Statistical methods for blinding assessment in randomized controlled trials
Victoria Landsman, PhD University of Toronto
In this talk, I will present key results from my recent work around statistical methods to assess blinding in randomized controlled trials (RCTs) with clustering (e.g., by center or practitioner) as well as meta-analyses of RCTs. I will define a Global Blinding Index (GBI) as a target parameter and summarize four possible estimators of the GBI. To this end, I will introduce a multinomial overdispersion model that requires the specification of the first two moments of the outcome and allows the estimation of the global parameters using generalized estimating equations (GEE). Similar models have been extensively studied for correlated normal and binomial data but much less so for correlated multinomial data. I will also summarize some important results from the simulation study designed to compare the performance of the four estimators. To conclude, I will describe the application of these methods to our latest meta-analysis study on the association of blinding and treatment effects in randomized controlled trials for back pain in the period 2000-2019.