Post-hoc analysis of randomised, placebo-controlled, double-blind study (MCI186-19) of edaravone (MCI-186) in amyotrophic lateral sclerosis

Abstract

Post-hoc analyses of the ALS Functional Rating Scale-Revised (ALSFRS-R) score data, the primary endpoint in the 24-week double-blind placebo-controlled study of edaravone (MCI186-19, NCT01492686), were performed to confirm statistical robustness of the result. The previously reported original analysis had used a last observation carried forward (LOCF) method and also excluded patients with fewer than three completed treatment cycles. The post-hoc sensitivity analyses used different statistical methods as follows: 1) including all patients regardless of treatment cycles received (ALL LOCF); 2) a mixed model for repeated measurements (MMRM) analysis; and 3) the Combined Assessment of Function and Survival (CAFS) endpoint. Findings were consistent with the original primary analysis in showing superiority of edaravone over placebo. We also investigated the distribution of change in ALSFRS-R total score across all patients in the study as well as which ALSFRS-R items and domains may have contributed to the overall efficacy findings. The distribution of changes in ALSFRS-R total score from baseline to the end of cycle 6 (ALL LOCF) shifted in favour of edaravone compared to placebo. Edaravone was descriptively favoured for each ALSFRS-R item and each of the four ALSFRS-R domains at the end of cycle 6 (ALL LOCF), suggesting a generalised effect of edaravone in slowing functional decline across all anatomical regions. The effect of edaravone appeared to be similar in patients with bulbar onset and limb onset. Together, these observations would be consistent with its putative neuroprotective effects against the development of oxidative damage unspecific to anatomical regions.

Keywords:

• Amyotrophic lateral sclerosis
• edaravone
• double-blind
• study MCI186-19
• post-hoc

Introduction

In the first phase III study of edaravone in ALS (MCI186-16), a double-blind 24-week trial, edaravone did not demonstrate statistical superiority to placebo for the primary efficacy analysis of change in the ALS Functional Rating Scale-Revised (ALSFRS-R) (1). Subsequent post-hoc exploratory analyses were conducted to identify a subpopulation in which efficacy of edaravone might be detectable within a 24-week study period. The subpopulations were defined with the following criteria: i) patients with scores of ≥2 points on all 12 items of the ALSFRS-R (broad retained functionality as measured by the ALSFRS-R); ii) a percent forced vital capacity (%FVC) of 80% or greater (almost normal respiratory function); iii) definite or probable ALS diagnosed by El Escorial revised criteria; and iv) within two years after the first symptom (2). The second 24-week, randomised, placebo-controlled, double-blind study of edaravone (MCI186-19) was prospectively designed to enrol patients meeting these criteria and demonstrated a significant difference from placebo in the change of the ALSFRS-R total score (2.49 ± 0.76 least square [LS] mean ± standard error [SE]; [p = 0.0013]) (3). This was the primary analysis for which data were imputed using the last observation carried forward (LOCF) method for patients with missing values, except that data from patients who discontinued before the end of cycle 3 were excluded. In order to examine robustness of the primary result, we here describe the post-hoc analyses we performed using all available patient data that included two distinct analysis methods: an analysis of variance model (ANOVA) with LOCF imputation that did not exclude any discontinuing patients (ALL LOCF) and a mixed model for repeated measurements (MMRM), which can address all available post baseline data. Analysis was also performed using the Combined Assessment of Function and Survival (CAFS) endpoint, which incorporates function and survival together and is a well-accepted non-parametric analysis method (4).

In the present post-hoc analyses, we also examined the distribution of patient changes in ALSFRS-R total score from baseline for each patient in the treatment groups and the distribution shift by edaravone treatment. The ALSFRS-R score consists of 12 items in four domains as follows: bulbar (items 1–3), fine motor (items 4–6), gross motor (items 7–9), and respiratory (items 10–12) (5). We looked for treatment effects on the individual domains and items and the extent to which these contributed to the change in ALSFRS-R total score. Since patterns of functional deterioration may potentially be different between patients with bulbar onset and patients with limb onset (6), analyses for these four domains stratified by initial symptom (bulbar versus limb onset) were also performed.

