J Moncrieff1, S Wessely, R Hardy. 1. Section of Epidemiology and General Practice, Institute of Psychiatry, London.
Abstract
BACKGROUND: Unblinding effects may introduce bias into clinical trials. The use of active placebos to mimic side-effects of medication may therefore produce more rigorous evidence on the efficacy of antidepressants. METHOD: Trials comparing antidepressants with active placebos were located. A standard measure of effect was calculated for each trial and weighted pooled estimates obtained. Heterogeneity was examined and sensitivity analyses performed. A subgroup analysis of in-patient and out-patient trials was conducted. RESULTS: Only two of the nine studies examined produced effect sizes which showed a consistent significant difference in favour of the active drug. Combining all studies produced pooled effect size estimates of between 0.41 (0.27-0.56) and 0.46 (0.31-0.60) with high heterogeneity due to one strongly positive trial. Sensitivity analyses excluding this and one other trial reduced the pooled effect to between 0.21 (0.03-0.38) and 0.27 (0.10-0.45). CONCLUSIONS: Meta-analysis is very sensitive to decisions about exclusions. Previous general meta-analyses have found combined effect sizes in the range 0.4-0.8. The more conservative estimates produced here suggest that unblinding effects may inflate the efficacy of antidepressants in trials using inert placebos.
BACKGROUND: Unblinding effects may introduce bias into clinical trials. The use of active placebos to mimic side-effects of medication may therefore produce more rigorous evidence on the efficacy of antidepressants. METHOD: Trials comparing antidepressants with active placebos were located. A standard measure of effect was calculated for each trial and weighted pooled estimates obtained. Heterogeneity was examined and sensitivity analyses performed. A subgroup analysis of in-patient and out-patient trials was conducted. RESULTS: Only two of the nine studies examined produced effect sizes which showed a consistent significant difference in favour of the active drug. Combining all studies produced pooled effect size estimates of between 0.41 (0.27-0.56) and 0.46 (0.31-0.60) with high heterogeneity due to one strongly positive trial. Sensitivity analyses excluding this and one other trial reduced the pooled effect to between 0.21 (0.03-0.38) and 0.27 (0.10-0.45). CONCLUSIONS: Meta-analysis is very sensitive to decisions about exclusions. Previous general meta-analyses have found combined effect sizes in the range 0.4-0.8. The more conservative estimates produced here suggest that unblinding effects may inflate the efficacy of antidepressants in trials using inert placebos.
Authors: Glen I Spielmans; Margit I Berman; Eftihia Linardatos; Nicholas Z Rosenlicht; Angela Perry; Alexander C Tsai Journal: PLoS Med Date: 2013-03-12 Impact factor: 11.069