OUP user menu

Prognostic value of apoptosis markers in advanced heart failure patients

Alexander Niessner, Philipp J. Hohensinner, Kathrin Rychli, Stephanie Neuhold, Gerlinde Zorn, Bernhard Richter, Martin Hülsmann, Rudolf Berger, Deddo Mörtl, Kurt Huber, Johann Wojta, Richard Pacher
DOI: http://dx.doi.org/10.1093/eurheartj/ehp004 789-796 First published online: 4 February 2009

Abstract

Aims Apoptosis plays an important role in the progression of heart failure (HF). The purpose of this study was to assess whether the pro-apoptotic molecules apoptosis-stimulating fragment (FAS, CD95/APO-1) and tumour necrosis factor-related apoptosis-inducing ligand (TRAIL) predict event-free survival of HF patients.

Methods and results We assayed soluble (s)FAS and sTRAIL levels in 351 patients with advanced HF. During the median follow-up time of 16 months, 175 patients (50%) experienced the composite endpoints: rehospitalization and death. The hazard increased with sFAS concentrations, with a hazard ratio of 2.3 comparing fourth and first quartiles. This association remained significant after adjustment for B-type natriuretic peptide (BNP) and other risk factors in a Cox regression model (P = 0.014). Patients with high sFAS but low BNP had a comparable event-free survival rate with those with elevated BNP only (P = 0.78). Conversely, high sTRAIL concentrations were related to a better prognosis. Particularly, the risk of mortality dropped by 70% in the fourth quartile of sTRAIL (P = 0.001, multivariable Cox regression model).

Conclusion sFAS is an independent risk predictor in advanced HF patients. It may be of particular value for the identification of high-risk patients in addition to BNP. Conversely, sTRAIL appears to be protective and could be an interesting therapeutic agent.

Keywords
  • TRAIL
  • FAS
  • Apoptosis
  • Heart failure
  • Cardiomyopathy

Introduction

Despite improved therapeutic options for heart failure (HF),1 >50% of those patients diagnosed with severe HF die within 1 year.2 B-type natriuretic peptide (BNP) has clearly improved the prediction of clinical events in HF patients.3 However, there is considerable variability of BNP levels in HF patients,4 and a refinement of prognosis would be desirable. A precise prediction of the clinical course is essential to select the appropriate treatment option ranging from pharmacological therapy up to heart transplantation.

Apoptosis plays a crucial role in the pathogenesis5 and progression6 of HF. Both ischaemic and idiopathic dilated cardiomyopathies are associated with apoptosis.7 Accordingly, the apoptotic activity measured in an ex vivo system predicted survival in HF patients.8 Thus, soluble markers of apoptosis reflecting the transition from a healthy into a failing myocardium could be early indicators of an unfavourable clinical course. So far, tumour necrosis factor-α (TNF-α) and its receptors have been the most thoroughly investigated soluble markers of apoptosis correlating with HF.911 However, blocking TNF-α in patients with HF resulted in no improvement.12,13 We sought to investigate the role of other mediators of apoptosis differentially inducing cell death. The pro-apoptotic protein apoptosis-stimulating fragment (FAS) may exert an adverse effect on the progression of HF.14 Whereas FAS ligand induces apoptosis when binding to cell-surface-bound FAS, soluble FAS (sFAS) may competitively inhibit the binding of FAS ligand to surface-bound FAS, thereby exerting a potential anti-apoptotic effect.15 TNF-related apoptosis-inducing ligand (TRAIL), another specific inducer of apoptosis, has been found to be elevated in HF patients.16 TRAIL particularly plays an important role in the development of plaque rupture causing myocardial infarction,1719 one of the major causes of HF. However, soluble TRAIL (sTRAIL) was inversely associated with coronary artery disease.20 Its role in HF remains to be elucidated. To assess the potential pathophysiological role and the predictive value of soluble markers of apoptosis in the progression of HF, we investigated their association with clinical characteristics, pharmacological treatment, and future clinical events in patients with advanced HF.

