × width2 × 0.4 (studies 2 and 3). All of the procedures were conducted in accordance with the Guide for the Care and Use of Laboratory Animals (Institute for Laboratory Animal Research, 1996) and Pfizer Animal Care and Use Committee guidelines.
Crizotinib Plasma Concentration Analysis.
A quantitative assay method to determine crizotinib plasma concentrations was previously reported (Yamazaki et al., 2008). In brief, mouse plasma samples were prepared by protein precipitation with a methanol/acetonitrile mixture (25:75, v/v). After the centrifugation, an appropriate volume of the resulting supernatant was analyzed by liquid chromatography-tandem mass spectrometry. The chromatography was performed with a Shimadzu high-performance liquid chromatography system, equipped with a binary solvent delivery system LC-10 ADvp and a controller SLC-10Avp (Shimadzu Scientific Instruments, Columbia, MD) using a reverse-phase column (Agilent XDB-C18, 2.1 × 50 mm, 5 μm; Agilent Technologies, Santa Clara, CA). Mass spectrometric analysis was performed on an API 4000 triple quadrupole mass spectrometer (Applied Biosystems/MDS Sciex, Foster City, CA) using turbo-ion spray ionization. Sample analysis was performed in the positive ionization, multiple reaction monitoring mode with unit resolution for the transitions of m/z 450 to 260 for crizotinib and m/z 377 to 348 for the internal standard. The calibration curve range was 1 to 1000 ng/ml. The back-calculated calibration standard concentrations were within ± 15% of their nominal concentrations with coefficients of variation of less than 15%. The precision and accuracy of the quality-control samples were within ± 15%.
A naive-pooled pharmacokinetic analysis was used to determine crizotinib pharmacokinetic parameters in mice because a subset of mice (n = 3/time points) was humanely euthanized at each time point to collect blood samples. Therefore all individual data at each dose were pooled together for the pharmacokinetic analysis as if they came from a single individual (Sheiner, 1984). This approach provided a better fit than a nonlinear pharmacokinetic model with Michaelis-Menten elimination (data not shown). Pharmacokinetic analysis was performed with a standard one-compartment model as implemented in NONMEM version VI (Beal and Sheiner, 1992). This model (subroutine ADVAN2 with TRANS2) was parameterized by using an absorption rate constant (ka, h−1), oral clearance (CL/F, L/h/kg), and oral volume of distribution (Vd/F, L/kg). Residual variability was characterized by a proportional error model. The pharmacokinetic parameters thus obtained were used to simulate plasma concentrations as a function of time after oral administration to drive the time-dependent pharmacodynamic models.
PK/PD modeling for the response of ALK phosphorylation in tumor to plasma concentration of crizotinib in studies 1 and 3 was performed by a link model (an effect-compartment model) (Sheiner et al., 1979). In brief, the effect site concentration of crizotinib (Ce, ng/ml) was expressed by the following differential equation: where ke0 is the rate constant for equilibration with the effect site (h−1), and Cp is the plasma concentration of crizotinib (ng/ml).
In the link model, the following equation was used to determine EC50 (the concentration causing one-half maximum effect, Emax) for the inhibition of ALK phosphorylation (E): where E0 is the baseline of ALK phosphorylation (ratio to control animals) and γ is the Hill coefficient.
An alternative model, the indirect response model, assumes that ALK phosphorylation at baseline is maintained by the balance of formation and degradation rates (Jusko and Ko, 1994). The addition of crizotinib was considered to inhibit the model's formation rate, because crizotinib was a competitive ATP-binding ALK inhibitor. Therefore the following differential equation was used to determine the EC50 required for the inhibition of ALK phosphorylation (R): where kin is the zero-order formation rate constant (h−1), and kout is the first-order degradation rate constant (h−1).
