The accuracy of sequential testing in the noninvasive diagnosis of coronary artery disease has been established in the symptomatic clinical populations, while little is known about its value when applied to low prevalence groups, such as totally asymptomatic men. To evaluate the accuracy of noninvasive sequential testing in the diagnosis of silent myocardial ischemia, data were collected from exercise electrocardiogram, <sup>201</sup>Tl perfusion scintigraphy and radionuclide angiography for 62 totally asymptomatic middle-aged men who underwent coronary arteriography because they were positive for two or more markers of myocardial ischemia as determined by a diagnostic screening of a nonbiased population consisting of 4,842 presumably healthy men aged 40-59 years (the ECCIS Project). The predictive value of serial testing procedures for significant coronary artery obstruction was 35%. Predictive values of an abnormal electrocardiogram associated with either an abnormal <sup>201</sup>Tl scintigram, an abnormal isotopic ventriculography, or both were 33, 38 and 31%, respectively. In asymptomatic middle-aged men, there is at least a 50% likelihood that an abnormal radionuclide test is a false-positive result, and the positive predictive value is not enhanced by the concordance of an abnormal <sup>201</sup>Tl scintigraphy with an abnormal isotopic ventriculography. Thus, the application of noninvasive sequential testing in screening for asymptomatic coronary artery disease is limited by its low predictive value in accordance with the Bayesian probability theory.