Clinical Research in Cardiology

, Volume 100, Issue 7, pp 587–594

Serum soluble E-selectin and NT-proBNP levels additively predict mortality in diabetic patients with chronic heart failure

Authors

    • Research Laboratory, 3rd Department of Internal Medicine, Szentágothai Knowledge CenterSemmelweis University
  • László Cervenak
    • Research Group of Inflammation Biology and ImmunogenomicsHungarian Academy of Sciences
  • Zsolt Förhécz
    • Research Laboratory, 3rd Department of Internal Medicine, Szentágothai Knowledge CenterSemmelweis University
  • Tímea Gombos
    • Research Laboratory, 3rd Department of Internal Medicine, Szentágothai Knowledge CenterSemmelweis University
  • Zoltán Pozsonyi
    • Research Laboratory, 3rd Department of Internal Medicine, Szentágothai Knowledge CenterSemmelweis University
  • Jan Kunde
    • Department of Global Medical AffairsBRAHMS GmbH
  • István Karádi
    • Research Laboratory, 3rd Department of Internal Medicine, Szentágothai Knowledge CenterSemmelweis University
  • Lívia Jánoskuti
    • Research Laboratory, 3rd Department of Internal Medicine, Szentágothai Knowledge CenterSemmelweis University
  • Zoltán Prohászka
    • Research Laboratory, 3rd Department of Internal Medicine, Szentágothai Knowledge CenterSemmelweis University
    • Research Group of Inflammation Biology and ImmunogenomicsHungarian Academy of Sciences
Original Paper

DOI: 10.1007/s00392-011-0283-6

Cite this article as:
Czúcz, J., Cervenak, L., Förhécz, Z. et al. Clin Res Cardiol (2011) 100: 587. doi:10.1007/s00392-011-0283-6

Abstract

Background

Neuroendocrine activation with endothelial dysfunction is a key pathophysiological process in chronic heart failure (CHF). Although increased soluble E-selectin (sE-selectin) levels predict adverse events in several forms of cardiovascular disease, there are only scarce data on its predictive value in CHF. The aim of our study was to investigate whether sE-selectin is a useful predictor of mortality in CHF patients and whether its predictive power is additive to that of NT-proBNP.

Methods

Plasma levels of sE-selectin were measured by ELISA in 192 CHF patients with clinical systolic heart failure. The study population was followed up for 14.9 months on average; 46 patients died during this period.

Results

Levels of sE-selectin were significantly higher in non-surviving patients than in survivors (p = 0.005) and significantly correlated with the following inflammatory markers: CRP (r = 0.242, p = 0.001), TNF-α (r = 0.201, p = 0.005), TNF-RII (r = 0.207, p = 0.004), and IL-6 (r = 0.339, p < 0.0001). According to Cox regression analysis of the prediction power of sE-selectin for all-cause mortality, high sE-selectin levels independently and significantly predicted short-term mortality in CHF (HR 1.47, 95% CI 1.103–1.956). Furthermore, sE-selectin predicted mortality in CHF patients with concomitant diabetes mellitus, as well as simultaneously elevated sE-selectin and NT-proBNP levels additively predicted mortality.

Conclusions

This study demonstrated a weak correlation of sE-selectin level with inflammatory markers and prediction of short-term mortality in diabetic CHF patients. Elevated serum sE-selectin levels and concomitantly increased NT-proBNP concentrations have additive predictive power in CHF. This suggests that parallel activation of various pathophysiological pathways confers increased risk of adverse outcome in CHF.

Keywords

Heart failuresE-selectinNT-proBNPInflammation

Introduction

Chronic heart failure (CHF) is a complex cardiovascular disorder affecting the musculoskeletal, renal, neuroendocrine, and immune systems in addition.

