Abstract

Aims

Multimarker approaches for risk prediction in coronary artery disease have remained inconsistent. We assessed multiple biomarkers representing distinct pathophysiological pathways in relation to cardiovascular events in stable angina.

Methods and results

We investigated 12 biomarkers reflecting inflammation [C-reactive protein, growth-differentiation factor (GDF)-15, neopterin], lipid metabolism (apolipoproteins AI, B100), renal function (cystatin C, serum creatinine), and cardiovascular function and remodelling [copeptin, C-terminal-pro-endothelin-1, mid-regional-pro-adrenomedullin (MR-proADM), mid-regional-pro-atrial natriuretic peptide (MR-proANP), N-terminal-pro-B-type natriuretic peptide (Nt-proBNP)] in 1781 stable angina patients in relation to non-fatal myocardial infarction and cardiovascular death (n = 137) over 3.6 years. Using Cox proportional hazards models and C-indices, the strongest association with outcome for log-transformed biomarkers in multivariable-adjusted analyses was observed for Nt-proBNP [hazard ratio (HR) for one standard deviation increase 1.65, 95% confidence interval (CI) 1.28–2.13, C-index 0.686], GDF-15 (HR 1.59, 95% CI 1.25–2.02, C-index 0.681), MR-proANP (HR 1.46, 95% CI 1.14–1.87, C-index 0.673), cystatin C (HR 1.39, 95% CI 1.10–1.75, C-index 0.671), and MR-proADM (HR 1.63, 95% CI 1.21–2.20, C-index 0.668). Each of these top single markers and their combination (C-index 0.690) added predictive information beyond the baseline model consisting of the classical risk factors assessed by C-index and led to substantial reclassification (P-integrated discrimination improvement <0.05).

Conclusion

Comparative analysis of 12 biomarkers revealed Nt-proBNP, GDF-15, MR-proANP, cystatin C, and MR-proADM as the strongest predictors of cardiovascular outcome in stable angina. All five biomarkers taken separately offered incremental predictive ability over established risk factors. Combination of the single markers slightly improved model fit but did not enhance risk prediction from a clinical perspective.

See page 2980 for the editorial comment on this article (doi:10.1093/eurheartj/ehq336)

Introduction

Increasingly, detailed knowledge of pathways contributing to myocardial infarction and heart failure, as well as the availability of new technologies, has led to the recognition of a growing number of novel biomarkers detectable in peripheral blood. In patients with angiographically confirmed stable angina, the risk of non-fatal myocardial infarction and cardiovascular death is considerable.1 From a clinical perspective, there is a need to improve risk stratification with biomarkers that provide long-term discriminative information beyond established cardiovascular risk factors. So far, these efforts have been disappointing, and we and others have concluded that most contemporary biomarkers only add moderate incremental prognostic information in the setting of primary prevention and in those with manifest disease.2–5 Further exploration of novel biomarkers and the use of newer analytical methods have been encouraged.6

We selected a broad panel of biomarkers to represent multiple pathways that have been invoked in coronary artery disease (CAD). Natriuretic peptides, in our study represented by N-terminal pro B-type natriuretic peptide (Nt-proBNP) and mid-regional pro atrial natriuretic peptide (MR-proANP), provide valuable information about cardiovascular stress.3,7 The availability of novel immunoassays for markers of vascular function with potent autocrine and paracrine actions such as C-terminal(Ct)-pro-endothelin-1, copeptin (a precursor of vasopressin), and mid-regional pro adrenomedullin (MR-proADM) permits a large-scale comparative study of these biomarkers with existing risk factors such as natriuretic peptides and the inflammatory marker C-reactive protein. Growth-differentiation factor (GDF)-15, a distant member of the transforming growth factor beta family, has recently been shown to be related to mortality and cardiovascular events in patients with CAD.8,9 Its prognostic ability in stable CAD in direct comparison with other biomarkers has not been established. Biomarkers that sensitively mirror impaired renal function such as cystatin C also have been noted to improve risk stratification in various settings.10

Finally, it remains unclear, whether a strategy that uses simultaneous assessment of several novel biomarkers, which have little interdependence, is superior to strategies that use single-biomarker assessments or established risk factors only for cardiovascular risk prediction.2,3

Methods

Study sample

AtheroGene is a prospective cohort study of consecutive patients with manifest CAD and at least one stenosis of 30% or more present in a major coronary artery.11 The combined outcome for this study was non-fatal myocardial infarction and cardiovascular mortality. Information on events was collected by regular follow-up questionnaires mailed to the participants and telephone interviews, and verified by death certificates and hospital or general practitioner charts. As the purpose of the analyses was to establish the incremental value of biomarkers in stable patients with proven CAD, patients with acute coronary syndrome (n = 282) or missing covariate data (n = 231) were excluded, leaving n = 1781 participants with complete data sets for the analysis. Further details of the study are provided in the Supplementary material.

The study was approved by the local Institutional Review Board, all participants gave written informed consent. The authors had full access to the data and take responsibility for its integrity.

Laboratory methods

Twelve biomarkers reflecting lipids [apolipoproteins (Apo) AI, Apo B100], renal function (serum creatinine, cystatin C), inflammation (C-reactive protein, GDF-15, neopterin12), vascular function (copeptin, Ct-pro-endothelin-1, MR-proADM), and haemodynamics and remodelling (MR-proANP, Nt-proBNP) were measured by commercially available assay systems or antibodies. Serum creatinine was measured by the modified Jaffe routine method. Further details on the tests used are provided in see Supplementary material online, Table S1.

Statistical methods

Biomarker concentrations were logarithmically transformed if the distribution strongly deviated from a normal distribution if it enhanced model fit. Associations of the biomarkers were assessed by Spearman correlation coefficients.

The prognostic abilities of all biomarkers were investigated through multivariable proportional hazards regression adjusting for age, sex and in a second model additionally for body mass index (BMI), LDL/HDL ratio, smoking, diabetes, hypertension, and number of diseased vessels. Hazard ratios (HRs) per one unit of the biomarker and for one standard deviation (SD) increase in (log-transformed) biomarker concentration along with the 95% confidence intervals (CIs) are provided.

