Introduction
Acute decompensated heart failure (ADHF) is the leading cause of hospitalization in adults over 65 years [
1]. Despite medical progress, ADHF is still the most costly cardiovascular disorder in Western countries and is associated with a very poor prognosis [
1‐
3].
Early prediction of a patient's clinical course is pivotal for selecting appropriate management strategies for patients with ADHF. However, risk stratification in these patients is still difficult. The tools used for the evaluation of disease severity and prognosis in the past have been criticized because epidemiological and clinical factors like age, New York Heart Association (NYHA) functional class, or Killip class were shown to be inadequately sensitive [
4]. Left ventricular ejection fraction (LVEF) determined by echocardiography was once considered a reliable surrogate prognostic marker [
5]. Recent reports, however, have demonstrated that about 50% of patients admitted with ADHF have a preserved LVEF [
6].
B-type natriuretic peptide (BNP) and N-terminal pro B-type natriuretic peptide (NT-proBNP) are quantitative markers of cardiac wall stress [
7,
8]. Both natriuretic peptides (NPs) have been shown to accurately mirror heart failure (HF) severity and to correlate well with NYHA classification [
9,
10]. BNP and NT-proBNP are cleaved in equimolar amounts from proBNP; thus, NP levels correlate with each other [
11]. Despite the considerable similarities between the two NPs, their different half-lives and different modes of degradation argue for a separate analysis and make a direct comparison indispensable.
In patients with HF, serial evaluations of BNP and NT-proBNP levels may be useful for guiding therapy decisions by indicating the need for treatment intensification [
12‐
18]. It is, however, unknown whether BNP and NT-proBNP differ in their utility to risk-stratify patients with ADHF. Also, little is known regarding the earliest time point for reliable assessment of treatment efficacy and prognosis. Therefore, the objectives of this study were (a) to define BNP and NT-proBNP plasma concentration profiles from admission to discharge in order to establish the more appropriate timing for these measurements, (b) to assess the role of BNP and NT-proBNP sequential measurement as a marker of clinical improvement of patients with ADHF in response to therapy, and (c) to compare the prognostic utility of BNP and NT-proBNP in this setting.
Discussion
In this study, we determined the prognostic value of serial BNP and NT-proBNP measurements and their accuracy to predict 1-year all-cause mortality, 30-day all-cause mortality, and 1-year HF hospitalization in patients presenting with ADHF. We report five major findings: First, BNP and NT-proBNP levels in 1-year as well as in 30-day non-survivors were higher at presentation and remain higher during the entire course of hospitalization. Second, in 1-year and 30-day survivors, BNP and NT-proBNP levels gradually decreased during the course of hospitalization, whereas in non-survivors, BNP and NT-proBNP levels demonstrated no significant change. Thereby, BNP levels decreased more rapidly than NT-proBNP between presentation and 24 hours. Accordingly, the accuracy of BNP and NT-proBNP to predict 1-year and 30-day mortality increased during the course of hospitalization. Third, at 24 hours, 48 hours, and discharge, BNP levels independently predicted 1-year and 30-day mortality in multivariate analysis whereas only pre-discharge NT-proBNP levels independently predict 1-year mortality. Fourth, neither BNP nor NT-proBNP at any determined point in time could reliably predict 1-year HF hospitalizations. Fifth, the accuracy of BNP to predict 1-year mortality by ROC analysis at 24 hours was comparable to values already obtained at 48 hours or at hospital discharge. This observation suggests that measurement of BNP at 24 hours may be suitable for early assessment of prognosis and consecutive intensification or change of treatment in those patients with continuously elevated levels. These findings are of major clinical importance.
According to other studies, NP levels were higher in patients who died or experienced cardiovascular events. In patients with a favorable outcome, NPs decreased during the course of hospitalization, presumably as a positive response to HF therapy [
15,
24,
28,
30]. This decline in NPs was delayed in comparison with improvement of clinical symptoms and hemodynamic parameters and usually was first observed at 24 hours after admission [
24‐
26,
28]. In patients with adverse outcome, NP levels remained elevated despite medical therapy, providing valuable prognostic information [
15,
24,
28,
30]. Most studies claimed that the best time to predict outcome by measurement of NP was prior to hospital discharge [
24,
27,
28]. Logeart and colleagues [
28] examined the prognostic value of serial BNP measurements in patients with ADHF and found elevated pre-discharge BNP levels to be the strongest independent predictor of death or readmission for HF. Comparable results were demonstrated by Cohen-Solal and colleagues [
31] in a large trial of ICU patients admitted with ADHF. In the latter study, a BNP decrease of greater than 30% between admission and day 5 independently predicted survival. In our study, we could confirm these results for 1-year survival for a BNP decrease of greater than 30% between admission and discharge (HR 0.42 [0.23 to 0.65],
P = 0.004) but not for NT-proBNP. Also, an NP decrease of greater than 30% between admission and 24 hours or between admission and 48 hours was not predictive for BNP or for NT-proBNP in our study. O'Brien and colleagues [
27] examined the prognostic value of admission and pre-discharge levels of NT-proBNP in patients presenting with ADHF. The main finding of this study was that only pre-discharge NT-proBNP levels independently predicted outcome, and this is consistent with our results.
