Intensive Care Med (2004) 30:2230–2236DOI 10.1007/s00134-004-2469-0 O R I G I N A L

Patrick RayMartine ArthaudYannick LefortSophie BirolleauCatherine BeigelmanBruno Riouthe EPIDASA Study Group

Usefulness of B-type natriuretic peptidein elderly patients with acute dyspnea

Received: 7 December 2003Accepted: 15 September 2004Published online: 22 October 2004� Springer-Verlag 2004

Presented, in part, at the 30th annual con-gress of the Soci�t� de R�animation deLangue Fran�aise, Paris, January 2002(abstract no. SP 94).

P. Ray ()) · B. RiouDepartment of Emergency Medicineand Surgery,Centre Hospitalo-Universitaire (CHU)Piti�-SalpÞtri�re Hospital,Assistance-Publique H�pitauxde Paris (AP-HP),Universit� Pierre et Marie Curie,47 boulevard de l’H�pital, 75013 Paris,Francee-mail: patrick.ray@psl.ap-hop-paris.frTel.: +33-1-42177249Fax: +33-1-42177264

M. ArthaudLaboratory of Emergency Biology,Centre Hospitalo-Universitaire (CHU)Piti�-SalpÞtri�re Hospital,Assistance-Publique H�pitauxde Paris (AP-HP),Universit� Pierre et Marie Curie,47 boulevard de l’H�pital, 75013 Paris,France

Y. Lefort · S. BirolleauDepartment of Pneumology,Centre Hospitalo-Universitaire (CHU)Piti�-SalpÞtri�re Hospital,Assistance-Publique H�pitauxde Paris (AP-HP),Universit� Pierre et Marie Curie,47 boulevard de l’H�pital, 75013 Paris,France

C. BeigelmanDepartment of Radiology,Centre Hospitalo-Universitaire (CHU)Piti�-SalpÞtri�re Hospital,Assistance-Publique H�pitauxde Paris (AP-HP),Universit� Pierre et Marie Curie,47 Boulevard de l’H�pital, 75013 Paris,France

Abstract Objective: Differentiatingcardiogenic pulmonary edema(CPE) from respiratory causes ofdyspnea is particularly difficult inelderly patients. The aim of our studywas to evaluate B-type natriureticpeptide (BNP) in patients older than65 years presenting with acute dys-pnea. Design: Prospective study.Setting: Medical emergency depart-ment of a 2000-bed urban teachinghospital. Patients: Patients aged over65 years presenting with acute dys-pnea and a respiratory rate more than25/min or a PaO2 below 70 mmHg,SpO2 less than 92%, PaCO2 higher

than 45 mmHg with pH less than7.35, were included. BNP levels,measured blind at admission werecompared with the final diagnosis(CPE or no CPE) as defined byexperts. Intervention: None.Measurements and results: Threehundred eight patients (mean age of80 years) were enrolled in the study.The median BNP was 575 pg/ml[95% confidence interval (CI): 410–898] in the CPE group (n=141) ver-sus 75 pg/ml (95% CI: 59–98) in theno CPE group (n=167) (p<0.001).The best threshold value of BNP was250 pg/ml, with a sensitivity andspecificity for CPE of 0.78 (95% CI:0.71–0.84) and 0.90 (95% CI: 0.84–0.93), respectively. The area underthe ROC curve was 0.874€0.081(p<0.001). The accuracy of BNP-assisted diagnosis was higher thanthat of the emergency physician(0.84 versus 0.77, p<0.05).Conclusion: Analysis of BNP isuseful in elderly patients with acutedyspnea, but the threshold value ishigher than that previously deter-mined.

