Clin Pharmacokinet DOI 10.1007/s40262-014-0143-4
ORIGINAL RESEARCH ARTICLE
Population Pharmacokinetics of Inhaled Umeclidinium and Vilanterol in Patients with Chronic Obstructive Pulmonary Disease Navin Goyal • Misba Beerahee • Chris Kalberg • Alison Church • Sally Kilbride • Rashmi Mehta
Ó Springer International Publishing Switzerland 2014
Abstract Background and objectives A fixed-dose combination of the bronchodilators umeclidinium and vilanterol is in development for the long-term, once-daily treatment of chronic obstructive pulmonary disease (COPD). We characterized the pharmacokinetics of umeclidinium and vilanterol in &1,635 patients with COPD, evaluating the impact of patient demographics and baseline characteristics on umeclidinium and vilanterol exposure. Methods Plasma concentrations of umeclidinium and vilanterol were evaluated in patients enrolled in two phase III, randomized, double-blind, parallel-group, placebo-controlled trials using inhaled umeclidinium/vilanterol combination therapy and inhaled umeclidinium and vilanterol monotherapies as treatments. Population-pharmacokinetic models were developed using non-linear mixed-effects
ClinicalTrials.gov identifiers NCT01313637; NCT01313650.
Electronic supplementary material The online version of this article (doi:10.1007/s40262-014-0143-4) contains supplementary material, which is available to authorized users. N. Goyal (&) GlaxoSmithKline Research and Development, Pharmacokinetics, 709 Swedeland Road, UW2431, King of Prussia, PA 19406, USA e-mail:
[email protected];
[email protected] M. Beerahee GlaxoSmithKline Research and Development, Stevenage, UK C. Kalberg A. Church R. Mehta GlaxoSmithKline, Respiratory Therapeutic Area, Research Triangle Park, NC, USA S. Kilbride GlaxoSmithKline, Stockley Park, Uxbridge, UK
analyses, performed using NONMEMÒ software. A likelihood-based approach was used to characterize the data below limit of quantification. Umeclidinium and vilanterol exposures at clinical doses were simulated based on the population model. Results For the umeclidinium and vilanterol populationpharmacokinetic analyses, 1,635 and 1,637 patients provided 8,498 and 8,405 observations, respectively. Umeclidinium and vilanterol pharmacokinetics were best described by a two-compartment model with first-order absorption. For umeclidinium, bodyweight, age, and creatinine clearance (CLCR) were statistically significant covariates for apparent inhaled clearance (CL/F); bodyweight was a statistically significant covariate for volume of distribution of central compartment (V2/F).The population parameter estimates namely CL/F and V2/F for umeclidinium were 218 L/h and 1,160 L and 40.9 L/h and 268 L for vilanterol. For vilanterol, bodyweight and age were statistically significant covariates for CL/F. The effect of covariates on umeclidinium and vilanterol systemic exposure was marginal. The population model indicates that a 10 % increase in bodyweight will result in a 2 % increase in CL/F for umeclidinium and vilanterol and 6 % increase in umeclidinium V2/F. A 10 % increase in age will provide a 7 and 4 % decrease in umeclidinium and vilanterol CL/F, respectively. A 10 % decrease in CLCR will result in a 3 % decrease in umeclidinium CL/F. Umeclidinium and vilanterol population-pharmacokinetic model-based systemic exposure predictions showed no pharmacokinetic interactions between umeclidinium and vilanterol when administered in combination. Conclusions There were no apparent pharmacokinetic interactions when umeclidinium and vilanterol were co-administered in patients with COPD. The effects of
N. Goyal et al.
patient demographics, including age, bodyweight, and CLCR, on umeclidinium or vilanterol systemic exposure were minimal, and therefore no dose adjustments are necessary.
Key Points Model-based exposure predictions in the monotherapy and combination therapy arms demonstrate lack of pharmacokinetic interaction when umeclidinium is co-administered with vilanterol. Bodyweight and age were significant covariates on umeclidinium and vilanterol apparent inhaled clearance (CL/F); bodyweight was found to be a significant covariate for umeclidinium volume of distribution of central compartment; creatinine clearance (CLCR) was found to be a significant covariate for umeclidinium but not vilanterol CL/F. Although bodyweight, age, and CLCR were significant covariates, the effects on systemic exposure were marginal, and therefore no dose adjustments are required.
1 Introduction Chronic obstructive pulmonary disease (COPD) is characterized by airflow limitation and reduced maximum expiratory flow that is not fully reversible [1, 2]. World Health Organization projections show that COPD will become the third leading cause of death globally by 2030 [3]. Current COPD treatment guidelines recommend the use of longacting muscarinic antagonists (LAMAs) and long-acting b2 agonists (LABAs) as maintenance bronchodilator therapies for COPD [2]. Umeclidinium is a LAMA in development as a once-daily monotherapy, and as a fixed-dose combination therapy with the LABA vilanterol, for the treatment of COPD. Randomized clinical trials have demonstrated sustained bronchodilator activity over 24 h with umeclidinium and vilanterol alone and in combination, in addition to improved health status and dyspnea scores, with an acceptable tolerability profile [4–6]. Inhalational therapies in respiratory disorders such as COPD and asthma aim to provide locally acting drugs at the site of action with minimal systemic exposure. This article reports the results of a population-pharmacokinetic analysis of pooled data from two 24-week, phase III, randomized, double-blind, placebo-controlled studies of umeclidinium/
vilanterol combination therapy, umeclidinium monotherapy, and vilanterol monotherapy in more than 1,600 patients with COPD [4, 6]. The principal aims of this analysis were to investigate the pharmacokinetics of umeclidinium and vilanterol and the effects of patient demographics on population-pharmacokinetic parameters that may impact their systemic exposure.
