evidence that carotid artery intima-media thickness (IMT) is also associated with increasedoccurrence of CHD. Apolipoprotein E (ApoE) has been widely studied in regard to its role in lipid transportand metabolism, but the role that variation in the APOE gene plays in relation to carotid artery IMT andrisk of incident CHD remains a subject of debate. The authors examined the effect of each APOE allele(e2, s3, s4) on LDL-C and carotid IMT,as well as the association with CHD risk, in 12,491 participants ofthe ARIC study. APOE s2, s3 and s4 relative allele frequencies were determined, respectively inEuropean-Americans (0.08, 0.77,0.15)and African-Americans (0.11, 0.67,0.22). Using Cox regression, itwas shown that these alleles did not predict incident CHD in either racial group. The APOE s2 allele wasassociated with lower LDL-C and the s4 allele with higher LDL-C in both European-Americans andAfrican-Americans. The APOE s2 and s4 alleles were associated with carotid IMT measures in both racialgroups, but after adjusting for lipid parameters, only the s4 allele was associated with carotid IMTmeasures in African-Americans. This study directly addresses a major question being addressed byProject 3: Will genetic variants that predict intermediate traits also predict endpoints? IV.C.4.2.2. Rare combinations of multiple common DMA sequencevariations predict high risk subgroups. SNPs have been hypothesized to explain the geneticpredisposition to CHD in the general population. Lack of evidence for a role of such variation is fosteringpessimism about the utility of genetic information in the practice of medicine. In this study we determinedthe utility of exonic and 5' SNPs in APOE and lipoprotein lipase (LPL) when considered singly and incombination for predicting incidence of CHD in 8,456 individuals from the general population during 24years of follow-up. In men,LPL D9N improved prediction of CHD (p=0.03) beyond smoking, diabetes andhypertension. The group of men heterozygous and homozygous for the rare D9N variant had a hazardratio of 1.69 (95% confidence interval =1.10-2.58) relative to the most common genotype. Pairwisecombinations of D9N with -219G>T in APOE and N291S and S447X in LPL significantly improved theprediction of CHD (p=0.05 in women, p=0.04 in men, p=0.03 in men, respectively) beyond smoking,diabetes and hypertension, and identified subgroups of individuals (n=6 to 94) with highly significanthazard ratios of 1.92to 4.35.These results were validated in a case-control study (n=8,806). This studyestablishes that combinations of SNPs in the APOE and LPL genes identify subgroups ofindividuals with substantially increased risk of CHDbeyond traditional risk factors. This result isconsistent with the heterogeneity of the etiology of CHD and the hypothesis that differentcombinations of variations in susceptibility genes are predictors in different subsets of those withCHD. IV.C.4.2.3. Using PRIM to identify SNPs that reproducibly predictvariation in disease risk beyond that afforded by established risk factors. All predictor variables arenot expected to be present and relevant in all cases of disease. When population strata are not defined,PRIM107'109 is an alternative strategy to traditional logistic regression for identifying partitions ofobservations with increased prevalence that are also predictive of the risk for unstudied members of thepopulation of inference. Multiple mutually exclusive partitions are produced, each grouping togetherobservations with equivalent values of significant predictor variables. The predictor variables will vary frompartition to partition. Details of PRIM and its application to AIMS 3.1-3.3 are presented in SectionIV.C.3.4.2.4 page 174. The results of our experience in the application of this analytical strategy to datacollected in a large population-based study of CHD events in Copenhagen, Denmark110 are summarizedin Figure IV.C.4.2.3.1, page 179.The prevalence of disease in the sample of 5,455 individuals studiedwas 0.095. We identified six partitions using traditional risk factors (age,gender, smoking, diabetes andhypertension) with prevalence estimates ranging from 0.034 to 0.195. Five of these partitions were furtherPHS 398/2590 (Rev.09/04, Reissued 4/2006) Page 178 Continuation Format Page
Showing the most recent 10 out of 76 publications