Keywords Acute dyspnea · Elderlypatients · Emergency department ·Cardiogenic pulmonary edema ·B-type natriuretic peptide

2231

Introduction

Heart failure is one of the major causes of admission tohospital in elderly patients and is associated with highmorbidity and mortality [1, 2]. Because of atypical clin-ical presentations such as cardiac asthma, differentiatingcardiogenic pulmonary edema (CPE) from respiratorycauses is frequently difficult in elderly patients [2, 3].However, a pre-hospital study demonstrated that theearlier the specific medications for CPE were adminis-tered the better was the prognosis [3]. Thus, numerousmethods have been proposed to differentiate respiratoryfrom cardiac dyspnea. Neither capnography, peak expi-ratory flow nor the dyspnea index have been shown to beof value [4–6]. Doppler echocardiography is time-con-suming, expensive, rarely available in the emergencydepartment and may not easily differentiate true diastolicdysfunction from physiologic aging modifications [7]. B-type natriuretic peptide (BNP) is a natriuretic factor se-creted by the cardiac myocytes, as a response to ven-tricular overload and activation of the renin-angiotensin-aldosterone system [8]. Previous studies have demon-strated, in a non-elderly population, that blood BNP levelis an excellent aid in diagnosing heart failure in acutedyspnea [9–13]. The BNP level is higher in elderly pa-tients [14]. So far no studies have specifically evaluatedBNP in acute dyspnea in elderly patients, where the assaycould be particularly valuable in diagnosing heart failureas the cause of dyspnea. Therefore, the aim of this studywas to evaluate the usefulness of BNP analysis in patientsolder than 65 years with acute dyspnea.

Patients and methods

This work was part of an epidemiological study of acutedyspnea in elderly patients (the EPIDASA study) [16].This was a single-center prospective study which tookplace between February 2001 and August 2002. The studywas approved by the ethics committee (CCPPRB Piti�-SalpÞtri�re Hospital, Paris). Waived informed consentwas authorized because no additional blood sampling wasrequired and routine care of the patient was not modified.

Patients

The criteria for inclusion in the study were: (1) presen-tation at the Emergency Department, (2) age over65 years and (3) acute dyspnea of less than 2-weeks du-ration as the prominent complaint. In addition, one of thefollowing criteria had to be fulfilled: (1) a respiratory ratehigher than 25/min; (2) PaO2 less than 70 mmHg; (3)PaCO2 more than 45 mmHg with pH below 7.35 or (4)peripheral oxygen saturation (SpO2) less than 92%. Be-cause atypical presentation of heart failure is frequent in

elderly patients, there were no exclusion criteria [2].Medical care, including emergency treatment, and ad-mission were decided by the emergency physician ac-cording to normal practice. Suspected diagnoses by theemergency physicians were based upon physical exami-nation and the results of routine blood tests, electrocar-diogram, chest X-ray and blood gas analysis. Creatinineclearance was estimated using the Cockcroft formula.

B-type natriuretic peptide assay

At admission a small sample of whole blood (250 �l) wascollected into tubes containing potassium ethylenedi-aminetetraacetic acid (EDTA). BNP assay was performedusing a rapid test (Triage BNP, Biosite, San Diego, CA;BMD, Marne la Vall�e, France), described in detailelsewhere [11, 13, 14, 15]. This method correlates a flu-orescence measurement to the BNP concentration by us-ing an internal calibration curve. Values may vary from5 to 1300 pg/ml. BNP levels were measured after theplasma sample had been centrifuged and plasma re-moved, divided into aliquots and frozen for several days(<2 weeks) at �30�C. The biochemist was unaware of thediagnosis of acute dyspnea and, conversely, the emer-gency physician and the expert panel were unaware of theBNP levels.

Confirmation of the diagnosis

For every patient the medical history (including previouscardiac and respiratory diseases, physical examinationfindings, arterial blood gas analysis while breathing roomair, 12-lead electrocardiogram and chest X-ray) and otherthe usual clinical tests were recorded. The diagnosissuspected by the emergency physician was classified intoone of four groups according to empirical probability: ascertain or very likely (high probability), possible, unlikely(low probability) and absent. The length of hospitalizationand the in-hospital mortality were reported.

Following the normal policy in our institution for acutedyspnea in patients older than 65 years, a thoracic highresolution computed tomography (HRCT) without con-trast iodine medium, a transthoracic Doppler echocardi-ography and a pulmonary function test or bedside spi-rometry were performed as quickly as possible. All tho-racic HRCTs were recorded less than 12 h after admis-sion. Echocardiography and pulmonary function testswere also performed as soon as possible during hospi-talization [16]. The final diagnosis was determined by twoindependent experts (either a pulmonologist, cardiologist,general medicine internist, geriatric or emergency physi-cian), who were blinded to the BNP level, using thecomplete medical chart including all initial data and these

2232

three investigations. In cases of disagreement between thetwo experts, a consensus was reached by a third expert.

