Make your own free website on

Basis for Disease Prevention
and Health Promotion

Cohort Studies

Meet the Author
Epidemiologic Approach
Classic Model
Natality & Mortality
Disease & Disability

Often it is impractical or unethical to randomly assign a group of people to be exposed to an independent variable we wish to study. Given what we already know about the harmful effects of smoking, we cannot ask people to smoke so that we can study yet another possible ill-effect. Likewise, it would be unethical to randomly assign people to be exposed to radiation or toxic wastes. We can, however, study people who have already been exposed to toxic wastes, whose job exposes them to radiation, or who choose to smoke.

Such studies are much weaker than randomized controlled trials because we are less certain that other possible causes are being excluded. People who have been exposed to toxic wastes may have been exposed in the same fashion to other causal factors. Smokers may be different from nonsmokers in ways that have a causal impact on the disease we are studying (e.g., they may be more inclined to take risks or less concerned about their health). Despite these weaknesses, cohort studies are still of considerable value, providing far stronger evidence than the cross-sectional surveys of descriptive epidemiology.

The Cohort Method

The term cohort originally referred to a group of persons born the same year in the same place -- the same nation, state, city, hospital, etc.. Cohorts were studied primarily for mortality data. Studies of births in cohorts of women born the same year have also been used for quite some time as a way to study fertility. Today the term birth cohort is used for cohorts of this original type.

The term cohort is used today for any group that is followed up over time. Subject are selected for inclusion in a cohort because of some current exposure to a possible cause. They are then followed up prospectively over time to measure the incidence of the disease under study. For this reason, cohort studies are also known as prospective studies and as incidence studies.

The Framingham Heart Study

One of the best known of all cohort studies is the Framingham Heart Study. This study was initiated in 1948 by the U.S. Public Health Service and researchers at the Harvard School of Public Health to study a variety of possible risk factors for heart disease. The town of Fram¬ingham in Massachusetts (population 28,000) was chosen as the site of the study because of its relatively stable population, availability of a local community hospital, proximity to Boston and Harvard, and cooperation in a previous community study. The study was originally planned as a 20-year followup but has now been underway almost twice that long.

A list of all local residents was developed as a sampling frame for random selection. Families were listed together within each precinct. Two out of every three families (6,507 persons) were invited to partic¬ipate. Of these, 4.469 persons agreed to participate. Because this was less than the researchers" goal of 5,000 participants, a group of 740 volunteers was added to make a total of 5,209 subjects.

All of these subjects were then given a physical examination that included measures of blood pressure and serum cholesterol. At this stage, 82 subjects were found to have symptoms indicative of previously undiagnosed heart disease and were removed from the sample, leaving a total of 5.127. These persons could be subdivided into subcohorts based on such characteristics as smokers versus nonsmokers or hypertensives versus normotensives. The cohort has been offered a relatively compre¬hensive physical examination every 2 years. This not only provides some outcome data but allows reassignment to subcohorts as conditions change and has allowed the addition of some new measurements of interest that were not part of the original study design.

Much of our current knowledge of risk factors for heart disease either came out of this study or was confirmed by the study. The find¬ings with regard to the role of serum cholesterol in heart disease were particularly important. In males aged 30 to 49, for instance, the risk of heart disease in those with serum cholesterol levels greater than 250 mg% is 4.1 times that of men with cholesterol levels below 190 mg%.

The existence of this long-standing cohort, on which numerous variables have been recorded, has permitted numerous studies that were not included in the original plan. For instance, although this cohort was created to study heart disease, the causes of all deaths in the cohort have been recorded, allowing studies of mortality of any type and not just heart disease mortality. For example, mortality due to all causes has been studied in relationship to alcohol consumption. Gordon and Kannel (1984) found that there is no association between drinking and subsequent mortality in women. Among men, although increased alcohol consumption is associated with increased mortality due to cirrhosis of the liver and cancer (especially stomach cancer), non-drinkers suffer higher mortality than drinkers, even the heaviest drinkers. Although contrary to many peoples' expectations, these findings are consistent with those from other cohort studies of drinking and mortality.