2 Methods 2.1 Study Designs and Populations Data from two 24-week, phase III, multicenter, randomized, double-blind, placebo-controlled, parallel-group studies were pooled (protocol number DB2113361 [ClinicalTrials.gov identifier NCT01313637]; protocol number DB2113373 [ClinicalTrials.gov identifier NCT01313650]). Study details have been published previously [4, 6]. Briefly, patients were randomized in a 3:3:3:2 ratio to receive 24 weeks of treatment with umeclidinium/vilanterol 125/25 mcg (delivering 113/22 mcg), umeclidinium 125 mcg (delivering 113 mcg), vilanterol 25 mcg (delivering 22 mcg), or placebo once daily (study DB2113361), or umeclidinium/vilanterol 62.5/25 mcg (delivering 55/22 mcg), umeclidinium 62.5 mcg (delivering 55 mcg), vilanterol 25 mcg (delivering 22 mcg), or placebo once daily (study DB2113373) via the ELLIPTAÒ dry powder inhaler. In each study, eligible patients were current or former cigarette smokers (C10 pack-years), C40 years of age, with a history of COPD, a post-albuterol/salbutamol forced expiratory volume in 1 s (FEV1)/forced vital capacity ratio \0.70, a post-albuterol/salbutamol FEV1 B70 % of predicted normal values (determined using National Health and Nutrition Examination Survey III [7, 8]), and a score of B2 on the modified Medical Research Council Dyspnoea Scale [9]. Patients were excluded if they had a current diagnosis of asthma or other known respiratory disorder, any clinically significant uncontrolled disease, an abnormal and significant electrocardiogram or 24-h Holter finding, or significantly abnormal clinical laboratory findings. The umeclidinium pharmacokinetic and vilanterol pharmacokinetic populations comprised all patients in the intentto-treat population (patients who were randomized and received at least one dose of study medication) from each study who had a pharmacokinetic sample taken and analyzed, and an umeclidinium and/or vilanterol concentration reported. The analyses included patients from the following six study arms: study DB2113361: umeclidinium 125 mcg, vilanterol 25 mcg, and umeclidinium/vilanterol 125/25 mcg; study DB2113373: umeclidinium 62.5 mcg, vilanterol 25 mcg and umeclidinium/vilanterol 62.5/25 mcg.
Pharmacokinetics of Inhaled Umeclidinium and Vilanterol
2.2 Pharmacokinetic Analyses Blood samples for plasma pharmacokinetic analyses were collected pre-dose and at one time point 1- to 15-min postdose on days 1, 56, and 168 (visits 2, 5, and 8). A subset of patients (13–14 % across treatment arms) provided preand post-dose serial samples at one time point at each of the following post-dose time intervals: 1–15 min, 20 min to 4 h, 4.5–15 h, 23–24 h, on each of days 1, 84, and 168 (visits 2, 6, and 8). The times of dose administration for day 56, 84, and 168 pre-dose samples were reported by the subject and recorded in an electronic case report form. The blood samples (approximately 2 mL) were collected in EDTA tubes, mixed by inversion, and immediately cooled to 2–4 °C. Samples were then centrifuged at 4 °C at 1,500g or 3,000 rpm for 10 min. The resultant plasma was transferred to polypropylene tubes and stored at -70 °C until transferred to the bio-analysis department at GlaxoSmithKline for analysis. The samples were shipped on dry ice. 2.3 Bioanalysis Plasma samples were analyzed for umeclidinium and vilanterol using validated analytical methods based on solid-phase extraction, followed by high-pressure liquid chromatography with tandem mass spectrometric analysis for detection analysis. Bioanalysis of pharmacokinetic samples was performed by Quotient Bioresearch Limited, Cambridgeshire, UK, using Analyst (Version 1.5.1) and Watson Laboratory Information Management System (version 7.2). Quality control (QC) samples in plasma were prepared at four different analyte concentrations and were stored with study samples. The QC samples were analyzed with each batch of study samples against separately prepared calibration standards. For the analysis to be acceptable, no more than one-third of the QC results were to deviate from the nominal concentration by more than 15 %, and at least 50 % of the results from each QC concentration were to be within 15 % of nominal. The applicable analytical runs met all predefined run acceptance criteria. The lower limit of quantification for umeclidinium and vilanterol in plasma was 10.0 pg/mL, and the higher limit of quantification was 2,000 pg/mL for umeclidinium and 1,000 pg/mL for vilanterol. 2.4 Population-Pharmacokinetic Modeling Population-pharmacokinetic models for umeclidinium and vilanterol were developed using non-linear mixed-effects analyses, performed using NONMEMÒ software [version 7.1.2] (ICON development solutions, Ellicott City, MD, USA). The population-pharmacokinetic modeling was