We further examined net reclassification improvement (NRI) for the risk categories 0–6, >6–20, and >20% for 3-year event-free survival and integrated discrimination improvement (IDI) by adapting recent approaches in order to make them suitable for survival data.6 For NRI we computed the predicted probability of an event over 3 years of follow-up for each individual using a Cox proportional hazards model based on the classical risk factors and categorized individuals according to their predicted risk. For reclassification analyses, the information of the novel biomarker was then added to the model. In individuals with a future event upward reclassification was indicated by adding +1, downward classification into a lower risk category was assigned −1 and missing reclassification was indicated by zero. These scores are summed and divided by the number of individuals in the event group. Individuals in the non-event group are reclassified the same way but assigned +1 for classification into a lower risk category, correct reclassification and −1 for reclassification into a higher risk rank and the overall net reclassification is derived as the NRI. The IDI is less intuitive. The advantage is that it does not depend on pre-specified risk categories. Integrated discrimination improvement evaluates reclassification as a continuous outcome across the spectrum of risk; a value of zero indicates no movement in predicted risk.

To derive the primary risk categories, we used the combination of the classical cardiovascular risk factors (age, sex, BMI, LDL/HDL ratio, smoking, diabetes, hypertension, and number of diseased vessels). We also present data on the additive value of the investigated biomarkers when used on top of the ESC (European Society of Cardiology) Score.13

For biomarkers with a significant IDI (P < 0.05), as an additional measure for assessing their discriminative ability C-indices were calculated and tested in bootstrap analysis.14 We evaluated models for the top biomarkers separately and for the combination of these biomarkers in linear regression models.

For the best biomarkers and their combination, we defined cut-off values for event-free survival using survival trees, generated by the rpart package, classification and regression trees (CART) in R.15 A binary tree is calculated from the whole group of predictors. The best predictor and its optimal cut value are chosen at each step. A log-rank statistic is used as a criterion for goodness of split for each subsequent node. In order to obtain a binary classification, only the first split was used. P-values for the likelihood-ratio-tests using the dichotomized variables are provided. We also present Kaplan–Meier curves for tertiles of the biomarker score (Figure 1).

Figure 1

Kaplan–Meier-survival curves for cardiovascular events by tertiles of the combination of the top biomarkers (score). The P-value is derived by log-rank test. Kaplan–Meier curves according to the cut-off values are provided in see Supplementary material online, Figure S1.

Figure 1

Kaplan–Meier-survival curves for cardiovascular events by tertiles of the combination of the top biomarkers (score). The P-value is derived by log-rank test. Kaplan–Meier curves according to the cut-off values are provided in see Supplementary material online, Figure S1.

P-values and CIs have been corrected for multiple testing (Bonferroni correction).

All analyses were carried out with R 2.10.1.16

Results

Baseline characteristics

Cardiovascular events were observed in 137 patients over a median follow-up of 3.6 (maximum 6.9) years. The baseline characteristics of the study participants are provided in Table 1. The median age was 63 years, 22.5% of study participants were female. Risk factor distribution was more unfavourable in individuals with cardiovascular event during follow-up. About half of the patients in the event group (51.8%) had a history of myocardial infarction. Approximately 80% of the sample had hypertension, one-fifth was diabetics, and more than half of the sample was treated with cardiovascular medication at baseline. Multivessel disease was present in ∼70% of the total sample and 80% of the event group. Most biomarker concentrations were higher in patients with cardiovascular events except for Apo AI, which tended to be lower in the event group.

Table 1

Baseline characteristics of the study sample

Variable All, n = 1781 No event, n = 1644 Event, n = 137 
Age, years 63 (56/69) 63 (55/69) 64 (60/70) 
Male sex, n (%) 1381 (77.5) 1283 (78.0) 98 (71.5) 
BMI, kg/m2 27.4 (25.0/30.1) 27.4 (25.0/30.1) 27.6 (25.0/29.8) 
Hypertension, n (%) 1402 (78.7) 1293 (78.6) 109 (79.6) 

 
Diabetes mellitus, n (%) 
 Diet 74 (4.2) 70 (4.3) 4 (2.9) 
 Oral medication 176 (9.9) 158 (9.6) 18 (13.1) 
 Insulin dependent 155 (8.7) 135 (8.2) 20 (14.6) 

 
Smoking, n (%) 
 Never 647 (36.3) 599 (36.4) 48 (35.0) 
 Former smoking 829 (46.5) 765 (46.5) 64 (46.7) 
 Current smoking 305 (17.1) 280 (17.0) 25 (18.2) 
Positive family history, n (%) 685 (38.5) 638 (38.8) 47 (34.3) 
History of myocardial infarction, n (%) 713 (40.0) 642 (39.1) 71 (51.8) 
Beta-blocker, n (%) 1181 (66.3) 1094 (66.5) 87 (63.5) 
ACE-inhibitors, n (%) 949 (53.3) 864 (52.6) 85 (62.0) 
Statins, n (%) 970 (54.5) 896 (54.5) 74 (54.0) 
LDL/HDL ratio 2.50 (1.86/3.19) 2.48 (1.85/3.18) 2.69 (2.19/3.31) 
Multivessel disease, n (%) 1280 (71.8) 1165 (70.9) 115 (83.9) 

 
Biomarkers 
 Apo AI, g/L 1.46 (1.29/1.65) 1.46 (1.30/1.66) 1.39 (1.23/1.59) 
 Apo B100, g/L 0.98 (0.82/1.14) 0.97 (0.82/1.14) 1.05 (0.85/1.18) 
 Creatinine, mg/dL 0.95 (0.83/1.08) 0.94 (0.83/1.07) 0.97 (0.84/1.17) 
 Cystatin C, mg/L 0.82 (0.72/0.96) 0.81 (0.72/0.95) 0.86 (0.72/1.09) 
 C-reactive protein, mg/L 2.19 (0.97/5.59) 2.09 (0.94/5.29) 3.86 (1.36/8.46) 
 GDF-15, ng/L 1155 (872/1616) 1133 (866/1564) 1516 (1025/2174) 
 Neopterin, nmol/L 6.23 (5.14/7.81) 6.19 (5.12/7.74) 7.01 (5.40/8.99) 
 MR-proADM, pmol/dL 0.54 (0.44/0.67) 0.53 (0.44/0.66) 0.64 (0.44/0.83) 
 Copeptin, pg/mL 7.25 (4.85/12.0) 7.19 (4.82/11.76) 8.29 (5.27/17.67) 
 Ct-pro-endothelin-1, pmol/L 57.3 (47.1/70.8) 57.1 (47.2/69.9) 62.3 (46.5/83.9) 
 MR-proANP, pg/mL 91 (61/143) 89 (61/138) 123 (70/212) 
 Nt-proBNP, pg/mL 200 (90/556) 191 (88/508) 501 (152/1288) 
Variable All, n = 1781 No event, n = 1644 Event, n = 137 
Age, years 63 (56/69) 63 (55/69) 64 (60/70) 
Male sex, n (%) 1381 (77.5) 1283 (78.0) 98 (71.5) 
BMI, kg/m2 27.4 (25.0/30.1) 27.4 (25.0/30.1) 27.6 (25.0/29.8) 
Hypertension, n (%) 1402 (78.7) 1293 (78.6) 109 (79.6) 

