Different control conditions can produce different effect estimates in psychotherapy trials for depression
Michopoulos I., Furukawa TA., Noma H., Kishimoto S., Onishi A., Ostinelli EG., Ciharova M., Miguel C., Karyotaki E., Cuijpers P.
Objectives: Control conditions’ influence on effect estimates of active psychotherapeutic interventions for depression has not been fully elucidated. We used network meta-analysis to estimate the differences between control conditions. Study Design and Setting: We have conducted a comprehensive literature search of randomized trials of psychotherapies for adults with depression up to January 1, 2019 in four major databases (PubMed, PsycINFO, Embase, and Cochrane). The network meta-analysis included broadly conceived cognitive behavior therapies in comparison with the following control conditions: Waiting List (WL), No Treatment (NT), Pill Placebo (PillPlacebo), Psychological Placebo (PsycholPlacebo). Results: 123 studies with 12,596 participants were included. The I-squared was 55.9% (95% CI: 45.9%; to 64.0%) (moderate heterogeneity). The design-by-treatment global test of inconsistency was not significant (P = 0.44). Different control conditions led to different estimates of efficacy for the same intervention. WL appears to be the weakest control (odds ratio of response against NT = 1.93 (1.30 to 2.86), PsycholPlacebo = 2.03 (1.21 to 3.39), and PillPlacebo = 2.66 (1.45 to 4.89), respectively). Conclusions: Different control conditions produce different effect estimates in psychotherapy randomized controlled trials for depression. WL was the weakest, followed by NT, PsycholPlacebo, and PillPlacebo in this order. When conducting meta-analyses of psychotherapy trials, different control conditions should not be lumped into a single group.