Author + information
- Susan P. Wright, MBChB*,
- Robert N. Doughty, MD, MRCP, FRACP*,* (, )
- Ann Pearl, MA (Hons), MBChB†,
- Greg D. Gamble, MSc*,
- Gillian A. Whalley, MHSci, DMU*,
- Helen J. Walsh, BSc*,
- Gary Gordon, MBChB, FCP(SA), FRACP*,
- Warwick Bagg, MD, FRACP*,
- Helen Oxenham, MBBS, MRCP*,
- Tim Yandle, PhD‡,
- Mark Richards, MD, PhD, DSc, FACC, FRACP, FRSNZ‡ and
- Norman Sharpe, MD, FRACP, FACC*
- ↵*Reprint requests and correspondence:
Dr. Robert N. Doughty, Department of Medicine, University of Auckland, Private Bag 92019, Auckland, New Zealand.
Objectives The purpose of this study was to determine the effect of amino-terminal pro-brain natriuretic peptide (N-BNP) on the diagnostic accuracy of heart failure (HF) in primary care.
Background The accurate diagnosis of patients with suspected HF presenting in primary care is difficult. Amino-terminal pro-brain natriuretic peptide is present in high levels in cardiac dysfunction and may improve the diagnostic accuracy of HF in primary care.
Methods The Natriuretic Peptides in the Community Study was a prospective, randomized controlled trial of the effect of N-BNP on the accuracy of HF diagnosis. Patients presenting to their general practitioner (GP) with symptoms of dyspnea and/or peripheral edema were included. The GPs formulated an initial diagnosis based on clinical assessment. All patients underwent a full cardiologic assessment that included echocardiography and N-BNP. Each patient was randomized to the BNP group (GP received the N-BNP result) or the control group (GP did not receive the N-BNP result). Patients were then reviewed by their GP, and their diagnosis was reviewed. The primary end point was the accuracy of the GPs' diagnoses compared with the panel standard.
Results A total of 305 patients were included; mean age was 72 years, 65% were female. Seventy-seven patients met the panel criteria for HF. The diagnostic accuracy improved 21% in the BNP group and 8% in the control group (p = 0.002). The main impact of N-BNP measurement on diagnostic accuracy was the GPs' correctly ruling out HF. The number needed to diagnose by N-BNP measurement was seven patients.
Conclusion This study demonstrates that N-BNP measurement significantly improves the diagnostic accuracy of HF by GPs over and above customary clinical review.
The accurate diagnosis of heart failure (HF) presenting in primary care is difficult. Clinical assessment of symptoms and signs has limited sensitivity and specificity (1). The clinical syndrome of HF is particularly difficult to diagnose in the elderly and in the presence of accompanying respiratory disease (1). For patients diagnosed with HF in primary care, only a minority have that diagnosis confirmed after cardiologic assessment (2). Hence, there is a need for a diagnostic test that can improve the accuracy of HF diagnosis in primary care.
Natriuretic peptides are released in response to increased intra-cardiac volume or pressure and can be assayed on venous blood samples. Brain natriuretic peptide (BNP), released mainly by the ventricle, is elevated in HF and offers promise as a diagnostic test in primary care. The amino-terminal pro-BNP (N-BNP) fragment, which is secreted in a 1:1 ratio with the bioactive 32-amino acid carboxy-terminal fragment (3)is present in higher levels in cardiac dysfunction (4), so it may be a superior diagnostic discriminator in patients with suspected HF.
Measurement of BNP is useful in the assessment of patients with acute dyspnea referred to the hospital (5–8)and patients with suspected HF referred to a rapid access hospital clinic (9). In newly symptomatic patients with suspected HF referred to a rapid access clinic, BNP had high negative predictive value, which is particularly important when community diagnosis depends on relatively non-specific clinical criteria (9).
However, not all community-based BNP studies are consistent with studies in the hospital or clinic setting. Observational and cross-sectional studies performed in primary care cohorts show poorer correlation of BNP with indices of HF such as the degree of left ventricular (LV) dysfunction (10)and prescription of frusemide (11). Brain natriuretic peptide can be elevated by many common conditions, including renal impairment (12), myocardial infarction (13), chronic lung disease (14), hypertension and LV hypertrophy (15), atrial fibrillation (AF) (16), valvular heart disease (17), thyroid disorders (18), and age itself (19,20). Brain natriuretic peptide may also be decreased by common cardiovascular medications, including loop diuretics (21,22)and angiotensin-converting enzyme (ACE) inhibitors (23,24), and thus may be normal in patients with compensated chronic HF.
