OUP user menu

Relationship between circulating endothelial cells and the predicted risk of cardiovascular events in acute coronary syndromes

Christopher J. Boos , Simren K. Soor , Delene Kang , Gregory Y.H. Lip
DOI: http://dx.doi.org/10.1093/eurheartj/ehm070 First published online: 21 April 2007


Aims The quantification of circulating endothelial cells (CECs) in whole blood is a novel marker of direct endothelial injury and shows promise as a potential biomarker of cardiovascular (CV) risk. The inter-relationship(s) between CECs and predicted CV risk has not been explored in large cohort of ‘high-risk’ patients. We hypothesized that there would be a significant relationship between increasing CEC counts and predicted CV risk in a broad spectrum of patients presenting with acute coronary syndrome (ACS).

Methods and results We studied 197 patients (aged 40–80 years) admitted with a confirmed diagnosis of unstable angina (UA), non-ST-elevation myocardial infarction (MI, NSTEMI), or ST-elevation MI (STEMI). CEC counts were performed on venous whole blood using the immunobead technique. Four well-validated ACS risk scores [(PURSUIT and TIMI for NSTEMI/UA) TIMI (STEMI) and GRACE (all ACS)] were calculated from the initial clinical history and electrocardiogram, as well as from values of laboratory parameters collected within 12 h of admission. We included a healthy control (HC) group of 50 matched patients in order to quantify the accuracy of CEC counts for the diagnosis of ACS and to compare disease vs. HC counts. CEC counts were significantly higher in the disease group when compared with the HC group. CEC counts significantly increased with increasing severity of disease (that is, UA vs. NSTEMI vs. STEMI; P = 0.002). CEC counts were higher among patients with clinical evidence of heart failure (Killip Class II–IV) when compared with those without (Killip Class I) on admission (P < 0.0001). There was a significant correlation between CEC counts and predicted CV risk for each of the four ACS risk scoring schemes (all P < 0.05). The area under the receiver-operating characteristic (ROC) curve (AUC) for the entire ACS cohort was 0.82 (95% CI: 0.76–0.88; P < 0.0001). A CEC count of ≥7/mL provided a positive predictive value of 90.6% (95% CI: 85.6–95.7%) and a negative predictive value of 53.5% (41.9–65.1%) for the diagnosis of MI (NSTEMI/STEMI) in the presence of an appropriate clinical presentation.

Conclusion There is a significant and positive correlation between increasing CECs and increasing CV risk in ACS. The diagnostic accuracy of CECs in this setting is only ‘moderate’. Whilst it is good at confirming the presence of MI, a CEC value of <7.0/mL is less reliable at confidently excluding patients without disease.

  • Circulating endothelial cells
  • Cardiovascular risk
  • Outcome
  • Myocardial infarction
  • Acute coronary syndromes
  • Risk scores


Acute coronary syndromes (ACS) are the most common cause of death in the western World.1,2 ACS represents a disease continuum ranging from unstable angina (UA), through non-ST-elevation myocardial infarction (MI, NSTEMI) with minimal myocardial necrosis and to ST-elevation myocardial infarction (STEMI) at the other extreme.3,4 Given the marked heterogeneity associated with this clinical entity, mortality may vary considerably across the three patient subgroups.3,4 Consequently, early risk stratification plays a pivotal role in patient management as early and aggressive intervention, improves clinical outcomes, particularly among patients at the highest risk.5,6

A number of validated multivariable cardiovascular (CV) risk scoring systems have been developed over the recent years that have been shown to accurately predict future major adverse CV events (MACE) including MI and CV death.712 These have been largely based on a combination of clinical history, examination, electrocardiogram (ECG), and laboratory tests. The Thrombolysis in Myocardial Infarction (TIMI) scoring systems for UA/NSTEMI7,8 and STEMI,9,10 respectively, and the Platelet glycoprotein IIb/IIIa in Unstable Angina: Receptor Suppression Using Integrilin (PURSUIT)11 scores are perhaps the most commonly described. The more recent Global Registry of Acute Coronary Events (GRACE) is unlike the other scoring systems, as it has been validated for use among the entire ACS spectrum.12

Endothelial damage/dysfunction plays a central role in the clinical manifestation of coronary atherosclerosis.13 The quantification of circulating endothelial cells (CECs) represents a novel and direct cellular marker of endothelial damage/dysfunction. Current consensus would support the concept that CECs are mature endothelial cells that have become mechanically detached from the mural endothelium in response to endothelial injury.14,15 In a recent study of 156 patients presenting with ACS, increased CECs were an independent predictor of 30 day and 1 year risk of death and MACE events.16 In a sub-study of 88 patients with UA and NSTEMI, higher CECs were noted among the patients with higher TIMI scores (≥4), supporting an association between increasing CECs and increasing CV risk.17

