A new exponential formula to characterize the human RR-QT relation was evaluated in comparison with Bazett's formula in 16 subjects: 10 healthy, normal men (ages 18 to 30 years) who exercised on a stationary bicycle, and 6 patients (ages 50 to 80 years; 2 women and 4 men) with rate-programmable VVI pacemakers whose rates were changed by an external programmer. The RR and QT intervals for heart rate in the range of 50 to 180 beats/min were measured from electrocardiographic tracings recorded at a paper speed of 100 mm/s. The data from each subject were fitted separately by 4 formulas by an appropriate regression analysis using a statistical package program: (F1) QT = A1 - B1*Exp(-k1*RR); (F2) QT = A2[1-Exp-(-k2*RR)]; (F3) QT = A3* square root (RR) + B3; and (F4) QT = A4* square root (RR), where all A, B, and k are regression parameters. The relative goodness of fit of data by the 4 formulas was assessed by the mean-squared residual and the Akaike Information Criterion using Wilcoxon signed-ranks tests. This analysis confirmed that F1 is the best model among the formulas tested and F4 (Bazett's formula) is the least acceptable for both exercised and paced groups. The deviations from Bazett's formula were more striking for the paced group than for the exercised group, as reflected by the mean-squared residual values for F4 (715 +/- 86 for the paced group vs 384 +/- 41 for the exercised group, p less than 0.005).(ABSTRACT TRUNCATED AT 250 WORDS)