Modeling and Simulation Analysis of Aprepitant Pharmacokinetics in Pediatric Patients With Postoperative or Chemotherapy-Induced Nausea and Vomiting

Introduction

Chemotherapy-induced nausea and vomiting (CINV) are common and distressing side effects¹² that can negatively impact health-related quality of life¹ and chemotherapy adherence.³ The risk for CINV varies according to age, with lower risk in younger (birth to 3 years) versus older children.⁴ To prevent CINV in children undergoing chemotherapy, guidelines developed by the Multinational Association of Supportive Care in Cancer/European Society for Medical Oncology recommend a combination of the neurokinin-1 receptor antagonist aprepitant plus a 5-HT3 antagonist and dexamethasone in patients receiving highly emetogenic chemotherapy, and a 5-HT3 antagonist plus dexamethasone in patients receiving moderately emetogenic chemotherapy.⁵

Aprepitant is a potent, selective, oral neurokinin-1 receptor antagonist approved in the United States and Europe for the prevention of CINV in pediatric patients ages 6 months or older and adults, and for postoperative nausea and vomiting (PONV) in adults.⁶⁷ Aprepitant is available as oral capsules or an oral suspension, which is prepared by reconstituting aprepitant powder in water.⁶⁷ Given the low solubility of aprepitant, fosaprepitant, a water-soluble prodrug of aprepitant, was developed to facilitate IV administration⁸; this IV formulation is also approved in the United States and Europe for prevention of CINV in adults.⁹¹⁰ After a 15-minute IV infusion of fosaprepitant in adults, the plasma elimination half-life averaged approximately 2.3 minutes, and the volume of distribution was approximately 5 L.¹¹ Plasma levels of fosaprepitant are generally below the level of detection (10 ng/mL) within 60 minutes after the infusion.⁸ Therefore, it has been concluded that fosaprepitant is rapidly and completely converted to aprepitant, and clinical efficacy following fosaprepitant administration is derived from circulating aprepitant exposures. Following oral administration and across the clinical dose range in adults, aprepitant exhibits nonlinear pharmacokinetics (PK), with a 14% (fasted state) or 26% (fed state) greater-than-dose-proportional plasma area under the concentration-time curve from time 0 to infinity (AUC_0-∞) between the 125- and 80-mg doses.¹² The apparent volume of distribution of aprepitant in adults at steady state is approximately 70 L, and greater than 95% is bound to plasma proteins.⁸ Aprepitant undergoes extensive metabolism, primarily by cytochrome P450 (CYP) 3A4, with minor metabolism by CYP1A2 and CYP2C19.⁸ The apparent terminal half-life ranged from approximately 9 to 13 hours in adults.⁸

The current analysis describes the development and validation of a population PK model based on data from 3 clinical studies, including one earlier phase 3 trial in adolescent patients,¹³ as well as 2 phase 1 trials (ClinicalTrials.gov identifiers: NCT00818259 and NCT00819039),¹⁴¹⁵ to guide an optimal dosing strategy for the oral administration of aprepitant in a pediatric population being treated for prevention of CINV.

Materials and Methods

Patients and Samples. Study 1 was a randomized, double-blind, placebo-controlled phase 3 trial of aprepitant for the prevention of CINV in adolescent patients aged 11 to 19 years (N = 50, study P097, ClinicalTrials.gov identifier NCT00080444).¹³ Patients with a Karnofsky performance score of >60% and confirmed malignancy who were scheduled to receive single- or multiday emetogenic chemotherapy or a previously intolerable chemotherapy due to CINV were included. Patients were excluded if they had a history of any major active (within 5 years) cardiac or vascular disorder; other pulmonary disease; major gastrointestinal abnormalities or peptic ulceration; neurologic, endocrine, hematologic, or renal disease; or major genitourinary disease. The current analysis includes only data from trial patients in the aprepitant treatment arm with PK samples (n = 18; aged 12–19 years). Aprepitant (125 mg on day 1, 80 mg on days 2 and 3) was administered as a capsule and taken orally 1 hour before chemotherapy, in combination with ondansetron (0.15 mg/kg IV, 3 doses) on days 1 and 2, and oral dexamethasone (8 mg on day 1, and 4 mg on days 2 to 4), which were both administered 30 minutes before chemotherapy. Blood samples were collected at predose (−2 hours), 1 hour (before chemotherapy infusion), and at 2, 3, 4, 8, 12, and 24 hours after the aprepitant dose, and then every 24 hours for the next 2 days.

