Reliability of bedside evaluation in determining left ventricular function: correlation with left ventricular ejection fraction determined by radionuclide ventriculography

Ninety-nine patients with chronic coronary artery disease were prospectively evaluated to determine the reliability of historical, physical, electrocardiographic and radiologic data in predicting left ventricular ejection fraction. The left ventricular ejection fraction measured by radionuclide angiography was normal (greater than or equal to 50%) in 44 patients (group 1) and abnormal (less than 50%) in 55 patients; 36 of those 55 patients had an ejection fraction between 30 and 49% (group 2) and the remaining 19 patients had an ejection fraction of less than 30% (group 3). The ejection fraction was correctly predicted in 33 of the 44 patients (75%) in group 1 and in 47 of the 55 patients (85%) with abnormal ejection fraction (groups 2 and 3), but the degree of ventricular dysfunction was correctly predicted in only 19 patients (53%) in group 2 and in only 9 patients (47%) in group 3. Stepwise linear regression analysis was performed. The single most predictive variable was cardiomegaly as seen on chest roentgenography (R2 = 0.52). Four optimal predictive variables--cardiomegaly, myocardial infarction as seen on electrocardiography, dyspnea and rales--could explain only 61% of the observed variables in left ventricular ejection fraction. Thus, radionuclide ventriculography adds significantly to the discriminant power of the clinical, radiographic and electrocardiographic characterization of ventricular function in patients with chronic coronary heart disease.