A Less Myopic View of the Virtues of Blinding (and of Tests for Blinding) in Clinical Trials
Abstract It is widely believed within medicine that, so far as single trials are concerned, the absolute top (‘platinum level’) evidence comes from trials that are not only randomized but also double-blind (meaning of course that neither the clinicians nor the participants involved know which arm of the trial—experimental treatment or control—any particular participant has been assigned to). The issue of randomization has been widely discussed; this talk concentrates on double-blinding. Blinding’s epistemic virtues seem evident, being concerned with the elimination of bias/confounding. Empirical research shows, however, that clinical trials are in fact seldom checked for blindness, and that, where they are checked, more often than not they fail: that is, the checks show that blinding has been lost. Given that the attempt to guard against the intrusion of bias was the main driving force behind EBM, its leaders might be expected to be in the forefront of those deploring this situation; but the reality is surprisingly different. Far from deploring the fact that clinical trials often end up unblinded, the ‘father of EBM’, David Sackett, went so far as to claim that checking at the end of a trial that blinding has been preserved amounts to ‘playing a mug’s game’. Moreover, a relatively recent revision of the EBM-inspired CONSORT guidelines for clinical trials eliminated all mention of such checks for continued blinding. This paper examines the complexities involved and tries to eliminate the (many) confusions. The result is an altogether more nuanced, but still positive, view of blinding and of end-of-trial tests for blindness—a view that mandates a re-revision of the CONSORT guidelines (possible REF ‘impact’!).