Author + information
- Received November 20, 2006
- Revision received January 3, 2007
- Accepted February 5, 2007
- Published online May 15, 2007.
- Gregg C. Fonarow, MD, FACC⁎,⁎ (, )
- William F. Peacock, MD†,
- Christopher O. Phillips, MD, MPH‡,
- Michael M. Givertz, MD, FACC§,
- Margarita Lopatin, MS‖,
- ADHERE Scientific Advisory Committee and Investigators
- ↵⁎Reprint requests and correspondence:
Dr. Gregg C. Fonarow, University of California, Los Angeles, Division of Cardiology, 10833 Le Conte Avenue, Los Angeles, California 90095.
Objectives This study was designed to determine whether admission B-type natriuretic peptide (BNP) levels are predictive of in-hospital mortality in acute decompensated heart failure (HF).
Background Levels of BNP have been demonstrated to facilitate the diagnosis of HF and predict mortality in chronic systolic HF.
Methods B-type natriuretic peptide levels within 24 h of presentation were obtained in 48,629 (63%) of 77,467 hospitalization episodes entered in ADHERE (Acute Decompensated Heart Failure National Registry). In-hospital mortality was assessed by BNP quartiles in the entire cohort and in patients with reduced (n = 19,544) as well as preserved (n = 18,164) left ventricular systolic function using chi-square and logistic regression models.
Results Quartiles (Q) of BNP were Q1 (<430), Q2 (430 to 839), Q3 (840 to 1,729), and Q4 (≥1,730 pg/ml). The BNP levels were <100 pg/ml in 3.3% of the total cohort. Patients in Q1 versus Q4 were younger, more likely to be women, and had lower creatinine and higher left ventricular ejection fraction. There was a near-linear relationship between BNP quartiles and in-hospital mortality: Q1 (1.9%), Q2 (2.8%), Q3 (3.8%), and Q4 (6.0%), p < 0.0001. B-type natriuretic peptide quartile remained highly predictive of mortality even after adjustment for age, gender, systolic blood pressure, blood urea nitrogen, creatinine, sodium, pulse, and dyspnea at rest, Q4 versus Q1 (adjusted odds ratio 2.23 [95% confidence interval 1.91 to 2.62, p < 0.0001]). The BNP quartiles independently predicted mortality in patients with reduced and preserved systolic function.
Conclusions An elevated admission BNP level is a significant predictor of in-hospital mortality in acute decompensated HF with either reduced or preserved systolic function, independent of other clinical and laboratory variables. (Registry for Acute Decompensated Heart Failure Patients; http://www.clinicaltrials.gov/show/NCT00366639; NCT00366639).
The natriuretic peptides are counter-regulatory hormones involved in volume homeostasis and cardiovascular remodeling. B-type natriuretic peptide (BNP) is a 32-amino-acid neurohormone synthesized in ventricular myocardium and released into the circulation in response to ventricular dilatation and pressure overload (1,2). B-type natriuretic peptide is derived from an intracellular 108-amino-acid precursor that is cleaved predominately into 2 fragments, yielding a 76-amino-acid N-terminal fragment (NT-proBNP) and BNP (1). Levels of BNP and NT-proBNP have been shown to be elevated in patients with left ventricular (LV) dysfunction and correlate with the New York Heart Association functional class (3). Clinical investigations of natriuretic peptides have focused on the diagnostic usefulness for heart failure (HF) and LV dysfunction and their prognostic usefulness in chronic HF, acute coronary syndromes, stable coronary artery disease, other medical conditions, and community cohorts (3–11).
