+Randomized controlled trials (RCTs) compare a drug’s efficacy to a placebo to isolate the effect of the drug above the psychological response to the belief of receiving treatment. In non-treatment-resistant depression (non-TRD), placebos have been shown to have a large effect that varies with the time of the investigation. No studies have yet quantified the magnitude of the placebo response in treatment-resistant depression (TRD) medication trials. This systematic review and meta-analysis quantified the placebo effect for different TRD treatments. The effect size was averaged as Hedges g for the placebo arms of 50 RCTs treating over 3000 TRD patients. The studies were assessed for bias. Subgroup analysis compared demographic, methodological, outcome and placebo type and duration data. The average effect of placebos was g = 1.05 and was similar for most placebo types. Further analysis of the funnel plot suggested some missing data that adjusted the effect size to g = 1.14. When removing the 19 high-risk of bias studies and only analyzing the 16 low-risk of bias RCTs, the g remained similar at 1.18. About 20% of patients responded to placebo and 13% went into remission. Furthermore, different modalities of placebo treatment did not differ significantly in response and remission rates. Sponsorship by industry, recent publication year, and open-label prospective treatment were associated with larger placebo effects, while numerous prior intervention failures was associated with a smaller placebo effect. A major concern of this meta-analysis was the heterogeneous classification of TRD by the included studies, causing exclusion of some large studies with more liberal TRD classification. One limitation of this study was the lack of psychotherapy and sham electroconvulsive therapy or magnetic seizure therapy placebos that might have met these inclusion criteria or have been conducted. A final limitation of this meta-analysis is that placebo groups occurred across many studies so none can be directly compared.
+RCTs of treatment-resistant unipolar depression with at least 1 placebo (inert pill placebo, parenteral placebo, sham device, or sham therapy) arm and a pharmacologic or somatic treatment arm were included. Waitlist or usual therapeutic treatment placebos were excluded. Data on demographics (age and sex, past depression history and response to therapy), methodology (augmentation vs monotherapy, number of centers, length of study, number of treatment arms, design, year of publication, and sponsorship, sample size), outcomes (mean depression rating, response vs remitting), and placebo characteristics (route, number of days of treatment). The quality of each included study was assessed. The effect of the placebo was estimated with Hedges g. 50 of the 11236 screened studies were included (N = 3228, 45.8±6.0 years, 54.8% women). Placebos had a pooled effect size of g = 1.05 (95%CI =0.91-1.18, I2 = 76.19%). The effect sizes were averaged for the pill placebo (g = 1.14, 95%CI = 0.99-1.30; I2 = 80.25%), parenteral placebo (g = 1.33, 95%CI = 0.63-2.04, I2 = 62.28%), liquid placebo (g = 0.45, 95%CI = −0.26-1.15; I2 = 0%), repetitive transcranial magnetic stimulation (g = 0.89, 95%CI = 0.63-1.15, I2 = 62.14%), transcranial direct current stimulation (g = 1.32, 95%CI = 0.53-2.11, I2 = 52.57%), and invasive brain stimulation (g = 0.86, 95%CI = 0.58-1.14, I2 = 15.48%). The funnel plot of the effect was symmetric (Egger test: z = 1.18; P = 0.23), but was asymmetric on deep analysis, suggesting 5 unpublished studies that when considered made the effect size g = 1.14 (95% CI = 1.00-1.29). 16 RCTs were at low risk of bias, while 16 were at high risk; when considering only the low risk of bias studies the effect size was similar (g = 1.18, 95%CI = 0.98-1.38). For the 42 studies reporting response rates and 25 reporting remission, they were 21.2±14.6% and 13.0±9.05%, respectively. Industry-sponsored (β = 0.34, 95%CI = 0.09-0.59; P = 0.007), recently published (β = 0.03, 95%CI = 0.003-0.05, P = 0.03), and open-label prospective treatment phase before randomization (β = 0.35, 95%CI = 0.11-0.59, P = 0.004) had larger placebo effects. Numerous failed interventions during the current depressive episode had a smaller placebo effect (β = −0.12, 95%CI = −0.23-−0.01, P = 0.03).