Ceruloplasmin and Heart Failure in the Atherosclerosis Risk in Communities (ARIC) Study

Abstract

Background

Ceruloplasmin (Cp) decreases nitric oxide bioavailability in blood and has been associated with cardiovascular disease (CVD) in clinical studies. We assessed the association between Cp and incident heart failure (HF), death and CVD in the Atherosclerosis Risk in Communities (ARIC) Study.

Methods and Results

Cp was measured at ARIC visit 4 (1996–1998). We studied 9,240 individuals without HF or CVD at ARIC visit 4, and followed them for a mean of 10.5 years. Genome-wide association study was performed to identify genetic determinants of Cp levels and evaluate their association with incident HF. Cp levels (mean±standard deviation) were higher in women vs men (335±79 vs 258±44 mg/L, p<0.0001), women on vs not on hormone-replacement therapy (398±89 vs 291±60 mg/L, p<0.0001) and African Americans vs Caucasians (299±63 vs 293±74 mg/L, p=0.0005). After adjusting for traditional risk factors, high-sensitivity C-reactive protein, N-terminal pro–B-type natriuretic peptide, and high-sensitivity cardiac troponin T, higher levels of Cp were associated with HF (hazard ratio [HR] 1.44, 95% confidence interval [CI] 1.13–1.83) and mortality (HR 1.38, 95% CI 1.11–1.63). A locus on the ceruloplasmin gene on chromosome 3 was significantly associated with Cp levels (normal 295.56±77.60mg/L, heterozygote 316.72±88.02mg/L; homozygote 331.04±85.40mg/L, p=8.3×10⁻) but not with incident HF. After adjustment for traditional risk factors Cp levels were also weekly associated with CVD.

Conclusions

Cp was associated with incident, HF mortality and CVD in the ARIC population. A single locus on chromosome 3 was associated with Cp levels but not with HF.

Ceruloplasmin (Cp) is an enzyme synthesized in the liver that enhances low-density lipoprotein oxidation and decreases nitric oxide (NO) bioavailability through NO oxidase activity^{1, 2}. NO also has an important role in cardiac contraction and alteration in endogenous NO secretion may contribute to heart failure (HF)^{3, 4}. Oxidized low-density lipoprotein has a central proatherogenic role in the arterial wall, whereas decreased NO bioavailability promotes endothelial dysfunction, which also leads to atherosclerosis². Cp is an acute-phase reactant, and its levels are correlated with other nonspecific inflammatory markers that are associated with cardiovascular disease (CVD) risk, such as high-sensitivity C-reactive protein (hs-CRP), leukocyte counts, and myeloperoxidase^{5, 6}. Based on these observations, an association between Cp and CVD has been hypothesized.

Several studies suggested that high Cp levels may be associated with CVD^7–11. One study on white men suggested an association of Cp with incident HF¹². To our knowledge, no studies in the literature have evaluated the association of Cp levels with HF and all-cause mortality in the overall population.

The goal of this study was to analyze the association of Cp levels with incident HF, and all-cause mortality in the biracial population of the Atherosclerosis Risk in Communities (ARIC) Study. We also aimed to identify the genetic variants associated with Cp levels through a genome-wide association study (GWAS) and to evaluate the associations of these genetic variants with incident HF. A secondary goal was to confirm the association of Cp levels with incident CVD events (a composite of incident CHD and stroke), in the ARIC population.

