To the Editor.
—In his Clinical Decision Making article on investigational treatments, Dr Eddy1 maintains a misconception that could have damaging consequences in small-population studies when he states that randomization is the best way to ensure similarity of the patient groups.As clinical trials experts, statisticians, and epidemiologists have stressed repeatedly, the purpose of randomization in randomized controlled trials is to minimize the effect of unconscious bias,2 to reduce the bias in assigning patients to treatment,3 to ensure the absence of confounding by selection bias,4 and to ensure that at the outset the trial groups differ only by chance.5 As they have warned, randomization will not ensure an equal distribution of risk factors or other relevant patient characteristics to each treatment group3 and does not maximize equality of the groups.5 Randomization is not the best way of producing equality,4 but it does
Burt RAP. Investigational Treatments: Process, Payment, and Priorities. JAMA. 1997;278(17):1402. doi:10.1001/jama.1997.03550170032019