Renal Dysfunction and Accuracy of N-Terminal Pro-B-Type Natriuretic Peptide in Predicting Mortality for Hospitalized Patients With Heart Failure

Abstract

Background: Renal dysfunction may confound the clinical interpretation of N-terminal pro-B-type natriuretic peptide (NT-proBNP) concentration. This study investigated whether renal dysfunction influences the prognostic accuracy of NT-proBNP in acute decompensated heart failure (ADHF).

Methods and Results: We studied 908 ADHF patients. The primary outcome was 12-month mortality. Interaction between estimated glomerular filtration rate (eGFR) and NT-proBNP in predicting mortality was tested with the likelihood ratio test. The patients were classified into 3 eGFR strata: ≥60, 30–59, and <30 ml·min^–1·1.73 m^–2. Cox models were used to calculate the adjusted hazard ratios (HR) for NT-proBNP, modeled as a dichotomous or categorized variable, within each level of eGFR. NT-proBNP was categorized using optimal cut-offs defined in ROC analysis for each eGFR level. A total of 234 patients (25.8%) died. Testing for interaction was not significant (χ²=0.29; P=0.5928). The adjusted HR for NT-proBNP >5,180 pg/ml was 2.09 (P<0.001) in the highest, 1.7 (P<0.001) in the intermediate, and 3.33 (P=0.010) in the lowest eGFR level. The adjusted HR for NT-proBNP above the optimal cut-offs defined on ROC analysis were 1.5 (P=0.239), 2.2 (P<0.001), and 3.24 (P=0.002), respectively. The models incorporating NT-proBNP as a dichotomous or categorized variable had equivalent C-statistics.

Conclusions: There was no evidence of interaction between eGFR and NT-proBNP in predicting mortality. The NT-proBNP cut-off of 5,180 ng/L provided independent prognostic information, irrespective of the level of residual renal function. (Circ J 2014; 78: 2439–2446)

Natriuretic peptides (NP) are valuable risk markers in a variety of clinical settings, including elevated risk factors, heart failure (HF), acute coronary syndromes, stable coronary artery disease, intensive care medicine, pulmonary embolism, and cardiac and vascular surgery.^1–7 In the setting of HF, NP testing has been widely adopted in order to improve diagnostic accuracy in dyspneic patients presenting to an emergency department; quantify the severity of illness; assess prognosis; and monitor disease progression, even though it is still under active investigation.⁸ NP testing is recommended in current disease-specific guidelines. Given that several factors may influence NP concentration,² potentially complicating their clinical interpretation in HF,^2,5,9,10 studies on the influence of these factors on the diagnostic and prognostic accuracy of NP have been conducted. In the Tsutamoto et al study, age, gender, body mass index (BMI), and atrial fibrillation were not independent predictors of brain natriuretic peptide (BNP) level when left ventricular ejection fraction (LVEF) and renal function were concomitantly evaluated.¹¹ Other studies provided evidence that NP remain prognostic independently of BMI and LVEF.^9,12,13

Renal dysfunction is highly prevalent in acute HF and strongly impacts on prognosis.^14–16 In patients with impaired renal function, NP concentration is markedly raised irrespective of whether HF is present or not,^17,18 suggesting that the N-terminal pro-B-type natriuretic peptide (NT-proBNP) prognostic cut-off may need to be adapted to the level of residual renal function to maximize prognostic accuracy in HF.^19,20 In the most recent Guidelines for the Management of Heart Failure,²¹ measurement of NP is recommended for “establishing prognosis or disease severity in acutely decompensated heart failure”. Clinicians, however, are still advised to be aware that impaired renal function is associated with elevated plasma levels of both BNP and NT-proBNP.²¹ The question of whether the NT-proBNP cut-off(s) needs to be adapted to the level of renal function to maximize prognostic accuracy has been incompletely addressed.

In the present study, we investigated the effect of renal function on the prognostic accuracy of NT-proBNP in hospitalized patients with acute decompensated HF (ADHF).

Methods

Nine hundred and twenty-nine patients hospitalized at the Cardiology Division of S. Maugeri Foundation (Bari, Italy), the heart failure unit of the Cardiovascular Department, Niguarda Cà Granda Hospital (Milan, Italy), and the Cardiology Division of S. Maugeri Foundation (Tradate, Varese, Italy) with worsening of chronic, established HF from January 2006 to December 2012 were retrospectively identified using a computer-generated list obtained from the administrative databases and by reviewing electronic and paper medical records. Patients with acute coronary syndrome or angina pectoris, hypertrophic cardiomyopathy, congenital heart disease, valvular heart disease, isolated right ventricular HF, or recent cardiac surgical or percutaneous procedures were excluded.²² Clinical and laboratory data were collected at admission. LVEF was assessed on 2-D echocardiography.²² The primary outcome was 12-month all-cause mortality. Death was ascertained by linking with the regional Health Information Systems or by telephone follow-up. The study was approved by the local Institutional Review Board.

