Abstract 13051: Unbiased Estimate of Heritability of CAD Before and After Adjustment for Traditional Risk Factors in Five NHLBI Cohorts

Abstract

Introduction: The relative contribution of traditional risk factors (TRFs) to the genesis of CAD remains uncertain. Cohort studies report a high population attributable risk for TRFs, but only a third of the GWAS loci for CAD mediate risk through TRFs. Furthermore, these loci explain a small fraction of the overall heritability of CAD based on twin studies making it unclear whether these patterns will persist and whether estimates from twin studies are inflated. Hypothesis: Unbiased estimates of narrow sense heritability ( h 2 ) of CAD using contemporary analytic techniques applied to very distantly related individuals prior to and after adjustment of TRFs have not yet been reported. We hypothesized that such estimates would confirm a high heritability of CAD and that a substantial proportion of the genetic variance would be independent of TRFs. Methods: We used ~7 million genome wide imputed single nucleotide polymorphisms (SNPs) and a mixed linear model implemented in Genome-wide Complex Trait Analysis to estimate the h 2 of CAD from genomic relatedness among participants of five prospective NHLBI cohorts - ARIC, CHS, FHS, MESA, and WHI. To guard against bias from differences in minor allele frequencies (MAF) as well as region-specific heterogeneity in linkage disequilibrium (LD) between variants used in our analyses and causal variants, we used MAF and LD stratified multicomponent genetic restricted maximum likelihood. We ran 2 models assuming a population prevalence of 8%. The first model was not adjusted for any covariates and the second was adjusted for all TRFs as well as use of antihypertensive and statin drugs. Results: We analyzed a total of 23,522 participants including 3520 that developed incident clinical CAD during follow up. CAD h 2 was 39% (±3.1%) in our unadjusted model and 32% (±3.2%) in our fully adjusted model. A large majority (~95%) of the heritability in both models was derived from common SNPs with MAF >1%. Conclusions: CAD is a moderately heritable trait with a large fraction of its genetic variance being a result of processes unrelated to TRFs. Prior family based estimates of the h 2 of CAD do not appear biased upwards. Consistent with many other complex traits, a majority of the genetic variance can be explained by common SNPs with very modest effects.