Therapeutic interventions for adolescent suicidal behavior and self-harm: an umbrella review for developing clinical practice guidelines in South Korea Authors: Sangha Lee, Jihye Song, Sooyeon Suh, Yunmi Shin Published: 2026 ABSTRACT AbstractBackgroundSouth Korea has the highest suicide rate among member countries of the Organisation for Economic Co-operation and Development, reaching 29.1 per 100,000 people in 2024, alongside a concurrent increase in the prevalence of adolescent non-suicidal self-injury (NSSI), estimated at 21.2%–24%, substantially exceeding the global average. Despite this public health crisis, evidence-based clinical practice guidelines specifically addressing adolescent suicidal behavior and NSSI in the Korean context remain lacking. This umbrella review synthesizes evidence on therapeutic interventions to inform the development of clinical practice guidelines.MethodsA systematic search was conducted across five databases (PubMed, Embase, CINAHL, Cochrane CENTRAL, Scopus) for systematic reviews published between January 2015 and December 2024. Eligibility was assessed using a predefined Population, Intervention, Comparison, Outcome framework targeting adolescents and young adults aged 10–24 years. Methodological quality was appraised using A Measurement Tool to Assess Systematic Reviews 2 (AMSTAR-2), and the certainty of evidence was graded using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework. Both quantitative and narrative systematic reviews were included.ResultsSix systematic reviews (published between 2018 and 2022) encompassing 178 unique primary studies met the eligibility criteria. AMSTAR-2 ratings identified one review as high quality, three as moderate quality, and two as critically low quality. High-certainty evidence supported combination therapy (fluoxetine plus cognitive behavioral therapy (CBT)) over fluoxetine monotherapy, demonstrating significantly improved treatment response rates (RR = 1.12) and reduced one-year recurrence rates (RR = 0.27). For psychotherapy as a stand-alone intervention, high-certainty evidence demonstrated a small effect on self-harm reduction (d = 0.13), and moderate-certainty evidence supported a reduction in suicidal ideation (d = 0.31). However, evidence for widely adopted models—including dialectical behavior therapy for adolescents, mentalization-based treatment for adolescents, attachment-based family therapy, and multisystemic therapy—was uniformly rated as very low certainty, revealing a substantial evidence–practice gap. A substantial proportion of the included reviews did not distinguish between NSSI and suicidal behavior in outcome reporting, limiting differential conclusions.ConclusionsThis umbrella review identified high-certainty evidence supporting fluoxetine plus CBT combination therapy as a first-line recommendation for adolescents with comorbid depression and suicidal risk. The evidence–practice gap for specific psychotherapy models underscores the need for methodologically rigorous primary research. Culturally adapted guidelines incorporating Korea-specific risk factors—including academic pressure, Confucian family values, and cyberbullying—are urgently needed to address the unique epidemiological and sociocultural context of Korean adolescents. This review was prospectively registered in INPLASY(registration number: INPLASY202630082, doi: 10.37766/inplasy2026.3.0082).Clinical trial numberNot applicable.Keywords: Adolescent suicide; Non-suicidal self-injury; Umbrella review; Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework; Clinical practice guidelines; South Korea Introduction Suicide is the second leading cause of death among young people aged 10–24 years globally, and its incidence is steadily increasing [1]. Between 2007 and 2019, suicide rates among adolescents and young adults increased markedly, and the COVID-19 pandemic further exacerbated this trend, with emergency department visits for suicide attempts rising substantially among youth [2]. These trends underscore that adolescent suicide prevention is a critical global public health priority. The situation is particularly concerning in South Korea, which has consistently recorded the highest suicide rate among member countries of the Organisation for Economic Co-operation and Development (OECD) since 2004 [3]. In 2024, the suicide rate reached 29.1 per 100,000 people—approximately 2.7 times the OECD average of 10.8—with a total of 14,872 suicide deaths, the highest figure since 2011 [4]. In contrast to the global downward trend in youth suicide rates, South Korea has experienced an increase in youth suicide. The suicide rate among those aged 10–14 years increased from 0.8 per 100,000 in 2000 to 1.9 in 2019, and among those aged 15–19 years, from 6.4 to 9.9 during the same period [5]. Of particular concern is the nearly threefold increase in the suicide rate among adolescent girls aged 10–14 years, from 1.2 in 2017 to 3.2 in 2022 [5]. Suicide is the leading cause of death among individuals aged 10–39 years and has remained the leading cause of death among those aged 10–24 years since 2011 [3, 4]. According to the Korea Youth Risk Behavior Survey, more than 10% of Korean adolescents reported seriously considering suicide within the past year, and suicide accounted for 43.7% of deaths among those aged 10–19 years [6]. In addition to suicide, non-suicidal self-injury (NSSI) has emerged as a major concern among adolescents. NSSI refers to deliberate self-harm without suicidal intent, and the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR) lists NSSI disorder as a condition for further study [7]. Globally, the prevalence of NSSI among adolescents is estimated at approximately 17.2%, whereas reported prevalence in South Korea ranges from 21.2% to 24%, substantially exceeding the global average [8, 9]. Although NSSI is defined as occurring without suicidal intent, it is classified as a high-risk behavior because it may increase suicide risk and lead to fatal outcomes. NSSI is also recognized as a significant predictor of subsequent suicide attempts [10]. Current state of the literature and research gaps The number of systematic reviews and meta-analyses examining the effectiveness of therapeutic interventions for adolescent suicide and self-harm has increased considerably. Various psychotherapeutic approaches have been investigated, including cognitive behavioral therapy (CBT), dialectical behavior therapy (DBT), interpersonal psychotherapy (IPT), family-based treatments, emotion regulation interventions, acceptance and commitment therapy (ACT), and multilevel school-based programs. Evidence regarding the efficacy and safety of pharmacological interventions has also accumulated [11, 12]. However, existing systematic reviews have predominantly focused on evaluating the effectiveness of specific treatment modalities in isolation—for example, CBT or DBT alone—thereby limiting comprehensive cross-modality comparisons. Previous umbrella reviews share similar limitations. Torralba-Suárez et al. (2023) conducted an umbrella review restricted to CBT and DBT for adolescent self-harm and suicidal behavior, excluding other psychotherapeutic modalities such as family therapy, IPT, and mentalization-based treatment for adolescents (MBT-A) [13]. Richardson et al. (2024) and McEvoy et al. (2023) performed umbrella reviews primarily focused on risk and protective factors, without directly comparing the effectiveness of treatment interventions [14, 15]. Furthermore, numerous existing reviews have analyzed suicidal behaviors and NSSI together, making it difficult to distinguish the differential effects of interventions on each outcome. Most critically, there is a notable lack of evidence synthesis to support the development of clinical practice guidelines that account for the unique epidemiological characteristics and sociocultural context of the Korean adolescent population. Despite having the highest suicide rate among OECD countries, comprehensive evidence-based guidelines specifically addressing adolescent suicide and NSSI remain lacking. Although international guidelines from the American Psychiatric Association (APA) [16], American Academy of Child and Adolescent Psychiatry (AACAP) [17], American Academy of Pediatrics (AAP) [18], Department of Veterans Affairs/Department of Defense (VA/DoD) [19], and Substance Abuse and Mental Health Services Administration (SAMHSA) [20] provide valuable frameworks, they were developed within their respective health care systems and cultural contexts, which may limit their direct applicability to Korean clinical settings. Study objectives To address the abovementioned gaps, the present study conducted an umbrella review to comprehensively evaluate the effectiveness of therapeutic interventions for suicidal behavior (suicidal ideation and suicide attempts) and NSSI in adolescents and young adults aged 10–24 years. An umbrella review is a research methodology that systematically synthesizes findings from existing systematic reviews and meta-analyses to provide the highest level of evidence on a given topic [21]. The specific objectives of this study were as follows: First, to comprehensively evaluate the effectiveness of psychotherapeutic interventions (including CBT, DBT, IPT, family-based therapy, MBT-A, ACT, and multilevel school-based interventions) for adolescent suicidal behavior and self-harm; second, to examine the efficacy and safety of pharmacological treatments (including antidepressants, antipsychotics, and mood stabilizers); third, to analyze the comparative effectiveness of psychotherapy versus pharmacotherapy; fourth, to evaluate the effectiveness of combined interventions (psychotherapy plus pharmacotherapy); and fifth, to establish an evidence base for the development of clinical practice guidelines applicable to the Korean adolescent population. Unlike previous umbrella reviews restricted to specific modalities, such as CBT and DBT, the present review encompasses the full spectrum of psychotherapeutic and pharmacological interventions, including combination treatments. Furthermore, it distinguishes between suicidal behavior and NSSI as separate outcomes and evaluates the comparative effectiveness of psychotherapy versus pharmacotherapy. By systematically applying assessments based on the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework to both quantitative and narrative reviews, this study aims to provide a transparent and rigorous evidence base for the development of clinical practice guidelines for adolescent suicide and self-harm in South Korea. Methods Study design and registration This study was conducted as an umbrella review following the Joanna Briggs Institute (JBI) methodology for umbrella reviews [22] and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [23]. The reporting of this umbrella review also adhered to the recommendations of the Preferred Reporting Items for an Overview of Systematic Reviews (PRIO-harms) tool [24]. The review protocol was developed a priori and guided all stages of the review process. PICO framework The research question was formulated using the PICO (Population, Intervention, Comparison, Outcome) framework as follows: Population (P): Adolescents and young adults aged 10–24 years who have experienced or are at elevated risk of suicidal behavior (suicidal ideation and suicide attempts) or NSSI. Intervention (I): Broadly defined psychotherapeutic interventions, including but not limited to CBT, DBT, IPT, family-based therapy, MBT-A, ACT, emotion regulation interventions, multisystemic therapy (MST), attachment-based family therapy (ABFT), and multilevel school-based interventions. Comparison (C): (1) Pharmacological treatments (e.g., antidepressants, antipsychotics, and mood stabilizers); (2) no intervention, waitlist control, or treatment as usual (TAU); (3) other psychotherapies; and (4) direct head-to-head comparisons between psychotherapeutic modalities. Outcome (O): Primary outcomes included reductions in suicide attempts, suicidal ideation, and the frequency and severity of NSSI. Secondary outcomes included treatment dropout rates, relapse rates, duration of treatment response maintenance, quality of life, and changes in depressive and anxiety symptoms. Eligibility criteria Inclusion criteria Studies were eligible for inclusion if they met all of the following criteria: (a) the study was a systematic review, with or without a meta-analysis; (b) the target population included adolescents or young adults (age range overlapping with 10–24 years); (c) the review examined the effectiveness of any therapeutic intervention (psychotherapeutic, pharmacological, or combined) for suicidal behavior or NSSI; (d) the primary outcomes included at least one of the following: suicidal ideation, suicide attempts, or NSSI; and (e) the study was published in English between January 2015 and December 2024. Exclusion criteria Studies were excluded if they met any of the following criteria: (a) the study was not a systematic review (e.g., a narrative review without a systematic search, scoping review, or primary study); (b) the study focused exclusively on adult populations (> 24 years) without reporting disaggregated data for the adolescent or young adult age range; (c) the study examined only risk factors, protective