Statistical Analysis

Data are reported as mean±SD for continuous variables or number (percentage) for categorical variables. Baseline data were 99.5% complete. Missing data included BMI (6.6%), total cholesterol (5.2%), and diastolic blood pressure at admission (1.8%). Baseline variables were compared using analysis of variance or chi-squared test. Multivariate linear regression analysis was used to confirm the independent association of log-transformed NT-proBNP concentration with estimated glomerular filtration rate (eGFR). To assess whether eGFR and NT-proBNP were independently associated with risk, a predictive model was developed using univariate and multivariate Cox proportional hazards regression with 1-year mortality as the dependent variable. Baseline characteristics associated with mortality on univariate analysis with P≤0.10 were retained for possible inclusion in the final model. We examined the strength and shape of the relationships of continuous variables with the log odds of the primary outcome including non-linear terms and using cubic spline plots. If the response appeared non-linear, appropriate transformations were applied. The following variables were examined: age (per 10-unit increase); gender; BMI; diabetes, chronic obstructive pulmonary disease (COPD), previous cerebrovascular events, previous coronary artery bypass grafting/percutaneous coronary angioplasty, chronic liver disease, known dysthyroidism, ischemic etiology, atrial fibrillation, New York Heart Association (NYHA) class IV symptoms, implanted cardioverter defibrillator, use of i.v. diuretics or inotropes, systolic blood pressure (SBP; per 10-unit increase), log-transformed eGFR, log-transformed NT-proBNP concentration, serum sodium, serum potassium, hemoglobin, LVEF (per 5-unit increase), and moderate-to-severe mitral or tricuspid regurgitation (TR). The patients who underwent heart transplantation (HT) or ventricular assist device (VAD) implantation during follow-up were censored at the time of the event in survival analysis. The distribution of NT-proBNP in non-survivors and survivors within each level of eGFR was compared using the 2-sample Kolmogorov-Smirnov test for equality of distribution functions. Statistical significance of the interaction between eGFR and NT-proBNP in predicting mortality was tested using the likelihood ratio test by including the 2 variables and their cross-product term in the same multivariate model.

To further investigate the effect of renal function on the association of NT-proBNP concentration with time to death, first, we stratified the patients according to eGFR into 3 clinical strata: <30 ml·min^–1·1.73 m^–2; 30–59 ml·min^–1·1.73 m^–2; or ≥60 ml·min^–1·1.73 m^–2.²³ We then calculated the hazard ratios (HR) with 95% confidence intervals (95% CI) for NT-proBNP concentration, modeled as a dichotomous variable or categorized on the basis of level of renal function, within each eGFR level using univariate and multivariate Cox proportional hazards models. To dichotomize NT-proBNP, we used a prespecified cut-off of 5,180 ng/L, the prognostic value of which has been identified and confirmed in previous studies.^22,24 To categorize NT-proBNP concentration according to the level of renal function, receiver operating characteristic (ROC) analysis was performed to identify the optimal cut-off (value with the highest sum of sensitivity and specificity²⁵) within each level of eGFR. To compare mortality rates between patient subgroups with NT-proBNP concentration above or below the cut-off of 5,180 ng/L within each level of eGFR, HT- and VAD-free survival curves were developed using Cox proportional hazards models. Finally, we substituted log-transformed NT-proBNP with NT-proBNP modeled as a dichotomous (model 1) or categorized variable (model 2) in the multivariate model and re-ran the analysis. The 2 models were compared using measures of global fit and discrimination. We calculated the Akaike information criterion (AIC) and the Bayesian information criterion (BIC), which are tools for model selection; the explained variation (R²), which measures “the proportion of the variation in the outcome accounted for through the prognostic index”;²⁶ and the Harrell C statistic, which is a measure of discrimination. As a sensitivity analysis, we also calculated HR for NT-proBNP cut-offs with 90% sensitivity within each level of eGFR.

Finally, we also investigated the relation between mortality and in-hospital changes in NT-proBNP and serum creatinine (SCr) concentration in 346 patients with available measurements at admission and discharge (within 48 h before discharge). According to previously published criteria,^27–30 SCr change was dichotomized at a threshold of ≥0.3-mg/dl increase and NT-proBNP change at a threshold of >30% decrease; with respect to decrease >30%, both a decrease <30% or an increase portend significantly poorer survival.^27,30 Thus, 4 patient subgroups could be distinguished: group 1, NT-proBNP decrease >30% and no SCr rise ≥0.3 mg/dl; group 2, no NT-proBNP decrease >30% and no SCR rise ≥0.3 mg/dl; group 3, NT-proBNP decrease >30% and SCr rise ≥0.3 mg/dl; group 4, no NT-proBNP decrease >30% and SCR rise ≥0.3 mg/dl. Survival curves were developed using multivariate Cox models. Adjusted HR with 95% CI were calculated; the group at lowest risk served as the reference group. Analysis was conducted using Stata 12 (StataCorp, College Station, TX, USA).

Results

Of the 929 patients, 21 (2.3%) were lost to follow-up, leaving 908 patients available for analysis.

