Our purpose was to determine the test performance characteristics of BNP and NT-proBNP in the diagnosis of heart failure for patients presenting to an emergency department or urgent care center. We searched Medline, Embase, AMED, Cochrane, Cochrane Database of Systematic Reviews, and CINAHL for English-language articles published between 1989 and June 2012. Studies were limited to those using FDA-approved assays. We examined test performance at three pre-specified cutpoints (manufacturers’ suggested, researchers’ optimal, and lowest) and considered the effect of age, gender, ethnicity and renal function. We used the QUADAS—2 tool to examine risk of bias and applicability, and the AHRQ Methods Guide to assess the strength of evidence. Seventy-six articles met our inclusion criteria, 37 examined BNP, 25 examined NT-proBNP, and 14 examined both. Pooled sensitivity and specificity for BNP at the three pre-specified cutpoints were 95, 91, and 95 % (sensitivity) and 55, 80, and 67 % (specificity), respectively. For NT-proBNP, sensitivity and specificity at the same cutpoints were 91, 90, and 96 % (sensitivity) and 67, 74, and 55 % (specificity). Both BNP and NT-proBNP perform well to rule out, but less well to rule in, the diagnosis of heart failure among persons presenting to emergency departments or urgent care centers. Both BNP and NT-proBNP levels are positively associated with age and negatively associated with renal function. However, the effect of these factors with respect to selecting optimal cutpoints is unclear. For BNP, 100 pg/mL appears to be a consensus cutpoint. No clear consensus has emerged for NT-proBNP, but the age-adjusted cutpoints of 450 pg/mL for <50 years, 900 pg/mL for 50–75 years and 1,800 pg/mL for >75 years appear promising and merit greater scrutiny and validation.