Case-Control Study

Case-control studies are a fundamental study design used in epidemiology. This nonexperimental, or observational, study design ascertains the differences in suspected exposures between individuals with a disease of interest (“cases”) and comparable individuals who do not have the disease (“controls”). Analysis yields an odds ratio (OR) that reflects the relative probabilities of exposure in the two populations. Case-control studies can be classified as retrospective or prospective, depending on when cases are identified in relation to measurement of exposures. The case-control study was first used in its modern form in 1926, but grew in popularity in the 1950s following the publication of several seminal case-control studies carried out by Sir Richard Doll and others that established the link between smoking and lung cancer.

Case-control studies are advantageous because they require smaller sample sizes and thus fewer resources and less time than other observational studies. This design is the only viable option for studying exposure related to rare diseases, for diseases that are a consequence of long exposure, and is an ideal design for conditions that potentially result from multiple exposures.

The primary challenges in designing a case-control study lie in appropriate selection of cases and controls, with the objective of minimizing confounding, or the differential distribution of exposures that are due to factors other than disease status. Steps can be taken in the study design and analysis to minimize confounding. A well-designed study must include controls that are selected from the same source population from which cases are selected. Additionally, cases and controls may be “matched” by relevant characteristics that might otherwise result in a biased estimate of the association between disease and exposure. With reference to analysis, multivariate analysis, usually logistic regression, allows the primary association between exposure and disease to be “adjusted” for the effect of measured confounders.

Bias might also result if exposures cannot be measured or recalled equally in both cases and controls. Choosing from a population with a disease different from the one of interest but of similar morbidity may minimize that probability, as these individuals are more likely to recall exposures or to have their information recorded to a level comparable to cases.

Care must also be taken in interpreting the results of a case-control study. Odds ratios are not equivalent to absolute risk difference, although given a rare-enough exposure, OR is considered to be nearly equal.