Impact of support levels on effectiveness and drop-out of internet-based interventions for depression: network...

Internet-based interventions vary with respect to the level of support provided, and the impact of support levels on outcomes has been unclear.

To evaluate the relative effectiveness and acceptability of support levels in internet-based cognitive-behavioural therapy (iCBT) for depression.

This network meta-analysis included randomised controlled trials of stand-alone iCBT for adults with elevated levels of depressive symptoms, identified via systematic searches of PubMed, EMBASE, PsycINFO and the Cochrane Library (1 January 2025). The primary outcome was post-intervention effectiveness. The secondary outcome was study drop-out risk. Risk of bias was assessed with Cochrane RoB-2. A frequentist random-effects model was conducted (preregistered at https://osf.io/amw4r).

We included 141 trials with 169 comparisons (n = 32 197). iCBT with therapeutic support had the greatest effect in terms of reducing depressive symptoms compared with care-as-usual (g = 0.42; 95% CI: 0.30 to 0.55). Such interventions outperformed offers with minimal coaching (encouragement only; g = 0.19, 95% CI: 0.03-0.35) and technical support (g = 0.27, 95% CI: 0.08-0.45) but had similar effects to those with full coaching (i.e. standardised feedback), automated support, on-demand support or no support. Interventions providing technical support represented the least effective iCBT format and were not statistically superior to care as usual (g = 0.15, 95% CI: -0.02 to 0.33). For acceptability, iCBT with minimal coaching showed the lowest drop-out rate (risk ratio = 1.13, (95% CI: 0.88-1.46), whereas technical support showed the highest (risk ratio = 1.6, 95% CI: 1.21-2.15). With pre-intervention human contact, all support levels were similarly effective; without it, therapeutic support outperformed other types of support (g = 0.32-0.68) and drop-out risks increased.

Low-intensity supported iCBT can be as effective as therapist-guided iCBT when initial human contact is present. Evidence regarding the potential harms of no-human support is needed before implementation.