Whether plasma levels of BNP are predictive of in-hospital mortality risk in patients hospitalized with acute decompensated HF has not been well studied. The prior studies that have examined the relationship between presentation levels of BNP or NT-proBNP and short-term mortality risk have been relatively small and not adequately powered to assess in-hospital mortality independent of other variables (12–18). The primary aim of this study was to assess the relationship of BNP to in-hospital mortality by using data from ADHERE (Acute Decompensated Heart Failure National Registry). This registry collects detailed hospitalization data from the initial presentation at the hospital or emergency department until discharge, transfer, or in-hospital death (19,20). Coming from a large observational database, these data reflect recent clinical characteristics and in-hospital outcomes for a broad cohort of patients hospitalized with acute decompensated HF (19–22).
Data used to determine risk associated with BNP levels were taken from ADHERE. This registry collects detailed hospitalization data from initial presentation in the hospital or emergency department until discharge, transfer, or in-hospital death (19,20). The ADHERE registry contains data on patients hospitalized with acute decompensated HF in community, tertiary, and academic centers from all regions of the country (20). For the purpose of the registry, acute decompensated HF is defined as new-onset decompensated HF or decompensation of chronic, established HF with symptoms sufficient to warrant hospitalization. The design, methods, and patient characteristics in ADHERE have been described previously (19). Briefly, medical records are reviewed at participating study sites, and data from consecutive eligible male and female patients ≥18 years of age at the time of hospitalization are entered in the registry using an electronic case report form incorporating real-time validity checking (19,20). These data include demographic information, medical history, baseline clinical characteristics, initial evaluation, treatment received, procedures performed, hospital course, and patient disposition. Standardized definitions are used for all patient-related variables, clinical diagnoses, and hospital outcomes. Importantly, registry participation does not require any alteration of treatment or hospital care, and entry of data into the registry is not contingent on the use of any particular therapeutic agent or treatment. Institutional review board approval is required for all participating centers; however, informed consent of individuals was not required for registry entry. To preserve patient confidentiality, direct patient identifiers are not collected, and data are reported only in aggregate format. Therefore, registry entries reflect individual hospitalization episodes, not individual patients, and multiple hospitalizations of the same patient may be entered into the registry as separate records.
In the beginning of 2003, ADHERE hospitals began employing the expanded range BNP testing (range 0 to 5,000 pg/ml). This current study analyzed ADHERE data from April 2003 through December 2004 (February 2005 data transfer). During this time frame 191 of 229 ADHERE hospitals reported having the capability of assessing BNP levels (176 BNP only, 15 BNP and NT-proBNP, 14 NT-proBNP only, and 24 none). B-type natriuretic peptide levels on presentation (first level obtained within 24 h of presentation) were analyzed by the local hospital laboratory and recorded in the medical record. During the analysis time frame, 48,629 (63%) of 77,467 patients episodes had BNP assessment. For the primary analyses, patients were grouped by BNP quartiles. Data were analyzed for the overall cohort as well as for those patients with reduced (left ventricular ejection fraction [LVEF] <40%) and preserved (LVEF ≥40%) systolic function. Analysis of patients with LVEF ≥50% was also performed. Data were also assessed for each weight category by body mass index (BMI) (kg/m2): underweight (BMI <18.5), healthy weight (BMI 18.5 to 24.9), overweight (BMI 25.0 to 29.9), and obese (BMI ≥30).
