Cognitive behavioral therapy for insomnia as a suicide prevention strategy: a protocol for a systematic review and meta-analysis

TL;DR

This study aims to evaluate if cognitive behavioral therapy for insomnia can reduce suicidal thoughts and behaviors in adults with insomnia.

Contribution

The study introduces a systematic review and meta-analysis protocol to assess the suicide prevention potential of cognitive behavioral therapy for insomnia.

Findings

Insomnia is an independent risk factor for suicidal ideation and behaviors.

Cognitive behavioral therapy for insomnia is the first-line treatment for chronic insomnia.

The study will quantify the effects of cognitive behavioral therapy on suicide-related outcomes.

Abstract

Insomnia is a prevalent and debilitating sleep–wake disorder, with growing evidence indicating that it is an independent risk factor for suicidal ideation and behaviors. Cognitive behavioral therapy for insomnia is the recommended first-line treatment for chronic insomnia. However, its effect on suicidal ideation and behaviors in those with insomnia has not been well-characterized. Thus, the aim of the planned meta-analysis is to quantify the effects of cognitive behavioral therapy for insomnia on suicidal ideation, suicidal behaviors, and suicide deaths in adults with insomnia with and without comorbidities. This study will include randomized controlled trials comparing cognitive behavioral therapy for insomnia with a control condition with no presumed strong effect on insomnia in individuals with insomnia according to standardized diagnostic criteria or a clinically relevant screening score. Both published and unpublished randomized controlled trials will be retrieved through a systematic search in major databases and trial registries. The primary outcomes include suicidal ideation, suicidal behaviors, and suicide deaths, assessed post-treatment and at follow-ups. We will only consider randomized controlled trials reporting suicide-related outcomes and/or enrolling participants with suicidal ideation or behaviors. For continuous data, a random-effects meta-analysis will be conducted to estimate (standardized) mean differences. In the case of categorical data, a random-effects logistic regression meta-analysis model will be used. The risk of bias of the primary outcomes will be evaluated using the Cochrane Risk of Bias 2 tool. The certainty of evidence will be assessed using the Grading of Recommendations, Assessment, Development and Evaluation framework.

Systematic review registration: PROSPERO-ID CRD420250628820.

Full text

Introduction

Insomnia is a highly prevalent and debilitating sleep–wake disorder, affecting approximately 1 in 10 individuals in the general population [1]. It is associated with reduced health-related quality of life [2], impaired work productivity [3] and an increased risk of developing psychiatric disorders [4]. Accumulating evidence indicates that insomnia is also a significant risk factor for suicidal ideation and behaviors across various age groups [5–10]. Thus, the effective treatment of insomnia is of high public health importance.

Cognitive behavioral therapy for insomnia (CBT-I) is the recommended first-line treatment for chronic insomnia among adults [11]. It is a multicomponent treatment including sleep hygiene education, relaxation strategies, cognitive restructuring, as well as sleep restriction and stimulus control as the core behavioral components. The aim of stimulus control is to strengthen the association between the bed and sleep, while sleep restriction involves restricting the time spent in bed to increase sleep efficiency. Recent meta-analyses have shown that CBT-I is effective for the treatment of insomnia in individuals with comorbid mental health and alcohol use disorders [12–14], where insomnia can exacerbate symptoms and increase the risk of relapse [15, 16]. While CBT-I may reduce comorbid symptom severity [12], there is a paucity of evidence synthesis addressing whether CBT-I may also reduce suicidal ideation and behaviors.

Since 2010, there has been a call to action to evaluate sleep interventions as a strategy for suicide prevention [17]. There has been a slow but steady response to this call, with initial observational studies suggesting that CBT-I may reduce suicidal ideation [18, 19]. Importantly, a recent randomized controlled trial (RCT) has found that insomnia remission may be a key mechanism through which CBT-I alleviates and prevents suicidal ideation [20]. Given recent meta-analytic evidence indicating that CBT-I is more effective than pharmacotherapy in achieving short- and long-term insomnia remission [21], CBT-I may be particularly well-suited for suicide prevention. One potential mechanistic pathway underlying the relationship between insomnia and suicidal phenomena is hyperarousal [22–24], which CBT-I has been hypothesized to reduce [25, 26].

