Effectiveness of app-based cognitive behavioral therapy for insomnia on preventing major depressive disorder in youth with insomnia and subclinical depression: A randomized clinical trial

Youth experience high incidence of mental disorders, with major depressive disorder (MDD) common and associated with long-term impairments. Insomnia is a modifiable, independent risk factor for depression in youth and adults. While CBT-I reduces depressive symptoms, few adequately powered studies have tested whether CBT-I prevents new-onset depressive disorder, especially in youth and in digital formats. This study tested whether a fully automated app-based CBT-I, compared with an attention-matched health education control, reduces the 12-month incidence of MDD in youth with insomnia disorder and subclinical depression, improves depressive and insomnia symptoms, increases remission of insomnia disorder, and whether insomnia improvements mediate depressive symptom changes.

Prior work shows insomnia predicts later depression (meta-analyses) and CBT-I improves insomnia and reduces depressive symptoms in adolescents and adults. However, adult prevention trials with digital CBT-I reduced depressive symptoms but did not prevent depressive disorder, whereas face-to-face CBT-I prevented incident and recurrent MDD in older adults. Digital CBT-I in youth has been less studied, often in Western samples and with modest sizes. There is a need to determine if digital insomnia treatment can prevent depressive disorder in youth with subclinical depression and insomnia, and to clarify mechanisms (e.g., insomnia as mediator).

Design: Randomized, assessor-blind, parallel-group controlled trial in Chinese youth aged 15–25 years with insomnia disorder (ICD-10) and subclinical depression (PHQ-9 >4 and <20). Recruitment: Universities, high schools, communities in mainland China and Hong Kong (Sept 9, 2019–Nov 25, 2022). Screening: Online ISI ≥15 and PHQ-9 criteria; diagnostic telephone MINI to confirm insomnia disorder and exclude current MDD or MDD within past 2 months. Ethics: Approved by CUHK-NTEC CREC (2019.044) and Peking University Sixth Hospital (Issue No. 21[2019]); informed consent/assent obtained. Registration: NCT04069247. Randomization and masking: 1:1 allocation to 6-week app-based CBT-I or 6-week app-based health education (HE), stratified by sex and insomnia severity (ISI 15–21 vs >21), block size 4. Allocation concealed from assessors; participants likely aware due to content. Telephone interviewers were masked; interrater reliability for MDD diagnosis κ=0.82 (12% sample). Interventions: CBT-I (esleep app) with six weekly modules (20–30 min each): (1) sleep overview, (2) sleep restriction with individualized weekly “sleep prescription” (5-h minimum; adjusted by sleep efficiency), (3) stimulus control, (4) cognitive restructuring of sleep-related cognitions, (5) structured worry time, (6) relapse prevention. Modules unlocked weekly; completion allowed within 12 weeks. Weekly individualized text reminders. Control: Attention-matched app-based HE including general sleep knowledge without active therapeutic components; weekly reminders. Assessments: Baseline, post-intervention, 6 and 12 months; plus post-session 2 and 4 symptom checks. Primary outcome: time to onset of MDD via MINI covering interval since last assessment. Secondary outcomes: PHQ-9/PHQ-8 (depressive symptoms), ISI (insomnia severity), remission of insomnia disorder (no impairment/distress, frequency <3/week ≥1 month, sleep med <1/week). Other prespecified outcomes: GAD-7, suicidal ideation (SSI-Current), multidimensional fatigue, reduced Morningness-Eveningness Questionnaire, DBAS-16, and 7-day sleep diary (TIB, TST, SOL, WASO, SE). Adverse events queried at each interview. Sample size: Assumed 1-year MDD incidence 16.5% in this mixed-history cohort, HR=0.5, α=0.05, power=80% → 282/arm; planned 940 allowing 40% attrition; amended to 708 (354/arm) given lower observed attrition (~20%). Statistical analysis: Intent-to-treat. Kaplan–Meier and Cox models for time to MDD (strata covariates included). GEE with weights for insomnia remission. Linear mixed-effects models (REML) for continuous outcomes; sensitivity analyses with LOCF for those reaching endpoint before 12 months. Mediation via SEM to test insomnia as mediator of depressive symptom change. Exploratory Cox models additionally adjusted for prior MDD, education, income, comorbid medical illnesses, and sleep-promoting medication use. Two-tailed α=0.05; Stata 17.0.