Data from ADHERE were used for analyses of clinical characteristics, treatments, and outcomes for patient episodes of HF grouped by quartiles of BNP. Analysis by BNP quartiles, continuous, and log-transformed BNP was performed. As log transformation produced similar findings, data are presented without log transformation for clarity. A relationship between mortality and BNP was examined using locally weighted smoother regression scatterplots (loess). Procedure PLSMO from Hmisc S-PLUS library (F.E. Harrell, Department of Biostatistics, Vanderbilt University, Nashville, Tennessee) was used to create the plot. The hypothesis of no differences in patients’ characteristics and outcomes among 4 BNP quartiles was tested using chi-square, analysis of variance, and Kruskal-Wallis tests as appropriate. The Kruskal-Wallis test was used to analyze outcome variables with skewed distribution. Two-sided p values were reported. Because of anticipated differences among the 4 BNP quartile groups in medical history and clinical characteristics at presentation, it was important to adjust the mortality comparison for relevant prognostic factors. Of 80 demographic, medical history, and initial evaluation variables collected in ADHERE, classification and regression tree analysis and/or logistic regression models previously identified 8 variables as the most important risk factors for in-hospital mortality (20,21,23). Mortality rates in the BNP quartiles were compared using logistic regression adjusted for these 8 variables: age, blood urea nitrogen (BUN), systolic blood pressure (SBP), diastolic blood pressure, creatinine, sodium, heart rate, and dyspnea at rest, as well as gender. A total of 1.2% of records were excluded for 1 or more missing values. Multivariable analysis using 48 of 51 variables previously described (21) resulted in little incremental improvement in prognostic value. To adjust for multiple comparisons (e.g., 3), only p values <0.017 were considered statistically significant using Bonferroni correction. With the large sample size, statistical significance with small differences in clinical variables was expected. Risk-adjusted mortality was also assessed separately in the cohort of patients with reduced and preserved LV systolic function (quantitative LVEF documented in 75.5%). The relationship between risk-adjusted mortality and BNP was also analyzed with BNP as a continuous variable. There was little variation in BNP by center, hospital type, or region; thus, hierarchic analysis was not performed. The area under the receiver-operator curve was used to assess the discrimination of the models. These analyses were performed using version 8.2 of SAS software (SAS Institute, Inc., Cary, North Carolina).
From April 2003 through December 2004, there were 48,629 (63%) of 77,467 patient episodes in 191 hospitals that had BNP assessment within 24 h of presentation. The BNP testing rate increased over time from 51.2% in second-quarter 2003 to 75.4% in fourth-quarter 2004 (p < 0.0001). The distribution of BNP for these acute HF hospitalizations is shown in Figure 1.B-type natriuretic peptide levels were <100 pg/ml in 3.3% of the total cohort. Quartiles (Q) of BNP were Q1 (<430 pg/ml), Q2 (430 to 839 pg/ml), Q3 (840 to 1,729 pg/ml), and Q4 (≥1,730 pg/ml).
Patient characteristics as stratified by BNP quartiles are shown in Table 1.Patients in BNP Q1 versus Q4 were slightly younger (mean age 71.5 vs. 73.9 years) and were more likely to be women (55% vs. 49%). A greater proportion of African Americans (22% vs. 16%) were in Q4 versus Q1. Patients in Q1 versus Q4 were less likely to have a history of HF and coronary artery disease and were half as likely to have a history of chronic renal insufficiency, but these patients had a greater prevalence of a history of diabetes (Table 1). Notably, there were few clinically relevant differences observed in the admission signs and symptoms of HF by BNP quartile. Somewhat higher systolic blood pressure and lower creatinine levels were associated with a lower BNP quartile. Patients in BNP Q4 were more than twice as likely to have moderate to severe impairment in LVEF (Q1 29% vs. Q4 70%).
In-hospital medication use by BNP quartiles is shown in Table 2.Most patients received intravenous diuretics (90% to 94%), and there were only small differences by BNP quartile. More than twice as many patients in BNP Q4 versus Q1 were treated with inotropic agents or nesiritide, whereas there was little difference in nitroglycerine use by BNP quartiles (Table 2). Use of beta-blockers and digoxin during hospitalization was modestly more frequent in patients in higher BNP quartiles. Use of ultrafiltration and dialysis was significantly more frequent in higher BNP quartile patients (Table 2).
