Population pharmacokinetic analysis to recommend the optimal dose of udenafil in patients with mild and moderate hepatic impairment

Abstract

Aims

The aim of this study was to develop a population pharmacokinetic (PK) model of udenafil and its active metabolite, DA‐8164, in healthy subjects and patients with hepatic impairment (HI) and to estimate the optimal dosing recommendations for patients with HI.

Methods

An open label, three parallel group, age and weight matched control study was conducted in 18 volunteers, six healthy subjects (n = 6) and patients with mild (Child–Pugh class A, n = 6) and moderate HI (Child–Pugh class B, n = 6). Serial blood samples were collected for up to 72 h after a single administration of udenafil 100 mg. A population PK model was developed using non‐linear mixed effects modelling (nonmem, ver. 7.2). The simulated data from the final PK model and original data of healthy subjects were compared to identify the optimal dose for patients with HI.

Results

A two compartment model for both udenafil and DA‐8164 best described the data. Prothrombin time on metabolic clearance of udenafil to DA‐8164 was included in the final model as a covariate. Compared with the AUC(0,t _last) value after administration of udenafil 100 mg to healthy subjects, the geometric mean ratios (95% confidence interval) after 100 mg and 75 mg udenafil administration were 1.21 (1.10, 1.32) and 0.74 (0.67, 0.81) in patients with mild HI, respectively. Meanwhile, those were 1.55 (1.43, 1.67) and 1.02 (0.92, 1.12) in patients with moderate HI, respectively.

Conclusions

This study suggests that the recommended doses of udenafil are 100 mg and 75 mg in patients with mild and moderate HI, respectively.

What is Already Known about this Subject

• Udenafil is a phosphodiesterase type 5 (PDE‐5)‐selective inhibitor and is used for the treatment of erectile dysfunction.

• Udenafil is mainly metabolized by the liver and udenafil and its active metabolite DA‐8164 strongly bind to plasma protein.

• Hepatic impairment affects pharmacokinetics, highlighting the need for dose adjustment.

What this Study Adds

• This study evaluated the exposure to udenafil in patients with hepatic impairment and optimized the dosage according to the severity of the liver disease using population pharmacokinetic modelling and simulation.

• A lower starting dose of 75 mg of udenafil is recommended for patients with moderate hepatic impairment.

Introduction

Erectile dysfunction (ED) is the persistent inability to attain and maintain an erection sufficient to permit satisfactory sexual performance. ED is the most common complaint among men and the prevalence of ED was suggested to be 5–20% in various epidemiologic studies 1. Phosphodiesterase type 5 (PDE‐5) inhibitors are used as the first line therapy for ED 1 and have improved the quality of life of men with ED owing to convenience, safety and efficacy of use for various aetiologies. The US Food and Drug Administration (FDA) and the European Medicine Agency have approved sildenafil, vardenafil and tadalafil for the treatment of ED 2, 3.

The liver is a major organ involved in the elimination of many drugs through a variety of oxidative and conjugative metabolic pathways and/or through biliary excretion of unchanged drug or its metabolites 4. Changes in the excretory and metabolic activities due to hepatic impairment (HI) can affect the absorption, distribution and elimination of the drug and its metabolite. Known PDE‐5 inhibitors are predominantly eliminated by hepatic metabolizing enzymes via the cytochrome P450 (CYP) isoform 3A4 5. Therefore, ED patients with HI, 50–60% prevalence reported among the patients with chronic liver disease 6, 7, may require dose adjustment of some PDE‐5 inhibitors 1, 2. Normally, lower doses are recommended as starting doses or maximum recommended doses for sildenafil, vardenafil and tadalafil for patients with HI 2.

Udenafil (Zydena^®; Dong‐A ST Co. Ltd, Seoul, Korea) is a selective inhibitor of PDE‐5 and is used for the treatment of ED. Udenafil is mainly metabolized by CYP3A4 to its active metabolite, DA‐8164 8. In addition, udenafil and DA‐8164 are strongly bound to plasma protein (93.9% and 98.8% in humans, respectively) 9. Therefore, HI can influence the pharmacokinetics (PK) of udenafil and its active metabolite, influencing the efficacy or safety of the drug.

