The scale and rigor of the analysis distinguish it from previous attempts to answer this question. The researchers, working under the umbrella of the SLEEPI project, or Systematic Living Evidence Evaluation of Psychotherapies for Insomnia, searched Medline, Embase, the Cochrane Library, PsycInfo and ClinicalTrials.gov from inception to 23 March 2025, without language restrictions, and supplemented the search with citation chasing and a living evidence database. Crucially, they did not rely solely on published papers. The team contacted the investigators of every eligible trial to request unpublished data, achieving a 100% response rate across 17 inquiries, a step designed to minimize bias from missing results. Of the 1,457 records screened, 19 parallel-group randomized controlled trials met the strict inclusion criteria, which required that CBT-I include at least one of the core behavioral components, stimulus control or sleep restriction, and that control conditions lack a presumed strong effect on insomnia.

The headline result concerns suicidal ideation severity, meaning the frequency, intensity or duration of suicidal thoughts. Across 16 trials contributing to the primary meta-analysis, CBT-I produced a ratio of means of 0.77, with a 95% confidence interval of 0.66 to 0.90 and a P value of 0.0009. In plain terms, average suicidal ideation scores in the treated groups were 23% lower than in control groups after the intervention ended. The heterogeneity statistic, I-squared, was 0%, indicating remarkable consistency across studies, and the 95% prediction interval of 0.65 to 0.91 suggests that a similar effect would be expected in future trials. Sensitivity analyses that excluded trials with a high overall risk of bias, more than 20% dropout, waitlist control groups or single-item measures of suicidal ideation all yielded similar results, and contour-enhanced funnel plots plus Egger’s regression test found no evidence of publication bias.

The picture becomes more complicated when the timeline extends beyond the end of treatment. Follow-up assessments were available in far fewer trials: only four to six trials contributed to each of the 1–3-month, 5–6-month and 12-month follow-up windows, with sample sizes ranging from 384 to roughly 2,500 participants. At none of these timepoints did the difference between CBT-I and control groups reach statistical significance, and the certainty of evidence ranged from very low to moderate. The authors are careful to note that this does not mean the benefit vanishes; it means the existing trials were not designed or powered to answer the durability question reliably. Long-term effects, particularly at the 1–3-month and 12-month marks, remain an open empirical issue.

Suicidal behaviors, by contrast, were vanishingly rare in the evidence base. Across the nine trials that reported on suicide attempts or preparatory behaviors, two participants in CBT-I groups reported a suicide attempt, compared with none in control conditions, and no suicide deaths were recorded in any trial. Five of these trials relied on adverse event reporting rather than structured interviews, a weakness with real consequences: an estimated 50–60% of individuals do not disclose their suicidal thoughts and behaviors to others, so passive reporting may miss incidents entirely. Hopelessness, assessed in only two trials, showed no difference between groups. The rarity of behavioral events means the meta-analysis simply cannot determine whether CBT-I prevents attempts or deaths, a limitation the authors acknowledge explicitly.

The treatment did deliver on its primary clinical targets. Post-intervention, CBT-I reduced insomnia severity by a mean difference of 4.09 points on the Insomnia Severity Index, with a confidence interval of 5.09 to 3.10 points below control, and reduced depression severity with a standardized mean difference of 0.42. Interestingly, meta-regressions found that the size of the depression improvement significantly moderated the size of the suicidal ideation reduction, but this association disappeared after excluding a single terminated trial identified as an outlier. Changes in insomnia severity and the percentage of women in each trial did not moderate the effect. Subgroup analyses showed that the benefit was not significantly moderated by age group, delivery format, population type, or whether trials selectively recruited people with current or recent suicidal ideation, though the authors caution that such trial-level analyses are vulnerable to ecological fallacy, and individual participant data meta-analyses will be needed to clarify whether insomnia remission itself is the active pathway.