The in-hospital mortality risk for the overall patient population was 3.6% (1,760 of 48,629). There was a near-linear relationship between BNP and in-hospital mortality: Q1 (1.9%), Q2 (2.8%), Q3 (3.8%), and Q4 (6.0%), p < 0.0001 (Fig. 2).For the 1,407 patients with BNP levels at or above 5,000 pg/ml, the in-hospital mortality was 8.5%. B-type natriuretic peptide quartiles were highly predictive of mortality even after adjustment for age, gender, SBP, BUN, creatinine, sodium, pulse, and dyspnea at rest, Q4 versus Q1, adjusted odds ratio 2.23 (95% confidence interval 1.91 to 2.62), p < 0.0001, as shown in Table 3.The area under the receiver-operator curve for the model was 0.77, indicating adequate model discrimination (Table 3). Addition of up to 40 additional demographic, medical history, outpatient medications, and initial assessment variables produced similar findings (area under the curve 0.79). As a continuous variable, BNP was also predictive of mortality (Fig. 3).In multivariable analysis, for every 400-U increase in BNP, the odds of risk-adjusted mortality were 9% higher (odds ratio 1.09, 95% confidence interval 1.08 to 1.11, p < 0.0001).
B-type natriuretic peptide quartiles also predicted mortality in patients with LVEF <40%, LVEF ≥40%, and LVEF ≥50% as shown in Figure 2. These findings for BNP and mortality were independent of other major risk factors in patients with and without preserved systolic function. When the analysis was confined to patients with very low LVEF (<15%, n = 1,574) the median BNP was 1,539 pg/ml, and higher BNP remained predictive of mortality (Q1, BNP <865 pg/ml, 1.5% vs. Q4, BNP >2,830 pg/ml, 7.1%, p = 0.0005). Of 38,242 hospitalizations with documented BMI, there were 1,194 (3%) underweight, 12,018 (31%) healthy weight, 11,046 (29%) overweight, and 13,984 (37%) obese. There was a strong inverse relationship between BMI and median BNP levels: underweight 1,380 (interquartile range 732 to 2,890 pg/ml), healthy weight 1,217 (613 to 2,354), overweight 891 (480 to 1,740), and obese 555 (294 to 1,103), p < 0.0001. For each BMI category, except for the underweight group, BNP levels remained a significant univariate and multivariate predictor of in-hospital mortality. B-type natriuretic peptide levels were also independently predictive of mortality in unique patients with new onset heart failure (n = 11,707).
The BNP quartile groups also predicted other clinical outcomes, including the need for mechanical ventilation, length of stay, time in the intensive care unit, and percent hospitalization in the intensive care unit, as shown in Table 4.Patients in higher BNP quartiles were more likely to require mechanical ventilation and be hospitalized in the intensive care unit. These patients also had longer length of stays. Patients with higher BNP levels on admission were also less likely to be asymptomatic at hospital discharge.
The ADHERE provides insights into the relationship among presenting characteristics, laboratories, and in-hospital outcomes in a broad population of patients recently hospitalized with acute decompensated HF. This analysis of >48,000 acute decompensated HF hospitalizations, in patients demographically and clinically similar to those seen in other large community- or Medicare-based evaluations (19,22), demonstrates that the risk of in-hospital mortality can be reliably estimated using BNP obtained on hospital admission. Overall, in-hospital mortality was 3.6%, but this mortality risk varied more than 3- to 4-fold on the basis of the patient’s initial BNP. B-type natriuretic peptide provides risk prediction independently of numerous other clinical and laboratory variables previously demonstrated to be predictive of in-hospital outcomes. This relationship between BNP and outcome persisted in hospitalized HF patients with both reduced and preserved systolic function and across BMI categories. Our findings confirm that BNP provides prognostic information among patients with cardiovascular disease, but they are the first to demonstrate this relationship with in-hospital mortality among patients hospitalized with acute decompensated HF.