 
Diabetes mellitus, n (%) 
 Diet 74 (4.2) 70 (4.3) 4 (2.9) 
 Oral medication 176 (9.9) 158 (9.6) 18 (13.1) 
 Insulin dependent 155 (8.7) 135 (8.2) 20 (14.6) 

 
Smoking, n (%) 
 Never 647 (36.3) 599 (36.4) 48 (35.0) 
 Former smoking 829 (46.5) 765 (46.5) 64 (46.7) 
 Current smoking 305 (17.1) 280 (17.0) 25 (18.2) 
Positive family history, n (%) 685 (38.5) 638 (38.8) 47 (34.3) 
History of myocardial infarction, n (%) 713 (40.0) 642 (39.1) 71 (51.8) 
Beta-blocker, n (%) 1181 (66.3) 1094 (66.5) 87 (63.5) 
ACE-inhibitors, n (%) 949 (53.3) 864 (52.6) 85 (62.0) 
Statins, n (%) 970 (54.5) 896 (54.5) 74 (54.0) 
LDL/HDL ratio 2.50 (1.86/3.19) 2.48 (1.85/3.18) 2.69 (2.19/3.31) 
Multivessel disease, n (%) 1280 (71.8) 1165 (70.9) 115 (83.9) 

 
Biomarkers 
 Apo AI, g/L 1.46 (1.29/1.65) 1.46 (1.30/1.66) 1.39 (1.23/1.59) 
 Apo B100, g/L 0.98 (0.82/1.14) 0.97 (0.82/1.14) 1.05 (0.85/1.18) 
 Creatinine, mg/dL 0.95 (0.83/1.08) 0.94 (0.83/1.07) 0.97 (0.84/1.17) 
 Cystatin C, mg/L 0.82 (0.72/0.96) 0.81 (0.72/0.95) 0.86 (0.72/1.09) 
 C-reactive protein, mg/L 2.19 (0.97/5.59) 2.09 (0.94/5.29) 3.86 (1.36/8.46) 
 GDF-15, ng/L 1155 (872/1616) 1133 (866/1564) 1516 (1025/2174) 
 Neopterin, nmol/L 6.23 (5.14/7.81) 6.19 (5.12/7.74) 7.01 (5.40/8.99) 
 MR-proADM, pmol/dL 0.54 (0.44/0.67) 0.53 (0.44/0.66) 0.64 (0.44/0.83) 
 Copeptin, pg/mL 7.25 (4.85/12.0) 7.19 (4.82/11.76) 8.29 (5.27/17.67) 
 Ct-pro-endothelin-1, pmol/L 57.3 (47.1/70.8) 57.1 (47.2/69.9) 62.3 (46.5/83.9) 
 MR-proANP, pg/mL 91 (61/143) 89 (61/138) 123 (70/212) 
 Nt-proBNP, pg/mL 200 (90/556) 191 (88/508) 501 (152/1288) 

Continuous variables are presented as median (25th/75th percentile). For discrete variables absolute (relative) frequencies of patients are given.

Correlations

Overall, partial correlations (age- and sex-adjusted) between biomarkers were weak to moderate. The strongest association between renal function, reflected by cystatin C, and biomarkers was observed for MR-proADM (r = 0.60) and GDF-15 (r = 0.50) (see Supplementary material online, Table S2). Between the biomarkers, correlations with r > 0.5 are observed within the cluster of neurohumoral and vasoactive biomarkers MR-proADM, Ct-pro-endothelin-1, and the inflammatory marker GDF-15. In addition, a positive association is observed between the biomarkers MR-proANP and Nt-proBNP (r = 0.58). The absolute value of correlations between apolipoproteins and other biomarkers was <0.2 except for C-reactive protein and Apo AI (r = −0.25).

Single biomarkers and cardiovascular outcome

The Cox proportional hazards assumption was not violated. Analyses of biomarker concentrations revealed a positive association with cardiovascular events for most of the markers investigated (Table 2). Overall, the strongest association with cardiovascular outcome was observed for Nt-proBNP, GDF-15, MR-proADM, MR-proANP, Ct-pro-endothelin, and cystatin C. Results remained similar after adjustment for classical risk factors and after additional adjustment for cardiac medication (ACE-inhibitors, beta-blockers, and statin intake). In a subgroup of 1214 participants with information on left ventricular ejection fraction, the additional adjustment for this variable did not markedly change the results (data not shown).

Table 2

Hazard ratios for cardiovascular outcome per unit of biomarker and per standard deviation of biomarker concentrations (n = 1781)

Variable Hazard ratio per unit increase 95% Confidence interval Hazard ratio per standard deviation 95% Confidence interval P-value 
Renal function 
 Creatinine 1.79 0.83–3.85 1.16 0.95–1.40 0.36 
1.76 0.73–4.21 1.15 0.93–1.43 0.78 
 Cystatin C 4.28 1.75–10.50 1.43 1.15–1.78 <0.0001 
3.77 1.48–9.61 1.39 1.10–1.75 0.0006 

 
Inflammation 
 C-reactive protein 1.24 1.04–1.46 1.33 1.06–1.67 0.004 
1.22 1.03–1.46 1.32 1.04–1.67 0.012 
 GDF-15 2.81 1.73–4.55 1.63 1.29–2.04 <0.0001 
2.67 1.59–4.47 1.59 1.25–2.02 <0.0001 
 Neopterin 2.22 1.20–4.10 1.34 1.07–1.68 0.002 
2.08 1.12–3.87 1.31 1.04–1.64 0.009 