Methods

Study population

Study participants were prospectively enrolled at all the six cardiology departments in Vienna as part of a disease management programme. Trained study personnel screened and included patients with advanced systolic HF according to the following criteria: (i) current hospitalization due to clinical signs and symptoms of cardiac decompensation, (ii) New York Heart Association functional classification class III or IV at the time of admission, and (iii) cardiothoracic ratio >0.5, and/or left ventricular ejection fraction <40%. B-type natriuretic peptide was used to assess the severity of HF. Exclusion criteria were non-cardiac diseases with a life expectancy <1 year (e.g. neoplasia) and refusal to informed consent. A sample size of 350 participants with 40% of patients experiencing an endpoint in the unexposed group allowed to detect a risk ratio of 1.5 (α = 0.05, power >90%). Clinical data are listed in Table 1. The study was approved by the institutional review board.

View this table:
Table 1

Baseline characteristics

QuartilesAll (n = 351)sFASsTRAIL
1234P-value1234P-value
4646–8807 pg/mL (n = 88)8821–10 638 pg/mL (n = 88)10 647–12 959 pg/mL (n = 88)12 967–22 457 pg/mL (n = 87)6.2–47.1 pg/mL (n = 88)47.5–63.2 pg/mL (n = 88)63.3–84.9 pg/mL (n = 88)85.2–205.8 pg/mL (n = 87)
Age, years75 (63–82)r = 0.27<0.001r = −0.070.19
Male gender, n (%)232 (66)60 (68)69 (78)44 (50)59 (68)0.00161 (69)58 (66)61 (69)52 (60)0.5
Myocardial infarction, n (%)158 (45)28 (32)38 (43)43 (49)49 (56)0.01; 0.00135 (40)45 (51)40 (45)38 (44)0.45; 0.74
BNP, pg/mL441 (231–842)r = 0.23<0.001r = −0.28<0.001
eGFR, mL/min/1.73 m251 (36–67)r = −0.40<0.001r = 0.060.3
BMI, kg/m226.1 (23.7–29.4)r = −0.150.005r = 0.060.3
Diabetes mellitus, n (%)140 (40)29 (33)42 (48)34 (39)35 (40)0.25; 0.5536 (41)36 (41)38 (43)30 (35)0.67; 0.44
Hypertension, n (%)243 (69)58 (66)62 (71)62 (71)61 (70)0.88; 0.4962 (70)53 (60)61 (69)67 (77)0.11; 0.23
Systolic BP, mmHg120 (110–130)r = −0.030.63r = −0.090.1
Heart rate, b.p.m.74 (66–85)r = −0.110.04r = −0.010.82
Atrial fibrillation, n (%)107 (30)19 (22)29 (33)32 (36)27 (31)0.20; 0.1533 (38)24 (27)25 (28)25 (29)0.49; 0.28
RAAS inhibitor, n (%)307 (87)84 (95)75 (85)80 (91)68 (78)0.004; 0.00373 (83)78 (89)76 (86)80 (92)0.31; 0.12
ACE-inhibitor, n (%)246 (70)68 (77)65 (74)63 (72)50 (57)0.028; 0.00662 (70)62 (70)54 (61)68 (78)0.16; 0.45
Beta-blocker, n (%)259 (73)68 (77)65 (74)65 (74)61 (70)0.73; 0.3161 (69)62 (70)65 (74)71 (82)0.22; 0.046
Diuretics, n (%)114 (33)31 (35)26 (30)25 (28)32 (37)0.51; 0.8333 (38)26 (30)24 (27)31 (36)0.44; 0.80
Statins, n (%)98 (27)28 (32)19 (22)23 (26)25 (29)0.56; 0.8225 (28)31 (35)19 (22)23 (26)0.17; 0.39
  • Continuous data are presented as median (inter-quartile range), and their association with apoptotic markers was assessed using Spearman-Rho correlation coefficient. Categorical data are analysed using (i) χ2 test and (ii) a test for linear association (Maentel–Haenszel χ2 test).

