To evaluate the variability in the reported diagnostic accuracy of the exercise electrocardiogram, we applied meta-analysis to 147 consecutively published reports comparing exerciseinduced ST depression with coronary angiography. These reports involved 24,074 patients who underwent both tests. Population characteristics and technical and methodologic factors, including publication year, number of electrocardiographic leads, exercise protocol, use of hyperventilation, definition of an abnormal ST response, exclusion of certain subgroups, and blinding of test interpretation were analyzed. Wide variability in sensitivity and specificity was found (mean sensitivity, 68%; range, 23-100%; SD, 16%; and mean specificity, 77%; range, 17-100%; SD, 17%). The four study characteristics found to be significantly and independently related to sensitivity were the treatment of equivocal test results, comparison with a "better" test such as thallium scintigraphy, exclusion of patients on digitalis, and publication year. The four variables found to be significantly and independently related to specificity were the treatment of upsloping ST depressions, the exclusion of subjects with prior infarction or left bundle branch block, and the use of preexercise hyperventilation. Stepwise linear regression explained less than 35% of the variance in sensitivities and specificities reported in the 147 publications. There is wide variability in the reported accuracy of the exercise electrocardiogram. This variability is not explained by information reported in the medical literature. (Circulation 1989;80:87-98) T he diagnosis of coronary artery disease in patients with chest pain syndromes is a major application of the exercise electrocardiogram (ECG) sion to determine the expected value and variability in this test's accuracy and determine which technical and methodologic factors independently affect the reported accuracy and its variability.
Methods
Literature ReviewThe Bibliography Retrieval Service and Medlars were used to search the National Library of Medicine data base for reports published after 1967 on the diagnostic accuracy of the exercise ECG when compared with coronary angiography. Though both exercise testing and angiography were used before 1967, few reports of their use in the same patients had appeared before that date. The search terms used were "exercise electrocardiogram" and "coronary artery disease." The bibliographies of three major textbooks on the subject were also searched.1-3 The bibliographies of review articles published between 1984 and 1987 and retrieved from the computer search were also scanned to complete the meta-analytic data base.The 325 publications resulting from these searches were screened for studies involving groups of 50 patients or more undergoing exercise electrocardiography with ST segment measurement and coronary angiography; for the latest report when more than one report from the same institution and same time period (within 3 years) was found, unless it could be verified that duplicatio...
