Introduction

Suicidal and self-injurious related behaviors (SSIRBs) is a devastating global health problem, particularly among patients with psychiatric disorders (PDs) [1, 2]. One or more than one in three patients diagnosed with mood disorders have attempted to suicide [3]. As a leading cause of adolescent death [4], approximately 11 to 43 percent of adolescents with PDs have been diagnosed with SSIRBs, such as non-suicidal self-injury (NSSI), suicidal ideation (SI), self-injurious behavior (SIB), suicidality and suicide attempts (SA), etc [5,6,7,8]. SSIRBs invariably lead to negative outcomes in adolescents with PDs [9], including alcohol and drug abuse [10], cognitive impairments, poor interpersonal relationships [11], and violent crimes [12], which even increase the medical burden [13]. The 2019 Global Burden of Disease Study found that self-harm contributes to 319.6 years of life lost, per one hundred thousand population [14]. PDs may contribute to the occurrence of SSIRBs [10, 15]. A systematic review reported that 58% of Chinese adolescents with major depressive disorder (MDD) had NSSI [16]. The suicide rate among patients with mood disorders was approximately 6–10%, 10 times higher than that of non-psychiatric patients [17, 18]. As the most extreme manifestation of PDs, SSIRBs not only increases the risk of PDs, but also aggravates the severity of PDs [19, 20]. The rapid socialization process, the distinct traditional Chinese culture and the highly unbalanced distribution of treatment resources have a certain influence on the occurrence of SSIRBs in Chinese adolescents with PDs [21,22,23,24]. Considering the above conditions, it is urgent to explore effective interventions for Chinese adolescents who experienced SSIRBs and PDs [3].

Early studies have shown that interventions for adolescents with PDs affected by SSIRBs can alleviate symptoms and reduce risk. For example, dialectical behavior therapy (DBT) showed positive improvements in emotional dysregulation, depression, and symptoms related to suicidal and self-injurious behaviors among adolescents with borderline personality disorders (BPD) and SSIRBs [25]. In addition, intermittent theta burst stimulation has been shown to reduce SI in adolescents with MDD [26]. On the other hand, anti-suicidal effects of medications (e.g. ketamine) have also been observed in adolescents with MDD and SSIRBs [27,28,29]. Overall, a variety of effective interventions were available for adolescents with PDs and SSIRBs. Numerous reviews and meta-analyses on individual interventions have been published [30,31,32,33]. However, there have been few reviews of drug therapy for PDs adolescents with SSIRBs. For example, a meta-analysis among adolescents reported that family therapy could significantly improve the outcome of SI rather than depression [30]. Another review showed that DBT was effective in simultaneously improving NSSI and depression in adolescents [31]. Also, two articles indicated the effectiveness of repetitive transcranial magnetic stimulation (rTMS) in treating MDD with SI [32, 34].

To date, two comprehensive systematic reviews have summarized the effectiveness of psychosocial interventions for NSSI or SSIRBs. Lu JJ et al. found cognitive behavioral therapy (CBT) to be the most common among psychosocial therapy, which is effective for SSIRBs in Chinese adolescents [35]. Qu DY et al. summarized the prevalence, risk factors, and interventions of NSSI among Chinese adolescents in a scoping review [1]. However, several considerations need to be made. First, more comprehensive databases and more precise search strategies should be used; Second, the existing systematic reviews targeting Chinese adolescents focused only on NSSI or SSIRBs but ignored comorbid PDs; Third, drug and physical therapies were not included.

Notably, this is the first study to examine a comprehensive systematic review and meta-analysis of interventions for SSIRBs in Chinese adolescents with PDs. Our study aims to systematically summarize the interventions (i.e., medication, physical therapy, psychosocial therapy) for SSIRBs in Chinese adolescents with PDs. This endeavor has the potential to develop more meaningful strategies for the treatment of SSIRBs associated with PDs. It is proving particularly valuable in promoting the integration of intervention methods into clinical practice and guiding the improvement of clinical guidelines.

Methods

This meta-analysis was pre-registered in the International Platform of Registered Systematic Review and Meta-analysis Protocols (INPLASY; registration number: 202350069) and conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.

Eligibility criteria and measurement

In accordance with the PICOS tool, the inclusion criteria were defined as follows: Participants (P): Chinese with PDs (e.g., MDD) and SSIRBs (e.g., NSSI, SI, SA); adolescents aged 18 years old and below or the sum of average age and SD ≤ 18 years old [36]. Intervention (I): psychosocial therapy (e.g., CBT); pharmacological therapy (e.g., antidepressants); physical therapy (e.g., rTMS). Comparison (C): no-treatment control or active control. Outcomes (O): effectiveness; and Study design (S): qualitative studies (QSs), randomized controlled trials (RCTs), clinical controlled trials (CCTs), pre-post studies, case reports (CRs). No language limitation was conducted. The primary outcomes included the average score and standard deviation (SD) derived from assessments using the SSIRBs scale, such as the Ottawa self-injury inventory (OSI). Secondary outcomes were the rate of adverse reactions (ADRs), effective rate, as well as mean and SD of scores on other symptom scales, such as self-rating depression scale (SDS); self-rating anxiety scale (SAS), etc.