  • FAS, apoptosis-stimulating fragment; TRAIL, TNF-related apoptosis-inducing ligand; eGFR, estimated glomerular filtration rate; BNP, B-type natriuretic peptide; ACE, angiotensin-converting enzyme; RAAS, renin–angiotensin–aldosterone system; Statins, hydroxymethylglutaryl coenzyme A reductase inhibitors.

Blood sampling and laboratory analysis

Biochemical parameters were analysed from venous blood samples obtained in the morning on the day of discharge before intake of medication. After centrifugation (2800 r.p.m., 20 min), EDTA plasma was stored at −80°C in multiple aliquots until analysed. sTRAIL and sFAS (both R&D Systems, Minneapolis, MN, USA) were determined by specific commercially available ELISA. Intra- and inter-assay coefficients were 3.4 and 7.5% for sTRAIL and 4.6 and 2.9% for sFAS, respectively. Concentrations of apparently healthy individuals were between 34 and 163 pg/mL for sTRAIL and between 5941 and 13460 pg/mL for sFAS. It has been described that sFAS is associated with age as well as with male gender in apparently healthy subjects.21 The minimum detectable dose was 2.9 pg/mL for sTRAIL and <20 pg/mL for sFAS. B-type natriuretic peptide was determined by a commercially available specific test (Viva, Bayer Health Care). The cut-off value for BNP was 100 pg/mL. All measurements were performed by staff unaware of the clinical data. Estimated glomerular filtration rate (eGFR) was calculated using the formula from the Modification of Diet in Renal Disease Study Group.22

Follow-up visits were arranged at the index hospital or participants were visited by trained study nurses at 1, 3, 4, 6, 12, and 24 months. The primary composite endpoints were all-cause mortality and rehospitalization for HF worsening requiring intravenous diuretic and/or inotropic support. The cause of admission was assessed by hospital reports. Mortality was confirmed by reviewing the death registry of ‘Oesterreichisches Melderegister’. For additional analysis, all-cause mortality alone was chosen as secondary endpoint.

Statistical methods

Continuous data which did not adhere to normal distribution are presented as median [inter-quartile range (IQR)] and were analysed using Kruskal–Wallis test. Correlations between variables were assessed using Spearman-Rho correlation coefficient. Categorical data were analysed using χ2 test and Maentel–Haenszel χ2 test for assessment of a linear trend. Kaplan–Meier curves (log-rank) and Cox proportional hazard regression models were used to evaluate the predictive value of biomarkers for event-free survival. The proportionality assumption for the Cox proportional hazard regression models was assessed using Nelson–Aalen plots and fulfilled for all variables included in the Cox regression models. A multivariable model was used to adjust for potential confounding baseline characteristics. The model encompassed demographics (age, sex), known predictors of clinical endpoints in patients with HF (prior myocardial infarction, diabetes mellitus, renal insufficiency, and BNP), and baseline characteristics associated with markers of apoptosis with a P ≤ 0.2 (using Spearman-Rho correlation for continuous variables and Maentel–Haenszel χ2 test for categorical variables, Table 1). Another multivariable model was used to adjust for therapy (as listed in Table 1). Interactions between apoptotic markers and all variables included in multivariable models were tested by entering interaction terms in the Cox proportional hazard regression models. To assess collinearity between BNP and apoptotic markers, we calculated variance inflation factors (VIF). Variance inflation factors very close to 1 (<1.1) between BNP and sFAS as well as sTRAIL indicated no evidence for collinearity. SPSS 15.0 was used for statistical analysis (SPSS Inc., Chicago, IL, USA). A two-sided P-value of ≤0.05 was considered statistically significant. Bonferroni–Holm correction was applied to adjust the α level for (i) testing two different biomarkers and two different endpoints in survival analysis and (ii) for performing six significance tests when comparing combined strata of sFAS and BNP.