factors, or screening tools without evaluating treatment effectiveness; (d) the study addressed NSSI as the sole outcome without any assessment of suicidal behavior, unless the intervention was explicitly designed to reduce NSSI; (e) the publication was a protocol, letter, commentary, or conference abstract; and (f) the publication was a case report. Information sources and search strategy Databases A systematic literature search was conducted across five electronic databases: PubMed, Embase, CINAHL, Cochrane CENTRAL, Scopus. The search covered the period from January 2015 to December 2024. Search strategy Two complementary search strategies were developed and combined using Boolean operators. Strategy 1 was designed to identify experimental and interventional studies, including systematic reviews and meta-analyses, addressing adolescent suicide. This strategy combined terms related to suicide (e.g., “suicide,” “suicidal ideation,” “suicide attempt,” “self-harm,” and “self-injury”) AND adolescents (e.g., “adolescent,” “youth,” “young adult,” “teen,” and “child”) AND study design or intervention filters (e.g., “clinical trial,” “randomized controlled trial,” “RCT,” “intervention,” “systematic review,” “meta-analysis,” “therapy,” “medication,” and “treatment”), while excluding observational study designs. Strategy 2 was designed to identify interventions shown to be effective for adolescent suicide using a broader combination of suicide-related terms AND adolescent-related terms AND effectiveness-related terms (e.g., “effectiveness,” “efficacy,” “effect,” and “outcome”). Strategy 3 combined the results of Strategies 1 and 2 using the OR operator to maximize sensitivity. Search terms were adapted to the controlled vocabulary and syntax requirements of each database (e.g., MeSH terms for PubMed, Emtree terms for EMBASE, and Subject Headings for CINAHL). The full electronic search strategy for each database is provided in Supplementary Material 1. Screening process A preliminary search in PubMed yielded 8,110 records. After expansion to all five databases, the combined results exceeded 30,000 records. The search strategy was subsequently refined, yielding a total of 4,596 unique records after deduplication using EndNote. Records were then classified by publication type using EndNote’s Smart Group function: records containing review-related terms (e.g., “review,” “literature review,” “narrative review”) in titles, keywords, abstracts, or notes were flagged as review articles, while systematic reviews, scoping reviews, and Cochrane records were retained. Case reports were similarly identified and excluded. This process removed 853 records, reducing the candidate pool to 3,743 records (2,225 original articles and 386 review articles). A total of 386 review articles formed the basis for umbrella review screening. Title and abstract screening of these 386 records was performed independently by two reviewers, followed by full-text assessment, with disagreements resolved through consensus discussion. Title and abstract screening were performed independently by two reviewers according to the predefined inclusion and exclusion criteria. Subsequently, full-text review was conducted to identify potentially eligible records. Disagreements at any stage were resolved through discussion and consultation with a third reviewer when necessary. The study selection process is presented in the PRISMA flow diagram in Fig. 1. PRISMA 2020 flow diagram illustrating the study selection process for the umbrella review Data extraction Data were extracted from each included systematic review using a standardized data extraction form. The following information was collected: Study/publication characteristics: first author, publication year, journal, country of origin, review type (systematic review with or without meta-analysis), number and design of included primary studies, total number of participants, databases searched and search period, and registration information. Population characteristics: age range, clinical characteristics (e.g., suicidal ideation, prior suicide attempts, NSSI, and psychiatric comorbidities), and setting (e.g., outpatient, inpatient, community, and school-based). Intervention and comparator details: types of interventions examined, specific treatment protocols, duration and frequency of treatment, and types of comparators. Outcome data: for systematic reviews with meta-analyses—pooled effect estimates (e.g., odds ratios, risk ratios, standardized mean differences, and Cohen’s d), 95% confidence intervals, heterogeneity statistics (I2, Q statistic, and p-values), and results of subgroup or sensitivity analyses; for systematic reviews without meta-analysis—direction and consistency of effects across included studies and key qualitative findings as reported by the review authors. Quality assessment reported in the review: any risk-of-bias assessment or quality appraisal conducted by the original review authors (e.g., Cochrane Risk of Bias tool and Jadad scale), and any GRADE assessment performed by the original authors. Data were simultaneously and independently extracted by two reviewers, with each reviewer’s extraction cross-checked by the other; discrepancies were resolved through consensus discussion. Assessment of primary study overlap To assess the extent of overlap among primary studies across included reviews, we calculated the Corrected Covered Area (CCA) using the method described by Pieper et al. (2014) [25]. The CCA quantifies the proportion of primary studies shared across reviews, corrected for the number of reviews included, and is interpreted as follows: slight (< 0.05), moderate (0.06–0.10), high (0.11–0.15), and very high (> 0.15). A citation overlap matrix was constructed to identify shared primary studies between each pair of reviews. Quality assessment of included reviews A measurement tool to assess systematic reviews 2 The methodological quality of each included systematic review was assessed using A Measurement Tool to Assess Systematic Reviews 2 (AMSTAR-2) [26]. AMSTAR-2 comprises 16 items that evaluate key methodological domains, including whether the research questions and inclusion criteria were established prior to conducting the review, the adequacy of the literature search, justification for excluding studies, risk-of-bias assessment of included studies, appropriateness of meta-analytic methods (where applicable), consideration of risk of bias when interpreting results, assessment of publication bias, and disclosure of conflicts of interest. Each item was rated as “Yes,” “Partial Yes,” or “No.” Based on the assessment, each review was assigned an overall confidence rating of high, moderate, low, or critically low, in accordance with the AMSTAR-2 algorithm. The seven critical domains (Items 2, 4, 7, 9, 11, 13, and 15) were given particular weight in determining the overall rating [26]. Two reviewers independently performed the AMSTAR-2 assessment, and discrepancies were resolved by consensus. Evidence grading The certainty of the evidence for each outcome was appraised using the GRADE framework [27]. Evidence from systematic reviews of randomized controlled trials (RCTs) started at a high certainty rating and could be downgraded based on the following five domains: Risk of bias was assessed based on the methodological limitations of the primary studies included in each review. For subjective outcomes (e.g., self-reported suicidal ideation and self-harm behaviors), the adequacy of blinding was given particular consideration, as lack of blinding in psychotherapy trials may introduce performance and detection bias. For objective outcomes (e.g., mortality and emergency department visits), blinding was considered less critical. Ratings were classified as “not serious,” “serious” (− 1), or “very serious” (− 2). Risk of bias was assessed based on the methodological limitations of the primary studies included in each review. For subjective outcomes (e.g., self-reported suicidal ideation and self-harm behaviors), the adequacy of blinding was given particular consideration, as lack of blinding in psychotherapy trials may introduce performance and detection bias. For objective outcomes (e.g., mortality and emergency department visits), blinding was considered less critical. Ratings were classified as “not serious,” “serious” (− 1), or “very serious” (− 2). Inconsistency was evaluated based on the variability of effect estimates across studies. For reviews with meta-analyses, statistical heterogeneity was assessed using the I² statistic, with thresholds of I² > 50% considered “serious” and I² > 75% considered “very serious” inconsistency. For reviews without a meta-analysis, inconsistency was assessed qualitatively by examining the direction and magnitude of effects across individual studies. Ratings were “not serious,” “serious” (− 1), or “very serious” (− 2). Inconsistency was evaluated based on the variability of effect estimates across studies. For reviews with meta-analyses, statistical heterogeneity was assessed using the I² statistic, with thresholds of I² > 50% considered “serious” and I² > 75% considered “very serious” inconsistency. For reviews without a meta-analysis, inconsistency was assessed qualitatively by examining the direction and magnitude of effects across individual studies. Ratings were “not serious,” “serious” (− 1), or “very serious” (− 2). Indirectness was assessed by evaluating whether the evidence directly addressed the PICO question of this umbrella review. Factors considered included differences in the population (e.g., age ranges extending beyond adolescents), interventions (e.g., modified protocols), comparators, and outcomes. Ratings were “not serious,” “serious” (− 1), or “very serious” (− 2). Indirectness was assessed by evaluating whether the evidence directly addressed the PICO question of this umbrella review. Factors considered included differences in the population (e.g., age ranges extending beyond adolescents), interventions (e.g., modified protocols), comparators, and outcomes. Ratings were “not serious,” “serious” (− 1), or “very serious” (− 2). Imprecision was evaluated for reviews with meta-analyses based on the width of confidence intervals relative to clinically meaningful thresholds and the total sample size. Narrow confidence intervals that did not cross the null value were considered precise. For reviews without meta-analyses, imprecision was assessed based on individual study sample sizes and the consistency of confidence intervals across studies. Ratings were “not serious,” “serious” (− 1), or “very serious” (− 2). Imprecision was evaluated for reviews with meta-analyses based on the width of confidence intervals relative to clinically meaningful thresholds and the total sample size. Narrow confidence intervals that did not cross the null value were considered precise. For reviews without meta-analyses, imprecision was assessed based on individual study sample sizes and the consistency of confidence intervals across studies. Ratings were “not serious,” “serious” (− 1), or “very serious” (− 2). Publication bias was assessed based on evidence reported in the original reviews, including funnel plot asymmetry, statistical tests (e.g., Egger’s test), and any noted limitations regarding unpublished studies. When insufficient information was available, publication bias was assessed based on the comprehensiveness of the search strategy and the inclusion of gray literature. Ratings were “not suspected,” “suspected” (− 1), or “strongly suspected” (− 2). Publication bias was assessed based on evidence reported in the original reviews, including funnel plot asymmetry, statistical tests (e.g., Egger’s test), and any noted limitations regarding unpublished studies. When insufficient information was available, publication bias was assessed based on the comprehensiveness of the search strategy and the inclusion of gray literature. Ratings were “not suspected,” “suspected” (− 1), or “strongly suspected” (− 2). The final certainty of evidence for each outcome was classified as high (⊕⊕⊕⊕), moderate (⊕⊕⊕⊝), low (⊕⊕⊝⊝), or very low (⊕⊝⊝⊝). For systematic reviews without meta-analyses, the GRADE assessment followed the guidelines provided by Murad et al. (2017) for rating certainty in the absence of a single estimate of effect [28]. In such cases, the direction and consistency of findings across individual studies, overall sample sizes, and the methodological quality of primary studies were considered when making judgments across all five GRADE domains. Data synthesis Given the heterogeneity in the types of interventions, comparators, outcomes, and review methodologies among the included systematic reviews, a narrative synthesis approach was adopted as the primary method for integrating findings across reviews. The results were organized thematically into two evidence summary tables. Within each table, findings are reported at the level of individual outcomes (suicidal ideation, suicide attempts, NSSI, depressive symptoms, etc.) and stratified by intervention type where possible. For reviews containing meta-analytic results, quantitative data—including effect sizes (standardized mean differences, odds ratios, or risk ratios), 95% confidence intervals, heterogeneity statistics, and p-values—were extracted and