Linearly scaled, rate-invariant normal limits for QT interval: eight decades of incorrect application of power functions.

INTRODUCTION: Normal limits for QT traditionally are derived as mean +/- 2*SD, with rate adjustment done by dividing QT values by power functions such as RR 1/2 (proportional scaling). METHODS AND
RESULTS: We evaluated procedures for deriving normal limits by comparing adjusted QT distributions versus heart rate using ECG data of 11,739 normal men and women aged > or = 40 years. QT decreased as predicted by many power functions with heart rate but its SD remained relatively unchanged. Consequently, proportional scaling induced rate-dependent distortion of normal limits. Furthermore, QT distributions by heart rate were variably skewed and non-normal. Therefore, normal limits expressed as mean +/- 2*SD were misleading. Omission of regression intercept was an additional reason for failure of Bazett's and Fridericia's formulas. Regularized normal limits for adjusted QT (Qta) were obtained with linear instead of proportional scaling of type QTa = QT + k1*(1 - RR(k2)), for instance, with k2 = 0.5, k1 = 0.360 for males and 0.353 for females, or with k2 = 0.42, k1 = 0.414 for males and 0.420 for females. With linear scaling, QTa = 460 msec was established as the upper 2% normal limit, with 32 msec as the limit for a significant QTa increase from reference ECG in serial comparison.
CONCLUSION: Traditional procedures for establishment of normal limits failed because of proportional scaling, assumption of normal QT distribution, or omission of regression intercept. Percentile distributions of linearly scaled adjusted QT produced regularized rate invariant normal limits within normal sinus rates.