Assessment of the accuracy of common clinical thresholds for cardiac morphology and function by transthoracic echocardiography

Abstract

We evaluated the ability of transthoracic echocardiography (TTE) to correctly identify abnormal left ventricular (LV) size, function, and mass when compared to cardiac magnetic resonance (CMR). Whilst numerous studies have compared TTE and CMR with respect to correlation between measurements and study reproducibility, few have employed categorical analysis relevant to clinical practice. Two hundred and fifteen consecutive patients who underwent both TTE and CMR were evaluated for the presence of abnormal LV size, systolic function, and mass. Abnormal LV systolic function was further categorized into grades (mild, moderate, and severe). Quantification of LV morphology and function was performed on TTE and CMR according to published guidelines. The level of agreement between TTE and CMR was compared across binary and categorical variables using Cohen's kappa. Compared to CMR, TTE demonstrated excellent agreement in identification of abnormal versus normal function (κ=0.87). However, agreement across grades of LV function was less strong (κ=0.63). Whilst agreement for identification of severe LV dysfunction was good (κ=0.68), this would still lead to misclassification of severe dysfunction in approximately one in seven cases. Agreement between TTE and CMR was moderate to good for identification of LV dilation (κ=0.43-0.63), but poor for identification of increased mass (κ=0.04). Whilst in clinical practice TTE performs well in identification of normal versus abnormal systolic function, it has substantial limitations across grades of dysfunction and in the assessment of LV size and mass. These limitations have important implications when considering management decisions for patients based on thresholds of LV morphology or function.