reported. For reviews employing narrative synthesis, the direction and consistency of effects were qualitatively summarized. Where multiple reviews reported on the same intervention–outcome combination, findings were compared across reviews, considering differences in inclusion criteria, search periods, quality ratings, and the degree of primary study overlap. Higher-quality reviews (as determined by AMSTAR-2 ratings) and reviews with more recent and comprehensive search strategies were prioritized. Results Study selection Through a systematic literature search across five databases (PubMed, Embase, CINAHL, Cochrane CENTRAL, Scopus), a total of 4,596 records were identified after deduplication. Following refinement of the search strategy and separation of review articles from original articles, 386 review articles were retained for umbrella review screening. After applying the predefined inclusion and exclusion criteria through title and abstract screening, followed by full-text assessment, six systematic reviews published between 2018 and 2022 met all eligibility criteria and were included in this umbrella review. Characteristics of included systematic reviews The six included systematic reviews were based on a total of 178 unique primary studies, although the reviews used aggregated data from 236 study reports, as Viswanathan et al. [29] included 80 studies derived from 106 publications. The primary study designs included RCTs, observational studies, and quasi-experimental designs. A detailed summary of the PICOS (Population, Intervention, Comparator, Outcome, Study Design) characteristics for each review is presented in Table 1. Table 1Characteristics of the included systematic reviewsAuthor (year of publication)Study designNumber ofincluded studiesAge cohortNationality/raceSearch datesNumber of databasessearched and listMeta-analysisInterventionMain resultLiu et al. (2022)RCT13AdolescentsDiagnosed with depression-2021.105; PubMed, web of science, Embase,China national knowledge infrastructure, WanFang databasesIncludedFluoxetine + CBT vs. fluoxetine aloneCombined therapy could1) improve the treatment effectiveness,2) lower the incidence of adverse reactions3) reduce the incidence of suicide or NSSI (potential—but unconfirmed—)4) reduce the one-year recurrence rateViswanathan et al. (2022)60 RCT,Controlled clinical trials,Observational studiesSystematic reviews80(106 publications)Children/adolescents(mean age ≤ 18 years)Very high HDI (2019 human development index)MDD (majority),Anxiety disorder,Or increased risk of suicide (suicidal ideation, history of suicidal attempts, deliberate self-harm)Depression:2015.01-2020.04Suicide:2012.06-2020.045; PubMed, Cochrane library, PsycINFO, CINAHL, clinicaltrials.govIncludedSuicide:Psychotherapy, counseling, support, or combined interventionsVs. Tau or attention controlDepression:Psychotherapy vs. Wait-list controls, tau, attention control or placeboPharmacotherapy vs. PlaceboCombination therapy (fluoxetine + CBT) vs. PlaceboCollaborative care vs. TAUBoth pharmacotherapy and psychotherapy treatments have some benefit for some depression outcomes; the evidence is limited for suicide risk interventions. Harms are rare in treatment studies but more frequent in pharmacotherapy arms when compared with placeboJoe et al. (2018)RCT6Adolescents(12–29 years)Black(African American)2000–20154; EbscoHost, ProQuest, SCOPUS, web of scienceNot includedABFT, SSRI, MST, good behavior game, headstrong program- MST, ABFT reduced suicidal ideation after treatment and after follow-up at 1 yearKothgassner et al. (2020)RCT25Adolescent(12–19 years)Majority of participants engaged in self-harm at least once1991.12-2019.122; Google Scholar, PubMedIncludedDBT-A, CBT, group therapy, family therapy, MBT-A, cognitive analytic therapy, interpersonal therapy, brief intervention + TAU, therapeutic assessment- treatment intervention fared slightly better than active controls in decreasing self-harm, suicidal ideation, and depressive symptoms- DBT-A and family-centered therapy(in particular, ABFT) showing small-to-moderate effects of self-harm and suicidal ideationFrey et al. (2020)16 RCT, 1 nonramdomized trial with comparison group, 1 post hoc secondary analysis of RCT, 3 pre-post-tx-group only, 1 comparing 2 treatment groups22Adolescent(10–19 years)With a past episode of either suicide ideation orBehavior or severe depression symptoms2010–20193; Cochrane Library,PubMed, PsycINFONot included(1) CBT+parent training (PA-CBT, I-CBT, F-CBT),(2) CBT-family-emergency (FISP),(3) CBT+systemic principles-emergency (FBCI),(4) CBT+systemic principles (safety),(5) DBT+family training (DBT-A),(6) systemic principles (ABFT, ABFT-LGB, shift),(7) psychoeducation (RAP-P, yst)Suicide-related outcomesLevel 1: DBT+family training, systemic principlesLevel 2: CBT+parent training, CBT+systemic principles, psychoeducationLevel 4: CBT+systemic principles-emergency, CBT family-emergencyJohnstone et al. (2022)6 RCT,Pre-post, uncontrolled studies,Quasi-experimental studies25Adolescent(10–19 years)With at least one or more DSM-V diagnostic criteria for borderline personality disorder (bpd) and self-harm/suicidal behavior in outpatient,Inpatient or correctional facilities.-2019.046; CINAHL,PsycINFO,PsycARTICLES,Psychological & behavioral sciences collection,PubMed,Web of science databasesNot includedDBT-a, MBT-a- DBT-A and MBT-A demonstrate significant improvements in suicidal ideation, suicidal attempts, self-harm, borderline personality disorder traits, depressive symptoms, trauma, behavior problems and general psychosocial functioning.- DBT-A with bpd traits who engage in repetitive self-harming behavior is more effective than MBT Characteristics of the included systematic reviews 5; PubMed, web of science, Embase, China national knowledge infrastructure, WanFang databases Combined therapy could 1) improve the treatment effectiveness, 2) lower the incidence of adverse reactions 3) reduce the incidence of suicide or NSSI (potential—but unconfirmed—) 4) reduce the one-year recurrence rate 60 RCT, Controlled clinical trials, Observational studies Systematic reviews 80 (106 publications) Children/adolescents (mean age ≤ 18 years) MDD (majority), Anxiety disorder, Or increased risk of suicide (suicidal ideation, history of suicidal attempts, deliberate self-harm) Depression: 2015.01-2020.04 Suicide: 2012.06-2020.04 Suicide: Psychotherapy, counseling, support, or combined interventions Vs. Tau or attention control Depression: Psychotherapy vs. Wait-list controls, tau, attention control or placebo Pharmacotherapy vs. Placebo Combination therapy (fluoxetine + CBT) vs. Placebo Collaborative care vs. TAU Adolescents (12–29 years) Black (African American) Adolescent (12–19 years) - treatment intervention fared slightly better than active controls in decreasing self-harm, suicidal ideation, and depressive symptoms - DBT-A and family-centered therapy(in particular, ABFT) showing small-to-moderate effects of self-harm and suicidal ideation Adolescent (10–19 years) With a past episode of either suicide ideation or Behavior or severe depression symptoms 3; Cochrane Library, PubMed, PsycINFO (1) CBT+parent training (PA-CBT, I-CBT, F-CBT), (2) CBT-family-emergency (FISP), (3) CBT+systemic principles-emergency (FBCI), (4) CBT+systemic principles (safety), (5) DBT+family training (DBT-A), (6) systemic principles (ABFT, ABFT-LGB, shift), (7) psychoeducation (RAP-P, yst) Suicide-related outcomes Level 1: DBT+family training, systemic principles Level 2: CBT+parent training, CBT+systemic principles, psychoeducation Level 4: CBT+systemic principles-emergency, CBT family-emergency 6 RCT, Pre-post, uncontrolled studies, Quasi-experimental studies Adolescent (10–19 years) With at least one or more DSM-V diagnostic criteria for borderline personality disorder (bpd) and self-harm/suicidal behavior in outpatient, Inpatient or correctional facilities. 6; CINAHL, PsycINFO, PsycARTICLES, Psychological & behavioral sciences collection, PubMed, Web of science databases - DBT-A and MBT-A demonstrate significant improvements in suicidal ideation, suicidal attempts, self-harm, borderline personality disorder traits, depressive symptoms, trauma, behavior problems and general psychosocial functioning. - DBT-A with bpd traits who engage in repetitive self-harming behavior is more effective than MBT Liu et al. (2022) [30] conducted a narrowly focused systematic review and meta-analysis of 13 RCTs examining combination therapy (fluoxetine plus CBT) compared directly with fluoxetine monotherapy for adolescents diagnosed with depression. The review searched five databases (PubMed, Web of Science, Embase, China National Knowledge Infrastructure, and WanFang) through October 2021. Key outcomes included treatment response rate, incidence of adverse reactions, and one-year recurrence rate. Viswanathan et al. (2022) [29] provided the largest and most methodologically comprehensive review, encompassing 80 studies (60 RCTs, controlled clinical trials, and observational studies) derived from 106 publications. Conducted for the United States Preventive Services Task Force (USPSTF), this review searched five databases (PubMed, Cochrane Library, PsycINFO, CINAHL, and ClinicalTrials.gov) and examined a broad population of children and adolescents (mean age ≤ 18 years) with major depressive disorder (MDD), anxiety disorders, or suicide risk (including suicidal ideation, suicide attempt history, and deliberate self-harm). The search period was January 2015 to April 2020 for depression and June 2012 to April 2020 for suicide-related outcomes. The interventions evaluated included psychotherapy, pharmacotherapy, and combination therapy compared with various comparators such as TAU or placebo conditions. Joe et al. (2018) [31] addressed a specific and often under-researched population by including six RCTs focusing exclusively on Black (African American) adolescents aged 12–29 years at risk of suicide. The review searched four databases (EBSCOhost, ProQuest, Scopus, and Web of Science) covering the years 2000–2015. The evaluated interventions included ABFT, MST, selective serotonin reuptake inhibitors (SSRIs), and the Good Behavior Game program. This review employed narrative synthesis without meta-analysis. Kothgassner et al. (2020) [11] included 25 RCTs focusing on adolescents aged 12–19 years with a history of self-harm. The review searched two databases (Google Scholar and PubMed) for articles published between December 1991 and December 2019. This review synthesized evidence for various psychotherapies—including dialectical behavior therapy for adolescents (DBT-A), CBT, group therapy, family therapy, MBT-A, and cognitive therapy—compared with active control conditions (TAU or enhanced usual care). Meta-analysis was performed for three primary outcomes: self-harm, suicidal ideation, and depressive symptoms. Frey et al. (2020) [32] synthesized 22 studies with mixed designs (16 RCTs, one nonrandomized trial with a comparison group, and other designs), targeting adolescents aged 10–19 years with a history of suicidal ideation or behavior, or severe depressive symptoms. The review searched three databases (Cochrane Library, PubMed, and PsycINFO) for articles published between 2010 and 2019. The distinctive contribution of this review was its specific focus on family-based interventions, including DBT-A with family training, various CBT models (PA-CBT, I-CBT, and F-CBT), and systemic principle models such as ABFT. Evidence was synthesized narratively using Southam-Gerow and Prinstein’s (2014) criteria to evaluate the strength of evidence. Johnstone et al. (2022) [33] included 25 studies (six RCTs, with the remainder comprising pre–post, uncontrolled, and quasi-experimental studies) targeting adolescents aged 10–19 years who met at least one DSM-5 diagnostic criterion for borderline personality disorder (BPD) with self-harm or suicidal behaviors. The review searched six databases (CINAHL, PsycINFO, PsycARTICLES, Psychological and Behavioral Sciences Collection, and others) through April 2019. This review specifically aimed to compare DBT-A and MBT-A through narrative synthesis. The heterogeneity in scope, target population, intervention focus, and synthesis methods among these six systematic reviews is a key characteristic of this umbrella review. Three reviews included meta-analyses (Liu et al., Viswanathan et al., and Kothgassner et al.), whereas three relied on narrative synthesis (Joe et al., Frey et al., and Johnstone et al.). This variability necessitates nuanced interpretation and synthesis of the findings rather than direct quantitative comparisons across reviews. Primary study overlap across included reviews Across the six included systematic reviews, a total of 97 primary study citations were identified, representing 72 unique primary studies. For Viswanathan et al. (2022), only primary studies directly relevant to the intervention–outcome combinations examined in this umbrella review were extracted for the overlap analysis, given the broad scope of that review. The CCA was 0.069, indicating moderate overlap (Pieper et al., 2014) [25]. The citation overlap matrix is presented in Table 2. Overlap was entirely absent between Liu et al. (2022) and all other reviews, reflecting that review’s distinct pharmacotherapy focus and