Family-based analysis of a myocardial infarction endophenotype: comparison of sampling designs

The power of linkage analysis of a quantitative disease endophenotype was compared for the following family selection designs: 1) Random samples: randomly chosen nuclear families, 2) "coronary artery calcification (CAC)" samples: selection of each nuclear family through a proband with abnormally high levels of the simulated quantitative endophenotype, CAC, and (3) "MI" samples: selection of each nuclear family through a disease affected proband, in this case a proband who had been simulated to have a myocardial infarction (MI) event.We assessed the power to detect linkage to five loci (two pairs of epistatic loci and one locus with an over-dominant allele) that were modeled as determinants of the simulated CAC levels. We did this using a Haseman-Elston regression-based linkage analysis of the adjusted CAC levels that considered each locus separately and then used a multiple regression extension of the Haseman-Elston method in which we considered the allele sharing at two true epistatic loci simultaneously and their interaction as possible factors related to the squared sibpair differences in adjusted CAC.Based on comparison of the mean square root of the LOD scores, there was no one sampling design that resulted in consistently greater power for these five loci. That is, we observed significant locus-by-sampling-design interaction (p < 0.0001). We noted however, that the largest average score was observed for the epistasis between tau3 and tau4 (mean > 1.8, SE = 0.06) in the MI-selected samples and the CAC-selected samples.