CADTH Reimbursement Review

Lemborexant (Dayvigo)

Sponsor: Eisai Limited

Therapeutic area: Insomnia

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: Submitted for Review

NOC = Notice of Compliance.

Source: Sponsor’s submission package for review of lemborexant.¹

Introduction

Insomnia disorder is the most common sleep disorder and, according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5), is described as being dissatisfied with the quality or quantity of sleep and having difficulty initiating and/or maintaining sleep, and is associated with daytime impairment.^2,3 The DSM-5 criteria include sleep disturbances occurring at least 3 nights per week for at least 3 months.^3,4 In Canada, the prevalence is estimated to be between 12% and 24%^5,6 and the incidence is estimated to be between 3.8% and 7.3% per year.^7,8 Older age, female sex, comorbid medical or psychiatric conditions, and social and environmental factors have been identified as common risk factors for insomnia.^2,3 Insomnia can be diagnosed as a disorder on its own or as a symptom associated with many other medical conditions and has been closely linked to reduced life expectancy and increased economic costs through lost productivity, workplace accidents and motor vehicle accidents (MVAs), and greater health care usage.^2,3,7-10

Per the Alberta Medical Association’s 2015 Assessment to Management of Adult Insomnia clinical practice guideline, insomnia can be treated and managed in a primary care setting or by a specialist when the problem cannot be diagnosed or if the insomnia is resistant to treatment.³ Nonpharmacologic treatment options such as sleep hygiene education and cognitive behavioural therapy for insomnia (CBT-I) are recommended as an initial treatment.^3,10,11 Pharmacologic treatments include benzodiazepine receptor agonists (BZRAs) — among them, benzodiazepines and Z-drugs (zolpidem tartrate [ZOL], eszopiclone, zaleplon, and zopiclone) — dual orexin receptor antagonists (DORAs), histamine receptor antagonists, and melatonin receptor agonists.¹² Off-label medications that have sedating effects (e.g., antidepressants, antihistamines, anticonvulsants, antipsychotic drugs) may be used in specific situations but are generally not recommended as a first-line treatment due to the lack of efficacy and safety evidence in this population.^3,12 Long-term use of hypnotics is discouraged as the studies supporting their approval were based on short-term use² and are not recommended as a first-line treatment option.^3,12

Lemborexant (LEM) is a DORA indicated for the treatment of insomnia, characterized by difficulties with sleep onset and/or sleep maintenance.¹ The recommended dosage is 5 mg once per night and may be increased to the maximum recommended dose of 10 mg based on clinical response and tolerability.¹³ LEM is taken orally once per night within a few minutes before going to bed, with at least 7 hours before planned awakening time.

The objective of this report is to perform a systematic review of the beneficial and harmful effects of lemborexant 5 mg (LEM5) or lemborexant 10 mg (LEM10) tablets taken orally once daily at bedtime for the treatment of patients with insomnia, characterized by difficulties with sleep onset and/or sleep maintenance.

Following the issuance of the draft CADTH Canadian Drug Expert Committee (CDEC) recommendation for LEM in August 2022, the following additional information was provided to CADTH:

• 1 unpublished manuscript by Drake et al.¹⁴ that provided evidence for LEM’s efficacy as measured by the Patient Global Impression–Insomnia (PGI-Insomnia) tool in the SUNRISE 2 study

• a periodic safety update report (PSUR)¹⁵ that included aggregate data from approximately 4,024 adult patients with insomnia disorder, irregular sleep-wake rhythm disorder (ISWRD), or healthy volunteers who were enrolled in the LEM clinical development program

• a published network meta-analysis (NMA) by De Crescenzo et al. (2022)¹⁶ comparing LEM to other pharmacologic treatments for the acute and long-term treatment of adults with insomnia.

These data were not included in the submission to CADTH and provided more information on the clinical meaningfulness of the subjective PGI-Insomnia instrument, and on LEM’s safety and comparison to other drugs for the treatment of insomnia. The information has been summarized and critically appraised as an addendum to the CADTH report in Appendix 6.

Stakeholder Perspectives

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

Patient Input

CADTH received input from 3 patient groups: the Mood Disorders Society of Canada (MDSC), Migraine Canada, and Menopause Chicks. The 3 groups conducted surveys and interviews with patients and caregivers to inform their input. Respondents from all 3 patient groups indicated that sleep problems significantly impacted their quality of life, energy level, cognitive function, mood the next day, and daytime activities. Most respondents reported having tried various treatments for sleep problems (e.g., benzodiazepines, Z-drugs). Many reported being dissatisfied and discontinued treatment due to side effects such as next-day sedation and cognitive impairment, and fear of developing a substance use disorder. The groups indicated the following as key outcomes for patients and caregivers: uninterrupted and restorative sleep, greater access to treatment, more effective treatment options, long-term effectiveness, fewer side effects, less stress and anxiety, improved productivity, and improved relationships. MDSC input included 3 respondents who had experience with LEM and accessed the drug through private health insurance. They described being able to manage their sleep problems without becoming dependent on the medication or experiencing serious side effects, and without feeling lethargic and sleepy the next morning.

Clinician Input

Input From the Clinical Expert Consulted by CADTH

The clinical expert consulted for this review stated that the current goals of treatment are to improve sleep initiation, maintenance, and terminal insomnia, leading to restorative sleep. According to the clinical expert, CBT-I is the first-line treatment for insomnia, but many patients have limited access to and/or success with this option; thus, pharmacotherapy is often used in addition to or in place of CBT-I. Per the expert, when a medication is used for extended periods of time, a patient may have a waning response or no response. At this point, it can be difficult to withdraw the therapy if the patient has developed tolerance to the drug and the patient may experience rebound insomnia. Another concern raised by the clinical expert was the risk of cognitive and behavioural changes the next morning that can lead to falls and other dangers, particularly for older adult patients.

The clinical expert suggested that LEM may be a first-line pharmacologic treatment for insomnia and noted that it would be necessary to determine how to optimally transition from other currently available medications (e.g., Z-drugs) to LEM. The expert indicated that most patients with insomnia may be candidates for treatment with LEM except for those who are pregnant, are nursing, or have narcolepsy. According to the clinical expert, patients are asked general questions about their sleep rather than have measurements taken in clinics. The expert thought that this practice does not tend to vary among physicians. The clinical expert indicated that few clinics are performing sleep studies for insomnia (except where another sleep disorder is suspected) because access is limited across Canada.

The expert noted that chronic insomnia generally does not go away and tends to worsen with age, menopause, or as a result of other major negative life events. The expert suggested that a patient may wish to trial being off medication; they can be supported by additional sleep management tools such as CBT-I. When deprescribing a medication, the clinical expert would observe for a return of symptoms or rebound insomnia. Per the clinical expert, insomnia is typically treated by family physicians or psychiatrists.

Clinician Group Input

CADTH received input from 2 clinician groups: the Canadian Consortium of Sleep and Sleep Interested Physicians (CCSSP) and the National Advisory Board. The clinician group input was aligned with that given by the clinical expert consulted by CADTH.

Drug Program Input

Input was obtained from the drug programs that participate in the CADTH reimbursement review process. Key issues raised by the drug plans included concerns over the most appropriate comparator to LEM, the necessity for patients to fail a treatment for insomnia before accessing LEM, the exclusion of comorbid conditions, and the use of polysomnography (PSG) for measuring efficacy. The clinical expert indicated the uncertainty of ZOL being the most appropriate comparator and how it compared to zopiclone, the latter of which was noted as being more commonly used in Canadian practice. The expert also stated that it would not be reasonable for patients to have to fail a less expensive drug with greater risk of harms before accessing safer alternatives if the 2 drugs had similar efficacy. Further, the clinical expert noted that there is currently a lack of evidence supporting the use of LEM in patients with comorbid conditions that were excluded from the SUNRISE 1 and SUNRISE 2 studies (based on Beck Depression Inventory [BDI-2], Beck Anxiety Inventory [BAI], and Apnea-Hypopnea Index [AHI] scores), but given that insomnia is closely linked to anxiety and depression, it would be very difficult to exclude these comorbidities until trials could be performed. Lastly, the expert stated that PSG results should not be used as a marker for efficacy or a necessary criterion for accessing LEM.

Clinical Evidence

Pivotal Studies and Protocol Selected Studies

Description of Studies

Two double-blind (DB), phase III, randomized controlled trials (RCTs) — SUNRISE 1 and SUNRISE 2 — were included in the systematic review of LEM.

The SUNRISE 1 trial (N = 1,006) was designed to assess the safety and efficacy of LEM5 and LEM10 for 30 days in females who were 55 years or older and males who were 65 years or older, all of whom had insomnia disorder according to the DSM-5. Comparators included zolpidem tartrate extended-release (ZOL-ER) 6.25 mg and appearance-matched placebos (PBOs) for both active compounds. The primary outcome was latency to persistent sleep (LPS) and key secondary outcomes were wake after sleep onset (WASO), wake after sleep onset in the second half of the night (WASO2H), and sleep efficiency as measured by PSG. Other secondary and exploratory outcomes important to the CADTH review included patient-reported outcomes such as the 3-level EQ-5D (EQ-5D-3L), Insomnia Severity Index (ISI), Fatigue Severity Scale (FSS), PGI-Insomnia, and quality of sleep questionnaires. Patients had a mean age of 63.9 (standard deviation [SD] = 6.81) years and 86.4% were female. At baseline, mean LPS was approximately 45 minutes, mean WASO was around 114 minutes, mean WASO2H was 77 minutes, and mean sleep efficiency was 68%.

The SUNRISE 2 trial (N = 971) was designed to assess the long-term safety and efficacy of LEM5 and LEM10 for up to 12 months in adults 18 years or older who had insomnia disorder according to the DSM-5. The first 6 months of the SUNRISE 2 study consisted of 3 treatment groups: LEM5, LEM10, and PBO. For the next 6 months of the trial, patients who were randomized to PBO were rerandomized to active treatment (i.e., LEM5 or LEM10) for the remainder of the trial. The primary outcome was subjective sleep onset latency (sSOL) and key secondary outcomes were subjective wake after sleep onset (sWASO) and subjective sleep efficiency as recorded in the sleep diary. Other secondary and exploratory outcomes important to the CADTH review were the same as the patient-reported outcomes listed for SUNRISE 1. Patients had a mean age of 54.5 (SD = 13.80) years and 68.2% were female. At baseline, mean sSOL was approximately 64 minutes, sWASO was 134 minutes, and subjective sleep efficiency was 62%.

Efficacy Results

Statistical testing was conducted based on a gatekeeping procedure for both studies. In SUNRISE 1, the following outcomes were controlled for multiplicity: LPS, sleep efficiency, WASO, and WASO2H. In SUNRISE 2, the following outcomes were controlled for multiplicity: sSOL, subjective sleep efficiency, and sWASO. Efficacy data have been summarized in Table 2 for SUNRISE 1 and Table 3 for SUNRISE 2.

Table 2: Summary of Key Efficacy Results From the SUNRISE 1 Study, FAS

C-SSRS = Columbia-Suicide Severity Rating Scale; CI = confidence interval; FAS = full analysis set; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; LPS = latency to persistent sleep; LSM = least squares mean; MMRM = mixed model of repeated measures; MVA = motor vehicle accident; NA = not applicable; NR = not reported; PBO = placebo; Q1 = first quartile; Q3 = third quartile; SAE = serious adverse event; SD = standard deviation; SE = standard error; sSOL = subjective sleep onset latency; sWASO = subjective wake after sleep onset; TEAE = treatment-emergent adverse event; vs. = versus; WASO = wake after sleep onset; WASO2H = wake after sleep onset in the second half of the night; WDAE = withdrawal due to adverse event; ZOL = zolpidem tartrate.

^aThe number of patients analyzed for the change from baseline at this time point (not the number of patients who provided data at this time point).

^bBased on an MMRM model with log transformation of LPS and factors of age group, region, treatment, visit (days 1 to 2, and days 29 to 30), and treatment-by-visit interaction as fixed effects, and the baseline LPS as a covariate. Missing values were imputed using multiple imputation and assumed to be missing not at random.

^cP value has been adjusted for multiple testing (i.e., the type I error rate has been controlled).

^dBased on an MMRM model with factors of age group, region, treatment, visit (days 1 to 2, and days 29 to 30), and treatment-by-visit interaction as fixed effects, and the baseline WASO as a covariate. Missing values were imputed using multiple imputation and assumed to be missing not at random.

^eBased on an MMRM model with factors of age group, region, treatment, visit (days 1 to 2, and days 29 to 30), and treatment-by-visit interaction as fixed effects, and the baseline WASO2H as a covariate. Missing values were imputed using multiple imputation and assumed to be missing not at random.

^fBased on an MMRM model with factors of age group, region, treatment, visit (days 1 to 2, and days 29 to 30), and treatment-by-visit interaction as fixed effects, and the baseline SE as a covariate. Missing values were imputed using multiple imputation and assumed to be missing not at random.

^gSafety results were based on the FAS population.

Source: SUNRISE 1 Clinical Study Report.¹⁷

Table 3: Summary of Key Efficacy Results From the SUNRISE 2 Study, FAS

AE = adverse event; C-SSR = Columbia-Suicide Severity Rating Scale; CCMV = complete case missing value; CI = confidence interval; FAS = full analysis set; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; LSM = least squares mean; MMRM = mixed model of repeated measures; MNAR = missing not at random; MVA = motor vehicle accident; NA = not applicable; NR = not reported; PBO = placebo; Q1 = first quartile; Q3 = third quartile; SAE = serious adverse event; SD = standard deviation; sSOL = subjective sleep onset latency; sWASO = subjective wake after sleep onset; vs. = versus; WDAE = withdrawal due to adverse event.

^aThe number of patients analyzed for the change from baseline at this time point (not the number of patients who provided data at this time point).

^bBased on an MMRM model with log transformation of sSOL and factors of age group, region, treatment, visit (first 7 nights, month 1, month 3, and month 6), and treatment-by-visit interaction as fixed effects, and the study baseline sSOL as a covariate. Missing values were imputed using multiple imputation and assumed to be MNAR or CCMV.

^cP value has been adjusted for multiple testing (i.e., the type I error rate has been controlled).

^dBased on an MMRM model with factors of age group, region, treatment, visit (first 7 nights, month 1, month 3, and month 6), and treatment-by-visit interaction as fixed effects, and the study baseline sWASO as a covariate. Missing values were imputed using multiple imputations and assumed to be MNAR or CCMV.

^eBased on an MMRM model with factors of age group, region, treatment, visit (first 7 nights, month 1, month 3, and month 6), and treatment-by-visit interaction as fixed effects, and the study baseline subjective sleep efficiency as a covariate. Missing values were imputed using multiple imputation and assumed to be MNAR or CCMV.

Source: SUNRISE 2 Clinical Study Report.¹⁸

Health-Related Quality of Life and Severity of Symptoms

The clinical expert emphasized the importance of patient-reported outcomes such as the EQ-5D-3L, ISI, FSS, PGI-Insomnia, and quality of sleep questionnaires. For the EQ-5D-3L visual analogue scale in the SUNRISE 1 trial, the least squares mean (LSM) treatment difference for LEM10 versus PBO was 2.53 (95% confidence interval [CI], 0.69 to 4.38) and for LEM5 versus PBO was 0.52 (95% CI, –1.32 to 2.37). In the SUNRISE 2 trial, the LSM treatment difference for LEM10 versus PBO was 1.04 (95% CI, –1.36 to 3.43) and for LEM5 versus PBO was –0.96 (95% CI, –3.31 to 1.39). All patient-reported outcome questionnaires were other secondary or exploratory outcomes and were not adjusted for multiplicity; therefore, definitive conclusions could not be made from the results.

