We assessed whether signal-averaged electrocardiography could identify patients with sustained ventricular arrhythmias in 41 patients with non-ischemic cardiomyopathy. Twelve of these patients presented with sustained ventricular arrhythmia and 29 patients had no history of sustained ventricular arrhythmias. The mean ejection fractions in the groups were 30 +/- 9% and 24 +/- 9%, respectively. Results were compared with signal-averaged electrocardiograms in 55 normal individuals. The filtered QRS duration was longest in patients with sustained ventricular arrhythmias (130.2 +/- 19.5 vs 105.0 +/- 13.1 msec in the group without sustained ventricular arrhythmia, p less than .001 and 95.9 +/- 9.1 in the normal group, p less than .001). The voltage in the last 40 msec of the filtered QRS was lower in the sustained ventricular arrhythmia group (11.3 +/- 9.3 microV) than the group without sustained ventricular arrhythmia (53.5 +/- 28.3 microV; p less than .001) or the normal group (53.7 +/- 25.2 microV; p less than .001). Eighty-three percent of patients in the sustained ventricular arrhythmia group had an abnormal signal-averaged electrocardiogram characterized by both a long filtered QRS duration and a late potential of low voltage level; only 2% of normal subjects and 14% of patients without sustained ventricular arrhythmias had an abnormal signal-averaged electrocardiogram. The signal-averaged electrocardiogram can identify patients with nonischemic congestive cardiomyopathy and sustained ventricular arrhythmias.