The present analysis used data from a phase III, randomized, double-blind, placebo-controlled, parallel-group trial of the safety and efficacy of two dose levels of daridorexant (NCT03545191), a dual orexin receptor antagonist that was recently approved for treating insomnia [8, 9]. Full details of the trial are reported elsewhere [10]. Briefly, the trial was conducted between 2018 and 2020 in North America, Europe, and Australia. Subjects were randomly assigned 1:1:1 to receive daridorexant 25 mg, daridorexant 50 mg, or placebo daily for 3 months double-blind. The double-blind treatment period was preceded and followed by run-in and run-out periods of daily single-blind placebo treatment. The run-out period was followed by a 23-day safety follow-up period or continuation to a 9-month extension trial (NCT03679884). Ethical approval was obtained for all study sites.

Adults aged ≥ 18 years with insomnia disorder according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) [11] and an Insomnia Severity Index (ISI) [12] score ≥ 15 at screening were enrolled. Each subject provided written informed consent. Subjects were excluded if they had an acute or unstable psychiatric condition (based on the Mini International Neuropsychiatric Interview [13]), a Mini Mental State Examination [14] score < 25 (subjects aged ≥ 50 years only), or periodic limb movement disorder, restless legs syndrome, circadian rhythm disorder, rapid eye movement sleep behavior disorder, or narcolepsy. Subjects with a history of sleep-related breathing disorders (including chronic obstructive pulmonary disease and sleep apnea) or self-reported suicidal ideation with intent/suicide attempts were also excluded.

Subjects completed the IDSIQ [7] daily in the evening throughout the study using an electronic diary. The IDSIQ comprises 14 items across three domains: Alert/Cognition (six items), Mood (four items), and Sleepiness (four items). Scores range from 0 to 60 for the Alert/Cognition domain, 0 to 40 for the Mood and Sleepiness domains, and 0 to 140 for the total score (see electronic supplementary material [ESM] Fig. 1 for the IDSIQ conceptual framework, and the online supplement of Hudgens et al. for a copy of the IDSIQ questionnaire) [7]. For each IDSIQ item, respondents were asked to indicate the average symptom severity or impact they experienced during the day on an 11-point numeric rating scale (NRS) ranging from 0 (not at all/none at all) to 10 (very/a lot), with a higher score indicating a greater severity or impact. The recall period was ‘today’ (during daytime hours on the day the IDSIQ was completed). Weekly average IDSIQ scores for baseline, month 1, and month 3 were based on the mean of daily entries in the 7 days prior to the visit. The IDSIQ was either completed in full or was marked as missing; programming of the eDiary in which subjects recorded IDSIQ scores did not permit partial completion. The prespecified scoring rule for calculating a weekly average IDSIQ score required subjects to have ≥ 2 days of data in each week; otherwise, the mean score was set to missing. Patients with < 2 days of diary data were omitted from the analysis. These criteria are based on a missing data simulation for the IDSIQ, conducted prior to this analysis, which showed that when the proportion of subjects with only two values for calculation of a weekly mean was varied from 0 to 100%, there was no effect on type I error probability and high power was maintained. Across all simulation scenarios, there was no bias in estimated mean treatment effect or variability, with minimal loss in precision as the number of subjects with two values increased. Weekly average scores for subjects with 2 or more days of data implicitly used mean imputation, where missing data points were given the same value as the mean of the non-missing data points of that same week [10].

The ISI [12] is a 7-item instrument for evaluating the symptoms and functional and emotional impacts of insomnia over the previous month. Each item is scored on a scale from 0 to 4, with a higher score indicating a greater symptom severity or impact. An ISI total score of 15–21 points indicates moderate insomnia, and a score of 22–28 points indicates severe insomnia. A 6-point decrease (improvement) in the ISI total reflects substantial reductions in feeling worn out/fatigued and being unable to think clearly [15, 16]. Subjects completed the ISI at screening, baseline, month 1, and month 3.

Subjects completed additional single-item measures weekly during treatment: the Patient Global Assessment of Disease Severity (PGA-S), Patient Global Impression of Severity (PGI-S), and Patient Global Impression of Change (PGI-C). The PGA-S was used to assess the severity of daytime symptoms and impacts of insomnia over the previous 7 days on a 6-point scale from ‘none’ to ‘very severe’. Similarly, the PGI-S assessed the severity of night-time insomnia symptoms over the previous 7 nights on a 4-point scale from ‘none’ to ‘severe’. Finally, two PGI-C scales were used to capture the severity of night-time insomnia symptoms over the previous 7 nights and daytime symptoms and impacts of insomnia over the previous 7 days, as compared with the week before starting treatment, on a 7-point scale from ‘very much better’ to ‘very much worse’.

PRO data for the weeks corresponding to baseline, month 1, and month 3 were used in the statistical analyses, which were performed using SAS® version 9.4 or later (SAS Institute Inc., Cary, NC, USA). The analyses used the full analysis set (FAS), which included all randomized subjects. Descriptive statistics were calculated for subject demographics and PRO scores.

Meaningful within-patient change for IDSIQ scores was estimated in accordance with US FDA recommendations [5] using anchor-based [17] and distribution-based approaches. As noted by the FDA, distribution-based methods do not directly consider the patient voice and are insufficient for identifying meaningful change [6]. Determination of thresholds for change was guided by the anchor-based approaches. Distribution-based estimates were considered supplemental and provided additional information about variability in the IDSIQ scores. In the anchor-based approach, Spearman correlation coefficients were first calculated for changes in scores for the IDSIQ and for potential anchors: PGA-S, PGI-S, PGI-C (daytime and night-time symptoms), and ISI total score. PRO measures with correlation coefficients ≥ 0.30 [17] were included in a subsequent anchor-based analysis. Mean and median IDSIQ score changes from baseline to month 1 and month 3 with 95% confidence intervals (CIs) were calculated for subjects who met prespecified thresholds defining clinically relevant score changes for the anchors. These clinically relevant score changes were a 1- or 2-point decrease (improvement) from baseline for the PGA-S and PGI-S, ‘a little better’ or ‘moderately better’ compared with the week before starting treatment for the PGI-C; and a 6-point decrease (improvement) from baseline for ISI total score [15, 16].

In the distribution-based analysis, the standard error of measurement (SEM) and standard deviations (SDs) of 0.25, 0.33, and 0.50 of the weekly average scores were calculated at baseline (SD only), month 1, and month 3. SEM [18,19,20] and 0.50 SD [21] have been proposed as useful measures of clinically meaningful change.

To obtain final meaningful within-patient change estimates, values obtained from the anchor-based approach were compared or ‘triangulated’ [17, 22, 23]. In this process, a descriptive non-inferential evaluation of meaningful within-patient change estimates was conducted to identify a similar set of values where the various estimates converged. This set of values was then used to identify a single meaningful within-patient change value.