OBJECTIVE--To study the effects of variability in the duration of the QT interval corrected for heart rate (QTc) on the occurrence of sudden death. DESIGN--Nested case-referent study. SUBJECTS--Cohort of 6693 consecutive patients who underwent 24 hour electrocardiography and were followed up for two years. Risk implications of QTc interval variables were studied in patients without evidence of cardiac dysfunction or of an intraventricular conduction defect (104 died suddenly and 201 patients were randomly drawn from the study cohort). MAIN OUTCOME MEASURES--Mean QTc interval duration and variation in QTc duration over time correlated with occurrence of sudden death. RESULTS--Patients with a prolonged mean QTc over 24 hours (> or = 440 ms) had a 2.3 times (95% confidence interval 1.3 to 4.5) higher risk of dying suddenly than patients with a normal mean QTc (400-440 ms); patients with a shortened mean QTc (< 400 ms) also had a higher risk (relative risk 2.4 (1.4 to 4.3)). Patients with low (< 20 ms) and high (> or = 25 ms) long term variation in QTc duration had an increased risk of dying suddenly compared with those with intermediate variation (20-25 ms) (relative risks 2.2 (1.2 to 4.2) and 2.3 (1.4 to 4.2) respectively). The relative risks for low and high short term variation were not significantly raised. CONCLUSIONS--A prolonged and a shortened mean QTc interval over 24 hours was associated with a more than twofold risk of sudden death compared with intermediate mean QTc values (400-440 ms). Neither short nor long term variability in QTc had a distinct relation with the risk of sudden death.