Heart rate variability in passive tilt test: comparative evaluation of autoregressive and FFT spectral analyses

The dynamic response of the autonomic nervous system during tilting is assessed by changes in the low (LF) and high frequency (HF) components of the RR series power spectral density (PSD). Although results of many studies are consistent, some doubts related to different methodologies remain. Specifically, the respective relevance of autoregressive (AR) and fast Fourier transform (FFT) methods is often questioned. Beat-to-beat RR series were recorded during 90 degrees passive tilt in 18 healthy subjects (29 +/- 5 years, eight females). FFT-based (50% overlap, Hanning window) and AR-based (Levinson-Durbin algorithm) PSDs were calculated on the same RR intervals. Powers in very low frequency (VLF: < 0.04 Hz), LF (0.04-0.15 Hz), and HF (0.15-0.40 Hz) bands were calculated either by spectrum integration (FFT and ARIN), by considering the highest AR component in each band (ARHP), or by summation of all AR components (ARAP). LF and HF raw powers (ms2) were normalized by total power (%P) and by total power after removal of the VLF component (nu). AR and FFT total powers were not different, regardless of body position. In supine condition, when compared to ARHP and ARAP, FFT underestimated VLF and overestimated LF, whereas in tilt position FFT overestimated HF and underestimated LF. However, supine/tilt trends were consistent in all methods showing a clear reduction of HF and a less marked increase of LF. Both normalization procedures provided a significant LF increase and further magnified the HF decrease. Results obtained with ARIN were remarkably close to those obtained with FFT. In conclusion, significant differences between AR and FFT spectral analyses do exist, particularly in supine position. Nevertheless, dynamic trends provided by the two approaches are consistent. Normalization is necessary to evidence the LF increase during tilt.