Author + information
- Received January 5, 2009
- Revision received May 21, 2009
- Accepted May 25, 2009
- Published online August 11, 2009.
- Toshimitsu Nozaki, MD⁎,
- Seigo Sugiyama, MD, PhD⁎,⁎ (, )
- Hidenobu Koga, MD, PhD⁎,
- Koichi Sugamura, MD⁎,
- Keisuke Ohba, MD⁎,
- Yasushi Matsuzawa, MD⁎,
- Hitoshi Sumida, MD, PhD†,
- Kunihiko Matsui, MD, PhD‡,
- Hideaki Jinnouchi, MD, PhD§ and
- Hisao Ogawa, MD, PhD⁎
- ↵⁎Reprint requests and correspondence:
Dr. Seigo Sugiyama, Department of Cardiovascular Medicine, Graduate School of Medical Sciences, Kumamoto University, 1-1-1 Honjo, Kumamoto City 860-8556, Japan
Objectives We investigated whether a multiple biomarkers strategy that includes plasma levels of endothelium-derived microparticles (EMP), reflecting endothelial dysfunction, can improve prediction of future cardiovascular events in patients at high risk for coronary heart disease (CHD).
Background Detailed risk stratification using multiple biomarkers can provide clinical benefits in high-risk patients. Endothelial dysfunction has been described as a predictor of cardiovascular complications.
Methods We measured 3 biomarkers in 488 consecutive patients with various CHD risks: B-type natriuretic peptide (BNP), high-sensitivity C-reactive protein (hsCRP), and EMP. We followed 387 stable patients at high risk for CHD and examined future cardiovascular events.
Results During a mean follow-up of 36 months, 55 patients developed cardiovascular events. Multivariate Cox proportional hazards analysis adjusted for established risk factors identified age, BNP, hsCRP, and EMP as significant and independent predictors of future cardiovascular events (age: hazard ratio [HR]: 1.042, 95% confidence interval [CI]: 1.007 to 1.080, p = 0.02; BNP: HR: 1.242, 95% CI: 1.004 to 1.536, p = 0.046; hsCRP: HR: 1.468, 95% CI: 1.150 to 1.875, p = 0.002; EMP: HR: 1.345, 95% CI: 1.094 to 1.652, p = 0.005). The C statistics for cardiovascular events increased when each biomarker or combinations of biomarkers were added to the Framingham risk model (C statistics: Framingham risk model alone 0.636, Framingham risk + BNP 0.695, Framingham risk + hsCRP 0.696, Framingham risk + EMP 0.682, and Framingham risk + BNP + hsCRP + EMP 0.763).
Conclusions The assessment of endothelial dysfunction by plasma levels of EMP can independently predict future cardiovascular events in patients at high risk for CHD. A multiple biomarkers strategy that includes endothelial dysfunction assessed by EMP can identify patients vulnerable to cardiovascular disease. (University Hospital Medical Information Network number: UMIN000000876)
The present cardiovascular risk stratification with established coronary risk factors cannot fully predict the development of cardiovascular events (1). Several biomarkers including B-type natriuretic peptide (BNP) and high-sensitivity C-reactive protein (hsCRP) have been reported to be useful for identifying the high-risk patients, independent of the established risk factors, and the multiple biomarkers strategy has been demonstrated to improve the risk stratification for cardiovascular events beyond the risk assessment based on established risk factors alone (2,3). Biomarkers reflecting different disease pathways may have the potential advantage of improving predictive power utility, and improvement of the assessment of cardiovascular risk with new biomarkers is desirable. It has been demonstrated that endothelial dysfunction is involved in the development of atherothrombogenic complications (4) and associated with future cardiovascular events in high-risk patients (5–7); however, it has not been incorporated into the previous multiple biomarkers strategy. Endothelial dysfunction can be clinically detected by measuring impairment of endothelium-dependent vasodilatation in response to acetylcholine during coronary angiography or by brachial artery flow-mediated vasodilation (5,8). These physiological tests are complex, operator dependent, and provide limited quantitative data (9,10).
