Estimating the Placebo Effect on Patient-Reported Outcomes in Sham-controlled Device Trials: Insights from REDUCE...

Patient-reported outcomes (PROs) are increasingly used as endpoints in clinical trials. However, the magnitude of observed changes in control arms attributable to placebo effects, as compared with other benefits of trial participation, has not been described.This study seeks to estimate the magnitude of the placebo effect by calculating changes in Kansas City Cardiomyopathy Questionnaire Overall Summary (KCCQ-OS) scores, which quantify the impact of heart failure on patients' symptoms, function, and quality of life, after participants were unblinded to treatment.

REDUCE LAP-HF II randomized participants to atrial shunt or sham procedure, with unblinding after 2 years. The KCCQ was collected at baseline, 2, and 3 years after randomization. KCCQ-OS change from baseline to 2 years (placebo effect plus other benefits from trial participation) and the change from 2 to 3 years (placebo effect loss after unblinding) were calculated in sham-treated patients using mean±SD, as were changes from 2 to 3 years in shunt-treated patients (placebo effect benefit).

The analytic cohort included 421 participants (median age 72 years, 65.6% female). Among sham-treated participants (N=182), the mean±SD KCCQ-OS 2-year improvement from baseline was +9.3±22.4 points, with a decrement after unblinding of -1.7±18.2 points from 2 to 3 years. Among shunt-treated patients (N=239), mean±SD KCCQ-OS 2-year improvement was +12.7±22.8 points, with an improvement after unblinding of +1.9±18.2 points. In a hypothetical unblinded trial where placebo effect benefit would be expected in the active intervention arm, and none in the untreated arm, the combined effects would be 3.6 points.

In a sham-controlled device trial that collected PRO data during blinded allocation to treatment or sham and after unblinding, the estimated mean placebo effect benefit and loss on the KCCQ-OS were small (≤2 points). Finding a modest placebo effect on PROs may increase confidence in their use as clinical trial outcomes.