Nonlinearities in heart rate variability

Jørgen K. Kanters

Laboratory of Experimental Cardiology, Dept. of Medical Physiology,
University of Copenhagen, Blegdamsvej 3C, Copenhagen, Denmark

Abstract: The heartbeat is not regular but varies in an apparent irregular matter. This variation termed heart rate variability (HRV) is not only due to changes in activity or respiration, but occurs even at rest or at night. Atropine abolishes most of these variations, indicating the significance of parasympathetic tone. The importance of HRV was demonstrated by Kleiger[1]. He showed that decreased HRV measured in the time domain as the standard deviation of 24-h RR-intervals in msec (SDNN), was a strong predictor of mortality after myocardial infarction. However the clinical utility of SDNN is limited, and only permits the identification of a high risk group, but lacks the sensitivity and the specificity to identify individual subjects at high risk for sudden death.
The drawback of time and frequency domain measures is that they only take linear information into account ignoring any form of nonlinear dynamics. Previous studies have shown that a small but significant amount of nonlinear dynamics exists in HRV[2]. Various methods exist to quantify nonlinearities in HRV. Strange attractors have been postulated by plotting a RR interval against the following one, revealing different patterns in healthy subjects and patients suffering sudden cardiac death. These patterns can be quantified by determining their dimension, ideally giving the minimum number of independent variables needed to describe HRV. Specific structures in HRV can be identified using nonlinear predictability or neural networks showing that some specific dynamics exist in HRV. Newer methods introducing nonlinear autoregressive modelling seems to be able to identify these characteristic patterns. The strength of these methods is the ability to discriminate between inherent dynamics and external perturbations (noise).
There has been much effort to apply these methods in the clinical setting. Preliminary studies has shown that dimension measures can distinguish patients that later developed ventricular fibrillation from those who did not, despite that the groups had the same SDNN. Unfortunately, we still need larger studies to establish the role of nonlinear dynamics in the clinical setting for predicting sudden cardiac death.

[1] R.E. Kleiger et al., Am J Cardiol, 59: 256-262, 1987.
[2] J.K. Kanters et al. J Cardiovasc Electrophysiol, 5:591-601, 1994