Background - An abnormal signal-averaged ECG (SAECG) is a noninvasive marker of the substrate of sustained ventricular tachycardia after myocardial infarction. We assessed its prognostic ability in patients with asymptomatic unsustained ventricular tachycardia, coronary artery disease, and left ventricular dysfunction. Methods and Results - A blinded core laboratory analyzed SAECG tracings from 1925 patients in a multicenter trial. Cox proportional hazards modeling was used to examine individual and joint relations between SAECG variables and arrhythmic death or cardiac arrest (primary end point), cardiac death, and total mortality. We also assessed the prognostic utility of SAECG at different levels of ejection fraction (EF). A filtered QRS duration >114 ms (abnormal SAECG) independently predicted the primary end point and cardiac death, independent of clinical variables, cardioverter-defibrillator implantation, and antiarrhythmic drug therapy. With an abnormal SAECG, the 5-year rates of the primary end point (28% versus 17%, P=0.0001), cardiac death (37% versus 25%, P=0.0001), and total mortality (43% versus 35%, P=0.0001) were significantly higher. The combination of EF <30% and abnormal SAECG identified a particularly high-risk subset that constituted 21% of the total population. Thirty-six percent and 44% of patients with this combination succumbed to arrhythmic and cardiac death, respectively. Conclusions - SAECG is a powerful predictor of poor outcomes in this population. The noninvasive combination of an abnormal SAECG and reduced EF may have utility in selecting high-risk patients for intervention.