Search strategy and selection criteria

Literature searching in twelve databases (PubMed, CINAHL, ScienceDirect, PsycINFO, EMBASE, Cochrane Library, Clinical Trial, Web of Science, CEPS, SinoMed, Wanfang and CNKI) was independently carried out by two groups of reviewers (group 1: J.J.L., J.H. and W.T.G.; group 2: Z.X.W., N.Y., Y.B.L. and J.X.G.). The literature searching was conducted from inception to January 31, 2023 and an updating was from January 31 to September 17, 2023. Followed by a review [37], subject and free terms were used: (“auto mutilat*“OR “cutt*“ OR “headbang*“ OR “overdos*“ OR “selfdestruct*“ OR “selfharm*“ OR “selfimmolat*“ OR “selfinflict*“ OR “selfinjur*“ OR “selfpoison*“ OR “suicid*“ OR “suicide, attempted” OR “suicidal ideation”) AND (“adolescent” OR “teen” OR “youth” OR “teenager”) AND (“China” OR “Chinese”). The search terms used for databases were recorded in the Supplementary Figs. 21–29. The titles and abstracts were independently reviewed and the full texts of relevant publications were scrutinized by the same two groups of reviewers. Any inconsistencies were resolved through consultation with a senior reviewer (WIP.P.). Additional studies were identified through manual search among citations in the included articles, previous systematic reviews, and meta-analyses [31, 38,39,40,41]. Moreover, we also searched conference papers from the 21st National Conference on Psychiatry and the 17th National Conference on Child and Adolescent Psychiatry of the Chinese Medical Association [42, 43].

Data extraction

Relevant data was independently extracted by two groups of reviewers based on a predesigned Excel data collection sheet. Data included: first author, year of survey and publication, survey province, study type, sampling method, sample size, types of interventions in the control and experimental groups, parameters of drug therapy and physical therapy, setting, intervention duration, types of SSIRBs, types of PDs, age range, mean and SD of participants age, number and proportion of males, definitions of SSIRBs and PDs, and measurements. According to a classification of psychological interventions [44], a new set of psychological interventions were defined as ten categories, including CBT, relationship-based interventions, systemic interventions, psychoeducation, group work with children, psychotherapy, counselling, peer mentoring, intensive service models, and activity-based therapies. Two reviewers (KIG.L. and W.W.R.) independently confirmed the accuracy of the data through a double-check process. Discrepancies were resolved through consultation with an additional reviewer (WIP.P.).

Quality assessment

RCTs were evaluated by the Jadad scale [45]. The overall score varied between 0 and 5 points. The Jadad score of 2 or lower was categorized as low quality, while those with 3 or higher were classified as high quality. CCT studies (0–16 points) and pre-post studies (0–12 points) were assessed using two different version of the National Heart, Lung, and Blood Institute (NHLBI) tailored quality assessment tool [46], which was widely used in previous systematic reviews [47,48,49]. The Critical Appraisal Skills Programme (CASP) of qualitative studies checklist was used (0–10 points) [50], which was commonly found in some early studies [51,52,53], while the JBI Critical Appraisal Checklist was utilized for CRs (0–8 points) [54]. Higher scores denoted superior reporting. The evidence level for primary and secondary outcomes was assessed using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) methodology. According to the evaluation rules, all outcomes can be classified into four categories: very low, low, moderate, or high [55, 56]. Also, the quality of our review was assessed using A MeaSurement Tool to Assess systematic Reviews, version 2 (AMSTAR 2) checklist [57], which included 16 items; each item is given one point if the criterion is met, or a zero point if the criterion is not met, is unclear, or is not applicable. Finally, a total score was categorized into four levels: critical low (0–4 points), low (5–8 points), moderate (9–11 points), and high (12–16 points) [58, 59]. Study quality was independently assessed by two reviewers (W.W.R. and J.J.L.). Discrepancies were resolved through consultation with an additional reviewer (KIG.L.).

Statistical analysis

Qualitative synthesis

The study and intervention characteristics, and outcomes of the effectiveness for multiple interventions were synthesized. Combination therapies are defined as the utilization of two or more different types of interventions, whereas physical therapy, psychosocial therapy and drug therapy were considered as a single intervention.

Meta-analysis

Due to the limited number of included studies, two or more RCT articles with the same characteristics were considered for the meta-analysis: 1. patients with SSIRBs; 2. patients with PDs; 3. relevant assessments. Without involving any comparative intervention, a no-treatment control is served as a neutral comparison for study groups receiving the treatment under investigation. In contrast, the active control involves the integration of a proven intervention into the control group, which is compared with an experimental treatment. Considering the differences in sampling methods, demographic profiles, and assessment tools across studies, the symptom estimates (e.g., self-injurious behavior score, depression score) were presented as standardized mean differences (SMDs) and rate [e.g., effective rate and incidence rate of ADRs] with risk ratio (RRs) and their 95% confidence intervals (CIs) by using the Der Simonian and Laird random-effects model [60]. Between-study heterogeneity was estimated using Cochran’s Q test and the I2 statistic, with an I2 ≥ 50% or Cochran’s Q of p < 0.05 indicating significant heterogeneity [61]. The Egger’s test, Begg’s test and the trim-and-fill method were used to assess publication bias when the number of literature was more than two [62]. The significance level was set at 0.05 (two-tailed). Sensitivity analyses were performed to examine the outlier studies. Meta-analyses were performed using Comprehensive Meta-Analysis software, Version 2.