The final diagnosis was classified as: CPE group or noCPE group. No specific guidelines were given to theexperts for the determination of the final diagnosis ofCPE or lung disease. Instead, the usual criteria were used,such as Framingham criteria and response to diuretic orvasodilator, and the results of Doppler echocardiographywere analyzed specifically for CPE [15]. The results fromspirometry, thoracic HRCT and response to bronchodi-lator or antibiotics were analyzed specifically for respi-ratory disorders [17]. Pulmonary embolism (PE) wasconfirmed by contrast-enhanced helical CT scan, nuclearlung scan or venous ultrasonography, as previously rec-ommended [18]. When acute right heart failure waspresent but without CPE, the patient was classified by theexperts as being in the no CPE group [19].

Statistical analysis

Data are expressed as means € standard deviation (SD)or median and 95% confidence interval (CI) for non-Gaussian variables. Comparison of two means was per-formed using the Student t-test; comparison of two me-dians using the Mann-Whitney test and comparison oftwo proportions using the Fisher exact method.

The reference method for diagnosis of CPE was theexpert panel. The receiver operating characteristic (ROC)curve was used to determine the best threshold value forBNP to predict CPE. The best threshold value was the onewhich minimized the distance to the ideal point (sensi-tivity = specificity =1) on the ROC curve. The area underthe ROC curve was also calculated. Sensitivity, speci-ficity, positive and negative predictive values, and accu-racy (defined as the sum of concordant cells divided bythe sum of all cells in the two-by-two table) and their 95%confidence intervals (CI) were calculated. The positiveand negative likelihood ratios and their 95% CIs werealso calculated. Comparisons of these diagnostic variablesbetween BNP and emergency physicians were performedusing the confidence interval method. We also used amultiple forward logistic regression combining all sig-nificant variables in the univariate analysis and BNP topredict the diagnosis of CPE.

In a preliminary analysis (n=100) the accuracy of emer-gency physicians and BNP measurement in diagnosingCPE were 0.75 and 0.84, respectively. Assuming an alpharisk of 0.05 and a beta risk of 0.20, we calculated thatat least 295 patients should be included in the study(Nquery Advisor 3.0 software, Statistical Solutions, Cork,Ireland).

All statistical comparisons were two-tailed and a pvalue of less than 0.05 was required to reject the nullhypothesis. Statistical analysis was performed using acomputer and NCSS 6.0 software (Statistical Solutions,

Cork, Ireland) to test the hypothesis that the accuracy ofBNP is superior to that of the emergency physician.

Results

From February 2001 to August 2002, BNP assays wereperformed in 313 patients. In five cases (1.6%) technicalproblems precluded the determination of BNP concen-tration. Thus, the data from 308 patients were analyzed.The mean age was 80€8 years; 70% of the patients wereolder than 70 years. The mean clearance of creatinine was54€35 ml/min. The main characteristics of patients withand without CPE are summarized in Table 1. The in-hospital mortality of patients with CPE was higher thanthat of patients without CPE (26 versus 8%, p<0.05).

Thoracic HRCT without contrast iodine medium,transthoracic Doppler echocardiography and pulmonaryfunction tests were performed in 144 (47%), 135 (44%)and 95 (31%) patients, respectively. Table 2 describes thecauses of acute dyspnea diagnosed by the expert panel.There was agreement between the two experts in 263cases (85%). The remaining 45 cases required a thirdexpert. Final diagnoses were classified as CPE (n=141,46%) or no CPE (n=167, 54%). The median value of BNPwas 575 pg/ml (95% CI: 410–898) in the CPE groupversus 75 pg/ml (95% CI: 59–98) (p<0.001) in the noCPE group. In the no CPE group, all the 17 patients withelevated BNP had respiratory disorders. The capacity ofBNP to differentiate CPE from other causes of dyspneawas assessed with a ROC curve analysis. The area underthe ROC curve was 0.874€0.081 (p<0.001). The bestthreshold value of BNP was 250 pg/ml (Table 3). Inmultiple logistic regression analysis, a value of BNP of250 pg/ml or higher was the strongest independent pre-dictor of CPE (Table 4). As shown in Table 5, BNPanalysis was more efficient than the emergency physicianin diagnosing CPE. Table 6 shows that, whatever theempirical probability estimated by the emergency physi-cian for CPE, an elevated level of BNP higher than250 pg/ml was more accurate.

