The utility of the signal-averaged electrocardiogram (SAECG) for predicting ventricular tachycardia (VT) induction in patients presenting with sustained VT or ventricular fibrillation (VF) while on an empirically chosen antiarrhythmic agent was assessed in 17 patients. At the time of presentation with a malignant arrhythmia, 12 patients were taking quinidine, three patients were taking procainamide, and two patients were taking flecainide. All patients underwent programmed ventricular stimulation when not taking antiarrhythmic drugs; 12 patients had no inducible sustained VT and five patients had inducible sustained monomorphic VT. The SAECG done in the control state without antiarrhythmic agents was negative for late potentials in 11 of 12 patients in the noninducible group and positive for late potentials in four of five patients in the inducible group (sensitivity = 80% and specificity = 92%). We conclude that in patients presenting with life-threatening ventricular arrhythmias while taking an antiarrhythmic drug, the SAECG distinguishes patients with possible proarrhythmic events from those who have the substrate for inducible sustained VT.