Results

Study characteristics

A total of 20,926 articles were retrieved from the databases and other sources. Following the removal of duplicate records and the use of automation tools for preliminary exclusion, 10,976 records would be used for screening at the first stage. Through initially reviewing the titles and abstracts, 716 records were identified and selected for full-text retrieval for the second stage of screening. With full-text screening, fifty-two articles contained fifty-three studies with 3709 participants (experimental group = 2034 adolescents vs control group = 1675 adolescents) (Fig. 1) [43, 63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112]. Out of ten articles from international databases, forty-two literatures were from Chinese databases. Twenty-eight (28/53, 53.8%) referred to NSSI. There was no eligible literature from Hong Kong and Macao, with the exception of one study from Taiwan. The majority of studies were distributed in Mainland China (Fig. 2a), mainly in coastal areas and publications showed a decreasing trend from coastal to inland areas. Generally, approximately 55% of therapies (29 in 53) were combination therapies. Physical therapy (e.g., rTMS and ECT) was included in 14 studies, while drug therapy (e.g., antidepressants and antipsychotics) was included in 37 of 53 studies. Additionally, 31 studies included seven psychosocial interventions [e.g., intensive service provision (ISP) and CBT] (Fig. 2b). Forty-eight studies were conducted for MDD adolescents, while 2 were conducted for autism spectrum disorders (ASD), 1 for first episode (FE)-BD, and 2 for multi-PDs. Only 11 studies reported medical records with a diagnosis of FE-MDD, with six of those studies utilized combination therapy. The most common therapy both in non-FE-MDD (32.4%) and in FE-MDD (36.4%) was pharmacological combined psychosocial therapy with no significant difference between two groups (X2 < 0.001, P = 1.000). Besides, physical therapy was exclusively employed as a monotherapy in non-FE-MDD, but not observed in patients with FE-MDD.

Fig. 1: PRISMA flow diagram.
figure 1

Note: A database search was combined with an independent search, yielding a total of 52 studies included in the review.

Full size image
Fig. 2: Visualization of intervention types and locations.
figure 2

a Provincial distribution of 51 included studies (52 included reports). Note: There is a study without reporting experimental address [112], and the distribution map is based on 52 reports from the remaining 51 studies. b Interventions for PDs with SSIRBs. Note: Mix(1) means multiple drugs; Mix(2) means PDs; This figure is presented based on the intervention strategies of the experimental group. ABT Activity-based therapy, AD Antidepressants, AE Antiepileptic, AP Antipsychotic, ASD Autism spectrum disorder, BD Bipolar disorder, CBT Cognitive-behavioral therapy, GWWC Group work with children, ISP Intensive service provision, MDD Major depressive disorder, MS Mood stabilizer, NSSI Nonsuicidal self-injury, PC Psychotherapy, PDs Psychiatric Q6 disorders, PM Pharmacological intervention, PH Physical intervention, PS Psychosocial intervention, PE Psychoeducational intervention, RBI Relationship-based intervention, SIB Self-injurious behavior, SI Suicide ideation, SA Suicide attempts, ST Suicidal tendencies, SSIRBs Suicidal and self-injurious related behaviors.

Full size image

Assessment quality and outcome evidence

Thirty-four RCTs used the Jadad scale, with 29 studies rated as high quality. The main reason for the low quality of the other 5 studies was the inappropriate method of randomization sequence. The quality of 11 CCTs ranged from 5–9 points, and the mean score was 7.2 points. One QS, 4 CRs and 2 pre-post studies were assessed, respectively. Nine studies accounted for more than half of the total score (Table 1). A high quality (12 points) of systematic review was identified by the AMSTAR-2 (Supplementary Table 1). Since all studies we included in the meta-analysis were RCTs, we initially rated four stars. There were some limitations in these RCTs, such as high heterogeneity, risk of bias, and small sample sizes. Given the large effect size, most of the evidence quality were at a medium to low level (Supplementary Table 2).

Table 1 Assessment of study quality characteristics.
Full size table

Major depressive disorder with single behavior

Non-suicidal self-injury

Single therapy

Nine studies were conducted during the COVID-19 epidemic. As of 2021, the number of publications in the next year was about twice as high as in the previous year. Patients were either from inpatient (IP) or outpatient (OP) settings, while 4 studies enrolled subjects without further explanation of the source. The duration of the intervention was between 4 and 12 weeks in eight studies. In four studies, three criteria were used to assess NSSI, namely the ANSAQ, adolescent self-harm scale (ASHS), and the diagnostic and statistical manual of mental disorders -V (DSM-V). In addition, 3 out of 9 studies diagnosed MDD in hospitals, 4 studies used 3 types of indicators, with 1 study using the international classification of diseases-10 (ICD-10), 1 study using the Chinese classification of mental disorders-3 (CCMD-3), and 2 studies using the DSM-V. Others used the mini international neuropsychiatric interview for children and adolescents (MINI-kid) and a guideline [113], respectively. Seven RCTs were high quality (mean score = 3), while one CCT study was rated as 7 and one pre-post study was rated as 8.