In this study, we sought to gain a greater conceptual insight into the relationship between CECs and CV risk. In this regard, we hypothesized that there would be a consistent link between increasing CEC counts and increasing calculated CV risk in a larger (and entirely different) cohort of patients presenting with ACS. Furthermore, we sought to investigate the predictive accuracy of CECs—based on the current consensus definition—in patients with a confirmed diagnosis of ACS.


Between May 2005 and August 2006, we performed an analysis of prospective data in patients aged 40–80 years old, admitted to our coronary care unit with ACS, and fitting our inclusion criteria. ACS was defined according to the European Society of Cardiology Guidelines.3,4 The diagnosis of STEMI was defined as the concurrence of prolonged chest pain or discomfort, with persistent ST-segment elevation of greater than 1 mm in two or more contiguous leads, or with presumed new left bundle branch block with cardiac enzymes [total creatine kinase (CK) and creatine kinase-MB fraction (CK-MB)] above twice the upper normal limit (or raised tropnin I >0.1 ng/mL).4 The diagnosis of NSTEMI included the presence of typical angina at rest associated with acute and transient ST-segment or T-wave changes with cardiac enzymes above twice the upper normal limit and/or raised troponin I levels (>0.1 ng/mL).3 Patients with clinical and/or ECG features of non-STEMI but with normal cardiac enzymes were classified as having UA.3

We excluded patients with any of the following: pregnancy, a history of liver disease, dialysis or with a serum creatinine >200 µmol/L, malignancy, recent (<3 months) major trauma, arterial or venous thrombo-embolic disease, patients with active infections, and/or a history of inflammatory or connective tissue disorders. We also excluded patients in which the presentation of chest pain was felt to be precipitated by trauma, surgery, hyperthyroidism, or anaemia.12

We included a cohort of ‘healthy controls’ (HCs) comparable with the disease group in terms of age, sex, and body mass index (BMI). HCs were recruited from a combination of hospital staff, relatives of patients with stable coronary artery disease (CAD), and following local advertisement. The rational for including this patient group was in order to give a comparative perspective of the difference in CEC counts among the ACS spectrum when compared with HCs, and furthermore, in order to calculate the diagnostic accuracy of CECs in ACS. HCs were identified by a detailed history and normal physical examination, with a normal baseline full blood count, renal function, fasting glucose and lipid profile. All blood samples for CEC isolation (see below for method) were collected within 24 h of admission, but following thrombolysis or primary percutaneous coronary intervention (PCI angioplasty ± stenting, where applicable).

Risk scores

The four CV risk scores—TIMI NSTEMI/UA7, PURSUIT,11 TIMI STEMI,9 and GRACE12 (Tables 1 and 2)—were calculated from the initial clinical history and ECG, as well as from the values of laboratory parameters collected within 12 h of admission. Risk scores were calculated retrospectively (using prospective data) and separately by two persons who were blinded to the CEC results (agreement >98%).

View this table:
Table 1

Calculation of risk scores using the TIMI STEMI, TIMI NSTEMI/UA, and PURSUIT system

View this table:
Table 2

Calculation of GRACE risk scores to predict the risk of in-hospital and 6-month mortality or myocardial infarction risk

For the PURSUIT risk score we used the simple score for death and MI, which is obtained as the sum of the points given to five predictive factors.11 For the TIMI NSTEMI/UA risk score all seven variables have the same magnitude with the result being the simple arithmetic sum of the number of variables present.7 For the variable of ‘known coronary artery stenosis >50%’ we instead attributed one point for a history of MI, PCI, or bypass surgery. This method has been validated by the original TIMI Investigators and is considered closer to ‘real-world’ practice.8,18,19 The TIMI risk score for STEMI is a simple weighted integer score based on eight clinical risk indicators.9 For the more recently developed GRACE risk score, we used the online downloadable programme designed for personal diary assistants (www.statcoder.com/grace.htm).12 We used the prediction models for in-hospital and 6-month risk of death or MI. The programme is unique in that it converts the relevant total score into an actual percentage risk for the particular event. Finally, Killip scores20 were calculated as follows: Class I, the absence of clinical heart failure (HF); Class II, mild clinical HF with rales involving one-third or less of the posterior lung fields and systolic blood pressure of 90 mmHg or higher; Class III, pulmonary oedema with rales involving more than one-third of the lung fields and systolic blood pressure of 90 mmHg or higher; and Class IV as cardiogenic shock with any rales and systolic blood pressure <90 mmHg.

