Population pharmacokinetic analysis of crizotinib in children with progressive/recurrent high-grade and diffuse intrinsic pontine gliomas

Abstract

Purpose:

Crizotinib, a potent oral tyrosine kinase inhibitor, was evaluated in combination with dasatinib in a phase 1 trial (NCT01644773) in children with progressive or recurrent high-grade and diffuse intrinsic pontine gliomas (HGG and DIPG). This study aimed to characterize the pharmacokinetics of crizotinib in this population and identify significant covariates.

Methods:

Patients (N=36, age range 2.9–21.3 years) were treated orally once or twice-daily with 100 to 215 mg/m² crizotinib and 50 to 65 mg/m² dasatinib. Pharmacokinetic studies were performed for crizotinib alone after the first dose and at steady state, and for the drug combination at steady state. Crizotinib plasma concentrations were measured using a validated LC-MS/MS method. Population modeling was performed (Monolix) and the impact of factors including patient demographics and co-medications were investigated on crizotinib pharmacokinetics.

Results:

Crizotinib concentrations were described with a linear two-compartment model and absorption lag time. Concomitant dasatinib and overweight/obese status significantly influenced crizotinib pharmacokinetics, resulting in clinically relevant impact (>20%) on drug exposure. Crizotinib mean apparent clearance (CL/F) was 66.7 L/h/m² after single-dose and decreased to 26.5 L/h/m² at steady state when given alone, but not when combined with dasatinib (mean 60.8 L/h/m²). Overweight/obese patients exhibited lower crizotinib CL/F and apparent volume V₁/F (mean 46.2 L/h/m² and 73.3 L/m²) compared to other patients (mean 75.5 L/h/m² and 119.3 L/m², p<0.001).

Conclusion:

A potential pharmacokinetic interaction was observed between crizotinib and dasatinib in children with HGG and DIPG. Further, crizotinib exposure was significantly higher in overweight/obese patients, who may require a dosing adjustment.

Introduction

High-grade gliomas (HGG), including diffuse intrinsic pontine gliomas (DIPG), are among the most aggressive and lethal brain tumors in children [1,2]. Standard of care therapy includes surgical resection and radiation therapy, but options remain limited. Maximum tumor resection is difficult to achieve because of high tumor vascularity, invasiveness, and location. Further, treating young children with radiation is challenged by the irreversible effect on neurocognition and quality of life [2]. New therapeutic combination strategies are desperately needed to improve the survival rates of these patients, which remain <30% for HGG and <10% for DIPG after 2 years [3].

Within the last 10 years, whole genome-wide studies have revealed several molecular characteristics of HGG and DIPG as potential therapeutic targets. Simultaneous amplification of multiple receptor tyrosine kinase (RTK) genes occurs in these tumors, the most common of which are the genes encoding the platelet-derived growth factor receptor α (PDGFRA) and the hepatocyte growth factor receptor c-MET (MET) [4,5]. These genetic abnormalities affect crucial cellular pathways (i.e., PDGF and c-MET) involved in cell proliferation, motility, angiogenesis, and dedifferentiation [6–8]. These findings provided a strong rationale for combining therapeutic agents targeting different RTK, such as crizotinib, a potent inhibitor of anaplastic lymphoma kinase (ALK) and c-MET, and dasatinib, a potent inhibitor of PDGFR and Src which is a downstream essential signaling effector for c-Met [9]. This drug combination was evaluated in a phase I trial in children with recurrent or progressive HGG and DIPG (NCT01644773) [10].

Crizotinib and dasatinib are both oral tyrosine kinase inhibitors metabolized by cytochrome P-450 (CYP) 3A4 enzymes and are thus subject to potential drug–drug interactions [11,12]. A decrease in crizotinib clearance after multiple dosing due to an autoinhibition of CYP3A4-mediated metabolism has been reported in adults [11]. Therefore, pharmacokinetic studies were conducted for crizotinib, as a single agent and in combination with dasatinib in this phase I study to describe the drug disposition in children with HGG and DIPG, and to evaluate for potential interactions between the two agents. This manuscript details the population pharmacokinetic modeling analysis performed for crizotinib from this pediatric phase I study. The objectives of this analysis were to characterize the pharmacokinetic profile of crizotinib, quantify the inter- and intra-patient pharmacokinetic variability within the study population, and identify patient covariates with significant impact on drug exposure.

