Family-Based Case–Control Studies

Case–control studies are used often in epidemiologic studies to investigate the association between disease and one or more risk factors. With increasing frequency, the set of risk factors being considered includes genotypes at one or more susceptibility, candidate, or marker loci (see Markers). The goals of association studies will differ, depending on the state of knowledge about a given disease. For example, once a susceptibility locus has been identified, the goals include estimating the relative risk and penetrance associated with specific mutations, and testing for interaction with environmental exposures or other genes [14] (see Disease–Marker Association; Gene–Environment Interaction). If a candidate locus has been identified on the basis of a biological hypothesis that relates gene function to phenotypic expression, then the primary goal is testing the null hypothesis of no association between the locus and disease. Finally, multiple tests of association with finely spaced markers in a chromosomal region that has been previously established to contain loci linked to the disease (see Linkage Analysis, Model-Based) may be used in the hope of detecting linkage disequilibrium with a disease locus. For diseases, candidate genes may be part of a larger hypothesized disease pathway that includes other genes and/or environmental exposures. For example, Gilliland et al. [13] have hypothesized that asthma and other respiratory phenotypes are related to air pollution through a pathway of oxidative stress that includes several genetic loci (see Causation). Even for BRCA1, a major susceptibility locus for breast cancer that substantially increases risk by itself, there is evidence for some effect modification by use of oral contraceptives [56]. It is therefore important in studies of candidate genes to consider not only the main effect of the gene but also its interactive role with other genes and/or environmental factors. Traditional unmatched or matched case–control studies [1] may not be optimal for the study of genes. A potential problem is that estimates of genetic effect are subject to confounding when cases and controls differ in their ethnic backgrounds. This phenomenon, also known as population stratification bias [2, 23], can occur when both disease risk and genetic mutation frequencies vary among ethnic groups (see Bias in Case–Control Studies; Disease–Marker Association). If all or part of the disease-risk variation is due to factors other than the candidate gene (e.g. environmental exposures, a second gene), and those other risk factors also vary among ethnic groups, then a spurious association with the candidate gene may occur simply due to the indirect correlation of its distribution with the other risk factor(s). A classic example of such confounding is the reported association between the Gm locus and non-insulin-dependent diabetes in American Indians that disappeared when the analysis was restricted to full-heritage Pima–Papago Indians [17]. To avoid the problem of population stratification bias, one can attempt to match cases to controls on ethnic background (see Matched Analysis; Matching). However, determination of ethnicity in a large-scale epidemiologic study is difficult, especially with the great diversity in cultural backgrounds that exists in the urban areas where studies are most likely to be conducted. It should be noted here that the degree to which population stratification can cause spurious findings for a candidate gene is the subject of current debate [57]. There has also been interest in using genetic markers to adjust for population stratification [32, 33, 37]. In this article we review family-based case–control designs that have as one of their primary features freedom from population stratification bias. These include the case–sibling and case–parent designs. We discuss methods for analyzing these designs that provide parameter estimates and hypothesis tests for genetic main effects and for gene–environment (G × E) or gene–gene (G × G) interactions (see Gene–Environment Interaction). We provide comparisons among the case–sib, case–parent, and standard case–control designs of the sample size requirements for testing these hypotheses. Finally, we review alternative study designs that make use of family data.