 
Cardiac remodelling and vascular function 
 Copeptin 1.37 1.01–1.85 1.27 1.01–1.61 0.035 
1.32 0.97–1.80 1.24 0.98–1.57 0.11 
 Ct-pro-endothelin 1.01 1.00–1.02 1.36 1.10–1.69 0.0005 
1.01 1.00–1.02 1.34 1.08–1.67 0.001 
 MR-proADM 3.83 1.79–8.20 1.67 1.25–2.24 <0.0001 
3.61 1.65–7.88 1.63 1.21–2.20 <0.0001 
 MR-proANP 1.90 1.29–2.79 1.50 1.17–1.91 <0.0001 
1.82 1.24–2.69 1.46 1.14–1.87 0.0001 
 Nt-proBNP 1.50 1.25–1.80 1.71 1.34–2.18 <0.0001 
1.46 1.20–1.77 1.65 1.28–2.13 <0.0001 

 
Lipid metabolism 
 Apo B100 1.09 0.70–1.70 1.03 0.88–1.21 1.0 
0.94 0.48–1.84 0.98 0.77–1.24 1.0 
 Apo A1 0.40 0.15–1.06 0.78 0.59–1.02 0.083 
0.51 0.17–1.54 0.83 0.62–1.13 1.0 
Variable Hazard ratio per unit increase 95% Confidence interval Hazard ratio per standard deviation 95% Confidence interval P-value 
Renal function 
 Creatinine 1.79 0.83–3.85 1.16 0.95–1.40 0.36 
1.76 0.73–4.21 1.15 0.93–1.43 0.78 
 Cystatin C 4.28 1.75–10.50 1.43 1.15–1.78 <0.0001 
3.77 1.48–9.61 1.39 1.10–1.75 0.0006 

 
Inflammation 
 C-reactive protein 1.24 1.04–1.46 1.33 1.06–1.67 0.004 
1.22 1.03–1.46 1.32 1.04–1.67 0.012 
 GDF-15 2.81 1.73–4.55 1.63 1.29–2.04 <0.0001 
2.67 1.59–4.47 1.59 1.25–2.02 <0.0001 
 Neopterin 2.22 1.20–4.10 1.34 1.07–1.68 0.002 
2.08 1.12–3.87 1.31 1.04–1.64 0.009 

 
Cardiac remodelling and vascular function 
 Copeptin 1.37 1.01–1.85 1.27 1.01–1.61 0.035 
1.32 0.97–1.80 1.24 0.98–1.57 0.11 
 Ct-pro-endothelin 1.01 1.00–1.02 1.36 1.10–1.69 0.0005 
1.01 1.00–1.02 1.34 1.08–1.67 0.001 
 MR-proADM 3.83 1.79–8.20 1.67 1.25–2.24 <0.0001 
3.61 1.65–7.88 1.63 1.21–2.20 <0.0001 
 MR-proANP 1.90 1.29–2.79 1.50 1.17–1.91 <0.0001 
1.82 1.24–2.69 1.46 1.14–1.87 0.0001 
 Nt-proBNP 1.50 1.25–1.80 1.71 1.34–2.18 <0.0001 
1.46 1.20–1.77 1.65 1.28–2.13 <0.0001 

 
Lipid metabolism 
 Apo B100 1.09 0.70–1.70 1.03 0.88–1.21 1.0 
0.94 0.48–1.84 0.98 0.77–1.24 1.0 
 Apo A1 0.40 0.15–1.06 0.78 0.59–1.02 0.083 
0.51 0.17–1.54 0.83 0.62–1.13 1.0 

Model 1 (upper row) is age- and sex-adjusted.

Model 2 (lower row) is additionally adjusted for BMI, LDL/HDL ratio, smoking, diabetes, hypertension, and number of diseased vessels.

MR-proANP, Copeptin, creatinine, C-reactive protein, GDF-15, and Nt-proBNP, MR-proADM, Cystatin C, and neopterin were logarithmically transformed. P-values and confidence intervals are Bonferroni corrected.

Risk information beyond classical risk factors

To further explore whether any of the biomarkers added to the predictive value of traditional risk factor screening, we assessed reclassification (Table 3). The strongest reclassification for single biomarkers with a NRI of over 10% was observed for GDF-15 (0.16), Nt-proBNP (0.15), MR-proANP (0.13), and cystatin C (0.12). All other biomarkers yielded a NRI of <0.1. An IDI with a significance of P < 0.05 was observed for MR-proADM (0.022), Nt-proBNP (0.022), GDF-15 (0.021), cystatin C (0.015), and MR-proANP (0.014). The strongest biomarkers, MR-proADM, GDF-15, Nt-proBNP, cystatin C, and MR-proANP entered further analyses.

Table 3

Net reclassification improvement and integrated discrimination improvement for all biomarkers

Biomarker NRI P-value Individuals with event
 
Individuals without event
 
IDI P-value R2 difference 
   Up Down Up Down    
Renal function 
 Creatinine 0.0084 1.0 78 49 0.0030 0.66 0.0016 
 Cystatin C 0.12 0.066 16 179 128 0.015 0.015 0.0082 

 
Inflammation 
 C-reactive protein 0.076 0.36 180 113 0.0063 0.18 0.0057 
 GDF-15 0.16 0.015 19 10 253 120 0.021 0.0004 0.014 
Neopterin 0.041 1.0 138 100 0.0087 0.24 0.0058 

 
Cardiac remodelling and vascular function 
 Copeptin 0.038 1.0 122 103 0.0052 0.25 0.0035 
 Ct-pro-endothelin 0.028 1.0 11 12 154 93 0.016 0.065 0.0074 
 MR-proANP 0.13 0.043 17 228 138 0.014 0.013 0.010 
 MR-proADM 0.087 0.78 15 11 223 138 0.022 0.001 0.012 
 Nt-proBNP 0.15 0.056 22 13 303 180 0.022 0.0009 0.017 

 
Lipid metabolism 
 Apo AI 0.048 0.064 105 69 0.0025 0.98 0.0017 
 Apo B100 −0.0006 1.0 −0.00001 1.0 0.00005 
Biomarker NRI P-value Individuals with event
 
