CADTH Reimbursement Review

Cenobamate (Xcopri)

Sponsor: Paladin Labs Inc.

Therapeutic area: Epilepsy, partial onset seizures

This multi-part report includes:

Clinical Review

Pharmacoeconomic Review

Stakeholder Input

Clinical Review

Executive Summary

An overview of the submission details for the drug under review is provided in Table 1.

Table 1: Background Information of Application Submitted for Review

NOC = Notice of Compliance.

Introduction

Epilepsy is a chronic neurologic disorder that affects the physical and mental health of patients and significantly interferes with daily activity as well as life expectancy.¹ The broad categories of epileptic seizures include partial onset (also known as focal), generalized, combined focal and generalized, and unknown onset.² In patients with partial onset seizures, only a portion of the brain, typically 1 lobe of 1 hemisphere, is affected by the seizure, while in patients with generalized seizures, large parts of both brain hemispheres are involved.³ The estimated prevalence of active epilepsy is 5.96 per 1,000 population (95% confidence interval [CI], 5.38 to 6.61) based on a meta-analysis of international studies.⁴ It is estimated that 300,000 people in Canada are living with epilepsy.⁵

Antiseizure medications (ASMs) are the most common treatments for seizures and are prescribed by physicians after discussing the risks and benefits of the medication with the patient or their caregiver. There are many different ASMs currently available in Canada. These differ in their mechanism of action, potential adverse-effect profiles, the types of seizures they are best at treating, and cost. Although ASMs help to control or reduce seizures, these drugs are not a cure for epilepsy. While the aim of treatment with ASMs is to eliminate seizures with no adverse effects, this may not be achieved in all patients;⁶ approximately 20% to 40% of patients are at risk of having refractory epilepsy.^7,8 Epilepsy is considered to be medically refractory (or drug-resistant) when a patient fails to achieve sustained freedom from seizures after adequate trials of 2 tolerated ASMs, either as monotherapy or in combination.⁷ Treatment options for patients with refractory epilepsy may be limited.

The objective of this report is to review and critically appraise the evidence submitted by the sponsor on the beneficial and harmful effects of cenobamate oral tablets, 12.5 mg to 200 mg, as adjunctive therapy in the management of partial onset seizures in adults who are not satisfactorily controlled with conventional therapy.

Stakeholder Perspectives

The information in this section is a summary of input provided by the patient and clinician groups who responded to CADTH’s call for input and from the clinical expert consulted by CADTH for the purpose of this review.

Patient Input

Six groups, including the Canadian Epilepsy Alliance, Epilepsy Toronto, Epilepsy South Central Ontario, Epilepsy Southwestern Ontario, the Epilepsy Association of Calgary, and the Edmonton Epilepsy Association, provided patient input for this review. These patient groups indicated that uncontrolled seizures and the adverse effects of ASMs affect patients’ daily activities, independence (e.g., they are not legally permitted to drive), and mental health (e.g., they experience higher risks of depression, anxiety, and suicidal ideation). The patient groups noted that whole families are affected. Patients are often unemployed or under-employed and negatively affected by other social determinants of health. The patient groups noted that the most important treatment outcome is seizure freedom, with an alternative expectation of reduced seizure frequency and/or severity. Patients and their families were also highly concerned about adverse effects of ASMs and interactions between these drugs. The patient groups indicated that new drugs could offer hope to patients who are close to giving up, and even a reduction in the absolute number of seizures can potentially improve overall quality of life.

Clinician Input

Input From the Clinical Expert Consulted by CADTH

The expert stated that treatment options are limited for the approximately 30% of patients with focal epilepsy who do not respond to ASMs, and that while ASMs may reduce seizure frequency, none offer a cure for epilepsy. Tolerability of ASMs can be an issue, and a patient’s comorbidities and concomitant medications may contribute to the development of adverse effects. The clinical expert stated that cenobamate is best suited to patients with focal epilepsy who have not responded to conventional ASMs. The expert also anticipates that it will be used as second- or third-line therapy, typically as add-on therapy. Treatment response would be demonstrated by seizure freedom or a reduction in seizure frequency as well as improved quality of life and acceptable tolerability. According to the clinical expert, patients prescribed cenobamate should be under the care of a neurologist or epileptologist in a community or hospital setting.

Clinician Group Input

One clinician group provided input for this review: the Canadian League Against Epilepsy, which has more than 125 health care-related members. The group highlighted that there is an unmet need for more effective treatments for patients with uncontrolled, focal onset seizures. Despite the availability of several ASMs, there has been no meaningful improvement in epilepsy treatment-related outcomes and no significant increase in seizure freedom rates in the past 20 years. The clinician group noted that cenobamate would likely be used in combination with other available treatments (i.e., as an add-on) and is unlikely be used as a monotherapy. However, if cenobamate proves to prevent seizures once added, a physician will occasionally try to minimize a patient’s ongoing treatment by weaning them off other ASMs. There was consistency between the views of the clinical expert consulted by CADTH and the Canadian League Against Epilepsy with regard to how response to treatment is assessed, reasons for discontinuing therapy, the treatment setting, and the specialists required to diagnose, treat, and monitor patients who may receive cenobamate.

Drug Program Input

Input was obtained from the drug programs that participate in the CADTH reimbursement review process. The following were identified as key factors that could potentially affect the implementation of a CADTH recommendation for cenobamate:

• relevant comparators

• consideration for initiating therapy

• consideration for prescribing therapy.

The clinical expert consulted by CADTH provided advice on the potential implementation issues raised by the drug program.

Clinical Evidence

Pivotal Studies and Randomized Controlled Trial Evidence

Description of Studies

The pivotal and randomized controlled trial (RCT) evidence included 2 double-blind (DB), placebo-controlled RCTs of add-on therapy with cenobamate in adults with uncontrolled partial seizures despite ongoing treatment with 1 to 3 ASMs.^9,10 In Study C013, 222 patients were randomized to cenobamate 200 mg daily or placebo; in Study C017, 437 patients were randomized to cenobamate 100 mg, 200 mg, or 400 mg once daily or placebo, in addition to the ASMs initiated before enrolment. Each study included an 8-week baseline period before randomization, a 6-week dose titration period, and a 6-week (Study C013) or 12-week (Study C017) maintenance period in which the dose of study drug remained stable. The primary end point was the percent change in seizure frequency per 28 days versus baseline for all simple partial motor, complex partial, and secondarily generalized seizures. The key secondary outcome was the proportion of patients who achieved at least a 50% reduction in partial seizure frequency versus baseline. In both studies, these end points were calculated based on the entire DB treatment period (titration and maintenance phases) and used to support regulatory approval in the US. Study C017 conducted alternate analyses based on seizure data from the maintenance phase only, and these data were used to support regulatory approval in Europe.

The mean age of patients enrolled in the pivotal trials ranged from 36.2 years (standard deviation [SD] = 11.3 years) to 40.9 years (SD = 12.4 years) across treatment groups. There were roughly equal proportions of men (47% to 54%) and women (46% to 53%) enrolled. In Study C013, the median baseline seizure frequencies per 28 days were 5.5 (range = 2 to 237) in the placebo group and 7.5 (range = 0 to 187) in the cenobamate 200 mg group. In Study C017, the median baseline seizure frequencies per 28 days ranged from 8.4 (range = 4 to 704) for placebo to 11.0 (range = 4 to 418) for the cenobamate 200 mg group. On average, the enrolled patients had been diagnosed with epilepsy for more than 20 years. Patients in Study C017 had previously been treated with a median of 3 prior ASMs (range = 1 to 9).

The CADTH review focused on data for the cenobamate 200 mg to 400 mg dosage range, as per the draft product monograph.

Efficacy Results

For the primary end point in Study C013, the cenobamate 200 mg group showed a median 55.6% reduction in partial seizure frequency per 28 days compared with a 21.5% reduction in the placebo group (P < 0.0001) (Table 2). In Study C017, the median percent reductions in seizure frequency per 28 days were 55.0%, 55.0%, and 24.0% in the cenobamate 200 mg, 400 mg, and placebo groups, respectively (with P < 0.001 favouring cenobamate versus placebo for both dosage groups).

For the responder analysis in Study C013, 50.4% of patients in the cenobamate 200 mg group and 22.2% of patients in the placebo group achieved at least a 50% reduction in seizure frequency during the DB treatment period (odds ratio = 3.94; 95% CI, 2.14 to 7.24; P < 0.0001; not controlled for type I error rate). In Study C017, 57.8%, 60.4%, and 21.7% of patients in the cenobamate 200 mg, 400 mg, and placebo groups, respectively, achieved at least a 50% reduction in seizure frequency per 28 days during the DB treatment phase. The analyses favoured the cenobamate groups versus placebo (both P < 0.001; not controlled for type I error rate) (Table 2). The alternate analyses, which were based on treatment response in the maintenance period only, reported that 56.1%, 64.2%, and 25.5% of patients achieved at least a 50% reduction in seizure frequency per 28 days in the cenobamate 200 mg, 400 mg, and placebo groups, respectively. For this alternate primary end point, the differences favoured the cenobamate 200 mg (P < 0.001) and 400 mg groups (P < 0.001) versus placebo.

The proportion of patients who achieved reductions of greater than or equal to 75%, greater than or equal to 90%, and 100% in seizure frequency during the DB period (Table 2) and maintenance period favoured the cenobamate 200 mg and 400 mg groups versus placebo (Appendix 1); however, there was no control of type I error rate for these analyses in either study, and in Study C013, these analyses were conducted post hoc. While these results are generally supportive of the efficacy of cenobamate, these data should be interpreted in light of the potentially inflated risk of type I error and risk of bias associated with post hoc analyses.

Of note, Study C017 also included a 100 mg cenobamate dosage group, which is half the Health Canada–recommended maintenance dose. Both the percent reduction in seizure frequency per 28 days and the proportion of patients who achieved at least a 50% reduction in seizure frequency favoured the cenobamate 100 mg group versus placebo; however, other secondary outcomes, such as the higher responder thresholds, failed to detect a difference between groups (Table 2).

Health-related quality of life (HRQoL) was not assessed in Study C013, and only descriptive data were available in Study C017 for approximately one-quarter of the patients enrolled. No meaningful change in HRQoL was observed in Study C017, based on data from the patient-weighted 31-item Quality of Life in Epilepsy Questionnaire (QOLIE-31-P).

Other outcomes of interest to patients, such as functional status, were not assessed, nor were seizure-free days or treatment retention, which were outcomes specified in the sponsor’s protocol.

Harms Results

During the 12-week to 18-week treatment periods, adverse events (AEs) were reported by 65%, 76%, and 90% of patients who received cenobamate 100 mg, 200 mg, and 400 mg, compared with 63% to 70% of patients who received placebo in studies C013 and C017. The most frequently reported AEs in the cenobamate groups were somnolence (19% to 37%), dizziness (18% to 33%), fatigue (11% to 24%), and diplopia (4% to 15%). In the placebo groups, somnolence was reported in 8% to 12% of patients, dizziness in 14% to 17% of patients, fatigue in 6% to 8% of patients, and diplopia in 2% to 3% of patients.

In Study C013, 5 patients (4%) in the cenobamate group and 3 patients (3%) in the placebo group stopped treatment due to AEs, whereas in Study C017, 11 patients (10%), 15 patients (14%), 22 patients (20%), and 5 patients (5%) stopped treatment due to AEs in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively.

No deaths occurred during the pivotal trials. Serious adverse events (SAEs) were reported in 3% to 9% of patients who received cenobamate and in 4% to 6% of patients who received placebo. One patient in the cenobamate 200 mg group of Study C017 developed a drug reaction with eosinophilia and systemic symptoms (DRESS) on day 24 that the investigator considered probably related to the study drug; the study drug was stopped. Other dermatologic reactions that led to study drug withdrawal were reported in 2 patients in the cenobamate 200 mg group. In Study C013, 1 patient who was randomized to the cenobamate group experienced a serious drug hypersensitivity reaction on day 1. The patient stopped cenobamate treatment and recovered after 22 days. In Study C013, suicidal ideation was reported by 1 patient in each treatment group, and in Study C017 by 2 patients in the cenobamate 100 mg group, and by 1 patient in the cenobamate 200 mg group. One patient in the cenobamate 100 mg group in Study C017 attempted suicide.

Table 2: Summary of Key Results From Pivotal Studies and RCT Evidence

ANCOVA = analysis of covariance; DB = double blind; CEN = cenobamate; mITT = modified intention to treat; RCT = randomized controlled trial; SAE = serious adverse event.

^aIn Study C013, the P value is based on a Wilcoxon rank sum test assessing if the median percent change for CEN is significantly different than for placebo. The Study C017 P value is based on a nonparametric ANCOVA model with terms for ranked baseline seizure rate and treatment group. The mITT population included all randomized patients who received at least 1 dose of the study drug and had any postbaseline seizure data.

^bPrimary end point in Study C013; no multiplicity adjustment required.

^cPrimary end point for the US regulatory agency in Study C017; the P value for each dosage group has been adjusted for multiple testing.

^dIn Study C013, the P value is based on logistic regression model with terms for treatment, country, and baseline seizure frequency (Wald chi-square test). In Study C017, the P value is based on Fisher’s exact chi-square test.

^eIn Study C013, greater than or equal to 75%, greater than or equal to 90%, and 100% of responder analyses were conducted post hoc.

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

Critical Appraisal

The risks of bias related to randomization, treatment allocation, and blinding in the pivotal trials were rated as low by the CADTH reviewer. At baseline, the patient characteristics appeared to be reasonably well balanced between groups within the studies. In Study C013, the proportions of patients who discontinued were similar in both groups (10% and 9%), but Study C017 showed differential losses to follow-up, with 12%, 18%, 27%, and 13% of patients stopping treatment in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively. As a result, it is not clear if the treatment groups remained balanced throughout the study. Both trials measured the change in partial seizure frequency as well as the proportion of responders (≥ 50%) as primary and secondary end points. The analyses included all simple partial motor, complex partial, and secondarily generalized seizures; the clinical expert agreed that this was appropriate for this population. While the 50% response threshold has been used in other ASM clinical trials and may be accepted as a minimal clinically important difference, the clinical expert indicated that higher response thresholds are desired, and the goal of therapy is seizure freedom. Other seizure response thresholds were tested in both studies, but these analyses were not controlled for type I error rate; in Study C013, these were conducted post hoc. Moreover, the responder analyses were conducted using the last observation carried forward for Study C013 and with no imputation for missing data in Study C017; thus, patients who withdrew early could be considered as treatment responders. Considering that patients who drop out are likely to have worse outcomes than those who continue, it is possible that the results of Study C017 may be biased in favour of cenobamate, due to the extent of the early withdrawals and the differential losses. However, the magnitude of any potential bias and the impact on the overall findings is unclear.

Neither of the pivotal trials was designed to test the impact of cenobamate on HRQoL or on patients’ ability to work or maintain independence. Although Study C017 collected data using the QOLIE-31-P instrument, this information was gathered only from approximately 25% of the patients enrolled and was reported descriptively. Moreover, the trials lacked an active comparator group and were 12 weeks to 18 weeks in duration; thus, the trials can only address short-term efficacy and safety versus placebo. The sample sizes and durations of the studies were insufficient to capture rare AEs.

No major issues were identified by the clinical expert on the generalizability of the pivotal studies. However, it should be noted that both studies had extensive exclusion criteria and were limited to adults up to 70 years of age. The trials used a more rapid titration schedule than has been recommended in the product monograph, which may have affected the occurrence of some AEs.

Long-Term Extension Studies

Description of Studies

Additional longer-term safety data were available from 2 single-arm, open-label extension (OLE) studies: Study C013 OLE and Study C017 OLE.^11,12 Patients who completed the randomized phase of Study C013 were eligible to enter the OLE phase (N = 149) and received open-label cenobamate at a daily maximum dose of 400 mg daily for up to 8.6 years. Only safety outcomes were assessed. Patients who completed Study C017 were eligible to enter its OLE phase (N = 356) and were transitioned to open-label cenobamate at a target dose of 300 mg once daily. In this phase, efficacy outcomes of seizure control up to 48 months and safety outcomes up to 6.4 years were assessed.

Efficacy Results

For the Study C017 OLE, Klein et al. (2022) reported interim efficacy outcomes based on a median exposure duration of 53.9 months (range = 1.1 months to 68.7 months), with retention rates at 12 months, 24 months, 36 months, and 48 months of 83%, 71%, 65%, and 62%, respectively.¹³ Among patients who remained on cenobamate, the treatment effects appear to be maintained for up to 48 months. The median 65.4% reduction in partial seizure frequency versus baseline was reported during the first 6 months of the OLE (interquartile range [IQR] = 52.0%; N = 354), with a 76.1% reduction (IQR = 44.8%) at month 43 to month 48 (N = 213).¹³ Of the 354 patients who entered the OLE, 10.2% achieved a 100% reduction in partial seizure frequency in the 36-month to 48-month interval.¹³

Harms Results

In both OLE studies, 89% of patients reported 1 or more treatment-emergent adverse events (TEAEs); dizziness (34% to 36%), somnolence (22% to 24%), and headache (17% to 28%) were the most common. Overall, 9% of patients stopped treatment due to AEs, and 22% to 26% of patients experienced an SAE. There were no cases of DRESS or any serious skin or subcutaneous tissue disorders reported during the OLEs. Four deaths (3%) were reported in the Study C013 OLE; these were due to sudden unexpected death in epilepsy (SUDEP), cardiac arrest, respiratory arrest, and completed suicide. In the Study C017 OLE, 6 patients (2%) died after experiencing myocardial infarction, cardiogenic shock, SUDEP, completed suicide, or sepsis.

Critical Appraisal

Limitations of the OLE studies include selection bias, lack of a control group, and lack of blinding. Given that completion of a pivotal trial was an eligibility criterion for the extension study, patients who discontinued those trials due to AEs or lack of response were excluded. This could have resulted in a population of patients who were more tolerant of cenobamate, which could lead to both a response bias (because those not responding to treatment are less likely to continue) and biased estimates related to AEs, potentially resulting in fewer and less severe AEs being reported. Without comparator groups, the interpretation of the results in relation to an appropriate comparator (e.g., another ASM) is limited. Unblinding of the cenobamate treatment in the OLE can bias the reporting of end points, particularly for any subjective measures, including AEs.