In eight studies, the scores of the Hamilton depression scale (HAMD) or a reduced rate were used to assess the effectiveness of the interventions. The remaining studies used the self-rating questionnaire for adolescent (SQAPMPU) and the screen for child anxiety related-emotional disorders (SCARED) to assess depressive and anxious symptoms, respectively. Only one study used transcranial direct current stimulation (tDCS) to the dorsolateral prefrontal cortex (DLPFC), with pseudo-stimulation as control. Compared to pre-intervention, the HAMD-17 score and ASHS decreased significantly (P < 0.05). And the effectiveness was greater in the tDCS group, and no ADR was reported.

Six of the nine studies focused on pharmacotherapy, which was a common approach. Apart from one no-treatment control, 2 studies with sertraline and 3 studies with sodium valproate (SV) were routinely considered as control groups. Two studies used magnesium valproate (MV) and sertraline, respectively, and no ADR was reported. However, four studies reported the occurrence of nausea and weakness after taking quetiapine fumarate (QF). In four studies, the risk of self-injury decreased significantly after the intervention (P < 0.05). Six studies reported a significant alleviation of depression, anxiety, and impulsivity after the intervention using HAMD-17, symptom checklist-90 (SCL-90), etc. (P < 0.05).

Two of nine studies were psychosocial interventions that could be assigned to more than one psychosocial category. CBT was flexibly combined with ISP, a psychoeducational intervention, and no ADR was reported. Zhu P et al. indicated a significant difference in the reduction rate of NSSI (experimental group: 14.29% vs control group: 46.51%, P < 0.01). The significant improvement was also seen in the adolescents’ depressive symptoms, self-efficacy, and life satisfaction (all P < 0.05). Lu HL et al. also reported a similar improvement effect in the teenagers’ mobile phone dependence, anxiety symptoms, and depressive symptoms (all P < 0.05).

Combined therapies

Seventeen studies were published between 2021 and 2023, of which Li HZ et al. conducted a comparison among 3 groups. Twelve high-quality RCTs with ratings between 3 and 5 were included. In addition, three CCTs received an average rating of 7, with one CR scoring 7. The study durations in hospital was between 2 and 12 weeks. Except for five studies that did not report diagnostic criteria, seven studies assessed MDD using DSM-V, while 5 studies used ICD-10. In addition, the functional assessment of self-mutilation (FASM), the self-injury behavior screening scale (SBSS), the OSI, and the ANSAQ were used to assess the risk of self-injury. Furthermore, 6 studies reported ADRs.

Twelve of the seventeen studies were psychosocial therapy in combination with pharmacotherapy, 3 studies used physical and pharmacological combination therapies. And 2 studies includes three types of therapies. Seventeen studies indicated a significant reduction in self-injury or suicide risk, and depression after the intervention. In addition, negative emotions (e.g., impulsivity and anxiety) as well as cognitive functioning and social support also were improved.

Meta-analysis

Compared with no-treatment control group, CBT in combination with antidepressants showed significant benefits for depression (SMD = −1.467; 95%CI = −2.492–−0.442, I2 = 88.39%, P < 0.001) and anxiety (SMD = −2.101; 95%CI = −3.869–−0.333, I2 = 95.18%, P < 0.001). Neither the Egger’s nor the Begg’s-tests (all P-values > 0.05) revealed any publication bias. In particular, the pooled SMD value remained consistent regardless of the exclusion of any individual study (Supplementary Figs. 12–18). Considering that there are two similar CCT studies, we conducted an additional sensitivity analysis to combine CCT and RCT studies and the findings indicated consistent conclusions. Sensitivity analysis found that CBT in combination with antidepressants indicated significant benefits for depression (SMD = −1.436; 95%CI = −2.158–−0.713, I2 = 85.33%, P < 0.001; Supplementary Figs. 19–20). However, Xi Y et al. used two depression scales. When the HAMD was replaced to Montgomery-Asberg depression rating scale (MADRS), the score was still significant (SMD = −1.254; 95%CI = −1.959–−0.550, I2 = 85.32%, P < 0.001). Neither the Egger’s nor the Begg’s-tests (all P-values > 0.05) revealed any publication bias.

However, CBT and antidepressants showed a significant effect compared to the active control on depression (SMD = −0.943; 95%CI = −1.414–−0.472, I2 = 33.38%, P = 0.212) and anxiety (SMD = −0.942; 95%CI = −1.323–−0.560, I2 = 0%, P = 0.592). When the HAMD-24 was replaced by the PHQ-9, there was still a significant difference, namely for depression (SMD = −0.905; 95%CI = −1.361–−0.450, I2 = 29.44%, P = 0.236). In addition, Peng HZ et al. replaced the HAMA-14 with generalized anxiety disorder scale-7, and found a similar result for anxiety (SMD = −0.971; 95%CI = −1.354–−0.589, P < 0.05, I2 = 0%, P = 0.592). Furthermore, no publication bias was detected by the Egger’s and the Begg’s-tests (all P-values > 0.05, Table 2). Moreover, significant advantages were found in emotion regulation (SMD = −0.832; 95%CI = −1.342–−0.322, I2 = 0%, P = 0.820).