Thus, the pathophysiology of CHF is extremely complicated [1]. Elevation of the plasma levels of cell adhesion molecules and other markers of endothelial activation/abnormalities have been demonstrated as potential manifestations of volume overload and systemic inflammation in CHF [2]. Endothelial dysfunction, shown to be present in CHF, can be assessed by in vivo techniques such as flow-mediated dilatation [3] and by measuring the changes of specific plasma markers, including von Willebrand factor (VWF) [4], endothelins [5], and soluble E-selectin [6]. The serum level of soluble E-selectin, an indicator of inflammatory endothelial activation, is elevated in a variety of cardiovascular diseases although data on its ability to predict adverse cardiovascular events are controversial [712]. E-selectin is expressed by activated endothelium exclusively; however, a circulating form of E-selectin (sE-selectin) can be detected in the plasma. This may result either from enzymatic cleavage or from shedding by damaged or activated endothelial cells. It is a key adhesion molecule with a fundamental role in endothelial progenitor cell-dependent reparative mechanisms in experimental ischemia. Additionally, it also contributes to leukocyte adherence to the endothelium in inflammatory processes. Acute and chronic inflammation may damage the endothelium and cytokines like IL-1 or TNF-α increase the level of sE-selectin via immunological activation, as shown by in vitro studies [8, 9].

Although multiple studies have demonstrated an elevated sE-selectin level in CHF patients as compared to healthy controls, less is known about its biological correlates and its relationship to the progression and outcomes of the disease [7, 1315]. In addition, a potential interaction between sE-selectin and NT-proBNP in predicting mortality in CHF has not been reported yet.

We hypothesized that in CHF characterized by systemic inflammation and endothelial activation, alterations of sE-selectin level may occur, as well as concomitant endothelial (sE-selectin) and neuroendocrine activation (NT-proBNP) additively predicts mortality. To test this hypothesis, we investigated the relationship between soluble E-selectin, NT-proBNP, and mortality in a sufficiently large, well-characterized, prospective cohort of CHF patients [16, 17].

Methods

Study cohort

The study was implemented in observance of the Helsinki Declaration, at the 3rd Department of Internal Medicine, Semmelweis University, according to a study protocol that had been approved beforehand by the supreme research ethics committee in Hungary. Consecutive patients with clinical signs of CHF, referred for transthoracic echocardiography, were screened for inclusion. All cardiology in- or outpatients with a left ventricular ejection fraction (LVEF) <45% who contributed written informed consent were included, regardless of the etiology of the disease. Patients with an acute, initial episode of HF were not enrolled. All inpatients hospitalized for decompensation of chronic HF were in a compensated state as regards fluid balance (since at least two consecutive days, as required for discharge from hospital). The exclusion criteria comprised co-existing chronic or acute infections. A total of 192 patients (143 men, 49 women) were enrolled between February 2005 and April 2007. At the baseline visit, a complete clinical record was compiled including detailed physical status and the results of routine laboratory tests. Following an overnight fast of minimum 6 h, blood samples were obtained between 8 and 10 AM by antecubital venipuncture into native, EDTA or sodium citrate anticoagulated tubes. The samples were processed to obtain aliquots of serum and plasma, and then stored at −70°C until laboratory analysis.

The patients were interviewed at visit 1 [14.9 ± 6.8 months (mean, SD) after study entry] to record all major clinical events and mortality.

Determination of soluble E-selectin

Levels of soluble E-selectin were determined in citrated plasma, using DuoSet ELISA Development System assay from R&D Systems (Europe Ltd., Abingdon, UK, Cat. No. DY724). The intra- and inter-assay coefficients of variation (CV) were 1.26 and 13.6%, respectively; the detection limit was 1.67 ng/mL.

Determination of other laboratory parameters

Levels of NT-proBNP (Biomedica ELISA kit Cat. No. BI-20852), serum TNF-α (R&D System high sensitivity ELISA kit (Cat. No. HSTA00C), soluble TNF-receptor-I (sTNF-RI) (R&D System Human sTNF RI/TNFRSF1A Quantikine ELISA kit (Cat. No. DRT100), sTNF-RII (R&D System Human sTNF RII/TNFRSF1B Quantikine ELISA kit (Cat. No. DRT200) and serum IL-6 (R&D System high sensitivity ELISA kit (Cat. No. HS600B) were measured according to the manufacturer’s instructions. Standard laboratory parameters were measured with a Roche Integra 800 (clinical chemistry, CRP), or a Cell-Dyn 3500 hematology analyzer (complete blood count).

