Breathing Not Properly Study. NEJM 2002; 347: 3.
Background: In an attempt by the Centers for Medicare and Medicaid Services to provide cost-effective treatment for patients with CHF, rapid and accurate differentiation of CHF from other causes of dyspnea must be accomplished.
Methods: Prospective, multi-center, International study. N = 1586. Exclusions include < 18y, clearly not CHF related dyspnea (trauma), AMI, ARI, and UA. BNP levels were blinded to the physician.
Gold Standard: Two cardiologists who independently classified the diagnosis as 1) dyspnea due to CHF, 2) acute dyspnea due to noncardiac causes in a patient with h/o LV dysfunction, or 3) dyspnea not due to CHF. They were allowed to use the Framingham CHF score (2 major or 1 major and 2 minor) and/or the NHANES CHF score (3+).
Baseline: Mean age 64 years, 56% men, 49% white, 45% black, 7% had an S3 canter, 43% had rales, 22% had JVD, and 42% had edema.
Results:
BNP: Area under the ROC = 0.91 (p < 0.001). Using 100 pg/mL had a sensitivity of 90%, a specificity of 76%, a PPV 79%, a NPV 89%, a positive likelihood ratio of 3.75, and a negative likelihood ratio of 0.13. Using a cutoff of 50 pg/mL, the NPV was 96%. The mean BNP differed per NYHA classes. The levels were approximately 244, 389, 640, and 817, for class I, II, III, & IV, respectively.
Application:
Using the Boston Heart Failure Criteria (Marantz. Circulation 1988; 77: 607), patients who have "possible" HF by the criteria will have an underlying incidence (pretest probability) of LVEF < 40%, 20-35% of the time. Using the +LR (3.75) of a BNP > 100 we see that the posttest probability in that patient is 40-60%. The presence of a JVD increases the posttest probability to about 80%, an S3 to 92%.
Absolute values are one means of assessing pts in HF with BNP (or even using pro-nt-bnp) but perhaps more useful is knowing what their 'baseline' BNP level is. If by lab hx you know they have a BNP of 500 at baseline, an admission value of 550, though significant by the above criteria (Boston Heart etc) has a dramatically reduced PPV vs. an admission BNP of 2000 in the same pt. Food for thought. Peace.
Your point is well taken.
However, the value of BNP, a marker of heart failure, depends on its use as a diagnostic tool, an indicator of prognosis, or as a guide for effective treatment.
As a diagnostic tool, we know from data that increasing levels are associated with increasing specificity and less sensitivity. In your example, a level of 500 is associated with a specificity of 93.9% and a PPV of 85.4%. And in the same paper, a level >800 pg/mL found in 47% of patients with heart failure, have a specificity of 96.4% and PPV of 88.2%, a negligible increase in predictive value (JACC 2005; 46: 838). The change in value does not change the fact that they have underlying CHF.
As a prognostic indicator, a study looking at the NT-proBNP levels of the Val-Heft study showed that absolute/relative changes in levels was associated with worse prognosis. The hazard ratios to compare risk of death by baseline NT-proBNP and changes over 4 months was 1.877 when it went from high to high, 1.699 when it went from low to high, 0.614 when it went from high to low, and 1 (reference) when it went from low to low (JACC 2008; 52: 997).
Gheorghiade has pointed out that reductions in BNP during treatment did not translate into improved outcomes. In the SURVIVE and REVIVE trials, reductions in BNP levels over 5 days by the use of levosimendan vs dobutamine or placebo (respectively) did not show a survival benefit and an increase in adverse events from the drug usage (SURVIVE. JAMA 2007; 297: 1883; REVIVE. AHA. Annual Scientific Session; Nov 13-16,2005; Dallas, Tx).
Diagnostically speaking, an absolute elevated level has a given likelihood for having CHF. A change from normal(low) levels to high levels is associated with worse prognosis but has little change in the predictive power of the existence of HF. From the STARS-BNP study, it suggested that decreased levels to normal (16% reached this level) may be associated with more aggressive treatments and better outcomes (JACC 2007; 49: 1733).