This study aimed to evaluate the exposure difference of udenafil between patients with HI and healthy subjects and to estimate the optimal dosage according to the severity of the liver disease. An open label, single dose, three parallel group, age and weight matched controls, PK study was performed in healthy subjects and in patients with mild and moderate HI. Based on the results, a population PK model of udenafil and its active metabolite, DA‐8164, was developed in healthy subjects and in patients with HI to recommend the optimal dosage regimen for patients with HI.

Methods

Clinical study, PK assessment and analyses

This study comprised of three groups, a mild HI group (Child‐Pugh classification A, n = 6), moderate HI group (Child‐Pugh classification B, n = 6) and healthy subjects (n = 6) whose age and weight were matched to those of the patients with moderate HI, in the range of ±10 years and ±10 kg, respectively. The age of the 18 subjects varied from 41–61 years. The following exclusion criteria were applied: history of cardiovascular and cerebrovascular disease, medical conditions that could impair participation and continuation in the study, creatinine clearance <40 ml min^–1 and the use of drugs that affect the PK of udenafil such as CYP3A4 and CYP2D6 inducers or inhibitors which are contraindicated for co‐administration with the PDE‐5 inhibitor. The subjects gave written informed consent to enrol in this study.

This open label, multicentre, three parallel group, age and weight matched control study was approved by four institutional review boards; the Institutional Review Board of Seoul National University Hospital, Seoul, Korea, Seoul Metropolitan Governance‐Seoul National University Boramae Medical Centre Institutional Review Board, Seoul, Korea, Institutional Review Board of Seoul National University Bundang Hospital, Seongnam, Korea and the Institutional Review Board of Asan Medical Center, Seoul, Korea. The clinical study was conducted in accordance with the Declaration of Helsinki, the International Conference on Harmonization Guidelines for Good Clinical Practice. This trial was registered in ClinicalTrials.gov (identifier: NCT00956306). Eighteen subjects were administered a single 100 mg oral dose of udenafil in the fasting state for at least 10 h, and serial blood samples were collected for 72 h; pre‐dose, 0.5, 1, 1.5, 2, 2.5, 3, 4, 6, 8, 12, 24, 32, 48 and 72 h after drug administration.

Liquid chromatography–tandem mass spectrometry was used for determination of udenafil and DA‐8164 plasma concentrations 10. The lower limit of quantification was 2 ng ml^–1 and the method was validated over a range of 2–2000 ng ml^–1. The accuracy for within‐ and between runs ranged from 94.6% to 102.3% and from 96.3% to 101.2%, respectively, and the precision for within and between runs ranged from 1.2% to 9.5% and from 1.7% to 4.3%, respectively.

The PK parameters of udenafil and DA‐8164 were determined by non‐compartmental analysis (NCA) using Phoenix 6.3 (Pharsight, Mountain View, CA, USA) and statistical analyses were performed using SAS software release 9.3 (SAS Institute, Cary, NC, USA). The peak concentration (C _max) and area under the concentration–time curve from 0 to the last time point (AUC(0,t _last)) were corrected with analysis of covariance for age and weight. For pairwise comparison of parameters among the groups, a post hoc least significant difference test was conducted. The results are presented as the ratio of geometric mean of AUC(0,t _last) and C _max with 95% confidence interval (CI).

Population pharmacokinetic model development

A population PK analysis was conducted using the first order conditional estimation method with η − ε interaction in nonmem version 7.2 (Icon Development Solutions, Ellicott City, MD, USA) with the G77 Fortran compiler. One or two compartment models with first order or zero order absorption were evaluated to identify the one that best described the plasma concentration–time profiles of udenafil and DA‐8164. The plasma concentration–time data of udenafil and DA‐8164 were modelled simultaneously. Since the fraction of udenafil metabolized to DA‐8164 and volume of distribution of DA‐8164 have not been reported in the literature and cannot be estimated using our data, central and peripheral volumes of distribution of udenafil were assumed to be equal to those of DA‐8164, respectively. The subroutine chosen to build the model was ADVAN6 TOL3 in the nonmem library. Because the molecular weights of udenafil (516.66 g mol^–1) and DA‐8164 (405.4 g mol^–1) were quite different 10, the molecular weight ratio of DA‐8164 to udenafil (R_pm) was multiplied by the turnover rate of udenafil to DA‐8164 in a differential equation equation (4). The differential equations for the structural model were as follows:

Depot:

$$\mathrm{\Delta }{\mathrm{A}}_1/{\mathrm{dt}=-\mathrm{k}}_{\mathrm{a}}\times {\mathrm{A}}_1$$ (1)

Central compartment (Parent drug, udenafil):

$${\mathrm{\Delta A}}_2{/\mathrm{dt}=\mathrm{k}}_{\mathrm{a}}{\times \mathrm{A}}_1{-\mathrm{CL}}_{\mathrm{pm}}{/\mathrm{F}\times \mathrm{C}}_2{-\mathrm{Q}}_{\mathrm{p}}{/\mathrm{F}\times \mathrm{C}}_2{+\mathrm{Q}}_{\mathrm{p}}{/\mathrm{F}\times \mathrm{C}}_3{-\mathrm{CL}}_{\mathrm{p}}{/\mathrm{F}\times \mathrm{C}}_2$$ (2)

Peripheral compartment (Parent drug, udenafil):

$${\mathrm{\Delta }A}_3{/\mathrm{dt}=\mathrm{Q}}_{\mathrm{p}}{/\mathrm{F}\times \mathrm{C}}_2{-\mathrm{Q}}_{\mathrm{p}}{/\mathrm{F}\times \mathrm{C}}_3$$ (3)

Central compartment (Metabolite, DA‐8164):

$${\mathrm{\Delta }A}_4{/\mathrm{dt}=\mathrm{CL}}_{\mathrm{pm}}{/\mathrm{F}\times \mathrm{C}}_2{\times \mathrm{R}}_{\mathrm{pm}}{-\mathrm{Q}}_{\mathrm{m}}{/\mathrm{F}·\mathrm{f}}_{\mathrm{m}}{\times \mathrm{C}}_4{+\mathrm{Q}}_{\mathrm{m}}{/\mathrm{F}·\mathrm{f}}_{\mathrm{m}}{\times \mathrm{C}}_5{-\mathrm{CL}}_{\mathrm{m}}{/\mathrm{F}·\mathrm{f}}_{\mathrm{m}}{\times \mathrm{C}}_4$$ (4)

Peripheral compartment (Metabolite, DA‐8164):

$${\mathrm{\Delta }A}_5{/\mathrm{dt}=\mathrm{Q}}_{\mathrm{m}}{/\mathrm{F}·\mathrm{f}}_{\mathrm{m}}{\times \mathrm{C}}_4{-\mathrm{Q}}_{\mathrm{m}}{/\mathrm{F}·\mathrm{f}}_{\mathrm{m}}{\times \mathrm{C}}_5$$ (5)

where A_n is the amount of drug in the n^th compartment, C _n is the concentration of the drug in the n^th compartment, k _a is the absorption rate constant, CL_pm/F is the metabolic clearance of udenafil to DA‐8164, CL_p/F is the apparent clearance of udenafil except in the metabolic pathway to DA‐8164, Q_p/F is the inter‐compartmental clearance of udenafil, CL_m/F·f _m is the clearance of DA‐8164 and Q_m/F·f _m is the inter‐compartmental clearance of DA‐8164.

Inter‐individual variability (IIV) of PK parameters was evaluated using the exponential terms and the PK parameters of the i^th subject (P_i) were described using the following equation:

$${\mathrm{P}}_{\mathrm{i}}=\mathrm{\theta }\mathit{exp}\left({\mathrm{\eta }}_{\mathrm{i}}\right)$$ (6)

where θ is the typical value of the PK parameters and η_i is a random variable of the i^th subject. Additive, proportional and combined (additive plus proportional) error models were compared for residual errors.

Model selection was based on the likelihood ratio test, Akaike information criterion and goodness‐of‐fit plots. A decrease in the objective function value (OFV) greater than 3.84 (α = 0.05, d.f. = 1) between two nested models was considered significant.

Covariate selection and model evaluation

The covariate model was developed sequentially with forward inclusion and backward deletion. The levels of significance used during the forward inclusion and backward deletion were 5% and 1%, respectively. The covariate screening procedure was conducted using numerical (generalized additive modelling) and visual methods (scatterplots for continuous variables and boxplots for categorical variables). Demographic and clinical variables such as age, height, body weight, creatinine clearance and prothrombin time were expressed as international normalized ratio (PT) and the serum albumin, total bilirubin, alanine transaminase, aspartate transaminase, presence of ascites and the study group were evaluated as potential covariates.