The sample sizes in the Study C013 OLE (N = 149, with 25% completing the study) and Study C017 OLE (N = 356, with 20% completing the study) may not be sufficient to detect rare AEs. In addition, there was wide variance in the follow-up durations for individuals. More common forms of morbidity (e.g., cardiac dysrhythmias) may not be easily identified as related to drug exposure.

Indirect Comparisons

Descriptions of Indirect Treatment Comparisons

Two indirect treatment comparisons (ITCs) were summarized and critically appraised for CADTH’s review of cenobamate. ITC 1 was a sponsor-submitted network meta-analysis (NMA). It was designed to assess the relative efficacy and safety of adjunctive therapy with cenobamate compared to brivaracetam, perampanel, lacosamide, eslicarbazepine acetate, and zonisamide in adult patients with partial onset seizures.^14,15 ITC 2 was an NMA conducted by the National Institute for Health and Care Excellence (NICE) to evaluate the relative efficacy and safety of ASMs used as add-on treatments for patients with partial onset seizures.¹⁶ Relevant treatments in ITC 2 included brivaracetam, carbamazepine, cenobamate, eslicarbazepine acetate, gabapentin, ganaxolone, lacosamide, lamotrigine, levetiracetam, oxcarbazepine, perampanel, phenytoin, pregabalin, primidone, retigabine, rufinamide, sodium valproate, topiramate, vigabatrin, and zonisamide.¹⁶

Efficacy Results

The sponsor-submitted NMA (i.e., ITC 1) included 22 studies for the combined efficacy outcome of a 50% or greater responder rate and seizure freedom.^14,15 The results suggest that patients receiving cenobamate may be more likely to achieve a 50% or greater reduction in seizures, or seizure freedom in the short-term, compared to those receiving any of the 5 ASMs in the NMA.^14,15

The ITC by NICE (ICT 2) included 99 studies for the analysis of seizure frequency reduction of greater than 50% reduction and 72 studies for the seizure freedom analysis.¹⁶ The results suggest that patients who received cenobamate may be more likely to achieve a greater than 50% reduction in seizure frequency in the short-term than patients on most other treatments, including eslicarbazepine acetate, retigabine, gabapentin, lacosamide, lamotrigine, levetiracetam, oxcarbazepine, perampanel, brivaracetam, pregabalin, rufinamide, zonisamide, and primidone.¹⁶ Relative treatment effects were not available for seizure freedom due to rarity of the event and insufficient statistical power.¹⁶

Harms Results

The sponsor-submitted NMA included 20 studies for the safety outcome of TEAEs leading to treatment discontinuation.^14,15 Compared to cenobamate, results for the 5 other ASMs had relatively wide 95% credible intervals (CrIs) that included the threshold of no difference; however, this does not indicate equivalence between ASMs for this outcome.^14,15 Due to the lack of precision in the estimates, firm conclusions cannot be made.

Harms data were reported numerically (not comparatively) by treatment in ITC 2 and do not inform relative safety for cenobamate versus other ASMs.¹⁶

Critical Appraisal

The main limitations of both ITCs were the inclusion of nonadult patients (cenobamate is indicated for adults), missing relevant comparator ASMs (mainly in ITC 1), and heterogeneity in patient and study characteristics across studies. Moreover, the limited reporting of trial data hindered the CADTH reviewers’ ability to assess the consistency assumption, particularly for ITC 2, and neither NMA conducted analyses to adjust for potential treatment-effect modifiers. Differences across studies in how efficacy outcomes were defined (e.g., seizure freedom) and the treatment period upon which these were based (i.e., the entire treatment period or excluding the dose titration period) were assessed as important potential sources of bias. The durations of the trials were variable (7 weeks to 24 weeks) and limited to a short time frame for this chronic condition. Given these limitations, there was uncertainty in the results of the NMAs that prevented firm conclusions from being drawn from the findings. Furthermore, there was a lack of safety data in the ITCs and no information about HRQoL or long-term outcomes reported; thus, it is not possible to estimate how cenobamate compares to other ASMs for these outcomes.

Studies Addressing Gaps in the Pivotal and RCT Evidence

Description of Studies

The sponsor submitted 1 prospective safety study (Study C021)¹⁷ and 3 additional studies to address gaps in the evidence.^18-20

Study C021 was a single-arm safety study of cenobamate as adjunctive therapy in patients with uncontrolled partial onset seizures despite treatment with 1 to 3 ASMs (N = 1,340). It consisted of a 12-month open-label treatment period that included a 12-week titration phase followed by an open-label maintenance phase, with a target daily dose of 200 mg and a maximum daily dose of 400 mg.¹⁷

The 3 additional studies submitted by the sponsor included Elizebath et al. (2021), a post hoc data analysis of HRQoL outcomes from all the patients enrolled in studies C013, C017 and C021 at a single US centre (N = 49);¹⁸ Connor et al. (2022), a real-world evidence study among adults with a developmental disability (N = 28);¹⁹ and Elliott et al. (2022), a real-world evidence study in patients with partial epilepsy who received cenobamate (N = 45, 13 of whom were adolescents aged 12 years to 17 years).²⁰

Efficacy Results

Elizebath et al. (2021) reported that the mean of overall QOLIE-31 scores (range = 1 to 100, with higher scores indicating better quality of life) measured for all patients who completed treatment was 67 (SD = 19; range = 32 to 97; n = 37) over a median treatment period of 5.6 years (range = 3 years to 8 years).¹⁸ Connor et al. (2022) reported that adjunctive cenobamate treatment, the mean number of focal seizures was reduced from 20.9 seizures per month at baseline to 4.1 seizures per month at 6 months’ follow-up (median = from 3.0 seizures at baseline to 0.5 seizures per month at 6 months’ follow-up); at 6 months, the greater than or equal to 50%, greater than or equal to 75%, greater than or equal to 90%, and 100% responder rates were 92.6%, 81.5%, 55.6%, and 48.2%, respectively.¹⁹ Elliott et al. (2022) reported that 60% of all the included patients achieved a reduction in seizures greater than or equal to 50%, and that among the 13 adolescents, 8 (61.5%) achieved a reduction in seizures greater than or equal to 50%.²⁰ No efficacy results were reported in Study C021.

Harms Results

In Study C021, TEAEs occurred in 91% of patients. The most frequently reported TEAEs were somnolence (31%), dizziness (29%), fatigue (19%), and headache (18%). Serious TEAEs occurred in 18% of the patients; the most frequently reported serious TEAE was seizures (2%). One hundred and 83 patients (14%) had at least 1 TEAE leading to discontinuation, most frequently due to dizziness (1.4%) and seizures (0.8%). At the data cut-off date, 10 deaths had been reported in Study C021, with causes of laryngospasm, glioblastoma, subdural hematoma, SUDEP, sudden death, traumatic intracranial hemorrhage, hypovolemic shock, pneumonia viral, status epilepticus, and cardiac arrest.¹⁷ One case of sudden death was deemed by the investigator to be remotely related to the study drug; the remainder were deemed to be unrelated.¹⁷ No case of DRESS was reported.¹⁷

Connor et al. (2022) reported that AEs occurred in 32% of patients and included dizziness (14%), drowsiness (117%), ataxia (7%), and behavioural “acting out” (4%).¹⁹ Elliott et al. (2022) reported that the most frequently reported AE was somnolence (18%).²⁰

Critical Appraisal

Limitations of the other studies addressing the evidence gaps include the relatively small sample sizes for the observational studies, selection bias, lack of a control group, lack of blinding, and probable lack of control for confounding in the studies. Due to these factors, CADTH could not determine whether the results of these studies were valid or reliable.

Conclusions

The direct evidence demonstrates that add-on therapy with cenobamate is superior to placebo in reducing the frequency of partial seizures in the short-term in adults whose partial onset seizures are inadequately controlled with up to 3 concomitant ASMs. In addition, a higher percentage of patients who received cenobamate appeared to achieve at least a 50% reduction in seizure frequency compared to placebo. While a 50% reduction in seizure frequency is an accepted end point for clinical trials, it may not be clinically relevant for all patients, given that the goal of therapy is the elimination of seizures. The pivotal studies examined higher seizure reduction thresholds (i.e., ≥ 75%, ≥ 90%, and 100%); however, due to statistical limitations (i.e., lack of control of the type I error rate) and risk of bias (i.e., analyses were conducted post hoc in Study C013), definitive inferences cannot be drawn from these data.

Direct comparative evidence for cenobamate versus other ASMs was unavailable. The indirect evidence from 2 NMAs suggest there may be a short-term benefit favouring cenobamate versus some ASMs for the proportion of patients achieving at least a 50% reduction in seizure frequency; however, the magnitude and clinical relevance of any benefit was unclear because of important limitations in the analyses. There was heterogeneity across the networks in the patient characteristics, in how outcomes were defined and analyzed, and in the durations of follow-up, creating uncertainty in the findings of the NMAs. Due to the uncertainty in the sponsor-submitted NMA, no conclusions could be drawn about the relative effects of cenobamate on seizure freedom.

The impact of cenobamate on HRQoL or other outcomes of importance to patients, such as the ability to work or live independently, is unknown because the placebo-controlled trials were not designed to test for these outcomes. Neither NMA assessed HRQoL or longer-term outcomes.

No conclusions could be drawn regarding the short-term comparative safety of cenobamate versus other ASMs due to the previously described limitations and the lack of precision in the estimates from the NMA. No new safety signals were identified from the longer-term single-arm studies; however, long-term comparative data were not available.

Introduction

The objective of this report is to review and critically appraise the evidence submitted by the sponsor on the beneficial and harmful effects of cenobamate 12.5 mg to 200 mg oral tablets as adjunctive therapy for the management of partial onset seizures in adults who are not satisfactorily controlled with conventional therapy.

Disease Background

The contents of this section have been informed by materials submitted by the sponsor and by clinical expert input. The following paragraphs have been summarized and validated by the CADTH review team.

Epilepsy is a chronic neurologic disorder that affects the physical and mental health of patients and significantly interferes with daily activity as well as life expectancy.¹ According to the International League Against Epilepsy, epilepsy is defined by any of the following conditions: at least 2 unprovoked seizures occurring more than 24 hours apart; 1 unprovoked seizure and a probability of further seizures that is similar to the general recurrence risk (i.e., ≥ 60% probability of further seizures) after 2 unprovoked seizures occurring over the next 10 years; or diagnosis of an epilepsy syndrome.²¹ The broad categories of epileptic seizures include partial onset (also known as focal), generalized, combined focal and generalized, and unknown onset.² In patients with partial onset seizures, only a portion of the brain, typically 1 lobe of 1 hemisphere, is affected by seizures, while in patients with generalized seizures, large parts of both brain hemispheres are involved.³ A partial onset seizure can be characterized as motor (with muscle activity changes) or nonmotor (without muscle activity changes) type.² Loss of consciousness (impaired awareness) is 1 of the symptoms in complex partial onset seizures (also known as focal impaired awareness) and generalized seizures.³

As the second most frequent neurologic disorder worldwide (after migraine), the prevalence of epilepsy is high.¹ The estimated prevalence of active epilepsy is 5.96 per 1,000 population (95% CI, 5.38 to 6.61), based on a meta-analysis of international studies published from 1985 to 2011.⁴ Prevalence estimates of active epilepsy in Canada range from 4 in 1,000 people based on the weighted Canadian Community Health Survey data (from 2010 to 2011) to 6.4 per 1,000 based on the Canadian Chronic Disease Surveillance System data (from 2019 to 2020).^4,22 It is estimated that 300,000 people in Canada are living with epilepsy.⁵ Approximately 20,000 Canadians (5,000 children and youths, and 15,000 adults)were diagnosed with epilepsy in 2013 to 14, or on average, 54 people per day.⁵ The incidence of epilepsy is highest in the first year of life and declines with increasing age.^1,4

Epilepsy is considered to be medically refractory epilepsy (also called drug-resistant epilepsy or intractable epilepsy) when a patient fails to achieve sustained seizure freedom after adequate trials of 2 tolerated ASMs, either as monotherapy or in combination.⁷ Approximately 20% to 40% of patients with epilepsy are at risk of having refractory epilepsy.^7,8

Standards of Therapy

Contents within this section have been informed by materials submitted by the sponsor and clinical expert input. The following paragraphs have been summarized and validated by the CADTH review team.

The main goals of therapy are to eliminate seizures or reduce their frequency and to improve patients’ safety, quality of life, and ability to work, attend school, drive, and maintain independence. Other goals are to reduce burden on caregivers, prevent premature death and SUDEP, avoid delayed remission, reduce AEs, and avoid invasive therapies (e.g., surgery).

ASMs are the most common treatment for seizures and are prescribed by physicians after discussing the risks and benefits of the medications with patients or caregivers. There are many different ASMs currently available within and outside Canada. These differ in terms of their mechanism of action, potential adverse-effect profiles, the types of seizures they are best at treating (i.e., focal versus generalized versus epileptic spasms), and cost. ASMs help to reduce seizure frequency, but are not a cure for epilepsy. Although the aim of treatment with ASMs is “no seizures, no adverse effects,” this may not be achieved in all patients.⁶ Monotherapy with an ASM is ideal; however, some individuals may require more than 1 ASM to control seizures.^23,24

The NICE guidelines²⁵ on the management of epilepsy recommend the following:

• Use a single antiseizure medication (monotherapy) to treat epilepsy whenever possible.

• Review the diagnosis of epilepsy if seizures continue despite an optimal dose of a first-line antiseizure medication.

• If first-line monotherapy is unsuccessful and epilepsy diagnosis remains confirmed, try monotherapy with another antiseizure medication, using caution during the changeover period:

  • Increase the dose of the second medicine slowly while maintaining the dose of the first medicine.

  • If the second medicine is successful, slowly taper off the dose of the first medicine.

  • If the second medicine is unsuccessful, slowly taper off the dose of the second medicine and consider an alternative.

• If monotherapy is unsuccessful, consider trying an add-on treatment.

• When starting an add-on treatment, carefully titrate the additional medicine and review treatment frequently, including monitoring for adverse effects, such as sedation.

• If trials of add-on treatment do not result in a reduction in seizures, use the regimen that provides the best balance between effectiveness and tolerability of side effects.²⁵

Approximately half of epilepsy patients will gain control of their seizures with an appropriate first ASM; in general, two-thirds of patients will be seizure-free on ASMs. However, more than one-third do not respond to ASMs and continue to have uncontrolled seizures.²⁶ Of those patients, only 5% per year will enter seizure remission as a result of changing from 1 current ASM to another.²⁷ According to the sponsor and clinical expert input, there are limited treatment options available for patients who are inadequately controlled using current ASMs.²⁸ Moreover, suboptimal seizure control with ASMs exposes these patients to increased risks (associated with uncontrolled seizures) and decreased HRQoL. There is a high risk of SUDEP in patients with uncontrolled seizures, particularly for those with focal to bilateral tonic-clonic seizures.^29,30

The clinical expert consulted for this review indicated that all the ASMs shown in Table 3 may be used as adjunctive therapy in patients with focal epilepsy. Other options for adjunctive therapy include off-label use of rufinamide or felbamate and zonisamide. These are not currently approved in Canada, but may be prescribed through the Health Canada Special Access Program. Nonpharmaceutical options for patients with treatment-resistant epilepsy include epilepsy surgery, neurostimulation, or ketogenic diet. Epilepsy surgery is an invasive and costly treatment that can be offered only to select patients with a single identifiable seizure focus located in an area of the brain that can be safely removed.³¹ Neurostimulation treatments are invasive, costly, palliative options that can reduce seizure frequency, but these rarely lead to seizure freedom.^32,33 The ketogenic diet is a palliative, arduous treatment that is poorly tolerated by adults.³⁴

Drug Under Review

Key characteristics of cenobamate are summarized in Table 3 along with other treatments available for partial onset seizures.

The precise mechanism of the therapeutic effects of cenobamate in patients with partial onset seizures is not known.³⁵ Cenobamate functions as a positive allosteric modulator of gamma-aminobutyric acid type A receptors at nonbenzodiazepine-binding sites.^36-38 Cenobamate also decreases excitatory currents by inhibiting the persistent component of the sodium current and enhancing the inactivated state of voltage-gated sodium channels.

Cenobamate is approved by Health Canada as adjunctive therapy in the management of partial onset seizures in adults with epilepsy who are not satisfactorily controlled with conventional therapy.³⁵ The usual maintenance dose of cenobamate is 200 mg once daily, and the maximum daily dose — if needed, based on clinical response and tolerability — is 400 mg once daily.³⁵ The dose of cenobamate should be titrated up in 2-week intervals (because of the potential for SAEs), starting with 12.5 mg once daily. It is recommended that discontinuation of cenobamate be undertaken gradually over a period of at least 2 weeks to minimize the potential for rebound seizures, unless safety concerns require abrupt withdrawal.³⁵

The dossiers for cenobamate were submitted to CADTH as a pre-Notice of Compliance submission, with Notice of Compliance received on June 12, 2023. Cenobamate underwent a standard review at Health Canada, and the sponsor has requested reimbursement as per the approved Health Canada indication.³⁹ Cenobamate has not been previously reviewed by CADTH. It received approval from the FDA in November 2019.⁴⁰

Table 3: Key Characteristics of Cenobamate and Other Drugs for Partial Onset Seizures

GABA = gamma-aminobutyric acid; GABA_A = gamma-aminobutyric acid type A; SV2A = synaptic vesicle glycoprotein.

^aHealth Canada–approved indication(s).

Source: Product monographs.^35,41-59

Clinical Evidence

The objective of CADTH’s clinical review is to review and critically appraise the clinical evidence submitted by the sponsor on the beneficial and harmful effects of cenobamate 12.5 mg, 25 mg, 50 mg, 100 mg, 150 mg, and 200 mg oral tablets as adjunctive therapy for the management of partial onset seizures in adults who are not satisfactorily controlled with conventional therapy. The focus will be on comparing cenobamate to relevant comparators and identifying gaps in the current evidence.

A summary of the clinical evidence included by the sponsor in the review of cenobamate is presented in 4 sections. CADTH’s critical appraisal of the evidence is included after each section. The first section, the Systematic Review, includes pivotal studies and RCTs that were selected according to the sponsor’s systematic review protocol. The second section includes sponsor-submitted, long-term extension studies. The third section includes indirect evidence from the sponsor. The fourth section includes additional studies that were considered by the sponsor to address important gaps in the pivotal and RCT evidence.

