Integrative pharmacokinetic–pharmacodynamic modeling and simulation of amenamevir (ASP2151) for treatment of recurrent genital herpes

Integrative pharmacokinetic–pharmacodynamic modeling and simulation of amenamevir (ASP2151) for treatment of recurrent genital herpes

Accepted Manuscript Integrative Pharmacokinetic-Pharmacodynamic Modeling and Simulation of Amenamevir (ASP2151) for Treatment of Recurrent Genital Her...

2MB Sizes 0 Downloads 16 Views

Accepted Manuscript Integrative Pharmacokinetic-Pharmacodynamic Modeling and Simulation of Amenamevir (ASP2151) for Treatment of Recurrent Genital Herpes Akitsugu Takada, Masataka Katashima, Atsunori Kaibara, Koji Chono, Kiyomitsu Katsumata, Taiji Sawamoto, Hiroshi Suzuki, Yoshitaka Yano PII:

S1347-4367(16)30029-5

DOI:

10.1016/j.dmpk.2016.05.005

Reference:

DMPK 115

To appear in:

Drug Metabolism and Pharmacokinetics

Received Date: 9 November 2015 Revised Date:

20 May 2016

Accepted Date: 27 May 2016

Please cite this article as: Takada A, Katashima M, Kaibara A, Chono K, Katsumata K, Sawamoto T, Suzuki H, Yano Y, Integrative Pharmacokinetic-Pharmacodynamic Modeling and Simulation of Amenamevir (ASP2151) for Treatment of Recurrent Genital Herpes, Drug Metabolism and Pharmacokinetics (2016), doi: 10.1016/j.dmpk.2016.05.005. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT 1

Integrative

Pharmacokinetic-Pharmacodynamic

Modeling

and

2

Amenamevir (ASP2151) for Treatment of Recurrent Genital Herpes

Simulation

of

3 Akitsugu Takada1, Masataka Katashima1, Atsunori Kaibara1, Koji Chono1, Kiyomitsu

5

Katsumata1, Taiji Sawamoto1, Hiroshi Suzuki1, Yoshitaka Yano2

6

RI PT

4

1

Astellas Pharma Inc., 2-5-1 Nihonbashi-Honcho, Chuo-ku, Tokyo, Japan

8

2

Kyoto Pharmaceutical University, 5 Misasagi-Nakauchicho, Yamashina-ku, Kyoto, Japan

SC

7

9 Corresponding author’s information

M AN U

10

Akitsugu Takada

12

Astellas Pharma Inc.

13

2-5-1 Nihonbashi-HonchoChuo-ku, Tokyo, 103-8411, Japan

14

E-mail: [email protected]

15

Tel: +81-3-3244-2579, Fax: +81-3-3243-5732

EP AC C

16

TE D

11

1

ACCEPTED MANUSCRIPT Abstract

18

Amenamevir is a novel drug that targets the viral helicase-primase complex. While

19

dose-dependent efficacy had been observed in non-clinical studies, no clear dose dependence

20

has been observed in humans. We therefore developed a pharmacokinetic/pharmacodynamic

21

(PK/PD) model to explain this inconsistency between species and to clarify the

22

immune-related healing of amenamevir in humans. The model consisted of a non-linear

23

kinetic model for a virtual number of virus plaques as a built-in biomarker. Lesion score was

24

defined as an endpoint of antiviral efficacy, and logit model analysis was applied to the

25

ordered-categorical lesion score. The modeling results suggested the time course profiles of

26

lesion score could be explained with the efficacy terms in the logit model, using change in

27

number of virus plaques as an indicator of the effects of amenamevir and time elapsed as an

28

indicator of the healing of the immune response. In humans, the PD effect was almost

29

dose-independent, and immune-related healing may have been the driving force behind the

30

reduction in lesion scores. Drug efficacy is occasionally masked in diseases healed by the

31

immune response, such as genital herpes. The PK/PD model proposed in the present study

32

must be useful for explanation the PK/PD relationship of such drugs.

33

EP

TE D

M AN U

SC

RI PT

17

Keywords

35

amenamevir, pharmacokinetic/pharmacodynamic, pharmacometrics, genital herpes, modeling

36

and simulation

37

AC C

34

2

ACCEPTED MANUSCRIPT 38

Introduction Recurrence of genital herpes is usually caused by HSV-2 and less commonly by HSV-1

40

[1]. Almost 50 million American adults and adolescents (20% of the total population) are

41

infected with genital herpes: it is one of the most common sexually transmitted diseases [2].

42

The recurrence of genital symptoms is very typical, with up to 80% of HSV-2-infected

43

patients experiencing recurrent outbreaks within the first 12 months following the first

44

episode.

RI PT

39

Amenamevir is a novel drug that targets the viral helicase-primase complex. This

46

protein complex is essential for herpes simplex virus (HSV) DNA synthesis, making

47

helicase-primase inhibitors such as amenamevir a potential treatment option for genital

48

herpes. A pharmacokinetic / pharmacodynamic (PK/PD) approach for anti-virus drug is

49

similar to the development of antibiotics, and in vitro IC50 is used to predict the clinical dose

50

in vivo animal model [3]. Amenamevir has demonstrated higher potency in vitro antiviral

51

activity against HSV-1 and HSV-2 than aciclovir [4].

TE D

M AN U

SC

45

A phase 2 study (ASP2151 CL-101 study) was performed in the United States to compare

53

the safety and efficacy of four different dose regimens of amenamevir with valaciclovir and

54

placebo in parallel in the acute treatment of recurrent genital HSV infection. At the first sign

55

of recurrence, patients self-initiated treatments with amenamevir 100, 200, or 400 mg daily

56

for 3 days, 1200 mg as a single dose; placebo for 3 days; or valaciclovir 500 mg twice daily

57

for 3 days. Dose and regimen were determined based on those results and PK profile of

58

amenamevir to maintain time above 200 ng/ml of plasma amenamevir concentration, T200 [5].