Study 2 was a multicenter, open-label phase 1 study of aprepitant and fosaprepitant dimeglumine in pediatric patients aged 6 months to 17 years, for the treatment of CINV (N = 104, study P134, ClinicalTrials. gov identifier NCT00818259).¹⁴ Patients scheduled to receive single-day or multiday moderately or highly emetogenic chemotherapy or a previously intolerable chemotherapy were included. Five different dosing regimens were evaluated, including a control regimen consisting of only ondansetron. The current analysis includes PK data from 62 pediatric patients (<12 years) in this study who received a 3-day oral aprepitant regimen equivalent to 125 mg on day 1 (administered approximately 105 minutes before chemotherapy initiation) followed by 80 mg on days 2 and 3; and 23 adolescent patients (aged 12 to 17 years) who received an adult 3-day regimen including 115 mg of fosaprepitant on day 1 (administered 75 minutes before chemotherapy initiation) followed by 80 mg of oral aprepitant on days 2 and 3. In the pediatric population, blood samples were collected before dose, and 1.5, 3, 4, 6, 8, and 24 hours after the aprepitant dose on day 1 and 24 hours after the aprepitant dose on days 2 and 3. In the adolescent population, blood samples were collected before dose, and at 0.25, 0.5, 0.75, 1, 1.5, 3, 4, 6, 8, and 24 hours after the fosaprepitant dose on day 1 and 24 hours after the aprepitant dose on days 2 and 3.

Study 3 was a multicenter phase 1 study of aprepitant in pediatric patients undergoing surgery, for the treatment of PONV (N = 55, study P148, ClinicalTrials. gov identifier NCT00819039).¹⁵ Pharmacokinetic data were included from 35 patients aged 6 months to 12 years who weighed ≥6 kg and received single doses of aprepitant as an oral suspension (1.2 mg/kg), and 9 adolescents aged 12 to 17 years who received a single 40-mg capsule. Blood samples were collected before dose, and at 1, 2, 3, 4, 8, 12, 24, and 48 hours after the aprepitant dose. The characteristics of patients included in the population PK analysis from the 3 clinical studies are shown in Table 1.

Table 1. Patient Characteristics Included in the Population Pharmacokinetic Analysis

View Fullscreen

Pharmacokinetic Assessments and Analysis. Plasma aprepitant concentrations were measured by high-performance liquid chromatography tandem mass spectrometry, with 1 assay used for study 1¹⁶ and a different assay for studies 2 and 3. Both assays were validated. Concentrations were log₁₀-transformed before modeling and simulation; concentrations below the assay quantification limit (<10 ng/mL) were set to 0 and were excluded from the log₁₀-transformed data. A sensitivity analysis was performed including data below the quantification limit and found no relevant differences from the final model (data not shown).

The data analysis population included all enrolled patients who received aprepitant or fosaprepitant and had at least 1 postdose study drug concentration measured in plasma. Pharmacokinetic data were modeled using nonlinear mixed-effects modeling (NONMEM, version 7.2), using the first-order conditional estimation without the interaction option, and stepwise covariate selection was performed using Perl-speaks-NONMEM (PsN), version 3.7.6. Statistical analysis, computations, and figures were produced using R (version 3.0.2) software.

Population PK Model Development. The population PK model was developed using data from the 3 clinical studies (Table 1), with the assumption that the bioavailability of IV fosaprepitant was 100% and that fosaprepitant was instantly and completely converted to the active metabolite aprepitant.

The 3 stages of base model development consisted of a structural model, an interindividual model, and a residual variability model. The initial base structural model, derived from preliminary data from the 3 clinical studies, was a 2-compartment linear model with first-order absorption and lag time (Figure S1). The structural model included an allometric scaling factor accounting for the effect of body size on clearance parameters (systemic clearance [CL] and intercompartmental clearance [Q]) and volume (central volume of distribution [V2] and peripheral volume of distribution [V3]),¹⁷ as well as a CYP3A4 maturation function to account for the effect of ontogeny (using age as the variable) on CL.¹⁸ A closed-form solution and simultaneous fit of oral aprepitant and IV fosaprepitant data and bioavailability estimation was allowed via the use of ADVAN4 and TRANS4 NONMEM subroutines. Interindividual variability was analyzed for each parameter using a log-normal random effects model, and the effect was considered significant if the change in objective function value (OFV) ≥3.84 (p ≤ 0.05). Residual variability was modeled using the log-additive error model. Stability of the base model during development was assessed by verifying that extreme pairwise correlations (>0.95) of parameter estimates were not encountered in the covariance matrix, and that the condition number of the correlation matrix of parameter estimates (i.e., the ratio of the largest to smallest eigenvalues), was less than 1000.

