Elevated natriuretic peptides provide strong prognostic information in patients with heart failure (HF). The role of novel biomarkers in HF needs to be established. Our objective was to evaluate the prognostic power of novel biomarkers, incremental to the N-terminal portion of the natriuretic peptide (NT-proBNP) in chronic HF. Concentrations of circulating NT-proBNP, growth differentiation factor 15 (GDF-15), high-sensitivity C-reactive protein (hs-CRP), galectin-3 (Gal-3), and high-sensitivity troponin T (hs-TnT) were measured and related to all-cause long-term mortality. Of 209 patients (age 71 ± 10 years, 73% male patients, 97% New York Heart Association class III), 151 (72%) died during a median follow-up of 8.7 ± 1 year. The calculated area under the curve for NT-proBNP was 0.63, GDF-15 0.78, hs-CRP 0.66, Gal-3 0.68, and hs-TnT 0.68 (all p <0.01). Each marker was predictive for mortality in univariate analysis. In multivariate analysis, elevated concentrations of GDF-15 (hazard ratio [HR] 1.41, confidence interval [CI] 1.1 to 178, p = 0.005), hs-CRP (HR 1.38, CI 1.15 to 1.67, p = 0.001), and hs-TnT (HR 1.27, CI 1.06 to 1.53, p = 0.008) were independently related to mortality. All novel markers had an incremental value to NT-proBNP, using the integrated discrimination improvement. In conclusion, in chronic HF, GDF-15, hs-CRP, and hs-TnT are independent prognostic markers, incremental to NT-proBNP, in predicting long-term mortality. In this study, GDF-15 is the most predictive marker, even stronger than NT-proBNP.
Concentrations of natriuretic peptides (NP) are useful in the diagnosis and management of heart failure (HF), and provide powerful prognostic information in these patients, independent of left ventricular ejection fraction. NPs are produced by the myocardium as a reaction to an increase in myocardial wall stress. However, concentrations of circulating NP do not necessarily provide reliable information about the mechanism, etiology, and intensity of myocardial distress. Furthermore, a high intraindividual variation of NP has been described in patients with stable chronic HF. Hence, there is a need for additive biomarkers with respect to pathophysiology, treatment effect, and prognosis. Several novel markers, such as growth differentiation factor-15 (GDF-15), high-sensitivity C-reactive protein (hs-CRP), galectin-3 (Gal-3), and high-sensitivity troponin T (hs-TnT) are being tested and introduced for their clinical use in chronic HF. However, the added value of these markers is still under debate, and long-term data are lacking. Therefore, we analyzed the power of these markers head to head, compared with and added to N-terminal pro-brain-type natriuretic peptide (NT-proBNP), with respect to all-cause mortality during long-term follow-up in a population with advanced chronic HF.
Methods
The present study was conducted as a substudy from the Deventer-Alkmaar Heart Failure study (DEAL-HF), which has been described elsewhere. In brief, 240 patients with typical signs and symptoms of chronic HF combined with findings of a reduced left ventricular ejection fraction (45%) or diastolic dysfunction, according to the 2001 guidelines for the diagnosis of HF of the European Society of Cardiology, were included. Main exclusion criteria were an expected survival of <1 year, kidney function replacement therapy, and planned hospitalization.
In the present multimarker study, a complete set of data was available of 209 patients at baseline, due to missing blood samples (n = 28) and loss to follow-up (n = 3). The study was approved by the local medical ethics committees and complied with the Declaration of Helsinki. All patients gave written informed consent.
Routine laboratory measurements and blood samples for biomarker analysis were obtained at baseline. Baseline was defined as the moment of signing the informed consent, with patients preferably in a stable condition. Blood samples were taken on the same day, just after signing the informed consent. Patients could be included before discharge after hospitalization for HF (31%) or at the out patient clinic (69%). EDTA plasma was separated and stored at −70°C. Renal function, assessed using the estimated glomerular filtration rate (ml/min/1.73m 2 ), was calculated using the Modification of Diet in Renal Disease equation.
Circulating concentrations of NT-proBNP, GDF-15, hs-CRP, Gal-3, and hs-TnT were analyzed according to the description of the manufacturer (NT-proBNP, Roche Diagnostics, Rotkreutz, Switzerland; GDF-15, Roche Diagnostics; hs-CRP, Roche Diagnostics; Gal-3, BG Medicine, Waltham, Massachusetts; hs-TnT, Roche Diagnostics).
GDF-15 was analyzed by electrochemiluminescence immunoassay (precommercial assay). The interassay coefficient of variation was 2.3% at 1,100 ng/l and 1.8% at 17,200 ng/L with a lower limit of detection level of <10 ng/l. The reference value for GDF-15 was 1,109 ng/l (97.5th percentile). Quality control data were acquired with spiked plasma.
