Author + information
- Received May 30, 2002
- Accepted July 12, 2002
- Published online November 20, 2002.
- Damien Logeart, MD*,* (, )
- Carole Saudubray, MD*,
- Pascale Beyne, MD†,
- Gabriel Thabut, MD‡,
- Pierre-Vladimir Ennezat, MD*,
- Christophe Chavelas, MD*,
- Caroline Zanker, MD§,
- Erik Bouvier, MD* and
- Alain Cohen Solal, MD, PhD*
- ↵*Reprint requests and correspondence:
Dr. Damien Logeart, Service de Cardiologie, Hôpital Beaujon, 100 Bd du Gal Leclerc, 92110 Clichy, France.
Objectives We compared the accuracy of B-type natriuretic peptide (BNP) assay with Doppler echocardiography for the diagnosis of decompensated congestive left-heart failure (CHF) in patients with acute dyspnea.
Background Both BNP and Doppler echocardiography have been described as relevant diagnostic tests for heart failure.
Methods One hundred sixty-three consecutive patients with severe dyspnea underwent BNP assay and Doppler echocardiogram on admission. The accuracy of the two methods for etiologic diagnosis was compared on the basis of the final diagnoses established by physicians who were blinded to the BNP and Doppler findings.
Results The final etiologic diagnosis was CHF in 115 patients. Twenty-four patients (15%) were misdiagnosed at admission. The BNP concentration was 1,022 ± 742 pg/ml in the CHF subgroup and 187 ± 158 pg/ml in the other patients (p < 0.01). A BNP cutoff of 300 pg/ml correctly classified 88% of the patients (odds ratio [OR] 85 [19 to 376], p < 0.0001), but a high negative predictive value (90%) was only obtained when the cutoff was lowered to 80 pg/ml. The etiologic value of BNP was low in patients with values between 80 and 300 pg/ml (OR 1.85 [0.4 to 7.8], p = 0.4) and also in patients who were studied very soon after onset of acute dyspnea. Among the 138 patients with assessable Doppler findings, a “restrictive” mitral inflow pattern had a diagnostic accuracy for CHF of 91% (OR 482 [77 to 3,011], p < 0.0001), regardless of the BNP level.
Conclusions Bedside BNP measurement and Doppler echocardiography are both useful for establishing the cause of acute dyspnea. However, Doppler analysis of the mitral inflow pattern was more accurate, particularly in patients with intermediate BNP levels or “flash” pulmonary edema.
Acute dyspnea is one of the main reasons for admission to emergency departments. Rapid diagnosis of decompensation of congestive left-heart failure (CHF) is important for prompt and appropriate treatment but is often difficult, especially in elderly or obese subjects, and when associated chronic pulmonary and cardiac diseases are present (1,2). The B-type natriuretic peptide (BNP) has been described as a powerful diagnostic test for heart failure (HF), even in the emergency setting (3,4). Ultra-rapid assays are now available and are suitable for bedside use (3–7). With the increasing use of rapid BNP assays it is important to determine diagnostic cutoff values. The BNP values increase with age (8)but also during right ventricle overload (9,10); the latter may be an important confounding factor, as the mean age of patients hospitalized for HF is about 70 to 75 years and chronic pulmonary disease is frequently associated (11). Finally, BNP levels are highly sensitive to drugs used in the treatment of HF (5). Bedside Doppler echocardiography is now routinely used in many intensive care units to manage patients with HF. It can be used to assess left ventricular (LV) systolic function and offers an indirect estimate of LV filling pressures (12,13). The value of bedside echocardiography for the etiologic diagnosis of dyspnea has not been adequately studied in the emergency setting.
We conducted a direct, prospective comparison of the role of bedside BNP and Doppler echocardiography in differentiating CHF from other causes in patients referred to a cardiopulmonary intensive care ward because of severe dyspnea.
The study protocol was approved by the local ethics committee. From June 1999 to June 2001, all patients presenting to the emergency room of our institution for acute severe dyspnea were enrolled, with the exception of patients with acute myocardial infarction, chest injury, or recent surgery. Among the 235 patients thus enrolled, 72 patients were subsequently excluded because of treatments (mechanical ventilation, diuretics, nitrates, or inotropic agents) started more than 2 h before arrival in our department, or because emergency echocardiography was not feasible. Finally, 163 patients were included.
At inclusion, the cause of acute dyspnea was assessed by the senior physician attending the patient on admission, on the basis of a physical examination, electrocardiogram, and chest X-ray examination.
