Data of clinical examination, exercise testing and exercise radionuclide angiography in 102 patients referred for assessment of chest pain was included in a logistic regression to optimize the diagnosis of coronary artery disease with coronary arteriography as the reference investigation. None of the patients had other cardiac problems or previous myocardial infarction. In the absence of symptoms, exercise testing was continued until at least 80% of the theoretical maximal heart rate was attained. Each patient was characterized by the value of the logistic function or probability of coronary artery disease. A threshold value corresponding to 80% sensitivity of ROC graphs was determined. The significant variables were: a clinical variable – the type of chest pain as assessed by the clinical history; two radionuclide angiographic variables – the ejection fraction at peak effort and the corrected variation of ejection fraction between rest and stress, that is not taking into account possible decreases at the last increment of exercise. Coronary patients can be identified with an 80% sensitivity and 77% specificity on these criteria. This specificity is greater than that obtained by clinical examination and exercise testing alone (65%). Exercise radionuclide angiography may therefore reduce the number of unnecessary coronary arteriographies.