Paul Erne, MD, Department of Cardiology, Kantonsspital Luzern, 6000 Luzern 16, Switzerland; e-mail: email@example.com
While low brain natriuretic peptide (BNP) values have been shown to improve diagnostic accuracy by excluding congestive heart failure in acute dyspnea, the meaning of excessively elevated BNP is not clear. This is a retrospective analysis of clinical, echocardiographic, and laboratory data in patients with BNP values >2500 ng/L. Sixty-seven patients (36 men) with a median age of 79 years were included. The median BNP level was 3118 ng/L (2506–5000 ng/L). Forty-six percent of the patients had no dyspnea or New York Heart Association class II dyspnea. Most patients had impaired renal function. BNP value did not correlate with ejection fraction (r=0.2; P=.15) and weakly correlated with left ventricular end-diastolic diameter index (r=0.26; P=.02). Very high BNP values only poorly correlated with dyspnea and traditional echocardiographic markers of congestive heart failure. The clinical usefulness of very high BNP values is questionable and needs evaluation in a prospective trial.
Brain natriuretic peptide (BNP) is a neuro-hormone that is secreted from both atria and ventricles. In healthy persons, BNP is mainly secreted in atria.1 In patients with heart failure, increased ventricular wall stretch resulting from volume and pressure overload leads to increased production of circulating cardiac natriuretic peptides (ie, A-type [ANP “atrial”] and B-type [BNP “brain”] peptides and their amino-terminal fragments of their prohormones, respectively).2,3 The addition of BNP testing to clinical judgment in the evaluation of acute dyspnea clearly improves diagnostic accuracy of congestive heart failure (CHF).4,5
Elevations of BNP6 and N-terminal prohormone brain natriuretic peptide (NT-proBNP) were shown to correlate with prognosis in patients with CHF.7–9 In contrast with these results, a recent study showed that lower rather than higher levels of B-type natriuretic peptides (NT-proBNP and BNP) predict short-term mortality in end-stage heart failure.10 Another study, which was designed to answer the question of whether BNP testing is less useful in patients with a history of CHF, showed that BNP levels >100 ng/L have a high sensitivity (100%) but low specificity (41%). Increasing the cutoff to 400 ng/L increased the positive predictive value somewhat but still missed some patients with heart failure as a cause.11
In our clinical practice, we have observed patients with high BNP values without clinical diagnosis of severe CHF, as well as patients with terminal heart failure with only slightly elevated BNP values. In a pooled analysis of 3 European epidemiologic studies (N=3051) the possible cause of high NT-proBNP values could not be determined in 5.9% of the patients with dyspnea. Sixty-four percent of the high BNP levels were associated with structural or functional cardiac abnormality or renal impairment.12
Because little is known about the characteristics of patients with very high BNP values, the aim of our study was to evaluate clinical, echocardiographic, and laboratory characteristics of hospitalized patients with very high BNP levels. We hypothesized that high BNP values weakly correlate with cardiac function and postulated that there is no good correlation between high BNP values and traditional Doppler echocardiographic markers of heart failure in these patients.
In this retrospective study, we analyzed the data of hospitalized patients treated in a single tertiary care institution in Switzerland. Data from all patients who presented at our hospital between February 2005 and March 2007 with a BNP level of >2500 ng/L were analyzed. The value of 2500 ng/L was chosen to represent the lower limit of “very high” BNP values. The upper range of BNP level by the method used was 5000 ng/L. In the time period mentioned above, 157 of a total of 5950 patients had a BNP measurement that was >2500 ng/L. Patient data were eligible for analysis if the patient had an echocardiographic study performed within 48 hours of BNP measurement. Clinical, laboratory, and Doppler echocardiographic data were retrieved by chart review.
Clinical Data. Hospitalization cause and the presence of symptoms and signs of cardiac failure at BNP measurement were registered. Dyspnea was classified according to New York Heart Association (NYHA) categories I through IV. In addition, in-hospital mortality data were obtained by reviewing hospital discharge reports. The etiology of the cardiac disease was characterized as coronary artery disease, valvular heart disease, hypertensive heart disease, dilating cardiomyopathy, or any combination of two or more thereof.
Laboratory Data. All samples were tested for BNP on an ADVIA Centaur analyzer (Siemens-Bayer Healthcare, Zurich, Switzerland). The ADVIA Centaur BNP assay is a 2-site, dual-monoclonal sandwich immunochemoluminescent assay. Creatinine level was determined enzymatically on a Roche Hitachi 917 system (Roche Diagnostics Division, Rotkreuz, Switzerland).