The non‐parametric bootstrap approach was used to evaluate the stability and reliability of the final PK model. The final population PK model was fitted repeatedly to 1000 resampled data sets. The medians and 95% CI of the parameters obtained from the bootstrap method were compared with the final PK parameter estimates. The median and 90% prediction interval of 1000 simulated data from the final model were overlaid with the observed data classified by the study group.

Simulation and statistical analysis for dose recommendation in patients with hepatic impairment

The plasma concentration–time data for 100 virtual subjects by each group were simulated from the final PK model. The maximum concentration (C _max) and area under the plasma concentration–time curve of the last quantifiable concentration (AUC(0,t _last)) of udenafil were calculated by non‐compartmental analysis from the simulated data.

Udenafil is known to exhibit dose non‐proportionality with both C _max and AUC(0,t _last) increasing supraproportionally with increased dosing 11. Because our study was performed with only one dose of udenafil 100 mg, linear models for C _max and AUC(0,t _last) were developed to correct the dose proportionality using individual C _max and AUC(0,t _last) from udenafil 25 mg to 100 mg in a previous single dose study 11. Data such as the dose and the C _max and AUC(0,t _last) values were log‐transformed prior to the analysis. Linear models for C _max and AUC(0,t _last) are as follows and presented in the supplementary material:

$$log\left({\mathrm{C}}_{max}\right)=1.485\times log\left(\text{dose}\right)-0.812$$ (7)

$$log\left({\mathrm{AUC}}_{\left(0,,,\text{tlast}\right)}\right)=1.700\times log\left(\text{dose}\right)+0.0316$$ (8)

When less than 100 mg udenafil was administered, the ratio of the PK parameter from the linear model to the PK parameter from the dose proportionality was multiplied by individual PK parameters (i.e. when udenafil 75 mg was administered, C _max = 0.87 [269.8/310.2] × individual C _max). The C _max and AUC(0,t _last) values obtained from these correction methods were compared with those after administration of udenafil 100 mg to healthy subjects. The geometric mean ratios (GMR) and the corresponding 90% CI for each HI patient group versus the healthy subject group were calculated. Statistical comparisons of PK parameters were conducted using WinNonlin (version 6.3.0, Pharsight Corporation, CA, USA).

Results

Pharmacokinetic results of clinical trial

The demographic and clinical data of the study participants are summarized in Table 1. The GMR (95% CI) of AUC(0,t _last) of udenafil were 1.05 (0.67, 1.64) and 1.49 (0.93, 2.38) in subjects with mild and moderate HI, respectively, compared with the values for healthy subjects. In addition, the GMR (95% CI) of C _max of udenafil were 0.69 (0.39, 1.24) and 0.97 (0.53, 1.78) in patients with mild and moderate HI, respectively, compared with those in healthy subjects. The metabolic ratios (AUC_DA‐8164 : AUC_udenafil, mean ± SD) were 1.01 ± 0.58, 0.69 ± 0.23, and 0.58 ± 0.11 in healthy subjects and mild and moderate HI patients, respectively. The impairment in the metabolic pathway of udenafil might have reduced the total clearance, leading to increased AUC(0,t _last) of udenafil in patients with HI. The severity of HI inhibiting the metabolic pathway seemed to contribute to higher AUC(0,t _last) of udenafil and a lower metabolic ratio. There was no significant association between HI and C _max of udenafil.

Table 1.

Demographic and clinical data of the study participants

	Healthy subjects (n = 6)	Patients with mild hepatic impairment (n = 6)	Patients with moderate hepatic impairment (n = 6)	Total
Age (years)	50.7 ± 7.63	52.7 ± 4.97	55.3 ± 5.96	52.9 ± 6.22
Height (cm)	171.8 ± 6.77	169.8 ± 6.97	168.3 ± 5.43	170.0 ± 6.21
Weight (kg)	65.7 ± 4.79	68.0 ± 7.90	67.0 ± 6.43	66.9 ± 6.18
CLcr (ml min–1) *	99.5 ± 17.8	99.5 ± 22.9	91.0 ± 17.1	96.7 ± 18.7
Albumin (g dl–1)	4.35 ± 0.27	3.95 ± 0.34	3.37 ± 0.29	3.89 ± 0.50§
PT †	0.97 ± 0.039	1.13 ± 0.13	1.33 ± 0.094	1.14 ± 0.18§
Total bilirubin (mg dl–1)	0.867 ± 0.44	1.20 ± 0.59	2.18 ± 0.48	1.42 ± 0.75§
Ascites ‡	0	0	2	2
Child–Pugh classification	−	A	B	−

^* Calculated using the Cockcroft–Gault equation.