59

Of 695 patients randomized to the treatments, data from 437 patients with recurrent infection

60

were analyzed. The primary efficacy endpoint was time to lesion healing which was shorter

61

in all amenamevir groups and the valaciclovir group compared with placebo, but a

AC C

EP

52

3

ACCEPTED MANUSCRIPT 62

dose-response relationship was not evident. All groups except amenamevir 100mg showed a

63

higher proportion of aborted lesions compared with placebo[6]. Population PK analysis of amenamevir was conducted to estimate patients’ PK profiles

65

using 957 plasma samples in 273 genital herpes subjects in CL-101study. The plasma

66

concentration-time courses for amenamevir were described using a one-compartment model

67

with first-order absorption. Exploratory PK and PD analysis was conducted to find an

68

appropriate PK-related parameter which correlates with PD instead of finding dose-PD

69

relationship, using individual PK parameters derived from the final model by the population

70

analysis. The variable T200 was found to correlate with time to lesion healing and viral

71

shedding, which is consistent with the in vivo results [7]. However, these correlations were

72

not statistically significant, because T200 was dose dependent parameter after all.

M AN U

SC

RI PT

64

In CL-101 study[6], dose-dependent efficacy was not evident in the primary analysis

74

using time to lesion healing, although dose-dependent efficacy had previously been suggested

75

in guinea pigs [8]. To discuss address this inconsistency between clinical and non-clinical

76

results, we conducted additional PK/PD modeling and simulations using the same component

77

model to for both humans and guinea pigs have conducted in the present study with a built-in

78

biomarker—number of virus plaques, which explains the time-dependent antivirus efficacy of

79

amenamevir—to combine the PK of amenamevir and lesion scores as an efficacy endpoint.

80

Furthermore, non-drug related such as immune response was added to explain the difference.

AC C

EP

TE D

73

81

In general, screening process of new drugs is conducted by using animal to confirm the

82

efficacy, safety and pharmacokinetic. After that, these endpoints are confirmed in clinical

83

trials. Amenamevir showed effective anti-virus action dose-dependently in guinea pigs but

84

not in humans as stated above. To fully understand the phenomenon is somewhat challenging

85

theme, although many drug candidate compounds may have same theme in drug development

86

stage. In this report, we tried to fill a gap between non-clinical and clinical results with the

4

ACCEPTED MANUSCRIPT 87

same PK/PD model both animal and human, which is bi-directional translational modeling

88

and simulation approach.

90

Materials and methods

91

Analysis data from clinical and non-clinical studies

92

RI PT

89

The present study is a population PK/PD analysis using the obtained PK and PD data in the previous clinical and non-clinical studies [4,8,6]. The scheme of the PK/PD model

94

applied in this study is shown in Figure 2. As mentioned in Introduction, the results of the

95

previous virus plaque study provided the possible relationship of viral dynamics with

96

amenamevir concentration [5], and efficacy such as suppression of lesion scores was

97

expected to correlate with amenamevir concentration. The present PK/PD model assumes that

98

the time time-dependent increase in number of virus plaques is reduced by amenamevir, and

99

the increase of lesion score is dependent on the virtual number of virus plaques. Data from

M AN U

SC

93

previous studies were reanalyzed in the present study as follows:

101

1) Virus plaque assay data were used for constructing virus plaque PD model component.

102

2) Plasma concentration data of amenamevir in guinea pigs and humans (reference) were

EP

103

TE D

100

used for constructing the PK model component. 3) Lesion score data in guinea pigs and humans were used for logit model analysis.

105

The experimental conditions for the previous studies will be described briefly in subsequent

106

sections.

AC C

104

107 108

Antiviral compound and sample measurement

109

Amenamevir (molecular weight, 482.55; international non-proprietary name, amenamevir)

110

was synthesized by Astellas Pharma Inc. (Tokyo, Japan). Sample measurement was

5

ACCEPTED MANUSCRIPT conducted using a validated liquid chromatography-tandem mass spectrometry (LC-MS/MS)

112

method based on FDA validation guidance at Covance Laboratories, Ltd[7]. The lower limit

113

of quantification (LLOQ) of this assay for amenamevir was 5 ng/mL when 0.1 mL plasma

114

was used. The LLOQ data were treated as zero and not used for modeling

115

Viruses and cell lines

116

HSV strains clinically isolated in the US were kindly provided by Dr. Nancy Sawtell

117

(Cincinnati Children's Hospital Medical Center, Cincinnati, OH, USA). Other viruses and cell

118

lines were provided by Rational Drug Design Laboratories (Fukushima, Japan). Human

119

embryonic fibroblast (HEF) cells and Vero cells were grown in Eagle’s minimum essential

120

medium supplemented with 10% fetal bovine serum (FBS), 100 units/mL penicillin G, and

121

100 µg/mL streptomycin (Invitrogen, Carlsbad, CA, USA). HSV-1 and HSV-2 were

122

propagated using HEF cells in maintenance medium containing 2% FBS.

123 Virus plaque reduction assay

TE D

124

M AN U

SC

RI PT

111

The antiviral activities of amenamevir against HSVs were tested using a plaque reduction

126

assay, as described previously [4,5] to determine the concentration and time dependent

127

effects of amenamevir. Briefly, HEF cells were seeded into multiwell plates and incubated

128

until they formed a monolayer. After the medium was removed, the cells were infected with

129

HSV-1, and the plates were further incubated for 1 h at 37°C. The cells were washed twice

130

with maintenance medium and then treated with the test compound until clear plaques

131

appeared. The cells were then fixed with 10% formalin in phosphate-buffered saline, stained

132

with a 0.02% crystal violet solution, and the number of plaques was determined under a light

133

microscope. Concentrations of amenamevir were 0.01, 0.03, 0.1, 0.3, 1, 3, 10 and 30 μM.

134

Duration times of the incubation were 6, 8 and 24 h. These data were included in the present

AC C

EP

125

6

ACCEPTED MANUSCRIPT 135

study to simulate the virtual time-course profiles of virus plaques in guinea pigs and humans

136

in combination with their PK data.

137 138

PK study in guinea pigs All animal experimental procedures were approved by the Animal Ethical Committee of

140

Yamanouchi Pharmaceutical Co., Ltd. (currently known as Astellas Pharma Inc.). Female

141

Hartley guinea pigs were purchased from Charles River Laboratories (Kanagawa, Japan).

