Efficacy of Omaveloxolone in Friedreich's Ataxia: Delayed‐Start Analysis of the MOXIe Extension

MOXIe was a two‐part study evaluating the safety and efficacy of omaveloxolone in patients with Friedreich's ataxia, a rare, progressive neurological disease with no proven therapy. MOXIe part 2, a randomized double‐blind placebo‐controlled trial, showed omaveloxolone significantly improved modified Friedreich's Ataxia Rating Scale (mFARS) scores relative to placebo. Patients who completed part 1 or 2 were eligible to receive omaveloxolone in an open‐label extension study.

A BS TRACT: Background: MOXIe was a two-part study evaluating the safety and efficacy of omaveloxolone in patients with Friedreich's ataxia, a rare, progressive neurological disease with no proven therapy. MOXIe part 2, a randomized double-blind placebocontrolled trial, showed omaveloxolone significantly improved modified Friedreich's Ataxia Rating Scale (mFARS) scores relative to placebo. Patients who completed part 1 or 2 were eligible to receive omaveloxolone in an open-label extension study. Objective: The delayed-start study compared mFARS scores at the end of MOXIe part 2 with those at 72 weeks in the open-label extension period (up to 144 weeks) for patients initially randomized to omaveloxolone versus those initially randomized to placebo. Methods: We performed a noninferiority test to compare the difference between treatment groups (placebo to omaveloxolone versus omaveloxolone to omaveloxolone) using a single mixed model repeated measures (MMRM) model. In addition, slopes of the change in mFARS scores were compared between both groups in the open-label extension.
Results: The noninferiority testing demonstrated that the difference in mFARS between omaveloxolone and placebo observed at the end of placebo-controlled MOXIe part 2 (À2.17 AE 1.09 points) was preserved after 72 weeks in the extension (À2.91 AE 1.44 points). In addition, patients previously randomized to omaveloxolone in MOXIe part 2 continued to show no worsening in mFARS relative to their extension baseline through 144 weeks.

Introduction
Friedreich's ataxia (FRDA) is a rare genetic neurodegenerative disorder affecting an estimated 5000 patients in the USA and 22,000 patients globally. 1 People with FRDA develop progressive difficulty with ambulation, coordination, and speech, losing the ability to walk, on average, 10 to 15 years after disease onset. 2,3 Ultimately, FRDA shortens life expectancy; the average age of death is approximately 37.5 years. 4,5 Currently, no approved disease-modifying therapy is available for the treatment of FRDA.
The pathophysiology of FRDA reflects mitochondrial dysfunction, impaired Nrf2 signaling, and decreased energy (adenosine triphosphate) production. [6][7][8][9] Omaveloxolone, a potent activator of Nrf2, restores mitochondrial function ex vivo in fibroblasts from people with FRDA. 10 Clinical data from the MOXIe study (ClinicalTrials.gov: NCT02255435) demonstrated that omaveloxolone treatment induces Nrf2 and exhibits pharmacological activity based on Nrf2-inducible biomarkers such as ferritin in those with FRDA. 1,11 This study includes two placebo-controlled parts and an open-label extension for patients who completed part 1 or part 2. Part 1 was a placebo-controlled, doseranging study that enrolled 69 individuals, whereas part 2 was a multicenter, randomized, placebo-controlled clinical trial that enrolled 103 people at 11 study sites in the USA, Europe, and Australia. 1,11 The open-label extension enrolled 149 patients (87% of those enrolled in part 1 or part 2) and remains ongoing.
In the prespecified primary analysis population for MOXIe part 2, treatment with omaveloxolone significantly improved neurological function, as measured by modified Friedreich's Ataxia Rating Scale (mFARS) scores, by À2.40 points relative to placebo at week 48 (P = 0.014; n = 82). 1 To evaluate the persistence of omaveloxolone's treatment effect, we performed a delayed-start analysis comparing the difference in mFARS scores at the end of the 48-week placebocontrolled period with the difference after 72 weeks in the open-label extension.

