The value of heart rate variability, ambulatory electrocardiographic (ECG) variables and the signal-averaged ECG in the prediction of arrhythmic events (sudden death or life-threatening ventricular arrhythmias) was assessed before hospital discharge in 416 consecutive survivors of acute myocardial infarction. During the follow-up period (range 1 to 1,112 days), there were 24 arrhythmic events and 47 deaths. The initial relation between several prognostic factors and arrhythmic events was explored with use of the Kaplan-Meier product limit estimates of survival function. Impaired heart rate variability less than 20 ms (p less than 0.0000), late potentials (p less than 0.0000), ventricular ectopic beat frequency (p less than 0.0000), repetitive ventricular forms (p less than 0.0000), left ventricular ejection fraction less than 40% (p less than 0.02) and Killip class (p less than 0.02) were identified as significant univariate predictors of arrhythmic events. When these variables were analyzed by using a stepwise Cox regression model, only impaired heart rate variability, followed by late potentials and repetitive ventricular forms remained independent predictors of arrhythmic events. The combination of impaired heart rate variability and late potentials had a sensitivity of 58%, a positive predictive accuracy of 33% and a relative risk of 18.5 for arrhythmic events and was superior to other combinations including those incorporating left ventricular function, exercise ECG, ventricular ectopic beat frequency and repetitive ventricular forms. These results suggest that a simple method of assessment based on heart rate variability and the signal-averaged ECG can select a small subgroup of survivors of myocardial infarction at high risk of future life-threatening arrhythmias and sudden death.