Infect Dis Ther (2018) 7:261–275 https://doi.org/10.1007/s40121-018-0197-y ORIGINAL RESEARCH Population Pharmacokinetic Analysis of Asunaprevir in Subjects with Hepatitis C Virus Infection . . . . Li Zhu Hanbin Li Phyllis Chan Timothy Eley . . . . Yash Gandhi Marc Bifano Mayu Osawa Takayo Ueno . . . Eric Hughes Malaz AbuTarif Richard Bertz Tushar Garimella Received: February 15, 2017 / Published online: March 27, 2018 The Author(s) 2018 Results: A two-compartment model with ﬁrst- ABSTRACT order elimination from the central compart- ment, an induction effect on clearance, and an Introduction: Asunaprevir (ASV) is a potent, absorption model consisted of zero-order release pangenotypic, twice-daily hepatitis C virus followed by ﬁrst-order absorption adequately (HCV) NS3 inhibitor indicated for the treatment described ASV PK after oral administration. A of chronic HCV infection. typical value for ASV clearance (CL/F) was 50.8 Methods: A population pharmacokinetic (PPK) L/h, increasing by 43% after 2 days to a CL/F of model was developed using pooled ASV con- 72.5 L/h at steady-state, likely due to auto-in- centration data from 1239 HCV-infected sub- duction of cytochrome P450 3A4 (CYP3A4). jects who received ASV either as part of the Factors indicative of hepatic function were DUAL regimen with daclatasvir or as part of the identiﬁed as key inﬂuential covariates on ASV QUAD regimen with daclatasvir and peg-inter- exposures. Subjects with cirrhosis had an 84% feron/ribavirin. increase in ASV area under the concentration time curve (AUC) and subjects with baseline aspartate aminotransferase (AST) above 78 IU/L had a 58% increase in area under the concen- Enhanced content To view enhanced content for this tration time curve (AUC). Asians subjects had a article go to https://doi.org/10.6084/m9.ﬁgshare. 46% higher steady-state AUC relative to White/ Caucasian subjects. Other signiﬁcant covariates Electronic supplementary material The online were formulation, age, and gender. version of this article (https://doi.org/10.1007/s40121- 018-0197-y) contains supplementary material, which is Conclusion: The current PPK model provided a available to authorized users. parsimonious description of ASV concentration data in HCV-infected subjects. Key covariates L. Zhu P. Chan T. Eley Y. Gandhi (&) identiﬁed in the model help explain the M. Bifano E. Hughes M. AbuTarif R. Bertz observed variability in ASV exposures and may T. Garimella guide clinical use of the drug. Bristol-Myers Squibb Research and Development, Lawrenceville, NJ, USA Funding: Bristol-Myers Squibb. e-mail: Yash.Gandhi@bms.com M. Osawa T. Ueno Keywords: Asunaprevir; Direct-acting Bristol-Myers Squibb K K, Tokyo, Japan antivirals; Hepatitis C infection; NS3 protease inhibitors; Population pharmacokinetics H. Li Quantitative Solutions, Menlo Park, CA, USA 262 Infect Dis Ther (2018) 7:261–275 recommended to be used in these subjects. INTRODUCTION Asunaprevir exposure was also higher with increased aspartate aminotransferase (AST), Chronic hepatitis C virus (HCV) infection, cur- alanine aminotransferase (ALT) and total rently estimated to affect about 3% of world bilirubin levels, and decreased albumin levels population, or 80–185 million people, frequently . Furthermore, cross-study comparisons of leads to hepatocellular carcinoma, cirrhosis, and steady-state ASV exposures from subjects with liver transplantation [1, 2]. Asunaprevir (for- HCV infection indicated approximately 2- to merly BMS-650032, ASV) is a tripeptidic acylsul- 3-fold higher exposure than those from healthy fonamide inhibitor of the HCV NS3/4A protease subjects . Inﬂammation, tissue damage and [3, 4]. Asunaprevir is approved for the treatment ﬁbrosis associated with chronic HCV infection of chronic HCV in multiple countries as part of could affect liver uptake of ASV and/or reduce the all-oral direct-acting antiviral (DAA) DUAL hepatic CYP3A4 activity, resulting in decreased combination regimen with the HCV NS5A inhi- clearance of the drug . bitor daclatasvir (DCV) or part of the QUAD Ethnic differences have been reported in ASV regimen with DCV and peginterferon/ribavirin exposure. In Phase II studies, ASV maximum (pegINF/RBV) in patients with genotype 1 or concentration (C ) and AUC in Japanese max genotype 4 HCV [5–8]. subjects were approximately double compared Asunaprevir has shown complex pharma- to White subjects . Data in Chinese and cokinetic (PK) characteristics . The dose–ex- Indian subjects also suggested that exposure in posure relationship for ASV is generally non- these populations was closer to that observed in linear. In multiple ascending dose (MAD) stud- Japanese subjects than observed in Caucasian ies, ASV administered using an early capsule subjects. Body weight may be a contributing formulation resulted in greater-than-dose-pro- factor; however, there was a high degree of portional increases in exposure from 10 to overlap in body weights between the Asian and 600 mg twice daily (BID) . White subjects in these clinical studies. Asunaprevir is eliminated primarily via Signiﬁcant food effects has been observed cytochrome P450 (CYP) 3A4-mediated hepatic with a tablet formulation used in Phase I/II metabolism, whereas distribution of asunapre- studies . A soft-gel capsule, developed as a vir to the liver occurs via organic anion-trans- food-effect-mitigating formulation, is able to porting polypeptide (OATP) [11–13]. provide signiﬁcantly higher exposures (approx- Asunaprevir displays time-dependent nonlinear imately 4- to 5-fold for C and 2-fold for AUC) max pharmacokinetics. Following multiple doses at with or without food than the tablet formula- 400 and 600 mg BID, ASV trough concentra- tion