JJ 12 BNP for Diagnosis of Acute CHF

ED physicians are only about 80% accurate in their diagnosis of acute CHF. Why? There is no single element of past medical history, presenting symptoms or physical exam findings that can reliably rule in or rule out acute CHF in the ED. Orthopnea, PND and weight gain are not especially helpful in making the diagnosis and even the lauded S3 gallop which most of us cannot identify on the best of ED shifts is not that helpful. The classic signs of CHF are often absent on CXR and interobserver agreement (whether you are an ED doc or a radiologist) on the diagnosis of CHF by CXR is enormous. Nonetheless, when all of these elements are put together, ED physician clinical gestalt is not bad at diagnosing CHF. But we could do better.

Enter BNP. BNP is currently in use in many EDs across North America and Europe. In this Journal Jam podcast we discuss the clinical utility of BNP and pro-NT-BNP in the work-up of the dyspneic ED patient. We ask the questions: does BNP add much beyond physician gestalt? Which patients might BNP be useful for? Should we abandon BNP as a dichotomous rule-in/rule-out variable and instead use it as a continuous variable? Does using BNP effect patient oriented outcomes? Is lung POCUS a better test? Are prediction models that include BNP useful? and many more….

Podcast: Play in new window | Download (Duration: 49:43 — 45.5MB)

Subscribe: Apple Podcasts

Podcast production by Anton Helman, Justin Morgenstern & Rory Spiegel. Editing and sound design by Anton Helman, March 2018.

Blog post by Anton Helman

Cite this podcast as: Helman, A, Morgenstern, J, Spiegel, R. BNP for Diagnosis of Acute CHF. Emergency Medicine Cases. March, 2018. https://emergencymedicinecases.com/epinephrine-cardiac-arrest/. Accessed [date].

BNP is a vasoactive hormone that is released by strained myocardium from a variety of causes. Since its discovery about 30 years ago, BNP has been utilized as a biomarker for the diagnosis and prognosis of acute heart failure both in the ED and the inpatient setting. Many factors potentially alter BNP:  age, kidney disease, obesity, hypertension, coronary artery disease, atrial fibrillation, and chronic respiratory disease. Many of the patients in which we have diagnostic uncertainty about CHF are older, have renal disease, are obese, hypertensive with respiratory or other cardiac disease. Nonetheless, there have been several observational studies and RCTs examining the role of BNP in the ED as a diagnostic aid and there remains controversy as to its utility. This Journal Jam podcast reviews these studies and suggests a clinical bottom line.

EBM topic highlighted in this Journal Jam: Spectrum Bias

Guest Peer Review of GASP4Ar Study by Chris Carpenter

Chris Carpenter MD, MSc, FACEP, FAAEM, AGSF,Emergency Physician, Barnes-Jewish Hospital
Associate Professor, Emergency Medicine, Washington University School of Medicine
Director, Evidence Based Medicine, Washington University School of Medicine

Distinguishing CHF from other causes of acute dyspnea in the decision-dense ED is sometimes challenging.  This diagnostic randomized controlled trial evaluates whether a decision aid using age, clinician gestalt, and NT-proBNP can improve diagnostic accuracy or alter measurable processes of care.

These investigators previously validated a 3-item (age, clinician pre-test probability, NT-proBNP) logarithmic equation designed to generate the probability that an acute dyspnea patient’s symptoms resulted from CHF. In this 4-site study, they now randomize clinicians to be informed of the instrument-based CHF probability or not after initial history/physical exam + CXR + EKG.  The primary outcome assessed was diagnostic accuracy and secondary outcomes included operational flow metrics (ED length of stay, ICU admission) and 60 day mortality.  Notably, these patients did not include the “slam dunk” subset of patients in whom clinicians are certain do not have CHF (clear pneumonia or COPD exacerbation) or definitely have CHF (peripheral edema, pulmonary edema, orthopnea) – eligible subjects had a pre-test probability b/w 20%-80% representing the indeterminate diagnostic zone with which clinicians struggle in making treatment and disposition decisions!

These patients are typical dyspnea patients evaluated in the ED.  Half had a history of CHF and half a history of COPD. With the receiver operating characteristic area under the curve of 0.93, the instrument displays excellent ability to distinguish those with CHF from those without.  The million dollar question, though, is does the instrument augment clinician’s gestalt?  Answer: There was no difference in clinicians’ diagnostic accuracy between those who were informed (77%) and those not informed (74%).  In addition, no clinically significant differences were observed in the ED length of stay or ICU admission rates.  The authors hypothesize that if the model thresholds of <20% or >80% probabilities had been used to define the presence of absence of acute heart failure that 48% of patients would have been redirected (alternative disposition or management decisions) with 95% accuracy, but that is not what occurred because clinicians rely upon myriad factors when determining management and disposition decisions for undifferentiated dyspnea patients including likelihood of alternative diagnoses, patient and family angst, patient goals of care, access to timely follow-up, referring consultants’ priorities and expectations, and malpractice concerns in the zero miss mentality of some healthcare systems.

Distinguishing CHF from other causes of acute dyspnea (COPD, VTE, deconditioning) remains a challenge for some patients.  This model is not a panacea to overcome diagnostic hurdles, but may provide a reasonable risk stratification instrument to facilitate shared decision making with patients and referring providers.  More diagnostic accuracy studies in different populations are still needed to assure that the performance of the model holds up across different populations.  An additional uncertainty is how hospitals not using NT-proBNP could use this model, or how to adjust the model for assay differences between different NT-proBNP machines as the authors did in this study?  If additional accuracy studies in different settings confirm the accuracy of this model and skeptics are subsequently convinced that the model supplements clinical gestalt, implementation science will be necessary to convert knowledge into bedside application as opposed to simply educating clinicians that the model exists.

 asdfasdf

Episode 4 – Acute Heart Failure main podcast and show notes