Two groups served as controls. A cohort of 30 patients age more than 65 years and hospitalized in the Medicine Department, with no cardiac or pulmonary disease and no history of hypertension or diabetes mellitus, served as a control group for BNP values; Doppler echocardiogram was performed in every case, and controls with any abnormality were excluded. The second control group consisted of 30 outpatients with chronic stable HF (New York Heart Association functional class II or III).
At inclusion, blood was immediately collected in tubes containing potassium ethylenediamine-tetraacetic acid (1 mg/ml blood) and plasma was stored at −80°C for blinded BNP assay using the Triage B-Type Natriuretic Peptide test (Biosite Diagnostics, San Diego, California), a point-of-care method based on fluorescence immunoassay. It is designed for quantitative determination of BNP in whole blood and plasma, and it has been previously characterized (6,14). Each sample was analyzed in duplicate (or in triplicate when the first two values differed by more than 10%). The measurable range of BNP concentrations with the Triage assay is 5 to 1,300 pg/ml. Samples were diluted with normal plasma when values exceeded 1,300 pg/ml.
Bedside Doppler echocardiography
Doppler echocardiograms were obtained at the bedside within 60 min following inclusion (30 ± 9, 10 to 60 min), with a Hewlett-Packard Sonos 1500 (Andover, Massachusetts) machine equipped with a 2.5-MHz probe. The data were recorded on videotapes for later analysis by cardiologists experienced in echocardiography. Echocardiographic examination and recording took about 15 min and included two-dimensional and M-mode examination, pulsed Doppler analysis of mitral inflow, and continuous Doppler analysis of tricuspid regurgitation. Pulsed Doppler analysis of mitral inflow yielded three patterns: 1) an “impaired relaxation” pattern (E/A ratio <1), suggesting no increase in LV filling pressures; 2) a “restrictive” pattern (suggesting an increase in LV filling pressures) when the E/A ratio was >2, or between 1 and 2 with deceleration time of E-wave (DT) <130 ms, or DT <130 ms alone in case of atrial fibrillation; and 3) a “normal” or “normalized” pattern when the E/A ratio was between 1 and 2 with the DT >130 ms. Particular care was taken to position the Doppler sample volume between the mitral leaflet tips where flow velocity is highest. When E and A waves were fused, slight sinocarotid massage was used to reduce the heart rate and separate the two waves. The mitral Doppler pattern was unavailable in 25 patients because of poor echogenicity, tachycardia, permanent pacing, or mitral prosthesis. Systolic pulmonary arterial pressure was calculated from the velocity of tricuspid regurgitation, when present. The left ventricular ejection fraction (LVEF) was estimated mainly by visual inspection.
The final diagnosis was determined for each patient by two cardiologists and one pneumologist, who were blinded to the results of BNP assay and Doppler echocardiography obtained on admission. All the patients’ medical records were reviewed. Confirmation of CHF was based on the generally accepted Framingham criteria, with corroborative information including the hospital course (response to diuretics, vasodilators, inotropic agents, or hemodynamic monitoring) and results of further cardiac tests such as Doppler echocardiography, cardiac catheterization, radionuclide ventriculography, and pulmonary functional tests. Patients were finally classified as CHF or non–CHF. This latter diagnosis included acute pulmonary embolism and acute primary lung disorders, with or without underlying LV dysfunction but with no evidence of CHF.
Categorical data are presented as numbers (percent), and continuous data as means + SD. The Students ttest and the Fisher exact test were used as indicated. Group comparisons of BNP values were made using analysis of variance (ANOVA) with the Newman-Keuls post hoc test; p values < 0.05 were considered significant. Log-transformed BNP values were used in these analyses to reduce the effects of the skewed distribution of BNP concentrations. The sensitivity, specificity, accuracy, and negative and positive predictive values of BNP assay and Doppler echocardiography for CHF were compared. We also computed receiver operating characteristic (ROC) curves to determine optimal BNP cutoffs. We then used stepwise multivariate logistic regression to determine whether BNP assay and Doppler echocardiography added independent diagnostic information to commonly collected clinical variables. First, we developed a model based on historical, physical, and radiological variables. Next, BNP and Doppler findings were added separately to the model, and improvements in the degree of fit were assessed using the likelihood ratio test. The analyses were performed using STATA 6.0 software for Windows (Stata Corporation, College Station, Texas).