Methods

Study design and treatment

The patients included in this study were treated on a dose-escalation phase I clinical trial (SJHG12; NCT01644773), conducted at St. Jude Children’s Research Hospital [10]. The trial was approved by the St. Jude Children’s Research Hospital Institutional Review Board and informed written consent was obtained from the legal guardians. The aims of the clinical trial were to find the maximum tolerable dosage (MTD) of crizotinib and dasatinib given in combination to patients with DIPG and other types of HGG and to identify the dose-limiting toxicities (DLT). Initially, crizotinib and dasatinib were orally administered twice daily (starting dosages of 130 mg/m² and 50 mg/m², respectively) and the DLT evaluation period lasted for 6 weeks. Drug administration was switched to once daily (starting dosages of 165 mg/m² and 50 mg/m², respectively) and the DLT evaluation period was shortened to 4 weeks as the initial regimen was poorly tolerable, even at the lowest planned dosage level [10]. Patients with recurrent or progressive disease were enrolled on Stratum A and patients who had completed local radiotherapy with or without chemotherapy were enrolled in Stratum B. The different dosing regimens used in the study are shown in Online Resource Table S1. Crizotinib was available as whole capsules or as a suspension. Dasatinib was available as film-coated tablets.

Pharmacokinetic study design and sample bioanalysis

Initial pharmacokinetic study design

During the first course of therapy, crizotinib and dasatinib were administered twice-daily, starting on days 1 and 15, respectively. Serial pharmacokinetic studies were conducted during course 1 days 1, 14, and 42. Single-dose and steady-state pharmacokinetic studies of crizotinib as a single agent were obtained pre-dose and at 1, 2, 4, 8, 24, and 48 hours after the first dose on day 1, and pre-dose and at 1, 2, 4, 8, and 24 hours after the morning dose on day 14. Steady-state pharmacokinetic studies of crizotinib and dasatinib in combination were obtained pre-dose and at 0.5, 1, 2, 4, and 8 hours after the morning dose of both agents on day 42 of therapy. Four doses of crizotinib were held for the studies (evening dose on day 1, both doses on day 2, and evening dose on day 14).

Modified pharmacokinetic study design

In the modified treatment plan, crizotinib and dasatinib were administered once-daily starting on days 1 and 3, respectively. Single-dose pharmacokinetic studies of crizotinib as a single agent were obtained pre-dose and at 1, 2, 4, 8, 24, and 48 hours after the first dose on day 1. Steady-state pharmacokinetic studies of crizotinib and dasatinib in combination were obtained pre-dose and at 1, 2, 4, 8, and 24 hours after the morning dose of both agents on day 14 of therapy. The dose of crizotinib on day 2 was held for the studies.

Sample bioanalysis

At each time point, 3 mL of whole blood were collected in K₂EDTA purple top vacutainer collection tubes. The samples were centrifuged at room temperature for 2 minutes at 10,000 × g within 30 minutes of collection to separate plasma. The plasma was stored at −80°C until analysis. Crizotinib plasma concentrations were determined using a validated liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/MS) method with a lower limit of quantitation (LLOQ) of 5 ng/mL [13]. Dasatinib plasma concentrations were also measured using a validated LC-ESI-MS/MS method with a LLOQ of 1 ng/mL [14].

Population pharmacokinetic modeling

Crizotinib and dasatinib plasma concentrations were analyzed separately using a nonlinear mixed-effects modeling approach [15]. The model parameters were assumed to be lognormally distributed and were estimated using the Stochastic Approximation Expectation Maximization algorithm in Monolix (version 2020R1. Antony, France: Lixoft SAS, 2020). Data below the LLOQ were handled as censored according to the Beal’s method M3 [16]. One- and two-compartment models with linear or nonlinear absorption and elimination were explored. For both drugs, the body surface area (BSA)-based dosages were used as model inputs.

Inter-individual and inter-occasion variabilities were tested and implemented on pharmacokinetic parameters using an exponential model. Additive, exponential, and combined residual error terms were explored. All random-effect parameters were assumed to be normally distributed. The selection of the best model was based upon the changes in the minimum objective function value, the precision of the parameter estimates (relative standard errors, RSE%), and typical goodness-of-fit plots (i.e., observations versus population and individual predictions, normalized prediction distribution errors (NPDE) versus time and population predictions). GraphPad Prism 9.1.1 (GraphPad Software Inc, La Jolla, CA, USA) was used for graphical presentations.

