Population Pharmacokinetics of Revefenacin in Patients with Chronic Obstructive Pulmonary Disease

All participants gave written informed consent before any study-related procedures, and the protocols were approved by the appropriate institutional review board [Electronic Supplementary Material (ESM)] for each study site and were carried out in concordance with ICH Guidelines for Good Clinical Practice [1‐3, 7]. Data for the analysis were obtained from three phase II studies (studies 1, 2, and 3) and two phase III studies (studies 4 and 5) in patients with COPD. The inclusion criteria for each study are outlined in the ESM. Participants in the phase II studies received revefenacin once daily as a nebulized solution at doses from 22 to 700 μg for 1, 7, or 28 days. Participants in the phase III studies received revefenacin once daily as a nebulized solution at 88 or 175 µg for 12 weeks. Pharmacokinetic sampling times are summarized in Table S1 of the ESM.

Plasma samples from the three phase II and the two phase III studies were analyzed at Q2 Solutions (formerly Quintiles, Inc. and Advion Bioservices, Inc., Ithaca, NY, USA) using validated liquid chromatography with tandem mass spectrometry methods. The lower limit of quantification for revefenacin and THRX-195518 in plasma was 0.005 ng/mL and 0.05 ng/mL for the phase II studies and 0.0005 and 0.005 ng/mL, respectively, for the phase III studies. The lower limit of quantification for revefenacin in the phase III studies represents a ten-fold increase in assay sensitivity relative to the assay used for the phase II studies to allow for the quantification of revefenacin in the trough samples collected in the phase III studies.

Data from all studies were pooled for the integrated population PK analysis using nonlinear mixed-effects modeling using the first-order conditional estimation method with interaction in NONMEM 7.2 (ICON, Dublin, Ireland) and PLT Tools (PLT Soft, San Francisco, USA). Models were compared using the mean value of the objective function computed as − 2 times the log-likelihood.

One-, two-, and three-compartment models were considered to describe revefenacin and THRX-195518 concentration–time data [8]. The formation rate of THRX-195518 was defined in the model to be a fraction of the clearance rate of revefenacin from the central compartment. All models assumed lognormal distributions of the individual PK parameters, with a mixed residual error model. A study-specific additive residual error for revefenacin was used for the observations from phase III because of the ten-fold difference sensitivity in the assay.

A sequential approach was used whereby the revefenacin data were initially fit to the revefenacin model. Then, the THRX-195518 data were fit to the combined model where PK parameters for revefenacin for each subject were fixed to the individually estimated post-hoc PK parameters from the revefenacin model. Inter-individual error terms for the revefenacin PK parameters were included in the fit to the THRX-195518 data to minimize the effect of bias.

A simultaneous modeling strategy was also evaluated in addition to the sequential approach; the parameters for revefenacin and THRX-195518 were simultaneously estimated to investigate the potential bias of the sequential approach because of the presence of shrinkage in the post-hoc individual parameters of the revefenacin model. However, the simultaneous approach introduced the potential issue of the propagation of uncertainty in model and parameter estimates from/to revefenacin to/from THRX-195518 with the known risk of increasing run-time and model instability.

Inter-occasion variability was evaluated by comparing the measured plasma concentrations of revefenacin and THRX-195518 to determine the presence of any differences between study visits. Concentrations below the limit of quantification were discarded in this analysis (M1) as the increased sensitivity of the revefenacin assay in the phase III studies allowed for the characterization of trough concentrations in all patients. A sensitivity analysis was conducted by repeating the analysis for revefenacin using the likelihood maximization (M3) method [9].

Intrinsic covariates of age, weight, sex, race, smoking status, creatinine clearance, liver enzymes (alanine aminotransferase, aspartate aminotransferase, and total bilirubin), body mass index, and baseline forced expiratory volume in 1 s (FEV₁) were tested on the PK parameters of apparent revefenacin clearance (CL/F), apparent revefenacin volumes of the central (V₁/F) and peripheral (V₂/F) compartments, apparent revefenacin intercompartmental clearance (Q/F), the fraction of CL/F that is metabolized to THRX-195518 (F_met), apparent THRX-195518 clearance (CL_met/F), apparent THRX-195518 volumes of the central (V₃/F) and peripheral (V₄/F) compartments, and apparent THRX-195518 intercompartmental clearance (Q_met/F).

Continuous covariates were normalized to the population median values and modeled using the general equation:where θ_i is the value of the parameter for the ith individual, θ_Typical is the typical value of the parameter in the population, Cov_i is the value of the covariate for the individual, Cov_Median is the median value of the covariate in the study population, and θ_eff is the effect of the covariate on the parameter.

