It is well-established that psychotherapies are effective in the treatment of depression. It remains unclear, however, whether the effects persist over the long term. We conducted a comprehensive meta-analysis examining the long-term effects of psychotherapies for depression compared with control conditions. We included randomized controlled trials (RCTs) comparing psychotherapies for adult depression with control conditions (care-as-usual, waitlist, or other) that reported outcomes at six months after randomization or later. The standardized mean difference (SMD) was the main outcome, but we also examined response rates in the treatment and control groups and the relevant risk ratio (RR). We evaluated the effectiveness of psychotherapies over time using a multivariate pooling model with cubic regression splines to allow for potentially non-linear interactions. Dependencies in the true effects were modeled using continuous-time autoregressive variance-covariance matrices. We included 191 RCTs with 209 comparisons between treatment and control groups, involving 33,691 participants (17,715 in the treatment, 15,976 in the control conditions). The trials encompassed 534 follow-up measurements, of which 52 (9.7%) occurred more than 12 months after randomization. The best fitting restricted cubic spline model for long-term SMDs (5 basis functions) indicated that the SMD increased from baseline to 0.55 at 12 weeks, then gradually decreased to 0.39 at around one-year follow-up, and remained largely stable thereafter. The SMD remained statistically significant up to 8.3 years (431 months) after randomization. The RR for response went up to 1.68 after 3 months and slowly decreased over time to 1.14 at 5-year follow-up. The RR remained significant up to 4.5 years (236 weeks). We conclude that psychotherapies have moderate to large effects at 3 months after randomization, which slowly decrease during the next 9 months to a moderate level, and remain stably moderate for several years.
Cuijpers et al. (Fri,) studied this question.