given with food . Subsequently, the tions were lower on Day 14 compared to Day 1, soft-gel capsule was used in Phase III studies suggesting auto-induction of CYP3A4. The without food restrictions and is the commercial magnitude of auto-induction appeared to be formulation . dose-dependent. At the 200 mg BID, the trough A population PK analysis was conducted concentration ratio of Day 14 to Day 1 was 1.9, using data pooled from select Phase II/III studies suggesting a more modest auto-induction. This to help better understand the complex PK pro- is consistent with the weak inductive effect of ﬁle and the exposure of ASV in subjects with asunaprevir on midazolam, a sensitive CYP3A4 chronic HCV infection at clinically relevant probe substrate . doses. More importantly, the analysis helps Hepatic function impairment had been explain the sources of variability in ASV expo- shown to signiﬁcantly impact the steady-state sures by investigating potential relationships PK of ASV . Asunaprevir exposures were between covariates and PK parameters of ASV in approximately 5- to 10- and 20- to 30-fold the target patient population. The model would higher in subjects with moderate (Child–Pugh also provide individual post hoc predicted B) and severe (Child–Pugh C) hepatic impair- exposure values for subsequent exposure–re- ment, respectively, and as such ASV is not sponse analyses. Infect Dis Ther (2018) 7:261–275 263 Model selection was based on signiﬁcant METHODS reduction in the objective function value (OFV; p \ 0.05), precision and plausibility of parame- Clinical Studies and Patient Population ter estimates, and goodness-of-ﬁt plots. Once the structural base model was identi- Asunaprevir concentration, time data were ﬁed, covariate evaluations were performed to obtained from 3 Phase II and 2 Phase III studies, assess the inﬂuence of key patient and treat- among which 1 Phase II and 1 Phase III studies ment characteristics on the PK of ASV. Covari- were in Japanese subjects (Supplemental Table 1 ate model building was conducted with the full in the Electronic Supplementary Material). model approach . Covariates were added to Asunaprevir was administered either as part of the base model in a univariate fashion; and the DUAL regimen with DCV or as part of the signiﬁcant covariate-PK relationships were QUAD regimen with DCV and pegINF/RBV. The evaluated using the log-likelihood ratio test asunaprevir doses investigated were 100, 200 (LRT). All signiﬁcant (p \ 0.05) covariates were and 600 mg, given once daily (QD) or BID, included in the full model. The selection of either in tablet or soft-gel formulation. Eligible covariates and likely impacted PK parameters subjects were male or female C 18 years of age, was based on clinical interest, pharmacological with chronic HCV genotype 1 and 4 infections. plausibility, prior knowledge and the availabil- Subjects with compensated cirrhosis and mild ity of data. As baseline AST and ALT were highly hepatic impairment were included. Subjects correlated (r = 0.86), only the most signiﬁcant could be treatment naive, partial, null respon- term was kept in the full model if both covari- ders or ineligible/intolerant to the pegINF/RBV- ates were found to be signiﬁcant in the uni- based therapy. variate screening. On-treatment AST and ALT values were also investigated as time-varying Compliance with Ethics Guidelines covariates on ASV CL/F, as treatments would This article is based on previously conducted normalize hepatic functions. Asunaprevir dose studies and does not contain any studies with was investigated as a covariate on the relative human participants or animals performed by bioavailability (F) based on the observed non- any of the authors. linear dose–exposure relationship in the MAD study. Population Pharmacokinetic Analysis The ﬁnal model was obtained using a step- wise backward elimination process to reach a Model Development parsimonious model. The covariate with the The population PK model development was smallest change in objective function was comprised of establishing a base, full and ﬁnal removed from the model, and the process model. The base structural model development repeated until all remaining covariates were was guided by the graphical evaluation of ASV signiﬁcant at p \ 0.001 based on LRT. Uncer- plasma concentration–time proﬁles. Also tainties in the ﬁnal model parameter estimates investigated as part of the base model were were reported with stratiﬁed nonparametric ﬁxed-effect parameters associated with ASV bootstrap 95% conﬁdence intervals (CI) . characteristics known to inﬂuence its PK: the The ﬁnal model was used for model application effect of auto-induction on ASV clearance (CL/ and for providing individual post hoc predicted F) and the effect of formulation. Inter-individ- exposure values for exposure–response analyses. ual variability for the PK parameters was mod- eled assuming a log-normal distribution. Model Evaluation Residual variability was investigated using Model evaluation was performed using good- either the proportional or the combined pro- ness-of-ﬁt diagnostic plots, prediction corrected portional plus additive error model for non- visual predictive checks (pcVPC) and shrinkage. transformed concentrations, and additive error As a graphic diagnostic tool, pcVPC was used to model for log-transformed concentrations. identify model misspeciﬁcation, especially with 264 Infect Dis Ther (2018) 7:261–275 respect to random effects, by comparing the potential correlations among covariates in the observed ASV concentration–time proﬁles with study population. the simulated median and distribution (5th to 95th percentile) . A total of 1000 trial Analysis Platforms