Novel diagnostic technologies are often implemented before evidence of their accuracy and clinical utility has been established in the population of interest. Ideally, new tests should be assessed in an unselected consecutive series of patients presenting with symptoms of the disease in primary care (25). The Natriuretic Peptides in the Community Study is the first prospective, randomized controlled trial to determine the effect of N-BNP on the accuracy of HF diagnosis in the community. Randomization of plasma N-BNP measurements allowed this test to be assessed against customary diagnostic practice in the “real world” setting of health care in the community.
Patients 40 years of age or more presenting to their general practitioner (GP) with symptoms of dyspnea and/or edema of recent onset were eligible for this study. Patients could have comorbid medical conditions and could have been receiving any treatment. Patients were excluded if they required urgent admission to the hospital or were unable to provide informed consent. Further exclusions were hospital admission with a diagnosis of HF, echocardiography for assessment of LV function, or outpatient cardiologic assessment for the evaluation of dyspnea or edema during the previous 12 months. The Auckland Ethics Committee approved the study. All patients provided written informed consent.
Eligible patients presenting to participating GPs with dyspnea and/or edema were referred to the study by their GPs. At this visit (Initial GP Visit, Fig. 1), the GPs recorded whether or not HF was suspected on the basis of patient history and clinical examination (the initial GP diagnosis). Commencement of any treatment was at GP discretion.
Patients were contacted within 24 h of the initial GP visit and booked to attend a study visit (Study Visit, Fig. 1). At this visit, each patient was assessed clinically by a cardiologist, and electrocardiography, chest radiograph, blood collection for N-BNP measurement, and transthoracic echocardiography (Philips HDI 5000, Bothell, Seattle, Washington) were performed. Blood was collected using standard venepuncture technique into tubes containing ethylenediaminetetraacetic acid. Samples were centrifuged and frozen at −70°C. The N-BNP was measured by radioimmunoassay at the CardioEndocrine Research Laboratory, Christchurch, New Zealand (4).
An independent study investigator performed computer randomization with the patient as the unit of randomization. Each patient was randomly allocated to either the BNP group or the control group. Patients in the BNP group had their N-BNP result faxed to their GPs with a standardized interpretative comment (Table 1). This included the laboratory normal range for N-BNP (0 to 50 pmol/l) and the range in which HF was very likely (>150 pmol/l) (5). Patients in the control group had a laboratory results sheet faxed to their GPs informing them that their patients' N-BNP results were not available.
At the time of this study, BNP measurement was not routinely available to GPs in Auckland. Before the study, all GPs received a standardized 30-min education session on the interpretation of N-BNP and were asked to apply the interpretative comment shown in Table 1.
Patients were then reviewed by their GPs (GP review visit, Fig. 1), either with their N-BNP result (BNP group) or according to customary clinical assessment (control group). A final diagnosis was made by the GP at this visit.
Panel standard diagnosis of HF
The definite reference test for the purposes of this study was a clinical diagnosis of HF made by a panel of four experts (three cardiologists and one GP) using the rigorous application of pre-defined criteria. Panel members were independent of all study procedures and were blinded to each patient's group allocation and N-BNP result. The panel reviewed all clinical data for each patient, including electrocardiogram, chest radiograph, and echocardiograph (but excluding natriuretic peptide measurements), and decided whether each individual met the case definition of HF. The panel decision was based on the European Society of Cardiology Working Group on HF diagnostic criteria (26). To meet the case definition of HF, patients had to have appropriate symptoms and clinical signs of pulmonary or peripheral congestion in the presence of objective evidence of cardiac dysfunction. If doubt remained, a beneficial response to treatment was considered. The panel standard used in this study was based on previous studies of the diagnosis of HF (9,27).
Study end point
The primary study end point was GP accuracy in the final diagnosis of the BNP and control groups. This was measured as percent correct final diagnoses in each group compared with the panel standard.
Preliminary estimates of the diagnostic accuracy of HF in general practice suggested that two groups of 100 patients each would be required to show a 20% absolute increase in diagnostic accuracy (assuming 5% significance and 80% power). However, it was understood that this estimate would need to be refined during the study, depending on the patients who were referred. Thus, a pre-planned re-estimate of the sample size was completed after the first 100 participants were recruited. To maintain an overall 5% significance level, the Bonferroni correction was applied to the critical value. For the main result to maintain a 5% statistical significance level after this interim analysis, a p value of <0.025 (for a two-tailed test) was required. This interim analysis resulted in an increase of the required sample size to 300 (150 patients in each group).