Participants
People in MOXIe were 16 to 40 years of age with genetically confirmed FRDA and baseline mFARS scores between 20 and 80. Individuals were excluded if they had uncontrolled diabetes, clinically significant cardiac disease, active infections, significant laboratory abnormalities, or interfering medical conditions. Those who developed diabetes or cardiac morbidity, such as arrhythmias, remained in the study unless they chose to withdraw. Subjects who completed MOXIe part 2, including 48 weeks of treatment and a follow-up safety visit at 52 weeks (4 weeks after last dose), could enroll in the open-label extension study (up to 144-168 weeks at March 24, 2022 database lock).
We defined the analysis populations for the present analysis in a manner consistent with MOXIe part 2; the full analysis set (FAS) included individuals without pes cavus, and the all-randomized population (ARP) included all who enrolled in part 2 of the trial. 1 Omaveloxolone-omaveloxolone refers to those who were randomized to omaveloxolone in MOXIe part 2 and then continued with omaveloxolone in the open-label extension. Placebo-omaveloxolone refers to the set of people originally randomized to placebo in part 2 who then initiated treatment with omaveloxolone in the open-label extension. Thus, people in the placebo-omaveloxolone group began treatment with omaveloxolone 52 weeks after those in the omaveloxolone-omaveloxolone group (Fig. 1).
All subjects provided written informed consent.

General Design
Day 1 of the extension was the same as the last visit in MOXIe part 2 (week 52). All people in the open- label extension received 150 mg omaveloxolone once daily. Subjects were scheduled for mFARS assessments in the extension at day 1 and every 24 weeks during treatment thereafter. Both subjects and examiners remained blinded to their original treatment group in the randomized and placebo-controlled part 2 throughout the open-label extension; thus, no difference in expectation bias should occur. Data were accrued through March 2022.
The primary endpoint was the difference in the initial placebo and initial omaveloxolone groups in the "delayed-start period" (extension week 72 change from baseline mFARS values) versus the initial omaveloxolone À placebo difference in the "placebocontrolled period" (MOXIe part 2 week 48 change from baseline mFARS values). This analysis used the FAS. As a sensitivity analysis, efficacy was assessed using the ARP.
We performed a noninferiority analysis using a single mixed model repeated measures (MMRM) model that included all available data from both the 48-week placebo-controlled period (part 2) and the open-label, delayed-start period (open-label extension) through extension week 144 (total follow-up week 196). 12,13 Analysis using MMRM without imputation assumes data are missing at random. This MMRM included treatment, time, and the interaction of treatment and time as fixed factors, as well as baseline mFARS (from part 2), study site, and the interaction of baseline mFARS and time as covariates. The difference between treatment groups (omaveloxolone-omaveloxolone versus placebo-omaveloxolone) was estimated using the MMRM model estimates at the end of the placebocontrolled period (Δ1; part 2, week 48) and in the openlabel, delayed-start period (Δ2; extension week 72). A decrease in mFARS represents improved function. Therefore, negative values for Δ1 and/or Δ2 represent improved function in the omaveloxolone-omaveloxolone versus placebo-omaveloxolone groups. The noninferiority test used a noninferiority margin equal to 50% of Δ1 (H0: Δ2 À Δ1 ≥ À0.5Δ1 versus Ha: Δ2 À Δ1 < À0.5Δ1). 12,13 The alternative hypothesis is equivalent to Δ2 < 0.5Δ1, which implies that the treatment effect at the end of the delayed-start period (Δ2) preserved more than 50% of the treatment effect after the placebo-controlled period (Δ1) or Δ2 did not lose more than 50% of Δ 1 . The upper bound of a one-sided 90% confidence interval [CI] was used for the noninferiority test. For this analysis, we first considered an unstructured covariance structure; however, the model did not converge. Therefore, a Toeplitz covariance structure, which assumes mFARS change from baseline values within a subject are correlated over time, was used.
To further characterize the rate of change in mFARS during open-label treatment, we used a random coefficients mixed model to fit change from baseline mFARS using all available data from the extension study through extension week 144 to estimate annualized slopes based on the part 2 randomized treatment group. The model included terms for baseline (day 1 of the extension), treatment, time, the interaction between treatment and time, and a random intercept.
All statistical analyses were performed using SAS (v9.4; SAS Institute, Cary, NC, USA). For safety data, the assessment of relation to drug is defined by the investigator, based on known effects of drug, temporal relationship to drug, and other factors.

