The Ability of Heart Failure Specialists to Accurately Predict NT-proBNP Levels Based on Clinical Assessment and a Previous NT-proBNP Measurement

Patient Characteristics and Study Protocol

We enrolled 160 consecutive patients >18 years with known HF attending the outpatient Heart Function, Post-Heart Transplant, and Maintenance Clinics at St. Paul's Hospital in Vancouver, British Columbia, Canada. Ethics approval for the study was obtained from the University of British Columbia/Providence Health Care Research Ethics Board. NT-proBNP levels were drawn from each HF patient upon arrival to clinic. Each patient was then assessed by one of five HF specialists as a regular office visit. All HF specialists were cardiologists trained specifically in HF and all patients were seen in St. Paul’s Hospital, a high volume tertiary care center. During this visit, physicians were asked to estimate the patient's current NT-proBNP level based upon their clinical assessment and all available information from the patient's chart, including a previous NT-proBNP level (if available). Actual and estimated NT-proBNP levels were then recorded by an independent physician, in addition to gender, age, previous NT-proBNP level (if available), current body mass index (BMI), most recent glomerular filtration rate (eGFR), and current New York Heart Association (NYHA) class. NT-proBNP levels were grouped into prognostic categories (<125, 125-1000, 1000-4998 or ≥4999 pg/mL) outlined by a recent study by Rothenburger et al. [10]. If the estimated and the actual NT-proBNP prognostic categories were the same, we considered this as an accurate estimation.

NT-proBNP Measurement

NT-proBNP levels were measured by the Elecsys® proBNP immunoassay on the Roche Elecsys® 1010 immunoassay analyzer.

Statistical Analyses

For univariate analysis, the Chi-square test was used to examine the association between correct physician NT-proBNP estimation and the following factors: gender, BMI <25 kg/m2, age <65, eGFR>60 mL/min./1.73m2, NYHA class I or II, the absence of a previous NT-proBNP measurement, and the category of the most recent prior NT-proBNP level (if available). For multivariable analysis, a logistic regression model was used to investigate the association between the prior NT-proBNP prognostic category and correct physician NT-proBNP estimation, adjusted for gender, BMI, age, eGFR, NYHA, and the time difference between the current test and prior test.

NT-proBNP/BNP in the Diagnosis, Prognostic Evaluation, and Management of Patients with CHF

To date, NT-proBNP/BNP has been studied in a variety of clinical settings. In the diagnosis of congestive HF, the use of NT-proBNP and BNP measurement in the emergency department has been well established [5,6]. In contrast, the use of neurohormones in the outpatient diagnosis of HF has been less well studied. In a study by Groenning et al., NT-proBNP measurement in the outpatient setting identified patients with symptoms of HF and a low left ventricular ejection fraction (LVEF) of less than 40% with a positive predictive value of 0.11 and a negative predictive value of 1.00 [4]. In a study by Wright et al. that recruited patients presenting to their general practitioner with dyspnea and/or edema, knowledge of the NT-proBNP level improved the diagnostic accuracy of heart failure from 8% (control group) to 21% (NT-proBNP group) (p=0.002)[3]. Finally, Zaphiriou et al. performed a study on 306 patients referred to a heart failure clinic by their general practitioners with suspected heart failure. They found that NT-proBNP and BNP were good rule-out tests with high negative predictive values (0.97 and 0.87, respectively) but low positive predictive values (0.44 and 0.59, respectively)[2].

Recent research suggests that serial NT-proBNP measurements may have a role in prognostic evaluation [7-2] and therapy optimization [13-16]. In prognostic evaluation, NT-proBNP/BNP has been found to be an independent predictor of sudden death [7], HF hospital admissions [4], and depressed left ventricular function [7]. In a recent systematic review of 19 studies using BNP to estimate the relative risk of death or cardiovascular events in HF patients, BNP was found to be a strong prognostic indicator [18]. Each 100 pg/mL increase in BNP was associated with a 35% increase in the relative risk of death. Further, Maisel et al. in the Rapid Emergency Department Heart Failure Outpatient Trial (REDHOT) study found that BNP is a potent predictor of the 90-day combined event rate of congestive HF visits or admissions and mortality in patients admitted from the emergency department with acute HF [19]. In the Valsartan Heart Failure Trial, HF patients with an initial BNP level greater than 97 pg/mL were 2.1 times more likely to die at 12 months [20]. As well, in the Australia/New Zealand Heart Failure Study, BNP was superior to LVEF in predicting mortality and HF in patients with ischemic LV dysfunction [9]. Above-median BNP levels conferred a 3-fold-increased risk of mortality. In pre-heart transplant assessment [10], an arbitrary cutoff point of >2000 pg/mL for patients with dilated cardiomyopathy and >1000 pg/mL for patients with LV dysfunction secondary to coronary artery disease included 90% of all patients who were clinically considered for cardiac transplantation. As well, BNP appears to be useful in predicting the progression of HF after cardiac resynchronization therapy (CRT) [21]. Specifically, patients with high BNP values after 1 month of CRT have worse prognosis during follow-up.