Patients and methods

Study design

MCI186-19 was a 24-week, placebo-controlled, double-blind study comparing 60 mg edaravone administered by infusion with matching placebo. In cycle 1, the study drug was administered for 14 consecutive days followed by a two-week drug-free period. In cycle 2 and thereafter, the study drug was administered for 10 of the 14 days followed by a two-week drug-free period. A more complete description of the study design has been previously published (3), as have the data pertaining to the primary and secondary efficacy measures and the safety findings (3). All patients provided written informed consent. The study was registered as NCT01492686 with clinicaltrials.gov.

Patients

The post-hoc analyses were performed in the full analysis set (FAS) population, which included all randomised patients who received study drug and had at least one post baseline efficacy data point. There were 69 patients in the edaravone group (16 with bulbar onset and 53 with limb onset) and 68 patients in the placebo group (14 with bulbar onset and 54 with limb onset). Of these, two patients in the edaravone group (one with bulbar onset and one with limb onset) and eight patients in the placebo group (three with bulbar onset and five with limb onset) discontinued treatment before the end of cycle 6.

Efficacy evaluation

Sensitivity analyses of the primary endpoint (change in ALSFRS-R total score from baseline to the end of cycle 6)

ANOVA was performed with treatment group, change in ALSFRS-R total score during the pre-observation period (–3/–4 or –2/–1), El Escorial revised Airlie House diagnostic criteria (definite or probable), and age (< 65 or ≥65 years) as fixed effects. This was the same statistical model as performed for the prespecified primary analysis (3), except that LOCF was applied to all randomised patients (ALL LOCF). MMRM analysis included treatment group, time, treatment group by time interaction, change in ALSFRS-R score during the pre-observation period, El Escorial revised Airlie House diagnostic criteria and age as fixed effects, and baseline ALSFRS-R score as a covariate. An unstructured covariance matrix was used to model the covariance of within-patient scores. The Kenward-Roger approximation was used to estimate the denominator degrees of freedom. We also performed CAFS analysis using the same statistical methodology as reported for the primary analysis of the EMPOWER study of dexpramipexole in ALS (7). CAFS Rank Scores were derived based on change in ALSFRS-R total score from baseline to the end of cycle 6. An analysis of covariance (ANCOVA) was performed with baseline ALSFRS-R score, duration of ALS, riluzole use, initial symptom (bulbar or limb), and treatment group.

Distribution of changes in the ALSFRS-R total score

The Hodges-Lehmann estimator was used to estimate median shift and the Wilcoxon rank-sum test was performed to determine its statistical significance.

Changes in each ALSFRS-R score item and four domains

The same ANOVA method with ALL LOCF that was used to analyse differences between treatment groups in total ALSFRS-R scores was applied to the changes of ALSFRS-R items and four domains.

Results

Efficacy results

Sensitivity analyses of ALSFRS-R total score

The post-hoc ANOVA with ALL LOCF and MMRM analyses have also been briefly described in the earlier report (3). Comparisons between treatment groups are shown in Table 1. The estimated difference in the least square mean with standard error between treatment groups was 2.37 ± 0.75 (p = 0.0019) using ANOVA with ALL LOCF and 2.81 ± 0.78 (p = 0.0004) using MMRM. These results were consistent with those obtained in the prespecified primary analysis, which used ANOVA with LOCF applied (2.49 ± 0.76 [p = 0.0013]) (3). Comparisons using the CAFS endpoint also showed a statistical difference between treatment groups, 41.64 ± 12.30 (p = 0.0009).

Table 1. Prespecified primary analysis and post-hoc sensitivity analyses of change in ALSFRS-R total score and CAFS from baseline to the end of cycle 6 (week 24).

Analysis method	Group	Number of patients analysed	Adjusted meanLS mean ± SE	Between-group differences in the adjusted meanLS mean ± SE (95% CI)
ANOVA with Prespecified LOCF^a (primary analysis)	Edaravone	68	−5.01 ± 0.64	2.49 ± 0.76 (0.99, 3.98)	p = 0.0013
	Placebo	66	−7.50 ± 0.66
ANOVA with ALL LOCF^b (sensitivity analysis)	Edaravone	69	−5.04 ± 0.64	2.37 ± 0.75 (0.89, 3.84)	p = 0.0019
	Placebo	68	−7.41 ± 0.65
MMRM (sensitivity analysis)	Edaravone	69	−4.56 ± 0.55	2.81 ± 0.78 (1.27, 4.35)	p = 0.0004
	Placebo	67^c	−7.37 ± 0.57
CAFS (sensitivity analysis)	Edaravone	69	–	41.64 ± 12.30 (17.31, 65.96)	p = 0.0009
	Placebo	68	–

ALSFRS-R = the Revised Amyotrophic Lateral Sclerosis Functional Rating Scale.