142

Amenamevir at doses of 0.3, 1.0, 3.0 mg/kg were administered as methylcellulose suspension

143

via oral to guinea pig (n = 3 for each group, 4 weeks of age). Plasma samples were obtained

144

at 0.25, 0.5, 1, 2, 4, 8, 12 and 24 h after dosing.

M AN U

145 146

SC

RI PT

139

PK study in humans

In the present study, the estimated population pharmacokinetic parameters for

148

amenamevir concentration data, which were collected at the time for screening and at the

149

clinical visits once daily through Days 1 to 4 in study CL-101[6], were used.

151

Lesion score measurement in guinea pigs

EP

150

TE D

147

Antiviral activity of ASP2151 against HSVs in guinea pigs was tested by the evaluation

153

of lesion score as described previously [8]. Briefly, Female guinea pigs (Hartley, aged 4

154

weeks at the time of viral infection) were intravaginally infected (designated as Day 0

155

post-infection) with a cotton swab saturated with PBS containing HSV-2 strain G, as

156

described previously [9]. For HSV-2 strain G, the virus pool contained 1.25 × 105 pfU/mL

157

and caused lesions in nearly 100% of control animals. Amenamevir at doses of 0 (placebo), 1,

158

3, 10, or 30 mg/kg was orally administered twice daily for 5 days starting 3 h after viral

159

inoculation as a prophylactic treatment, or 4 days after viral inoculation as a therapeutic

AC C

152

7

ACCEPTED MANUSCRIPT 160

treatment (n = 10 for each group). The disease profile was monitored daily for 21 days and

161

was scored on a 0-6 composite scale based on the severity of vaginitis and neurological

162

symptoms according to the following criteria: Score 0: no signs of infection

164

Score 1: localized, barely perceptible small vesicles

165

Score 2: small or large vesicles involving 10% to 50% of the area

166

Score 3: small or large vesicles involving 50% to 100% of the area

167

Score 4: small ulcers involving 10% to 50% of the area

168

Score 5: severe ulcers involving 50% to 100% of the area

169

Score 6: hind limb paralysis or death

170

The scores of genital herpes in guinea pigs were determined daily. In the PK/PD analysis,

171

the scores on Day 3 were used as the baseline scores to make the modeling easier because of

172

no lesion symptoms until Day 3 after virus allocation. We assumed that disease did not alter

173

the PK of amenamevir.

174

Lesion score measurement in humans

TE D

M AN U

SC

RI PT

163

The data for times to healing of all lesions were obtained as the primary efficacy endpoint

176

of Study CL-101 [6]. Genital herpes recurrence was defined as herpes recurrence below the

177

umbilicus and above the knees. Times to healing (h) of all lesions were determined by the

178

study investigator as the times from therapy initiation to re-epithelialization of all lesions,

179

excluding aborted lesions. Aborted lesions were defined by the presence of prodromal

180

symptoms including pain, tingling, itching, and burning, but lesions failed to develop beyond

181

the macule/papule stage to the vesicular/ulcerative stage. Healed lesions were defined as the

182

absence of crusts, depressions, erosions or ulcerations. Residual erythema in absence of the

183

preceding is defined as healed. Existing symptom lesions were classified by the site clinician

184

during the genital examinations on Days 1 - 6, and also on Days 8 and 10 only if lesions had

AC C

EP

175

8

ACCEPTED MANUSCRIPT not healed by Day 6 and Day 8, respectively. If a patient had existing symptom lesions at the

186

visit on Day 17, an additional genital examination for lesion classification was performed.

187

The classification of the existing symptom lesions were categorized in the similar way for

188

guinea pig as follows for a purpose of logit analysis in this study: Score 0: healed lesion

190

Score 1: crust

191

Score 2: vesicle / pustule / ulcer

192

Score 3: macule / papule

SC

189

RI PT

185

The changes of the scores are unidirectional in the order of 0, 3, 2, 1 and 0. Patients

194

whose scores were maintained to be 0 during the study period were treated as aborted lesions.

195

Mean time course profiles of observed lesion scores are shown in Fig. 1a for guinea pigs and

196

Fig. 1b for humans. The present study was conducted in accordance with the ethical

197

principles stated in the Declaration of Helsinki, Good Clinical Practice, International

198

Conference on Harmonization guidelines, and applicable laws and regulations.

TE D

M AN U

193

199

201

Analysis models and simulation data

The analysis conditions in this study are described briefly in this section.

EP

200

Population modeling

203

Logit model analysis was performed via the non-linear mixed effects model [10] using

204

NONMEM Version VI Level 1.0 (Icon Clinical Research, North Wales, PA, USA).

205

Graphical processing of the NONMEM output was performed with SAS Version 8.2, Release

206

8.02 (SAS Institute Inc., USA). Model selection was based on the goodness-of-fit criteria

207

(log-likelihood difference (l.l.d.) calculated as a difference of objective functions) with visual

208

inspection of the diagnosis plots. A value of l.l.d. more than 3.84 between two models with a

209

1 degree of freedom difference was considered to be significant (p<0.05). First-order

AC C

202

9

ACCEPTED MANUSCRIPT 210

conditional estimation with interaction (FOCE-INTERACTION) [11] was used in

211

NONMEM execution.

212 Model validation

214

Visual predictive check [12] was performed using the final models for guinea pig PK and

215

virus plaque profiles, where 1,000 hypothetical amenamevir concentrations and the numbers

216

of virus plaques were simulated at each time point. The median and 95% prediction interval

217

were constructed from the resulting predictions at each time point. Plots overlaying the

218

medians, prediction interval with observed amenamevir concentrations or observed the

219

numbers of virus plaques were created.