Sleep Latency (Sleep Onset): Latency to Persistent Sleep, Subjective Sleep Onset Latency

In SUNRISE 1, the geometric LSM treatment ratio for LPS for LEM10 compared to PBO was 0.72 (95% CI, 0.63 to 0.83; P < 0.0001) and for LEM5 compared to PBO was 0.77 (95% CI, 0.67 to 0.89; P = 0.0003).

In SUNRISE 2, the geometric LSM treatment ratio for LPS for LEM10 compared to PBO was 0.70 (95% CI, 0.61 to 0.81; P < 0.0001) and for LEM5 compared to PBO was 0.73 (95% CI, 0.64 to 0.84; P < 0.0001).

Waking After Sleep Onset (Sleep Maintenance): WASO, WASO2H, sWASO

In the SUNRISE 1 study, the LSM treatment differences for WASO for LEM10 versus PBO was –25.35 minutes (95% CI, –31.36 minutes to –19.34 minutes; P < 0.0001) and for LEM5 versus PBO was –23.96 minutes (95% CI, –29.98 minutes to –17.95 minutes; P < 0.0001). The LSM treatment differences for WASO2H for LEM10 versus ZOL was –8.00 minutes (95% CI, –12.53 minutes to –3.47 minutes; P = 0.0005) and for WASO2H for LEM5 versus ZOL was –6.65 minutes (95% CI, –11.15 minutes to –2.15 minutes; P = 0.0038).

In the SUNRISE 2 study, the LSM treatment differences for LEM10 versus PBO was –12.67 minutes (95% CI, –22.38 minutes to –2.96 minutes; P = 0.0105) and for LEM5 versus PBO was –17.47 minutes (95% CI, –27.31 minutes to –7.64 minutes; P = 0.0005).

Sleep Efficiency: Sleep Efficiency and Subjective Sleep Efficiency

In SUNRISE 1, the LSM treatment differences for LEM10 versus PBO was 8.03% (95% CI, 6.57% to 9.49%; P < 0.0001) and for LEM5 versus PBO was 7.07% (95% CI, 5.61% to 8.54%; P < 0.0001).

In SUNRISE 2, the LSM treatment differences for LEM10 versus PBO was 4.67% (95% CI, 2.37% to 6.96%; P < 0.0001) and for LEM5 versus PBO was 4.55% (95% CI, 2.24% to 6.86%; P = 0.0001).

Harms Results

Harms data have been summarized in Table 2 for SUNRISE 1 and Table 3 for SUNRISE 2.

Adverse Events

In the SUNRISE 1 study, approximately one-third of patients experienced a treatment-emergent adverse event (TEAE). Rates were relatively similar among all groups: 82 (30.6%) patients, 74 (27.8%) patients, 93 (35.4%) patients, and 53 (25.4%) patients in the LEM10, LEM5, ZOL, and PBO groups, respectively. In the SUNRISE 2 study, more than half of the patients experienced a TEAE and rates were similar among all groups: 187 (59.6%) patients, 192 (61.1%) patients, and 200 (62.7%) patients in the LEM10, LEM5, and PBO groups, respectively. The most common events in both studies were headache, somnolence, and nasopharyngitis.

Serious Adverse Events

Serious adverse events (SAEs) were rare in both studies. In SUNRISE 1, 2 (0.8%) patients and 4 (1.5%) patients in the LEM5 and ZOL groups, respectively, experienced at least 1 SAE. No patients in the LEM10 or PBO groups reported an SAE. In SUNRISE 2, 9 (2.9%) patients, 7 (2.2%) patients, and 5 (1.6%) patients in the LEM10, LEM5, and PBO groups, respectively, experienced at least 1 SAE. No SAEs occurred in more than 1 patient per treatment group.

Withdrawals Due to Adverse Events

In general, there were few withdrawals from treatment due to adverse events (AEs) in both studies. In SUNRISE 1, 3 (1.1%) patients, 2 (0.8%) patients, 7 (2.7%) patients, and 2 (1.0%) patients in the LEM10, LEM5, ZOL, and PBO groups, respectively, stopped treatment due to an AE. No events occurred in more than 1 patient in any treatment group. In SUNRISE 2, 26 (8.3%) patients, 13 (4.1%) patients, and 12 (3.8%) patients in the LEM10, LEM5, and PBO groups, respectively, stopped treatment due to an AE. The following events occurred in more than 1 patient in any group: headache, somnolence, nightmare, and palpitations.

Mortality

No deaths were reported in either the SUNRISE 1 or SUNRISE 2 trial.

Notable Harms

In both studies, rates of somnolence were numerically higher among patients who received LEM10 compared to LEM5 (7.1% versus 4.1% in the SUNRISE 1 study and 13.1% versus 8.6% in the SUNRISE 2 study) and were greater than rates in either the ZOL group (1.5%) or PBO group (1.9%) in SUNRISE 1 or the PBO group (1.6%) in SUNRISE 2.

Falls were rare in SUNRISE 1 (1.5% in the LEM5 group and 0% for all other groups) and in SUNRISE 2 (1.6% in each of the LEM10 and LEM5 groups and 3.1% for the PBO group).

A road traffic accident (MVA for the CADTH systematic review protocol) was described for 1 patient who received ZOL in SUNRISE 1; no other treatment groups reported an MVA. In SUNRISE 2, 1 patient each in the LEM10 and PBO groups reported an MVA and no MVAs were reported in the LEM5 group.

Reports of hallucinations were rare in both studies. In SUNRISE 1, tactile hallucination was reported for 1 patient who received LEM10 (0 patients for all other treatments). In SUNRISE 2, hypnagogic hallucination was reported for 3 patients (2 patients who received LEM10 and 1 patient who received LEM5) while hypnopompic hallucination was reported for 1 patient who received LEM10. The PBO group did not have any reports of hallucinations.

No intentional overdoses were reported in the SUNRISE 1 trial. Intentional overdoses were reported for 2 patients who received LEM5, 1 patient who received PBO, and 0 patients who received LEM10 in the SUNRISE 2 trial.

Neither trial reported impaired driving or workplace accidents.

Postural instability was assessed in SUNRISE 1. On days 2 to 3, the LSM treatment difference between the LEM10 group and the PBO group was 2.91 units (95% CI, –0.28 units to 6.10 units) and between LEM5 and PBO was 2.49 units (95% CI, –0.70 units to 5.67 units). The LSM treatment difference between the LEM10 group and the ZOL group was –4.29 units (95% CI, –7.32 units to –1.26 units) and between the LEM5 group and the ZOL group was –4.71 units (95% CI, –7.73 units to –1.70 units). On days 30 to 31, the LSM treatment difference between the LEM10 group and the PBO group was –0.58 units (95% CI, –3.68 units to 2.53 units) and between the LEM5 group and the PBO group was –0.71 units (95% CI, –3.80 units to 2.38 units). The LSM treatment difference between the LEM10 group and the ZOL group was –2.57 units (95% CI, –5.53 units to 0.39 units) and between the LEM5 group and the ZOL group was –2.70 units (95% CI, –5.64 units to 0.23 units). Since there was no adjustment for multiplicity, the results are uncertain.

Impaired attention was assessed in the SUNRISE 1 study using 2 components of the Cognitive Performance Assessment Battery (CPAB): power of attention and continuity of attention. On both days 2 to 3 and days 30 to 31, the mean change from baseline for power of attention decreased for the PBO group and increased for the LEM10, LEM5, and ZOL groups. On days 2 to 3, the mean change from baseline for continuity of attention increased for the LEM5 group and decreased for the LEM10, PBO, and ZOL groups. On days 30 to 31, the mean change from baseline for continuity of attention decreased for all groups. Since there was no adjustment for multiplicity for this outcome, conclusions based on the results cannot be drawn with certainty.

In SUNRISE 1, differences in reports of rebound insomnia during the follow-up period between the treatment groups were not tested statistically. The frequency of rebound insomnia appeared similar between the LEM10 and LEM5 groups based on sSOL measures (17% to 21% for the first 7 nights of follow-up and 22% to 24% for the last 7 nights of follow-up) and proportions were numerically higher for the ZOL and PBO groups (23% to 27% for the first 7 nights and 23% to 27% for the last 7 nights). The results using sWASO measures were similar for the LEM10 and LEM5 groups (16% to 19% for the first 7 nights and 18% to 19% for the last 7 nights) as well as the ZOL and PBO groups (15% to 22% for the first 7 nights and 18% for the last 7 nights). Rates of rebound insomnia were generally lower in SUNRISE 2 and similar between the LEM10 and LEM5 groups based on sSOL measures (11% to 12% for the first 7 nights and 9% to 12% for the last 7 nights) and sWASO measures (12% to 14% for both the first and last 7 nights).

Rates of withdrawal symptoms were similar among the LEM10, LEM5, ZOL, and PBO groups in SUNRISE 1 (10.0%, 11.6%, 14.7%, and 14.1%, respectively) and for the LEM10 and LEM5 groups in SUNRISE 2 (16.8% and 20.7%, respectively).

In both the SUNRISE 1 and SUNRISE 2 studies, suicidal ideation was reported in no more than 3 patients in any treatment group at any postbaseline time point.

Critical Appraisal

Both the SUNRISE 1 and SUNRISE 2 trials appeared to have appropriate methods for blinding to the assigned treatment, randomization with stratification, and adequate power for the primary and secondary outcomes. Adjustments for multiplicity were made for all primary and key secondary outcomes and the type I error was controlled for in both studies. All primary and key secondary outcomes were objective PSG measures (SUNRISE 1) or subjective measures based on sleep diary responses (SUNRISE 2). The sponsor noted the importance of having objective outcomes to assess the physiologic effect of the medication along with subjective outcomes to measure the patient’s perception of the medication’s effect. According to the clinical expert, PSG results may not always be interpreted meaningfully or consistently when compared side-by-side with subjective or patient-reported outcomes and consequently, may not be the most meaningful marker of efficacy. The expert further emphasized that insomnia is a subjective issue; therefore, patient-reported outcomes and perceptions of sleep changes may be more appropriate for assessing treatment effect. With all subjective measures, there is a risk of bias that cannot be measured and leads to uncertainty of how meaningful the results are. The direction of the treatment effect for objective and subjective measures aligned in the SUNRISE 1 study, which contributes to the certainty of the results. The numerically higher rates of discontinuations in the LEM10 group compared to the LEM5 group may bias the results, though the magnitude and direction of bias is unknown. There was some amount of missing data at postbaseline visits for all outcomes in both trials, particularly for long-term end points in the SUNRISE 2 study; this prevents strong conclusions from being made. Prespecified subgroup analyses by age were considered exploratory, may not have been powered to detect a treatment difference, and were not adjusted for multiplicity, and there was variability in the change from baseline results (noted by large SDs and interquartile ranges [IQRs]). For responder analyses, SUNRISE 1 was not powered to detect a treatment difference and neither trial was adjusted for multiplicity. As a result of these limitations, conclusions could not be drawn from either the subgroup or responder analysis results.

In general, the clinical expert consulted for this review confirmed that the populations of the SUNRISE 1 and SUNRISE 2 trials were similar to those of patients seen in Canadian clinics and that the trial results would be generalizable with some limitations. There was a large proportion of individuals screened out before randomization, thus producing a study population that may not adequately represent the broader Canadian population with insomnia who would otherwise be eligible for treatment with LEM. Eligibility for SUNRISE 1 was restricted to females 55 years or older and males 65 years or older and the clinical expert stated that the generalizability of the results would be limited to patients matching these demographics. Further, both SUNRISE 1 and SUNRISE 2 excluded a number of comorbid conditions (e.g., based on BDI-2 and BAI scores) and had AHI cut-offs that the clinical expert suggested may have captured individuals with insomnia related to mild sleep apnea rather than psychophysiological insomnia. It is uncertain how applicable the trial results would be to patients with the excluded comorbidities or different AHI scores. No dose changes were allowed during the studies and it is noted in the Health Canada product monograph that the recommended nightly dose for LEM is 5 mg, which may be increased to 10 mg based on clinical response and tolerability.¹⁹ ZOL was the active comparator in SUNRISE 1 and discussions with the clinical expert and a representative from the Canadian public drug plans indicated that ZOL is not publicly funded by any drug plans in Canada and is less commonly used for the treatment of insomnia; therefore, it is uncertain if it is the most appropriate comparator for a Canadian setting. Patients were required to maintain a sleep diary throughout the course of both studies, and the clinical expert stated that a sleep diary may not be required in clinical practice. Thus, screening out patients who could not comply with completing a daily sleep diary excluded patients who could be candidates for LEM. Most outcomes identified in the input received by CADTH from patient groups aligned with efficacy and harms outcomes in the studies, though there are still gaps in the evidence for the use of LEM in patients with comorbid conditions and alongside other medications.

Indirect Comparisons

Description of Studies

One sponsor-submitted indirect treatment comparison (ITC) (NMA 1) and a published NMA (NMA 2) were included. NMA1 is a systematic review with 11 studies, that evaluated the efficacy and safety of LEM in patients with insomnia by comparing it to relevant drugs in Canadian public formularies (e.g., zopiclone, temazepam, triazolam, flurazepam, nitrazepam) with respect to clinical end points evaluated objectively (by PSG) or subjectively (patient-reported). The clinical end points included LPS, WASO, sleep efficiency, and total sleep time (TST) with subjective assessment in these same end points. Harms related to the use of LEM were also evaluated in an ITC analysis, including treatment discontinuations, somnolence, dizziness, headache, and in a posthoc analysis, the risk of falls. NMA 2 is a published comparative efficacy analysis by McElroy et al. (2021) aimed at evaluating the efficacy and safety of LEM against other insomnia treatments through a systematic literature review and NMA.²⁰ The ITC search strategy included RCTs of drugs used in adults with primary insomnia (not all reimbursed in Canada). The drugs of interest in NMA 2 were LEM, suvorexant, benzodiazepines, Z-drugs (zolpidem, eszopiclone, zaleplon, zopiclone), trazodone, and ramelteon. Of these, only zolpidem, zopiclone, eszopiclone, trazodone, triazolam, and temazepam were available in the body of evidence from NMA 2 and of interest to this CADTH reimbursement review.

Efficacy Results

Indirect evidence from the NMAs suggest that for the end point of LPS, LEM5 is superior to triazolam and PBO but no evidence of a difference between LEM5 and LEM10 was observed. With LEM5 as a reference, LPS was longer in the PBO group (mean difference = –19.1 minutes [95% CI, –3.20 minutes to –35.0 minutes], in which a negative value in the mean difference implies improvement in favour of LEM5). Patients treated with LEM5 had a reduction in LPS when compared to triazolam 0.5 mg (mean difference = –34.1 minutes; 95% CI, –5.47 minutes to –62.8 minutes). The results for LEM10 were consistent with LEM5. In NMA 2, LEM showed a reduction in LPS when compared to PBO (–18.6 minutes [95% credible interval or CrI, –29.0 minutes to –10.9 minutes]), ZOL-ER (–13.4 minutes [95% CrI, –24.4 minutes to –4.9 minutes)], and triazolam (–23.2 minutes [95% CrI, –38.8 minutes to –9.6 minutes]).

For the end point of WASO, the evidence suggests LEM5 and LEM10 were superior to PBO, but the evidence against triazolam was very imprecise in detecting a difference in PSG-assessed WASO between LEM5, LEM10, and triazolam 0.5 mg. In NMA 2, for WASO, LEM was superior to PBO and zolpidem tartrate immediate-release (ZOL-IR), with an average reduction of 21.3 minutes (95% CrI, –29.6 minutes to –10.1 minutes) and 19.6 minutes (–31.9 minutes to –0.3 minutes), respectively.

