Current algorithms used for assessing risk of coronary heart disease (CHD) events are based on established clinical risk factors, yet these algorithms fail to predict many CHD events (1- 2). Because genetics influences predisposition to CHD, evaluating genetic polymorphisms that are associated with risk of CHD may help to improve CHD risk assessment (3- 4). Studies that investigate the association between CHD and genetic polymorphisms may test polymorphisms in candidate genes, which provide biologic plausibility to any observed association. An alternative approach is to test polymorphisms in an unbiased collection of genes (5).

As the technology for genotyping polymorphisms has improved, the number of polymorphisms reported to be associated with CHD has increased; however, many of these associations are not confirmed in subsequent studies (6). Confirmation of associations between genetic polymorphisms and CHD can be difficult to achieve for a variety of reasons. The typically modest risk for CHD associated with each genetic polymorphism can lead to false-positive associations in initial studies or to false-negative associations in confirmation studies. Additionally, the case-control design of an initial study can affect subsequent confirmation, because unanticipated biases in the selection of subjects for case-control studies can lead to spurious associations. Thus, polymorphisms reported to be associated with CHD should be further investigated in multiple studies, including prospective studies, which may avoid some of the inherent bias of case-control studies.

We selected 35 polymorphisms, which had been associated with cardiovascular disease in previous studies, for further investigation. Ten of these polymorphisms are in 9 candidate genes and have been reported to be associated with cardiovascular disease or intermediate phenotypes in 2 studies. Four of these 10 polymorphisms are in genes related to hemostasis: rs1126643 in ITGA2 (integrin alpha 2) has been associated with nonfatal myocardial infarction (MI) in younger patients (7- 8); rs1866389 in THBS4 (thrombospondin 4) with premature MI (9- 10); rs1799768 in SERPINE1 (plasminogen activator inhibitor-1) with MI and with high plasma plasminogen activator inhibitor-1 activity (11- 12); and rs1799963 in F2 (coagulation factor II) with MI in young women (13- 14). Two of the 10 polymorphisms are in genes related to hypertension: a 278-basepair insertion–deletion polymorphism in ACE (angiotensin I-converting enzyme) has been associated with MI (15- 16); and rs5186 in AGTR1 (angiotensin II receptor type 1), with cardiovascular risk independent of blood pressure (17- 18). The remaining 4 of these 10 polymorphisms are in genes involved in vascular biology: rs1041981 in LTA (lymphotoxin alpha) has been associated with MI (19- 20); rs3025058 in MMP3 (matrix metallopeptidase-3) with progression of coronary atherosclerosis (21- 22); rs429358, the ε4 polymorphism in APOE (apolipoprotein E), with CHD (23- 24); and rs405509 in the APOE promoter with increased risk of MI (25) and coronary artery disease (23). The other 25 of the 35 polymorphisms had been previously found to be associated with MI in a case-control study that tested putative functional single nucleotide polymorphisms (SNPs) present in an unbiased collection of noncandidate genes (26).

In the present study, we asked if these 35 polymorphisms were associated with MI in the placebo arm of the CARE (Cholesterol and Recurrent Events) study and with CHD in the placebo arm of the WOSCOPS (West of Scotland Coronary Prevention Study) trial and if the risk associated with these polymorphisms could be reduced by pravastatin treatment.

We conducted genetic association studies in populations derived from 2 prospective trials that assessed the effect of pravastatin in the prevention of MI and CHD: the CARE and WOSCOPS trials. The inclusion criteria for the CARE study, a secondary prevention trial, and the WOSCOPS, a primary prevention trial, have been described elsewhere (27- 29).

The CARE study included 3,847 Caucasian patients, and the present study comprised 2,913 Caucasian patients who provided deidentified deoxyribonucleic acid (DNA) for genetic studies. We used a composite end point of confirmed fatal or nonfatal MI to focus on the genetics of MI (nonfatal MI constituted 86.7% of the events in this composite end point). The DNA was collected approximately 1.5 years after the start of the CARE trial. Therefore, the proportion of patients with fatal MI was higher in those who did not provide DNA than in those who did provide DNA; however, there was no significant difference for the end point of the present genetic study (fatal or nonfatal MI) between those who provided DNA and those who did not (30). The CARE trial genetic cohort was restricted to Caucasian patients, because the number of non-Caucasian patients (7.3% of the CARE trial population) did not provide sufficient power for a separate analysis.