based on the principles outlined in the US Food and Drug Administration and European Medicines Agency population-pharmacokinetic regulatory guidance [10, 11]. A likelihood-based approach was used to account for plasma samples with drug concentrations below the limits of quantification (BLQ) (all analyzed non-quantifiable samples except day 1, pre-dose were assumed to be [0 and \10 pg/mL) [12]. R (The R Foundation for Statistical Computing Version 2.15) was used for data handling, data manipulation, processing, and generation of graphs. The observed analyte concentration-time profiles from the subset of patients with serial sampling were used to determine the initial population-pharmacokinetic model, with no covariates included in the structural model. Actual sampling times were used in the dataset. Goodness-of-fit (GOF) plots, objective function values (OFVs), residual plots, standard error (SE) of parameters, and distribution of individual population-pharmacokinetic parameters were used to evaluate the overall fit of the structural model to the data. The structural base model was also used to estimate the population-pharmacokinetic parameters and perform sensitivity analyses by excluding data from certain centers and/or patients when deemed necessary. The Stochastic Approximation Expectation Maximization (SAEM) with interaction method was used in NONMEM 7.1 for umeclidinium and vilanterol population-pharmacokinetic analyses; under this method, the BLQ data were considered to be censored. The F_FLAG functionality in NONMEM 7.1 was used to implement the M3 method that maximizes the likelihood of all data while considering the BLQ data as censored data. The OFVs were obtained with the IMPMAP step (EONLY = 1) after the SAEM estimation. The OFVs follow an approximate chi square distribution and were used to compare nested models along with other criteria listed above. The final structural model was then carried forward for covariate analysis. Covariates were evaluated for their potential impact on umeclidinium and vilanterol pharmacokinetic parameters, including age, bodyweight, sex, race, percent predicted baseline FEV1, treatment effect, concomitant administration of inhaled corticosteroids, postalbuterol/salbutamol reversibility, post-albuterol/salbutamol and ipratropium reversibility, baseline creatinine clearance (CLCR), and smoking status. The continuous covariates (e.g., age, bodyweight, CLCR) were added to the model as a power function by centering these covariates on the typical or population median values for that particular covariate, as shown in Eq. 1, where hi is the individual parameter modeled as a function of the population parameter estimate h1pop centered around the median value of a covariate of interest. The categorical covariates were modeled as shown in Eq. 2, where the fractional change in the parameter value relative to the reference group can be estimated.
N. Goyal et al.
hi ¼ h1pop ðCOVind =MedianÞh2
ð1Þ
hi ¼ h1pop ð1 þ h2 COVind Þ
ð2Þ
Potential covariate relationships were explored graphically using individual inter-individual variability (ETA) vs. covariate plots. If a trend or correlation was observed, the covariate was tested by adding it to the structural model. If the updated model had a lower OFV ([3.84 points for chi square distribution and one degree of freedom [df = 1] compared with the original model, which represents p = 0.05) and/or the trend in the ETA vs. the covariate plot disappeared, that covariate was included and tested with other significant covariates in the final model. Change in OFV at p = 0.01 was used to evaluate the final model by backward elimination. If the OFVs increased by more than 6.62 points (difference of 6.62 points for chisquare distribution and with df = 1 represents p = 0.01), the covariate was retained in the model. If the 90 % confidence interval (CI) of the exponent in Eq. 1 included 0, then the covariate had no significant effect on that particular population-pharmacokinetic parameter. Performance of the final model was evaluated by visual predictive check (VPC) [13]. When evaluating model performance with VPCs, the observed data were differentiated based on the dose administered to the patient in the clinic and the patient-reported non-clinic dose.
The final population-pharmacokinetic model was used to simulate area under the plasma concentration-time curve (AUC) and the maximum plasma concentration (Cmax), to predict drug exposure at clinical doses. This involved estimating the AUC for each patient by dividing the analyte dose by the post hoc inhaled clearance. Cmax for each analyte was obtained by simulating individual concentration-time profiles using the parameter and variability estimates from the final population-pharmacokinetic model. The impact of significant covariates on drug exposure was evaluated to assess if any dose adjustment would be deemed necessary.
