- Top of page
- Disclosure of Conflict of Interest
- Supporting Information
Summary. Background: Thrombosis and inflammation are critical in stroke etiology, but associations of coagulation and inflammation gene variants with stroke, and particularly factor VII levels, are inconclusive. Objectives: To test the associations between 736 single-nucleotide polymorphisms (SNPs) between tagging haplotype patterns of 130 coagulation and inflammation genes, and stroke events, in the 5888 participants aged ≥ 65 years of the observational Cardiovascular Health Study cohort. Patients/Methods: With 16 years of follow-up, age-adjusted and sex-adjusted Cox models were used to estimate associations of SNPs and FVIIc levels with future stroke. Results: Eight hundred and fifteen strokes occurred in 5255 genotyped participants without baseline stroke (748 ischemic strokes; 586 among whites). Among whites, six SNPs were associated with stroke, with a nominal P-value of < 0.01: rs6046 and rs3093261 (F7); rs4918851 and rs3781387 (HABP2); and rs3138055 (NFKB1A) and rs4648004 (NFKB1). Two of these SNPs were associated with FVIIc levels (units of percentage activity): rs6046 (β = − 18.5, P = 2.38 × 10−83) and rs3093261 (β = 2.99, P = 3.93 × 10−6). After adjustment for age, sex, race, and cardiovascular risk factors, the association of FVIIc quintiles (Q) with stroke were as follows (hazard ratio; 95% confidence interval): Q1, reference; Q2, 1.4, 1.1–1.9); Q3, 1.1, 0.8–1.5); Q4, 1.5, 1.1–2.0); and Q5, 1.6, 1.2–2.2). Associations between SNPs and stroke were independent of FVIIc levels. Conclusions: Variations in FVII-related genes and FVIIc levels were associated with risk of incident ischemic stroke in this elderly cohort, suggesting a potential causal role for FVII in stroke etiology.
Stroke is a major cause of morbidity and mortality in the developed world; in the USA, one in six men and one in five women suffer a stroke in their lifetime, with stroke being responsible for 17% of all deaths . Thrombosis plays a key role in ischemic stroke; after disruption of the vessel wall, thrombus is formed, and either disrupts blood flow at the site of injury or breaks off and embolizes to where the occlusion occurs [2,3]. Inflammation is also associated with ischemic stroke pathophysiology, and may relate to changes in the composition of blood or of the vessel wall . Risk factors for stroke are not as well characterized as for myocardial infarction (MI), and few prospective studies have evaluated associations of hemostatic and inflammation biomarkers with stroke risk .
Proteins related to hemostasis and thrombosis have long been prime biomarker candidates for stroke risk, but many are difficult to measure, owing to high within-person variability and difficulty in standardizing assays. Thus, measurement of gene variants may reveal associations that cannot be determined by assessment of phenotypes.
We studied polymorphisms in genes related to hemostasis and inflammation in relation to stroke risk in the Cardiovascular Health Study (CHS) cohort, and evaluated whether the protein products of genes related to stroke (where possible) were associated with stroke and other cardiovascular disease (CVD) outcomes. We also assessed whether these protein products mediated any of the associations between the single-nucleotide polymorphisms (SNPs) and CVD. Findings may provide insights into the pathophysiology of stroke that can be exploited for risk stratification, new interventions for primary prevention, or perhaps novel treatment approaches.
- Top of page
- Disclosure of Conflict of Interest
- Supporting Information
Through June 2005 (up to 15 years of follow-up), the 5759 participants without baseline stroke from the original and minority cohorts experienced 815 incident strokes: 748 were ischemic and 106 were hemorrhagic (the overlap resulted from incident strokes of different types). Through the follow-up period, the 4382 CHS white participants without baseline stroke experienced 713 incident strokes: 586 ischemic and 79 hemorrhagic. Through June 2005, there were 736 incident MIs in the entire cohort.