Endothelium-derived microparticles (EMP) are small membrane-shed vesicles generated from endothelial cell surfaces in response to cellular activation or injury/apoptosis, and can potentially reflect endothelial dysfunction (11,12). Recently, we reported that CD144-EMP is derived selectively from human endothelial cells (13) and that circulating plasma CD144-EMP levels correlate significantly with coronary endothelial dysfunction and are significantly elevated in patients with type 2 diabetes and atherosclerosis (13). Although EMP are still only used for research purpose and in specialized laboratories because of their elusive nature and difficult assessment due to very small size (14), these findings underscore the potential application of CD144-EMP as a quantitative biomarker of endothelial dysfunction.
We hypothesized that the addition of a quantitative measure of endothelial dysfunction to a multiple biomarkers strategy could improve the prediction of future cardiovascular events. The hypothesis was tested by investigating the utility of plasma CD144-EMP levels for prediction of future cardiovascular events in stable patients at high risk for coronary heart disease (CHD), and examined the usefulness of the modified multiple biomarkers strategy, including endothelial dysfunction assessed by EMP, to predict cardiovascular complications.
In this prospective study, we screened 519 consecutive Japanese patients between May 2003 and August 2007 at Kumamoto University Hospital. Patients with severe valvular heart disease requiring surgical intervention within 1 month, scheduled for coronary revascularization, active infection, or malignant disease were excluded from the study (n = 31). The 488 patients who fulfilled the study criteria were divided into the following 4 groups: low-risk patients who had no or 1 CHD risk factor, patients with multiple risk factors without documented coronary artery disease (CAD), patients with documented CAD at stable condition (stable-CAD), and patients with acute coronary syndromes (ACS) (Fig. 1).Stable-CAD represented patients with angiographically documented organic coronary stenosis of >50% by quantitative coronary angiography in major coronary arteries. Risk factors for CHD were defined as age ≥65 years (15); current smoking; family history of ischemic heart disease; hypertension (>140/90 mm Hg or taking antihypertensive medication) (16); dyslipidemia (high-density lipoprotein [HDL] cholesterol <40 mg/dl, low-density lipoprotein [LDL] cholesterol ≥140 mg/dl, triglycerides ≥150 mg/dl, or receiving lipid-lowering treatment); diabetes mellitus (DM) (17); body mass index ≥25.0 kg/m2(16); hsCRP ≥2.0 mg/l; or chronic kidney disease (estimated glomerular filtration rate [eGFR] <60 ml/min/1.73 m2). The glomerular filtration rate was estimated using the modified formula of Modification of Diet in Renal Disease study equation, which was proposed by the Japanese Society of Nephrology (18). This study protocol was conducted in accordance with guidelines approved by the ethics committee at our institution.
Measurement of plasma levels of CD144-EMP and blood parameters
Blood samples were withdrawn by venipuncture into vacutainer tubes containing sodium citrate after a 12-h overnight fast for stable patients and on admission to the emergency room for ACS patients, before any mechanical intervention. Fresh plasma was assayed immediately for CD144-EMP by flow cytometry using the method described previously (13,14). We verified plasma levels of CD144-EMP with standard plasma for each sample. Standard plasma were subdivided into 1-use volume and stocked at −80°C. One thawing of stock plasma did not affect CD144-EMP levels. We measured hsCRP by a nephelometry with BN II (Siemens, Berlin, Germany) and BNP by a fluorescence enzyme immunoassay with AIA-21 (Tosoh Bioscience, Tokyo, Japan). Total cholesterol, HDL cholesterol, triglyceride, LDL cholesterol, and creatinine concentrations were determined by routine laboratory methods.