Covariate analysis

A covariate analysis was performed to identify patient factors that explain the variability of the pharmacokinetic parameters of crizotinib. The potential covariates of interest included sex, age, BSA, total body weight, height, body mass index (BMI) percentile calculated using the Centers for Disease Control and Prevention’s gender-specific BMI-for-age growth charts established for children [17], albumin, aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, bilirubin, serum creatinine, estimated glomerular filtration rate (eGFR) using the reduced equation using on height and serum creatinine as previously described [18], drug formulation (suspension versus whole capsule), and concomitant medications taken within 24 hours of the pharmacokinetic study. The concomitant medications included dasatinib, loperamide, ondansetron, sulfamethoxazole–trimethoprim, and dexamethasone. These medications were tested as a categorical variable (i.e., yes/no). Dasatinib individual exposures as steady-state area under the curve (AUC_ss,0–24h) were also tested as a continuous covariable. Weight groups were also defined based upon the patients’ BMI percentiles. Underweight, normal weight, overweight, and obesity were defined as BMI percentile <5%, 5–84.9%, 85–94.9%, and ≥95%, respectively [17]. Continuous and categorical covariates were implemented on pharmacokinetic parameters using exponential models. A classic forward/backward stepwise approach was used to select significant covariates. The statistical criteria used for the forward and backward steps were a decrease in the objective function value of at least 3.84 (p<0.05 based on the χ² test) and 6.63 (p<0.01 based on the χ² test), respectively.

The final pharmacokinetic model including covariates was re-assessed using the same graphical and statistical criteria as described above. The addition of correlation coefficients between model parameters was also explored.

Model internal evaluation

The final pharmacokinetic model developed for crizotinib was evaluated using non-parametric bootstraps and prediction-corrected visual predictive checks (pcVPC) [19,20]. Non-parametric bootstraps were performed to evaluate the precision of the model parameter. The original dataset was replicated 200-times using random sampling with replacement and the model was run for each replicate dataset. The mean and 90% confidence interval of each parameter estimate were calculated and compared to the original model parameters. The bootstraps were performed with the R-based Rsmlx 3.0 package (http://rsmlx.webpopix.org).

To build the pcVPC, a total of 500 dataset replicates from the original dataset were generated and simulated conditionally on the final model parameters. The 5^th, 50^th, and 95^th prediction percentiles of the model simulations along with their 90% confidence interval were overlaid with the observed data for visual comparison. The median of the population prediction within each time bin was used to correct the data and predictions to account for the different crizotinib dosages and the impact of significant covariates.

Clinical impact of covariates on crizotinib exposure

Model simulations were conducted to further evaluate the effects of the selected covariates on crizotinib exposure at steady state after multiple dosing of 215 mg/m² crizotinib once-daily, which was the MTD defined in this clinical study [10]. The effects of the significant covariates were explored one at a time. Virtual patient populations (N=500 patients) were created from the original dataset and had differing combinations of the values of the covariates retained in the final model. Simulations were performed including the interpatient and residual variabilities. Crizotinib steady-state AUC_ss,0–24h and maximum concentration (C_MAX,ss) values were calculated for each virtual patient and compared to the reference population simulated using the median covariate values. The geometric mean ratio and 90% confidence interval were calculated for each covariate and summarized using Forest plots. A covariate was considered as clinically relevant when the geometric mean ratio and 90% confidence interval were outside the 80% to 125% boundaries of the reference population estimate.

Results

Patient and data information

The first eight patients were enrolled in the study with the initial treatment design (i.e., twice-daily drug administration). The remaining 28 patients received crizotinib and dasatinib once-daily. Single-dose pharmacokinetic studies of crizotinib as a single agent were performed for all the patients enrolled in the study on course 1 day 1. Steady-state pharmacokinetic studies for single-agent crizotinib were conducted for 7 patients. Steady-state pharmacokinetic studies for crizotinib and dasatinib in combination were performed for 31 patients. A total of six patients went off study prior to collecting steady-state pharmacokinetic data. The patient characteristics including the clinical covariates are summarized in Table 1.