reliance on Chinese-language RCTs not covered elsewhere. Similarly, Joe et al. had minimal overlap with other reviews (one shared study), given its exclusive focus on Black male adolescents. The greatest overlap was observed between Viswanathan et al. (2022) and Kothgassner et al. (2020), which shared 12 primary studies—primarily RCTs evaluating psychotherapeutic interventions. Diamond et al. (2010) and Mehlum et al. (2014) were the most frequently shared primary studies, each appearing across four of the six reviews. Overall, the moderate level of overlap is consistent with the complementary scope of the included reviews, and the low overlap of the pharmacotherapy review (Liu et al.) and the population-specific review (Joe et al.) supports the validity of the overall synthesis. Table 2Citation overlap matrix of included systematic reviewsLiu 2022Viswanathan 2022Joe 2018Kothgassner 2020Frey 2021Johnstone 2022Liu 2022—Viswanathan 20220—Joe 201801—Kothgassner 20200121—Frey 20210516—Johnstone 202202034—Numbers indicate the count of shared primary studies between each pair of reviews. The upper triangle is left blank to avoid redundancy. The value 12 (bold red) indicates the highest pairwise overlap, observed between Viswanathan et al. (2022) and Kothgassner et al. (2020). CCA = 0.069, indicating moderate overlap (Pieper et al., 2014 [25]) Citation overlap matrix of included systematic reviews Numbers indicate the count of shared primary studies between each pair of reviews. The upper triangle is left blank to avoid redundancy. The value 12 (bold red) indicates the highest pairwise overlap, observed between Viswanathan et al. (2022) and Kothgassner et al. (2020). CCA = 0.069, indicating moderate overlap (Pieper et al., 2014 [25]) Methodological quality of included reviews (AMSTAR-2) The methodological quality of the six included reviews was assessed using the AMSTAR-2 instrument. The full item-by-item ratings are presented in Table 3. The overall quality was highly variable: one review (16.7%) was rated as “high,” three reviews (50.0%) as “moderate,” and two reviews (33.3%) as “critically low.” Table 3AMSTAR quality assessmentAMSTAR-ITEMQ1Q2Q3Q4Q5Q6Q7Q8Q9Q10Q11Overall QualityLiu et al. (2022)-++--++++++ModerateViswanathan et al. (2022)+++++++++++HighJoe et al. (2018)-++--+----+Critically LowKothgassner et al. (2020)-+++-+--+++ModerateFrey et al. (2020)--+--+--+--Critically LowJohnstone et al. (2022)--+--++++-+Moderate AMSTAR quality assessment High quality Viswanathan et al. (2022) [29] was the only review to receive a “high” quality rating, satisfying all 11 AMSTAR-2 criteria. This review demonstrated exceptional methodological rigor across all domains, including a preregistered protocol (posted on the USPSTF website for public comment); independent dual study selection and data extraction; a comprehensive literature search across five databases, including trial registries for unpublished studies; a complete list of included and excluded studies; appropriate risk-of-bias assessment with quality-stratified interpretation of results; appropriate meta-analytic methods; and formal assessment of publication bias using funnel plots. Preregistration of the protocol and inclusion of ClinicalTrials.gov for unpublished literature were notable strengths that distinguished this review from the others. Moderate quality Liu et al. (2022) [30], Kothgassner et al. (2020) [11], and Johnstone et al. (2022) [33] were rated as “moderate” in quality. Although all three reviews met most methodological criteria, they shared common weaknesses. None of the three preregistered their review protocols (Q1). Liu et al. and Johnstone et al. did not provide a list of excluded studies (Q5). Liu et al. assessed risk of bias using the Jadad scoring system and reported publication bias using funnel plots; however, the use of the Jadad scale—rather than more comprehensive tools such as the Cochrane Risk of Bias tool—may represent a less rigorous approach. Kothgassner et al. did not perform a formal risk-of-bias assessment of the included studies (Q7, Q8), although publication bias was assessed using Egger’s regression. Johnstone et al. performed quality assessment using the National Institutes of Health (NIH) quality assessment tool, but two independent reviewers reported some discrepancy regarding whether the quality assessment findings were adequately incorporated into the review conclusions (Q8). A notable divergence between the two reviewers was observed for Johnstone et al.’s review on several items (Q2, Q7, Q8, Q9), reflecting genuine ambiguity in the reporting of these methodological elements. Critically low quality Joe et al. (2018) [31] and Frey et al. (2020) [32] were rated as “critically low” in quality, exhibiting critical methodological flaws that substantially limit confidence in their findings. Both reviews failed to preregister protocols (Q1), provide lists of excluded studies (Q5), and assess publication bias (Q10). Joe et al. additionally did not conduct a risk-of-bias assessment of the included studies (Q7), did not use quality findings to inform conclusions (Q8), and did not employ appropriate methods for combining individual study results (Q9). Frey et al. showed additional weaknesses, including the absence of a clearly described dual-reviewer study selection process (Q2), no formal risk-of-bias assessment (Q7, Q8), and no disclosure of conflicts of interest (Q11). A critical cross-cutting finding from the quality assessment was the absence of publication bias assessment in five of the six reviews (all except Viswanathan et al.). Only Viswanathan et al. used funnel plots, and Liu et al. reported funnel plot assessment, whereas Kothgassner et al. used Egger’s regression. This limitation raises concerns about the potential overestimation of effect sizes due to the nonpublication of smaller, nonsignificant studies—the “file drawer problem.” This methodological weakness directly informed the GRADE assessments, particularly for narrative reviews (Joe et al. and Johnstone et al.), where publication bias was judged as “Suspected.” Notably, the two reviews rated as “critically low” in quality (Joe et al. and Frey et al.) largely correspond with the reviews that did not include quantitative meta-analyses and whose narrative syntheses were ultimately graded as “Very Low” certainty evidence. This pattern suggests that lower methodological quality often coincides with a reduced ability to provide high-certainty, quantitatively synthesized conclusions. Synthesis of intervention efficacy and certainty of evidence (GRADE) The findings from the six systematic reviews are synthesized below by intervention type and outcome domain. The certainty of evidence for each outcome was assessed using the GRADE framework. The summary of findings for reviews with meta-analysis is presented in Table 4, and the summary of findings for reviews without meta-analysis is presented in Table 5. Table 4GRADE evidence summary (with meta-analysis)Systematic ReviewTotal Number ofStudiesMeta ResearchQuality AssessmentResult SummaryRoBInconsistencyIndirectenessImprecisionPublication BiasIntervention/ControlNumber ofStudiesNumber of ParticipantsOutcomeEffect Estimate (95% CI)p-valueHeterogeneity (I²)p-valueSummary of FindingsEvidence LevelLiu et al. (2022)13IncludedNot seriousNo serious inconsistencyNo seriousindirectnessNo seriousimprecisionUndetectedFluoxetine + CBT vs. Fluoxetine alone7628Response rateRR 1.12 (1.04, 1.21)0.0%0.598Significantly higher response rate for Fluoxetine + CBT compared with Fluoxetine monotherapyⴲⴲⴲⴲHighNot seriousNo serious inconsistencyNo seriousindirectnessNo seriousimprecisionUndetectedFluoxetine + CBT vs. Fluoxetine alone6850Recurrence rate (1year)RR 0.27 (0.16, 0.45)5.8%0.379Combined treatment significantly reduces recurrence rate compared with fluoxetine aloneⴲⴲⴲⴲHighNot seriousNo serious inconsistencyNo seriousindirectnessSerious imprecisionUndetectedFluoxetine + CBT vs. Fluoxetine alone7604Adverse reactionsRR 0.62 (0.40, 0.96)0.0%0.901Combined treatment reduces adverse reaction rates but not statistically significantⴲⴲⴲⵔModerateSeriousNo serious inconsistencyNo seriousindirectnessSerious imprecisionUndetectedFluoxetine + CBT vs. Fluoxetine alone3568Suicide/NSSIRR 0.94 (0.74, 1.20)0.0%0.442Combined treatment significantly reduces suicide/NSSI ratesⴲⴲⵔⵔLowSeriousVery serious inconsistencyNo seriousindirectnessNo seriousimprecisionUndetectedFluoxetine + CBT vs. Fluoxetine alone9568HAM-DSMD − 1.01 (-1.39, -0.63)78.5%≤ 0.001Combined treatment significantly improves HAM-D scoresⴲⴲⵔⵔLowNot seriousNo serious inconsistencyNo seriousindirectnessSerious imprecisionUndetectedFluoxetine + CBT vs. Fluoxetine alone3568CDRS-RSMD − 0.10 (-0.26, 0.07)24.0%0.268Combined treatment significantly improves CDRS-R scoresⴲⴲⴲⵔModerateNot seriousSerious inconsistencyNo seriousindirectnessSerious imprecisionUndetectedFluoxetine + CBT vs. Fluoxetine alone3504CGISMD − 0.22 (-0.54, 0.10)72.1%0.028Combined treatment improves CGI scores but not statistically significantⴲⴲⵔⵔLowViswanathan et al., 202280IncludedNot seriousSeriousinconsistencyNo seriousindirectnessSerious imprecisionUndetectedYouth-nominated support team, Psychotherapy (motivational interviewing, DBT, IPT-A-IN)4644Beck Hopelessness ScaleRR -2.35 (-4.06, -0.65)0.00721%0.135Interventions may be effective in alleviating feelings of despair and suicidal thoughts, the evidence is limited.ⴲⴲⵔⵔLowNot seriousNo serious inconsistencyNo seriousindirectnessSerious imprecisionUndetectedPsychotherapy (CBT, DBT)3978Self-harm-related hospital visit ratesRR = 0.998 (0.67, 1.50)68.4%0.28No clear evidence that interventions reduce self-harm related hospital visitsⴲⴲⴲⵔModerateSeriousNo serious inconsistencyNo seriousindirectnessVery serious imprecisionUndetectedPsychotherapy (Family therapy, DBT, developmental group therapy)3972Mean number of self-harm eventsMD = -0.76 (-2.15, 0.63)0.28379.1%0.042No clear evidence that interventions reduce self-harm related hospital visitsⴲⴲⵔⵔLowSeriousVery serious inconsistencyNo seriousindirectnessVery serious imprecisionUndetectedPsychotherapy (Group psychotherapy,family therapy,mentalization-based treatment,developmental group therapy)51040Proportion with self-harm eventsRR = 0.88 (0.62, 1.24)0.34546.0%0.001No evidence of reduction in number of self-harm eventsⴲⵔⵔⵔVery LowSeriousNo serious inconsistencyNo seriousindirectnessVery serious imprecisionUndetectedPsychotherapy (Attachment-based family therapy, group psychotherapy, group therapy, youth-nominated support team, motivational interviewing, DBT, developmental group therapy)71111SIQ(Suicidal Ideation Questionnaire)/SIQ-JR(Suicidal Ideation Questionnaire-Junior)SMD = -0.176 (-0.36, 0.008)0.06144.80.092Limited evidence for clear reduction in suicidal ideationⴲⴲⵔⵔLowNot seriousSerious inconsistencyNo seriousindirectnessSerious imprecisionUndetectedGroup psychotherapy; group therapy; developmental group therapy; psychoeducation for parents4509Functional Status (HoNOSCA)MD = -0.398 (-2.55, 1.76)0.71756.1%0.08Group-based psychotherapeutic interventions show no significant effect on functional statusⴲⴲⵔⵔLowNot seriousNo serious inconsistencyNo seriousindirectnessSerious imprecisionUndetectedTherapeutic assessment; individual and family DBT; group therapy3195Functional Status (CGAS)MD = 1.302 (-2.51, 5.12)0.50429.5%0.24Therapeutic assessment and DBT show no significant effect on functional statusⴲⴲⴲⵔModerateSeriousNo serious inconsistencyNo seriousindirectnessSerious imprecisionUndetectedPsychotherapy(CBT, interpersonal psychotherapy)4471BDI-IISMD = -0.58 (-0.83, -0.34)significant0.0%Not reportedInterventions were found to be effective in improving depressive symptomsⴲⴲⵔⵔLowSeriousNo serious inconsistencyNo seriousindirectnessSerious imprecisionUndetectedPsychotherapy(CBT, interpersonal psychotherapy)3262HAM-DMD = -2.25 (-4.09, -0.42)significant0.0%Not reportedInterventions were found to be effective in improving depressive symptomsⴲⴲⵔⵔLowSeriousNo serious inconsistencyNo seriousindirectnessSerious imprecisionUndetectedPsychotherapy(CBT, family CBT)3471CDRS-RMD = 0.76 (-0.97, 2.48)not significant0.0%Not reportedPsychotherapy shows no significant effectiveness in improving depressive symptomsⴲⴲⵔⵔLowNot seriousSeriousinconsistencyNo seriousindirectnessNo seriousimprecisionUndetectedPharmacotherapy(Escitalopram, fluoxetine)3793CDRS-RMD = -3.76 (-5.95, -1.57)significant49.0%Not reportedPharmacotherapy shows clear effectiveness in improving depressive symptomsⴲⴲⴲⵔModerateNot seriousVery serious inconsistencyNo seriousindirectnessSeriousUndetectedPsychotherapy (CBT)4395Loss of diagnosisRR = 1.73 (1.00, 3.00)significant81.0%Not reportedThe pooled estimate and results favored the treatment arms, but not statistically significantⴲⴲⵔⵔLowNot seriousNo serious inconsistencyNo seriousindirectnessSeriousUndetectedPharmacotherapy(Escitalopram, fluoxetine)3793Remission fromdepressionRR = 1.20 (1.00, 1.45)not significant0.0%Not reportedNo significant effect on remission from depressionⴲⴲⴲⵔModerateNot seriousSeriousinconsistencyNo seriousindirectnessSeriousUndetectedPsychotherapy(CBT, IPT)4601Functioning (CGAS)MD = 1.52 (-1.54, 4.58)not significant66.0%Not reportedNo significant improvement in functioning (CGAS)ⴲⴲⵔⵔLowSeriousNo serious inconsistencyNo seriousindirectnessNo