- Participants: 708 randomized (CBT-I n=354; HE n=354); 57% female; mean age 22.1 (SD 1.9) years; 48% with prior MDD; 20% using sleep-promoting meds at baseline. Follow-up completion: post 90% (634/708), 6-month 80% (565/708), 12-month 71% (500/708); no group differences. Engagement: ≥1 session login ~94% both groups; completed all 6 sessions 84% (CBT-I) and 85% (HE). No intervention-related adverse events. - Primary outcome (MDD incidence over 12 months): 10% (37/354) CBT-I vs 18% (62/354) HE; HR 0.58 (95% CI 0.38–0.87), p=0.008. Cumulative 1-year incidence: 12% (9%–17%) CBT-I vs 21% (17%–27%) HE. NNT at 1 year = 10.9 (6.8–26.6). Adjusted exploratory Cox (adding prior MDD, education, income, medical comorbidity, sleep meds): HR 0.60 (0.40–0.90), p=0.01. Effects consistent across subgroups and in those with persistent insomnia (≥3 months). - Insomnia remission (clinical interview): Post-intervention 52% CBT-I vs 28% HE; RR 1.83 (1.49–2.24), p<0.001. At 6 months: 56% vs 44%; RR 1.27 (1.08–1.48), p=0.003. At 12 months: 57% vs 48%; RR 1.17 (1.01–1.35), p=0.03. - Symptom severity: Mixed-effects models showed greater reductions with CBT-I. • PHQ-9: Post-session 4 adjusted difference –1.0 (–1.6 to –0.4), d=0.50, p=0.002; post-intervention –1.0 (–1.6 to –0.5), d=0.53, p<0.001; 6 months –0.8 (–1.4 to –0.2), d=0.42, p=0.009; 12 months –0.5 (–1.1 to 0.2), p=0.14. PHQ-8 results similar. • ISI: Post-session 4 –1.7 (–2.4 to –1.0), d=0.66, p<0.001; post-intervention –2.0 (–2.7 to –1.3), d=0.78, p<0.001; 6 months –1.1 (–1.9 to –0.4), d=0.45, p=0.002; 12 months –0.8 (–1.6 to –0.1), d=0.32, p=0.03. • Sensitivity (LOCF): sustained between-group differences through 12 months for both PHQ and ISI. - Mediation: Insomnia symptom improvement mediated 74% (post-session 4 → PHQ-8 at post) and 78% (post → PHQ-8 at 6 months) of depression improvement; no significant residual direct effects. Reverse models indicated direct effects of intervention on insomnia independent of depression (81% at post; 57% at 6 months). - Other outcomes: CBT-I improved SOL, TIB, SE, anxiety (GAD-7), fatigue, dysfunctional sleep beliefs, and morningness shift at post-intervention, with maintenance to at least 6 months (anxiety effects not sustained). No significant between-group differences in WASO, TST, or suicidal ideation.

The trial demonstrates that targeting insomnia with a fully automated app-based CBT-I can prevent incident MDD in youth at risk due to insomnia and subclinical depression. The reduction in MDD incidence (HR ~0.58) aligns with prevention magnitudes seen in face-to-face CBT-I in older adults and web-based preventive interventions in adults, extending evidence to a youth, largely non-Western cohort. Mediation findings support a causal pathway wherein improvements in insomnia drive reductions in depressive symptoms, although shared mechanisms (hyperarousal, rumination, worry) may also contribute. Clinically, app-based CBT-I yielded robust and sustained benefits on insomnia severity and remission, and improved multiple sleep and daytime functioning metrics, while depressive symptom advantages persisted to 6 months. The relatively low attrition and high adherence suggest feasibility and acceptability in digitally native youth, and weekly reminders may enhance engagement. Integrating digital CBT-I into primary and student health services could improve access and support depression prevention at scale.

App-based CBT-I effectively prevents onset of major depression and improves insomnia outcomes in youth with insomnia and subclinical depression. The findings support incorporating digital CBT-I into clinical and community practice to enhance sleep and mental health in youth. Future research should evaluate implementation in routine care, assess booster strategies to maintain long-term gains, consider integrating circadian-focused components, and stratify by prior depression to differentiate prevention of first-onset vs recurrent MDD.

- Mixed sample with and without prior MDD history (48% with prior MDD), which may affect generalizability; stratified randomization by prior history was not performed. - MINI diagnostic interviews were conducted by telephone rather than face-to-face; although reliable in research, this may differ from in-person assessments. - Potential partial unmasking of assessors due to participants’ responses could introduce bias. - Control condition included general sleep knowledge; thus, it was not a completely inactive control and showed improvements, potentially reducing between-group differences. - Small adolescent subsample (<18 years; n=10) limits generalizability to younger adolescents.