Results

Patients

Out of 462 patients assessed for eligibility between July 2003 and September 2004, 441 met inclusion criteria of advanced HF. A total of 360 patients agreed to participate in the study. Values of apoptosis markers were available for 351 patients. The median BNP level was 441 pg/mL (IQR 231–842 pg/mL). Forty-five per cent of patients had a prior myocardial infarction. No patient had an acute coronary syndrome leading to hospitalization.

Baseline characteristics

Table 1 illustrates baseline characteristics in relation to sFAS and sTRAIL concentrations. sFAS was associated with age (r = 0.27, P < 0.001) and BNP levels (r = 0.23, P < 0.001, Figure 1). Moreover, a history of prior myocardial infarction was more frequently observed in patients with higher sFAS values (for trend, P = 0.001). In contrast, the eGFR (r = −0.40, P < 0.001), the body mass index (BMI, r = −0.15, P = 0.005), and the heart rate (r = −0.11, P = 0.04) declined with increasing concentrations of sFAS. Interestingly, there was an inverse association between the use of renin–angiotensin–aldosterone system (RAAS) inhibitors and quartiles of sFAS (for trend, P = 0.003). In contrast to sFAS, sTRAIL was not related to any baseline characteristics except for an inverse correlation of sTRAIL with BNP (r= −0.28, P < 0.001, Figure 1). Furthermore, sTRAIL was associated with the use of beta-blockers (for trend, P = 0.046).

Figure 1

Scatter plots illustrating the relationship between apoptotic markers and B-type natriuretic peptide.

Follow-up

During a median follow-up time of 16 months (IQR 12–19 months), 175 patients (50%) experienced the composite endpoint. During the follow-up period, 93 patients (26%) died.

Univariate analysis

The hazard of the composite endpoint increased with concentrations of sFAS (P < 0.001, Table 2) and was elevated by 2.3-fold in the fourth quartile compared with the first quartile (Figure 2). To determine a potential additive prognostic value of sFAS and BNP, we assessed event-free survival in combined strata of BNP and sFAS. Heart failure patients with both markers below the median had a 12 month event-free survival of 77% (Figure 3). Those patients with either sFAS or BNP above the median had a 12 month event-free survival of 52 and 53%, respectively (log-rank, P = 0.78 between strata with one marker elevated; P ≤ 0.001 between strata with one vs. no elevated marker, significant after Bonferroni–Holm correction). The stratum with both markers above the median showed the worst outcome with a 12 month event-free survival of only 36% (log-rank, P = 0.041 compared with high sFAS and low BNP, P = 0.018 compared with low sFAS and high BNP, not significant after Bonferroni–Holm correction). Soluble apoptosis-stimulating fragment was also a significant predictor of the secondary endpoint all-cause mortality (P < 0.001, Table 2) with a hazard ratio (HR) of 2.4 in the fourth quartile.

Figure 2

Survival curves according to quartiles of markers of apoptosis. Kaplan–Meier plots showing the crude cumulative survival free of rehospitalization for heart failure worsening and all-cause mortality according to quartiles of soluble apoptosis-stimulating fragment (sFAS) (upper panel) and soluble tumour necrosis factor-related apoptosis-inducing ligand (sTRAIL) (lower panel); log-rank test for the overall comparison among groups.

Figure 3

Survival curves according to the combined strata of soluble apoptosis-stimulating fragment (sFAS) and B-type natriuretic peptide (BNP). Kaplan–Meier plots showing the crude cumulative survival free of rehospitalization for heart failure worsening and all-cause mortality according to the combined strata of sFAS and B-type natriuretic peptide; +/− sFAS, above vs. below median soluble apoptosis-stimulating fragment (10 638 pg/mL); +/− BNP, above vs. below median B-type natriuretic peptide (440 pg/mL); log-rank test for the overall comparison among groups.

