Increased repolarization variability has been observed in various cardiac conditions. However, data on its relation to heart rate variability and on its value in predicting adverse outcomes in high risk patients are limited. Forty-seven patients with decreased left ventricular function and ICDs had high resolution 10-minute ECG recordings and were followed for 781 +/- 258 days (mean +/- SD) on average. The interval from the R peak to the T wave peak with maximum amplitude (RTmax) and from the R peak to the T wave offset (RToff) were determined automatically on a beat-to-beat basis. Temporal beat-to-beat RTmax and RToff variability were analyzed using traditional summary statistics, a complexity measure (approximate entropy [ApEn]), and the short-term scaling exponent (alpha1). Eight (17%) patients died and 16 (34%) patients experienced death/appropriate ICD shock during follow-up. RTmax-ApEn was significantly higher in patients who died compared with patients who survived (1.24 +/- 0.13 vs 1.01 +/- 0.21, respectively, P=0.008). When RTmax-ApEn was tested together with the alpha1 of the RR intervals, occurrence of ventricular arrhythmias before ICD implantation, and beta-blockers usage in the Cox regression analysis, it still independently predicted mortality; hazard ratio=3.36 (1.28-8.83, 95% CI, P=0.014) for every 0.10-increase in RTmax-ApEn. None of the repolarization variability parameters independently predicted death/appropriate ICD shocks. Increased temporal complexity of repolarization (RTmax-ApEn) independently predicts mortality in ICD patients.