PK/PD modeling for the tumor growth inhibition (TGI) to crizotinib plasma concentration in studies 2 and 3 was performed by a modified indirect response model as reported previously (Yamazaki et al., 2008). In brief, crizotinib plasma concentration inhibited tumor growth, assuming the effect of crizotinib ultimately decreased the tumor growth rate: where T is tumor volume, ktg is the first-order tumor growth rate constant (h−1), ktd is the first-order tumor death rate constant (h−1), and Tss represents the maximum sustainable tumor volume (carrying capacity).
The ratio of T/Tss approximates zero when T is relatively small, meaning the net growth rate is approximately first-order (i.e., exponential growth). The tumor growth thereafter approaches zero when the ratios of T/Tss reach unity. Thus, the logistic model is applicable when tumor growth starts to slow down at later growth stages of the study period. When the estimate of Tss was much greater than the observed maximum tumor volume, tumor growth simply followed exponential growth curve in a study period. Therefore, the above TGI model could be simplified to the following differential equation:
In the present study, the logistic model was used in study 2, whereas the exponential growth model was used in study 3, because each respective model provided a better fit to the individual tumor growth curves (data not shown). This difference may simply reflect tumor growth dynamics that differ among xenograft models. Hill coefficients (γ) were fixed to be unity in both studies.
All PK/PD modeling analyses were performed with NONMEM version VI and S-Plus 6.2 (Insightful Corporation, Seattle, WA). The NONMEM subroutine ADVAN6 was used for the link model, whereas the ADVAN8 was used for the indirect response and TGI models. The initial conditions at time 0 for the gastrointestinal tract compartment, ALK phosphorylation ratio, and tumor volume were the dose amount (mg/kg), the ALK baseline ratio (i.e., unity), and the measured initial individual tumor volume (mm3), respectively. Residual variability was characterized by a proportional error model. In the TGI model, an interanimal variability on ktg or ktd was estimated by using an exponential variance model. Model selection was based on a number of criteria such as the NONMEM objective function values (OFVs), estimates, standard errors, and scientific plausibility, as well as exploratory analysis of standard goodness-of-fit plots. The difference in the OFV between two nested models was compared with a χ2 distribution in which a difference of 6.63 was considered significant at the 1% level (Wählby et al., 2001). The final model parameter estimates were evaluated by running a bootstrap procedure with 5000 datasets (Efron and Tibshirani, 1993). The parametric statistics of the parameters (median and 10th and 90th percentiles) thus generated were compared with the final parameter estimates generated by the NONMEM analysis.
PK/PD Simulation in Patients.
Crizotinib plasma concentrations were first simulated as a function of time in patients following the clinically recommended dosage of crizotinib, twice-daily doses of 250 mg (500 mg/day), for 14 days, using a one-compartment pharmacokinetic model with CL/F of 70 L/h, Vd/F of 1500 liters, and ka of 0.75 h−1. These pharmacokinetic parameters were adjusted from the clinically observed single-dose pharmacokinetic parameters to simulate comparable steady-state plasma concentrations to the clinically observed results reported previously (Tan et al., 2010): the differences in steady-state maximum plasma concentrations and area under the plasma concentration-time curves between the simulated (342 ng/ml and 3570 ng · h/ml, respectively) and observed values (368 ng/ml and 3641 ng · h/ml, respectively) were within 10%. Based on the simulated crizotinib plasma concentrations, the crizotinib-mediated inhibitions of ALK and MET phosphorylation in tumors were simulated by using the pharmacodynamic parameters determined in the present study for ALK inhibition and the previous study for MET inhibition (Yamazaki et al., 2008). Because the crizotinib clinical studies were conducted mainly in patients with advanced NSCLC who were positive for EML4-ALK fusion (Kwak et al., 2010), the pharmacodynamic parameters estimated from an EML4-ALK-positive H3122 NSCLC xenograft model were used for the PK/PD simulations in patients. The EC50 values from the nonclinical xenograft models were corrected for the difference in plasma protein binding between humans and mice, thus assuming that the unbound EC50 values were comparable between the patients and xenograft mouse model. The PK/PD simulation was performed by using NONMEM version VI with the subroutine ADVAN8. The initial condition at time 0 for the gastrointestinal tract compartment was the dose amount (i.e., 250 mg), whereas those for the ALK and MET phosphorylation were their baselines (i.e., unity).