Statistical analysis

As the majority of variables were not normally distributed, measured values are presented as medians with 25th–75th percentile or as absolute numbers (percentages) for descriptive purposes. Non-parametric tests were used for group comparisons. Continuous variables were compared between the two groups with the Mann–Whitney U test, whereas categorical variables were compared with Pearson’s χ2 test. Spearman’s rank-order correlation coefficients were calculated to estimate interrelationship between sE-selectin and other variables.

Multivariate Cox proportional hazard models were fitted to assess the effect of sE-selectin on the primary outcome event. Survival times were measured from inclusion into the study to the endpoint, which was all-cause mortality. Patient characteristics and laboratory markers (as presented in Table 1)—together with the SD-standardized variants of continuous variables—were evaluated using a multitude of univariate Cox regressions. The best predictor (according to χ2 values) from each group of variables of clinically valid pathological pathways was included in the multivariate models to adjust for known effects and the study variables. The results of the Cox regression model are presented as hazard ratios standardized on the SD of the predictors with corresponding 95% confidence intervals (CI), as well as χ2 and p values of likelihood ratio tests. Age was analyzed as a time-dependent covariate. To study interaction (additive) effects between sE-selectin and NT-proBNP, we divided the study population into four groups according to high or low values of the variables. Cut points with optimum sensitivity and specificity for sE-selectin and NT-proBNP were obtained with ROC analysis.
Table 1

Baseline clinical characteristics of CHF patients

Characteristics

All patients (n = 192)

Non-survivors (n = 46)

Survivors (n = 146)

p value

Age (years)

69.4 (59.1–77.3)

73.4 (62.5–79.7)

68.9 (58.9–76.6)

0.201

Females/males, % (n)

25.5/74.5 (49/143)

19.6/80.4 (9/37)

27.4/72.6 (40/106)

0.288

BMI (kg/m2)

26.6 (24.3–30.9)

25.4 (23–28)

27.2 (24.5–31.2)

0.014

NYHA, n (%) I/II/III/IV

37/63/69/23 (19/33/36/12)

2/8/22/14 (4/17/48/30)

35/55/47/9 (24/38/32/6)

<0.0001

NT-proBNP (pmol/mL)

0.691 (0.370–1.565)

1.596 (0.886–3.597)

0.530 (0.330–1.252)

<0.0001

LVEF (%)

34 (27–40)

28.5 (22.8–37.4)

35 (29–41)

0.003

Ischemic etiology, % (n)

59.2% (113)

63% (29)

58% (84)

0.539

Peripheral edema, n (%)

81 (42.2)

31 (67.4)

50 (34.2)

<0.0001

Diabetes mellitus, % (n)

37.5 (72)

47.8 (22)

34.2 (50)

0.097

Heart rate (1/min)

80 (70–90)

80 (70–84)

79 (68–92)

0.805

Systolic BP (mmHg)

120 (110–140)

120 (100–130)

124 (115–140)

0.009

Diastolic BP (mmHg)

80 (70–80)

70 (60–80)

80 (70–80)

0.006

Sodium (mmol/L)

140 (137–142)

138 (135–141)

141 (138–143)

0.003

Creatinine (μmol/L)

97 (78–138)

129 (80–199)

93.5 (78–112)

0.002

BUN (μmol/L)

8.6 (6.3–12.6)

12.8 (8.6–22.2)

7.8 (5.7–10.3)

<0.0001

Bilirubin (μmol/L)

13.4 (8.9–20.8)

16.8 (9.2–30.8)

12.6 (8.8–19)

0.031

HDL-cholesterol (mmol/L)

1.16 (0.95–1.39)

1.07 (0.91–1.36)

1.2 (0.98–1.41)

0.255

LDL-cholesterol (mmol/L)

2.29 (1.77–2.96)

1.92 (1.43–2.51)

2.4 (1.93–3.02)

0.003

VLDL-cholesterol (mmol/L)

0.54 (0.42–0.77)

0.47 (0.36–0.72)

0.57 (0.43–0.77)

0.059

AST (U/L)

23 (18–33)

23.5 (18.8–32.0)

23 (18–33)

0.921

ALT (U/L)

23 (17–36)

21 (13–34)

23 (18–38)

