6.1 Cohort studies

Choosing the cohort

A cohort is chosen for the exposure it offers and for the ease of following it. Rare exposures, typically occupational or medical, are studied in special-exposure cohorts: workers in one industry, patients given a drug. Common exposures such as smoking or diet are studied in general-population cohorts, which are chosen for completeness of follow-up rather than for being representative: nurses, doctors, civil servants, residents of one town. A cohort need not be representative of any population to give valid comparisons between its exposed and unexposed members; it needs the two to be comparable.

The comparison group may be internal (members of the same cohort with less or no exposure) or external (the general population, through national rates). External comparison is cheap but suffers from the healthy worker effect: people in employment are healthier than the general population, which includes those too ill to work, so an occupational cohort usually shows lower mortality than national rates even when the work is harmful. Internal comparisons avoid this.

Follow-up

Losses to follow-up threaten validity in the way Proposition 2.5.1 described: if the reasons for loss are related to both exposure and outcome, the risks in those followed are biased. Small losses that are unrelated to outcome cost only precision. The practical aim is near-complete follow-up, through record linkage to death and cancer registers where possible, and a check that the people lost resemble those followed in the characteristics recorded at entry.

Questions on this section

Stuck on something here? Ask below and it stays attached to this topic.