seriousimprecisionUndetectedPharmacotherapy(Escitalopram, fluoxetine)3793Functioning (CGAS)MD = 2.60 (0.78, 4.42)not significant0.0%Not reportedPooled results for CGAS indicated statistically significant differences favoring pharmacotherapyⴲⴲⴲⵔModerateKothgassner et al. (2020)25IncludedNot seriousNo serious inconsistencyNo serious indirectnessNo Serious imprecisionUndetectedDBT-A, CBT, Group Therapy, Family Therapy, MBT-A, Cognitive Analytic Therapy, Brief Intervention + TAU, Therapeutic Assessmentvs. TAU or EUC172534Self-harmd = 0.13 (0.04, 0.22)0.0042.51%0.258Treatment interventions show small effect in reducing self-harm compared with active controlsⴲⴲⴲⴲHighNot seriousVery serious inconsistencyNo serious indirectnessNo Serious imprecisionUndetectedDBT-A, CBT, Group Therapy, Family Therapy, Brief Intervention + TAU, Interpersonal Therapyvs. TAU or EUC192542Suicidal ideationd = 0.31 (0.12, 0.50)0.00178.51%0.0000Treatment interventions show moderate effect in reducing suicidal ideation compared with active controlsⴲⴲⴲⵔModerateNot seriousSerious inconsistencyNo serious indirectnessNo Serious imprecisionUndetectedDBT-A, CBT, Group Therapy, Family Therapy, MBT-A, Brief Intervention + TAU, Interpersonal Therapyvs. TAU or EUC172071Depressive symptomsd = 0.22 (0.07, 0.38)0.00656.70%0.002Treatment interventions show moderate effect in reducing depressive symptoms compared with active controlsⴲⴲⴲⵔModerate GRADE evidence summary (with meta-analysis) No serious indirectness No serious imprecision ⴲⴲⴲⴲ High No serious indirectness No serious imprecision ⴲⴲⴲⴲ High No serious indirectness ⴲⴲⴲⵔ Moderate No serious indirectness ⴲⴲⵔⵔ Low No serious indirectness No serious imprecision ⴲⴲⵔⵔ Low No serious indirectness ⴲⴲⴲⵔ Moderate No serious indirectness ⴲⴲⵔⵔ Low Serious inconsistency No serious indirectness ⴲⴲⵔⵔ Low No serious indirectness ⴲⴲⴲⵔ Moderate No serious indirectness ⴲⴲⵔⵔ Low No serious indirectness Psychotherapy (Group psychotherapy, family therapy, mentalization-based treatment, developmental group therapy) ⴲⵔⵔⵔ Very Low No serious indirectness ⴲⴲⵔⵔ Low No serious indirectness ⴲⴲⵔⵔ Low No serious indirectness ⴲⴲⴲⵔ Moderate No serious indirectness Psychotherapy (CBT, interpersonal psychotherapy) ⴲⴲⵔⵔ Low No serious indirectness Psychotherapy (CBT, interpersonal psychotherapy) ⴲⴲⵔⵔ Low No serious indirectness Psychotherapy (CBT, family CBT) ⴲⴲⵔⵔ Low Serious inconsistency No serious indirectness No serious imprecision Pharmacotherapy (Escitalopram, fluoxetine) ⴲⴲⴲⵔ Moderate No serious indirectness ⴲⴲⵔⵔ Low No serious indirectness Pharmacotherapy (Escitalopram, fluoxetine) Remission from depression ⴲⴲⴲⵔ Moderate Serious inconsistency No serious indirectness ⴲⴲⵔⵔ Low No serious indirectness No serious imprecision Pharmacotherapy (Escitalopram, fluoxetine) ⴲⴲⴲⵔ Moderate DBT-A, CBT, Group Therapy, Family Therapy, MBT-A, Cognitive Analytic Therapy, Brief Intervention + TAU, Therapeutic Assessment vs. TAU or EUC ⴲⴲⴲⴲ High DBT-A, CBT, Group Therapy, Family Therapy, Brief Intervention + TAU, Interpersonal Therapy vs. TAU or EUC ⴲⴲⴲⵔ Moderate DBT-A, CBT, Group Therapy, Family Therapy, MBT-A, Brief Intervention + TAU, Interpersonal Therapy vs. TAU or EUC ⴲⴲⴲⵔ Moderate Table 5GRADE evidence summary (without meta-analysis)Systematic ReviewTotal Number ofStudiesGRADE DomainQuality AssessmentResult SummaryJudgmentConcerns AboutCertainty DomainsIntervention/ControlNumber ofStudiesNumber of ParticipantsOutcomeSummary of FindingsEvidence LevelJoe et al. (2018)6RoBAll of the studies included were RCTsNot seriousMulti-component interventions vs. various control groups62763Suicidal ideation/attemptSome interventions (MST, ABFT) show effectiveness but evidence is limited ⴲⵔⵔⵔ Very Low InconsistencyThere are not enough studies to adequately assess consistency. Although ABFT showed effectiveness in reducing suicidal ideation, the evidence is based on only two RCTs. Similarly, MBT demonstrated effectiveness in reducing suicidal behavior, but this conclusion rests on only one RCT. Therefore, consistency is difficult to evaluate.SeriousIndirectenessThere are the inconsistencies in how suicidal behavior was measured across the studies. It is not clear if therewill be long-term treatment benefits, because this informationwas not given in these articles. Subgroup analyses by gender and ethnicity were not possible; therefore, applicability to the target population is limitedSeriousImprecisionNo single effect estimate or confidence interval was provided because quantitative synthesis was not performed. Among the six RCTs, four had very small sample sizes, and each intervention was represented by only a very limited number of studiesSeriousPublication BiasPublication bias may be suspected because the body of evidence is composed primarily of small positive studies and does not include both positive and negative studiesSuspectedFrey et al. (2021)22RoB17 of the 22 studies (16 RCTs and one post-hoc analysis of an RCT) show a low risk of bias. However, 5 studies have a high risk of bias due to their use of a nonrandomized comparison group or a pre- and post-treatment design without a control group.SeriousFamily-based interventions vs. various control groups222252Suicidal ideation/attemptWell-established: DBT+Family Training, Systemic Principles; Probably efficacious: CBT+Parent Training ⴲⵔⵔⵔ Very Low InconsistencyThere was inconsistency in the direction of effects; For example, within DBT + family training interventions, some trials showed superiority over control, while others found no significant group differencesSeriousIndirectnessVariation in outcome measures; most studies used validated tools, but not all assessed both suicidal ideation and behavior. Most studies compared interventions only to treatment as usual or slightly enhanced care, limiting conclusions about the added value of family involvementSeriousImprecisionMost individual trials enrolled fewer than 200 participants. Some studies showed small reductions, whereas others found non-significant results, likely due to the small sample sizes leading to wide confidence intervals that encompassed both meaningful benefits and no effectSeriousPublication BiasPublication bias was considered unlikely, as the evidence base included both positive and negative trials and the literature search was comprehensive.UnsuspectedJohnstone et al. (2022)25RoBMost of the studies included were not RCTs, but rather pre-post or quasi-experimental designs. Only 6 of the 25 total studies were RCTs, and most of their control groups were usual care. The quality of the included studies was rated as ‘Fair’ or ‘Poor’. The small sample sizes and potential for bias were also noted.SeriousDBT-A vs. MBT-A251539Suicidal ideation/attempt, BPD symptomDBT-A more effective for repetitive self-harm behaviors; MBT-A research very limited ⴲⵔⵔⵔ Very Low InconsistencyThe review reported mixed findings across studies. For instance, the effect of MBT-A on self-harm varied: one study found it more effective than usual care, while another found no significant difference. The effect of DBT-A on parental depressive symptoms was also inconsistent. Additionally, while DBT-A studies generally showed a reduction in hospitalizations, the effect was not greater than that of control groups.Serious InconsistencyIndirectnessThe included studies showed significant heterogeneity in treatment duration, intensity, methods, and follow-up periods. In addition, two of the MBT-A studies did not assess mentalization ability, and most studies did not distinguish between self-harm with or without suicidal intent. This inconsistency reduces the directness of the findings.Serious IndirectnessImprecisionNo single effect estimate or confidence interval was provided because quantitative synthesis was not performed. The small sample sizes are a key concern for imprecision. For example, the review included 21 DBT-A studies and only 4 MBT-A studies, which likely fails to meet the optimal information size.Serious ImprecisionPublication BiasThe review’s inclusion of a disproportionate number of DBT-A studies (21) compared with MBT-A studies (4) suggests a potential for publication bias. The review did not explicitly state whether it included both positive and negative findings.Suspected GRADE evidence summary (without meta-analysis) ⴲⵔⵔⵔ Very Low There are the inconsistencies in how suicidal behavior was measured across the studies. It is not clear if there will be long-term treatment benefits, because this information was not given in these articles. Subgroup analyses by gender and ethnicity were not possible; therefore, applicability to the target population is limited ⴲⵔⵔⵔ Very Low ⴲⵔⵔⵔ Very Low Combination therapy (Fluoxetine + CBT) versus monotherapy All findings in this domain are derived from the “moderate” quality meta-analysis by Liu et al. (2022) [30], which compared fluoxetine plus CBT combination therapy with fluoxetine monotherapy for adolescents with depression. Seven distinct outcomes were examined. Treatment response and recurrence High-certainty evidence (⊕⊕⊕⊕) supports that combination therapy is superior to monotherapy for improving treatment response rates. The meta-analysis of seven studies (N = 628) demonstrated a statistically significant benefit (RR = 1.12, 95% CI [1.04, 1.21]; I2 = 0%, p = 0.598). The evidence was not downgraded in any GRADE domain: risk of bias was not serious, inconsistency was absent (I2 = 0%), indirectness was not serious, imprecision was not serious, and publication bias was not detected. Similarly, high-certainty evidence (⊕⊕⊕⊕) suggests that combination therapy significantly reduces the one-year recurrence rate compared with monotherapy. The meta-analysis of six studies (N = 850) showed a substantial reduction (RR = 0.27, 95% CI [0.16, 0.45]; I2 = 5.8%, p = 0.379). No downgrading was applied in any GRADE domain. Safety Moderate-certainty evidence (⊕⊕⊕⊝) indicates that combination therapy significantly reduces the incidence of adverse reactions compared with monotherapy. The meta-analysis of seven studies (N = 604) showed a significant reduction (RR = 0.62, 95% CI [0.40, 0.96]; I2 = 0%, p = 0.901). This evidence was downgraded by one level for “Serious Imprecision,” reflecting the wide confidence interval, with the upper bound (0.96) approaching the null value, and the relatively small total sample size. Suicidality/NSSI Low-certainty evidence (⊕⊕⊝⊝) was found for the suicidality/NSSI outcome. The meta-analysis of three studies (N = 568) found no statistically significant difference between combination therapy and monotherapy (RR = 0.94, 95% CI [0.74, 1.20]; I² = 0%, p = 0.442). This evidence was downgraded by two levels for “Serious Risk of Bias” and “Serious Imprecision.” The nonsignificant result is consistent with Liu et al.’s characterization of this effect as “potential” rather than definitive. Depressive symptom scales For the Hamilton Depression Rating Scale (HAM-D), low-certainty evidence (⊕⊕⊝⊝) showed a statistically significant improvement with combination therapy. The meta-analysis of nine studies (N = 568) yielded a large effect (SMD = − 1.01, 95% CI [− 1.39, − 0.63]), but heterogeneity was substantial (I² = 78.5%, p ≤ 0.001). This evidence was downgraded by two levels for “Serious Risk of Bias” and “Very Serious Inconsistency,” the latter reflecting the high I² value that exceeded the 75% threshold for very serious inconsistency. For the Children’s Depression Rating Scale–Revised (CDRS-R), moderate-certainty evidence (⊕⊕⊕⊝) showed no statistically significant difference (SMD = − 0.10, 95% CI [− 0.26, 0.07]; I² = 24%, p = 0.268). This evidence was downgraded by one level for “Serious Imprecision.” For the Clinical Global Impression (CGI), low-certainty evidence (⊕⊕⊝⊝) showed no statistically significant difference (SMD = − 0.22, 95% CI [− 0.54, 0.10]; I² = 72.1%, p = 0.028). This evidence was downgraded by two levels for “Serious Inconsistency” and “Serious Imprecision.” Psychotherapy for suicidal ideation and self-harm The evidence for psychotherapy in reducing suicidal behavior and self-harm presents a complex and, at times, contradictory picture, derived from multiple reviews with variable methodological quality and conflicting results. A. Quantitative Evidence from Kothgassner et al. (2020) The “moderate” quality review by Kothgassner et al. [11] meta-analyzed 25 RCTs comparing various psychotherapies (DBT-A, CBT, group therapy, family therapy, MBT-A, and others) with active control conditions (TAU or enhanced usual care). For self-harm, high-certainty evidence (⊕⊕⊕⊕) demonstrated a small but statistically significant effect favoring intervention over control conditions (d = 0.13, 95% CI [0.04, 0.22]; 17 studies, N = 2,534; I² = 2.51%, p = 0.258). No downgrading was applied: risk of bias was not serious, inconsistency was minimal (I² = 2.51%), indirectness was not serious, imprecision was not serious, and publication bias was not detected. For suicidal ideation, moderate-certainty evidence (⊕⊕⊕⊝) showed a moderate effect (d = 0.31, 95% CI [0.12, 0.50]; 19 studies, N = 2,542; I² = 78.51%, p < 0.001). This evidence was downgraded by one level for “Very Serious Inconsistency” (I² = 78.51%), reflecting the substantial statistical heterogeneity likely arising from the clinical heterogeneity associated with pooling distinct interventions into a single effect estimate. For