The present genetic study of the WOSCOPS trial was derived from a previously described prospective nested case-control study, which included all of the 580 on-trial CHD events from the WOSCOPS cohort as case subjects and 1,160 control subjects matched to case subjects by age and smoking (29). The present genetic study of the WOSCOPS trial included all patients in the nested case-control study for whom sufficient DNA was available for genotype analysis: 481 case and 1,086 control subjects. Because the control subjects had been matched to case subjects for age and current smoking status, the present genetic study used the same CHD end point as that reported in the prospective nested case-control study (death from coronary heart disease, nonfatal MI, or revascularization procedures). The baseline characteristics for patients included in these genetic studies of CARE and WOSCOPS are presented in (Table 1). All patients enrolled in CARE and WOSCOPS provided written informed consent. The CARE study was approved by the institutional review boards of the Brigham and Women’s Hospital and all participating centers. The WOSCOPS study was approved by the ethics committees of the University of Glasgow and all participating health boards.

Genotypes were determined using multiplexed polymerase chain reaction-based amplification of genomic DNA followed by multiplexed allele detection using oligonucleotide ligation as described by Iannone et al. (31) (Online Appendix). Primer sequences are available upon request. This multiplex panel was initially intended to test 10 polymorphisms in 9 candidate genes (Table 2) that others had previously reported to be associated with cardiovascular disease in 2 studies. We investigated the 278-basepair insertion–deletion polymorphism in ACE using an assay for an SNP (rs4344) reported to be in strong linkage disequilibrium with this insertion–deletion polymorphism (32). Because the multiplex technology could accommodate 35 SNP assays in a single panel, we included 25 assays for putative functional SNPs that had been found to be associated with disease in primary or subgroup analyses of a single case-control study of MI, study 1 in Shiffman et al. (26). Most of these 25 SNPs were not in typical cardiovascular disease candidate genes.

To test for association between CHD and other SNPs that may be in linkage disequilibrium with the KIF6 Trp719Arg SNP, we selected 27 tagging SNPs, including Trp719Arg, in the regions flanking the Trp719Arg SNP using pairwise tagging in Tagger (33) as implemented in Haploview (34) (Online Appendix).

All reported p values are 2-sided. Differences between baseline characteristics of patients were assessed by t tests (continuous variables) or by chi-square tests (discrete variables). We assessed deviation from Hardy-Weinberg expectations using an exact test in the CARE trial cohort and in the control patients of the WOSCOPS trial (35). The power to detect significant associations with disease in the CARE and WOSCOPS trials for each of the 35 genetic polymorphisms was estimated by simulation (Online Appendix) and is shown in (Table 1) of the Online Appendix.

Cox proportional hazards models in the CARE trial and conditional logistic regression models in the WOSCOPS trial were used to assess the association of genotype with incident disease in the placebo arms (Wald tests) and were also used to assess the effect of pravastatin compared with placebo in subgroups defined by genotypes. Likelihood ratio tests were used to evaluate potential interactions between genotype and each conventional risk factor in separate regression models that included an interaction term between the risk factor and genotype.

Absolute risk reduction was estimated in the CARE trial at 5 years of follow-up using Kaplan-Meier estimates of MI-free survival of subgroups defined by KIF6 genotype. Absolute risk and absolute risk reduction in the WOSCOPS trial was projected for the first 4.9 years of follow-up using estimates of the genotype frequencies, an estimate that assumed that the outcome and treatment arm-specific genotype frequencies of subjects included in the nested case-control study were equal to the corresponding treatment arm-specific genotype frequencies in the WOSCOPS trial cohort. The SAS version 9 software (SAS Institute, Cary, North Carolina) was used for all regression models and for generating Kaplan-Meier estimates of survival. We used the Fisher method (36) of combining p values to assess the evidence for association from the combined CARE and WOSCOPS trials. In this method, under the null hypothesis of no association, the statistic S = −2 (ln [p1] + ln [p2]) and has a chi-square distribution with 4 degrees of freedom, where p1 and p2 are the p values from 2 degrees of freedom genotypic tests of association in the CARE and WOSCOPS trials, respectively. We then applied a Bonferroni correction to the combined p value of each SNP to adjust for the multiple hypothesis tests performed.