The proportion of correct diagnoses of HF in the BNP and control groups was compared using the Fisher exact test. The Fisher exact test was also used to examine the change in diagnostic accuracy before (Initial GP Visit) and after (GP Review Visit) N-BNP results were made available. Comparisons of continuous variables were performed using the Student ttest (normally distributed variables) and Wilcoxon's test (non-normally distributed variables). In multivariate analysis, the change in correct diagnosis between first and second GP visits was modeled using logistic regression to test the effect of N-BNP results on the accuracy of diagnosis after adjusting for the potentially confounding effect of treatment with loop diuretics, ACE inhibitor medications, and time between GP visits. Number needed to diagnose analysis was performed using change in the absolute numbers of correct diagnoses. All tests were two-tailed, and a 5% significance level was maintained throughout.
Ninety-two GPs referred a total of 319 patients with symptoms of dyspnea and/or edema to the study (Fig. 1), of whom 307 consented to the study. One patient later withdrew consent, and one patient was unable to attend a study visit. The remaining 305 patients are included in this report. The BNP and control groups were comparable at baseline, except for a minor difference in mean age (p = 0.04) (Table 2). The mean age of the overall group was 72 years (range 40 to 95 years), and 65% were female. A total of 49% of patients presented with dyspnea only, 12% with edema only, and 39% with both symptoms. Although the majority appeared to have New York Heart Association class II symptoms, functional ability assessed by a 6-min walk distance was markedly impaired (mean 262 meters), reflecting an elderly cohort of patients with multi-system disease. Comorbid diagnoses were common (Table 2), and 35 patients (11%) were in AF.
The median delay between a patient's initial GP visit and attendance at the study visit was seven days (interquartile range [IQR] 5 to 11). The median delay between patient attendance at the study visit and attendance at the GP review visit was 24 days (IQR 18 to 29).
GP diagnosis of HF and the panel standard
The initial diagnostic impression of referring GPs based on patient history and examination was that 215 (70%) patients had HF. However, only 77 patients (25%) were allocated a diagnosis of HF by the panel. The consistency of the panel standard diagnosis of HF was assessed by the representation to the panel of a random sample of 10% of patients; kappa coefficient = 0.81. The receiver operating characteristic curve for N-BNP is shown in Figure 2, with area under the curve = 0.85. The sensitivity and specificity were maximized at a cutoff of N-BNP 100 pmol/l.
The effect of N-BNP on diagnostic accuracy (Table 3)
At the GP review visit, each GP reviewed the patient either with the patient's N-BNP result (BNP group) or using customary clinical assessment (control group). The provisional diagnosis made at the initial GP visit was then reviewed. The diagnostic accuracy in the N-BNP group improved 21% (from 49% diagnoses correct to 70% correct). This compared with an improvement of 8% in the control group (from 52% diagnoses correct to 59% correct, p = 0.002, Table 3, Fig. 2). The increase in diagnostic accuracy in the BNP group was mainly due to an increase of GPs correctly ruling out HF (Table 3). In the BNP group, 30 patients initially thought to have HF were correctly re-characterized as not having HF when clinically reviewed with the N-BNP result, compared with 14 patients correctly re-characterized in the control group using customary clinical review (Table 3). Hence, the N-BNP results improved accuracy in the diagnosis of HF by GPs, when compared with customary clinical assessment.
The GPs involved in the study referred a median of two patients (range 1 to 14), with a median interval between referral of 184 days (IQR 50 to 355) over the recruitment period of 24 months. Of the GPs who referred more than two patients to the study (n = 57, 62%), the median interval between referral of patients to the study was 42 days (IQR 12 to 106). Trends in diagnostic accuracy in this group of GPs were examined for a learning effect over the course of the study. Diagnostic accuracy did not change over the course of the study for these GPs. In addition, patterns of diagnostic accuracy in the subgroup of GPs with patients randomized to the BNP group and then patients randomized to the control group during the study (n = 7) were examined for contamination. Diagnostic accuracy remained unchanged over the course of the study in this group of GPs (p = 0.63).
Multivariate logistic regression was performed to examine if N-BNP still improved diagnostic accuracy when patients had been commenced on loop diuretics or ACE inhibitor medications at the initial GP visit, or if patients were being treated with these medications before this visit. The N-BNP had a clear beneficial effect on the diagnostic accuracy after controlling for treatment (p = 0.012). In addition, in a multivariate model the delay between the initial GP visit and the study visit did not alter the effect of N-BNP on diagnostic accuracy (p = 0.76).