First, we compared plasma levels of CD144-EMP among low-risk patients (CHD risk factor ≤1), multiple risk patients (CHD risk factors ≥2), stable-CAD, and ACS patients. Second, patients with multiple risk factors or stable-CAD were categorized as high-risk patients for CHD and followed up every month at the outpatient department until July 2008 or at end point (Fig. 1). The end point was cardiovascular death, nonfatal myocardial infarction, unstable angina, ischemic stroke, or coronary revascularization to new lesions. Cardiovascular events were documented by phone calls to the patients or their families, followed by a review of medical records, electrocardiogram, ultrasound echocardiogram, and cardiac enzyme data. Cardiovascular death was defined as death due to myocardial infarction, congestive heart failure, or documented sudden cardiac death. Diagnosis of ischemic stroke was made if the patient had clinical and radiological evidence of stroke without intracranial hemorrhage. For subjects experiencing more than 2 acute events, only the first event was considered in the analysis. Revascularization therapy based only on angiographic data, including percutaneous coronary intervention-mediated restenosis, was not counted as a cardiovascular event. We used the previously reported cutoff values of 52.6 pg/ml (19) and 2.0 mg/l (20), and the median levels for BNP, hsCRP, and CD144-EMP, respectively, to divide our follow-up population into 2 groups: the high-level group and low-level group for the particular parameter.
Results were expressed as mean ± SD or as frequencies (percentages), while BNP, hsCRP, and CD144-EMP levels were expressed as median and interquartile range. The frequencies of risk factors and medications were compared between 2 groups by using chi-square analysis. Continuous variables were compared between 2 groups by the unpaired ttest or Mann-Whitney Utest, as appropriate. Data of the 4 groups were compared by 1-way analysis of variance, Kruskal-Wallis test, and chi-square analysis. Survival analysis was performed using the Kaplan-Meier method and assessed with the log-rank test.
The predictive value for cardiovascular events was assessed by Cox proportional hazards regression. The following variables were incorporated first into the univariate model: age, sex, current smoking, hypertension, DM, body mass index, HDL cholesterol, LDL cholesterol, eGFR, BNP, hsCRP, and CD144-EMP. Variables with p values <0.20 were then entered into a forward stepwise multivariate Cox proportional hazards analysis. In this model, we evaluated the effect of the biomarkers, BNP, hsCRP, and CD144-EMP, according to quintile increment in biomarkers levels.
Proportional hazards assumption was confirmed by Schoenfeld's test. Estimates of the C statistic for Cox proportional hazards regression models were calculated (21). The comparison of C statistics after the addition of the biomarkers to the model with Framingham risk was estimated (22). We also examined whether the addition of various combinations of biomarkers improved the discriminatory power of the model.
We assessed the calibration of Cox regression models by the Grønnesby and Borgan (23) calibration test, which compares the number of events that are expected based on estimation from 5 risk score groups. To evaluate whether the global model fit improved after the addition of the biomarkers, we performed likelihood ratio tests.
The statistical analyses were carried out using SPSS version 15.0J for Windows (SPSS Inc., Chicago, Illinois), STATA version 10.0 (StataCorp LP, College Station, Texas), and SAS version 9.1.3 (SAS Institute Inc., Cary, North Carolina). Statistical significance was defined as a value of p < 0.05 from 2-sided tests.
Enrollment, classification, and follow-up of patients
We screened 519 patients, but 31 patients were excluded (Fig. 1). Data of the remaining 488 patients were subjected to analysis. In this study population, 387 patients at high risk for CHD were followed up, and the data of 378 patients (multiple risk factors, n = 167; stable-CAD, n = 220) were available for analysis of cardiovascular events while 9 patients were lost to follow-up (Fig. 1). The follow-up period was 1 to 62 months (mean 36 months).
