Use of Placebo and Nonoperative Control Groups in Surgical Trials: A Systematic Review and Meta-analysis

IMPORTANCE: Nonspecific effects, particularly placebo effects, are thought to contribute significantly to the observed effect in surgical trials. OBJECTIVE: To estimate the proportion of the observed effect of surgical treatment that is due to nonspecific effects (including the placebo effect). DATA SOURCES: Published Cochrane reviews and updated, extended search of MEDLINE, Embase, and CENTRAL until March 2019. STUDY SELECTION: Published randomized placebo-controlled surgical trials and trials comparing the effect of the same surgical interventions with nonoperative controls (ie, no treatment, usual care, or exercise program). DATA EXTRACTION AND SYNTHESIS: Pairs of authors independently screened the search results, assessed full texts to identify eligible studies and the risk of bias of included studies, and extracted data. The proportion of all nonspecific effects was calculated as the change in the placebo control divided by the change in the active surgery and pooled in a random-effect meta-analysis. To estimate the magnitude of the placebo effect, we pooled the difference in outcome between placebo and nonoperative controls and used metaregression to estimate the association between the type of control group and the treatment effect (difference between the groups), adjusting for risk of bias, sample size, and type of outcome. MAIN OUTCOMES AND MEASURES: Between- and within-group effect sizes expressed as Hedges g. RESULTS: In this review, 100 trials were included comprising data from 62 trials with placebo controls (3 also included nonoperative controls), and 38 trials with nonoperative controls (32 interventions; 10 699 participants). Risk of bias across trials was comparable except for performance and detection bias, which was high in trials with nonoperative controls. The mean nonspecific effects accounted for 67% (95% CI, 61% to 73%) of the observed change after surgery; however, this varied widely between different procedures. The estimated surgical placebo effect had a standardized mean difference (SMD) of 0.13 (95% CI, −0.26 to 0.51). Trials with placebo and nonoperative controls found comparable treatment effects (SMD, −0.09 [95% CI, −0.35 to 0.18]; 15 interventions; 73 between-group effects; adjusted analysis: SMD, −0.11 [95% CI, −0.37 to 0.15]). CONCLUSIONS AND RELEVANCE: In this review, the change in health state after surgery was composed largely of nonspecific effects, but no evidence supported a large placebo effect. Placebo-controlled surgical trials may be redundant when trials with nonoperative controls consistently report no substantial association from surgery compared with nonoperative treatment.