MPH 510 Module 2 Cohort Studies and Relative Risk Example

Following the Doctors: Cohort Design and Measures of Association in the Smoking and Mortality Studies

Student Name

Master of Public Health Program, Aspen University

MPH 510: Epidemiology in Public Health

Instructor Name

Month Day, Year

Following the Doctors: Cohort Design and Measures of Association in the Smoking and Mortality Studies

A cohort study follows people who differ in an exposure and compares how often they develop disease. Because exposure is measured before disease occurs, the design establishes the order of events, one of the strongest pieces of evidence for causation. This paper explains the prospective cohort design and its measures of association through one of the most influential studies in public health history, the British doctors study of smoking and mortality.

The British Doctors Study

In 1951, Richard Doll and Austin Bradford Hill sent a short questionnaire on smoking habits to British doctors and then tracked deaths through official records. Doctors were chosen because they were easy to follow and their causes of death were likely to be accurately recorded. The preliminary report, after about 29 months of follow-up, found 36 deaths from lung cancer among male doctors aged 35 and older, none of them in nonsmokers, and a death rate that rose steadily with the amount smoked (Doll & Hill, 1954).

Strengths of the Design

Measuring smoking before deaths occurred removed the risk that people with illness would recall their habits differently, a weakness of case-control studies. The cohort also allowed the investigators to study many outcomes at once, including heart disease and other cancers. The main costs were time and size: rare outcomes require large cohorts followed for years.

Incidence and Relative Risk

The basic measure in a cohort study is incidence, the number of new cases divided by the population at risk over a period. Relative risk compares incidence in the exposed with incidence in the unexposed. When the ratio equals 1, exposure makes no difference; a ratio greater than 1 signals excess risk among the exposed. The worked example uses a composite cohort of 20,000 men followed for ten years.

A Worked Example

The table presents the composite data and calculations.

Interpreting the Measures

A relative risk of 11.7 means smokers in the composite cohort were nearly twelve times as likely to develop lung cancer as nonsmokers. The attributable risk expresses the excess in absolute terms: about 11 extra cases per 1,000 smokers over ten years. The attributable fraction among smokers suggests that about nine in ten of their cases would not have occurred without smoking, and the population attributable fraction estimates how much of the disease in the whole cohort could be prevented if smoking were eliminated.

Dose-Response

The preliminary report's most persuasive finding was the gradient: lung cancer death rates rose from zero in nonsmokers through light, moderate and heavy smokers (Doll & Hill, 1954). A dose-response relationship is hard to explain by chance or by a single confounder and is one of the viewpoints Hill later listed for judging causation.

Fifty Years Later

The study continued for half a century. The 50-year report on 34,439 male doctors found that men who smoked throughout adult life died about ten years earlier than lifelong nonsmokers, and that stopping at age 60, 50, 40 or 30 gained about 3, 6, 9 or almost 10 years of life expectancy respectively (Doll et al., 2004). The long follow-up turned an association into a detailed map of smoking's harms and of the benefits of quitting.

From Association to Causation

An association does not prove causation. Hill offered nine viewpoints for weighing such evidence, among them how strong and consistent the association is, whether exposure comes first, whether more exposure brings more disease and whether the link makes biological sense (Hill, 1965). The smoking studies satisfied nearly all of them: a strong association, consistent findings across countries and designs, exposure preceding disease, a dose-response gradient, biological plausibility from carcinogens in smoke and falling risk after quitting.

Limitations

The British doctors were not representative of the general population, being mostly men of higher social class. Smoking was self-reported at intervals rather than measured continuously. Loss to follow-up was low but not zero. Still, the study's internal validity, the comparison of smokers with nonsmokers within the same group, was strong, and its findings were replicated widely.

Cohort Studies Today

Modern cohorts, such as those following nurses, veterans or entire birth years, use the same logic with biological samples, repeated measurements and linkage to electronic records. They support studies of diet, physical activity, environmental exposures and genetics, and they help evaluate public health interventions by comparing outcomes among people who did and did not receive them.

Retrospective Cohorts

Not every cohort study must wait years for outcomes. A retrospective cohort uses records collected in the past, such as employment files or medical charts, to define exposure and then traces outcomes up to the present. Occupational studies of asbestos workers used this approach. The logic is the same, exposure defined before outcome, but the data already exist, which saves time at the cost of less control over measurement.

Confidence Intervals

Every relative risk is an estimate with uncertainty. A 95% confidence interval marks out the plausible span for the true value; when that span sits entirely above or below 1, the association is statistically significant at the conventional level. Large cohorts produce narrow intervals. In the composite example, with 148 cases, the interval around a relative risk of 11.7 would be wide in absolute terms but would still lie far above 1.

Public Health Uses of the Measures

Relative risk speaks to causation and to individual risk, which is why clinicians use it to counsel patients. Attributable risk and the population attributable fraction speak to public health impact, telling planners how many cases a program could prevent. A modest relative risk for a very common exposure can produce more cases than a large relative risk for a rare one, which is why both kinds of measures matter.

Conclusion

The British doctors study shows why the cohort design is central to epidemiology: measuring exposure first, following people over time and comparing incidence yields relative and attributable risks that describe both the strength and the public health impact of an exposure. Combined with Hill's viewpoints, cohort evidence can support confident judgments about cause and prevention.

References

Doll, R., & Hill, A. B. (1954). The mortality of doctors in relation to their smoking habits: A preliminary report. BMJ, 1(4877), 1451-1455. https://doi.org/10.1136/bmj.1.4877.1451

Doll, R., Peto, R., Boreham, J., & Sutherland, I. (2004). Mortality in relation to smoking: 50 years' observations on male British doctors. BMJ, 328(7455), Article 1519. https://doi.org/10.1136/bmj.38142.554479.AE

Hill, A. B. (1965). The environment and disease: Association or causation? Proceedings of the Royal Society of Medicine, 58(5), 295-300. https://doi.org/10.1177/003591576505800503