Of the 736 SNPs examined, six were associated with stroke, with nominal P-values < 0.01 (Table 1). One SNP (rs6046) was located within the coding region and another (rs3093261) within the 3′-flanking region of F7 (between F7 and the FX gene) on chromosome 13. Two SNPs (rs4918841 and rs3781387) were located within introns of HABP2 on chromosome 10. Two SNPs were located near inflammation-related genes: rs3138055 within the 3′-flanking region of I-kappa-B-alpha (NFKBIA) on chromosome 14, and rs4648004 within the intron of NF-kappa-B p50 subunit (NFKB1) on chromosome 4. Tables S2 and S3 present the findings from the SNP look-up in the CHARGE Consortium. Within the CHARGE Consortium, none of the six SNPs that we identified was replicated for all-cause stroke or ischemic stroke; however rs3138055 (or any proxy for this SNP) was not assessed in the CHARGE Consortium, and rs4648004 was measured through a proxy SNP only (rs3960787, P = 0.02 for ischemic stroke).
Table 1. Coagulation and inflammation gene single-nucleotide polymorphisms (SNPs) associated with risk of incident stroke in Cardiovascular Health Study whites
|SNP no.||Gene symbol||Gene name||Chromosome||Alleles||Location||Nominal P-value*|
|rs4918851||HABP2||Factor VII-activating protease||10||A/C||Intron||0.0007|
|rs3781387||HABP2||Factor VII-activating protease||10||G/A||Intron||0.0051|
|rs4648004||NFKB1||NF-kappa-B p50 subunit||4||G/A||Intron||0.0092|
Table 2 presents age-adjusted and sex-adjusted hazard ratios (HRs) for associations of the six SNPs with ischemic and hemorrhagic stroke among white participants. With a corrected P-value of 0.004 (α = 0.05/12; six SNPs × two outcomes), three of the six SNPs were associated with ischemic stroke: both of the F7 SNPs (rs6046, HR 0.73, P = 0.002; rs3093261, HR 1.21, P = 0.002) and one of the HABP2 SNPs (rs4918841, HR 0.75, P = 0.0002). None of the six SNPs was associated with hemorrhagic stroke or MI (in an exploratory analysis; all P-values > 0.004; Table 2).
Table 2. Age-adjusted and sex-adjusted hazard ratios (HRs) of single-nucleotide polymorphism (SNPs) for incident stroke among Cardiovascular Health Study whites by stroke type
|Gene symbol||SNP no.||Ischemic stroke (n = 586)||Hemorrhagic stroke (n = 79)||Myocardial infarction (n = 669)|
|HR** (95% CI)||P-value*||HR** (95% CI)||P-value*||HR** (95% CI)||P-value*|
|HABP2||rs4918851||0.75 (0.64–0.87)||0.0002||1.08 (0.73–1.60)||0.69||1.01 (0.88–1.15)||0.94|
|F7||rs6046||0.73 (0.60–0.89)||0.002||0.74 (0.44–1.25)||0.26||1.01 (0.85–1.19)||0.94|
|F7||rs3093261||1.21 (1.07–1.36)||0.002||0.88 (0.64–1.22)||0.46||0.99 (0.88–1.11)||0.88|
|HPBP2||rs3781387||0.77 (0.63–0.93)||0.006||0.91 (0.55–1.51)||0.71||1.01 (0.86–1.19)||0.91|
|NFKB1A||rs3138055||0.83 (0.72–0.95)||0.007||0.82 (0.57–1.18)||0.29||1.00 (0.89–1.14)||0.95|
|NFKB2||rs4648004||1.14 (1.01–1.28)||0.03||1.40 (1.02–1.92)||0.04||0.95 (0.85–1.07)||0.40|
Table 3 shows that the two F7 SNPs (rs6046 and rs3093261) were associated with FVIIc levels in CHS whites; each copy of the minor allele of rs6046 was associated with an 18.5% point lower FVII level (P = 2.38 × 10−83), and each copy of the minor allele of rs3093261 was associated with a 2.99% point higher FVIIc level (P = 3.93 × 10−6). HABP2 SNPs were not associated with FVIIc levels. The two F7 SNPs (rs6046 and rs3093261) were associated with FVIIc quintiles (both P < 0.001).
Table 3. Unadjusted correlation of stroke-associated single-nucleotide polymorphisms (SNPs) and baseline factor VIIc level among Cardiovascular Health Study whites (N = 4286)
|Gene||SNP no.||β (%)||SE||R2||P-value*|
|F7||rs6046||− 18.5||< 0.01||0.084||2.38 × 10−83|
|F7||rs3093261||2.99||0.64||0.005||3.93 × 10−6|
Given the association of the F7 SNPs with stroke, we assessed the association of FVIIc levels with CVD risk factors (Table 4). FVIIc was lower with greater age, in males than in females, and in blacks than whites. FVIIc was higher with hypertension, higher HDL and LDL cholesterol, greater BMI, presence of CKD, and higher CRP. FVIIc was not associated with smoking status or diabetes.
