Pharmacokinetic profile of edaravone: a comparison between Japanese and Caucasian populations

Abstract

Background: Amyotrophic lateral sclerosis (ALS) affects persons of all races, and there continues to be a need for effective therapies to treat the disease.

Objective: To compare the pharmacokinetics (PK) of edaravone between Japanese and Caucasian populations. Methods: Data from five PK studies among Japanese and Caucasian healthy volunteers were pooled and evaluated. In population PK (PPK) modelling, compartment models and other models with linear elimination were evaluated for appropriateness. Covariate effects by race, sex, weight, and age were investigated to explain variability in PK parameters. Simulations of the final PPK model were performed using a virtual population based on ALS clinical trials. Results: The analysis included 86 subjects. A three-compartment model with Michaelis-Menten plus linear elimination was selected as the best fit model. Race was statistically detected as a covariate for the second peripheral volume of distribution (V2), indicating a 26% increase for Caucasian subjects compared to Japanese subjects. However, based on simulation of PPK model for a virtual ALS population, the small difference of V2 was associated with a difference of C_tau around 1 ng/mL after infusion, which was minimal compared to C_max of approximately 1000 ng/ml. Conclusion: The PPK analyses demonstrated no clinically relevant difference in the PK profiles of edaravone by race, sex, weight, or age.

Keywords:

• Amyotrophic lateral sclerosis
• pharmacokinetic
• edaravone

Introduction

Edaravone (MCI-186), hypothesised as a free radical scavenger, has been extensively examined in in vitro and in vivo animal and human studies over the last 30 years. Initially, edaravone was developed for treatment of acute ischaemic stroke (AIS). In Japan, edaravone was approved in 2001 for AIS with a dosing regimen of 30 mg over 30 min by infusion, twice daily up to 14 d (1). After approval in Japan, edaravone was also explored in Europe for intensive treatment of AIS using a continuous infusion dosing regimen more suitable for medical practice in that region (2). Thus, the pharmacokinetics (PK) of edaravone has been studied among several different dosing regimens in both Japanese and Caucasian populations.

The PK of single and multiple doses of edaravone in Japanese and Caucasian healthy volunteers have been previously examined (Table 1). While there are no direct PK data for the proposed dose for ALS treatment (60 mg/60-min infusion once daily with a two-week period of treatment followed by a two-week period of no drug), or for comparison between Japanese and Caucasian populations, there are enough data to conduct a population PK (PPK) analysis and simulate the PK profile of edaravone. The results of PPK analyses to compare the PK profile of edaravone between Japanese and Caucasian populations are described here.

Table 1. Summary of PK studies in healthy volunteers and demographics summary.

					Race		Sex		Age
MCI186 Study	Bioanalytical Method (LLOQ)	Treatment Regimen	Dosing	Duration of Dosing	JP	CA	M	F	Median (range), years	Weight Median (range), kg
J01	GC/MS (0.1 ng/mL)	Single dose IV infusion	0.2, 0.5, 1.0, or 1.5 mg/kg for 40 min; 2.0 mg/kg for 3 h	1 day, 1 dose	30	0	30	0	28.5 (24–44)	66.05 (53.7–85.0)
		Multiple dose IV infusion	1.0 mg/kg for 40 min	7 days, 7 doses
J10	GC/MS (1.0 ng/mL)	Multiple dose IV infusion	0.5 mg/kg for 30 min twice daily	2 days, 4 doses	10	0	10	0	50.0 (25–71)	62.35 (56.8–74.2)
J14	LC-MS/MS (2.0 ng/mL)	Single dose IV infusion	120 mg/day for 48 h	2 days, 1 dose	7	0	7	0	23.0 (20–28)	62.70 (50.9–75.4)
E01	GC/MS (5.0 ng/mL)	Single dose IV infusion	0.6 or 1.8 mg/kg for 6 h	1 day, 1 dose	0	20	10	10	60.5 (52–75)	71.65 (49.9–95.0)
E02	LC-MS/MS (2.0 ng/mL)	Bolus + continuous IV infusion	0.05 mg/kg (3 min) + 0.125 mg/kg/h for 23.950.1 mg/kg (3 min) + 0.25 mg/kg/h for 23.95 h; or0.2 mg/kg (3 min) + 0.5 mg/kg/h for 23.95 h	1 day, 1 dose	0	30	15	15	62.0 (51–70)	72.35 (59.3–99.9)