Methods

Study population

The ARIC Study is a prospective population study of CVD in 15,792 middle-aged adults recruited from four U.S. communities in 1987–1989. Of the 11,656 eligible participants in visit 4 (1996–1998), 11,484 had Cp data available. Individuals with prevalent HF (n=406), CHD (n=973) or ischemic stroke (n=268) at visit 4 or those with missing data for any variable used in the statistical models (n=523) were excluded from the study. We also excluded individuals who were not Caucasian or African American (n=27) and any African American participants at the Minnesota and Washington County field centers because of small enrollment numbers (n=31). Therefore, 9,240 individuals were included in our analysis. Incident HF hospitalizations before visit 4 were determined through diagnosis codes from hospital discharges. Prevalent CHD and stroke were defined as self-reported myocardial infarction or stroke before visit 1, or silent myocardial infarction (diagnosed by electrocardiographic changes), validated myocardial infarction, coronary revascularization, or stroke between visits 1 and 4. Medical history, demographic data, anthropometric data, blood pressure measurements, and fasting lipid assessments were obtained during visit 4 at the same time as the blood draw for Cp measurement. Research protocols were approved by each ARIC Field Center's institutional review board, and all participants provided written informed consent.

Outcomes

We investigated the associations of Cp with incident HF, CHD and CVD and all-cause mortality events in the overall population and stratified by race and gender. Incident HF was defined as the first HF hospitalization or HF coded as the underlying cause of death and was noted by review of local hospital discharge lists that showed HF in any position and county death certificates. Hospitalization for HF (ICD-9 code 428) was determined through diagnosis codes from hospital discharges, while deaths were coded as HF (ICD-9 and ICD-10, code 428 and 150). We also analyzed the association of Cp levels with incident CHD and stroke as separate outcomes. Incident CHD events included fatal CHD, definite or probable myocardial infarction, a silent MI between examinations by electrocardiography, and coronary revascularization. Incident stroke events included only ischemic strokes, defined as validated definite or probable hospitalized embolic or thrombotic strokes. The methods of assessing incident CHD events and strokes in the ARIC Study have been described previously^13–15. Incident CVD events were a composite of incident CHD and incident ischemic stroke after ARIC visit 4. We analyzed the association of Cp with the etiology of HF (ischemic and non-ischemic). Individuals who had the diagnosis of CHD (as defined above) and who also had HF (as defined above) were included in the ischemic–HF group. Individual with the diagnosis of HF but no diagnosis of CHD were included in non-ischemic heart failure group. The mean follow-up time was 10.5±2.2 years for all outcomes, 10.5±2.5 years for HF, 10.8±2.2 years for all-cause mortality and 10.5±2.4 years for CVD.

Biomarker assays

Plasma Cp levels were measured by immunoturbidimetric assay using an automated chemistry analyzer (Olympus AU400e, manufacturer Olympus Life Science Research Europa GmbH). Highly lipemic samples, which may produce excessively high scatter signals, were avoided. The Cp turbidimetric procedure was calibrated every 14 days by using Olympus Serum Protein Multi-calibrator 2 (Cat #ODR3023), which was traceable to IFCC International Reference Preparation CRM470 (RPPHS). During operation of the Olympus analyzer, at least two levels of immunology control material were tested a minimum of once a day. The within-run precision was <5%, and total precision was <10%. N-terminal pro–B-type natriuretic peptide (NT-proBNP) was measured by using an electrochemiluminescent immunoassay on an automated Cobas e411 analyzer (Roche Diagnostics, Indianapolis, IN) with lower limit of detection ≤5 pg/mL and coefficient of variation 3.5–4.7%¹⁶. hs-CRP levels were measured by using an immunonephelometric assay on a BNII autoanalyzer (Siemens Healthcare Diagnostics, Deerfield, IL) with a reliability coefficient of 0.99^{17, 18}. Cardiac troponin T (cTnT) levels were measured by using a novel precommercial highly sensitive assay, Elecsys Troponin T (Roche Diagnostics), on an automated Cobas e411 analyzer with a lower limit of detection of 0.003 µg/L¹⁹.

