Fig. 2: Risk of bias graph.

Review authors’ judgements about each risk of bias item presented as percentages across all included studies. Regarding selection bias, 80% of the included studies had a low risk of bias for random allocation, and almost 50% had low risk of bias for allocation concealment. Sixty-five percent of the studies had a high risk of performance bias, due to the inability to blind participants given to the inherent characteristics of the intervention (digital versus waitlist control or standard therapies), which also influenced blinding for outcome assessment especially on self-reported rating scales (52% of the studies had low risk of detection bias). For over 25% of the studies, there was unclear report on how they handled missing data (attrition bias) and 30% had unclear reporting bias (selective outcome reporting) due to the absence of a registered protocol (Fig. 2). These risks of bias are commonly reported by other meta-analyses analyzing digital therapies. Seven out of the 76 studies (9%) were classified as overall low risk of bias (no bias detected). Risk of bias rating for each study is reported in Supplementary Fig. 1.