CA: Caucasian; F: female; GC/MS: gas chromatography/mass spectrometry; IV: intravenous; JP: Japanese; LC-MS/MS: liquid chromatography-tandem mass spectrometry; LLOQ: lower limit of quantitation; M: male.

Methods

Data

Data from the active treatment arm (edaravone) of five PK studies in Japanese or Caucasian healthy volunteers were pooled for the PPK analysis (Table 1). All study participants gave written informed consent and met the inclusion and exclusion criteria for the study in which they were enrolled. Dosing regimens for edaravone are described in Table 1. Plasma samples were assayed for unchanged edaravone using gas chromatography/mass spectrometry (GC/MS) or liquid chromatography-tandem mass spectrometry (LC-MS/MS). The bioanalytical method and lower limit of quantitation (LLOQ) of edaravone in plasma used for each study are described in Table 1. The following data were collected during screening or prior to the first dose of edaravone: sex, race (Japanese or Caucasian), age, and body weight. Data utilised in the creation of the analysis datasets included dosing information (amount, route, timing), PK sampling information (time relative to dosing, concentration of edaravone), treatment assignment, and demographic data.

Population PK modelling

Population modelling was performed using the computer programme NONMEM (version 7, level 3) with first-order conditional with interaction estimation method used during all stages of model development. For each analysis, NONMEM computed the value of the objective function (VOF), a statistic that is proportional to minus twice the log likelihood of the data. In the case of hierarchical models, the change in the VOF produced by the inclusion of a parameter was asymptotically χ²-distributed, with the number of degrees of freedom equal to the number of parameters added to or deleted from the model. The general procedure followed for the development of a PK model included: (i) exploratory data analysis; (ii) base structural model development; (iii) evaluation of covariate effects (sex, race, age, and weight); (iv) model refinement; and (v) model evaluation using a visual predictive check (VPC) and comparison of exposures calculated using observed and model predicted concentrations.

Exploratory data analyses and data visualisation techniques were used to understand the informational content of the dataset with respect to the anticipated model, search for extreme values and potential outliers, assess possible trends in the data, and determine if any errors were made in the manipulation of the data and creation of the analysis datasets. This exploratory analysis was also used to confirm the appropriateness of the models tested and verify model assumptions.

Based on preliminary examination of summarised concentration-time data, mamillary, two-compartment and three-compartment models were assessed, and other models with linear and/or Michaelis-Menten elimination were evaluated for appropriateness in describing the PK data. The various population PK models evaluated were described by the estimation of mean structural model parameters (e.g. volumes of distribution in central and peripheral compartments, clearance [CL], and intercompartmental CL), magnitude of interindividual variability (IIV) in these parameters, and magnitude of residual variability (RV). Following development of an appropriate base structural model, covariates were evaluated for their ability to explain variability in the PK model parameters: age, years; weight, kg; sex (0 = male, 1 = female); and race (0 = Japanese, 1 = Caucasian). Graphical and statistical approaches (forward selection followed by backward elimination) were used to develop the covariate models and to assess the mathematical forms of their relationships and statistical significance.

Virtual simulations

Simulations were performed to compare anticipated edaravone exposures by race. A virtual population of 2000 individuals with ALS (1000 Japanese and 1000 Caucasian) was generated assuming normal distributions and using mean and standard deviation of age and weight for each race with considering gender ratio (Table 2). Japanese demographics were based on edaravone ALS (3–7) clinical studies and Caucasian demographics were based on publications for recent ALS clinical trials conducted in the EU and North America (8–15).

