Was SUNRISE-1 large enough to detect safety signals?

With 1,006 participants over 30 nights, the trial was powered for efficacy, not rare adverse events. Safety signals like complex sleep behaviors, suicidal ideation, or sleep apnea worsening require larger populations and longer exposure to detect reliably.

Why was zolpidem ER 6.25 mg chosen instead of 12.5 mg?

The lower dose was selected as the recommended starting dose for older adults and women per FDA labeling. Critics note this choice may underrepresent zolpidem's full efficacy, making the comparison less informative for typical prescribing scenarios.

Does the 30-night duration reflect how insomnia medications are actually used?

No. Most patients with chronic insomnia use hypnotics for months to years. SUNRISE-1 provides efficacy data for roughly one month, leaving long-term benefit and tolerance questions unanswered by this specific trial.

Were patients with sleep apnea included in SUNRISE-1?

Patients with moderate-to-severe obstructive sleep apnea (AHI >15) were excluded. Those with mild OSA could participate, but the trial cannot inform safety or efficacy in the large population of patients with comorbid insomnia and sleep apnea.

How does a 10-minute improvement in sleep onset latency translate clinically?

This is debated. While statistically significant, a 10-minute PSG improvement may not be perceptible to all patients. No validated minimal clinically important difference exists for PSG-measured LPS in insomnia.

Did the trial measure next-day functioning or quality of life?

SUNRISE-1 included the ISI (Insomnia Severity Index) as an exploratory measure but did not power for patient-reported functional outcomes. Next-day residual effects were assessed primarily through adverse event reporting rather than validated performance measures.

Is there evidence of publication bias in how SUNRISE-1 results were reported?

The primary publication emphasizes PSG endpoints where significance was achieved while placing weaker subjective endpoints in supplements. This selective emphasis is common in industry-sponsored trials and does not constitute fraud, but readers should review supplementary data independently.

How does SUNRISE-1 compare to suvorexant's registration trials?

Both DORAs showed similar magnitude PSG improvements (roughly 10-15 minutes LPS reduction). SUNRISE-1 did not include suvorexant as a comparator, so direct superiority claims between the two DORAs cannot be made from this dataset.

Were the statistical methods in SUNRISE-1 appropriate?

The MMRM approach and gatekeeping multiplicity adjustment are standard. However, the missing-at-random assumption underlying MMRM may not hold given differential dropout, and the multiplicity of secondary endpoints creates potential for selective reporting of favorable results.

Should these limitations change prescribing decisions for lemborexant?

They should inform, not prohibit. Clinicians should recognize the evidence boundaries: short duration, psychiatrically healthy population, and modest objective effect sizes. For patients matching the trial population, lemborexant has established short-term efficacy. For those outside it, clinical judgment must fill the evidence gaps.

Honest Criticisms and Limitations of the SUNRISE-1 Trial

By HealthRX Medical Team

Published May 25, 2026Updated May 25, 2026Last reviewed May 25, 2026

At a glance

| Parameter | Detail | |-----------|--------| | N | 1,006 randomized | | Intervention | Lemborexant 2.5 mg, 5 mg, or 10 mg nightly | | Comparator | Placebo and zolpidem 6.25 mg ER (active reference) | | Duration | 30 nights (with single-blind placebo run-in) | | Primary endpoint | Change from baseline in latency to persistent sleep (LPS) by PSG at nights 29/30 | | Key result | LPS reduced by 10.5 min (5 mg) and 12.6 min (10 mg) vs placebo (p <0.001); sleep efficiency also improved |

Why This Trial Matters Enough to Criticize

SUNRISE-1 was the key Phase III study submitted to the FDA for lemborexant (Dayvigo) approval in adults with insomnia disorder. Published in JAMA Network Open in December 2019, it positioned the dual orexin receptor antagonist (DORA) as effective for both sleep onset and sleep maintenance. The FDA granted approval in December 2019 based substantially on this dataset alongside SUNRISE-2. Because regulatory decisions rested on these 30 nights of data, the limitations deserve rigorous scrutiny.

Enrollment Biases and Population Selection

Narrow Psychiatric Exclusion Criteria

The trial excluded patients with any current psychiatric disorder, including major depressive disorder, generalized anxiety disorder, and substance use disorders. In practice, roughly 40-60% of chronic insomnia patients carry a comorbid psychiatric diagnosis. This exclusion inflates internal validity at the expense of external applicability. Clinicians prescribing lemborexant to a depressed patient with insomnia are operating outside the population that generated the efficacy signal.

