An orexin-2 blocker beat zolpidem on night 13 in a dose-finding insomnia trial

A dose-finding trial of seltorexant, an experimental selective orexin-2 receptor antagonist, reported results in JAMA Psychiatry on 13 August [s1]. The headline comparison is against zolpidem, one of the most widely prescribed hypnotics, and the interesting part is not what happened on the first night but what happened on the thirteenth.

The design

The study was randomised, double-blind, and both placebo- and active-controlled, run at 55 sites in six countries [s1]. Participants were adults aged 18 to 64 and older adults aged 65 to 85, all with insomnia — an Insomnia Severity Index score of 15 or above — and no psychiatric comorbidity [s1]. They were randomised 1:1:1:1:1 to nightly oral seltorexant at 5 mg, 10 mg or 20 mg, to placebo, or to zolpidem at 5 to 10 mg, for 14 days [s1].

Outcomes were measured by polysomnography, not by self-report. The primary and key secondary outcomes were the dose-response relationship on night 1 for latency to persistent sleep (how long it took to fall properly asleep) and for wake after sleep onset over the first six hours [s1]. Because both measures were skewed at baseline, results were log-transformed and reported as back-transformed least-squares mean ratios between groups — a ratio below 1 means less time awake [s1].

In total 364 participants received treatment, with a mean age of 57.8 years and 246 (67.6%) female: 71 on seltorexant 5 mg, 74 on 10 mg, 71 on 20 mg, 75 on placebo and 73 on zolpidem [s1].

What the trial found

On night 1, the dose-response relationship for sleep latency was significant across all four prespecified models (trend test t statistics ≥ 3.99; adjusted P values < .001) [s1]. Compared with placebo, seltorexant produced a least-squares mean ratio of 0.64 (90% CI, 0.51 to 0.81) at 10 mg and 0.51 (90% CI, 0.41 to 0.64) at 20 mg [s1]. Against zolpidem, the 20 mg dose gave a ratio of 0.71 (90% CI, 0.57 to 0.88) [s1].

Wake after sleep onset showed the same pattern on night 1: ratios versus placebo of 0.68 (90% CI, 0.55 to 0.85) at 10 mg and 0.60 (90% CI, 0.48 to 0.74) at 20 mg, with a significant dose-response relationship [s1].

The night-13 result is the one that separates the drugs. Improvements in both measures were maintained at 10 mg and 20 mg of seltorexant but diminished for zolpidem [s1]. By night 13, compared with zolpidem, seltorexant improved sleep latency by 30% at 10 mg and 28% at 20 mg, and the 20 mg dose improved wake after sleep onset by 31% [s1].

Treatment-emergent adverse events were lower on the combined seltorexant doses (73 of 216, 33.8%) than on placebo (37 of 75, 49.3%) or zolpidem (31 of 73, 42.5%) [s1]. Two participants had serious treatment-emergent adverse events during the double-blind phase, one in the seltorexant 20 mg group and one in the zolpidem group [s1]. Four participants — three on 5 mg and one on 20 mg — discontinued because of asymptomatic electrocardiogram-related adverse events [s1].

The six-year gap

The trial ran from November 2017 to April 2019 and was analysed in August 2019 [s1]. It was published in August 2025. The paper states the reason directly: the timeline for submission of the data for publication was impacted by internal strategic decision-making [s1].

That disclosure is worth naming rather than skipping past. A six-year delay between analysis and publication is a live issue in evidence synthesis — meta-analyses and guideline committees can only weigh trials they can see, and a result that sits unpublished for six years is invisible to them for six years. The disclosure is unusually explicit; the underlying pattern is not unusual.

What it does not establish

This was a dose-finding study with polysomnographic endpoints over 14 nights [s1]. Two weeks is short relative to how long people actually take hypnotics, and objective sleep-laboratory measures are not the same thing as feeling rested or functioning better the next day. The trial excluded people with psychiatric comorbidity [s1], which excludes a large share of the population that presents with chronic insomnia in practice. And seltorexant is not an approved medicine; nothing here describes an available treatment option.

The other August insomnia paper

A systematic review and meta-analysis published in Sleep Health on 7 August pooled 16 studies covering 27,789 participants to examine the association between insomnia symptoms and chronotype — whether someone is naturally an early or late type, measured here with the Morningness-Eveningness Questionnaire [s2]. Compared with morning types, evening chronotypes had a substantially higher likelihood of insomnia (odds ratio, 3.47; 95% CI, 2.50 to 4.83; P < .00001) and higher Insomnia Severity Index scores (mean difference, 3.00; 95% CI, 1.70 to 4.30) [s2]. Intermediate chronotypes sat between the two (odds ratio, 1.61; 95% CI, 1.24 to 2.09) [s2].

Those are cross-sectional and prospective observational studies with moderate-to-high heterogeneity, so they establish association rather than direction of cause [s2]. Their relevance here is as a reminder that insomnia is not one condition with one lever. Pharmacology addresses sleep initiation and maintenance; the timing mismatch between a person's internal clock and their required schedule is a separate contributor that a hypnotic does not touch.

The two papers landed within a week of each other and point at the same gap: measurable improvements in sleep architecture are easier to produce than durable improvements in the lived experience of insomnia.

Sources

• Efficacy and Safety of Seltorexant in Insomnia Disorder: A Randomized Clinical Trial — JAMA Psychiatry, 2025-08-13
• Association between insomnia symptoms and chronotype—A systematic review and meta-analysis — Sleep Health, 2025-08-07