In patients presenting with wheezing found during thephysician’s examination (n=94), the BNP level washigher in patients with CPE (cardiac asthma) than thosewith wheezing secondary to obstructive lung disease, thevalues being, respectively, 611 pg/ml (95% CI: 370–709)versus 56 pg/ml (95% CI: 48–97) (p<0.001). In the noCPE group, patients with acute right heart failure deter-mined clinically (jugular venous pressure elevation orabdominal hepato-jugular sign) and/or with echocardio-graphic acute cor pulmonale had a higher BNP level(125 pg/ml (95% CI: 75–152), n=52) than patients with-out acute right heart failure (59 pg/ml (95% CI: 41–88),n=115), but the difference was not significant.

A value of BNP of 250 pg/ml or higher was associat-ed with greater in-hospital mortality (25 versus 12%,

2233

p<0.001). The best threshold value of BNP to predict thein-hospital mortality was 250 pg/ml, with an area underthe ROC curve of 0.666€0.124 (p<0.05), a sensitivity of0.63 (95% CI: 0.49–0.75) and a specificity of 0.63 (95%CI: 0.57–0.69).

Discussion

The main finding of our study is that the BNP level is asensitive and specific marker of CPE in elderly patients(mean age 80 years) with acute dyspnea. However, wealso demonstrated that the threshold value is higher inpatients older than 65 years. Previous studies reported thatacute left heart failure, frequently presenting as acutedyspnea, is one of the first causes of hospitalization in

Table 1 Main characteristics ofpatients without (no CPE) andwith cardiogenic pulmonaryedema (CPE)

No CPE CPE

(n=167) (n=141)

Demographic dataAge (years) 78€7 83€8*Patients over 75 years 104 (62%) 110 (78%)*Female 84 (50%) 72 (51%)

Previous medical historyCardiac insufficiency 15 (9%) 48 (34%)*Chronic respiratory insufficiency 44 (26%) 32 (23%)Venous thromboembolic disease 23 (14%) 13 (9%)

Criteria of severityUse of accessory muscles 29 (17%) 45 (32%)*Cyanosis 12 (7%) 18 (13%)RV failure signs 56 (33%) 78 (55%)*Ventilatory rate >30/min 21 (13%) 34 (24%)*

Renal functionUrea (mmol/l) 9.8€10.9 12.1€12.7Creatinine clearance (ml/min) 61€31 46€26*

Arterial blood gaspH 7.44€0.06 7.43€0.08PaO2 (mmHg) 65€16 66€21SaO2 (%) 91€8 91€8PaCO2 (mmHg) 41€14 41€12Bicarbonates (mmol/l) 26.8€5.7 26.4€5.1

Data are means € SD, median [95% confidence interval] or number of patients (percentage)* p<0.05 versus no CPERV right ventricular; failure signs including pedal edema, SpO2 peripheral oxygen saturation, PaO2partial oxygen pressure, SaO2 arterial oxygen saturation, PaCO2 partial carbon dioxide pressure

Table 2 Diagnosis of the caus-es of the acute dyspnea in el-derly patients, by the expertteam (n=308)

Diagnosis Number of patients (%)

Cardiogenic pulmonary edema 141 (46%)Infectious pneumonia 108 (35%)Acute exacerbation of chronic respiratory insufficiency 102 (33%)Pulmonary embolism 59 (19%)Asthma 12 (4%)Others 68 (22%)

Because several causes may occur in the same patient, the total percentage does not equal 100%. Otherdiagnostics included severe sepsis other than acute pneumonia (n=15), malignancy (n=10), pleuraleffusion (n=5), tense ascites (n=4), neurologic diseases (n=2), anemia (n=1), acute pancreatitis (n=1),unknown diagnostics (n=2)

Table 3 Comparison of the ef-ficiency of various thresholdvalues of B-type natriureticpeptide (BNP) in the diagnosisof cardiogenic pulmonary ede-ma in acute dyspnea in patientsolder than 65 years

