According to the American Heart Association, an estimated 6.2 million Americans over the age of 20 years have congestive heart failure (CHF), and the lifetime risk of developing heart failure from ages 45 to 95 years is 20% to 45%, depending on a variety of risk factors including gender and race.1 An estimated, 16.1% of heart failure cases in the Emergency Department (ED) present with the symptom of dyspnea, yet dyspnea may be due to many etiologies such as chronic obstructive pulmonary disease, pneumonia, myocardial infarction, atrial fibrillation or flutter, malignancy, and pulmonary embolism.2 The diagnosis of CHF can be challenging considering the extensive differential diagnosis of its nonspecific presenting symptoms and lack of a diagnostic gold standard.3,4 The series of tests needed to make an accurate diagnosis of CHF are expensive, and the cost associated with its management is expensive compared with other cardiac diseases.5,6 For these reasons, the quick and accurate diagnosis of CHF on presentation can help reduce the cost and improve patient outcomes.7,8
Two tests commonly used in the initial evaluation of a patient presenting with dyspnea are chest radiography (CXR) and brain natriuretic peptide (BNP) testing, which are often ordered within the same period of time but have substantial differences in accuracy. CXR has been shown to have a moderate sensitivity and specificity of 74.54% and 86.2%, respectively, for the evaluation of CHF, while the sensitivity and specificity of serum BNP levels are 97.5% and 65%, respectively.9,10 Hence, the strong negative predictive value of BNP for CHF makes it an effective “rule-out” tool.
A normal BNP result in the presence of a CHF diagnosis on CXR represents a diagnostic dilemma. This scenario may be due to overdiagnosis of CHF on CXR or a false-negative BNP result. The purpose of the present study is to further characterize this population in the ED setting to facilitate the “rule-out” of heart failure.
MATERIALS AND METHODS
After Institutional Review Board approval, 49,188 adult patients, aged 21 years and older, were identified who presented to the EDs of a single multisite urban academic health system from January 2018 through December 2018. Of those patients, 2,661 received a CXR within 24 hours of initial presentation and had a laboratory BNP test ordered within 24 hours of the CXR. In all, 862 patients (Fig. 1) met these criteria and had a normal BNP result, defined by < 300 pg/mL, which served as our reference.11 Eight patients had equivocal CXR results and were excluded from the analysis. A total of 854 patients were included in the study.
Data collection was performed using Looking Glass Clinical Analytics (Streamline Health) software that surveys demographic, clinical, and administrative data in the electronic medical record (EMR). Data compiled from the EMR for analysis in this study includes demographic information (age, sex, race), body mass index (BMI), complete CXR reports, selected laboratory results (BNP, white blood cell count, creatinine, blood urea nitrogen, Troponin T, and creatine kinase), and comorbidity information. Comorbidity information was collected according to the Charlson Comorbidity Index and its individual components, which provides an index and percent 10-year survival estimate based on a series of 22 disease conditions, taking into account the severity.12
Each CXR report was reviewed by the study investigators and assessed for the presence or absence of a CHF diagnosis. Patients were considered to have a diagnosis of CHF on CXR if the impression explicitly stated the diagnosis of heart failure or used a wording that is generally accepted to qualify as mild, moderate, or severe equivalent of CHF, such as pulmonary vascular congestion or pulmonary edema. Impressions with a diagnosis that was unclear were determined to be an “equivocal” diagnosis. The language for an equivocal diagnosis may include “cannot exclude,” “hazy” lung markings, or opacities without a specified diagnosis, or other ambiguous wording determined collectively by the research team. All patients with diagnoses other than CHF or normal studies were assigned to the group without a CHF diagnosis, “No CHF.” Patients with an equivocal CHF diagnosis on CXR were excluded from analysis.
The cohort was divided into 2 categories based on the CXR impression: CHF or no CHF. Portable versus nonportable radiographic technique was recorded.
Statistical analysis was performed using R 3.6.3. Patient characteristics between the CHF on CXR group and no CHF group were compared using the Fisher exact test for categorical variables and the Welch T test for age and BMI. Wilcoxon rank-sum tests were used for comparing lab tests and the Charlson score. All P-values reported are 2 sided. Logistic regression models were used to model the joint effects of the variables on CHF. When lab tests were included as co-variants in a regression model, log-transformation was applied except for BNP.
A total of 854 patients presenting to the ED had a CXR within 24 hours and normal serum BNP level < 300 pg/mL. CXR demonstrated CHF in 8.5% (73/854) of patients with a normal BNP result and no CHF in 91.5% (781/854). The patient population is described in Table 1.