Included Studies

Clinical evidence from the following are included in the CADTH review and appraised in this document:

• Two pivotal trials included in the systematic review

• Two long-term extension studies

• Two ITCs

• One prospective safety study and 3 additional studies addressing gaps in the evidence.

Of note, the sponsor included 3 trials in its summary of the pivotal and RCT evidence (studies C013, C017, and C021). However, Study C021, an open-label single-arm safety study, was not considered a pivotal trial by Health Canada and did not meet the study design criteria for inclusion in the systematic review section of the CADTH report. Study C021 was summarized in the Studies Addressing Gaps in the Pivotal and RCT Evidence section of this report because it provides additional safety data as well as evidence to support the dose titration schedule listed in the product monograph.

Pivotal Studies and RCT Evidence

Contents within this section have been informed by materials submitted by the sponsor. The following information has been summarized and validated by the CADTH review team.

Description of Studies

Characteristics of the included studies are summarized in Table 4.

Table 4: Details of Pivotal Studies and RCT Evidence Identified by the Sponsor

ASM = antiseizure medication; BMI = body mass index; CNS = central nervous system; DB = double blind; EEG = electroencephalogram; QOLIE-31-P = patient-weighted 31-item Quality of Life in Epilepsy Questionnaire; RCT = randomized controlled trial; ROW = rest of world.

Note: Four additional reports were included: FDA Clinical Review,⁶² FDA Statistical Review,⁶³ Chung et al. (2020),⁶⁰ and Krauss et al. (2020).⁶¹

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

Two pivotal, phase II, DB, randomized, placebo-controlled trials (studies C013 and C017) met the inclusion criteria for the sponsor’s systematic review.

The primary objective of study C013 was to evaluate the efficacy of cenobamate when titrated to 200 mg per day from 50 mg per day in reducing seizure frequency in adults with partial onset seizures whose seizures are not fully controlled despite treatment with 1 to 3 concomitant ASMs. The secondary objectives were to assess the safety and tolerability of cenobamate. Patients who met the enrolment criteria underwent a pre-treatment period to determine the baseline rate of partial seizures. This baseline period could include data from the patient’s own seizure diary completed before enrolment (for 4 weeks) plus another 4 weeks’ prospective data capture, or 8-weeks of prospective follow-up, as part of the baseline study period. Patients were eligible for randomization if they experienced at least 3 seizures per 28 days during the baseline period and must not have a seizure-free interval greater than 21 days despite ongoing therapy with 1 to 3 ASMs. Allocation to treatments was conducted using an interactive web response system based on a computer-generated randomization schedule stratified by country. A total of 222 patients were assigned 1:1 to either cenobamate 200 mg or placebo once daily for 12 weeks (including a 6-week dose titration period and a 6-week maintenance period). The primary outcomes were the percent changes in partial seizure frequency relative to baseline. The study was conducted between July 2011 and June 2013 and included patients from the US, India, Republic of Korea, and Poland.

The objective of Study C017 was to determine the effective dose range and the safety and tolerability of cenobamate as adjunctive therapy for the treatment of partial seizures in adults. The study design for Study C017 is shown in Figure 1. Patients who met the enrolment criteria underwent an 8-week baseline pre-treatment period to assess the frequency of partial seizures. Patients were eligible for randomization if they experienced at least 8 seizures during the baseline period despite ongoing therapy with 1 to 3 ASMs. They must have experienced at least 3 partial seizures during each of the 2 consecutive 4-week baseline periods and must not have had a seizure-free interval greater than 25 days. A total of 437 patients were randomized (1:1:1:1) to placebo, cenobamate 100 mg, cenobamate 200 mg, or cenobamate 400 mg once daily for a total of 18 weeks (a 6-week dose titration period, then a 12-week maintenance period). The computer-generated randomization sequence was based on permuted blocks of 4 stratified by country, with allocation completed centrally using an interactive web response system. The primary outcomes were the percent change in partial seizure frequency relative to baseline and the proportion of patients with at least a 50% reduction in partial seizure frequency. The trial was conducted from July 2013 to June 2015 in Europe, Asia, US, New Zealand, Australia, and South Africa. No patients were enrolled in Canada.

For both studies, patients who completed the trials were eligible to enter an extension phase and receive open-label cenobamate.

Study C017 included a cenobamate 100 mg dosage group. This dosage is lower than the recommended target dose for cenobamate.³⁵ Given that cenobamate was still under review by Health Canada when this report was drafted, and that not all patients may be able to tolerate the 200 mg daily dose, it was determined that a summary of the 100 mg dose group was warranted; thus, these data have been included in CADTH’s clinical review.

Figure 1: Study C017 Design

DB = double blind; OLE = open-label extension.

Source: Clinical Study Report for Study C017.⁹

Populations

Inclusion and Exclusion Criteria

Both pivotal trials included adults who were diagnosed with partial epilepsy according to the International League Against Epilepsy’s classification of epileptic seizures and who had uncontrolled seizures despite treatment with 1 to 3 ASMs at stable doses for at least 12 weeks before screening. In Study C013, patients were required to meet the criteria for treatment-resistant partial epilepsy, defined as having failed treatment with at least 2 ASMs at doses that were at least 50% of the WHO–defined daily dose for at least 3 months.

To be eligible for enrolment in the pivotal trials, patients had to experience a minimum number of partial seizures during the 8-week baseline period:

• In Study C013, the minimum was at least 3 seizures per month (simple partial with motor component, complex partial, or secondarily generalized), with no 21-day seizure-free period.

• In Study C017, the minimum was at least 8 seizures (simple partial with motor component, complex partial, or secondarily generalized), without a seizure-free interval greater than 25 days, and at least 3 seizures per 4-week block.

Patients were excluded from the studies if they had other forms of epilepsy, seizure clusters, or had experienced status epilepticus in the past year. In addition, those with a history of allergic reactions to ASMs, impaired renal or liver function, clinically significant electrocardiogram abnormalities, or cardiac disorders were prohibited from entering the studies. Patients with a neurologic or central nervous system disease, psychiatric or psychotic illness, history of alcohol or drug abuse, or suicidality were also excluded. The prohibited medications are described in the Interventions section.

Interventions

Study Drug

In Study C013, patients were randomized to receive blinded cenobamate 200 mg once daily or placebo for a total of 12 weeks. The study drug was supplied as identical-looking capsules containing cenobamate or placebo in blister packages. The initial dose of cenobamate was 50 mg once daily. This was increased in 2-week intervals until the target dose (200 mg) was reached or until the patient experienced adverse effects (Table 5). Patients who reported adverse effects were not titrated to the higher dose until the next study visit, at which point the investigator had the choice to keep the patient on the same dose or to increase or decrease the dose. Patients who had doses reduced due to adverse effects could not be titrated to a higher dose at the next study visit. The study drug dose could not be titrated up during the maintenance phase, but dose reductions for tolerability were allowed.

In Study C017, patients were randomized to DB treatment with placebo, cenobamate 100 mg, 200 mg, or 400 mg once daily for a total of 18 weeks. All patients and study personnel were blinded to treatment allocation. The study drug was supplied as identical-looking tablets containing cenobamate 50 mg, cenobamate 100 mg, or placebo in blister packs. For the first 6 weeks, daily doses of the study drug were increased each week by 50 mg or 100 mg until the target dose was achieved (Table 5). Patients with significant tolerability issues during week 1 were discontinued from treatment. From week 2 to week 6, if patients did not tolerate the drug, they were allowed to reduce the dose to the previous week’s dosage for the next 7 days to 13 days, after which the upward titration could continue (until week 6) at the discretion of the investigator. If the new upward titration was not tolerated, the daily dose could be reduced by 50 mg (until the end of week 8). No further up-titration was allowed after week 6, and consecutive dose reductions were not allowed.

Those who completed the studies had the option to enter the OLE and were either directly converted to open-label cenobamate or underwent a 2-week DB dose conversion phase. Patients who withdrew or were not continuing in the OLE underwent a 1- or 3-week taper period, with the final visit conducted 2 weeks or 3 weeks after the last dose of study drug.

Concomitant Medications

During the DB treatment period in studies C013 and C017, patients continued taking the background 1 to 3 ASMs that they had been receiving for the 12 weeks before the study. Vagal nerve stimulation was not counted as an ASM; it was allowed to continue during the trial, provided patients’ parameters remained stable. Benzodiazepines taken at least once per week for epilepsy, anxiety, or sleep disorder were counted as 1 ASM.

Patients were allowed to receive intermittent benzodiazepines (other than diazepam) as rescue therapy once during the baseline period and twice during the treatment phase.

During the studies, the doses of the background ASMs were to remain stable. In Study C017, the investigator could modify the timing or amount of an individual dose of background ASMs to improve tolerability, but the total daily dosage and dosage frequency had to remain unchanged.

Prohibited Medications

During the trial and in the 30 days before the baseline phase, patients were prohibited from receiving diazepam, phenytoin, phenobarbital, or metabolites of these drugs, as well as vigabatrin (in the year before screening). Due to potential interactions between drugs, patients were prohibited from taking clopidogrel, fluvoxamine, amitriptyline, clomipramine, bupropion, methadone, ifosfamide, cyclophosphamide, efavirenz, natural progesterone (both trials), and omeprazole (Study C013 only).

Table 5: Dose Titration Schedule — Study C013 and Study C017

^aAt the start of Study C017, the titration schedule increased the dose in 100 mg increments. However, following a blinded review of tolerability, the protocol was amended, and doses were increased in 50 mg increments after the first 46 patients were treated.

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

Outcomes

A list of efficacy end points assessed in this clinical review report are provided in Table 6. These end points are further summarized in this section (Table 7). Summarized end points are based on those included in the sponsor’s Summary of Clinical Evidence along with any identified as important to this review according to stakeholders, such as the clinical expert, clinician groups, or patient groups.

In both studies, the seizure events included in the analyses were all simple partial motor (type B, or focal aware motor), complex partial (type C, or focal impaired awareness), and secondarily generalized seizures (type D, or focal to bilateral tonic-clonic), excluding simple partial seizures without a motor component (type A, or focal aware nonmotor). Data on the number and type of seizures experienced were recorded daily by patients using a seizure diary. Seizure diaries were collected and reviewed at each study visit, and new diaries were dispensed. The seizure rate was computed based on days without missing data, assuming that days with missing data would have the same seizure rate as days with nonmissing data. The baseline seizure rate was calculated by counting the number of seizures during the baseline period (56 days) and dividing the result by the number of days in the interval with nonmissing seizure data, then multiplying by 28. The same method was used to calculate the seizure rate during the DB treatment period. The percent change in seizure rate was calculated using the following formula: DB period seizure frequency per 28 days minus the baseline seizure frequency per 28 days, with the result divided by the baseline seizure frequency per 28 days and multiplied by 100. The responder analyses were based on the proportion of patients who achieved at least a 50%, 75%, 90%, or 100% reduction in seizure frequency versus baseline.

The primary outcome in Study C013 was the percent change from baseline in partial onset seizure frequency per 28 days for the DB treatment period (Table 6). The secondary outcome was the proportion of patients who achieved at least a 50% reduction in seizure frequency during the DB treatment period.

In Study C017, 2 sets of primary and key secondary outcomes were defined in the study’s protocol to meet the regulatory requirements of different countries (Table 6). The changes in seizure frequency or responder rate for Europe, Australia, New Zealand, and South Africa were calculated based on the subset of patients who entered the maintenance period (i.e., the modified intention-to-treat-in-maintenance-period [mITT-M] population) based on seizure events that occurred during maintenance therapy only (last 12 weeks). The seizure frequency or responder rates for the US and rest of the world were based on all patients (i.e., the modified intention-to-treat [mITT] population) using seizure data reported during the entire DB treatment period (induction and maintenance).

Both studies also analyzed other seizure response thresholds (defined by cut-offs of ≥ 75%, ≥ 90%, and 100% reduction in partial seizures) for the DB treatment period (mITT population), and for the maintenance period only in the subgroup of patients who took at least 1 dose of cenobamate during the maintenance period (i.e., subgroup 5 in Study C013). In Study C013, these analyses were conducted post hoc.

In Study C017, the QOLIE-31-P version 2.0 was completed by patients at baseline and at the end of the maintenance period. QOLIE-31-P captures overall quality of life, seizure worry, emotional well-being, energy and fatigue, cognitive, medication effects, and social functioning. All subscales and the total score range from 0 to 100, with higher scores indicating better function (Table 7). In Study C017, QOLIE-31-P was completed only by English-speaking patients in the US, UK, and Australia. In addition, patients were prohibited from completing the questionnaire if they were cognitively impaired and unable to understand the scale, were unable to read, or had not fully recovered from a recent seizure and were unable to reschedule the study site visit to complete the questionnaire. The sponsor defined the minimally important change as at an increase of least 11.8 points versus baseline in the weighted overall score.⁹ No HRQoL data were collected in Study C013.

In both studies, AEs were defined as any symptoms, signs, illnesses, or experiences that developed or worsened in severity during the study and were captured up to 30 days after the last dose of the study drug. In Study C017, seizures were not considered AEs unless these occurred with a measurable increase over the patients’ typical seizure frequency or duration, or included multiple seizures in a pattern that was distinguishable from the usual seizure pattern.

SAEs included events that were fatal, life-threatening, required or prolonged a hospital stay, resulted in persistent or significant disability or incapacity, congenital anomaly, or birth defect, or were an important medical event. The Columbia-Suicide Severity Rating Scale was used to assess for suicidal ideation or behaviour in both studies.

The sponsor listed seizure-free days, treatment retention or adherence, and tolerability as outcomes of interest in its systematic review protocol; however, these outcomes were not measured in either study. In addition, neither study assessed patients’ ability to work, attend school, or maintain independence, which were outcomes of importance to patient groups.

Table 6: Outcomes Summarized From Pivotal Studies and RCT Evidence Identified by the Sponsor

DB = double blind; HRQoL = health-related quality of life; mITT = modified intention to treat; mITT-M = modified intention to treat in maintenance period; NA = not applicable; QOLIE-31-P = patient-weighted 31-item Quality of Life in Epilepsy Questionnaire; RCT = randomized controlled trial; ROW = rest of world.

^aThe durations of the DB treatment period were 12 weeks in Study C013 and 18 weeks in Study C017. The maintenance periods were 6 weeks in duration for Study C013 and 12 weeks for Study C017, excluding the first 6 weeks of treatment, during which the dose of study drug was being titrated.

^bThe analysis of partial seizures was based on the total number of seizures (simple partial with motor component, complex partial [unique seizure types to be combined], and secondarily generalized tonic-clonic). Partial seizures without a motor component (i.e., type A) were not analyzed.

^cStatistical testing was adjusted for multiple comparisons using a hierarchical testing procedure for the cenobamate 200 mg, 400 mg, and 100 mg dose groups (in that order) vs. placebo.

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

Table 7: Summary of Outcome Measures and Their Measurement Properties

AE = adverse event; ASM = antiseizure medication; EEG = electroencephalogram; ILAE = International League Against Epilepsy; MID = minimal important difference; NA = not applicable; POS = partial onset seizure; QOLIE-31 = 31-item Quality of Life in Epilepsy Questionnaire; QOLIE-89 = 89-item Quality of Life in Epilepsy Questionnaire; QOLIE-31-P = patient-weighted 31-item Quality of Life in Epilepsy Questionnaire.

Statistical Analysis

In both studies, seizure frequency was calculated from the patient-reported seizure diary data based on days without missing data (i.e., both missing interim data from patients who continued in the trial, and missing data from a certain date for patients who dropped out before the end of the trial). It was assumed that days with missing data would have the same seizure rate as days with nonmissing data. The sponsor indicated that there were no missing data for the primary end point of Study C013, whereas in Study C017, end point data were missing for 6 patients (5.6%), 3 patients (2.7%), 6 patients (5.4%), and 4 patients (3.7%) in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively.⁸⁴

In Study C013, the primary outcome was tested based on a nonparametric Wilcoxon rank sum test (i.e., if the median percent change from baseline in the cenobamate was different from the median percent change from baseline in the placebo group). This model was selected after reviewing the results of prespecified model-fitting analyses that were run to understand the distributional characteristics of the seizure frequency data. Due to the nonparametric nature of the data, the Wilcoxon rank sum test was used for the primary analysis. The responder analyses were tested using a logistic regression model that included country and baseline seizure frequency as covariates (i.e., the Wald chi-square test). The study used the last observation carried forward for missing efficacy data. Sensitivity analyses were conducted based on the per-protocol and completer populations. There was no description in the CSR of any steps taken to control the type I error rate for the secondary and other end points.

In Study C017, the percent change in seizure rate per 28 days was analyzed using an analysis of covariance (ANCOVA) model for the ranked values of the primary end point, including the ranked baseline seizure rate as a covariate. Although randomization was stratified by country, country was not included as a covariate in the model. The CSR states that omitting country as a covariate did not affect the validity of the analysis and may not have reduced variability because each region was to capture patients who were relatively homogenous in response. Summary tables for the actual (not ranked) primary efficacy variable were presented, and a sensitivity analysis using a parametric ANCOVA model was performed. The responder rate was summarized using count and percentage of patients achieving at least a 50% response to treatment and analyzed using a Fischer’s exact chi-square test. Alternate responder thresholds (i.e., at least 75%, 90%, or 100% response) were analyzed using the same methods. Sensitivity analyses were conducted for the seizure frequency and responder analyses based on the titration period only and the first and last 6 weeks of the maintenance period only (additional details in Table 8). Sensitivity analyses were also conducted based on predefined regions based on their perceived level of epilepsy care and expertise.

In Study C017, the testing strategy for the primary efficacy end point was to compare each of the cenobamate dosage groups with the placebo group. Due to multiple treatment comparisons, a step-down procedure was used to ensure that the overall type I error rate was controlled at the 5% level. Each of the cenobamate dosage groups was compared with the placebo group at a 2-sided 0.05 level according to the following hierarchy: 200 mg group, then 400 mg group, then 100 mg group. First, the 200 mg dosage group was compared with the placebo group at a 2-sided 0.05 level. If no statistically significant difference was detected between the 200 mg dosage group and the placebo group, the procedure was to stop, and it would be concluded that none of the cenobamate dosages were efficacious. Testing proceeded through the hierarchy if the previous dosage group was statistically significant. The CSR and statistical analysis plan contained conflicting information about whether the key secondary outcomes were included in the hierarchical testing procedure. Based on CADTH’s interpretation of the information available, it was assumed that only the primary outcome was controlled for multiple testing.