BNP value (pg/ml) 100 150 200 250 300 350 400

Sensitivity 0.90 0.85 0.82 0.78 0.72 0.67 0.60Specificity 0.59 0.71 0.84 0.90 0.92 0.92 0.95Positive predictive value 0.65 0.71 0.81 0.87 0.88 0.88 0.90Negative predictive value 0.88 0.85 0.85 0.83 0.80 0.77 0.74Accuracy 0.72 0.78 0.83 0.84 0.82 0.80 0.79

2234

elderly patients, with a high mortality and readmissionrate [1–3]. Usually, physical examination, laboratorytests, electrocardiogram and chest X-ray are sufficient forthe diagnosis of CPE but, in elderly patients, the clinicalor radiological signs may be obscured by the presence ofaging changes or other diseases [2]. Furthermore, autopsystudy in elderly patients has demonstrated that CPE wasone of the main primary causes of deaths and was eitherover- or under-estimated [20]. Several reasons explainthese difficulties. CPE frequently presents as wheezing,which may also be the result of obstructive airways dis-ease, and the classical radiological signs of CPE aresometimes confusing [21]. Doppler echocardiography israrely available in an emergency care setting and is rarelyperformed in elderly patients [2, 19, 22]. Furthermore,abnormal diastolic filling assessed by the Doppler ratioE/A is common in elderly patients and does not alwayscorrelate with heart failure [7–10]. Inappropriate treat-ment of ventricular dysfunction can lead to unnecessarymorbidity and mortality [3].

B-type natriuretic peptide is a 32-aa polypeptide se-creted by cardiac myocytes in response to increasingventricular stretch or activating the renin-angiotensin-al-dosterone system, either during systolic or diastolic heartfailure [8, 23]. Several studies in patients younger than65 years have demonstrated the usefulness of BNP indiagnosing left heart failure [10–13]. In our study in-cluding older patients, at a cut-off value of 250 pg/ml,BNP analysis had a good sensitivity, specificity and ac-curacy for the diagnosis of CPE in an emergency de-partment. Nevertheless, our threshold value of BNP ishigher than that determined in previous studies [10–13].Several reasons may explain this discrepancy. Our studyincluded very elderly patients (mean age 80 years) andprevious studies suggested that BNP increased with theage of the population [8, 14]. Conversely, a recent studyfailed to show an effect of age on plasma BNP, but itincluded only middle-aged patients (<64 years) [24].Previous studies determined a threshold value of BNP of100 pg/ml, but included patients younger than 65 years[10–13]. In contrast, another study including older pa-tients (mean age �70 years) found a cut-off value of BNPbetween 208 and 300 pg/ml [19, 25].

In the largest multicenter study to evaluate BNPin acute dyspnea, the authors purposed a cut-off valueof BNP of 100 pg/ml, with a sensitivity of 90% andan accuracy of 83% [13]. In fact, the analysis of theirROC curve reveals that the best threshold value whichshould have been defined by the highest accuracy of84% was close to 150 pg/ml (with a lower sensitivity of85%), and thus closer to our cut-off value. Anotherpossible explanation for our higher cut-off value of

Table 4 Multiple forward logistic regression analysis of factorsused for differentiating between patients with cardiogenic pulmo-nary edema (CPE) and those without CPE

Odds ratio (95% CI) p value

BNP �250 pg/ml 24.4 (12.0–49.6) <0.001Lower extremity edema 4.6 (2.0–10.6) <0.001Rales 3.1 (1.6–6.0) 0.001History of cardiac disease 2.9 (1.4–6.1) 0.005

CI confidence interval

Table 5 Comparison of the ef-ficiency of the emergency phy-sician and the B-type natriureticpeptide (BNP) level (value�250 pg/ml)

Emergency physician BNP level

Sensitivity 0.74 (0.66–0.80) 0.78 (0.71–0.84)Specificity 0.80 (0.73–0.85) 0.90 (0.84–0.93)*Positive predictive value 0.75 (0.68–0.82) 0.87 (0.80–0.91)*Negative predictive value 0.78 (0.71–0.84) 0.83 (0.77–0.88)Accuracy 0.77 (0.72–0.81) 0.84 (0.80–0.88)*Positive likelihood ratio 3.62 (2.64–4.97) 7.66 (4.84–12.12)Negative likelihood ratio 0.33 (0.25–0.44) 0.24 (0.18–0.33)