Table 1:

Demographics, laboratory values, and co-medications of patient participants

Patient characteristics	Results
Male/female n (%)	11 (31) / 25 (69)
aEthnicity n (%)
White / Black / unknown	25 (69) / 6(17) / 5 (14)
Age (years)	10.6 (2.9 – 21.3)
Height (cm)	139 (89.7 – 131)
Weight (kg)	39.7 (13.6 – 109)
bWeight category n (%)
Normal / Overweight / Obese	17 (47) / 9 (25) / 10 (28)
Body surface area (m2)	1.3 (0.6 – 2.4)
Albumin (g/dL)	4.2 (2.4 – 5.2)
Alkaline phosphatase (U/L)	111 (30 – 357)
Aspartate aminotransferase (U/L)	25.5 (7 – 88)
Alanine aminotransferase (U/L)	24 (13 – 147)
Bilirubin (mg/dL)	0.2 (0.1 – 0.9)
Creatinine (mg/dL)	0.49 (0.16 – 1.32)
Calculated eGFR (mL/min/1.73 m2)	93.8 (47.8 – 210.4)
Stratum A / B n (%)	25 (69) / 11 (31)
Crizotinib formulation
Suspension / Whole capsule	24 (67) / 12 (33)
cConcomitant treatment n (%)
Loperamide	19 (26)
Ondansetron	32 (43)
Sulfamethoxazole-trimethoprim	21 (29)
Dexamethasone	11 (15)

Data are reported as frequency (percentage) or median (range).

^aEthnicity was self-declared and not genotypically determined.

^bWeight groups were defined based upon the CDC gender-specific BMI-for-age growth charts established for children [15]. Normal, overweight, and obesity correspond to BMI percentiles <5%, 5–84.9%, 85–94.9%, and ≥95%, respectively.

^cFrequencies were calculated based upon the total number of pharmacokinetic studies (N=74)

Dasatinib population pharmacokinetics

Steady-state dasatinib pharmacokinetic profiles obtained from 31 patients receiving the drug combination were analyzed using a population-based modeling approach. No data were below the LLOQ. A linear two-compartment model with a delayed absorption using a lag-time best described the concentration-time data. Mean±sd dasatinib apparent oral clearance (CL/F) estimate was 73.6±38.6 L/h/m². The pharmacokinetic data, model structure, parameter estimates, and goodness-of-fit plots are reported as supplementary files (Online Resource Figures S1a, S2, S3 and Table S2). The pharmacokinetic model was used to calculate dasatinib exposure for each patient. Individual AUC_ss,0–24h were generated by integrating the predicted dasatinib concentration-time curves. After repeated 50 and 65 mg/m² dasatinib dosages, mean±sd dasatinib plasma AUC_ss,0–24h were 708.2±288 and 732.6±296 h·ng/mL, respectively.

Crizotinib population pharmacokinetics and covariate analysis

Single-dose and steady-state crizotinib pharmacokinetic profiles after administration of crizotinib alone or in combination with dasatinib were simultaneously analyzed using a population-based modeling approach. About 5% of the data were below the LLOQ, all during course 1 day 1, and were censored as previously described. Crizotinib observed plasma concentrations are depicted in Online Resource Figure S1b–c. As for dasatinib, a linear two-compartment model with a delayed absorption using a lag-time best described crizotinib disposition (Online Resource Figure S2). Inter-individual variability was estimated on all parameters except the peripheral clearance and volume (Q/F and V₂/F). Additionally, inter-occasion variability was estimated on the absorption lag-time (T_lag), central CL/F, and apparent central volume (V₁/F). The shrinkage values associated with T_lag, absorption rate constant ka, CL/F, and V₁/F estimates from the base model (i.e., without covariate) were all below 55%.