Research on CBT-I as a suicide prevention strategy has recently gained significant momentum, particularly through RCTs contributing to the evidence base [20, 27–29]. However, access to and scalability of CBT-I remain limited due to a paucity of trained providers. Digital formats of CBT-I have emerged as viable alternatives to face-to-face delivery [30], with recent research showing promising results for digital CBT-I in reducing the risk of suicidal ideation [20]. This suggests that CBT-I not only holds promise for suicide prevention but also holds potential for greater scalability and accessibility, which could further strengthen suicide prevention efforts.

Despite these promising developments, uncertainty remains regarding the overall effect of CBT-I on suicidal ideation and behaviors. A recent meta-analysis reported a modest effect of sleep interventions in reducing suicidal ideation. However, these findings were limited by substantial heterogeneity among the relatively few included studies, which may be due to the broad inclusion of various sleep disorders (e.g. nightmares, hypersomnia, insomnia), as well as a mix of pharmacological and behavioral interventions and different age groups (youths and adults) [31]. A preliminary search has also identified relevant newly published trials [32, 33] and an unpublished, completed trial [34] for CBT-I. Building on these recent advances, the planned meta-analysis aims to update the current evidence by incorporating new publications and unpublished data, while narrowing the focus to CBT-I. Specifically, we aim to quantify the effects of CBT-I, compared with control conditions, on suicidal ideation, suicidal behaviors, and suicide deaths among adults with insomnia with and without comorbidities.

Methods

The information in this protocol is reported in accordance with the guidelines of the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) [35, 36]. The completed PRISMA-P checklist is provided in Supplement 1. The study was prospectively registered on PROSPERO (CRD420250628820). In the event of amendments to the study protocol, each amendment will include the date, a description of the change, and the rationale.

Selection criteria

We will include published and unpublished records reporting on RCTs that compare CBT-I in any delivery format to control conditions with no presumed strong effect on insomnia (i.e. placebo, sleep hygiene, self-help, waitlist, no treatment) in parallel-group designs. CBT-I must include stimulus control or sleep restriction as behavioral components, in accordance with findings from a recent network meta-analysis identifying these components as essential for achieving insomnia remission [37]. CBT-I may be combined with other treatments (e.g. specific to a psychiatric disorder), provided that the control group receives the same treatment. We will consider adults (≥18 years) of any gender with insomnia with or without comorbidity according to diagnostic criteria (i.e. the Diagnostic and Statistical Manual of Mental Disorders [38, 39], the International Classification of Sleep Disorders [40] or Diseases [41, 42]), or a validated, clinically meaningful screening score. We will only consider studies reporting suicide-related outcomes and/or enrolling participants with suicidal ideation or behaviors. Disagreements regarding study selection will be resolved by consensus or, if necessary, with the involvement of another member of the review team.

Search strategy

The following databases and registers were systematically searched to identify records published from inception through March 21, 2025: Medline, Embase, Cochrane Library, PsycInfo, and ClinicalTrials.gov. No language restrictions were applied. An RCT filter was applied to all databases except the Cochrane Library. The search will be complemented by screening the reference lists of the included studies and of relevant systematic reviews on this topic [31, 43]. Additionally, forward citation searches (i.e. identifying studies that have cited the included studies and relevant systematic reviews on this topic) will be performed to capture any further eligible records. Search strings and the number of hits for each database/register are provided in Supplement 2.

Primary and secondary outcomes

The primary outcomes include suicidal ideation, suicidal behaviors, and suicide deaths, assessed post-treatment and at follow-up(s). The secondary outcomes include insomnia severity, also measured post-treatment and at follow-up(s), as well as depression severity and hopelessness, which are key correlates of suicidal ideation, behaviors, and suicide deaths [44].

Regarding the primary outcomes, suicidal ideation refers to thoughts of self-harm with the intent to die and includes self-reported or clinician-rated thoughts of suicide, assessed via clinical records, structured interviews, or validated instruments (e.g. Beck Scale for Suicide Ideation [45], Columbia-Suicide Severity Rating Scale [46]). Suicidal behaviors encompass any reported suicide attempts or preparatory behaviors, based on clinical records, structured interviews, or validated instruments (e.g. Columbia-Suicide Severity Rating Scale [46]). Suicide deaths, typically verified through medical records or death registries, will be included when reported. For the secondary outcomes, all assessments should be conducted using validated instruments (e.g. Insomnia Severity Index [47] for insomnia severity).