View this table:
Table 2

The effect of apoptosis markers on the survival of heart failure patients in Cox proportional hazards models

HRs (95% CI)P-value
sFAS
 All-cause mortality
  Univariate1.0000944 (1.0000414–1.0001474)<0.001
  Adjusted for baseline characteristicsa1.0000587 (0.9999984–1.0001189)0.056
  Adjusted for therapyb1.0000666 (1.0000122–1.0001209)0.016c
Composite endpoint
  Univariate1.0000911 (1.0000516–1.0001306)<0.001
  Adjusted for baseline characteristicsa1.0000526 (1.0000074–1.0000978)0.014c
  Adjusted for therapyb1.0000704 (1.0000297–1.0001111)0.001c
sTRAIL
 All-cause mortality
  Univariate0.983 (0.975–0.991)<0.001
  Adjusted for baseline characteristicsa0.986 (0.978–0.994)0.001c
  Adjusted for therapyb0.985 (0.977–0.993)<0.001c
 Composite endpoint
  Univariate0.992 (0.987–0.997)0.003
  Adjusted for baseline characteristicsa0.996 (0.990–1.001)0.12
  Adjusted for therapyb0.992 (0.987–0.997)0.003c
  • HRs refer to an increase of 1 pg/mL of the respective apoptosis marker. CI, confidence interval; FAS, apoptosis-stimulating fragment; TRAIL, TNF-related apoptosis-inducing ligand.

  • aAdjusted for demographics (age, male gender) and HF characteristics associated with prognosis (prior myocardial infarction, diabetes mellitus, eGFR, and B-type natriuretic peptide) and additional HF characteristics associated with sFAS (BMI, heart rate, atrial fibrillation) or sTRAIL (systolic blood pressure).

  • bAdjusted for RAAS inhibitors, beta-blockers, diuretics, and statins.

  • cSignificant after adjusting α levels of multivariable models for multiple testing (two biomarkers and two endpoints) by Bonferroni–Holm correction.

In contrast to sFAS, higher sTRAIL concentrations were associated with a lower risk of the composite endpoint (P = 0.003) with an HR of 0.5 comparing the fourth with the first quartile of sTRAIL (Table 2, Figure 2). In particular, the hazard of all-cause mortality decreased with increasing concentrations of sTRAIL (P < 0.001, Table 2) with a reduced risk of mortality by 70% in the fourth quartile.

Multivariable analysis

sFAS remained a significant risk predictor of the composite endpoint after adjustment for demographics and prognostic markers including BNP (P = 0.014, significant after Bonferroni–Holm correction, Table 2). sFAS significantly increased the predictive value of the multivariable model (likelihood ratio test, P = 0.017). No significant interaction was found between sFAS and any of the clinical characteristics including BNP (n.s., data not shown). The association of sFAS with all-cause mortality lost significance in the multivariable model (Table 2). Adjustment for therapy did not change the significant results of the univariate analyses (Table 2).

sTRAIL remained a strong inverse predictor of all-cause mortality after adjustment for demographics and HF characteristics (P = 0.001, significant after Bonferroni–Holm correction, Table 2). However, the association of sTRAIL with the composite endpoint lost significance in the multivariable model (P = 0.12, Table 2). No interaction was observed between sTRAIL and any of the clinical characteristics including BNP (n.s., data not shown). Adjustment for therapy did not change the significant results of the univariate analyses (Table 2).

Discussion

This study demonstrates that the risk of a clinical event increases with sFAS concentrations in patients with advanced HF with a 2.4-fold higher risk in the fourth quartile compared with the first quartile. Our results are in accordance with the previous clinical studies linking sFAS to the presence and severity of HF.2325 In the present study, sFAS was further associated with clinical characteristics related to a poor prognosis of HF patients, such as age and renal insufficiency. In parallel to other cytokines, high sFAS concentrations were also associated with a low BMI in our study population and may contribute to cachexia in HF patients.26 Although sFAS was related to a history of myocardial infarction, no variation of the prognostic value of sFAS was found among different aetiologies of cardiomyopathy.