Recently, Di Somma and colleagues [
32] could demonstrate that ADHF patients with a discharge BNP level of less than 300 pg/mL and a percentage decrease during hospitalization of greater than 46% had a better outcome compared with patients with a discharge BNP level of greater than 300 pg/mL or a percentage decrease of less than 46% or both. In this study, ROC curves for percentage decrease of BNP levels at 24 hours after hospitalization reliably predicted adverse events (
P < 0.001), corroborating the results of our study. The area under the curve (AUC) for percentage decrease at discharge in their study was, however, higher compared with percentage decrease at 24 hours.
The clinical value of outcome measurements performed during a late stage of hospitalization or prior to hospital discharge has limitations. Important decisions regarding patient management and treatment strategy, including consultation by a cardiologist, ICU admission, and non-invasive ventilation, must be taken into account at an early stage of hospitalization. A reliable risk stratification parameter that is available earlier could help to mitigate the dismal outcome of patients with ADHF by treatment intensification. Our data suggest that the 1-year prognostic accuracy of BNP levels measured at 24 hours is comparable to levels obtained prior to discharge, which are widely accepted to be excellent [
24,
27]. Thus, the most significant change in BNP levels affecting 1-year prognostic value seems to occur during the first 24 hours, reflecting a satisfactory response to HF therapy. Simultaneously, owing to their delayed kinetic, NT-proBNP levels in survivors decline more slowly than BNP levels during the first 24 hours. This finding is supported by other studies [
27,
30]. Di Somma and colleagues [
33] has demonstrated a more rapid decline of BNP compared with NT-proBNP in response to therapy in ADHF patients. Bayés-Genís and colleagues [
30] examined the prognostic value of the percentage decrease of NT-proBNP during the course of hospitalization in patients with ADHF. In that study, no significant change in NT-proBNP levels during the first 24 hours was observed, confirming the delayed kinetic of NT-proBNP during early hospitalization. This finding is supported by a study by Metra and colleagues [
24], who determined serial measurements of NT-proBNP at 6, 12, 24, and 48 hours and at discharge in consecutive patients with ADHF. The earliest significant decline of NT-proBNP levels was observed at 48 hours, followed by stable NT-proBNP levels during the remaining hospitalization. Di Somma and colleagues [
34] found a decrease of NT-proBNP of 18.8% during the first 24 hours in a comparable setting.
Several small studies have compared the diagnostic accuracy of BNP and NT-proBNP [
33,
35‐
37]. Unfortunately, it is unknown whether BNP and NT-proBNP differ in their utility to risk-stratify patients with ADHF. In our study, no significant difference between the areas under the ROC at the different measured time points was identified between BNP and NT-proBNP. However, at 24 hours, only BNP levels independently predicted mortality by multivariate analysis, suggesting a more sensitive response in patients with a favorable outcome at this early time point. Whether early risk stratification that is based on persistently elevated BNP levels and that is followed by treatment intensification has the potential to improve patient outcome needs to be confirmed in larger prospective trials.
The mean decreases of BNP levels between presentation and 24 hours in 1-year survivors of our study were 34% for BNP and 27% for NT-proBNP levels. We believe that this rapid change in BNP levels, reflecting an adequate response to HF therapy, is a very important, early risk stratification and therapy guidance tool. A lack of this response, given optimal medical treatment, implies a more complex and therapy-refractory disease associated with an adverse long-term outcome. Accordingly, if this change does not occur, treatment intensification should be the consequence. In patients with a comparable decrease in BNP levels (roughly 30% between admission and 24 hours), we would expect a favorable outcome; however, future prospective studies have to evaluate a distinct cut-point to allow a more precise recommendation.
Interestingly, in our study, neither BNP nor NT-proBNP at any determined time point was able to reliably predict 1-year readmission for HF. Previously published studies presuming this finding - including those of Cheng and colleagues [
38], who used BNP, or Bettencourt and colleagues [
15], who used NT-proBNP - used combined endpoints consisting of all-cause mortality and readmission for HF.