Based on preliminary searches, suicidal ideation and the secondary outcomes are commonly measured using continuous outcome measures. Thus, continuous data will be prioritized over categorical data in the analyses for these outcomes. Suicidal behaviors and suicide deaths are typically measured as categorical variables, and therefore, categorical data will be prioritized for these outcomes.

Record management and screening

Records will first be imported into EndNote (Clarivate, Version 21, 2023) [48]. and deduplicated using the software’s built-in functions. The deduplicated records will then be exported from EndNote to an Excel spreadsheet for formal screening.

Two reviewers will independently screen the titles and abstracts of all studies and assess whether studies are thematically relevant to the relationship of interest. At this stage, essential eligibility criteria, for which information is typically available in abstracts, will be checked, including study design (RCT), age group (adults), intervention type (CBT-I), and type of control group. Records will initially be coded as clearly irrelevant (excluded prior to full-text review) or potentially relevant (proceeding to full-text review).

At the full-text review stage, potentially relevant records will be comprehensively assessed to ensure that all inclusion/exclusion criteria listed under Selection criteria are met. We will also verify that outcome data are available in the correct format for analysis and ensure that no duplicate studies are included. For unpublished studies, we will also report whether the study is ongoing and note the estimated completion date to determine the feasibility of inclusion in the meta-analysis. Finally, records will be coded as either included or excluded, with reasons documented for each exclusion.

Data extraction

Two reviewers will independently extract data from all included trials for the following variables: authors, year of publication, number of participants in the CBT-I/control groups at each timepoint, age of participants at baseline, percentage of women at baseline, race/ethnicity, number of follow-up assessments, time between baseline and each follow-up, criteria for insomnia, outcome measures, CBT-I components, number of CBT-I sessions, treatment duration, and details regarding control conditions. This information will be summarized narratively in the text and/or presented in tabular form.

Outcome data will be extracted independently by two reviewers using a pre-defined Excel sheet for the meta-analyses. Depending on the frequency of continuous versus categorical data reporting, we may opt to harmonize the data to enhance statistical power and reduce the number of analyses by reaching out to study authors. For instance, if most studies reported continuous data and it is feasible to obtain continuous data from studies that reported categorical data (e.g. those that transformed continuous data into binary outcomes using cut-off scores), we will request these data from the principal investigators.

Principal investigators of both unpublished and published studies will be contacted up to three times, with a 2-week interval between each attempt, to request unpublished data or address other relevant inquiries.

Statistical analysis

For pairwise comparisons informed by ≥3 RCTs, we will perform a meta-analysis if the RCTs are sufficiently homogeneous with respect to design and comparator. For continuous data, mean differences (MDs) or standardized mean differences (SMDs), along with 95% confidence intervals (CIs), between the CBT-I group versus control groups will be calculated for the outcomes post-treatment and at follow-up(s). MDs will be used when all studies measured an outcome on the same scale, while SMDs will be used when multiple scales are used across studies. In the case of SMDs, Hedges’ g will be used as a measure of effect size. We will rely on post- and (follow-up-) means, instead of a change score between baseline and post- (and follow-up-) means. Based on the assumption that there is a distribution of true effect sizes rather than a single true effect size, a random-effects model will be used when pooling the primary studies, in line with the recommendation by Borenstein et al. (2010) [49]. Forest plots will be presented to visualize the results. The degree of heterogeneity between studies will be assessed using I^2^, which describes the proportion of total variation in estimated effect sizes attributable to differences among the studies. An I^2^ value of 50% or higher is commonly considered an indicator of heterogeneity. As an additional measure of heterogeneity and to estimate the range of effect sizes in future studies, corresponding 95% prediction intervals will be calculated [50, 51].

For categorical data (yes/no), odds ratios (ORs) and 95% CIs will be estimated using a random-effects logistic regression model. Suicide-related outcomes, particularly suicide deaths, are expected to be rare (i.e. low counts). For rare events, the Mantel–Haenszel method will be used, which avoids continuity corrections that might bias results [52, 53]. To evaluate the robustness of the results, we will compare the results of the inverse variance model (which assumes a common treatment effect) and those of the Mantel–Haenszel method. If notable discrepancies arise between the methods, only the Mantel–Haenszel method will be used. Forest plots will be used for the visualization of the results, applying a 0.5 continuity correction for studies with zero events in one treatment arm. We will report the I^2^ statistic, along with its 95% CI, as an indicator of heterogeneity for all analyses. Prediction intervals will be calculated as an additional indicator for the degree of heterogeneity and to estimate the effect size range in future studies [50, 51].