0.105

ALP (U/L)

80 (64–103)

95 (72–149)

78 (63–100)

0.002

Gamma-GT (U/L)

65 (34–123)

95 (36–139)

60 (31–115)

0.121

Fasting blood glucose (mmol/L)

5.6 (4.9–7.3)

6.4 (5.2–8.4)

5.5 (4.9–6.9)

0.007

GFR (mL/min)

65.5 (45–85)

46 (30–75)

69 (50–86)

0.002

WBC (109/L)

7.27 (6.18–8.48)

7.57 (6.67–9.07)

7.13 (5.99–8.29)

0.049

Hb (g/L)

142 (129–154)

137 (115–150)

143 (132–154)

0.020

PLT (109/L)

195 (161–231)

193 (149–234)

196 (166–231)

0.580

CRP (mg/L)

6.4 (3.0–14.6)

9.2 (3.3–21.8)

5.4 (3.0–11.0)

0.022

Albumin (g/L)

41 (38–44)

39 (35–41)

41 (39–44)

<0.0001

Total protein (g/L)

72 (67–77)

71 (64–77)

72 (67–77)

0.115

IL-6 (pg/mL)

10.2 (6.1–16.8)

15.5 (10.5–27.8)

8.8 (4.9–14.1)

<0.0001

TNF-α (pg/mL)

2.4 (1.4–3.7)

3.0 (2.1–5.0)

2.2 (1.4–3.5)

0.021

sE-selectin (ng/mL)

18.5 (12.6–23.7)

22.1 (16–26.9)

17.5 (12.2–22.5)

0.005

Numbers in the table are medians (and interquartile ranges) or numbers of patients (percentages)

NYHA New York Heart Association, BMI body mass index, NT-proBNP N-terminal pro-brain-natriuretic-peptide, LVEF left ventricular ejection fraction, BP blood pressure, BUN blood urea nitrogen, AST aspartate aminotransferase, ALT alanine aminotransferase, ALP alkaline phosphatase, Gamma-GT gamma glutamyl transpeptidase, GFR glomerular filtration rate, WBC white blood cell count, Hb hemoglobin, PLT platelet count, CRP C-reactive protein, IL-6 interleukin-6, TNF-α tumor necrosis factor-α

Mann–Whitney test or Pearson χ2 test

A significance level of 0.05 was used throughout the study.

Statistical analyses were carried out using SPSS v13.01 (Apache Software Foundation, USA) and GraphPad Prism v4.03 (GraphPad Software, San Diego, CA, USA) software.

Results

Baseline characteristics

Baseline characteristics of CHF patients, stratified according to all-cause mortality during the follow-up period, are shown in Table 1. All patients had clinical (symptomatic) left ventricular systolic dysfunction (LVEF < 45%). The cohort was properly characterized in order to obtain detailed information on cardiac function, electrolyte and fluid homeostasis, renal or hepatic impairment, hemopoietic and endothelial dysfunction, as well as inflammation. The patients received the following treatments: loop diuretics 146 (76%), an ACE-inhibitor or angiotensin receptor blocker (ARB) 133 (69%), β blockers 131 (68%), statins 76 (40%), and aspirin 75 (40%). The study population (n = 192) was dominated by males (74.5%), and its mean age was 68.2 ± 11.5 years. The proportion of patients with NYHA class I to II CHF was 52%, whereas of those with class III to IV disease was 48% (Table 1). The patients who died during follow-up were older, belonged to a higher NYHA class, developed peripheral edema more frequently, had a reduced LVEF along with significantly lower systolic and diastolic blood pressure. Plasma NT-proBNP, fasting blood glucose, creatinine, blood urea nitrogen, inflammatory markers (CRP, TNF-α, IL-6), and soluble E-selectin levels were significantly higher, compared to survivors (Table 1). Although diabetes mellitus (DM) accompanying CHF occurred more frequently in the non-survivor group (48 vs. 34% among survivors), this difference was not statistically significant (Table 1).