Coffee Consumption and Cancer Mortality

In 1966, a prospective study of 26,020 male life insurance policy-holders was initiated. A dietary frequency-of-use survey was mailed to this population and was completed and returned by 17,818 men, who became the cohort under study. Cancer mortality among heavy coffee users was compared to all other subjects. The hypothesized relation¬ship between heavy coffee drinking and pancreatic cancer was not found. No positive relationships were found between coffee consump¬tion and any form of cancer except lung cancer. Even after controlling for age, urban/rural residence, and cigarette use, the relative risk of death due to lung cancer was 7.33 times as great for heavy coffee drinkers. This compares to a relative risk of 9.6 for cigarette smokers, controlling for the influence of age, urban/rural residence, and coffee consumption. Persons who drink five or more cups of coffee per day and smoke a pack or more of cigarettes per day are 40.37 times more likely to die of lung cancer.

Historical Cohorts

The major disadvantage of cohort studies is the amount of time it takes to conduct them. We often want answers now that could only be provided by a twenty-year prospective study (RCT or cohort). Sometimes we can obtain those answers sooner by way of a historical cohort study -- also known as a retrospective cohort study or, rather confusingly, as a prospective study done retrospectively. The opportunity to conduct such a study is present when we can find some record that allows us to assign some population to groups based on their exposure to a possible cause at some time in the past and to follow those people until the present time, keeping track of the incidence of the disease under study.

Radiation and Mortality

The end of World War II thrust us into the Atomic Age with little idea of what the health consequences of radiation might be. Although it was known that large doses of radiation can be lethal, the cumulative effects of smaller doses were unknown. Unfortunately, to study this question through standard cohort or experimental methods would have required decades and the answers then might come too late.
Seltser and Sartwell (1965) devised a historical cohort study to assess the possible hazards of small doses of radiation. They studied mortality among members c F three professional organizations. The three or¬ganizations were the Radiologic Society of North America (founded in 1915), the American College of Physicians (founded in 1915), and the American Academy of Ophthalmology and Otolaryngology (founded in 1921). Members were studied from the time they joined their professional organization until 1958.

As radiologists, the first group had substantial exposure daily to radiation, especially during this period when modern precautions were not taken. The second group, being composed of internists, was likely in those days to have quite a bit of x-ray exposure but far less than the radiologists. The third group, being eye, ear, nose, and throat (ENT) specialists, made little if any use of x-rays and thus served as a low-risk cohort.

Mortality was studied for the periods 1935-1944, 1945-1954, and 1955-1958. In all three periods, mortality rates were highest for the radiologists and lowest for the ENT specialists. Radiologists, for instance, showed mortality due to leukemia that was 2.5 times as great as that of the ENT specialists. Radiation exposure was clearly associated with greater mortality due to cancer, heart disease, kidney disease, and all causes combined. The precautions taken by modern radiologists and Radiologic technicians owe much to this study.

Recommended Reading

Cook, N. R., & Ware, J. H. (1983). Design and analysis methods for longitudinal research. Annual Review of Public Health, 4, 1-23.

Dawber, T. R., Kannel, W. B., & Lyell, L. P. (1963). An approach to longitudinal studies in a community: The Framingham Study. Annals of the New York Academy of Sciences, 107, 593-599.

Kandel, D. B. (Ed.). (1978). Longitudinal research on drug use: Empirical findings and methodological issues. New York: John Wiley & Sons.

Stallones, R. A. (1966). Prospective epidemiologic studies of cerebrovascular disease. In Cerebrovascular disease epidemiology: A workshop (Public Health Monograph No. 76). Washington, DC: U.S. Government Printing Office.


(c) David F. Duncan, 2007
originally published by Macmillan Publishing Co., 1988

Powered by WebRing.