M AN U

220

SC

RI PT

213

PK model in guinea pigs

222

We used a linear single-compartment model with an absorption lag time and first-order

223

absorption as a guinea pig PK model by visual inspection of the observed data. The

224

pharmacokinetic parameters of ka (absorption rate constant), V/F (apparent distribution

225

volume) and CL/F (apparent total clearance) were defined, where F is the oral bioavailability

226

fraction. Mixed-effect modeling was applied using NONMEM, where log-normal distribution

227

for the inter-individual variability was assumed for each pharmacokinetic parameter, in

228

Equation 1:

AC C

EP

TE D

221

229

Pj = θ j ⋅ exp(ηj)

230

where Pj is a parameter of interest in jth subject, θj is the population mean of the

231

corresponding parameter, ηj is the random variable which gives the inter-individual

232

variability of Pj from θj; ηj was assumed to be normally distributed with a mean 0 and a

233

variance ω2 . Residual error was assumed to be described by a proportional error model

234

given by Equation. 2;

Eq. 1

10

ACCEPTED MANUSCRIPT 235

Yij = Cij (1 + εij)

Eq. 2

236

where Yij is the ith observed concentration in jth subject, Cij is the predicted concentration

237

by the model, and εij was the random variable for residual error which is assumed to be

238

normally distributed with a mean 0 and a variance σ 2 .

RI PT

239 PK model in humans

241

PK model analysis of amenamevir in patients with genital herpes had already been performed

242

using a single-compartment model with first-order absorption[7]. As there were some

243

extremely low concentration data points (LCPs) during the absorption phase in the plasma

244

concentration data, another first-order absorption rate constant (ka,LCP) was used in addition to

245

ka to explain the profile of these LCPs. On NONMEM analysis, the estimated mean

246

(and %RSE) values of apparent clearance (CL/F) was found to be 13.8 (4.28%) L/h, apparent

247

distribution volume (V/F) was 143 (4.31%) L, absorption rate constant (ka) was 0.874

248

(15.70%) h-1, and the absorption rate constant for LCP (ka,LCP) was 0.00107 (63.30%) h-1.

249

Relative bioavailability (assuming a bioavailability of 1.0 at 100 mg dose) at a dose of

250

200 mg was 0.982 (6.02%), 400 mg was 0.874 (5.74%), and 1200 mg was 0.706 (6.15%).

M AN U

TE D

EP

251

SC

240

Simulation of PK profiles in guinea pigs and humans

253

Simulations of plasma amenamevir profiles in guinea pigs and humans were conducted using

254

the population mean PK parameters for guinea pigs and the post-hoc PK parameter estimates

255

for humans to simulate time course profiles of the numbers of virus plaques in those species.

256

The data point treated as LCP were not included in the simulation processes in this study

257

because our modeling method using LCP concept was to explain the outliers (i.e. LCP) and

258

obtain a better estimate for Ka using non-LCP data only. The numbers of LCP data point was

259

33 in totally 928 points (3.56% total), and we think the effect of omitting the LCP on

AC C

252

11

ACCEPTED MANUSCRIPT 260

simulation results is negligible. Time-courses of simulated concentration are shown in

261

Figures 5a and 6a, and simulation conditions were as follows: Guinea pig = 0, 1, 3, 10, and

262

30 mg/kg BID for 5 days;

263

Human = 0, 100, 200, and 400 mg QD for 3 days or 1200 mg as a single dose.

RI PT

264 PD model for virus plaque

266

A two-compartment model was applied to explain the concentration and time-dependent

267

changes in the number of virus plaques, as shown in Figure 2. A virus cycle was incorporated

268

by defining the amenamevir effective/non-effective compartments, which are connected by

269

first-order rate constants kinact and kact [13]. Increase of virus plaque was assumed as the

270

first-order rate constant (kin), it was fixed to 0.0569 h-1 (= ln(60) / 72), which gives the

271

number of virus plaques to be 60 pFU at 72 h based on the study setting. A non-linear drug

272

effect was assumed using the Michaelis-Menten form with maximum drug effect (Emax) and a

273

Michaelis constant (EC50).

TE D

M AN U

SC

265

The mass-balance equations for the number of virus plaques are given as follows.

275



 = k × V1 − k  × V2 + k  × V3 −  

Eq. 3

276



= k  × V2 − k  × V3

Eq. 4

277

In Eq. 3-4, V(1) represents the number of virus plaques in a hypothetical input

278

compartment, V(2) is the amenamevir effective compartment, V(3) is the amenamevir

279

ineffective compartment; the initial value of V(1) is 60, while those of V(2) and V(3) are 0.

280

Plasma concentration of amenamevir is given by CP. No inter-individual variability was

281

assumed in this case and an additive error model for residual variability was assumed as

282

given in Equation 5;

283

Yij = Cij + εij







×

AC C



EP

274

Eq. 5

12

ACCEPTED MANUSCRIPT 284

where Yij and Cij are the observed and model predicted numbers of plaques for the ith

285

sample in jth subject, εij is the residual error which is assumed to be normally distributed with

286

a mean 0 and a variance σ 2 .

287 Simulation of virus plaque profile

289

Simulations of virtual numbers of virus plaque profiles in guinea pigs and humans were

290

conducted using the simulated amenamevir concentration and the estimated virus kinetic

291

parameters. Simulation settings were as follows.

SC

RI PT

288

- Initial administration amount of number of virus plaques was set to 60 pFU

293

- Time of virus plaque administration was defined as Day 0 (0 h)

294

- Starting time of amenamevir administration was Day 1 (24 h)

295

Lesion scores in human on Day 1 were regarded as the initial scores, this is because the

296

Phase 2 study (CL-101) was designed as the a self-initiation study, where patients

297

self-initiated treatment and returned to the clinic within 24 h after the initial dose of study

298

drug (Day 1). Time-course of virus plaques is shown in Figures 5b and 6b.

299

TE D

M AN U

292

PK/PD Model

301

Logit model analysis of lesion score

AC C

EP

300

302

Logit model analysis was applied to the ordered-categorical lesion score data. Let Yijk =

303

Y(tijk) be the lesion score at the kth time point tijk, in the ith individual of the jth treatment group,

304

and Yijk takes the values of 0, 1, 2, 3, or 4 in the guinea pig study, and 0, 1, 2, or 3 in the

305

human study. The scores 5 and 6 were not included in the analysis for guinea pig because of

306

no observed data for these scores. A logit model for the probability of having a lesion score

307

Yijk that was equal to or less than a given score yijk = 0, 1,.., n can be expressed as a

308

cumulative distribution function as follows [14,15].