TABLE 1 -
Characteristics of the Study Population
||No. Patients (N=854)
|Mean age (y)
|Sex, n (%)
|Race, n (%)
| Black or African
| Other/not given
|Ethnicity, n (%)
| Not Spanish/Hispanic/Latino
| Patient declined
The comparisons in patient demographic and clinical characteristics between the CHF and no CHF groups are shown in Table 2.
TABLE 2 -
Patients With Normal BNP (<300 pg/mL)
||CHF on CXR (n=73)
||No CHF on CXR (n=781)
|Sex, n/N (%)
|Race, n (%)
| Black or African
| Other/not given
|Ethnicity, n (%)
| Not Hispanic/Spanish/Latino
| Patient declined
|Laboratory value (median)
| BNP (pg/mL)
| WBC (109 cells/L)
| Creatinine (mg/dL)
| BUN (mg/dL)
| Troponin T (ng/mL)
| CK total (U/L)
| CK (using max) (U/L)
|Charlson avg. value
|Age-related risk avg.
|Charlson w/ age average
|Charlson comorbidity variables, n/N (%)
| Cerebrovascular disease
| Chronic pulmonary disease
| Rheumatic disease
| Peptic ulcer disease
| Mild liver disease
| Diabetes w/o complications
| Diabetes w/ complications
| Renal disease
| Any malignancy
| Moderate/severe liver disease
| 10 y survival (%)
|Portable CXR, n/N (%)
Figures 2 and 3 are radiographs from the current study who had discordantly positive CXR and negative BNP results.
No significant difference was found for age, gender, or race between the CHF and no CHF groups. BMI was higher in the CHF versus the no CHF group (mean 32.9 vs. 29.8 kg/m2, respectively, P=0.0205).
Charlson Comorbidity Index was higher in patients in the CHF than the no CHF group (mean 5.0 vs. 4.1, respectively, P=0.016). Compared with patients in the no CHF group, patients in the CHF group were more than twice as likely to have a history of CHF than patients in the no CHF group (odds ratio [OR]=2.72, 95% confidence interval [CI]: 1.54-4.72, P=0.0004), and were more than twice as likely to have a history of diabetes with complications (OR=2.53, 95% CI: 1.42-4.41, P=0.0009).
Among the laboratory tests, BNP levels were higher in the CHF versus no CHF group [median 164 vs. 98 pg/mL, respectively, P=4.91×10−5), as were troponin T levels (mean 0.52 vs. 0.19 ng/mL, P=9.8×10−4) and white blood cell count (9.98×103 cells/µL vs. 8.41×103 cells/µL, P=0.0003).
Among all the factors, we found that portable CXRs are most associated with the CHF group (OR=4.65, 95% CI: 2.32-10.32, P=5.91×10−7), accounting for 86.3% (63/73) of CXRs in the CHF group underwent portable CXRs, compared with 57.5% (449/781) in the no CHF group. Only 10 of 73 patients in the CHF group did not undergo portable CXR.
Further analysis of the relationship of BMI to other variables demonstrated that there was no association between the portable technique and BMI (P=0.815). A higher BMI was related to a history of CHF (P=4.72×10−7) and inversely related to higher serum BNP levels (P=2.42×10−5), suggesting that a higher BMI is associated with lower BNP.
Table 3 shows the results of a logistic regression model that estimates the joint effects of the 5 variables on the diagnosis of CHF on CXR in the setting of a normal serum BNP level: portable radiograph technique (P=5.39×10−5), increasing BNP levels (P=000447), increasing BMI (P=0.008159), history of diabetes with complications (P=0.01704), and history of CHF (P=0.032943). Each of the lab values (WBC, creatinine, BUN, troponin) were added to the logistic regression model, but were not independently significant. This can be explained by the fact that these variables were associated with one or more variables already in the model. Neither gender nor race was statistically significant when added to the model.
TABLE 3 -
Relationship Between CHF on CXR, Clinical Variables, and Portable Radiography
|Congestive heart failure
|Diabetes with chronic complications
Among ED patients with normal BNP results, CHF was concordantly ruled out by CXR in 91.5% of patients, and discordantly diagnosed by CXR in 8.5%. The group with CHF on CXR differed notably from those with concordantly negative BNP and CXR results. Patients with a discordant diagnosis of CHF on CXR had higher BMIs (32.9 vs. 29.8 kg/m2) and were more than twice as likely to have a history of CHF or diabetes with complications (OR of 2.72 and 2.53, respectively), had higher serum BNP levels (median 164 vs. 98 pg/mL), and more frequently underwent portable CXR (86.3% vs. 57.5%, OR of 4.65). Hence, the combination of concordantly negative BNP and CXR results are reassuring in ruling out CHF. Discordant results remain a diagnostic dilemma. Results of the present study suggest that these cases are due to both false-positive CXR results and false-negative BNP results.