Numbers in parentheses are the 95% confidence intervals* p<0.05 versus emergency physician

Table 6 Comparison of the efficiency of the B-type natriuretic peptide level (value �250 pg/ml) in the diagnosis of cardiogenicpulmonary edema (CPE) according to the probability of that diagnosis quoted by the emergency physician

Empirical probabilityof CPE estimated byemergency physicians

Number ofpatients

Prevalence ofCPE

Positive likelihoodratio

Negative likelihoodratio

Accuracy

Absent 78 0.22 7.2 (4.4–11.9) 0.2 (0.1–0.4) 0.86 (0.76–0.92)Unlikely 92 0.19 8.3 (4.2–16.3) 0.1 (0–0.5) 0.89 (0.81–0.94)Possible 67 0.60 8.8 (2.3–34) 0.4 (0.2–0.6) 0.76 (0.65–0.85)Very likely 71 0.90 5.9 (0.9–36.4) 0.2 (0.1–0.4) 0.85 (0.74–0.91)

Data are associated with their (95% confidence interval)

2235

References

1. Gottdiener JS, Arnold AM, AurigemmaGP, Polak JF, Tracy RP, Kitzman DW,Gardin JM, Rutledge JE, Boineau RC(2000) Predictors of congestive heartfailure in the elderly: the cardiovascularhealth study. J Am Coll Cardiol3:1628–1637

2. Lien CT, Gillespie ND, Struthers AD,McMurdo ME (2002) Heart failure infrail elderly patients: diagnostic diffi-culties, co-morbidities, polypharmacyand treatment dilemmas. Eur J HeartFail 4:91–98

3. Wuerz RC, Meador SA (1992) Effectsof pre hospital medications on mortalityand length of stay in congestive heartfailure. Ann Emerg Med 21:669–674

4. Brown LH, Gough JE, Seim RH (1998)Can quantitative capnometry differen-tiate between cardiac and obstructivecauses of respiratory distress? Chest113:323–326

5. McNamara RM, Cionni DJ (1992)Utility of the peak expiratory flow ratein the differentiation of acute dyspnea.Cardiac vs pulmonary origin. Chest101:129–312

6. Ailani RK, Ravakhah K, DiGiovine B,Jacobsen G, Tun T, Epstein D, West BC(1999) Dyspnea differentiation index. Anew method for the rapid separation ofcardiac vs pulmonary dyspnea. Chest116:1100–1104

7. Marantz PR, Tobin JN, Derby CA,Cohen MV (1994) Age-associatedchanges in diastolic filling: DopplerE/A is not associated with congestiveheart failure in the elderly. South Med J87:728–735

8. Levin ER, Gardner DG, Samson WK(1998) Natriuretic peptides N Engl JMed 339:321–328

9. Davis M, Espiner E, Richards G,Billings J, Town I, Neill A, Richards M,Turner J, Yandle T (1994) Plasma brainnatriuretic peptide in assessment ofacute dyspnoea. Lancet 343:440–444

10. Cabanes L, Richaud-Thiriez B, Fulla Y,Heloire F, Vuillemard C, Weber S,Dusser D (2001) Brain natriuretic pep-tide blood levels in the differential di-agnosis of dyspnea. Chest 120:2047–2050

11. Dao Q, Krishnaswamy P, Kazanegra R,Harrison A, Amirnovin R, Lenert L,Clopton P, Alberto J, Hlavin P, MaiselAS (2001) Utility of B-natriuretic pep-tide in the diagnosis of congestiveheart failure in an urgent-care setting.J Am Coll Cardiol 37:379–385

12. Morrison LK, Harrison A, Krish-naswamy P, Kazanegra R, Clopton P,Maisel A (2002) B-type natriureticpeptide predicts future cardiac events inpatients presenting to the emergencydepartment with dyspnea. J Am CollCardiol 39:202–209

13. Maisel AS, Krishnaswamy P, NowakRM, McCord J, Hollander JE, Duc P,Omland T, Storrow AB, Abraham W,Wu A, Clopton P, Steg G, Westheim A,Wold Knudsen C, Perez A, KazanegraR, Herrmann H, McCullough P (2002)Rapid measurement of B-type natri-uretic peptide in the emergency diag-nosis of heart failure. N Engl J Med347:161–167

