To determine to what extent the diagnostic accuracy of stress testing is influenced by the prevalence of coronary-artery disease, we correlated the description of chest pain, the result of stress testing and the results of coronary arteriography in 1465 men and 580 women from a multicentered clinical trial. The pre-test risk (prevalence of coronary-artery disease) varied from 7 to 87 per cent, depending on sex and classification of chest pain. A positive stress test increased the pre-test risk by only 6 to 20 per cent, whereas a negative test decreased the risk by only 2 to 28 per cent. Aothough the percentage of false-positive results differed between men and women (12 +/- 1 per cent versus 53 +/- 3 per cent P less than 0.001), this difference was not seen in a subgroup matched for prevalence of coronary-artery disease. We conclude that the ability of stress testing to predict coronary-artery disease is limited in a heterogeneous population in which the prevalence of disease can be estimated through classification of chest pain and the sex of the patient.