The final diagnosis was CHF in 115 cases and non-CHF in 48 cases. Diagnosis of CHF was due to coronary artery disease, hypertension, valve disease, and dilated cardiomyopathy in 49, 45, 20, and 42 cases, respectively. Non–CHF was due to acute pulmonary embolism in 11 cases and primary lung disorders in 37 cases (decompensated chronic obstructive pulmonary disease or severe emphysema, pneumonia, pulmonary fibrosis, or severe asthma in 27, 5, 3, and 2 cases, respectively). The interval between the onset of acute dyspnea and inclusion in the study ranged from 2 to 48 h (<6 h in 18 patients). Acute dyspnea was generally associated with severe clinical manifestations, as 90% of patients were admitted to the intensive care unit and 21% of these required mechanical ventilation. The main clinical characteristics of patients are shown in Table 1, according to the final diagnosis. The initial diagnosis was wrong in 24 patients (15%): CHF was wrongly diagnosed in 13 patients and missed in 11 patients. In addition, 22 patients (13%) had a “doubtful” initial diagnosis.
BNP and Doppler echocardiographic findings
Figure 1Ashows box plots of log BNP values in each final diagnostic group. Mean BNP concentrations were 1,022 ± 742 pg/ml in the acute CHF group, 187 ± 158 pg/ml in the non–CHF group, 44 ± 39 pg/ml in the control group, and 144 ± 142 pg/ml in the stable chronic HF group. The intergroup difference was significant (p < 0.001, ANOVA); both the mean and the median were significantly higher in the patients with CHF (p < 0.001). No significant difference was observed between the other groups. Figure 1Bshows log BNP concentrations in various subgroups. The BNP values in the non–CHF group were significantly higher than in the control group without heart or lung disease (p < 0.05). As shown in Figure 1B, BNP concentrations were similar in patients with normal and abnormal LVEF in the CHF group. In contrast, BNP values were significantly higher when LVEF was abnormal in the non–CHF group and also in patients with pulmonary embolism (223 ± 99 vs. 138 ± 102 pg/ml and 294 ± 245 vs. 155 ± 106 pg/ml, respectively, p < 0.05).
The main Doppler and echocardiographic findings are presented in Table 2. Systolic LV dysfunction, defined by an LVEF of <0.45, was found in 75 patients with CHF (65%) and in 7 patients with other causes of dyspnea (14%). The E/A ratio and the DT were significantly higher and lower, respectively, in the patients with CHF. Finally, a “restrictive” mitral pattern was observed in 85 (89%) of the patients with CHF and in only 3 (7%) of the patients with other etiologic diagnoses. Systolic pulmonary artery pressures were not significantly different among the various groups.
Etiologic diagnosis of acute dyspnea
None of the clinical, electrocardiographic, or radiographic variables studied had a diagnostic accuracy of more than 80% for CHF. Table 3shows the sensitivity, specificity, positive and negative predictive values, and accuracy of the most relevant BNP cutoff values, LVEF values, and mitral Doppler patterns. The most accurate BNP cutoff was 300 pg/ml (88% accuracy); this had a satisfactory positive predictive value (94%), but a negative predictive value of only 75%. The best negative predictive value (93%) was obtained with a BNP cutoff of 80 pg/ml. Figure 2shows the ROC curve illustrating these BNP cutoff values; the area under the ROC curve was 0.93. Single LVEF values were poorly predictive of the final diagnosis: using LVEF <0.45 as the criterion for CHF, only 71% patients were correctly classified. The best diagnostic performance was obtained with Doppler analysis of mitral inflow: the “restrictive” pattern had an accuracy of 91%. This result was similar whatever the LVEF. Multivariate analysis demonstrated the independent predictive value of five clinical and radiographic parameters for CHF (Table 4). Both BNP and the Doppler mitral pattern added significant incremental predictive diagnostic value to the clinical variables, but BNP only predicted CHF when levels were above 300 pg/ml (odds ratio [OR] 85, 95% confidence interval [CI] [19.2 to 376.4], p < 0.0001). The “restrictive” mitral pattern was strongly predictive of CHF (OR 481.7, 95%CI [77 to 3,011], p < 0.0001).
