Published ahead of print on July 13, 2006, doi:10.1164/rccm.200603-380OC
© 2006 American Thoracic Society doi: 10.1164/rccm.200603-380OC
Association of Left-Heart Dysfunction with Severe Exacerbation of Chronic Obstructive Pulmonary DiseaseDiagnostic Performance of Cardiac BiomarkersIntensive Care Unit, Department of Cardiology, and Biochemistry Laboratory, Fattouma Bourguiba University Hospital, Monastir, Tunisia Correspondence and requests for reprints should be addressed to Fekri Abroug, M.D., ICU, CHU F. Bourguiba, 5000 Monastir, Tunisia. E-mail: f.abroug{at}rns.tn
Rationale: Cardiac biomarkers are used to distinguish acute dyspnea due to left-heart dysfunction from that of pulmonary origin. However, they have not been assessed in the specific setting of acute exacerbation of chronic obstructive pulmonary disease (AECOPD), where they might be released without left-heart impairment. Objective: To assess the accuracy of troponin T and of amino-terminal pro–brain natriuretic peptide (NT-proBNP) in the diagnosis of AECOPD associated with left ventricular (LV) dysfunction. Methods: Both biomarkers were measured in 148 consecutive patients on intensive care unit admission for AECOPD. A panel of physicians adjudicated blindly the cause of AECOPD to be unlikely, possibly associated, or definitely associated with LV dysfunction. Measurements and Main Results: The final diagnosis was AECOPD definitely associated with acute left-heart dysfunction in 31.1%, possibly associated with LV dysfunction in 13.5%, and probably not associated with LV dysfunction in 55.4%. Both NT-proBNP and troponin T levels were significantly different among the three groups. The area under the receiver operating characteristic curve was greater for NT-proBNP (0.95 vs. 0.67). A cutoff of 1,000 pg/ml was accurate to rule out left-heart involvement in AECOPD (sensitivity, 94%; negative predictive value, 94%; negative likelihood ratio, 0.08). A cutoff of 2,500 pg/ml had the best operating characteristics to rule in the diagnosis (positive likelihood ratio, 5.16). Left-heart involvement in AECOPD was the only variable independently associated with increased secretion of NT-proBNP (odds ratio, 74; 95% confidence interval, 15–375; p = 0.0001). Conclusion: NT-proBNP and troponin T are useful in excluding AECOPD associated with left ventricular dysfunction. NT-proBNP was the more accurate of the two.
Key Words: B-type natriuretic peptide • chronic obstructive pulmonary disease • exacerbation • left ventricular dysfunction • troponin Chronic obstructive pulmonary disease (COPD) is a major cause of chronic morbidity and mortality worldwide. It is the fifth leading cause of death worldwide (1). Acute exacerbation of COPD (AECOPD) accounts for large amounts of the morbidity and mortality attributed to COPD (2). AECOPD refers to the exaggeration of COPD symptoms: aggravation of dyspnea; an increase in expectoration volume; and a change in the appearance of sputum, which becomes purulent (3). Although dominated by bacterial or viral infection, etiologies of AECOPD remain unrecognized in as much as one-third of these patients (4). Accordingly, acute left ventricular (LV) dysfunction in patients with otherwise known or unknown left-heart chronic disease is suspected as a cause of exacerbation in many such patients. Nevertheless, in many of these situations, LV dysfunction might be associated without being the cause of the exacerbation. Yet, a diagnosis of LV dysfunction in patients with dyspnea is currently challenging emergency department physicians, because bedside clinical assessment has a poor performance record and cardiac function tests with enough accuracy to diagnose LV dysfunction (in particular, echocardiography) are not always possible, because they are unavailable or difficult to interpret (5). This is particularly true in patients with COPD, in whom echocardiography is not often feasible for technical reasons. Yet, the prevalence of LV dysfunction is probably high in COPD because this condition shares many risk factors with coronary disease: age, male predominance, cigarette smoking, and so on (6). Hence, apart from patients with previously known chronic heart disease, automatically prompting a consideration of the possibility of LV dysfunction as a cause of the AECOPD, the actual prevalence of AECOPD associated with LV dysfunction remains unknown. Biomarkers such as troponin or natriuretic peptides have the advantage of being easy to obtain at affordable cost. Troponin I levels have been shown to have prognostic benefit in patients admitted to the intensive care unit (ICU) for AECOPD (7). On the other hand, brain natriuretic peptide (BNP) and amino-terminal pro–brain natriuretic peptide (NT-proBNP) have been shown to perform well in distinguishing between dyspnea of cardiac origin dyspnea of pulmonary causes in patients attending the emergency department (8–13). However, extrapolation of these results to the specific context of AECOPD should not be straightforward because BNP secretion might be secondary either to LV stress or to hypoxemia, as well as to pulmonary hypertension or right ventricular (RV) stress (14, 15). Indeed, all these conditions might be present in a patient with AECOPD. Only one study so far has suggested that BNP secretion originating from the LV and that originating from the right side of the heart might add up (16). To the best of our knowledge, no previous study has assessed the clinical performance of natriuretic peptide doses in the diagnosis of LV dysfunction associated with AECOPD. We therefore undertook the current study to assess the diagnostic performance of troponin T and NT-proBNP levels in the diagnosis of AECOPD associated with LV dysfunction.
