Cohort Study

A cohort study defines groups by exposure and follows them forward to see what happens. Because participants are counted before their outcomes occur, incidence can be measured directly and expressed as a risk ratio or a hazard ratio. A prospective cohort recruits and then waits; a retrospective cohort reconstructs the same structure from records already collected, which is faster and cheaper but limited to what was recorded.

Long-running cohorts have supplied much of what is known about chronic disease, and the Framingham Heart Study, begun in 1948, generated the modern concept of a cardiovascular risk factor. Cohorts assembled from national dispensing and claims databases are now the main source of post-approval evidence about how drugs behave in the populations trials excluded, including older patients, those with renal impairment, and people taking several other medicines.

The comparison with a randomised trial is both the point and the limitation. A cohort reaches sample sizes and durations no trial will fund, and it studies real prescribing rather than protocol-driven treatment. But exposure is chosen by patients and clinicians, so treated and untreated groups differ at baseline in ways adjustment and propensity methods reduce without eliminating, and unmeasured confounding is the permanent residual.

Two artefacts recur often enough to check by default. Immortal time bias appears when a period during which the outcome could not have occurred is assigned to the exposed group, manufacturing an apparent benefit; the new-user design exists to prevent it. Healthy-adherer bias appears because people who take medication as directed also do other things that lower risk, which is why good adherence to placebo predicts better outcomes in randomised trials.