Table 2. Demographic definitions for simulated ALS patient populations.

	Japanese n = 1000	Caucasian n = 1000
Sex (male)	62%	59%
Weight	57.9 + 10.69 × random normal (0,1) restricted to a range of 35 to 109 kg	77.5 + 15.5 × random normal (0,1) restricted to a range of 39 to 116 kg
Age	59 + 9.9 × random normal (0,1) restricted to a range of 28 to 75 years	56 + 11.7 × random normal (0,1) restricted to a range of 20 to 79 years

NONMEM was used to generate individual single-dose and steady-state measures of edaravone exposures (C_max, C_tau, and AUC_0–24 h), by integration of the model-predicted concentration equation for each subject based on the population PK model and the simulated individual-specific model parameters. A dose of 60 mg edaravone infused over 1 h daily for 14 d was simulated for the virtual ALS population of Japanese and Caucasian subjects using the final PPK model. An identical approach was also conducted for a dose of 30 mg and 120 mg of edaravone. Summary statistics and box plots of exposures stratified by race were generated for each dosing regimen.

Results

From the active treatment arms (edaravone) of the five studies, a total of 97 subjects were included in the source PK dataset (excluding samples collected prior to the first dose of edaravone). For E01 study, the 0.6 mg/kg treatment group had 64 below the lower limit of quantification (BLQ) samples (34%). As the 0.6 mg/kg/6-h infusion group was the only treatment group with a substantial number of BLQ samples and because the treatment group represented a dose well below the planned clinical dose of 1 mg/kg/h, the decision was made to remove this treatment group from the analysis (M1 method as recommended by Beal (16)). Additionally, one subject from Study E02 was identified as a graphical outlier and removed from the analysis dataset. Thus, the final PPK analysis dataset included 86 subjects.

Basic demographic characteristics are summarised in Table 1. Briefly, there was a near-equal distribution of Japanese and Caucasian subjects, 54.7% vs. 45.3%. The mean age (SD) of subjects was 45.8 (17.4) years, and approximately three-quarters of subjects were male (76.7%).

On average, subjects had approximately 16 samples each with a measurable concentration following a single dose of edaravone. Following the first dose of edaravone, sample times with measurable concentrations ranged from 0.1 to 57 h after the start of infusion and the sample times following seven doses of edaravone ranged from 0 to 24 h after the start of infusion.

Concentration-time profiles showed edaravone to decline in a multi-phasic manner post-infusion (Figure 1). Concentration-time profiles during longer infusions (24–48 h) also showed multi-peak behaviour during infusion (Figure 1(c)). Additionally, plots of dose-normalized edaravone concentrations indicated a small separation of the treatment groups (0.5 to 1 times the standard deviation) at specific intervals for select dosing regimens, suggesting the PK of edaravone may exhibit non-linerarity (Figure 2).

Figure 1. Edaravone concentration vs time (since start of infusion) plots. (a) 40-minute infusion (log-linear scale). (b) 24-hour and 48-hour infusions (log-linear scale). (c) 48-hour infusion (linear scale).

Figure 2. Dose-normalized edaravone concentration vs time (since start of infusion) plots. (a) 40-minute infusion (log-linear scale). (b) 24-hour infusion (log-linear scale).