Gal-3 concentrations were measured by enzyme-linked immunosorbent assay; the lower limit of detection was 1.13 ng/ml. The 90th, 95th, and 97.5th percentile of the normal reference interval were 17.6, 20.3, and 22.1 ng/ml, respectively. Imprecision studies demonstrated that the total coefficient of variation was <10% at a low concentration of 6 ng/ml, 7% near the midlevel concentration of 21 ng/ml, and 15% at the high level of 70 ng/ml.
High-sensitivity CRP and NT-proBNP concentrations were measured by a electrochemiluminescence immunoassay using an Elecsys (Roche Diagnostics). The lower limit of detection level for hs-TnT was ≤3 pg/ml, for hs-CRP ≤0.15 mg/L, and for NT-proBNP 20 pg/ml. The coefficient of variation for hs-TnT was ≤5% in the range 25 to ≤100 pg/ml; for hs-CRP, ≤5% for values >1.0 mg/L; and for NT-proBNP, 5% for values >100 pg/ml.
The end point of the current study was all-cause mortality. Patients were followed up to 10 years after randomization at the outdoor patient clinic. In case of no show, information regarding survival was obtained from the hospital system, relatives, or general practitioner. Baseline medication was up-titrated according to the 2001 guidelines.
Data are expressed as mean ± SD when normally distributed, as median with interquartile range when distribution was skewed and as frequencies and percentages. The intergroup differences were tested using Student t test, Mann-Whitney U test, or Pearson chi-square test when appropriate. For further analyses, logarithmic transformation was performed to achieve a normal distribution for skewed variables. To assess the ability of NT-proBNP, GDF-15, hs-CRP, Gal-3, and hs-TnT in predicting all-cause mortality, areas under the curve (AUCs) of receiver operating characteristics (ROC) curves were constructed. The statistical significance of differences in AUCs was estimated using the approach by DeLong et al. Optimal cut-off points were calculated using ROC curves. Log-rank tests were used to assess the significance between the concentrations of GDF-15, hs-CRP, Gal-3, and hs-TnT, above and below the optimal cut-off point.
Unadjusted hazard ratios (HR) of log-transformed biomarkers were calculated for univariate Cox regression analyses (depicted as per SD increase). In consecutive multivariate models, GDF-15, hs-CRP, Gal-3, and hs-TnT were adjusted for age and gender, renal function, HF etiology, NT-proBNP concentrations, and all other univariate significant biomarkers. Finally, the net reclassification improvement (NRI) and the integrated discrimination improvement (IDI), described by Pencina et al, were also calculated. The aim of the NRI was to examine the prognostic discrimination of integrated discrimination (ID) on top of clinical risk factors on the primary end point. Clinical risk factors included age, gender, renal function, HF etiology, and NT-proBNP concentrations. We used risk categories of <25%, 25% to 50%, and >50%.
All tests were 2-sided, and a p value <0.05 was considered statistically significant. All statistical analyses were performed using STATA version 11.0 (StataCorp LP, College Station, Texas) and SPSS version 18.0 (SPSS, Chicago, Illinois).
Results
Characteristics of the study population are described in Table 1 . Medical care was provided according to the guidelines of the European Society of Cardiology prevailing at the time of inclusion and execution of the study with optimal application of therapy (baseline medication, Table 1 ). At baseline, beta blockers were prescribed in 60% of the patients. This figure went up to 69% after 1 year of follow-up. Almost all patients used a blocker of the renin-angiotensin system (96%) and diuretic therapy (97%) at baseline. Nonsurvivors were older, more often male, more frequently had ischemic etiology and diabetes mellitus, and had lower sodium and estimated glomerular filtration rate.
