What are likelihood ratios in the evaluation of brain natriuretic peptide (BNP)?

Updated: Jan 08, 2018
  • Author: Donald Schreiber, MD, CM; Chief Editor: Erik D Schraga, MD  more...
  • Print


BNP levels in several pilot studies had a strong correlation with the severity of illness and were very reliable in differentiating heart failure from pulmonary disease. The recommended thresholds of less than 100pg/mL to rule out heart failure and more than 500pg/mL to rule in heart failure have been estimated to have the following likelihood ratios (LRs): LR-negative = 0.13 and LR-positive = 8.1. These different cutoffs create an intermediate range of 100-500pg/mL with LR-positive of only 1.9. Therefore, an intermediate BNP result alone cannot be used to rule in or rule out heart failure.

To explore the effect of clinical variables on intermediate BNP values, the Breathing Not Properly group calculated LR, accounting for a history of heart failure and/or coronary artery disease (CAD), lower-extremity edema, pulmonary edema, cephalization of the pulmonary vasculature, or cardiomegaly. In all patients, the LR-positive for a history of heart failure alone was 4.6. [35] As expected, the cumulative LR-negative was 0.02 for BNP < 100 pg/mL, no history of heart failure, and 0 or 1 clinical feature. [4, 5]

At the opposite extreme, a BNP level of greater than 500 pg/mL, a history of heart failure, and 2 or more clinical features had an LR-positive of 37. Other LRs were as follows:

  • A midrange BNP level with a history of heart failure - Had a modest cumulative LR of 4.3, which was not significantly different from the LR of heart failure alone [4, 5]

  • Any 2 or more clinical features, a heart failure history, and a mid-range BNP level - Had a cumulative LR-positive of 10

  • A mid-range BNP level, no history of heart failure, and 0 or 1 clinical criterion - Had an intermediate cumulative LR-negative of 0.7, which was not clinically useful

These data emphasize that an intermediate BNP result that is in conjunction with 2 or more clinical features and a history of heart failure is reasonably predictive of an ultimate diagnosis of an exacerbation of heart failure. However, an intermediate BNP result without these criteria (ie, no history of heart failure and 0 or 1 clinical variable) does not help to rule in or rule out heart failure as the etiology of the patient's dyspnea. Care must be taken in interpreting these intermediate results. The use of serial BNP measurements for monitoring the treatment of heart failure is not well established. [36, 37, 38]

A study by Alehagen et al found that among elderly patients with symptoms of heart failure, elevated concentrations of copeptin and the combination of elevated concentrations of copeptin and NT-proBNP were associated with an increased risk of all-cause mortality. [39]

Steinhart et al derived and validated a diagnostic prediction model for acute heart failure that incorporates clinical assessment and NT-proBNP. Variables used to predict acute heart failure were age, pretest probability, and log NT-proBNP. Validation of the model in 1073 patients showed that LRs for acute heart failure with NT-proBNP were 0.11 for cut-point values of less than 300pg/mL, increasing to 3.43 for values 2700-8099pg/mL and to 12.80 for values 8100pg/mL or higher.

When the model was applied to external data, 44% of patients who had been clinically classified as having intermediate probability of acute heart failure were appropriately reclassified to either low- or high-probability categories, with negligible (< 2%) inappropriate redirection. [40]

Did this answer your question?
Additional feedback? (Optional)
Thank you for your feedback!