Table 1.
Formal description of individual models.
Fig 1.
Comparison of data fit among discussed models.
The fit of the M0 model, which is used as a reference, is zero by definition. The vertical bars indicate 95% confidence intervals for the fit means. The symbols indicate statistical significance of the pairwise comparisons with paired t-test: * p<0.001; ** p<0.0001; # p<0.01, † p<0.05.
Fig 2.
A Data from a single patient derived from the M0 model. In addition to the QT/RR data (red), corrected QT intervals based on Bazett (black) and Fridericia (green) formulas are displayed along with the corresponding QTc/RR regression lines. Note the markedly negative slope of both QTc regression lines, indicating poor performance of both formulas, namely overcorrection for fast heart rates. The regression lines are extended to a region without data points for clarity of labeling only. B QT/RR data are compared to QTc/RRc plot from the same patient derived from the Mpopul model. The slope of the regression line derived from Mpopul (blue) is smaller and the data fit is better than for QTcB and QTcF. Similar to the Bazett, Fridericia, and Framingham formulas, QTc derived from Mpopul has only one parameter (QTc). In contrast, the regression line slope and the IR parameter are also individually determined in the related Mexp120 model.
Fig 3.
A The performance of individual correction formulas using standard deviation of 20 QTc values from each patient. The standard deviation is the highest for the Bazett formula and the lowest for the Mpopul formula. The error bars denote standard deviations. The symbol ** indicates p < 10−4. B QTc values from a 60 s recording from an individual patient demonstrating visually the difference in consistency between the Mpopul and Bazett corrections. C The distribution of differences between QTc values calculated by Bazett (blue) or Fridericia (orange) formulas from the QTc values calculated by the best model (Mpopul). The histogram is derived from 600 manually measured QT intervals (20 measurements in each patient). The QTcB and QTcF values differ from the best QTc estimate by > 30 ms in 54% and 20% of measurements, respectively. The data in this Figure indicate that the Mpopul formula performs better than the alternatives when manual QT measurement are used as the gold standard.