Genotyping

Genome-wide genotyping of single-nucleotide polymorphisms (SNPs) was performed using the Affymetrix Genome-Wide Human SNP Array 6.0 (Santa Clara, CA). Study participants who refused DNA testing, unintentional duplicate samples with higher missing rates, suspected contaminated samples, samples with genotype mismatch with 47 previously genotyped SNPs, and genetic outliers based on identity-by-state statistics and EIGENSTRAT principal components analysis were excluded. Additionally, monomorphic SNPs, SNPs with no chromosome location, and SNPs with call rate <95%, minor allele frequency <1%, or Hardy–Weinberg equilibrium p<10⁻⁶ were also excluded. In addition to SNPs genotyped using the Affymetrix 6.0 chip, among individuals of European ancestry in our study sample, genotypes were also imputed for 2.4 million autosomal SNPs identified in HapMap CEU samples. For this study, imputation genotypes were not determined in ARIC participants of African ancestry. Imputation results were summarized as an allele dosage, which was defined as the expected number of copies of the minor allele at each SNP.

For the association of the SNP with the incident HF we used the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium database, which includes participants from ARIC Study, the Cardiovascular Health Study, the Framingham Heart Study, and the Rotterdam Study. Detailed description regarding the genotyping in CHARGE was described elsewhere²⁰.

Statistical methods

The distributions of continuous variables were evaluated to assess normality. For this analysis, we modeled Cp both as continuous and categorical variables. As a categorical variable, quartile measures were used as cut points to obtain four groups (Supplemental Table 1). Associations between Cp and clinical outcomes (HF, all-cause mortality and CVD) were determined using Cox proportional-hazards modeling, in both unadjusted and adjusted models. Incident HF, CHD, CVD, mortality and stroke event rates were each calculated using time-to-event methods. For all survival analyses, the follow-up time was defined as the period from entry into the study (visit 1) to the first respective event hospitalization, death, or up to the time an individual left the study. Additionally, we used linear terms using quartile number to obtain a p-value for trend and compare any linear trend in risks with each increasing quartile. The basic model (model 1) adjusted for age (continuous), gender, race, and ARIC center as potential confounders. Model 2 included all components of model 1 and additionally adjusted for traditional cardiovascular risk factors including total cholesterol, high-density lipoprotein cholesterol, systolic blood pressure, antihypertensive medication use, smoking status (current versus not current), presence of diabetes mellitus (defined as fasting blood glucose >126 mg/dL, self-reported diagnosis by a physician, or diabetes medication use), and body mass index. Because previous studies suggested that Cp levels are influenced by hormones, model 3 included all components of model 2 plus hormone-replacement therapy (HRT) use. Model 4 included all components of model 3 plus biomarkers associated with HF or CVD: hs-CRP, NT-proBNP, hs-cTnT, and estimated glomerular filtration rate (eGFR, estimated by Modification of Diet in Renal Disease formula). In all models, the second, third and fourth quartiles were compared to the first quartile.

We also performed quartile analyses stratified by race and gender. The adjustment models used were the same as those for the whole population (race was not included in race-specific analyses, and gender was not included in gender-specific analyses). In each subgroup, we additionally tested for interactions between gender and race by adding an interaction term of the two to each of the Cox proportional-hazards models and found no significant interaction in any of the groups.

A GWAS for serum Cp levels, both unadjusted and adjusted for age, race, gender, and ARIC center, was performed with PLINK (version 1.07) under dominant models (i.e., 0, 1, or 2 risk-raising alleles). To test the association of genetic variants with clinical outcomes (HF, all-cause mortality, CVD, CHD and stroke,) separately, Cox proportional-hazards modeling was performed, producing both unadjusted and adjusted (for age, race, gender, and ARIC center) models. To assess the power of our analysis to detect the increase in risk for events per allele, we performed a power calculation using Sean Purcell’s Genetic Power Calculator²¹.

