This study is a systematic review and meta-analysis conducted to objectively evaluate the effects of CBT-I on menopausal insomnia by analyzing randomized controlled trials (RCTs). The study was registered with PROSPERO (CRD42024604013) and was reported in accordance with the PRISMA 2020 guidelines [17].

The patients, intervention, comparison, outcome, and study design (PICOS) framework was used to structure the clinical question for this systematic review. The patients (P) were: (1) women aged 40–60 years in the early to late stages of menopause, and (2) those who, by self-report, experienced early, middle, or late insomnia at least three times per week for more than 3 months despite having an adequate sleep environment, resulting in functional impairment; or those with an Insomnia Severity Index (ISI) score ≥7 or a Pittsburgh Sleep Quality Index (PSQI) score ≥5. The interventions (I) involved CBT-I, including components such as stimulus control therapy, sleep restriction therapy, cognitive therapy, and sleep hygiene education. The comparison (C) group consisted of participants who received no intervention or routine care, such as general menopausal care, menopause education, or sleep hygiene education. The outcomes (O) included sleep quality as the primary outcome and insomnia severity as the secondary outcome. The study design (S) was limited to RCTs.

A comprehensive literature search and analysis were conducted for studies published up to October 2024. The international databases searched included PubMed, EMBASE, the Cochrane Central Register of Controlled Trials (CENTRAL), and CINAHL. To include East Asian studies, regional databases from China and Korea were also searched. The Chinese databases included the China National Knowledge Infrastructure (CNKI), VIP (Chinese Science and Technology Journals Database), and Wanfang (Chinese academic journals and theses). The Korean databases included RISS (academic research portal), NDSL (science and technology literature), DBpia (journal articles and theses), KISS (multidisciplinary academic literature), and ScienceON (science and technology database). Both peer-reviewed journal articles and academic theses were retrieved. Studies published in English, Korean, and Chinese were included to ensure comprehensive coverage, as research on menopausal insomnia and CBT-I has been actively conducted in these languages. To enhance search sensitivity, gray literature—including theses, reports, and conference presentations—was manually searched using Google Scholar in addition to the electronic databases. Search terms were developed using a combination of MeSH terms and free-text keywords, applying Boolean operators (AND/OR) and truncation where appropriate. For Participants (P), the keywords included “Insomnia,” “Sleep,” “Disorder,” “Problem,” “Menopause,”* and “Middle age.” For Interventions (I), the terms “Cognitive behavioral therapy [MeSH],” “Cognitive behavior therapy,” and “CBT” were used. For Outcomes (O), the keywords “Sleep quality,” “PSQI,” “Pittsburgh Sleep Quality Index,” “VSH,” “Verran and Snyder-Halpern,” “ISI,” and “Insomnia Severity Index” were applied. The final search string was constructed using the following structure: Participants: ((Insomnia) OR ((Sleep) AND ((Disorder) OR (Problem)))) AND ((Menopause*) OR (Middle age)), Intervention: (Cognitive behavioral therapy) OR (Cognitive behavior therapy) OR (CBT), Outcome: (Sleep quality) OR (PSQI) OR (Pittsburgh Sleep Quality Index) OR (VSH) OR (Verran and Snyder‐Halpern) OR (ISI) OR (Insomnia Severity Index). Advanced search options were used where available to optimize search sensitivity and specificity according to the features of each database (Supplementary Table 1).

The inclusion criteria for this review were: (1) intervention studies examining the effects of CBT-I on menopausal insomnia; (2) RCTs; (3) studies in which the control group either received no intervention or only general education, allowing the specific effect of CBT-I to be isolated; and (4) studies where the intervention provided to the control group was also applied to the experimental group, ensuring comparability between conditions. The exclusion criteria were: (1) preclinical or animal studies and studies involving children; (2) studies not published in Korean, English, or Chinese; (3) observational studies or review articles; (4) studies incorporating additional pharmacological interventions beyond participants’ usual medications; and (5) studies in which the interventions for the experimental and control groups were not equivalent, making it impossible to isolate the specific effects of CBT-I. No disagreements occurred between the researchers during study selection.