Model development

A three-compartment model with Michaelis-Menten plus linear elimination and fluctuating V_max was selected as the best fit model for the concentration-time profiles of edaravone (Figure 3, Table 3). The observe versus model-predicted concentration plots showed good correspondence (Figure 4), with the exception of a small number of C_max values (>2800 ng/ml) which were underpredicted. All parameters were well estimated (standard error of the mean ≤34%) and the interindividual variability of clearance (CL), central volume of distribution (V1), second peripheral volume of distribution (V2), and third peripheral volume of distribution (V3) was small (≤26% coefficient of variation). Gender, age or weight did not affect any PK parameters. Race was only statistically detected as a covariate for V2, indicating a 26% increase for Caucasian subjects compared to Japanese subjects. In model refinement process, the model reducing the number of RV error models by grouping the assays with an LLOQ of 0.1 and 2.0 ng/ml was successful. And the simplified model was selected as the final population PK model for edaravone. A prediction-corrected VPC analysis was performed to evaluate the predictive performance of the final PPK model. The VPCs also indicated that the model was able to simulate concentration-time profiles that reasonably corresponded to the distribution of observed concentration-time profiles (data not shown).

Figure 3. Three-compartment Michaelis-Menten plus linear elimination model diagram. A: amount of drug in the central compartment (mg); A₅₀: amount of drug needed to achieve 50% of V_max; CL: clearance (L/h); Q2: intercompartmental clearance (L/h); Q3: intercompartmental clearance 2 (L/h); R₀: zero-order infusion rate (mg/h); V1: central volume of distribution (L); V_max: maximum nonlinear elimination rate (mg/h) either constant or fluctuating; V2: peripheral volume of distribution 2 (L); V3: peripheral volume of distribution 3 (L).

Figure 4. Scatterplot of observed versus model-predicted concentrations of edaravone.

Table 3. Parameter estimates from the final edaravone PPK model.

	Final parameter estimate		Interindividual variability/residual variability
Parameter	Typical value	% SEM	Magnitude	% SEM
CL
Clearance (L/h)	23.0	8.26	20.7% CV	22.9
V1
Central volume of distribution (L)	14.7	4.44	23.4% CV	26.8
Q2
Intercompartmental clearance (L/h)	4.79	5.71	Not estimated	Not estimated
V2
Peripheral volume of distribution (L)	32.6	6.40
V2 proportional shift for Caucasian	0.262	21.9	16.0% CV	26.7
Q3
Intercompartmental clearance 2 (L/h)	26.8	3.14	Not estimated	Not estimated
V3
Peripheral volume of distribution 2 (L)	25.1	3.49	26.0% CV	22.3
A50
Amount to achieve 50% VMM (mg)	6.84	19.9	Not estimated	Not estimated
VMM
Midpoint maximum elimination rate (mg/h)	17.9	17.9	Not estimated	Not estimated
VMMF
Fractional amplitude VMM (unitless)	0.174	18.9	Not estimated	Not estimated
Log RV for Studies J01, J14, and E02	0.0199	7.74	0.141 SD	Not applicable
Log RV for Study J10	0.0665	20.4	0.258 SD	Not applicable
Log RV for Study E01	0.0492	34.0	0.222 SD	Not applicable
Minimum value of the objective function = −3482.919

%CV: coefficient of variation expressed as a percentage; SD: standard deviation; %SEM: percent standard error of the mean (standard error/parameter estimate ×100).

V_max = VMM + VMM . VMMF .

V_max is the maximum nonlinear elimination rate; VMM is the midpoint of V_max; VMMF is the amplitude of V_max expressed as a fraction of VMM; and t is continuous time (h).

Simulation in a virtual ALS population

Based on the final PPK model, the small difference of V2 was associated with a difference of C_tau around 1 ng/ml after infusion (Table 4). This difference was minimal compared to C_max (approximately 1000 ng/ml), and it did not result in the accumulation of drug concentration after multiple dosing. Simulations of 14 daily doses administered to a virtual ALS population showed that the estimated C_max and AUC values were less than 0.1% different for Japanese and Caucasian patients (Table 4, Figure 5).

Figure 5. Simulated edaravone concentration versus time since start of infusion (Virtual ALS Patients, 60 mg/60 minutes daily for 14 days). For display purpose, the time since start of infusion of each sample for Caucasian subjects was offset by 0.15 hours.