Table 4. Baseline associations of factor VIIc quintiles with cardiovascular risk factors and stroke-associated single-nucleotide polymorphisms
| ||FVIIc quintiles||P-value (trend)|
|FVII range (%)||41–99||100–113||114–126||127–144||145–346|| |
|Number of strokes||136||173||145||172||189|| |
|Average follow-up (years)||10.5||11.2||11.6||11.5||11.7|| |
|Age (years ± SD)||73.3 ± 5.8||73.1 ± 5.6||72.9 ± 5.6||72.8 ± 5.5||72.0 ± 5.2||< 0.001|
|Male, n (%)||759 (66)||622 (55)||475 (42)||362 (31)||225 (19)||< 0.001|
|Black, n (%)||283 (25)||213 (19)||143 (13)||137 (12)||99 (8)||< 0.001|
|Current smoking, n (%)||136 (12)||145 (13)||135 (12)||147 (13)||125 (11)||0.33|
|Diabetes, n (%)||347 (30)||305 (27)||327 (29)||350 (30)||375 (32)||0.19|
|Hypertension, n (%)||475 (41)||474 (42)||463 (41)||551 (47)||578 (49)||< 0.001|
|Systolic BP (mmHg ± SD)||135 ± 22||136 ± 22||136 ± 21||137 ± 23||139 ± 22||< 0.001|
|HDL cholesterol (mg dL−1 ± SD)||51 ± 14||53 ± 15||54 ± 15||56 ± 16||57 ± 18||< 0.001|
|LDL cholesterol (mg dL−1 ± SD)||118 ± 33||124 ± 31||132 ± 34||136 ± 36||140 ± 39||< 0.001|
|BMI (kg m−2 ± SD)||26.1 ± 4.5||26.4 ± 4.7||26.5 ± 4.7||26.8 ± 4.7||27.5 ± 4.8||< 0.001|
|Chronic kidney disease*, n (%)||63 (5.7)||55 (5.1)||58 (5.3)||68 (6.0)||99 (8.7)||< 0.001|
|CRP (mg L−1 ± SD)||3.6 ± 6.9||3.1 ± 6.6||3.5 ± 6.4||3.5 ± 5.0||4.3 ± 5.4||0.001|
|rs3093261 (minor allele frequency)||0.29||0.35||0.38||0.38||0.41||< 0.001|
|rs6046 (minor allele frequency)||0.25||0.15||0.10||0.08||0.05||< 0.001|
Table 5 presents associations between FVIIc quintiles and ischemic stroke, hemorrhagic stroke and MI in the three models described in Methods. Quintile 1 served as the reference group. The top two quintiles of FVIIc were associated with ischemic stroke in all models, with a 36% increased risk for both quintiles 4 and 5 vs. quintile 1 in Model C. There was no significant association in any model between FVIIc quintiles and hemorrhagic stroke or MI. Results were similar when estimated glomerular filtration rate was coded as a continuous variable.