13

ACCEPTED MANUSCRIPT 309 310 311

Fy; θ, η = Pr&Y ≤ y; θ, η) =

*+,-./;0,12

3*+,-./;0,12



Eq. 6

where θ is a vector of parameters and η is a random variable. Based on the cumulative probability function, the probability of having a single lesion score was given by Equation 7: Pr&Y = y) = Pr&Y ≤ y) − Pr&Y ≤ y − 1),

313

where Pr&Y < 0) = 0 and Pr&Y ≤ n) = 1.

314

The logit, g(yijk; θ, η) was defined as follows:

315

gy; θ, η = ∑ :<3 θ: × Q: y + Eff

316

for yijk = n, where Q: ?y @A B = C

317

In guinea pigs, lesion scores increased almost monotonically in the placebo group,

318

however it began to decrease after Day 2 in amenamevir groups and clear dose dependence

319

was observed as shown in Figure 1a. In contrast, lesion scores in humans decreased

320

monotonically over time in both amenamevir and placebo groups, and dose dependence was

321

not clear as shown in Figure 1b. These data suggest that the decrease in lesion scores could

322

not be explained only by the changes of virus plaques reflecting healing of amenamevir

323

(Virus Plaque component) and thus the elapsed time reflecting the healing of the immune

324

response (Time component) was included into the model in this study. For the term of

325

describing drug efficacy (Eff), following model was tested. The term Eff was expressed by

326

the addition of Virus Plaque component and Time component.

RI PT

312

AC C

EP

TE D

M AN U

1, y @A ≥ m − 1 . 0, otherwise

SC

Eq. 7

327

Eff = β3 × Virus Plaque + β × Time+η

328

In these models, number of virus plaques (given by ‘Virus Plaque’) and the elapsed time

329

from virus plaque increase (given by ‘Time’) were incorporated as fixed effects. This model

330

was built theoretically, no covariates step was conducted. Maximum Likelihood Estimation

331

was used for the fitting and posterior prediction, and individual estimated lesion scores for ith

Eq. 8

14

ACCEPTED MANUSCRIPT 332

subject at time t (Lesion scorei,t) were calculated based on the estimated parameters by

333

following equation. Lesion score , = ∑ :
335

Results

336

PK model in guinea pigs

RI PT

334

As shown in Figure. 3, result of VPC shows that most of the observed values were within

338

95% prediction interval. Estimated population mean parameters of the PK model for guinea

339

pig PK are summarized in Table 1.

SC

337

340 PD model for virus plaques

342

Estimated population mean parameters of the PD model for the virus plaque profiles are

343

summarized in Table 1. The observed plaque counts versus amenamevir concentration are

344

plotted with the 95% prediction interval in Figure 4. The estimated EC50 was 127 ng/mL,

345

which is a little smaller than the in vitro expected effective concentration of 200 ng/mL [5].

346

As the result, in vitro anti-virus effect of amenamevir was consistent with the virus kinetic

347

model.

348

Logit model for lesion score

TE D

EP

Final models for the term logit lesion score in guinea pigs and humans are obtained as

AC C

349

M AN U

341

350

follows:

351

Guinea pig:

For Lesion score =

0; Logit = 6.76 + Eff.[ * , . 1; Logit = 6.76 + 3.11 + Eff.[ * , . = 9.87 + Eff.[ * , . 2; Logit = 6.76 + 3.11 + 4.17 + Eff.[ * , . = 14.04 + Eff.[ * , .

15

ACCEPTED MANUSCRIPT 3; Logit = 6.76 + 3.11 + 4.17 + 5.89 + Eff.[ * , . = 19.93 + Eff.[ * , . 352 353

Where Eff.[ * , . = −0.265 × Virus Plaque − 0.0334 × Time

Eq.10

Human:

0; Logit = −5.21 + Eff`[:

1; Logit = −5.21 + 1.86 + Eff`[: = −3.35 + Eff`[:

RI PT

For Lesion score =

2; Logit = −5.21 + 1.86 + 3.34 + Eff`[: = −0.01 + Eff`[:

356

SC

355

Where Eff`[: = −0.0247 × Virus Plaque + 0.0424 × Time

Eq.11

In these models, a beneficial effect means a decrease of lesion scores while a worsening

M AN U

354

effect means an increase of lesion scores.

While the fixed effect in Eff of Virus Plaque was negative in both species, the fixed

358

effect in Eff of the Time component by the immune system was negative in guinea pigs but

359

positive in humans. In guinea pigs, both the Virus Plaque and Time components worsened the

360

lesion score. In humans, Virus Plaque worsened the lesion score, whereas immune system

361

improved. Estimated profiles of mean lesion scores were consistent with the observed values

362

in both species (Fig. 5(c) and Fig. 6(c)). Some simulations were performed to show the

363

probabilities above each lesion score in guinea pig (Fig. 5(d)-(g)) and in human (Fig. 6(d)-(f)),

364

where dose and time dependent profiles are shown. Fitting results are summarized in Table 2,

365

and no 95% confidence intervals (CIs) for any fixed effects included 0, indicating that all

366

parameters were significant.

367

Discussion

AC C

EP

TE D

357

368

In the present study, an empirical PK/PD model, as shown in Figure 2 for the

369

helicase-primase inhibitor in genital herpes patients was developed. In this model, the time

370

course profiles of lesion scores was not directly dependent on the amenamevir PK, but it was

16

ACCEPTED MANUSCRIPT 371

mainly dependent on the virtual number of virus plaques which suggests a time-dependent

372

anti-virus mechanism of amenamevir. The PK analysis of amenamevir concentration in guinea pig suggested liner PK profile as

374

shown in Table 1 and Figure 3. In contrast, the CL/F estimated in our previous study in

375

humans [7] suggested dose-dependent bioavailability as F decreased as dose increased. A

376

possible reason is the difference of the dosage form between the species; amenamevir was

377

administered in methylcellulose solution to guinea pigs and in tablet form to humans.

378

Amenamevir is poorly soluble, which strongly affects its passive diffusivity (data not shown).

