Randomized phase 2 study of perampanel for sporadic amyotrophic lateral sclerosis

Eligibility criteria for registration included age 40–78 years, “clinically definite ALS”, “clinically probable ALS”, or “clinically probable-laboratory supported ALS” as specified in the revised El Escorial Airlie House diagnostic criteria [22]; a maximum sum of the three respiratory items of the ALS functional rating scale-revised (ALSFRS-R) scores of 12 points; ≤ 2 years from disease onset; and the ability to visit the study site for outpatient treatment. None of the patients had a family history of ALS. Inclusion criteria after the observation period were: the ALSFRS-R score during the observation period must have decreased by 2–5 points; patients had not initiated riluzole therapy during the observation period; patients had not undergone riluzole dose escalation or resumed administration of riluzole therapy after previous dose reduction or discontinuation; and patients had not newly initiated edaravone therapy during the observation period. Exclusion criteria were tracheostomy; noninvasive positive pressure ventilation; % forced vital capacity (%FVC) < 80%; progressive bulbar palsy type; cognitive impairment; severe renal, hepatic, cardiovascular, or hematological disease; malignancy; pregnancy or the possibility of becoming pregnant; participation in another clinical study within 12 weeks before starting the observation period; and prior or current perampanel therapy.

The trial was conducted at 12 sites from April 2017 to January 2020. It included a 12-week observation period, followed by a randomized, double-blind, placebo-controlled, three-arm, 48-week trial period. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guidelines.

Following completion of the 12-week observation period, the patients started study treatment with perampanel 2 mg or placebo tablet once a day. The number of tablets administered per day was increased over 4 weeks to 4 placebo tablets/day in the placebo arm, 2 placebo tablets and 2 perampanel 2 mg tablets/day in the 4 mg arm, and 4 perampanel 2 mg tablets/day in the 8 mg arm. The full doses were administered daily for 44 weeks.

ALSFRS-R score, manual muscle testing (MMT) grading scale [23], and %FVC measurements were scheduled at screening and at day 0, 4 weeks, every 12 weeks, and at the followup visit. The tested muscles were as follows: bilateral biceps brachii, triceps brachii, quadriceps femoris, and hamstrings. Risk factors for suicidal behavior evaluated by the Columbia Suicide Severity Rating Scale [24] were reported at screening, and following every visit. The placebo and active tablets (Eisai Co., Ltd., Ibaraki, Japan) were quality-assured, packaged, labeled, stored at room temperature, and distributed to each study site.

The primary efficacy outcome was the change in ALSFRS-R score from baseline to week 48 after randomization. The major secondary efficacy outcomes were ALSFRS-R total score (the sum of all 12 items), the ALSFRS-R score for each item, the change in ALSFRS-R score for each item, MMT grading scale [23], %FVC at 4 weeks and then every 12 weeks, and time to death or disease progression after starting the treatments. Disease progression was defined as an event as follows: inability to walk without assistance, loss of function of both upper limbs, tracheostomy, respirator use, tube feeding, or death.

Safety outcomes were determined by site investigators reporting adverse events (AEs), vital sign measurements, body weight, laboratory tests, and risk factors for suicidal behavior. AEs were reviewed by all investigators.

The sample size in this study for each group was calculated referring to the results of an edaravone study [25]. That phase 3 study, comparing edaravone with placebo in ALS patients, reported a change of − 5.0 ± 5.3 (mean ± SD) in ALSFRS-R in the edaravone arm (n = 68) and − 7.5 ± 5.4 in the placebo arm (n = 66) (p = 0.0013) [25]. This study was designed to detect a dose–response relationship in changes in ALSFRS-R based on the hypothesis that perampanel would demonstrate benefits on a level similar to that of edaravone (a 3 ± 5-point maximum difference). A sample size of at least 20 patients was needed for each treatment group to have 70% power to detect a dose–response relationship with a one-sided significance level of 5% in the ALSFRS-R scores among the three groups at 48 weeks.

For the primary efficacy analysis, ALSFRS-R scores were analyzed for all patients in the intention-to-treat population. The means of changes from baseline values were calculated and analyzed using a restricted maximum likelihood (REML)-based repeated measures approach combined with the Newton Raphson algorithm. The analytical model included fixed categorical effects of treatment, visit, and treatment-by-visit interaction [26]. Akaike Information Criteria were used to choose the working correlation structures for modeling the within‐patient errors among an unstructured covariance structure, autoregressive (1) structure, and a compound symmetry structure. The Kenward–Roger approximation was used to estimate denominator degrees of freedom. The least square means of change from baseline in ALSFRS-R score during double-blind treatment and its 95% CIs were estimated and compared among the three groups. The contrast coefficients corresponding to dose–response relationships (linear dose–response relationship, dose–response curve peaking at 4 mg, dose–response curve demonstrating efficacy only at 8 mg, and dose–response curve demonstrating efficacy only at 4 mg) were tested at a one-sided significance level of 5%. The contrast test with the smallest p value was selected as the true dose–response relationship among the three groups. As a sensitivity analysis for the primary analysis, analysis of covariance was also performed for the primary endpoint after the missing data were imputed by last observation carried forward methods. Secondary endpoints were analyzed similarly through the linear mixed-effects model for repeated measures. Time to disease progression after starting the treatments, were plotted using the Kaplan–Meier method and compared among groups using log-rank tests stratified by the minimization factors. A trajectory analysis [27] was performed for exploratory analysis to identify any subgroups appearing over the trajectory of ALSFRS-R scores from baseline to 48 weeks.

