In a group of 193 postinfarction patients with ventricular premature beats on a resting 12-lead ECG, the dispersion of QT and JT intervals was calculated as a difference between maximum and minimum QT and JT intervals in ventricular premature beats. During a follow-up of 38 +/- 17 months, death from all causes was noted in 56 patients. Univariate predictors of mortality included QT dispersion >/=100 ms, JT dispersion >/=100 ms, left ventricular ejection fraction <40%, complete bundle branch block, 'R-on-T' index of ventricular premature beats <1 and age of patients >60 years. At multivariate Cox proportional hazards survival analysis, only QT dispersion >/=100 ms, left ventricular ejection fraction <40% and complete bundle branch block had an independent relation to postinfarction mortality. The final model selected increased QT dispersion as the prognostic factor which was the most strongly related to mortality (chi(2) = 23.60, p = 0.0000).