On the insufficiency of reporting masking.

Villamar et al. (1) noted that masking was reported at a higher rate in 2010 than it was in 2000, but lamented that other aspects of CONSORT were not similarly improved. To place the reporting of masking in context, imagine a scenario in which the rehabilitation of patients with knee injuries was considered complete and successful as soon as these patients reported feeling better. Imagine a police department reporting lower crime rates on the basis of fewer among the accused reporting their guilt of the crimes of which they stand accused. These analogies are used not to discredit the reporting of masking; this is certainly part of the equation, necessary, but not in any way sufficient (2): “The attempt to mask in no way equates to successful masking, and an uncorroborated statement with no test of the success of masking is simply not credible. Masking is not a binary phenomenon; it can be partially successful, and the extent to which it is successful determines the extent to which the trial is likely to be biased. The claim by itself does not ensure any level of success.” One testable consequence of unmasking is blown allocation concealment (at least when restricted randomization is used, as it is in pretty much every randomized trial in practice) (3). This is why a test of the success of allocation concealment doubles also as a test of the success of masking (as opposed to just the attempt to mask, which is generally all that is meant when masking is reported), and must be a part of any serious evaluation of the success of masking. The Berger-Exner test of selection bias (4) is an objective test of the success of allocation concealment, and is subject to neither manipulation nor recall bias. The absence of any mention of this analysis means that we do not really know how successful the masking was in trials in 2000 or in 2010, and one can only hope that future trials do better.