The covariate analysis resulted in the selection of four patient covariates: crizotinib formulation (suspension versus whole capsule), patient age, CDC weight category (normal versus overweight/obese), and dasatinib concomitant treatment (i.e., yes or no). Crizotinib T_lag was significantly higher in patients taking crizotinib as a whole capsule (mean 1.21 h) compared to patients taking the suspension (mean 0.45 h, p<0.0001, Fig. 1a). Including the effect of drug formulation explained most of the inter-individual variability initially observed on T_lag (~85%). Patient age was continuously associated with crizotinib ka, with older children exhibiting lower ka values (r²=0.48, p<0.0001, Fig. 1b). The analysis suggested higher crizotinib BSA-based CL/F and V₁/F with lower BMI percentiles (Online Resource Figure S4). Including the effect of BMI percentiles as a continuous covariate on CL/F and V₁/F was not significant; thus, it was implemented using the corresponding weight groups as a categorical covariate (Fig. 1c–d). Mean crizotinib CL/F and V₁/F were 44.2 L/h/m² and 63.5 L/m² in overweight/obese patients, approximately half of the values estimated in patients with a normal weight (mean CL/F of 75.4 L/h/m², p=0.0015 and V₁/F of 136.4 L/m², p=0.0008). Last, an effect of dasatinib concomitant medication was observed on both crizotinib ka and CL/F. In presence of dasatinib, crizotinib ka was significantly lower (mean 0.071 versus 0.088 /h, p=0.0004, Fig. 1b). Steady-state crizotinib CL/F values for patients receiving the drug combination were similar to the single-dose CL/F values (mean 60.8 versus 66.7 L/h/m², p>0.99, Fig. 1e). However, steady-state crizotinib CL/F values for patients receiving crizotinib alone were significantly lower than the single-dose CL/F values (mean 26.5 L/h/m², p=0.0053, Fig. 1e). As shown on Fig. 1f, in patients receiving single-agent crizotinib, steady-state CL/F were systematically lower than single-dose CL/F for each patient, as opposed to patients receiving the drug combination. Instead of using a categorical covariate (i.e., yes/no) for concomitant dasatinib, we also explored the associations between individual steady-state crizotinib CL/F and model-predicted dasatinib AUC_ss,0–24h. However, no significant relationship was found (Online Resource Figure S5).

Fig. 1.

Crizotinib pharmacokinetic parameters and covariate relationships. Distribution of absorption lag-time in patients taking crizotinib as suspension versus whole capsule (a). Scatterplot of absorption rate ka versus age and co-dasatinib (b). Distribution of apparent oral clearance CL/F (c) and volume V1/F (d) in normal weight patients versus overweight/obese patients. Distribution of CL/F after single-dose versus at steady state with or without co-dasatinib (e). Individual changes in CL/F from single-dose to steady-state with or without co-dasatinib (f). In panels a, c, d, and e, the distribution of variables is shown using Tukey boxplots.

The final equations describing the relationships between covariates and crizotinib pharmacokinetic parameters are reported below:

$$T_{lag,i}=T_{lag,pop}\cdot \left(e^{WC\cdot {\beta }_{WC}}\right),$$

$$k{a}_i=k{a}_{pop}\cdot \left(e^{Age\cdot {\beta }_{Age}}+e^{Das\cdot {\beta }_{Das}}\right),$$

$$CL/F_i=CL/F_{pop}\cdot \left(e^{WTg\cdot {\beta }_{\mathit{\text{WTgcl}}}}+e^{SSa\cdot {\beta }_{SSa}}\right),$$

$$V_1/F_i=V_1/F_{pop}\cdot e^{WTg\cdot {\beta }_{\mathit{\text{WTgv}}1}},$$

where WC refers to whole capsule formulation (1 for whole capsule, 0 for suspension), Das refers to concomitant dasatinib (1 for co-treatment, 0 for crizotinib alone), WTg is the weight group category (1 if overweight or obese, and 0 if normal weight), and SSa refers to steady-state status with crizotinib alone as a single agent. The β terms represent the estimated covariate coefficients.

All the final model parameter estimates along with the mean and 90% confidence intervals of the non-parametric bootstraps are reported in Table 2. The model parameters were adequately estimated with RSE% below 50%, except for the inter-individual variability of V₁/F (57.5%) and T_lag (159%). The population parameter estimates were similar to the mean of the non-parametric bootstrap estimates. The goodness-of-fit plots and pcVPC for the final pharmacokinetic model including the selected covariates are shown in Fig. 2. No significant bias was observed, and the central tendency and variability of the data were adequately predicted by the model. Individual single-dose and steady-state crizotinib AUC and C_MAX were derived from the final pharmacokinetic model for each patient and reported by dosage in Table 3.

Table 2.

