| Both a covariate-based linkage analysis and a linkage disequilibrium mapping study were designed to investigate genetic susceptibility for inflammatory bowel disease (IBD). A novel application of a propensity score (PS, conditional predicted probability of being affected given the observed covariate data) was tested to incorporate multiple covariate data with a single scalar covariate. Simulations to assess power and type I error rates, and analyses with the PS in two genome wide linkage studies were performed. Simulation results suggest that including the PS consistently increased power compared to including no covariates and to including multiple covariates used in defining the PS. Type I error rates were inflated by 20% for each additional covariate. Increasing sample size reduced the inflated rates, suggesting a non-asymptotic likelihood ratio distribution for a sample size of less than 1,000 families (requiring permutation-based significance values). Results from the linkage analysis for alcoholism (covariate data on both affected and unaffected individuals) suggest that using covariates did not always improve the power to detect linkage, but using the PS generally resulted in the highest number of significant marker loci identified compared to its corresponding multiple covariate analysis. For IBD, which consists of two subtypes, Crohn's disease (CD) and ulcerative colitis (UC), no unaffected covariate data were available, so a subtype-specific PS was instead considered for linkage analysis. The IBD3 region, containing the HLA genes on chromosome 6p21, was identified as the most significant linkage region for IBD with no covariates analyzed. Incorporating smoking status at diagnosis increased the linkage evidence on chromosome 5q for IBD, and a subtype-specific PS identified new loci for both UC and CD. Secondly, a linkage disequilibrium mapping approach was designed in UC families to narrow the previously identified region of linkage in IBD3 using 37 microsatellite markers. Association analyses were performed using family-based association tests (FBAT). Several haplotypes (ranging from 65kb to 528 kb in size) were identified that independently spanned the Class I, Class II, and Class III HLA genes. Using both approaches, new evidence of genetic susceptibility for IBD were revealed. |