Individuals without event
 
IDI P-value R2 difference 
   Up Down Up Down    
Renal function 
 Creatinine 0.0084 1.0 78 49 0.0030 0.66 0.0016 
 Cystatin C 0.12 0.066 16 179 128 0.015 0.015 0.0082 

 
Inflammation 
 C-reactive protein 0.076 0.36 180 113 0.0063 0.18 0.0057 
 GDF-15 0.16 0.015 19 10 253 120 0.021 0.0004 0.014 
Neopterin 0.041 1.0 138 100 0.0087 0.24 0.0058 

 
Cardiac remodelling and vascular function 
 Copeptin 0.038 1.0 122 103 0.0052 0.25 0.0035 
 Ct-pro-endothelin 0.028 1.0 11 12 154 93 0.016 0.065 0.0074 
 MR-proANP 0.13 0.043 17 228 138 0.014 0.013 0.010 
 MR-proADM 0.087 0.78 15 11 223 138 0.022 0.001 0.012 
 Nt-proBNP 0.15 0.056 22 13 303 180 0.022 0.0009 0.017 

 
Lipid metabolism 
 Apo AI 0.048 0.064 105 69 0.0025 0.98 0.0017 
 Apo B100 −0.0006 1.0 −0.00001 1.0 0.00005 

NRI stands for net reclassification improvement, IDI for integrated discrimination improvement. The basic model comprised age, sex, BMI, LDL/HDL ratio, smoking, diabetes, hypertension, and number of diseased vessels. P-values are Bonferroni corrected.

Comparison of single biomarkers and a multimarker panel

The combination of the top biomarkers (Nt-proBNP, GDF-15, MR-proANP, cystatin C, MR-proADM) reclassified 30 individuals with events into a higher risk category, 15 into a lower risk group. Of the event-free subgroup 169 patients were upwardly classified, 341 were reclassified in the correct direction. The NRI was 0.24, P = 0.0001 and the IDI 0.036, P < 0.0001. Correct reclassification was thus achieved in the majority of individuals with events as well as for down-classification of participants without event during follow-up.

We computed the C-index for risk prediction models based on estimated 3-year event-free survival. The information obtained through the individual biomarkers was assessed on top of the basic model (C-index basic model 0.656, 95% CI 0.608–0.705) (Figure 2). The C-index improved for all biomarkers and the combination of the top five biomarkers. Adding the information on all the top five biomarkers to the baseline model increased the C-statistic from 0.656 to 0.690 (95% CI 0.640–0.739). The effect estimate is comparable with the best single biomarkers on top of the classical risk factors, e.g. the addition of Nt-proBNP to the baseline model increased the C-index to 0.686 (95% CI 0.637–0.735). The C-index for the combination of all biomarkers was not significantly different from the C-index resulting from single biomarkers.

Figure 2

Hazard ratios and 95% confidence intervals for cut-off values for event-free survival determined from survival trees for single markers and for the strongest biomarkers combined. P-values derived from Cox regression analyses were all <0.0001. The C-index for continuous biomarker concentrations was calculated on top of the basic model comprising the classical risk factors (age, sex body mass index, LDL/HDL ratio, smoking, diabetes, hypertension, and number of diseased vessels), C-index 0.656. The coefficients used to derive the combined cut-off were log-transformed N-terminal-pro-B-type natriuretic peptide 0.2702, log-transformed growth-differentiation factor-15 0.4595, log-transformed mid-regional-pro-atrial natriuretic peptide 0.0119, log-transformed cystatin C 0.2527, and log-transformed mid-regional-pro-adrenomedullin 0.2310.

Figure 2

Hazard ratios and 95% confidence intervals for cut-off values for event-free survival determined from survival trees for single markers and for the strongest biomarkers combined. P-values derived from Cox regression analyses were all <0.0001. The C-index for continuous biomarker concentrations was calculated on top of the basic model comprising the classical risk factors (age, sex body mass index, LDL/HDL ratio, smoking, diabetes, hypertension, and number of diseased vessels), C-index 0.656. The coefficients used to derive the combined cut-off were log-transformed N-terminal-pro-B-type natriuretic peptide 0.2702, log-transformed growth-differentiation factor-15 0.4595, log-transformed mid-regional-pro-atrial natriuretic peptide 0.0119, log-transformed cystatin C 0.2527, and log-transformed mid-regional-pro-adrenomedullin 0.2310.

Cut-off analyses for risk stratification

To give an idea of the strength of the single biomarkers when used as dichotomized variables, we calculated cut-offs. Cut-off values for discrimination of high- and low-risk individuals were defined using survival trees. Results confirmed the single biomarkers Nt-proBNP, GDF-15, MR-proANP, cystatin C, and MR-proADM and the combination of the five biomarkers as strong risk indicators (Figure 2). Results were comparably strong when the top biomarkers were added to risk estimates derived from the ESC Score (see Supplementary material online, Table S3). Since HRs and cut-offs may vary by end-point we also present the final analyses separately for cardiovascular death (see Supplementary material online, Table S4). As the cut-offs are optimal for the used data sets, the P-values were expected to be small.

Discussion

We evaluated the prognostic value of 12 contemporary biomarkers reflecting atherosclerosis and heart failure associated pathways in a prospective cohort of stable angina patients. Our data indicate that the most informative biomarkers that offer incremental predictive ability over traditional risk factors were concentrations of circulating Nt-proBNP, GDF-15, MR-proANP, MR-proADM, and—to a slightly lesser extent in this data set—cystatin C. A multimarker approach combining these biomarkers was most strongly related to outcome, added incremental risk information, and showed the highest model stability and precision. However, the combined measurement did not improve risk stratification or reclassification compared with the strongest single biomarkers from a clinical perspective. Based on event-free survival probabilities, Nt-proBNP and GDF-15 in direct comparison proved to be the strongest single markers with substantial reclassification. In contrast to data derived in initially healthy individuals,5 the top biomarkers achieved correct reclassification in both, patients with future events in whom the majority was upwardly reclassified and in individuals without cardiovascular events during follow-up who were predominantly reclassified in lower risk categories. Results were similar when the ESC Score was used as the basic model for primary risk classification.

