Preferences, Equipoise, and Blinding

Unique challenges for behavioral clinical trials are posed by expectancy bias resulting from the lethal combination of visible treatment status and visible beliefs about these treatments. In contrast to double-blind drug trials where pre-existing beliefs are neutralized by treatments that look identical, behavioral trials evaluate randomized arms that often look different, thus limiting options for design control of biases resulting from preferences. Blinding outcomes assessors is a necessary, but insufficient, approach to minimize expectancy bias in a behavioral trial. Three additional targets should be considered. The mindset of the principal investigator should be scientific and objective, characterized by equipoise and freedom from a preference for a particular trial result. Design control can be maximized by a push to extend the blind to as many entities as possible, by the use of neutral names for trial arms, and by blinding to trial hypotheses but not to aims. Prevention and early detection of adverse implications of expectancy bias can be accomplished by ongoing assessment of risk of bias in such areas as differential retention, adherence, and co-interventions.