Relationship between sE-selectin levels and clinical/laboratory variables

In Table 2, patients are grouped according to tertiles of sE-selectin level. Patients in the uppermost sE-selectin tertile (≥21.83 ng/mL) belonged to a higher NYHA functional class and were characterized by a higher prevalence of diabetes and CHF complications (peripheral edema and pulmonary congestion) when compared with patients in the other two tertiles. There was a positive, linear association between sE-selectin levels and NYHA functional class (Fig. 1; Kruskal–Wallis test p = 0.039). Comparing sE-selectin level with drug therapies (ACEIs or ARBs, β blockers, statins, and aspirin), we did not find any association (data not shown). The 133 (69%) inpatients had a significantly (p = 0.001) higher sE-selectin level (median 19.76 IQ range 13.70–26.02 ng/mL), compared to outpatients (median 15.21, 11.40–18.5 ng/mL). Furthermore, high sE-selectin levels were associated with the presence of peripheral and pulmonary edema (Table 2). WBC and the levels of inflammatory markers (CRP, TNF-α, TNF-RI, TNF-RII, IL-6) were significantly higher in the uppermost sE-selectin tertile than in the middle or low tertiles. These results were confirmed by Spearman’s rank correlation analyses (Table 3) testing the association between serum sE-selectin as continuous variable and some of the risk factors listed in Table 1. Notable correlations were found between sE-selectin and WBC (r = 309, p < 0.0001), as well as IL-6 (r = 339, p < 0.0001) level, whereas the correlations observed for blood glucose and inflammatory markers (CRP, TNF-α, IL-1-β) were though significant statistically but very weak in biological terms. We did not find significant correlation for NT-proBNP level.
Table 2

Baseline characteristics of the cohort, stratified by sE-selectin values (n = 192)

 

sE-sel ≤ 14.80 ng/mL

sE-sel 14.80-21.83 ng/mL

sE-sel ≥ 21.83 ng/mL

p value*

Median

Interquartile range

Median

Interquartile range

Median

Interquartile range

Age (years)

73

63–78

71

60–79

65

57–75

0.044

Female/male, % (n)

22/78 (14/50)

30/70 (19/45)

25/75 (16/48)

0.686

NYHA (I/II/III/IV) (%)

27/31/33/9

22/39/28/11

9/28/47/16

0.008

Heart rate (1/min)

72

65–86

80

70–92

80

70–90

0.313

Systolic BP (mmHg)

122

115–138

120

110–139

120

110–140

0.936

Diastolic BP (mmHg)

80

70–80

80

70–80

71

66–80

0.445

BMI (kg/m2)

27

24–30

26

24–29

28

24–32

0.340

LVEF (%)

34

28–41

34

27–41

34

24–40

0.509

HbA1c (%)

5.98

5.69–7.15

7.27

5.88–10.60

7.19

6.27–8.75

0.065

Fasting blood glucose (mmol/L)

5.22

4.80–6.55

5.61

5.02–6.88

6.47

5.24–8.28

0.002

Sodium (mmol/L)

142

139–143

140

138–142

139

135–141

0.001

WBC (109/L)

6.6

5.8–7.8

7.3

6.2–8.2

8.0

6.7–9.7

0.001

NT-proBNP (pmol/mL)

0.652

0.380–1.385

0.664

0.358–1.556

0.929

0.376–1.889

0.522

Peripheral edema, % (n)

33 (21)

41 (26)

53 (34)

0.020

Pulmonary congestion, % (n)

25 (16)

41 (26)

59 (38)

<0.0001

Diabetes mellitus, % (n)

27 (17)

34 (22)

52 (33)

0.004

Current smoking, % (n)

6 (4)

13 (8)

21 (13)

0.017

IL-6 (pg/mL)

6.95

4.76–11.31

10.79

6.69–17.06

14.19

9.57–21.47

<0.0001

TNF-α (pg/mL)

2.11

1.37–3.39

2.54

1.33–3.54

2.78

1.79–4.68

0.058

TNF-RI (ng/mL)

6.17

4.21–10.01

5.15

3.17–7.05

7.18

3.74–10.42

0.019

TNF-RII (ng/mL)

3.55

2.54–4.88

4.06

3.07–5.62

4.73

3.50–6.50

0.005

CRP (mg/L)

4.65

2.5–8.2

7.43

1.8–15.3

9.67

3.90–17.33

0.012

Creatinine (μmol/L)

