In epidemiology, the term "case" refers to the person with a certain disease or health-related condition (including death), and "control" refers to the person without the condition. A case-control study is an observational study that compares attributes of sampled cases and controls.
The following diagram shows the basic design of a typical case-control study. First, the investigator identifies a group of people with the disease (cases) and another comparison group without the disease (controls). For both groups, each person's past exposure to a certain attribute (life style, environmental factor, genetic trait, etc.) is then determined by means of questionnaire, chart review, and/or biospecimen analysis. Then, the frequencies or levels of the past exposures in the case group are compared with that in the control group. If the cases have a higher amount of exposure than controls, the exposure may be associated with the disease.
|Case Control Study Design|
To determine if herpes simplex virus 1 (HSV1) infection was associated with the development of oral cancer, Starr et al (Cancer Res. 2001 Dec 1;61(23):8459-64) first identified 259 people with oral cancer (cases) and 444 without oral cancer (controls). The investigators then measured the HSV1 antibody levels of each individual in both the case and control groups. Of the 259 cases, 190 (73%) showed antibodies to HSV1, while 287 out of 444 controls (64%) had HSV1 antibodies. Since the percentage of those with HSV1 antibodies is greater in cases than in controls (73% vs. 64%), HSV1 infection might be positively associated with oral cancer.
The above example contains a simplified analytical reasoning for the disease-exposure association. A standard measure to examine this association in a case-control study is the 'odds ratio,' the definition and calculation of which can be found in the following web sources. In addition, the exposure status need not be dichotomous; cases and controls can be assigned into multiple (or continuous) exposure categories. Such exposure classification will require a more elaborate analytical approach, which is also provided in the web resources below.
Case-control studies can be subject to bias. Cases and controls may recall past events differently. Moreover people with disease may have different "exposure" information from people without disease, not because the exposure causes the disease but because the disease causes the exposure! Thus, some case-control studies typically may not be able to differentiate cause and effect. Finally, picking controls requires thorough considerations. Ideally, controls need to be representative of the population from which cases are drawn. Some investigators compare several different types of control groups (hospital controls, population controls, family controls) to the cases in the hopes that all of the comparisons will yield similar results. Although bias exists in case-control studies, they are often the studies possible for rare diseases. A well-designed case-control study that minimizes the potential pitfalls can indeed demonstrate the association between an exposure and a disease.
Odds ratio calculation (Blackwell Publishing)
Fundamentals of epidemiology (Univ. of North Carolina, Chapel Hill)