| Variable | Total (n = 209) | Survivors (n = 58) | Nonsurvivors (n = 151) | p Value |
|---|---|---|---|---|
| Age (yrs) | 71 ± 10 | 67 ± 11 | 72 ± 9 | <0.001 |
| Men | 73% | 62% | 77% | 0.02 |
| HF etiology, ischemic | 66% | 50% | 72% | 0.003 |
| Body mass index (kg/m 2 ) | 26 ± 5 | 27 ± 5 | 26 ± 5 | 0.105 |
| NYHA class, III/IV | 97%/3% | 98%/2% | 96%/4% | 0.418 |
| Diabetes mellitus | 27% | 17% | 32% | 0.028 |
| Chronic obstructive pulmonary disease | 27% | 29% | 26% | 0.405 |
| Hypercholesterolemia | 48% | 50% | 48% | 0.119 |
| Sodium (mmol/L) | 138 ± 3 | 139 ± 3 | 138 ± 3 | 0.026 |
| Hemoglobin (mmol/L) | 8.4 ± 1.0 | 8.5 ± 0.8 | 8.3 ± 1.0 | 0.242 |
| Estimated glomerular filtration rate (ml/min/1.73m 2 ) | 52 ± 14 | 55 ± 13 | 50 ± 15 | 0.038 |
| NT-proBNP (pg/ml) ∗ | 1,771 (1,008–4,483) | 1,729 (720–3,144) | 2,373 (1,093–5,813) | 0.004 |
| GDF-15 (ng/L) | 1,606 (1,087–2,412) | 1,082 (802–1,502) | 1,900 (1,357–2,671) | <0.001 |
| hs-CRP (mg/dl) | 4.7 (1.9–11.2) | 3.1 (1.4–5.4) | 5.8 (2.0–14.9) | 0.003 |
| Gal-3 (ng/ml) | 17.6 (13.3–21.4) | 15.0 (11.8–18.0) | 18.5 (13.9–22.3) | <0.001 |
| hs-TnT (ng/ml) | 25.6 (17.2–39.3) | 18.8 (10.4–30.8) | 28.2 (18.9–1,900) | <0.001 |
| Use of angiotensin converting enzyme inhibitors | 86% | 90% | 84% | 0.306 |
| Use of angiotensin receptor blockers | 11% | 8% | 11% | 0.578 |
| Use of beta blockers | 64% | 67% | 62% | 0.502 |
| Use of diuretics | 97% | 96% | 97% | 0.757 |
| Use of statins | 41% | 52% | 36% | 0.044 |
| Use of spironolactone | 31% | 24% | 34% | 0.187 |
The median follow-up for survivors was 8.4 (interquartile range 7.8 to 9.8) years. In total, 151 (72%) patients died. Nonsurvivors had significantly higher baseline concentrations of all biomarkers. For each individual marker, the AUC was plotted against the AUC of NT-proBNP ( Figure 1 ). Only GDF-15 was significantly better than NT-proBNP in predicting mortality ( Table 2 ; p <0.001). Figure 2 depicting Kaplan-Meier survival curves show that all biomarkers significantly predicted outcome (all p <0.001).
| Novel Marker vs NT-proBNP | p Value | AUC for Combination of Markers | Incremental Power of Novel Marker | p Value |
|---|---|---|---|---|
| GDF-15 vs NT-proBNP | <0.001 | GDF-15 + NT-proBNP = 0.78 | NT-proBNP vs GDF-15 + NT proBNP | <0.001 |
| hs-CRP vs NT-proBNP | 0.488 | hs-CRP + NT-proBNP = 0.68 | NT-proBNP vs hs-CRP + NT-proBNP | 0.068 |
| Gal-3 vs NT-proBNP | 0.279 | Gal-3 + NT-proBNP = 0.69 | NT-proBNP vs Gal-3 + NT-proBNP | 0.039 |
| hs-TnT vs NT-proBNP | 0.235 | hs-TnT + NT-proBNP = 0.68 | NT-proBNP vs hs-TnT + NT-proBNP | 0.067 |
The power of each novel marker incremental to NT-proBNP was calculated ( Table 2 ). GDF-15 and Gal-3 showed to be of significant additive value when combined with NT-proBNP ( Table 2 ; p <0.001 and p = 0.039, respectively) whereas the other markers only showed marginal changes.
Univariate and multivariate ( Table 3 ) Cox-proportional hazard regression models were conducted for each variable. GDF-15 (p = 0.005), hs-CRP (p = 0.001), and hs-TnT (p = 0.008) were significant independent predictors for all-cause mortality, whereas Gal-3 was not (p = 0.638).
| Variable | GDF-15 (per SD) | Z | p Value | hs-CRP (per SD) | Z | p Value | Galectin-3 (per SD) | Z | p Value | hs-TnT (per SD) | Z | p Value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | HR (95% CI) | HR (95% CI) | HR (95% CI) | |||||||||
| Univariate | 1.69 (1.43–1.98) | 6.31 | <0.001 | 1.55 (1.31–1.83) | 5.15 | <0.001 | 1.41 (1.20–1.64) | 4.30 | <0.001 | 1.53 (1.32–1.78) | 5.65 | <0.001 |
| Model 1 | 1.56 (1.31–1.86) | 4.94 | <0.001 | 1.47 (1.26–1.73) | 4.70 | <0.001 | 1.30 (1.10–1.53) | 3.11 | 0.002 | 1.45 (1.24–1.71) | 4.54 | <0.001 |
| Model 2 | 1.45 (1.18–1.79) | 3.45 | 0.001 | 1.42 (1.19–1.70) | 3.93 | <0.001 | 1.27 (1.04–1.57) | 2.35 | 0.019 | 1.36 (1.14–1.62) | 3.39 | 0.001 |
| Model 3 | 1.41 (1.11–1.78) | 2.81 | 0.005 | 1.38 (1.15–1.67) | 3.41 | 0.001 | 1.06 (0.84–1.32) | 0.47 | 0.638 | 1.27 (1.06–1.53) | 2.63 | 0.008 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