Four hundred and nine patients (45%) had eGFR ≥60 ml·min^–1·1.73 m^–2; 399 (44%) in the range 30–59 ml·min^–1·1.73 m^–2; and 100 (11%), <30 ml·min^–1·1.73 m^–2. Baseline characteristics across eGFR strata are reported in Table 1. Median NT-proBNP was 4-fold higher among the patients with eGFR <30 ml·min^–1·1.73 m^–2 than among those with eGFR ≥60 ml·min^–1·1.73 m^–2. Two hundred and thirty-four patients (25.8%) died and 89 (9.8%) underwent HT or VAD implantation within 1 year. Fifty-three patients (13%) in the highest eGFR level died, 126 (31.6%) in the intermediate level, and 55 (55%) in the lowest level.

Table 1. Baseline Characteristics vs. eGFR

	All (n=908)	eGFR (ml·min–1·1.73 m–2)			P-value
		>60 (n=409 [45%])	30–59 (n=399 [44%])	<30 (n=100 [11%])
Male	76.9	81.7	73.9	69.0	0.004
Age (years)	66±14	60±15	70±11	72±10	<0.001
BMI (kg/m2)	27±6	27±6	27±6	27±5	0.930
Hypertension	49.7	41.6	55.6	59.0	<0.001
Diabetes	31.6	25.4	33.3	50.0	<0.001
COPD	23.6	19.3	28.1	23.0	0.013
Previous CVE	8.7	6.6	9.5	14.0	0.046
Cancer	6.1	6.4	5.0	9.0	0.291
Previous CABG/PTCA	33.3	26.4	39.6	36.0	<0.001
Chronic liver disease	5.9	3.9	7.3	9.0	0.044
Known dysthyroidism	18.5	13.4	21.6	27.0	0.001
Ischemic etiology	48.0	37.9	55.1	61.0	<0.001
NYHA IV at admission	36.1	29.6	39.8	48.0	<0.001
Atrial fibrillation	34.8	28.4	41.1	36.0	0.001
ICD	60.1	61.6	59.4	57.0	0.651
SBP (mmHg)	111±19	110±19	112±18	111±18	0.305
DBP (mmHg)	70±9	70±10	69±9	69±8	0.110
I.v. inotropes	25.9	24.0	23.6	43.0	<0.001
I.v. diuretics	76.1	73.8	77.2	81.0	0.256
SCr (mg/dl)	1.46±0.71	1.02±0.17	1.54±0.34	2.97±0.95	<0.001
eGFR (ml·min–1·1.73 m–2)	58±24	79±17	46±9	22±6	<0.001
NT-proBNP (ng/L), median (IQR)	3,401 (1,343–7,922)	2,207 (997–5,061)	3,948 (1,668–8,212)	8,814 (4,999–21,403)	<0.001
Serum sodium (mmol/L)	139±4.7	139.1±4.4	139.3±4.8	137.8±5.5	0.025
Serum potassium (mmol/L)	4.3±0.5	4.2±0.5	4.3±0.5	4.5±0.6	<0.001
Hb (g/dl)	12.6±1.9	13.2±1.9	12.4±1.9	11.2±1.7	<0.001
LVEF (%)	29±11	29±11	30±11	28±10	0.141
LVEF <0.40	756 (83.3)	343 (83.9)	325 (81.5)	88 (88)
LVEF <0.30	519 (57.2)	248 (60.6)	210 (52.6)	61 (61)
MR moderate or severe	40.8	39.4	39.8	50.0	0.140
TR moderate or severe	31.2	25.1	33.3	48.0	<0.001

Data given as mean±SD, %, n (%) or median (IQR).

BMI, body mass index; CABG, coronary artery bypass grafting; COPD, chronic obstructive pulmonary disease; CVE, cerebrovascular event; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; Hb, hemoglobin; ICD, implantable cardioverter defibrillator; LVEF, left ventricular ejection fraction; MR, mitral regurgitation; NT-proBNP, N-terminal pro-B-type natriuretic peptide; NYHA, New York Heart Association; PTCA percutaneous coronary angioplasty; SCr, serum creatinine; SBP, systolic blood pressure; TR, tricuspid regurgitation.

On multivariate regression analysis, eGFR was inversely correlated with log-transformed NT-proBNP (adjusted β coefficient=–0.81; P<0.001).

Median NT-proBNP across eGFR strata for patients who survived or died are shown in Figure 1. The median NT-proBNP concentration was higher in the patients who died than in the survivors across all eGFR strata. The difference in cumulative distribution functions of NT-proBNP between non-survivors and survivors, as assessed with the 2-sample Kolmogorov-Smirnov test, was highly significant in each level of eGFR (P<0.001 for all 3 strata). The D statistic was 0.31 for eGFR ≥60 ml·min^–1·1.73 m^–2; 0.35 for eGFR 30–59 ml·min^–1·1.73 m^–2; and 0.49 for eGFR <30 ml·min^–1·1.73 m^–2.

Figure 1.

Median N-terminal pro-B-type natriuretic peptide (NT-proBNP) concentration vs. estimated glomerular filtration rate (eGFR). Boxes, interquartile range; vertical lines, 5th and 95th percentiles.

