The ability of exercise radionuclide angiography to predict the risk of having significant left main or three vessel coronary artery disease was examined in 681 patients who underwent both radionuclide and coronary angiography. There were significant differences in multiple variables between patients with or without such disease. Logistic regression analysis identified seven variables as independently predictive of the presence of left main or three vessel disease. Using these variables, low, intermediate and high probability groups could be identified. The four most important variables--the magnitude of exercise ST segment depression, peak exercise ejection fraction, peak exercise rate-pressure product and sex of the patient--can provide practical estimates of the risk of having left main or three vessel disease. Exercise radionuclide angiography can provide a clinically useful noninvasive estimate of the risk of having significant left main or three vessel disease.