replicates were simulated using actual covari- ates and dose regimens, the ﬁnal model The population PK analysis was performed parameter estimates, and simulated subject- using NONMEM (v.7.1), compiled using Intel speciﬁc random effects and residual errors. FORTRAN Compiler (v.10.1; Intel). The ﬁrst- Shrinkage was assessed for inter-individual order conditional estimation method was used variability (g) and for residual error (e). High g throughout the model-building process. Perl- shrinkage (e.g., [ 30%) would indicate lack of speaks-NONMEM (PsN, v.3.2 or later, http:// information for a reliable estimate of the inter- psn.sourceforge.net/) was used to aid the model individual variability . development. Diagnostic graphics, exploratory analyses, and post-processing of NONMEM Model Application output were performed using S-Plus (v.8.1;, The ﬁnal model was used to investigate the TIBCO Software, Palo Alta, CA, USA), or R (v.3.0 effects of covariates on the PK of ASV. The or later, http://www.r-project.org). impact of signiﬁcant covariates on ASV popu- lation PK parameters was examined using forest plots. Pharmacokinetic parameters, determined RESULTS at the 5th and 95th percentiles of a continuous covariate, or at different levels of a categorical A total of 9496 ASV concentration–time data from 1236 subjects were included in the popu- covariate, were compared with typical PK esti- mates. Effect of covariate at the extreme values lation PK analysis dataset. The majority of sub- and associated 95% CI, when wholly contained jects had 4–9 samples. Key demographics and baseline characteristics are summarized in within the 80–125% boundaries of the typical PK estimates, would suggest a lack of clinical Table 1. The majority of subjects were White (58.3%) or Asian (34.2%, among which 21.5% relevance. The impact of signiﬁcant covariates on ASV PK exposures at steady-state was also were Japanese). The average baseline values for AST and ALT were 62 and 73 U/L, respectively, evaluated using simulation. All subjects were simulated to receive the 100-mg BID soft-gel which were approximately two times the upper end of the normal range. Average ALT and AST Phase 3/commercial formulation, except when evaluating the effect of formulation, where values decreased by approximately 50% and were approaching the upper end of the normal range subjects were simulated to receive their actual formulation at 100 mg BID. The contribution of over the initial 6 weeks. Most subjects (69.7%) each covariate independently to the overall received the soft-gel capsule formulation at 100 mg BID as part of the DUAL regimen. variability of ASV PK exposures at steady-state was determined using the 5th and 95th per- centile values of a continuous covariate, or at Structural Model different levels of a categorical covariate, while ﬁxing other covariates to their respective typical The ﬁnal structural model was a two-compart- values in the population. The results were ment model with a zero-order release from the illustrated using tornado plots. Furthermore, formulation followed by ﬁrst-order absorption ASV steady-state exposures were determined into the central compartment and ﬁrst-order using all observed covariate values in the data- elimination. The auto-induction of ASV CL/ set and were summarized by quantiles of con- F was modeled with a step function occurring tinuous covariates or levels of categorical 2 days after the initial dose. A change point at covariates to allow the assessment of the impact approximately 6 days after the initial dose was of each covariate while taking into account also investigated, but found to result in a greater Infect Dis Ther (2018) 7:261–275 265 Table 1 Summary of key demographic and baseline Table 1 continued characteristics Covariate (n = 1239) Median (min, max) for Covariate (n = 1239) Median (min, max) for continuous or n (%) for continuous or n (%) for categorical covariates categorical covariates DCV/PegIFN/RBV 211 (17.0%) Age (years) 57 (18–79) DCV/ASV/PegIFB/RBV or 73 (5.9%) Weight (kg) 70 (36–124) DCV/ASV/RBV (QUAD) Baseline AST (U/L) 51 (13–595) DCV daclatasvir, ASV asunaprevir, PegIFN pegylated Baseline ALT (U/L) 60 (7–475) interferon, RBV ribavirin, ALT alanine aminotransferase, AST aspartate aminotransferase, QD once daily, BID twice Creatinine clearance (mL/ 101 (40–286) daily min) Gender OFV. Inter-individual random effects were included as a diagonal block on CL/F, central Male/female 609 (49.2%)/630 (50.8%) (V /F) and peripheral (V /F) volume of distri- c p Race bution, and the absorption rate constant (K ). White/Black/Asian/other 722 (58.3%)/74 (6.0%)/ Residual variability was modeled using an additive error model with log-transformed ASV 423 (34.2%)/20 (1.6%) concentrations. Cirrhosis No/yes/missing 996 (80.4%)/242 Covariate Models (19.5%)/1 (0.1%) Following the univariate covariate search pro- Patient type cess, age, weight, gender, race, baseline AST, ALT Non-responder/null/partial 435 (35.1%) and creatinine clearance, cirrhosis, patient type, PegIFN/RBV ineligible/ 389 (31.4%) and host CC genotype were found to be signiﬁ- cant covariates on CL/F whereas age, gender, intolerant baseline AST and ALT, and cirrhosis were signif- Naive 415 (33.5%) icant for V /F. Additionally, weight, formulation Virus genotype and ASV dose were signiﬁcant covariates for V /F, K and F, respectively. As the inclusion of base- 1A/1B/4/1 192 (15.5%)/1026 line AST resulted in greater reduction in the OFV, (82.8%)/18 (1.5%)/3 baseline ALT were excluded from subsequent (0.2%) evaluations. All other variables were included in the full model. Following the stepwise backward Formulation elimination process, covariates excluded from