Comparison of CD144-EMP levels
All clinical factors except the frequency of current smoking were significantly different among patients with various CHD risk. The plasma levels of CD144-EMP increased significantly with increased coronary risk factors and with complicated clinical manifestations (patients at low-risk: n = 51, median [interquartile range], 0.303 [0.142 to 0.367] × 106; multiple risk factors: n = 167, 0.508 [0.387 to 0.681] × 106; stable-CAD: n = 220, 0.604 [0.449 to 0.795] × 106; ACS: n = 50, 0.983 [0.718 to 1.150] × 106/ml, p < 0.001) (Fig. 2).LDL cholesterol, eGFR, and hsCRP were higher in ACS than stable-CAD (ACS vs. stable-CAD: LDL cholesterol: 121.2 ± 30.0 mg/dl vs. 110.8 ± 32.7 mg/dl, eGFR: 65.7 ± 20.8 ml/min/1.73 m2vs. 58.9 ± 21.4 ml/min/1.73 m2, and hsCRP: 2.2 [0.7 to 7.8] mg/l vs. 1.2 [0.5 to 3.6] mg/l). Moreover, CD144-EMP levels were significantly higher in ACS patients than in stable-CAD patients (Fig. 2).
Baseline clinical features of patients at high risk for CHD
Table 1summarizes the baseline clinical features of patients at high risk for CHD (multiple risk factors or stable-CAD; follow-up population). The mean age was 66.9 years and 61.4% were men. Plasma levels of CD144-EMP correlated weakly with hsCRP (r = 0.16, p = 0.002) and did not correlate with BNP (r = 0.08, p = 0.14). Multivariate logistic regression analysis identified male sex and DM as significant risk factors of high EMP levels (above median) (men: hazard ratio [HR]: 1.685, 95% confidence interval [CI]: 1.076 to 2.639, p = 0.02; DM: HR: 1.551, 95% CI: 1.009 to 2.386, p = 0.046).
Cardiovascular events and biomarker levels
We recorded 55 cardiovascular events in patients at high risk for CHD during the follow-up period. Patients of the high EMP group developed significantly more cardiovascular events than the low EMP group during the follow-up (Table 2).Specifically, the incidences of cardiovascular death and ACS were significantly higher in the high-EMP group than in the low-EMP group (Table 2). Kaplan-Meier analysis based on high and low levels of biomarkers showed a significantly higher probability of cardiovascular events in the presence of high levels of BNP, hsCRP, and EMP during the follow-up (log-rank test: BNP p < 0.001, hsCRP p < 0.001, and EMP p < 0.001) (Figs. 3Ato 3C).
Cox proportional hazard analysis and C statistics for cardiovascular events
Univariate and multivariate Cox proportional hazards analysis for cardiovascular events showed that age, BNP, hsCRP, and CD144-EMP were independent predictors of future cardiovascular events in patients at high risk for CHD (age: HR: 1.042, 95% CI: 1.007 to 1.080, p = 0.02; BNP: HR: 1.242, 95% CI: 1.004 to 1.536, p = 0.046; hsCRP: HR: 1.468, 95% CI: 1.150 to 1.875, p = 0.002; EMP: HR: 1.345, 95% CI: 1.094 to 1.652, p = 0.005) (Table 3).Framingham risk was not incorporated into multivariate analysis because it was constructed by the same variables in univariate analysis. Framingham risk was confirmed to be a significant factor by univariate analysis in the present study (HR: 1.043, 95% CI: 1.011 to 1.076, p = 0.008). We then estimated the C statistic of Framingham risk alone. Separate incorporation of each biomarker into the Framingham risk model showed that all biomarkers increased the C statistic for prediction of cardiovascular events (C statistics: Framingham risk alone 0.636, Framingham risk + BNP 0.695, Framingham risk + hsCRP 0.696, and Framingham risk + EMP 0.682) (Table 4).Moreover, we examined the additive usefulness of EMP in multiple biomarkers strategy based on Framingham risk and BNP, hsCRP, or both. EMP increased the C statistics in multiple biomarkers strategy (C statistics: Framingham risk + BNP 0.695, Framingham risk + BNP + EMP 0.741; Framingham risk + hsCRP 0.696, Framingham risk + hsCRP + EMP 0.734; and Framingham risk + BNP + hsCRP 0.732, Framingham risk + BNP + hsCRP + EMP 0.763) (Table 4). The p value for the Schoenfeld's tests indicated that proportional hazards assumptions were appropriate (p = 0.70). We also confirmed good calibration for the model in patients at high risk for CHD by Grønnesby and Borgan (23) statistics (p = 0.34). Furthermore, models that included all biomarkers had better global fit than models with only Framingham risk, as evaluated by the likelihood ratio test (p = 0.02).
