blinding

Blinding, also called masking, is a guard against a very human problem: people who know which treatment someone is on tend to see what they expect to see. Simon Day and Douglas Altman, writing in the BMJ, define it as keeping study participants, those involved with their management, and those collecting and analysing the data unaware of the assigned treatment, so that they should not be influenced by that knowledge.

More than one group can be blinded, and the useful question is always which. The Centre for Evidence-Based Medicine's catalogue of bias lists five: the participants, the clinicians caring for them, the people collecting the data, the people assessing the outcomes, and the people analysing them. Labels like single blind and double blind are used loosely enough that they are easily confused, so a well reported trial names who was kept unaware rather than counting them.

A trial in which nobody is blinded is called open-label, and that is sometimes unavoidable: two treatments given in different ways, or a drug with an unmistakable side effect, will announce themselves. The usual repair is to blind the assessment even where the treatment itself cannot be hidden, handing the scans or the records to independent reviewers who do not know who received what. Day and Altman argue that blind assessment of the outcome may matter more than blinding the treatment, above all where measuring that outcome involves any judgement.