Studies were selected based on the core research question and the predefined inclusion and exclusion criteria. The PRISMA 2020 flow diagram (Figure 1) presents the detailed study selection process. A total of 1,720 records were initially retrieved through the search strategy. After removing 488 duplicates, 1,232 articles were screened by two researchers (authors HJM and MHH) according to the inclusion and exclusion criteria. As a result, 1,195 articles that did not meet the criteria were excluded, and 36 studies were initially selected for full-text review. Of these, eight were excluded because they were unpublished, available only as abstracts, or lacked sufficient methodological detail. The remaining 29 articles were re-evaluated by two researchers using the same criteria. Among these, 18 were excluded because they were not RCTs, did not apply a single intervention to the experimental group, or did not apply a uniform control condition to the comparison group. Ultimately, 11 studies [3,8-25] were included in the qualitative synthesis. Of these, nine studies were incorporated into the meta-analysis, as they employed consistent intervention methods and outcome variables, clearly reported sample sizes for both groups, and provided post-intervention outcome data suitable for quantitative analysis.

The methodological quality of the included RCTs was evaluated using the Risk of Bias 2.0 (RoB 2.0) tool developed by the Cochrane Collaboration [26]. This updated instrument assesses the potential for bias across five core domains that may influence study outcomes: (1) bias arising from the randomization process, (2) bias due to deviations from intended interventions, (3) bias due to missing outcome data, (4) bias in the measurement of outcomes, and (5) bias in the selection of reported results.

Each domain includes a structured set of signaling questions, with responses categorized as “yes,” “probably yes,” “probably no,” “no,” or “no information.” An embedded algorithm interprets these responses to classify the risk of bias for each domain as low risk, some concerns, or high risk. The overall risk of bias for each study was determined by aggregating judgments across the five domains: studies rated as low risk in all domains were classified as having a “low overall risk”; studies with at least one domain rated as “some concerns” were categorized accordingly; and those with one or more domains rated as “high risk” were classified as having a “high overall risk of bias” [27].

Reviewers (authors HJM, SNY, and MHH) independently assessed each study using the RoB 2.0 tool. Any discrepancies between reviewers were resolved through discussion and consensus. This process ensured methodological rigor and reliability in quality assessment. The results of the evaluations were visualized using the robvis tool (http://mchuinlu.shinyapps.io/robvis/).

The characteristics of the studies included in this systematic review were analyzed, and data were extracted based on the following variables: first author, publication year, country, sample size, participant age, type of intervention, type of control intervention, primary outcome variables, and the authors’ conclusions (Table 1).

Given the observed heterogeneity among studies in intervention methods, duration, and insomnia severity, a random-effects model was employed to account for both within- and between-study variance.

Data were analyzed using Review Manager (RevMan) version 5.4 (Cochrane Collaboration). As the outcomes reported in the included studies were continuous variables, effect sizes were calculated using the number of participants (n), mean values, and standard deviations for both intervention and control groups. For insomnia severity, all studies utilized a consistent measurement instrument—the ISI—allowing for the calculation of mean differences (MDs) between groups. In contrast, sleep quality was assessed using various validated instruments, including the PSQI, Groningen Sleep Quality Scale, and others; therefore, standardized mean differences (SMDs) were calculated to synthesize these outcomes.

Among the 11 included studies, approximately half were conducted in the United States [3,14,15,18-25], while the others were conducted in Egypt [15,18], China [3,21], Iran [19], and South Korea [20]. About half of the studies were published before 2020 [22-25], with the remainder published thereafter. As presented in Table 2, face-to-face CBT interventions were predominant, being applied in eight studies [3,14,18-23], whereas two studies employed remote delivery [15,25] and one study utilized both delivery modes [24]. The number of CBT sessions ranged from 4 to 12, with most delivered once per week. Session duration varied from 20 to 120 minutes, with 50–60 minutes being the most common; four studies [3,14,21,23] did not report session length. Follow-up periods ranged from immediate post-intervention to 52 weeks.

As shown in Figure 2, the risk of bias arising from the randomization process was rated as low in nine studies (81.8%) [14,15,18-20,22-25] and as having some concerns in two studies (18.2%) [3,21]. For bias due to missing outcome data, eight studies (72.7%) [3,14,15,18,19,22,24,25] were rated as low risk, while three studies (27.3%) [20,21,24] were assessed as high risk. Bias related to deviations from intended interventions, outcome measurement, and selection of reported results was rated as low risk across all included studies.

Overall, the methodological quality of the included studies was considered acceptable, with most domains demonstrating a low risk of bias. However, the presence of high risk in the domain of missing outcome data in three studies suggests that some caution is warranted when interpreting the pooled findings. Despite this, the consistency of results across studies with low risk supports the robustness of the meta-analytic conclusions.