Following the guidelines of the Kidney Disease Outcomes Quality Initiative,13 the Modification of Diet in Renal Disease formula14 was used to estimate renal function, which was divided in 5 stages using the estimated glomerular filtration rate (GFR). Stage I (GFR, 90–130 mL/min/1.73 m2), stage II (GFR, 60–90 mL/min/1.73 m2), stage III (GFR, 30–60 mL/min/1.73 m2), stage IV (GFR, 15–30 mL/min/1.73 m2), and stage V (GFR, <15 mL/min/1.73 m2). BNP values were then compared among the different GFR groups.
Doppler Echocardiographic Data. Conventional 2-dimensional transthoracic echocardiography was performed in apical 4-chamber view and parasternal short- and long-axis views. Ejection fraction (EF) was obtained using the modified Simpson method.15 According to the European Society of Cardiology guidelines, systolic heart failure was defined as presence of CHF symptoms and an EF <50%.16 The following Doppler echocardiographic parameters were obtained and analyzed for correlation with BNP values: interventricular septum (IVS) in mm, left ventricular end-diastolic diameter (LVEDD) index in mm/m2, left atrial (LA) diameter index in mm/m2, and LA and right atrial (RA) surface index in cm2/m2. Left ventricular mass index was calculated following the recommendations of Devereux and associates.17
Diastolic heart failure was defined following the guidelines from the European Study Group on Diastolic Heart Failure: presence of clinical signs of CHF and evidence of abnormal left ventricular relaxation; pseudonormalization; and restrictive physiology, indicating elevated left ventricular filling pressure with documentation of normal or only mildly impaired systolic function.18
Statistical Analysis. For statistical analysis, StatView software (version 5.0;SAS Institute, Cary, NC) was used. Correlation was tested using Fisher's R to Z test. A P value <.05 was considered significant.
A total of 67 patients (36 men, 31 women) with a BNP value >2500 ng/L were included in the study. Forty-three patients (64%) were hospitalized because of CHF, 24 patients (36%) for reasons other than CHF.
Baseline characteristics are presented in the Table. Median age was 79 years (range, 47–90 years), and the median BNP value was 3118 ng/L (range, 2506–5000 ng/L). Age weakly correlated with BNP values (r=0.36; P=.002). Median BNP level was slightly higher in men (3282 ng/L) than in women (2967 ng/L).
Table Table. Baseline Characteristics of the Study Population
Body mass index, kg/m2
Heart rate on admission, bpm
Estimated creatinine clearance, mL/min
LVM index, g/m2
LVEDD index, mm/m2
LA diameter, mm
LA surface index, cm2/m2
RA surface index, cm2/m2
Abbreviations: BNP, brain natriuretic peptide; bpm, beats per minute; IVS, interventricular septum; LA, left atrial; LVEDD, left ventricular end-diastolic diameter; LVEDVI, left ventricular end-diastolic volume index; LVEF, left ventricular ejection fraction; LVM, left ventricular mass; RA, right atrial.
All patients had symptoms and/or signs of CHF at the time of BNP measurement. Fifteen patients (22%) had no dyspnea, 16 (24%) had NYHA class II dyspnea, 20 (30%) had NYHA class III dyspnea, and 16 (24%) had NYHA class IV dyspnea.
Twelve patients (18%) died during hospitalization. Patients with in-hospital death had higher median BNP values (3961 vs 2982 ng/L; range, 2652–5000 vs 2506–5000 ng/L). The difference was significant (P=.01).
Etiology of Cardiac Failure. Nineteen patients (28%) had a combination of several causes of heart failure (mostly coronary artery disease and valvular disease), and 16 patients (24%) had manifesting coronary artery disease that was judged as the main cause of cardiac decompensation. Isolated hypertensive heart disease was present in 9 patients (13%), and valvular heart disease as the main cause was present in 10 patients (15%). Eight patients (12%) had dilated cardiomyopathy. The etiology of the cardiac failure could not be determined in 5 patients (7%).
BNP and Renal Function. The median creatinine value was 128 mmol/L (range, 38–1004 µmol/L). Eleven patients (16%) had stage I kidney function, 16 patients (24%) had stage II, 27 patients (40%) had stage III, 6 patients (9%) had stage IV, and 7 patients (10%) had stage V. Figure 1 shows the median BNP values in different kidney function stages. BNP increases with deteriorating kidney function. There was a weak but significant positive correlation between BNP and creatinine values (r=0.25; P=.04).