Tablet/soft-gel 376 (30.3%)/863 (69.7%) the ﬁnal model were weight on CL/F, creatinine Randomized dose clearance on CL/F, patient type on CL/F, host CC genotype on CL/F, age on V /F, and baseline AST 100/200/600 mg 863 (69.7%)/288 and AST ratio on V /F. The ﬁnal model did not (23.2%)/88 (7.1%) converge due to rounding errors; however, Dosing frequency bootstrap means were similar, and 95% conﬁ- dence intervals included each parameter esti- QD/BID 53 (4.3%)/1186 (95.7%) mate. The ﬁnal model parameters are provided in Treatment type Table 2. The covariate–PK parameter relation- ships in the ﬁnal model are shown as follows: DCV/ASV (DUAL) 955 (77.1%) 266 Infect Dis Ther (2018) 7:261–275 Table 2 Summary of population pharmacokinetic param- Table 2 continued eter estimates from the ﬁnal model Parameter name Parameter Standard 95% CI a b Parameter name Parameter Standard 95% CI estimates error a b estimates error (RSE%) (RSE%) CL * cirrhosis - 0.378 0.037 (3.7) - 0.447, Fixed effects - 0.298 CL/F (L/h) 50.8 4.1 (8.0) 43.3, 59.1 K * tablet - 0.503 0.071 (7.1) - 0.635, V /F (L) 47.6 5.3 (10.9) 39.4, 60.3 c - 0.356 K (1/h) 0.484 0.036 (7.4) 0.42, a Relative 0.65 0.09 (9.0) 0.473, 0.563 F1 * 600 mg 0.825 Q/F (L/h) 21.6 1.9 (8.6) 18.3, 25.8 Random effects V /F (L) 561 45 (8) 483, 661 p CL/F 0.168 0.011 (6.5) 0.145, (0.41) 0.186 D1 (h) 1.12 0.07 (6.1) 1.00, 1.28 V /F 2.19 (1.48) 0.163 (7.5) 1.877, CL/ 0.355 0.072 (7.2) 0.199, 0.5 2.519 F * induction K 0.300 0.041 (13.6) 0.229, Relative - 0.215 0.028 (2.8) - 0.265, (0.548) 0.395 F * tablet - 0.154 V /F 0.777 0.171 (22.3) 0.444, D1 * tablet 0.864 0.085 (8.5) 0.7, 1.03 (0.881) 1.101 CL * age - 0.341 0.068 (6.8) - 0.477, Residual error - 0.203 e 0.386 0.012 (3.2) 0.36, V * female - 0.608 0.123 (12.3) - 0.857, (0.621) 0.407 - 0.36 Estimates referenced to a non-cirrhotic, male, 70-kg, CL * female - 0.117 0.028 (2.8) - 0.168, 55-year-old White subject with baseline AST of 60 IU/L - 0.058 receiving the soft-gel prior to induction V * weight 1.42 0.416 (41.6) 0.686, Bootstrap statistics derived from 449 out of 500 samples. 2.242 RSE equals estimate/mean 9 100 for non-transformed parameters, and equals SE 9 100 for log-transformed CL * Black 0.0386 0.071 (7.1) - 0.095, parameters Race 0.188 Random effect and residual error estimates include the estimated variance and its square root CL * Asian - 0.255 0.032 (3.2) - 0.32, Race - 0.198 0:355IðTime [ 48Þ 0:341 CL * Other - 0.0678 0.103 (10.3) - 0.276, CL=F ¼ 50:8 e ðAge=55Þ 0:117Female 0:0386Black Race 0.123 e e 0:255Asian 0:0678Otherrace e e CL * baseline - 0.46 0.028 (2.8) - 0.512, 0:458 ðBaseline AST=60Þ AST - 0.403 0:291 ðAST=Baseline ASTÞ CL * AST -- 0.29 0.019 (1.9) - 0.325, 0:378Cirrhosis e ; - 0.251 V * cirrhosis - 0.835 0.194 (19.4) - 1.251, 0:608Female 0:835Cirrhosis V =F ¼ 47:6 e e ; - 0.475 Infect Dis Ther (2018) 7:261–275 267 1:42 gender on CL/F lay within the 80–125% bound- V =F ¼ 561 ðWeight=70Þ ; aries. Relative bioavailability was lower in subjects 0:503Tablet Ka ¼ 0:484 e ; receiving the tablet relative to the soft-gel for- 0:864Tablet D ¼ 1:12 e ; mulation, but increased with the 600-mg tablet 0:215Tablet 0:65Doseð600 mgÞ dose relative to the other tablet doses. The dura- F ¼ e e : tion of absorption increased with the tablet for- mulation. The effect of covariates on other population PK parameters should be interpreted with caution given the high shrinkage values Model Evaluations associated with these parameters. Goodness-of-ﬁt diagnostic plots show good Effect of Covariates on ASV Steady-State agreement between the predicted and the Exposures observed ASV concentrations, and that condi- tional-weighted residuals are unbiased over time The impact of each covariates independently on and population predicted concentrations ASV steady-state AUC based on the ﬁnal model is (Fig. 1). The random effect for CL/F, V /F, V / c p shown in Fig. 4a. Baseline AST, cirrhosis, and F and K showed a symmetric, bell-shaped dis- change in AST from baseline had moderate effects, tribution that was centered near zero. The pcVPC each covariate contributing 30–50% change in plots in Fig. 2 showed good agreement between AUC. The effect of Asian race was modest; Asians the median of the data and the predicted medi- have a 29.2% increase in AUC relative to the ans and distributions; the variability appears to White/Caucasians. Other covariates such as age, be slightly under-predicted in the ﬁrst 4 h, and gender, and weight had a minimal effect (\ 20%). slightly over-predicted around 12 h. However, The predicted steady-state AUC was 19% lower Fig. 2b, c shows consistency of the model with with the tablet than with the soft-gel formulation, the pre- and post-induction periods with the which was less than expected given known dif- 200-mg tablet. The shrinkage value for the ferences in relative bioavailability between the residual error was 11%, indicative of the model’s two formulations.A similar trend was observed for adequacy in providing individual predictions for ASV steady-state minimum concentration (C ), min subsequent exposure–response analysis. Shrink- although the magnitude of covariate effects was age on CL/F was 13%, suggesting each subject’s generally larger (Fig. 4b). CL/F is well estimated and can be used for further The impact of select covariate on ASV steady- exploration of covariate-CL/F relationships. state exposures when taking into account the Shrinkage values on V /F, K and V /F were 39, 42 c a p potential correlations amongst covariates in the and 65%, respectively, suggesting signiﬁcant patient population is shown in Fig. 5 and amounts of shrinkage in the empirical Bayes Table 3. Asunaprevir steady-state AUC increased estimates of