Table 4. Simulated exposures to edaravone (mean (SD), virtual ALS patients, dosing daily for 14 days).

	Japanese n = 1000 subjects		Caucasian n = 1000 subjects
Simulated exposure measure	Day 1	Day 14	Day 1	Day 14
30 mg/60 min
Cmax (ng/mL)	470.0 (52.7)	470.5 (52.8)	469.2 (54.1)	469.9 (54.3)
Ctau (ng/mL)	0.32 (0.14)	0.33 (0.16)	0.47 (0.19)	0.51 (0.22)
AUC0–24 h (ng*h/mL)	599.5 (85.2)	602.1 (86.1)	597.8 (87.3)	602.6 (89.0)
60 mg/60 min
Cmax (ng/mL)	1048.1 (114.2)	1049.3 (114.8)	1046.6 (117.0)	1048.6 (117.6)
Ctau (ng/mL)	0.73 (0.32)	0.77 (0.35)	1.07 (0.43)	1.17 (0.50)
AUC0–24 h (ng*h/mL)	1367.0 (191.2)	1373.5 (193.2)	1362.3 (194.6)	1374.3 (198.6)
120 mg/60 min
Cmax (ng/mL)	2293.5 (247.8)	2296.9 (248.6)	2289.8 (252.0)	2294.4 (253.8)
Ctau (ng/mL)	1.70 (0.75)	1.79 (0.82)	2.48 (0.98)	2.73 (1.14)
AUC0–24 h (ng*h/mL)	3143.9 (456.4)	3160.9 (462.0)	3130.3 (460.9)	3161.1 (471.3)

AUC_0–24 h: area under the concentration-time curve from 0 to 24 hours; C_max: maximum concentration; C_tau: concentration at 24 hours; n: number of subjects; SD: standard deviation.

While the majority of the dosing regimens included in the analysis dataset were weight-based, weight was not found to be a statistically significant predictor of PK parameters for edaravone (Figure 6). The box plots of the model-predicted exposures versus weight quartiles for the analysis population also indicated that variability in PK exposure was not related to weight. The range of exposures for the lowest weight subjects and the highest weight subjects were similar.

Figure 6. Boxplot of model-predicted exposures versus weight covariate (virtual ALS patients, 60 mg/60 minutes daily for 14 days). Boxes are the 25th, 50th, and 75th percentiles; whiskers are the 5th to 95th percentiles. Asterisks show data points outside this range. [or] indicates the respective endpoint is included in the interval. (or) indicates the respective endpoint is not included in the interval.

Simulations of 14 daily doses administered to a virtual ALS patient population showed that the complete dose amount was excreted over a 24-h period without accumulation of AUC after multiple doses, and estimated that the mean amount excreted via the non-linear pathway (expressed as a percentage of the total excretion) was 47.4% for the 60-mg infusion over 60-min regimen.

Based on the simulation from analysis, the PK of edaravone was nearly linear over a dose range of 30 mg–120 mg (infused over 60 min/subject) as indicated by a 2.2 to 2.3 times greater AUC when the dose was doubled.

The half-lives of each elimination phase (α, β, and γ) after dosing at 60 mg/60 min/subject were calculated using the mean of the simulated plasma concentration of edaravone by time in virtual ALS populations (1000 patients for Japanese and Caucasian, respectively). The calculated half-lives of the α, β, and γ elimination phases in Japanese subjects were 0.15, 0.86, and 4.41 h, and those in Caucasian subjects were 0.15, 0.88, and 6.34 h, respectively.

Discussion

Based on recently published literature of clinical trials for ALS in the EU and North America, 90% or more of patients identified as Caucasian or white (8–15). In the development programme of edaravone, no direct PK data were obtained for the proposed dose for ALS treatment (60 mg infusion for 60 min once a day) or for comparison between Japanese and Caucasian populations. However, the PK of edaravone has been evaluated with various dosing regimens, 30-min to 48-h infusions, in both Japanese and Caucasian subjects as shown in Table 1. With this extensive PK dataset, we considered that population PK approach would be adequate to evaluate the similarity or difference in PK of edaravone between Japanese and Caucasian populations, and simulate the PK curve of edaravone for the proposed dosing regimen for ALS treatment.

