Abstract

ObjectiveA statistical modelling study investigated whether incorporating the curvatures of QT/RR patterns into the individual-specific QT heart rate correction increases QTc data accuracy. MethodsRepeated ECG readings were available from 4 different drug-free recordings made in 176+176 healthy female and male subjects (aged 32 ± 10 and 33 ± 8 years, respectively). In each subject, up to 1440 ECG readings were made of QT intervals and of the corresponding QT/RR hysteresis corrected RR intervals. The QT/RR patterns of each study participant was fitted with 12 different regression formulae that corresponded to differently curved physiologically plausible QT/RR profiles. In each subject, each of the regression fits was converted into a QT heart rate correction formula and the optimum model that fitted the data of the subject best was identified. Correction formulae were applied to modelled QT/RR data with RR intervals between 400 ms and 1600 ms. Differences in QTc intervals calculated by the correction formulae corresponding to the individually optimum QT/RR regression models and by the same type of regression in all study subjects were statistically summarised in females and males. ResultsCompared to the individually curvature optimised QTc heart rate correction formulae, formulae of the different regression models overestimated or underestimated the QTc values when applied on all study subjects. At RR of 500 ms, the model assuming linear QT/RR relationship led to errors of −5.01 ± 6.63 ms and of −4.80 ± 7.23 ms in females and males, respectively. At the same RR interval level, the model assuming the linear relationship between the logarithms of QT and RR intervals led to errors of +11.51 ± 6.36 ms and of +15.09 ± 7.61 ms in females and males, respectively. ConclusionThe differences in the curvatures of QT/RR patterns should be considered in the optimisation of subject-specific heart rate corrections. Forcing an arbitrary simple regression model on the QT/RR patterns of different subjects may lead to appreciable errors in QTc estimates. The frequently used linear and log-linear regression models were among the least precise if used without checking their appropriateness in individual subjects.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call