^† Promthrombin time expressed as international normalized ratio.

^‡ Clinical diagnosis (accumulation of fluid in the peritoneal cavity).

^§ Each measure was compared among the study groups; P value <0.05 in the Kruskal–Wallis test

Final pharmacokinetic model

A two compartment model with first order absorption and elimination best described udenafil kinetics, with first order metabolism to DA‐8164. The plasma concentration–time data for DA‐8164 was also described by a two compartment model (Figure 1). A proportional error model was shown to be sufficient to explain the residual variability for both udenafil and DA‐8164. The final parameter estimates, the corresponding relative standard errors and the results of bootstrap including median and coefficient variation are presented in Table 2.

Figure 1.

Pharmacokinetic model structure of udenafil. It was assumed that the central and peripheral volumes of distribution for udenafil (V _p/F and V _p2/F) are equal to the central and peripheral volumes of distribution for DA‐8164, respectively. ALAG absorption lag time; CL_m/F·f _m clearance of DA‐8164; CL_p/F clearance of udenafil except the metabolic pathway to DA‐8164; CL_pm/F metabolic clearance of udenafil to DA‐8164; k _a absorption rate constant; M udenafil metabolite (DA‐8164); P parent drug (udenafil); Q_m/F·f _m inter‐compartmental clearance of DA‐8164; Q_p/F inter‐compartmental clearance of udenafil

Table 2.

Final parameter estimates and bootstrap results

Parameter	Description (units)	Estimate	RSE (%)	Bootstrap median (95% CI)	CV (%)
Structural model
CLpm/F = θ1 × (PT/1.13) ^ θ10
θ1	Metabolic clearance of udenafil to DA‐8164 (l h–1)	35.7	25.3	34.1 (22.0, 44.1)	18.3
θ10	Exponent of PT on CLpm/F (normalized by 1.13)	−1.65	33.6	−1.66 (−3.43, –0.729)	−37.5
CLp/F	Clearance of udenafil except the metabolic pathway to DA‐8164 (l h–1)	3.62	19.4	3.94 (2.56, 5.43)	20.3
CLm/F·fm	Clearance of DA‐8164 (l h–1)	36.5	26.2	34.5 (21.9, 48.3)	19.6
ka	Absorption rate constant (h−1)	0.326	19.4	0.333 (0.192, 0.477)	22.8
Vp/F	Central volume of distribution of udenafil and DA‐8164 (l)	44.1	29.0	43.0 (22.8, 65.7)	25.8
Vp2/F	Peripheral volume of distribution of udenafil and DA‐8164 (l)	588	7.21	588 (500, 665)	7.3
Qp/F	Inter‐compartmental clearance of udenafil (l h–1)	61.7	15.7	61.1 (40.3, 102)	26.9
Qm/F·fm	Inter‐compartmental clearance of DA‐8164 (l h–1)	11.4	25.4	11.0 (7.08, 17.1)	23.6
ALAG	Absorption lag time (h)	0.314	14.8	0.311 (0.128, 0.369)	21.3
Inter‐individual variability
ωCLpm/F	Inter‐individual variability of CLpm/F (%)	35.1	24.8	32.6 (17.6, 49.7)	25.4
ωCLm/F·fm	Inter‐individual variability of CLm/F·f m (%)	54.4	23.8	53.8 (31.5, 79.5)	23.0
ωka	Inter‐individual variability of k a (%)	64.1	25.9	56.1 (28.8, 81.7)	25.0
ωVp/F	Inter‐individual variability of V p/F (%)	75.6	25.7	66.3 (35.1, 121)	32.2
ωVp2/F	Inter‐individual variability of V p2/F (%)	23.8	21.1	27.1 (14.1, 37.3)	23.4
ωQp/F	Inter‐individual variability of Qp/F (%)	51.3	34.6	47.5 (0.3, 110)	53.6
ωALAG	Inter‐individual variability of ALAG (%)	20.7	65.4	22.4 (0.2, 119)	87.1
Residual error
σprop,p	Proportional error of plasma udenafil (%)	21.6	16.2	21.0 (14.3, 28.6)	17.5
σprop,m	Proportional error of plasma DA‐8164 (%)	23.1	8.18	22.7 (19.2, 27.0)	8.6