Table 5. Adjusted hazard ratios for factor VIIc quintiles and cardiovascular outcomes
|Outcome||Model||FVIIc quintile||P-value (trend)|
|Ischemic stroke (n = 748)||A||Reference||1.32 (1.02–1.71)||1.11 (0.84–1.45)||1.40 (1.07–1.81)||1.60 (1.22–2.08)||0.001|
|B||Reference||1.31 (1.01–1.69)||1.05 (0.80–1.38)||1.32 (1.01–1.73)||1.46 (1.11–1.92)||0.016|
|C||Reference||1.29 (1.00–1.66)||1.06 (0.81–1.38)||1.36 (1.04–1.77)||1.36 (1.03–1.79)||0.03|
|Hemorrhagic stroke (n = 106)||A||Reference||1.43 (0.77–2.67)||0.81 (0.40–1.65)||0.81 (0.40–1.67)||1.03 (0.51–2.08)||0.51|
|B||Reference||1.50 (0.81–2.79)||0.80 (0.39–1.66)||0.91 (0.44–1.90)||1.06 (0.51–2.22)||0.61|
|C||Reference||1.38 (0.74–2.59)||0.65 (0.30–1.40)||0.66 (0.30–1.45)||0.72 (0.32–1.59)||0.11|
|Myocardial infarction (n = 736)||A||Reference||0.95 (0.75–1.18)||1.01 (0.80–1.27)||1.07 (0.85–1.35)||1.12 (0.88–1.42)||0.22|
|B||Reference||0.89 (0.71–1.11)||0.95 (0.75–1.20)||0.98 (0.77–1.25)||0.98 (0.78–1.27)||0.81|
|C||Reference||0.86 (0.68–1.08)||0.94 (0.75–1.19)||0.95 (0.74–1.22)||0.96 (0.74–1.25)||0.95|
In the traditional CVD risk factor model (Model B), we assessed the mediation by the F7 SNPs (rs6046 and rs3093261) on the association between FVIIc quintiles and our primary outcome, stroke, by adding both FVIIc quintiles and the SNPs in the same proportional hazard model (Table 6). The fourth (HR 1.46; 95% confidence interval [CI] 1.08–1.98) and fifth (HR 1.58; 95% CI 1.15–2.16) quintiles of FVIIc were associated with stroke when no SNPs were added into the model. When rs6046 (the SNP with the largest association with FVIIc level) was added to the model, the HRs for stroke for the fourth (HR 1.28, 95% CI 0.93–1.76) and fifth (HR 1.34; 95% CI 0.96–1.87) quintiles of FVIIc were lower. With singular adjustment for rs3093261 alone, only minimal changes in the FVIIc HRs were observed. When both SNPs were added to the model, FVIIc quintiles were no longer associated with stroke risk (Table 6). The associations of rs6046 (HR 0.71; 95% CI 0.58–0.87) and rs3093261 (HR 1.17; 95% CI 1.04–1.32) changed minimally when FVIIc quintiles were added into the model (Model B): for rs6046, HR 0.75 (95% CI 0.60–0.93); and for rs3093261, HR 1.15 (95% CI 1.02–1.30).
Table 6. Impact of factor VII gene single-nucleotide polymorphisms on association of FVIIc with stroke in whites
|Model*||HR (95% CI) for stroke by FVIIc quintile|
|Model B alone||Reference||1.42 (1.06–1.91)||1.12 (0.82–1.53)||1.46 (1.08–1.98)||1.58 (1.15–2.16)|
|Model B + rs6046||Reference||1.31 (0.97–1.77)||1.00 (0.72–1.37)||1.28 (0.93–1.76)||1.34 (0.96–1.87)|
|Model B + rs3093261||Reference||1.39 (1.03–1.87)||1.09 (0.79–1.48)||1.41 (1.04–1.92)||1.52 (1.11–2.08)|
|Model B + rs6046 + rs3093261||Reference||1.23 (0.97–1.56)||0.92 (0.71–1.18)||1.15 (0.89–1.48)||1.21 (0.92–1.57)|
- Top of page
- Disclosure of Conflict of Interest
- Supporting Information
Of 736 SNPs from 130 candidate inflammation and hemostasis genes, six were associated with stroke among whites in this elderly cohort: two SNPs were located within or near F7 (rs6046 and rs3093261); two were within introns of HABP2 (rs4918841 and rs3781387); and two were near inflammation-related genes (rs3138055, NFKBIA; and rs4648004, NFKB1). The two F7 SNPs were associated with FVIIc levels (rs6046 and rs3093261). FVIIc levels were also significantly associated with ischemic stroke, but not with MI or hemorrhagic stroke. Addition of both F7 SNPs into a model for total stroke attenuated the association of FVIIc with stroke, but FVIIc did not attenuate the association of the two F7 SNPs with stroke.
These data extend those from prior studies on the genetics of stroke risk. In a recent genome-wide association study from the CHARGE Consortium of 2 194 468 SNPs, only two SNPs on chromosome 12, near NINJ2, were associated with stroke, with genome-wide significance . NINJ2 encodes an adhesion molecule expressed in glia. The CHARGE Consortium meta-analysis included 19 602 Caucasian individuals and 1544 incident stroke cases (including data from the CHS). Although the CHS was part of the CHARGE Consortium, its dataset used a shorter follow-up period than in this analysis, and only included 459 incident stroke events from the CHS. None of the associations in the current study of stroke was close to having genome-wide significance; the lowest P-value was 0.02, for rs3960787 (a proxy for rs4648004, NFKB1), with all other P-values being > 0.10. Furthermore, in our protocol, we did not assess any SNPs from NINJ2, as we concentrated on 736 SNPs from 130 inflammation-related and hemostasis-related genes. Not surprisingly, the genome-wide association study did not detect any of the associations that we observed, as our lowest P-value was 7 × 10−4 for rs4918851, which was well above the threshold selected for the genome-wide statistical significance in the CHARGE Consortium (P < 5 × 10−8).