Natriuretic peptides have been shown to predict prognosis in patients with acute coronary syndromes, stable coronary artery disease, chronic HF, and a variety of other medical conditions, as well as in community cohorts (6–11). There have been relatively few studies assessing the role of natriuretic peptide testing for predicting prognosis in patients presenting with acute decompensated HF. Harrison et al. (24) found that BNP levels measured in 325 patients presenting to the emergency department with dyspnea were predictive of cardiac events (cardiac death, emergency department visit, or admission for a cardiac cause) over the ensuing 6 months. The REDHOT (Rapid Emergency Department Heart Failure Outpatient Trial) was a 10-center study assessing the relationship between BNP levels and 90-day outcomes of patients presenting to the emergency department with shortness of breath (13). The BNP levels upon presentation were predictive of 90-day mortality rates, but multivariable analysis included only New York Heart Association functional class and disposition. In another analysis of 96 patients hospitalized with acute HF, admission NT-proBNP levels were not predictive of subsequent events (16). A larger report compared differences in NT-proBNP levels among 1,256 patients with and without acute HF and the relationship between NT-proBNP levels and HF symptoms in a pooled analysis of 3 studies and 1 registry cohort (18). The investigators found that in the acute HF subgroup, a presenting NT-proBNP concentration >5,180 pg/ml was predictive of death by 76 days. Finally, in a recent analysis of the PRIDE (ProBNP Investigation of Dyspnea in the Emergency Department) study of 599 patients who presented to the emergency department with dyspnea, it was observed that an elevated NT-proBNP concentration at presentation was predictive of death by 1 year after presentation, independent of other prognostic variables (17). These prior studies have mostly involved a mixed population of patients presenting with dyspnea, including patients not ultimately diagnosed as having heart failure; have involved relatively few study sites; have been inadequately powered; and as a result they have not adequately investigated whether natriuretic peptides are predictive of in-hospital mortality risk.
Because multiple risk factors can exist in the same patient, to be meaningful, analysis of the predictive value of a potential biomarker must consider its relationship to other prognostic factors and predictive models (20). As higher BNP levels are associated with other clinical presentation and laboratory variables that are indicative of higher risk, it could be speculated that BNP levels, although providing prognostic information in isolation, do not provide meaningful incremental risk prediction beyond traditional assessment. In this study, patients in the higher BNP quartiles were slightly older, and had a higher BUN, a lower admission SBP, and a lower LVEF than those in lower BNP quartiles. However, even after adjustment for multiple other prognostic factors previously identified by classification and regression tree analysis and/or logistic regression models as the most important risk factors for in-hospital mortality in acute decompensated HF and stratification for LV systolic function, the relationship between BNP and mortality persisted. This study demonstrates that assessment of admission BNP levels can provide prognostic information regarding in-hospital mortality in a broad population hospitalized with HF, which is incremental to clinical assessment and prior risk prediction models.
National guidelines for acute coronary syndromes recommend the use of the biomarkers cardiac troponin and BNP for prognosis and risk stratification (25). Current HF guidelines, however, recommend consideration of BNP testing solely for diagnostic purposes, and only in patients where the initial diagnosis is uncertain (26). These new findings suggest that assessment of presentation BNP levels in patients hospitalized with HF provides independent prognostic information regarding in-hospital mortality and other clinical outcomes and may be useful to stratify risk in HF patients. In light of the value of BNP and NT-proBNP testing in providing short-term and long-term prognostic information in patients presenting with acute decompensated HF, routine assessment of BNP or NT-proBNP levels may be considered (17,18). Unlike LVEF, which is frequently not available to clinicians for hours to days after hospital admission, BNP levels are generally available within 15 min to a few hours of presentation and are predictive of subsequent outcome in patients with both reduced and preserved systolic function. This incremental prognostic information provided by BNP has the potential to aid in medical decision making. Patients judged to be at higher risk may receive higher level monitoring and earlier, more intensive treatment for acute decompensated HF. Conversely, patients estimated to be at lower risk may receive lower level monitoring and be managed less intensively. However, prospective studies will be necessary to validate this hypothesis.