The potential effect of covariates on aprepitant PK was evaluated using the base model. Assessed covariates included age (ontogeny, as a maturational function, as suggested by Johnson et al¹⁸), sex, race, formulation (IV, oral capsule, or oral suspension), and indication (CINV or PONV; Table 1), and these effects were assessed simultaneously (full covariate model).

Continuous covariates were assessed using: P* = θ_x (covariate/covariate median)^θy, where P* is a typical value of a PK parameter and θ_x and θ_y are fixed-effect parameters.

Categoric covariates were assessed using: where Q is an index variable.

A stepwise addition/elimination procedure was used to determine whether covariates in the full covariate model would be included in the final model, based on the change in OFV and the associated probability (p value). Covariate effects were included in the forward-addition step if the p value for the nested model was less than 0.01 (for a single degree of freedom, ΔOFV >6.63), and in the backward-elimination step if the p value was less than 0.001 (for a single degree of freedom, ΔOFV >10.83). Covariates were included in the final model if they met the predefined statistical criteria and had relative standard errors <40%.

Model Assessment. Diagnostic plots were used to visually assess goodness of fit, including population predictions (PRED) versus observed values (DV) and individual predictions (IPRED), as well as conditional weighted residuals (CWRES) versus time, time after last dose, and PRED, and individual weighted residuals versus IPRED.¹⁹ Nonparametric bootstrap and prediction-corrected visual predictive checks (n = 1000) with appropriate data stratification (formulation, age group) were used to evaluate the final population model. Prediction-corrected visual predictive checks allow for accurate visual presentation of simulation results, without loss of power due to data stratification.¹⁹ For visual predictive checks, the 5th, 50th (median), and 95th percentiles (and their corresponding 95% CIs) of the distributions of simulated aprepitant concentration-time profiles and observed percentiles by time since last dose were compared. The geometric mean estimates of clearance (CL) and central volume of distribution (V2) in each age group were tested to meet the recommended 60% to 140% CIs around point estimates.²⁰ Shrinkage estimates were calculated for all variability components.

Simulations. The final population PK model was used to simulate aprepitant exposures following dosing with various regimens and formulations in 2 age groups: 0 to 12 years (pediatrics) and 12 to 19 years (adolescents/young adults), using data from studies 1 to 3. Each age group was assigned to a 3-day oral dosing regimen of aprepitant: oral suspension 3 mg/kg on day 1 and 2 mg/kg on days 2 and 3 for patients aged 6 months to 12 years, and oral capsules 125 mg/kg on day 1 and 80 mg on days 2 and 3 for patients aged 12 to 19 years. Aprepitant exposure for each dose/age group was simulated with uncertainty using the final model bootstrap results (n = 1000). Pharmacokinetic parameters (C_max, t_max, C_24hr, C_48h, C_72hr, AUC_0–24hr) were computed using noncompartmental analysis based on simulated rich concentration-time profiles, and partial areas under the concentration-time curve (AUC_0–24hr) were computed using the linear trapezoidal rule. To assess whether pediatric dosing strategies adequately produced aprepitant exposures comparable to those observed in adults, PK parameters for each pediatric age group were visually assessed and compared with population estimates from healthy adults and adult cancer subjects treated in two previous studies.²¹²²

Results

Patient Samples. The full data set for the population PK analysis included 147 patients with a total of 1642 samples, of which 1326 were measurable for aprepitant concentrations. The average age of patients was 8.1 years (range, 6 months to 19 years). Approximately half of the patients were male, most (76%) were white, and 70% had CINV (Table 1).

Development of the Population PK Model. The final population PK model was a 2-compartment model with linear elimination and first-order absorption, with allometric scaling for clearance and volume parameters, and a CYP3A4 maturation component for systemic clearance. Dose affected systemic clearance; clearance decreased above the reference dose of 80 mg and increased at doses lower than 80 mg, and this was included in the model as a power function: CL = 6.32 × (dose/80)^−0.394.

