OBJECTIVE This study sought to develop a regression-based method for characterization of QT interval behavior during exercise and to define the normal range of the resulting "dynamic" measures of repolarization during submaximal treadmill testing in men and women. BACKGROUND The Bazett-corrected QT (QTc) interval during exercise has been used as a marker for ischemic disease, arrhythmogenic substrate and the long QT syndrome. However, recent studies indicate that the QTc interval is nonlinear with respect to heart rate during exercise, making the end-exercise QTc interval dependent on peak work load achieved. In contrast, the unadjusted QT interval measured from QRS onset to T wave offset (QTo) and from QRS onset to T wave peak (QTm) appears to vary linearly with heart rate during gently graded effort. METHODS The QT interval relation to heart rate and cycle length was examined by linear regression in 50 normal men (mean age 48 years) and 30 normal women (mean age 51 years), all of whom had normal rest electrocardiograms. The QTo and QTm measurements were made from digitized lead V5 complexes averaged by computer over 20-s periods, at upright control and after seven 2-min stages of the Cornell modification of the Bruce treadmill protocol (work load equivalent to Bruce stage 3). RESULTS For each subject, regression of QTo (ms) versus heart rate (beats/min) resulted in a slope (reflecting the "dynamic" change in QTo during effort), an adjusted intercept (reflecting QTo extrapolated to a heart rate of 60 beats/min) and a significant correlation coefficient (r) value. Under these conditions, mean +/- SD (5th to 95th percentile) values for men were -1.45 +/- 0.34 ms/beat per min (-0.90, "less dynamic" to -1.96, "more dynamic") for the slope; 403 +/- 21 ms (365 to 431) for the adjusted intercept; and -0.93 +/- 0.06 (-0.81 to -0.99) for r. Values for women were more dynamic, with a mean slope of -1.74 +/- 0.32 ms/beat per min (-1.23 to -2.18, p < 0.0005 vs. men) and higher adjusted intercept of 426 +/- 23 ms (392 to 462, p < 0.0001 vs. men) at similar strength of correlation (r = -0.95 +/- 0.06). Corresponding normal data were also tabulated for QTm behavior and QT-RR interval behavior during exercise. CONCLUSIONS These data provide a "dynamic" definition of normal and abnormal repolarization and describe normal limits for the linear relations of the QTo and QTm intervals with respect to heart rate and cycle length during submaximal exercise in normal men and women.