97

81–112

92

78–136

102

78–145

0.465

Albumin (g/L)

41

39–44

41

38–43

41

37–44

0.274

Total protein (g/L)

71

66–76

72

67–77

73

67–77

0.709

NYHA New York Heart Association, BP blood pressure, BMI body mass index, LVEF left ventricular ejection fraction, HbA1c hemoglobin A1c, WBC white blood cell count, NT-proBNP N-terminal pro-brain-natriuretic-peptide, IL-6 interleukin-6, TNF-α tumor necrosis factor-α, TNF-R tumor necrosis factor receptor, CRP C-reactive protein

*Kruskal–Wallis ANOVA or Pearson χ2 test

https://static-content.springer.com/image/art%3A10.1007%2Fs00392-011-0283-6/MediaObjects/392_2011_283_Fig1_HTML.gif
Fig. 1

Box-and-whiskers plot (median, IQ range and range is indicated) of soluble E-selectin levels versus NYHA class. Kruskal–Wallis ANOVA test (p = 0.039)

Table 3

Correlation between sE-selectin and selected clinical variables in patients with chronic heart failure (n = 192)

 

sE-selectin

r

p value

NT-proBNP (pmol/mL)

0.051

0.479

Fasting blood glucose (mmol/L)

0.261

<0.0001

WBC (109/L)

0.309

<0.0001

CRP (mg/L)

0.242

0.001

IL-6 (pg/mL)

0.339

<0.0001

Spearman’s rank correlation coefficients with p values are presented, entries typeset in bold denote statistically significant difference after Bonferroni correction (p < 0.01)

NT-proBNP N-terminal pro-brain-natriuretic-peptide, WBC white blood cell count, CRP C-reactive protein, IL-6 Interleukin-6

Serum sE-selectin level and survival

During the follow-up period (mean 14.9 ± 6.8 months), 46 patients died, and the 12-month mortality rate was 0.17. In a Cox univariate survival analysis, the risk of all-cause mortality increased with serum sE-selectin level [HR (95% CI) for sE-selectin as a continuous variable was 1.030 (0.999–1.062), p = 0.055]. The relative risk of death was higher in patients with an sE-selectin level above 20.1 ng/mL [an optimum cut point determined by ROC analysis, sensitivity 0.609, specificity 0.685 (AUC 0.639, 95% CI of AUC 0.548–0.73)], as compared to patients with lower sE-selectin levels (Fig. 2; log-rank test p = 0.002). This relationship was also present in diabetics [n = 72, HR 3.58 (1.33–9.66)], but was not significant in the non-diabetic subgroup [HR 1.73 (0.75–3.94)].
https://static-content.springer.com/image/art%3A10.1007%2Fs00392-011-0283-6/MediaObjects/392_2011_283_Fig2_HTML.gif
Fig. 2

Kaplan–Meyer survival curves of serum sE-selectin versus all cause mortality. The study cohort was divided into two subsets of those with a sE-selectin level of 20.1 ng/mL or higher (broken line) and of those with a level less than 20.1 ng/mL (solid line). Log-rank statistic = 10.611, p = 0.001

The independent prognostic value of serum sE-selectin (as a continuous variable for all-cause mortality) was further explored by multivariate Cox regression analysis after adjustment for age, sex, glomerular filtration rate, and hemoglobin level. Soluble E-selectin independently predicted mortality in CHF patients (HR 1.469, 95% CI 1.103–1.956, p = 0.009) (Table 4).
Table 4

Multivariate Cox regression analysis of serum sE-selectin as a continuous variable for all cause mortality (n = 192)

Variable

HR

95% CI

χ2 change

p value

sE-selectin (ng/mL)

1.469

1.103–1.956

6.57

0.01

Age (years)

1.004

1.000–1.007

5.82

0.016

Sex (male)

1.821

0.856–3.871

2.67

0.102

Glomerular filtration rate (mL/min)

0.988

0.696–1.404

0.005

0.945

Hemoglobin (g/L)