posterior individual parameters. from - 25.6 to 57.7% relative to the median value in subjects with baseline AST from the Effect of Covariates on Population PK ﬁrst to the fourth quartile (Fig. 5a). Cirrhotic Parameters subjects had an 84.3% increase in AUC relative to non-cirrhotic subjects (Fig. 5b). Japanese and The impact of covariates retained in the ﬁnal non-Japanese Asians had an approximately model on the population PK parameters is shown 40–45% increase in AUC relative to the White/ in Fig. 3. Asunaprevir CL/F decreased with Caucasian subjects (Fig. 5c). The tablet formu- increasing baseline AST, increasing AST relative to lation (100 mg BID with actual observed patient baseline, and cirrhosis, with the 95% CIs for these characteristics in the dataset) produced an effects lay outside the 80–125% region. The CL/ approximately 40% lower AUC compared to the F decreased with increasing age and decreased in soft-gel formulation (Fig. 5d). The ASV steady- Asians compared to Caucasians, with the 95% CIs state AUC changed more modestly with age, overlapped the 80–125% region. The effect of 268 Infect Dis Ther (2018) 7:261–275 Fig. 1 Goodness-of-ﬁt diagnostic plots for the ﬁnal model. CWRES conditional weighted residual. Dashed lines the lowest (local regression smoother) trend lines gender and weight, suggesting a lack of clinical with ﬁrst-order elimination from the central signiﬁcance (data not shown). compartment, an induction effect on clearance, and an absorption model consisted of zero- order release followed by ﬁrst-order absorption DISCUSSION adequately described ASV PK after oral admin- istration. For a typical White, non-cirrhotic The purpose of this analysis was to characterize 70-kg, 55-year-old male with baseline AST of the population PK of ASV in patients with 60 IU/mL receiving the 100 mg BID of the soft- chronic HCV infection, who are treatment- gel formulation, ASV CL/F was 50.8 L/h, naive, non-responders and ineligible/naive increasing by 43% after 2 days to a CL/F of 72.5 intolerant to IFN-based therapy. This analysis L/h at steady-state, most likely due to auto-in- provides a quantitative description of ASV PK duction of CYP3A4. Since there were limited and, in particular, evaluated the impact of key ASV PK data available during the ﬁrst 7 days covariates such as hepatic function, race, gen- after the start of dosing in the PPK dataset, the der, age, weight and formulation on ASV steady- dynamics of the induction process could not be state exposures. A two-compartment model estimated. Therefore, a change in clearance Infect Dis Ther (2018) 7:261–275 269 bFig. 2 Prediction-corrected visual predictive check plots for the ﬁnal model. Circles are observed ASV plasma concentrations, and dashed red lines represent 5th, 50th and 95th percentiles of the observed values. The blue ﬁeld represents 95% CI of the model simulated median and the red ﬁelds 95% CI of the 5% tile (lower) and 95% tile (upper) of the simulation-based prediction intervals. Data and predicted values are binned, and plotted at the center points of each bin. a All studies (n = 1236); b subjects who received the tablet formulation at 200 mg BID, pre- induction (n = 94); c subjects who received the tablet formulation at 200 mg BID, post-induction (n = 286). A corresponding pcVPC plot for the soft-gel formulation was not produced due to the lack of PK samples during the ﬁrst week of dosing when auto-induction presumably developed Fig. 3 Effect of covariates on asunaprevir population pharmacokinetic parameters of the ﬁnal model. PK parameters were referenced to a non-cirrhotic, male, 70-kg, 55-year-old White subject with baseline AST of 60 IU/L receiving the soft-gel prior to induction. The horizontal axis represents percent change in PK parameter relative to the reference. The thick horizontal lines indicate the PK parameter change at the 5th and 95th percentile of the continuous covariates. The points indicate the PK parameter change attributed to the respective continuous covariate value or categorical covariate. The thin horizon- tal lines represent the 5th to 95th percentile conﬁdence interval in the estimated value. Vertical lines indicate the reference PK parameter value (100%) and 80% and 125% changes with time was investigated with a step function as part of the base model. 270 Infect Dis Ther (2018) 7:261–275 However, it was not included in the ﬁnal model due to its high correlation with baseline AST. These results are consistent with the pharma- cokinetic proﬁle of ASV which is transported into the liver by OATP1B1, and is extensively metabolized in the liver via the CYP3A4-medi- ated oxidative pathway and eliminated primar- ily through the feces. The effect of AST on CL/ F has been previously reported for drugs that are extensively metabolized in the liver, such as sildenaﬁl and tacrolimus [22, 23]. Cirrhosis changes the architecture of the liver leading to changes in blood ﬂow, protein binding and drug-metabolizing enzymes . Reduction in the enzyme level and activity due to the loss of hepatic function could result in decreased clearance of drugs that are primarily metabo- lized by the liver. In the hepatic impairment study conducted in non HCV-infected subjects, ASV exposures were signiﬁcantly higher in subjects with moderate and severe hepatic impairment. In the hepatic impairment study, there were no signiﬁcant changes in ASV exposure in subjects with mild hepatic impair- ment compared to matched controls with nor- mal hepatic function, whereas ASV AUC was 10- and 32-fold higher in subjects with moderate and severe hepatic impairment, respectively. The results from the population PK analysis therefore provided further evidence that