For the evaluation of objectively measured sleep efficiency, LEM5, LEM10, flurazepam 30 mg, and triazolam 0.5 mg could be compared in the NMA. The mean differences in sleep efficiency for LEM5 and LEM10 compared with PBO were 7.62% (95% CI, 5.93% to 9.31%) and 8.80% (95% CI, 7.06% to 10.5%), respectively, where higher values mean improvement in favour of LEM. The mean differences in sleep efficiency for LEM5 and LEM10 compared with triazolam were 9.62% (95% CI, 4.92% to 14.3%) and 10.8% (95% CI, 6.09% to 15.5%), respectively. Effect estimates for the comparisons between LEM5 and LEM10 and flurazepam were too imprecise to draw a conclusion.

TST data were only available for LEM5 and LEM10 against PBO. The NMA showed that effect estimates were imprecise in detecting a difference between LEM5 and LEM10 (mean difference = –4.65 minutes; 95% CI, –2.45 minutes to 11.8 minutes). Although relative to PBO, LEM5 and LEM10 were associated with an improvement of 34.8 minutes (95% CI, 27.4 minutes to 42.4 minutes) and 39.5 minutes (95% CI, 32.1 minutes to 46.9 minutes), respectively.

In the second NMA, for the subjective outcomes, LEM was superior to PBO in all end points evaluated but not against eszopiclone for the subjective quality of sleep (sQUAL) end point (mean difference = –0.6 [95% CrI, –0.9 to –0.2]). For the rest of the comparisons, no evidence of effect was observed.

Harms Results

The evidence from the ITC suggested that LEM10 had an increased risk of drug discontinuation when compared to LEM5 (odds ratio [OR] = 1.99; 95% CI, 1.06 to 3.74) and PBO, but not against the other comparators. The odds of discontinuations were less for PBO compared with LEM10 (OR = 0.48; 95% CI, 0.25 to 0.91). The effect estimates for all other comparisons with LEM10 were too imprecise to draw a conclusion.

For the end point of somnolence, when compared with LEM5, PBO had reduced odds of somnolence (OR = 0.25; 95% CI, 0.12 to 0.52); similarly, the odds were reduced in the triazolam 0.25 mg group (OR = 0.31; 95% CI, 0.14 to 0.69) and zopiclone group. Increased odds of somnolence for LEM10 were observed against PBO, triazolam 0.25 mg, and zopiclone 7.5 mg. For other comparisons (i.e., LEM5 and LEM10 versus flurazepam, temazepam, and triazolam 0.125 mg), the effect estimates were too imprecise to detect a difference.

For the end point of dizziness, the effect estimates were too imprecise to draw conclusions about the effect of LEM5 or LEM10 compared with PBO, flurazepam, triazolam, or zopiclone.

For the end point of headache, the results were also too imprecise to detect a difference in the odds of headache between LEM5 and each of the alternatives. This situation also occurred in the comparison of LEM10 to the other comparators.

In an additional posthoc analysis (Bucher ITCs) of studies reporting falls, using LEM5 as the reference, no evidence of any difference could be detected when comparing to patients receiving triazolam, flurazepam, lorazepam, trazodone, benzodiazepines, and Z-drugs due to the high imprecision of the results. When LEM10 was used as the reference, the odds of falls were higher with triazolam, flurazepam, and lorazepam when compared to LEM10. When the entire class of benzodiazepines and trazodone were compared to LEM10, the odds of falls were also higher with the former drugs. Even though these findings suggested that LEM10 is associated with a reduction in the odds of falls when compared to other drugs, the results were very imprecise with wide CIs, which made it difficult to draw conclusions.

In the second NMA evaluated in this CADTH report, there was no evidence of increased or decreased odds for presenting SAEs for patients receiving LEM when compared to relevant comparators, mainly due to wide CrIs. Similarly, no difference was observed in the odds of withdrawals due to AEs or in the odds of falls. LEM was associated with lower odds of dizziness compared with ZOL-IR, ZOL-ER, and eszopiclone. LEM, however, increased the odds of somnolence when compared to PBO, ZOL, and eszopiclone.

Critical Appraisal

The results from both NMAs have uncertainty due to the risk of bias in the individual studies (i.e., unclear randomization, allocation concealment, and baseline imbalances), heterogeneity in the pairwise comparisons, and suspected publication bias. Furthermore, the evidence is imprecise in most of the effect estimates from both NMAs, with wide CIs that could include an appreciable threshold of benefit or harm. In the NMAs, there were concerns of incomplete information and discrepancies in the included studies that could affect the plausibility of the transitivity assumption. Overall, the populations included in the individual studies of the network are generalizable to the Canadian population without comorbid conditions.

Other Relevant Evidence

Description of Studies

Study 312 was an open-label, phase IIIb, pilot, multicentre trial (N = 53) investigating next-dose transition from ZOL to LEM for the treatment of insomnia. The trial included adult patients with insomnia who were receiving ZOL as monotherapy, and who agreed to substitute it with LEM.

Efficacy Results

Overall, 81.1% of patients successfully transitioned to LEM at the end of the 2-week titration period and entered study 312’s extension phase. A total of 15 (48.4%) patients had an LEM dose increase from 5 mg to 10 mg while 5 (22.7%) patients had an LEM dose decrease from 10 mg to 5 mg at the end of the 2-week titration period.

At the end of the titration period, patients in both the LEM5 and LEM10 groups indicated that LEM had a positive effect on sleep, time to fall asleep, and TST. A large proportion of patients felt that the medication was “too weak.” In the overall trial population, the mean change in ISI total score was –4.6 (SD = 6.26) and the mean change in the quality of sleep score was –0.19 (SD = 0.92) at the end of the titration period.

In the overall trial population, at baseline and at the end of the titration period, the mean sleep efficiency scores were 79.03% (SD = 85.4%) and 80.17% (SD = 8.49%), respectively. The mean WASO scores were 80.90 (SD = 33.23) minutes and 83.92 (SD = 35.44) minutes, respectively, and the mean TST scores were 403.44 (SD = 62.07) minutes and 412.11 (SD = 60.17) minutes, respectively.

Harms Results

Of the 53 patients enrolled in the core study, 20 (37.7%) patients experienced at least 1 TEAE (5 patients from the LEM5 group and 15 patients from the LEM10 group). The most common TEAEs were abnormal dreams (7.5%) and somnolence (7.5%). In the overall trial population, 13% of patients withdrew from treatment due to an AE. No deaths or SAEs leading to study drug discontinuation were reported in the study.

Critical Appraisal

Study 312 had an open-label design, small sample size, short study duration, no formal statistical testing, subjective secondary end points, and 50% screening failures, and chose ZOL as the previous treatment to transition from (rather than other treatments that may have been more relevant to Canadian clinical practice); this prevented conclusions from being drawn and limited generalizability to the Canadian population with insomnia.

Conclusions

Both LEM10 and LEM5 demonstrated statistically significant improvements in sleep onset, sleep maintenance, and sleep efficiency among adults with insomnia relative to PBO in the SUNRISE 1 and SUNRISE 2 trials. In SUNRISE 1, all primary and key secondary outcomes were for objective PSG measures and were clinically meaningful based on suggested thresholds for a clinically important effect identified from the literature — except for WASO2H, which did not have a minimal important difference (MID) or threshold identified. In SUNRISE 2, all primary and key secondary outcomes were for subjective sleep diary measures, and none reached the suggested thresholds for a clinically important effect identified from the literature. Numerical differences in ISI item 4 to item 7 and FSS indicated a benefit from treatment with LEM over PBO, but due to the lack of adjustment for multiple comparisons and no established MID for ISI item 4 to item 7 and FSS, conclusions could not be drawn from these outcomes. Other patient-reported outcomes such as the EQ-5D-3L, PGI-Insomnia, and quality of sleep scale were exploratory, were not controlled for multiplicity, and did not have MIDs for populations with insomnia; therefore, no conclusions could be drawn regarding whether treatment with LEM conferred a benefit for these outcomes. The clinical expert consulted for this review emphasized the importance of subjective outcomes and how patients perceive changes in their sleep, and sponsor-submitted posthoc analyses of SUNRISE 2 data lent support to the meaningfulness of the PGI-Insomnia and ISI subjective outcomes. However, the sleep improvements based on PSG and sleep diary outcomes observed in the SUNRISE 1 and SUNRISE 2 studies did not appear to be reflected in the results from the patient-reported questionnaires (i.e., ISI, FSS, and perceived sleep quality).

It is uncertain if the differences in treatment effect observed in SUNRISE 1 and SUNRISE 2 would be experienced by and be meaningful to all patients who could be treated with LEM. There did not appear to be a clear difference in efficacy between LEM10 and LEM5, although differences in efficacy between the 2 doses were not tested statistically in the trials. The product monograph’s recommendation to start with LEM5 and the higher rates of somnolence and treatment discontinuations for LEM10 would be considerations when prescribing or increasing a patient’s dose. Overall, the clinical expert indicated that LEM demonstrated a minor harms profile with few SAEs, low rates of withdrawals from treatment due to AEs, and no deaths. Patients with specific comorbid conditions such as sleep apnea, anxiety, and depression were excluded from the studies based on the exclusion criteria for AHI, BAI, and BDI-2 scores. Further research showing adequate efficacy and safety is needed to inform broader treatment with LEM. Indirect evidence suggested LEM is superior to triazolam 0.5 mg for LPS and sleep efficiency, is superior to ZOL for WASO and LPS, and is inferior to eszopiclone in sleep quality; however, there is uncertainty in this conclusion due to limitations of the NMAs and differences in conclusions for some outcome comparisons across NMAs. Evidence was very uncertain or absent for the other comparisons and end points due to imprecision, risk of bias, and heterogeneity in the pairwise comparisons. Additionally, evidence for transitioning from ZOL to LEM from the pilot study 312 was limited by the study design, small sample size, and large proportion of screening failures, preventing conclusions from being drawn. Overall, LEM appears to be effective relative to PBO for important sleep-related outcomes, and long-term results from SUNRISE 2 supported those observed during the first 6 months of treatment. There were no serious safety concerns, but efficacy for health-related quality of life (HRQoL) and subjective appraisals of symptoms and sleep quality were less certain due to limitations of the trials. ZOL may not have been the most relevant comparator in Canadian practice, and comparisons with ZOL in the trial (except for WASO2H) were not controlled for multiplicity; therefore, there is uncertainty in the conclusions about the comparative efficacy. LEM appeared to be more effective than ZOL for the outcome of WASO2H, which was noted as being important for sleep maintenance, though there was no MID identified and there was uncertainty if the change observed in SUNRISE 1 was clinically meaningful. Moreover, there was no direct evidence for efficacy or harms relative to other relevant comparators that may be used in Canadian clinical practice, and the indirect evidence was uncertain. Owing to its superiority over PBO, LEM may be another treatment option for patients with insomnia.

Introduction

Disease Background

Insomnia disorder is the most common sleep disorder and, according to the DSM-5, is described as being dissatisfied with the quality or quantity of sleep as well as having difficulty initiating and/or maintaining sleep, and is associated with daytime impairment.^2,3 The DSM-5 criteria include sleep disturbances occurring at least 3 nights per week for at least 3 months causing significant distress or impairment in daily functioning, having adequate opportunity for sleep, ruling out other sleep disorders, and excluding the possibility of being caused by coexisting mental disorders or medical conditions, or by a substance.⁴

According to the Alberta Medical Association’s 2015 Assessment to Management of Adult Insomnia clinical practice guideline, there are no validated questionnaires for the diagnosis of sleep disorders and a diagnosis should be made based on clinical judgment.³ There are instruments for assessing insomnia and related conditions that can support a diagnosis such as the Sleep Disorders Assessment Questionnaire, ISI, and screening tools for psychiatric and medical disorders that affect sleep. A sleep diary may be used in clinical practice to track a patient’s perception of their sleep outcomes, particularly when CBT-I is used to treat insomnia.

Estimates for the prevalence and incidence of insomnia vary based on the population studied and how insomnia has been defined.⁸ The prevalence of insomnia is estimated to be between 12% and 24% based on surveys of adults in Canada.^5,6 There is additional uncertainty in the numbers, given that individuals with insomnia may not immediately seek advice from their primary care physicians until symptoms are bothersome; thus, many individuals are not diagnosed.^2,21 A Canadian study (N = 3,073) investigating the incidence of insomnia found that 3.8%, 9.3%, and 13.9% of adults without insomnia at baseline developed an insomnia disorder at the 1-year, 3-year, and 5-year follow-up time points of the study, respectively.⁷ Another study of adults in Quebec found that 30.7% of survey respondents who reported no insomnia symptoms at baseline had developed symptoms within 1 year while 7.3% had developed an insomnia disorder in that time.⁸

Older age, female sex, comorbid medical or psychiatric conditions, and social and environmental factors have been identified as common risk factors for insomnia.^2,3 It has been reported that adults aged between 30 years and 59 years are almost twice as likely to have insomnia compared to adults aged 18 years to 29 years.^2,3 Furthermore, natural age-related changes and medical issues increase the risk of insomnia among older adults. It has been estimated that females are 1.2 to 1.5 times more likely to have insomnia than males.^2,3 When compared to people without insomnia, patients with insomnia have consistently demonstrated impaired HRQoL on different generic health questionnaires — typically on multiple domains assessed by the instrument with some variability by age.²² This trend was also observed when comparing patients with different severity of symptoms where those with insomnia disorder scored lower than those with mild or occasional insomnia symptoms.²² Insomnia can be diagnosed as a disorder on its own or as a symptom associated with other conditions such as chronic pain, heart disease, respiratory disease, gastrointestinal problems, cancer, and diabetes.^3,8 The negative impact insomnia has on HRQoL was found to have a compounding effect on poor sleep for patients with existing conditions such as multiple sclerosis, Parkinson disease, and cancer.²² Additionally, insomnia has been closely linked to increased risks of mental health, occupational, and medical issues; reduced life expectancy; and increased economic costs through lost productivity, workplace accidents and MVAs, and greater health care usage.^2,3,7,9,10

Standards of Therapy

Per the Alberta Medical Association’s clinical practice guideline, insomnia can be treated and managed in a primary care setting.³ A patient may be referred to a physician or medical centre that specializes in treating sleep disorders when the problem cannot be diagnosed or treated in the primary setting or if the insomnia is resistant to treatment or worsens. Referrals may also depend on the availability of specialists, with the possibility of long wait times.

According to the clinical expert CADTH consulted for this review, the primary goals of treatment are to improve sleep initiation and maintenance and reduce terminal insomnia, and for sleep to be more restorative. These goals were consistent with those noted by the clinician groups that provided input for the review, with the addition of improved related daytime function. The group emphasized that simply treating nocturnal symptoms is inadequate if overall daytime function is not also improved.

Nonpharmacologic treatment options such as sleep hygiene education (e.g., changing sleep habits, relaxation techniques) and CBT-I are recommended as an initial treatment of insomnia.^3,10,11,23 It is recommended that behavioural therapy be introduced before or concurrently with pharmacotherapy, and that the former be continued even if medication is stopped. The clinical expert consulted for this review described how CBT-I works to target the underlying mechanisms of insomnia while pharmacotherapy treats the symptoms. It has been noted that CBT-I may not be accessible to all patients due to cost or lack of availability in the patient’s region,^10,23 although the clinical expert suggested that access to CBT-I is less of an issue now with the availability of CBT-I courses and treatment through online and book formats. Other issues with CBT-I are that patients may be unwilling or unable to participate, or patients may be unresponsive to treatment.¹⁰ When these barriers have been identified, pharmacologic treatment can be considered.

