Twelve RCTs (n=2,862) were included. Sample sizes of the studies ranged from 96 to 376.
Three site selection variables had significant positive associations with placebo response in univariate analyses: number of patients randomised (12 studies); number of study sites (12 studies); and number of participants per study site (11 studies). The mean Children's Depression Rating Scale - Revised (CDRS-R) score at baseline had a significant inverse association with placebo response in univariate analyses. In multiple linear regression analysis, only number of study sites remained significantly associated with placebo response (β=0.59, t=2.46, df=6, p=0.05, partial r=0.71).
Placebo response was not significantly different between children and adolescents, but with the exclusion of one fluoxetine trial in sensitivity analysis, younger children had higher placebo response rates than older adolescents (response rate in children 54.3% versus response rate in adolescents 44.9%; Χ2=5.87; p=0.02), but this relationship was not found when other trials were excluded in sensitivity analysis.
Higher placebo response rates in more recent studies were associated with an increasing trend toward large multisite trials and by publication delays and failures to publish some negative trials.