An interim analysis was run in both studies, but there were no plans to modify the studies based on these data. These analyses were conducted by independent statisticians, and all study personnel were unaware of the results. The sponsor stated that there was no need to adjust the P value to account for this interim analysis.

Sample Size and Power Calculation

For the pivotal trials, a sample size of 100 patients per treatment group was estimated to provide 80% power to detect a difference of 16% in the percent reduction in seizure frequency between an active dosage group and placebo group at a 2-sided significance level of 0.05, assuming an SD of 40% using a 2-sample Wilcoxon rank sum test (Study C013) or an independent samples t test (Study C017). Potential losses to follow-up were not considered in the power calculations. No sample-size calculations were conducted for the primary responder rate end point of greater than or equal to 50% in Study C017.

Subgroup Analyses

In Study C013, subgroup analyses were conducted based on the completer population, per-protocol population, and 3 groups of patients who received at least 1 dose of the study drug during a specific period of the protocol: seizure frequency data only beyond visit 7 (last 8 weeks of treatment); seizure frequency data only beyond visit 9 (6-week analysis); and seizure frequency data only beyond visit 5 (10-week analysis).

No subgroup analyses were conducted in Study C017.

Table 8: Statistical Analysis of Efficacy End Points

ANCOVA = analysis of covariance; DB = double blind; LOCF = last observation carried forward; mITT = modified intention to treat; mITT-M = modified intention to treat in maintenance period; NA = not applicable; NR = not reported; QOLIE-31-P = patient-weighted 31-item Quality of Life in Epilepsy Questionnaire.

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

Analysis Populations

The efficacy outcomes in both studies were analyzed based on the mITT population, which included all randomized patients who had taken at lease 1 dose of the study drug and had any postbaseline seizure data (Table 9). Study C017 also analyzed efficacy end points based on the mITT-M population, which included all patients in the mITT population who received at least 1 dose of the study drug and had seizure data during the maintenance period. Subgroup 5 of study C013 included all patients who received at least 1 dose of the study drug during the maintenance period.

The safety population included all randomized patients who received at least 1 dose of the study drug.

Table 9: Analysis Populations of Study C013 and Study C017

ITT = intention to treat; mITT = modified intention to treat; mITT-M = modified intention to treat in maintenance period.

^aIn the Clinical Study Report for Study C013, this population is described as the ITT population; however, for consistency with Study C017, it has been labelled mITT in this CADTH report.

Source: Clinical Study Report for Study C013,¹⁰ Clinical Study Report for Study C017.⁹

Results

Patient Disposition

In Study C013, 285 patients were screened and 222 patients (78%) were randomized. The reasons for screening failure were not reported. Among the patients randomized, 10% in the cenobamate group stopped treatment compared with 9% in the placebo group. In both groups, 4% of patients stopped due to AEs and 4% stopped due to withdrawal by patient (Table 10).

Of the 533 patients screened for Study C017, 437 patients (80%) were randomized, with 9% excluded because they did not meet the inclusion criteria, 7% excluded due to exclusion criteria, and another 2% not assigned to treatment (no reason reported) (Table 10). The proportion of patients who discontinued treatment varied across groups, with the higher-dose groups reporting more patients who stopped therapy (12%, 18%, 27%, and 13% in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively). The most common reasons for stopping therapy were AEs, which were reported by 11%, 14%, 21%, and 5% of patients in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively.

Table 10: Summary of Patient Disposition From Pivotal Studies and RCT Evidence Submitted by the Sponsor

DB = double blind; CEN = cenobamate, ITT = intention to treat; mITT = modified intention to treat; mITT-M = modified intention to treat in maintenance period; NA = not applicable; NR = not reported.

^aAll randomized patients who received at least 1 dose of study drug.

^bAll randomized patients who received at least 1 dose of study drug and had data for at least 1 postbaseline seizure. This population is described as the ITT population in Study C013, but is referred to in this CADTH report as mITT to be consistent across studies.

^cAll randomized patients who completed the titration period, received at least 1 dose of study drug, and had data for at least 1 postbaseline seizure during the maintenance period.

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

Baseline Characteristics

The baseline characteristics outlined in Table 11 were limited to those most relevant to this review or that were believed to affect the outcomes or interpretation of the study results.

In general, the patient demographics were similar between groups within the trials. The mean age of patients enrolled in the pivotal trials ranged from 36.2 years (SD = 11.3 years) to 40.9 years (SD = 12.4 years) across treatment groups. There were roughly equal proportions of men (47% to 54%) and women (46% to 53%) enrolled. In Study C017, most patients were white (82% to 86%), with fewer patients who were Asian (8% to 10%) and Black (< 1% to 4%) or who were another race or whose race was unknown (2% to 5%). In Study C013, 50% to 53% of patients were white, 41% to 43% of patients were Asian, 2% to 3% of patients were Black, and 4% of patients were other races or their race was unknown.

Most patients had been diagnosed with epilepsy many years ago, ranging from 22.6 years (SD = 13.7 years) to 25.5 years (SD = 13.4 years) ago. The most common seizure types reported by patients were complex partial (74% to 84% of patients) and partial onset with secondary generalization (55% to 65% of patients). The baseline rate of seizures was highly skewed, and there were some potential imbalances between groups. In Study C013, the median baseline seizure frequency per 28 days was 5.5 (range = 2 to 237) in the placebo group and 7.5 (range = 0 to 187) in the cenobamate 200 mg group. In Study C017, the median baseline seizure frequency per 28 days ranged from 8.4 (range = 4 to 704) for placebo to 11.0 (range = 4 to 418) in the cenobamate 200 mg group. In Study C017, patients had tried 1 to 9 prior ASMs, with a median of 3 ASMs in all groups. Overall, 14% of patients had received 1 ASM before enrolment, while 28%, 24%, and 35% of patients had received 2, 3, or more than 3 ASMs, respectively. In both studies, 1% to 3% of patients had a vagal nerve stimulator implanted.

Table 11: Summary of Baseline Characteristics in Pivotal Studies and RCT Evidence Submitted by the Sponsor

ASM = antiseizure medication; CEN = cenobamate; NR = not reported; SD = standard deviation; VNS = vagal nerve stimulator.

Note: Data were based on the safety population in both studies.

^aPatients may be reported in more than 1 category. Seizure history information was missing for 30 patients in Study C017. Partial onset seizure type may also be described as focal aware nonmotor, focal aware motor, focal impaired awareness, and focal to bilateral tonic-clonic.

^bCalculated by the number of seizures over the baseline period divided by number of days in the interval multiplied by 28 (modified intention-to-treat population).

^cOne patient with only simple partial seizures with sensory component was randomized and treated in error. This was a major protocol deviation.

^dASMs taken any time before the start of the study; these might or might not have been ongoing during the study.

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017;⁹ additional data from sponsor.⁸⁴

Exposure to Study Treatments

In Study C013, the mean duration of exposure was similar in the cenobamate group (86.3 days [SD = 21.1 days]) and placebo group (86.3 days [SD = 17.2 days]) (Table 12). The modal dose was not reported for Study C013; however, data on the highest cenobamate dose achieved during the titration period were available for patients who completed the study (N = 102). Most patients achieved a maximum dose of 200 mg daily (67%), with 25%, 5%, and 4% achieving a daily dose of 150 mg, 100 mg, and 50 mg, respectively. Chung et al. (2020) reported that 59 patients out of 102 patients (58%) completed the study on the 200 mg dose.⁶⁰

In Study C017, the mean duration of exposure was highest for the placebo group (119.6 days [SD = 24.1 days]) and lowest for the cenobamate 400 mg group (108.8 days [SD = 34.8 days]); however, the median duration of exposure was the same for all groups (126 days) (Table 12). Dose reductions were more common in the cenobamate 400 mg group, with 52% of patients requiring a reduction compared with 11%, 22%, and 2% of patients in the cenobamate 100 mg, cenobamate 200 mg, and placebo groups, respectively. The mean modal daily doses were 95.4 mg (SD = 14.6 mg), 184.5 mg (SD = 34.4 mg), and 289.6 mg (SD = 120.1 mg) in the cenobamate 100 mg, 200 mg, and 400 mg dosage groups, respectively.

Table 12: Summary of Patient Exposure From Pivotal Studies and RCT Evidence Submitted by the Sponsor

CEN = cenobamate; NA = not applicable; NR = not reported; SD = standard deviation.

Note: Data were based on the safety population in both studies.

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

During Study C013, fewer patients were taking 1 concomitant ASM (14%) versus 2 ASMs (47%) or 3 ASMs (39%). In Study C017, 27%, 48%, and 24% of patients were taking 1, 2, or 3 ASMs during the study. Table 13 shows that there was some variation between groups in the distribution of patients receiving 1, 2, or 3 ASMs in Study C017.

In the pivotal trials, the most common ASMs (used by at least 10% of patients) were levetiracetam, lamotrigine, carbamazepine, lacosamide, valproate sodium and valproic acid, oxcarbazepine, topiramate, and clobazam (Table 13).

Table 13: Summary of Concomitant ASMs From Pivotal Studies and RCT Evidence Submitted by the Sponsor

ASM = antiseizure medication; CEN = cenobamate; NR = not reported.

Note: Data were based on the safety population in both studies.

^aASMs ongoing at the start of the study and continued during the study.

^bASMs used in 10% or more of all patients.

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

Efficacy

Seizure Frequency

The percent reduction in partial seizure frequency per 28 days for the DB treatment period relative to baseline (mITT population) was the primary outcome in Study C013 and a co-primary outcome in Study C017 (Table 14). In Study C013, the cenobamate 200 mg group showed a median 55.6% reduction in partial seizure frequency per 28 days compared with a 21.5% reduction in the placebo group (P < 0.0001). In Study C017, the median percent reductions in seizure frequency per 28 days were 35.5%, 55.0%, 55.0%, and 24.0% in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively. The P values were 0.007 or less, with all cenobamate dosage groups favoured versus placebo.

In Study C017, a secondary analysis of the change in seizure frequency based on the maintenance period (last 12 weeks for the mITT-M population) reported median reductions of 41.5%, 56.5%, 63.0%, and 27.0% in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively. The P values for both the cenobamate 200 mg and 400 mg groups versus placebo were less than 0.001, but for the cenobamate 100 mg group, the P value was 0.054. In a subgroup analysis in Study C013, the seizure frequency during the maintenance period (last 6 weeks) showed median 63.4% and 29.7% seizure reductions in the cenobamate and placebo groups, respectively (P < 0.0001). These analyses were not controlled for type I error rate in either study; thus, any statistically significant findings should be interpreted in light of the potential for inflated type I error rate.

For Study C013, sensitivity analyses based on the per-protocol and completer populations showed findings that were consistent with the primary analysis. In Study C017, sensitivity analyses based on region, follow-up periods (titration period, first and last 6 weeks of maintenance period), using a parametric ANCOVA model, and including patients who stopped treatment during the titration period (for the analysis of maintenance period treatment effects) did not consistently detect a difference in seizure frequency between the cenobamate 100 mg group versus placebo. The differences in seizure frequency for the 200 mg and 400 mg dosage groups favoured cenobamate versus placebo.

The percent change in seizure frequency per 28 days for each type of partial seizures is shown in Appendix 1, Table 36. The reductions in seizures (simple partial with motor component, complex partial, and secondary generalized tonic-clonic) favoured the cenobamate 200 mg and 400 mg dosage groups versus placebo. These analyses were controlled for type I error rate.

Table 14: Summary of Seizure Frequency Results From Pivotal and RCT Evidence

DB = double blind; CEN = cenobamate; mITT = modified intention to treat; mITT-M = modified intention to treat in maintenance period; RCT = randomized controlled trial; SD = standard deviation.

^aOne patient who had 0 relevant partial seizures during the baseline period was excluded from the analysis.

^bIn Study C013, the P value was based on a Wilcoxon rank sum test assessing whether the median percent change for CEN was significantly different from the median percent change for placebo. In Study C017, the P value was based on a nonparametric ANCOVA model with terms for ranked baseline seizure rate and treatment group.

^cPrimary end point in C013; no multiplicity adjustment required.

^dPrimary end point for the US regulatory agency in Study C017; the P value for each dosage group has been adjusted for multiple testing.

^eSeizure frequency during the maintenance period of Study C013 (last 6 weeks of treatment) was based on subgroup 5 (i.e., patients who received the study drug during the maintenance period and may or may not have completed the study). For Study C017, seizure frequency during the maintenance period (i.e., the last 12 weeks) was based on the mITT-M population (i.e., all randomized patients who completed the titration period, received at least 1 dose of the study drug in the maintenance period, and had any data related to a maintenance period seizure).

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

Seizure-Free Days

This outcome was not reported in either study.

Seizure Freedom

The proportions of patients who achieved a 100% reduction in seizure frequency were lower for the analyses based on the entire treatment period (titration and maintenance; refer to Table 15) than for the analyses that were based on the maintenance period only (Appendix 1, Table 37). In Study C013, 10 patients (8.8%) versus 1 patient (0.9%) achieved a 100% reduction in seizures in the DB treatment period (P = 0.0148). Based on the maintenance period only, 30 patients (28.3%) in the cenobamate 200 mg group and 9 patients (8.8%) in the placebo group reported a 100% reduction in partial seizure frequency per 28 days (P = 0.0001). Both analyses were conducted post hoc, and there was no control for multiple testing.

In Study C017, 0 patients who received placebo, 2 patients in the 100 mg group (1.9%; P = 0.50), 8 patients in the 200 mg group (7.3%; P = 0.007), and 7 patients in the 400 mg cenobamate group (6.3%; P = 0.014) achieved a 100% reduction in seizures for the DB treatment period. The maintenance period analysis reported that 1 patient (1.0%) in the placebo group achieved a 100% reduction in seizure frequency compared to 4 patients (3.9%; P = 0.37), 11 patients (11.2%; P = 0.002), and 20 patients (21.1%; P < 0.001) in the cenobamate 100 mg, 200 mg, and 400 mg groups, respectively. These analyses were not controlled for multiple testing.

Seizure Frequency — Responder Analyses

The proportion of patients who achieved at least a 50%, 75%, or 90% reduction in partial seizure frequency per 28 days was reported based on the DB treatment period (Table 15) and maintenance period only (Appendix 1, Table 37). For both studies, the proportion of responders was generally higher for the analyses based on the maintenance period than the analyses based on the entire DB treatment period.

In Study C017, 40.7%, 57.8%, 60.4%, and 21.7% of patients in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively, achieved at least a 50% reduction in seizure frequency per 28 days during the DB treatment phase (Table 15). All analyses favoured cenobamate versus placebo (P = 0.003, P < 0.001, and P < 0.001, respectively; not controlled for type I error rate). In Study C013, 50.4% of patients in the cenobamate 200 mg group and 22.2% of patients in the placebo group achieved at least a 50% reduction in seizure frequency during the DB treatment period (odds ratio = 3.94; 95% CI, 2.14 to 7.24; P < 0.0001; not controlled for type I error rate).

The alternate analyses, which were based on treatment response in the maintenance period only in Study C017, reported that 40.2%, 56.1%, 64.2%, and 25.5% of patients achieved at least a 50% reduction in seizure frequency per 28 days in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively. For this co-primary end point, the differences favoured the cenobamate 100 mg (P = 0.036), 200 mg (P < 0.001), and 400 mg groups (P < 0.001) versus placebo. The sensitivity analyses showed results that were consistent with the primary analysis. In Study C013, 50.0% and 21.6% of patients in the cenobamate 200 mg and placebo groups achieved at least a 50% reduction in seizure frequency per 28 days during the maintenance period (P < 0.001; not controlled for type I error rate) (Appendix 1, Table 37).

The proportion of responders who achieved a reduction in seizure frequency of greater than or equal to 75% and greater than or equal to 90% during the DB and maintenance periods favoured the cenobamate 200 mg and 400 mg groups versus placebo, but did not favour the cenobamate 100 mg group versus placebo. Of note, there was no control of type I error rate for these analyses in either study. In Study C013, these analyses were conducted post hoc.

Table 15: Summary of DB Treatment Period Responder Results From Pivotal and RCT Evidence

DB = double blind; CEN = cenobamate; CI = confidence interval; mITT = modified intention to treat; NR = not reported; OR = odds ratio; RCT = randomized controlled trial.

^aIn Study C013, the OR (95% CI) and P values were based on a logistic regression model with terms for treatment, country, and baseline seizure frequency (Wald chi-square test). In Study C017, the P values were based on Fisher’s exact chi-square test.

^bGreater than or equal to 75%, greater than or equal to 90%, and 100% responder analyses were conducted post hoc in Study C013.

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

Health-Related Quality of Life

No data on HRQoL were available in Study C013. Study C017 reported descriptive data for the QOLIE-P-31 instrument that were collected from 116 patients enrolled in the US, UK, or Australia (i.e., 27% of the total study population). The QOLIE-31-P is scored from 0 to 100, with higher scores indicating better quality of life. The CSR stated that an 11.8-point increase in the total score represents the minimal important change. The mean within-group change from baseline ranged from –6.2 to 0.6 points in the cenobamate groups, compared with 3.8 points in the placebo group. The CSR states that the changes observed were not clinically relevant.

Treatment Retention or Adherence

Neither study reported data on treatment retention or adherence.

Harms

Refer to Table 17 for harms data.

Table 16: Change From Baseline in QOLIE-31-P Score — Study C017

CEN = cenobamate; QOLIE-31 = 31-item Quality of Life in Epilepsy Questionnaire; QOLIE-31-P = patient-weighted 31-item Quality of Life in Epilepsy Questionnaire; SD = standard deviation.

Note: Data were based on a subgroup of the mITT population in the US, UK, and Australia who completed the QOLIE-31-P questionnaire.

^aThe minimally important change was defined as a change from baseline plus or minus 11.8 at the end point for the weighted overall QOLIE-31 score.

Source: Clinical Study Report for Study C017.⁹

Table 17: Summary of Harms — Pivotal and RCT Evidence

CEN = cenobamate; DRESS = drug reaction with eosinophilia and systemic symptoms; NR = not reported; RCT = randomized controlled trial; SAE = serious adverse event.

Note: Data were based on the safety population in both studies.