Estimates of absolute risk reduction and differences in absolute risk reduction were evaluated for statistical significance in the CARE trial by assuming the difference in the estimate divided by its standard error approximates a standard normal distribution under the null hypothesis and by Monte Carlo simulation (100,000 iterations) in the WOSCOPS trial.

In this genetic study of CARE and WOSCOPS, we investigated 35 genetic polymorphisms (Table 2). The genotype distributions for these polymorphisms were all in accord with Hardy-Weinberg expectations (p > 0.05) after Bonferroni correction for testing 35 polymorphisms.

Four SNPs (in KIF6, THBS4, CALCOCO2, and LILRA4) were associated with MI in the placebo arm of the CARE trial and 2 SNPs (in KIF6 and ARL5C) were associated with CHD in the placebo arm of the WOSCOPS trial (Table 2). The SNPs in KIF6 (p = 0.001) and ARL5C (p = 0.009) remained significantly associated with disease after combining the evidence for association from both the CARE and the WOSCOPS trials; however, only the KIF6 association remained significant (p = 0.04) after applying a Bonferroni multiple testing correction to the combined p values of the 35 SNPs tested in the present study.

The KIF6 SNP (rs20455) is a Trp719Arg polymorphism in the gene encoding kinesin-like protein 6, a member of the superfamily of molecular motors that are involved in intracellular transport (37). This gene was not an a priori candidate gene for CHD; the association between the KIF6 Trp719Arg polymorphism and CHD was identified in an investigation of polymorphisms in an unbiased collection of genes. In the placebo arm of the CARE trial, carriers of the KIF6 719Arg allele had a hazard ratio (HR) for recurrent MI of 1.50 (95% confidence interval [CI] 1.05 to 2.15) (Table 3) in a model adjusted for age, gender, smoking, history of hypertension, history of diabetes, body mass index, low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C). This HR for the 719Arg allele was similar to that of some of the conventional risk factors tested in the CARE trial (Figure 1). In the placebo arm of the WOSCOPS trial, we found that carriers of the KIF6 719Arg allele had an odds ratio (OR) for CHD of 1.55 (95% CI 1.14 to 2.09) (Table 3) in a model adjusted for history of hypertension, history of diabetes, body mass index, LDL-C, and HDL-C (case and control subjects were matched for age and smoking status).

The genotype frequencies for KIF6 Trp719Arg were 40.6%, 46.5%, and 12.8%, for Trp/Trp, Arg/Trp, and Arg/Arg, respectively, in the CARE trial cohort. In the WOSCOPS trial control group, the frequencies were 44.2%, 43.6%, and 12.2%, respectively. The KIF6 Trp719Arg SNP was not associated with any covariates used for adjustment (p > 0.13), nor did we observe any consistent interaction between genotype and the covariates used for adjustment in the CARE and WOSCOPS trials (data not presented).

The Trp719Arg SNP is located on human chromosome 6 at position 39,433,056 and corresponds to the first nucleotide of codon 719 in the full-length transcript NM_145027 (National Center for Biotechnology Information SNP database, build 36.1). To explore associations between disease and SNPs that might be in linkage disequilibrium with KIF6 Trp719Arg, we genotyped 26 other SNPs in the regions flanking the Trp719Arg SNP. We found that no SNP or haplotype in the KIF6 region was more significantly associated with both recurrent MI in the CARE trial and with CHD in the WOSCOPS trial than the KIF6 Trp719Arg SNP alone (Online Appendix).

