Doubts have been expressed about the clinical usefulness of time domain analysis of the signal averaged electrocardiogram in patients with prolonged QRS complex duration. We studied 147 patients using a signal averaged ECG (40-250 Hz) whose QRS complex was longer than 100 ms. A baseline electrophysiology study was also performed in 128 of these patients. Seventy-seven patients had a minor (QRS < 120 and > 100 ms) conduction defect. Thirty-seven of these 77 had either induced or spontaneous sustained ventricular tachycardia (group I) and 40 had no sustained ventricular tachycardia (group II). Seventy patients had a major (QRS > or = 120 ms) conduction defect, 44 of whom had sustained ventricular tachycardia (group A). The remaining 26 without this condition formed Group B. Group I compared to group II patients had a longer filtered QRS duration (120.8 +/- 14 vs 104.5 +/- 9.5 ms, P < 0.001), a longer low amplitude signal duration (41 +/- 12.1 vs 31 +/- 12.6 ms, P < 0.0001) and a lower root mean square of the last 40 ms of the filtered QRS complex (27 +/- 29.8 vs 35 +/- 25.3 microV, P = ns). Group A compared to group B had a longer filtered QRS duration (157.7 +/- 20.2 vs 140.7 +/- 15.7 ms, P < 0.001), a longer low amplitude signal duration (57.3 +/- 24.9 vs 37.8 +/- 20.3 ms P < 0.001) and a lower root mean square of the last 40 ms of the filtered QRS complex (14.3 +/- 11.2 vs 22.0 +/- 10.5 microV, P < 0.01). Using conventional late potential criteria, the sensitivity and specificity of the signal averaged ECG for the detection of sustained ventricular tachycardia patients with a minor conduction defect were 89% and 75%, respectively. The same criteria applied to patients with a major conduction defect were sensitive (sensitivity: 87%) but non-specific (specificity: 50%). However, by using modified late potential criteria, such as the presence of two of any of the following three signal averaged parameters: filtered QRS duration > or = 145 ms, low amplitude signal duration > or = 50 ms, root mean square of the last 40 ms of the filtered QRS complex < or = 17.5 microV, we derived a non-optimal but still acceptable combination of sensitivity (68%) and specificity (73%). We conclude that traditional late potential criteria can be applied in patients with a minor conduction defect, but modification of these criteria is necessary to derive useful clinical information for risk stratification of patients with a QRS complex duration > or = 120 ms.