Alan Maisel, MD, Veterans Affairs Medical Center, Cardiology 111-A, 3350 La Jolla Village Drive, San Diego, CA 92161 E-mail: email@example.com
This study aimed to assess whether the combination of a B-type natriuretic peptide (BNP) level with various noninvasive hemodynamic parameters can help physicians more quickly and accurately diagnose congestive heart failure and determine the type of left ventricular dysfunction present in patients presenting to the emergency department with dyspnea. Subjects were 98 men (aged 64.57±1.23 years) that presented to the VA San Diego Healthcare System. Hemodynamic parameters were measured using impedance cardiography, and BNP levels were quantified using a rapid immunoassay. All patients with a BNP <100 pg/mL (n=37) had no evidence of congestive heart failure 97% of the time. In those with a BNP >100 pg/mL (601 ±55 pg/mL; n=61), a cardiac index of 2.6 L/min/m2 is 65% sensitive and 88% specific in determining systolic dysfunction. In patients with a BNP >100 pg/mL, a multivariate model consisting of noninvasive hemodynamic measurements was able to predict cardiac deaths, readmissions, and emergency department visits within 90 days with 83% accuracy. The authors conclude that, in patients presenting to an emergency department with dyspnea, the addition of impedance cardiography measurements to BNP level measurements will more effectively diagnose congestive heart failure and determine both the type of heart dysfunction and the severity of illness.
Correctly diagnosing congestive heart failure (CHF) in patients presenting with acute dyspnea will not only allow a patient to accrue the benefits of improved survival and increased well-being secondary to taking medications such as angiotensin-converting enzyme inhibitors and β blockers,1 but will also help physicians avoid a misdiagnosis that could place a patient at risk for increased morbidity and mortality.2 In addition, delineation of whether impaired vs. preserved systolic function is present should allow further tailoring of treatment according to the main pathophysiologic disturbances, as well as allowing the identification of a group of patients with a poorer prognosis.
B-type natriuretic peptide (BNP), which is secreted primarily from the left ventricle in response to volume expansion and pressure overload,3 has been useful in diagnosing CHF in patients presenting to the emergency department (ED) with dyspnea.4,5 Though highly prognostic in this setting, BNP levels alone cannot differentiate between patients with and without systolic dysfunction.
Impedance cardiography (ICG) estimates stroke volume by measuring changes in thoracic electrical impedance over time during the cardiac cycle. A low-energy, high-frequency alternating current is introduced into the thorax via sensors placed on both sides of the neck and lower thorax. Blood is the most conductive substance in the thorax, therefore the pulsatile changes of aortic blood volume correlate to changes in resistance to the electrical current (impedance) over time. By filtering out any effects of respiration on the electrical impedance signal, measured changes in impedance are solely attributed to the cardiac component. Hemodynamic measurements obtained by ICG have previously been shown to be highly reproducible. Verhoeve et al.6 demonstrated both intraday (cardiac index [CI], r=0.96; stroke index [SI], r=0.97; thoracic fluid content [TFC], r=0.99) and interday reproducibility in 96 cardiac rehabilitation patients. Greenberg et al.7 had similar findings in clinically stable heart failure patients.
The purpose of this study was to assess whether the addition of hemodynamic measurements to BNP levels could improve physicians' ability to diagnose CHF and determine a patient's prognosis in the ED, as well as help delineate the etiology of failure (preserved vs. systolic dysfunction).
Study Population. The study was approved by the Institutional Review Board of the University of California and conducted between June 2001 and April 2002. A convenience sample of 98 patients at the VA San Diego Healthcare System was enrolled in the study. Eligible patients presented to the urgent care center or ED with acute shortness of breath. Patients under age 18 years or those receiving continuous inotrope infusions on an outpatient basis, as well as the patients whose dyspnea was clearly not secondary to CHF (trauma or cardiac tamponade) were excluded.
Written informed consent was obtained from each patient. Subsequently, the patient had a blood sample collected to determine its concentration of BNP and noninvasive hemodynamic monitoring was performed via ICG to determine the patient's CI, SI, TFC, acceleration index (ACI), and left cardiac work index (LCWI). Data pertaining to the patient's physical exam findings, medical history, and workup were also recorded. ED physicians were blinded to the ICG measurements when determining each patient's diagnosis and appropriate course of treatment. Upon request, physicians were told if the BNP level was >100 pg/mL. Retrospective chart reviews were performed at 90 days to determine if the patients had reached any of the following end points: cardiac death, readmission, or visit to the ED within 90 days.