3 Results 3.1 Patients Data for study DB2113361 were collected at 153 centers in 14 countries from March 22, 2011 to April 19, 2012; data for study DB2113373 were collected at 163 centers in 13 countries from March 30, 2011 to April 5, 2012. For the umeclidinium and vilanterol population-pharmacokinetic analyses, 1,635 and 1,637 patients provided 8,498 and 8,405 observations, respectively. Baseline patient demographics and characteristics for the umeclidinium and vilanterol datasets are summarized in Table 1. Approximately
Table 1 Demographics and baseline patient characteristics Characteristic Age, years
Umeclidinium N = 1,635 64 (40–93)
Vilanterol N = 1,637 63 (40–88)
Race, n (%) White/Caucasian/European heritage
1,416 (87)
1,419 (87)
Sex, n (%) Female Male Body mass index (kg/m2)
501 (31)
520 (32)
1,134 (69)
1,117 (68)
26.1 (13.9–56.7)
26.2 (13.3–48.5)
Current smokera
825 (50)
812 (50)
Former smoker
810 (50)
825 (50)
514 (31) 889 (54)
535 (33) 872 (53)
Smoking status, n (%)
Reversibility at screeningb, n (%) Reversibility to albuterol and ipratropium testc, n (%) % predicted FEV1
48.4 (13.1–76.0)
48.8 (13.4–75.8)
Baseline FEV1 (L)
1.18 (0.33–3.39)
1.19 (0.33–3.39)
Creatinine clearance (mL/min)
88.5 (19.6–256.3)
89.3 (19.6–256.1)
Values are expressed as median (range) unless specified otherwise FEV1 forced expiratory volume in 1 s a
A patient was classed as a current smoker at a visit unless they had not smoked in the 6 months prior to that visit
b
Reversibility was an increase in FEV1 of C12 % and C200 mL following administration of albuterol
c
Reversibility was an increase in FEV1 of C12 % and C200 mL following administration of both albuterol and ipratropium
Pharmacokinetics of Inhaled Umeclidinium and Vilanterol
70 % of patients were male (69 % and 68 % in the umeclidinium and vilanterol datasets, respectively) and predominantly of White Caucasian/European heritage (87 % in both datasets). Median age was 64 years in the umeclidinium dataset and 63 years in the vilanterol dataset. 3.2 Umeclidinium and Vilanterol PopulationPharmacokinetic Analyses
and proportional) was used for umeclidinium and vilanterol. As none of the covariates had more than 1–2 % missing data, no multiple imputations were performed for any missing covariates; the population median values were used for missing covariate values, as discussed in the analysis plan. Approximately 20–25 % of the samples had concentrations BLQ. 3.2.1 Sensitivity Analyses
Both umeclidinium and vilanterol pharmacokinetics were best described by a two-compartment model with first-order absorption (Table 2). The sparse sampling scheme and variability associated with inhaled drug administration likely contributed to the relatively high inter-individual variability estimates. The SEs on most parameters were relatively low. The run time for each model was approximately 4–5 h. Although most subjects provided two samples over 24 h on three different occasions, about 13–14 % subjects provided five samples over 24 h on three occasions. A biphasic profile was observed for umeclidinium and vilanterol from these serial samples. The two-compartment body model also provided a better fit with lower OFVs (p \ 0.001) as compared with one-compartment body model. The initial phase I studies for umeclidinium and vilanterol in healthy volunteers with rich sampling demonstrated biphasic profiles as well (data on file). While high inter-subject variability was observed, the shrinkage on the ETA terms had a range between 13 % for CL/F and 29 % for umeclidinium and between 14 % for CL/F to 29 % for V3/F (volume of distribution of second peripheral compartment) for vilanterol. A combined residual error model (additive
Approximately 4–5 % of pharmacokinetic samples obtained in the 23- to 24-h post-dose window had unexpectedly high concentrations of umeclidinium and vilanterol, with some values higher than the 0- to 15-min post-dose sample in the same patient. Such observations occurred only with data obtained on the second and third pharmacokinetic occasions (pharmacokinetic sampling days 56, 84, and 168). This was observed for umeclidinium and vilanterol across all treatment arms. To gauge the effect of these samples on pharmacokinetic parameter estimates with subject-reported dosing times, a decision was made to perform sensitivity analyses by fitting the base model with and without such data. To avoid any selection bias while performing the sensitivity analyses, the structural model was run with and without all subject-reported dosing time samples, irrespective of whether they were higher than expected or within the expected range. Similar sensitivity analyses were also performed by excluding data from centers where such an event occurred. The population-pharmacokinetic parameter estimates obtained from the model that excluded these patients or centers from the dataset were close to estimates obtained
Table 2 Umeclidinium and vilanterol population-pharmacokinetic parameters from final model Parameter
Umeclidinium
Vilanterol
Population estimate (% RSE)
Inter-subject variability % CV (% RSE)
Population estimate (% RSE)
Inter-subject variability % CV (% RSE)
CL/F (L/h)
218 (2.3)
42.5 (9.0)
40.9 (1.4)
30.8 (29.7)
V2/F (L)
1,160.0 (2.8)
32.1 (8.7)
268.0 (2.0)
26.9 (26.9)
Q/F (L/h)
873.0 (4.7)
68.3 (9.6)
118.0 (4.6)
65.1 (6.8)
V3/F (L)
30,200 (22.1)
113.0 (30.9)
1,240.0 (6.3)
42.2 (45.0)
ka (h-1)
39.1 (43.7)
210.0 (14.0)
18.8 (12.8)
173.0 (11.5) N/A
Bodyweight exponent on CL/F
0.15 (50.3)
N/A
0.192 (23.0)
Age exponent on CL/F
-0.731 (19.7)
N/A
-0.398 (23.0)
Creatinine clearance exponent on CL/F
0.271 (26.4)
N/A
N/A
N/A
N/A
N/A
N/A
Bodyweight exponent on V2/F
0.616 (9.4)
Additive residual error (pg/mL)
0.7 (17.3)
0.9 (24.4)
Proportional residual variability (% CV)
40.9 (8.6)
41.4 (22.8)
CL/F apparent clearance of parent drug following oral (non-intravenous) dosing, CV coefficient of variation, ka absorption rate constant of parent drug, N/A not applicable, Q/F apparent inter-compartment clearance of parent drug, RSE relative standard error, V2/F apparent volume of distribution of parent drug after inhalation in the central compartment, V3/F apparent volume of distribution of parent drug after inhalation in the peripheral compartment