depressive symptoms, moderate-certainty evidence (⊕⊕⊕⊝) showed a moderate effect (d = 0.22, 95% CI [0.07, 0.38]; 17 studies, N = 2,071; I² = 56.7%, p = 0.002). This evidence was downgraded by one level for “Serious Inconsistency” (I² = 56.7%). B. Quantitative Evidence from Viswanathan et al. (2022) The “high” quality review by Viswanathan et al. [29] presented a more cautious picture regarding the effectiveness of psychotherapy for self-harm and suicidal ideation outcomes. For the proportion of adolescents reporting a self-harm event, very low-certainty evidence (⊕⊝⊝⊝) showed no significant reduction (RR = 0.88, 95% CI [0.62, 1.24]; five studies, N = 1,040; I² = 79.14%, p = 0.001). This evidence was downgraded by three levels for “Serious Risk of Bias,” “Very Serious Inconsistency” (I² = 79.14%), and “Very Serious Imprecision.” This three-level downgrade renders the result highly uncertain. For the mean number of self-harm events, low-certainty evidence (⊕⊕⊝⊝) showed no significant reduction (MD = − 0.76, 95% CI [− 2.15, 0.63]; three studies, N = 972; I² = 68.39%, p = 0.042), downgraded by two levels for “Serious Risk of Bias” and “Very Serious Imprecision.” For self-harm–related hospital visit rates, moderate-certainty evidence (⊕⊕⊕⊝) showed no clear evidence of reduction (RR = 0.998, 95% CI [0.67, 1.50]; three studies, N = 978; I² = 21%, p = 0.28), downgraded by one level for “Serious Imprecision.” For suicidal ideation measured by the Suicidal Ideation Questionnaire (SIQ)/Suicidal Ideation Questionnaire–Junior (SIQ-JR), low-certainty evidence (⊕⊕⊝⊝) showed limited evidence of a clear reduction (SMD = − 0.176, 95% CI [− 0.36, 0.008]; seven studies, N = 1,111; I² = 44.8%, p = 0.092), downgraded by two levels for “Serious Risk of Bias” and “Very Serious Imprecision.” For the Beck Hopelessness Scale, low-certainty evidence (⊕⊕⊝⊝) showed a significant reduction (RR = − 2.35, 95% CI [− 4.06, − 0.65]; four studies, N = 644; I² = 46%, p = 0.135), downgraded by two levels for “Serious Inconsistency” and “Serious Imprecision.” For functional status measured by HoNOSCA, low-certainty evidence (⊕⊕⊝⊝) showed no significant effect (MD = − 0.398, 95% CI [− 2.55, 1.76]; four studies, N = 509; I² = 56.1%, p = 0.08), downgraded by two levels for “Serious Inconsistency” and “Serious Imprecision.” Similarly, for CGAS, moderate-certainty evidence (⊕⊕⊕⊝) showed no significant effect (MD = 1.302, 95% CI [− 2.51, 5.12]; three studies, N = 195; I² = 29.5%, p = 0.24), downgraded by one level for “Serious Imprecision.” C. Discrepancy Between Kothgassner et al. and Viswanathan et al. A notable finding of this umbrella review is the discrepancy between the two quantitative reviews regarding the effectiveness of psychotherapy for self-harm outcomes. Kothgassner et al. reported high certainty evidence for a small but significant effect (d = 0.13), whereas Viswanathan et al. found no evidence of reduction with very low to moderate certainty. This discrepancy likely stems from differences in inclusion criteria (Kothgassner et al. included only RCTs of adolescents with self-harm history, whereas Viswanathan et al. included broader study designs and populations), comparator definitions (active controls vs. TAU/placebo conditions), and outcome operationalization. These divergent findings underscore the sensitivity of effect estimates to methodological choices and highlight the importance of specifying population and comparator characteristics when interpreting evidence. D. Narrative Evidence for Specific Psychotherapy Models The efficacy of specific named-brand therapies was addressed in three narrative reviews, all of which yielded very low certainty evidence (⊕⊝⊝⊝) upon GRADE assessment. An important observation of this umbrella review is the significant discrepancy between the clinical conclusions of these reviews and the methodological certainty of the evidence supporting these conclusions. Frey et al. (2020) [32]: This “critically low” quality review concluded that DBT with family training and systemic principles (e.g., ABFT) are “well-established,” and CBT with parent training is “probably efficacious” for suicidal ideation and behavior. However, the GRADE certainty was rated as very low (⊕⊝⊝⊝) based on 22 studies (N = 2,252). This rating resulted from downgrades in four GRADE domains: “Serious Risk of Bias” (five of 22 studies were non-randomized or non-controlled); “Serious Inconsistency” (mixed results, with some trials reporting no effect while others reported significant benefits); “Serious Indirectness” (variation in outcome measures, with most studies using only TAU as a comparator); and “Serious Imprecision” (most individual trials enrolled fewer than 200 participants). Publication bias was judged as “Unsuspected” given the inclusion of both positive and negative findings. Johnstone et al. (2022) [33]: This “Moderate” quality review concluded that both DBT-A and MBT-A demonstrate significant reductions in suicidal ideation, BPD symptoms, and self-harm, and that DBT-A is more effective than MBT-A for repetitive self-harm behaviors. However, the GRADE certainty was rated as very low (⊕⊝⊝⊝) based on 25 studies (N = 1,539). This rating resulted from downgrades in all five GRADE domains: “Serious Risk of Bias” (most included studies were not RCTs; only six of 25 were randomized); “Serious Inconsistency” (mixed findings for MBT-A effects and DBT-A hospitalization outcomes); “Serious Indirectness” (high heterogeneity in treatment duration, intensity, and failure to distinguish suicidal intent); “Serious Imprecision” (small sample sizes, with a disproportionate 21 DBT-A vs. four MBT-A studies limiting comparability); and “Suspected Publication Bias” (disproportionate inclusion of DBT-A studies suggesting possible selection bias). Joe et al. (2018) [31]: This “critically low” quality review concluded that MST and ABFT showed effectiveness in reducing suicidal ideation among Black male adolescents, but that the evidence base was too limited to provide unequivocal guidance. The GRADE certainty was rated as very low (⊕⊝⊝⊝) based on six RCTs (N = 2,763). Despite the “Not Serious” rating for risk of bias (all included studies were RCTs), the evidence was downgraded in four other domains: “Serious Inconsistency” (insufficient number of studies to assess consistency; ABFT evidence from only two RCTs and MST from one); “Serious Indirectness” (inconsistent measurement of suicidal behavior across studies and lack of long-term outcome data); “Serious Imprecision” (four of six RCTs had very small sample sizes); and “Suspected Publication Bias” (the evidence base was composed of small, predominantly positive studies). Pharmacotherapy and psychotherapy for depressive symptoms Evidence for the treatment of depressive symptoms in adolescents at risk of suicide was derived from the “High” quality review by Viswanathan et al. (2022) [29] and the “Moderate” quality review by Kothgassner et al. (2020) [11]. A. Psychotherapy for Depressive Symptoms From Viswanathan et al., psychotherapy (CBT and IPT) was found to be effective in improving depressive symptoms as measured by the Beck Depression Inventory-II (BDI-II) (SMD = − 0.58, 95% CI [− 0.83, − 0.34]; four studies, N = 471) and the HAM-D (MD = − 2.25, 95% CI [− 4.09, − 0.42]; three studies, N = 262). Both outcomes were rated as low certainty evidence (⊕⊕⊝⊝), downgraded for “Serious Risk of Bias” and “Serious Imprecision.” By contrast, psychotherapy (CBT and family CBT) showed no significant effectiveness on the CDRS-R (MD = 0.76, 95% CI [− 0.97, 2.48]; three studies, N = 471), also rated as low certainty evidence (⊕⊕⊝⊝). Psychotherapy (CBT) was significantly associated with loss of depression diagnosis (remission) (RR = 1.73, 95% CI [1.00, 3.00]; four studies, N = 395), rated as low certainty evidence (⊕⊕⊝⊝), downgraded for “Very Serious Inconsistency” (I² = 81%) and “Serious Imprecision.” From Kothgassner et al., psychotherapy interventions showed a moderate effect on depressive symptoms (d = 0.22, 95% CI [0.07, 0.38]; 17 studies, N = 2,071; I² = 56.7%), rated as moderate certainty evidence (⊕⊕⊕⊝), downgraded by one level for “Serious Inconsistency.” B. Pharmacotherapy for Depressive Symptoms From Viswanathan et al., pharmacotherapy (escitalopram and fluoxetine) showed a clear effect in improving depressive symptoms on the CDRS-R (MD = − 3.76, 95% CI [− 5.95, − 1.57]; three studies, N = 793; I² = 49%), rated as moderate certainty evidence (⊕⊕⊕⊝), downgraded by one level for “Serious Inconsistency.” Pharmacotherapy also showed a potential improvement in functioning as measured by the CGAS (MD = 2.60, 95% CI [0.78, 4.42]; three studies, N = 793; I² = 0%), rated as moderate certainty evidence (⊕⊕⊕⊝), downgraded by one level for “Serious Risk of Bias.” However, pharmacotherapy did not show a significant effect on remission from depression (RR = 1.20, 95% CI [1.00, 1.45]; three studies, N = 793; I² = 0%), rated as moderate certainty evidence (⊕⊕⊕⊝), downgraded for “Serious Imprecision.” C. Comparative Synthesis: Psychotherapy versus Pharmacotherapy for Depression A comparative interpretation of Viswanathan et al.’s findings reveals a nuanced and clinically important pattern. Pharmacotherapy demonstrated clear effectiveness in improving symptom scores on the CDRS-R (moderate certainty), but did not significantly improve remission rates (moderate certainty). Conversely, psychotherapy significantly improved remission from depression (low certainty) but showed mixed results across different symptom measures (low certainty). Psychotherapy also showed no significant improvement in functioning as measured by the CGAS (MD = 1.52, 95% CI [− 1.54, 4.58]; four studies, N = 601; I² = 66%), rated as low certainty evidence (⊕⊕⊝⊝), while pharmacotherapy showed a marginally significant improvement in CGAS functioning (moderate certainty). This divergent pattern suggests that pharmacotherapy and psychotherapy may exert their therapeutic effects through different mechanisms—with pharmacotherapy yielding more robust symptom reduction on clinician-rated scales, while psychotherapy may be more effective in achieving diagnostic-level recovery. These complementary profiles provide a rationale for the use of combination treatment approaches. Summary of evidence by certainty level Across all the outcomes evaluated in this umbrella review, the distribution of evidence certainty was as follows: High certainty (⊕⊕⊕⊕): Three outcomes—treatment response rate for fluoxetine + CBT combination versus fluoxetine monotherapy (RR = 1.12), one-year recurrence rate reduction for the same combination (RR = 0.27), and self-harm reduction with psychotherapy versus active controls (d = 0.13). These represent the most robust findings of this umbrella review. Moderate certainty (⊕⊕⊕⊝): Eight outcomes, including adverse reaction reduction with combination therapy, suicidal ideation reduction with psychotherapy (d = 0.31), depressive symptom reduction with pharmacotherapy (CDRS-R), and several outcomes from Viswanathan et al. relating to self-harm hospital visits, functional status, and remission. Low certainty (⊕⊕⊝⊝): Eleven outcomes, predominantly from Viswanathan et al. and Liu et al., covering suicidal ideation (SIQ/SIQ-JR), depressive symptoms on various scales (BDI-II, HAM-D), hopelessness, functioning, and the suicidality/NSSI outcome for combination therapy. Very low certainty (⊕⊝⊝⊝): Four outcomes—all three narrative review conclusions (Joe et al., Frey et al., Johnstone et al.) regarding specific psychotherapy models, and the proportion with self-harm events from Viswanathan et al. These findings should be interpreted with considerable caution. The distribution of evidence certainty across all evaluated outcomes is summarized in Fig. 2. Summary of GRADE evidence certainty levels across all evaluated outcomes Discussion Summary of principal findings This umbrella review synthesized evidence from six systematic reviews published between 2018 and 2022, encompassing 178 unique primary studies involving adolescents and young adults with suicidal behavior or NSSI. The findings present a nuanced evidence landscape, in which the certainty of evidence varies substantially across intervention types, outcomes, and review methodologies. Three key findings emerged. First, the strongest evidence identified in this umbrella review supports combination therapy (fluoxetine plus CBT) over fluoxetine monotherapy for adolescents with depression. High certainty evidence demonstrated improved treatment response rates (RR = 1.12) and reduced one-year recurrence rates (RR = 0.27), with additional moderate certainty evidence for reduced adverse reactions. These findings, derived from Liu et al. (2022) [30], represent the most robust evidence base in this review. Second, the evidence for psychotherapy as a stand-alone intervention for suicidal ideation and self-harm is promising but inconsistent. Kothgassner et al. (2020) [11] reported high certainty evidence for a small effect on self-harm reduction (d = 0.13) and moderate certainty evidence for suicidal ideation reduction (d = 0.31), whereas Viswanathan et al. (2022) [29]—the review of the highest quality—found no significant evidence for self-harm reduction with very low to moderate certainty. This discrepancy highlights the critical influence of methodological