In the analytic method of epidemiological study called a cohort study, subsets of a defined population are identified and categorized on the basis of exposure to known levels of a risk factor that is believed to be associated with a disease outcome such as coronary heart disease or cancer. The numbers of persons in the total population and the numbers in each subset are known. All are followed over a period, usually years or even decades, and the disease outcomes are recorded and counted at specified periods. These outcomes may be the incidence of diagnosed disease, and/or deaths certified due to the disease being studied, as well as deaths due to other causes. The total numbers, or the number of person-years of observation, must be large enough to generate stable rates, so that the rates can be compared in subsets of the total population that have been exposed to different levels of risk. Hypotheses about causes
The effort and expense required to conduct cohort studies have been justified by the results of several well-known studies. One of these is the Framingham study, which began in 1948 and still continues. It is a study of samples of the population of Framingham, Massachusetts, in which several risk factors associated with coronary heart disease, other cardiovascular diseases, and more recently, several other chronic diseases, have been assessed. This and several other cohort studies have clarified our understanding of the principal risk factors for coronary heart disease, such as elevated serum lipids, high blood pressure, and cigarette smoking. Other well-known cohort studies include the long-term follow-up of a cohort of male British doctors who were first asked about their smoking habits in 1951. After 20 years, the death rates from lung cancer, other respiratory system cancers, chronic obstructive lung disease, and coronary heart disease all showed significant differences related to smoking habits among this large cohort (Doll and Peto 1976).
Several cohort studies of cancer risks associated with exposure to ionizing radiation have made use of existing data to shorten considerably the many years of observation that would otherwise be required to demonstrate and measure levels of risk. This has been made possible by the existence of good medical records of past diagnostic X-rays that exposed people to low doses of radiation. After case-control studies had revealed evidence suggesting that the use of diagnostic X-rays during pregnancy might increase the risk of cancer in childhood, several cohort studies were set up to confirm or refute this evidence. MacMahon and others used the medical records of over three quarters of a million women in New England to determine the amount of diagnostic radiation to which they had been exposed during pregnancy, and ascertained the incidence and mortality rates from leukemia and other cancers, including cancers of the brain, bone, and kidney, in the first eight to ten years of their children's lives. They found a significantly higher rate among children who had been prenatally exposed to small doses of diagnostic X-rays, and also observed a dose-response relationship, meaning that there were higher rates among children whose mothers had two or more X-rays than in children whose mothers had only one X-ray. This method is known as an historical cohort study. Other historical cohort studies have shown that repeated chest X-rays (or fluoroscopic screenings) increase the risk of breast cancer many years later.
A by-product of cohort studies is the use of some of the persons studied to conduct one or more case-control studies that are "nested" within the total cohort population. This has the advantage of offering a more rapid answer to questions that have arisen in the course of the cohort study, and also eliminates some of the common biases, such as differential recall of relevant facts by cases and controls, encountered in other varieties of case-control study.
As noted above, cohort studies are more powerful than case-control studies but they have some disadvantages. Strengths include the following: complete data on cases, stages, exposures; can study more than one effect of exposure; can calculate and compare rates and risks; choice of factors available for study; quality control of data; can accommodate "nested" case-control study. Weaknesses include the following: must study large numbers; usually takes many years, even decades; circumstances may change during study; expensive in money, skilled staff required; incomplete control of extraneous factors; rarely possible to study disease mechanism.
Cohort studies are sometimes called prospective or longitudinal studies. It is important to emphasize that a cohort study, like a case-control study, is not an experiment, but merely observes the subjects of the study without intervening—except to ask questions or conduct physical examinations and laboratory tests at various intervals.
The analysis of results is generally a simple matter of calculating and comparing rates, which are commonly expressed in terms of person-years of observation—if one person is observed for ten years, this is ten person-years of observation; two years observation of five persons is also ten-person years. The use of person-years is a convenient way to generate larger numbers for calculation of rates that are more stable than with smaller numbers.
JOHN M. LAST
Breslow, N. E., and Day, N. E. (1987). Statistical Methods in Cancer Research, Vol. 2: The Design and Analysis of Cohort Studies. Lyon: International Agency for Research on Cancer.
Doll R., and Peto, R. (1976). "Mortality in Relation to Smoking: 20 Years Observations of Male British Doctors." British Medical Journal 2:1525–1536.
Kelsey, J. E.; Whittemore, A. S.; Evans, A. S.; and Thompson, D. (1996). Methods in Observational Epidemiology, 2nd edition. New York: Oxford University Press.
Rothman, K. J., and Greenland, S. (1998). "Cohort Studies." In Modern Epidemiology, 2nd edition, ed. K. J. Rothman and S. Greenland. Philadelphia, PA: Lippincott-Raven.