Table 2 Effectiveness of interventions targeting MDD with NSSI/SIB.
Full size table

In comparison with no-treatment control group, the aggregated SMD value of depressive symptom scores was found [−1.587 (95% CI: −2.505, −0.670), P < 0.05]. Considerable heterogeneity was reported (I2 = 90.45%, P < 0.05), as well as no publication bias (Begg’s test: Z = 1.567, P = 0.117; Egger’s test: Z = 4.746, P = 0.132). The combined SMD of anxiety symptom scores was found [−1.925 (95% CI: −2.700, −1.150), P < 0.05]. The results suggested high heterogeneity (I2 = 85.23%, P = 0.001). Nevertheless, no publication bias was detected (Begg test: Z = 1.567, P = 0.117; Egger test: Z = 1.053, P = 0.484). In addition, the combined RR value for clinical effectiveness was found [−1.204 (95% CI: 1.084, 1.338), P < 0.05]. No heterogeneity was found (I2 = 0%, P = 0.810). However, we found publication bias (Begg’s test: Z = 1.567, P = 0.117; Egger’s test: Z = 17.263, P = 0.037), with an adjusted RR value [1.167 (95% CI: 1.069, 1.273), P < 0.05]. Significant relief of impulsivity was found [−2.439 (95% CI: −2.748, −2.094), P < 0.05], while no heterogeneity was detected (I2 = 0%, P = 0.989). In addition, statistically significant positive effects were observed for SIB [−2.466 (95% CI: −3.305, −1.628), P < 0.05], with heterogeneity (I2 = 88.36%, P = 0.017).

Contrary to our expectations, no significant difference was observed in the incidence rate of ADRs after QF and SV, with nausea/vomiting [1.194 (95% CI: 0.357, 3.992), I2 = 0%, P = 0.668], dizziness/vertigo [0.864 (95% CI: 0.292, 2.556), I2 = 0%, P = 0.704], xerostomia [1.043 (95% CI: 0.352, 3.092), I2 = 0%, P = 0.729]. A publication bias was found in nausea/vomiting (Begg’s test: Z = 1.567, P = 0.117; Egger’s test: Z = 34.540, P = 0.018). In addition, increased BMI [1.217 (95% CI: 0.382, 3.870), P > 0.05] and drowsiness/drowsiness/fatigue [1.150 (95% CI: 0.281, 4.713), P > 0.05] were not detected with significance, nor was heterogeneity (P > 0.05, Table 2). The overall SMD value remained consistent regardless of the exclusion of any individual study (Supplementary Figs. 2–11).

Suicidal ideation

As shown in Table 3, six of the seven studies were conducted between 2017 and 2022, with four published in 2023 and two in 2022. For one study in 2018, the date of the study was not specified. Four studies were IPs and one study was OPs. The mean trial duration was approximately 3.3 weeks and ranged from 1 to 6 weeks. For the assessment of MDD, 7 studies used ICD-10, MINI-kid, DSM-V or DSM-IV, respectively. To define SI, 4 studies used the Beck scale for suicide ideation (BSSI); 1 study used the self-rating idea of suicide scale (SIOSS-26); 1 study used both the Columbia suicide severity rating scale (C-SSRS) and the BSSI; and 1 study used the SI item of HAMD. With an average rating of over 3, 3 RCTs were high quality, while another RCT was rated as 2. In addition, 1 CR was rated as 8, while the other 2 CCTs were rated as 7 and 8, respectively.

Table 3 Characteristics of included studies.
Full size table

With the exception of one study that only used esketamine, the others were combination therapies. The HAMD was widely used to assess the severity of depressive symptoms, while suicide risk was assessed using the BSSI, HAMD-SI and C-SSRS, etc. ECT was applied in 2 studies and resulted in significant improvement in SI and depressive symptoms. In one study, it was observed that high-frequency rTMS showed greater efficacy in the early improvement of MDD than low-frequency rTMS (χ2 = 8.167, P < 0.01). The difference in SI between the two groups was statistically significant (low-frequency group: 36.7% vs. high-frequency group: 63.3%, P < 0.05). No ADR was reported. Similarly, Pan F et al. reported positive effect of high-frequency rTMS for depression and suicide risk, with 2 participants experiencing hypomania. Three adolescents experienced drowsiness after each rTMS but without other subjective side effects.

In comparison with sertraline, DBT was combined with sertraline in two studies, which showed a significant therapeutic effect, and no ADR was reported. The total BSSI score, SI intensity and suicide risk decreased after the intervention (P < 0.01). In addition, two studies showed very similar efficacy on depressive symptoms, with 1 study (experiment vs. control = 89.47% vs. 63.16%, P < 0.05), and the other (experiment vs control = 92.86% vs. 70.00%, P < 0.05).

The use of midazolam was associated with fewer adverse effects, particularly nausea, dissociation, etc., by Zhou YL et al. However, significant differences were observed in the mean changes of C-SSRS scores for ideation and intensity from baseline to day 6 between the esketamine group and the midazolam group.[ideation, 2.6 (SD = 2.0) vs. 1.7 (SD = 2.2), P < 0.01; intensity, 10.6 (SD = 8.4) vs. 5.0 (SD = 7.4), P < 0.01]. The response rates of antidepressants at 4 weeks post-treatment between esketamine and midazolam were 61.5% versus 52.5%. No significant difference in mania symptoms between the two groups was detected (χ2 = 0.384, P > 0.05).

Self-injurious behavior

Five studies have been published in the last three years. Two of the four RCTs were rated as high quality, the other 2 RCTs received 2 each, while the only CCT was rated 9. Three studies used psychosocial therapy alone, while the other two studies used antidepressants with ISP or high frequency rTMS. All studies were conducted in hospital. HAMD was used in 4 studies, while only 1 study used SDS. SIB lasting longer than 5 days or 6 weeks was categorized as SIB in two of five studies, while the others were identified as medical records. In addition, three studies used CCMD-2-R and ICD-10. Overall, both depression and self-injury were alleviated after the intervention. In addition, the combination of ISP and relationship-based intervention resulted in higher adherence compared to the control group.

