Cohort Study — Longitudinal Observational Design
- Introduced by
- Doll & Hill (British Doctors Study, 1951); Snow (cholera, 1854)
- Year
- Mid-20th century (formal epidemiological design codified ~1950s)
- Type
- Observational longitudinal study design
- Family
- Clinical / epidemiology
What does it do?
A cohort study assembles a group of individuals who share a common starting point — typically freedom from the outcome of interest — and follows them over time to observe who develops the outcome.
By comparing incidence rates between exposed and unexposed subgroups, researchers can estimate relative risk and absolute risk differences. Cohort studies are the gold-standard observational design for measuring disease incidence and establishing temporal relationships between exposure and outcome.
The intuition behind the method
Think of a cohort study as a prospective documentary: you recruit a cast of participants before the plot unfolds, record who is exposed to a particular factor, and then watch over months or years to see who gets sick. Because you observe exposure before outcome, the timeline of causation is clearer than in a case-control study. The trade-off is time and cost — rare diseases may require enormous cohorts tracked for decades before enough events accumulate to draw reliable conclusions.
History
The cohort design has roots in John Snow's observations on cholera in mid-19th-century London, but the formal epidemiological cohort study was established in the early 1950s through landmark investigations such as the British Doctors Study by Richard Doll and Austin Bradford Hill (1951–1954), which demonstrated the causal relationship between cigarette smoking and lung cancer. The Framingham Heart Study, launched in 1948, exemplified the long-term prospective cohort approach to cardiovascular risk factors. Methodological foundations were codified by Brian MacMahon, Thomas Pugh, and later Kenneth Rothman.
When to use it
Use a cohort study when the research question concerns disease incidence, temporal sequence between exposure and outcome, or the natural history of a condition — and when randomisation is unethical or impractical. Cohort designs are particularly well suited to studying rare exposures (because participants can be selected on the basis of exposure status), multiple outcomes from a single exposure, and long-term effects. Do not use a cohort study when the outcome is very rare (requiring an enormous sample to observe enough events), when the study budget or timeline cannot support prolonged follow-up, or when a case-control or cross-sectional design can answer the question more efficiently.
Limitations and considerations
- Inefficient for rare outcomes — enormous sample sizes and long follow-up periods may be needed before sufficient events accumulate.
- Loss to follow-up can introduce selection bias if those who drop out differ from those who remain.
- Long duration and large sample requirements make cohort studies expensive and logistically demanding.
- Confounding by unmeasured variables cannot be eliminated through design alone, unlike in randomised trials.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Establishes temporal sequence — exposure is measured before outcome, supporting causal inference more convincingly than case-control or cross-sectional designs.
- Allows calculation of absolute incidence rates and relative risk, not just odds ratios.
- A single cohort can yield data on multiple outcomes simultaneously.
- Prospective measurement of exposure reduces recall bias compared with retrospective designs.
- Well suited to studying rare exposures by enriching the cohort with exposed individuals.
Applications
- Quantifying the long-term cardiovascular effects of dietary patterns, physical activity, or lipid-lowering therapy.
- Estimating the incidence of occupational diseases (e.g., mesothelioma in asbestos-exposed workers).
- Studying the natural history of chronic conditions such as diabetes, HIV, or Alzheimer's disease.
- Evaluating vaccine effectiveness and adverse events in post-authorisation surveillance.
- Investigating the effects of environmental exposures (air pollution, radiation) on population health over decades.
How it works
How it works
The investigator identifies and recruits a population at risk — individuals who do not yet have the outcome of interest. Entry criteria may be defined by age, geography, occupation, or enrollment in a health system. The inception cohort is enrolled at a common starting point, and baseline characteristics are recorded.
Exposure status (e.g., smoking, diet, medication use, occupational agent) is assessed at baseline and, in many designs, reassessed at regular intervals throughout follow-up. Accurate exposure measurement at this step determines the quality of the entire study; misclassification here biases risk estimates toward the null.
Participants are monitored over time through periodic assessments, registry linkage, administrative data, or active contact. Loss to follow-up is the main threat to internal validity at this stage: if those who leave the study differ systematically from those who remain, the estimated incidence may be biased.
Incident outcome events (new diagnoses, deaths, hospitalisations) are identified through clinical examination, medical records, disease registries, or linkage with vital statistics. Clear, pre-specified outcome definitions minimise outcome misclassification.
Incidence rates are calculated for exposed and unexposed groups. The primary effect measure is the relative risk (RR = incidence in exposed / incidence in unexposed) or, when follow-up time varies, the rate ratio. Adjusted estimates are obtained via multivariable regression (e.g., Poisson, Cox) to control for confounding.
Formula
RR = (Ie / Iu)
Common pitfalls
- Ignoring time-varying exposures — treating a baseline exposure as fixed when it changes over follow-up (e.g., smoking cessation) can misclassify exposure and bias results; use time-updated analyses or marginal structural models.
- Inadequate control of follow-up time — comparing raw event counts rather than person-time incidence rates will produce misleading results when follow-up duration varies across participants.
- Immortal time bias — periods of follow-up during which a participant cannot experience the outcome (e.g., time before a drug prescription is filled) must not be assigned to the exposed group or risk estimates will be artificially favourable.
- Healthy worker or healthy user effect — cohorts recruited from occupational or clinical settings may be systematically healthier than the general population, limiting external generalisability.
Frequently asked questions
What is the difference between a prospective and a retrospective cohort study?
In a prospective cohort study, investigators recruit participants and then follow them forward in time, collecting exposure and outcome data as events occur. In a retrospective (historical) cohort study, both exposure and outcome have already happened; the investigator reconstructs the cohort from existing records. The analytic logic is identical, but retrospective designs are faster and cheaper while being more susceptible to information bias from incomplete historical records.
How large does a cohort study need to be?
Sample size depends on the expected incidence of the outcome, the size of the effect you wish to detect, the duration of follow-up, and the desired statistical power. For common outcomes a few hundred participants may suffice; for rare outcomes with modest effect sizes, tens of thousands of participants followed for decades may be necessary. Power calculations should be performed before recruitment begins.
How is a cohort study different from a randomized trial?
In a randomised trial the investigator controls exposure assignment, which eliminates confounding by design and allows causal conclusions. In a cohort study exposure is determined by natural circumstances, so unmeasured confounders may distort the exposure-outcome association. Cohort studies are observational and therefore occupy a lower position in the hierarchy of evidence for causal questions, but they are feasible when randomisation would be unethical or impractical.
What effect measure does a cohort study produce?
The primary effect measure is the relative risk (risk ratio) when cumulative incidence is used, or the incidence rate ratio when person-time denominators are used. Cohort studies can also estimate the risk difference (attributable risk), which is directly interpretable for public health impact. Odds ratios from logistic regression are sometimes reported for convenience but approximate the risk ratio only when the outcome is rare.
How should loss to follow-up be handled?
First, minimise loss through active retention strategies. Report the proportion lost and compare baseline characteristics of completers versus non-completers. Analyse using intention-to-follow-up principles, and conduct sensitivity analyses (e.g., best-case and worst-case scenario imputations) to assess the robustness of findings. Inverse probability weighting can correct for informative censoring when predictors of dropout are measured.
Related methods
Sources
Rothman, K. J., Greenland, S., & Lash, T. L. (2008). Modern Epidemiology (3rd ed.). Lippincott Williams & Wilkins.
ISBN: 978-0781755641Copied.Doll, R., & Hill, A. B. (1954). The mortality of doctors in relation to their smoking habits: a preliminary report. British Medical Journal, 1(4877), 1451–1455.
Open DOI referenceCopied.