Doppler echocardiography in patients with inconclusive BNP values
Relevant negative and positive predictive BNP values were scattered between 80 and 300 pg/ml, respectively; within this range, BNP was poorly predictive of the final diagnosis (OR 1.85, 95% CI [0.4 to 7.8], p = 0.4) (Table 4) and added nothing to the clinical diagnosis. Among the 40 (25%) patients with such intermediate BNP levels (Fig. 3), 12 were misdiagnosed by the attending physician at admission, and only 6, 3, and 6 patients were correctly identified by BNP cutoff values of 100, 200, and 300 pg/ml, respectively. In contrast, a “restrictive” Doppler mitral pattern correctly classified 33 of these patients (4 had no assessable Doppler analysis of mitral inflow) and corrected 10 of the 12 clinical misdiagnoses. The BNP assay was also of limited value when blood was sampled very soon after the onset of acute dyspnea. Among the 14 patients in whom blood was sampled <4 h after onset, and who had a final diagnosis of CHF, admission BNP levels were below 100 pg/ml in 4 patients and between 100 and 300 pg/ml in 6, whereas the Doppler mitral pattern was “restrictive” in 11 of these patients.
This is the first direct comparison of the etiologic value of BNP assay and Doppler echocardiography in patients presenting with acute dyspnea. Our results show that: 1) both bedside BNP assay and Doppler echocardiography add important diagnostic information to clinical findings in emergency patients with acute dyspnea; and 2) the diagnostic value of BNP is poor at levels between 80 and 300 pg/ml and also in patients with “flash” pulmonary edema, whereas the Doppler mitral inflow pattern is highly relevant.
Rapid etiologic diagnosis of acute dyspnea is often difficult (1), because of the nonspecific nature of signs (especially in elderly and obese patients and subjects with a previous history of cardiac or pulmonary disease), together with the poor accuracy of electrocardiograms and chest radiography, and frequent early and blind prescription of nitrates, diuretics, aerosols, and antibiotics, often in combination. In our study, 15% of patients were misdiagnosed at admission, a rate close to that observed in another recent study of emergency patients with dyspnea (4).
Natriuretic peptide concentrations are powerful biochemical markers of HF (3,15)and they correlate well with invasively measured LV filling pressures (16,17). Bedside BNP assay has been facilitated by the advent of new rapid tests such as the Triage kit used here. The results obtained with these assays are close to those given by the time-consuming radioimmunologic reference method (14). Dao et al. (4)recently reported the good diagnostic value of a rapid point-of-care BNP test in emergency patients presenting with dyspnea. They found that a BNP concentration of 80 pg/ml in whole blood had a negative predictive value of 98% and an accuracy of 95% for CHF, and they concluded that BNP might be the test of choice for the differential etiologic diagnosis of dyspnea. Nevertheless, Morrison et al. (6)reported a wider dispersion of BNP levels in patients with pulmonary diseases, and especially in patients with acute pulmonary embolism.
Using the same assay, we confirm that BNP measurement accurately discriminates between acute dyspnea due to CHF and to other causes. However, while a BNP cutoff of 80 pg/ml had a strong negative predictive value (93%), acceptable sensitivity and specificity, and a good positive predictive value were only obtained with a cutoff of 300 pg/ml. The diagnostic value of BNP fell at values between 80 and 300 pg/ml (i.e., in 26% of our patients). Many of these patients had a final diagnosis of non–CHF; underlying LV dysfunction and/or severe pulmonary disease or pulmonary embolism could explain the high BNP levels found in some patients with acute dyspnea due to causes other than CHF. Indeed, BNP concentrations can be elevated in patients with compensated LV dysfunction or right ventricular overload (9). This difference in cutoff values with previous studies may be due to differences in the study populations. Our population comprised patients with severe dyspnea admitted to an intensive care unit, whereas a number of the patients in the San Diego group study were not admitted (4,6).
In contrast to our study, the final diagnoses of CHF in these studies included not only LV HF but also pedal edema, due to pulmonale cor, for example. Our study also shows for the first time that low BNP values can be observed in patients with CHF, especially those with “flash” pulmonary edema and when blood is assayed very soon after the onset of dyspnea. These “false-negative” results may be explained by the lag-time between acute LV overload and BNP secretion into the circulation (18), and also by the absence of LV systolic dysfunction.
