What is a case-control study?

Setia describes the design as one in which participants are chosen by outcome: people with the outcome are cases, those without are controls, and exposure is then assessed in both groups. STROBE adds that the case sample may be all or a large fraction of the available cases, while the control sample is usually only a small fraction of the people without the outcome; controls represent the population from which the cases arose. It says case and control selection is crucial to interpreting results and that controls should in general mirror the source population of the cases.

The effect measure is the odds ratio, the odds of exposure in cases divided by that in controls. Because investigators choose how many cases and controls to include, Setia states that incidence and prevalence cannot be estimated, and that the odds ratio approximates relative risk only under some assumptions. STROBE explains that, depending on how cases and controls were sampled, the odds ratio is read as a risk ratio, a rate ratio or a prevalence odds ratio.

According to Setia, the design suits rare outcomes and long latent periods but not rare exposures, is prone to selection and recall bias, and may leave unclear which came first, exposure or outcome. STROBE describes recall bias as arising when cases recall or report past exposures more or less accurately or willingly than controls, and notes that selection bias may arise if the probability of being included as a case or control is associated with exposure.