Finally, in therapy optimization, Troughton et al. demonstrated that titration of therapy guided by NT-proBNP levels in symptomatic HF patients decreased hospital re-admission rates and cardiac mortality at 9.5 months follow-up, as compared to conventional follow up [14]. BNP may guide the introduction of medications in HF. Richards et al. demonstrated that carvedilol reduced mortality rates in HF to a greater extent in patients with higher pretreatment BNP levels (9). More recently, in the Systolic Heart Failure Treatment Supported By BNP (STARS-BNP) Multicenter Randomized Trial patients randomized to the BNP guided treatment group suffered fewer events (death or hospitalizations) for HF than in the control group (p<0.001) [16].

The above studies suggest that knowledge of an initial NT-proBNP and perhaps repeat NT-proBNP levels may guide diagnosis, prognosis and therapy management. The most recent HF guidelines, specifically, the 2007 Canadian Cardiovascular Society Consensus Conference Recommendations on Heart Failure, support the initial measurement of NT-proBNP/BNP for diagnosis (class I recommendation) and prognostic stratification (class IIa recommendation). The guidelines do not however support sequential measurements of BNP/NT-proBNP levels in the outpatient setting (class IIb recommendation) [17]. To date, no study has demonstrated that routine measurement of NT-proBNP levels in the outpatient setting is cost-effective. In our study, HF specialists were relatively accurate at estimating NT-proBNP levels when a prior NT-proBNP measurement was available, irrespective of the time interval between measurements. In contrast, they were less accurate at predicting NT-proBNP levels in new patients with no prior NT-proBNP level and in patients with patients with a prior NT-proBNP level ≥4999 pg/mL.

Predictors of NT-proBNP Estimation Accuracy

Recent studies suggest that NT-proBNP levels are altered in overweight individuals. Specifically, NT-proBNP levels are significantly lower in overweight and obese individuals with HF compared with lean patients [22,23]. Age and gender alter NT-proBNP levels. Redfield et al. found in a recent study that BNP increased significantly with age and was higher in women than men [24]. Renal dysfunction also alters NT-proBNP levels in that eGFR and BNP are inversely related [25].

We hypothesized that age>65 years, female gender, elevated BMI (>25 kg/m²), renal dysfunction (eGFR <60mL/min./1.73m²), NYHA class I or II, and absence of a prior NT-proBNP level would be predictive of NT-proBNP estimation accuracy. Surprisingly, we found that a prior NT-proBNP measurement was the only factor significantly associated with the accuracy of physician estimation. Age >65 years, female gender, elevated BMI (>25 kg/m²), renal dysfunction (eGFR <60mL/min./1.73m²), NYHA class I or II were not associated with accurate NT-proBNP estimations. In our study, physicians had access to all patient records in the heart failure clinic, including GFR, BMI, etc. We believe that HF specialists purposefully altered their NT-proBNP estimation as a result of this knowledge enabling them to guess correctly.

Interestingly, in patients with a prior NT-proBNP level, accuracy reduced greatly as the prior NT-proBNP levels increased. Compared to patients with a prior NT-proBNP level <125 pg/mL, physicians were 95% less likely to get a correct estimation, in patients with the highest prior NT-proBNP level (≥4999 pg/mL). Karabulut et al. found that while NT-proBNP levels increase significantly with each increasing class of disease, there is a wider range of NT-proBNP levels in NYHA classes III and IV [26]. This may account for our relative inaccuracy at estimating NT-proBNP levels in higher NYHA classes. In these patients, it may be useful to measure NT-proBNP levels on a more routine basis.