Results

Cp was measured in 9,240 individuals (mean age 62.6±5.6 years) and ranged from 26.9 mg/L to 852.0 mg/L. In race-adjusted models, women had significantly higher Cp levels than men (335±79 mg/L vs 258±44 mg/L, p<0.0001), and in gender-adjusted models, mean Cp levels were higher in African Americans than Caucasians (299±63 mg/L vs 293±74 mg/L, p=0.0005). Also in race-adjusted analyses, women on HRT had significantly higher Cp levels than women not on HRT (388±89 mg/L vs 291±60 mg/L, p<0.0001). The distribution of Cp levels by gender is presented in Figure 1, and the baseline demographic characteristics by Cp quartiles for the overall population are shown in Table 1. Cp levels were positively correlated with hs-CRP (r=0.2, p<0.001) and negatively correlated with hs-cTnT (r=−0.12, p<0.001). No significant correlation was seen with NT-proBNP.

Figure 1. Gender-specific histogram for distribution of ceruloplasmin.

Table 1.

Baseline characteristics of the overall population by ceruloplasmin (Cp) quartiles

	Quartile 1	Quartile 2	Quartile 3	Quartile 4	P
Cp (mg/L)	<249.0	249.0–<285.8	285.8–<338.3	≥338.3
N	2305	2313	2312	2310
Demographics
Age (years)	62	63	62	61	<0.0001
Female (%)	22	47	72	93	<0.0001
African American (%)	13	19	28	25	<0.0001
BMI	28.1	27.9	28.0	27.2	0.009
Medical History
Diabetes mellitus (%)	16	16	16	11	<0.0001
HTN medication (%)	37	35	39	41	0.0006
Systolic blood pressure (mmHg)	124	125	126	125	0.004
Diastolic blood pressure (mmHg)	71	71	71	70	<0.0001
Menopausal (% of women)	89	94	92	89	0.42
Smokers, current (%)	13	15	16	15	0.02
Statin use (%)	8	9	9	8	0.49
HRT use (% of women)	14	12	17	46	<0.0001
Laboratory data
LDL cholesterol (mg/dL)	119	125	126	116	0.004
Total cholesterol (mg/dL)	192	200	206	206	<0.0001
HDL cholesterol(mg/dL)	42	45	48	57	<0.0001
Triglycerides (mg/dL)	121	118	117	126	<0.0001
hs-CRP (mg/L)	1.4	1.8	2.6	4.5	<0.0001
Creatinine (mg/dL)	0.8	0.7	0.7	0.6	<0.0001
hs-cTnT (µg/L)	0.006	0.005	0.004	0.003	<0.0001
NT-proBNP (pg/mL)	49.8	57.8	65.7	84.4	<0.0001
GGT (U/L)	23	22	22	18	0.07
ALT (U/L)	14	14	13	11	<0.0001
AST (U/L)	18	18	18	17	0.0002

BMI=body mass index, HTN=hypertension), HRT=hormone-replacement therapy, LDL=low-density lipoprotein, HDL=high-density lipoprotein, hs-CRP=high-sensitivity C-reactive protein, hs-cTnT=high-sensitivity cardiac troponin T, NT-proBNP=N-terminal pro–B-type natriuretic peptide, GGT=gamma glutamyl transpeptidase, ALT=alanine aminotransferase, AST=aspartate aminotransferase

Associations of Cp levels with incident HF, mortality and CVD

The relations between Cp level (as a continuous variable) and HF, mortality and CVD after adjustment for age, gender, and race are presented in Figure 2. During a mean follow-up of 10.5 years, a total of 752 individuals were hospitalized for HF, 1275 had died and 1234 had CVD events. Cp level was associated with HF and all-cause mortality in the model adjusted for traditional risk factors and HRT as well as in the fully adjusted model when hs-CRP, NT-proBNP, hs-cTnT, and eGFR were added (hazard ratio [HR] 1.14 per standard deviation [SD], 95% confidence interval [CI] 1.04–1.24 for HF; HR 1.12 per SD, 95% CI 1.04–1.20 for all-cause mortality; Supplemental Table 2). After adjustment for traditional risk factors and HRT, Cp was associated with CVD and stroke but not with CHD. The relations were not significant in the fully adjusted model.

Figure 2. Relation of ceruloplasmin with cardiovascular disease, heart failure, and all-cause mortality with adjustment for age, gender, and race.