This is a prospective study conducted between October 1, 2001, and March 31, 2005. All consecutive patients admitted to the medical ICU for severe acute exacerbation of COPD were considered for inclusion. The study protocol was approved by the institutional review board of our hospital (Fattouma Bourguiba University Hospital, Monastir, Tunisia), and written, informed consent was obtained from all next of kin.
Study Population
Noninclusion criteria.
Protocol All included patients underwent transthoracic echocardiographic examination on the first day of ICU admission, performed by one of two echocardiographers. M-mode and two-dimensional and color Doppler imaging were obtained with commercially available instruments operating at 2.0 to 3.5 MHz. Two-dimensional imaging examinations were performed in standard and parasternal views. Pulsed Doppler spectral recordings were obtained in the apical four-chamber view from a sample volume positioned at the tips of the mitral leaflets. LV systolic and diastolic volumes and ejection fraction were derived from biplane apical views according to a modified Simpson's rule algorithm. Left-atrium and LV dimensions were measured from M-mode images. The transmitral pulsed Doppler velocity recordings from three consecutive cardiac cycles were used to derive the following measurements: early diastolic inflow (E) and late atrial inflow (A) velocities, deceleration time (DT), and LV isovolumetric relaxation time (IVRT). All echocardiographic data were copied to a VHS videotape for subsequent playback, analysis, and measurement by both echocardiographers in order to reach a consensus concerning all the study patients. Echocardiographers were blinded to BNP levels. This allowed for the following echographic classification:
Right-heart catheterization was performed only in mechanically ventilated patients who experienced difficult weaning with strong clinical and echocardiographic indications of a decompensated left-heart origin. Data, both measured by Swan-Ganz catheter and derived, were obtained while patients were under controlled ventilation and while breathing spontaneously.
NT-proBNP and Cardiac Troponin Testing NT-proBNP and cardiac troponin T were determined by quantitative electrochemiluminescence assay (Elecsys proBNP and Elecsys Troponin; Roche Diagnostics, Indianapolis, IN) on an Elecsys 2010 analyzer (Roche Diagnostics) according to established methods.
Determination of the Presence of LV Dysfunction By using all available data without proBNP and troponin testing, patients were stratified through a consensus of all four physicians to one of three categories as follows: (1) AECOPD unlikely to be associated with left ventricular dysfunction; (2) AECOPD possibly associated with LV dysfunction; or (3) AECOPD definitely associated with LV dysfunction. In the first category, we included patients fulfilling all the following criteria: no history of, or risk factors for, LV disease, with no abnormalities pertaining to LV dysfunction in the physical examination or on chest radiograph; a normal echocardiograph; and no difficulties during the mechanical ventilation weaning process. At the opposite, AECOPD definitely associated with LV dysfunction included patients who usually had an initial physical examination including relevant signs of LV dysfunction (pulmonary rales, third heart sound, etc.), and pulmonary congestion on chest X-ray. All patients in this category had to exhibit LV systolic dysfunction or a restrictive pattern of diastolic dysfunction on echocardiographic examination. Many of these patients who were mechanically ventilated also experienced weaning difficulties potentially attributable to LV dysfunction (increase in pulmonary artery occlusion pressure levels greater than 10 mm Hg when shifting from mechanical ventilation to spontaneous ventilation). AECOPD possibly associated with LV dysfunction corresponded to the remaining intermediate situations.