0.639

0.475–0.860

8.200

0.004

HR hazard ratio/category change, CI confidence interval

In a multivariate Cox regression analysis, we tested the interaction of sE-selectin and NT-proBNP levels in predicting all-cause mortality in patients with CHF. Cut points for sE-selectin and NT-proBNP were calculated with ROC analysis, which yielded a sE-selectin level of 20.1 ng/mL and an NT-proBNP level of 0.553 pmol/mL (sensitivity 0.891, specificity 0.507) as optimum cut-off levels. Patients with high sE-selectin and high NT-proBNP levels had an increased risk for mortality [17.87 (95% CI 4.18–76.35) times higher] than those with lower levels of both variables (Fig. 3; Table 5). However, analyzing only the NT-proBNP levels in the adjusted model, the hazard ratio was 7.08 (2.74–18.30), which is very similar to that of patients with a high NT-proBNP and a low sE-selectin level [6.58 (1.49–29.01) ng/mL]. Thus, these two variables (i.e. sE-selectin and NT-proBNP) are strongly additive in predicting mortality from CHF. It should be noted, however, that there was no correlation between sE-selectin and NT-proBNP levels (r = 0.051, p = 0.479) (Table 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs00392-011-0283-6/MediaObjects/392_2011_283_Fig3_HTML.gif
Fig. 3

Kaplan–Meyer survival curves of patients with different sE-selectin and NT-proBNP levels. Cut points for sE-selectin and NT-proBNP were calculated with ROC analysis. A sE-selectin level of 20.1 ng/mL (sensitivity 0.609, specificity 0.685) and an NT-proBNP of 0.553 pmol/mL (sensitivity 0.891, specificity 0.507) were identified as optimum cut-off levels (Log-rank p < 0.0001)

Table 5

Multivariate (adjusted for age, sex, glomerular filtration rate and hemoglobin level) Cox regression analysis of sE-selectin and NT-proBNP levels analyzed as categorical predictors for prediction of all cause mortality in patients with chronic heart failure (n = 192)

Variable

HR

95% CI of HR

p value

sE-selectin ↓ NT-proBNP ↓

1 (reference)

  

sE-selectin ↑ NT-proBNP ↓

2.82

0.46–17.24

0.261

sE-selectin ↓ NT-proBNP ↑

6.58

1.49–29.00

0.013

sE-selectin ↑ NT-proBNP ↑

17.87

4.18–76.35

<0.0001

Cut points calculated by ROC analysis were 20.1 ng/mL for sE-selectin and 0.553 pmol/mL for NT-proBNP (see in the text)

HR hazard ratio/category change, CI confidence interval

Discussion

This study shows that, in our prospective CHF cohort, serum sE-selectin predicted all-cause mortality in diabetic patients, irrespective of other important risk factors and co-variates. In a Cox regression model, we demonstrated that high sE-selectin and NT-proBNP levels additively predict the probability of a fatal outcome in CHF. Furthermore, we found that diabetic patients have a higher sE-selectin level compared to non-diabetics, and importantly, high sE-selectin levels are known to be associated with inflammatory markers.

In CHF patients, sE-selectin level was found elevated in comparison to healthy controls [7, 13, 15]. This study extended these observations by showing that increased sE-selectin levels predict all-cause mortality in CHF, especially in patients with accompanying diabetes mellitus. This finding corroborates the results of Kistorp et al. [14], who also showed that in CHF elevated sE-selectin levels are characteristic of the subgroup of patients with DM. This association was also observed in the study by Chong et al. [12]; however, in that study, higher sE-selectin levels were not typical of CHF outpatients with only a few disease symptoms, as compared to healthy controls. However, our study population is entirely different from the cohort reported by Chong et al. since we included more inpatients with advanced heart failure.