Doppler echocardiography, and especially Doppler analysis of the mitral inflow pattern, was highly accurate in distinguishing between acute dyspnea due to CHF and forms due to other causes. To our knowledge, ours is the first study to show such diagnostic value in the emergency setting. Several studies have shown strong correlations of the E/A ratio, DT, and the mitral inflow pattern with LV end diastolic pressure and pulmonary capillary wedge pressure at rest (12,13). However, these relationships were mainly observed in patients with LV systolic dysfunction. In our study, Doppler analysis of the mitral inflow pattern remained relevant in patients with a preserved LVEF, possibly because of the patients’ advanced age, the acute setting, and the use of simple criteria (“restrictive” vs. “nonrestrictive” patterns). More refined Doppler analysis can be obtained with pulsed Doppler analysis of pulmonary venous flow, color M-mode analysis of LV inflow, and Doppler tissue imaging of the mitral annulus (19,20). However, these measurements are time-consuming and require a well-trained operator, making them poorly suited to the emergency setting and dyspneic patients: in contrast to Doppler analysis of the mitral inflow pattern, pulmonary venous flow was only assessable in about 25% of our patients (data not shown). It should be noted that our Doppler studies were performed in severely dyspneic patients in whom treatment had been started <2 h previously. Subsequent identical Doppler examination at the echo lab, after diuretic and vasodilator therapy, and clinical improvement, often showed markedly different findings with, for example, an E/A ratio <1 in most of the patients with a preserved LVEF, as well as in elderly patients without CHF. Finally, only four patients with CHF had an E/A ratio <1 at admission; this could be explained by abrupt changes in load during aggressive initiation of treatment or by marked, isolated impairment of relaxation (21).
The relatively poor diagnostic value of the LVEF compared to Doppler mitral analysis is not surprising: CHF and pulmonary edema are often found in patients (who are frequently old) with a preserved LVEF (22); this was the case in 35% of our patients. Finally, echocardiographic machines usually have limited availability in the emergency care setting, but the emergence of portable ultrasound imagers (23), including Doppler devices, could permit their more widespread use.
This study included patients with particularly severe dyspnea; whether or not our results can be extrapolated to patients with milder dyspnea remains to be shown. In addition, CHF was the most frequent final diagnosis, and patients with non–CHF diagnoses had severe lung disease, with probably frequent chronic and/or acute pulmonale cor. This could partly explain the relatively high optimal BNP cutoff of 300 pg/ml obtained here, as the prevalence of right ventricular or LV failure was lower in other studies. Finally, the high diagnostic value of Doppler echocardiography found here may be markedly decreased if the examination is performed later after arrival or initiation of treatment.
This study suggests that both BNP assay and Doppler echocardiography can be used for the diagnosis of CHF in acutely dyspneic patients. Bedside BNP assay is simple and repeatable. Doppler analysis of mitral inflow is at least as effective as BNP assay in this setting, but was not feasible in 15% of our patients. We propose a diagnostic algorithm (Fig. 4), in which bedside BNP assay is first used to rule out CHF (values <80 pg/ml) or to confirm CHF (values >300 pg/ml). In patients with “intermediate” BNP levels (i.e., between 80 and 300 pg/ml), Doppler echocardiography is used to confirm or rule out CHF.
The BNP assays were gifts from Biosite Diagnostics (San Diego, California). We are grateful to David Young for his help in restyling the manuscript.
☆ This work was supported in part by a grant from the French Federation of Cardiology, Paris, France.
Gottlieb C. Friesinger II, MD, is the guest editor.
- analysis of variance
- B-type natriuretic peptide
- decompensated congestive left-heart failure
- confidence interval
- deceleration time of the mitral E-wave
- heart failure
- left ventricular
- left ventricular ejection fraction
- odds ratio
- receiver operating characteristic
- Received May 30, 2002.
- Accepted July 12, 2002.
- American College of Cardiology Foundation
- Remes J.,
- Miettinen H.,
- Reunanen A.,
- Pyoorala K.
- Dao Q.,
- Krishnaswamy P.,
- Kazanegra R.,
- et al.
- Cheng V.L.,
- Krishnaswamy P.,
- Kazanegra R.,
- Garcia A.,
- Gardetto N.,
- Maisel A.S.
- Morrison K.L.,
- Harrison A.,
- Krishnaswamy P.,
- Kazanegra R.,
- Clopton P.,
- Maisel A.S.
- Nagaya N.,
- Nishikimi T.,
- Okano Y.,
- et al.
- Bellotti P.,
- Badano L.P.,
- Acquarone N.,
- et al.
- Nishimura R.A.,
- Tajik A.J.
- Morita E.,
- Yasue H.,
- Yoshimura M.,
- et al.
- Garcia M.J.,
- Thomas J.D.,
- Klein A.L.
- Ommen S.R.,
- Nishimura R.A.,
- Appleton C.P.,
- et al.
- Vasan S.,
- Larson M.G.,
- Benjamin E.J.,
- et al.
- Roelandt J.R.T.C.