Characterization of patients with HF
The characteristics of the 77 patients with a confirmed diagnosis of HF are shown in Table 4. The patients with HF were more likely to be male and to have diabetes, AF, or a history of myocardial infarction but not hypertension. The patients with HF were more likely to be receiving frusemide and ACE inhibitor medication at the study visit. Echocardiographic parameters (LV size, function, and LV mass) and cardio-thoracic ratio were significantly different for patients with HF (Table 4). Forty-three (56%) of the patients with HF had normal LV systolic function. Patients with HF had significantly higher mean N-BNP, 286 pmol/l (SD 319) compared with 61 pmol/l (SD 67) in those patients without HF.
Heart failure is difficult to diagnose in primary care (28). Patients are often elderly with comorbidity, symptoms may be mild, routine clinical assessment lacks specificity, and echocardiography may not be universally available (29). These factors contribute to considerable diagnostic problems that are more important and frequent than often recognized. Over-diagnosis of HF in the community is a well-documented phenomenon, with only a quarter to a third of patients with suspected HF by GPs in several studies having HF confirmed on further cardiologic assessment (1,2,9). In the current study, GPs suspected that 70% of patients had HF, compared with the final panel diagnosis of definite HF in only 25% of patients.
The study demonstrated that the availability of N-BNP to GPs resulted in an absolute improvement in diagnostic accuracy of 21% compared with only 8% with clinical review alone (p = 0.002). The main impact was in enabling GPs to correctly rule out HF. This is consistent with the finding of high negative predictive value of BNP in previous studies (8,9). This clearly has important implications for the improvement in the accuracy of diagnosis of HF in primary care.
Previous studies examining the potential role of BNP measurement in the diagnosis of HF have concluded that it is likely to be of benefit (6–9). However, these studies had several limitations, including not recruiting consecutive symptomatic patients (6)and not comparing BNP with standardized criteria for the clinical diagnosis of HF (30). Several studies were conducted in the acute-care hospital setting and may not be generalizable to primary care (6–8). Screening studies have indicated that false positive results may present problems in elderly primary care populations in which other medical conditions known to elevate natriuretic peptide levels are common (10). Importantly, none of the previous studies involved random allocation of the N-BNP result to the physician. This study is the first prospective, randomized controlled trial of the effectiveness of N-BNP in a community care setting. This study design has extended the previous studies of BNP in the assessment of patients with suspected HF to evaluate the impact of this test on clinical decision-making; a final step in the evidence base of clinical diagnosis (31). The study was conducted in the day-to-day setting of primary health care delivery in which other diagnostic strategies commonly used by GPs, such as empirical prescribing of treatment and subsequent clinical reassessment, continued to operate.
New diagnostic tests must be assessed against agreed clinical diagnostic criteria for HF rather than associated surrogates such as an arbitrary level of LV ejection fraction. The trial used a rigorous clinical diagnosis of HF made by a blinded, independent panel according to standard criteria (26). By showing a clear improvement in the accuracy of diagnosis of HF in primary care, this study moves the use of BNP from the field of observational studies to that of community-based clinical trials.
The simplicity of a single blood test that can be taken without special preparation or sampling conditions makes widespread use of this test feasible in clinical practice, including primary care settings. Although N-BNP improved diagnostic accuracy in this study, other tests such as echocardiography may still be required once HF has been diagnosed (32). However, the use of N-BNP may reliably rule out HF, allowing the targeted use of tests such as echocardiography in clinical settings in which availability is limited. The availability of portable BNP meters designed for point-of-care use and centralized, automated N-BNP assays using pre-existing laboratory platforms indicates that BNP measurement in many settings is becoming common and turn-around time more rapid.
This study examined a cohort of ambulatory symptomatic patients in primary care. The study design ensured that all patients attended a review visit with their referring GPs as part of the study. Clinical review may not be universal practice in primary care, but the study design combined the need for a standardized study protocol in both control and BNP groups while operating in the context of day-to-day health care delivery.
In clinical practice, natriuretic peptide measurement would occur at initial presentation, unlike this study, in which it was performed at a study visit a median of seven days later. However, N-BNP had a clear beneficial effect on the diagnostic accuracy of HF despite this delay and despite the initiation of empirical therapy in many patients.