Crizotinib final pharmacokinetic parameter estimates and non-parametric bootstraps

Parameter (unit)	Estimate (RSE%)	Non-parametric bootstraps Mean [90% CI]
Fixed-effect parameters
Absorption lag time Tlag (h)	0.36 (24.1)	0.32 [0.11–0.49]
Formulation effect βWC on Tlag	1.1 (25.6)	1.3 [0.73–2.52]
Absorption rate constant ka (/h)	0.12 (14.7)	0.13 [0.11–0.16]
Co-dasatinib effect βDas on ka	−0.24 (32)	−0.27 [−0.44 to −0.094]
Age effect βAge on ka	−0.035 (25.3)	−0.036 [−0.049 to −0.021]
Central clearance CL/F (L/h/m2)	67.4 (14.8)	68.1 [48.8–88.4]
Steady-state alone effect βSSa on CL/F	−1.05 (19.0)	−1.06 [−1.32 to −0.72]
Weight group effect βWTgcl on CL/F	−0.49 (40.1)	−0.52 [−0.81 to −0.19]
Central volume V1/F (L/m2)	88.4 (27.8)	110.1 [60.1–187.5]
Weight group effect βWTgv1 on V1/F	−0.77 (45.6)	−0.91 [−1.71 to −0.31]
Peripheral clearance Q/F (L/h/m2)	16.6 (21.5)	18.1 [13.3–24.8]
Peripheral volume V2/F (L/m2)	917 (33.6)	1047 [348.3–2155]
Random-effect parameters
IIV Tlag (sd)	0.060 (159)	0.21 [0.043–0.67]
IIV ka (sd)	0.16 (26.4)	0.15 [0.057–0.25]
IIV CL/F (sd)	0.47 (18.5)	0.46 [0.36–0.60]
IIV V1/F (sd)	0.53 (57.5)	0.38 [0.052–0.77]
IOV Tlag (sd)	0.58 (27.4)	0.57 [0.31–0.93]
IOV CL/F (sd)	0.36 (15.4)	0.35 [0.24–0.50]
IOV V1/F (sd)	0.83 (24.6)	0.81 [0.50–1.1]
Proportional residual error	0.19 (6.1)	0.19 [0.17–0.2]

RSE%: relative standard errors, CI: confidence interval, IIV: inter-individual variability, IOV: inter-occasion variability. The weight group covariate was defined upon the calculated age and gender-specific body mass index percentiles and includes two categories: normal weight (BMI percentiles ≤85%, reference), and overweight/obese (BMI percentiles >85%).

Fig. 2.

Diagnostic plots for the crizotinib population pharmacokinetic model.

Observations versus population (a) and individual (b) predictions. Normalized prediction distribution errors [NPDE] versus time (c) and population predictions (d). Prediction-corrected visual predictive checks after single-dose (e) and at steady state (f). In panels e and f, solid lines and shaded areas represent the 5^th, 50^th, and 95^th percentiles and their corresponding 90% confidence intervals of the model simulations.

Table 3.

Summary of crizotinib exposure values

Occasion	Dosage (mg/m2)	CMAX (ng/mL)	AUC0-t (h·ng/mL)
Single-dose as single agent	100	65.4 (41.9–89.8)	1010 (567–1495)
	130	79.6 (74.7–134)	1527 (1044–2497)
	165	163 (57.0–550)	2391 (797–8960)
	215	301 (72.2–636)	4296 (1398–9604)
Steady-state as single agent	100	373 (348–445)	5608 (4552–7987)
	130	401 (250–620)	6344 (4659–10088)
Steady-state in combination with dasatinib	100	101 (98.7–223)	1109 (790–2243)
	165	184 (90.6–457)	2870 (1422–6443)
	215	308 (97.3–419)	4939 (1689–6537)

Data are reported as median (range)

Area under the curve (AUC): time t refers to 48h for single-dose data and to 24h for steady-state data.

Clinical impact of selected covariates on crizotinib exposure

Model-based simulations were further conducted to evaluate the influence of the significant covariates on crizotinib AUC_ss,0–24h and C_MAX,ss after an average daily dosage of 215 mg/m². The reference population was generated by setting each significant covariate to the median or nominal values of those in the study population. The typical patient was 11-years-old, had a normal weight, and was taking crizotinib as a suspension and as a single agent. Virtual populations were generated with the following changes in covariates: crizotinib whole-capsule formulation, minimal age of 2.9 years, maximal age of 21.3 years, overweight/obese weight group, and concomitant dasatinib.

Relative changes in crizotinib steady-state exposure when covariates were varied one at a time are shown using forest plots (Fig. 3). Patient age and drug formulation did not have a clinically relevant impact on crizotinib exposure (i.e., the exposure ratios were within the 0.80–1.25 range). However, the weight group category and concomitant dasatinib treatment significantly impacted crizotinib clinical exposure. Overweight or obese patients had ~1.5-fold and 1.6-fold increase in AUC_ss,0–24h and C_MAX,ss relative to a patient with a normal weight. Concomitant dasatinib resulted in ~0.64-fold and 0.69-fold decrease in AUC_ss,0–24h and C_MAX,ss.