Source: Clinical Study Report for Study C013;¹⁰ Clinical Study Report for Study C017.⁹

Adverse Events

TEAEs were reported by 65%, 76%, and 90% of patients who received cenobamate 100 mg, 200 mg, and 400 mg compared with 63% to 70% of patients who received placebo in studies C013 and C017. The most frequently reported AEs in the cenobamate groups were somnolence (19% to 37%), dizziness (18% to 33%), fatigue (11% to 24%), and diplopia (4% to 15%). In the placebo groups, somnolence was reported in 8% to 12% of patients, dizziness in 14% to 17% of patients, fatigue in 6% to 8% of patients, and diplopia in 2% to 3% of patients.

Serious Adverse Events

SAEs were reported in 3% to 9% of patients who received cenobamate and in 4% to 6% who received placebo.

In Study C013, 2 patients who were receiving cenobamate experienced an SAE (drug hypersensitivity reaction; urinary tract infection), and 1 other patient had status epilepticus just after the end of the taper-off period for cenobamate. Four patients in the placebo group experienced SAEs of status epilepticus (2 patients), major seizure attack, and hospitalization for an abnormal electrocardiogram results detected at screening.

In Study C017, SAEs (reported in at least 2 patients) included suicidal ideation and seizure among those who received cenobamate 100 mg, and ataxia, dizziness, and nystagmus among those who received cenobamate 400 mg. One patient attempted suicide in the cenobamate 100 mg group.

Withdrawals Due to Adverse Events

In Study C013, 5 patients (4%) in the cenobamate group and 3 patients (3%) in the placebo group stopped treatment due to AEs. No specific events were reported in more than 1 patient. In the cenobamate group, the events that led to discontinuation included tachycardia, abdominal pain, gastroesophageal reflux, drug hypersensitivity, nystagmus, aggression, depression, and dyspnea. In the placebo group, the events were altered state of consciousness, ataxia, dyskinesia, grand mal seizure, partial seizures, status epilepticus, somnolence, and tremor.

In Study C017, 11 patients (10%), 15 patients (14%), 22 patients (20%), and 5 patients (5%) stopped treatment due to AEs in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively. The most frequently reported AEs leading to withdrawal from the cenobamate groups were ataxia, dizziness, somnolence, nystagmus, and vertigo.

Mortality

No deaths were reported during the DB treatment period of either study.

Notable Harms

CADTH identified DRESS, dermatologic reactions, arrhythmias (short QT syndrome), and suicidal ideation or behaviour as AEs of interest to this review.

One patient in the cenobamate 200 mg group of Study C017 developed a DRESS reaction on day 24 that the investigator considered probably related to the study drug; the study drug was stopped. The patient was also receiving divalproex sodium, levetiracetam, and zonisamide. Other dermatologic reactions that led to study drug withdrawal were reported in 2 patients in the cenobamate 200 mg group. One patient experienced a pruritic rash and pyrexia (concomitant ASMs were clonazepam, clobazam, and lacosamide), and the second patient experienced a rash (concomitant ASMs were valproic acid, levetiracetam, and lacosamide).

In Study C013, 1 patient who was randomized to cenobamate experienced a serious drug hypersensitivity reaction on day 1 of treatment, characterized by reddening of the palms and soles and itching of the ears. The investigator considered this reaction to be definitely related to the study drug. The patient stopped cenobamate treatment and recovered after 22 days.

In Study C013, 3 patients in the cenobamate group reported AEs of atrial flutter, tachycardia, and palpitations, and 1 patient in the placebo group reported palpitations. In Study C017, 2 patients in the cenobamate group reported bradycardia or sinus bradycardia; 2 patients in the 400 mg group reported palpitations; and 2 patients in the placebo group reported sinus bradycardia or tachycardia.

Suicidal ideation was reported by 1 patient in each treatment group in Study C013. In Study C017, suicidal ideation was reported by 2 patients in the cenobamate 100 mg group and by 1 patient in the cenobamate 200 mg group. One patient in the group receiving cenobamate 100 mg in Study C017 attempted suicide.

Critical Appraisal

Internal Validity

The sponsor’s Summary of Clinical Evidence did not meet CADTH’s expectations for clarity, transparency, completeness, or accuracy, and was rated as poor quality by the CADTH reviewer. Moreover, the CSRs for the pivotal trials lacked clarity and had conflicting descriptions of key elements of the study design and methods. Thus, some aspects of the studies’ conduct remain uncertain. This hindered the reviewer’s ability to assess the risk of bias.

Systematic Review

The protocol for the systematic review included the relevant patient population and intervention; however, only 4 drugs plus placebo were included as comparators (eslicarbazepine, brivaracetam, perampanel, and lacosamide). Although the third-generation ASMs are important comparators for cenobamate, the clinical expert consulted by CADTH noted that other drugs listed in Table 3 may also be used as adjunctive therapy in patients with uncontrolled partial seizures. Thus, the focus of the systematic review was considered too narrow.

The outcomes selected were relevant, but the sponsor did not identify any specific AEs of interest. Based on expert input, CADTH identified DRESS, dermatologic reactions, arrhythmias, and suicidal ideation or behaviour as notable AEs.

There were discrepancies in the study design criteria applied in the systematic review. As a result, it is unclear if the criteria were limited to phase II to IV RCTs, phase III RCTs, or phase III and IV RCTs. As noted earlier, the sponsor included Study C021 in its systematic review. However, this study did not meet the study design criteria outlined in the sponsor’s protocol.

Several issues were identified with the literature search conducted by the sponsor. As a result, not all relevant trials may have been identified. Many key search terms were not truncated (i.e., the search used “epilepsy” instead of “epilep*”); thus, the search would have missed terms such as epileptic, epilepsies, and so on. In addition, the search did not include all possible terms for cenobamate. The search included extraneous terms (focal, partial, local, and simple), which likely increased the number of irrelevant results rather than focusing on the population of interest. The description of the methods states that the search was limited to humans, but no terms were included in the search to limit the results. In addition, the reported search strategy contained an error that made the search impossible to execute. Thus, CADTH was unable to verify that the search output would match the numbers reported in the sponsor’s PRISMA chart. Based on these issues, the reported methods were not consistent with CADTH standards, according to the Peer Review of Electronic Search Strategies criteria.

Pivotal Trials

The risks of bias related to randomization, treatment allocation, and blinding in the pivotal trials were rated as low by the CADTH reviewer. There were no major discrepancies in the frequency of specific AEs that may have led to significant unblinding. At baseline, the patient characteristics appeared to be reasonably well balanced between groups within the studies. There was some variation in the median baseline rate of partial seizures, but the impact of these differences is unclear. It was noted that the baseline seizure frequency was highly skewed, with a large range in values. In Study C013, the proportion of patients who discontinued was similar in the cenobamate 200 mg and placebo groups (10% and 9%), but Study C017 showed differential losses to follow-up, with 12%, 18%, 27%, and 13% of patients stopping treatment in the cenobamate 100 mg, 200 mg, 400 mg, and placebo groups, respectively. As a result, it is not clear if the treatment groups remained balanced throughout the study.

Both trials measured the change in partial seizure frequency as well as the proportion of responders (≥ 50%) as primary and secondary end points. The analyses included all simple partial motor, complex partial, and secondarily generalized seizures, which the clinical expert agreed was appropriate for this population. Seizure data were collected from patient diaries. There is evidence that seizures reported by patients were undercounted compared with objective measures, such as electroencephalogram monitoring.^68-71 Under-reporting may be related to lack of awareness of an event (e.g., due to the type of seizure or to seizures occurring at night) and to other factors related to the disease itself (as opposed to lack of motivation to record events).^68-71,85 However, in the clinical trials, under-reporting is expected to occur in both the active and control groups; thus, it may not bias the findings.

While the 50% response threshold has been used in other ASM clinical trials and may be accepted as a minimal clinically important difference, the expert indicated that generally higher response thresholds are desired and that the goal of therapy is seizure freedom. Other seizure response thresholds were tested in both studies, but these analyses were not controlled for type I error and were conducted post hoc in Study C013. Moreover, the responder analyses were conducted using the last observation carried forward for Study C013 and with no imputation for missing data in Study C017; thus, patients who withdrew early could be considered as treatment responders. Patients who achieved a 100% reduction in seizure frequency were interpreted as achieving seizure freedom. However, only 3 types of partial seizures were included in the analysis. This means that patients could have other types of seizures, including generalized seizures, and still be deemed seizure-free. Similarly, those who stopped treatment early for intolerable AEs could also be classified as responders. The use of a “pragmatic intention-to-treat” analysis, in which patients who withdrew were considered nonresponders, may have been a preferred approach.⁷²

Given the skewed nature of the seizure frequency data, the use of nonparametric statistical testing methods was reasonable. In Study C017, randomization was stratified by country, but country was not included in the nonparametric ANCOVA model. The CSR states that omitting country as a covariate did not affect the validity of the analysis and may not have reduced variability because each region was to capture patients who were relatively homogenous in response. The responder analyses also did not adjust for country or baseline seizure rate. It is unclear if the assumption of homogeneity is valid, but CADTH was unable to assess the potential impact that excluding these covariates may have had on the findings.

The calculation of seizure rates was based on patient-reported diary data, excluding any days with missing data. This method assumed that days with missing data would have the same seizure rate as days with nonmissing data, both for missing interim days and for patients who discontinued the study early. The sponsor indicated that there were limited missing data from the primary efficacy analyses;⁸⁴ however, it is unclear if this accounts for missing data after dropout for patients who withdrew early. As noted previously, there were differential discontinuation rates across groups in Study C017, with most of the differences explained by withdrawals due to AEs. It is not clear that patients who withdrew from the study early would have had the same seizure frequency as patients who remained in the trial. While the CSR states that several sensitivity analyses were conducted to assess the impact of missing data, these analyses focused on different follow-up periods (i.e., first or last 6 weeks of the maintenance phase), but did not impute any missing data. The sponsor was asked to provide additional information on the rationale for the sensitivity analyses and how these could be used to confirm that the missing data assumption was valid. Based on the information presented,⁸⁴ the CADTH review team concluded that the analyses did not fully test the missing data assumption because none took a conservative approach (e.g., nonresponder imputation, baseline value carried forward). In general, patients who drop out are likely to have worse outcomes than those who continue; therefore, given the extent of the early withdrawals and the differential losses in Study C017, it is possible that the results may be biased in favour of cenobamate. However, the extent of any potential bias and the impact on the overall findings are unclear.

Study C017 had 2 preplanned statistical analyses with different primary and key secondary outcomes to meet regulator requirements in the US and Europe. The US-based efficacy end points were based on the total DB treatment period (i.e., titration and maintenance, mITT population), whereas the European outcomes were based on the maintenance period only for patients who completed the titration phase (mITT-M). The outcomes based on the DB treatment period were closer to an intention-to-treat approach; however, the clinical expert consulted stated that both methods may be used to assess treatment effects in clinical practice. In Study C017, the multiplicity adjustment appears to apply to the primary end points only; thus, depending on whether the US or European statistical analysis is preferred, only the comparisons for the 3 cenobamate dosage groups versus placebo for the percent change in seizure frequency (mITT population) or the greater than or equal to 50% responder analysis (mITT-M population) would have a family-wise type I error rate controlled at 0.05. In Study C013, there was no control of multiplicity for secondary end points; thus, the family-wise type I error rate may be elevated. An interim analysis was run in both studies, and the CSRs state that there were no plans to modify the studies based on these data; thus, there was no need to adjust the P value to account for the interim analyses. Regarding Study C013, the FDA Statistical Review report states that it was designed as a proof of concept study, and the lack of prespecification of the unblinded interim analysis (e.g., timing, firewall, details of analyses to be conducted) potentially jeopardized the study’s credibility (although the reviewer did not find changes in the study conduct before and after the analysis).⁶³ In addition, the statistical review states that the “ambiguity in the primary analysis adds complexity to the evaluation of the efficacy evidence,” but deemed that the statistical issues did not affect the overall conclusions.⁶³ Given the design and analysis limitations of Study C013, the CADTH review focused on Study C017.

Neither of the pivotal trials was designed to test the impact of cenobamate on HRQoL. Although Study C017 collected data using the QOLIE-31-P, this information was gathered only from approximately 25% of the patients enrolled and was reported descriptively. Moreover, the trials lacked an active comparator group and were 12 weeks to 18 weeks in duration; thus, these can only address short-term efficacy and safety versus placebo. The sample sizes and durations of the studies were insufficient to capture rare AEs.

External Validity

Of the patients screened for entry, 80% were included in the pivotal trials, with limited details available on the characteristics of those not enrolled. Both trials excluded older patients (i.e., aged > 65 years or > 70 years) in addition to patients with comorbid psychiatric or neurologic conditions, substance use disorder, suicidal ideation or behaviour, or a history of drug hypersensitivity reactions, seizure clusters, or status epilepticus. As a result, the treatment effects of cenobamate in these patients is unknown. On average, the patients enrolled were in their late thirties, were predominantly white, and had had a diagnosis of epilepsy for more than 20 years. Those enrolled were required to be experiencing at least 3 to 4 partial seizures per month; however, the baseline seizure frequency was highly variable, with some patients experiencing more than 200 seizures per month. Neither study included patients from Canada, but the clinical expert consulted did not think there were important regional differences that would limit the generalizability of the results to patients in Canada with partial onset seizures.

The pivotal trials used a more rapid titration schedule than has been recommended in the product monograph, which may have affected the occurrence of some AEs. Study C017 included a 100 mg dosage group (i.e., half the recommended maintenance dose), which may be subtherapeutic. But in clinical practice, not all patients may be able to tolerate 200 mg daily and may receive lower doses. In Study C017, 14% of patients had tried 1 ASM before enrolment. For these patients, cenobamate would be considered second-line therapy (prior ASM data were not reported for Study C013). During the trials, it was most common for patients to be receiving 2 or more concomitant ASMs (73% to 86%) versus 1 ASM (14% to 27%), and few patients had a vagal nerve stimulator implanted (1% to 3%). The clinical expert stated that the concomitant ASMs used were consistent with clinical practice. However, no dosing information was reported to determine whether dosages were similar to practice in Canada.

Of note, the studies were 12 weeks to 18 weeks in duration, which the expert considered to be short for a chronic condition that often requires lifelong therapy. Based on these data, the longer-term efficacy and safety are uncertain.

Long-Term Extension Studies

Contents within this section have been informed by materials submitted by the sponsor. The following subsections have been summarized and validated by the CADTH review team.

Description of Studies

Two long-term extension studies were included in this review: Study C013 OLE^11,86 and Study C017 OLE.^12,13 There were no comparator groups in the OLE studies. Moreover, no efficacy results were reported in the Study C013 OLE. Both studies reported safety results. The CADTH review team summarized the study design and results to provide information as supportive evidence.

The Study C013 OLE was a long-term extension of Study C013, a 2-arm (cenobamate 200 mg once daily or placebo) RCT with a DB treatment period of 12 weeks.⁸⁶ Patients who completed the randomized phase of Study C013 were eligible to enter the OLE phase and received open-label cenobamate at a daily maximum dose of 400 mg daily. Adults with uncontrolled partial onset seizures (N = 149) were followed-up up to 8.6 years. The only safety outcomes assessed were TEAEs.

The Study C017 OLE was a long-term extension of Study C017, which was a 4-arm (cenobamate 100 mg, 200 mg, or 400 mg once daily or placebo), dose-response trial with a DB treatment period of 18 weeks (a 6-week titration period and a 12-week maintenance period).¹² Patients who completed Study C017 were eligible to enter the OLE phase and were converted to open-label cenobamate at a target dose of 300 mg once daily. Efficacy outcomes of seizure control for up to 48 months of the OLE and safety outcomes for up to 6.4 years of cenobamate treatment in adults with uncontrolled partial onset seizures (N = 356) were assessed.

Profiles of the Study C013 OLE and the Study C017 OLE are summarized in Table 18.

Table 18: Summary of Study Characteristics for Study C013 OLE and Study C017 OLE

DB = double blind; OLE = open-label extension; SAE = serious adverse event; TEAE = treatment-emergent adverse event.

Source: Clinical Study Report for Study C013 OLE;¹¹ Clinical Study Report for Study C017 OLE.¹²

Populations

The inclusion and exclusion criteria for the OLE studies were the same as for the pivotal studies, but without the specification of seizure frequency at baseline (i.e., adults with uncontrolled partial onset seizures). Of the patients who entered the Study C013 OLE (n = 149), 48.3% were men, which is slightly lower than the proportion of men at baseline in Study C013 (113 patients of 222 patients [50.9%] were men). The mean age of patients in the Study C013 OLE was 37.6 years (SD = 10.9 years), similar to the mean age of 36.9 years (SD = 11.3 years) at baseline in Study C013. The proportion of men among the patients who entered the Study C017 OLE (185 patients out of 356 patients [52.0%]) was slightly higher than that at baseline in Study C017 (221 patients of 437 patients were men [50.6%]). The mean age of the patients who entered the Study C017 OLE (39.6 years; SD = 11.7 years) was similar to that at baseline in Study C017 (mean = 39.8 years; SD = 11.8 years).

Interventions

The initial dose of cenobamate was 100 mg/day. This was followed by increments of 50 mg/day biweekly in the Study C013 OLE¹¹ and by 100 mg/day weekly in the Study C017 OLE (later amended to 50 mg/day),¹² as tolerated. The target dose of cenobamate in the Study C017 OLE was 300 mg per day, and the maximum dose was 400 mg/day.^11,12 In the OLE for Study C013, the maximum daily dose was initially 200 mg per day. However, 2 years into the study, it was changed to 400 mg per day.

During the OLE periods, investigators could adjust the dosage and add or remove concomitant ASMs as clinically indicated. However, monotherapy with cenobamate was not allowed.

Outcomes

The safety outcomes were same for both OLE studies, including incidences of TEAEs, serious TEAEs, and TEAEs leading discontinuation of the study medication.^11,12 No seizure diary data were collected during the OLE of Study C013. The Study C017 OLE also reported the efficacy outcomes as single-arm cohort results without a comparison group, including the incidence of seizure freedom, time to seizure freedom, and responder rates of greater than or equal to 50%, greater than or equal to 75%, and greater than or equal to 90%.¹²

Statistical Analysis

The numbers and proportion of patients with outcomes were reported using descriptive statistical methods. Outcomes by subgroups were statistically tested and reported when applicable.

Characteristics of the patients who enrolled in the 2 OLE studies are summarized in Table 19.