ANOVA = analysis of variance.

CAFS = the Combined Assessment of Function and Survival.

CI = confidence interval.

LOCF = last observation carried forward.

LS Mean = least-squares means.

MMRM = mixed model for repeated measures.

SE = standard error.

^a LOCF was applied to the patients who completed cycle 3 (reached 81 days after treatment initiation).

^b LOCF was applied to all randomised patients.

^c One patient in the placebo group, who lacked ALSFRS-R post baseline data needed for MMRM analysis was excluded.

Distribution of changes in ALSFRS-R total score

The distributions of changes in ALFRS-R total score from baseline to the end of cycle 6 (ALL LOCF) in the placebo and the edaravone groups are shown in Figure 1. Changes in ALSFRS-R from baseline to the end of cycle 6 (week 24) tended to be fewer (closer to zero) with edaravone treatment than with placebo. Four patients (5.9%) in the placebo group and nine patients (13.0%) in the edaravone group maintained their functionality (0 point change [or 1 point increase]). Nine patients (13.2%) in the placebo group and 27 patients (39.1%) in the edaravone group showed unchanged functionality or a small (1 or 2 point) deterioration at the end of cycle 6. The distribution shift was statistically confirmed using the Hodges-Lehmann estimator (2, 95% CI 1.00–3.00) and Wilcoxon rank-sum test (p = 0.0034).

Figure 1. Distributions of changes in ALSFRS-R total score from baseline to the end of cycle 6 (ALL LOCF).

One dot represents one patient. ALSFRS-R = the Revised Amyotrophic Lateral Sclerosis Functional Rating Scale. LOCF = last observation carried forward.

Changes in each ALSFRS-R score items and 4 domains

Edaravone was descriptively favoured for each item of the ALSFRS-R at the end of cycle 6 (ALL LOCF) as shown in Figure 2. The largest difference (LS mean ± SE) was observed for item 9 climbing stairs (0.51 ± 0.15). Edaravone appeared to be descriptively favoured in all four domains of ALSFRS-R with the largest observed magnitude of difference in the gross motor domain (1.05 ± 0.32), followed by the bulbar (0.53 ± 0.28), fine motor (0.50 ± 0.33) and respiratory (0.28 ± 0.15) domains as shown in Figure 3(a).

Figure 2. Changes in each item of ALSFRS-R score from baseline to the end of cycle 6 (ALL LOCF).

ALSFRS-R = the Revised Amyotrophic Lateral Sclerosis Functional Rating Scale. LOCF = last observation carried forward. LS mean = least squares mean. SE = standard error.

Figure 3. (a) Changes in four domains of the ALSFRS-R score from baseline to the end of cycle 6 (ALL LOCF). (b) Changes in four domains of ALSFRS-R score from baseline to the end of cycle 6 in patients with limb onset (ALL LOCF). (c) Changes in four domains of ALSFRS-R score from baseline to the end of cycle 6 in patients with bulbar onset (ALL LOCF).

ALSFRS-R = the Revised Amyotrophic Lateral Sclerosis Functional Rating Scale. LOCF = last observation carried forward. LS mean = least squares mean SE = standard error.

When we investigated four-domain data stratified by the onset of disease (bulbar versus limb), patients with bulbar onset on placebo showed the largest decline in bulbar function while patients with limb onset on placebo showed the largest decline in gross or fine motor function. However, regardless of the onset, edaravone appeared to be descriptively favoured in all four domains of the ALSFRS-R (Figure 3(b, c)).

