The idea of observing people in their natural setting in order to gain insight into health determinants goes back a very long way. Hippocrates (circa 400 BCE to circa 370 BCE) urged us to consider the health of individuals in relation to “the mode in which the inhabitants live, and what are their pursuits, whether they are fond of drinking and eating in excess, and given to indolence, or are fond of exercise and labor and not given to excess in eating and drinking” (Hippocrates, 400 BCE). Here we recognize the seed of the observational cohort design, that is, comparing the long term health experiences of groups of individuals based on personal characteristics and exposures.

First century (CE) Roman authors noted differences in the patterns of diseases among various worker groups. Specific references to the diseases of slaves, sulfur workers, blacksmiths, and miners were made by the poets Pliny, Martial, and Luca (Rosen, 1993).

However, it was not until the 18th century when the focus on the health of worker cohorts met an important milestone with the publication of the treatise De Morbis Artificum Diatriba (“Diseases of Workers,” 1713) by Bernardino Ramazzini (1633–1714). De Morbis documented the effects of dozens of hazardous environmental exposures, such as specific chemicals, dusts, and abrasives. In addition, it documented the ill-effects of specific occupational practices and lifestyles. For example, the sedentary lifestyle and constant bent work posture of cobblers was cited by Ramazzini (1713) as a cause of ill-health.

In 1755, the surgeon Percival Pott (1713–1788) observed enormously elevated rates of scrotal cancer in chimney sweeps. He attributed this to the lodgment of soot in the rugae of the scrotum of the chimney sweeps, which is perhaps the first identification of an environment carcinogen.

Eighteenth century French physicians such as Louis and Pinel brought observational cohort studies into a clinical setting. In one study, Pierre Charles Alexandre Louis (1787–1872) observed superior cure rates in pneumonia patients who experienced delayed bloodletting compared with those who experienced early treatment. Phillippe Pinel (1745–1826) used the clinical histories of patients with mental illnesses to demonstrate superior cure rates at institutions that practiced humane methods of treatment compared with the standard inhumane practices of the time.

The Victorian physician and statistician William Farr (1807–1883), who is usually associated with vital statistics studies in open populations, also understood the importance of the longitudinal observation of individuals, declaring individual subjects in studies of health “should be followed from the beginning to the end; every death or recovery should be recorded.” Farr (1838) applied this concept in studying the prognosis of smallpox patients using sophisticated survival analysis techniques (Gerstman, 2003).

The 20th century brought with it the development of the modern cohort study. The British scientist Janet Elizabeth Lane-Claypon (1877–1967) reported on the longitudinal results of weight gain in infants fed either boiled cows’ milk or human breast milk (1912). One year later, the German physician Wilhelm Weinberg (1862–1937) published the results of a large retrospective cohort study comparing the health experiences of 18 212 children whose fathers and mothers had previously died of tuberculosis with that of an “nonexposed” cohort of 7574 children of parents who died of non-tubercular causes (Morabia and Guthold, 2007).

In 1914, Joseph Goldberger (1847–1929) published cohort observations on pellagra, noting the absence of pellagra in nurses and ward attendants at hospitals in which 98 cases of pellagra had occurred in the patient population. Although the common belief at the time was that pellagra was infectious, this suggested to Goldberger that pellagra was not contagious.

It was not until 1935, however, that the first recorded use of the term cohort made its appearance in epidemiology when Wade Hampton Frost referred to rates of tuberculosis in generational cohorts (Doll, 2001). Frost’s generational cohort studies are discussed in the last section of this chapter.

By the middle of the 20th century, the epidemiologic transition made it clear that large-scale long-term follow-up studies were needed in order to untangle the causes of chronic diseases in human populations. These modern cohort studies initially addressed cigarette-related diseases, cancers, and heart disease. One such study from this era—The Framingham Heart Study—was introduced as Illustrative Example 7.1. Another historically important cohort study from this era was the British Doctors Study.

Before embarking on a cohort investigation, it is essential to obtain the cooperation of the study population. One of the reasons the landmark Framingham Heart Study was based in Framingham was because of its supportive population and health-care system (Dawber et al., 1963). The same can be said of The British Doctors Study. As another example, the Nurses’ Health Study selected this population for long-term follow-up because nurses represented a large group of cooperative, health-conscious women who would be compliant and relatively easy to follow (Belanger et al., 1978).

However, even with a highly cooperative population, a certain percentage of individuals will refuse to participate. The aforementioned British Doctors, Framingham, and Nurses cohort studies had, for example, initial nonresponse rates that were all around 30%, (Doll and Peto, 1976; Dawber et al., 1963; Belanger et al., 1973).

In addition, subjects who enroll in a study may become lost to follow-up, refuse to continue in the study, or die from unrelated causes during the course of the study. These individuals are referred to as withdrawals. For example, the Framingham study began with 5209 subjects attending the initial set of examinations. Over the next ten years there were 950 withdrawals leaving the cohort with 4259 study subjects (Framingham Heart Study, 2011). This amounts to a ten-year withdraw rate of about 18%.

Nonresponses and withdrawals raise a potential for a “selection bias” in a cohort study. However, if nonresponse and withdrawal are independent of the exposures and diseases being studied, observed association will remain unbiased.

Once enrolled in a cohort study, each study subject is followed until:

• the subject withdraws from the study
  
• the study ends, or
  
• the study outcome is experienced.

The method of identifying the cases in a cohort study will depend greatly on the type of disease or health outcome being studied and the diagnostic and technologic resources that are available. Cases are ascertained based on criteria referred to as the case definition. The case definition is then uniformly applied to screen for and confirm cases using relevant diagnostic and other information. See Chapter 16 for additional details regarding the construction of case definitions.

The objective of a cohort study is to provide accurate information about the independent effects of study exposures on health outcomes. To accomplish this goal, like-to-like comparisons are necessary. Without like-to-like comparisons, differences in incidence found at the end of the study could be due to factors other than the study exposure. This phenomenon is known as confounding.³

Conceptually, the ideal nonexposed cohort would consist of the same individuals as the exposed cohort had these individuals been nonexposed. Of course this is counterfactual, that is, not possible in fact. Nevertheless, this counterfactual ideal defines a way to think about suitable comparisons. To address this issue of comparability, cohort studies compare the distribution of risk factors and other attributes in exposure groups at the onset of the follow-up period. An example follows.