The impact of active placebo control groups on estimated intervention effects in pharmacological randomized trials: a meta-epidemiological study

Background: In a randomized trial comparing a drug with perceptible adverse effects (side effects) against standard placebo, patients may guess their allocated intervention and become unblinded. Active placebo controls are designed to imitate the drug’s perceptible effects while being therapeutically inactive, which in theory reduces the risk of bias due to unblinding. Therefore, the estimated intervention effects of drugs could differ depending on the choice of placebo control. A previous systematic review of randomized within-trial comparisons of active placebo versus standard nonactive placebo indicated
that the estimated intervention effects may be smaller, on average, when comparing a drug against active placebo, but the result was imprecise.

Objectives: To estimate the impact of active placebo control interventions, compared with standard placebo control interventions, on estimated effects of experimental interventions in pharmacological randomized clinical trials and to explore causes for heterogeneity
Methods: We conducted a meta-epidemiological study of meta-analyses that included pharmacological randomized trials with standard placebo control as well as trials with active placebo control. We searched for systematic reviews with eligible meta-analyses in PubMed, Embase, the Cochrane Library, Google Scholar, and PROSPERO. We also searched studies citing known active placebo-controlled trials. We removed overlapping trials between the eligible meta-analyses. The primary analysis was the ratio of odds ratios (ROR) of standard placebo-controlled trials versus active placebo-controlled trials for outcomes of benefit. We conducted a frequentist 2-step analysis by comparing trials
within each meta-analysis before summarizing results across meta-analyses. We explored heterogeneity by outcome type, study design, matching adequacy of the active placebo, and risk of therapeutic effect of the active placebo. We also checked the robustness of our findings in sensitivity analyses. In secondary analyses, we investigated outcomes of harms, attrition, and co-intervention use.

Results and conclusions: We have currently included 65 systematic reviews with 64 meta-analyses for the primary analysis. The number is subject to change after updating the search and removal of meta-analyses due to overlapping trials. We will present the analysis results and conclusions at the Summit.

Patient and public involvement: None