For the categorical analysis, HRs are presented with the first quartile as reference (Table 2). Individuals with Cp in the highest quartile had significantly higher risk for HF hospitalization than those in the lowest quartile (HR=1.71, 95% CI 1.33–2.17) after adjusting for traditional risk factors and HRT The association remained significant in the fully adjusted model (HR 1.44 95% CI 1.13–1.83). After adjustment for traditional risk factors Cp was associated with both ischemic HF (HR=1.76, 95% CI 1.19–2.61) and non-ischemic HF (HR=1.67, 95% CI 1.22–2.29). After adjustment for hs-cTnT, hs-CRP and pro-BNP the association remain significant only for non-ischemic HF (HR=1.44, 95% CI 1.05–1.98). (Supplemental Table 3). Individuals with Cp in the highest quartile also had greater all-cause mortality than those in the lowest quartile after adjusting for traditional risk factors and HRT (HR=1.57, 95% CI 1.30–1.89) as well as in the fully adjusted model (HR 1.38 95% CI 1.11–1.63) (Table 2). Cp was also associated with non-cardiovascular death even in the most complex models (HR 1.63 95% CI 1.25–2.14) but not with the cardiovascular death. (Supplemental Table 4). In the overall population, Cp level was associated with incident CVD after adjusting for traditional CVD risk factors (HR 1.21, 95% CI, 1.00–1.49) but not in the fully adjusted model. Cp was not associated with CHD or stroke as separate outcomes (Table 2).

Table 2.

Categorical analysis of relation of Cp with incident HF, all-cause mortality, CVD, CHD and stroke

	Quartile 1	Quartile 2	Quartile 3	Quartile 4	p for trend
Cp range (mg/L)	<249.0	249.0–<285.8	285.8–<338.3	≥338.3
HF
Events/# at risk	176/2305	194/2313	191/2312	191/2310
Unadjusted	1.00	1.11 (0.91–1.35)	1.06 (0.87–1.30)	1.07 (0.88–1.29)	0.76
Model 1	1.00	1.12 (0.92–1.37)	1.10 (0.88–1.37)	1.36 (1.07–1.71)	<0.001
Model 2	1.00	1.22 (0.99–1.49)	1.18 (0.95–1.47)	1.63 (1.28–2.07)	0.07
Model 3	1.00	1.21 (0.99–1.49)	1.19 (0.96–1.48)	1.71 (1.33–2.17)	<0.001
Model 4	1.00	1.21 (0.99–1.50)	1.13 (0.91–1.41)	1.44 (1.13–1.83)	0.03
ALL-CAUSE MORTALITY
Events/# at risk	326/2305	341/2313	318/2312	290/2310
Unadjusted	1.00	1.04 (0.89–1.21)	0.97 (0.83–1.11)	0.86 (0.73–1.01)	0.07
Model 1	1.00	1.13 (0.97–1.32)	1.21 (1.01–1.41)	1.44 (1.20–1.73)	<0.001
Model 2	1.00	1.16 (0.99–1.35)	1.21 (1.02–1.41)	1.45 (1.21–1.74)	<0.001
Model 3	1.00	1.16 (0.99–1.35)	1.21 (1.02–1.41)	1.57 (1.30–1.89)	<0.001
Model 4	1.00	1.15 (0.98–1.31)	1.15 (0.96–1.38)	1.38 (1.11–1.63)	<0.001
CVD
Events/# at risk	384/2305	306/2313	300/2312	244/2310
Unadjusted	1.00	0.78 (0.68–0.91)	0.75 (0.64–0.87)	0.59 (0.50–0.70)	<0.0001
Model 1	1.00	0.91 (0.79–1.06)	1.08 (0.92–1.28)	1.15 (0.95–1.39)	0.14
Model 2	1.00	0.92 (0.78–1.09)	1.01 (0.86–1.19)	1.21 (1.00–1.46)	0.08
Model 3	1.00	0.89 (0.77–1.05)	1.01 (0.86–1.19)	1.21 (1.00–1.49)	0.08
Model 4	1.00	0.88 (0.75–1.02)	0.96 (0.81–1.13)	1.06 (0.83–1.30)	0.30
CHD
Events/# at risk	326/2305	257/2313	234/2312	177/2310
Unadjusted	1	0.77 (0.66–0.91)	0.68 (0.58–0.82)	0.52 (0.42–0.61)	0.27
Model 1	1	0.93 (0.79–1.10)	1.07 (0.89–1.28)	1.08 (0.87–1.35)	0.14
Model 2	1	0.92 (0.78–1.09)	1.00 (0.83–1.19)	1.15 (0.93–1.42)	0.12
Model 3	1	0.92 (0.78–1.09)	1.01 (0.84–1.19)	1.15 (0.93–1.44)	0.08
Model 4	1	0.91 (0.77–1.07)	1.04 (0.80–1.24)	1.10 (0.82–1.38)	0.26
STROKE
Events/# at risk	82/2305	62/2313	80/2312	80/2310
Unadjusted	1.00	0.75 (0.53–1.09)	0.96 (0.70–1.30)	0.96 (0.70–1.31)	0.88
Model 1	1.00	0.76 (0.55–1.10)	1.01 (0.73–1.41)	1.26 (0.88–1.81)	0.60
Model 2	1.00	0.77 (0.56–1.08)	1.00 (0.72–1.39)	1.33 (0.93–1.91)	0.48
Model 3	1.00	0.77 (0.56–1.08)	1.00 (0.71–1.39)	1.33 (0.93–1.92)	0.48
Model 4	1.00	0.77 (0.51–1.12)	1.00 (0.69–1.41)	1.09 (0.75–1.58)	0.72

