Abstract 4359: Adjusting for Major Clinical Covariates Improves the Diagnostic Performance of Brain Natriuretic Peptide for Distinguishing Cardiac from Non-cardiac Dyspnea
AHA 2008 Abstract #1
Introduction: Dyspneic patients with a BNP <100 pg/mL are unlikely to have heart failure (HF). Age, atrial fibrillation (AF), renal dysfunction, and body mass index (BMI) affect BNP levels independent of HF. We hypothesized that BNP cutoff levels could be adjusted in the above clinical subgroups to more accurately identify dyspneic patients with a non-cardiac etiology.
Methods: Patients presenting to the Salt Lake City VA Medical Center with dyspnea in 2007 were eligible. Data were collected by retrospective chart review. Physicians blinded to BNP values adjudicated outcomes. Dyspnea was attributed to HF if Framingham criteria for HF were met. The diagnostic characteristics of a BNP of 100 pg/mL were determined for the entire cohort and for the following subgroups: age > 75 years, history of AF, creatinine > 2 mg/dL, and BMI °¥ 30 kg/m2. Adjusted BNP cutoff values for each subgroup were determined that retained the sensitivity of the BNP cutoff value of 100 pg/mL in the entire cohort.
Results: 349 patients met inclusion criteria. The adjusted BNP cutoff values that correspond to 91% sensitivity (the sensitivity of the BNP cutoff value of 100 pg/mL in the entire cohort) in the different subgroups were higher in those > 75 years of age, those with AF, and those with creatinine > 2 mg/dL but lower for the subgroup with BMI °¥ 30 kg/m2 (Table⇓). Due to improved specificity of these subgroup-specific BNP cutoff values, 22 (14%) patients > 75 years old, 5 (4%) patients with AF, and 12 (23%) patients with a creatinine > 2 mg/dL could be correctly reclassified as having non-cardiac dyspnea. The cutoff adjusted for BMI identified an additional 7 (4%) of obese patients as having HF but lacked sufficient specificity to increase overall diagnostic accuracy in this subgroup.
Conclusions: Adjusting for the presence of clinical covariates may improve the accuracy of BNP in distinguishing cardiac from non-cardiac dyspnea. This approach is undergoing validation.