Evolution of epidemiologic methods and concepts in selected textbooks of the 20th century.

Textbooks are an expression of the state of development of a discipline at a given moment in time. By reviewing eight epidemiology textbooks published over the course of a century, we have attempted to trace the evolution of five epidemiologic concepts and methods: study design (cohort studies and case-control studies), confounding, bias, interaction and causal inference. Overall, these eight textbooks can be grouped into three generations. Greenwood (1935) and Hill (first edition 1937; version reviewed 1961)'s textbooks belong to the first generation, "early epidemiology", which comprise early definitions of bias and confounding. The second generation, "classic epidemiology", represented by the textbooks of Morris (first edition 1957; version reviewed 1964), MacMahon & Pugh (first edition 1960; version reviewed 1970), Susser (1973), and Lilienfeld & Lilienfeld (first edition 1976; version reviewed 1980), clarifies the properties of cohort and case-control study designs and the theory of disease causation. Miettinen (1985) and Rothman (1986)'s textbooks belong to a third generation, "modern epidemiology", presenting an integrated perspective on study designs and their measures of outcome, as well as distinguishing and formalizing the concepts of confounding and interaction. Our review demonstrates that epidemiology, as a scientific discipline, is in constant evolution and transformation. It is likely that new methodological tools, able to assess the complexity of the causes of human health, will be proposed in future generations of textbooks.