Methods: We conducted a systematic review and network meta-analysis (NMA). Medline, Embase, PsycINFO and Cochrane Central Register of Controlled trials were searched to 4th April 2018. Educational setting-based, universal or targeted, interventions for prevention of anxiety and/or depression in CYP aged 4-18 were included. Interventions aimed at promoting mental health, positive psychology or emotional well-being were excluded. Randomised and quasi-randomised, passive- and active-controlled studies were included. Main outcomes were post-intervention self-report anxiety and depression, wellbeing, suicidal ideation/self-harm and inequalities. We assessed risk of bias following the Cochrane Handbook for Systematic Reviews. We estimated standardised mean differences (SMD) using random effects NMA conducted in a Bayesian framework. The study is registered with PROPSERO CRD42016048184.
Findings: 1511 full-text articles were independently screened for inclusion by two reviewers, from which 137 studies of 56,620 participants were included. Only 20 studies were assessed as being at low risk of bias for both random sequence generation and allocation concealment. There was weak evidence that cognitive behavioural interventions may reduce anxiety in primary and secondary settings. In universal secondary settings mindfulness/relaxation-based interventions (SMD -0.65, [95% CrI -1.14 to -0.19], showed a reduction in anxiety symptoms relative to usual curriculum. There was no evidence that any one type of intervention was effective for depression in universal or targeted, primary or secondary settings. Comparison-adjusted funnel plots suggest the presence of small-study effects for the universal secondary anxiety analysis.
Interpretation: Considering unclear risk of bias and likely small study effects for anxiety, we conclude there is little evidence that educational setting-based interventions addressing the prevention of mental disorders are effective. These results contrast with other recently published meta-analyses. Post-hoc sensitivity analyses suggest this may be driven by differential control group effects masked by ‘lumping’ in pairwise meta-analysis. Future research could consider multi-level, systems-based interventions as an alternative to the ‘downstream’ interventions considered here.