Model 1 is adjusted for age, gender, race, and ARIC site

Model 2 is adjusted for all variables in model 1 plus total cholesterol, HDL cholesterol, systolic blood pressure, antihypertensive medication use, smoking status, and presence of diabetes mellitus (fasting blood glucose >126 mg/dL, self-reported diagnosis by a physician, or diabetes medication use), and BMI

Model 3 is adjusted for all variables in model 2 plus HRT use

Model 4 is adjusted for all the factors included in model 3 plus hs-CRP, NT-proBNP, hs-cTnT, and eGFR

Presented as hazard ratio 95% confidence interval.

We also performed a subgroup analysis by race and gender (Supplemental Tables 5A–D). There was no interaction between Cp and gender for risk of incident HF (p=0.22). In the gender stratified analysis Cp was associated with HF in both women and men in the models adjusted for traditional risk factors and in the fully adjusted models (HR 1.43 95% CI 1.06–1.93 for women and HR 1.38 95% CI 1.00–1.88 for men, Supplemental Tables 5C and 5D). There was no interaction between Cp and race for risk of incident CHD (p=0.96) or HF (p=0.55). In the Caucasian subgroup, after adjustment for traditional risk factors and HRT use, Cp level in the highest quartile was significantly associated with incident HF (HR=1.78, 95% CI 1.32–2.40) and all-cause mortality. Higher Cp levels were also associated with CVD, CHD and stroke. In the fully adjusted models the associations remained significant for all outcomes except CHD (Supplemental Table 5A).

GWAS for serum Cp levels

We next performed race stratified GWAS in 8,094 subjects in ARIC. Serum Cp levels were primarily associated with a single locus (rs1307255) on chromosome 3, which contains the CP gene (Figure 3). In the race-specific analysis (Table 3), Cp levels in TT homozygotes and heterozygotes (GT) were significantly higher than in GG homozygotes (the normal variant) in both Caucasians and African Americans. Although the minor allele frequency was smaller in Caucasians (0.07) than in African Americans (0.42), the impact of the allele on Cp level was higher in Caucasians than in African Americans (respective mean increase per allele 18 mg/L and 8.5 mg/L), and an interaction between race and alleles on Cp were statistically significant (p<0.0001)

Figure 3. Q-Q and Manhattan plots from genome-wide association study (GWAS) for serum ceruloplasmin levels in Caucasians.