It can be challenging to make an accurate diagnosis of CHF in the emergency setting. The current practice for diagnosing CHF in a patient presenting with dyspnea includes a history, physical examination, and performing imaging and laboratory tests.10,13,14 CXR is commonly used in the initial workup of CHF and may demonstrate multiple findings, some of which are nonspecific.9,10,15,16 The limited accuracy of CXRs poses a problem in terms of the ability to efficiently diagnose heart failure.9,10 This problem is further exacerbated by the lack of inter-reader agreement when analyzing images.17 In a previous study, the CXR was shown to have a moderate sensitivity and specificity of 74.54% and 86.2%, respectively, representing a suboptimal ability to diagnose CHF.9,10 Other notable imaging modalities used in the workup for heart failure are expensive, associated with risks (angiography), and may take a significant amount of time to coordinate and perform (echocardiography, cardiac computed tomography, or magnetic resonance imaging), whereas BNP blood testing can be obtained with ease.4,11 However, it is widely accepted that BNP values should not be used as a standalone method for diagnosing heart failure and should be used in coordination with a detailed history and physical examination.9,10,16,18
Evidence demonstrating the strength of the BNP lab test, relative to other cardiac biomarkers such as troponin and creatine kinase, in diagnosing heart failure have been well studied.9 The ICON-RELOADED study found that BNP testing has a negative predictive value of 98% when the BNP level is less than 300 pg/mL.12 Therefore, 2% of patients with a normal BNP may have a false-negative result. It is unclear how these false-negative cases of CHF would be distributed among the patients in our study. In the best case for radiologists, all patients with false-negative BNP testing would be attributed to the patients with a diagnosis of CHF on CXR (8.5% of patients in the current study). If this were the case, this would still result in 6.5% of patients with an overdiagnosis of CHF, which may lead to these patients receiving unnecessary testing and treatment, or a delay in an accurate diagnosis.
Although the benefits of using BNP testing to rule out heart failure are numerous, there are special considerations when interpreting BNP serum levels. The exact pathophysiology is not certain, but it is known that lower BNP levels may be found in obesity.19,20 It is unclear if this accounts for the relationship we found between BMI and a diagnosis of CHF on CXR in the setting of a normal BNP. This relationship may also reflect the challenging interpretation of chest radiography in obese patients, particularly if those patients receive portable studies.
Lower BNP levels have also been reported in the acute onset of heart failure, or in the setting of ventricular inflow obstruction, for example, in hypertrophic obstructive cardiomyopathy, mitral valve stenosis, or atrial myxoma.16 One study further shows that 1 in 5 patients with symptomatic chronic heart failure presents with BNP levels below diagnostic values, potentially representing those who have a good response to optimal therapy.21 Our study found that CHF on CXR patients were more than twice as likely to have a history of CHF and had a higher median troponin T level, suggesting cardiomyopathy may be present. It is possible that these patients are experiencing cardiac disease that radiologists can detect radiographically, but are not reflected by elevations in BNP. Of note, elevations in BNP due to noncardiac factors have been described in old age, pulmonary disease, pulmonary hypertension, and renal failure.22,23
Limitations of the present study include utilizing data from a single hospital system in an urban setting. The data collection using the Looking-Glass system is dependent on the completeness of the EMR having all the information entered in the record, and information that is omitted, such as the patient’s race or past medical history leading to lapses in our cohort characteristics. Another limitation is the lack of a diagnostic gold standard. Diagnosing heart failure based on CXR also has limitations in that different practices and terminology are used in depicting signs of CHF on imaging. Lack of inter-reader agreement also poses limitations in using CXR to diagnose heart failure in our patients. An important factor not accounted for was whether the BNP result was available before CXR interpretation, and whether the radiologists reviewed BNP results as part of their routine practice.
In conclusion, a normal serum BNP was concordant with CXR results, adding diagnostic confidence in ruling out CHF in a large majority of ED patients (91.5%). Patients with higher BMI, a history of CHF or diabetes with complications, increasing serum BNP level, and those who underwent portable radiographs were more likely to be diagnosed with CHF on CXR in the setting of a normal serum BNP. These patients require further clinical evaluation to facilitate an accurate diagnosis, to improve patient care, and to reduce the economic burden associated with CHF.
The authors acknowledge Eran Bellin, MD, PhD, for his development of Looking Glass and the associated staff for their assistance in creating the study cohort.
1. Benjamin EJ, Mutner P, Alonso A, et al. Heart disease and stroke statistics-2019 update: a report from the American Heart Association. Circulation. 2019;139:e56–e528.
