Assessment of suicidality in trials of psychological interventions for depression: a meta-analysis.

Abstract

Background: Psychological interventions that are efficacious as treatments for depression could indirectly affect suicide-related outcomes. We examined suicidal thoughts and behaviours as eligibility criteria, outcomes, and adverse events across trials of psychotherapy for depression.

Methods: We used a publicly available meta-analytic database developed through systematic searches (updated as of May 1, 2023) to identify randomised controlled trials in which a psychological intervention for depression was compared with an inactive or non-specific control condition in adults with depression and in which any suicide-related outcomes were reported. We also identified studies in which suicide risk was an exclusion criterion. We excluded inpatient studies and trials of unguided digital interventions or collaborative care that included a psychological component. Pairs of reviewers worked independently to select studies and extract data. In a random-effects meta-analysis with robust variance estimation, we assessed the effect of the psychological intervention on suicide outcomes in trials in which suicide was explicitly assessed as an outcome with clinical scales with established psychometric properties. Risk of bias was assessed with the Cochrane risk-of-bias tool (version 2).

Findings: Of the 469 randomised trials we identified in which a psychological intervention was compared with an inactive control in people with depression, 251 excluded people judged at risk of suicide. Any assessment of suicide was included in only 45 trials, 12 of which assessed suicidal ideation or risk as an outcome. These 12 trials included 3930 participants, 2795 (71%) of whom were female and 1135 (29%) of whom were male; data for age and ethnicity were not consistently reported. Psychological interventions for depression were associated with a small reduction in suicidal ideation and risk in 11 trials (one trial reported only follow-up data) after the intervention (standardised mean difference -0·31 [95% CI -0·60 to -0·03]) but not at follow-up (-0·49 [-1·31 to 0·32]). Suicide-related adverse events were reported in 25 trials, and suicide-related serious adverse events (eg, suicide attempts, deaths by suicide) were reported in 13 trials. Heterogeneity was substantial across all analyses, and prediction intervals crossed zero.

Interpretation: Trials of psychological interventions for depression rarely report assessments of suicide. Psychological interventions might reduce suicidal ideation in patients with depression, but more randomised controlled trials are required to clarify this effect. Monitoring and reporting of suicide-related adverse events should be improved in trials of psychological interventions for depression, and future trials should incorporate outcomes related to suicidal thoughts or behaviours.

Funding: None.

Translation: For the Spanish translation of the abstract see Supplementary Materials section.