A) The p-values obtained from the GWAS deviate from that expected by chance, suggesting true associations. B) Serum ceruloplasmin levels were primarily controlled by a single locus (rs1307255) on chromosome 3.

Table 3.

The association of rs13072552 with Cp levels in ARIC

African American population (N=1708)
	MAF	GG (N=572)	GT (N=824)	TT (N=312)	p
Unadjusted	0.42	307.35±71.07	308.82±70.04	324.36±63.76	1.0×10−3
Adjusted	0.42	307.23±70.05	308.88±69.90	324.26±63.28	1.1×10−3
Caucasian population (N=6386)
	MAF	GG (N=5502)	GT (N=846)	TT (N=38)	P
Unadjusted	0.07	295.55±76.66	316.69±87.88	331.64±86.87	9.9×10−14
Adjusted	0.07	295.56±77.60	316.72±88.02	331.04±85.40	8.3×10−13

Presented as mean±standard deviation. The adjusted model includes: age, gender, and ARIC center. MAF=minor allele frequency

Associations of rs13072552 with incident HF

We next investigated the relation between the rs13072552 variant and HF in CHARGE. Because of differences in the impact of the SNP on Cp levels and the differences in the minor allele frequency in African Americans versus Caucasians, we performed a race-stratified analysis. In CHARGE, 2526 incident HF events (12%) occurred in 20 926 Caucasian participants and 466 events (16%) in 2895 African-American participants. The mean follow-up was 11.5 years for Caucasians and 13.7 years for African Americans. This genetic variant was not significantly associated with HF in Caucasians (HR= 1.03, 95% CI 0.92–1.15 for the presence of either 1 or 2 copies of the allele) or African Americans (HR=0.97, 95% CI 0.85–1.11). Among the whole ARIC study sample, each SD of Cp (79 mg/L) was associated with a 14% increase in risk for HF and the increased HF risk per allele was 4%. In light of our observations that the impact of the SNP was only 18 mg/L in Caucasians with a low minor allele frequency, our power calculations indicated we had only 18% power for Caucasians to detect a 4% increase in risk for HF in CHARGE; therefore we were underpowered to show a significant association between the SNP and incident HF.

Discussion

We investigated the association of Cp levels with incident HF hospitalization all-cause mortality and CVD in the biracial cohort of the ARIC study. Our study is the largest prospective study that shows that high levels of Cp, an inflammatory plasma protein, are associated with HF, death and CVD. Plasma Cp levels are associated with a locus on chromosome 3 which contains the CP gene and are strongly influenced by race, gender, and HRT use.

Prior smaller studies have shown an association of Cp levels with CVD^{5, 7, 8, 11, 22} and with HF in Caucasian men¹². Our study extends these findings to the large, middle aged, biracial cohort of men and women in the ARIC study. In this population, plasma Cp levels were associated with incident HF, all-cause mortality and CVD. The strongest association we observed were with HF and all-cause mortality; these associations persisted even after adjusting for other biomarkers known to have a role in HF prediction, such as NT-proBNP, hs-cTnT, and hs-CRP¹⁹. The risk for incident HF in the overall ARIC population was similar to that reported by Engstrom et al¹² in Caucasian men (mean±SD age 46.8±3.7 years) at high risk for CVD. Additionally we show that the association remains significant even after adjustment for hs-CRP, pro-BNP and hs-cTnT. In comparison to his study, our study is larger, biracial, and included both genders and an older population (mean±SD age 62.6±5.7 years). We also found that Cp levels were associated with incident CVD events in the overall population after adjustment for traditional cardiovascular risk factors, but this association was not as strong as the one seen with HF and mortality and did not remain significant in the fully adjusted model. We found no association between Cp and incident CHD or stroke in the overall population. We found an association between higher Cp levels and incident CHD events in Caucasians which confirms the findings of Tang et al.⁵. In addition, we also found that in the Caucasian subgroup in ARIC, higher Cp levels were also associated with incident stroke and incident CVD events. Studies in animals and humans suggest that Cp levels are influenced by hormones^{23, 24}. Prior observations in very small numbers of patients found higher mean Cp levels in women compared with men and in women taking HRT compared with those who did not¹⁰. We extended these findings to the general population and also found significantly higher mean Cp levels in African Americans compared with Caucasians.

