[Cohort study and multivariate analysis: an epistemological and historical analysis of the Framingham heart study]

Begun in 1947 and still ongoing, the epidemiological study of heart disease known as the Framingham study was one of the first prospective studies based on a large cohort and has rapidly been considered as the prototype and model for the cohort study. Nevertheless, an examination of its history reveals that the protocol does not at all correspond to today's standards for this type of study. How, then, can we account for the remarkable reputation of this study?

This paper consists in an epistemological and historical analysis of the Framingham study that provides some of the answers to this question. In my treatment of the study's methodology, I focus on the issue of how the study population was constituted, and the manner in which the multiple factor analyses were conducted, two issues that are now central to cohort studies and more generally to analytic epidemiology.

I show how the study population of Framingham and its long-term follow-up have contributed significantly to the interpretation of the cohort as a sort of "population-laboratory". The data generated by this study, which have been very widely used by epidemiologists and other researchers, are unparalleled in terms of the amount of detailed clinical information available for such a long follow-up period. Furthermore, multivariate statistical modelling, which has become a standard statistical tool for clinical as well as epidemiological studies was introduced in the context of this study to improve the identification of significant factors in the simultaneous analysis of multiple correlations. Multivariate analysis has since proved crucial in shaping the epidemiological concept of "risk factor" and in analysing multifactorial diseases. Indeed, I suggest that the modern idea of multifactorial diseases depends on the adaptation of this statistical method.

Thus, the Framingham study played a leading role not only in remodelling epidemiology after Second World War, in particular because of its contribution to the establishment of the cohort study as a standard method of investigation in etiological research, but also in constituting the "risk factor approach" to disease.