Linkage studies of complex diseases have so far had limited success in producing significant and replicable results, in part owing to genetic heterogeneity. We recently reported the results of a large genome-wide linkage scan for coronary artery disease (CAD) based on 1933 families. The greatest evidence for linkage was to a region of chromosome 2, with a logarithm of odds (LOD) score of 1.86, based on the non-parametric SALL statistic, which did not reach genome-wide significance (P>0.3). Inclusion of a covariate in linkage analysis can be a powerful method of accounting for disease heterogeneity. As CAD is a heterogeneous disease, we carried out a linkage analysis of chromosome 2 incorporating covariates. Increased evidence for linkage was found when hypercholesterolemia was considered (LOD score including covariate of 4.4) reaching genome-wide significance as assessed by simulation (P=0.04). Results showed that the original evidence for linkage was largely attributable to the subset of 108 non-hypercholesterolemic affected sibling pairs. In separate linkage analyses of subsets of hypercholesterolemic and non-hypercholesterolemic sibling pairs, the maximum LOD scores were 1.09 in the former group and 3.74 in the latter. This result illustrates the potential to increase the power of linkage analysis in the presence of heterogeneity by inclusion of covariates. This linked locus on chromosome 2 should now be investigated further to identify the gene(s) influencing risk of CAD in subjects with a normal level of total cholesterol. Candidate genes include the interleukin 1 cluster and two potential regulators of high-density lipoprotein cholesterol level, PLA2R1 and OSBPL6.