Multivariate analysis of 11 clinical variables was performed in 104 patients with sustained, symptomatic ventricular tachycardia (VT) or ventricular fibrillation treated with amiodarone to determine variables predictive of subsequent cardiac arrest or sudden death. Twenty-five patients (24%) had fatal or nonfatal cardiac arrest after 7.3 +/- 6.2 months (mean +/- standard deviation) of therapy. Multivariate analysis identified an ejection fraction of less than 0.40, syncope or cardiac arrest before amiodarone therapy, and VT (3 or more consecutive ventricular premature complexes) during predischarge ambulatory electrocardiographic monitoring as variables associated with a high risk of subsequent fatal or nonfatal cardiac arrest (p less than 0.03). Patients who had these 3 clinical variables had a much higher predicted incidence of cardiac arrest at 6 months (62%) and 12 months (76%) than did patients with an ejection fraction greater than 0.40, without syncope or cardiac arrest before amiodarone therapy, and without VT during predischarge ambulatory electrocardiographic monitoring (2% and 5%, respectively) (p less than 0.02). Risk stratification using clinical variables can predict which patients are at high risk of recurrent cardiac arrest or sudden death during amiodarone therapy.