A prospective assessment of several clinical variables, left ventricular function indexes, Holter recording characteristics and signal-averaged electrocardiogram (ECG) for their value in predicting the inducibility of sustained ventricular tachyarrhythmias was carried out in a consecutive series of 105 patients with nonsustained ventricular tachycardia (VT). The patients were divided into 3 groups based on the results of programmed electrical stimulation: group 1, 22 patients with induced sustained monomorphic VT; group 2, 14 patients with induced ventricular fibrillation (VF) and group 3, 69 patients with no induced sustained VT/VF. Left ventricular ejection fraction less than 0.40, history of syncope/presyncope and abnormal signal-averaged ECG were significantly more common in group 1 than in group 3. No significant difference was found between groups 2 and 3. The sensitivity, specificity and predictive accuracy of the signal-averaged ECG for the induction of sustained monomorphic VT were 64, 89 and 84%, respectively. Using stepwise discriminant function analysis, the signal-averaged ECG was found to be the single most accurate screening test to predict the inducibility of sustained VT in patients with nonsustained VT and its value was independent of the etiology of heart disease and the length of spontaneous runs. Because of the very high specificity and negative predictive accuracy, patients with normal signal-averaged ECGs may not require invasive evaluation.