Abstract
Although the short-term efficacy of internet-delivered cognitive-behavioral therapy (i-CBT) is well-established, its long-term efficacy remains understudied. Robust variance estimation meta-analysis was thus conducted across guided and self-guided i-CBT, synthesizing data from 154 randomized controlled trials (N = 45,335) with ≥ 12-month follow-ups. For binary outcomes, guided (52.3% vs. 38.6%; log-risk ratio [LOG-RR] = 1.15 95% confidence interval [1.04, 1.26]) yielded higher remission, reliable improvement, and response rates, and lower suboptimal treatment outcome rates (9.3% vs. 10.8%; LOG-RR = 0.63 [0.45, 0.80]) than treatment-as-usual, active controls, and waitlists at ≥12 months. Insufficient studies precluded testing the efficacy between self-guided i-CBT and controls for binary outcomes. For baseline-to-12-month dimensional outcomes, guided i-CBT produced greater reductions in anxiety, depressive, post-traumatic stress disorder (PTSD) symptoms, and repetitive negative thinking (Hedge's g = -1.86 to -0.31), and self-guided i-CBT yielded stronger reductions in depressive symptoms (g = -0.51) than all controls. For outcome scores aggregated at ≥ 12-month follow-ups, guided i-CBT alleviated anxiety, depression, distress, insomnia, PTSD symptoms, role impairment, emotion regulation, and quality of life (g = -0.31 to 0.26), and self-guided i-CBT yielded lower anxiety and depressive symptoms (g = -0.16 to -0.09) than all controls. No significant differences in efficacy emerged between guided and self-guided i-CBT when sufficient studies existed for a meta-analysis. There was no evidence for publication bias. Long-term efficacy was similar to short-term efficacy for most outcomes. Implementing scalable i-CBTs should entail transparency about their long-term benefits and drawbacks.