Data Sharing
Data collected for the study, including individual patient data, will not be made available.

Disposition and Baseline Characteristics
From the FAS (n = 82), 73 individuals enrolled in the extension study, including 39 people who were randomized to placebo in MOXIe part 2 (placeboomaveloxolone group) and 34 randomized to omaveloxolone in MOXIe part 2 (omaveloxoloneomaveloxolone group). At extension week 72, subjects completed a combined total follow-up of up to 124 weeks (2.4 years) when accounting for the initial 52 weeks of follow-up in the double-blind study (Fig. 1). Within the FAS, 75 subjects had mFARS assessments at week 48 in MOXIe part 2, and 31 had mFARS assessments at extension week 72. A notable number of mFARS values were missing at weeks 48 and 72 of the extension study, primarily because of coronavirus disease 2019 (COVID-19) pandemic-related interruptions in on-site clinic visits. However, most patients remained in the study and in the extension study beyond week 72 and subsequently returned to the clinic for mFARS assessments in greater numbers. Only 10 patients in the FAS discontinued treatment during the extension study, including 8 in the placebo-omaveloxolone group and 2 in the omaveloxolone-omaveloxolone group.
Demographics and baseline characteristics were generally similar across treatment groups. Select demographic and baseline characteristic data for the FAS are summarized in Table 1. The mean baseline age AE standard deviation (SD) was 23.6 AE 7.8 years in the placebo-omaveloxolone group and 24.2 AE 6.5 years in the omaveloxolone-omaveloxolone group. The majority of subjects were non-Hispanic/Latino white individuals in both treatment groups. In the placeboomaveloxolone group, 28/42 (67%) subjects were male, and 13/42 (31%) were less than 18 years old at baseline. These values are approximately two times higher than the values summarizing the number of male (16/40, 40%) and pediatric (7/40, 18%) individuals in the omaveloxolone-omaveloxolone group. The mean age AE SD at FRDA onset was 15.1 AE 5.3 years in the placebo-omaveloxolone treatment group and 15.9 AE 5.7 years in the omaveloxolone-omaveloxolone treatment group. The majority of subjects were ambulatory in both the placebo-omaveloxolone (93%) and the omaveloxolone-omaveloxolone (93%) treatment groups. The mean baseline mFARS score was 38.8 in the placebo-omaveloxolone group and 40.9 in the omaveloxolone-omaveloxolone group. The mean GAA1 (shorter of the two FXN intron 1 GAA repeats) repeat length (with longer repeat length associated with worse disease severity) at baseline was 694 in the placebo-omaveloxolone group and 739 in the omaveloxolone-omaveloxolone group. More omaveloxolone-omaveloxolone subjects (19/40 [48%]) had a history of cardiomyopathy than those in the placebo-omaveloxolone group (12/42 [29%]). Thus, although the distribution of most baseline characteristics was similar between treatment groups, compared with the placebo-omaveloxolone group, the omaveloxoloneomaveloxolone group had slightly more advanced disease, with higher average baseline mFARS scores, longer GAA1 repeat lengths, and a greater proportion of individuals with a history of cardiomyopathy.