Our results on natriuretic peptides confirm findings for Nt-proBNP3,17,18 and demonstrate a comparable strength of association for MR-proANP. Similarly to the natriuretic peptides, GDF-15 is associated with outcome in heart failure and in patients with CAD.8,9 GDF-15 has been detected in high concentrations in autopsy samples after acute myocardial infarction.19 Along these lines, GDF-15 concentrations have been found to closely correlate with CAD severity and impaired systolic function in patients with acute coronary syndrome, suggesting that the circulating concentrations of this stress-inducible cytokine may integrate information on vascular and cardiac pathologies.20 Consistent with this hypothesis, GDF-15 concentrations were related to inflammatory markers and Nt-proBNP in AtheroGene. Notably, GDF-15 added prognostic information to these markers in patients with acute coronary syndrome and heart failure, as well as in the present patient cohort, indicating that this marker may provide insights into additional disease pathways.

Adrenomedullin is widely expressed, suggesting a crucial role in the control of central body functions such as endothelial regulation of vascular tone and blood pressure,21 modulation of cardiac output,22 protection against organ damage in hypoxia or sepsis, and fluid and electrolyte homoeostasis.23–25 Plasma concentrations of adrenomedullin are elevated in a number of cardiovascular pathologies including hypertension, diabetes, heart failure, or acute coronary syndrome.26–29 Whereas adrenomedullin concentrations are difficult to assess in vivo, the identification of the more stable MR-proADM with favourable analytic characteristics allows the application in epidemiological studies. Recent reports highlight the prognostic power of MR-proADM after acute myocardial infarction,26 and its relation to cardiovascular events in high-risk patients.30 The present data on MR-proADM have two major implications. First, we were able to demonstrate that MR-proADM is strongly associated with cardiovascular outcome in stable angina; and second, the inclusion of MR-proADM improves the model accuracy comparable with Nt-proBNP, one of the strongest predictors of outcome in secondary prevention.3 A single MR-proADM determination provides information for cardiovascular risk stratification incremental to that obtained from risk models simultaneously incorporating the classical risk factors, and similar accuracy of risk information obtained from the combination of several biomarkers of different pathophysiological backgrounds.

Our biomarker evaluation accounts for renal function by incorporation of cystatin C, which is well correlated with a glomerular filtration rate even in patients with only minor renal impairment.31 Kidney disease may affect both biomarker concentrations and cardiovascular outcome.32 All top biomarkers revealed relevant bivariate correlations with cystatin C. However, single-marker analysis and the multimarker model suggest that the top biomarkers, natriuretic peptides, GDF-15, and MR-proADM, offer predictive information beyond cystatin C.

Clinical application and decision-making is based on cut-off values. Therefore, we dichotomized the single biomarkers and their combination by identifying an optimal cut-off for the prediction of cardiovascular events. Interestingly, the statistically optimal cut-offs correspond comparatively well with recently observed thresholds of concentrations for the longer known biomarkers such as Nt-proBNP and cystatin C (Figure 2). The confirmation and refinement of cut-off values might help to identify patients at high risk for cardiovascular events in the clinical setting.

Further establishment of the pathophysiological role of MR-proADM and GDF-15 is necessary to improve our understanding of the underlying mechanisms that lead to the additional risk information provided by these novel markers and to assess their potential for clinical application.

Limitations

The number of events in this study is relatively small, which may limit our ability to detect small effects and to compare individual biomarkers head-to-head. In particular, the strength of association may differ by end-point (non-fatal myocardial infarction, cardiovascular death). But the number of outcomes in the two subgroups was too small to derive stable risk estimates in separate analyses and results have to be interpreted with caution. In confirmatory analyses, we observed similar effect sizes and directions as reported in the literature, which supports the reliability of our current results. To enhance comparability of the biomarkers, we only used individuals with complete biomarker data for analyses. The exclusion of participants with missing values may have introduced a bias. The maximum values of the C-indices for single markers and the combination of the top markers did not exceed 0.70, the range when model discrimination is considered satisfactory.33 This finding may, in part, be due to the underlying expected event rate in a stable CAD sample and the overall predictability of events. Of note, the basic model comprising classical cardiovascular risk factors only achieved a C-index of 0.656. In contrast to initially healthy samples, validated risk scores for patients with manifest disease and stable angina do not exist. Our efforts can be viewed as an attempt at stratifying patients by combining some of the known variables strongly related to outcome to provide the basis for the assessment of additional, novel biomarkers. We used risk categories derived from primary prevention settings. These may not be the most appropriate classifications in manifest CAD, and at this point, therapeutic implications cannot be directly derived from these categories, thus precluding cost-benefit analyses of the additional measurement of the novel biomarkers. Overall, our analyses are exploratory and need confirmation in independent samples.

Strengths of the study are clearly the homogeneous prospective cohort with almost complete long-term follow-up and the availability of a large panel of contemporary biomarkers some of which were, to our knowledge, evaluated in a stable angina cohort for the first time. We were able to demonstrate that among a panel of novel and established cardiovascular risk indicators, single biomarkers such as Nt-proBNP, GDF-15, MR-proANP, cystatin C, and MR-proADM provided information on cardiovascular outcome in stable angina with similar accuracy as the combination of the five strongest biomarkers combined on top of classical risk factors. In conclusion, we can state that the robust neurohormonal precursor Nt-proBNP remains one of the strongest and best examined biomarkers in cardiovascular risk prediction. None of the other promising novel markers proved to be clearly superior. From a clinical perspective, the measurement of one of the top biomarkers provided comparative information to the determination of all five markers and a multiple biomarkers approach combining the presented biomarkers does not seem to be favourable. Whether the additional measurement of any of the top biomarkers affects clinical decision-making and outcome has to be investigated in future studies.

Supplementary material

Supplementary material is available at European Heart Journal online.

Funding

The AtheroGene study is supported by a grant of the ‘Stiftung Rheinland-Pfalz für Innovation,’ Ministry for Science and Education (AZ 15202–386261/545), Mainz; R.S. is funded by a German Research Foundation Research Fellowship SCHN 1149/1-1. This work was further supported by research grants from the Brandenburg Ministry of Economics, Germany, and the European Regional Development Fund (EFRE/ERDF).