All analyses, except for the primary analysis, were two-sided with a p value < 0.05 considered statistically significant. Analyses were performed with SAS software, version 9.4 (SAS Institute, Cary, NC, USA). PROC TRAJ macro was used for the trajectory analysis (http://​www.​andrew.​cmu.​edu/​user/​bjones/​) [28]. This study is registered with ClinicalTrials.gov, number NCT03019419.

The intention-to-treat-population included 65 patients. The baseline demographics and clinical characteristics were similar among the three groups (Table 1). Approximately 75% of the patients were diagnosed as having “clinically probable ALS”, or “clinically probable-laboratory supported ALS”. Most of the patients were being treated with riluzole (88%) and/or edaravone (80%).

The discontinuation rate of the trial increased in a dose-dependent manner, with 18 of 22 patients (82%) in the placebo group, 14 of 22 patients (64%) in the 4 mg perampanel group, and 7 of 21 patients (33%) in the 8 mg perampanel group completing the 48-week trial. Of the 26 patients who did not complete the trial, one patient died of ALS, 11 had AEs, 11 declined to participate further, and 3 did not meet continuation criteria (one patient: impossible to take tablet orally; 2 patients: unable to visit the study site) (Fig. 1).

Evaluation of the dose–response relationship did not reach statistical significance (one-sided p = 0.463 for a dose–response curve demonstrating efficacy only at 4 mg, Table S-1). The estimated mean change of ALSFRS-R total scores from baseline to week 48 was − 9.0 in the placebo group (n = 18), − 13.4 in the 4 mg perampanel group (n = 14), and − 17.4 in the 8 mg perampanel group (n = 7) (p = 0.01 for the comparison between the placebo group and the 8 mg perampanel group) (Table 2). The changes in the ALSFRS-R score from the baseline were larger in the 8 mg perampanel groups compared to the placebo group at 12, 36, and 48 weeks (Fig. 2a; Table 3). The change of bulbar subscores in the 8 mg perampanel group were significantly higher than those of the control group (p = 0.0206); there were no significant differences in upper limb, lower limb, or respiratory subscores among placebo and perampanel groups (Table 2). The Kaplan–Meier curve for time to disease progression did not differ among groups (log-rank; p = 0.276, Fig. 2b). In addition, the ALS Milano-Torino Staging (MiToS) system [29], which correlates well with functional stage and QOL in ALS, was retrospectively evaluated using the ALSFRS-R score. MiToS staging at week 48 was not significantly different among the three groups (data not shown).

In the subgroup analysis (changes in ALSFRS-R scores during the observation period, sex, age, and the use of riluzole or edaravone), the patients with small but significant decreases (− 3 to − 2) in ALSFRS-R score during the observation period (p = 0.0467), and those taking riluzole and/or edaravone (p = 0.0374), had a significant decrease in the score if they went on to take 8 mg perampanel, suggesting high-dose perampanel may promote disease progression in patients taking riluzole and/or edaravone. (Tables S-2 and S-3).

The changes in ALSFRS-R scores varied more than expected within each group, and some patients in each group experienced changes in ALSFRS-R scores ranging from 3 to − 5 after randomization through the 48 weeks (Fig. 2c). Because of the wide variations in the changes of ALSFRS-R scores over the treatment period, trajectory analysis was performed in the placebo and perampanel groups. In the trajectory analysis, the perampanel groups were merged because of their high dropout rates. The results suggested three different patterns of progression (Fig. 2d). It is of note that the pattern of small changes (red line) in the perampanel group was nonprogressive as compared to the corresponding pattern in the placebo group that showed a steady decline. On the other hand, the pattern of large changes (blue line) in the perampanel groups showed a faster decline than that in the placebo group. To elucidate the differences in clinical features of subsets according to the trajectory analysis, we retrospectively compared the changes in ALSFRS-R scores during the observation period and the baseline characteristics between nonprogressive and progressive subsets on trajectory analysis (i.e. the small change subset vs. the intermediate and large change subset) (Table S-4). In the placebo group, the small change subset (nonprogressive) showed a higher total ALSFRS-R score at baseline (p = 0.0407) and smaller changes in score during the observation period (p = 0.0105) than the progressive subset (Table S-4). In the perampanel group, the small change subset (nonprogressive) had lower subscores in the lower limbs in ALSFRS-R at baseline (p = 0.0081) and longer duration from disease onset (p = 0.0007) than the progressive subset (Table S-4).

Changes in MMT grading from baseline to week 48 showed no significant difference among the three groups except that the deterioration in MMT grading of the left hamstring muscles was significantly smaller in the 8 mg perampanel group than that in the placebo group (Table 2). When comparing the placebo group to the perampanel group (4 mg and 8 mg), the change in MMT grading of the hamstrings on both sides was significantly smaller in the perampanel group than in the placebo group (Fig. 3). There was no statistically significant difference in the mean decline of %FVC after 48 weeks among the three groups (Table 2).

The percentage of patients with reported serious AEs was 13.6% in the placebo group, 27.3% in the 4 mg perampanel group, and 47.6% in the 8 mg perampanel group (Table 4), more frequent in the 8 mg perampanel group than in the placebo group (p = 0.0483). One patient in the 4 mg perampanel group died of ALS. Treatment-related AEs such as dizziness and excessive sleepiness were reported more frequently in the 8 mg perampanel group (16 events) and 4 mg perampanel group (11 events) than in the placebo group (one event). There were no significant differences in the results of laboratory tests, vital signs, ECGs, or chest X-rays among the three groups.