Send to

Choose Destination
Ann Pharmacother. 2003 Dec;37(12):1891-9.

Meta-analysis of placebo rates in major depressive disorder trials.

Author information

Department of Pharmacoepidemiology and Pharmacotherapy, Utrecht Institute for Pharmaceutical Sciences, Utrecht, Netherlands.



Placebo effects in major depressive disorder (MDD) have received much interest in the medical literature. However, few quantitative analyses have been done in homogeneous populations.


To determine efficacy rates for placebo in patients with MDD; to quantify the correlation between efficacy and publication year, as well as between placebo and drug response rates.


Searching MEDLINE (1966-December 2000), EMBASE (1998-February 2001), HealthSTAR (1975-December 2000), and Cochrane (1980-December 2000) databases, randomized, placebo-controlled trials were retrieved including patients with MDD as defined by Diagnostic and Statistical Manual of Mental Disorders, 3rd and 4th editions criteria, Hamilton Rating Scale for Depression score >/=18 or Montgomery-Asberg Depression Rating Scale score >/=16, reporting successes as 50% decreases in scores after 6-8 weeks of treatment. Response rates were summarized using a random effects meta-analysis for per protocol (PP) and intent-to-treat (ITT) results.


We included 24 of 134 potential studies examining 4459 patients, 1786 on placebo and 2673 on an antidepressant. Placebo response rates were 45.5% (PP) and 26.9% (ITT). Correlations were significant between year and rates (PP rho 0.448, p = 0.042; ITT rho 0.557; p = 0.006), but not for active drugs. Placebo and drug rates were correlated (PP r 0.397, p = 0.020; ITT r 0.539; p = 0.002).


These placebo rates confirm those reported previously, but were from a homogeneous population. Although statistically significant, the correlation between drug and placebo rates was lower than others reported. During the study period, placebo rates increased linearly; active drugs did not. Correlations between placebo and drug response rates reflected moderate to strong effect sizes. We suggest that current methodology has been unsuccessful in achieving unbiased double-blind conditions not influenced by extra-trial factors, including time.

[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Atypon
Loading ...
Support Center