It is unclear whether sudden or nonsudden death can be predicted independently from other risk factors for coronary artery disease (CAD). Therefore, this investigation was undertaken to measure 12-lead QRS voltage sum, a recently proposed (Am J Cardiol 1985;55:485-494) index of left ventricular (LV) hypertrophy, and its ability to predict either subsequent sudden (less than 2 hours) or nonsudden CAD death during 20 to 23 years of follow-up in 1,588 middle aged men (40 to 61 years old) from 2 cohorts of the Italian section of the Seven Countries Study who were free of demonstrable CAD (at entry examination in 1962). The Sokolow-Lyon and the modified Sokolow-Lyon indexes, 2 standard electrocardiographic methods to detect LV hypertrophy were also measured and compared. During follow-up, 67 patients died suddenly and 87 died a nonsudden CAD death. In the Cox proportional-hazards model, age, mean blood pressure, heart rate, body mass index, cholesterol, physical activity, smoking habit, ST-T alterations (Minnesota codes 4.1 to 4.3 together with 5.1 to 5.3) and the 3 electrocardiographic indexes, all measured at the time of enrollment into the study, were included. The 12-lead QRS voltage sum retained significant and independent relation to sudden death (t = 2.00); Sokolow-Lyon index entered the Cox solution for nonsudden CAD death but the association was inverse (t = -2.10). ST-T alterations were significantly associated only with nonsudden CAD death (t = 2.19). Thus, in addition to several known risk factors, measurement of 12-lead QRS voltage sum in middle-aged men without clinical evidence of heart disease may help identify subjects at an increased risk of sudden death; nonsudden CAD death is predicted by Sokolow-Lyon index and by ST-T alterations. The usefulness of these indexes needs to be tested in different populations.