Plasma concentration-time courses of crizotinib in both H3122 and Karpas299 xenograft mouse models after repeated oral administration of crizotinib were adequately described by a one-compartment model. The observed and model-fitted plasma concentrations of crizotinib in both xenograft models are shown in Fig. 1. Pharmacokinetic parameter estimates for ka, CL/F, and Vd/F were 0.076 to 1.8 h−1, 0.76 to 5.3 L/h/kg, and 0.52 to 17 L/kg, respectively (Table 1). The CL/F values tended to be higher at the lower doses than at the higher doses, suggesting nonlinear pharmacokinetics at the dose range of 25 to 200 mg/kg. The observed dose-dependent pharmacokinetics could be, in part, caused by an inhibition of crizotinib hepatic/intestinal clearance at higher doses because crizotinib was reported to be a substrate and inhibitor of CYP3A isozymes (Tan et al., 2010; Johnson et al., 2011). The standard errors of the majority of pharmacokinetic parameter were relatively small (coefficients of variation <40%). Residual variability was estimated to be 28, 17, and 8% in studies 1, 2, and 3, respectively, and OFVs were 682, 566, and 475, respectively. Final parameter estimates (median values) from the bootstrap procedure were 0.078 to 1.8 h−1, 0.77 to 5.5 L/h/kg, and 0.54 to 17 L/kg for ka, CL/F, and Vd/F, respectively. Thus, the final parameter estimates (50th percentile) for the bootstrap validation were in good agreement with the estimates of the final pharmacokinetic model (< ±3%).
PK/PD Relationships for ALK and TGI
PK/PD Modeling for ALK Inhibition.
Crizotinib plasma concentrations slowly declined in both H3122 and Karpas299 xenograft models after reaching the maximum concentrations at 1 to 4 h postdose, whereas the inhibition of ALK phosphorylation was sustained throughout most of the dosing interval of 24 h, especially at higher doses. Thus it was important to incorporate a time-delay (hysteresis) factor between the crizotinib plasma concentrations and the ALK inhibition into PK/PD modeling. The observed and link model-fitted ALK phosphorylation-time profiles, along with the predicted crizotinib concentrations in plasma and effect compartment, are graphically presented in Fig. 2. The link model reasonably fit the time courses of ALK inhibition with EC50 of 233 and 666 ng/ml in studies 1 and 3, respectively (Table 2). The standard errors of EC50, ke0, and γ were, respectively, 66, 42, and 20% of the estimates in study 1 and 12, 30, and 8% in study 3. OFVs were −153 in study 1 and −178 in study 3. In contrast to the link model, an indirect response model did not fit the time courses of ALK inhibition well in studies 1 and 3, providing higher OFVs of −132 and −162, respectively. Final parameter estimates of EC50 from the bootstrap procedure were 233 and 667 ng/ml in studies 1 and 3, respectively, and other final PK/PD parameter estimates were also in good agreement with the estimates of the final model (< ±1%).
PK/PD Modeling for TGI.
The TGI model fit the individual tumor growth curves well during crizotinib repeated-dose treatment in both H3122 and Karpas299 xenograft models (studies 2 and 3, respectively) (Fig. 3). The EC50 values were estimated to be 255 and 875 ng/ml in studies 2 and 3, respectively (Table 3). OFVs were 4011 in study 2 and 2377 in study 3. Final parameter estimates of EC50 for the bootstrap procedure were 257 and 881 ng/ml in studies 2 and 3, respectively, and other final parameter estimates of the PK/PD parameters were also in good agreement with the estimates of the final model (< ±1%). The concentration-response curves for ALK and TGI using a sigmoidal Emax model with the estimated pharmacodynamic parameters (EC50, Emax, and γ) from H3122 and Karpas299 xenograft mouse models are graphically presented in Fig. 4, A and B, respectively. The concentration-response curves for MET and TGI in a GTL16 xenograft model from a previous study (Yamazaki et al., 2008) are shown in Fig. 4C for comparison. The EC50 values for ALK inhibition were comparable with those for TGI in both H3122 and Karpas299 xenograft models (233 versus 255 and 666 versus 875 ng/ml, respectively) (Table 4). On the other hand, the EC50 for MET inhibition (19 ng/ml) in a GTL16 xenograft model was approximately 10-fold lower than that for TGI (213 ng/ml), which was approximately comparable with the EC90 for MET inhibition (167 ng/ml) (Table 4).