CI confidence interval; CV coefficient of variation; PT prothrombin time expressed as international normalized ratio; RSE relative standard error

In the covariate analysis, PT on the CL_pm/F was the only covariate incorporated into the final PK model. The inclusion of PT on the CL_pm/ F decreased the OFV by 9.05 and improved the model fit. The equation for the relationship between PT and CL_pm/F is as follows:

$${\mathrm{CL}}_{\mathrm{pm}}/\mathrm{F}=35.7\times \left(\mathrm{PT}/1.13\right)^\left(–1.65\right)$$ (9)

The exponent of PT normalized to the median value (1.13) was −1.65, which indicates a decrease in the CL_pm/F with increase in PT. For patients with moderate HI (mean value of PT = 1.33), the typical CL_pm/F value was expected to be 27.3 l h^–1, which was smaller than that (45.9 l h^–1) of healthy subjects (mean value of PT = 0.97).

The basic goodness‐of‐fit plots for the final PK model are shown in Figure 2. The medians and 95% CI obtained from the bootstrap procedure are summarized in Table 2. The median values were highly consistent with the estimates of the final model. Visual predictive checks (VPC) stratified by the study group for the final population PK model of udenafil are shown in Figure 3. In addition, VPCs of DA‐8164 are presented in the supplementary material. In the VPCs, the majority of the data was evenly distributed around the prediction median. Furthermore, the model prediction intervals, as well as the medians, corresponded adequately to the observed data. These results confirmed the suitability of the model for demonstrating the effects of HI on udenafil PK and validity of the assumption that V _d,DA‐8164 and V _d,udenafil were equal.

Figure 2.

Basic goodness‐of‐fit plots for the final pharmacokinetic model of udenafil. Black line, line of identity; grey line, locally weighted regression smooth line. CWRES conditional weighted residuals; IWRES individual weighted residuals

Figure 3.

Visual predictive check of the final pharmacokinetic model. (A) healthy subjects, (B) patients with mild hepatic impairment, (C), patients with moderate hepatic impairment (― predictive median, ⋯ 90% prediction interval) and (D) predictive medians classified by study group ( healthy subjects, mild HI patients, moderate HI patients)

Statistical analysis for dose recommendation in patients with hepatic impairment

The comparison results of C _max and AUC(0,t _last) between patients with HI and healthy subjects are presented in Table 3. When udenafil 100 mg was administered to patients with mild HI, the GMRs (95% CI) for the AUC(0,t _last) and C _max were 1.21 (1.10, 1.32) and 1.13 (1.01, 1.25), respectively, compared with udenafil 100 mg in healthy subjects. For patients with moderate HI, GMRs (95% CI) for the AUC(0,t _last) after 100 mg and 75 mg udenafil administration were 1.55 (1.43, 1.67) and 1.02 (0.92, 1.12), respectively. The exposure to udenafil after administration of udenafil 100 mg in healthy subjects was similar with that after 100 mg udenafil administration in patients with mild HI and that after 75 mg udenafil administration in patients with moderate HI (Table 3).

Table 3.

Prediction results for pharmacokinetic parameters using the simulated data

PK parameter	Patients with hepatic impairment	Dose (mg)	Udenafil
			GMR	95% CI
AUC(0,tlast)	Mild	100	1.21	1.10, 1.32
		75	0.74	0.67, 0.81
	Moderate	100	1.55	1.43, 1.67
		75	1.02	0.92, 1.12
Cmax	Mild	100	1.13	1.01, 1.25
		75	0.74	0.66, 0.82
	Moderate	100	1.21	1.03, 1.39
		75	0.83	0.74, 0.92

AUC(0,t _last), area under the plasma concentration–time curve of the last quantifiable concentration; CI confidence interval; C _max maximum plasma concentration; GMR ratio of geometric mean parameters for patients with hepatic impairment relative to healthy subjects; PK pharmacokinetic