Pharmacologic treatments for insomnia can be categorized based on 4 different mechanisms of action: BZRAs, DORAs, histamine receptor antagonists, and melatonin receptor agonists.¹² The BZRA group includes benzodiazepine hypnotics (e.g., flurazepam, temazepam, triazolam) and non-BZRAs, also known as Z-drugs (e.g., zopiclone, zolpidem); DORAs include suvorexant and LEM; histamine receptor antagonists include low-dose doxepin; and melatonin receptor agonists include ramelteon.^12,23 Of the medications listed, suvorexant and ramelteon are not available in Canada. Off-label medications that have sedating effects (e.g., antidepressants, antihistamines, anticonvulsants, antipsychotic drugs) may also be used in specific situations but are generally not recommended as a first-line treatment due to the lack of efficacy and safety evidence in this population.^3,12 This was also noted in the clinician group input submitted to CADTH. The Alberta Medical Association’s clinical practice guideline emphasizes a short initial treatment duration (i.e., limited prescription without refills) of the lowest possible dose with follow-up within 2 weeks to 4 weeks to assess adverse effects and whether there is a need for continued treatment. For sleep onset insomnia (trouble falling asleep, but not maintaining sleep), non-BZRAs or melatonin receptor agonists are recommended.¹² For sleep maintenance or mixed insomnia (sleep onset and sleep maintenance), DORAs, non-BZRAs, and low-dose doxepin would be considered acceptable.¹² Tapering medication may take weeks to months and is recommended to improve the transition to on-treatment or off-treatment, avoid rebound insomnia, and promote long-term success.³ Despite the fact that insomnia can be chronic and unremitting, long-term use of these medications, particularly hypnotic drugs, is discouraged as the studies supporting their approval were based on short-term use.² Specifically with the older benzodiazepine hypnotics, their use as a first-line treatment option is not recommended due to their long half-life, their higher risk of tolerance and dependence, and the availability of safer options.^3,12 Resistant insomnia and insomnia that occurs with other comorbidities may warrant the long-term use of medications.³ For the extended use of sedatives, it is recommended that follow-up visits occur every 3 months to 4 months until the insomnia is stable, after which reassessment can be made at 6 months. Over-the-counter medications with sedative effects such as those containing diphenhydramine or doxylamine and alcohol are not recommended for the treatment of insomnia.^3,12 Studies of alternative therapies for treating insomnia were noted to have methodological limitations that have prevented the use of their findings.² The clinical expert also noted that cannabinoids have gained widespread use — again, with little clinical trial data. The expert suggested that the widespread prevalence of over-the-counter medications, alcohol, and cannabinoids, as well as the off-label use of antidepressants and antipsychotic drugs, suggests that no single drug or mechanism of action can resolve the majority of problems with insomnia.

Drug

The key characteristics of LEM, Z-drugs, benzodiazepines, doxepin, antidepressants, and antipsychotic drugs are summarized in Table 4.

LEM (Dayvigo) is indicated for the treatment of insomnia, characterized by difficulties with sleep onset and/or sleep maintenance.¹ The drug is available in 5 mg and 10 mg tablets to be taken orally once per night within a few minutes before going to bed, with at least 7 hours before planned awakening time.¹³ The recommended dose is 5 mg once per night, although the dose may be increased to the maximum recommended dose of 10 mg based on clinical response and tolerability. The drug is a competitive antagonist of both orexin receptor 1 (OX1R) and orexin receptor 2 (OX2R), with a higher affinity for OX2R. The orexin neuropeptide signalling system is a central promoter of wakefulness and blocking the binding of wake-promoting neuropeptides orexin A and orexin B to receptors OX1R and OX2R is thought to suppress wake drive.

LEM underwent a standard review at Health Canada and was issued a Notice of Compliance on November 4, 2020. The sponsor has requested reimbursement as per the approved Health Canada indication.¹ LEM has not been previously reviewed by CADTH.

Stakeholder Perspectives

Patient Group Input

This section was prepared by CADTH staff based on the input provided by 3 patient groups. The full patient group submissions are included in the stakeholder input section at the end of this report.

The patient and caregiver input received for this review was collected by 3 groups: MDSC, Migraine Canada, and Menopause Chicks. The input provided by MDSC included 1,249 respondents and was sourced from a survey conducted from September 21 to October 7, 2021, interviews with 3 respondents with insomnia, several family members, 2 clinicians, and social media interactions. The input provided by Migraine Canada consisted of 1,385 respondents and was sourced from 2 online surveys conducted in the fall of 2021 and in March 2022. The input provided by Menopause Chicks consisted of 1,027 respondents and was sourced from an online survey conducted in the fall of 2021, and a focus group with 4 participants.

Respondents from all 3 patient groups indicated that sleep problems significantly impacted their quality of life, energy level, cognitive function, and mood the next day. More than 70% of respondents to the surveys reported sleep dissatisfaction and insufficient sleep duration, and more than 60% reported some degree of disruption of daytime activities. The daytime activities that were most commonly impacted included the ability to work, conduct household chores, exercise, fulfill family obligations, and spend time with family and friends. Moreover, due to the perceived stigma of insomnia, many reported delaying getting a diagnosis and seeking treatment for fear of what other people might think. According to the patient input received, most respondents have tried various treatments for sleep problems (e.g., benzodiazepines, Z-drugs), especially respondents with insomnia symptoms, mental illness, and pain. Many of them were dissatisfied and discontinued the treatment due to side effects, such as next-day sedation and cognitive impairment, and fear of developing a substance use disorder.

Respondents from all 3 groups reported that they would like a treatment to address the following key outcomes that are important to patients, family members, and caregivers: uninterrupted and restorative sleep, greater access to treatment, more effective treatment options, long-term treatment effectiveness, fewer side effects, less stress and anxiety, improved productivity, and improved relationships with family members and colleagues. The input received from MDSC included 3 respondents with experience with LEM for poor sleep and/or insomnia; all accessed the drug through private health insurance. They indicated that they were able to manage their sleep problems without becoming dependent on the medication or experiencing serious side effects, and without feeling lethargic and sleepy the next morning.

Clinician Input

Input From the Clinical Expert Consulted by CADTH

All CADTH review teams include at least 1 clinical specialist with expertise regarding the diagnosis and management of the condition for which the drug is indicated. Clinical experts are a critical part of the review team and are involved in all phases of the review process (e.g., providing guidance on the development of the review protocol, assisting in the critical appraisal of clinical evidence, interpreting the clinical relevance of the results, providing guidance on the potential place in therapy). The following input was provided by 1 clinical specialist with expertise in the diagnosis and management of insomnia.

Unmet Needs

The clinical expert consulted for this review stated that the current goals of treatment are to improve sleep initiation, maintenance, and terminal insomnia, leading to restorative sleep. First-line treatment for insomnia is CBT-I according to the clinical expert, yet many patients have limited access to and/or success with this option. As a result, pharmacotherapy is often used in addition to or in place of CBT-I. The clinical expert described that CBT-I targets the underlying mechanisms of insomnia, whereas medications tend to target the symptoms.

As per the clinical expert, patients may have no response or a waning response to medications when used for extended periods of time, at which point it can be difficult to withdraw the therapy if the patient has developed a tolerance to it and experiences rebound insomnia. Another concern raised by the clinical expert was the risk of cognitive and behavioural changes the next morning that can lead to falls and other dangers, particularly for older adult patients.

Place in Therapy

The clinical expert suggested that LEM may be a first-line pharmacologic treatment for insomnia and noted that it would be necessary to determine how to optimally transition from other currently available medications (e.g., Z-drugs) to LEM.

Patient Population

The expert indicated that most patients with insomnia may be candidates for treatment with LEM except for those who are pregnant, nursing, or have narcolepsy.

Assessing Response to Treatment

Per the clinical expert, patients are asked generally if they are sleeping better and longer, and feel refreshed, rather than performing tests and having measurements taken in clinics. The expert suspected that methods for the evaluation of treatment response do not vary among physicians. The clinical expert indicated that few clinics are performing sleep studies (except where another sleep disorder is suspected), particularly since access is limited in many clinics across Canada.

Discontinuing Treatment

The expert noted that chronic insomnia generally does not go away unless there is successful behavioural intervention or another major illness overtakes it. Furthermore, the clinical expert reported that insomnia tends to worsen with age, menopause, or major negative life events. It was suggested by the expert that a patient may wish to trial being off the medication and can be supported by additional sleep management tools such as CBT-I. When deprescribing a medication, the clinical expert would observe the patient for a return of symptoms or rebound insomnia.

Prescribing Conditions

Per the clinical expert, insomnia is typically treated by family physicians and few patients are referred to sleep specialists, for instance, if another comorbid sleep condition is suspected.

Clinician Group Input

This section was prepared by CADTH staff based on the input provided by 2 clinician groups. The full clinician group submissions are included in the stakeholder input section at the end of this report.

Clinician group input on the review of LEM for the treatment of insomnia was received from 2 clinician groups: the CCSSP and the National Advisory Board, comprising Canadian family physicians and psychiatrists with expertise in the management of insomnia.

The clinician groups identified that current pharmacotherapy options for treating insomnia disorder include benzodiazepines and Z-drugs, which can pose a significant risk of adverse effects such as delirium, falls, MVAs, cognitive impairment, dependance, and tolerance in some populations, including older adults and patients with comorbid conditions. Moreover, both clinician groups noted that due to a lack of safe on-label pharmacologic options and limited access to CBT-I for treating insomnia, physicians have resorted to prescribing off-label sedating drugs, including antipsychotic drugs, antidepressants, and anticonvulsants for long periods that are generally not supported by guidelines. The CCSSP noted that all current pharmacologic treatments should be considered symptomatic as they do not modify the underlying mechanisms of insomnia. Additionally, both groups noted the widespread use of over-the-counter medications by patients in combination with alcohol and cannabis to improve sleep, which can have harmful effects on their mental and physical health. Thus, new drugs with good evidence of improved efficacy, tolerability, and safety for the treatment of insomnia are critical.

According to the clinician groups, patients who are most likely to respond better to treatment with LEM are older adults with chronic insomnia as well as patients with comorbidities of major depressive disorder, anxiety, attention-deficit/hyperactivity disorder, post-traumatic stress disorder, obstructive sleep apnea, and fibromyalgia. Both groups agreed that responses to treatment are typically assessed by improvement in sleep onset, LPS, sleep efficiency, and TST, as well as daytime alertness, and functioning. Both groups noted that the ISI and Epworth Sleepiness Scale are the most frequently used patient-reported outcomes to document treatment response and remission from insomnia in clinical trials. The CCSSP also indicated that response should be assessed every 2 weeks to 4 weeks when treatment is initiated, and then every 3 months to 6 months if a patient remains stable.

Both groups agreed that the novel mechanisms of LEM would improve the evolving paradigm of insomnia medications with its high pharmacologic specificity, improved safety, tolerability, and efficacy profiles. The clinician groups indicated that LEM can be considered as first-line treatment for all patients with insomnia disorder who are offered pharmacotherapy and clinicians have felt that the clinical trial data generally aligned with their experiences with LEM in practice. The clinician groups stated that the prescription of LEM should not be limited to sleep specialists and that discontinuing treatment may be considered if there are concerning AEs, the risk of drug-drug interactions, or a lack of treatment response.

Drug Program Input

The drug programs provide input on each drug being reviewed through CADTH’s reimbursement review processes by identifying issues that may impact their ability to implement a recommendation. The implementation questions and corresponding responses from the clinical expert consulted by CADTH are summarized in Table 5.

Table 5: Summary of Drug Plan Input and Clinical Expert Response

BIA = budget impact analysis; CDEC = CADTH Canadian Drug Expert Committee; LEM = lemborexant; PSG = polysomnography; vs. = versus; Z-drugs = zolpidem, eszopiclone, zaleplon, and zopiclone; ZOL = zolpidem tartrate.

Clinical Evidence

The clinical evidence included in the review of LEM (Dayvigo) is presented in 3 sections. The first section, the systematic review, includes pivotal studies provided in the sponsor’s submission to CADTH and Health Canada, as well as those studies that were selected according to an a priori protocol. The second section includes indirect evidence from the sponsor and indirect evidence selected from the literature that met the selection criteria specified in the review. The third section includes sponsor-submitted long-term extension studies and additional relevant studies that were considered to address important gaps in the evidence included in the systematic review.

Following the issuance of the draft CDEC recommendation for LEM in August 2022, the following additional information was provided to CADTH: 1 unpublished manuscript by Drake et al.¹⁴ that provided evidence for LEM’s efficacy by the PGI-Insomnia in the SUNRISE 2 study; a PSUR¹⁵ that included aggregate data from approximately 4,024 adult patients with insomnia disorder, ISWRD, or healthy volunteers who were enrolled in the LEM clinical development program; and a published NMA by De Crescenzo et al. (2022)¹⁶ comparing LEM to other pharmacologic treatments for the acute and long-term treatment of adults with insomnia. These data were not included in the submission to CADTH and provided more information on the clinical meaningfulness of the subjective PGI-Insomnia instrument, and on LEM’s safety and comparative evidence versus other drugs for the treatment of insomnia. The information has been summarized and critically appraised as an addendum to the CADTH report in Appendix 6.

Systematic Review (Pivotal and Protocol Selected Studies)

Objectives

To perform a systematic review of the beneficial and harmful effects of LEM5 or LEM10 tablets taken orally once daily at bedtime for the treatment of patients with insomnia, characterized by difficulties with sleep onset and/or sleep maintenance.

Methods

Studies selected for inclusion in the systematic review included pivotal studies provided in the sponsor’s submission to CADTH and Health Canada, as well as those meeting the selection criteria presented in Table 6. Outcomes included in the CADTH review protocol reflect outcomes considered to be important to patients, clinicians, and drug plans.

Table 6: Inclusion Criteria for the Systematic Review

AE = adverse event; ESS = Epworth Sleepiness Scale; FSS = Fatigue Severity Scale; HRQoL = health-related quality of life; ISI = Insomnia Severity Index; MVA = motor vehicle accident; RCT = randomized controlled trial; SAE = serious adverse event; SDLP = standard deviation of lateral position; vs. = versus; WDAE = withdrawal due to adverse event.

^aOff-label use.

The literature search for clinical studies was performed by an information specialist using a peer-reviewed search strategy according to the PRESS Peer Review of Electronic Search Strategies checklist.³²

Published literature was identified by searching the following bibliographic databases: MEDLINE All (1946–) via Ovid and Embase (1974–) via Ovid. All Ovid searches were run simultaneously as a multifile search. Duplicates were removed using Ovid deduplication for multifile searches, followed by manual deduplication in Endnote. The search strategy consisted of both controlled vocabulary, such as the National Library of Medicine’s MeSH (Medical Subject Headings), and keywords. The main search concept was lemborexant. Clinical trials registries were searched: the US National Institutes of Health’s ClinicalTrials.gov, the WHO’s International Clinical Trials Registry Platform search portal, Health Canada’s Clinical Trials Database, and the European Union Clinical Trials Register.

No filters were applied to limit the retrieval by study type. Retrieval was not limited by publication date or by language. Conference abstracts were excluded from the search results. Refer to Appendix 1 for the detailed search strategies.

The initial search was completed on March 24, 2022. Regular alerts updated the search until the meeting of CDEC on July 28, 2022.

Grey literature (literature that is not commercially published) was identified by searching relevant websites from the Grey Matters: A Practical Tool For Searching Health-Related Grey Literature checklist.³³ Included in this search were the websites of regulatory agencies (US FDA and European Medicines Agency). Google was used to search for additional internet-based materials. Refer to Appendix 1 for more information on the grey literature search strategy.

These searches were supplemented by contacting the manufacturer of the drug for information regarding unpublished studies.

Two CADTH clinical reviewers independently selected studies for inclusion in the review based on titles and abstracts, according to the predetermined protocol. Full-text articles of all citations considered potentially relevant by at least 1 reviewer were acquired. Reviewers independently made the final selection of studies to be included in the review, and differences were resolved through discussion.

Findings From the Literature

A total of 6 reports of 2 studies^34-39 were identified from the literature for inclusion in the systematic review (Figure 1). The included studies are summarized in Table 7. A list of excluded studies is presented in Appendix 2.