Selection bias is a potential problem in every diagnostic study. Participating GPs were encouraged to refer all patients presenting with dyspnea and/or edema to minimize selection bias and allow the evaluation of the diagnostic test in a consecutive sequence of primary care patients. The patients recruited in this study were typical of those in primary care with a provisional diagnosis of HF. Approximately half were women, mean age was in the mid 70s, and comorbidity was common.
The Natriuretic Peptides in the Community study is the first prospective, randomized controlled trial showing the positive effect of N-BNP measurement on the diagnostic accuracy of HF in primary care. The change in the percentage of improvement in correct diagnoses of the BNP group (21%) compared with the control group (8%) allows calculation of the number needed to diagnose: measurement of N-BNP in seven patients with a provisional diagnosis of HF is needed to re-characterize one patient correctly.
This study shows that N-BNP measurement significantly improves the diagnostic accuracy of HF by GPs. Amino-terminal pro-brain natriuretic peptide is particularly important in decreasing the over-diagnosis of HF that occurs in primary care. This study now allows the evidence-based recommendation that availability of N-BNP measurement to GPs will significantly improve the diagnostic accuracy of HF in primary care.
We acknowledge the contribution of Ms. J. Pomfret and Ms. S. Muncaster who contacted the patients, arranged patient appointments, provided clinical care, and supervised data collection. We acknowledge the participation of Auckland general practitioners who participated in this study (the Natriuretic Peptides in the Community Study GP Collaborative Group). The full list of GPs appears online.
For the full listing of Auckland general practitioners who participated in this study, please see the online appendixfor this article in the November 5, 2003 issue of JACCat http://www.cardiosource.com/jacc.html.
☆ This trial was supported by grants from the Health Research Council of New Zealand and the New Zealand National Heart Foundation. S. P. Wright is supported by a Health Research Council of New Zealand Fellowship in clinical research. Previous support (S.P.W.) includes Cardiac Society of Australia and New Zealand Research Scholarship and the Royal Australasian College of Physicians Pfizer Medical Research Fellowship. R. N. Doughty was the recipient of the New Zealand Heart Foundation Bank of New Zealand Senior Fellowship. A. M. Richards is the holder of the National Heart Foundation (New Zealand) Chair of Cardiovascular Studies. Gottlieb C. Friesinger, II, MD, acted as the Guest Editor for this manuscript.
- angiotensin-converting enzyme
- atrial fibrillation
- brain natriuretic peptide
- general practitioner
- heart failure
- interquartile range
- left ventricular
- amino-terminal pro-brain natriuretic peptide
- American College of Cardiology Foundation
- Wheeldon N.M.,
- MacDonald T.M.,
- Flucker C.J.,
- McKendrick A.D.,
- McDevitt D.G.,
- Struthers A.D.
- Dao Q.,
- Krishnaswamy P.,
- Kazanegra R.,
- et al.
- Cowie M.R.,
- Struthers A.D.,
- Wood D.A.,
- et al.
- Smith H.,
- Pickering R.M.,
- Struthers A.,
- Simpson I.,
- Mant D.
- Hetmanski D.J.,
- Sparrow N.J.,
- Curtis S.,
- Cowley A.J.
- Buckley M.G.,
- Sethi D.,
- Markandu N.D.,
- Sagnella G.A.,
- Singer D.R.,
- MacGregor G.A.
- Richards A.M.,
- Nicholls M.G.,
- Yandle T.G.,
- et al.
- Rossi A.,
- Enriquez-Sarano M.,
- Burnett J.C.,
- Lerman A.,
- Abel M.D.,
- Seward J.B.
- McClure S.J.,
- Caruana L.,
- Davie A.P.,
- Goldthorpe S.,
- McMurray J.J.
- Kohno M.,
- Yokokawa K.,
- Kano H.,
- Minami M.,
- Hanehira T.,
- Yoshikawa J.
- Murdoch D.R.,
- McDonagh T.A.,
- Byrne J.,
- et al.
- Sackett D.L.,
- Haynes R.B.
- Task Force on Heart Failure of the European Society of Cardiology
- Cowie M.R.,
- Wood D.A.,
- Coats A.J.S.,
- et al.
- Hobbs F.D.
- Hobbs F.D.R.,
- Jones M.I.,
- Allan T.F.,
- Wilson S.,
- Tobias R.
- Landray M.J.,
- Lehman R.,
- Arnold I.
- Knottnerus J.A.,
- van Weel C.,
- Muris J.W.M.
- Task force for the diagnosis and treatment of chronic heart failure