Other studies of the SNPs and genes that we identified here have reported mixed results. F7 encodes FVII, an essential component of hemostasis at sites of vessel injury. The rs6046 SNP has long been known to affect FVII levels, and has been studied in relation to risk of stroke, MI and venous thrombosis in case–control studies [18–21]. Although the effects of rs6046 on FVIIc levels are well documented, the reason for the lower FVIIc levels is unclear. The minor haplotype of rs6046 results in a guanine to adenine substitution in codon 353 of the F7 gene, resulting in glutamine to arginine substitution . Whether this SNP in and of itself reduces FVIIc levels or is a marker for a hypofunctioning phenotype requires further investigation . In the Framingham Heart Study, using a stepwise selection algorithm to select SNPs associated with FVII levels, rs6046 fell out of the model with inclusion of rs1755685 (within the 5′-flanking/promoter region of F7). These SNPs are in linkage disequilibrium (r2 = 0.79), and the stepwise selection algorithm makes no assumptions on function, but only determines which variable produces a better model fit for analysis . In a recent meta-analysis of 1537 cases of ischemic stroke and 3133 controls, the rs6046 SNP showed no overall association with stroke (odds ratio 0.9; 95% CI 0.4–1.9), although some of the individual studies did show associations . In terms of MI, in the Framingham Heart Study, with only 155 CVD events (the majority of which were MIs), rs6046 was associated with FVII levels, but no F7 SNPs were associated with the combined CVD endpoint . In another case–control study, rs6046 was associated with MI in patients with established coronary artery disease . In a recent analysis of the Women’s Genome Health Study, rs6046 was associated with a lower risk of idiopathic venous thrombosis in women . Differences among studies may reflect heterogeneity of the phenotypes of stroke, MI, and venous thrombosis. In our elderly population, we hypothesize that stroke may represent a less heterogeneous phenotype than in younger populations. SNP rs3093261 is in the 3′-flanking region of F7, between F7 and the FX gene, with an r2 = 0.137 between rs6046 and rs3093261 . When both rs6046 and rs3093261 were in the same model for stroke risk, rs3093261 was not significantly associated with stroke (P = 0.13), and so this association may represent a weak linkage with rs6046 or another SNP. We are not aware of any other data for rs3093261, FVII levels, and CVD risk.
The association of FVII with stroke has been controversial. Theoretically, FVII is a strong hemostasis candidate protein for vascular disease risk . In primary hemostasis, FVII plays a key role in vascular wall injury, and exogenous administration of activated FVII leads to thrombotic complications, including stroke, MI, and venous thrombosis [26–28]. Epidemiologic studies, including in the CHS, have shown mixed results for the association between FVIIc and CVD risk [2,29,30]. The first study to suggest an association was the Northwick Park Study, which showed an association between fatal CVD and FVII levels . Prior analyses in the CHS have had fewer stroke events or combined stroke and transient ischemic attack as an endpoint [2,31]. FVII levels in the Atherosclerosis Risk in Communities cohort (a younger cohort) were not associated with stroke risk for 268 ischemic strokes . Our analysis revealed a modest association between FVII and stroke, and many other studies would be underpowered according to the association that we observed, and have often used a combined CVD endpoint (which may attenuate associations, owing to the lack of association of FVIIc with MI seen here). Although arterial diseases in diverse vascular beds share many common risk factors, emerging evidence suggests that each has a unique risk factor profile. Furthermore, the quality of the literature examining novel risk factors and biomarkers for stroke is sparse as compared with the equivalent literature for MI . Our current analysis represents the largest prospective study that we are aware of relating FVII levels and cardiovascular risk in an elderly population. Further basic science studies are needed to determine the reasons for differing risk factors in diverse vascular beds.
