The development of cohort studies in epidemiology: A review

Abstract

An historical outline of the evolution of cohort (or incidence) studies spans well over 100 years, from the work of Farr and Snow in the 1850s, through an appraisal of analytical methods in 1977, after which the literature mushroomed. Since the early 1950s, analysis has conventionally taken the form of comparing subcohorts that had suffered varying degrees of exposure to factor(s) under investigation. For this purpose the “subject-years” approach to data reduction has now become virtually universal. Usually, some population's mortality (or morbidity) experience is used as reference, but there is continuing controversy over the choice of reference population, while difficulties arise in relation to study intervals, periods over which exposures should be measured, etc. The material for analysis becomes age- and periodic-specific ratios of disease, which, collapsed over ages and periods, lead to Standardized Mortality (or Morbidity) Ratios. For the analysis itself, Poisson regression models are efficient. From the late 1970s, analysis by case-referent methods has become common; here, the debate centres on how closely, and in what ways, referents should be matched with the cases. Logistic regression is the most common form of analysis. As there have been excellent recent summaries of methods of analysis (for both approaches), little emphasis is placed here on those aspects of development. Comparisons are made of research designs, and some possibilities for future development are outlined.