Because of potential differences in relative liver/body size, allometric scaling was incorporated using the fixed exponent of 0.75 for clearance parameters (CL and Q) and 1 for volume parameters (V2 and V3), as reported by Albers et al.¹⁷

The CYP3A4-mediated hepatic metabolism is the predominant clearance route of aprepitant,²³ and to account for the effect of ontogeny on CL,¹⁸ a maturation function for gut CYP3A4 was incorporated as: CL = CL × 0.639 × age/(2.4 + age) + 0.42.

Other maturation functions (both hepatic¹⁸²⁴) were tested to evaluate the appropriateness of the maturation component used. However, when comparing these alternatives with the original model, only marginal differences in CL were observed, and most PK parameters differed by <1%, suggesting data in the model were relatively insensitive to various maturation functions used to describe CL.

During the assessment of the potential effect of covariates on PK parameters in the base model, PONV initially appeared to affect clearance. However, based on the visual predictive check, incorporation of PONV into the clearance function overestimated clearance at higher aprepitant dose levels, which appeared attributable to a lower dose of aprepitant in the PONV population (maximum dose, 1.2 mg) compared with the CINV population (maximum dose, 4.1 mg). Therefore, effect of dose on clearance was investigated and was determined to improve model fit, and its inclusion resulted in no effect of PONV on clearance.

The final model equations for typical values (tv) of CL, Q, V2, and V3, along with parameter estimates, are shown in Table 2.

Table 2 Typical Population Pharmacokinetic (PK) Parameters of Aprepitant in Pediatric Patients in the Final Population PK Model

View Fullscreen

Model Assessment. Basic goodness-of-fit diagnostic plots of aprepitant concentrations versus population predictions and individual predictions for the final model showed the absence of model misspecifications, and the residual error model appeared adequate, as demonstrated by no apparent trend in the CWRES plots (Figure 1).

View Figure

• Download as Powerpoint
• Download Figure

Precision was generally high, with relative standard error estimates <20% across all PK parameters (Table 2). Shrinkage estimates were relatively low for CL (17%) and higher for intercompartmental clearance (51%), indicating that the current data set had more information, enabling more accurate estimation of individual CL values compared with the individual intercompartmental clearance values, which require richer data for better estimation.

The final population PK model was validated using prediction-corrected visual predictive check with stratification by age group (Figure 2). The observed median (solid black line) and 10th and 90th percentiles (dotted black lines) of observed data generally fell within the areas representing the respective prediction intervals of the simulated data.

View Figure

• Download as Powerpoint
• Download Figure

Typical systemic CL and V2 of aprepitant in patients were 6.3 L/hr and 43 L, respectively, using the final population PK model. Estimated CL and V2 were 1.81 L/hr and 6.15 L, 2.28 L/hr and 9.06 L, 3.93 L/hr and 18.34 L, and 4.58 L/hr and 33.19 L, respectively, for the age groups of 6 months to 2 years, 2 to 6 years, 6 to 12 years, and 12 to 19 years (Table 3). By both variance-covariance matrix and nonparametric bootstrap analyses, the 95% CIs for the geometric mean estimates of systemic CL and V2 were within the predefined and US Food and Drug Administration–recommended 60% to 140% criteria²⁰ within each age group.

Table 3 Main Population Parameter Estimates of Aprepitant in Pediatric Patients and 95% CI by Age Group Computed From the Final Model Variance-Covariance Matrix, and the Final Model Nonparametric Bootstrap

View Fullscreen

Simulations. As described in Figure 3A and B, the predicted concentration-time profiles adequately described the observed data, using model-based simulation with uncertainty (n = 1000 samples). The observed peak (C_max) and total (AUC_0–24hr) exposure to aprepitant were in agreement with the median model-predicted PK parameters, differing only by 6.5% and 1.6%, respectively, for pediatric patients (6 months to 12 years), and 6.0% and 2.4%, respectively, for adolescent patients (12–19 years; Supplemental Table S1). Aprepitant trough concentrations on day 1 (C_24hr) were generally well predicted, with a difference between the observed value and the predicted value of 11.7% for pediatric patients and 0.6% for adolescents. After 24 hours, the aprepitant concentrations were more variable, with differences between median predicted and observed trough concentrations on days 2 and 3 of 23% to 123% in pediatric patients and 28% to 30% in adolescents. Simulated median values for PK parameters within each pediatric age group (0.5–2 years, 2–6 years, and 6–12 years) given a 3-day weight-based regimen of 3 mg/kg (day 1), 2 mg/kg (day 2), and 2 mg/kg (day 3) oral suspension were within the range of observed values for adolescent patients given a 3-day fixed-dose regimen (125, 80, and 80 mg) of oral capsules, and within the range of exposures observed in healthy adults and adult cancer patients given a 3-day fixed-dose regimen of oral aprepitant capsules (125, 80, and 80 mg) in 2 previous studies²¹²² (Figure 4).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Discussion