choices on effect estimation. Third, specific psychotherapy models such as DBT-A, MBT-A, ABFT, and MST, despite their widespread clinical adoption and favorable narrative conclusions from individual reviews, were uniformly supported by only very low certainty evidence upon GRADE assessment. This finding represents a significant gap between clinical perception and the methodological rigor of the underlying evidence base. Interpretation of key findings Strength of combination therapy evidence The high certainty evidence supporting fluoxetine plus CBT combination therapy aligns with the landmark Treatment for Adolescents with Depression Study (TADS), which demonstrated the superiority of combination treatment over monotherapy and placebo [34, 35]. The present umbrella review extends this evidence by confirming, through pooled data from 13 RCTs, that the benefit of combination therapy is observed not only in acute symptom reduction but also in sustained outcomes, including reduced one-year recurrence rates. The substantial magnitude of recurrence reduction (RR = 0.27) is particularly clinically significant, suggesting that integrating CBT and pharmacotherapy may confer enduring protective effects. However, the low certainty evidence for the suicidality/NSSI outcome within the same review (RR = 0.94, non-significant) warrants caution. This finding may reflect the well-documented paradox in adolescent depression treatment: while combination therapy effectively treats the underlying depressive disorder, its direct impact on suicidal behavior remains uncertain [36]. The relatively small number of studies (k = 3) addressing this specific outcome underscores the need for adequately powered trials with suicidality as a primary endpoint rather than a secondary safety measure. It is therefore critical to distinguish between these two endpoints. The high-certainty evidence supporting fluoxetine plus CBT pertains primarily to depressive symptom reduction and treatment response rates (RR = 1.12) and one-year recurrence prevention (RR = 0.27). Evidence for direct effects on suicidality and NSSI remains of low certainty and was derived from only three studies in which suicidality was a secondary safety measure rather than a primary endpoint. These two endpoints are not interchangeable, and clinicians should interpret the first-line recommendation for fluoxetine plus CBT as supported by strong evidence for depression outcomes, while acknowledging that its specific effect on suicidal behavior requires further investigation. Psychotherapy paradox: divergent evidence across reviews The divergent findings of Kothgassner et al. and Viswanathan et al. regarding the effectiveness of psychotherapy for self-harm merit careful consideration, as both reviews examined overlapping but distinct bodies of evidence. This discrepancy could be attributed to several methodological factors. Kothgassner et al. restricted their study inclusion to RCTs of adolescents with documented self-harm history and used active controls as comparators, potentially selecting a population more likely to demonstrate treatment effects. Meanwhile, Viswanathan et al. employed broader inclusion criteria, encompassing multiple study designs and a wider population of at-risk youth, which may have diluted the observed effects due to greater clinical heterogeneity. This pattern is consistent with the broader literature on psychotherapy, in which effect sizes tend to be smaller in pragmatic trials with heterogeneous populations than in efficacy trials with selected samples [37]. Both perspectives are informative for clinical decision-making: the findings of Kothgassner et al. suggest that targeted psychotherapy can produce meaningful improvements in self-harm among selected populations, whereas those of Viswanathan et al. temper expectations for population-level effectiveness. Evidence−practice gap in specific psychotherapy models Perhaps the most consequential finding of this umbrella review is the uniformly very low certainty evidence supporting specific psychotherapy models (DBT-A, MBT-A, ABFT, MST) that are considered “well-established” in clinical practice. This evidence−practice gap raises fundamental questions about how clinical recommendations are formulated and communicated. The narrative reviews by Frey et al. (2020) [32], Johnstone et al. (2022) [33], and Joe et al. (2018) [31] each concluded that certain interventions are effective, yet these conclusions were based on evidence with serious limitations across multiple GRADE domains. The classification of DBT with family training as “well-established” by Frey et al., for example, was derived from studies with mixed results, small sample sizes, and varied outcome measures—all of which contributed to the very low GRADE rating. Similarly, Johnstone et al.’s conclusion that DBT-A is more effective than MBT-A for repetitive self-harm was based on a markedly unequal evidence base (21 DBT-A vs. four MBT-A studies), raising concerns about comparability. However, this finding does not necessarily imply that these interventions are ineffective. Rather, it reflects the inherent methodological challenges in psychotherapy research: the impossibility of true double blinding, difficulty of standardizing complex interventions across sites, heterogeneity of outcome measures, and ethical constraints that limit the use of no-treatment control groups in high-risk populations [38]. Therefore, the very low certainty ratings should be interpreted as an indication that higher-quality evidence is needed and not that these interventions should be abandoned. This distinction is particularly important for clinicians and guideline developers who may conflate low GRADE ratings with evidence of ineffectiveness. In psychotherapy research, structural constraints—including the impossibility of participant blinding and the inherent heterogeneity of complex interventions—systematically lower GRADE ratings regardless of clinical efficacy. A decision to use an intervention with very low certainty evidence may still be clinically justified when the need is urgent, no higher-certainty alternative exists, and the intervention has established theoretical rationale and favorable clinical consensus. Complementary profiles of pharmacotherapy and psychotherapy for depression Viswanathan et al.’s findings revealed a clinically meaningful differential pattern between pharmacotherapy and psychotherapy for adolescent depression. Pharmacotherapy (escitalopram and fluoxetine) demonstrated clear effectiveness in improving clinician-rated depressive symptoms (CDRS-R: MD = − 3.76, moderate certainty) but did not significantly improve remission rates. Conversely, psychotherapy (CBT) significantly improved remission from depression (RR = 1.73, low certainty) but showed inconsistent results across different symptom measures. This complementary pattern suggests that pharmacotherapy and psychotherapy may operate through different therapeutic mechanisms—with pharmacotherapy providing more rapid and measurable symptom reduction on standardized scales, and psychotherapy facilitating deeper cognitive and behavioral changes that support diagnostic-level recovery [39]. This interpretation further reinforces the rationale for combination treatment, consistent with the high certainty evidence for fluoxetine plus CBT identified in this review. Comparison with previous umbrella reviews The findings of this umbrella review both align with and extend the conclusions of previous umbrella reviews in this field. Torralba-Suárez et al. (2023) [13], who restricted their review to CBT and DBT, concluded that both therapies showed positive results for suicide prevention but noted high heterogeneity across reviews. Our findings are consistent with this assessment while extending the evidence base to include family-based therapies, MBT-A, pharmacotherapy, and combination treatments. The present review also identified the evidence−practice gap for specific psychotherapy models that was not explicitly addressed in Torralba-Suárez et al.’s review. Richardson et al. (2024) [14] and McEvoy et al. (2023) [15], both of whom focused on risk and protective factors for self-harm and suicidality, identified psychiatric disorders, prior self-harm, and family dysfunction as key risk factors. Although these reviews did not directly evaluate treatment effectiveness, their identification of modifiable risk factors aligns with the therapeutic targets of the interventions assessed in our review—particularly family-based interventions targeting relational dysfunction and CBT addressing cognitive vulnerability. Compared with the umbrella review by Quinlivan et al. (2025) [39] examining psychosocial interventions for self-harm in inpatient settings, our review adds evidence specific to adolescent populations across diverse settings and includes pharmacotherapy and combination treatment evidence not covered in their review. Their findings that CBT reduces repeat self-harm and DBT decreases self-harm frequency are broadly consistent with Kothgassner et al.’s findings, although our GRADE assessment suggests more caution regarding the certainty of evidence for specific modalities. A unique contribution of the present review is the systematic application of GRADE assessments to both quantitative and narrative systematic reviews, enabling a direct comparison of evidence certainty across methodologically diverse reviews. This approach revealed that reviews employing meta-analysis consistently achieved higher certainty ratings than those relying on narrative synthesis, suggesting that future evidence synthesis efforts should prioritize quantitative methods where feasible. Implications for clinical practice in South Korea Evidence-based recommendations Based on the findings of this umbrella review, several evidence-based recommendations are proposed for clinical practice guidelines in South Korea. The strongest recommendation can be made for fluoxetine plus CBT combination therapy as a first-line treatment for adolescents with depression and co-occurring suicidal risk. The high certainty evidence for improved treatment response and reduced recurrence, combined with moderate certainty evidence for reduced adverse reactions, provides a robust foundation for this recommendation. This is particularly relevant in the Korean context, where depression is the most common psychiatric comorbidity among adolescents with suicidal behavior [6]. A conditional recommendation can be made for structured psychotherapy (particularly DBT-A and CBT-based approaches) as an intervention for self-harm and suicidal ideation in adolescents. While the evidence is mixed across reviews, the balance of evidence—including high certainty evidence from Kothgassner et al. for self-harm reduction and moderate certainty for suicidal ideation reduction—supports the use of these interventions in clinical settings, particularly for adolescents with documented self-harm history. For pharmacotherapy (escitalopram and fluoxetine) to address adolescent depression, the moderate certainty evidence for symptom improvement supports its use, with the caveat that symptom reduction on standardized scales does not necessarily translate to diagnostic remission. Close monitoring of suicidality during the initial phase of pharmacotherapy remains essential, consistent with existing black box warnings [40]. Considerations for the korean healthcare context Several contextual factors must be considered when translating these findings into Korean clinical practice guidelines. First, the vast majority of the evidence base is derived from Western populations, and the cultural specificity of psychotherapeutic interventions—particularly those involving family dynamics, communication styles, and therapeutic alliance—may require adaptation for Korean adolescents and families [41, 42]. The Confucian cultural values that shape family relationships in Korea, including hierarchical family structures and an emphasis on academic achievement, may both facilitate (through family engagement) and complicate (through reluctance to discuss emotional distress openly) the delivery of family-based and individual psychotherapies. Second, the structure of the Korean healthcare system, including the coverage policies of the national health insurance system for psychotherapy, may influence the feasibility of delivering manualized psychotherapy protocols such as DBT-A or CBT. Currently, access to structured psychotherapy for adolescents varies considerably between urban and rural settings, and the availability of trained therapists may represent a significant implementation barrier [43]. Third, the unique epidemiological profile of adolescent suicide in South Korea—including the disproportionately high female adolescent suicide rate compared with other OECD countries and the strong association with academic stress and cyberbullying—suggests that intervention strategies may need to address culturally specific risk factors that are not fully captured in the international evidence base [25]. In this regard, Baldini et al. (2025) recently proposed a comprehensive, multi-sectoral framework for adolescent suicide prevention that integrates universal screening, targeted psychological interventions, policy reform, and digital mental health tools within