N. Goyal et al.
(a)
90% PI - weight 70 kg Pred median - weight 70 kg 90% PI - weight 140 kg Pred median - weight 140 kg
90% PI - weight 70 kg Pred median - weight 70 kg 90% PI - weight 140 kg Pred median - weight 140 kg
1000
UMEC conc (pg/mL)
UMEC conc (pg/mL)
1000
100
10
1
100
10
1
UMEC 62.5 0
5
10
15
20
UMEC 125 25
0
5
Time after dose (hr)
(b)
90% PI - age 60 yrs Pred median - age 60 yrs 90% PI - age 80 yrs Pred median - age 80 yrs
15
100
10
1
10
1
0
5
10
15
20
UMEC 125 25
0
5
Time after dose (hr)
90% PI - CLCR 110 ml/min Pred median - CLCR 110 ml/min 90% PI - CLCR 110 ml/min Pred median - CLCR 110 ml/min
1000
10
15
10
1
100
10
1
UMEC 62.5 5
10
15
Time after dose (hr)
25
90% PI - CLCR 110 ml/min Pred median - CLCR 110 ml/min 90% PI - CLCR 110 ml/min Pred median - CLCR 110 ml/min
1000
100
0
20
Time after dose (hr)
UMEC conc (pg/mL)
UMEC conc (pg/mL)
25
100
UMEC 62.5
(c)
20
90% PI - age 60 yrs Pred median - age 60 yrs 90% PI - age 80 yrs Pred median - age 80 yrs
1000
UMEC conc (pg/mL)
UMEC conc (pg/mL)
1000
10
Time after dose (hr)
20
UMEC 125 25
0
5
10
15
20
25
Time after dose (hr)
Fig. 1 Comparison of systemic umeclidinium exposure at steady state due to changes in a bodyweight, b age, and c creatinine clearance. Conc concentration, CLCR creatinine clearance, PI prediction interval, pred predicted, UMEC umeclidinium
Pharmacokinetics of Inhaled Umeclidinium and Vilanterol
(a)
90% PI - weight 70 kg Pred median - weight 70 kg 90% PI - weight 140 kg Pred median - weight 140 kg
1000
VI conc (pg/mL)
1000
VI conc (pg/mL)
(b)
90% PI - age 60 yrs Pred median - age 60 yrs 90% PI - age 80 yrs Pred median - age 80 yrs
100
10
1
100
10
1
0
5
10
15
20
25
0
Time after dose (hr)
5
10
15
20
25
Time after dose (hr)
Fig. 2 Comparison of systemic vilanterol exposure at steady state due to changes in a bodyweight and b age. Conc concentration, PI prediction interval, pred predicted, VI vilanterol Table 3 Umeclidinium and vilanterol population-pharmacokinetic model-based systemic exposure predictions Study
Treatment
Dose
Geometric means (95 % CI) AUCss (pgh/mL)
Cmax-ss (pg/mL)
Umeclidinium DB2113373
Umeclidinium/vilanterol 62.5/25 (n = 410)
62.5 lg
307.6 (293.2–322.7)
68.5 (65.2–71.9)
Umeclidinium 62.5
62.5 lg
317.6 (303.1–333.5)
70.3 (67.0–73.8)
125 lg
627.5 (597.8–658.9)
138.0 (131.6–144.9)
125 lg
622.9 (593.4–653.0)
138.8 (132.2–146.0)
25 lg
612.3 (588.6–636.7)
128.2 (122.1–134.6)
25 lg
612.8 (589.3–637.3)
128.2 (122.0–134.6)
25 lg
616.7 (592.1–642.1)
128.4 (122.3–135.0)
25 lg
610.5 (586.7–635.1)
128.2 (122.0–134.9)
(n = 417) DB2113361
Umeclidinium/vilanterol 125/25 (n = 402) Umeclidinium 125 (n = 406)
VilanteroI DB2113373
Umeclidinium/vilanterol 62.5/25 (n = 410) Vilanterol 25 (n = 421)
DB2113361
Umeclidinium/vilanterol 125/25 (n = 402) Vilanterol 25 (n = 404)
AUCss area under the concentration-time curve at steady state, CI confidence interval, Cmax-ss maximum concentration at steady state
that included all such data; \5 % change was observed in the key parameters, including apparent inhaled clearance (CL/F), volume of distribution (V2/F), and the absorption rate constant (ka). The variability estimates were higher when the data were excluded from the analyses. The SE on the population parameters such as CL/F, V2/F changed by \5 % while the change in SE on variability estimates had a range between 3 and 25 %. As the overall fit of the model to the data and population-pharmacokinetic parameters
remained relatively unchanged with the exclusion of data, the entire dataset was used for modeling purposes. None of the available pharmacokinetic concentrations were excluded from final analyses. 3.2.2 Covariate Analyses Covariates such as age, bodyweight, sex, study, CLCR, smoking status, and ethnicity were evaluated by including
N. Goyal et al. UMEC 125 COMBO
(a)
UMEC 125 MONO
UMEC 62.5 COMBO
UMEC 62.5 MONO
UMEC conc (pg/mL)
500
100 50
10 0
5
10
15
20
25
0
5
10
15
20
25
Time relative to last dose (hrs)
(b)
VI COMBO (DB2113361)
VI MONO (DB2113361)
VI COMBO (DB2113373)
VI MONO (DB2113373)
VI conc (pg/mL)
500
100 50
10 0
5
10
15
20
25
0
5
10
15
20
25
Time relative to last dose (hrs) Fig. 3 Observed a umeclidinium and b vilanterol plasma concentration vs. time data. Conc concentration, MONO monotherapy, COMBO combination therapy, VI vilanterol, UMEC umeclidinium
them one at time in the base models for umeclidinium and vilanterol. Covariates that resulted in increased OFVs or the 90 % CI for the covariate included 0 or 1 (with regard to Eqs. 1 and 2, respectively) were excluded from the model. The ETA vs. covariate plots for the covariates also confirmed their exclusion. The ETA vs. covariate plots were also used in conjunction with the OFVs. The statistically significant covariates were then evaluated with forward addition and backward elimination. Bodyweight, age, and CLCR were statistically significant covariates for CL/F of umeclidinium. Bodyweight was also a significant covariate for umeclidinium V2/F. For every 10 % increase in bodyweight, CL/F of umeclidinium increased by approximately 2 %. A 10 % increase in age (from 60 years old) resulted in an approximate 7 % decrease in CL/F of umeclidinium. With every 10 % decrease in
