Annals of Internal Medicine | 2019
Review: In acute dyspnea, lung US has higher sensitivity than chest X-ray for detecting cardiogenic pulmonary edema
Abstract
Question In adults presenting with dyspnea, how do point-of-care lung ultrasonography (LUS) and chest radiography (CXR) compare for diagnosing cardiogenic pulmonary edema? Review scope Included studies evaluated both LUS and CXR as initial assessments in adults presenting to any clinical setting with acute dyspnea, compared LUS and CXR results with a reference standard for diagnosing cardiogenic pulmonary edema, and provided sufficient data to calculate sensitivity and specificity as outcomes. The reference standard diagnosis was based on independent expert clinical review of medical records, or echocardiography results plus brain-type natriuretic peptide (BNP) criteria. Review methods MEDLINE; EMBASE/Excerpta Medica; Cochrane Library; and gray literature including conference proceedings, ClinicalTrials.gov, and ProQuest Dissertations and Theses were searched to May 2018 for prospective cohort studies. 6 studies (n =1827, mean or median age 71 to 81 y in 5 studies, 46% to 54% women in 4 studies) met inclusion criteria: 4 were done in emergency departments and 2 in internal medicine wards after presentation to an emergency department. 2 studies blinded LUS evaluation to all clinical information, and 4 blinded LUS evaluation to clinical information other than bedside information that could not be blinded (e.g., patient appearance). In all 6 studies, CXR interpretation was not clearly blinded to clinical information and the reference standard diagnosis was not blinded to CXR results. {In 5 of 6 studies, the reference standard diagnosis was blinded to LUS evaluation; the reference standard in the other study included evaluation of echocardiography and CXR results and BNP criteria}*. Main results 20% to 62% of patients were diagnosed with cardiogenic pulmonary edema across studies. Pooled diagnostic test characteristics are in the Table. The relative sensitivity ratio for LUS compared with CXR was 1.2, 95% CI 1.08 to 1.34; the relative specificity ratio was 1.0, CI 0.90 to 1.11. Conclusion In adults presenting with dyspnea, point-of-care lung ultrasonography had higher sensitivity than, and similar specificity to, chest radiography for diagnosing cardiogenic pulmonary edema. Pooled diagnostic characteristics of lung ultrasonography and chest radiography for diagnosing cardiogenic pulmonary edema in adults with dyspnea Diagnostic tests Sensitivity (95% CI) Specificity (CI) LR+ LR Lung ultrasonography 0.88 (0.75 to 0.95) 0.90 (0.88 to 0.92) 8.6 0.14 Chest radiography 0.73 (0.70 to 0.76) 0.90 (0.75 to 0.97) 7.4 0.30 Abbreviations defined in Glossary. Results based on 6 studies (n =1827), with reference standard diagnosis based on independent expert clinical review of medical records, or echocardiography results plus brain-type natriuretic peptide criteria. Commentary Acute heart failure represents a wide spectrum of disease severity and can be difficult to diagnose in some patients (1). LUS has the potential to improve diagnostic accuracy for acute heart failure compared with CXR, with the benefits of decreased cost and rapid bedside availability. The meta-analysis by Maw and colleagues compared the accuracy of LUS with CXR and deemed LUS superior, with a positive likelihood ratio of 8.6 compared with 7.4 for CXR and a negative likelihood ratio of 0.1 compared with 0.3 for CXR. However, confidence in these pooled estimates is limited due to high risk of bias in some domains of included studies and inconsistent results. The reference standard for diagnosing acute heart failure probably incorporated CXR findings in all studies; therefore, CXR accuracy may be overestimated (2). 3 of 6 studies used convenience sampling, and some used only experienced US operators. Applicability may be limited because the accuracy of US is operator-dependent. Although no statistical measure of heterogeneity was provided, the sensitivity of LUS varied widely, from 0.58 to 0.97. Maw and colleagues suggest that different positivity thresholds could explain this apparent heterogeneity. Disease spectrum may also have contributed to between-study differences, but there was insufficient information to determine this. Limitations of the reviewed studies raise serious questions about the results. However, it is not clear that those limitations would favor LUS over CXR, and with experienced operators, LUS probably offers better overall accuracy than CXR. LUS is relatively simple to perform and interpret and can be reasonably incorporated into the evaluation of acute heart failure by experienced users.