a culturally sensitive approach [44]. This framework resonates with the Korean context, where coordinated action across healthcare systems, schools, and communities—anchored by culturally adapted interventions—is particularly critical given the unique socioecological risk factors facing Korean adolescents. Implications for future research Several critical research gaps were identified through this umbrella review. First, adequately powered RCTs with suicidal behavior (suicide attempts, suicidal ideation) as primary outcomes are urgently needed. The current evidence base predominantly derives suicidality data from secondary endpoints in depression treatment trials, limiting the specificity and power of these findings. Second, head-to-head comparisons between specific psychotherapy models (e.g., DBT-A vs. CBT, DBT-A vs. MBT-A) using standardized outcome measures and adequate sample sizes are needed to resolve the current uncertainty regarding differential effectiveness. Johnstone et al.’s comparison of DBT-A and MBT-A, although valuable, was limited by the small number of MBT-A studies and the predominance of non-randomized designs. Third, research specifically addressing the effectiveness of suicide and self-harm interventions in Korean and East Asian adolescent populations is critically needed. None of the included reviews specifically examined outcomes in Korean populations, and the applicability of Western-developed interventions to the Korean context remains empirically untested. Fourth, long-term follow-up studies are needed to assess the durability of treatment effects. Most studies included in the reviews assessed outcomes at relatively short follow-up periods, and the sustainability of treatment gains beyond 12 months remains poorly understood. Fifth, future systematic reviews in this field should adopt more rigorous methodological standards, including pre-registration of review protocols, comprehensive publication bias assessments, and standardized outcome definitions. The finding that five of the six reviews in this umbrella review failed to adequately assess publication bias is a significant concern that should be addressed in future evidence synthesis efforts. Strengths and limitations Strengths This umbrella review has several methodological strengths. First, it is the first umbrella review to comprehensively examine the full spectrum of therapeutic interventions—psychotherapy, pharmacotherapy, and combination treatment—for adolescent suicidal behavior and self-harm, with explicit application to the development of Korean clinical practice guidelines. Previous umbrella reviews were either restricted to specific modalities (e.g., CBT and DBT only) or focused on risk and protective factors rather than treatment effectiveness. Second, the systematic application of GRADE assessments to both quantitative and narrative reviews provides a standardized and transparent evaluation of evidence certainty that enables direct comparisons across methodologically diverse reviews. The application of Murad et al.’s (2017) framework for GRADE assessment in the absence of pooled estimates [28] represents a methodologically sound approach for evaluating narrative reviews within the GRADE framework. Third, the use of AMSTAR-2 for quality assessment by two independent reviewers, including detailed item-by-item reporting with supporting evidence from each review, enhanced the transparency and reproducibility of the quality assessment process. Fourth, the review scope—covering 2015–2024 across five databases—ensured comprehensive coverage of the most recent evidence while maintaining a manageable and focused evidence base. Limitations Several study limitations must be acknowledged. First, the final inclusion of only six systematic reviews may limit the comprehensiveness of the evidence synthesis. While this number is consistent with published umbrella reviews in similarly specialized fields, it constrains the ability to draw conclusions for intervention−outcome combinations that were addressed by only one or two reviews. Second, a substantial proportion of the included systematic reviews did not distinguish between NSSI and suicidal behavior (suicidal ideation and suicide attempts) in their outcome reporting, combining these conceptually distinct constructs into a single composite measure. This limits the ability of the present umbrella review to draw separate conclusions regarding the differential effects of interventions on NSSI versus suicidal behavior, despite the conceptual distinction emphasized in the introduction of this review. Third, the substantial heterogeneity among the included reviews—in terms of populations, interventions, comparators, outcomes, and study designs—precluded a quantitative meta-analytic synthesis at the umbrella review level. While the narrative synthesis approach adopted is appropriate for this level of heterogeneity, it introduces a degree of subjectivity in the interpretation and weighting of the findings. Fourth, language restriction to English-language publications may have excluded relevant systematic reviews published in Korean or other languages, potentially omitting evidence from Korean and East Asian populations that would be directly applicable to the development of Korean clinical practice guidelines. Fifth, the GRADE assessments conducted in this umbrella review represent a secondary assessment applied at the umbrella review level, which may differ from the assessments that would be obtained if GRADE were applied directly to the primary studies. Therefore, certainty ratings should be interpreted as reflecting confidence in the evidence as synthesized and reported by the original systematic reviews, rather than as definitive assessments of the underlying primary evidence. Sixth, the search was restricted to English-language international databases and did not include Korean-language bibliographic databases such as the Research Information Sharing Service (RISS), Korean studies Information Service System (KISS), or Korean Citation Index (KCI). Given that this review explicitly aims to inform Korean clinical practice guidelines, the exclusion of domestic databases represents a meaningful limitation. Korean-language primary studies and grey literature that may capture context-specific intervention outcomes remain unexamined, and future guideline development efforts should incorporate systematic searches of these resources. Seventh, the citation overlap matrix revealed moderate overlap (CCA = 0.069) across included reviews, with the highest overlap observed between Viswanathan et al. (2022) and Kothgassner et al. (2020), which shared 12 primary studies. While this level of overlap is within an acceptable range for umbrella reviews, some degree of double-counting of evidence—particularly from RCTs evaluating psychotherapeutic interventions—cannot be excluded and should be considered when interpreting the synthesized findings. Furthermore, several conclusions regarding specific psychotherapy modalities, including DBT-A and MBT-A, rest on one or two underlying reviews, and should therefore be interpreted with appropriate caution. The synthesis findings represent the best available evidence rather than definitive conclusions. This umbrella review provides a comprehensive synthesis of the available evidence on therapeutic interventions targeting suicidal behavior and self-harm in adolescents and young adults. The strongest evidence supports fluoxetine plus CBT combination therapy for adolescents with depression, with high-certainty evidence for improved treatment response and reduced recurrence. Psychotherapy demonstrates promising but variable effectiveness for reducing self-harm and suicidal ideation, with evidence certainty ranging from high to very low depending on the specific review and outcome measure. Specific psychotherapy models widely adopted in clinical practice (DBT-A, MBT-A, ABFT) are currently supported by only very low-certainty evidence, highlighting an urgent need for higher-quality primary research. For the development of clinical practice guidelines in South Korea, these findings support a stepped-care approach in which combination therapy serves as the cornerstone for adolescents with depression and suicide risk, supplemented by structured psychotherapy (particularly DBT-A and CBT-based approaches) for self-harm reduction. A proposed evidence-informed stepped-care algorithm for clinical implementation in Korean settings is presented in Fig. 3. Evidence-informed stepped-care algorithm for adolescent suicidal behavior and self-harm intervention in Korean clinical practice The translation of this evidence into Korean clinical practice will require careful consideration of cultural adaptation, healthcare system feasibility, and the development of Korea-specific outcomes research. Given that South Korea has the highest suicide rate among OECD countries, with a disproportionately increasing trend among adolescents, the establishment of evidence-based clinical practice guidelines is not merely an academic exercise but a public health imperative. Acknowledgements The authors gratefully acknowledge the assistance of Yujin Kwon, Medical Librarian at Ajou University Medical Center, for expert guidance in developing and refining the systematic search strategy across multiple databases. Abbreviations American Academy of Child and Adolescent Psychiatry American Academy of Pediatrics Attachment-Based Family Therapy Acceptance and Commitment Therapy A Measurement Tool to Assess Systematic Reviews 2 American Psychiatric Association Beck Depression Inventory-II Borderline Personality Disorder Cognitive Behavioral Therapy Children’s Depression Rating Scale-Revised Children’s Global Assessment Scale Clinical Global Impression Dialectical Behavior Therapy Dialectical Behavior Therapy for Adolescents Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision Enhanced Usual Care Food and Drug Administration Grading of Recommendations Assessment, Development and Evaluation Hamilton Depression Rating Scale Human Development Index Health of the Nation Outcome Scales for Children and Adolescents Interpersonal Psychotherapy Joanna Briggs Institute Mentalization-Based Treatment for Adolescents Major Depressive Disorder Multisystemic Therapy Non-Suicidal Self-Injury Organisation for Economic Co-operation and Development Population, Intervention, Comparison, Outcome Population, Intervention, Comparator, Outcome, Study Design Preferred Reporting Items for Systematic Reviews and Meta-Analyses Preferred Reporting Items for an Overview of Systematic Reviews Randomized Controlled Trial Risk of Bias Risk Ratio Substance Abuse and Mental Health Services Administration Suicidal Ideation Questionnaire Suicidal Ideation Questionnaire – Junior Standardized Mean Difference Selective Serotonin Reuptake Inhibitor Treatment for Adolescents with Depression Study Treatment As Usual United States Preventive Services Task Force Department of Veterans Affairs / Department of Defense Author contributions S.L. conceived and designed the study, developed the review protocol, conducted the systematic literature search, performed independent screening, full-text eligibility assessment, data extraction, AMSTAR-2 quality appraisal, GRADE evidence evaluation, and data synthesis, and drafted and revised the manuscript. J.S. independently performed title and abstract screening, full-text eligibility assessment, and data extraction, conducted independent AMSTAR-2 assessment, and resolved discrepancies through consensus discussion. S.S. provided methodological expertise in GRADE evidence grading and evidence synthesis, contributed to the interpretation of findings and development of clinical recommendations, and critically revised the manuscript for important intellectual content. Y.S. acquired funding, supervised the overall study, provided clinical expertise in adolescent psychiatry, contributed to the contextual interpretation of findings for Korean clinical practice and the development of the stepped-care algorithm, and reviewed the manuscript for clinical accuracy. All authors read and approved the final manuscript. Funding This research was supported by a grant of the Mental Health related Social Problem Solving Project, Ministry of Health & Welfare, Republic of Korea (grant number : RS-2024-00409433). Data availability All data generated or analysed during this study are included in this published article. The datasets supporting the conclusions of this article are derived from previously published systematic reviews, which are cited in the reference list. Declarations Ethics approval and consent to participate Not applicable. This study is an umbrella review of previously published systematic reviews and did not involve the collection of primary data from human participants, human tissue, or animal subjects. No ethics approval or participant consent was required. Consent for publication Not applicable. This manuscript does not contain data from any individual person. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement All data generated or analysed during this study are included in this published article. The datasets supporting the conclusions of this article are derived from previously published systematic reviews, which are cited in the reference list.