Estimation of treatment allocation in a randomised, double-blinded, placebo-controlled trial

AIM OF THE STUDY: The internal validity of double blinding in randomised placebo-controlled trials (RCTs) has become a target of criticism. The goal of this study was to investigate (a) how accurately the patients and their treating physicians were able to guess their assigned treatment, and (b) predictors for an accurate guess. METHODS: Data on treatment estimation from patients (n = 382) and their physicians (n = 358 guesses) in an RCT investigating the role of adjunct prednisone for community-acquired pneumonia in a tertiary care setting were analysed. At discharge, patients and their physicians had to guess whether they had been assigned to the prednisone or to the placebo group. The alternative possibility was "uncertain". Percentages and confidence intervals (CIs) were calculated for the proportion of patients guessing correctly. Chance finding was defined as having 50% or less correct guesses. To test for predictors for prednisone treatment guess, a mixed effects logistic regression model was performed. RESULTS: In the prednisone group, 28.9% (55/190; 95% CI 22.6-36.0%) of the patients made a correct guess and the majority (61.6%, 117/190) was uncertain. In the placebo group, 13.0% (25/192; 95% CI 8.8-18.8%) guessed correctly, with the majority being uncertain (69.8%, 134/192). Physicians guessed correctly in 48.3% (87/180, 95% CI 40.8-55.9%) of cases in the prednisone group and in 66.3% (118/178, 95% CI 58.8-73.2%) of cases in the placebo group, which was above chance for the placebo group. The physicians were uncertain in 21.7% (39/180) of cases in the prednisone group, and in 15.2% (27/178) of cases in the placebo group. Significant predictors for guessing prednisone were the occurrence of hyperglycaemia (odds ratio [OR] 3.77, 95% CI 2.39-5.95; p<0.001) and a shorter time to clinical stability (OR 0.95, 95% CI 0.91-0.99; p = 0.02). CONCLUSIONS: We confirmed that patient blinding was achieved in this study. Physicians made correct guesses more often than patients. Treatment estimation by both patients and physicians was led not only by the expectations of treatment effects of the study drug but also by known side effects of prednisone.