Results

Patient Disposition

Of the patients initially randomized in Study C013 and Study C017, 67.1% (149 patients of 222 patients in the Study C013) and 81.5% (356 patients of 437 patients in Study C017), respectively, entered the OLE studies. (None of the 43 patients from the study site in India participated in the Study C013 OLE.) Details of patient disposition to cenobamate are summarized in Table 20.

Table 19: Summary of Baseline Characteristics for Study C013 OLE and Study C017 OLE

ASM = antiseizure medication; BMI = body mass index; OLE = open-label extension; SD = standard deviation.

Note: Patient characteristics were collected at baseline in the OLEs of Study C013 and Study C017 for the safety population (i.e., patients who entered the OLE phase of the pivotal studies C013 and C017).

^aPatients may be reported in more than 1 category.

^bIn the Study C013 OLE, valproate included valproate semisodium, valproate sodium, and valproic acid. In the Study C017 OLE, valproate included all forms of valproate, valproic acid, or divalproex sodium.

Source: Clinical Study Report for Study C013 OLE;¹¹ Clinical Study Report for Study C017 OLE.¹²

Table 20: Patient Disposition for Study C017 OLE and Study C013 OLE

EAP = Extended Access Program; NR = not reported; OLE = open-label extension.

^aBecause India did not approve the OLE, all 43 patients from that site were excluded from it.

^bOf the 360 patients who completed the DB Study C017,⁶¹ 356 patients entered the OLE. For the 4 who did not, reasons included adverse events (N = 1), withdrawal by patient (N = 1), withdrawal by sponsor due to noncompliance (N = 1), and reason unspecified (N = 1). One patient did not have any dose data recorded, leaving 355 in the safety population.⁸⁷

Source: Clinical Study Report for Study C013 OLE,¹¹ Clinical Study Report for Study C017 OLE.¹²

Exposure to Study Treatments

Details of patient exposure to cenobamate are summarized in Table 21.

Table 21: Patient Exposure for Study C017 OLE and Study C013 OLE

NR = not reported; OLE = open-label extension; SD = standard deviation.

^aAnalyzed for safety (N = 355); analyzed for efficacy (modified intention-to-treat population, N = 354).

Source: Clinical Study Report for Study C013 OLE,¹¹ Clinical Study Report for Study C017 OLE.¹²

Efficacy

The efficacy of cenobamate was not analyzed for patients in the Study C013 OLE.

The efficacy of cenobamate was reported for a prospective, single-arm cohort of the study population in the Study C017 OLE. Klein et al. (2022) reported interim efficacy outcomes based on a median exposure duration of 53.9 months (range = 1.1 months to 68.7 months), with retention rates at 12 months, 24 months, 36 months, and 48 months of 83%, 71%, 65%, and 62%, respectively.¹³ Klein et al. (2022) reported that “the percent of observed patients achieving 100% seizure reduction at consecutive 12-month intervals increased from 13.3% (36/271) during >12-24 months to 16.4% (36/220) during the last 12-month interval, >36-48 months” (observed case data).¹³ Of the 354 patients who entered the OLE, 10.2%, 9.6%, and 10.2% achieved a 100% reduction in seizure frequency in the intervals of greater than 12 months to 24 months, greater than 24 months to 36 months, and greater than 36 months to 48 months, respectively.¹³

Kelin et al. (2022) also reported that, “Among the patients in each 12-month interval group, the median IQR durations of 100% seizure reduction for the entire study were 48.0 months (20.1 months), 47.2 months (18.3 months), and 45.1 months (27.4 months). The median modal daily dose for patients with 100% seizure reduction at each 12-month interval was 300 mg (IQRs ranging from 50 to 100 mg).”¹³ Further, “Median percent reduction in seizure frequency during the first 6 months of the OLE for all cenobamate OLE patients was 65.4% and was similar among patients originally treated with cenobamate or placebo in the double-blind study”¹³ and the “median percent reduction in seizure frequency over baseline for all cenobamate OLE patients increased with each 6-month OLE interval, up to 76.1% (IQR 44.8%) at months 43–48.”¹³

During the greater than 36 months to 48 months interval, seizure frequency reductions greater than or equal to 50%, greater than or equal to 75%, and greater than or equal to 90% were achieved in 76.1% of patients (168 patients out of 220 patients), 51.8% of patients (114 patients out of 220 patients), and 39.1% of patients (86 patients out of 220 patients), respectively.¹³ Among the 354 patients who entered the OLE, 47.5%, 32.2%, and 24.3% achieved seizure frequency reductions of greater than or equal to 50%, greater than or equal to 75%, and greater than or equal to 90% during the 36-month to 48-month interval.¹³

Harms

According to the CSR for the Study C013 OLE, TEAEs occurred in 88.6% of patients over a median cenobamate treatment duration of 87.0 months, with 22.8% of events (34 out of 149) being mild, 50.3% (75 out of 149) being moderate, and 15.4% (23 out of 149) being severe in severity. SAEs occurred in 26.2% of patients (39 patients out of 149 patients) in the OLE. The most frequently occurring SAEs (experienced by at least 1% of patients) were convulsion (4.0%), inguinal hernia (1.3%), vomiting (1.3%), sepsis (1.3%), and osteoarthritis (1.3%). TEAEs leading to discontinuation were reported in 9.4% of patients (14 patients out of 149 patients); those leading to discontinuation in greater than 1% of patients were fatigue (1.3%) and ataxia (1.3%). Four deaths were reported during the Study C013 OLE. One of these patients had received placebo during the DB treatment period and died after experiencing respiratory arrest on day 2,543; the remaining 3 patients died after experiencing SUDEP on day 267, cardiac arrest on day 697, and completed suicide on day 132, respectively. The SUDEP and completed suicide were deemed by the investigator to be remotely related to the study drug; the respiratory arrest and cardiac arrest were deemed unrelated.^11,39

According to the CSR for the Study C017 OLE, TEAEs occurred in 89.0% of patients over a median exposure duration of 63.8 months, with 21.1% of events (75 out of 356) being mild, 45.5% (162 out of 356) being moderate, and 22.5% (80 out of 356) being severe in severity. There were no cases of DRESS or any serious skin or subcutaneous tissue disorders. Thirty-two patients (9.0%) had at least 1 TEAE leading to discontinuation, most frequently (≥ 0.5%) due to dizziness (0.8%), somnolence (0.6%), balance disorder (0.6%), and depression (0.6%). In total, 6 deaths were reported in the Study C017 OLE, with all considered by investigators to be unrelated to the study drug.¹² Causes of death were myocardial infarction, cardiogenic shock, SUDEP, completed suicide, and sepsis.

Harms data for Study C013 OLE and Study C017 OLE are summarized in Table 22.

Table 22: Harms Results for Study C013 OLE and Study C017 OLE – Long-Term Extension Studies

NR = not reported; OLE, open-label extension; SAE = serious adverse event; TEAE = treatment-emergent adverse event.

^aOne event of complex partial seizures and 1 event of seizure cluster were reported. No clear information was provided regarding whether these TEAEs occurred in the 1 same patient or in 2 different patients.

^bThe most common TEAEs that led to treatment discontinuation are listed.

Source: Clinical Study Report for Study C013 OLE¹¹ and Clinical Study Report for Study C017 OLE.¹²

Critical Appraisal

Internal Validity

Neither of the extension studies had an active comparator or placebo group. Thus, the safety data in both studies and the efficacy data in Study C017 OLE could not be used to draw any conclusion in relation to an appropriate comparator, such as another ASM. Furthermore, the open-label study design in the OLE can bias the reporting of end points, particularly subjective measures, including AEs, SAEs, and TEAEs. Given that the completion of a pivotal trial was an eligibility criterion for the extension study, patients who discontinued those trials due to AEs or lack of response were excluded. This exclusion could result in a population of patients who were more tolerant of cenobamate, potentially leading to a response bias because those not responding to treatment are less likely to continue. Having patients who are more tolerant of cenobamate can also lead to biased estimates related to AEs, potentially resulting in fewer and less severe AEs being reported.

The sample sizes in the Study C013 OLE (N = 149) and Study C017 OLE (N = 356) may not be sufficient to detect rare AEs. Approximately 25% of patients in the Study C013 OLE and 20% of patients in the Study C017 OLE completed the studies, and there was wide variance in follow-up durations for individuals. Also, more common forms of morbidity (e.g., cardiac dysrhythmias) may not be easily identified as related to drug exposure versus as “natural” events unrelated to the drug. These factors may affect the internal validity of the study results.

External Validity

The patients enrolled in the 2 OLE studies were from multiple sites in different countries. Although the sample size of Study C013 was relatively small, with a median study drug exposure of 87 months, the OLE provided for longer follow-up for AE assessment versus the DB study. Nevertheless, because the proportion of patients who adhered to the study drug during the longer follow-up period was not reported, the study drug exposure among the patients in the 2 OLE studies was uncertain. There were no study sites in Canada for the OLEs of Study C013 or Study C017. On the other hand, no particular evidence indicating a difference between the study population and patients in Canada was identified in consultation with the clinical expert.

Indirect Evidence

Contents within this section have been informed by materials submitted by the sponsor. The following subsections have been summarized and validated by the CADTH review team.

Objectives for the Summary of Indirect Evidence

The clinical trials included in the pivotal and RCT evidence did not provide direct evidence of the comparative efficacy and safety of cenobamate versus other adjunctive ASMs. The objective of this section is to summarize and critically appraise 1 NMA submitted by the sponsor and 1 NMA conducted by NICE that assessed the relative efficacy and safety of cenobamate versus other therapies in the treatment of adult patients with focal onset seizures (FOS).^14-16 This summary also informs the pharmacoeconomic evaluation.

Description of Indirect Comparisons

The sponsor-submitted NMA (ITC 1) was based on a systematic review to assess the clinical efficacy and safety of cenobamate compared to relevant comparator therapies in patients with FOS.^14,15 The NMA was originally conducted and submitted as part of the cenobamate evidence package submission to NICE. Treatments included cenobamate, brivaracetam, perampanel, lacosamide, eslicarbazepine acetate, and zonisamide.

The NICE-conducted NMA (ITC 2) was based on FOS and generalized tonic-clonic seizures; the latter are not relevant to the CADTH reimbursement review and will not be further discussed.¹⁶ Relevant treatments included brivaracetam, carbamazepine, cenobamate, eslicarbazepine acetate, gabapentin, ganaxolone, lacosamide, lamotrigine, levetiracetam, oxcarbazepine, perampanel, phenytoin, pregabalin, primidone, retigabine, rufinamide, sodium valproate, topiramate, vigabatrin, and zonisamide. Carisbamate, loreclezole, losigamone, selurampanel, and tiagabine were also included as comparators in the NMA, but do not have a Health Canada Notice of Compliance and are not reported on in this review.

The selection criteria and methods for both NMAs are presented in Table 23.

Table 23: Study Selection Criteria and Methods for ITCs Informing the CADTH Review

AE = adverse event; ASM = antiseizure medication; CDSR = Cochrane Database of Systematic Reviews; CENTRAL = Cochrane Central Register of Controlled Trials; DARE = Database of Abstracts of Reviews of Effects; DRESS = drug reaction with eosinophilia and systemic symptoms; FOS = focal onset seizure; HRQoL = health-related quality of life; HTA = Health Technology Assessment; ICTRP = International Clinical Trials Registry Platform; ITC = indirect treatment comparison; ITC 1 = indirect treatment comparison 1; ITC 2 = indirect treatment comparison 2; NICE = National Institute for Health and Care Excellence; NIH = National Institutes of Health; NMA = network meta-analysis; OLE = open-label extension; RCT = randomized controlled trial; ROBINS-I = Risk Of Bias In Nonrandomized Studies of Interventions; ROBIS = Risk of Bias Assessment Tool for Systematic Reviews; SAE = serious adverse event; TEAE = treatment-emergent adverse event.

Note: Details shown in the table have been taken from the sponsor’s Summary of Clinical Evidence.

Source: Sponsor’s Summary of Clinical Evidence; sponsor-submitted ITC report; NICE NMA.^14-16

ITC Design – ITC 1 (Sponsor NMA)

Objectives

The objective of ITC 1 was to conduct a systematic review and NMA to estimate the relative efficacy and safety of cenobamate versus comparator ASMs that are relevant to drug plans in Canada for the treatment of FOS in adult patients with epilepsy.^14,15

Study Selection Methods

A literature search of the Embase, MEDLINE, Cochrane Central Register of Controlled Trials, National Institutes of Health Clinicaltrials.gov, WHO ICTRP, Cochrane Database of Systematic Reviews, and Epistemonikos databases was conducted for studies published from 1999 to December 2019.^14,15 The initial search was updated in October 2020 and again in September 2021. The search plan included both electronic searching and handsearching.

Studies included in ITC 1 may have enrolled patients as young as 12 years old with FOS.^14,15 The list of comparators included 4 established third-generation ASMs and zonisamide. The combined efficacy outcome was achieving a 50% or greater responder rate and seizure freedom. The studies used data from the maintenance period (or treatment period, if the former was not reported). The safety outcome was the relative proportion of patients experiencing at least 1 TEAE leading to discontinuation. Both RCTs and OLEs were eligible for inclusion in the NMA. Only articles published in English were included. Patients with primary generalized epilepsy and status epilepticus were excluded from the sponsor’s NMA report.

An internal validation check was conducted to ensure that the search strategy identified all relevant records.^14,15 The sponsor’s systematic review search strategy was compared to strategies used in recently published systematic reviews in this therapeutic area. Handsearching was performed to identify other studies of interest and included searching review articles, reference lists of included full-text publications, and free text keyword searching in internet search engines.

Records retrieved from the searches were added to an Endnote library, and duplicate records were removed.^14,15 Two reviewers were involved in study selection, and disagreements were resolved through consensus or by a third reviewer. Data were extracted and checked for errors (details not available). One reviewer performed a quality assessment of the included studies using the Cochrane Risk of Bias 2.0 tool for RCTs and the Cochrane Risk Of Bias In Nonrandomized Studies of Interventions tool tailored for OLEs.^88,89

ITC Analysis Methods

The analysis methods for ITC 1 are presented in Table 24.

The NMA was conducted using Bayesian Markov chain Monte Carlo methods with both random-effects (base-case) and fixed-effect (supportive analysis) models and noninformative priors.^14,15 A multinomial likelihood model with a probit link function was used for the combined efficacy analysis, while a binomial likelihood model with logit function was used for the safety analysis. Model fit was assessed using the deviance information criterion and residual deviance information. Convergence was assessed using trace plots, Gelman-Rubin plots, statistical analyses, and assessment of the effective sample sizes, while correlation was assessed through autocorrelation plots. For the efficacy outcome, the random-effects model was run with 700,000 iterations, 100,000 burn-ins, and a thinning factor of 6, while the fixed-effect model was run with 400,000 iterations, 100,000 burn-ins, and a thinning factor of 4. For the safety outcome, the random-effects model was run with 300,000 iterations, 100,000 burn-ins, and a thinning factor of 70, while the fixed-effect model was run with 100,000 iterations, 25,000 burn-ins, and a thinning factor of 10.

The results formed open-loop networks with ASMs linked through placebo.^14,15 All maintenance doses licensed by the European Medicines Agency were pooled into a single treatment arm.

The 2 efficacy outcomes (i.e., 50% or greater responder rate and seizure freedom) were combined, and the analysis was run as a multinomial probit model that included an adjustment for the placebo rate.^14,15 Median relative risks with 95% CrIs were reported for the efficacy outcome. The analysis for discontinuations due to TEAEs was run using a binomial model with a logit link. Odds ratios with 95% CrIs were reported for the safety outcome. Data for the maintenance period were reported unless unavailable, in which case the treatment period was used. The last observation carried forward for the mITT maintenance population was used to ensure that the populations between efficacy outcomes aligned for the combined analysis.

To assess homogeneity, a pairwise meta-analysis was performed for each comparison in the network, and the I² statistic and chi-square test P value were calculated.^14,15 A placebo adjustment for the efficacy outcome was included in the model to adjust for cross-study heterogeneity of placebo group response rates using meta-regression methods. A feasibility assessment was conducted to investigate potential sources of heterogeneity and the implications on the NMA. No sensitivity or subgroup analyses were performed.

Because there were no closed loops in the NMA, consistency between direct and indirect comparisons could not be evaluated.

Table 24: ITC 1 Analysis Methods

ASM = antiseizure medication; DIC = deviance information criterion; EMA = European Medicines Agency; ITC = indirect treatment comparison; ITC 1 = indirect treatment comparison 1; LOCF = last observation carried forward; mITT = modified intention to treat; NMA = network meta-analysis; SD = standard deviation; TEAE = treatment-emergent adverse event; vs. = versus.

Note: Details from the table have been taken from the sponsor’s Summary of Clinical Evidence.

Source: Sponsor’s Summary of Clinical Evidence and sponsor-submitted ITC report.^14,15

Results of ITC 1

Summary of Included Studies

The systematic review for ITC 1 identified 2,462 studies. Of these, 269 were retrieved for full-text review, and 56 RCTs and 18 OLEs were considered.^14,15 From this group of selected studies, a feasibility assessment was conducted, and 22 phase II and phase III placebo-controlled studies were included in the ITC. All 22 studies were included in the NMA efficacy analysis, and 20 were included in the safety analysis. (One brivaracetam study and 1 lacosamide study reported combined safety results for multiple seizure populations or treatment groups; therefore, these were excluded).

Characteristics of the included studies are presented in Appendix 1, Table 38.

Sample sizes ranged from 157 patients to 768 patients.^14,15 There were 2 cenobamate studies (the 2 pivotal trials described in the CADTH main report), 6 brivaracetam studies, 4 lacosamide studies, 4 eslicarbazepine acetate studies, 4 perampanel studies, and 2 zonisamide studies. Most studies were multinational, except for 2 that were conducted in the US only. Where reported, titration periods ranged from 0 weeks to 8 weeks, and maintenance periods ranged from 4 weeks to 18 weeks. Total treatment times (whose definition varied among studies, but was typically the titration and maintenance periods together) ranged from 7 weeks to 24 weeks, where reported.

Baseline patient characteristics are presented in Appendix 1, Table 39. Reporting of baseline characteristic data for studies was limited, and it is unclear what age ranges were included in the studies (e.g., if patients under 18 years were included or not); however, the mean or median age for the studies ranged from 32 years to 40 years.^14,15 Proportions of males and females were generally balanced in the studies. The duration of epilepsy ranged from 13.7 years to 25.3 years; it was not reported in 3 studies. Baseline median seizure frequency varied among studies, from 0 seizures to 15 seizures per 28 days; it was not reported in 1 study. The baseline seizure rate was highest in studies of perampanel versus other ASMs. The number of prior ASMs used was rarely reported, and when it was, the definition (e.g., the time period used to define prior ASM use) was inconsistent; therefore, these data could not be assessed. All studies included patients who were using at least 1 ASM at baseline (information was not reported for 2 studies), which was considered a proxy for failed treatment, with most patients being on 2 medications.

