Analysis of fetal heart rate variability in frequency domain: methodical considerations

Koome et al. (2014) investigated the possibility of inferring the status of fetal autonomic nervous system activity from heart rate (FHR) variability (fHRV) analysis in frequency domain. The authors provide evidence that this is not possible. However, a number of crucial methodological issues need to be resolved before this conclusion can be drawn.

First, information on the fetal ECG sampling rate is not provided, but it is crucial for a correct estimation and analysis of HRV (1996; Karin et al. 1993). The time scale of subtle fHRV events requires the temporal resolution of R peak detection in the QRS complex to be within <1 ms (Karin et al. 1993).

Second, the low-frequency to high-frequency ratio (LF/HF) of fHRV spectral power is presented and used as the measure of autonomic balance. This mathematical approach has been challenged because it does not reflect the underlying autonomic nervous system physiology (Eckberg, 1997). The authors themselves discuss this as an explanation for why the measure failed. Low-frequency spectral power reflects both sympathetic and parasympathetic influences, while HF spectral power reflects parasympathetic influences; consequently, changes in the ratio do not purely reflect the balance between sympathetic and parasympathetic activity. To assess autonomic balance, one would need to use direct measurements of sympathetic activity, such as muscle or renal sympathetic nerve activity, and relate these to changes in the HF power spectrum.

Lastly, there are a number of concerns with the authors’ approach to frequency domain analysis, as follows. First, the adult HF band range was taken (Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology, 1996) rather than the fetal one (Karin et al. 1993, 1996; Van Leeuwen et al. 2003). Thus, the conclusion that no developmental change in HF frequency band was present is premature, as it was probably low-pass filtered above the 0.4 Hz cut-off chosen by the authors. Second, the fHRV intervals of 120 s may not have captured well the intermittent respiratory activity contributing to the higher frequency spectral power of fHRV and the LF and very LF band contributions (Karin et al. 1993, 1996). It is not clear how the authors validated these stationarity and physiological aspects. Third, while Van Leeuwen et al. (2003) likewise defined HF as <0.4 Hz, they did look at higher frequencies and reported a developmental increase of spectral power in those frequencies. Fourth, were the frequency domain measures normalized in relationship to the integration band width prior to statistical analysis? This is crucial for the assessment of developmental effects when the overall variability and spectral power change (Karin et al. 1993, 1996). Fifth, David et al. (2007) have shown in human fetuses that if mean FHR drops below 130 beats min⁻¹, aliasing of the respiratory frequency contained in the FHR signal will occur; breathing faster than 50% of mean FHR will exceed the Nyquist critical value of ~1.2 Hz and result in aliasing when estimating the HF component of fHRV. To rule this out, the authors need to assess the fetal respiratory frequency spectrum and FHR range in 0.8 gestation group.

The above critique demonstrates why fHRV analysis in the frequency domain has a low appeal for animal model-based or human fHRV monitoring. The expertise required to dissect modulatory contributions of the parasympathetic branch of the autonomic nervous system to fHRV versus the sympathetic modulatory contributions restricts the physiological applicability and dissemination of this approach. For these reasons, the time domain fHRV measures should be preferred, because they have been widely validated for detection of physiological and pathophysiological phenomena and are safer to use (Garzoni et al. 2013). Adherence to a fetal equivalent of the adult HRV Task Force is needed (Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology, 1996) to promote and maintain a gold standard in validation of fHRV methodology at the fetal stage of development in animal and human studies. This is a conditio sine qua non to foster the development of objective and reliable fetal monitoring technologies relying on estimation of sympathetic and parasympathetic modulation of fHRV.