Statistical Analysis
During the study period, 238 patients with severe AECOPD were admitted for the first time to the ICU. Of these, 90 were not included because of the following reasons: AECOPD was due to infectious pneumonia (n = 57), pneumothorax (n = 15), pulmonary embolism diagnosed on computerized tomography scan in all cases (n = 6), and poor echogenicity (n = 12). The remaining 148 patients were included in the study. They were mainly men (n = 120) and had a median age of 68 yr (interquartile range, 15), with a history of smoking in 83% of patients. Table 1 reports baseline clinical characteristics of the study patients.
Echocardiographic Findings Seventy-five patients (51%) had LV echocardiographic dysfunction, whether systolic (n = 17), diastolic (n = 48), or both (found in 10 patients). LV diastolic dysfunction was of the impaired relaxation type (n = 46) or restrictive in pattern (n = 12).
Patient Management
Association of AECOPD with LV Dysfunction According to Expert Classification
Distribution of Biomarkers among Study Groups Both NT-proBNP and troponin T biomarkers were significantly different among groups (p < 0.0001 for both; Figure 1). For both biomarkers, Tukey post hoc analysis showed a statistically significant difference between patients with an adjudicated final diagnosis of AECOPD unlikely to be associated with LV dysfunction and those with AECOPD definitely associated with LV dysfunction. A weak although statistically significant correlation was found between NT-proBNP and troponin T (R = 0.4; p = 0.0001).
Role of Biomarkers to Rule Out the Association of AECOPD with LV Dysfunction To rule out the association of AECOPD with LV dysfunction, we compared NT-proBNP and troponin T levels between patients in whom AECOPD was probably not associated with LV dysfunction versus the remaining patients (those with possible or definite LV dysfunction). Differences among these groups were statistically significant for NT-proBNP (p < 0.0001) and troponin T (p < 0.0004). ROC analysis (Figure 2) demonstrated NT-proBNP to have higher accuracy to rule out the association of AECOPD with LV dysfunction, as indicated by a higher area under the ROC curve (area under curve [AUC], 0.95 vs. 0.67 for NT-proBNP and troponin T, respectively). The optimal cut point for proBNP was 1,000 pg/ml, which had the following operative characteristics for ruling out the association of AECOPD with LV dysfunction: sensitivity, 94%; specificity, 77%; negative predictive value, 94%; positive predictive value, 78%; negative likelihood ratio, 0.08; and positive likelihood ratio, 4.18 (with 85% of well-classified patients).
Role of Biomarkers to Rule In the Association of AECOPD with LV Dysfunction Cardiac biomarkers were then compared between the two subgroups of patients with AECOPD definitely associated with LV dysfunction versus the remaining patients (those with AECOPD unlikely or only possibly associated with LV dysfunction). Differences were statistically significant for both biomarkers (p = 0.0001 and 0.002 for NT-proBNP and troponin T, respectively). NT-proBNP was more accurate to rule in AECOPD associated with LV dysfunction as reflected by an ROC curve with a higher AUC (0.91 vs. 0.69 for NT-proBNP and troponin T, respectively). A proBNP cut point of 2,500 pg/ml had the following operative characteristics: sensitivity, 77%; specificity, 86%; positive predictive value, 71%; negative predictive value, 89%; negative likelihood ratio, 0.27; and positive likelihood ratio, 5.16.
Predictors of NT-proBNP Increase
In the multivariate analysis, we assessed the respective contribution of RHF and AECOPD associated with LV dysfunction in the prediction of an elevated level of NT-proBNP (defined as a proBNP level of 1,000 pg/ml or more). Only AECOPD associated with LV dysfunction was independently associated with increased secretion of NT-proBNP (odds ratio, 74; 95% confidence interval, 15–375; p = 0.0001).