BNP and Doppler Echocardiographic Parameters. The median EF was 35% (range, 10%–75%), and there was no correlation with BNP values (r=−0.21; P=.09). The scattergram of BNP values and corresponding EFs is shown in Figure 2, and BNP values according to different categories of CHF are shown in Figure 3. Twenty-five patients (37%) had isolated systolic heart failure, 10 patients (15%) had isolated diastolic heart failure, and 32 patients (48%) had a combination of both. Median BNP values were slightly higher in systolic (3138 ng/L) than in diastolic (2846 ng/L) heart failure. Patients with a combination of both systolic and diastolic heart failure had the highest BNP values (3281 ng/L).
BNP levels did not correlate with IVS thickness (r=−0.02; P=.85), left ventricular mass index (r=0.1; P=.41), left ventricular end-diastolic volume index (r=0.15; P=.27), LA index (r=0.21; P=.13), LA surface index (r=0.21; P=.13), or RA surface index (r=0.27; P=.05). BNP values weakly correlated with LVEDD index (r=0.28; P=.02).
We evaluated clinical, Doppler echocardiographic, and laboratory parameters in patients with clinical symptoms and/or signs of CHF and excessively high BNP values (>2500 ng/L). The median age of our study population was 79 years (range, 53–90 years). The older age of this population is not surprising, because the incidence of heart failure increases with age (prevalence, 6%–10% in persons older than 65 years).19
Our results can also be seen in accordance with earlier studies that showed that age can markedly affect the circulating levels of BNP.20,21 In our population, age weakly correlated with BNP values. Besides age, renal function can also affect BNP values.22 The median creatinine value in our study was 129 µmol/L, and the majority of our patients (59%) had impaired renal function (stage III-V). BNP levels increased with worsening renal function.
The pathophysiologic mechanism of heart failure is thought to influence the levels of BNP. A subanalysis of the ProBNP Investigation of Dyspnea in the Emergency Department (PRIDE) study23 demonstrated that BNP and NT-proBNP levels were significantly lower in nonsystolic heart failure. In our study population, isolated diastolic heart failure was present in 15% of patients. The majority of our patients had a combination of systolic and diastolic heart failure (48%), and these patients had the highest BNP levels. The fact that a combination of systolic and diastolic heart failure produces the highest BNP values was also shown in earlier studies.24 There was no correlation of EF (a marker of systolic heart failure) with absolute BNP values. The other echocardiographic markers we analyzed (IVS, left ventricular mass index, left ventricular end-diastolic volume index, LA index, LA surface index, RA surface index, LVEDD index) did not correlate or only weakly correlated with BNP levels.
Because we did not obtain early transmitral velocity/tissue Doppler mitral annular velocity (E/Ea), we cannot compare BNP values and this parameter in our study. Earlier studies showed that E/Ea is more specific than BNP levels in detecting clinical CHF25 and high pulmonary capillary wedge pressure.26
Of interest, 24 patients (46%) had no dyspnea or dyspnea on heavy exertion (NYHA class I or II) while having excessively high BNP values. BNP values were similar in all NYHA groups (Figure 4). These results can be seen in accordance to Hogenhuis and colleagues,27 who did not find a correlation between BNP plasma levels and 6-minute walk test results,
Our subgroup of patients with very high BNP values shares the same characteristics (older age, combination of systolic and diastolic heart failure, renal failure) as the average patient presenting with symptoms and signs of CHF. That 46% had no dyspnea or NYHA class II dyspnea while having excessively high BNP seems surprising and important. While low BNP values in the presence of dyspnea have been well documented in clinical trials for the exclusion of CHF, the meaning of very high BNP values in patients with or without dyspnea has not been analyzed in a prospective trial.
Our results need reconfirmation and further exploration of the mechanisms in a prospective study with a larger sample size and inclusion of ambulatory patients.
Limitations. The retrospective character and the relatively small sample size are the limiting factors in this study. In addition, we analyzed a subselected group of patients with high BNP values, so it must be taken into account that correlation coefficients can be weaker in this study, and comparison with other studies should be done with caution.
The absence of correlation between BNP level and EF in the very high BNP value subset does not imply that there is not a correlation between BNP and EF in the broader population. We used NYHA functional classification to simply classify the extent of heart failure in our patients. Because no objective information was acquired (eg, exercise test results), it cannot be excluded that some of the patients of NYHA functional class II were actually more limited.