Experimental studies have stressed the importance of FAS protein expression in ischaemic, inflammatory, and dilated cardiomyopathies.2729 However, experimental findings indicated a protective effect of sFAS in the myocardium. sFAS inhibited apoptosis of muscle cells in cell culture30 and improved the survival of animals with HF.14 The complex pathogenesis of HF in men may explain discrepant clinical results. Production of sFAS might primarily reflect the activation of cytotoxic lymphocytes which contribute to the development of cardiomyopathy.31,32 Accordingly, sFAS correlated with the soluble IL2 receptor, a marker of activated lymphocytes,23 and was elevated in auto-immune diseases15 triggered by cytotoxic lymphocytes. In addition to the potential effects of sFAS on the myocardium, circulating sFAS may also contribute to multi-organ dysfunction in HF patients.33

Of note, sFAS correlated with BNP. A stretch-induced expression of FAS protein and BNP in myocytes may be a potential explanation for this observation.34,35 Despite this correlation, the risk prediction of sFAS remained significant after adjustment for BNP and other HF characteristics. High sFAS indicated a poor prognosis even in patients with low BNP. Patients with high sFAS but low BNP had a comparable risk with those with elevated BNP only. Transformation of the myocardium at the cellular levels through apoptosis may precede elevated myocardial strain with additional secretion of BNP. However, patients with high sFAS reflecting ongoing apoptotic activity as well as high BNP indicating elevated myocardial strain had the highest risk of a clinical event. Thus, high sFAS may help to identify high-risk patients who may benefit from an intensified treatment independently of BNP. Interestingly, we found lower levels of sFAS in patients with RAAS inhibitors. We could not confirm previous findings linking the administration of beta-blockers to lower sFAS levels.36 Prospective intervention studies are needed to show whether sFAS-guided treatment may improve the prognosis of HF patients. In summary, sFAS is a promising biomarker with pathophysiological relevance and prognostic potential and may possibly be modulated by medical HF treatment.37 In particular, sFAS may help to identify patients with elevated risk despite low levels of BNP.

In contrast to sFAS, sTRAIL was inversely associated with events in advanced HF patients. Unlike membrane-bound TRAIL,18 sTRAIL has been shown to protect muscle cells expressing death receptors 4 or 5 from apoptosis. It even tended to induce muscle cell proliferation.38 But recombinant sTRAIL is also known for its ability to kill cancer cells in vitro. However, sTRAIL needs to be cross-linked to exert a pro-apoptotic effect. The balance between pro- and anti-apoptotic effects of TRAIL is further influenced by the presence of soluble decoy receptors including osteoprotegerin abundantly expressed in HF patients.39 In parallel to our findings, sTRAIL (but not sFAS) was reduced in patients with acute coronary syndrome.20 However, we did not find any interaction between the cause of HF and the prognostic value of sTRAIL. The strong inverse association of sTRAIL with mortality with a 70% reduction in the highest quartile of sTRAIL underlines the clinical importance of this biomarker. One may speculate that low sTRAIL is specifically associated with sudden cardiac death and to a lesser extent with worsening of HF leading to rehospitalization. Of note, treatment with beta-blockers preventing sudden cardiac death was associated with higher concentrations of sTRAIL.

In summary, sFAS was strongly related to an unfavourable prognosis. sFAS improved prediction beyond BNP. In contrast to sFAS, sTRAIL showed a protective effect with a markedly reduced mortality in patients with high levels of this biomarker. There is a need for further clinical studies in order to validate the present results and to define cut-off values before the implementation of these results into clinical practice. sTRAIL could not only be of interest for risk prediction but also as therapeutic agent. Animal studies are needed to prove a net protective effect of sTRAIL before progressing with intervention studies. In conclusion, we found discrepant effects of soluble markers of apoptosis on the prognosis of advanced HF patients indicating a complex interaction between mediators of apoptosis and progression of HF. These results emphasize the importance of fully understanding pathophysiological processes before selecting specific pathways as therapeutic target.

Funding

This work was supported by the Ludwig Boltzmann Cluster for Cardiovascular Research and by the Association for the Promotion in Research in Arteriosclerosis, Thrombosis and Vascular Biology.

Conflict of interest: none declared.

References

View Abstract