What is a cohort study?

STROBE describes a cohort study as one in which investigators record participants and their exposures at baseline, follow them for a period, and then assess who develops the outcome. According to Setia, nobody has the outcome at the start and participants are selected on exposure status. The comparison is between exposed and unexposed people or between exposure levels, and several outcomes can be examined in one cohort (STROBE).

In a closed cohort, which starts with a fixed number of participants, STROBE says both cumulative incidence (risk) and incidence rates can be estimated; comparing exposed with unexposed groups then yields a risk ratio or a rate ratio. The same document notes that “relative risk” can stand for either. Setia points out that exposure is measured before the outcome, so time order is clear.

A cohort is prospective when data are collected going forward and retrospective when they come from existing records; Setia adds that mixed designs exist. Setia calls prospective cohorts time-consuming and costly, and retrospective cohorts quick and relatively inexpensive but warns that their measurements may not be very accurate or what the study needs; Euser and colleagues add that these measurements were often made for another purpose, such as patient care. Euser and colleagues consider cohorts suited to rare exposures, whereas Setia says they may be inefficient for rare outcomes. Confounding and selection bias (Euser and colleagues) and losses to follow-up (Setia) can distort results.