Why we need to pursue both universal and targeted prevention to reduce the incidence of affective and psychotic disorders: Systematic review and meta-analysis

Abstract

The effectiveness of universal preventive approaches in reducing the incidence of affective/psychotic disorders is unclear. We therefore aimed to synthesise the available evidence from randomised controlled trials. For studies reporting change in prevalence, we simulated all possible scenarios for the proportion of individuals with the disorder at baseline and at follow-up to exclude them. We then combined these data with studies directly measuring incidence and conducted random effects meta-analysis with relative risk (RR) to estimate the incidence in the intervention group compared to the control group. Eighteen studies (k=21 samples) were included investigating the universal prevention of depression in 66,625 individuals. No studies were available investigating universal prevention on the incidence of bipolar/psychotic disorders. 63 % of simulated scenarios showed a significant preventive effect on reducing the incidence of depression (k=9 - 19, RR=0.75-0.94, 95 %CIs=0.55-0.87,0.93-1.15, p=0.007-0.246) but did not survive sensitivity analyses. There is some limited evidence for the effectiveness of universal interventions for reducing the incidence of depression but not for bipolar/psychotic disorders.

Keywords: Bipolar; Depression; Intervention; Meta-analysis; Prevention; Psychosis; Universal.

Fig. 1

Overview of simulation approach for meta-analysis for the primary outcome. We know the number of individuals with the disorder at baseline, who should be excluded, and while we know the number of individuals with the disorder at follow-up, we do not know if any of these are the same individuals with the disorder at baseline. Therefore, we tested all combinations of the percentage of individuals with a disorder at baseline who have not dropped out of the study and still meet criteria at follow-up. This allowed us to conduct a meta-analysis of incident cases, combined with studies that directly measured incidence, and a relative risk was calculated.

Fig. 2

Study selection.

Fig. 3

Summary of simulation meta-analytic results. As several included studies reported prevalence at baseline and follow-up, rather than incidence, there were some cases where negative within-group changes occurred (due to individuals dropping out of the study or remitting). We were therefore unable to use conventional meta-analysis. Instead, we simulated the proportion of cases at baseline who did not drop out or remit over the course of follow-up. The resulting number of individuals was subtracted from the cases at follow-up. Those remaining were considered incident cases. The proportion of excluded cases who dropped out or remitted is shown on the x-axis. The corresponding relative risk (RR) from that relevant meta-analysis is shown on the y-axis. RRs below 1 indicate that the intervention group had a lower risk of developing the condition than the control group. As impossible scenarios (where there were more cases than those remaining at follow-up) were excluded from each analysis, the number of studies included in the analysis (k) is shown at the top of the graph. The blue dots represent significant preventive effects, the red dots non-significant effects and the line shows the trend modelled with Loess smoothing and standard errors.

Fig. 4

Forest plot for primary sensitivity analysis restricted to studies investigating universal preventive interventions for reducing the incidence of depression.