ABSTRACT
Case-control studies can sample more than one case group and/or more than one control group for a number of reasons. Sometimes multiple case or control groups may be sampled to examine the robustness of an observed association between exposure and disease. As one example, in a clinic-based study of alcohol consumption and lung cancer, one might use both bone-fracture controls and acne controls to protect against the possibility that, unbeknownst to investigators, one or the other of these conditions is associated with alcohol use. As another example, if the association of prior vasectomy with prostate cancer is stronger for less aggressive disease, this may be because vasectomized men are more likely to be screened and diagnosed, and not because vasectomy causes cancer (Stanford et al., 1999). In addition, when different subtypes of disease lead to different treatment options and have different prognoses, it can be of public health importance to determine whether an exposure is associated with the incidence of all, or only some, subtypes of the disease. For example, Britton et al. (2002) compared joint hormone receptor positive and hormone receptor negative breast cancer cases to population controls.
