RCT

From EvaluationWiki

Randomly controlled trials, a putative paradigm for evidential sources in the social, health, and behavioral sciences today. Best seen as one more legacy of the neo-positivist legacy, based on a simplistic analysis of causation. Often incorrectly claimed to be only sure way to substantiate causal claims, whereas in fact they can be easily established in many other ways, e.g., by observation in some cases (as in the everyday observation that pressure on the brake pedal did in fact slow a car down), by many quasi-experimental designs, and by solid theoretical underpinning plus the elimination of other possible causes in a particular case (e.g., an autopsy, an airline crash investigation, an epidemic). The RCT design is sometimes referred to as 'the gold standard' in experimental design (Tom Cook, co-author with Don Campbell of the original work on quasi-experimental designs, explicitly condemns this description, mainly for reason (ii) below). It is a long way from the gold standard, for three reasons in particular: (i) it is a zero-blind design which means the Hawthorne effect is not eliminated (cf. double-blind design); (ii) it is extremely difficult to prevent both cross-contamination of the two groups and differential attrition, each of which invalidates the design; (iii) it does not control for genetic differences, which a twin study using identical twins does control. A gold standard study would be a typical pharmaceutical study, which is always double-blind, but using identical twins, on a drug with quick effects, e.g., an analgesic for post-operative pain. It would be expensive and difficult, but not in the least impossible.