Figure 1: Flow Diagram for Inclusion and Exclusion of Studies

Table 7: Details of Included Studies — SUNRISE 1 AND SUNRISE 2

AHI = Apnea-Hypopnea Index; BAI = Beck Anxiety Inventory; BDI-2 = Beck Depression Inventory; CBT = cognitive behavioural therapy; DB = double-blind; DSM-5 = Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition; ECG = electrocardiogram; EQ-5D-3L = 3-level EQ-5D; ESS = Epworth Sleepiness Scale; FSS = Fatigue Severity Scale; IRLS = International Restless Legs Scale; ISI = Insomnia Severity Index; LEM = lemborexant; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; LPS = latency to persistent sleep; NREM = nonrapid eye movement; PBO = placebo; PD = pharmacodynamic; PGI-Insomnia = Patient Global Impression–Insomnia; PK = pharmacokinetic; PSG = polysomnography; QTcF = a prolonged corrected QT interval by Fridericia’s formula interval; RCT = randomized controlled trial; REM = rapid eye movement; SDSB = Sleep Disorders Screening Battery; sSOL = subjective sleep onset latency; STOP-Bang = Snoring, Tiredness, Observed apnea, blood Pressure, Body mass index, Age, Neck circumference, and Gender; sTST = subjective total sleep time; sWASO = subjective wake after sleep onset; T-BWSQ = Tyrer Benzodiazepine Withdrawal Symptom Questionnaire; TST = total sleep time; vs. = versus; WASO = wake after sleep onset; WASO1H = wake after sleep onset in the first half of the night; WASO2H = wake after sleep onset in the second half of the night; WPAI: GH = Work Productivity and Activity Impairment Questionnaire: General Health; ZOL = zolpidem tartrate.

^aInsomnia disorder according to the DSM-5:⁴ “(1) Complained of dissatisfaction with night time sleep, in the form of difficulty staying asleep and/or awakening earlier in the morning than desired despite adequate opportunity for sleep (note that if the complaint was limited to difficulty initiating sleep, the patient was not eligible); (2) Frequency of complaint 3 or more times per week; (3) Duration of complaint greater than or equal to 3 months; (4) Associated with complaint of daytime impairment.”

^bExclusionary score on SDSB as follows: (1) STOP-Bang Sleep Apnea Questionnaire score greater than or equal to 5; (2) IRLS score greater than or equal to 16; (3) ESS score greater than 15 (scores of 11 to 15 required excessive daytime sleepiness to be recorded in the patient's medical history).

^cAll medications were taken orally, once daily at bedtime.

^dDuring the placebo run-in phase of the SUNRISE 1 study, all patients received 1 LEM-matched PBO tablet and 1 ZOL-matched PBO tablet in a single blind manner. During the placebo run-in phase of the SUNRISE 2 study, all patients received 1 LEM-matched PBO tablet in a single blind manner.

^eOutcome comparing LEM10 and LEM5 to ZOL. US-only outcome.

^fResponders analysis in the SUNRISE 1 study: (1) objective sleep onset response was defined as LPS less than or equal to 20 minutes (provided mean baseline LPS was greater than 30 minutes); (2) subjective sleep onset response was defined as sSOL less than or equal to 20 minutes (provided mean baseline sSOL was greater than 30 minutes); (3) objective sleep maintenance response was defined as WASO less than or equal to 60 minutes (provided mean baseline WASO was greater than 60 minutes and was reduced by greater than 10 minutes compared to baseline); and (4) subjective sleep maintenance response was defined as sWASO less than or equal to 60 minutes (provided mean WASO was greater than 60 minutes and was reduced by greater than 10 minutes compared to baseline).

^gRebound insomnia assessed using the sleep diary in the SUNRISE 1 study: (1) the change from baseline for sSOL on each of the first 3 nights, the mean sSOL of the first 3 nights, the mean sSOL of the first 7 nights, and the mean sSOL of the second 7 nights of the follow-up period; (2) the change from baseline for sWASO on each of the first 3 nights, the mean sWASO of the first 3 nights, the mean sWASO of the first 7 nights, and the mean sWASO of the second 7 nights of the follow-up period; (3) the proportion of patients whose sSOL was longer than at screening at the following time points during the follow-up period by at least 5 minutes: each of the first 3 nights, the mean of the first 3 nights, the mean of the first 7 nights, and the mean of the second 7 nights; and (4) the proportion of patients whose sWASO was higher than at screening at the following time points during the follow-up period by at least 5 minutes: each of the first 3 nights, the mean of the first 3 nights, the mean of the first 7 nights, and the mean of the second 7 nights.

^hResponders analysis in the SUNRISE 2 study: (1) sleep onset responder defined as sSOL at baseline was greater than or equal to 30 minutes and mean sSOL at 6 months was less than or equal to 20 minutes; (2) sleep maintenance responder defined as sWASO at baseline was greater than or equal to 60 minutes and mean sWASO at 6 months was less than or equal to 60 minutes and showed a reduction of greater than 10 minutes compared to baseline.

ⁱRebound insomnia assessed using the sleep diary in the SUNRISE 2 study: (1) the change from screening of sSOL on each of the first 3 nights, the mean sSOL of the first 7 nights, and the mean sSOL of the second 7 nights of the follow-up period; (2) the change from screening of sWASO on each of the first 3 nights, the mean sWASO of the first 7 nights, and the mean sWASO of the second 7 nights of the follow-up period; (3) the proportion of patients whose sSOL was longer than at screening for each of the first 3 nights, or whose mean sSOL was longer than at screening for the first 7 nights or the second 7 nights of the follow-up period; (4) the proportion of patients whose sWASO was higher than at screening for each of the first 3 nights, or whose mean sWASO was higher than at study baseline for the first 7 nights or the second 7 nights of the follow-up period.

^jPersistence of effect assessment: (1) the mean change from study baseline of sSOL, of subjective sleep efficiency, of sWASO, and of sTST at month 3, month 6, month 9, and month 12 compared to month 1; (2) the mean change from period 2 baseline (month 6) of sSOL, of subjective sleep efficiency, of sWASO, and of sTST at month 9 and month 12 compared to month 7 (the first month of treatment in period 2); (3) the mean change from study baseline and period 2 baseline (as appropriate) of sSOL, subjective sleep efficiency, sWASO, and sTST at 3 months of exposure and 6 months of exposure compared to 1 month of exposure.

Sources: SUNRISE 1 Clinical Study Report¹⁷ and SUNRISE 2 Clinical Study Report.¹⁸

Description of Studies

Details of the 2 pivotal studies included are summarized in Table 7 and diagrams of the trial designs for SUNRISE 1 and SUNRISE 2 are depicted in Figure 2 and Figure 3, respectively.

Figure 2: SUNRISE 1 Study Design

# = Cognitive Drug Research posture and Cognitive Performance Assessment Battery assessments in the morning following the polysomnography assessment; BL = baseline; D = day; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; PBO = placebo; PSG = polysomnography; ZOL = zolpidem tartrate.

^a All PSG visits required an overnight stay in the clinic. At least 2 nights must have intervened between the second BL PSG and BL (D1).

Source: SUNRISE 1 Clinical Study Report.¹⁷

SUNRISE 1 Study

SUNRISE 1 (N = 1,006) was a DB, phase III, multicentre (including 5 sites in Canada) RCT designed to assess the safety and efficacy of LEM5 and LEM10 compared with PBO and ZOL for 30 days in females aged 55 years or older and males aged 65 years or older, all of whom had insomnia disorder.^17,37 Patients were randomized in a 5:5:5:4 ratio to LEM5, LEM10, ZOL, or PBO. A computer-generated randomization scheme was used and randomization was performed centrally with an interactive voice and web response system. Randomization was stratified by country and age group (55 years to younger than 65 years versus 65 years and older). The trial end date was January 30, 2018.

A prespecified interim analysis was conducted by an independent statistician after 50% of patients had been randomized and either completed the day 31 evaluations or discontinued from the trial to assess if there was a statistically significant difference between LEM10 and ZOL on WASO2H at days 29 to 30. No other end points, treatment groups, or time points were analyzed, and it was expected that there was no impact on the type I error rate. The final Clinical Study Report was dated November 29, 2018.

Figure 3: SUNRISE 2 Study Design

R = randomization.

Source: SUNRISE 2 Clinical Study Report.¹⁸

SUNRISE 2 Study

SUNRISE 2 (N = 971) was a DB, phase III, multicentre (including 4 sites in Canada) RCT designed to assess the long-term safety and efficacy of LEM5 and LEM10 compared with PBO for up to 12 months in adults 18 years or older who had insomnia disorder.^18,35,39 Patients were randomized in an approximately 1:1:1 ratio to LEM5, LEM10, or PBO in period 1 (first 6 months). Patients randomized to PBO were rerandomized approximately 1:1 at the end of 6 months to either LEM5 or LEM10 in period 2 (second 6 months). A computer-generated randomization scheme was used and randomization was performed centrally with an interactive voice and web response system. Both randomization and rerandomization (for period 2) were stratified by country and age group (18 years to younger than 65 years versus 65 years and older). The PBO-controlled period 1 portion of the trial was completed on May 31, 2018, while the trial end date was January 8, 2019. The report containing the first 6 months of trial data was dated December 4, 2018, while the report containing data for 12 months was dated March 25, 2019.

Patients were made aware that they all would receive PBO at some point during the trial and active treatment for at least 6 months but were not told when either would take place.

SUNRISE 1 and SUNRISE 2 Studies

Both studies consisted of a prerandomization phase and randomization phase.

The prerandomization phase was a maximum of 35 days and included the screening, run-in, and baseline periods. During the screening period, participants had to have completed the electronic sleep diary within 1 hour of waking for at least 7 consecutive mornings as well as performed additional screening evaluations (e.g., postural stability, CPAB, PSG). During the run-in period, participants received PBO tablets that were taken once per night until baseline (day 1), completed the electronic sleep diary for at least 7 consecutive mornings, and did additional baseline evaluations. During the baseline period, participants completed patient-reported questionnaires (e.g., ISI, FSS, EQ-5D-3L, Columbia-Suicide Severity Rating Scale [C-SSRS]) before being randomized. Assessments and the administration of PBO during the prerandomization phase were to exclude patients who could not comply with study procedures (e.g., completing the sleep diary and other study restrictions) as well as exclude patients who demonstrated a PBO response.

The randomization phase included the treatment period (30 days in the SUNRISE 1 study and at least 6 months and up to 12 months in the SUNRISE 2 study) and follow-up period (a 14-day treatment-free interval before the end-of-study visit). In SUNRISE 1, patients completed the in-clinic assessments (i.e., postural stability, CPAB, and PSG) on day 2, day 3, day 29, and day 30. In SUNRISE 2, patients completed the ISI, FSS, C-SSRS, EQ-5D-3L, and PGI-Insomnia at month 1, month 3, month 6, month 9, and month 12. The sleep diary was completed each day during the treatment and follow-up periods in both studies. The Tyrer Benzodiazepine Withdrawal Symptom Questionnaire (T-BWSQ) and C-SSRS were completed at the end-of-study visit.

Populations

Inclusion and Exclusion Criteria

SUNRISE 1 and SUNRISE 2 Studies

To participate, patients in either trial must have been adults (aged 55 years or older for females and 65 years or older for males in the SUNRISE 1 study only) with a diagnosis of insomnia disorder according to the DSM-5. Patients in SUNRISE 1 must have had an ISI score of at least 13 while those in the SUNRISE 2 study must have had a score of at least 15 to participate. Key exclusion criteria included diagnosis of a sleep-related breathing disorder, periodic limb movement disorder, restless legs syndrome, circadian rhythm sleep disorder, narcolepsy, depression, anxiety, or the use of any prohibited over-the-counter concomitant medications or insomnia treatments within 1 week or 5 half-lives of beginning the trial. Additionally, patients in SUNRISE 1 must have had an AHI score greater than 15 whereas in SUNRISE 2, patients aged 18 years to 64 years must have had an AHI score of at least 10 and patients 65 years and older must have had an AHI score of at least 15.

Baseline Characteristics

Baseline characteristics are summarized in Table 8.

Table 8: Summary of Baseline Characteristics — SUNRISE 1 AND SUNRISE 2 Studies, FAS

BAI = Beck Anxiety Inventory; BDI-2 = Beck Depression Inventory; FAS = full analysis set; FSS = Fatigue Severity Scale; ISI = Insomnia Severity Index; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; LPS = latency to persistent sleep; NA = not applicable; PBO = placebo; SD = standard deviation; sSOL = subjective sleep onset latency; sTST = subjective total sleep time; sWASO = subjective wake after sleep onset; TST = total sleep time; WASO = wake after sleep onset; WASO2H = wake after sleep onset in the second half of the night; ZOL = zolpidem tartrate.

^aPatients in the SUNRISE 1 study were aged 55 years or older.

^bBaseline sleep diary variables were analyzed with data handling rules applied.

Sources: SUNRISE 1 Clinical Study Report¹⁷ and SUNRISE 2 Clinical Study Report.¹⁸

SUNRISE 1 Study

The mean age of all randomized patients in SUNRISE 1 was 63.9 (SD = 6.81) years and most patients were female (83.6%) and white (72.3%). More patients were aged between 55 years and 64 years (55.0%) compared to 65 years and older (45.0%). Mean and median ages were similar among the treatment groups and the proportion of females was similar among the groups. There was a smaller proportion of patients who were white and a larger proportion of patients who were Black in the ZOL group compared to the other treatments. Overall, the mean and median LPS were 44.50 (SD = 35.47) minutes and 34.25 minutes, respectively, while the mean sleep efficiency was 68.27% (SD = 10.87%), and the mean WASO was 113.69 (SD = 39.09) minutes. Across treatment groups, most characteristics were balanced with a few exceptions. The baseline mean sSOL was longest in the LEM5 group at 65.79 (SD = 43.53) minutes and shortest in the PBO group at 55.90 (SD = 37.39) minutes. The baseline mean sWASO was longest in the LEM10 group at 173.35 (SD = 83.45) minutes and shortest in the LEM5 group at 166.76 (SD = 82.05) minutes. The baseline mean subjective total sleep time (sTST) was longest in the PBO group at 276.23 (SD = 87.65) minutes and shortest in the LEM10 group at 266.10 (SD = 92.16) minutes.

SUNRISE 2 Study

The mean age of all randomized patients in SUNRISE 2 was 54.5 (SD = 13.80) years and most patients were female (68.2%) and white (71.5%). More patients were younger than 65 years (72.4%) compared to those 65 years or older (27.6%). Mean and median ages were similar among the treatment groups and the proportion of females was similar among the groups. Overall, the mean and median sSOL were 63.73 (SD = 44.94) minutes and 55.21 minutes, respectively, while the mean subjective sleep efficiency was 62.17% (SD = 17.77%) and the mean sWASO was 134.02 (SD = 83.35) minutes. Across treatment groups, most characteristics were balanced, though the baseline mean sTST was longest in the LEM5 group at 315.52 (SD = 93.50) minutes and shortest in the PBO group at 304.25 (SD = 91.46) minutes.

Interventions

The doses used in both the SUNRISE 1 and SUNRISE 2 studies were LEM5 and LEM10. PBOs were matched in appearance to the active drugs. All study drugs were taken orally in tablet form immediately before bed (i.e., when the patient intended to sleep) for the entirety of the trial.

Patients were restricted to consuming no more than 4 cups of caffeinated beverages or 400 mg of caffeine per day. They were to avoid caffeine after 13:00 on days when a PSG reading took place (for SUNRISE 1) and after 18:00 during the rest of the trial. Alcohol was restricted to no more than 2 drinks per day, not within 3 hours of bedtime, and not on days when a PSG reading took place (for SUNRISE 1). Other prohibited therapeutics included strong and moderate CYP3A inhibitors and all CYP3A inducers, all treatments for insomnia, and any medications with the purposes of inducing sleep or wakefulness, or with known sedating or alerting effects.

