The aim of this study was to determine whether the diagnostic capability of radionuclide angiography (RNA) in detecting coronary artery disease (CAD) might be improved by using several indices of left ventricular (LV) function in concert. Three different models (rest data, exercise data, and rest plus exercise data) were derived by stepwise multivariate discriminant analysis of RNA data in 65 normal volunteers and 111 patients with CAD and normal ejection fraction (EF) at rest. The model with only resting indices yielded a diagnostic capability comparable to the simple measure of EF response to exercise (area under receiver operating characteristic curve = 89% and 91%, respectively). Both the exercise and rest plus exercise models gave better results (area = 94% and 97%, respectively), but only the rest plus the exercise model was better than the EF response alone (p less than 0.001). Thus (a) if resting studies alone are performed, the diagnostic potential of RNA may be improved by combining several indices of resting function; and (b) combined rest and exercise data may improve the sensitivity of RNA in detecting CAD over what could be obtained with the EF response to exercise alone.