Meta-analysis

Only the depressive symptoms were summarized in two studies with no-treatment control were aggregated, which included a total of 93 samples. The combined SMD value of depressive symptom score was [−1.647 (95% CI: −2.474, −0.820)]. Significant heterogeneity was observed (I2 = 81.10%, P = 0.021, Table 2). The exclusion of a single study did not change the stability of the aggregated SMD value (Supplementary Fig. 1).

Suicidal tendencies

Four studies were included, three of which had a duration of 2 years and one of which had a duration of 3 years. One study was published in 2013, the others were published in the last 3 years. One study used the DSM-V, while three studies were based on medical records. In the absence of a suitable randomization method, the quality of two RCTs was 2 and 3, respectively. Nevertheless, two CCT studies both scored 7.

Psychosocial therapy was used alone. Wang CL et al. showed significant difference in HAMD and HAMA scores between two groups at both 2 and 4 weeks after the intervention (P < 0.01). Zhu L et al. used cognitive correction and behavioral shaping nursing to intervene with MDD patients with suicidal tendencies. Symptoms, psychological status, and cognitive functioning improved after the interventions(P < 0.05). Shao HH et al. demonstrated the combination of CBT and ISP could significantly reduce depressive symptoms and suicide-related score (P < 0.001). In addition, the combination of CBT and ISP showed significant improvement in quality of life [psychological (χ2 = 14.83, P < 0.001), physical (χ2 = 10.35, P < 0.005), physiological (χ2 = 10.92, P < 0.001) and social functions (χ2 = 15.61, P < 0.001)]. Su M et al. reported that quality of life, sleep quality and depression improved significantly after implementation of the clinical characteristics analysis and nursing strategies (CCANS)(P < 0.05). Specifically, compared with control group, CCANS significantly reduced rates of cutting wrist (9.52% vs. 0%, P < 0.05), jumping (14.29% vs. 4.76%, P < 0.05), poison ingestion (14.29% vs. 4.76%, P < 0.05) and overall suicidal rate (38.09% vs. 9.52%, P < 0.05). Importantly, no ADR was reported in these 4 studies.

Suicidality

Quan LJ et al. conducted a high quality RCT of FE-MDD, with sensory integration therapy and sertraline in 2020. The study lasted a total of 12 weeks and several evaluations were conducted. A significant difference in ISI score was found between 2 groups after the intervention (experimental:4.52 ± 1.02 vs. control: 5.27 ± 1.06, P < 0.01). The positive number of SI decreased after sensory integration therapy (baseline vs. week 4 vs. week 8 vs. week 12 = 40 vs. 24 vs. 15 vs 5), with a significant difference found in total ADRs, nausea (experimental: 2/40 vs. control: 4/40), drowsiness (experimental: 1/40 vs. control: 3/40), and dizzy (experimental: 3/40 vs. control: 5/40).

Major depressive disorder with multi-behaviors

Five studies applied a single therapy for MDD with multiple behaviors, including 3 of 5 simultaneously studied SSIRBs. Five different scales were used to assess the severity of SSIRBs, including BSSI, SIOSS, clinical global impression scale severity (CGI-S), suicide-visual analog scale (SVAS). Due to the high dropout rate, one CCT was rated as 5 [102]. In addition, the lack of rigorous double-blind studies and randomization methods were the main reasons for the low quality (2 vs. 3) in two RCTs. One pre-post study was rated as 8, while one CR was rated as 7 and one CCT was rated as 5. Four out of five studies were conducted in hospital. The mean trial duration ranged from 10 days to 2 months.

Non-major depressive disorder with SSIRBs

Five studies were conducted: two for ASD with SIB, one for FE-BD with NSSI, one for PDs with SI and one for PDs with SIB. Two of five studies used DSM-V and DSM-IV. On average, one article has been published every two years since the study was established in 2016. The quality of two studies was 9 for QS and 5 for CR, respectively. The other 2 studies were rated as high quality according to the Jadad scale (mean score = 3). One CCT was rated as 8. One study did not specify the source of the sample, the other 4 studies were from IPs or OPs. The average intervention duration was 7.3 weeks, while 2 studies did not specify the duration.

Four out of five studies were single therapy, with only one study using rTMS and SSRIS. Only one study adopted a novel online psychoeducational intervention for MDD adolescents. Duan SQ et al. found that transcripts of semi-structured interviews reduced instances of deliberate self-harm by providing acceptable support to adolescents. One CR reported that traditional Chinese medicine (TCM) and the five elements of music therapy significantly improved the child’s sleep and emotions, and SIB was also alleviated. Li XD et al. found that aripiprazole had a significant effect on alleviating the occurrence of NSSI in FE-BD compared to lithium (week 8: experimental: 1/38 vs. control: 8/38, P < 0.05) with similar incidence rates of ADRs. Only one study used rTMS with different intensities to affect the left or right DLPFC. Good relief effects were observed for SI, with 22 of 29 adolescents recovering. In contrast, a negative correlation was observed between improvements in HAMD total score and HAMD-SI score (r = −0.094, P = 0.629).