Multivariate predictors of 1-year mortality were age (HR, 1.22; 95% CI: 1.07–1.39; P=0.003), COPD (HR, 1.42; 95% CI: 1.07–1.89; P=0.015), NYHA class IV (HR, 1.71; 95% CI: 1.29–2.27; P<0.001), SBP (HR, 0.9; 95% CI: 0.82–0.98; P=0.013), log-transformed eGFR (HR, 0.67; 95% CI: 0.5–0.88; P=0.005) and NT-proBNP (HR, 1.52; 95% CI: 1.31–1.76; P<0.001), serum sodium (HR, 0.96; 95% CI: 0.94–0.98; P=0.001), hemoglobin (HR, 0.85; 95% CI: 0.79–0.92; P<0.001), LVEF (HR, 0.88; 95% CI: 0.81–0.96; P=0.005), and moderate-severe TR (HR, 1.47; 95% CI: 1.13–1.91; P=0.005).

The likelihood ratio test comparing the model including log-transformed eGFR and log-transformed NT-proBNP to the model including their cross-product term yielded χ²=0.29 (P=0.5928), indicating a lack of statistically significant interaction. The inclusion of the interaction term in the model did not produce any change in the predictive power. The C-statistic was 0.805 for both models.

Figure 2 shows ROC curves for cut-offs selected on ROC analysis. The unadjusted and adjusted HR for the NT-proBNP cut-off of 5,180 ng/L and the optimal cut-offs selected according to the level of renal function across eGFR strata are reported in Table 2. Table 2 also shows that renal dysfunction was associated with an increased adjusted risk for 12-month mortality regardless of whether NT-proBNP concentration was below or above the cut-off of 5,180 ng/L. Among the patients with NT-proBNP <5,180ng/L, 12-month adjusted mortality was 10.1% (95% CI: 7.2–14.2), 21.1% (95% CI: 16.3–27.0), and 22.1% (95% CI: 10.9–43.6) in those with eGFR ≥60, 30–59, and <30 ml·min^–1·1.73 m^–2, respectively. Among the patients with NT-proBNP >5,180 ng/L, the corresponding mortalities were 28.6% (95% CI: 20.2–39.5), 48.9% (95% CI: 41.4–57.0), and 68.5% (95% CI: 57.7–78.9). Figure 3 shows HT- and VAD-free adjusted survival curves. Measures of model fit and discrimination for the model with NT-proBNP concentrations modeled as a dichotomous (model 1) or categorized variable (model 2) are reported in Table 3. The 2 models had comparable measures of global fit and discrimination The HR for death associated with NT-proBNP cut-offs with 90% sensitivity within each level of eGFR was not significant.

Figure 2.

Receiver operating characteristic curves for N-terminal pro-B-type natriuretic peptide cut-off. AUC, area under the curve; CI, confidence interval; eGFR, estimated glomerular filtration rate.

Table 2. HR for 1-Year Mortality vs. eGFR

	NT-proBNP (ng/L)	n	12-month mortality, % (95% CI)	Univariate analysis		Multivariate analysis
				HR (95% CI)	P-value	HR (95% CI)	P-value
Dichotomous cut-off
Entire cohort	≤5,180	573	15.4 (12.6–18.7)	4.09 (3.13–5.35)	<0.001	2.19 (1.64–2.92)	<0.001
	>5,180	335	47.9 (42.5–53.6)
eGFR (ml·min–1·1.73 m–2)
≥60	≤5,180	308	10.1 (7.2–14.2)	3.35 (1.94–5.76)	<0.001	2.09 (1.14–3.85)	0.018
	>5,180	101	28.6 (20.2–39.5)
30–59	≤5,180	237	21.1 (16.3–27.0)	2.89 (2.02–4.14)	<0.001	1.70 (1.17–2.49)	0.006
	>5,180	162	48.9 (41.4–57.0)
<30	≤5,180	28	22.6 (10.9–43.6)	4.84 (2.07–11.31)	<0.001	3.33 (1.33–8.33)	0.010
	>5,180	72	68.5 (57.7–78.9)
Categorized cut-offs
Entire cohort	≤	468	13.0 (10.2–16.5)	4.08 (3.04–5.49)	<0.001	2.19 (1.59–3.01)	<0.001
	>	440	42.9 (38.2–47.8)
eGFR (ml·min–1·1.73 m–2)
≥60	≤2,847	235	7.7 (4.9–12.1)	3.56 (2.0–6.35)	<0.001	1.5 (0.76–2.96)	0.239
	>2,847	174	24.1 (18.0–31.8)
30–59	≤3,539	190	15.8 (11.3–21.9)	3.77 (2.5–5.69)	<0.001	2.2 (1.42–3.42)	<0.001
	>3,539	209	47.6 (41.0–54.8)
<30	≤8,140	43	29.3 (17.8–45.7)	4.07 (2.14–7.74)	<0.001	3.24 (1.56–6.73)	0.002
	>8,140	57	75.4 (63.8–85.7)

CI, confidence interval; HR, hazard ratio. Other abbreviations as in Table 1.

