Bias in Treatment Assignment in Controlled Clinical Trials

Abstract

To the Editor: Like Chalmers et al., we believe in "the necessity for blinding the randomization process in a controlled trial" in order to minimize bias (Dec. 1 issue).* However, we found their paper to be an even more dramatic example of the insidiousness of this problem in the clinical literature than they might have intended. The authors themselves have fallen prey to the same shortcomings — namely, the use of maldislributed variables and historical controls — that they decry. One can see this by first examining their Table 2, in which some "types of treatments studied" are maldistributed: for . . .