Discussion

To the best of our knowledge, this was the inaugural systematic review and meta-analysis to summarize the characteristics of interventions for Chinese PDs adolescents with SSIRBs. Geographically distributed along the coast, all studies were located far from underdeveloped areas, highlighting the uneven distribution of mental health resources in China [23]. Nevertheless, almost all of the literature was published in the past four years, with 2 studies in 2020, 9 studies in 2021, and 19 studies in 2022. Notably, the growth trend of publications indicates a tremendous research enthusiasm during the COVID-19 pandemic [114]. ISP and CBT were the most common psychosocial strategies, while the most commonly used medication was antidepressants. In addition, rTMS was the most common physical therapy.

Given the prevailing global circumstances, each of the three major intervention strategies has its own merits and drawbacks. The Times and the Guardian noted that “antidepressants do more harm than good” and “psychiatric drugs are doing us more harm than good” [115]. And antidepressants have been given a black box warning by the Food and Drug Administration, suggesting that they may have an increased risk of suicide in adolescents with PDs [116]. All phenomena have also had unseen negative effects on drug treatment. As a substitute for drug therapy, psychosocial therapy not only circumvents the potential hazards arising from insufficient evidence of efficacy but also mitigates the occurrence of many ADRs which would be a risk factor for COVID-19 complications [117]. From another perspective, the delayed therapeutic response of psychological therapy was also a recognized disadvantage [92]. As a newly explored intervention, physical therapy has been gradually promoted in recent years, whose advantages are fewer ADRs and rapid response [80, 118]. Therefore, three types of existing interventions were comprehensively integrated in our study, which could facilitate the optimization of resource allocation and the improvement of effectiveness.

Psychosocial intervention

Psychosocial therapy was used in 31 studies, representing 7 of 10 major categories of psychological interventions [44]. Our study suggests that psychosocial therapy was effective, which is consistent with international studies for adolescents with PDs and SSIRBs [11, 30]. However, as the majority of studies involved combination therapies and considered psychosocial therapy as a complementary approach [119,120,121], it is a challenging to precisely determine the source of therapeutic effectiveness [122].

To date, the effectiveness of CBT and antidepressants in PDs or SSIRBs has been widely recognized [123,124,125]. Our study suggests that CBT in combination with antidepressants can alleviate symptoms of depression, anxiety, and difficulty in emotion regulation, which is superior to active control. In addition, our result indicated that the experimental group with a no-treatment control seemed to have a stronger effect. Although it has been suggested in the past that patients with and without antidepressant medication derived similar benefits from CBT in terms of anxiety [126, 127], the combination of CBT and sertraline was more effective in relieving anxiety and depression than either treatment alone [128, 129], which is consistent with our findings. The decrease in NSSI and alleviation of depression in adolescents were reported after DBT, which was considered as comprehensive CBT [31]. Another RCT confirmed the positive efficacy of DBT in combination with medication in reducing SA in adolescents with BD [130]. Similar to their findings, our study also found that CBT with antidepressants could decrease self-injury behaviors score in Chinese adolescents with PDs. Lu JJ and colleagues have reported that CBT is the most commonly used intervention for the treatment of SSIRBs in Chinese adolescents [35]. However, our results suggest that ISP has an almost equivalent status with CBT in adolescents with PDs and comorbid SSIRBs. CBT focuses on an individual’s psychological and behavioral patterns to achieve a transformation personal control [35]. ISP achieves therapeutic goals by providing comprehensive and highly focused support with attachment and object relations [131]. Additionally, the cultivation of skills not reflected in CBT is integrated into this process and includes the development of self-esteem and the navigation of interpersonal relationships [44]. On the other hand, more conflictual relationships were observed in the families of adolescents with PDs and SSIRBs [132]. Ebrahimi et al. applied parent–child interactions and observed a reduction in depressive symptoms [133], which was also found in our study that demonstrated the positive effectiveness against depression in PGPC. In summary, psychosocial intervention is a promising therapeutic approach.

Online interventions are very popular, especially during the COVID-19 pandemic. Buronfosse et al. demonstrated the effects of hotlines in reducing self-aggressive behavior in patients diagnosed with BPD [134]. Effectiveness in suicide prevention and symptoms improvement has also been reported [135,136,137]. Although SMS text messaging interventions were first introduced as a productive novel approach in the treatment of Chinese PDs adolescents with SSIRBs [94], the effectiveness was similar to previous studies, with SMS text messaging interventions showing promising potential due to their cost-effectiveness, low-intensity, and widespread acceptability [94, 135]. The importance of psychosocial therapy is essential that it goes beyond symptom management and addresses the complex interplay of psychological, social, and environmental factors. It is an integral part of the holistic treatment of adolescents struggling with mental health problems associated with SSIRBs.

