OBJECTIVE Our aim was to study the relationship between T-wave alternans (TWA) and rate-response (restitution) of repolarization in subjects with and without ventricular systolic dysfunction. BACKGROUND T-wave alternans is a promising predictor of sudden death, yet the mechanisms linking it with human ventricular arrhythmias are unclear. From theoretic considerations, we hypothesized that abnormal TWA is linked with steep restitution of action potential duration (APD) and that both predict arrhythmic outcome. METHODS We studied 53 subjects with left ventricular ejection fraction (LVEF) < or =40% and 18 control subjects. At electrophysiologic study, we recorded APD at 90% repolarization (APD(90)) in the right (n = 62) or left (n = 9) ventricle during pacing while measuring TWA from the body surface. RESULTS As expected, TWA (at <109 beats/min) was more likely to be abnormal in study than in control subjects (p < 0.01). However, study (LVEF 28 +/- 8%) and control (LVEF 58 +/- 12%) subjects did not differ in APD(90) restitution slope maxima (1.2 +/- 0.6 vs. 1.3 +/- 0.6, respectively; p = 0.82) or numbers with steep slope (>1; 58% vs. 67%). T-wave alternans and simultaneous APD alternans always occurred at diastolic intervals where APD restitution was not steep (p < 0.001), and there was no relationship between maximum restitution slope and TWA magnitude. Over 829 +/- 473 days, TWA (p = 0.02), but not restitution slope >1, predicted ventricular arrhythmias in subjects with LVEF < or =40%. CONCLUSIONS The mechanism by which TWA predicts arrhythmic mortality does not reflect the maximum slope of ventricular APD restitution. Better understanding of the mechanisms underlying TWA may enable improved prediction and prevention of ventricular arrhythmias.