ASV exposures were highly inﬂuenced by the degree of hepatic impairment. Worsening of hepatic Fig. 4 Effect of covariates on asunaprevir steady-state function may lead to increased ASV exposures, exposures. The black bar represents the 5th to 95th which may in turn impact the safety proﬁle of percentile range of the exposures calculated using individ- ASV. However, it should be noted that, in the ual PK parameter estimates. The impact of each covariate current dataset, ASV exposures in subjects with on exposure was calculated using the 5th to 95th percentile cirrhosis and those with elevated AST levels range of each continuous covariate, or the covariate were limited to \ 2-fold of the population category, while ﬁxing other covariates to the respective median exposure value, and were well within typical values in the population the range of observed exposures in Phase II/III studies. This modest change is consistent with the fact that only subjects with compensated Factors indicative of hepatic function were cirrhosis and hepatic function no worse than identiﬁed as key inﬂuential covariates on ASV exposures. Subjects with cirrhosis had an Child–Pugh A (mild impairment) were included in the Phase II/III population. Therefore, a approximately 84% increase in ASV AUC (Fig. 5b) and subjects with baseline AST in the potential limitation of the model is to extrapo- late ASV exposures in patients with cirrhosis fourth quartile (above 78 IU/L) had an approx- and worsening hepatic function (Child–Pugh B imately 58% increase in AUC (Fig. 5a). Baseline and C). In addition, it is known that ASV ALT was also identiﬁed as a signiﬁcant covariate exposures are generally higher in subjects with on ASV CL/F in the univariate screening step. Infect Dis Ther (2018) 7:261–275 271 Fig. 5 Effect of select covariates on asunaprevir steady- extend to 1.5 times the interquartile range. The notch state AUC. Steady-state exposure was calculated using indicates an approximate 95% conﬁdence interval on the individual PK parameter estimates. The box shows the median, and is calculated as 1.58 times the interquartile median, 25th and 75th percentile per group. The whiskers range normalized to the number of data points chronic HCV infection compared to those in with the DUAL regimen, only 3% of subjects healthy subjects, highlighting the role of hep- had experienced Grade 3/4 AST or ALT eleva- atic function in the PK of ASV. Although the tions, consistent with ﬁndings from a previous number of subjects with cirrhosis was limited, exposure–response analysis that suggested the difference in plasma exposure in subjects minimal increase in the probability of liver with compensated cirrhosis did not appear to be safety events, with an approximate doubling of clinically meaningful, as virological outcomes ASV AUC at the clinically relevant dose [25, 26]. in these subjects were comparable to those in On the other hand, subjects receiving the DUAL subjects without cirrhosis. In Phase III studies treatment had in general reduced AST levels as a 272 Infect Dis Ther (2018) 7:261–275 Table 3 Effect of selected covariates on asunaprevir steady-state AUC Covariate name Covariate categories n Simulated ASV steady-state AUC % Difference from (ng h/mL) reference (%) Median 95% PI Baseline AST 1st quartile (median = 28 IU/L) 321 1162 568, 2278 - 25.6 2nd quartile (median = 42 IU/L) 299 1310 729, 2872 - 16.1 3rd quartile (median = 62 IU/L) 309 1706 882, 3743 9.3 4th quartile (median = 104 IU/L) 310 2462 1250, 5064 57.7 All (median = 51 IU/L) 1239 1561 729, 4042 – Cirrhosis No 996 1406 695, 3241 – Yes 242 2591 1206, 5251 84.3 Race White 722 1388 662, 3735 – Black 74 1389 670, 3111 0.1 Japanese 267 1945 986, 4393 40.1 Non-Japanese Asian 156 2023 956, 4125 45.7 Formulation Tablet 376 1039 532, 2725 - 40.3 Soft-gel capsule 863 1739 791, 4188 – PI prediction interval result of viral clearance. Covariate effect of the approximately 21.5% subjects were Japanese, time-varying AST change from baseline sug- and Japanese subjects had a higher proportion gested that, for subjects with the largest reduc- of females (approximately 65% compared to tion in AST during treatment (43% decrease 37%) and higher median age (approximately from baseline), ASV CL/F increased by 27%. As 62 years compared to 55 years) than non-Ja- the subjects’ hepatic function normalized, there panese subjects. Advanced age had been esti- was a lowering effect on ASV exposure which mated to independently contribute to thus further mitigated the risk of safety events. approximately 9 and 16% higher ASV steady- Race was identiﬁed as a signiﬁcant covariate state AUC and C , respectively (Fig. 4). min on ASV CL/F, with approximately 29% lower Females were estimated to have 12 and 19% CL/F in Asian subjects when race was assessed as higher ASV steady-state AUC and C , min an independent covariate. However, when the respectively. effect of race on ASV AUC was evaluated after The effect of formulation on the bioavail- accounting for potential correlations between ability of ASV suggested that the bioavailability covariates, Asians in general and Japanese sub- of the Phase III/commercial soft-gel capsule jects had approximately 46 and 55% higher formulation was approximately 1.3-fold higher steady-state AUC and C , respectively, relative than the tablet formulation used in the Phase II min to the White/Caucasian subjects. The overall studies. Based on the relative bioavailability higher ASV exposures in Asians and Japanese study in healthy subjects, it was expected that subjects were likely contributed by the different ASV bioavailability following the soft-gel Phase age and gender distribution among the study III/commercial