James Brophy, MD, PhD
Can clinical assessment plus levels of N-terminal pro–B-type natriuretic peptide (NT-proBNP) accurately predict a diagnosis of acute heart failure (AHF) in emergency department (ED) patients with dyspnea?
2 cohort studies, 1 for derivation (Improved Management of Patients with Congestive Heart Failure [IMPROVE-CHF]) and 1 for validation (N-Terminal Pro-BNP Investigation of Dyspnea in the Emergency Department [PRIDE]).
7 EDs in Canada and 1 ED in Boston, USA.
483 patients in the derivation cohort (mean age 70 y, 52% men) and 573 patients in the validation cohort (mean age 63 y, 51% men) who presented to the ED with undifferentiated dyspnea. Exclusion criteria included acute myocardial infarction, renal failure, malignancy, or obvious alternate cause of dyspnea.
A mathematical combination of patient age, the ED physician’s estimated pretest probability of AHF, and serum NT-proBNP level. The formula is posttest probability = 1/(1 + ex), where e = 2.71828 and x = 8 + (0.011 × age) – (5.9 × pretest probability) – (2.3 × log NT-proBNP) + (0.82 × pretest probability × log NT-proBNP).
Diagnosis of AHF, as determined independently by 2 cardiologists (blinded to NT-proBNP level) using all available information up to 60 days.
The prevalence of adjudicated AHF was 46% in the derivation cohort and 35% in the validation cohort. When NT-proBNP was divided into 5 levels, the likelihood ratios (LRs) for AHF were 0.11 for < 300 pg/mL, 0.34 for 300 to 899 pg/mL, 1.35 for 900 to 2699 pg/mL, 3.43 for 2700 to 8099 pg/mL, and 12.8 for ≥ 8100 pg/mL. The c statistic (area under the receiver-operating characteristic curve) for the model predicting AHF was 0.91 for the derivation cohort and 0.97 for the validation cohort. In the validation cohort, clinical assessment alone accurately predicted the absence of AHF in 89% of patients classified as low risk (≤ 20%) and the presence of AHF in 95% of patients classified as high risk (≥ 80%). The model reclassified risk for 25% of patients, but reclassification was accurate for only 68% of these (Table). In the group with an intermediate pretest risk, 45% were reclassified with 98% accuracy. The net reclassification improvement was 0.23, and the integrated discrimination improvement was 0.11 (P < 0.001).
A mathematical model combining clinical assessment and levels of N-terminal pro–B-type natriuretic peptide accurately predicted a diagnosis of acute heart failure in emergency department patients with dyspnea.
A mathematical model combining clinical assessment and levels of N-terminal pro–B-type natriuretic peptide to predict acute heart failure in 573 emergency department patients with dyspnea*
*Low risk = ≤ 20%; intermediate risk = 21% to 79%; high risk = ≥ 80%.
†Based on clinical assessment alone.
‡Patients with AHF reclassified to a higher risk group or patients without AHF reclassified to a lower risk group.
Brophy J. A model combining clinical assessment and NT-proBNP level accurately predicted heart failure in ED patients with dyspnea. Ann Intern Med. ;152:JC1–13. doi: 10.7326/0003-4819-152-2-201001190-02013
Download citation file:
Published: Ann Intern Med. 2010;152(2):JC1-13.
Cardiology, Heart Failure, Pulmonary/Critical Care.
Results provided by:
Copyright © 2018 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use