Physical intervention

In our studies, three types (i.e., ECT, tDCS, and rTMS) of non-invasive brain stimulation (NBS) were applied [138], which were feasible and demonstrated to be effective in Chinese adolescents. Although Bloch and colleagues reported that rTMS treatment did not significantly alleviate the severity of SI in MDD adolescents [139], most of studies have confirmed the role of rTMS in alleviating SI in MDD adults [140, 141], our results supported this effectiveness on Chinese adolescents [72, 77, 79]. The factors leading to the differences could be the effects of comorbidities and previous history of ECT were not excluded by Bloch et al. whose study included only 9 participants. However, there is a recognized mechanism that may explain the efficacy of ECT. Cortical inhibition may be enhanced by rTMS, possibly by modulating GABAB receptor-mediated activity, leading to a reduction in SI among depression patients [77, 142]. A weak positive effect on stress-related emotions was found after tDCS treatment [143]. While tDCS was able to significantly alleviate depressive symptoms in our study, which was also shown by Charvet et al. [144]. Furthermore, the alleviation of suicidal symptoms and depression after ECT was demonstrated in our study, which was also found in previous results [145, 146]. The pathogenesis of PDs with SSIRBs has been linked to neurochemical metabolic processes [147], HPA axis dysfunction, and psychosocial factors [148]. By applying external magnetic fields or electric currents to the brain, NBS induces changes in neuronal excitability, thereby affecting cerebral metabolism and neuronal electrical activity [149], with the aim of alleviating symptoms [64, 68].

Contact dermatitis was observed after tDCS with a parameter control above or below 2 mA [150], while no ADR was observed at a current of 1–2 mA. Appropriate parameters may be an associated factor for ADRs. In addition, physical therapy combined with drug treatment had a lower proportion of ADRs than the group receiving drug treatment alone [92, 103]. Therefore, NBS was one of the options for the treatment of Chinese PDs adolescents with SSIRBs. Furthermore, ECT was traumatizing, while both tDCS and ECT had difficulties in accurately determining the location of the stimulus effect. However, today’s technology can analyze data in real time and automatically adapt the stimuli to the behavioral state of the brain [151]. This can not only improve controllability and safety, but also increase confidence in the treatment. In the future, the combination of artificial intelligence and targeted electrical brain stimulation offers endless possibilities [152, 153].

Pharmacotherapy

This is the inaugural meta-analysis investigating the effectiveness of pharmacotherapy (antipsychotics and antiepileptics, antidepressants and CBT) in adolescents with MDD and NSSI. Our study showed that QF and sodium valproate (SV) had positive effects on the relief of depressive symptoms, anxiety, impulse symptoms, and self-injury symptoms, as well as safety, compared to SV alone.

There is ample evidence in the literature that defects in gamma-aminobutyric acid (GABA) transmission are associated with MDD [154], while the main neurophysiological basis of MDD is related to dopamine [155]. SV not only prevents the degradation of GABA by inhibiting GABA aminotransferase [156], but also increases the activity of glutamic acid decarboxylase [157], which is the rate-limiting enzyme for the synthesis of GABA [158]. Thus, SV increases brain GABA concentrations and modulates the neuronal. QF is an antagonist with moderate affinity for adrenergic a1 and a2 receptors, serotonergic 5-hydroxytryptamine 2 receptors (5-HT2 receptors) and dopaminergic D2 receptors; the affinity for serotonergic 5-HT1A receptors is low [159]. By antagonizing 5-HT2A receptors, acting as a partial agonist of 5-HT1A receptors, and antagonizing α2 adrenoceptors, QF increases the release of dopamine from the prefrontal cortex [160]. Based on the potential mechanisms, SV and QF may have an effect on improving anxiety and depressive symptoms [161]. Furthermore, in practice, SV has been shown to be the first choice for the drug treatment of BD, while QF served as a complementary strategy for PDs due to its safety [161]. Previous studies have reported the role of QF in reducing impulsivity and depression [162, 163], which was also reflected in our results. Our study also found a reduction in self-injury scores after taking QF and SV, whereas other studies have reported a similar role for these drugs [164, 165]. In addition, our study showed that the incidence of ADRs was slightly higher than when taking SV alone, but no significant difference, suggesting that the ADRs caused by low-dose QF were still within an acceptable range. The efficacy, optimal dosage, and compliance of other medications (e.g., ketamine) should be further investigated.

Strengths and limitations

This systematic review and meta-analysis included databases from international and Chinese sources to conduct a comprehensive literature search and use of sophisticated analyses. We described the characteristics of interventions in Chinese PDs adolescents with SSIRBs, and presented meta-analyses for both NSSI and SIB. Our study also included psychosocial, pharmacologic, and physical treatments. In addition, our review also included gray papers from top conferences in Chinese psychiatry.

However, it is important to note some limitations. First, the results should be interpreted with caution due to the small number of studies. Second, except for MDD adolescents with NSSI or SIB, our study did not conduct a meta-analysis of interventions for other SSIRBs in adolescents with PDs due to insufficient data. Third, our study was limited to focusing primarily on interventions and overlooking preventive strategies while using a promising risk calculator for early detection of SA in BD [166]. Fourth, optimal doses and medication compliance were not analyzed due to insufficient data. Finally, comparisons in our study were not possible due to the different components and efficacy of pharmacotherapy in the included studies. Since the articles studied were primarily from Mainland China, the generalizability of these results to other ethnicities may be limited.

Conclusions

This systematic review described the main characteristics, safety and effectiveness of interventions in Chinese PDs adolescents with SSIRBs. Single therapy and combination therapies have shown varying degrees of safety and effectiveness in relieving symptoms. These findings expanded the means and theoretical basis of mental health treatment to provide benefits for future health care utilization and the economy as a whole. Larger extensive, multicenter RCTs with large sample sizes are needed to evaluate efficacy and safety.