formulation at 100 mg BID populations. In the current dataset, would be approximately 2-fold higher relative Infect Dis Ther (2018) 7:261–275 273 to the Phase II tablet at 200 mg BID. In the included Phase II/III studies and may guide current dataset, subjects received a wide range clinical use of the drug. The results of the of the tablet doses/regimens including 200 mg analysis were used for the development of the QD, 200 mg BID, 600 mg QD and 600 mg BID. exposure–response analyses. As reported previously , and identiﬁed in the current analysis, ASV exposure increased more than proportionally with an increase in dose ACKNOWLEDGEMENTS from 200 mg BID to 600 mg BID of the Phase II tablet, especially in Japanese subjects, which may have also contributed to the lower estimate Funding. The studies described in this report of the difference in bioavailability between the and the article processing charges were funded soft-gel and tablet formulations. When ASV by Bristol-Myers Squibb. 100 mg BID of the soft-gel capsule was directly compared to the tablet at the same dose, simu- Authorship. All authors had full access to all lations using the ﬁnal model suggest that the of the data in this study and take complete tablet had approximately 40% lower AUC, responsibility for the integrity of the data and which would translate into an approximate accuracy of the data analysis. All named authors difference of 1.7-fold in bioavailability between meet the International Committee of Medical the formulations, which was closer to the Journal Editors (ICMJE) criteria for authorship expected difference. for this article, take responsibility for the Other covariates such as age, gender, and integrity of the work as a whole, and have given weight which were retained in the ﬁnal model their approval for this version to be published. had statistically signiﬁcant effects; however, with marginal importance, i.e. \ 20% effect on Disclosures. Yash Gandhi is an employee of ASV steady-state exposure. In addition to the Bristol-Myers Squibb Research and Develop- ment. Tushar Garimella is an employee of Bris- above-described covariates the full model iden- tiﬁed the following covariates as signiﬁcant tol-Myers Squibb Research and Development. Li (p \ 0.05): sex, weight, patient type and CC Zhu is an employee of Bristol-Myers Squibb genotype on CL/F and baseline AST and time- Research and Development. Richard Bertz is a varying AST on V /F. These covariates were former employee of Bristol-Myers Squibb and dropped from the ﬁnal model during the back- was employed by Bristol-Myers Squibb when ward elimination process. Other covariates, the studies and analyses were conducted. including OATP transporters and comedica- Richard Bertz is currently employed at Univer- tions, were not identiﬁed as signiﬁcant in the sity of Pittsburgh (Pittsburgh, PA, USA). Marc Bifano is a former employee of Bristol-Myers univariate screening, and were not included in the full model, and therefore it can be inferred Squibb and was employed by Bristol-Myers that they did not help explain the observed PK Squibb when the studies and analyses were variability in ASV. conducted. Marc Bifano is now retired. Phyllis Chan is a former employee of Bristol-Myers Squibb and was employed by Bristol-Myers CONCLUSION Squibb when the studies and analyses were conducted. Phyllis Chan is currently employed In conclusion, the current ASV PPK model pro- at Genentech Inc (South San Francisco, CA, vided a parsimonious description of the con- USA). Timothy Eley is a former employee of centration data in subjects with chronic HCV Bristol-Myers Squibb and was employed by infection. The key covariates identiﬁed and Bristol-Myers Squibb when the studies and described in the model, such as hepatic func- analyses were conducted. Timothy Eley is cur- tion status, race and formulation, had statistical rently employed at Arbutus Biopharma Corpo- and biological plausibility that help explain the ration (Warminster, PA, USA). Eric Hughes is a observed variability in ASV exposures in the former employee of Bristol-Myers Squibb and 274 Infect Dis Ther (2018) 7:261–275 3. McPhee F, Sheaffer AK, Friborg J, Hernandez D, Falk was employed by Bristol-Myers Squibb when P, Zhai G, et al. Preclinical proﬁle and characteri- the studies and analyses were conducted. Eric zation of the hepatitis C virus NS3 protease inhi- Hughes is currently employed at Novartis bitor asunaprevir (BMS-650032). Antimicrob Agents International AG (Basel, Switzerland). Malaz Chemother. 2012;56:5387–96. AbuTarif is a former employee of Bristol-Myers 4. Gentile I, Buonomo AR, Zappulo E, Minei G, Mor- Squibb and was employed by Bristol-Myers isco F, Borrelli F, Coppola N, Borgia G. Asunaprevir, Squibb when the studies and analyses were a protease inhibitor for the treatment of hepatitis C conducted. Malaz AbuTarif is currently infection. Ther Clin Risk Manag. 2014;26(10):493–504. employed at Daiichi Sankyo Inc (Parsippany, NJ, USA). Mayu Osawa is an employee of Bristol- 5. Manns M, Pol S, Jacobson IM, Marcellin P, Gordon Myers Squibb K.K. Takayo Ueno is an employee SC, Peng CY, et al. All-oral daclatasvir plus of Bristol-Myers Squibb K.K. Hanbin Li is cur- asunaprevir for hepatitis C virus genotype 1b: a multinational, phase 3, multicohort study. Lancet. rently employed by Certara and was funded by 2014;384:414–29. Bristol-Myers Squib to carry out the modeling analyses reported in this paper. 6. Bronowicki JP, Ratziu V, Gadano A, Thuluvath PJ, Bessone F, Martorell CT, et al. Randomized trial of Compliance with Ethics Guidelines. This asunaprevir plus peginterferon alfa and ribavirin for previously untreated genotype 1 or 4 chronic hep- article is based on previously conducted studies atitis C. J Hepatol. 