Conflict of interest: T.K. and K.C.W. have filed a patent and have a contract with Roche Diagnostics to develop a GDF-15 assay for cardiovascular applications. There is no conflict of interest present for any of the other co-authors.

Acknowledgements

We are indebted to Beatrix C. Mattes and Moira M. Pryde for their help with the preparation of the manuscript and Margot Neuser for her graphical work. The testkits for copeptin, Ct-pro-endothelin-1, MR-proADM, MR-proANP, and neopterin were provided by B.R.A.H.M.S, Hennigsdorf, and for Nt-proBNP by Roche Diagnostics, Mannheim at no cost. We thank Viviane Nicaud, MS for her statistical assistance.

References

1
Daly
CA
De Stavola
B
Sendon
JL
Tavazzi
L
Boersma
E
Clemens
F
Danchin
N
Delahaye
F
Gitt
A
Julian
D
Mulcahy
D
Ruzyllo
W
Thygesen
K
Verheugt
F
Fox
KM
Predicting prognosis in stable angina–results from the Euro heart survey of stable angina: prospective observational study
BMJ
 , 
2006
, vol. 
332
 (pg. 
262
-
267
)
2
Wang
TJ
Gona
P
Larson
MG
Tofler
GH
Levy
D
Newton-Cheh
C
Jacques
PF
Rifai
N
Selhub
J
Robins
SJ
Benjamin
EJ
D'Agostino
RB
Vasan
RS
Multiple biomarkers for the prediction of first major cardiovascular events and death
N Engl J Med
 , 
2006
, vol. 
355
 (pg. 
2631
-
2639
)
3
Blankenberg
S
McQueen
MJ
Smieja
M
Pogue
J
Balion
C
Lonn
E
Rupprecht
HJ
Bickel
C
Tiret
L
Cambien
F
Gerstein
H
Munzel
T
Yusuf
S
Comparative impact of multiple biomarkers and N-terminal pro-brain natriuretic peptide in the context of conventional risk factors for the prediction of recurrent cardiovascular events in the Heart Outcomes Prevention Evaluation (HOPE) Study
Circulation
 , 
2006
, vol. 
114
 (pg. 
201
-
208
)
4
Olsen
MH
Hansen
TW
Christensen
MK
Gustafsson
F
Rasmussen
S
Wachtell
K
Ibsen
H
Torp-Pedersen
C
Hildebrandt
PR
N-terminal pro-brain natriuretic peptide, but not high sensitivity C-reactive protein, improves cardiovascular risk prediction in the general population
Eur Heart J
 , 
2007
, vol. 
28
 (pg. 
1374
-
1381
)
5
Melander
O
Newton-Cheh
C
Almgren
P
Hedblad
B
Berglund
G
Engstrom
G
Persson
M
Smith
JG
Magnusson
M
Christensson
A
Struck
J
Morgenthaler
NG
Bergmann
A
Pencina
MJ
Wang
TJ
Novel and conventional biomarkers for prediction of incident cardiovascular events in the community
JAMA
 , 
2009
, vol. 
302
 (pg. 
49
-
57
)
6
Pencina
MJ
D'Agostino
RB
Sr
D'Agostino
RB
Jr
Vasan
RS
Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond
Stat Med
 , 
2008
, vol. 
27
 (pg. 
157
-
172
)
7
Wang
TJ
Larson
MG
Levy
D
Benjamin
EJ
Leip
EP
Omland
T
Wolf
PA
Vasan
RS
Plasma natriuretic peptide levels and the risk of cardiovascular events and death
N Engl J Med
 , 
2004
, vol. 
350
 (pg. 
655
-
663
)
8
Wollert
KC
Kempf
T
Peter
T
Olofsson
S
James
S
Johnston
N
Lindahl
B
Horn-Wichmann
R
Brabant
G
Simoons
ML
Armstrong
PW
Califf
RM
Drexler
H
Wallentin
L
Prognostic value of growth-differentiation factor-15 in patients with non-ST-elevation acute coronary syndrome
Circulation
 , 
2007
, vol. 
115
 (pg. 
962
-
971
)
9
Kempf
T
Sinning
JM
Quint
A
Bickel
C
Sinning
C
Wild
PS
Schnabel
RB
Lubos
E
Rupprecht
HJ
Munzel
T
Drexler
H
Blankenberg
S
Wollert
KC
Growth-differentiation factor-15 for risk stratification in patients with stable and unstable coronary heart disease: results from the AtheroGene study
Circ Cardiovasc Genet
 , 
2009
, vol. 
2
 (pg. 
286
-
292
)
10
Zethelius
B
Berglund
L
Sundstrom
J
Ingelsson
E
Basu
S
Larsson
A
Venge
P
Arnlov
J
Use of multiple biomarkers to improve the prediction of death from cardiovascular causes
N Engl J Med
 , 
2008
, vol. 
358
 (pg. 
2107
-
2116
)
11
Rupprecht
HJ
Blankenberg
S
Bickel
C
Rippin
G
Hafner
G
Prellwitz
W
Schlumberger
W
Meyer
J
Impact of viral and bacterial infectious burden on long-term prognosis in patients with coronary artery disease
Circulation
 , 
2001
, vol. 
104
 (pg. 
25
-
31
)
12
Zouridakis
E
Avanzas
P
Arroyo-Espliguero
R
Fredericks
S
Kaski
JC
Markers of inflammation and rapid coronary artery disease progression in patients with stable angina pectoris
Circulation
 , 
2004
, vol. 
110
 (pg. 
1747
-
1753
)
13
Conroy
RM
Pyorala
K
Fitzgerald
AP
Sans
S
Menotti
A
De
BG
De
BD
Ducimetiere
P
Jousilahti
P
Keil
U
Njolstad
I
Oganov
RG
Thomsen
T
Tunstall-Pedoe
H
Tverdal
A
Wedel
H
Whincup
P
Wilhelmsen
L
Graham
IM
Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project
Eur Heart J
 , 
2003
, vol. 
24
 (pg. 
987
-
1003
)
14
Harrell
FE
Jr
Califf
RM
Pryor
DB
Lee
KL
Rosati
RA
Evaluating the yield of medical tests
JAMA
 , 
1982
, vol. 
247
 (pg. 
2543
-
2546
)
15
Therneau
TM
Atkinson
B
rpart: Recursive Partitioning. R package version 3.1–42