The overall risk of bias was assessed using the Cochrane Risk of Bias 2.0 tool for RCTs.⁸⁸ According to the sponsor, most studies had an overall low risk of bias. The exceptions were 1 lacosamide study and 3 eslicarbazepine acetate studies that had some concerns with risk of bias and 1 zonisamide study that had a high risk of bias.^14,15 The main reasons for the concerns were risk of bias arising from the randomization process and missing outcomes data.

The assessment of homogeneity for ITC 1 is presented in Table 25.

Table 25: Assessment of Homogeneity in ITC 1

ASM = antiseizure medication; ITC = indirect treatment comparison; ITC 1 = indirect treatment comparison 1; OLE = open-label extension; RCT = randomized controlled trial; TEAE = treatment-emergent adverse event.

Note: Details from the table have been taken from the sponsor’s Summary of Clinical Evidence.

Source: Sponsor’s Summary of Clinical Evidence and sponsor-submitted ITC report.^14,15

Results

Figure 2 presents the network, number of studies, and corresponding populations for each drug contributing to the combined efficacy outcome (i.e., 50% or greater responder rate and seizure freedom). No additional trials were excluded from the analysis. Figure 3 presents the network, number of studies, and corresponding populations for each drug contributing to the safety outcome (i.e., TEAEs leading to discontinuation). Two studies were excluded from the network due to combined reporting of safety results for multiple seizure populations or treatment groups.^14,15

For pairwise comparisons, I² statistic values ranged from | || | for the 50% or greater responder rate outcome, from | || | for the seizure-free outcome, and from | || | for the safety outcome.^14,15 P values from the chi-square tests were greater than 0.05 for all comparisons.

Figure 2: Overall Network for Combined 50% or Greater Responder Rate and Seizure Freedom Analysis

ITC = indirect treatment comparison.

Source: Sponsor’s Summary of Clinical Evidence and sponsor-submitted ITC report.^14,15

Figure 3: Overall Network for the Proportion of Patients Experiencing at Least 1 TEAE Leading to Discontinuation Analysis

ITC = indirect treatment comparison.

Source: Sponsor’s Summary of Clinical Evidence and sponsor-submitted ITC report.^14,15

Efficacy Outcome: 50% or Greater Responder Rate and Seizure Freedom

The results of the models indicated that patients receiving cenobamate were more likely to achieve a 50% or greater reduction in seizures in the short-term compared to any of the 5 ASMs in the NMA using the random-effects model (Table 26).^14,15 The fixed-effect model results supported the primary analysis. Similarly, the random-effects model suggested that treatment with cenobamate was more likely to result in seizure freedom compared to the other interventions, a conclusion that was supported by the fixed-effect model results (Table 26). Figure 4 and Figure 5 present the relative risks and 95% CrIs for the random-effects model results as forest plots for the efficacy outcomes.

Table 26: Relative Risk for 50% or Greater Responder Rate and Seizure Freedom

CrI = credible interval; DIC = deviance information criterion; ITC = indirect treatment comparison; NA = not applicable; RR = relative risk; SD = standard deviation.

Notes: Bolded values indicate a CrI that does not cross 1.0. RR of less than 1 favours cenobamate vs. the comparator. The random-effects model (base case) was run with 700,000 iterations, 100,000 burn-ins, and a thinning factor of 6. The fixed-effect model (supportive analysis) was run with 400,000 iterations, 100,000 burn-ins, and a thinning factor of 4. The mean and precision used in the fixed effect and random effects from the baseline model were | || || || || || |, respectively. The covariate for placebo response was 0.805. Details from the table have been taken from the sponsor’s Summary of Clinical Evidence.

Source: Sponsor’s Summary of Clinical Evidence and sponsor-submitted ITC report.^14,15

Figure 4: Forest Plot for 50% or Greater Responder Rate

Figure was redacted per the sponsor request.

Source: Sponsor’s Summary of Clinical Evidence and sponsor-submitted ITC report.^14,15

Figure 5: Forest Plot for Seizure Freedom

Figure was redacted per the sponsor request.

Source: Sponsor’s Summary of Clinical Evidence and sponsor-submitted ITC report.^14,15

Safety Outcome: Proportion of Patients Experiencing at Least 1 TEAE Leading to Discontinuation

Compared to cenobamate, the results for the comparators had relatively wide 95% CrIs that included the threshold of no difference (1.0) using the random-effects model (Table 27).^14,15 The fixed-effect model results were consistent with the primary analysis. Due to the lack of precision in the estimates, firm conclusions cannot be made. Figure 6 presents the ORs and 95% CrIs for the random-effects model results as a forest plot for the safety outcome.

Table 27: Odds Ratio for the Proportion of Patients Experiencing at Least 1 TEAE Leading to Discontinuation

CrI = credible interval; DIC = deviance information criterion; ITC = indirect treatment comparison; NA = not applicable; OR = odds ratio; SD = standard deviation; TEAE = treatment-emergent adverse event.

Notes: Bolded values indicate a CrI that does not cross 1.0. An OR of greater than 1 favours cenobamate vs. the comparator. The random-effects model (base case) was run with 300,000 iterations, 100,000 burn-ins, and a thinning factor of 70. The fixed-effect model (supportive analysis) was run with 100,000 iterations, 25,000 burn-ins, and a thinning factor of 10. The predictive mean and SD used in the fixed effect and random effects from the baseline model were | || || || ||, respectively. Details from the table have been taken from the sponsor’s Summary of Clinical Evidence.

Source: Sponsor’s Summary of Clinical Evidence and sponsor-submitted ITC report.^14,15

Figure 6: Forest Plot for the Proportion of Patients Experiencing at Least 1 TEAE Leading to Discontinuation

Figure was redacted per the sponsor request.

Source: Sponsor’s Summary of Clinical Evidence and sponsor-submitted ITC report.^14,15

Critical Appraisal of ITC 1

Overall, the methods used in the sponsor’s literature search strategy and screening appeared to be adequate. Study selection was performed in duplicate; however, some details of data extraction were unknown (e.g., the number of reviewers involved and error checking). An internal validation check was performed by comparing the sponsor’s search strategy to those used in published systematic reviews to confirm that all relevant records were captured. An appraisal of the risk of bias conducted using the Cochrane Risk of Bias 2.0 tool found that 4 of the included studies had some concerns for bias (due to missing data or arising from the randomization process), and 1 zonisamide study had a high risk of bias (in the selection of reported results). However, no studies were excluded based on the risk of bias assessment, and there was no plan to investigate the potential impact of the 4 studies in question.

The sponsor conducted a feasibility assessment for the study inclusion and exclusion criteria. Based on input from clinical experts, overall, the approach to this assessment appeared to be reasonable. With 1 exception, no studies were excluded based on their eligibility criteria. Therefore, some studies may have included pediatric patients, who are not included in the proposed Health Canada indication for cenobamate. It was unclear what proportion of patients in the studies were adolescents and what effect their inclusion had on the results; however, the clinical expert consulted by CADTH indicated there would be issues with generalizing from children to adults. ITC 1 included only third-generation ASMs and zonisamide, which the CADTH review team and clinical expert noted was missing other possible treatments for this population and indication. Although zonisamide is not available in Canada, the clinical expert and clinician group input for this review stated that it can be obtained through Health Canada’s Special Access Program for some patients. The clinician group that submitted input for the cenobamate review estimated that 2% to 5% of patients may require drugs accessed through the Special Access Program.

The sponsor stated that seizure freedom was defined differently among studies; this difference has the potential to bias the findings. Some studies used a conservative definition of seizure freedom (i.e., defining it as the absence of any seizures during follow up, and any patients who withdrew early were considered nonresponders) compared to that used in the cenobamate studies. In the cenobamate trials seizure freedom was defined as a 100% reduction in 3 types of partial seizures compared to the baseline rate, based on patients who completed the study as well as those who withdrew early. In addition, it was unclear if the other efficacy and safety outcomes were defined and measured consistently across studies. Results were analyzed from either the maintenance period or the total treatment period, but were not consistent for all studies in ITC 1. This inconsistency is an important source of heterogeneity. The analyses based on the maintenance period may show a greater reduction in seizure frequency than based on the entire treatment period. The maintenance phase analyses exclude the time period when the drug has not yet reached the target dose and may still be subtherapeutic. Another source of heterogeneity was the difference in the durations of the clinical trials. The treatment periods had variable durations across studies, ranging from 7 weeks to 24 weeks. Moreover, the clinical expert consulted by CADTH indicated that the studies were short in duration, considering that epilepsy is a chronic condition that often requires lifelong treatment.

Based on the sponsor’s feasibility assessment, 7 studies were excluded from the NMA with reasonable explanations. As previously mentioned, 1 zonisamide study was excluded for having an entirely Asian population and being relatively small (N = 104). Three dose-escalation studies were also excluded for their short durations at a stable dose (1 eslicarbazepine acetate and 2 perampanel studies), and 2 zonisamide studies were excluded for using dosages outside of the maintenance range approved by the European Medicines Agency. Lastly, 1 zonisamide study was excluded because it was a single-arm study. As noted, most of the exclusions were for zonisamide studies, which reduces the available evidence for comparisons to this drug to 2 studies.

Twenty-two studies contributed to the efficacy analysis, resulting in a sparse network with no closed loops. The sponsor’s report stated that OLEs were included in the network, but it is unclear how the results from these studies could have been included in the analyses. A random-effects model was the base case. The sponsor stated that this was justified was due to potential heterogeneity among studies, and the CADTH review team agreed that it was an appropriate model choice. Results from the fixed-effect model were reported as a supportive analysis and were consistent with those from the random-effects model. The use of a multinomial model to assess different response levels in the same analysis (i.e., 50% response and seizure freedom) was reasonable (and was the method preferred by NICE in its review of cenobamate).⁹⁰ However, this model assumes a common relative treatment effect for both response thresholds (i.e., the inference for comparator A versus B is the same for a 50% response as for seizure freedom). Whether this assumption is clinically plausible is unclear.

The variability in baseline patient characteristics (particularly the proportion of white patients, mean BMI, duration of epilepsy, seizure frequency per 28 days, and number of ASMs at baseline) suggests that patients differed across studies and that the similarity assumption was likely violated. The duration of epilepsy among the studies in ITC 1 ranged from 13.7 years to 25.3 years, and the clinical expert explained that patients who have had epilepsy for longer have likely been on more ASMs and could have more refractory disease than those more recently diagnosed. The number of prior ASMs was not available due to incomplete reporting in the studies and inconsistent definitions, but the CADTH review team noted this was an important treatment-effect modifier and indicator of disease severity that should be adjusted for in the analysis. Other effect modifiers that the CADTH review team and the clinical expert identified include age, sex, cause of seizures, baseline number of seizures, and duration of active epilepsy. However, these were not adjusted for in any analyses.

The sponsor concluded that there was no statistically significant heterogeneity among pairwise comparisons; however, the results for the I² statistic and chi-square test P value are inconclusive evidence for homogeneity. There were differences in the definition of drug-resistant epilepsy, baseline patient characteristics, doses used, outcome definitions, and treatment durations that indicated heterogeneity in the populations and study characteristics. Data available for different doses were pooled by drug in the NMA and included only licensed maintenance doses, which may have decreased both the number of studies contributing to the analysis and the heterogeneity between studies. However, sensitivity analyses by dose may uncover possible inconsistencies in treatment effects across dosage groups. The study publication dates covered nearly 15 years, and it is uncertain if standard of care or patients themselves have changed in that time. The model took into account variable placebo responses in an attempt to adjust for cross-study heterogeneity; however, this is only a proxy adjustment, and it is unknown if or how adequately it adjusts for important prognostic factors or effect modifiers in this patient population.

No direct evidence comparing active treatments was identified in ITC 1. Therefore, consistency was not assessable.

Detailed harms data (aside from discontinuations due to AEs) and outcomes for HRQoL were not included in the analysis, yet both of these were noted by the clinical expert and the clinician and patient groups as important to patients with epilepsy. Considering that time on treatment for ASMs must be long enough to sufficiently assess the outcome (for instance, 1-year for seizure freedom, as suggested in the NICE technology appraisal of cenobamate), follow-up times were relatively short in the studies, and long-term comparative efficacy and safety are unknown.⁹⁰ Of the 22 included studies, only 3 had study locations in Canada, while 3 were “global” (i.e., may have had locations in Canada), which limits the generalizability of the results to clinical practice in Canada. However, the clinical expert was of the opinion that, based on the reported patient characteristics and their breadth, the populations could be similar to patients with epilepsy living in Canada.

Due to the notable heterogeneity in study characteristics and populations, the incomplete list of interventions relevant to FOS treatment in Canada, and the lack of head-to-head studies, there is uncertainty in the sponsor’s NMA results, making it difficult to draw meaningful conclusions.

ITC Design – ITC 2 (NICE NMA)

Objectives

The objective of ITC 2 was to compare the efficacy and summarize the safety data of add-on ASMs in pediatric and adult patients with FOS.¹⁶ The NMA conducted by NICE defined add-on therapy as ASMs “prescribed to people that have failed to respond to 1 or more ASM or had refractory epilepsy.”¹⁶

Study Selection Methods

The Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, Embase, and MEDLINE databases were searched for studies up to February 3, 2021.¹⁶ Reference lists from systematic reviews were checked to identify studies not captured in the search.

Studies in ITC 2 enrolled patients with FOS with or without other generalized seizure types, such as absence or myoclonus seizures.¹⁶ All ages were eligible except infants younger than 28 days with acute symptomatic seizures. The analysis did not limit by any ASM type as long as the drug was used to treat the population of interest and data for different doses were pooled by drug. The 2 efficacy outcomes were a greater than 50% reduction in seizure frequency and seizure freedom. Data were taken from the treatment or maintenance periods. Safety outcomes were reported for the proportion of patients who experienced a trial-defined AE or SAE or who discontinued treatment due to an AE, and for all-cause mortality. HRQoL and AEs were outcomes of interest, but heterogeneity in data collection and reporting prevented the pooling of results or direct comparisons between drugs. Only RCTs published in English were eligible for inclusion. Studies with mixed populations of patients with epilepsy and patients without epilepsy were excluded (except where there was a subgroup analysis for the former group). Populations that mixed patients with FOS and those with generalized tonic-clonic seizures were included in the analysis because it was reasoned that most patients would have FOS.

Ten percent of the references were dual-weeded by 2 independent reviewers, with disagreements resolved through discussion until a consensus was reached or a third reviewer consulted. The remaining 90% were screened by 1 reviewer.¹⁶ Data were extracted by the same 2 reviewers, and disagreements were resolved in a similar manner. Quality assessment was performed by 1 reviewer using the Cochrane Risk of Bias 2.0 tool.⁸⁸

ITC Analysis Methods

The analysis methods for ITC 2 are presented in Table 28.

The NMA for ITC 2 was conducted similarly to ITC 1 (i.e., Bayesian Markov chain Monte Carlo methods with noninformative priors), but used only random-effects models for the analysis.¹⁶ It was believed that the same true effect size was unlikely for all studies; thus, a fixed-effect model would be inappropriate. This choice was also supported by previous NMAs reporting better goodness of fit with random-effects models over fixed-effect models. Goodness of model fit was assessed by comparing the posterior mean of the residual deviance (i.e., the magnitude of difference between model-predicted data and observed data) and the number of data points in the model. The deviance information criterion (equal to the sum of the number of effective parameters and posterior mean deviance) was also used to assess model fit. Lastly, the posterior median for the between-study SD (which measures the heterogeneity of treatment effects) was used to compare models.

For the Bayesian model, 100,000 burn-in simulations were run to allow for model convergence, then 60,000 simulations were run to estimate outputs.¹⁶ History and kernel density plots were assessed visually for convergence of the 2 chains.

Each outcome was a separate network, with each treatment a separate node, and results were pooled by treatment.¹⁶ Larger nodes indicated more patients randomized to that treatment, and the thickness of the line joining comparators indicated the number of trials for the direct comparison.

The efficacy outcomes were similar to those in ITC 1 (i.e., a 50% or greater decrease in seizure frequency and seizure freedom); however, it is unknown if the outcomes were defined exactly the same way across studies or between the 2 ITCs. Estimates of relative treatment effect were presented as odds ratios with 95% CrIs.¹⁶ Summaries for AEs were noncomparative due to heterogeneity in definitions, measurements, and reporting. Follow-up times were not detailed in the analysis, but it was indicated that all data were pooled by treatment.

Study characteristics and baseline patient characteristics were not well reported, and it is unclear if or how homogeneity was assessed among studies in the NMA.

Heterogeneity in the placebo response rate for efficacy outcomes was assessed using Cochran’s Q test of homogeneity.¹⁶ Placebo response was investigated using Spearman correlation coefficients between placebo response rate and study publication date (proxy for when a trial began) to detect placebo drift, and between placebo response rate and relative effectiveness to determine the direction of the drift. The sensitivity analysis excluding studies that excluded women of child-bearing age was not conducted because the networks did not include at least 75% of the treatments for the primary analysis. For the same reason, subgroup analyses for previous treatment and patients with treatment-resistant epilepsy were not conducted.

Consistency was assessed using an unrelated mean-effects model (i.e., comparison of fit for a model that assumes consistency against a model that allowed for inconsistency) and deviance plots (i.e., posterior mean deviance of data points from a model with consistency plotted against those from a model with inconsistency).¹⁶ Potential inconsistencies were further explored using node-splitting methods. Inconsistency was suspected if the point estimates from the direct evidence fell outside of the 95% CrI of the corresponding indirect estimates. Instances of inconsistency were checked for data accuracy. Loops with inconsistencies had baseline characteristics compared to identify possible differences in treatment-effect modifiers.

Table 28: ITC 2 Analysis Methods

AE = adverse event; CrI = credible interval; DIC = deviance information criterion; HRQoL = health-related quality of life; ITC 2 = indirect treatment comparison 2; NICE = National Institute for Health and Care Excellence; NMA = network meta-analysis; OR = odds ratio; RCT = randomized controlled trial; SAE = serious adverse event; SD = standard deviation.

