In Reply: In response to Ms Strite and Dr Stuart, the nondifferential error assumption we referred to entails equal probabilities of missed outcomes among randomization groups, conditional on study participant characteristics that are associated with outcome rates. We agree that a less stringent criterion without this conditioning could lead to systematic bias in summary measures.
It is not clear how Strite and Stuart imagine that the balance between treatment groups resulting from randomization itself might be disrupted. Perhaps they have in mind allocation concealment, which is typically used to describe tamper-proof methods of randomization. Allocation concealment is different from whether the trial is conducted in an open or a blinded fashion, and it is important in all trials, open or blinded.
Bruce M. Psaty, Ross L. Prentice. Importance of Blinding in Randomized Trials—Reply. JAMA. 2010;304(19):2127–2128. doi:10.1001/jama.2010.1622