We have previously reported that the 94Asn allele of the gene FCAR on chromosome 19 was associated with MI in the CARE trial and with CHD in the WOSCOPS trial (30). The risk associated with the KIF6 SNP seems to be independent of that associated with the FCAR SNP, because the risk estimates for carriers of the KIF6 risk allele were essentially unchanged after adjustment for FCAR Asp92Asn and traditional risk factors: In the CARE trial the HR for MI was 1.52 (95% CI 1.06 to 2.17); in the WOSCOPS trial the OR for CHD was 1.55 (95% CI 1.14 to 2.10). There was no indication of an interaction between KIF6 Trp719Arg and FCAR Asp92Asn (p = 0.48 in the CARE trial; p = 0.98 in the WOSCOPS trial).

None of the 10 polymorphisms previously reported to be associated with cardiovascular disease in multiple studies was associated with CHD in the placebo arm of the WOSCOPS trial, and only the Ala387Pro SNP in THBS4 (encoding thrombospondin 4) was associated with recurrent MI in the placebo arm of the CARE trial (Table 4). Compared with THBS4 Ala387 homozygotes, homozygotes for 387Pro had a hazard ratio for MI of 2.07 (95% CI 1.16 to 3.67) (Table 4). For 5 of these 10 polymorphisms (in THBS4, ITGA2, MMP3, SERPINE1, and F2), the power to detect association was greater than 80% in both the CARE and the WOSCOPS trials. For the polymorphism in LTA the power was 62% in the CARE trial and 72% in the WOSCOPS trial. The power to detect association was <60% in both studies for the other 4 SNPs. The power to detect association with disease for all the polymorphisms studied is shown in (Table 1) of the Online Appendix.

Because the KIF6 Trp719Arg SNP was associated with both recurrent MI in the CARE trial and CHD in the WOSCOPS trial, we asked whether carriers of the KIF6 719Arg risk allele benefited from pravastatin treatment. In the CARE trial, pravastatin treatment reduced the relative risk of MI by 37% among carriers of the 719Arg risk allele (adjusted HR 0.63, 95% CI 0.46 to 0.87) (Table 5), and among carriers in the WOSCOPS trial pravastatin treatment resulted in an OR for CHD of 0.50 (95% CI 0.38 to 0.68) (Table 5). Thus, in the present genetic study of the CARE trial absolute risk reduction by pravastatin was 4.89% (95% CI 1.82% to 7.97%) (Table 6) for carriers of the 719Arg risk allele and 1.39% (95% CI −1.94% to 4.72%) for noncarriers; when genotype was not considered in the present study of the CARE trial the absolute risk reduction was 3.47% (95% CI 1.19 to 5.74). In the present genetic study of the WOSCOPS trial, projected absolute risk reduction by pravastatin was 5.49% (95% CI 3.52% to 7.46%) (Table 6) for carriers of the 719Arg risk allele and 0.09% (95% CI −1.97% to 2.14%) for noncarriers; when genotype was not considered in the present genetic study the absolute risk reduction was 3.48% (95% CI 2.51 to 5.36).

Because carriers of the KIF6 719Arg allele benefited from pravastatin treatment in both the CARE and the WOSCOPS trials, we asked if pravastatin benefit differed between carriers and noncarriers. We observed a significant interaction between genotype and treatment in the WOSCOPS trial (interaction p = 0.01) (Table 5) but not in the CARE trial (p = 0.39).

We found that the Trp719Arg SNP in KIF6 was associated with risk of recurrent MI in the CARE trial and odds of CHD in the WOSCOPS trial, and this association with risk remained significant after correcting for multiple testing. Carriers of the KIF6 719Arg risk allele had an adjusted HR of 1.50 for recurrent MI in the CARE trial and an adjusted OR of 1.55 for CHD in the WOSCOPS trial. The KIF6 719Arg risk allele has also been recently shown to be associated with CHD in the ARIC (Atherosclerosis Risk in Communities) study (38).