R port by Brian Ripley
 , 
2008
16
R Development Core Team
R: A Language and Environment for Statistical Computing
R Foundation for Statistical Computing
 , 
2009
Vienna, Austria
17
Bibbins-Domingo
K
Gupta
R
Na
B
Wu
AH
Schiller
NB
Whooley
MA
N-terminal fragment of the prohormone brain-type natriuretic peptide (NT-proBNP), cardiovascular events, and mortality in patients with stable coronary heart disease
JAMA
 , 
2007
, vol. 
297
 (pg. 
169
-
176
)
18
Schnabel
R
Rupprecht
HJ
Lackner
KJ
Lubos
E
Bickel
C
Meyer
J
Munzel
T
Cambien
F
Tiret
L
Blankenberg
S
Analysis of N-terminal-pro-brain natriuretic peptide and C-reactive protein for risk stratification in stable and unstable coronary artery disease: results from the AtheroGene study
Eur Heart J
 , 
2005
, vol. 
26
 (pg. 
241
-
249
)
19
Kempf
T
Eden
M
Strelau
J
Naguib
M
Willenbockel
C
Tongers
J
Heineke
J
Kotlarz
D
Xu
J
Molkentin
JD
Niessen
HW
Drexler
H
Wollert
KC
The transforming growth factor-beta superfamily member growth-differentiation factor-15 protects the heart from ischemia/reperfusion injury
Circ Res
 , 
2006
, vol. 
98
 (pg. 
351
-
360
)
20
Lind
L
Wallentin
L
Kempf
T
Tapken
H
Quint
A
Lindahl
B
Olofsson
S
Venge
P
Larsson
A
Hulthe
J
Elmgren
A
Wollert
KC
Growth-differentiation factor-15 is an independent marker of cardiovascular dysfunction and disease in the elderly: results from the Prospective Investigation of the Vasculature in Uppsala Seniors (PIVUS) Study
Eur Heart J
 , 
2009
, vol. 
30
 (pg. 
2346
-
2353
)
21
Lainchbury
JG
Cooper
GJ
Coy
DH
Jiang
NY
Lewis
LK
Yandle
TG
Richards
AM
Nicholls
MG
Adrenomedullin: a hypotensive hormone in man
Clin Sci (Lond)
 , 
1997
, vol. 
92
 (pg. 
467
-
472
)
[PubMed]
22
Parkes
DG
May
CN
Direct cardiac and vascular actions of adrenomedullin in conscious sheep
Br J Pharmacol
 , 
1997
, vol. 
120
 (pg. 
1179
-
1185
)
23
Nishida
H
Sato
T
Miyazaki
M
Nakaya
H
Infarct size limitation by adrenomedullin: protein kinase A but not PI3-kinase is linked to mitochondrial KCa channels
Cardiovasc Res
 , 
2008
, vol. 
77
 (pg. 
398
-
405
)
[PubMed]
24
Dackor
R
Caron
K
Mice heterozygous for adrenomedullin exhibit a more extreme inflammatory response to endotoxin-induced septic shock
Peptides
 , 
2007
, vol. 
28
 (pg. 
2164
-
2170
)
25
Sandner
P
Hofbauer
KH
Tinel
H
Kurtz
A
Thiesson
HC
Ottosen
PD
Walter
S
Skott
O
Jensen
BL
Expression of adrenomedullin in hypoxic and ischemic rat kidneys and human kidneys with arterial stenosis
Am J Physiol Regul Integr Comp Physiol
 , 
2004
, vol. 
286
 (pg. 
R942
-
R951
)
[PubMed]
26
Khan
SQ
O'Brien
RJ
Struck
J
Quinn
P
Morgenthaler
N
Squire
I
Davies
J
Bergmann
A
Ng
LL
Prognostic value of midregional pro-adrenomedullin in patients with acute myocardial infarction: the LAMP (Leicester Acute Myocardial Infarction Peptide) study
J Am Coll Cardiol
 , 
2007
, vol. 
49
 (pg. 
1525
-
1532
)
27
Randa Abdel
KM
Grace
BD
Badawi
NE
Plasma levels of adrenomedullin and atrial natriuretic peptide in patients with congestive heart failure of various etiologies
Ital J Biochem
 , 
2007
, vol. 
56
 (pg. 
18
-
27
)
[PubMed]
28
Gegenhuber
A
Struck
J
Dieplinger
B
Poelz
W
Pacher
R
Morgenthaler
NG
Bergmann
A
Haltmayer
M
Mueller
T
Comparative evaluation of B-type natriuretic peptide, mid-regional pro-A-type natriuretic peptide, mid-regional pro-adrenomedullin, and Copeptin to predict 1-year mortality in patients with acute destabilized heart failure
J Card Fail
 , 
2007
, vol. 
13
 (pg. 
42
-
49
)
29
Kato
J
Kitamura
K
Uemura
T
Kuwasako
K
Kita
T
Kangawa
K
Eto
T
Plasma levels of adrenomedullin and atrial and brain natriuretic peptides in the general population: their relations to age and pulse pressure
Hypertens Res
 , 
2002
, vol. 
25
 (pg. 
887
-
892
)
30
Nishida
H
Horio
T
Suzuki
Y
Iwashima
Y
Kamide
K
Kangawa
K
Kawano
Y
Plasma adrenomedullin as an independent predictor of future cardiovascular events in high-risk patients: comparison with C-reactive protein and adiponectin
Peptides
 , 
2008
, vol. 
29
 (pg. 
599
-
605
)
31
Pucci
L
Triscornia
S
Lucchesi
D
Fotino
C
Pellegrini
G
Pardini
E
Miccoli
R
Del
PS
Penno
G
Cystatin C and estimates of renal function: searching for a better measure of kidney function in diabetic patients
Clin Chem
 , 
2007
, vol. 
53
 (pg. 
480
-
488
)
32
Go
AS
Chertow
GM
Fan
D
McCulloch
CE
Hsu
CY
Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization
N Engl J Med
 , 
2004
, vol. 
351
 (pg. 
1296
-
1305
)
33
Cook
NR
Use and misuse of the receiver operating characteristic curve in risk prediction
Circulation
 , 
2007
, vol. 
115
 (pg. 
928
-
935
)

Supplementary data

Comments

0 Comments