Figure 3.

Heart transplantation (HT)- and ventricular assist device (VAD)-free survival curves stratified by estimated glomerular filtration rate (eGFR; ml·min^–1·1.73 m^–2) and N-terminal pro-B-type natriuretic peptide (NT-proBNP; ng/L).

Table 3. Measures of Global Fit and Discrimination

	Model 1	Model 2
R2 (95% CI)	0.442 (0.380–0.498)	0.458 (0.395–0.515)
BIC	2,880.94	2,873.82
AIC	2,832.90	2,825.78
C-statistic (95% CI)	0.801 (0.72–0.8267)	0.801 (0.781–0.8265)

AIC, Akaike information criterion; BIC, Bayesian information criterion. Other abbreviation as in Table 2.

Survival and Change in NT-ProBNP and SCr

Compared to the patients without available NT-proBNP and SCr measurements at discharge, those with available data had higher NT-proBNP at admission (P=0.003) and more frequently had NYHA class IV symptoms (P<0.0001) and severe left ventricular (LV) dysfunction (LVEF <0.30; P=0.003) and required i.v. diuretic treatment (P<0.0001). Figure 4 shows adjusted HT- and VAD-free Kaplan-Meier survival curves according to changes in NT-proBNP and SCr from admission to discharge. The HR with 95% CI adjusted for significant covariates, including NT-proBNP and SCr at admission, are reported in Table 4.

Figure 4.

Adjusted heart transplant- and ventricular assist device-free Kaplan-Meier survival curves according to changes in N-terminal pro-B-type natriuretic peptide and serum creatinine (SCr) from admission to discharge.

Table 4. HR and 95% CI for NT-ProBNP and SCr Change

	Group 1 (reference)	Group 2	Group 3	Group 4
NT-proBNP change	Decrease >30%	No decrease >30%	Decrease >30%	No decrease >30%
SCr change (mg/dl)	No SCr rise ≥0.3	No SCr rise ≥0.3	SCr rise ≥0.3	SCr rise ≥0.3
n	127	162	26	31
Unadjusted HR (95% CI)	1.00	2.34 (1.49–3.68)	1.13 (0.46–2.73)	3.62 (1.98–6.60)
P-value		<0.001	0.794	<0.001
Adjusted HR (95% CI)	1.00	2.37 (1.47–3.83)	0.93 (0.37–2.32)	2.55 (1.34–4.86)
P-value		<0.001	0.881	0.005

Abbreviations as in Tables 1,2.

Discussion

We investigated the interaction between renal function and NT-proBNP in predicting mortality for patients hospitalized with ADHF. The major findings are: (1) there was no statistically significant interaction between eGFR and NT-proBNP in relation to 1-year mortality risk prediction; (2) irrespective of the level of renal function, the NT-proBNP cut-off of 5,180 ng/L retained a significant prognostic value; and (3) use of a triple cut-off based on eGFR did not improve the performance of the predictive model.

As expected, we found an inverse correlation between eGFR and NT-proBNP. Median NT-proBNP increased with decreasing eGFR both in non-survivors and survivors, with a steep increase in the lowest eGFR level (<30 ml·min^–1·1.73 m^–2). Of interest, the trend of the differences in cumulative distribution of NT-proBNP between non-survivors and survivors across eGFR strata suggests a decreasing overlap of NT-proBNP with advancing renal dysfunction. No statistically significant interaction, however, between eGFR and NT-proBNP and 1-year mortality, was seen.

After adjusting for the covariates independently associated with mortality risk, the NT-proBNP cut-off of 5,180 ng/L was associated with a significantly increased hazard for death across all 3 strata of renal function. Conversely, when multiple cut-offs selected on ROC analysis according to level of renal function were applied, that for the subgroup with preserved renal function (≥60 ml·min^–1·1.73 m^–2) was not independently associated with increased mortality. It should, however, be considered that the lower event rate in this subgroup may have biased the estimate for the optimal cut-off. The hazard for death associated with elevated NT-proBNP was greater in the lowest (eGFR <30 ml·min^–1·1.73 m^–2) than in the intermediate or highest eGFR level. In the lowest eGFR level, the HR for death was >3, meaning that the patients with severely impaired renal function and elevated NT-proBNP concentration had a >3-fold likelihood of dying within 1 year compared with their counterparts with NT-proBNP below the cut-off. Visual inspection of the survival curves also shows that the difference in mortality between patients with NT-proBNP above or below the cut-off progressively increased with declining renal function. It is relevant to note that survival curves began to diverge very early, especially in the subgroup with severe renal dysfunction, and continued to separate throughout the follow-up. Moreover, it should be noted that the area under the curve for optimal cut-offs selected on ROC analysis progressively increased across decreasing eGFR (Figure 2). Collectively, these data indicate that the discriminative value of NT-proBNP in predicting death improves with declining renal function.

Although changes in AIC and BIC suggest that the predictive model incorporating NT-proBNP cut-offs adapted to the level of residual renal function performed better than the same model including the single cut-off of 5,180 ng/L, the hazard of death conferred by elevated NT-proBNP in the entire cohort, the explained variation, and the discriminative ability, which remains the primary criterion to assess the performance of a predictive model,³¹ were virtually equivalent for the 2 models.