CLCR from 110 mL/min, CL/F of umeclidinium decreased by approximately 3 %. No significant differences were found between healthy volunteers and subjects with severe renal impairment on umeclidinium pharmacokinetics in a renal impairment study (data on file). Apparent umeclidinium V2/F increased by approximately 6 % for every 10 % increase in bodyweight after 70 kg. The resulting marginal changes in exposures did not warrant any dose adjustments (Fig. 1). Bodyweight and age were statistically significant covariates for CL/F of vilanterol. For every 10 % increase in bodyweight the CL/F of vilanterol increased by approximately 2 %. With a 10 % increase in age from 60 years, the CL/F of vilanterol decreased by approximately 4 %. The resulting marginal change in exposure did not warrant any dose adjustments (Fig. 2).
Pharmacokinetics of Inhaled Umeclidinium and Vilanterol DB2113361 – 125
(a)
Clinic dose Non-clinic dose 90% PI Pred median
1000
UMEC conc (pg/mL)
1000
UMEC conc (pg/mL)
DB2113373 – 62.5
Clinic dose Non-clinic dose 90% PI Pred median
100
10
100
10 0
5
10
15
20
25
0
5
Time after dose (hr) DB2113361
1000
15
20
25
DB2113373
Clinic dose Non-clinic dose 90% PI Pred median
Clinic dose Non-clinic dose 90% PI Pred median
1000
VI conc (pg/mL)
VI conc (pg/mL)
(b)
10
Time after dose (hr)
100
100
10
10
0
5
10
15
20
25
Time after dose (hr)
0
5
10
15
20
25
Time after dose (hr)
Fig. 4 Assessment of model performance for a umeclidinium and b vilanterol by visual predictive check with simulations from final models. Conc concentration, PI prediction interval, pred predicted, VI vilanterol, UMEC umeclidinium
Lung function status in patients with COPD (or other respiratory disorders) based on reversibility to beta agonists or muscarinic antagonists, or baseline FEV1 may impact the amount of drug inhaled by an individual thus indirectly influencing the systemic exposures of inhaled drugs. Smoking may also affect the lung function and impact the pharmacokinetics of drugs via inhibition of certain clearance pathways such as cytochrome P450 (CYP)1A2 [14]. As there was no apparent relationship between individual estimates of CL/F or V2/F for umeclidinium or vilanterol, and sex, post-albuterol/salbutamol reversibility, postipratropium response, percent predicted baseline FEV1, or smoking status, these covariates were not tested or included in the model.
Simulations based on the umeclidinium and vilanterol population-pharmacokinetic model demonstrated that the effects of bodyweight, age, and CLCR covariates on umeclidinium systemic exposure and bodyweight and age on vilanterol systemic exposure were marginal, therefore, no dose adjustments were deemed necessary based on these demographic characteristics. The parameter estimates from the final population-pharmacokinetic model are listed in Table 2 and the GOF plots are shown in Supplementary Figure 1. Along with the observed concentration-time profiles, umeclidinium and vilanterol population-pharmacokinetic model-based systemic exposure predictions showed that there was no pharmacokinetic interaction between
N. Goyal et al. 0.6
(b)
95% predicted proportion Observed proportion
Proprtion below LOQ
Proprtion below LOQ
(a)
0.4
0.2
0.0
0.6
95% predicted proportion Observed proportion
0.4
0.2
0.0 0
5
10
15
20
25
Time after dose (hr)
0
5
10
15
20
25
Time after dose (hr)
Fig. 5 Assessment of model performance for a umeclidinium and b vilanterol below the LOQ data by visual predictive check with simulations from final models. LOQ lower limit of quantification
umeclidinium and vilanterol when administered as a combination (Table 3; Fig. 3). 3.3 Model Performance The model was able to predict the majority of the data well for both umeclidinium and vilanterol (Fig. 4). Although most of the observed concentrations were adequately explained by the model predictions, there were some unexpectedly high concentrations observed in about 4–5 % of patients in the 23- to 24-h post-dose window (Fig. 4). Details for this observation are presented in the discussion section. As a significant fraction of collected pharmacokinetic data (20–25 %) was BLQ, the model performance was also evaluated in predicting the BLQ data. The model performed reasonably well in describing proportion of BLQ data; limited over-predictions were made for observations around the 23- to 24-h post-dose window (Fig. 5) for umeclidinium and vilanterol. Sensitivity analyses demonstrated that some observed trough concentrations with patient-reported dosing time did not significantly affect the population-pharmacokinetic parameters.