Source: NICE NMA.¹⁶

Results of ITC 2

Summary of Included Studies

In total, 18,066 studies were identified from the literature search, of which 448 were retrieved for full-text review.¹⁶ After screening, 109 RCTs were included: 99 studies (27,686 patients and 26 treatments) contributed to the greater than 50% reduction in seizure frequency analysis, and 72 studies (20,826 patients and 21 treatments) contributed to the seizure freedom analysis. The placebo response meta-analyses included 73 of the 99 studies (25,824 patients) for the greater than 50% reduction in seizure frequency outcome and 41 of 72 studies (13,579 patients) for the seizure freedom outcome. Both cenobamate pivotal trials described in the CADTH main report were included in the analyses.^60,61

Study and baseline patient characteristics were not described in the NMA, making it difficult to assess heterogeneity among studies. The majority of studies were placebo-controlled, and 10 studies included an active comparator.¹⁶

Overall, 43 studies were judged to have a low risk of bias; 41 studies had some concerns for bias; and 26 had a high risk of bias.¹⁶ Greater concerns for bias were attributed mainly to 4 of the 5 domains: bias arising from the randomization process; bias due to deviations from intended interventions; bias due to missing outcome data; and bias arising from measurement of the outcome.

The assessment of homogeneity for ITC 2 is presented in Table 29.

Table 29: Assessment of Homogeneity for ITC 2

AE = adverse event; ASM = antiseizure medication; CI = confidence interval; FOS = focal onset seizure; ITC 2 = indirect treatment comparison 2; NICE = National Institute for Health and Care Excellence; NMA = network meta-analysis; NR = not reported; RCT = randomized controlled trial.

Source: NICE NMA.¹⁶

Results

Figure 7 and Figure 8 present the networks for greater than 50% reduction in seizure frequency and seizure freedom, respectively. Results were reported relative to placebo. Less than 10% of all evidence included in the NMA was direct evidence between active comparators.¹⁶

Efficacy Outcome: Greater Than 50% Reduction in Seizure Frequency

Odds ratios with 95% CrIs based on adjusted data for treatments compared to cenobamate are presented in Table 30. The results of the analyses indicate that patients who received cenobamate were more likely to achieve a greater than 50% reduction in seizure frequency than those on most other treatments, including eslicarbazepine acetate, retigabine, gabapentin, lacosamide, lamotrigine, levetiracetam, oxcarbazepine, perampanel, brivaracetam, pregabalin, rufinamide, zonisamide, and primidone.¹⁶ The relative treatment effects for cenobamate versus other comparators had wide CrIs that contained 1.0 (no difference).

Efficacy Outcome: Seizure Freedom

Relative treatment effects were not available for seizure freedom. It was reported that studies were generally underpowered for this outcome due to the rarity of the event, particularly for patients randomized to placebo, and CrIs were wide for available data.¹⁶

Figure 7: Network for Greater Than 50% Reduction in Seizure Frequency

NICE = National Institute for Health and Care Excellence; NMA = network meta-analysis.

© NICE [2022] Epilepsies in children, young people and adults: [F] Add-on therapy for generalized tonic-clonic and focal onset seizures. Available from https://www.nice.org.uk/guidance/ng217/evidence/f-addon-therapy-for-generalisedtonicclonic-and-focal-onset-seizures-pdf-398366282815 All rights reserved.

Subject to Notice of rights.

NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication.

Source: NICE NMA.¹⁶

Safety Outcomes

Safety data were limited and reported by treatment, and do not inform relative safety for cenobamate versus other ASMs. For cenobamate, 547 patients (63.4%) experienced an AE during the study; 204 patients (23.6%) discontinued treatment due to an AE; and 3 patients (0.35%) died.¹⁶

Critical Appraisal of ITC 2

In general, the ITC 2 search strategy methods were adequate, a prespecified protocol was used for the systematic review, and the choice of databases seemed appropriate. Only 10% of the literature was screened by 2 reviewers, while the remaining 90% was screened by 1 reviewer. The data extraction and verification process was performed by the same 2 reviewers. The Cochrane Risk of Bias 2.0 tool was used to evaluate the selected studies: nearly a quarter had a high risk of bias and more than a third had some concerns for bias. No studies were excluded based on the risk of bias assessment, and there was no reported plan to investigate the potential impact of these studies.

Figure 8: Network for Seizure Freedom

NICE = National Institute for Health and Care Excellence; NMA = network meta-analysis.

© NICE [2022] Epilepsies in children, young people and adults: [F] Add-on therapy for generalized tonic-clonic and focal onset seizures. Available from https://www.nice.org.uk/guidance/ng217/evidence/f-addon-therapy-for-generalisedtonicclonic-and-focal-onset-seizures-pdf-398366282815 All rights reserved.

Subject to Notice of rights

NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication.

Source: NICE NMA.¹⁶

Table 30: Comparative Efficacy for Treatments Relative to Cenobamate for Greater Than 50% Reduction in Seizure Frequency

CrI = credible interval; NICE = National Institute for Health and Care Excellence; NMA = network meta-analysis; OR = odds ratio.

Note: Bolded values indicate a CrI that does not cross 1.0. An OR greater than 1 favours cenobamate vs. the comparator.

^aAdjustments made for baseline placebo response rate using meta-regression.

^bValues were presented as the intervention relative to cenobamate in the original NICE report.

© NICE [2022] Epilepsies in children, young people and adults: [F] Add-on therapy for generalized tonic-clonic and focal onset seizures. Available from https://www.nice.org.uk/guidance/ng217/evidence/f-addon-therapy-for-generalisedtonicclonic-and-focal-onset-seizures-pdf-398366282815 All rights reserved.

Subject to Notice of rights.

NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication.

Source: NICE NMA.¹⁶

ITC 2 excluded newborns aged younger than 28 days, and — as with ITC 1 — some studies likely included nonadults. According to the clinical expert consulted by CADTH, there may be generalizability issues with extrapolating to adults from data that included pediatric patients (given that cenobamate is indicated for adults). Combined FOS and generalized tonic-clonic seizure populations were also included in the analyses. The authors of ITC 2 deemed this inclusion to be acceptable because the latter group of patients would have made up only a small proportion of the mixed populations. ITC 2 focused on “add-on therapy,” which was defined similarly to the expected place in therapy for cenobamate (i.e., not a first-line treatment). Compared to ITC 1, there were no restrictions on the type of ASM in ITC 2, and more interventions were included, including some drugs not available in Canada. The broader evidence base also included direct comparisons between active treatments, something that was lacking in ITC 1. All doses were pooled by treatment, an approach that the authors of ITC 2 justified because it allowed for drug titration and changes in dose for tolerability and AEs. While this better reflects clinical practice, it could also include doses outside of the Health Canada–indicated ranges for these ASMs. No sensitivity analyses were performed to detect potential differences in treatment effects across doses. The clinician group that submitted input for the CADTH review of cenobamate indicated that ganaxolone and zonisamide (included in ITC 2) are not available in Canada. However, given that patients may access them through Health Canada’s Special Access Program, these drugs have been included in the CADTH report. The efficacy outcomes in ITC 2 were similar to those of ITC 1 in name (i.e., a 50% reduction in seizure frequency and seizure freedom), but it is unclear if the exact definitions were consistent between the ITCs or among the included studies. The authors of ITC 2 justified the choice of outcomes based on their common inclusion in studies and because a 50% reduction would likely lead to a meaningful improvement in HRQoL. Additionally, these outcomes are recommended as trial outcomes by the Commission on Antiepileptic Drugs of the International League Against Epilepsy and were confirmed as being important to patients by the clinical expert and based on the patient and clinician group input.¹⁶ However, it was noted in the NICE technology appraisal of cenobamate that a 50% reduction in seizure frequency is not as meaningful to patients as seizure freedom. This is because the aim is to regain independence, and a reduction in the number of seizures may not result in complete patient independence.⁹⁰ As with ITC 1, data were taken from either the maintenance or treatment period, and the same issues apply with measuring seizure reduction depending on whether the titration period is included or not.⁹⁰

The authors of ITC 2 assessed the certainty of the evidence for each outcome and concluded that it was very low to moderate for the greater than 50% reduction in seizure frequency outcome. The certainty of evidence was downgraded due to indirectness, wide CrIs, and the inclusion of drugs that were used off-label for FOS. There was low to moderate certainty for the seizure freedom outcome (evidence was downgraded due to indirectness [few events and wide CrIs]). Safety outcomes were not formally assessed due to inconsistent definitions and reporting.

In total, more studies were included in ITC 2, making the networks larger than in ITC 1, yet there were still few direct comparisons of active treatments. For the ASMs common to the 2 networks, ITC 2 had 7 studies not included in ITC 1, while ITC 1 had 1 study not included in ITC 2. Therefore, it is possible that important studies were missed from either network of evidence. The authors of ITC 2 reasoned that it was unlikely there would be a single true effect size for all studies; they chose a random-effects model over a fixed-effect model, which the CADTH review team agreed was appropriate. The ITC 2 authors also noted that previous NMAs for FOS reported better fit with a random-effects model. No preplanned subgroup or sensitivity analyses were conducted because their expected networks did not include at least 75% of the original network’s treatments.

Baseline patient characteristics were not reported in ITC 2, and it was not possible to assess the similarity assumption.

Patient and study characteristics were not reported in ITC 2, making it challenging to compare the heterogeneity of the studies. A possible source of heterogeneity is the wide range in study publication dates (1990 to 2020); it is possible that management of FOS and patients with FOS have changed in that time. Statistical heterogeneity between studies was assessed for placebo response rates for each outcome and was significant for the greater than 50% reduction in seizure frequency outcome (P < 0.001), but not for seizure freedom (P = 0.30). The results of the NMA indicated a changing placebo response rate over time. This finding may also suggest that the study populations are different and support the likeliness of heterogeneity.⁹⁰ Although placebo response was adjusted for in ITC 2, it is unknown if this was an adequate adjustment for patient characteristics and effect modifiers that may bias the results. Moreover, no separate analyses were performed to adjust for the impact of the previously mentioned potential treatment-effect modifiers.

Less than 10% of all evidence in ITC 2 was direct evidence between active comparators. The results of the consistency assessment were not reported.

It was unclear how many studies took place in Canada, and the lack of reporting of patient characteristics makes it difficult to know how applicable the results are to practice in Canada. HRQoL was noted as important to patients, according to the stakeholder input for this CADTH review, but due to the use of various HRQoL instruments and a general lack of outcome reporting, results for HRQoL were not presented in ITC 2. Safety data were reported as number of patients and proportions without comparisons, and the variation in the definitions of harms in the trials makes it difficult to draw conclusions from these results. There is a need for better and long-term HRQoL and safety data as well as direct comparisons between ASMs.

Relative treatment effects for cenobamate had relatively wide 95% CrIs for a number of comparisons, and while several CrIs excluded the threshold of no difference, including those for comparators in ITC 1, there were other comparisons where the 95% CrIs included the null. This uncertainty in the results, along with the previously mentioned limitations in the analysis, makes it challenging to draw any meaningful conclusions with confidence.

Studies Addressing Gaps in the Pivotal and RCT Evidence

Contents within this section have been informed by materials submitted by the sponsor. The following subsections have been summarized and validated by the CADTH review team.

Six studies were submitted by the sponsor to address gaps in the pivotal and RCT evidence, including a study reporting the outcome of quality of life in epilepsy,¹⁸ a study among adults with a developmental disability,¹⁹ and 4 other real-world evidence studies.^20,91-93 Of these, 3 real-world evidence studies are not included in this report because the studies are available only as posters and contain insufficient information to critically appraise these studies.^20,91-93

Additionally, 1 sponsor-submitted phase III safety trial has been summarized in this section: Study C021 is an open-label, multicentre, long-term safety and pharmacokinetic study of cenobamate as adjunctive therapy in patients with partial onset seizures, with a larger sample size.^17,94

Description of Studies

A summary of the studies from sponsor’s submission is shown in Table 31.

Table 31: Summary of 3 Sponsor-Submitted Studies Addressing Gaps in the Evidence

AE = adverse event; ASM = antiseizure medication; QOLIE-31 = 31-item Quality of Life in Epilepsy Questionnaire; RCT = randomized controlled trial; SD = standard deviation.

^aOne patient reported 2 AEs: drowsiness and ataxia.

Source: Elizebath et al. (2021);¹⁸ Connor et al. (2022);¹⁹ Elliott et al. (2022).²⁰

Study C021

The objective of Study C021 was to evaluate the safety and pharmacokinetics of cenobamate and concomitant ASMs when administered as adjunctive therapy for the treatment of partial onset seizures. The additional objective of this study was to characterize the rate of DRESS using a lower starting dose and a slower titration rate. According to studies of healthy volunteers receiving phenytoin and phenobarbital, cenobamate increases plasma levels of phenobarbital and phenytoin levels. A portion of this open-label safety and pharmacokinetics study was designed to understand the impact of adding cenobamate to an ASM regimen, including either phenytoin or phenobarbital in patients with partial onset epilepsy. In addition, this open-label and pharmacokinetics study enrolled patients taking concomitant ASMs.³⁹

Study C021 enrolled patients with poorly controlled partial seizures. It consisted of a 21-day screening period and a 12-month, open-label treatment period. The treatment period included a 12-week titration phase followed by an open-label maintenance phase. For patients discontinuing, it consisted of a taper period and a follow-up visit.

Populations

According to the eligibility criteria, patients had to:

• be adults aged 18 years to 70 years with a diagnosis of focal (partial onset) epilepsy, according to the International League Against Epilepsy’s classification of epileptic seizures

• be experiencing uncontrolled focal seizures despite treatment with at least 1 ASM within the past 2 years

• have an electroencephalographic reading consistent with the diagnosis of focal epilepsy and a CT or MRI scan performed within the previous 10 years to rule out a progressive cause of epilepsy

• be currently taking a stable dose of 1 to 3 concomitant ASMs for at least 3 weeks.

The first patient in Study C021 was screened on November 3, 2015, and the last patient visit was on March 31, 2021.³⁹

To note, there were no sites in Canada listed as investigation sites for this study.

Interventions

Most patients were supplied with open-label cenobamate 12.5 mg, 25 mg, 50 mg, and 100 mg tablets to be taken orally once daily. Patients in the US were also supplied with 150 mg and 200 mg tablets. The study drug could be taken with or without food. The target dose was 200 mg per day. After reaching the target dose, all patients were allowed to titrate up to a maximum dose of 400 mg per day.

Study patients were supplied with cenobamate at daily doses of 12.5 mg in weeks 1 and 2, 25 mg in weeks 3 and 4, 50 mg in weeks 5 and 6, 100 mg in weeks 7 and 8, 150 mg in weeks 9 and 10, and 200 mg in weeks 11 and 12 during a titration phase; they could receive a maximum dose of 400 mg per day for more than 12 months in the maintenance or dose-optimization phase. The titration phase included 7 study visits (visit 2 to visit 8), and the maintenance phase included up to 6 visits (visit 9 to visit 14). During the first 16 weeks of treatment, visits occurred every 2 weeks. The follow-up visit (visit 15) was to occur 14 days after the last dose. Patients who benefited from treatment may have continued beyond year 1, with visits occurring every 3 months.^39,94

Outcomes

The safety outcomes and assessment intervals included:

• frequency and severity of reported AEs and SAEs and their relation to cenobamate treatment

• physical examinations to identify evidence of hypersensitivity signs (every 2 weeks during the first 4 months of treatment, at year 1, and yearly thereafter)

• in-depth reviews of all hypersensitivity reactions monthly to screen for DRESS.

The end point measurement was at 12 months. After 12 months, patients were re-evaluated and could continue at the discretion of investigator.³⁹

Statistical Analysis

The study planned to enrol at least 1,000 patients and expose them to cenobamate for at least 6 months. At least 20 patients taking phenytoin and at least 20 patients taking phenobarbital were to be enrolled. Additional patients taking ASMs other than phenytoin or phenobarbital were also enrolled to expose at least 1,000 patients for 6 months to further evaluate the long-term safety of cenobamate and the drug interactions with concomitant ASMs.³⁹

Descriptive statistics were used to summarize safety data.³⁹

The pharmacokinetic analyses are beyond the scope of the CADTH review; thus, these are not summarized in the current report. Also, post hoc analyses of cenobamate efficacy (seizure outcomes) were conducted among some subgroups of patients. For example, a subset population in Study C021 who had focal aware motor, focal impaired awareness, or focal to bilateral tonic-clonic seizures (N = 240) were evaluated for efficacy data in a post hoc analysis.⁹⁵ These efficacy outcomes are not included in the current report because these were not prespecified in the study protocol of Study C021.¹⁷

Results

Patient Disposition

All patients who successfully met the entry criteria and provided informed consent to participate in the study were considered as enrolled patients. Safety-evaluable patients were defined as all patients enrolled in the study who received at least 1 dose of study drug medication.

Baseline Characteristics

The baseline characteristics of patients at baseline in Study C021 are summarized in Table 32.

Table 32: Summary of Baseline Characteristics of Study C021 (Safety Population)

ASM = antiseizure medication; BMI = body mass index; SD = standard deviation.

^aThe safety population was defined as patients who received greater than or equal to 1 dose of cenobamate.

^bN = 1,336 for this variable.

^cPatients may be reported in more than 1 category.

^dBaseline ASMs were defined as ASMs that started before and were ongoing at the time of the first dose of cenobamate.

^eConcomitant ASMs were defined as ASMs that started before and were ongoing at the time of the first dose of cenobamate or that started after the first dose of cenobamate.

Source: Clinical Study Report for Study C021.¹⁷

Exposure to Study Treatments

Details of patient disposition and exposure to cenobamate are summarized in Table 33. Of the 1,484 patients screened for Study C021, 1,345 patients (91%) were enrolled in the study.⁹⁴

Table 33: Patient Disposition for Study C021

EAP = Expanded Access Program.

^aIncluded patients who completed the study as per end of study status case report form.

^bReceived cenobamate through the EAP, which distributes non-commercial drugs outside the US, or the Navigator Program, which is for commercial distribution of drugs to patients inside the US.

^cThe safety population was defined as patients who received 1 or more doses of cenobamate.

Source: Clinical Study Report for Study C021.¹⁷

Exposure to Study Treatments

Details of patient exposure to cenobamate are summarized in Table 34.

Table 34: Patient Exposure in Study C021