A three-compartment model with Michaelis-Menten plus linear elimination was selected as the best fit model to present the PK of edaravone. Because plasma concentration of edaravone over a broad range of dosing regimens were used to develop the PPK model, non-linearity model was included to fit the observed data. However, the PK of edaravone was nearly linear over a dose range of 30 –120 mg (infused over 60 min/subject) as indicated by a 2.2 to 2.3 times greater AUC when the dose was doubled. A fluctuation was observed in plasma concentration in the 24-h and 48-h infusions. The reproducibility of the plasma concentration of edaravone was improved by incorporating fluctuation in V_max. Also, the incorporation of the fluctuation in V_max reduced the residual variabilities and interindividual variability in CL compared to the model without the fluctuation. Based on the final PPK model, gender, age or weight did not affect any PK parameters. Race was only statistically detected as a covariate for the second volume of distribution (V2), indicating a 26% increase for Caucasian subjects compared to Japanese subjects. However, this difference was minimal compared to C_max, and there was no difference in simulated AUC or C_max between the populations.

There is a potential limitation of data used in the PPK analyses regarding an effect of body weight on PK since the doses in four out of five studies were adjusted by subject body weight. Therefore, the evaluation of relationships between body weight or body mass index (BMI) and PK parameters (AUC) in this PPK modelling may have a limitation. Study J14 was the only study where a fixed dose of 240 mg edaravone was administered without adjustments with body weight to decide dose level per each subject. There appeared to be no tendency for AUC to decrease with increasing body weight (range 50.9–75.4) or BMI (range 19–24) (Figure 7).

Figure 7. Relationship between body weight/body mass index and pharmacokinetic parameters (120 mg/subject/24 hr for 2 days) in Japanese healthy male adults.

Edaravone is metabolised into its sulphate and glucuronide (17,18). Neither the sulphate nor the glucuronide have radical scavenging activities (19). For the glucuronide conjugate reaction, multiple uridine diphosphate glucuronosyltransferase (UGT) isozymes including UGT1A9, with the highest contribution, are involved (20). Edaravone and its metabolites are excreted into urine. While the main metabolite in plasma was the sulphate conjugate, the main metabolite in human urine is the glucuronide because it is deconjugated and then reconjugated with glucuronide in the human kidney. Since edaravone is metabolised with sulphate conjugation and glucuronide conjugation involving multiple UGT enzymes, it is theoretically unlikely that edaravone metabolism is different between Japanese and Caucasian populations.

The PPK analyses demonstrated no clinically relevant difference in PK profiles between the Japanese and Caucasian subjects. Neither C_max nor AUC of edaravone appear to be related to age, gender, or weight. Taken together, there is PK evidence to suggest that dose adjustments of edaravone are not likely warranted between Japanese and Caucasian individuals.

Declaration of interest

YN, SK, AK, and MS are employees of MTPC. KT is an employee of MTDA. JP is an employee of MTPC and MTDA. The authors take full responsibility for the content of and the decision to submit this manuscript but thank Teresa A. Oblak of Covance Market Access Services Inc. for providing research support, coordination assistance, and editorial contributions.

Mitsubishi Tanabe Pharmaceutical Corporation (MTPC) was responsible for oversight of the PPK analysis performed by Cognigen Corporation. Mitsubishi Tanabe Pharma Development America (MTDA) provided medical writing support for this article. The ALSFTD supplement, Edaravone (MCI-186) in Amyotrophic Lateral Sclerosis (ALS), was funded by Mitsubishi Tanabe Pharma America, Inc.