Plasma sE-selectin level is a potential marker of endothelial cell damage or activation since the concentration of sE-selectin appears to correlate with its expression on the surface of human umbilical vein endothelial cells [18]. This is supported by our current results because increased sE-selectin levels were observed both in patients with edematous complications and after the decompensation of heart failure. Thus, volume overload is well reflected by this biomarker. However, the release of sE-selectin may result also from shedding, mediated by E-selectin-cleaving enzymes. By cleaving additional adhesion molecules, this process could contribute directly to increased permeability and edema formation. Cytokines, such as IL-1 or TNF-α—known to increase in acute or chronic inflammation and also in CHF—seem to increase sE-selectin level [9], possibly through inducing E-selectin-cleaving enzymes. Sustained endothelial activation and shedding of adhesion molecules could lead to endothelial dysfunction and eventually to irreversible endothelial damage if the stimulus persists [19]. This possibility is supported by the observed positive correlation between sE-selectin and inflammatory markers including CRP, TNF-α, soluble TNF receptors, IL-6, and white blood cell counts. These findings are in agreement with previous results, inasmuch as the plasma levels of inflammatory cytokines, adhesion molecules, and markers of endothelial dysfunction were reported to be higher in CHF patients than in normal controls [2, 20, 21]. Progressive CHF is associated with a more intense inflammatory reaction and more severe endothelial dysfunction—the significant association demonstrated between sE-selectin level and NYHA functional class (in agreement with Kistorp et al. [14]) fits into this pathomechanism. Furthermore, our observation showing multiple associations between inflammatory markers and sE-selectin levels indicates that inflammation is the primary mechanism leading to the shedding of E-selectin from endothelial cells. Since increased inflammation is linked to the metabolic syndrome and diabetes [22], the association between inflammation and elevated sE-selectin levels may explain why the latter are predictive of mortality in CHF patients with DM.

Additionally, the plasma concentrations of soluble E-selectin were analyzed in relation to associated co-morbidities and risk factors. The presence of peripheral edema, pulmonary congestion, diabetes mellitus, and current smoking was associated with higher sE-selectin levels. Since E-selectin is a leukocyte adhesion molecule specific to endothelial cells, the latter may be stimulated when granulocytes adhere to the endothelium. This process may induce the rounding up of endothelial cells and facilitate leukocyte transmigration, as well as the diffusion of various chemotactic substances [23]. This is in agreement with our result that peripheral edema and pulmonary congestion develop more frequently in patient with high sE-selectin level.

An important, new finding of our study is that, in the CHF prospective cohort, elevated sE-selectin and NT-proBNP levels additively predicted all-cause mortality in diabetic patients. Similar to the results of a previous study [15], we did not observe a significant correlation between sE-selectin and NT-proBNP levels. Thus, the additive predictive power of NT-proBNP and soluble E-selectin levels may be attributed to the fact that the two affected, parallel pathways (i.e. neuroendocrine and endothelial dysfunction) converge to determine the clinical prognosis jointly in these patients. In support of the additive predictive power of these markers, Potapov et al. [11] showed that, in inotrope-dependent heart failure, the combination of high BNP and sE-selectin levels had an increased predictive power in estimating the onset of clinical deterioration 1 day before its actual occurrence. Therefore, our current results may pinpoint and emphasize the importance of using multimarker prediction strategies in patients with CHF [24] and fuel further investigation of the hypothesis by future studies in this field.

A potential limitation of our study is the heterogeneity of the cohort, caused by the recruitment of outpatients, as well as of hospitalized patients. However, since patients were included after full compensation of an acute decompensation of CHF, we are convinced that this factor had no significant influence on our results.

In conclusion, this study demonstrated a weak correlation of sE-selectin level with inflammatory markers and the prediction of short-term mortality in diabetic CHF patients. Elevated serum sE-selectin levels with concomitantly increased NT-proBNP concentrations have additive predictive power in CHF. This indicates that parallel activation of different pathophysiological pathways in CHF confers an increased risk of adverse clinical events. These observations can be utilized for designing further clinical studies evaluating multimarker prediction strategies in CHF.

Acknowledgments

We are grateful to our patients who consented to participate in this study. The authors appreciate the skillful technical assistance of Holeczky Rudolfné, Szigeti Antalné, Korponai Gézáné, Sturmann Piroska and Kókai Márta. This study was supported by the following grants: Hungarian Scientific Research Fund (OTKA T046837, NF72689, ZP), and National Development Agency TÁMOP 4.2.2-08/01/KMR-2008-0004 (Semmelweis Bridge Project).

Conflict of interest

No financial conflicts of interest are present with regard to this manuscript by the authors.

Copyright information

© Springer-Verlag 2011