2. Berliner D, Schneider N, Welte T, et al. The differential diagnosis of dyspnea. Dtsch Arztebl Int. 2016;113:834–845.
3. Dao Q, Krishnaswamy P, Kazanegra R, et al. Utility of B-type natriuretic peptide in the diagnosis of congestive heart failure
in an urgent-care setting. J Am Coll Cardiol. 2001;37:379–385.
4. Montez J, Henderson S. Great tool or gold standard? B-type natriuretic peptide and congestive heart failure
. West J Emerg Med. 2011;12:107–108.
5. Mosterd A, Hoes AW. Clinical epidemiology of heart failure
. Heart (British Cardiac Society). 2007;93:1137–1146.
6. O’Connell JB. The economic burden of heart failure
. Clin Cardiol. 2000;23:Iii6–Iii10.
7. Di Somma S, Magrini L, Pittoni V, et al. In-hospital percentage BNP reduction is highly predictive for adverse events in patients admitted for acute heart failure
: the Italian RED Study. Crit Care. 2010;14:R116.
8. Moe GW, Howlett J, Januzzi JL, et al. N-terminal pro-B-type natriuretic peptide testing improves the management of patients with suspected acute heart failure
: primary results of the Canadian prospective randomized multicenter IMPROVE-CHF study. Circulation. 2007;115:3103–3110.
9. Mant J, Doust J, Roalfe A, et al. Systematic review and individual patient data meta-analysis of diagnosis of heart failure
, with modelling of implications of different diagnostic strategies in primary care. Health Technol Assess (Winchester, England). 2009;13:1–207; iii.
10. Sartini S, Frizzi J, Borselli M, et al. Which method is best for an early accurate diagnosis of acute heart failure
? Comparison between lung ultrasound, chest x-ray and NT pro-BNP performance: a prospective study. Intern Emerg Med. 2017;12:861–869.
11. Januzzi JL Jr, Camargo CA, Anwaruddin S, et al. The N-terminal Pro-BNP investigation of dyspnea in the emergency department (PRIDE) study. Am J Cardiol. 2005;95:948–954.
12. Januzzi JL Jr, Chen-Tournoux AA, Christenson RH, et al. N-terminal Pro-B-type natriuretic peptide in the emergency department: the ICON-RELOADED study. J Am Coll Cardiol. 2018;71:1191–1200.
13. Knudsen CW, Omland T, Clopton P, et al. Diagnostic value of B-type natriuretic peptide and chest radiographic findings in patients with acute dyspnea. Am J Med. 2004;116:363–368.
14. Mak G, Ryder M, Murphy NF, et al. Diagnosis of new onset heart failure
in the community: the importance of a shared-care approach and judicious use of BNP. Irish J Med Sci. 2008;177:197–203.
15. Spevack DM, Bowers J, Banerjee A, et al. Diagnostic accuracy of Doppler echocardiography for determining left ventricular diastolic pressure elevation: prospective comparison to chest radiography, serum B-type natriuretic peptide, and chest auscultation. Echocardiography (Mount Kisco, NY). 2008;25:946–954.
16. Wang CS, FitzGerald JM, Schulzer M, et al. Does this dyspneic patient in the emergency department have congestive heart failure
? JAMA. 2005;294:1944–1956.
17. Cardinale L, Volpicelli G, Lamorte A, et al. Revisiting signs, strengths and weaknesses of standard chest radiography in patients of acute dyspnea in the emergency department. J Thorac Dis. 2012;4:398–407.
18. Packer M. Should B-type natriuretic peptide be measured routinely to guide the diagnosis and management of chronic heart failure
? Circulation. 2003;108:2950–2953.
19. Clerico A, Giannoni A, Vittorini S, et al. The paradox of low BNP levels in obesity. Heart Fail Rev. 2012;17:81–96.
20. Madamanchi C, Alhosaini H, Sumida A, et al. Obesity and natriuretic peptides, BNP and NT-proBNP: mechanisms and diagnostic implications for heart failure
. Int J Cardiol. 2014;176:611–617.
21. Tang WH, Girod JP, Lee MJ, et al. Plasma B-type natriuretic peptide levels in ambulatory patients with established chronic symptomatic systolic heart failure
. Circulation. 2003;108:2964–2966.
22. Lewis RA, Durrington C, Condliffe R, et al. BNP/NT-proBNP in pulmonary arterial hypertension: time for point-of-care testing? Eur Respir Rev. 2020;29:156.
23. Nakane T, Kawai M, Komukai K, et al. Contribution of extracardiac factors to the inconsistency between plasma B-type natriuretic peptide levels and the severity of pulmonary congestion on chest x-rays in the diagnosis of heart failure
. Intern Med (Tokyo, Japan). 2012;51:239–248.