Why would fixing sleep touch suicidal thinking? The authors situate the finding within established psychological theories of suicide. Insomnia may act as an upstream driver, feeding feelings of defeat and entrapment described in the integrated motivational–volitional model, or fostering the social disconnection central to the interpersonal theory of suicide. Other candidate mechanisms include depression, nightmares, hyperarousal and the specific effects of being awake during the biological night. That last point has an epidemiological backbone: incident suicide risk has been found to be roughly threefold higher between midnight and 6 a.m. than at any other time of day, after adjusting for population sleep–wake timing. The recent ‘Mind after Midnight’ hypothesis proposes that nighttime wakefulness degrades mood regulation, attention, reward processing and executive function, heightening vulnerability. By stabilizing sleep–wake patterns, CBT-I may interrupt these pathways, a mechanism associated in prior work with reduced psychological distress.

The public health logic of the finding may be its most striking feature. Only about 30% of people experiencing suicidal thoughts, plans or attempts in the past year are estimated to seek or engage with mental health services, with perceived stigma a major barrier. Insomnia carries far less stigma, so offering effective sleep treatment through primary care or digital platforms could reach people who would never present for suicide-focused care. The trials in the review split roughly evenly between therapist-delivered CBT-I and self-guided digital formats, and the meta-analysis found no evidence that the two delivery formats differed in their effects on suicidal ideation severity. Fully automated digital CBT-I, the authors argue, is well suited to earlier preventive stages, where even modest reductions in suicidal ideation across large populations could yield meaningful impact, and recent evidence indicates culturally tailored digital programs improve engagement.

Yet the authors draw a firm ethical line around self-guided formats. In the absence of direct clinical oversight, digital CBT-I is most appropriate when embedded in broader care pathways that include safety monitoring and clear referral routes to crisis services, and it should not stand alone for populations at elevated suicide risk. Engagement itself may falter in groups such as people with severe mental illness, where digital poverty and intervention complexity pose barriers; for settings like psychiatric inpatient units, therapist-delivered CBT-I remains the more feasible choice. The authors also note that CBT-I remains underutilized in routine practice, where sleep medications dominate, and that while some evidence suggests zolpidem may mitigate suicidal ideation in severe insomnia, medications should be used only short term given poor tolerability and unproven long-term efficacy.

The limitations of the evidence base shape what can responsibly be concluded. Only three trials selectively enrolled participants with current or recent suicidal ideation, and individuals at high suicide risk, such as those with a recent attempt, were generally excluded, which may have masked larger effects. Notably, one adapted face-to-face CBT-I trial in psychiatric inpatients, where elevated risk was not an exclusion criterion, produced the largest reduction in suicidal ideation severity of any included study, at 78%, hinting that bigger benefits may be achievable in higher-risk groups. Among trials reporting race and ethnicity, participants were predominantly White, between 52% and 100%, limiting generalizability at a time when ethnic minority and indigenous populations face elevated suicide risk. Suicidal ideation also fluctuates over hours, so assessments at fixed timepoints may miss short-term change, and the authors recommend ecological momentary assessment and evaluation of the suicide crisis syndrome in future work. The bottom line stands on its own terms: high-certainty evidence shows CBT-I modestly reduces suicidal ideation severity after treatment, and larger trials in genuinely high-risk populations are now needed to determine whether that translates into fewer suicidal behaviors.

Subject of Research: Effects of cognitive behavioral therapy for insomnia on suicidal ideation and behaviors in adults

Article Title: Systematic review and meta-analysis of cognitive behavioral therapy for insomnia as a suicide prevention strategy

Article References: Türkmen, C., Schneider, C. L., Furukawa, Y., Saksvik, S. B., Vethe, D., van Dalfsen, J. H., McCall, W. V., Pigeon, W. R., Tubbs, A. S., Perlis, M. L., Riemann, D., Spiegelhalder, K., Türkmen, C., Hara, S., Takashina, H. N., Imai, K., Ishii, A., Ito, M., Kawai, K., … SLEEPI Investigators (2026). Systematic review and meta-analysis of cognitive behavioral therapy for insomnia as a suicide prevention strategy. Nature Mental Health. https://doi.org/10.1038/s44220-026-00743-0

Image Credits: AI Generated

DOI: 10.1038/s44220-026-00743-0

Keywords: cognitive behavioral therapy for insomnia, CBT-I, suicidal ideation, suicide prevention, meta-analysis, randomized controlled trials, insomnia, mental health, sleep interventions, digital therapeutics, evidence certainty, Nature Mental Health