PK/PD Simulation for ALK and MET Inhibition in Patients
The crizotinib-mediated inhibitions of ALK and MET phosphorylation in tumors were simulated in a population of patients following the recommended clinical dosage of crizotinib, twice-daily doses of 250 mg (500 mg/day) for 14 days (Fig. 5). The simulated MET inhibition rapidly reached near-complete inhibition (∼95%), whereas the simulated ALK inhibition was approximately 75% at steady state. The simulated ALK and MET inhibitions in patient tumors were relatively sustained throughout the dosing interval, which seemed to be largely caused by the relatively small ke0 values (0.030 and 0.14 h−1 for ALK and MET, respectively) combined with the twice-daily dosing regimen. Overall, the simulation results indicated that crizotinib could significantly inhibit the phosphorylation of ALK (>70%) and MET (>90%) in patient tumors during the treatment of the recommended clinical dosage of 250 mg twice daily.
The present nonclinical PK/PD study demonstrated that crizotinib was associated with the inhibition of ALK phosphorylation in H3122 NSCLC and Karpas299 ALCL xenograft mouse models. The estimated in vivo EC50 in a H3122 NSCLC xenograft model was 19 nM free (Table 4), which was 3-fold lower than the in vitro EC50 of 56 nM (Zou et al., 2011). The EC50 in a Karpas299 ALCL xenograft model was 53 nM free, which was 1.5-fold higher than the in vitro EC50 of 35 nM (Christensen et al., 2007; Zou et al., 2011). Thus, the estimated EC50 values for ALK inhibition in these cancer cell lines were relatively consistent between in vitro and in vivo studies. In the previous study using a GTL16 gastric carcinoma xenograft model (Yamazaki et al., 2008), the in vivo EC50 for MET inhibition was 1.5 nM free, which was 7-fold lower than the in vitro EC50 of 11 nM (Zou et al., 2007). Crizotinib showed relatively high nonspecific binding (approximately 90%) in hepatic microsomes and hepatocytes, along with plasma protein binding of 91 to 96% across species (Yamazaki et al., 2011). Therefore, the correction of nonspecific binding in the cell-based assay systems might be required to further evaluate the EC50 difference between in vitro and in vivo. In addition, it has been reported that the expression levels of drug-metabolizing enzymes and transporters could be altered after inoculation of tumor cells into athymic mice, which might cause the discrepancies between in vivo and in vitro activities of antitumor drugs (Sugawara et al., 2010). Crizotinib is a substrate of CYP3A isozymes and the multidrug-resistance transport protein P-glycoprotein (Johnson et al., 2011). Therefore, the in vivo and in vitro EC50 differences of crizotinib might, in part, be caused by the changes of the expression levels of drug-metabolizing enzymes and transporters in human tumor xenografts implanted subcutaneously into athymic or immunodeficient mice.