Diagnosis. To determine the diagnosis for the purpose of this study, a cardiologist blinded to the hemodynamic parameters reviewed the patient's medical record and independently classified the diagnosis as acute dyspnea either due to CHF or to other causes. The cardiologist had access to the same information the ED physician did, as well as the results of any further workup done after the patient left the ED. Patients diagnosed with CHF were further classified into two groups: those with impaired left ventricular systolic function and those with preserved left ventricular function based on echocardiography performed within the preceding or subsequent 6 months to their visit to the ED. Patients with an ejection fraction of ≤45% were defined as having systolic dysfunction, and those with an ejection fraction >45% were classified as having preserved systolic function.
Measurement of BNP. Blood samples were collected in 5-mL tubes containing potassium ethylenediaminetetraacetic acid (1 mg/mL) and BNP was measured using the Triage BNP (Biosite Inc., San Diego, CA) assay. The Triage BNP test is a rapid, fluorescence immunoassay that quantitatively measures BNP in whole blood and plasma specimens. When possible, BNP levels were measured in whole blood within 4 hours. When this was not possible, samples were spun down, frozen, and then analyzed 1–2 days later. Details regarding precision, interference, and stability have been described previously.8
Noninvasive Hemodynamic Monitoring. ICG measurements were performed using the BioZ ICG Monitor (CardioDynamics, San Diego, CA). Two sets of sensor patches were placed on the patient, one pair placed 180° apart at the root of the neck, and the other pair placed at the midaxillary line at the level of the xiphoid process. A high-frequency, low-amplitude alternating current was introduced and the measurements were electronically recorded for each patient. After all data were collected, a proprietary algorithm (ZMARC; CardioDynamics, San Diego, CA) was utilized to calculate continuous measurements of stroke volume and a cascade of measured and derived hemodynamic parameters (Table I).
Table I. Parameters of Impedance Cardiography
Cardiac output normalized for body surface area
An index of the peak acceleration of blood flow in the aorta
Men: 70–150/100 sec Women: 90–170/100 sec
Thoracic fluid content
The net content of the fluid in the thorax
Men: 30–50 k/W Women: 21–37 k/W
Left cardiac work index
The amount of work the left ventricle performs each minute when ejecting blood
Statistics. Group comparisons of BNP and CI values were made using the Mann-Whitney test. Results are expressed as mean ± standard error of the mean. The diagnostic utility of BNP and CI independently in determining CHF was compared through the use of receiver operating characteristic (ROC) curves.
The same statistical approach was used to assess the utility of BNP and CI independently in identifying the presence or absence of systolic heart dysfunction in patients with BNP levels >100 pg/mL. Results are expressed in terms of the area under the curve (AUC) and 95% confidence intervals are shown for this area. Sensitivity, specificity, and accuracy were computed for BNP and CI by use of a selection of possible cut points.
Logistic regression was used in the multivariate approach for evaluating the ability of TFC, ACI, LCWI to predict cardiac deaths, readmissions, and ED visits within 90 days in patients with BNP levels >100 pg/mL.
The characteristics of the 98 patients included in this study are shown in Table II and the signs, symptoms, and results of any diagnostic studies performed are displayed in Table III. ED diagnosis, disposition, and 90-day end points for the patients enrolled in this study are shown in Table IV.
Table II. Characteristics of the 98 Patients
%of Total Patients
Congestive heart failure
Chronic obstructive pulmonary disease
Coronary artery disease/angina
Valvular heart disease
Any cardiac surgery
Table III. Patient Symptoms, Physical Exam Findings, and Treatments/Studies
%of Total Patients
Recent weight gain
Paroxysmal nocturnal dyspnea
JVP >6 cm
Abnormal heart sounds
JVP=jugular venous pressure; PMI=point of maximal intensity
Table IV. Emergency Department (ED) Diagnosis, Disposition, and 90-Day Follow-Up
Patients diagnosed with CHF (n=57) had a mean BNP concentration of 630±57 pg/mL while the non-CHF group (n=41) had a mean BNP concentration of 44±9 pg/mL (p<0.001). Within the CHF group, patients with systolic dysfunction (n=33) had a mean BNP concentration of 698±73 pg/mL and in those with preserved systolic function (n=24) the mean BNP concentration was 535±88 pg/mL (p=0.13).