We examined the effect modification of interaction among all biomarkers and found that there was an interaction term between EMP and hsCRP (p = 0.03).
We demonstrated that circulating plasma levels of CD144-EMP in patients at high risk for CHD were independent predictors of future cardiovascular events. We also found that the addition of multiple biomarkers, including endothelial dysfunction, as assessed by CD144-EMP, to the Framingham risk model improved classification of risk, as evidenced by a substantial increase in the C statistics. Thus, quantitative evaluation of cardiovascular risk leading to atherothrombogenic complications from multiple aspects that include endothelial dysfunction can be clinically useful and valuable in patients at high risk for CHD.
Although the mean age of the study population and combination of biomarkers were issues of concern in the study design, the multiple biomarkers strategy, which is based on adding several biomarkers to the prediction model, including the established risk factors, is useful for risk stratification of cardiovascular events (2,3). It has already been demonstrated that BNP and hsCRP are independent predictors in healthy subjects (24,25) and CHD patients (26,27), and are significant biomarkers that improve C statistics for death and cardiovascular events (2,3). Endothelial dysfunction has also been recognized as an independent predictor of future cardiovascular events (5–7). Despite the pathophysiological significance of endothelial dysfunction in cardiovascular medicine, one cannot clinically assess coronary endothelial dysfunction because the available method is complex and invasive. It is probably for this reason that endothelial dysfunction was not incorporated into the multiple biomarkers strategy. In addition to the use of coronary reactivity to acetylcholine or brachial artery flow-mediated vasodilation, endothelial dysfunction can be assessed by measuring circulating levels of intercellular adhesion molecule 1, E-selectin (28), and von Willebrand factor (29). Soluble biomarkers offer the advantage of convenience and quantitative assessment; however, there is little evidence at present that such markers can accurately predict future cardiovascular events. Because the aforementioned molecules can be produced from cells other than endothelial cells such as leukocytes (28) and platelets, we need to identify a highly specific soluble biomarker that reflects endothelial dysfunction and can predict the prognosis of CHD patients.
Microparticles are released from various circulating blood cells and have many pathophysiological properties, as procoagulants and messengers (11). Microparticles detected by CD144 antigens (vascular endothelial cadherin), which are endothelial cell-type specific transmembrane adhesion molecules located only on the endothelium, exist in human plasma and are derived selectively from human endothelial cells, and their plasma levels can be a clinically specific marker for endothelial dysfunction (13,30). In the present study, we used the CD144-EMP assay to quantitate endothelial dysfunction. Although the clinical significance of measurement of microparticles has not been established yet, as stated in the preceding text, the method used for measurement of CD144-EMP is more specific, safe, simple, and rapid. Moreover, the fact that plasma levels of CD144-EMP independently predicted future cardiovascular events in the present study indicates that measurement of plasma CD144-EMP levels could be potentially useful for risk assessment of endothelial dysfunction with potential cardiovascular complications.
Endothelial dysfunction is one component of vulnerable plaques and closely associated with the occurrence of ACS (31). Vulnerable plaques are characterized by a thin fibrous cap with a large lipid core and superficial erosion of the luminal endothelium. Severe endothelial dysfunction may predispose to vulnerable endothelium, and the main feature of endothelial vulnerability is probably endothelial erosion. A vulnerable endothelium can promote atherothrombogenic complications through endothelial erosion, but there are no reliable methods for evaluating the risk of endothelial vulnerability, including endothelial erosion (31,32). Therefore, cardiovascular risk stratification that includes evaluation of endothelial dysfunction is a sound approach. Analysis of the risk in different disease pathways is important, and we propose that evaluation of endothelial dysfunction could be an important and clinically useful strategy. Based on the concept of vascular protection, a specific and quantifiable marker that can monitor endothelial dysfunction is necessary, as is the need to design intensive treatment to improve endothelial dysfunction.