Isolation of circulating endothelial cells

Our technique of CEC quantification and its validation has been previously published in detail, and is based on an updated baseline consensus definition.2123 In brief, 1 mL of whole blood is incubated with CD146 (endothelial-associated marker) conjugated immunomagnetic beads and suspended in a known quantity of buffer. After mixing for ≥30 min, CD146+ cells are separated using several washing steps in a magnet. Fluorescein isothiocyanate (FITC)-labelled Ulex Europeus lectin (an endothelial-specific marker) is added to the remaining cell/bead suspension with a further mixing (≥30 min) step in darkness. The cells are then washed again and suspended in a known volume of buffer solution and viewed under fluorescence microscopy. CECs are defined as CD146 rosetted cells, bearing greater than or equal to four beads, sized approximately 10–50 µm in diameter with positive staining for FITC-labelled Ulex Europeus lectin (Figure 1). Bland–Altman analysis24 (i.e. individual differences vs. the mean CEC count) was performed in order to assess the intra- and inter-observer agreement in CEC isolation methodology among paired blinded observers. All assessments of variability were performed on a minimum of 50 patients representing the three main patient disease groups (HCs, stable CAD, and patients with ACS) over a 4-week period. The intra- and inter-observer agreements [percentage within 2 standard deviations (SD) of the mean difference] were 94.4% and 90.4%, respectively.

Figure 1

Comparative circulating endothelial cell counts (CEC) among the healthy controls (HC) and the three acute coronary syndrome subgroups of unstable angina (UA), non ST-elevation myocardial infarction (NSTEMI), and ST-elevation myocardial infarction (STEMI). Data are expressed as the median with IQRs. Results expressed as Kruskal–Wallis test (overall P < 0.0001) with Dunn post hoc test.

Power calculation

We performed our power calculation using the Simple Interactive Statistical Analysis package (http://home.clara.net/sisa). There have been no studies that have sought to investigate actual correlations between CECs and predicted CV risk. Based on our previous work of CECs in ACS,16,17,22 we calculated that a sample size of >125 patients would be sufficient to demonstrate a significant correlation between CECs and predicted CV risk at a two-sided alpha level of 0.05 at a correlation coefficient (r) of >0.22. In order to gain an even greater perspective on the relationship between increasing CECs and CV risk, the risk scores were categorized into three risk groups. These three groups (representing increasing CV risk) are based on previously published data (TIMI NSTEMI/UA and PURSUIT,19 TIMI STEMI,25 and GRACE26) have been linked to worsening clinical outcomes.9,19,25 In addition, we calculated that an average sample size of 30 patients per subgroup would provide at least an 80% power to detect >50% difference in median CEC counts between the highest and lowest tier of CV risk.

Statistical analysis

Data were analysed using GraphPad InStat version 3.05 (GraphPad Software, San Diego, CA, USA; www.graphpad.com). After determination of normality (on visual inspection of the data and using the Kolmogorov–Smirnov test) for all continuous data sets, appropriate parametric, and non-parametric tests were utilized.

All continuous data are presented as mean (SD or 95% CI) when normally distributed or as the median (IQR, inter-quartile range) for non-parametric data. The Kruskal–Wallis test (non-parametric) was used for between-group comparisons in CEC counts, with a Dunn post hoc test, where appropriate. For two group comparisons of continuous data, an unpaired t test and Mann–Whitney test were used for parametric and non-parametric data, respectively. As CECs were not normally distributed, the relationship between individual CEC counts and the predicted score for each patient was assessed using Spearman rank correlation (95% CI).

Multiple linear regression analysis was performed to determine the influence of multiple baseline factors on CEC counts. The following variables were assessed: age, sex, ethnicity, weight, systolic blood pressure, heart rate, previous history of hypertension, CAD, HF, diabetes mellitus, family history of CAD, current smoking status, diagnostic category (UA, NSTEMI, STEMI), Killip classification, CK and troponin levels, and previous use of aspirin. In addition, we included the time from admission to venous sampling and as well as adjusting for the effects of early intervention with PCI performed prior to venous sampling, as these have been previously shown to influence CEC levels. The rational for inclusion of these factors were that they represented the majority of individual factors contained within the four risk scoring systems, while maintaining a minimum of ≥10 patients per variable adjusted for. Area under the ROC curves (AUC 95% CI) was calculated (GraphPad Prism version 4.00; www.graphpad.com) by comparing a HC cohort to the ACS group in order to calculate the predictive accuracy of CECs for the diagnosis of ACS. From these ROC curves, we assessed the optimum ‘cut-off’ CEC value (with 95% CI) (thus, a balance of sensitivity vs. specificity) for ACS diagnosis and calculated the positive and negative predictive values (PPV and NPV, respectively) for this cut-off. A two-tailed P-value <0.05 was considered statistically significant for all comparisons.


