Session P86.6

Dynamic Properties of QT Intervals

J Halamek*, P Jurak, V Vondra, J Lipoldova,
P Leinveber, M Plachy

Institute of Scientific Instruments
Brno, Czech Repub.

Aim: Dynamic properties of QT intervals are critical for arrhythmia diagnosis and predicting untoward effects of drugs. The analysis of QT properties is conditioned by the definition of model of QT/RR coupling.
Methods: We tested 4 models: i) Transfer function model (TRF) with 3 parameters; ii) The dependency on weighted sum of previous RR intervals with constant weighting (CW); iii) Model with exponential weighting (EXP); iv) The EXP model supplemented with direct coupling between RR and QT (DCEXP). The tilt test (supine for 10 min, then tilted for 10 min and supine again for 5 min) of 18 healthy subjects was used and the error signal, i.e. the difference between measured QT and model QT was analyzed.
Results: RMS of error signals (mean±STD over subjects) were 3.95±1.5; 4.58±1.4; 4.31±1.4 and 3.94±1.5 ms at TRF; CW; EXP and DCEXP models respectively. The correlation coefficients between error signals and RR intervals were 0.009±0.009; 0.24±0.1; 0.25±0.1 and 0±0 at TRF, CW, EXP and DCEXP models respectively.
Discussion: 1) CW and EXP models have significantly higher RMS of error signal and correlation coefficients between error signal and RR intervals, i.e. some important QT parameter is missing. The shape of step response does not correspond with measured step response by MR. Franz, (1988). 2) DCEXP may be a choice of QT/RR models. It has a sufficient number of parameters and RMS and correlation coefficients are minimal. But DCEXP is based on the assumption of known step response of QT intervals (similar to MR. Franz) and on the shape of weighting coefficients. 4) The TRF model does not assume any shape of step response and the shapes given by analysis are similar to measured by MR. Franz. The control system analysis is the background of this model and the number of parameters was given by optimization of the order of transfer function. 5) Linear, dynamic coupling between QT/RR was assumed in the analysis and the minimization of RMS and correlation coefficient was achieved in TRF and DCEXP models. The analysis of possible nonlinearity must be done on significantly larger amount of measurements. 6) The QT analysis should be based on measurements with significant RR excitation, otherwise the signal-to-noise ratio of analyzed data is low and resulting QT parameters are inaccurate.
Conclusion: The complete set of QT dynamic and static parameters is very important for patient diagnosis and drugs influence test. Such set may be established by the proper model of QT/RR coupling only. The TRF model is the best present choice according to our opinion.

(Abstract Control Number: 5)