2014;61:1220–7. and does not contain any studies with human participants or animals performed by any of the 7. Zeuli JD, Adie SK, Rizza SA, Temesgen Z. Drugs authors. today (Barc), Asunaprevir plus daclatasvir for the treatment of chronic hepatitis C virus infection. Drugs Today (Barc). 2015;51(11):629–43. Data Availability. The datasets during and/ or analyzed during the current study are avail- 8. Yang SS, Kao JH. Asunaprevir-containing regimens able from the corresponding author on reason- for the treatment of hepatitis C virus infection. Expert Rev Gastroenterol Hepatol. 2015;9(1):9–20. able request. 9. Eley T, Garimella T, Li W, Bertz RJ. Asunaprevir: a Open Access. This article is distributed review of preclinical and clinical pharmacokinetics under the terms of the Creative Commons and drug–drug interactions. Clin Pharmacokinet. Attribution-NonCommercial 4.0 International 2015;54(12):1205–22. License (http://creativecommons.org/licenses/ 10. Eley T, He B, Huang SP, Li W, Pasquinelli C, by-nc/4.0/), which permits any noncommer- Rodrigues AD, Grasela DM, Bertz RJ. Pharmacoki- cial use, distribution, and reproduction in any netics of the NS3 protease inhibitor, Asunaprevir medium, provided you give appropriate credit (ASV, BMS-650032), in phase I studies in subjects with or without chronic hepatitis C. Clin Pharma- to the original author(s) and the source, provide col Drug Dev. 2013;2(4):316–27. a link to the Creative Commons license, and indicate if changes were made. 11. Eley T, Garimella T, Li W, Bertz RJ. Asunaprevir: a review of preclinical and clinical pharmacokinetics and drug–drug interactions. Clin Pharmacokinet. 2015;54:1205–22. REFERENCES 12. Gong J, Eley T, He B, Arora V, Philip T, Jiang H, Easter J, Humphreys WG, Iyer RA, Li W. Charac- terization of ADME properties of [(14)C]asunaprevir 1. Gower E, Estes CC, Hindman S, Razavi-Shearer K, (BMS-650032) in humans. Xenobiotica. Razavi H. Global epidemiology and genotype dis- 2016;46(1):52–64. tribution of the hepatitis C virus infection. J Hepa- tol. 2014;61(S1):S45–57. 13. Eley T, Gardiner D, Persson A, He B, You X, Shah V, et al. Evaluation of drug interaction potential of the 2. Alter MJ, Kruszon-Moran D, Nainan OV, et al. The HCV protease inhibitor asunaprevir (ASV; BMS- prevalence of hepatitis C virus infection in the 650032) at 200 mg twice daily in metabolic cocktail United States, 1988 through 1994. N Engl J Med. and P-glycoprotein (P-gp) probe studies in healthy 1999;341:556–62. Infect Dis Ther (2018) 7:261–275 275 volunteers [abstract]. Hepatology. 2011;54:abstract 19. Parke J, Holford NH, Charles BG. A procedure for no. 381. generating bootstrap samples for the validation of nonlinear mixed-effects population models. Com- 14. Eley T, He B, Chang I, Colston E, Child M, Bedford put Methods Programs Biomed. 1999;59(1):19–29. W, et al. The effect of hepatic impairment on the pharmacokinetics of asunaprevir, an HCV NS3 20. Bergstrand M, Hooker AC, Wallin JE, et al. Predic- protease inhibitor. Antivir Ther. 2015;20(1):29–37. tion-corrected visual predictive checks for diagnos- ing nonlinear mixed-effects models. AAPS J. 15. Morcos PN, Moreira SA, Brennan BJ, Blotner S, 2011;13:143. Shulman NS, Smith PF. Inﬂuence of chronic hep- atitis C infection on Asunaprevir Preclinical and 21. Savic RM, Karlsson MO. Importance of shrinkage in Clinical PK and Drug Interactions 1221 cytochrome empirical bayes estimates for diagnostics: problems P450 3A4 activity using midazolam as an in vivo and solutions. AAPS J. 2009;11(3):558–69. probe substrate. Eur J Clin Pharmacol. 2013;69:1777–84. 22. Milligan PA, Marshall SF, Karlsson MO. A popula- tion pharmacokinetic analysis of sildenaﬁl citrate in 16. Eley T, Chan P, Huang S, Wind-Rotolo M, Del- patients with erectile dysfunction. Br J Clin Phar- macol. 2002;53(Suppl 1):45S–52S. monte T, He B, et al. Asunaprevir in Japanese sub- jects in phase 2: exposure-safety versus US/EU- based subjects and preliminary assessment of cor- 23. Oteo I, Lukas JC, Leal N. Tacrolimus pharmacoki- netics in the early post-liver transplantation period relation with single nucleotide polymorphisms and clinical applicability via bayesian prediction. (SNPs) in liver uptake transporters. Hepatology. Eur J Clin Pharmacol. 2013;69(1):65–74. 2012;56(Suppl 1):1070A–1A. 24. Elbekai RH, Korashy HM, El-Kadi AOS. The effect of 17. Eley T, Chan P, Sverdlov O, He B, Bedford B, Kan- liver cirrhosis on the regulation and expression of doussi H, et al. Improved bioavailability and miti- drug metabolizing enzymes. Curr Drug Metab. gated food effect for asunaprevir utilizing a lipid- based formulation: similar exposure with 100 mg 2004;5(2):157–67. twice-daily soft-gel capsule relative to 200 mg twice daily of phase 2 tablet [abstract/poster no. A-1247]. 25. Chan P, et al. Exposure-response analyses of In: 52nd Interscience conference antimicrobial asunaprevir in combination with agents chemotherapy. San Francisco; 9–12 Sep daclatasvir ± peginterferon/ribavirin among 2012. patients with genotype 1 chronic HCV infection: dose selection for phase 3 clinical trials. J Hepatol. 18. Marc Gastonguay. full covariate models as an 2013;58(Suppl1):S328–9. alternative to methods relying on statistical signif- icance for inferences about covariate effects: a 26. Gentile I, Zappulo E, Buonomo AR, Scotto R, Borgia review of methodology and 42 case studies. Page 20 G. Asunaprevir for hepatitis C: a safety evaluation. (2011) Abstr 2229. http://www.page-meeting.org/ Expert Opin Drug Saf. 2015;14(10):1631–46. ?abstract=2229. Accessed 21 Mar 2018.
Infectious Diseases and Therapy
– Springer Journals
Published: Mar 27, 2018