A weak correlation between EMP and hsCRP resulted in statistical modification of interaction between EMP and hsCRP. EMP levels correlated to some extent with various inflammatory markers, because inflammatory cytokines can induce the release of EMP, and the latter, in turn, promote endothelial injury, leading to endothelial dysfunction (11).
One limitation of the present study is the relatively small number of patients in a single center. However, this should result in underestimation, stressing the need for further multicenter studies in a larger population to confirm the present results. There is no consensus about measurement of EMP for assessment of endothelial damage and prothrombotic state at this stage, and microparticles are still used only for research purposes. There is a need to standardize the EMP assay for the development and establishment of routine clinical tests, because measurement of CD144-EMP could be potentially useful for the evaluation of endothelial dysfunction. The number of this study population was not estimated by power calculation. It is effective and necessary to have a plan for the number of patients required for a prospective study.
Endothelial dysfunction leading to cardiovascular complications can be assessed quantitatively by measurement of plasma levels of CD144-EMP. Moreover, a multiple biomarkers strategy that includes endothelial dysfunction assessed by CD144-EMP can provide better risk stratification of cardiovascular events and, hence, more thorough clinical assessment of patients who might benefit from more aggressive treatment strategies that improve prognosis.
This study was supported, in part, by grants-in-aid for Scientific Research (#C19590869 to Dr. Sugiyama) from the Ministry of Education, Science, and Culture, Japan; Advanced Education Program for Integrated Clinical, Basic and Social Medicine, Graduate School of Medical Sciences, Kumamoto University in Kumamoto, Japan (Support Program for Improvement of Graduate School Education, MEXT, Japan); and Kimura Memorial Heart Foundation Bayer Grant for Clinical Vascular Function 2008, Kurume, Japan.
- Abbreviations and Acronyms
- acute coronary syndromes
- B-type natriuretic peptide
- coronary artery disease
- coronary heart disease
- confidence interval
- diabetes mellitus
- estimated glomerular filtration rate
- endothelium-derived microparticle(s)
- high-density lipoprotein
- hazard ratio
- high-sensitivity C-reactive protein
- low-density lipoprotein
- Received January 5, 2009.
- Revision received May 21, 2009.
- Accepted May 25, 2009.
- American College of Cardiology Foundation
- Widlansky M.E.,
- Gokce N.,
- Keaney J.F. Jr..,
- Vita J.A.
- Schachinger V.,
- Britten M.B.,
- Zeiher A.M.
- Chan S.Y.,
- Mancini G.B.,
- Kuramoto L.,
- Schulzer M.,
- Frohlich J.,
- Ignaszewski A.
- Modena M.G.,
- Bonetti L.,
- Coppi F.,
- Bursi F.,
- Rossi R.
- Corretti M.C.,
- Anderson T.J.,
- Benjamin E.J.,
- et al.
- Deanfield J.E.,
- Halcox J.P.,
- Rabelink T.J.
- Kuvin J.T.,
- Karas R.H.
- VanWijk M.J.,
- VanBavel E.,
- Sturk A.,
- Nieuwland R.
- Koga H.,
- Sugiyama S.,
- Kugiyama K.,
- et al.
- Grundy S.M.,
- Balady G.J.,
- Criqui M.H.,
- et al.
- Pearson T.A.,
- Blair S.N.,
- Daniels S.R.,
- et al.
- Omland T.,
- Sabatine M.S.,
- Jablonski K.A.,
- et al.
- Koenig W.,
- Sund M.,
- Frohlich M.,
- et al.
- Amabile N.,
- Guerin A.P.,
- Leroyer A.,
- et al.
- Naghavi M.,
- Libby P.,
- Falk E.,
- et al.
- Sugiyama S.,
- Kugiyama K.,
- Aikawa M.,
- Nakamura S.,
- Ogawa H.,
- Libby P.