This analysis evaluated the population PK of aprepitant in children and adolescents using a large data set (147 patients) from 3 clinical studies. A population PK model for a pediatric aprepitant regimen was developed, consisting of first-order absorption, 2-compartment distribution, and linear elimination, with an allometric scaling factor for clearance and volume parameters, and a CYP3A4 maturation factor to account for the effect of ontogeny (using age as the variable) on clearance. Covariate analysis initially suggested that indication (PONV) affected clearance (1.44× increase); however, this was likely confounded by the lower dose received by the PONV versus the CINV population, because inclusion of the dose effect on CL gave an adjusted PONV estimate of 1.01× increase. Doses below 80 mg increased CL (1.25× at a dose of 46 mg), whereas doses above 80 mg decreased CL (0.8× at a dose of 141 mg), and incorporation of the dose effect on clearance improved the overall model fit. CYP3A4 is the main route of aprepitant metabolism,²³ and given that 80% of the data are based on oral administration, a pre-systemic (gut) CYP3A4 maturation function adequately described the data; however, it should be noted that the maturation effect was minor and likely attributable to a patient population that exceeds 6 months of age and whereby maturation of CYP3A may approximate those in adults. We also assessed whether the inclusion of a hepatic CYP3A maturation function would alter the results we observed using the gut function. Differences in CL were small and inconsistent, and most differences in predicted PK were within <1% of each other. We therefore found that the original model including a gut CYP3A maturation function adequately described the PK of aprepitant across the pediatric continuum evaluated. Adequacy of the final population PK model, which included fixed scaling factors for body weight (allometry) and ontogeny (CYP3A4 maturation), was assessed using multiple methods, including goodness-of-fit plots, visual predictive checks, shrinkage, and overparameterization for diagnostic purposes. These multiple approaches did not change the results, indicating that the integrated model can be applied in the setting of CINV and/or PONV and following IV (fosaprepitant) and/or oral (aprepitant) administration across pediatric patients. Importantly, estimates from the final population PK model demonstrated that for clearance and volume of distribution, the 95% CIs for the geometric mean estimates were within the 60% to 140% criterion proposed by the US Food and Drug Administration within each pediatric age group, meeting expectations for the precise estimation of important PK parameters and justifying the PK characterization of this oral aprepitant pediatric dose regimen.²⁰ Furthermore, median estimated PK parameters for oral suspension in pediatric age groups were within the range reported for oral capsules in adolescent and adult patients, supporting comparable exposure with these pediatric dose regimens to studies of adolescents and adults in which clinical efficacy and safety of aprepitant has been demonstrated.

Clinical data in pediatric patients provides evidence of clinical efficacy of aprepitant for the prevention of CINV using different dosing regimens for children aged 6 months to <12 years (individual weight-based dosing) and adolescents aged 12 to 17 years (fixed adult dosing).²⁵ In this phase 3 trial, adolescents were given adult doses of oral aprepitant capsules based on the results of a phase 3 trial demonstrating that plasma drug concentrations were not significantly lower in adolescents compared with historical data from adults.¹³ Clinical results from this current study validated this approach, because the proportions of patients achieving complete response were similar between these 2 age groups (6 months to 12 years and 12–17 years).²⁵ However, complete response rates were lower than those reported for adults in other trials.^26–29

As with all modeling and simulation studies, several limitations must be considered when evaluating the results. In the current study, there were differences in model-predicted and observed trough concentrations, although this may be explained by the high variability of observed data after 24 hours. Of note, 19% of samples overall were below the assay detection limit, which was higher than would be expected. Finally, patients with CINV and PONV were pooled in the analysis, as were the oral (aprepitant) and IV (fosaprepitant) formulations; however, by incorporation of additional components into the model, any confounding resulting from these variables was minimized.

Conclusion

The current analysis demonstrates the importance of PK population modeling studies for estimating drug exposure in pediatric patients. Dosing by individual weight (i.e., weight-based) is a valid approach to aprepitant dosing in pediatric patients, whereas fixed dosing can be used in adolescents.