Although the discovery of genetic polymorphisms that are associated with CHD may aid in the assessment of an individual’s risk of disease, a therapy that specifically counteracts the mechanism of action of the deleterious gene variant is unlikely to be immediately available. However, carriers of the deleterious gene variant might benefit from aggressive treatment of modifiable CHD risk factors. And this may be the case for carriers of the KIF6 risk allele: In both the CARE and the WOSCOPS trials, carriers of the 719Arg allele significantly benefited from pravastatin treatment. Among carriers of the KIF6 719Arg allele, pravastatin treatment in the CARE trial resulted in an absolute risk reduction of 4.9% and a relative risk reduction of 37% and in the WOSCOPS trial resulted in a projected absolute risk reduction of 5.5% and a relative risk reduction of 50%.

KIF6 encodes a kinesin, a class of motor proteins involved in the intracellular transport along microtubules; the cargos transported include membrane organelles, protein complexes, and mRNAs (37). Kinesins consist of a conserved motor domain that propels the kinesin along microtubules in an ATP dependent manner and a nonconserved tail domain that binds to its cargo (37). The tail domains contain coiled-coil structures that facilitate protein-protein interactions (39), and the KIF6 tail domain contains an alpha-helical region that is predicted by the COILS program to form a coiled-coil structure (40). The Trp719Arg polymorphism is in this predicted coiled-coil structure; therefore, the Trp719Arg polymorphism, a nonconservative amino acid change that replaces a nonpolar residue with a basic residue, might affect the cargo binding of the kinesin encoded by KIF6. Several kinesins have been implicated in the pathogenesis of chronic diseases, such as neurodegenerative diseases, type 2 diabetes, and Alzheimer’s disease (41); however, the role of KIF6 and the Trp719Arg SNP in cardiovascular disease remains to be elucidated.

Of the 10 polymorphisms previously reported to be associated with cardiovascular disease, only the THBS4 Ala387Pro SNP (9) was associated with recurrent MI in the CARE trial, and the risk allele in the CARE trial was the same as the previously reported risk allele. Although this SNP was not significantly associated with CHD in the WOSCOPS trial, the OR for CHD was higher in carriers of the THBS4 387Pro allele than in noncarriers.

One limitation of the present study is that only a small number of women were enrolled in the CARE trial and none were enrolled in the WOSCOPS trial; therefore, the association of KIF6 Trp719Arg with cardiovascular risk should be investigated in cohorts that are adequately powered for analysis in women. Additional limitations are that the end points and entry criteria of the CARE and WOSCOPS trials differed and that for 4 of the 10 polymorphisms previously reported in multiple studies to be associated with cardiovascular disease (in ACE, APOE, and AGTR1), power to detect association was <60%. Finally, because KIF6 was not an a priori candidate gene for cardiovascular disease, the role of kinesin-like protein 6 and the Trp719Arg polymorphism in the pathogenesis of cardiovascular disease remains to be determined in future studies.

Of the 35 polymorphisms tested, only the KIF6 719Arg allele was associated with both MI in the CARE trial and CHD in the WOSCOPS trial. In both the CARE and the WOSCOPS trials, carriers of the 719Arg allele received significant and substantial absolute risk reduction from pravastatin treatment.

The authors express their gratitude to the CARE and WOSCOPS patients and to Drs. John Sninsky and Koustubh Ranade for helpful comments on this manuscript, to Joe Catanese, Sheng-Yung Chang, David Lew, Dr. Diane Leong, and the Celera high throughput group for generating genotyping data, to Drs. Anand Chokkalingam, Lynn Ploughman, and Kit Lau for their input into statistical analysis, and to Dr. David Ross, Dr. Goran Gogic, Joel Bolonick, Migdad Machrus, Daniel Civello, Alla Smolgovsky, David Wolfson, and Kevin Tsai for computational biology support.

For supplemental procedure descriptions and a table, please see the online version of this article.

Association of the Trp719Arg Polymorphism in Kinesin-like Protein 6 With Myocardial Infarction and Coronary Heart Disease in 2 Prospective Studies: The CARE and the WOSCOPS Trials