Categorical covariates were modeled using the general equation:where K_ind is an indicator variable representing one form of the categorical variable, e.g., men are coded as 1 and women as 0.

Model selection was based on a comparison of the objective function value and visual inspection of goodness-of-fit plots. Significant parameter covariate relationships identified were included in an initial full PK model (a decrease of 6.63 points in the value of the objective function equivalent to a two-sided α = 0.01). Covariates were subsequently excluded from the model using a stepwise deletion method in which the statistical significance of each parameter-covariate relationship was tested using a likelihood ratio test (an increase of 10.83 points in the value of the objective function, equivalent to a two-sided α = 0.001). Evaluation of the model fit was conducted using a bootstrap analysis (200 replicates) and a visual predictive check of the final model using 1000 simulated data sets.

The steady-state exposure in each subject in the two phase III studies was simulated using the individual post-hoc parameter estimates from the final population PK model to compare exposures between smokers and nonsmokers, patients with and without concomitant medications with known potential drug–drug interactions, and between patients with and without LABA/ICS use. Nonlinear mixed-effects modeling was used to simulate the steady-state plasma concentrations at 0, 0.01, 0.25, 0.5, 1, 2, 4, 6, 8, 12, 16, and 24 h after the start of nebulization using individually estimated PK parameters (CL/F, V₁/F, Q/F, V₂/F) for each subject as determined by the final population PK model. The dose and interval were obtained from the study data. The duration of nebulization was not recorded in the phase III studies and was assumed to be 10 min for all patients based on observed dosing durations in prior studies. Steady-state exposure (AUC_0–24) was calculated using linear trapezoidal approximation and C_max was the maximum observed post-dose concentration in the simulated plasma concentration–time profile.

The baseline demographics and clinical characteristics of patients in this analysis are listed in Table 1. The PK analysis population comprised 935 patients aged 41–88 years, weight 38.5–192 kg, estimated creatinine clearances from 22 to 151 mL/min, and baseline FEV₁ values from 44 to 2962 mL. The study population was 52.2% men; 46% of patients were current smokers; 32.5% of patients in the PK analysis population were receiving concomitant LABA/ICS therapy. Pharmacokinetic sampling times used in the analysis are described by study in Table S1 of the ESM.

Systemic revefenacin pharmacokinetics was best described by a two-compartment model with fixed first-order absorption (K_a) from the dosing depot (representing the lung). Terms used include relative bioavailability (F1) and distribution and elimination parameterized by CL/F, V₁/F, Q/F, and V₂/F with inter-individual variability (IIV) terms on CL/F, V₁/F, Q/F, V₂/F, and F1 with a combination additive and proportional residual error model for phase II study data and a separate proportional residual error model for phase III study data (Fig. 1). The F1 in the structural model included a separate categorical covariate term to represent observations from Study 1 to reflect the observed lower exposures, and a continuous covariate term to represent the relationship between dose and bioavailability.

In the final model (Table 2), the population estimate of CL/F was 668 L/h with an IIV of 56.2%; the typical V₁/F was estimated to be 867 L with an IIV of 26.9%; Q/F was estimated to be 2607 L/h with an IIV of 31.0%; and the population value of V₂/F was estimated to be 15,495 L/h with an IIV of 52.2%. First-order absorption was fixed at 200/h, the effect of Study 1 on F1 was 55.3% with an IIV of 33.7%, and the effect of the dose on F1 was 0.0987.

The analysis was repeated using the maximum likelihood (M3) method to assess the potential risk of bias. A sensitivity analysis conducted on the final structural model to compare the parameter estimates using the M1 and M3 methods did not show significant differences in the parameter estimates, indicating the absence of bias in the parameter values estimated with the M1 method. The corresponding η-shrinkage for the revefenacin model was 13.5% for CL/F, 42.5% for V₁/F, 33.7% for Q/F, and 46.8% for V₂/F.

The appearance of the metabolite THRX-195518 in the systemic circulation was assumed to be a fraction of the revefenacin cleared from the systemic circulation and is structurally unidentifiable. It was not possible to simultaneously estimate F_met, V₃/F, and CL_met/F based on the measured concentration as the total amount of THRX-195518 formed per dose is unknown. The approach used to address the parameter identifiability problem was to estimate CL_met/F and V₃/F while fixing F_met. The value of F_met was chosen to match the total fraction of the dose recovered as THRX-195518 in the urine and feces (21%) following a single IV dose in the human absorption, distribution, and metabolism, and excretion study [6].