The role of Cp in cardiovascular pathophysiology is not well described. Cp has been shown to have important oxidase activities, including the capacity to consume NO catalytically through NO–oxidase activity, resulting in decreased NO bioavailability in plasma. Both in vitro and in vivo studies have shown that NO–oxidase activity is decreased in plasma after Cp immunodepletion and in humans with aceruloplasminemia^{1, 25}. Over the past several decades, clinical and experimental studies have provided substantial evidence that oxidative stress, defined as an excess production of reactive oxygen species relative to antioxidant defense, is enhanced in HF^26–28. Animal studies have shown that NO and NO synthases play a key role in normal cardiac physiology^{3, 4}. NO also has a protective role in the ischemic and/or failing heart; this protective role is mediated by several mechanisms, including the stimulation of soluble guanylyl cyclase, which leads to a decrease in the concentration of intracellular Ca²⁺ and the inhibition of oxidative stress. Therefore, superoxide generated during oxidant stress may exert cytotoxic effects, both directly and through the rapid scavenging of NO and the formation of the highly reactive oxidant peroxynitrite (ONOO⁻), which is formed at diffusion-limited rates from interaction of NO with superoxide²⁹. It is possible that high levels of Cp may decrease the available NO in the heart through increased NO–oxidase activity, and as a result, enhanced oxidative stress causes more dysfunction, which would explain the association between higher Cp levels and incident HF. Prevention of HF is clearly an important goal not only to reduce pain suffering and death from HF but also to reduce the high cost of treatment once patient become symptomatic enough to require hospitalization. We hypothesize that the use of new biomarkers will not only help in regards to risk stratification but in addition may help to identify individuals to whom specific treatment strategies are indicated. Cp is an inexpensive easy to perform automated assay that should be examined to see if therapies which target oxidative stress pathways have benefit in the subpopulation of individuals at increased risk of HF who have high levels of Cp.

One of the novel aspects of our study is that in addition to testing the association of Cp levels with incident HF, we also performed a GWAS analysis to determine if certain SNPs are associated with the same outcome. Our GWAS identified a locus on chromosome 3 in the CP gene that controls Cp levels, rs13072552, which is the same locus found by Tang et al⁵. This SNP is intronic and it was associated with increased Cp levels in both Caucasians and African Americans. Although minor allele frequency was much greater in African Americans, the impact of this SNP on Cp levels was much higher in Caucasians. When we analyzed the association of rs13072552 with incident HF, we found no significant association; however, in spite of using the CHARGE database, our power calculations suggest that we had insufficient power to rule out the possibility that this SNP is associated with incident HF and that genetically elevated levels of Cp are in the causal pathway. Much larger studies are needed to test this association.

Limitations

A limitation in our study is insufficient power in the GWAS analysis to rule out an association between rs13072552 and clinical outcomes. Another limitation of our study is that the measurements were performed at only one time point using frozen plasma samples.

Conclusion

Cp levels were higher in women than men and were associated with HRT use. Higher plasma Cp levels were associated with incident HF, mortality and CVD in the overall ARIC population, but not with CHD or stroke as separate outcomes. Cp levels were associated with incident HF, death CVD, CHD and stroke in Caucasians. Genetic variation of rs13072552 in the CP gene was associated with plasma Cp levels but not with HF, death, CVD, CHD or stroke in the ARIC population.