BACKGROUND:It has been postulated that increased inter-lead differences in QT interval (QT dispersion) and the maximum QT interval (QTmax), in the standard 12-lead electrocardiogram (ECG), may be associated with an increased risk of cardiac death. The aims of this study were to assess the relationship between QT dispersion and QTmax, corrected and uncorrected for heart rate, and the risk of cardiac death. DESIGN:Nested case-control study within the Caerphilly prospective cohort study. METHODS:We studied 2512 men who participated in phase 1 of the Caerphilly study between 1979 and 1983. After a mean follow up of 7.1 years, 218 men had died from coronary heart disease and these men were compared with 218 age-matched controls. RESULTS:Data are presented on 422 patients with ECG suitable for analysis, 207 cases and 215 controls. Four trained observers measured the QT intervals and the reliability of each observer was estimated using repeat measurements on a randomly chosen sub-sample of ECGs. Median corrected QT dispersion and corrected QTmax were significantly higher in cases than in controls (51.9 versus 47.7 ms [P=0.01] and 430 versus 421 ms [P<0.001] respectively). In univariate analyses by quartiles of corrected QT dispersion and corrected QTmax, increased risk was largely confined to the upper quartile of the distribution with these subjects having twice the risk of those in the lower quartile [odds ratio (OR) 2.14, 95% confidence interval (CI) 1.2-3.7 and 2.56 (95% CI 1.5-4.5) respectively]. In logistic regression analysis, adjusted for age, smoking, body mass index, hypertension, history of myocardial infarction and ECG Minnesota code, we observed an increased risk in the upper quartile of the corrected QT dispersion relative to the other three quartiles combined [adjusted OR=1.74 (P=0.03)]. The magnitude of this association was increased in analyses based on the data from the most reliable observers. The association between corrected QTmax and cardiac death was attenuated in multivariate analysis. The findings in relation to both uncorrected QT dispersion and uncorrected QTmax were similar, i.e., consistent with a significant independent effect of QT dispersion but not QTmax for cardiac death in multivariate analysis. CONCLUSION:The data suggest that QT dispersion is an independent predictor of cardiac death provided it can be measured with sufficient reliability. The association is non-linear with increased risk largely confined to the upper quartile of the distribution. The QT maximum is not an independent predictor of cardiac death.