SUNRISE 1 Study

ZOL was the active comparator in the trial. According to the sponsor, ZOL was available as both 6.25 mg and 12.5 mg doses for the trials, and the former was selected.¹⁷ The 6.25 mg dose is recommended for females and older adult patients.¹⁷ Health Canada product monographs exist for ZOL sublingual orally disintegrating tablets in 5 mg and 10 mg doses, with the lower dose being recommended for females and older adult patients.²⁵

During the single blind run-in period, patients received 1 LEM-matched PBO tablet and 1 ZOL-matched PBO tablet. During the DB treatment period, all patients took 2 tablets each night based on the treatment group to which they were randomized:

• LEM5 — 1 LEM 5 mg tablet and 1 ZOL-matched PBO tablet

• LEM10 — 1 LEM 10 mg tablet and 1 ZOL-matched PBO tablet

• ZOL — 1 ZOL 6.25 mg tablet and 1 LEM-matched PBO tablet

• PBO — 1 LEM-matched PBO tablet and 1 ZOL-matched PBO tablet.

SUNRISE 2 Study

During the single blind run-in period, patients received 1 LEM-matched PBO. During the DB treatment period 1, all patients took 1 tablet each night based on the treatment group to which they were randomized:

• LEM5 — 1 LEM 5 mg tablet

• LEM10 — 1 LEM 10 mg tablet

• PBO — 1 LEM-matched PBO tablet.

During the DB treatment period 2, all patients took 1 tablet each night based on the treatment group to which they were randomized at the beginning of the trial, with patients in the PBO group rerandomized to active treatment (i.e., LEM5 or LEM10).

Outcomes

A list of efficacy end points identified in the CADTH systematic review protocol that were assessed in the clinical trials included in this review are summarized in Table 9. A detailed discussion and critical appraisal of the outcome measures is provided in Appendix 4.

Table 9: Summary of Outcomes of Interest Identified in the CADTH Review Protocol

EQ-5D-3L = 3-level EQ-5D; FSS = Fatigue Severity Scale; HRQoL = health-related quality of life; ISI = Insomnia Severity Index; LPS = latency to persistent sleep; NR = not reported; PGI-Insomnia = Patient Global Impression–Insomnia; sSOL = subjective sleep onset latency; sTST = subjective total sleep time; sWASO = subjective wake after sleep onset; TST = total sleep time; WASO = wake after sleep onset; WASO2H = wake after sleep onset in the second half of the night.

^aPrimary and key secondary outcomes were included in the multiple testing procedure.

Efficacy Outcomes

Health-Related Quality of Life

The EQ-5D-3L is a generic instrument and was used in both trials to measure HRQoL. The 5 domains include mobility, self-care, usual activities, pain/discomfort, and anxiety/depression along with a visual analogue scale where patients rate their health from 0 = worst imaginable health state to 100 = best imaginable health state.^17,40,41 No information on the validity, reliability, responsiveness, or MID of the EQ-5D-3L for patients with insomnia was identified from the literature.

Severity of Symptoms

The ISI, FSS, and PGI-Insomnia were used in both trials to measure the severity of insomnia. The ISI is a 7-item patient-reported questionnaire that captures the nature, severity, and impact of insomnia.^17,42 The 7 dimensions are severity of sleep onset, sleep maintenance, early morning awakening problems, sleep dissatisfaction, interference of sleep difficulties with daytime functioning, noticeability of the sleep problems by others, and distress caused by the sleep difficulties. Each item is answered based on a 5-point Likert scale, where 0 = no problem to 4 = very severe problem, for a total score ranging from 0 to 28, with a higher score indicating more acute symptoms of insomnia. Item 4 to item 7 were analyzed as a measure of daytime functioning.¹⁷ Strong convergent validity was demonstrated between the ISI items and subjective sleep measures as well as between the ISI total score and the Pittsburgh Sleep Quality Index (PSQI) total score and the 12-item Short Form (12) Health Survey mental component.⁴² There was weak to moderate correlation between the ISI total score and the Short Form (12) Health Survey physical component⁴² and PSG-derived sleep measures at baseline and posttreatment.⁴³ Criterion validity was demonstrated with the ISI being able to identify insomnia among 183 patients evaluated for treatment.⁴² The ISI was found to have high internal consistency, with moderate to strong correlations between individual items and the total ISI score.^42,44 Several studies found that the ISI is sensitive to detecting changes in perceived sleep difficulties with treatment.^42-44 A 7-point reduction in the ISI total score corresponded to moderate improvements, a 9-point reduction corresponded to marked improvements,⁴² and a 6-point reduction was suggested to be a clinically meaningful improvement in patients with primary insomnia.⁴⁵ There was a lack of evidence assessing the psychometric properties and MID for only item 4 to item 7.

The FSS is a 9-item patient-reported questionnaire that captures a patient’s agreement with each statement from 1 = strongly disagree to 7 = strongly agree.¹⁷ The FSS responses are summed for a total score and higher scores indicate greater fatigue. Construct validity was demonstrated with moderate to strong correlation between the FSS and fatigue as measured on a visual analogue scale as well as the Epworth Sleepiness Scale in patients with sleep-wake disorders.⁴⁶ The FSS showed excellent internal consistency and strong test-retest reliability in patients with sleep-wake disorders.⁴⁶ No information on the responsiveness or MID of the FSS for patients with insomnia was identified from the literature.

The PGI-Insomnia is a patient-reported questionnaire that evaluates the patient’s perception of how the study drug has impacted their sleep.¹⁷ Three items assess the drug’s effects (i.e., helped or worsened sleep, decreased or increased time to fall asleep, and increased or decreased TST) and 1 item assesses the patient’s perceived appropriateness of the drug strength. The former 3 items are rated on a 3-point scale, where 1 = positive effect, 2 = neutral effect, and 3 = negative effect, and the latter item involves rating the medication on a separate 3-point scale, where 1 = too strong, 2 = just right, and 3 = too weak. No information on the validity, reliability, responsiveness, or MID of the PGI-Insomnia for patients with insomnia was identified from the literature.

Objective Sleep Latency, Sleep Maintenance, Total Sleep Time, and Sleep Efficiency

PSG was used in the SUNRISE 1 trial to objectively measure aspects of sleep. PSG readings scored sleep parameters and sleep architecture in 30-second epochs using standard criteria performed by trained scorers. In SUNRISE 1, PSG parameters were averaged for pairs of consecutive days when readings took place. The parameters included LPS (a measure of sleep latency), WASO (a measure of sleep maintenance), WASO2H, TST, and sleep efficiency. Briefly, LPS was defined as the number of minutes from lights off to the first epoch of 20 consecutive epochs of nonwakefulness; WASO was the number of minutes spent awake from the onset of persistent sleep until lights on; WASO2H was the number of minutes spent awake during the interval from 240 minutes after lights off (i.e., second half of the night) until lights on; TST was the number of minutes of sleep from sleep onset until terminal awakening; and sleep efficiency was the proportion of time spent asleep per time in bed, calculated as TST divided by the interval from lights off until lights on. The sponsor suggested the following thresholds as clinically meaningful: a decrease in LPS of 15 minutes, a decrease in WASO of 20 minutes or a decrease of 10 minutes for each half of the night, and an increase in sleep efficiency by 5%.¹⁷ These thresholds were similar to those suggested in the literature.¹⁰

Evidence from the literature suggested that LPS derived from PSG does not accurately distinguish between primary insomnia and normal sleepers.⁴⁷ There was strong test-retest reliability when comparing the mean of LPS measurements over 3 nights in patients with insomnia.⁴⁸ No information on the responsiveness to change of LPS for patients with insomnia was identified from the literature. A mean change of at least a 10-minute difference between treatment versus PBO groups was suggested as a threshold for clinical importance.¹⁰

Construct validity was demonstrated between PSG-derived and actigraphy-derived measures of WASO,⁴⁹ but it was found that PSG-derived measures do not accurately distinguish between primary insomnia and normal sleepers.⁴⁷ There was adequate test-retest reliability when comparing the mean of WASO measurements over 3 nights in patients with insomnia.⁴⁸ No information on the responsiveness of WASO for patients with insomnia was identified from the literature. A mean change of at least a 20-minute difference between treatment versus PBO groups was suggested as a threshold for clinical importance.¹⁰ No information on the validity, reliability, or responsiveness of WASO2H for patients with insomnia was identified from the literature.

Moderate construct validity was demonstrated between PSG-derived and self-reported measures of TST,⁵⁰ but it was found that PSG-derived measures do not accurately distinguish between primary insomnia and normal sleepers.⁴⁷ There was adequate test-retest reliability when comparing the mean of TST measurements over 3 nights in patients with insomnia.⁴⁸ No information on the responsiveness of TST for patients with insomnia was identified from the literature. A mean change of at least a 20-minute difference between treatment versus PBO groups was suggested as a threshold for clinical importance.¹⁰

Strong construct validity was demonstrated between PSG-derived and actigraphy-derived sleep efficiency,⁵¹ but it was found that PSG-derived measures do not accurately distinguish between primary insomnia and normal sleepers.⁴⁷ No information on the reliability or responsiveness of sleep efficiency for patients with insomnia was identified from the literature. A mean change of at least a 5% difference between treatment versus PBO groups was suggested as a threshold for clinical importance.¹⁰

Subjective Sleep Latency, Sleep Maintenance, Total Sleep Time, Sleep Efficiency, and Quality of Sleep

The patient-recorded electronic sleep diary was used to track subjective sleep measures in both the SUNRISE 1 and SUNRISE 2 studies. The diary was completed in the first hour upon waking throughout the trials and included sSOL (sleep latency), sWASO (sleep maintenance), sTST, subjective sleep efficiency, and quality of sleep. The parameters were defined as follows: sSOL was the estimated number of minutes from the time that the patient attempted to sleep until sleep onset; sWASO was the sum of the estimated minutes of wake during the night after initial sleep onset until when the patient stopped trying to sleep for the night, operationalized as the time the patient got out of bed for the day; sTST was the derived minutes of sleep from sleep onset until when the patient stopped trying to sleep for the night; and subjective sleep efficiency was the proportion of sTST per subjective time spent in bed, calculated as the interval from the time that a patient reported attempting to sleep until the time the patient stopped trying to sleep for the night, operationalized as the time the patient got out of bed for the day, and time spent asleep derived from subjective time spent in bed minus sWASO. Quality of sleep of the previous night was rated on a scale from 1 to 9, where 1 was extremely poor and 9 was extremely good.

Strong convergent validity was demonstrated between the sSOL and the “difficulty falling asleep” item of the ISI.⁴² Moderate convergent validity was demonstrated between the sSOL and the PSQI⁵² as well as the actigraphy-derived SOL.⁵³ Discriminant validity has been demonstrated between patients with insomnia and controls.⁵⁴ No information on the reliability or responsiveness of sSOL for patients with insomnia was identified from the literature. A mean change of at least a 20-minute difference between treatment versus PBO groups was suggested as a threshold for clinical importance.¹⁰

Strong convergent validity was demonstrated between the sWASO and the ISI “difficulty of staying asleep” item.⁴² Moderate convergent validity was demonstrated between the sWASO and the PSQI⁵² as well as the actigraphy-derived WASO.⁵³ Discriminant validity has been demonstrated between patients with insomnia and controls.⁵⁴ No information on the reliability or responsiveness of sWASO for patients with insomnia was identified from the literature. A mean change of at least a 30-minute difference between treatment versus PBO groups was suggested as a threshold for clinical importance.¹⁰

Moderate to strong convergent validity was demonstrated between the sTST and the total ISI score,⁴² actigraphy-derived TST,⁵³ and PSG-derived TST.⁵⁰ Discriminant validity has been demonstrated between patients with insomnia and controls.⁵⁴ No information on the reliability or responsiveness of sTST for patients with insomnia was identified from the literature. A mean change of at least a 30-minute difference between treatment versus PBO groups was suggested as a threshold for clinical importance.¹⁰

Strong convergent validity was demonstrated between the subjective sleep efficiency and the total ISI score⁴² as well as the PSQI.⁵² Weak convergent validity was demonstrated between subjective sleep efficiency and actigraphy-derived sleep efficiency.⁵³ Discriminant validity has been demonstrated between patients with insomnia and controls.⁵⁴ No information on the reliability or responsiveness of subjective sleep efficiency for patients with insomnia was identified from the literature. A mean change of at least a 10% difference between treatment versus PBO groups was suggested as a threshold for clinical importance.¹⁰

Responder analyses were conducted on days 1 to 2 and days 29 to 30 using PSG (i.e., objective responders) and during the first 7 nights of treatment and last 7 nights of treatment using the sleep diary (i.e., subjective responders) for both sleep onset and sleep maintenance in the SUNRISE 1 study and at 6 months and 12 months using the sleep diary in the SUNRISE 2 study. Objective sleep onset response was defined as LPS being 20 minutes or less, provided the mean baseline LPS was greater than 30 minutes. Subjective sleep onset response was defined as sSOL being 20 minutes or less, provided the mean baseline sSOL was greater than 30 minutes. Objective sleep maintenance response was defined as WASO being 60 minutes or less, provided the mean baseline WASO was greater than 60 minutes and was reduced by more than 10 minutes compared to baseline. Subjective sleep maintenance response was defined as sWASO being 60 minutes or less, provided the mean baseline sWASO was greater than 60 minutes and was reduced by more than 10 minutes compared to baseline.

Harms Outcomes

The incidence and seriousness of AEs, withdrawals due to AEs, and deaths were reported for the safety population during the treatment and follow-up periods in both trials. AEs, SAEs, and protocol-defined notable harms were described based on preferred term and associated system organ class. Notable harms from the CADTH systematic review protocol included cognitive or psychomotor impairment (e.g., somnolence, impaired driving performance, postural instability or falls, impaired attention, MVA, workplace accidents), rebound insomnia, withdrawal symptoms or dependence, drug misuse, suicidal ideation, and hallucinations.

Postural stability was assessed using the Cognitive Drug Research posture device, which measures directional trunk movements or body sway through a cable attached to the patient’s waist and connected to an ataxiameter.¹⁷ Tests were performed within 5 minutes of waking and patients were instructed to stand as still as possible with their eyes closed for 1 minute. Body sway was measured in units of one-third degree of the angle of arc where a higher number indicated greater sway and less postural stability. The sponsor suggested that a 7-unit difference between LEM and PBO groups would be clinically meaningful.¹⁷ No information on the validity, reliability, or responsiveness of postural stability measures for patients with insomnia was identified by the CADTH review team from the literature. A 7-unit change in body sway from baseline (or a 35% increase) has been associated with an alcohol dose of 0.5 g per kilogram body mass,⁵⁵ leading to impairment on tasks requiring postural stability.⁵⁶

Impaired attention was captured using the CPAB, a study-specific battery of assessments.¹⁷ Tests were performed after the postural stability test (i.e., soon after waking). The entire CPAB consisted of 9 tasks: simple reaction time, choice reaction time, digit vigilance, immediate word recall, delayed word recall, word recognition, picture recognition, numeric working memory, and spatial working memory. Scores could be calculated for 4 domains, 2 of which were related to attention: power of attention and continuity of attention. For power of attention scores, a lower number is better, while for continuity of attention scores, a higher number is better. No information on the validity, reliability, responsiveness, or MID of the CPAB for patients with insomnia was identified from the literature.

