We assessed the usefulness of ventricular late potentials (VLP) for detecting ventricular tachycardia (VT), and of the relationship between VLP and left ventricular dysfunction. The subjects consisted of 21 patients with dilated cardiomyopathy (DCM) not accompanied with ventricular conduction disturbances. Signal-averaged electrocardiograms were recorded by using VCM-3000. The signals of 200 or more beats were averaged through the 40-300 Hz band-pass filter until the noise level was reduced to less than one microvolt (microV). The duration (fQRSd) and the root mean square voltage (LP40) for the last 40 msec of the filtered QRS complex were calculated. The left ventricular diastolic dimension (LVDd), systolic dimension (LVDs) and the ejection fraction (EF) were obtained by echocardiography. Five of 21 patients had VT. fQRSd was significantly longer in patients with VT than in those without VT (158.8 +/- 8.6 vs 126.5 +/- 10.4 msec, p < 0.01). LP40 was significantly less in patients with VT than in those without VT (8.7 +/- 2.7 vs 24.1 +/- 12.3 microV, p < 0.01). If the presence of "fQRSd > or = 120 msec" or "LP40 < or = 20 microV" was defined as VLP positive, as in patients with old myocardial infarction, the criteria sensitivity was 100% and their specificity was 25% for detecting VT. If the presence of both "fQRSd > or = 135 msec" and "LP40 < or = 20 microV" was defined as VLP positive, the criteria sensitivity was 100%, and their specificity, 88%.(ABSTRACT TRUNCATED AT 250 WORDS)