This study compares the accuracy of the routine exercise electrocardiogram, exercise score, and exercise echocardiography for the diagnosis of coronary artery disease (CAD) in women. Seventy women with a pretest probability of 53 +/- 30% for CAD were stressed using a maximal symptom-limited bicycle exercise protocol. Significant ST-segment change was defined by a depression of > 0.1 mV 0.06 second after the J point. The exercise score was calculated from ST response, heart rate, and workload using an equation derived from a multivariate model. A positive stress echocardiogram was defined by development of a new or worsening wall motion abnormality. The results were compared with the presence or absence of significant (> 50% diameter) stenoses at angiography. Exercise echocardiography identified 29 of the 33 patients (88%) with CAD, compared with 22 (67%) using ST analysis alone, and 20 (61%) using the exercise score (both p < 0.05 vs exercise echocardiography). The specificity of exercise echocardiography (84%) and the multivariate score (73%) were comparable, and exceeded that of the ST analysis (51%) in 37 patients without CAD (p < 0.01). The accuracy of exercise echocardiography (86%) exceeded that of the exercise score (67%, p = 0.01) and ST analysis (59%, p < 0.01). Among all 70 patients, an intermediate (20% to 80%) probability of coronary disease was identified in 21 patients on the basis of exercise echocardiography, in 38 based on the multivariate score, and in 38 based on the ST analysis alone. Exercise echocardiography is more sensitive than the exercise score, and more sensitive and specific than ST-segment analysis for the diagnosis of CAD in women.(ABSTRACT TRUNCATED AT 250 WORDS)