Rebound insomnia was assessed based on sSOL and sWASO responses during the follow-up period. Rebound insomnia was suggested if the sSOL or sWASO was at least 5 minutes longer during the follow-up period compared to screening.¹⁷

Withdrawal symptoms or dependence was assessed using the T-BWSQ at the end of the study. Patients responded to each listed symptom as being present or absent and the severity, where 0 = no, 1 = yes (moderate), and 2 = yes (severe).^17,57 The responses were summed for a total score. No information on the validity of the T-BWSQ for patients with insomnia was identified from the literature. High interitem associations were found for the scores of 3 perceptual items: sensitivity to smell, touch, and the perception of things moving when they were still; meanwhile, items such as sensitivity to noise, depression, and shaking were generally independent.⁵⁷ The T-BWSQ scores during the period of withdrawal were greater in patients taking higher drug doses and for longer periods of time.⁵⁷ A mean change of at least 3 points indicates a patient experienced withdrawal symptoms.⁵⁷

Suicidal ideation was captured using the C-SSR, which assesses a patient’s degree of suicidality (suicidal ideation and behaviour).^17,58 Four constructs are measured, including the severity of ideation rated on a 5-point ordinal scale, the intensity of ideation containing 5 items rated on a 5-point ordinal scale, behaviour rated on a nominal scale, and lethality rated on a 6-point ordinal scale for actual lethality or on a 3-point ordinal scale for potential lethality of attempts.⁵⁸

Statistical Analysis

The statistical analysis of efficacy end points conducted in the SUNRISE 1 and SUNRISE 2 studies is summarized in Table 10.

Sleep diary data handling rules were applied to sleep diary data when responses were illogical. The 3 main errors were related to a.m. and p.m. and 24-hour clock use, incorrectly choosing the hour versus minutes, and reporting a final wake time that was later than the time the patient got out of bed for the day. The data presented in the CADTH review are with data handling rules applied.

Table 10: Statistical Analysis of Efficacy End Points — SUNRISE 1 AND SUNRISE 2 Studies

AR(1) = autoregressive; CCMV = complete case missing value; CMH = Cochran-Mantel-Haenszel; FAS = full analysis set; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; LPS = latency to persistent sleep; MAR = missing at random; MMRM = mixed model of repeated measures; MNAR = missing not at random; PBO = placebo; PP = per-protocol; sSOL = subjective sleep onset latency; sWASO = subjective wake after sleep onset; TOEP = Toeplitz covariance matrix; UN = unstructured covariance; vs. = versus; WASO = wake after sleep onset; WASO2H = wake after sleep onset in the second half of the night; ZOL = zolpidem tartrate.

Primary Outcome

SUNRISE 1 Study

The primary end point was the change from baseline for mean LPS on days 29 to 30 for LEM10 and LEM5 compared to PBO. The null hypothesis was that there was no difference in the mean change from baseline for LPS on days 29 to 30 for LEM10 versus PBO, then LEM5 versus PBO, while the alternative hypothesis was that a treatment difference existed.

The geometric LSM treatment ratio of changes from baseline have been reported in the CADTH review for active treatment (i.e., LEM or ZOL) versus PBO of the visit per baseline ratio in each group (i.e., [visit per baseline ratio of LEM10] versus [visit per baseline ratio of PBO]).

SUNRISE 2 Study

The primary end point was the mean change from study baseline in sSOL at month 6. The null hypothesis was that there was no difference in the mean change from study baseline at month 6 for LEM10 versus PBO, then LEM5 versus PBO, while the alternative hypothesis was that a treatment difference existed.

Similar to SUNRISE 1, the geometric LSM treatment ratio of changes from baseline have been reported.

Power Calculation

SUNRISE 1 Study

The sample size was estimated for each comparison of LEM10 versus PBO and LEM5 versus PBO for the mean change from baseline for the primary end point (LPS) based on a 2-sided t test at the 0.05 alpha level for each comparison. Based on a previous dose-finding study of LEM,⁵⁹ it was determined that a sample size of 250 patients for LEM5, 250 patients for LEM10, and 200 patients for PBO had at least 95% power for detecting estimated treatment differences of –1.15 between LEM10 and PBO, and –0.75 between LEM5 and PBO for log-transformed LPS values.

Power estimates were made for the key secondary end points (WASO, WASO2H, and sleep efficiency). It was determined that a sample size of 250 patients for each of the LEM5, LEM10, and ZOL groups and 200 patients for the PBO group had at least 95% power for detecting a statistically significant difference between LEM and PBO for change from baseline for sleep efficiency and at least 80% power for detecting a statistically significant difference between LEM10 and ZOL or PBO for change from baseline for WASO2H and WASO based on a 2-sided, 2-sample t test at a significance level of 5%. For sleep efficiency, the estimated treatment differences were 7% and 5% for LEM10 versus PBO and LEM5 versus PBO, respectively. For WASO2H, the estimated treatment difference was –11 minutes between LEM10 and ZOL while for WASO, the estimated treatment difference was –15 minutes between LEM10 and PBO.

SUNRISE 2 Study

Sample size calculations were conducted in a similar manner as in the SUNRISE 1 study. It was determined that a sample size of 300 patients per treatment group at a 5% (2-sided) level of significance had greater than 90% power for detecting an estimated treatment difference in sSOL of –8.7 minutes when comparing a dose of LEM to PBO.

For key secondary end points, a sample size of 300 patients in each group had more than 99% power for detecting an estimated treatment difference in subjective sleep efficiency of 5.5% and 90% power for detecting an estimated treatment difference in sWASO of –11.4 minutes when comparing LEM to PBO. A total of 900 patients provided greater than 99% power for detecting a treatment difference in sleep onset responder rates of 16% and sleep maintenance responder rates of 24.4% compared with PBO.

Statistical Testing

SUNRISE 1 Study

Statistical analyses were conducted after study completion and database locks. Continuous variables were reported using descriptive statistics (e.g., number of observations, mean, SD, median). Categorical variables were reported as numbers and percentages.

All statistical testing was performed at the 5% level of significance (2-sided). Where statistical comparisons were not defined, pairwise comparisons were tested. Efficacy end points were calculated as the average of the pairs of values (e.g., average of day 1 and day 2 [days 1 to 2]).

To control the overall type I error at the 0.05 significance level, a gatekeeping testing procedure was used where the first primary efficacy end point comparison (i.e., LPS for LEM10 versus PBO) was performed at the 0.05 significance level and subsequent testing proceeded only if the previous test was statistically significant. Tests were conducted in the following order:

• LPS — LEM10 versus PBO

• LPS — LEM5 versus PBO

• sleep efficiency — LEM10 versus PBO

• sleep efficiency — LEM5 versus PBO

• WASO — LEM10 versus PBO

• WASO2H — LEM10 versus ZOL

• WASO — LEM5 versus PBO

• WASO2H — LEM5 versus ZOL.

The primary outcome was analyzed using a mixed model of repeated measures (MMRM) analysis with factors of age (55 years to younger than 65 years and 65 years or older), geographic region (North America and Europe), treatment, visit (days 1 to 2 and days 29 to 30), and treatment-by-visit interaction as fixed effects, and baseline LPS as a covariate based on the full analysis set (FAS). The unstructured covariance matrix was used in the analysis, or the autoregressive covariance matrix was in the case of nonconvergence of the unstructured covariance matrix. The LPS data are nonnormally distributed, and a log transformation was performed. LSMs, 95% CIs for treatment differences, and P values were calculated.

SUNRISE 2 Study

Data reporting and statistical testing were conducted in a similar manner as SUNRISE 1. End points used different baselines and were specified as the baseline for period 1 (month 1 to month 6) or period 2 (month 7 to month 12). Efficacy end points were calculated as the average of 7-day sleep diary parameters (e.g., the average of the first 7 days of treatment). Exposure data for duration of exposure were grouped by treatment (i.e., LEM5 and LEM10). Each group included (1) period 1 patients in the on-treatment FAS using the change from study baseline and (2) period 2 patients who previously received PBO on the on-treatment FAS using the change from period 2 baseline for the respective doses. Hypothesis testing was not performed for efficacy end points beyond the month 6 visit or for other secondary or exploratory end points.

To control the overall type I error at the 0.05 significance level, a gatekeeping testing procedure was used where the first primary efficacy end point comparison (i.e., sSOL for LEM10 versus PBO) was performed at the 0.05 significance level and subsequent testing proceeded only if the previous test was statistically significant. Tests were conducted in the following order:

• sSOL — LEM10 versus PBO

• sSOL — LEM5 versus PBO

• subjective sleep efficiency — LEM10 versus PBO

• subjective sleep efficiency — LEM5 versus PBO

• sWASO — LEM10 versus PBO

• sWASO — LEM5 versus PBO.

The primary outcome was analyzed using an MMRM analysis with adjustments made for the corresponding study baseline value, region (North America, Europe, New Zealand, and Asia), age group (younger than 65 years, and 65 years or older), treatment, time (first 7 nights, month 1, month 2, month 3, month 4, month 5, and month 6), and the interaction of treatment by time. The unstructured covariance matrix was used in the analysis, the Toeplitz covariance matrix was used in the case of nonconvergence of the unstructured covariance matrix, and the autoregressive covariance matrix was used in the case of nonconvergence of Toeplitz covariance matrix. The sSOL data are nonnormally distributed, and a log transformation was performed. LSMs, 95% CIs for treatment differences, and P values were calculated.

Data Imputation

SUNRISE 1 Study

Missing data were imputed using a pattern-mixture model using multiple imputation assuming data were not missing at random (MAR) and using the complete case missing value (CCMV) pattern (i.e., patients who completed all primary efficacy assessments without missing values). Missing values were imputed using all available values.

SUNRISE 2 Study

The imputation of missing data was conducted in a similar manner as in SUNRISE 1.

Subgroup Analyses

SUNRISE 1 Study

Prespecified subgroups included age groups (aged 55 years to younger than 65 years, 65 years to younger than 75 years, and 75 years or older); alternative age groups (aged 55 years to younger than 65 years and 65 years or older); sex (male or female); race (white, Black, Asian, or other); geographic region (North America or Europe); and body mass index (less than 18.5 kg/m², 18.5 kg/m² to less than 25 kg/m², 25 kg/m² to 30 kg/m², and greater than 30 kg/m²). No hypothesis testing was performed in the subgroup analyses. The CADTH review identified subgroups based on age and previous treatment as being relevant.

SUNRISE 2 Study

Prespecified subgroups were similar to those in SUNRISE 1 with the exception of age groups (younger than 65 years, 65 years to younger than 75 years, and 75 years or older); alternative age groups (younger than 65 years, and 65 years or older); and region (North America, Europe, New Zealand, and Asia). No hypothesis testing was performed in the subgroup analyses.

Sensitivity Analyses

SUNRISE 1 Study

Several sensitivity analyses were performed on the primary efficacy analyses:

• the per-protocol (PP) analysis set with multiple imputations for missing data

• a completer analysis for patients who completed all primary efficacy assessments and had no missing visits without imputations for missing data

• an as-treated analysis based on the actual treatment received regardless of randomization with multiple imputations for missing data

• an analysis assuming missing data were MAR without imputations for missing data

• an analysis assuming missing data were missing not at random (MNAR) using CCMV-4, wherein up to 4 monotone missing patterns were used for missing value imputation

• a tipping point analysis where a range of shifts was used in the multiple imputations of missing data, assuming MNAR to identify at what point the treatment effect would change the results from statistically significant to nonsignificant.

SUNRISE 2 Study

The following sensitivity analyses were performed on the primary efficacy analyses:

• an analysis assuming missing data were MNAR using CCMV-7, wherein up to 7 monotone missing patterns were used for missing value imputation

• a similar tipping point analysis as in SUNRISE 1

• an analysis using nonlog-transformed data with multiple imputations for missing data

• an analysis using CCMV-7 on nonlog-transformed data

• a tipping point analysis on nonlog-transformed data.

Supplementary analyses were performed on the primary efficacy analyses and were similar to those in SUNRISE 1 for the PP analysis, the completer analysis, the MAR assumption analysis, and the MAR assumption analysis on nonlog-transformed data.

Secondary Outcomes

SUNRISE 1 Study

Key secondary end points included the change from baseline for mean sleep efficiency on days 29 to 30 for LEM10 and LEM5 compared to PBO, the change from baseline for mean WASO on days 29 to 30 for LEM10 and LEM5 compared to PBO, and the change from baseline for mean WASO2H on days 29 to 30 for LEM10 and LEM5 compared to ZOL.

The same MMRM model was used with factors of age group (aged 55 years to younger than 65 years, and 65 years or older), geographic region, treatment, visit (days 1 to 2 and days 29 to 30), and treatment-by-visit interaction as fixed effects, and baseline as a covariate based on the FAS. Similar to the primary outcomes, missing data were imputed using a pattern-mixture model with the same missing data assumptions and imputation methods.

Responder analyses were conducted using the Cochran-Mantel-Haenszel test and adjusted for age group. Missing data were considered nonresponders and no multiplicity adjustments were made.

SUNRISE 2 Study

Key secondary end points included the mean change from study baseline in subjective sleep efficiency at month 6 and the mean change from study baseline of sWASO at month 6.

As with SUNRISE 1, an MMRM model was used with similar adjustments made for the corresponding study baseline value, region, age group (younger than 65 years, and 65 years or older), treatment, time (first 7 nights, month 1, month 2, month 3, month 4, month 5, and month 6) and the interaction of treatment by time. Missing data were imputed in a similar manner as in SUNRISE 1.

Responder analyses were conducted in a similar manner as in SUNRISE 1 with adjustments made for age group and region.

Analysis Populations

SUNRISE 1 Study

The FAS (N = 1,006) consisted of all randomized patients who received at least 1 dose of study drug and had at least 1 postdose primary efficacy measurement. Unless otherwise specified, all efficacy end points were summarized and analyzed using the FAS. The summaries for the FAS and PP analysis set were based on patients “as randomized.”

The safety analysis set (N = 1,006) consisted of all randomized patients who received at least 1 dose of study drug and had at least 1 postdose safety assessment. The summary for the safety analysis set was based on patients “as treated.”

The PP analysis set (N = 992) consisted of all randomized patients who received a protocol-assigned study drug and did not have a major protocol deviation that was likely to affect the primary or key secondary efficacy end points. These included being randomized during the run-in period before randomization visit 5, and deviations of treatment assignment, treatment administration, and/or dispensing (e.g., patients who received the wrong dose or completed PSG assessment on visit 7 and visit 8 without being dosed). Most patients (more than 95%) were included in the PP analysis set and the most common reason for exclusion was due to the study drug not being administered or an incorrect study drug being dispensed (12 [1.2%] patients). One patient randomized to the LEM10 group received PBO and was included in the PBO group in the safety analysis set.

SUNRISE 2 Study

The FAS (N = 949) consisted of all randomized patients who received at least 1 dose of study drug and had at least 1 postdose primary efficacy measurement. Efficacy analyses for period 1 were performed on the FAS unless otherwise specified. The on-treatment FAS was made up of patients who received at least 1 dose of LEM and had at least 1 postdose primary efficacy measurement. The FAS and PP analysis set were based on patients “as randomized.”

The safety analysis set (N = 947) consisted of all randomized patients who received at least 1 dose of study drug and had at least 1 postdose safety assessment. The on-treatment safety analysis set was made up of patients who received at least 1 dose of LEM and had at least 1 postdose safety assessment. The safety analysis set was based on patients “as treated.”

The 6-month completer analysis set (N = 639) consisted of all randomized patients in the FAS who had all efficacy assessments up to and including month 6 (i.e., week 1 and month 1 to month 6 visits) without missing primary or key secondary efficacy assessments at any of these visits.

The PP analysis set (N = 921) consisted of all randomized patients who sufficiently complied with the protocol and included all patients in the FAS who sufficiently complied with the protocol during period 1. The criteria for exclusion in this analysis set were violated inclusion or exclusion criteria, duplicate randomization, a missing primary efficacy assessment, prohibited concomitant medication, a study drug not being administered, an incorrect study drug kit dispensed, and noncompliance with study medication. Most patients (more than 94%) were included in the PP analysis set and the most common reason for exclusion was due to less than 80% compliance (15 patients; 1.5%).

Results

Patient Disposition

Patient disposition is summarized in Table 11.