379

When an allometric scaling [16] was applied, CL/F in humans was estimated about 0.16

380

L/h/kg, smaller than in guinea pig (2.04 L/h/kg) although the reason of the difference

381

between species was unclear. We focused on the relationship between amenamevir

382

concentrations and virus plaque data in the present study.

M AN U

SC

RI PT

373

We used the virtual number of virus plaques obtained via plaque reduction assay as a

384

marker to explain amenamevir PK/PD mechanism. Goodness of fit of the virus plaque

385

modeling were not enough acceptable especially in the higher concentration range at 6 or 8

386

hrs exposure. Several models to explain this, i.e. some models including not only the time

387

dependent component but also the dose (concentration) dependent component were tested to

388

try to improve the modeling result, however, no clear improvement was obtained(Data not

389

shown). We could not yet find the reason of this discrepancy at the higher concentration, but

390

we conclude that the current model is sufficient to simulate the efficacy of twice a day or

391

daily dosing of amenamevir because some acceptable result was obtained in the condition of

392

24hrs exposure which showed the threshold concentration to maintain the amenamevir

393

efficacy throughout the day (Fig.4c). In this our analysis, the virus plaque time profile in vivo

394

was simulated based on the virus plaque kinetic model, which was built based on in vitro data.

395

Several points remain unclear, however—namely the utility of the same viral kinetic model

AC C

EP

TE D

383

17

ACCEPTED MANUSCRIPT 396

between species, whether or not plasma concentration is the best surrogate marker in these

397

species, and differences in amenamevir efficacy for viral kinetics between species. The categories for lesion scores in human were not determined in the clinical study

399

protocol, and we originally defined four categories as shown in the present study. Scores in

400

humans showed monotonical change, i.e. the scores started from 3 and decreased along with

401

the lesion healing. Assuming that most patients showed lesion healing without recurrence and

402

therefore the lesion scores tended to monotonically decrease, the definition of lesion scores

403

seem acceptable for modeling purposes. We did not include data from patients whose lesion

404

scores remained 0 throughout the study period (aborted lesion) in the model analysis because

405

for a patient whose lesion score was 0 during the study, as we were unable to determine

406

whether or not the lack of any symptoms was due to the drug’s effects. Therefore, the PD

407

model of the present analysis was built only for patients who developed the symptoms.

M AN U

SC

RI PT

398

Finally, the lesion scores in guinea pig and human could be explained by the similar

409

models including with two fixed-effect parameters, i.e. amount of virus plaques and the

410

elapsed time. Our previous study regarding virus plaques showed that the continual exposure

411

of amenamevir above a certain concentration is necessary to prevent the virus re-production

412

[5], and this was confirmed by another study with multiple dose design which is usually used

413

in the antibiotics area [17]. During the clinical development of amenamevir, the value for

414

EC50 obtained in the non-clinical studies could be directly extrapolated into the clinical study

415

and is used for the dose rationale [5,6] As results, the non-clinical EC50 without a cure effect

416

was under-estimated, we were unable to detect a clear dose relationship in the clinical study.

AC C

EP

TE D

408

417

In the present study, we developed similar PK/PD models in both guinea pigs and

418

humans with Virus Plaque and Time components to explain the time-course profiles of lesion

419

scores. A virtual kinetic profile of virus plaque was incorporated into the model in order to

420

connect the PK profile of amenamevir with the lesion score profiles, and the terms in the

18

ACCEPTED MANUSCRIPT 421

logistic model consisting of the number of virus plaques and the elapsed time well explained

422

the dose- and time-dependent PD profiles. These results suggest that the virtual number of

423

virus plaques can be used as a built-in biomarker. While the fixed effect in Eff of Virus Plaque was negative in both species, the fixed

425

effect in Eff of the Time component for the immune system was negative in guinea pigs but

426

positive in humans. HSV-2 damages the central nervous system, which in turn affects the

427

immune system [18]. Our present findings suggest that the immune system might be

428

weakened by virus infection in guinea pigs, although evidence for this is insufficient at

429

present. The differences in results for the Time Component between species may have been

430

due to the different experimental conditions and different responses of immune systems. In

431

the guinea pig study, animals were infected with a lethal amount of HSV to ensure herpes

432

infection, and lesion severity increased with time. In humans, the immune system may work

433

adequately to reduce lesion severity.

M AN U

SC

RI PT

424

The effect of the Virus Plaque component was deemed to be large in guinea pigs, as

435

efficacy was clearly dose-dependent in the amenamevir groups while the effect was saturated

436

in the placebo group. In contrast, the effect of the Virus Plaque component was relatively

437

small in humans, and the drug effect for lesion scores was smaller than in guinea pigs.

EP

TE D

434

In humans, the PD effect was almost dose-independent, and immune system-related

439

healing was likely the driving force behind reductions in lesion scores. These findings

440

suggest that the drug effect may be masked in diseases healed by the immune response, such

441

as genital herpes. Therefore, the PK/PD model proposed in the present study will be

442

particularly useful for explaining the PK/PD relationship of drugs used to treat self-cured

443

diseases. In antibiotics and antiviral drug kinetic analyses, drug-bacteria (or virus) interaction

444

is assumed to be independent of in vivo conditions, such as host species. Here, we assumed

445

that the kinetic parameters for virus plaque data obtained in in vitro experiments could be

AC C

438

19

ACCEPTED MANUSCRIPT applied to both guinea pigs and humans. In addition, as a practical problem, we cannot obtain

447

the virus kinetic data in human and the difference of it between the species is difficult to be

448

evaluated.

449

In the development of drugs for diseases with natural healing, if healing does not happen in

450

the animal disease model, its efficacy may differ from clinical efficacy.

451

In the non-clinical studies from the perspective of the prediction of efficacy, animal model

452

without natural healing which can confirm the drug power clearly is suitable, however, it may

453

misjudge the clinical endpoint. For example, even when the development of animal models,

454

natural healing should be considered.