14. Clerico A, Del Ry S, Maffei S, PronteraC, Emdin M, Giannessi D (2002) Thecirculating levels of cardiac natriuretichormones in healthy adults: effect ofage and sex. Clin Chem Lab Med40:371–377

BNP is renal dysfunction. In our study, the mean clear-ance of creatinine was 54 ml/min and it was significantlylower in the CPE group (Table 1). A recent study sug-gested that the optimal cut-off value for BNP may beinfluenced by a glomerular filtration rate lower than60 ml/min [26]. Other methodological reasons may beinvolved. Previous studies included either left or rightheart failure in the heart failure group [13]. Conversely,we included the 52 patients with confirmed isolatedacute right heart failure (clinically or by echocardio-graphy) in the no CPE group. BNP levels tend to behigher in this group than those in the group of patientswithout CPE and without acute right heart failure,thus underestimating the diagnostic performance of BNP[19]. As statistical significance was not reached, wecannot confirm this hypothesis.

Few studies have evaluated BNP level in heart failurein elderly patients and none has specifically evaluated thediagnostic performance of BNP in acute dyspnea [27–31].To our knowledge, our study is the first to demonstratethe value of BNP in diagnosing CPE in patients olderthan 65 years with acute dyspnea. Emergency physiciansshould be alerted to the fact that the threshold value ofBNP may be higher in elderly patients. As previousstudies have suggested, we also confirmed that BNP is ofprognostic importance in patients with left heart failure[28, 31, 32].

We agree with other authors that the plasma BNPmeasurement should not be used in isolation without theclinical context [33, 34]. Although the confidence inter-vals of positive likelihood ratio are large (Table 6), wealso confirm that the measurement of BNP constitutesan important addition to the diagnosis of CPE in elderlypatients [33, 35]. Our study suggests that, in clinicalpractice, BNP analysis should be used in every breathlesselderly patient, whatever the empirical probability quotedby the emergency physician. Using BNP assays, heartfailure appears to be very unlikely below a plasma con-centration of 100 pg/ml, whereas CPE is highly likelyabove a plasma concentration of 400 pg/ml in the elderlypatient. In patients with intermediate values, we suggestthat full cardiological assessment and Doppler echocar-diography are required because CPE is still possible. Inthis situation, however, respiratory diseases should alsobe considered [34].

Acknowledgements The authors would like to thank the physiciansand nursing staff working in the emergency department and all thephysicians of the EPIDASA study group: Dr M.-H. Becquemin,Dr J. Boddaert, Dr R. Isnard and Dr A. Teixeira, for their coop-eration and support. We also thank Dr D.J. Baker (Department ofAnesthesiology, CHU Necker-Enfants Malades, Paris, France) forreviewing the manuscript.

This study was supported solely from departmental sources. Thetest and kits for BNP assay were provided free of charge by BiositeUSA/BMD France.

2236

15. Ho KK, Pinsky JL, Kannel WB, Levy D(1993) The epidemiology of heart fail-ure: the Framingham Study. J Am CollCardiol 2:6A-13A

16. Ray P, Birolleau S, Lefort Y, Beigel-man C, Becquemin M-H, Isnard R,Boddaert J, Teixeira A, Riou B (2002)Epidemiological study of acute dyspneain elderly patients, in an emergencydepartment. The EPIDASA study. Am JRespir Crit Care Med 165:A265

17. American Thoracic Society (1995)Standards for the diagnosis and care ofpatients with chronic obstructive pul-monary disease. Am J Respir Crit CareMed 152 (Suppl):S77–S120

18. Task force report (2000) Guidelines ondiagnosis and management of acutepulmonary embolism. Eur Heart J21:1301–1336

19. Logeart D, Saudubray C, Beyne P,Thabut G, Ennezat PV, Chavelas C,Zanker C, Bouvier E, Solal AC (2002)Comparative value of Doppler echo-cardiography and B-type natriureticpeptide assay in the etiologic diagnosisof acute dyspnea. J Am Coll Cardiol40:1794–1800

20. Gross JS, Neufeld RR, Libow LS,Gerber I, Rodstein M (1988) Autopsystudy of the elderly institutionalizedpatient. Review of 234 autopsies. ArchIntern Med 14:173–176