The present findings strongly suggest that higher NT-proBNP concentration in hospitalized patients with ADHF and markedly decreased eGFR do not merely result from decreased renal filtration or impaired renal clearance but are mostly associated with disease severity and prognosis. This is in keeping with the mechanistic study by van Kimmenade et al indicating that, in patients with eGFR <60 ml·min^–1·1.73 m^–2, NT-proBNP concentration “may be more determined by cardiac production than renal clearance”,³² and with other studies assessing the prognostic value of NP in dialysis patients.²⁰ The finding that the “triple cut-point strategy”²⁴ based on eGFR level did not improve the predictive accuracy of the multivariate model supports the use of a single NT-proBNP cut-off for risk prediction. In addition, the use of a prespecified cut-off may allow reduction of the potential bias in selecting optimal cut-off value(s) with a data-driven approach.³³

Finally, we also examined the relationship between mortality and changes in NT-proBNP and SCr concentration from admission to discharge. It is generally accepted that worsening renal function (WRF), defined as an increase in SCr ≥0.3 mg/dl during hospitalization, negatively impacts on survival in ADHF.³⁴ The present data suggest that the impact of WRF on mortality risk is influenced by the pattern of change in NT-proBNP concentration. Indeed, WRF was not related to increased mortality risk unless associated with persistently elevated or increased NT-proBNP. Among the patients who had a decrease in NT-proBNP >30% during hospitalization, those with or without WRF had similar survival. These findings, obtained in a subset of 346 patients with available data at discharge, however, need to be confirmed by larger studies.

Previous Studies

The prognostic accuracy of NT-proBNP in ADHF patients with concomitant renal dysfunction has been addressed in a few studies. deFilippi et al studied 831 dyspneic patients presenting to emergency departments, of whom 437 (53%) had decompensated HF.¹⁷ Among HF patients with renal dysfunction, progressively higher NT-proBNP concentration remained predictive of increased 1-year mortality. van Kimmenade et al analyzed the association of NT-proBNP and renal function with mortality in 720 patients with acute HF enrolled in the International Collaborative on NT-proBNP (ICON) study.³⁵ They found that the combined use of eGFR and NT-proBNP allows identification of patients at highest risk of mortality. In that study, however, the length of follow-up was limited to 60 days and only 84 fatal events were recorded. Thus, confirmatory data are warranted. In addition, the question of whether the use of NT-proBNP cut-offs adapted to level of renal function allows maximization of prognostic accuracy has been incompletely addressed.

The present findings are in line with the results obtained by deFilippi et al and van Kimmenade et al.^17,35 NT-proBNP was independently associated with an increased mortality risk in patients with renal dysfunction, especially in those with severe renal dysfunction. In addition, we showed that, although median NT-proBNP was 4-fold higher among the patients with severely impaired renal function than among those with normal renal function, there was no need to adapt NT-proBNP cut-off to the level of residual renal function to maximize prognostic accuracy. Perhaps, the present data may also have implications for patients with end-stage renal disease in whom a prevalence of HF of 31–40% has been reported.²⁰

Study Limitations

This was a retrospective study. Data were derived from a cohort of patients with predominant systolic HF, most of whom had severe LV systolic dysfunction. Given that patients with reduced LVEF may reasonably have higher NT-proBNP concentration than those with preserved LVEF within the same range of renal function, the results may not be generalizable to HF patients with preserved LV systolic function. Women represented only 23.1% of the study sample. This is consistent with the finding that the majority of acute HF patients with LV systolic dysfunction are men.³⁶ We used all-cause mortality instead of cardiovascular mortality as the primary outcome; this may be another limitation. All-cause death, however, was recognized as a “harder end point, relatively unbiased, easily ascertained, and the most valid”.³⁷ Moreover, establishing the cause of death in observational studies can be very difficult and inaccurate,³⁷ making the outcome of cardiac death susceptible to misclassification bias.^37,38 Finally, it should be considered that, in patients hospitalized with worsening HF and reduced LVEF, cardiovascular death accounts for nearly 90% of all-cause deaths occurring in the months following discharge.³⁹

Conclusions

There was a lack of interaction between renal dysfunction and NT-proBNP in predicting 1-year mortality for patients hospitalized with ADHF. A single NT-proBNP cut-off provided important prognostic information, irrespective of the level of residual renal function. Stratification of NT-proBNP concentration based on clinical renal function level did not result in improved predictive accuracy for 1-year mortality.

Disclosures

No conflict of interest to be declared.