Our study shows that 31% of patients admitted to the ICU for severe AECOPD had an exacerbation definitely associated with left-heart dysfunction. In 55.4%, the exacerbation was deemed unlikely to be associated with LV dysfunction. Both biomarkers assessed in the current study were discriminant of the association of AECOPD with LV dysfunction, with NT-proBNP having the best operative characteristics. NT-proBNP performed better to rule out than to rule in an association of AECOPD with LV dysfunction. An NT-proBNP level of 1,000 pg/ml or less was the optimal cutoff to rule out an association of AECOPD with LV dysfunction (sensitivity, 94%; negative predictive value, 94%; negative likelihood ratio, 0.08). Although RHF usually evokes elevated levels of NT-proBNP, association of LV dysfunction further increases NT-proBNP levels, suggesting the addition of BNP secretions from right and left ventricles. However, the increase in NT-proBNP levels recorded here is not accounted for by RHF per se. In patients such as those included here, some degree of LV dysfunction might be present during AECOPD with only marginal impact on clinical status and response to treatment. In fact, LV dysfunction might be present without being the cause of the exacerbation, at least in some patients. This is the reason why we prefer the term AECOPD associated with LV dysfunction instead of LV-related AECOPD. Operative characteristics of a test rely heavily on the "gold standard" used. For diagnosis of LV and RV dysfunction in the current study, we used a combination of clinical and echocardiography findings. Indeed, the accuracy of the diagnosis of congestive heart failure by clinical means and standard testing might be inadequate (20). This is the reason why all patients had an echocardiographic examination exploring both LV systolic and diastolic function. Those patients in whom echocardiographic examination was not possible because of poor echogenicity were not included in the study. It is noteworthy that the methodology we used in our study is the one implemented in all basic studies on the validation of BNP dose as a diagnostic test of dyspnea of cardiac or pulmonary origin (8, 9, 11, 12, 21). In addition, the final diagnosis of the association of LV dysfunction with AECOPD had to be adjudicated through a consensus of all four physicians. Regarding the diagnosis of RHF, it should be acknowledged that it is usually easier than that of left-heart failure. Natriuretic peptide levels are currently widely used in the diagnosis of left-heart dysfunction (22). In addition, they have prognostic value in this setting, and serve as a guide to titrate treatments in patients with left-heart failure (23–25). Natriuretic peptides have also been used in emergency departments to distinguish acute dyspnea related to left-heart failure from that of pulmonary origin. Numerous studies have validated the use of natriuretic peptide levels in this indication (8, 9, 11, 12, 16, 21). Levels of 100 pg/ml for BNP and 450 pg/ml for NT-proBNP were accurate in ruling in the diagnosis of acute congestive heart failure (26). On the other hand, levels of 50 pg/ml for BNP and 300 pg/ml for NT-proBNP were accurate in ruling out the diagnostic of acute congestive heart failure in emergency department attendees with dyspnea (26). The extrapolation of these findings to patients with AECOPD has two drawbacks. First, in these patients there is usually a supraphysiologic secretion of natriuretic peptides because of the presence of hypoxemia, pulmonary hypertension, and RV dysfunction (14, 15, 27). Indeed, each of these factors, usually present in patients with AECOPD, has been individually linked to secretion of natriuretic peptides. The second drawback to the extrapolation of natriuretic peptide findings to patients with COPD results from the unproved hypothesis concerning whether natriuretic peptide liberation is additive when both left ventricle and right ventricle contribute to the natriuretic peptide level. Our findings provide part of the answer to these questions. There is, indeed, a supraphysiologic release of natriuretic peptides in AECOPD because these levels are elevated in decompensated COPD patients without LV dysfunction. However, when LV dysfunction is also present, natriuretic peptide levels are much more elevated. Hence, this biomarker might be considered an accurate indicator of the association of LV dysfunction in such patients. Accordingly, the threshold identified in our study to rule out LV involvement (1,000 pg/ml) is greater than is usually recommended in patients without COPD. The threshold we determined to rule in the diagnosis (2,500 pg/ml) is also more elevated than that recommended in patients without COPD. Numerous studies have shown that in primary or secondary pulmonary hypertension, there is a supraphysiologic secretion of natriuretic peptides. Nagaya and coworkers showed that in patients with RV pressure overload due to primary pulmonary hypertension, atrial natriuretic protein and BNP levels are higher than in patients with RV volume overload due to atrial septal defect (28). Hence, coexistence of pulmonary hypertension with RV distension elicits larger amounts of natriuretic peptide liberation. Moreover, atrial