Cohort Study

A cohort study is an observational design that starts from exposure status: it identifies a group of people who are free of the outcome, classifies them by whether they are exposed to a factor of interest, and follows them over time to compare how often the outcome develops in each group. Because it observes outcomes after exposure is recorded, it can directly estimate incidence and relative risk, making it a central design for studying the determinants of disease.

Definition

A cohort study classifies a population free of the outcome by exposure status and follows it over time to compare the incidence of the outcome between exposed and unexposed groups, estimating the risk ratio or incidence rate ratio as the measure of association.

Scope

The entry covers exposure-based sampling, the prospective and retrospective variants of the design, the risk ratio and incidence rate ratio as its natural measures, and its principal limitations — loss to follow-up, confounding, and the cost of long observation. It treats the cohort study as a methodological topic within epidemiologic study designs, not as clinical guidance.

Key concepts

• Exposure-based sampling
• Prospective versus retrospective cohorts
• Incidence and person-time
• Risk ratio and incidence rate ratio
• Loss to follow-up
• Confounding
• Closed and open (dynamic) cohorts

Mechanisms

Investigators define a cohort that is free of the outcome and record each member's exposure status, then follow the cohort forward, counting new (incident) outcomes over person-time at risk. Because exposure is ascertained before the outcome occurs, the temporal sequence is clear and reverse causation is reduced. The design directly yields incidence in each group, and dividing these gives the risk ratio or incidence rate ratio. In a prospective cohort the investigator assembles the group and follows it into the future; in a retrospective (historical) cohort, both exposure and outcome have already occurred and are reconstructed from existing records, trading speed for dependence on the quality of those records. Validity depends on complete follow-up and on controlling confounders, since exposure is observed rather than assigned.

Clinical relevance

Cohort studies generate much of the evidence on risk factors, prognosis, and the natural history of disease, and appraising them is part of evidence-based practice. They are a reference design for understanding how incidence and relative risk are estimated from observation over time; they describe how evidence is generated and are not a basis for individual diagnostic or treatment decisions.

Epidemiology

Cohort studies are well suited to rare exposures and to questions requiring incidence or multiple outcomes from a single exposure, but they are inefficient for rare diseases or those with long latency, where they demand large samples or extended follow-up. The British Doctors Study by Doll and Hill, which followed physicians by smoking status from 1951 onward, is a landmark prospective cohort that quantified the mortality associated with smoking.

Evidence & guidelines

Reporting of cohort studies is standardized by the STROBE statement for observational research, and well-conducted cohorts are generally placed above case-control and cross-sectional studies in evidence hierarchies for questions of risk and prognosis, while remaining below randomized trials for questions of intervention effect because exposure is not randomized.

History

The prospective cohort approach matured in the mid-twentieth century alongside large studies of chronic disease, including the British Doctors Study of smoking and mortality begun by Doll and Hill in 1951 and the Framingham Heart Study of cardiovascular risk. These investigations demonstrated that following exposed and unexposed groups over time could quantify the determinants of common diseases, establishing the cohort study as a pillar of modern epidemiology.

Debates

How damaging is loss to follow-up?

Participants who drop out may differ systematically from those who remain, biasing estimates; how much attrition a cohort can tolerate, and how to handle it analytically, is a recurring methodological concern.

Can a cohort study support causal claims despite residual confounding?

Because exposure is observed rather than randomized, unmeasured and residual confounding can remain even after statistical adjustment, so how confidently a cohort association can be read as causal is contested.

Key figures

• Richard Doll
• Austin Bradford Hill
• Kenneth Schulz
• David Grimes
• Kenneth Rothman

Seminal works

• doll-hill-1954
• grimes-schulz-2002-cohort

Frequently asked questions

What is the difference between a prospective and a retrospective cohort study?

In a prospective cohort the investigator enrolls people and follows them forward in time to observe outcomes; in a retrospective (historical) cohort the exposures and outcomes have already happened and are reconstructed from existing records. Both classify subjects by exposure and compare outcome incidence.

Why can a cohort study estimate risk when a case-control study cannot?

A cohort follows a defined group over time and counts new outcomes, so it can measure incidence directly and compute risk ratios. A case-control study samples on outcome and so cannot measure incidence, which is why it reports odds ratios instead.