The Evolution of Epidemiologic Research

Abstract

Fhe term was first coined to describe several changes in the way science I (particularly physics) was conducted in industrial nations during and after World War II. These changes included the assembly of staffs (usually multidisciplinary) to work in laboratories, using machines, all requiring budgets. The proximal stimulus for this was the belief that this was the only way to make rapid progress in developing a variety of critical defense-related products for the war effort, including the proximity fuse, radar, and the atomic bomb. This general approach, however, continued after the war, extending to particle research in general, and into other areas of physics as well (eg, laser technology). As time went on, other disciplines developed significant big science components as well, perhaps most notably astronomy. These trends have only recently entered the biologic sciences with the most graphic example being the changes in genomic research over the last 15 years. So how did big science come to epidemiology? The process has been somewhat different for classic epidemiology and molecular epidemiology. For the classic ap proaches, the need developed in a gradual and subtle manner. Up until 25 to 30 years ago, most epidemiologic research could be described as a cottage industry. There were a number of programs, but most were comprised of a small number of investigators with expertise largely limited to epidemiology, medicine, and biostatistics. Each investigator worked on his own studies, which were generally small (a study of 300 cases of a disease was viewed as a large study) using a limited study team and run on a relatively small budget. The principal investigator tended to design a study, develop the interview and abstract forms, train and supervise the data collectors (often doing much of it himself), code the data, and conduct the analyses on a calculator or perhaps through simple computer programs he could write himself. Although all this sounds rather quaint to young investigators today, it actually worked quite well, and this was a productive era for meaningful epidemiologic discov eries. The success of these approaches was largely due to the objectives of the investi gations. For the most part, the interest was in identifying large risks associated with obvious and easily measured exposures with a focus only on main effects. In my own area of cancer epidemiology, this was the era when many of the major risks with tobacco and alcohol consumption, ionizing radiation, occupational chemical exposures, medicinal agents, and reproductive factors were elucidated. Although many studies continue to focus on these objectives, over the last 30 years, there has been an increasing interest in identifying relatively low levels of risk, often associated with low-level or difficult-to measure exposures, and a concern about effects in subgroups (effect modification). This has resulted in the need for larger studies (now a study of 300 cases is often considered small), larger, more interdisciplinary study teams, and sophisticated, technology-intensive analytic methods-all of which require substantially larger budgets.