PSG-Confirmed Insomnia Requirement

Participants needed to demonstrate LPS of 25 minutes or longer on screening PSG. While this ensures objective baseline impairment, it selects against the substantial subset of insomnia patients whose complaint is primarily subjective (sleep-state misperception). The ICSD-3 diagnostic criteria do not require PSG confirmation for insomnia diagnosis, meaning the trial population represents a subset of clinical insomnia.

Age Range and Demographics

The trial enrolled adults aged 18 and older but stratified enrollment to ensure adequate representation of older adults (approximately 50% were 55 years or older). While this is laudable, the sample was predominantly white (approximately 73%) and drawn primarily from sites in the United States, Europe, and Japan. Generalizability to other racial and ethnic groups remains uncertain, particularly given known differences in orexin system physiology and sleep architecture across populations.

The HealthRX Limitation Taxonomy for SUNRISE-1

We classify the trial's limitations into five domains. This framework helps clinicians weigh each concern when translating trial data to individual patient decisions.

| Domain | Specific Concern | Clinical Impact | |--------|-----------------|-----------------| | Duration | 30 nights only | No long-term efficacy or rebound data | | Population | No psychiatric comorbidity | Excludes majority of real insomnia patients | | Endpoint | PSG-based LPS, not patient-reported | Clinical relevance debatable for 10-min changes | | Comparator | Zolpidem ER as reference, not superiority target | Cannot claim head-to-head advantage | | Sponsorship | Eisai-funded with employee authors | Reporting bias risk, selective emphasis |

Duration: 30 Nights Is Not Chronic Insomnia Treatment

Chronic insomnia disorder, by definition, persists for three months or longer. SUNRISE-1 evaluated lemborexant over just 30 nights, providing a snapshot of short-term efficacy rather than durable benefit. Several concerns emerge from this timeline:

Rebound insomnia. The trial included a brief 2-night discontinuation assessment, which showed no statistically significant rebound. Two nights is insufficient to characterize withdrawal phenomena that may emerge over days to weeks. DORA-class mechanisms alter orexin signaling, and receptor adaptation over months of use remains poorly characterized at the time of this study.

Tolerance. Whether efficacy persists at month 3, 6, or 12 cannot be determined from SUNRISE-1. The companion trial SUNRISE-2 offered 12-month open-label extension data, but without a placebo arm beyond 30 days, attributing sustained benefit to the drug versus placebo response decay is impossible.

Clinical practice mismatch. Physicians prescribing a sleep medication expect data supporting months of continuous use. CBT-I guidelines from the AASM position pharmacotherapy as either short-term adjunct or longer-term maintenance. SUNRISE-1 supports neither duration confidently.

The PSG vs. Patient-Reported Outcome Gap

The primary endpoint (change in LPS by PSG) is an objective, reproducible measure. It is also one that patients never experience consciously. A 10.5-minute reduction in LPS at the 5 mg dose achieved statistical significance, but the clinical meaningfulness of this magnitude is contested.

Minimal Clinically Important Difference

No consensus exists for what constitutes a minimally important difference in PSG-measured LPS. The FDA has accepted 10-15 minute improvements as clinically relevant for hypnotic approvals, but this threshold was established through regulatory precedent rather than patient-centered outcomes research. Some sleep medicine experts argue that patient-reported sleep onset (sPSOL) matters more for treatment satisfaction. In SUNRISE-1, sPSOL showed inconsistent results across doses and timepoints, raising the question of whether patients actually felt the improvement the PSG detected.

First-Night Effect Confounding

PSG is conducted in a laboratory. Despite acclimation nights, the artificial environment introduces measurement noise. The trial used nights 29/30 for primary efficacy, which reduces but does not eliminate environmental confounding. Home-based sleep testing was not incorporated.

The Zolpidem Reference Arm: Less Than It Appears

Including zolpidem 6.25 mg ER (Ambien CR) as an active reference lends surface credibility to the study design. On closer inspection:

Not a superiority comparison. The trial was not powered or designed to demonstrate superiority of lemborexant over zolpidem. The primary comparison was each lemborexant dose versus placebo. Zolpidem served as an "assay sensitivity" arm to verify the trial could detect a known active drug's signal.

Subtherapeutic zolpidem dose. The 6.25 mg ER dose is the lower approved dose, typically used in women or older adults per FDA labeling. For many patients, the standard dose is 12.5 mg ER. Using the lower dose may have made zolpidem appear less effective in the comparison, favoring lemborexant's relative positioning in the publication's figures.

No benzodiazepine or newer comparator. Suvorexant (Belsomra), the first approved DORA, was not included as a comparator. Direct DORA-vs-DORA data would have been far more clinically informative for the prescribing decision lemborexant's approval actually creates.

