OBJECTIVE Previous studies have shown the usefulness of dobutamine echocardiography to differentiate dilated cardiomyopathy (DC) from ischemic left ventricular dysfunction (ILVD), but no studies have been made using exercise echocardiography (EE). We hypothesized that most patients with DC have some contractile reserve and experience an increase in left ventricular ejection fraction (LVEF) during exercise, as opposed to patients with ILVD. Differences in response to EE may be useful to clinically differentiate between these two entities. METHODS Between 1 March 1995 and 1 March 2001, we performed 4,133 EE studies on 3,830 patients. Of 289 patients (8%) with moderate or severe LV dysfunction (biplane LVEF < 41% and left ventricular end-diastolic diameter > 5.2 cm), 207 were excluded: 111 for a history of myocardial infarction; 28 for scarring on echocardiography (regional akinesia/dyskinesia with thinning and/or increased brightness); 13 for previous revascularization procedures; 9 for aortic valve disease; 11 for a known cause of cardiomyopathy; and 35 for not undergoing angiography. The study group was therefore composed of 82 patients who were encouraged to perform maximal treadmill EE. EE criteria for ILVD were either impaired regional wall motion (RWM) or a decrease/no change in LVEF from baseline to peak exercise, while criteria for DC were RWM improvement/no change and LVEF increase. The ILVD group was formed by 39 patients with stenosis >/= 70% diameter stenosis of a major epicardial coronary artery or major branch vessel. The remaining 43 patients constituted the DC group. RESULTS The number of coronary risk factors (ILVD 2.0 1.1; DC 1.9 1.1), baseline LVEF (ILVD 30 7; DC 30 8), and exercise-induced angina (ILVD 23%; DC 14%) did not differ between groups (p = NS). ILVD patients achieved less Mets (6.6 3.1 vs 8.3 2.8; p < 0.05), had a lower heart rate x systolic blood pressure product (22 5 vs 27 7; p < 0.001), and developed regional and/or global LV dysfunction more frequently (79 vs 28%; p < 0.001). Sensitivity, specificity, positive and negative predictive values and global accuracy for ILVD detection were 79% (95% CI: 70-88), 72% (95% CI: 63-81), 72% (95% CI: 63-81), 79% (95% CI: 67-85), and 76% (95% CI: 69-83), respectively. CONCLUSIONS Global and/or regional LV function impairment with exercise is accurate in identifying patients with ILVD. This method could reduce the need for invasive procedures.