All analyses will be conducted in R using the meta package [54]. Although we anticipate conducting meta-analyses for the primary and secondary outcomes, we will provide a systematic narrative synthesis if quantitative synthesis is not appropriate (e.g. due to insufficient data). Information will be presented in the text and in tabular form to summarize and explain the characteristics and findings of the included studies. The narrative synthesis will assess the relationships and results within and between the included studies.

Subgroup analyses

Assuming that sufficient data are available, meta-regression will be used to explore potential subgroup differences based on study characteristics, including age group, psychiatric comorbidities, CBT-I delivery format (e.g. in-person vs. digital), treatment setting (e.g. outpatient vs. inpatient), and whether studies specifically recruited participants with suicidal ideation and behaviors or assessed these outcomes as a secondary endpoint.

Sensitivity analyses

To assess the robustness of all synthesized results, we will conduct sensitivity analyses excluding studies at high overall risk of bias.

Publication bias

Provided that ≥10 studies are included [55], the presence of potential publication bias (or “small-study effects”) will be assessed by examining asymmetry in a contour-enhanced funnel plot [56] using the primary outcomes.

Risk of bias

Two reviewers will independently assess the risk of bias of the primary outcomes in each study using the Cochrane Risk of Bias 2 (RoB 2) tool [57]. The assessment will cover the following domains:

(1) Bias arising from the randomization process,

(2) Bias due to deviations from the intended interventions,

(3) Bias due to missing outcome data,

(4) Bias in measurement of the outcome and

(5) Bias in selection of the reported result.

Each domain will be rated as “low risk of bias”, “some concerns”, or “high risk of bias”. The overall risk of bias of each primary outcome will be determined by the least favorable rating among the domains. The assessments will be managed using the RoB 2 Excel tool. Disagreements will be resolved by consensus or, if necessary, with the involvement of another member of the review team.

Certainty of evidence

Two reviewers will independently assess the certainty of the evidence for the primary outcomes using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach [58]. Based on GRADE guidelines [59], the following factors will be considered: (1) risk of bias (study limitations), (2) inconsistency of results (heterogeneity in the meta-analysis), (3) indirectness (relevance to the research question), (4) imprecision (small sample sizes, wide CIs), and (5) publication bias. The certainty in the body of evidence will be rated as high, moderate, low, or very low. Justifications will be provided for decisions to downgrade or upgrade the certainty of the evidence, as well as for the importance rating of each outcome. Disagreements will be resolved by consensus or, if necessary, with the involvement of another member of the review team.

Discussion

The planned meta-analysis will offer a synthesis of the current evidence on the effects of CBT-I on suicide-related outcomes in individuals with insomnia. Given the strong and growing evidence linking insomnia with suicidal ideation and behaviors [5–10], the findings could have important implications for clinical practice and suicide prevention strategies. Notably, CBT-I in digitally delivered formats (digital CBT-I) offers a scalable, low-cost, and accessible alternative to face-to-face CBT-I [60], making it a promising component of broader public health approaches to suicide prevention. The planned meta-analysis may also play a critical role in identifying gaps in the literature, highlighting important methodological considerations, and informing the design of future studies.

Yet, it is important to acknowledge that RCTs in this specific area of research have only recently begun to emerge [27], and the current evidence base may still be limited in both size and scope. Another important consideration is that studies may differ in whether they recruited participants specifically for suicidal thoughts and behaviors (STBs) or measured STBs as a secondary outcome. These differences may influence the results, as the former may involve higher-risk populations with more severe psychiatric symptoms, while the latter may include less severe cases and thus be more prone to floor effects. To address this, we plan to conduct a meta-regression comparing these study types, provided that a sufficient number of studies are available. If not, this limitation will be clearly acknowledged in the interpretation of the results.

Supplementary Material

CBT-I_suicideprev_Supplement_zpaf070