Double Standards, Scientific Methods, and Epidemiologic Research

Abstract

Because suspected noxious agents can seldom be tested experimentally in people, epidemiologic investigation of the causes of disease depends on observational procedures, of which the most prominent types are cohort and case–control studies. In these studies the compared agents cannot be assigned experimentally, but many other basic experimental principles that could be employed are often overlooked, including verification of quality and accuracy in raw data, avoidance of major biases in comparison, vigilance in checking for methodologic errors, and maintenance of a careful distinction between research data that generate hypotheses and the new data needed to test these hypotheses. The problems that result when these principles are overlooked can be alleviated if etiologic research is performed and evaluated with the same standards used in other branches of science. (N Engl J Med. 1982; 307:16117.)