4 Aug, 15 | by EBM
By: Dr. Geoffrey Modest
A potentially very useful multicenter study was just published finding that lung ultrasound (LUS) was better than chest xray (CXR) in the diagnosis of acute decompensated heart failure (ADHF) in patients presenting to the emergency room (ER) with dyspnea (see CHEST 2015; 148(1): 202 – 210). This is an increasingly common issue, with evidently very different therapeutic approaches depending on the etiology.
–1005 patients from 7 Italian ERs presented with acute dyspnea, and the ER MD was asked to categorize the dyspnea diagnosis as either from ADHF or noncardiogenic dyspnea, based on their initial clinical assessment (this was not prescribed but left up to the ER MD, to simulate a real-world situation) and after LUS. All patients also had CXR’s. After patient discharge, independent reviewers determined the cause of the dyspnea.
–544 (51.4%) patients were in community hospitals, 461 (45.9%) academic medical centers. Median age 77, 46.2% women, 41.8% smokers, 40.7% with COPD, 63.5% hypertensive, 20.2% with congestive heart failure (CHF), 25.8% with ischemic cardiomyopathy/CAD, 27.9% diabetic. Most on lots of meds. So, pretty sick group overall
–final diagnosis of ADHF was in 463 patients (46%), with very high agreement between the 2 physician adjudicators of the diagnosis. 81.7% of the patients were admitted to the hospital.
–clinical workup and LUS alone were much more accurate than CXR (p<0.01):
–I included below 2 of their summary figures. The one to the left is the summary sensitivity, specificity, etc in the whole group (top) and in those who had BNP/NT-pro-BNP sent. The figure on the right is the ROC curves, showing that clinical evaluation plus LUS were remarkably accurate: they achieved sensitivity, specificity and AUC values for diagnosis of ADHF of 97.0%, 97.4%, and 0.972!!! CXR had sensitivity as low as 69.5% (which is higher than in some studies…..) — so, absence of CXR findings does not even come close to excluding a diagnosis of ADHF!!
–the net reclassification improvement (NRI), an estimate of the % of subjects moving from one diagnosis to another, of the LUS approach vs standard clinical workup was 19.1% (p<0.01)
–the accuracy of BNP/NT-pro-BNP values (available on subgroup of 486 patients) was significantly lower than the LUS approach (p<0.01): the sensitivity/specificity was 85%/61.7% with AUC of 0.733, again being not nearly as sensitive as LUS.
–there was near perfect intraobserver agreement in reading LUS (0.97) among experts, and a really good agreement (0.92) for ER MDs with limited LUS training (<10 examinations performed)
One of the strengths of this study is that it incorporated the clinical pretest probability (the “LUS-implemented” approach), which was much more accurate than LUS alone, which was better than CXR (though, to be fair, they did not look at a “CXR-implemented approach”, but the stand-alone CXR was still really inferior to the stand-alone LUS). In this study LUS was done as a rapid, bedside test, with a positive test for ADHF defined as “bilateral presence of 2 or more zones showing the presence of at least 3 B-lines: vertical, hyperechoic reverberation artifacts extending from the pleural line to the bottom of the screen”, these artifacts are presumably from interstitial fluid in the lung. This was a stricter definition of a “positive” LUS than in prior studies, which could explain the greater accuracy here — prior studies have found more limited LUS accuracy (as low as 54%) in ADHF diagnosis. One potential bias in this study is that the same MD who made the initial clinical diagnosis also performed the LUS and was therefore not blinded to the results of the clinical workup in interpreting the LUS.
So, bottom line: this study comes on the tail of one earlier this year showing better accuracy of LUS than CXR for pneumonia (see here). Together, these studies reinforce the potential to use a noninvasive and nonradiation-associated evaluation for both CHF and pneumonia. Although this was an ER-based study, it is likely to be applicable to patients who come into primary care with decompensated heart failure, though it would be great to have a community-based study done — the majority of patients with ADHF I see at the health center are not sent to the ER or admitted. I treat these patients based on clinical assessment and CXR, with very close followup, and they do fine clinically. But I imagine that the availability of LUS would allow for a more assured diagnosis.