Discussion

This study was designed to evaluate the PK differences in patients with mild and moderate HI relative to the PK in healthy subjects. Based on the GMR of udenafil exposure among the groups, the impact of HI on drug exposure was significant in patients with moderate HI but not in those with mild HI. However, it could not be concluded, according to the FDA guidelines 4, that there was no effect of mild HI. The 90% CI for AUC(0,t _last) of udenafil was 72–151%, which was not within the 80–125% limit. This was probably because of the small sample size. In addition, there was no obvious relation between HI and C _max in the impaired patients. This might be because of relatively large inter‐individual variability in either the absorption rate or volume of distribution derived from a single dose study with a small number of subjects. The sample size in this study was six in each group, smaller than that specified by the FDA guidelines (eight subjects in each group). The number of subjects in this trial was not enough to yield conclusive results, but it was judged sufficient to explore the safety and to detect a trend in the effect of HI on drug exposure.

To overcome this limitation, modelling and simulation approaches were employed. The population PK model was developed using data from healthy subjects as well as patients with mild and moderate HI. The goodness‐of‐fit plot (Figure 2) indicated that the model clearly explained the PK of udenafil and DA‐8164. According to the simulated data obtained after single administration of 100 mg udenafil, the predictive results presented the effect of HI on udenafil exposure (Table 3). Values of the GMR of AUC(0,t _last) and C _max of udenafil increased and the 95% CI range became narrower. In other words, the trend of HI effect on AUC(0,t _last) and C _max was more obvious compared with that obtained from the clinical trial.

Serum albumin, total bilirubin and PT were employed to the Child–Pugh score as clinical measures of liver disease. The results of the clinical trial showed strong correlations among three measures, i.e. albumin vs. PT, PT vs. total bilirubin and total bilirubin vs. albumin. Thus, we applied these three measures one by one as covariates in the covariate analysis and selected PT as a covariate to show best improvement in the model fit based on decrease in OFV.

In the covariate analysis, the significant differences in serum albumin concentration among the groups could not describe sensitively the effect of HI on PK of udenafil. It is reported that udenafil binds plasma proteins with high affinity and has a low extraction ratio 12. This corresponded with the results of an earlier study, which reported that total drug exposure is independent of protein binding for all drugs with low extraction ratio 13. Bilirubin would be an insensitive marker for drug elimination capacity because its abnormality is related to extensive and complex processes in the intra‐hepatic and extra‐hepatic lesions 14. This might explain why the contribution of total bilirubin to the improvement of the model was less than that of PT even though it showed significant differences among the groups in our study. The present study showed that PT, known as a useful marker of liver synthetic function 15, was relatively sensitive to the effect of HI on metabolic clearance of udenafil to DA‐8164. However, hepatic dysfunction affected the drug absorption, distribution, metabolism and excretion. Thus, PT is not useful as a simple marker for hepatic dysfunction and cannot be used to quantitate the effect of HI on drug exposure.

For dose optimization, it is desirable to establish an exposure–response relationship. However, until date, an exposure–response relationship has not been established clearly for PDE‐5 inhibitors including udenafil. In this study, we recommended an optimal dose of udenafil for HI patients on the assumption of a positive drug exposure–response relationship. Accordingly, pharmacological activity could be proportional to the sum of each AUC multiplied by an unbound fraction and a potency for the inhibition of PDE‐5 of udenafil or DA‐8164. In this process, the exposure to udenafil was used only as drug exposure, although the DA‐8164 data were useful in the development of a population pharmacokinetic model. The unbound fraction of DA‐8164 was one‐fifth of that of udenafil 9 and the metabolic ratios (AUC_DA‐8164 : AUC_udenafil) were between 0.6–0.7 in mild and moderate HI patients and 1.0 in healthy subjects. In addition, the in vitro potency of DA‐8164 for the inhibition of PDE‐5 was half of that of udenafil 8. Thus, the pharmacological activity of DA‐8164 might be less than 1/10–1/20 of udenafil and the effect of DA‐8164 on the pharmacological activity might be marginal. Consequently, the dose optimization was conducted based on udenafil exposure.

To determine the optimal dose with comparable exposure to udenafil among the groups, a dose–exposure relationship also needed to be established. Udenafil is known to exhibit dose non‐proportionality and, thus, the linear model for C _max and AUC(0,t _last) was developed using data from a previous single ascending dose study 11. The optimal dose was estimated based on drug exposure (AUC(0,t _last) and C _max) obtained from simulation using the parameter estimates of the final PK model and linear model for correcting the dose proportionality.