Patients diagnosed with CHF had a mean CI of 2.71±0.08 L/min/m2 while the non-CHF group had a mean CI of 2.88±0.12 L/min/m2 (Figure 1; p=0.19). In the CHF group, the patients with systolic dysfunction had a mean CI of 2.54±0.11 L/min/m2 and those with preserved systolic function had a mean CI of 2.94±0.10 L/min/m2 (p<0.003).
An ROC curve showing the sensitivity and specificity of using BNP values alone vs. a CI measurement alone in the diagnosis of CHF is shown in Figure 2. BNP alone was a sensitive and specific test for the presence of CHF (AUC=0.979; p<0.001). CI, when used alone, was neither sensitive nor specific for diagnosing CHF (AUC=0.58; p=0.196).
A second ROC curve, in which the sensitivity and specificity of using BNP values alone vs. a CI measurement alone in determining the presence vs. absence of systolic dysfunction, is also shown in Figure 2. BNP values used alone were not helpful in determining the presence of systolic dysfunction (n=61; AUC=0.64; p=0.78). In patients with BNP levels >100 pg/mL, the addition of a CI measurement markedly improved the ability to differentiate patients with preserved systolic function from those with impaired systolic function (AUC=0.735; p<0.003). A cut-point of 2.6 L/min/m2 for CI was 65% sensitive, 88% specific, and 75% accurate in determining impaired systolic function in patients with CHF (Table V).
Table V. Cardiac Index (CI) in Determining Systolic Dysfunction in Patients Wity B-Type Natriuretic Peptide >100
Positive Predictive Value (%)
Negative Predictive Value (%)
Multivariate analysis was performed for all variables applicable to the diagnosis of CHF, including various ICG measurements obtained upon patient enrollment. These were analyzed for their potential ability to predict the severity of illness in patients with BNP >100 pg/mL. End points used to define severity of CHF included cardiac death, readmission, and an ED visit within 90 days from enrollment in the study. The best clinical predictors of more severe illness, as defined above, were LCWI, ACI and TFC, and these together were able to predict increased severity of disease with 89% specificity and 80% accuracy in patients with BNP >100 pg/mL.
Establishing CHF as the cause of dyspnea in patients presenting to the ED is an important, but not always easy, task9–16 and the ability to predict a patient's prognosis in the ED would aid immeasurably in triaging and treating patients appropriately. Although echocardiography is considered the gold standard in diagnosing left ventricular dysfunction, it is costly and has limited availability in urgent-care settings. Also, echocardiography may not always reflect an accurate condition.17
BNP is a 32 aa polypeptide containing a 17 aa ring structure common to all natriuretic peptides.18,19 Ventricular cells secrete BNP in response to the high ventricular filling pressures.3 The point-of-care BNP whole blood test has been approved and shown to be effective in diagnosing dyspneic patients in an ED.4,5,20,21 In the multi-center Breathing Not Properly trial,20 BNP levels were shown to be 90% sensitive and 76% specific at a cutoff of 100 pg/mL for the diagnosis of CHF. Furthermore, BNP outscored the Framingham criteria for CHF, clinical judgment by the ED physician, and chest x-ray in this setting.
BNP levels measured in the ED among patients presenting with dyspnea have also been shown to be predictive of subsequent adverse cardiac outcomes, including recurrent CHF and death from CHF.22 The high negative predictive value of BNP that may be especially useful in settings with acute dyspnea was demonstrated in this, as well as in other studies.23 Plasma BNP, measured at initial presentation, provides prognostic information in patients with chronic heart failure, even in those receiving therapy with a β blocker and an angiotensin-converting enzyme inhibitor,24 and in those with asymptomatic or minimally symptomatic left ventricular dysfunction.25–28 If there is a limitation of BNP levels in this setting, it is the lower specificity, especially in older patients, and an inability to separate patients with CHF and preserved systolic function from those with CHF and systolic dysfunction.