A total of 197 ACS patients and 50 HCs were included in the study. The baseline characteristics of the overall cohort as well as the detailed characteristics of the ACS group are shown in Tables 3 and 4, respectively. The mean age was 62.5 (11.2) years (73% male) (Table 3). The mean (SD) time from admission to sample collection was 13.3 (5.6) hours. From the entire cohort of ACS patients, 34% had a diagnosis of STEMI, 42.6% with NSTEMI, and 23.4% with UA (Table 3). CEC counts were significantly higher in ACS group (overall), and among the differing ACS subtypes (UA, NSTEMI, and STEMI) compared with HCs (Table 3). There was a significant increase in CEC counts in relation to severity of disease diagnostic category (Table 4 and Figure 1).

View this table:
Table 3

Baseline characteristics and circulating endothelial cell counts (CEC) and of entire cohort of acute coronary syndrome (ACS) patients

View this table:
Table 4

Baseline characteristics and circulating endothelial cell counts (CEC) for the three component patient groups comprising the overall acute coronary syndrome cohort

Circulating endothelial cells and increasing cardiovascular risk scores

There was a significant increase in CEC counts with increasing CV risk among each of the four risk schemes (Table 5 and Figure 2). CEC counts were significantly higher among patients with clinical HF on admission (Killip Class II–IV) compared with those without (Killip Class I), respectively [15 (9.3–22.0) vs. 8 (5–12) cells/mL; P < 0.0001]. There was a non-significant trend to higher CEC counts among patients with a previously known diagnosis of HF [11.3 (7.8–19.3) vs. those without, 8.5 (6.0–15.0); P = 0.06].

Figure 2

Relationship between circulating endothelial cell counts (CEC) and calculated cardiovascular risk score. TIMI, Thrombolysis in Myocardial Infarction; PURSUIT, Platelet glycoprotein IIb/IIIa in Unstable Angina: Receptor Suppression Using Integrilin. Data are expressed as the median with IQRs. P-value refers to the results of Kruskal–Wallis with Dunn post hoc test.

View this table:
Table 5

Comparative circulating endothelial cell counts with increasing cardiovascular risk prediction


There were significant correlations (Spearman rank; 95%CI) between CEC counts and predicted CV risk for each of the four risk scoring schemes as follows: PURSUIT (r = 0.22; 0.04–0.38; P = 0.01); TIMI NSTEMI/UA (r = 0.23; 0.05–0.4; P = 0.009); TIMI STEMI (r = 0.44; 0.22–0.47; P<0.0001); and for GRACE [risk of in-hospital MI/death (r = 0.36; 0.23–0.48; P < 0.0001) and for the risk of 6-month death/MI (r = 0.32; 0.19–0.45; P < 0.0001)] (Figure 3).

Figure 3

Relationship between circulating endothelial cell (CEC) counts and predictive cardiovascular risk (%) using the GRACE scoring system. (A) Risk (%) of in-hospital (IH) death or myocardial infarction (MI). (B) Six-month risk of death or MI. GRACE, Global Registry of Acute Coronary Events. Data are expressed as the median with IQRs. P-value refers to the results of Kruskal–Wallis with Dunn post hoc test.

Multiple regression analyses

Multiple regression analyses was performed to assess the influence of defined baseline and clinical factors contained predominantly within the four scores (see methods) on the dependent continuous variable of CEC counts. This revealed a significant overall model fit (R2 = 19.8%; P = 0.017) (Table 6). Only the admission Killip score (P = 0.0002) and diagnostic category (P = 0.017) remained as a significant predictor of CEC counts.