Regarding the PK/PD relationships of biomarker inhibition to antitumor efficacy, the EC50 for ALK inhibition was consistent with the EC50 for TGI in both H3122 NSCLC and Karpas299 ALCL xenograft models (19 versus 20 nM free and 53 versus 70 nM free, respectively). This relationship suggests that >50% inhibition of ALK phosphorylation would be required for significant antitumor efficacy (>50%). Therefore, the EC50 value for ALK inhibition could be considered as a minimum target efficacious concentration in the clinic. This is, to our knowledge, the first report describing the PK/PD relationships of drug concentration to ALK inhibition and TGI. In our previous study (Yamazaki et al., 2008), the PK/PD relationship of MET inhibition to antitumor efficacy in a GTL16 xenograft model was characterized with a similar PK/PD modeling approach. Unlike in the present study, it was the EC90 (13 nM free) for MET inhibition that was comparable with the EC50 (17 nM free) for TGI (Table 4), suggesting that near-complete inhibition of MET phosphorylation (>90%) would be required for significant antitumor efficacy (>50%). A similar crizotinib PK/PD relationship with MET inhibition to TGI was also suggested in a U87MG glioblastoma xenograft mouse model (Yamazaki et al., 2008). Therefore, the crizotinib PK/PD relationships of target modulation to antitumor efficacy in the xenograft models seem to be different between ALK and MET inhibition. These considerations are graphically summarized in Fig. 4. The concentration-response curves for ALK and TGI were similar in magnitude and had comparable EC50 values. In contrast, the concentration-response curve for TGI was shifted to the right compared with the MET response curve, resulting in approximately 10-fold difference in the EC50 values between MET inhibition and TGI. The difference in these response curves between ALK-TGI and MET-TGI in the nonclinical xenograft models seems to suggest that targeting ALK may be more effective than MET to achieve similar levels of antitumor efficacy in cancer patients. Such an extrapolation, of course, assumes similar PK/PD relationships from nonclinical xenograft models to patients. Although xenograft mouse models are extensively used as the most common nonclinical antitumor efficacy model (Kelland, 2004; Burchill, 2006; Hollingshead, 2008), there are obviously several assumptions associated with such an extrapolation of the PK/PD relationship, which should be addressed.
One of the main assumptions is that the tumor environment in subcutaneous tumor xenograft models is similar to that in human tumors, which also presumes a similar drug distribution between xenograft mouse models and human tumors. In other words, the plasma concentrations of crizotinib (and other inhibitors) required for biomarker response (or target modulation) and antitumor efficacy is assumed to be equivalent between the xenograft models and patient tumors. In this context, we assumed that unbound plasma concentrations of crizotinib were efficaciously equivalent between xenograft mouse models and patients. Therefore, in addition to accounting for species-related difference in pharmacokinetics, only the difference in crizotinib plasma protein binding between mice and humans was accounted for in the simulation of crizotinib-mediated ALK and MET inhibition in patients (Fig. 5). Crizotinib extensively distributed into the tumors of the xenograft mouse models with an approximate tumor/plasma area under the concentration-time curve ratio of 4 at steady state (in-house data), despite its being a substrate of P-glycoprotein. This observation seems to be consistent with the significant biomarker inhibition and TGI observed in the xenograft models. Marked differences in tumor growth rate between xenograft models and patients may also have a significant impact on an evaluation of antitumor efficacy (Komlodi-Pasztor et al., 2011). Up to this point, the promising clinical response by crizotinib as a single agent has been reported in EML4-ALK-positive patients with NSCLC (Kwak et al., 2010): an overall response rate of 57% (confirmed partial and complete responses) and a rate of stable disease of 33% (stable disease plus unconfirmed partial responses) in 85 patients, the majority of whom had received multiple previous therapies. These clinical responses of crizotinib in the ALK-positive patients with NSCLC seem to derive from the crizotinib ALK inhibition activity, because tumors from 33 patients with available tissues were negative for MET amplification (Kwak et al., 2010). These clinical results therefore seem to be consistent with the present PK/PD simulation in patients, where ALK inhibition by crizotinib at steady state was more than 70% (Fig. 5). We could also expect that crizotinib would show promising clinical response in MET-positive patients because the simulated MET inhibition by crizotinib at steady state was more than 90% in patients (Fig. 5). A recent report indicated that a patient with NSCLC with MET amplification, but no ALK rearrangement, achieved a rapid and durable response to crizotinib (Ou et al., 2011); however, extensive clinical results have not been reported yet.