The model for THRX-195518 used the post-hoc individually estimated revefenacin PK parameters for each individual subject as the basis for the kinetics of THRX-195518 formation in the dataset. THRX-195518 pharmacokinetics was best described by a two-compartment model (Fig. 1) with a fixed value for F_met and PK characteristics parameterized by CL_met/F, V₃/F, Q_met/F, and V₄/F with IIV terms on CL_met/F and V₃/F with a combination additive and proportional residual error model for phase II study data and a separate proportional residual error model for phase III study data (Table 2). Inter-individual variability in the PK parameters CL_met/F and V₃/F was moderate (36.0% and 52.1%). The corresponding η-shrinkage for the THRX-195518 model was 30.5% for CL_met/F.

Covariates of age, weight, sex, race, smoking status, creatinine clearance, body mass index, and baseline FEV₁ were tested for significance on the PK parameters CL/F, V₁/F, Q/F, V₂/F, CL_met/F, F_met, Q_met/F, and V₄/F. Additionally, the effects of baseline FEV₁ were tested on the PK parameters CL/F, V₁/F, Q/F, and V₂/F because of the potential impact of respirational capacity on the kinetics of revefenacin absorption. The effect of liver enzymes was included by the addition of alanine aminotransferase, aspartate aminotransferase, and total bilirubin as covariates on the formation and clearance of THRX-195518. Hepatobiliary elimination of revefenacin was assumed from the results of the absorption, distribution, and metabolism, and excretion study indicating the clearance of revefenacin via hepatobiliary elimination [6] and data from the hepatic impairment study demonstrating increased exposures to THRX-195518 in patients with moderate hepatic impairment [10].

The covariate analysis identified age as a statistically significant covariate on CL/F, and body weight was a significant covariate on Q/F of revefenacin. Age was identified as a statistically significant covariate on CL_met/F, and body weight as a statistically significant covariate on F_met.

There is considerable overlap in the steady-state revefenacin PK profile and exposure over the entire age and weight range of patients in the study (Fig. 2a). The AUC_0–24 in the median (64-year-old) subject following a 175-µg dose is predicted to be 0.332 ng·h/mL (covariance 74.8%). The corresponding predicted revefenacin exposures are 24% lower in a younger 40-year-old subject, and 18% higher in an older 85-year-old subject. The corresponding predicted revefenacin exposures are 3% lower in a 50-kg subject and the same as a 150-kg subject. A comparison of the individually predicted revefenacin exposures of all patients in the phase III patients does not indicate any significant differences across different age groups and weights (Fig. 2b). The effect of age and weight on revefenacin exposure is therefore considered to be minimal and does not warrant any dose adjustment. Effects of age and weight on THRX-195518 pharmacokinetics are further described in the ESM (Fig. S1a and b).

The effect of age, weight, sex, smoking status, and concomitant LABA/ICS therapy on exposure (AUC_0–24 and C_max) was evaluated by comparing the individually predicted revefenacin and THRX-195518 exposures of all patients in the phase III studies. No clinically meaningful effect of age, weight, sex, smoking status, and concomitant LABA/ICS therapy was observed on exposure (AUC_0–24 and C_max) (Fig. 3a, b), and therefore, these intrinsic and extrinsic factors do not warrant any dose adjustment.

The final model was re-estimated using a lower and higher value for the rate of absorption, K_a, (100 and 1000/h vs 200/h in the final model) and did not result in an improvement of the fit based on the value of the − 2 log-likelihood function (p < 0.1). The revefenacin and THRX-195518 AUC_0–24 and C_max values predicted using the population PK model with and without the inclusion of quantifiable pretreatment values showed minimal differences, with a point estimate of 0.0072 ng/mL for the residual additive error. The inclusion of the quantifiable pretreatment values reduced the mean predicted steady-state revefenacin AUC_0–24 estimates by 8% at the 88-µg dose (0.170–0.157 ng h/mL) and by 3% at the 175-µg dose (0.329–0.319 ng h/mL), and the THRX-195518 AUC_0–24 estimates were reduced by 7% at the 88-µg dose (0.426–0.396 ng h/mL) and by 1% at the 175-µg dose (0.859–0.849 ng h/mL). The mean predicted steady-state C_max estimates for revefenacin and THRX-195518 were nearly identical. Further analysis was conducted on data from the phase II studies only, which did not have quantifiable pretreatment concentrations. The resulting covariate effects determined by the phase II only model and phase II/III combined model were in agreement. Therefore, the inclusion of quantifiable pretreatment concentrations in the dataset did not significantly alter the results of the analysis.