Table 11: Patient Disposition — SUNRISE 1 and SUNRISE 2 Studies

FAS = full analysis set; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; NA = not applicable; PBO = placebo; PP = per-protocol; ZOL = zolpidem tartrate.

^aDiscontinued from period 1 in the SUNRISE 2 study.

^bAs reported on the patient disposition case report form. Multiple “other” reasons for discontinuation may have been checked on the case report form; therefore, percentages may add up to more than the overall percentage of patients who discontinued.

^cCorresponding adverse events leading to withdrawal from the study or study drug were reported on the adverse events case report form.

^dOne patient who was randomized to LEM10 received PBO instead; therefore, this patient was counted in the PBO group rather than the LEM10 group.

Sources: SUNRISE 1 Clinical Study Report¹⁷ and SUNRISE 2 Clinical Study Report.¹⁸

SUNRISE 1 Study

In the SUNRISE 1 trial, 3,537 individuals were screened, of whom 2,531 (71.6%) were screening failures. The main reasons for screening failures were not meeting eligibility criteria (65.1%) and consent withdrawal (4.4%). Additional population-level data were not available to indicate which eligibility criteria were most commonly not met. In total, 1,006 patients were randomized in the treatment period. All patients who were randomized were treated with study medication. Overall, 962 (95.6%) patients completed the trial with similar completion rates across treatment groups ranging from 93% in the ZOL group to 97% in the LEM5 group.

SUNRISE 2 Study

In the SUNRISE 2 trial, 2,059 individuals were screened, of whom 1,088 (52.8%) were screening failures. The main reasons for screening failures were not meeting eligibility criteria (45.5%) and consent withdrawal (4.3%). Additional population-level data were not available to indicate which eligibility criteria were most commonly not met. In total, 971 patients were randomized in treatment period 1. All but 12 patients who were randomized were treated with study medication. In total, 734 (75.6%) patients completed period 1, with a higher proportion of discontinuations in the LEM10 group (25.1%) compared to either the LEM5 group (16.0%) or the PBO group (18.1%). The most common reasons for discontinuation were withdrawal of consent and patient choice.

Exposure to Study Treatments

Exposure to study medications is summarized in Table 12 and Table 13 for the SUNRISE 1 study and SUNRISE 2 study, respectively.

Table 12: Treatment Exposure — SUNRISE 1 Study, Safety Analysis Set

LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; PBO = placebo; SD = standard deviation; ZOL = zolpidem tartrate.

Source: SUNRISE 1 Clinical Study Report.¹⁷

Table 13: Treatment Exposure — SUNRISE 2 Study, On-Treatment Safety Analysis Set

LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; SD = standard deviation.

Note: One month was defined as at least 23 days, 3 months was defined as at least 83 days, 6 months was defined as at least 173 days, 9 months was defined as at least 263 days, and 12 months was defined as at least 353 days in the SUNRISE 2 study. Data were reported for period 2 (i.e., after patients who received PBO in period 1 were rerandomized to either LEM5 or LEM10).

Source: SUNRISE 2 Clinical Study Report.¹⁸

Treatment compliance was assessed based on the amount of study drug returned at the end of the treatment periods and most patients (more than 95% of patients in SUNRISE 1 and more than 92% and 73% of patients in SUNRISE 2 period 1 and period 2, respectively) were between 80% and 100% compliant with the study medications during the trial, with similar compliance among treatment groups.

In SUNRISE 1, most patients (at least 93%) had completed at least 4 weeks of treatment in the trial and the mean duration of exposure for patients in the safety analysis set was similar among all treatment groups.

In SUNRISE 2, most patients (at least 76.7%) had completed at least 6 months of active treatment in the trial; the proportion of patients was lower among those in the LEM10 group (76.7% of patients) compared to the LEM5 group (83.7%). The mean duration of exposure for patients in the on-treatment safety analysis set was higher in the LEM5 group at 260.5 (SD = 116.34) days compared to the LEM10 group at 245.2 (SD = 122.09) days.

Protocol Deviations

Protocol deviations for both pivotal trials are summarized in Table 14.

SUNRISE 1 Study

Overall, 114 (11.3%) patients had at least 1 major protocol deviation. The most common protocol deviations were related to inclusion criteria (3.8%), study procedures or assessments (3.6%), study drug administration or dispensing (2.6%), and exclusion criteria (1.9%).

SUNRISE 2 Study

Overall, 27 (2.8%) patients had at least 1 major protocol deviation. The most common protocol deviations were related to prohibited concomitant medication (0.7%), study procedures or assessments (0.7%), and visit scheduling (0.5%).

Table 14: Protocol Deviations — SUNRISE 1 AND SUNRISE 2 Studies, FAS

FAS = full analysis set; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; PBO = placebo; ZOL = zolpidem tartrate.

Sources: SUNRISE 1 Clinical Study Report¹⁷ and SUNRISE 2 Clinical Study Report.¹⁸

Efficacy

Only those efficacy outcomes and analyses of subgroups identified in the CADTH systematic review protocol are reported as follows. Refer to Appendix 3 for additional detailed efficacy data.

Health-Related Quality of Life: EQ-5D-3L and Visual Analogue Scale

SUNRISE 1 Study

Most patients in the trial reported having “no problems” for any of the dimensions at either baseline or day 31 (Table 15). At most, there were 3 patients in any treatment group who reported “extreme problems” for a dimension at any time. For the EuroQol Visual Analogue Scale (EQ VAS) at day 31, the LSM treatment differences for LEM10 versus PBO and LEM5 versus PBO were 2.53 points (95% CI, 0.69 points to 4.38 points) and 0.52 points (95% CI, –1.32 points to 2.37 points), respectively, and for LEM10 versus ZOL and LEM5 versus ZOL were 3.10 points (95% CI, 1.35 points to 4.85 points) and 1.09 points (95% CI, –0.65 points to 2.83 points), respectively.

SUNRISE 2 Study

Similarly, most patients in the SUNRISE 2 trial reported having “no problems” for any of the dimensions at baseline, month 1, month 3, and month 6 (Table 16). At most, there were 9 patients in any treatment group who reported “extreme problems” for a dimension at any time. For the EQ VAS at month 1, the LSM treatment differences for LEM10 versus PBO and LEM5 versus PBO were 1.26 points (95% CI, –0.81 points to 3.32 points) and 0.62 points (95% CI, –1.43 points to 2.67 points), respectively, while at month 3 they were 0.86 points (95% CI, –1.42 points to 3.15 points) and 1.03 points (95% CI, –1.23 points to 3.29 points), respectively, and at month 6 they were 1.04 points (95% CI, –1.36 points to 3.43 points) and –0.96 points (95% CI, –3.31 points to 1.39 points), respectively.

Long-term EQ-5D results based on 3 months, 6 months, 9 months, and 12 months of exposure to LEM showed that most patients reported “no problems” for any of the dimensions (Table 62). In general, the long-term results for EQ VAS scores supported the results observed in the first 6 months.

Table 15: Change From Baseline EQ-5D-3L and Visual Analogue Scale (Exploratory Outcome) — SUNRISE 1 Study, FAS

ANCOVA = analysis of covariance; CI = confidence interval; EQ-5D-3L = 3-level EQ-5D; FAS = full analysis set; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; LSM = least squares mean; NA = not applicable; PBO = placebo; SD = standard deviation; SE = standard error; VAS = visual analogue scale; ZOL = zolpidem tartrate.

^aThe number of patients analyzed for the change from baseline at this time point (not the number of patients who provided data at this time point).

^bBased on an ANCOVA model with factors of age group, region, treatment, and the baseline VAS as covariates. Missing values were not imputed.

^cP value was not adjusted for multiple testing (i.e., the type I error rate was not controlled).

Source: SUNRISE 1 Clinical Study Report.¹⁷

Table 16: Change From Baseline EQ-5D-3L and Visual Analogue Scale (Exploratory Outcome) — SUNRISE 2 Study, FAS

CI = confidence interval; EQ-5D-3L = 3-level EQ-5D; FAS = full analysis set; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; LSM = least squares mean; MMRM = mixed model of repeated measures; PBO = placebo; SD = standard deviation; VAS = visual analogue scale.

^aThe number of patients analyzed for the change from baseline at this time point (not the number of patients who provided data at this time point).

^bBased on an MMRM model with factors of age group, region, treatment, visit (month 1, month 3, and month 6), and treatment-by-visit interaction as fixed effects, and study baseline VAS score as a covariate. Missing values were not imputed and assumed to be missing at random.

^cP value was not adjusted for multiple testing (i.e., the type I error rate was not controlled).

Source: SUNRISE 2 Clinical Study Report.¹⁸

Severity of Symptoms: Insomnia Severity Index, Fatigue Severity Scale, and Patient Global Impression–Insomnia

SUNRISE 1 Study

For ISI item 4 to item 7 at day 31, the LSM treatment differences for LEM10 versus PBO and LEM5 versus PBO were –1.08 points (95% CI, –1.71 points to –0.46 points) and –1.10 points (95% CI, –1.73 points to –0.47 points), respectively, and for LEM10 versus ZOL and LEM5 versus ZOL were 0.33 points (95% CI, –0.26 points to 0.92 points) and 0.32 points (95% CI, –0.28 points to 0.91 points), respectively (Table 17).

Table 17: Change From Baseline ISI Item 4 to Item 7 and FSS Total Score (Other Secondary Outcomes) — SUNRISE 1 Study, FAS

ANCOVA = analysis of covariance; CI = confidence interval; FSS = Fatigue Severity Scale; ISI = Insomnia Severity Index; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; LSM = least squares mean; NA = not applicable; PBO = placebo; SD = standard deviation; SE = standard error; ZOL = zolpidem tartrate.

^aThe number of patients analyzed for the change from baseline at this time point (not the number of patients who provided data at this time point).

^bBased on an ANCOVA model with factors of age group, region, treatment, and the baseline ISI as covariates. Missing values were not imputed.

^cP value was not adjusted for multiple testing (i.e., the type I error rate was not controlled).

^dBased on an ANCOVA model with factors of age group, region, treatment, and the baseline FSS as covariates. Missing values were not imputed.

Source: SUNRISE 1 Clinical Study Report.¹⁷

For FSS total scores at day 31, the LSM treatment differences for LEM10 versus PBO and LEM5 versus PBO were –1.17 points (95% CI, –3.26 points to 0.93 points) and –1.26 points (95% CI, –3.35 points to 0.82 points), respectively, and for LEM10 versus ZOL and LEM5 versus ZOL were –0.27 points (95% CI, –2.26 points to 1.71 points) and –0.37 points (95% CI, –2.35 points to 1.60 points), respectively (Table 17).

According to the PGI-Insomnia, most patients among all treatment groups indicated that the treatment they received had a positive effect in helping them sleep (64% for each of LEM10 and LEM5, 72% for ZOL, and 42% for PBO), decreased the time to fall asleep (65% for LEM10, 60% for LEM5, 63% for ZOL, and 43% for PBO), and increased the TST (62% for each of LEM10 and LEM5, 71% for ZOL, and 44% for PBO) (Table 19). More than half of the patients in the active treatment groups responded that the medication strength was “just right” (56% for LEM10, 52% for LEM5, 52% for ZOL, and 39% for PBO) compared to patients who received PBO, where more responded that the treatment was “too weak.” Overall, it appeared that patients were more satisfied with the active treatments than PBO, though this outcome was not adjusted for multiplicity and conclusions cannot be drawn with certainty.

SUNRISE 2 Study

For ISI item 4 to item 7, the LSM treatment differences for LEM10 versus PBO and LEM5 versus PBO at month 1 were –0.94 points (95% CI, –1.51 points to –0.38 points) and –0.71 points (95% CI, –1.27 points to –0.15 points), respectively, while at month 3 were –1.36 points (95% CI, –1.96 points to –0.76 points) and –1.16 points (95% CI, –1.75 points to –0.57 points), respectively, and at month 6 were –1.32 points (95% CI, –1.92 points to –0.71 points) and –1.30 points (95% CI, –1.90 points to –0.71 points), respectively (Table 18).

For FSS total scores, the LSM treatment differences for LEM10 versus PBO and LEM5 versus PBO at month 1 were –2.04 points (95% CI, –3.83 points to –0.25 points) and –1.66 points (95% CI, –3.44 points to 0.12 points), respectively, while at month 3 were –3.04 points (95% CI, –4.91 points to –1.18 points) and –2.18 points (95% CI, –4.02 points to –0.34 points), respectively, and at month 6 were –2.56 points (95% CI, –4.57 points to –0.54 points) and –2.50 points (95% CI, –4.48 points to –0.52 points), respectively (Table 18).

For the PGI-Insomnia, most patients among all treatment groups indicated that the treatment they received had a positive effect in helping them sleep (more than 59%, 65%, and 67% of patients at month 1, month 3, and month 6 in the LEM groups, respectively, versus 34%, 41%, and 45% of patients for PBO at the same time points, respectively), decreased the time to fall asleep (more than 61%, 68%, and 72% of patients at month 1, month 3, and month 6 in the LEM groups, respectively, versus 40%, 42%, and 46% of patients for PBO at the same time points, respectively), and increased the TST (more than 53%, 55%, and 58% of patients at month 1, month 3, and month 6 in the LEM groups, respectively, versus 36%, 39%, and 40% of patients for PBO at the same time points, respectively) (Table 20). Around half of the patients in the LEM groups responded that the medication strength was “just right,” with more than 43% at month 1, 49% at month 3, and 53% at month 6 versus 29%, 34%, and 36% of patients in the PBO group at the same time points, respectively. Similarly, patients treated with LEM appeared to be more satisfied compared to patients treated with PBO, though the results are not adjusted for multiplicity.

Long-term ISI item 4 to item 7 and FSS total score results based on 3 months, 6 months, 9 months, and 12 months of exposure to LEM supported those observed during the first 6 months of treatment (Table 63).

The long-term PGI-Insomnia results based on 3 months, 6 months, 9 months, and 12 months of exposure to LEM were consistent with those during the main trial (Table 64).

Table 18: Change From Baseline ISI Item 4 to Item 7 and FSS Total Score (Other Secondary Outcomes) — SUNRISE 2 Study, FAS

CI = confidence interval; FAS = full analysis set; FSS = Fatigue Severity Scale; ISI = Insomnia Severity Index; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; LSM = least squares mean; MMRM = mixed model of repeated measures; PBO = placebo; SD = standard deviation.

Note: Data were reported with sleep diary handling rules to address potential errors and illogical values.

^aThe number of patients analyzed for the change from baseline at this time point (not the number of patients who provided data at this time point).

^bBased on an MMRM model with factors of age group, region, treatment, visit (month 1, month 3, and month 6), and treatment-by-visit interaction as fixed effects, and study baseline ISI score as a covariate. Missing values were not imputed and assumed to be missing at random.

^cP value was not adjusted for multiple testing (i.e., the type I error rate was not controlled).

^dBased on an MMRM model with factors of age group, region, treatment, visit (month 1, month 3, and month 6), and treatment-by-visit interaction as fixed effects, and study baseline FSS score as a covariate. Missing values were not imputed and assumed to be missing at random.

Source: SUNRISE 2 Clinical Study Report.¹⁸

Table 19: PGI-Insomnia (Exploratory Outcome) — SUNRISE 1 Study, FAS

FAS = full analysis set; LEM5 = lemborexant 5 mg; LEM10 = lemborexant 10 mg; NA = not applicable; PBO = placebo; PGI-Insomnia = Patient Global Impression–Insomnia; vs. = versus; ZOL = zolpidem tartrate.

^aP value was based on chi-square test comparing “positive” to other categories combined. The same test comparison was done with “just right” to other categories combined.

^bP value was not adjusted for multiple testing (i.e., the type I error rate was not controlled).

Source: SUNRISE 1 Clinical Study Report.¹⁷

Table 20: PGI-Insomnia (Exploratory Outcome) — SUNRISE 2 Study, FAS