OBJECTIVE: To test in a prospective study the hypothesis that increased QT dispersion in resting 12-lead ECG is a predictor of sudden cardiac death. DESIGN: A nested case-control study during a mean (SD) follow up time of 6.5 (2.8) years. SETTING: A prospective, placebo controlled, coronary prevention trial with gemfibrozil among dyslipidaemic middle aged men in primary (occupational) health care units: the Helsinki heart study. PATIENTS: 24 victims of fatal myocardial infarction, 48 victims of sudden cardiac death without acute myocardial infarction, and their matched controls. MAIN OUTCOME MEASURES: QT dispersion in baseline and pre-event electrocardiograms. RESULTS: At study baseline, QT dispersion was similar in all victims and controls. When estimated from the pre-event ECG on average 14 months before death, the risk of sudden cardiac death in the highest QTPEAK (up to the peak of the T wave) dispersion tertile (> or = 50 ms) was 6.2-fold (95% confidence interval 1.7 to 23.5) compared with the risk in the lowest tertile (< or = 30 ms), and 4.9-fold (1.2 to 19.5) after adjustment for the presence of left ventricular hypertrophy, while QTPEAK dispersion could not predict fatal myocardial infarction. QTEND dispersion (up to the end of the T wave) in pre-event ECGs could not discriminate victims of either sudden cardiac death or fatal myocardial infarction from their matched controls. CONCLUSIONS: In middle aged men with a normal conventional QT interval in 12-lead resting ECG, increased QTPEAK dispersion is an independent risk factor for sudden cardiac death, but not for fatal myocardial infarction.