M AN U

455

SC

RI PT

446

Conclusions

457

This PK/PD modeling approach based on bi-directional translational approach is useful for

458

not only new candidate exploration in the non-clinical stage but also further application in

459

clinical data analysis. We believe that this kind of modeling and simulation approach will

460

give some suggestions especially as a unique PK/PD modeling approach connecting the

461

non-clinical and clinical data during the HSV drug development.

EP

462

TE D

456

Conflicts of interest

464

Financial support for this study was provided by Astellas Pharma Inc. The authors report no

465

conflicts of interest.

466

AC C

463

20

ACCEPTED MANUSCRIPT 467 References:

469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514

1. Stanberry L, Cunningham A, Mertz G, Mindel A, Peters B, Reitano M, Sacks S, Wald A, Wassilew S, Woolley P (1999) New developments in the epidemiology, natural history and management of genital herpes. Antiviral Res 42 (1):1-14. doi:S0166-3542(99)00004-2 [pii] 2. Snoeck R, De Clercq E (2002) New treatments for genital herpes. Curr Opin Infect Dis 15 (1):49-55 3. Reddy MB, Morcos PN, Le Pogam S, Ou Y, Frank K, Lave T, Smith P (2012) Pharmacokinetic/Pharmacodynamic predictors of clinical potency for hepatitis C virus nonnucleoside polymerase and protease inhibitors. Antimicrob Agents Chemother 56 (6):3144-3156. doi:10.1128/AAC.06283-11 AAC.06283-11 [pii] 4. Chono K, Katsumata K, Kontani T, Kobayashi M, Sudo K, Yokota T, Konno K, Shimizu Y, Suzuki H (2010) ASP2151, a novel helicase-primase inhibitor, possesses antiviral activity against varicella-zoster virus and herpes simplex virus types 1 and 2. J Antimicrob Chemother 65 (8):1733-1741. doi:10.1093/jac/dkq198 dkq198 [pii] 5. Katsumata K, Chono K, Kato K, Ohtsu Y, Takakura S, Kontani T, Suzuki H (2013) Pharmacokinetics and pharmacodynamics of ASP2151, a helicase-primase inhibitor, in a murine model of herpes simplex virus infection. Antimicrob Agents Chemother 57 (3):1339-1346. doi:10.1128/AAC.01803-12 AAC.01803-12 [pii] 6. Tyring S, Wald A, Zadeikis N, Dhadda S, Takenouchi K, Rorig R (2012) ASP2151 for the treatment of genital herpes: a randomized, double-blind, placebo- and valacyclovir-controlled, dose-finding study. J Infect Dis 205 (7):1100-1110. doi:10.1093/infdis/jis019 jis019 [pii] 7. Takada A, Katashima M, Kaibara A, Sawamoto T, Zhang W, Keirns J (2014) Statistical analysis of Amenamevir (ASP2151) between pharmacokinetics and clinical efficacies with non-linear effect model for the treatment of Genital Herpes. Clin Pharmacol Drug Dev. doi:DOI: 10.1002/cpdd.108 8. Katsumata K, Chono K, Sudo K, Shimizu Y, Kontani T, Suzuki H (2011) Effect of ASP2151, a herpesvirus helicase-primase inhibitor, in a guinea pig model of genital herpes. Molecules 16 (9):7210-7223. doi:10.3390/molecules16097210 molecules16097210 [pii] 9. Miller RL, Imbertson LM, Reiter MJ, Gerster JF (1999) Treatment of primary herpes simplex virus infection in guinea pigs by imiquimod. Antiviral Res 44 (1):31-42. doi:S0166-3542(99)00052-2 [pii] 10. Beal L, Sheiner B (1989) NONMEM Users Guide NONMEM Project Group, UCSF. 11. Wang Y (2007) Derivation of various NONMEM estimation methods. J Pharmacokinet Pharmacodyn 34 (5):575-593. doi:10.1007/s10928-007-9060-6 12. Post TM, Freijer JI, Ploeger BA, Danhof M (2008) Extensions to the visual predictive check to facilitate model performance evaluation. J Pharmacokinet Pharmacodyn 35 (2):185-202. doi:10.1007/s10928-007-9081-1 13. Lobo ED, Balthasar JP (2002) Pharmacodynamic modeling of chemotherapeutic effects: application of a transit compartment model to characterize methotrexate effects in vitro. AAPS PharmSci 4 (4):E42. doi:10.1208/ps040442 14. Sheiner LB (1994) A new approach to the analysis of analgesic drug trials, illustrated with bromfenac data. Clin Pharmacol Ther 56 (3):309-322

AC C

EP

TE D

M AN U

SC

RI PT

468

21

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

15. Liu CY, Sambol NC (1999) Pharmacodynamic analysis of analgesic clinical trials: nonlinear mixed-effects logistic models. J Biopharm Stat 9 (2):253-270. doi:10.1081/BIP-100101175 16. Holford N, Heo YA, Anderson B (2013) A pharmacokinetic standard for babies and adults. J Pharm Sci 102 (9):2941-2952. doi:10.1002/jps.23574 17. Frimodt-Moller N (2002) How predictive is PK/PD for antibacterial agents? Int J Antimicrob Agents 19 (4):333-339. doi:S0924857902000298 [pii] 18. Black PH (1994) Central nervous system-immune system interactions: psychoneuroendocrinology of stress and its immune consequences. Antimicrob Agents Chemother 38 (1):1-6

AC C

515 516 517 518 519 520 521 522 523 524 525 526

22

ACCEPTED MANUSCRIPT Tables Table 1 Estimated Population parameters of amenamevir in a guinea pig PK model and Virus Plaque PD model

PD model for Virus Plaque (Common parameters in Guinea pigs and humans) )

95% CIb (Lower – Upper) 1.93 – 2.15 3.05 – 4.73 1.02 – 2.50 0.154 – 0.204 5.9 – 14.0 8.61 – 18.6 9.2 – 28.9 5.3 – 12.7 13.6 – 45.4 112 –246 – 0.764 – 0.984 65.3 – 189 7.0 – 10.2

RI PT

CL/F (L/h/kg) V/F (L/kg) ka (h-1) Lag Time (h) ηCL/F (CV%) ηka (CV%) ηLagTime (CV%) εc kinact (h-1) kact (h-1) kin (h-1) Emax (pFU) EC50 (ng/mL) εc

