In order to begin searching for new markers for safe exercise training in patients with cardiac diseases, we tested the sensitivity and reliability of the short-term variability of repolarization (STV(QT)) in comparison with QT interval, QTc, and T(peak)-T(end) interval (T(p-e)) in patients with cardiac diseases. Nine patients (8 men, 1 woman; 58 ± 10 years) were enrolled. The cardiac rehabilitation (CR) program consisted of walking, bicycling on an ergometer, and calisthenics for 30-50 minutes/session and 3-5 sessions/week for 3 months. ECGs of 31 consecutive sinus beats were obtained before and after the CR program. RR and QT intervals were measured in the aVL lead. The mean orthogonal distance from the diagonal to the points of the Poincaré plots was determined using the following equation; STV(QT) [= Σ |QT(n+1)-QT(n)/(30 × 2(1/2))], as a marker of temporal dispersion of repolarization. Also, T(p-e) of 5 consecutive beats was measured as a marker of spatial dispersion. No fatal arrhythmias were observed in the CR. No significant difference was observed in the RR or QT interval between at baseline and at the end of the CR program. Meanwhile, QTc, STV(QT) and T(p-e) decreased significantly from 429 ± 27 to 400 ± 17 (P < 0.01), from 6.8 ± 1.3 to 4.7 ± 1.4 msec (P < 0.001), and from 74.8 (61.2/79.1) to 64.8 (51.4/70.7) msec (median (25th/75th percentile), P < 0.01), respectively. STV(QT) together with T(p-e) and QTc may reflect the time-courses of safe exercise training.