Comparison of Six Commonly Used QT Correction Models and Their Parameter Estimation Methods

Abstract

This paper compares six commonly used QT correction models and three available parameter estimation methods using five indices for QTc evaluation based on real and simulated electrocardiograph (ECG) datasets. The results show that the golden section approach always finds the correction factor making QTc interval uncorrelated to heart rate for all six formulas. However, the correction formulas derived from mixed model sometimes fail to make QTc interval invariant of heart rate. The performance of an individual least-square regression method lies between the golden section iteration approach and the mixed model in terms of QTc–RR relationship.

Key Words:

• Clinical trial
• Golden section
• QT correction model
• QTc
• Thorough QT study

Notes

Note. M1 = Model 1: Linear model: QTc = QT + α(1 − RR)

M2 = Model 2: Hyperbolic model: QTc = QT − α(1/RR − 1)

M3 = Model 3: Parabolic model: QTc = QT/RR ^α

M4 = Model 4: Logarithmic model: QTc = QT − αln(RR)

M5 = Model 5: Shifted logarithmic model: QTc = ln(e ^QT  + α(1 − RR))

M6 = Model 6: Exponential model: QTc = QT − α(e ^−RR  − 1/e)

NA, not applicable; RMSE, root mean squared error. Absolute values of ρ and slope are used for calculating their mean values.

Note. M1 = Model 1: Linear model: QTc = QT + α(1 − RR)

M2 = Model 2: Hyperbolic model: QTc = QT − α(1/RR − 1)

M3 = Model 3: Parabolic model: QTc = QT/RR ^α

M4 = Model 4: Logarithmic model: QTc = QT − αln(RR)

M5 = Model 5: Shifted logarithmic model: QTc = ln(e ^QT  + α(1 − RR))

M6 = Model 6: Exponential model: QTc = QT − α(e ^−RR  − 1/e)

NA, not applicable; RMSE, root mean squared error. Absolute values of ρ and slope are used for calculating their mean values.

Note. M1 = Model 1: Linear model: QTc = QT + α(1 − RR)

M2 = Model 2: Hyperbolic model: QTc = QT − α(1/RR − 1)

M3 = Model 3: Parabolic model: QTc = QT/RR ^α

M4 = Model 4: Logarithmic model: QTc = QT − αln(RR)

M5 = Model 5: Shifted logarithmic model: QTc = ln(e ^QT  + α(1 − RR))

M6 = Model 6: Exponential model: QTc = QT − α(e ^−RR  − 1/e)

NA, not applicable; RMSE, root mean squared error. Absolute values of ρ and slope are used for calculating their mean values.