natriuretic protein and BNP levels each correlated with mean pulmonary artery pressure, right atrial pressure, RV end-diastolic pressure, and pulmonary resistance (28). BNP levels have also been shown to be higher in patients with acute pulmonary embolism, especially in those with echocardiographic evidence of RV dysfunction (29–32). Studies suggest that BNP levels can be used in pulmonary embolism to identify patients with RV overload. Binder and coworkers have validated a strategy of stratification of pulmonary embolism severity on the basis of RV dysfunction based on echocardiographic findings and BNP levels. An NT-proBNP threshold of 1,000 pg/ml had the best operative characteristics to rule out a risk of poor outcome (29), and higher BNP levels were associated with increased mortality of patients with acute pulmonary embolism. It is noteworthy that the threshold proposed by these authors is similar to that found in our study. Moreover, a case report showed that the clinical and echocardiographic improvement associated with successful treatment of massive pulmonary embolism with fibrinolytics was associated with a reduction of the initially elevated BNP level (30). All these studies suggest that clinical situations evoking RV dysfunction are actually associated with increased BNP release. The severity of pulmonary hypertension and of RV distension (whether measured hemodynamically or echocardiographically) might even be estimated by BNP levels, which might also reflect treatment efficacy. How should BNP levels be interpreted in patients with coexisting RV and LV dysfunction? Comparison of LV dysfunction and RV dysfunction shows little evidence that the BNP plasma levels are different. BNP levels reported in RV are usually equivalent to those elicited either by LV systolic or diastolic dysfunction (9, 28). Few studies have so far suggested that BNP secretions might be additive in the setting of distension of both ventricles. A scintigraphic study by Mariano-Goulart and coworkers has shown higher BNP levels in patients with both RV and LV failure than in patients with LV failure alone (33). Along with these findings, our study suggest that LV and RV release of BNP might be additive. In an ancillary study from the Breathing Not Properly multicenter study, McCullough and coworkers explored the accuracy of BNP to distinguish new-onset heart failure in the subset of patients with COPD and/or asthma, who presented to the emergency department with dyspnea (16). The diagnosis of congestive heart failure was adjudicated by independent cardiologists who were blinded to BNP results. Eighty-seven of 417 patients (20.9%) were found to have congestive heart failure. These patients had significantly higher levels of BNP (587 ± 426 vs. 108.8 ± 22 pg/ml, respectively). A cut point of 100 pg/ml was accurate to rule out the diagnosis of congestive heart failure (negative likelihood ratio, 0.09). Along with our study, this study documents the accuracy of natriuretic peptides in the diagnosis of exacerbation of asthma/COPD from cardiac origin. However, these patients were not patients with severe exacerbations characterized by severe hypoxemia, pulmonary hypertension, and RV distension. Whether our results would hold true had we used BNP instead of NT-proBNP is now well documented. Both peptides are derived from the 134-amino acid precursor preproBNP (26). They have been shown to be closely correlated to each other and exhibit parallel changes across a broad spectrum of age, renal function, and LV ejection fraction (34). However, they are not interchangeable because the level of NT-proBNP tends to be about 10-fold higher than that of BNP (26). In addition to NT-proBNP, our study assessed the accuracy of another cardiac biomarker (troponin T) in the diagnosis of AECOPD associated with LV dysfunction. The rationale of this evaluation is that an RV lesion is usually associated with ventricular distension. Indeed, Baillard and coworkers have shown that as much as 18% of patients with AECOPD requiring admission to the ICU and ventilatory assistance had significantly elevated troponin I levels (greater than 0.05 ng/ml) (7). These authors reported that troponin I levels carry prognostic value, to the extent that it was similar to that of SAPS II (new Simplified Acute Physiology Score); indeed, both variables estimated in-hospital mortality well. Our study confirms that troponin levels are elevated in AECOPD. They extend previous conclusions by suggesting that troponin release occurs as a consequence of coexisting LV dysfunction. Indeed, significant levels of troponin were found only in patients with definite AECOPD associated with LV dysfunction. Nevertheless, NT-proBNP performed better than troponin T as a marker of AECOPD associated with left-heart dysfunction.
The authors thank Professor Christian Brun Buisson (Hôpital Henri Mondor, Créteil, France) for help in preparation of the manuscript.
Originally Published in Press as DOI: 10.1164/rccm.200603-380OC on July 13, 2006 Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript.
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
2-agonist bronchodilators. In intubated patients, the weaning process was started when usual weaning criteria were present (
