OBJECTIVE To determine the incremental value of exercise radionuclide angiography for identification of severe coronary artery disease. METHODS Retrospective analysis comparing logistic regression models. METHODS A tertiary care referral center. METHODS Three hundred and ninety-one consecutive patients who had normal resting electrocardiograms (ECGs) and no digoxin therapy within the previous week. RESULTS The exercise ECG model, consisting of magnitude of ST depression, exercise heart rate, and patient gender, was highly predictive of three vessel or left main coronary artery disease (chi 2 = 100, P less than 0.0001). The model correctly classified 60% of the study group which included 56 patients with and 179 without severe disease. The addition of radionuclide angiographic variables improved the predictive power of the model (chi 2 = 124, P less than 0.0001). However, the exercise radionuclide angiographic variables increased the number of patients who were correctly classified by only 11 and the percentage by 3% (to a total of 63% of the study group). CONCLUSIONS The modest additional advantage provided by exercise radionuclide angiography for identification of three vessel or left main coronary artery disease in patients with normal resting ECGs would not appear to justify its routine use for this purpose. Before this conclusion is used as a guide for clinical practice, our results should be prospectively confirmed in a separate sample of patients in another institution.