References

• 1.    Wang  TJ,  Larson  MG,  Levy  D,  Benjamin  EJ,  Leip  EP,  Omland  T, et al. Plasma natriuretic peptide levels and the risk of cardiovascular events and death. N Engl J Med 2004; 350: 655–663.
• 2.    Daniels  LB,  Maisel  AS. Natriuretic peptides. J Am Coll Cardiol 2007; 50: 2357–2368.
• 3.    Di Angelantonio  E,  Chowdhury  R,  Sarwar  N,  Ray  KK,  Gobin  R,  Saleheen  D, et al. B-Type natriuretic peptides and cardiovascular risk: Systematic review and meta-analysis of 40 prospective studies. Circulation 2009; 120: 2177–2187.
• 4.    Maisel  A,  Mueller  C,  Adams  K Jr,  Anker  SD,  Aspromonte  N,  Cleland  JG, et al. State of the art: Using natriuretic peptide levels in clinical practice. Eur J Heart Fail 2008; 10: 824–839.
• 5.    Thygesen  K,  Mair  J,  Mueller  C,  Huber  K,  Weber  M,  Plebani  M, et al. Recommendations for the use of natriuretic peptides in acute cardiac care: A position statement from the Study Group on Biomarkers in Cardiology of the ESC Working Group on Acute Cardiac Care. Eur Heart J 2012; 33: 2001–2006.
• 6.    Levy  WC. Can B-type natriuretic peptides replace heart failure risk models? Eur J Heart Fail 2008; 10: 224–225.
• 7.    Pascual-Figal  DA,  Manzano-Fernández  S,  Boronat  M,  Casas  T,  Garrido  IP,  Bonaque  JC, et al. Soluble ST2, high-sensitivity troponin T- and N-terminal pro-B-type natriuretic peptide: Complementary role for risk stratification in acutely decompensated heart failure. Eur J Heart Fail 2011; 13: 718–725.
• 8.    Richards  AM,  Troughton  RW. Use of natriuretic peptides to guide and monitor heart failure therapy. Clin Chem 2012; 58: 62–71.
• 9.    van Veldhuisen  DJ,  Linssen  GC,  Jaarsma  T,  van Gilst  WH,  Hoes  AW,  Tijssen  JG, et al. B-Type natriuretic peptide and prognosis in heart failure patients with preserved and reduced ejection fraction. J Am Coll Cardiol 2013; 61: 1498–1506.
• 10.    Januzzi  JL Jr. Natriuretic peptides, ejection fraction, and prognosis: Parsing the phenotypes of heart failure. J Am Coll Cardiol 2013; 61: 1507–1509.
• 11.    Tsutamoto  T,  Wada  A,  Sakai  H,  Ishikawa  C,  Tanaka  T,  Hayashi  M, et al. Relationship between renal function and plasma brain natriuretic peptide in patients with heart failure. J Am Coll Cardiol 2006; 47: 582–586.
• 12.    Schou  M,  Gustafsson  F,  Kistorp  CN,  Corell  P,  Kjaer  A,  Hildebrandt  PR. Effects of body mass index and age on N-terminal pro brain natriuretic peptide are associated with glomerular filtration rate in chronic heart failure patients. Clin Chem 2007; 53: 1928–1935.
• 13.    Christenson  RH,  Azzazy  HM,  Duh  SH,  Maynard  S,  Seliger  SL,  deFilippi  CR. Impact of increased body mass index on accuracy of B-type natriuretic peptide (BNP) and N-terminal proBNP for diagnosis of decompensated heart failure and prediction of all-cause mortality. Clin Chem 2010; 56: 633–641.
• 14.    Butlet  J,  Chirovsky  D,  Phatak  H,  McNeill  A,  Cody  R. Renal function, health outcomes, and resource utilization in acute heart failure: A systematic review. Circ Heart Fail 2010; 3: 726–745.
• 15.    Shirakabe  A,  Hata  N,  Kobayashi  N,  Shinada  T,  Tomita  K,  Tsurumi  M, et al. Prognostic impact of acute kidney injury in patients with acute decompensated heart failure. Circ J 2013; 77: 687–696.
• 16.    Sato  N,  Kajimoto  K,  Keida  T,  Mizuno  M,  Minami  Y,  Yumino  D, et al. Clinical features and outcome in hospitalized heart failure in Japan (from the ATTEND Registry). Circ J 2013; 77: 944–951.
• 17.    deFilippi  CR,  Seliger  SL,  Maynard  S,  Christenson  RH. Impact of renal disease on natriuretic peptide testing for diagnosing decompensated heart failure and predicting mortality. Clin Chem 2007; 53: 1511–1519.
• 18.    Forfia  PR,  Watkins  SP,  Rame  JE,  Stewart  KJ,  Shapiro  EP. Relationship between B-type natriuretic peptides and pulmonary capillary wedge pressure in the intensive care unit. J Am Coll Cardiol 2005; 45: 1667–1671.
• 19.    Palazzuoli  A,  Masson  S,  Ronco  C,  Maisel  A. Clinical relevance of biomarkers in heart failure and cardiorenal syndrome: The role of natriuretic peptides and troponin. Heart Fail Rev 2014; 19: 267–284.
• 20.    Wang  AY. Clinical utility of natriuretic peptides in dialysis patients. Semin Dial 2012; 25: 326–333.