Estimate (%RSEa) 2.04 (2.76%) 3.89 (11.1%) 1.76 (21.4%) 0.179 (7.15%) 10.7 (35.3%) 14.5 (33.1%) 21.4 (41.6%) 9.76 (35.9%) 29.5 (27.6%) 179 (19.0%) 0.0569 (Fixed) 0.874 (6.40%) 127 (24.8%) 8.75 (18.7%)

SC

PK model for Guinea pig

Parameter

M AN U

Model

AC C

EP

TE D

CI: confidence interval, CL/F: oral clearance, V/F: volume of distribution, ka: absorption rate, η: inter-individual variability, ε: intra-individual variability, kinact: inactivation ratio , kact: activation ratio, kin: increase ratio , Emax: maximum drug effect, EC50: Michaelis constant a: %RSE is percent relative standard error (100% × Standard Error / Estimate) b: 95% CI = T ± 1.96 × Standard Error c: given as standard deviation

ACCEPTED MANUSCRIPT Table 2 Estimated parameters of the logistic regression model for lesion score in guinea pigs and humans

θScore=0 θScore=1 θScore=2 θScore=3 ηc

RI PT

β2

Estimate (%RSEa) -0.0247 (29.5%) 0.0424 (6.0%) -5.21 (6.8%) 1.86 (7.3%) 3.34 (6.0%)

Human 95% CIb (Lower – Upper) -0.0390, -0.0104 0.0374, 0.0474 -5.90, -4.52 1.60, 2.12 2.94, 3.74

SC

β1

Estimate (%RSEa) -0.265 (10.6%) -0.0334 (14.6%) 6.76 (15.4%) 3.11 (13.2%) 4.17 (11.2%) 5.89 (12.4%) 1.62 (31.7%)

M AN U

Parameter

Guina pig 95% CIb (Lower – Upper) -0.320 -0.210 -0.0430, -0.0238 4.72, 8.80 2.30, 3.92 3.25, 5.09 4.46, 7.32 0.996, 2.07

1.36 (16.2%)

1.12, 1.56

AC C

EP

TE D

CI: confidence interval, β1: effect of virus plaque, β2: time effect for healing, θScore=x: logit value for score x, η: inter-individual variability a: %RSE is percent relative standard error (100% × Standard Error / Estimate) b: 95% CI = T ± 1.96 × Standard Error c: given as standard deviation

ACCEPTED MANUSCRIPT Figure captions Figure 1: Time course profiles of mean observed lesion scores in guinea pigs (a) and humans (b). (a) Closed circle: Placebo, open triangle: 1 mg/kg, closed triangle: 3 mg/kg, open square: 10 mg/kg, closed square: 30 mg/kg; treatment duration was 5 days from Day 1.

RI PT

(b) Closed circle: Placebo, open triangle: 100 mg, closed triangle: 200 mg, open square: 400 mg, closed square: 1200 mg; treatment duration was 3 days from Day 1.

SC

Figure 2: Overview of the PK/PD model

M AN U

Figure 3: Results of population PK modeling and visual predictive check in guinea pigs. (a) Dose = 0.3 mg/kg, (b) 1.0 mg/kg, (c) 3.0 mg/kg. Solid line: median, filled region: 95% prediction interval.

TE D

Figure 4: Results of population PD modeling and visual predictive check for virus plaque data. (a) Amenamevir duration time = 6 h, (b) 8 h, (c) 24 h. Solid line: median, filled

EP

region: 95% prediction interval.

AC C

Figure 5: Results of model predicted time course profiles in guinea pigs. Closed circle: Placebo, open triangle: 1 mg/kg, closed triangle: 3 mg/kg, open square: 10 mg/kg, closed square: 30 mg/kg; treatment duration was 5 days from Day 1. (a), (b) Simulated time-course profiles of plasma concentration and virus plaque. (c) Observed (plots) and model-predicted (lines) time-course profiles of lesion scores in guinea pigs. Predictions are given as surface of lesion scores (z-axis) as a function of time (x-axis) and dose (y-axis). Symbols show the observed values. (d) to (g) Predicted probability surfaces for lesion scores (z-axis) as a

ACCEPTED MANUSCRIPT function of day (x-axis) and dose (y-axis) in guinea pigs. (d) Pr{Y>=1}, (e) Pr{Y>=2}, (f) Pr{Y>=3}, (g) Pr{Y=4}.

Figure 6: Results of model predicted time course profiles in humans. Closed circle:

RI PT

Placebo, open triangle: 100 mg, closed triangle: 200 mg, open square: 400 mg, closed square: 1200 mg; treatment duration was 3 days from Day 1. (a), (b) Simulated time-course profiles of plasma concentration and virus plaque. (c) Observed (plots) and model-predicted (lines)

SC

time-course profiles of lesion scores in humans. Predictions are given as surface of lesion scores (z-axis) as a function of time (x-axis) and dose (y-axis). Symbols show the observed

M AN U

values. (d) to (f) Predicted probability surfaces for lesion scores (z-axis) as a function of day

AC C

EP

TE D

(x-axis) and dose (y-axis) in human. (b) Pr{Y>=1}, (c) Pr{Y>=2}, (d) Pr{Y=3}.

ACCEPTED MANUSCRIPT Figure 1

M AN U

SC

RI PT

a)

AC C

EP

TE D

b)

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Figure 2

ACCEPTED MANUSCRIPT Figure 3

SC

RI PT

a)

EP AC C

c)

TE D

M AN U

b)

ACCEPTED MANUSCRIPT Figure 4

SC

RI PT

a)

EP AC C

c)

TE D

M AN U

b)

ACCEPTED MANUSCRIPT Figure 5 a)

SC

RI PT

b)

EP AC C

d)

TE D

M AN U

c)

e)

ACCEPTED MANUSCRIPT g)

AC C

EP

TE D

M AN U

SC

RI PT

f)

ACCEPTED MANUSCRIPT Figure 6 b)

SC

RI PT

a)

AC C

d)

EP

TE D

M AN U

c)

e)

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

f)