Fig. 3.

Forest plots of covariate effects on steady-state crizotinib exposure.

Crizotinib exposure was measured as area under the curve AUC_ss,0–24h and maximum concentration C_MAX after multiple doses of 215 mg/m² once-daily. Each covariate effect is represented by the fold change in parameter relative to the reference, calculated as the geometric mean ratio (GMR) and 90% confidence interval (CI). The solid vertical line indicates the ratio of 1 which corresponds to the typical patient (11-years old, normal weight group, taking crizotinib as suspension, without dasatinib).

Discussion

A population pharmacokinetic analysis of crizotinib as a single agent or in combination with dasatinib was performed for the first time in children and young adults with HGG and DIPG. Crizotinib disposition was described using a linear two-compartment model with a delayed absorption using a lag-time. Drug formulation and patient age were found to influence the drug absorption process but with no clinically relevant impact on crizotinib systemic exposure. Patients considered overweight and obese had significantly lower drug apparent central clearance and volume. Additionally, a potential interaction between crizotinib and dasatinib was observed with a significant change in drug absorption rate and clearance at steady state in the presence of concomitant dasatinib. Both patient weight group and concomitant dasatinib had a clinically relevant impact on drug exposure.

The pharmacokinetic properties of crizotinib in adult patients have been well characterized and reviewed [11]. Overall, crizotinib exhibits high pharmacokinetic variability, with a clearance after a single dose ranging from 86.6 to 170.5 L/h (i.e., ~ 51 to 100 L/h/m²) but from 60.1 to 81 L/h (i.e., ~ 35.4 to 47.6 L/h/m²) after repeated dosing. A potential autoinhibition of the CYP3A-mediated crizotinib metabolism was suggested to explain the decreased drug clearance at steady state [11]. The same pattern was observed in the population pharmacokinetic analysis reported by Wang and colleagues who studied 1214 patients (age 19–83 years) diagnosed with non-small cell lung cancer [21]. Crizotinib was dosed at 250 mg twice daily (~150 mg/m² in an average adult with a BSA of 1.7 m²). Crizotinib concentration-time data were best described by a linear two-compartment model with a time-dependent clearance. Crizotinib mean CL/F was estimated at 136 L/h (i.e., ~80 L/h/m²) after the first dose and the mean decreased to 76 L/h (i.e., ~44.7 L/h/m²) at steady state. The authors also reported Asian race, female sex, body weight, creatinine clearance, and total bilirubin significantly influenced CL/F, although none of those covariates were clinically relevant.

A few studies have evaluated crizotinib pharmacokinetics in children with various solid tumors and reported similar parameters to those observed in adults. In particular, Balis and colleagues studied 64 children with refractory solid tumors receiving varying crizotinib dosages (100 to 365 mg/m²/dose twice daily) and formulations (powder in capsule or bottle, and oral solution) [22]. After 215 mg/m², mean±sd single-dose and steady-state AUC were 2820±1890 and 5630±1370 h·ng/mL, respectively, with a steady-state CL/F of 39.1±9.54 L/h/m². Across all dosages, the accumulation index at steady state for crizotinib AUC was 4.9. Greengard and colleagues published the safety, tolerability, and pharmacokinetics of crizotinib in combination with cytotoxic chemotherapy in 44 children with solid tumors or anaplastic large cell lymphoma [23]. At the 215 mg/m² twice daily dosage, the authors reported the mean (CV%) steady-state CL/F as 44.4 (46%) L/h/m², and AUC_0–12h as 5935 (62%) h·ng/mL. In these pediatric studies, the single-dose crizotinib CL/F values were not reported.

The results of our pharmacokinetic study seem to agree with the previously published reports. In our analysis, the overall mean steady-state crizotinib CL/F was ~60 L/h/m² and was similar to single-dose CL/F (mean 66.7 L/h/m²). When only considering the patients that were not taking dasatinib for the pharmacokinetic study, we found that the steady-state CL/F decreased compared with the single-dose values, with a mean of 26.5 L/h/m². The corresponding steady-state AUC values were between four and fivefold higher than those obtained after a single dose (Table 3). These results may suggest that, in the absence of dasatinib co-treatment, an autoinhibition of CYP3A4-induced metabolism occurs in children and young adults and leads to a decrease in crizotinib clearance over time. This putative autoinhibition was not observed in patients receiving dasatinib at steady state. Those patients exhibited similar CL/F values and similar AUC after single-dose and at steady state (Table 3). These results suggest a drug–drug interaction between crizotinib and dasatinib, most likely related to CYP3A4-mediated metabolism, which affects both drugs [11,12]. As shown in Fig. 3, concomitant dasatinib results in significantly lower crizotinib exposure, which could potentially impact the drug efficacy.

