Twelve RCTs including 2,334 participants were included.
Study quality was reasonable to good. All prevented foreknowledge of treatment assignment. Most outcomes were self-reported. The loss to follow-up ranged from 3 to 34%.
A meta-analysis of all 12 studies showed a significant difference in ES in favour of internet-based CBT compared with control and statistically significant heterogeneity between the studies (random-effects model, ES 0.51, 95% CI: 0.28, 0.74). A pooled analysis of the 2 prevention studies (n=352) showed no significant difference between intervention and control. Heterogeneity between the 10 treatment studies was statistically significant and the ES in favour of internet-based CBT remained significant.
Further subgroup analysis of the treatment intervention studies showed a statistically significant small ES for depression (mixed-effects analysis, ES 0.32, 95% CI: 0.08, 0.57; 4 studies). Heterogeneity between the studies was still significant. The subgroup analysis of anxiety treatment showed a statistically significant large ES in favour of treatment (fixed-effect and mixed-effects analysis, ES 0.96, 95% CI: 0.69, 1.22; 6 studies) and no significant heterogeneity between the studies.
The pooled ES of treatment interventions was large in studies with therapist support (fixed-effect and mixed-effects analysis, ES 1.00, 95% CI: 0.75, 1.24; 5 studies) and small and just reached statistical significance in studies without therapist support (mixed-effects analysis, ES 0.26, 95% CI: 0.08, 0.44; 5 studies).