There are certain classic EBM papers that at some point you have to read. And ‘Empirical Evidence of Bias Dimensions of Methodological Quality Associated With Estimates of Treatment Effects in Controlled Trials’, bit of a mouthful, is certainly one of them.
Allocation concealment is a technique used to prevent selection bias in a randomized controlled, preventing researchers from unconsciously or consciously influencing which group a participant goes in to.
The absence of allocation concealment therefore biases the treatment effect; but the question is, by how much?
The study published in JAMA, in 1995, assessed the methodological quality of 250 controlled trials from 33 meta-analyses. Note the eminent authors on this one. In the trials in which concealment was either inadequate or unclear, the estimates of treatment effects were significantly larger: odds ratios were exaggerated by 41% for inadequately concealed trials and by 30% for unclearly concealed trials.
This effect was noted in a similar follow up study published in the, Lancet, in 1998. This study selected 11 meta-analyses including 127 RCTs on a variety of treatments, and in the trials that used inadequate or unclear allocation concealment the estimates of effect were much the same as the earlier study, 37% larger.
Therefore, when you are asked, exactly by how much does a lack of allocation concealment bias the treatment effect? Then answer, about 40%, which would get rid of a lot of treatment effects.
Other Blogs to read in the EBM papers to read series include: The concept of risk