View this table:
Table 6

Multiple regression analysis demonstrating the influence of multiple variables on the dependent variable of circulating endothelial cell (CEC) counts

Receiver-operating characteristic curve analyses

Comparing the entire ACS cohort with the HC group gave an AUC (95% CI) of 0.82 (0.76–0.88; P < 0.0001). For the diagnostic sub-categories of ACS, the AUCs were as follows: 0.75 (0.65–0.84; P < 0.0001) for UA, 0.82 (0.75–0.89; P < 0.0001) for NSTEMI, 0.88 (0.82–0.94) for STEMI, and 0.85 (0.79–0.90; P < 0.0001) for MI (NSTEMI/STEMI) (Figure 4). A CEC count ≥7/mL (95% CI 6.15–8.20%) equated to a PPV of 92.4% (95% CI 88.3–96.5%) and a corresponding NPV of 44.8% (34.4–55.3%) for confirming the diagnosis of ACS in the presence of an appropriate clinical presentation. For the diagnosis of MI (NSTEMI/STEMI) this cut-off gave a PPV of 90.6% (85.6–95.7%) and an NPV of 53.5% (41.9–65.1). However, for UA, the PPV and NPV were only 71.4% (56.4–82.8%) and 65.6% (52.7–76.4%), respectively.

Figure 4

Receiver-operator characteristic curves illustrating the predictive accuracy of circulating endothelial cell (CEC) counts for the diagnosis of myocardial infarction.


This is the first study to investigate the relationship between CECs and predicted CV risk across a wide spectrum of differing CV risk scoring systems. We observed a consistent relationship between increasing CEC counts and escalating CV risk among a large cohort of patients presenting with ACS. Moreover, we found that a laboratory ‘cut-off’ of 7.0 CEC/mL, while good at identifying patients with genuine CAD and confirmed MI [(NSTEMI/STEMI), a CEC count of <7/mL was rather weak at excluding patients without disease. Furthermore, there was weak, yet statistically significant, correlation between CECs and predicted CV risk among patients with UA/NSTEMI using the PURSUIT and TIMI risk scoring system. However, there was a moderate, yet again significant, correlation between CEC and predicted risk among STEMI patients using the TIMI STEMI risk score and for the entire ACS cohort using the GRACE risk score.

There have been seven previous smaller studies (20–156 patients), that have consistently demonstrated higher CEC counts in patients with ACS when compared with matched controls.16,17,22,2730 In the only study to investigate the relationship between CECs and CV outcomes, to date, CEC counts were independent predictors of death and MACE (composite of CV death, non-fatal MI, or refractory angina with ECG ST-T segment changes requiring hospitalization or urgent coronary revascularization) at both 30 days and 1 year.16 In addition, there has only been one previous study that has investigated the relationship between CECs and predicted CV risk.17 In this study, significantly higher CEC counts were noted among patients with higher TIMI (≥4) vs. lower (<4) TIMI scores in a cohort of 88 patients with UA or NSTEMI.17 Our data confirms this and is consistent with the findings of these two previous studies. Moreover, we have expanded upon this previous work in a much larger (and entirely different) cohort of ACS patients, covering the entire spectrum of ACS diagnoses (i.e. UA, NSTEMI, and STEMI), utilizing four differing CV scoring systems, and a more updated method of CEC isolation.2123

Our observations of higher CECs among patients with clinical evidence of HF (Killip Class II–IV) on admission are also in keeping with the results of a recent study31 in which significantly higher CEC counts were present among patients with both acute and chronic HF (n = 60), compared with matched HCs. Indeed, HF is linked to worsening endothelial dysfunction and the degree of endothelial damage is proportional to HF.31,32 However, in our study the presence of chronic HF failed to significantly influence CEC scores in the final regression model, after accounting for the interaction of a number of additional patient-related variables.

The four risk scoring systems used in this study represent the most widely used and validated risk scoring schemes for patients presenting with ACS. The discriminatory accuracy of the TIMI UA/NSTEMI,7,19 PURSUIT,11,19 TIMI STEMI,9 and GRACE12,19 scores for the prediction of a variety of MACE, including death and re-infarction, has been well demonstrated at a number of follow-up time points. In our study, the diagnostic accuracy of CEC counts for the diagnosis of UA was ‘fair’ and only moderate for the confirmation of MI. Our results are in keeping with the only previous paper to investigate the diagnostic accuracy of CECs in ACS which studied 60 patients presenting with NSTEMI only.29 This paper by Quilici et al.29 noted an AUC of 0.82–0.87 (PPV of 100% and NPV of 58.8% at a cut-off of >3 CEC/mL) for the combination of CECs and elevated troponin for the final diagnosis of MI. This compares well with our PPV of 90.6% and an NPV of 53.5%, respectively. In contrast to their study, we have included a larger patient cohort and have incorporated the new consensus definition of ACS.3,4 Furthermore, we have used an updated method of CEC isolation.2123

The diagnostic accuracy of CEC, using the current immunobead method, would appear to under-perform that of other more established and cardio-specific biomarkers, such as brain natriuretic peptides33 and ischaemia-modified albumin.34 However, a detailed comparison of CEC counts to other biomarkers in ACS risk-stratification is beyond the scope of this article. Furthermore, CECs are a highly novel cellular marker and a rapidly evolving research growth area, for which, unlike the above-mentioned markers, an automated method has not yet been developed.