Furthermore, the comparison between ALK and MET response curves in Fig. 4 might hint that a combination approach of a MET inhibitor with another tyrosine kinase inhibitor, e.g., EGFR inhibitor, would be a promising therapeutic strategy for patients with cancer, although several assumptions would be required for the extrapolation of the PK/PD relationships, as mentioned above. Emerging data (e.g., systems biology/pharmacology) support the view of an extensive and intricate signaling cross-talk and scaffold networks within cancer cells during tumorigenesis and tumor progression (Rikova et al., 2007; Guo et al., 2008). These networks can also possibly undergo adaptive changes during long-term therapy where sustained inhibition is maintained such as an acquired resistance against EGFR inhibitor caused by MET amplification (Engelman et al., 2007; McDermott et al., 2010). This suggests a multitargeted approach may be better for certain cancer therapies including overcoming acquired treatment resistance. It is noteworthy that some clinical MET inhibitor candidates are considered to be highly selective, whereas others are multitargeted inhibitors (Faoro et al., 2009; Tirgan et al., 2009). It remains to be seen whether highly selective MET inhibitors or multitargeted inhibitors (or a combination of selective MET inhibitor with other tyrosine kinase inhibitors) will provide a better success in the clinic.
In conclusion, the PK/PD relationships among crizotinib systemic concentration, ALK or MET inhibition, and TGI in human tumor xenograft models were well characterized in a quantitative manner by using PK/PD modeling (Fig. 6). The present modeling efforts suggests that >50% ALK inhibition would be required for significant antitumor efficacy (>50%). Accordingly, we proposed that the EC50 value for ALK inhibition could be considered a minimum target efficacious concentration in the clinic. PK/PD simulation based on the results from a NSCLC xenograft model suggests that crizotinib could inhibit ALK phosphorylation by more than 70% in patients with cancer who follow the clinically recommended dosing regimen (250 mg twice a day), which seems to be consistent with the reported clinical response results (Kwak et al., 2010). Overall, the present PK/PD results will be helpful in understanding PK/PD relationships of crizotinib and guiding dose escalation or de-escalation to achieve or maintain efficacious exposure to crizotinib in patients with cancer. It would be interesting to compare clinical biomarker responses between ALK and MET in crizotinib-treated patients, which has not been reported to date.
Participated in research design: Yamazaki, Zou, and Christensen.
Conducted experiments: Shen, Zou, Lee, and Li.
Performed data analysis: Yamazaki and Vicini.
Wrote or contributed to the writing of the manuscript: Yamazaki, Vicini, Shen, Christensen, Smith, and Shetty.
We thank Hunter and David Paterson for the animal experiments, Darian Bartkowski, Leslie Nguyen, Ravi Rahavendran, and Sylvia Vekich for the bioanalytical assay, and Theodore Johnson, Kirk Kozminski and Justine Lam for the PK/PD discussion (all Pharmacokinetics, Dynamics, and Metabolism Department, Pfizer Worldwide Research and Development, San Diego, CA); and Jean Cui, Michelle Tran-Dube, Pei-Pei Kung, and Mitch Nambu (Pfizer Worldwide Research and Development, San Diego) for the chemical synthesis.
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
- epidermal growth factor receptor
- anaplastic large cell lymphoma
- anaplastic lymphoma kinase
- effect-site concentration of crizotinib
- oral clearance
- plasma concentration of crizotinib
- crizotinib (PF02341066)
- ALK phosphorylation baseline
- maximum effect
- echinoderm microtubule-associated protein-like 4
- Hill coefficient
- absorption rate constant
- rate constant for equilibration with the effect site
- first-order tumor death rate constant
- first-order tumor growth rate constant
- hepatocyte growth factor receptor
- non–small-cell lung cancer
- objective function value
- tumor volume
- tumor growth inhibition
- maximum sustainable tumor volume
- oral volume of distribution
- study 1
- study 2
- study 3.
- Received October 12, 2011.
- Accepted November 29, 2011.
- Copyright © 2012 by The American Society for Pharmacology and Experimental Therapeutics