Blinded assessment of treatment effects utilizing information about the randomization block length.

It is essential for the integrity of double-blind clinical trials that during the study course the individual treatment allocations of the patients as well as the treatment effect remain unknown to any involved person. Recently, methods have been proposed for which it was claimed that they would allow reliable estimation of the treatment effect based on blinded data by using information about the block length of the randomization procedure. If this would hold true, it would be difficult to preserve blindness without taking further measures. The suggested procedures apply to continuous data. We investigate the properties of these methods thoroughly by repeated simulations per scenario. Furthermore, a method for blinded treatment effect estimation in case of binary data is proposed, and blinded tests for treatment group differences are developed both for continuous and binary data. We report results of comprehensive simulation studies that investigate the features of these procedures. It is shown that for sample sizes and treatment effects which are typical in clinical trials, no reliable inference can be made on the treatment group difference which is due to the bias and imprecision of the blinded estimates. (c) 2009 John Wiley & Sons, Ltd.