Is psychotherapy effective for reducing suicide attempt and non-suicidal self-injury rates? Meta-analysis and meta-regression of literature data

Objective: To determine the efficacy of psychotherapy interventions for reducing suicidal attempts (SA) and non-suicidal self-injury (NSSI). Methods: Meta-analysis of randomized controlled trials (RCTs) comparing psychotherapy interventions and treatment as usual (TAU; including also enhanced usual care, psychotropic treatment alone, cognitive remediation, short-term problem-oriented approach, supportive relationship treatment, community treatment by non-behavioral psychotherapy experts, emergency care enhanced by provider education, no treatment) for SA/NSSI. RCTs were extracted from MEDLINE, EMBASE, PsycINFO and Cochrane Library and analyzed using the Cochrane Collaboration Review Manager Software and Comprehensive Meta-analysis. Results: In the 32 included RCTs, 4114 patients were randomly assigned to receive psychotherapy (n = 2106) or TAU (n = 2008). Patients who received psychotherapy were less likely to attempt suicide during the follow-up. The pooled risk difference for SA was 0.08 (95% confidence intervals = 0.04 to -0.11). The absolute risk reduction was 6.59% (psychotherapy: 9.12%; TAU: 15.71%), yielding an estimated number needed to treat of 15. Sensitivity analyses showed that psychotherapy was effective for SA mainly in adults, outpatients, patients with borderline personality disorder, previously and non-previously suicidal patients (heterogeneous variable that included past history of SA, NSSI, deliberate self-harm, imminent suicidal risk or suicidal ideation), long- and short-term therapies, TAU only as a control condition, and mentalization-based treatment (MBT). No evidence of efficacy was found for NSSI, with the exception of MBT. Between-study heterogeneity and publication bias were detected. In the presence of publication bias, the Duval and Tweedie's "trim and fill" method was applied. Conclusion: Psychotherapy seems to be effective for SA treatment. However, trials with lower risk of bias, more homogeneous outcome measures and longer follow-up are needed. (C) 2016 Elsevier Ltd. All rights reserved.