Efficacy
Results of the noninferiority testing using a single MMRM model demonstrated that the difference in  mFARS between omaveloxolone and placebo observed at the end of the placebo-controlled portion (LS mean difference = À2.17 AE 1.09) was preserved at the end of the delayed-start period (LS mean difference = À2.91 AE 1.44). The upper limit of the 90% CI for the test statistic (Δ 2 -0.5 Â Δ 1 ) was less than zero (À0.09), demonstrating significant evidence of noninferiority (Table 2). Similar trends were observed in the ARP, although the noninferiority criteria were not met ( Table 2). The graphical representation of changes from baseline in mFARS for omaveloxolone and placebo groups shows that the separation at the end of the placebocontrolled period is maintained in the open-label period (through extension week 144). Apart from the data at extension week 48, which had many missing mFARS assessments because of COVID-19 pandemic-related interruptions, nearly parallel trajectories were seen between the placebo-omaveloxolone group and the omaveloxolone-omaveloxolone group (Fig. 2). The difference at extension week 72 persisted through week 120 and in the limited number of subjects who have presently returned for week 144. In addition, the mFARS scores of people in the omaveloxoloneomaveloxolone group were maintained relative to their extension baseline through extension week 120.
Annualized slopes during the open-label period using all available data from the extension study through week 144 for the omaveloxolone-omaveloxolone group was 0.45 AE 0.63 (95% CI: À0.82, 1.71) and the placebo-omaveloxolone group was 0.76 AE 0.28 (95% CI: 0.21, 1.31). There was no evidence of a significant difference between the two groups (difference: À0.31 AE 0.71, 95% CI: À1.72, 1.10; P = 0.66). There was no evidence of convergence between the two groups. The annual slope for each group in MOXIe Extension is less than the expected worsening of about two points per year based on natural history data. 14 The resulting parallel trajectories between both treatment groups demonstrate a lack of convergence between the groups that is consistent with omaveloxolone treatment altering disease course.

Safety
The longer-term safety profile of omaveloxolone in the extension study was similar to that seen in MOXIe Parts 1 and 2, and omaveloxolone was generally well tolerated in the extension study. No deaths were reported. Serious adverse events were reported in 13 (8.7%) patients; of these, 8 (7.5%) individuals were in the placeboomaveloxolone group and 5 (11.6%) were in the omaveloxolone-omaveloxolone group. All of the serious adverse events were considered by the investigator to be unrelated to study drug, and none resulted in permanent discontinuation of study drug (Table 3). An increased number of upper respiratory infections was noted in part 1, but not part 2, of MOXIe and were not clearly noted in the extension cohort, although there was no clear control group to precisely define this. The most common treatment-emergent adverse events were increased alanine aminotransferase and coronavirus infection. Other adverse events occurring in ≥10% of patients in either group and treatment-emergent serious adverse events are shown in Table 3. Elevations in aminotransferase increases were not associated with elevations in total bilirubin, and no subject met Hy's law criteria.

Discussion
The results of the delayed-start analyses indicate a persistent benefit of omaveloxolone treatment on disease course. Those who received omaveloxolone during the placebo-controlled, double-blind period (MOXIe part 2) experienced a sustained benefit that could not In addition, the delayedstart analysis suggests that the improvement in neurological function persists to some degree with ongoing therapy, suggesting that earlier treatment with omaveloxolone might provide greater benefit than delayed therapy in this study population. The ongoing difference in groups in the later time points of the delayed-start analysis supports this idea. A challenge in conducting studies of investigational therapies for mitochondrial diseases is the complicating factor of hope bias. Although the possible presence of such bias existed in the placebo-omaveloxolone group during the extension study (an early placebo effect in mFARS was observed in MOXIe part 2 that was maximal by week 12 but resolved by week 36), the noninferiority criterion was still met using the MMRM, which included all data up through 144 weeks of the extension study. The 72-week duration of the delayed-start period for the primary noninferiority analysis also allowed those with a delayed start enough time to experience potential symptomatic effects of omaveloxolone, because the maximal effects of omaveloxolone on subjects in the placebocontrolled MOXIe part 2 occurred at week 24.
In the open-label phase, the safety profile was similar to that seen in MOXIe Parts 1 and 2, without any new safety risks identified. 1,11 The later portion of the extension study was conducted during the global COVID-19 pandemic. Because the mFARS examination is an inperson assessment, this led to a large number of missed mFARS assessments. This primarily impacted study visits occurring at or beyond 48 weeks in the extension study, but a substantial majority of subjects had at least one inperson visit during the later time points. In addition, one major reason for loss of visits was the inability to travel to the primary site, a reason that most likely is random. Thus, missing visits most likely lead to a loss of sensitivity but are unlikely to systematically bias the study.
Limitations of the study include missing mFARS data, small sample size, and the possibility of unmeasured confounding effects. Collectively, the safety and efficacy results of this study support that omaveloxolone is generally safe and well tolerated and indicate a persistent effect of omaveloxolone on disease course in FRDA.