There were some notable differences beyond NPs between 1-year survivors and non-survivors in our study, including lower BMI and eGFR levels and higher cTn and ASAT levels in non-survivors. More 1-year survivors were treated with beta-blocker and ARB, diuretics, or aspirin.
Obese HF patients have a better outcome compared with patients with low BMI [
39,
40]. The exact mechanism of this survival benefit linked to higher BMI is yet unknown. Suitable explanations for this paradoxical finding may include an increased neurohumoral and cytokine activation found in patients with advanced HF, leading to higher levels of tumor necrosis factor (TNF) and other inflammatory cytokines [
41,
42]. TNF and inflammatory cytokines may contribute to myocardial damage and thus to a higher mortality [
41,
42]. Adipose tissue was demonstrated to produce soluble TNF receptors, which might counteract the harmful property of TNF-α on the myocardium cells [
43].
Renal dysfunction is a strong and independent predictor of prognosis in the general population as well as in patients with ADHF [
44]. The underlying pathophysiology is multifactorial and most probably associated with decreased renal perfusion, atherosclerosis, inflammation, endothelial dysfunction, neurohormonal activation, and in particular venous congestion [
45,
46].
cTn levels are known to be elevated in a considerable proportion of patients with ADHF (6% to 10% using standard and 92% using high-sensitivity assays) independently of concomitant acute coronary syndrome [
47,
48]. The mechanisms underlying cTn release in ADHF remain speculative and include subendocardial ischemia leading to myocyte necrosis, cardiomyocyte damage from inflammatory cytokines or oxidative stress, hibernating myocardium, or apoptosis [
49]. cTn has excellent predictive properties in patients with ADHF [
48,
50,
51]; thus, not surprisingly, in our cohort, elevated cTn levels are highly predictive for adverse outcome. At 24 hours, cTn levels and age were even better prognosticators of 1-year mortality compared with BNP.
Liver function test abnormalities are common in patients with HF and independently predict adverse outcome [
52‐
54]. In a
post hoc analysis of the CHARM (Candesartan in Heart Failure: Assessment of Reduction in Mortality) study, Allen and colleagues [
54] demonstrated that elevated total bilirubin was the strongest liver function test predictor of cardiovascular death or HF hospitalizations. In our study, ASAT levels were higher in 1-year non-survivors whereas no difference in total bilirubin or albumin could be observed between 1-year survivors and non-survivors. Since patients with cardiogenic shock were not included in our study, passive hepatic congestion due to increased central venous pressure remains the most suitable explanation for this finding.
There was also a notable difference in 'life-saving' discharge medication between 1-year survivors and non-survivors in our study. One-year survivors received more beta-blocker and ARB compared with non-survivors, whereas treatment with ACE (angiotensin-converting enzyme) inhibitors was comparable. Treatment with beta-blocker and ARB is known to improve outcome in patients with HF; accordingly, our results are consistent with these findings [
55,
56].
Limitations
Some limitations of our study need to be mentioned. First, our study may have been too small to reach statistical significance for the comparison between BNP and NT-proBNP for 30-day all-cause mortality at all of the different measurement time points. Second, as we recruited consecutive patients, there may be some interindividual heterogeneity regarding doses of nitrates and diuretics applied as treatments were individualized for each patient. Third, as with all observational studies, we can only hypothesize that patient management could be improved by the clinical use of this monitoring tool.
Competing interests
CM has received research support from the Swiss National Science Foundation (PP00B-102853), the Swiss Heart Foundation, the Novartis Foundation, the Krokus Foundation, Abbott (Abbott Park, IL, USA), AstraZeneca (London, UK), Biosite (San Diego, CA, USA), Brahms (Hennigsdorf, Germany), Roche (Basel, Switzerland), Siemens (Munich, Germany), and the Department of Internal Medicine of University Hospital Basel as well as speaker honoraria from Abbott, Biosite, Brahms, Roche, and Siemens. All other authors declare that they have no competing interests.
Authors' contributions
MN made substantial contributions to the conception and design of the study, acquisition of data, analysis and interpretation of data, and drafting of the manuscript. TB contributed to the acquisition of data, the conception and design of the study, and critical revision of the manuscript. MP, TR, RT, HU, TS, NA, MR, JM, CH, and SS contributed to the acquisition of data and the critical revision of the manuscript. CM made substantial contributions to the conception and design of the study, analysis and interpretation of data, and drafting and critical revision of the manuscript. All authors read and approved the final manuscript.