Usefulness of adjusting for clinical covariates to improve the ability of B-type natriuretic peptide to distinguish cardiac from noncardiac dyspnea.

Certain clinical characteristics affect brain natriuretic peptide (BNP) levels independently of clinical heart failure (HF). However, it is unclear how to adjust the diagnostic cutoffs of BNP for these variables. We hypothesized that adjusting for important covariates would improve the diagnostic accuracy of BNP for HF in the emergency room setting. We included patients presenting with dyspnea at the Salt Lake City Veterans Affairs Medical Center. Physicians unaware of the BNP values adjudicated the outcome as dyspnea due to HF or noncardiac dyspnea. Subgroup analyses and logistic regression analysis were used to adjust the BNP cutoffs. The mean age of the study population (n = 335) was 72 +/- 11 years. A BNP of 100 pg/ml had a sensitivity of 91%, and a BNP of 400 pg/ml had a specificity of 92%. The covariates age, history of atrial fibrillation, creatinine, and body mass index affected BNP levels independently of HF. The subgroup-specific BNP cutoff that maintained 91% sensitivity was 184 pg/ml for patients > or =75 years, 150 pg/ml for those with atrial fibrillation, and 449 pg/ml for patients with a creatinine > or =2 mg/dl. These subgroup-specific cutoffs improved specificity compared to a cutoff of 100 pg/ml. The regression model that adjusted BNP improved the reclassification of patients as having cardiac or noncardiac dyspnea compared to the conventional BNP cutoffs. Of the patients without HF, 11% were correctly reclassified as having noncardiac dyspnea (p = 0.003). In conclusion, adjusting BNP levels for clinical covariates improves its diagnostic performance.