4 Discussion Administering therapies in respiratory disorders via inhalation is aimed at providing a drug at the site of action while limiting the unwanted systemic drug exposure. While limited, the systemic drug exposure does exist and it should be characterized to understand any changes that may occur by patient demographics or concomitant therapy. This article describes the population-pharmacokinetic models developed to characterize the pharmacokinetic and systemic
exposure in patients with COPD, for the combination LAMA/LABA, umeclidinium/vilanterol; these methods have also been used in population-pharmacokinetic studies in patients with asthma receiving the inhaled corticosteroid, fluticasone propionate [15]. Results of these analyses show that both umeclidinium and vilanterol pharmacokinetics are adequately described by a two-compartment pharmacokinetic model with first-order absorption. Bodyweight and age were important factors to investigate, as previous studies have shown lean bodyweight to be a significant covariate on the relative bioavailability of the LAMA/ LABA combination glycopyrronium/indacaterol (QVA149), and an impaired bronchodilator response to b-agonists has been reported in healthy elderly men and women [15, 16]. In the present study, bodyweight and age were significant covariates on umeclidinium and vilanterol CL/F based on covariate analysis. Bodyweight was also a significant covariate on umeclidinium V2/F. CLCR was a covariate of interest, as COPD is predominantly observed in older patients, and aging is associated with a decline in renal function [17]. CLCR was found to be a significant covariate for umeclidinium CL/F but not vilanterol CL/F. For every 10 % decrease in CLCR from 110 mL/min, the decrease in umeclidinium CL/F was marginal, therefore, no dose adjustment was deemed necessary based on renal function. It should also be noted that recent studies in patients with severe renal impairment or moderate hepatic impairment have shown that the systemic exposure of umeclidinium and vilanterol was not significantly different to that observed in healthy volunteers [18, 19]. No significant effects of other covariates such as sex, post-albuterol/salbutamol reversibility, post-ipratropium response, percent predicted baseline FEV1, smoking status,
Pharmacokinetics of Inhaled Umeclidinium and Vilanterol
and inter-individual variability on the pharmacokinetics of umeclidinium or vilanterol were observed in the present study. The hepatic route has been determined as a major route of elimination for umeclidinium with in vitro data indicating umeclidinium as a metabolic substrate for the CYP2D6 isozyme. Vilanterol elimination predominantly involves the CYP3A4 pathway. Although bodyweight, age, and CLCR were significant covariates, the effects on systemic exposure were marginal, indicating that no dose adjustments are required. The majority of the parameters and the associated interindividual variability in the model were estimated with reasonable precision; the only parameter with very high inter-individual variability was ka. This may be explained by the insufficient spread of data in the absorption phase, as each patient had only one sample in the 0- to 15-min postdose interval with the other sample in the 23- to 24-h postdose sampling window. Future studies should attempt to employ sampling strategies that can contribute to a more robust estimation of ka. It is important to understand that the absorption occurs rapidly with inhaled drugs and can substantially vary between individuals. The exposure predictions using the current model are reliable to assess any differences in exposures across treatment arms. Pharmacokinetic parameters were similar for umeclidinium and vilanterol administered as monotherapies or in combination with one another. The absence of any trends in these parameters across the monotherapy and combination treatment arms indicates that there is little or no pharmacokinetic interaction when umeclidinium is co-administered with vilanterol. The results from the model-based simulations clearly demonstrate the lack of any pharmacokinetic interaction between umeclidinium and vilanterol with mono or combination therapy. The observed concentration-time profiles for umeclidinium and vilanterol are also similar across the mono and combination therapy arms (Figure 3). The exposures (AUC and Cmax) simulated from the model are as expected and in line with the observed data. The unexpectedly higher concentrations observed in 4–5 % of patients during the 23- to 24-h post-dose window may be related to inaccurate recording of the dosing time reported for some pre-dose samples, or inaccurate recording of the pharmacokinetic sampling time for some patients. The times of dose administration for day 56, 84, and 168 pre-dose samples were reported by the patient and recorded in the electronic case report form. Therefore, given the small window for obtaining two samples within the nominal sampling windows of pre-dose and 0- to 15-min post-dose, errors in sampling time may have occurred. As sensitivity analyses demonstrated no significant impact on pharmacokinetic parameters when the data were excluded, the decision was made to retain these data
in the final analyses. The unexpected high concentrations observed may explain why the model over-predicted BLQ data in the 23- to 24-h time window as seen in Fig. 5. It is not uncommon to observe a significant fraction of systemic pharmacokinetic samples with drug concentrations BLQ with inhalation formulations. The primary objective of inhalation formulations in diseases such as COPD and asthma is to ensure local drug delivery to the site of action while limiting unwarranted systemic exposures. Nevertheless, it is important to account for such observations in model development. Different approaches that account for the BLQ data have been employed and discussed in detail in the literature [12, 20, 21]. With improvements in NONMEM software, the likelihood-based method (M3) was employed in the current analysis [12]. Using this methodology, the likelihood of a BLQ observation being BLQ is calculated while simultaneously estimating the expected concentration of other data. This was employed using the F_FLAG option available within NONMEM (Version C6). The population modeling exercise characterized the pharmacokinetics of umeclidinium and vilanterol in adult subjects. The model can be readily used to design future studies with optimal sampling schemes. This will allow reliable measurement of systemic exposure of umeclidinium and/or vilanterol in clinical trials conducted for different indications, pediatric trials, clinical studies in special populations, or other specific drug-drug interaction studies. These models can also be used in clinical studies with combinations of umeclidinium and/or vilanterol with other drugs to be administered via inhalation.