HABP2 encodes a heterodimeric serine protease that has a variety of effects on coagulation and inflammation genes, including cleaving fibrinogen and activating pro-urokinase and FVII . HABP2 has been a target of study for CVD risk, with the Marburg I and II polymorphisms being associated with stroke, MI, and venous thrombosis [32–34]. Marburg I and II SNPs were not evaluated here, as their gene frequencies were < 0.05, but we found that two SNPs within introns of HABP2 were associated with stroke risk. Neither of these SNPs has specifically been linked with stroke risk previously, and they are not in linkage disequilibrium . Several hypotheses link variants of HABP2 with vascular diseases: one espouses a decoupling of activation of the fibrinolytic system (pro-urokinase to urokinase) with activation of FVII and fibrinogen, and another postulates increased smooth muscle cell proliferation . Neither of the HABP2 SNPs studied here was associated with FVII levels after adjustment for multiple testing. The protein product of HABP2 can also be measured in blood, but was unavailable for this study, so we could not assess the correlation between HABP2 SNPs, FVII-activating protease levels, and stroke risk .
We observed nominal associations between stroke and two SNPs near inflammation-related genes: rs4648004 (NFKB1, p50 subunit) and rs3138055 (NFKBIA). Nuclear factor kappaB is a complex multimeric transcription factor regulating hundreds of inflammation and cellular apoptotic proteins, and is upregulated in brain ischemia models . NFKBIA encodes a component of the inhibitor of kappaB kinase . Animal models suggest that these proteins play a key role in cerebral ischemia, and specific inhibitors have been investigated in stroke models, with mixed results .
Despite our study having a well-characterized cohort with a large number of validated strokes, a major limitation was the small number of non-whites; our results may not generalize to other races or ethnicities. Furthermore, we were limited by the need to balance between the number of SNPs and genes assessed and the potential for false-positive results. For SNP discovery, we used P < 0.01; this is greater than in the most conservative method (Bonferroni correction), which would require P < 6.7 × 10−5. The application of a Bonferroni threshold, which does not factor in the correlation between the test statistics for the SNPs that are in linkage disequilibrium, is strictly agnostic regarding biological information about the genes being studied. We note that three of the genes (F7, HABP2, and NFKB1) that contain, or are near, SNPs that were nominally associated with stroke have strong circumstantial evidence relating them to CVD. We also note that there were strong internal consistencies between the genotypes of the F7 SNPs and the phenotypes of the F7 gene product and stroke; rs6046 was associated with lower FVIIc levels and a lower risk of ischemic stroke, and rs3093261 was associated with higher FVIIc levels and a higher risk of ischemic stroke. When both F7 SNPs were added into a model with FVIIc levels, the association of FVIIc with stroke was partly mediated by the genotypes. The genotypes probably better reflect FVIIc levels over time than the one-time determination of FVIIc levels.
In summary, six SNPs within four hemostasis-related and inflammation-related genes (F7, HABP2, NFKBIA, and NFKB1) were nominally associated with stroke risk. Two F7 SNPs that were associated with stroke were also associated with FVIIc levels, which in turn were associated with ischemic stroke risk. The F7 SNPs mediated the association of FVIIc levels with stroke. The consistency and the rigorous methods used to acquire these data provide strong supporting evidence that, in this elderly population, FVII may play an etiologic role in stroke. Future work needs to confirm these findings and further elucidate the role of inflammation and hemostasis in stroke risk.
- Top of page
- Disclosure of Conflict of Interest
- Supporting Information
The authors acknowledge the role of the Neurology Working Group of the CHARGE Consortium in providing data for independent verification of the genetic association findings in this article. In addition to the CHS, CHARGE Consortium members include the National Heart, Lung, and Blood Institute Atherosclerosis Risk in Communities Study, National Institute on Aging Iceland Age, Gene/Environment Susceptibility Study, Framingham Heart Study, and Netherlands Rotterdam Study. The research reported in this article was supported by the National Institute on Aging (AG-023629). The CHS was supported by contract numbers N01-HC-85079 to N01-HC-85086, N01-HC-35129, N01 HC-15103, N01 HC-55222, and N01-HC-75150, N01-HC-45133, grant number U01 HL080295 from the National Heart, Lung, and Blood Institute, and grants R01 HL-071862 and R01 HL59367, with an additional contribution from the National Institute of Neurological Disorders and Stroke. Additional support was provided through R01 AG-15928, R01 AG-20098 and AG-027058 from the National Institute on Aging, R01 HL-075366 from the National Heart, Lung and Blood Institute, and the University of Pittsburgh Claude. D. Pepper Older Americans Independence Center (P30-AG-024827).