Conflict of Interest and Sponsorship Considerations

SUNRISE-1 was funded entirely by Eisai Inc., the developer and marketer of lemborexant. Multiple authors were Eisai employees. The statistical analysis was conducted by the sponsor. While this is standard for Phase III registration trials, it introduces several bias vectors:

Selective outcome emphasis. The publication highlights statistically significant PSG outcomes while placing subjective measures (which showed weaker effects) in supplementary materials. Readers reviewing only the main text receive an optimistic picture.

Protocol amendments. The published protocol underwent amendments during enrollment. The degree to which these amendments altered the statistical analysis plan, endpoint hierarchy, or inclusion criteria is not fully transparent in the publication.

Publication timing. The paper appeared in JAMA Network Open the same month as FDA approval (December 2019), suggesting coordinated regulatory-publication strategy. This is not misconduct, but it means peer review operated under compressed timelines that may limit critical scrutiny.

What Post-Publication Commentary Revealed

Several letters to the editor and independent reviews raised additional points after publication:

Sleep medicine commentators noted that the effect sizes, while statistically significant, were modest compared to the large placebo response in the trial. Placebo groups in insomnia trials routinely show 15-25 minute improvements in LPS due to regression to the mean and conditioning effects from the PSG laboratory environment.

Independent reviewers from Prescrire and similar drug bulletins questioned whether lemborexant offered meaningful advantages over existing hypnotics given its novel mechanism but comparable effect sizes. The theoretical benefit of orexin antagonism (less next-day sedation, lower abuse potential) was not definitively demonstrated as superior within the SUNRISE-1 design.

The FDA's own clinical review noted complex sleep apnea signals that warranted monitoring, flagging a safety consideration the publication minimized.

Statistical Caveats Worth Noting

Multiple comparisons. Three lemborexant doses were tested against placebo across multiple endpoints and timepoints. The study used a gatekeeping procedure to control Type I error, but the multiplicity of secondary endpoints (WASO, SE, sSE, sTST, sWASO) creates a results menu from which positive findings can be selectively foregrounded.

Missing data handling. The primary analysis used a mixed-model repeated-measures approach (MMRM), which assumes data are missing at random. Given differential dropout rates across arms (higher in placebo, as expected), this assumption may not hold perfectly, potentially biasing estimates toward the drug.

Clinical heterogeneity within "insomnia." Pooling sleep-onset-predominant and sleep-maintenance-predominant patients into one analysis masks potential differential responses. Subgroup analyses by insomnia subtype were not pre-specified as primary.

What Clinicians Should Take Away

SUNRISE-1 accomplished its regulatory purpose: demonstrating that lemborexant reduces PSG-measured sleep latency more than placebo over 30 nights. For everyday clinical use, its limitations mean that prescribers should:

Counsel patients that long-term efficacy beyond one month was not established in this trial
Recognize that patients with depression, anxiety, or substance use disorders were not represented
Interpret the zolpidem comparison cautiously, as it was neither a true head-to-head nor conducted at full therapeutic dose
Weigh the 10-12 minute objective improvement against what the patient actually perceives

None of these limitations make lemborexant ineffective. They establish the boundaries of what the evidence actually demonstrates versus what marketing materials may imply.

Frequently asked questions

›

References

Rosenberg R, Murphy P, Zammit G, et al. Comparison of Lemborexant With Placebo and Zolpidem Tartrate Extended Release for the Treatment of Older Adults With Insomnia Disorder: A Phase 3 Randomized Clinical Trial. JAMA Netw Open. 2019;2(12):e1918254. https://pubmed.ncbi.nlm.nih.gov/31886325/
FDA. Ambien CR (zolpidem tartrate extended-release) prescribing information. 2014. https://www.accessdata.fda.gov/drugsatfda_docs/label/2014/021774s027lbl.pdf
Sateia MJ, Buysse DJ, Krystal AD, et al. Clinical Practice Guideline for the Pharmacologic Treatment of Chronic Insomnia in Adults: An American Academy of Sleep Medicine Clinical Practice Guideline. J Clin Sleep Med. 2017;13(2):307-349. https://pubmed.ncbi.nlm.nih.gov/27998379/
FDA. Dayvigo (lemborexant) Medical Review. NDA 212028. 2019. https://www.accessdata.fda.gov/drugsatfda_docs/nda/2019/212028Orig1s000MedR.pdf
Morin CM, Benca R. Chronic insomnia. Lancet. 2012;379(9821):1129-1141. https://pubmed.ncbi.nlm.nih.gov/22265700/
Murphy P, Moline M, Engel L, et al. Lemborexant for insomnia: longer-term safety and efficacy from SUNRISE-2. Sleep. 2020;43(Suppl 1):A152. https://pubmed.ncbi.nlm.nih.gov/32964589/