There are three PDE‐5 inhibitors approved by the FDA: sildenafil, vardenafil and tadalafil. The effects of mild and moderate HI on the PK of each drug have been evaluated in clinical trials 2. The recommended doses for sildenafil and vardenafil were half of the commonly used starting dose for patients with moderate HI, based on increases in AUC and C _max values 16. For tadalafil, an effect of HI on PK was not observed and it does not need adjustment for starting dose for mild and moderate HI patients 17. A previous study evaluated the efficacy and safety of udenafil in 110 patients with ED after multiple administrations of either 100 mg or 200 mg of udenafil for 12 weeks 18. After 12 weeks of treatment, the udenafil treatment group showed significant improvements in efficacy compared with the placebo group 18. The 200 mg udenafil group showed greater trends for improvements across all the efficacy variables than did the 100 mg udenafil group. However, they were not statistically significant. 18. The previous study demonstrated a favourable safety profile in the 100 mg and 200 mg udenafil treatment groups 18. In addition, udenafil was reported to be safe and well tolerated in dosages ranging from 100 mg to 300 mg 11. The values of AUC and C _max reported in the previous study 11 were much higher than those in the present clinical trial. These results indicate that udenafil would show a comparable efficacy and safety profile in the recommended dosages of udenafil, 100 mg and 75 mg for patients with mild and moderate HI, respectively.

We have demonstrated that population PK modelling and simulation could be an effective strategy to overcome the limitation of clinical trials in a special population with a small number of subjects and to integrate the data from several clinical trials. This could lead to disease stratification and optimal dose recommendation. The present study suggests that udenafil 100 mg can be an effective and well‐tolerated therapy for patients with mild HI and udenafil 75 mg can be recommended as the initial dosing for patients with moderate HI.

Competing Interests

All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare that this study was sponsored by Dong‐A ST Co., Ltd., Seoul, Korea. MYB is an employee of Dong‐A ST Co. Ltd, Seoul, Korea. AK, JL, DS, YJJ, J‐YC and IJJ have had no financial relationships with any organizations that might have an interest in the submitted work in the past 3 years.

The authors would like to thank the subjects and staff who participated in the clinical trial and Hyo‐Suk Lee, MD, PhD for his valuable contribution as the principal investigator in planning and completing this clinical trial. We would also like to thank Sook‐Hyang Jeong, MD, PhD (Department of Internal Medicine, Seoul National University Bundang Hospital, Seoul National University College of Medicine, Seongnam, Korea) and Young‐Suk Lim, MD, PhD (Department of Internal Medicine, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea) for participating in the multicentre clinical trial as co‐investigators.

Contributors

AK conducted the clinical study with healthy subjects, participated in data analysis and interpretation and wrote the clinical study report. JL performed the population pharmacokinetic modelling and simulation. AK and JL contributed equally while writing the manuscript. DS helped in the study planning and conducted the clinical study. YJJ was a co‐investigator responsible for conducting the clinical study with patients and reviewed the final clinical study report. MYB was the clinical trial manager for Dong‐A ST Co. Ltd., Seoul, Korea. IJJ was a co‐investigator in the clinical study using healthy subjects and was responsible for the overall study design, approval of the clinical study report and final data interpretation. J‐YC was the bioanalytical laboratory head and was responsible for sample analysis. IJJ is the corresponding author for this manuscript, because the manuscript focused on pharmacokinetic analysis, interpretation, and modelling and simulation. All authors have reviewed and approved the manuscript.

Supporting information

Figure S1 Linear model for adjusting the dose proportionality of udenafil

Figure S2 Visual predictive check of the final pharmacokinetic model: DA‐8164 (A) and all subjects (B). Predictive medians classified by study group

Supporting info item

Kim, A. , Lee, J. , Shin, D. , Jung, Y. J. , Bahng, M. Y. , Cho, J. ‐Y. , and Jang, I. ‐J. (2016) Population pharmacokinetic analysis to recommend the optimal dose of udenafil in patients with mild and moderate hepatic impairment. Br J Clin Pharmacol, 82: 389–398. doi: 10.1111/bcp.12977.