Placebo Effects in Antidepressant Trials Across Four Decades: A Comparison of Binary and Continuous Outcomes

Abstract

Background: Previous work suggested placebo response rate in antidepressant trials to be stable since the 1970s. The present analysis aimed to compare placebo response with placebo remission (both binary outcomes) and placebo efficacy (continuous outcome, based on the Hamilton Depression Scale), thereby considering previously suggested (but not assessed) nonlinear time trends and accounting for previously suggested (but not properly adjusted for) small-study effects (SSE). Methods: Random-effects meta-regression was conducted to estimate the effects of study year. Potential nonlinear effects of study year were modeled using restricted cubic splines (RCS). Relative importance of SSE compared to other trial-level covariates (study centers, dosing schedule, study length) was assessed using multimodel averaging. Findings: Multimodel averaging suggested any nonlinear time trends in placebo response (0·07 [-0·05-0·18] 95% CI) and remission (0·00 [-0·01-0·02] 95% CI) to be sufficiently explained by SSE and other trial-level covariates, with between-trial heterogeneity being reduced to I2 =55%/672%, respectively. This suggested stable rates across years, in line with previous work. By contrast, efficacy (continuous) showed an increasing trend from 1979 to 2914 (0·11 [0·02-0·21] 95% CI) with substantial between-trial heterogeneity being unexplained (I2 =85%). Interpretation: The present analysis reports differences in the time trends between binary and continuous placebo outcomes across decades. The findings are discussed regarding their clinical relevance because increasing placebo effects have been suspected of contributing to so-called failed antidepressant trials. Funding Statement: The authors stated: No funding. Declaration of Interests: The author declares no competing interests. Ethics Approval Statement: The authors declared: Not applicable.