We assessed the value of positron emission tomography to predict long-term outcome in patients with diabetes and ischemic left ventricular (LV) dysfunction. Circumferential profiles of nitrogen-13 ammonia (NH3) and fluorine-18 fluorodeoxyglucose uptakes were obtained in 61 patients who had diabetes and ischemic LV dysfunction. Patient profiles were compared with those from a normal database. NH3 and fluorine-18 fluorodeoxyglucose defect sizes and extent of perfusion-metabolism mismatch (percentage of myocardium with fluorine-18 fluorodeoxyglucose uptake minus NH3 uptake >2 SD above the normal difference) were determined. Patients were followed every 6 months. Over a mean follow-up of 4.3 years, cardiac death occurred in 52% of patients who underwent revascularization and 61% of those who underwent medical therapy (p = 0.69). No clinical or imaging variables predicted cardiac death in patients who underwent revascularization. In those who received medical therapy, mismatch in > or =3% of the left ventricle (risk ratio 4.0, p = 0.01) was the only multivariate predictor of cardiac death. Revascularization improved survival of patients who had mismatch of > or =3% at 4 years (p = 0.003) and at 8 years (p = 0.012) of follow-up. Patients who had mismatch > or =3% and ejection fraction <30% had the greatest improvement in survival with revascularization compared with medical therapy (p <0.0001). Revascularization also improved 4-year survival of patients who had NH3 perfusion defects of > or =25% of the left ventricle (p = 0.02). In conclusion, mismatch identifies medically treated patients who have diabetes and LV dysfunction, who are at high risk for cardiac death. Intermediate- and long-term survival of patients who have diabetes and mismatch may be improved with revascularization, and those who have significant mismatch and severe LV dysfunction have the greatest benefit.