• 21.    Yancy  CW,  Jessup  M,  Bozkurt  B,  Butler  J,  Casey  DE Jr,  Drazner  MH, et al. 2013 ACCF/AHA guideline for the management of heart failure: A report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines. Circulation 2013; 128: e240–e319, doi:10.1161/CIR.0b013e31829e8776.
• 22.    Scrutinio  D,  Ammirati  E,  Guida  P,  Passantino  A,  Raimondo  R,  Guida  V, et al. Clinical utility of N-terminal pro-B-type natriuretic peptide for risk stratification of patients with acute decompensated heart failure: Derivation and validation of the ADHF/NT-proBNP risk score. Int J Cardiol 2013; 168: 2120–2126.
• 23.   K/DOQI clinical practice guidelines for chronic kidney disease: Evaluation, classification, stratification. Am J Kidney Dis 2002; 39: S1–S266.
• 24.    Januzzi  JL,  van Kimmenade  R,  Lainchbury  J,  Bayes-Genis  A,  Ordonez-Llanos  J,  Santalo-Bel  M, et al. NT-proBNP testing for diagnosis and short-term prognosis in acute destabilized heart failure: An international pooled analysis of 1256 patients. The International Collaborative of NT-proBNP Study. Eur Heart J 2006; 27: 330–337.
• 25.    Zou  KH,  O’Malley  AJ,  Mauri  L. Receiver-operating characteristic analysis for evaluating diagnostic tests and predictive models. Circulation 2007; 115: 654–657.
• 26.    Royston  P. Explained variation for survival models. Stata J 2006; 6: 83–96.
• 27.    Bettencourt  P,  Azevedo  A,  Pimenta  J,  Friões  F,  Ferreira  S,  Ferreira  A. N-terminal-pro-brain natriuretic peptide predicts outcome after hospital discharge in heart failure patients. Circulation 2004; 110: 2168–2174.
• 28.    Damman  K,  Valente  MA,  Voors  AA,  O’Connor  CM,  van Veldhuisen  DJ,  Hillege  HL. Renal impairment, worsening renal function, and outcome in patients with heart failure: An updated meta-analysis. Eur Heart J 2014; 35: 455–469.
• 29.    Givertz  MM,  Postmus  D,  Hillege  HL,  Mansoor  GA,  Massie  BM,  Davison  BA, et al. Renal function trajectories and clinical outcomes in acute heart failure. Circ Heart Fail 2014; 7: 59–67.
• 30.    Carrasco-Sánchez  FJ,  Pérez-Calvo  JI,  Morales-Rull  JL,  Galisteo-Almeda  L,  Páez-Rubio  I,  Barón-Franco  B, et al. Heart failure mortality according to acute variations in N-terminal pro B-type natriuretic peptide and cystatin C levels. J Cardiovasc Med (Hagerstown) 2014; 15: 115–121.
• 31.    Pencina  MJ,  D’Agostino  RB,  D’Agostino Jr  RB,  Vasan  RS. Evaluating the added predictive ability of a new marker: From area under the ROC curve to reclassification and beyond. Stat Med 2008; 27: 157–172.
• 32.    van Kimmenade  RR,  Januzzi  JL Jr,  Bakker  JA,  Houben  AJ,  Rennenberg  R,  Kroon  AA, et al. Renal clearance of B-type natriuretic peptide and amino terminal pro-B-type natriuretic peptide: A mechanistic study in hypertensive subjects. J Am Coll Cardiol 2009; 53: 884–890.
• 33.    Leeflang  MMG,  Moons  KGM,  Reitsma  JB,  Zwinderman  AH. Bias in sensitivity and specificity caused by data-driven selection of optimal cutoff values: Mechanisms, magnitude, and solutions. Clin Chem 2008; 54: 729–737.
• 34.    Greenberg  B. Acute decompensated heart failure: Treatments and challenges. Circ J 2012; 76: 532–543.
• 35.    van Kimmenade  RR,  Januzzi  JL Jr,  Baggish  AL,  Lainchbury  JG,  Bayes-Genis  A,  Richards  AM, et al. Amino-terminal pro-brain natriuretic peptide, renal function, and outcomes in acute heart failure: Redefining the cardiorenal interaction? J Am Coll Cardiol 2006; 48: 1621–1627.
• 36.    Gheorghiade  M,  Pang  PS. Acute heart failure syndromes. J Am Coll Cardiol 2009; 53: 557–573.
• 37.    Hachamovitch  R,  Di Carli  MF. Methods and limitations of assessing new noninvasive tests: Part II Circulation 2008; 117: 2793–2801.
• 38.    Lauer  MS,  Blackstone  EH,  Young  JB,  Topol  EJ. Cause of death in clinical research: Time for a reassessment? J Am Coll Cardiol 1999; 34: 618–620.
• 39.    O’Connor  CM,  Miller  AB,  Blair  JEA,  Konstam  MA,  Wedge  P,  Bahit  MC, et al. Causes of death and rehospitalizations in patients hospitalized with worsening heart failure and reduced left ventricular function: Results from Efficacy of Vasopressin Antagonism in Heart Failure Outcome Study with Tolvaptan (EVEREST) program. Am Heart J 2010; 159: 841–849.