Abstract
BACKGROUND: To assess the accuracy of findings from the clinical history, symptoms, signs and diagnostic tests (ECG, CXR and natriuretic peptides) in relation to the diagnosis of left ventricular systolic dysfunction (LVSD) in a primary care setting. METHODS: Diagnostic accuracy systematic review, we searched Medline (1966 to March 2008), EMBASE (1988 to March 2008), Central, Cochrane and ZETOC using a diagnostic accuracy search filter. We included cross-sectional or cohort studies that assess the diagnostic utility of clinical history, symptoms, signs and diagnostic tests, against a reference standard of echocardiography. We calculated pooled positive and negative likelihood ratios and assessed heterogeneity using the I2 index. RESULTS: 24 studies incorporating 10,710 patients were included. The median prevalence of LVSD was 29.9% (inter-quartile range 14% to 37%). No item from the clinical history or symptoms provided sufficient diagnostic information to "rule in" or "rule out" LVSD. Displaced apex beat shows a convincing diagnostic effect with a pooled positive likelihood ratio of 16.0 (8.2-30.9) but this finding occurs infrequently in patients. ECG was the most widely studied diagnostic test, the negative likelihood ratio ranging from 0.06 to 0.6. Natriuretic peptide results were strongly heterogeneous, with negative likelihood ratios ranging from 0.02 to 0.80. CONCLUSION: Findings from the clinical history and examination are insufficient to "rule in" or "rule out" a diagnosis of LVSD in primary care settings. BNP and ECG measurement appear to have similar diagnostic utility and are most useful in "ruling out" LVSD with a normal test result when the probability of LVSD is in the intermediate range.