5 Conclusion The analysis adequately characterized the population pharmacokinetics of umeclidinium and vilanterol when given as monotherapy or co-administered as a fixed-dose combination in patients with COPD. The impact of patient and disease characteristics on umeclidinium or vilanterol systemic exposure was minimal and no dose adjustments are necessary. Acknowledgments Editorial support was provided by Joanne Parker of Fishawack Indicia Ltd, funded by GlaxoSmithKline. Ethical standards All patients were required to give written, informed consent. The studies were approved by local review boards/ ethics committees and were conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. Financial disclosures
This study was funded by GlaxoSmithKline.
Conflict of interest All authors are employees of GlaxoSmithKline and own stock/stock options in GlaxoSmithKline.
N. Goyal et al.
References 1. Celli BR, MacNee W. ATS ERS Task Force. Standards for the diagnosis and treatment of patients with COPD: a summary of the ATS/ERS position paper. Eur Respir J. 2004;23:932–46. 2. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease. 2011. http://www. goldcopd.org/uploads/users/files/GOLD_Report_2011_Feb21.pdf. Accessed 28 Jan 2014. 3. World Health Organisation. Chronic obstructive pulmonary disease (COPD): burden of COPD. 2011. http://www.who.int/ respiratory/copd/burden/en/index.html. Accessed 28 Jan 2014. 4. Donohue JF, Maleki-Yazdi MR, Kilbride S, et al. Efficacy and safety of once-daily umeclidinium/vilanterol 62.5/25 mcg in COPD. Respir Med. 2013;107:1538–46. 5. Feldman G, Walker RR, Brooks J, et al. Safety and tolerability of the GSK573719/vilanterol combination in patients with COPD. Am J Respir Crit Care Med. 2012;185:A2938. 6. Celli B, Crater, G, Kilbride, S, et al. Once-daily umeclidinium/ vilanterol 125/25 mcg in COPD: a randomized, controlled study. Chest. 2014; in press. 7. Hankinson JL, Kawut SM, Shahar E, et al. Performance of American Thoracic Society-recommended spirometry reference values in a multiethnic sample of adults: the Multi-Ethnic Study of Atherosclerosis (MESA) lung study. Chest. 2010;137:138–45. 8. Hankinson JL, Odencrantz JR, Fedan KB. Spirometric reference values from a sample of the general U.S. population. Am J Respir Crit Care Med. 1999;159:179–87. 9. Manali ED, Lyberopoulos P, Triantafillidou C, et al. MRC Chronic Dyspnea Scale: relationships with cardiopulmonary exercise testing and 6-minute walk test in idiopathic pulmonary fibrosis patients: a prospective study. BMC Pulm Med. 2010;10:32. 10. European Medicines Agency. Guideline on reporting the results of population pharmacokinetic analyses. 2007.
11. Food and Drug Administration. Guidance for industry: population pharmacokinetics. 1999. 12. Ahn JE, Karlsson MO, Dunne A, et al. Likelihood based approaches to handling data below the quantification limit using NONMEM VI. J Pharmacokinet Pharmacodyn. 2008;35:401–21. 13. Holford N. The visual predictive check: superiority to standard diagnostic (Rorschach) plots. Presented at the Population Approach Group in Europe (PAGE), Pamplona, Spain. 2005; Abstract 738. 14. Zevin S, Benowitz NL. Drug interactions with tobacco smoking: an update. Clin Pharmacokinet. 1999;36:425–38. 15. Bieth B. Population pharmacokinetics of QVA149, the fixed dose combination of indacaterol maleate and glycopyrronium bromide in chronic obstructive pulmonary disease (COPD) patients. Presented at the Population Approach Group in Europe (PAGE), Glasgow Scotland. 2013; Abstract 2800. 16. Connolly MJ, Crowley JJ, Charan NB, et al. Impaired bronchodilator response to albuterol in healthy elderly men and women. Chest. 1995;108:401–6. 17. Rowe JW, Andres R, Tobin JD, et al. The effect of age on creatinine clearance in men: a cross-sectional and longitudinal study. J Gerontol. 1976;31:155–63. 18. Kelleher D, Hardes K, et al. Effect of severe renal impairment (SRI) on umeclidinium (UMEC) and vilanterol (VI) pharmacokinetics (PK). Eur Respir J. 2013;42(Suppl. 57):881s. 19. Mehta R, Hardes K, Kelleher D, et al. Effect of moderate hepatic impairment (MHI) on umeclidinium (UMEC) and vilanterol (VI) pharmacokinetics (PK). Eur Respir J. 2013;42(Suppl. 57):751s. 20. Beal SL. Ways to fit a PK model with some data below the quantification limit. J Pharmacokinet Pharmacodyn. 2001;28: 481–504. 21. Bergstrand M, Karlsson MO. Handling data below the limit of quantification in mixed effect models. AAPS J. 2009;11:371–80.