While we were able to observe the effect of dasatinib on crizotinib pharmacokinetics, we were not able to identify a potential effect of crizotinib on dasatinib pharmacokinetics. The study was designed to only collect steady-state pharmacokinetic data for dasatinib in combination with crizotinib. Therefore, no covariate analysis was performed for dasatinib. Nonetheless, the data were analyzed using a population modeling approach. The mean±sd predicted dasatinib CL/F was 73.6±38.6 L/h/m² after 50 or 65 mg/m² dosages. Current published dasatinib pharmacokinetic data remain scarce with only single-dose parameter values reported in children [12]. One study in adults shows much higher CL/F values ranging from ~127.6 to 410 L/h/m² across 70–120 mg doses given twice-daily [24].

As noted previously, Wang and colleagues reported a statistically significant relationship between extremes in patient body weight and crizotinib CL/F with patients who were considered overweight/obese (>100 kg) having a 9% higher CL/F and low body weight patients (<40 kg) having a 9% lower CL/F. Overall, these effects were considered minimal, and the authors recommended no change in starting crizotinib dosage. While we did observe that patient weight was significantly related to CL/F in our patient population, the relationship was opposite of what was observed in adults. The overweight/obese children had a lower CL/F and higher AUC_ss,0–24h as compared to normal weight children. Our finding of a decreased CL/F in overweight/obese children is not unexpected since crizotinib is primarily metabolized by CYP3A4/5 enzymes, which have been previously shown to be less active in obese subjects [25,26]. It should be noted that the observed weight effect on both CL/F and V₁/F may reflect an actual weight effect on crizotinib bioavailability rather than on the clearance and volume of distribution. Additionally, changes in the absorption rate (ka) given differences on the activity of CYP3A4/5 enzymes between obese subjects and normal weight patients could explain the observed changes on crizotinib CL/F and V₁/F. In this study, we could not establish any relationships between crizotinib exposure and effect, i.e., antitumor activity or toxicities. Therefore, no dosing adjustment is proposed for overweight and obese patients. However, this specific population should be monitored when receiving crizotinib, as patients are likely to exhibit much higher drug exposures as shown in Fig. 3.

While crizotinib is hepatically metabolized by CYP3A4/5 enzymes, previous studies of crizotinib disposition in patients with mild hepatic impairment suggested no dosage adjustment was necessary [27]. Wang and colleagues concluded that in patients with mild or moderate hepatic dysfunction (i.e., total bilirubin ≤ 2.1 mg/dl) the starting crizotinib dosage of 250 mg twice daily was appropriate. Unpublished data from the sponsor suggested that mild and moderate renal impairment did not have a clinically significant effect on crizotinib AUC, whereas severe renal impairment increased AUC by 79% [21]. It is likely that in our patient population with a relatively normal and narrow range of hepatic and renal function, we would not have identified creatinine clearance or bilirubin as significant covariates.

The present study had several limitations. Despite a reasonable number of pediatric patients enrolled in the pharmacokinetic analysis, the sample size was not large enough to create a validation group, and no external dataset was available to validate the model. Because of toxicities, the treatment design was modified which led to a change in the pharmacokinetic study design. As a result, only seven patients had steady-state pharmacokinetic data available for crizotinib as a single agent, while all patients had steady-state data for the drug combination. We recognize that overall, the relatively small sample size might have restricted our ability to identify finer trends in the data. Nonetheless, we have showed relevant findings for this young population that agree with previously published reports.

In addition to monitoring overweight and obese patients potentially at risk of high crizotinib exposures, our results also showed that when crizotinib is administered with dasatinib, lower crizotinib exposures are observed. Similar results might be expected for the combination of crizotinib with other tyrosine kinase inhibitors that have similar metabolic profiles as dasatinib. This possible interaction could lead to lower crizotinib exposures and could favor higher crizotinib dosing regimens, however, increased toxicity of the combination might limit the ability to dose increase crizotinib.