21. Hermann PG, Khan A, Kollman CE,Rojas KA, Carnody DP, BodenheimerM (1990) Limited correlation of leftventricular end diastolic pressure withradiographic assessment of pulmonaryhemodynamics. Radiology 174:721–724

22. Hood S, Taylor S, Roeves A, CrookAM, Tlusty P, Cohen J, Kaddoura S,Hemingway H (2000) Are there age andsex differences in the investigation andtreatment of heart failure? A popula-tion-based study. Br J Gen Pract50:559–563

23. Maisel AS, Koon J, Krishnaswamy P,Kazenegra R, Clopton P, Gardetto N,Morrisey R, Garcia A, Chiu A, DeMaria A (2001) Utility of B-natriureticpeptide as a rapid, point-of-care test forscreening patients undergoing echocar-diography to determine left ventriculardysfunction. Am J Heart 141:367–374

24. Loke I, Squire I, Davies J, Ng L (2003)Reference ranges for natriuretic pep-tides for diagnostic use are dependenton age, gender and heart rate. Eur JHeart Fail 5:599–606

25. Lainchbury J, Campbell E, Frampton C,Yandle T, Nicholls M, Richards A(2003) Brain natriuretic peptide and N-terminal brain natriuretic peptide in thediagnosis of heart failure in patientswith acute shortness of breath. J AmColl Cardiol 42:728–735

26. McCullough PA, Duc P, Omland T,McCord J, Nowak RM, Hollander JE,Herrmann HC, Steg PG, Westheim A,Knudsen CW, Storrow AB, AbrahamWT, Lamba S, Wu AH, Perez A,Clopton P, Krishnaswamy P, KazanegraR, Maisel AS, Breathing Not ProperlyMultinational Study Investigators(2003) B-type natriuretic peptide andrenal function in the diagnosis of heartfailure: an analysis from the BreathingNot Properly Multinational Study. Am JKidney Dis 41:571–579

27. Smith H, Pickering RM, Struthers A,Simpson I, Mant D (2000) Biochemicaldiagnosis of ventricular dysfunction inelderly patients in general practice: ob-servational study. BMJ 320:906–908

28. Wallen T, Landahl S, Hedner T, NakaoK, Saito Y (1997) Brain natriureticpeptide predicts mortality in the elderly.Heart 77:264–267

29. Yamada Y, Goto J, Yokota M (1997)Brain natriuretic peptide is a sensitiveindicator of impaired left-ventricularfunction in elderly patients with car-diovascular disease. Cardiology88:401–407

30. Ueda R, Yokouchi M, Suzuki T, OtomoE, Katagiri T (2003) Prognostic valueof high plasma brain natriuretic peptideconcentrations in very elderly persons.Am J Med 114:266–270

31. Tamura K, Takahashi N, Nakatani Y,Onishi S, Iwasaka T (2001) Prognosticimpact of plasma brain natriuretic pep-tide for cardiac events in elderly pa-tients with congestive heart failure.Gerontology 47:46–51

32. Cheng V, Kazanagra R, Garcia A,Lenert L, Krishnaswamy P, Gardetto N,Clopton P, Maisel A (2001) A rapidbedside test for B-type peptide predictstreatment outcomes in patients admittedfor decompensated heart failure: a pilotstudy. J Am Coll Cardiol 37:386–391

33. McCullough PA, Nowak RM, McCordJ, Hollander JE, Herrmann HC, StegPG, Duc P, Westheim A, Omland T,Knudsen CW, Storrow AB, AbrahamWT, Lamba S, Wu AH, Perez A,Clopton P, Krishnaswamy P, KazanegraR, Maisel AS (2002) B-type natriureticpeptide and clinical judgment in emer-gency diagnosis of heart failure: analy-sis from Breathing Not Properly (BNP)Multinational Study. Circulation106:416–422

34. De Denus S, Pharand C, Williamson D(2004) Brain natriuretic peptide in themanagement of heart failure: the ver-satile neurohormone. Chest 125:652–668

35. Fischer JE, Bachmann LM, Jaeschke R(2003) A readers’ guide to the inter-pretation of diagnostic test properties:clinical example of sepsis. IntensiveCare Med 7:1043–1051