Placebo Control and Blinding in Randomized Trials of Procedural Interventions
Christopher A. Rajkumar,Katharine E. Thomas,Michael Foley,Sashiananthan Ganesananthan,Holli Evans,Florentina Simader,Sharan Syam,Daniel Nour,Catherine Beattie,Caitlin Khan,Rohin K. Reddy,Fiyyaz Ahmed-Jushuf,Darrel P. Francis,Matthew Shun-Shin,Rasha K. Al-Lamee
DOI: https://doi.org/10.1001/jamasurg.2024.0718
2024-04-17
JAMA Surgery
Abstract:Importance Unlike medications, procedural interventions are rarely trialed against placebo prior to becoming accepted in clinical practice. When placebo-controlled trials are eventually conducted, procedural interventions may be less effective than previously believed. Objective To investigate the importance of including a placebo arm in trials of surgical and interventional procedures by comparing effect sizes from trials of the same procedure that do and do not include a placebo arm. Data Sources Searches of MEDLINE and Embase identified all placebo-controlled trials for procedural interventions in any specialty of medicine and surgery from inception to March 31, 2019. A secondary search identified randomized clinical trials assessing the same intervention, condition, and end point but without a placebo arm for paired comparison. Study Selection Placebo-controlled trials of anatomically site-specific procedures requiring skin incision or endoscopic techniques were eligible for inclusion; these were then matched to trials without placebo control that fell within prespecified limits of heterogeneity. Data Extraction and Synthesis Random-effects meta-regression, with placebo and blinding as a fixed effect and intervention and end point grouping as random effects, was used to calculate the impact of placebo control for each end point. Data were analyzed from March 2019 to March 2020. Main Outcomes and Measures End points were examined in prespecified subgroups: patient-reported or health care professional–assessed outcomes, quality of life, pain, blood pressure, exercise-related outcomes, recurrent bleeding, and all-cause mortality. Results Ninety-seven end points were matched from 72 blinded, placebo-controlled trials (hereafter, blinded) and 55 unblinded trials without placebo control (hereafter, unblinded), including 111 500 individual patient end points. Unblinded trials had larger standardized effect sizes than blinded trials for exercise-related outcomes (standardized mean difference [SMD], 0.59; 95% CI, 0.29 to 0.89; P < .001) and quality-of-life (SMD, 0.32; 95% CI, 0.11 to 0.53; P = .003) and health care professional–assessed end points (SMD, 0.40; 95% CI, 0.18 to 0.61; P < .001). The placebo effect accounted for 88.1%, 55.2%, and 61.3% of the observed unblinded effect size for these end points, respectively. There was no significant difference between unblinded and blinded trials for patient-reported end points (SMD, 0.31; 95% CI, −0.02 to 0.64; P = .07), blood pressure (SMD, 0.26; 95% CI, −0.10 to 0.62; P = .15), all-cause mortality (odds ratio [OR], 0.23; 95% CI, −0.26 to 0.72; P = .36), pain (SMD, 0.03; 95% CI, −0.52 to 0.57; P = .91), or recurrent bleeding events (OR, −0.12; 95% CI, −1.11 to 0.88; P = .88). Conclusions and Relevance The magnitude of the placebo effect found in this systematic review and meta-regression was dependent on the end point. Placebo control in trials of procedural interventions had the greatest impact on exercise-related, quality-of-life, and health care professional–assessed end points. Randomized clinical trials of procedural interventions may consider placebo control accordingly.
surgery