Potential limitations of the current analysis must be acknowledged. Real-world practice information can be both an advantage and a disadvantage of analyses based on registry data. This real-world study used results of various commercially available BNP assays rather than results from a single central core laboratory. Although this methodology may introduce great variability to BNP testing results, this approach makes these findings more applicable to clinical practice. It should also be noted that to the extent that different commercially available BNP assays produce results that vary, the specific BNP level cutpoints for risk stratification may vary. These study results can be influenced by differences in disease assessment, background medical therapy for heart failure, in-hospital treatments, and documentation patterns at participating institutions. However, ADHERE reflects patients cared for by thousands of clinicians at hundreds of hospitals across the country, thus providing an excellent opportunity to adjust for this variation and determine the added risk prediction provided by BNP (20). Conversely, it must be considered that these findings may not apply to patients who are cared for in settings that deviate substantially from those in ADHERE.
Other factors to consider include that each patient’s actual risk may be influenced by many factors not measured or considered in this analysis. It thus must be emphasized that BNP levels, use of other predictive variables, and application of risk prediction models enhance, but are not intended to replace, physician assessment. Because the ADHERE registry does not contain specific patient identifiers, information regarding patient status following hospital discharge is not available. Thus, the effects of BNP on after-discharge mortality risks cannot be determined in this study. Similarly, because of the lack of patient identifiers, the analyzed cohorts may contain multiple admissions for the same patient. This should not have influenced the study results, as the principal outcome parameter, in-hospital mortality as it relates to the admission BNP level, is specific to individual hospitalization episodes and analysis confined to unique new-onset heart failure patients in this study yielded similar findings (20). Despite these potential limitations, the current analysis of the ADHERE registry represents the largest analysis to date of the predictive capability of a biomarker in patients hospitalized with cardiovascular disease.
An elevated admission BNP level is a significant independent predictor of in-hospital mortality in acute HF. B-type natriuretic peptide level is predictive of in-hospital mortality in patients with either reduced or preserved LVEF independent of other clinical and laboratory variables. In light of the demonstration of incremental prognostic information, the BNP assay may be considered for inclusion as part of the assessment of patients presenting with acute decompensated HF. Further research will be necessary to determine whether patients with higher admission BNP will benefit from more intensive monitoring and aggressive treatment strategies.
ADHERE and this study were funded by Scios, Inc., Mountain View, California.
- Abbreviations and Acronyms
- body mass index
- B-type natriuretic peptide
- blood urea nitrogen
- heart failure
- left ventricular
- left ventricular ejection fraction
- N-terminal pro-B-type natriuretic peptide
- Received November 20, 2006.
- Revision received January 3, 2007.
- Accepted February 5, 2007.
- American College of Cardiology Foundation
- Bruneau B.G.,
- Piazza L.A.,
- de Bold A.J.
- Cowie M.R.,
- Jourdain P.,
- Maisel A.,
- et al.
- Anand I.S.,
- Fisher L.D.,
- Chiang Y.T.,
- et al.
- Gustafsson F.,
- Steensgaard-Hansen F.,
- Badskjaer J.,
- et al.
- Bettencourt P.,
- Azevedo A.,
- Pimenta J.,
- et al.
- Maisel A.,
- Hollander J.E.,
- Guss D.,
- et al.
- Van Kimmenade R.R.,
- Januzzi J.L. Jr..,
- Baggish A.L.,
- et al.
- Cheng V.,
- Kazanagra R.,
- Garcia A.,
- et al.
- Januzzi J.L. Jr..,
- Van Kimmenade R.R.,
- Lainchbury R.,
- et al.
- Adams K.F.,
- Fonarow G.C.,
- Emerman C.L.,
- et al.
- Yancy C.W.,
- Lopatin M.,
- Stevenson L.W.,
- de Marco T.,
- Fonarow G.C.
- Abraham W.T.,
- Adams K.F.,
- Fonarow G.C.,
- et al.
- Antman E.M.,
- Anbe D.T.,
- Armstrong P.W.,
- et al.
- Hunt S.A.,
- Abraham W.T.,
- Chin M.H.,
- et al.