Recently, it was demonstrated that the QT-RR relationship pattern varies significantly among healthy individuals. We compared the intra- and interindividual variations of the QT-RR relationship. Twenty-four-hour 12-lead digital electrocardiograms (ECGs; SEER MC, GE Marquette; 10-s ECG recorded every 30 s) were obtained at baseline and after 24 h, 1 wk, and 1 mo in 75 healthy subjects (42 women, 33 men, age 27.9 +/- 9.6 vs. 26.8 +/- 7.5 yr, P = not significant). QT interval was measured automatically in each ECG by six different algorithms, and the mean of the six measurements was analyzed. In each recording of each individual, QT-RR relationship was assessed by 10 different regression models including linear (QT = beta + alpha x RR) and parabolic (QT = beta x RR(alpha)) models. Standard deviations (SDs) of regression parameters alpha and beta of consecutive recordings of each individual were compared with SD of the individual means. Intrasubject stability and interindividual variability were further tested by ANOVA. With all models, intraindividual SDs of the regression parameters were highly significantly smaller than SD of individual means (P < 10(-5)-10(-9)). The intrasubject stability was further confirmed by ANOVA (P < 10(-19)-10(-30)). The QT-RR relationship exhibits substantial intersubject variability as well as a high intrasubject stability. This has practical implications for a precise estimation of the heart rate-corrected QT interval in which optimized subject-specific rate correction formulas should be used.