The imperfections of our assay relate to a number of factors and limitations that must be acknowledged. For example, elevated cardiac troponins, with or without dynamic ECG changes in a patient with typical chest pain history are both sensitive and specific for CAD and myonecrosis.3,4 However, while CEC have been shown to be macrovascular in origin in ACS16,28 they may represent more widespread, and perhaps generalized, endothelial injury rather than merely the local release of endothelial cells from one or more culprit coronary arteries. It must be emphasized that the ‘cut-off’ value used in this study only relates to the study population in this paper and is critically influenced by disease prevalence. Secondly, the use of a HC, rather than disease control is likely to increase the diagnostic accuracy. Nevertheless, CECs have been consistently shown to correlate with a variety of markers of generalized endothelial damage/dysfunction including soluble E-selectin, von Willebrand factor, and flow-mediated dilatation.1517,30,31 This is important as generalized endothelial damage/dysfunction, in itself is an independent predictor of worsening CV outcomes among patients presenting with ACS.35

Secondly, the isolation and enumeration of CECs is technically difficult, with notable intra- and inter-observer variability.22 The later, partly, relates to the relatively low CEC numbers in the circulation and to limitations in the immunobead method itself. CEC cells are thought to represent terminally differentiated, largely non-viable, mature endothelial cells, that become detached in response to the endothelial injury.14 The mechanism behind endothelial cell detachment and their circulation in blood as ‘CECs’ are highly complex and include a response to mechanical injury, alterations in cellular adhesion molecules, and cytoskeletal proteins, defective binding to anchoring matrix proteins and cellular apoptosis.14,15 Hence, they are considered to be an entirely different cell population to that of endothelial progenitor cells (EPCs) which are a heterogeneous population of endotheloid cells (involved in vascular repair and angiogenesis) at various stages of maturation.14,15 Despite these apparent differences in the two cell types, we cannot fully exclude the possibility that we may have included a small number of mature CD146 bearing EPCs. However, data from Nakatani et al.36 has shown that only 4.4% of isolated CECs, using the immunobead method, could have been defined as EPCs on counter-staining. This distinction is important, as EPC counts have also been shown to be independently predictive of MACE.37,38 Nevertheless, we have used an entirely different method (immunobead compared with flow cytometry for EPCs) designed to capture a distinct cell population. Moreover, our method of CEC isolation is current and has been supported by a recent Pan European consensus.23 There is clearly a need for further research investigating the inter-relationship between endothelial injury and repair (CECs vs. EPCs) in both disease and health.

There are a number of further limitations that must be acknowledged. While adjusting for the effects of early PCI (n = 25) and the effects of time from admission to CEC isolation, these are important potential confounders that must be acknowledged. The calculation of CV risk scores used in this paper is based on an arbitrary and point-in-time clinical ‘snapshot’ on the patient at admission to hospital. These scores do not reflect the changing dynamics of a patient: it is well known that a patient's clinical state and calculated risk score can rapidly change over a short period of time in response to both treatment and/or clinical deterioration. Finally, despite clear hypotheses, and the use of robust post hoc tests, we cannot fully exclude the possibility of type 1 error due to multiple testing. Accepting these inherent limitations, the scoring systems used in this paper are well validated and have been consistently shown to be predictive of future CV death and MACE. The relationship between CEC counts and actual ‘long-term’ MACE on an expanded patient cohort is ongoing and will be reported in the distant future.

In this paper, we have provided evidence to support a consistent, yet only moderate relationship between CEC counts and predicted CV disease risk. The quantification of CECs from blood is a definite growth area in vascular biology. Owing to their endothelial origin, they are novel and direct markers of endothelial damage/dysfunction, and thus, have great potential as a future clinical biomarker of endothelial injury. Unfortunately, current limitations in methodology and definition have, at present, precluded its progression into the clinical arena, and their present utility as a cardiospecific biomarker, appears limited. There is clearly a need for improved, well-validated, and automated methods for their enumeration.

Conflict of interest: none declared.


View Abstract