Approaches for Conducting Large Cohort Studies

Numerous cohort studies have been initiated over the past 50 years to address lifestyle and biochemical determinants of specific diseases. A key feature of these studies is that information on the primary exposures and potential confounders is required from the individual subjects themselves, rather than from routinely collected data. In this brief overview we will discuss some of the theoretical and practical issues to be considered in the development and conduct of a large modern cohort study. Earlier cohorts, such as those of occupational and medical exposures including the health effects of ionizing radiation (1, 2), frequently did not require contact with the participants, and, thus, could address only very specific issues and were limited in ability to account for confounding. These studies will not be considered in detail here. Some of the first modern cohort studies were established specifically to evaluate one or a few factors, such as the health effects of smoking (3-5) or physical activity (6, 7). Still, other simple variables were collected that have provided valuable opportunities to examine the effects of obesity (8), alcohol consumption (9, 10), and oral contraceptives (11, 12). Early population-based cohorts included the work of Archie Cochrane who, in 1950, began the study of Rhondda Fach which encompassed eight mining towns. Of the adult population, 89 percent were x-rayed, including some 2,764 miners and 3,055 ex-miners (13, 14). Though the initial focus was on causes of pneumoconiosis and progressive massive fibrosis among miners living in these communities (14), rheumatoid arthritis was examined (15), and by the first follow-up 2'/2 years after the baseline data collection, psychiatric conditions were added as a focus of disability (16). Cardiovascular risk factors and endpoints were studied from early in the life of this