ICG has also been shown to aid in the diagnosis and prognosis determination of patients with CHF. These tests appear to more accurately estimate hemodynamic measurements than clinical estimates,29 and indices have been established through various studies comparing estimates of cardiac output by ICG to thermodilution or the gold standard of direct Fick.7,30,31 Marrocco et al.32 showed significant differences in mean measurements of CI, SI, ejection fraction, and LCWI in patients with cardiogenic vs. pulmonary causes for dyspnea. Sensitivity associated with use of the single parameter CI for this determination was 100%, although specificity was low (48%). In a similar study of 38 patients, Springfield et al.33 used a combination of parameters (CI and systolic time ratio) and improved overall accuracy, demonstrating sensitivity (92%) and specificity (88%) for cardiac vs. noncardiac etiology of dyspnea.
ICG has also been shown to have some prognostic use in the ED setting. In 50 heart failure patients presenting to an urban ED setting, an initial low CI (CI <2.2 L/min/m2) or a low stroke volume (SI <60 mL) was associated with an increased length of stay and increased hospital charges when compared with CHF patients with normal central hemodynamic values.34 Another study of 58 CHF patients presenting to an ED demonstrated that patients with an increased CI after 60 minutes of resuscitation trended toward both a reduced hospital stay and hospital charges compared with patients without an improved CI.35
Use of ICG Along With BNP in Dyspnea. This is the first study to suggest that two noninvasive point-of-care tests, BNP and ICG, might provide complementary and additive information in patients presenting with dyspnea. While BNP levels by themselves can accurately diagnose CHF, BNP could not differentiate patients with impaired vs. preserved left ventricular function. However, in patients with BNP levels >100 pg/mL, CI measurements could differentiate the two different CHF states. A CI of 2.6 L/min/m2 was 74% accurate in diagnosing systolic dysfunction in patients with a BNP >100 pg/mL.
The importance of differentiating patients with impaired vs. preserved left ventricular function cannot be overstated, especially in the ED setting. Patients presenting with preserved systolic function often have CHF precipitated by ischemia, abnormally high blood pressure, or atrial fibrillation, all of which would receive different workups and treatment than is used for patients with systolic dysfunction. Patients with systolic dysfunction are more likely to be admitted to the hospital and to require parenteral vasodilator and/or inotropic therapy.
Additionally, the use of BNP and ICG in the ED setting might help physicians profile higher risk patients. The results of this study demonstrated that low LCWI in patients with BNP >100 pg/mL was 89% specific and 71% accurate in predicting subsequent adverse outcomes over a 6-month follow-up period. A multivariate model using low LCWI, low ACI, and high TFC in patients with abnormal BNP levels showed an 89% specificity in predicting death, ED visits, or hospital readmission due to CHF or other cardiac causes over a 90-day follow-up period. Thus, the combination of a high BNP and a low LCWI may delineate patients with more severe disease, allowing for a more expedient triage and workup.
Study Limitations. This was an observational study, performed in a convenience sample of strictly male patients at a VA San Diego Healthcare System. Thus the results of this study should be confirmed in a broader population base. Also, the patients studied were those who came to the ED with current symptoms and therefore these results may not be easily applicable to the outpatient setting.
In patients presenting to the ED with dyspnea, the addition of noninvasive hemodynamic measurements to a BNP level give evaluating physicians a higher level of differentiation and prognostic capability. In the future, the combination of BNP and ICG may provide a surrogate end point for the evaluation of various treatments of heart failure. Falling BNP levels with treatment is associated with falling wedge pressures, a lower readmission rate to the hospital, and a better prognosis.21,27 Rising cardiac outputs with treatment should provide additional information regarding drug efficacy and patient outcomes.
Acknowledgments and Disclosures: The authors thank the physicians and nursing staff working in the emergency department of the Veterans Affairs Medical Center for their cooperation and support. Kits for BNP tests were supplied by Biosite Inc., San Diego, CA. BioZ device was loaned from CardioDynamics, San Diego, CA, for ICG monitoring studies. Dr. Maisel is a consultant for Biosite, Inc.