Discussion

The LOCF method may underestimate deteriorating function in an active treatment group if patients in that group are withdrawing from the study at a greater rate (or withdraw earlier) than those in a comparator group, resulting in a bias against the estimate of treatment effect (8). In the present study, however, the discontinuation rate before the end of cycle 6 in the 24-week double-blind period was low (edaravone 2.9%, placebo 11.8%) and, therefore, the effect using alternative statistical methods for missing data did not differ greatly from the previous estimate calculated for the primary analysis. ANOVA using ALL LOCF might here be expected to be a conservative approach (underestimating the difference between groups) because more patients in the placebo group than in the edaravone group discontinued before the study end. Sensitivity analyses with ANOVA using data imputation applying the ALL LOCF approach and MMRM analyses without data imputation showed similar results to those of the original prespecified comparison.

The CAFS, a composite measure of ALSFRS-R change and death was used as the primary endpoint analysis in the EMPOWER study of dexpramipexole in ALS as reported by Cudkowicz et al. (7). Although no deaths occurred during 24 weeks of the current study, the CAFS was chosen for our post-hoc analysis, being a well-accepted non-parametric method for use as a sensitivity analysis for the primary analysis. The CAFS analysis, which used all available data in this study, likewise showed a benefit of edaravone after 24 weeks of treatment compared with placebo.

The distributions of changes in ALSFRS-R total score (ALL LOCF) from baseline to the end of cycle 6 appeared to shift with edaravone treatment, with 39% of patients in the edaravone group showing minimal functional decline (+1 to –2 change) after six cycles compared with 13% of patients in the placebo group. In this population, defined by the inclusion and exclusion criteria of the study, the overall distribution shift suggests that the effect of slowing decline in ALSFRS-R, the primary efficacy endpoint in this study, was driven by the majority of patients responding to treatment rather than a response to treatment that was limited to a specific subtype of patient.

Edaravone was descriptively favoured at the end of six cycles (using ALL LOCF to impute missing data) for all 12 items of the ALSFRS-R and for all of the four domains of the ALSFRS-R. The smallest difference was in the respiratory subdomain, an effect that might have been due to a study entry requirement of normal respiratory function. Interestingly, the effect of edaravone on four domains appeared to be similar in patients with bulbar onset and limb onset. As such, the differences between placebo and edaravone in change from baseline to the end of cycle 6 (ALL LOCF) in ALSFRS-R total score appeared to be similar in patients with bulbar onset (2.42 ± 1.40, LS mean ± SE) and in patients with limb onset (2.31 ± 0.88). Several limitations must be considered in a post-hoc analysis such as this. First, these results were demonstrated in a subpopulation well-defined by the inclusion and exclusion criteria of this study (3). Therefore, the efficacy of edaravone in a wider ALS population that would include patients with advanced disease was not investigated in this study. Secondly, there were no deaths during the 24-week study period because the current study enrolled patients with high levels of general baseline functionality and, therefore, no effect on survival could be measured within 24 weeks. Thirdly, in conducting a series of sensitivity analyses, no adjustments were made for statistical multiplicity. Despite this limitation, results of the sensitivity analyses appearing to show treatment benefit of edaravone were consistent with those of the prespecified primary analysis (3).

In conclusion, the consistent results of post-hoc analyses using different approaches demonstrate statistical robustness of the prespecified primary analysis. In addition, slowing in functional decline over 24 weeks was found across all anatomical regions assessed. This observation would be consistent with its putative neuroprotective effects against the development of oxidative damage unspecific to anatomical regions.

Declaration of interest

Mr. Takahashi is an employee of Mitsubishi Tanabe Pharma Corporation (MTPC). Mr. Takei, Dr. Liu, Ms. Tsuda are employees of Mitsubishi Tanabe Pharma Development America (MTDA). Dr. Palumbo is an employee of MTPC and MTDA. The edaravone (MCI-186) clinical trials were funded by Mitsubishi Tanabe Pharma Corporation. The ALSFTD supplement, Edaravone (MCI-186) in amyotrophic lateral sclerosis (ALS), was funded by Mitsubishi Tanabe Pharma America, Inc. The authors alone are responsible for the content of this paper.

Acknowledgements

We thank David Hartree under contract with Mitsubishi Tanabe Pharma America, Inc. for medical writing assistance during second and subsequent drafts of the manuscript. We thank David Schoenfeld of Massachusetts General Hospital for statistical advice on MMRM for sensitivity analysis and Benjamin Rix Brooks of Carolinas HealthCare System for advice for subdomain analysis.