Spectral analysis of heart rate variability: interchangeability between autoregressive analysis and fast Fourier transform

Interchangeability between fast Fourier transform (FFT) and autoregressive (AR) analysis was assessed on series of 256 R-R intervals recorded in 56 seated subjects and in 15 men performing an orthostatic test. Low- (LF) and high-frequency (HF) components were calculated and expressed both in absolute (square milliseconds) and normalized units (NU). During orthostatic stress, the same upward trend for LF square milliseconds and LF/HF ratio and downward trend for HF square milliseconds and HF NU were observed with FFT and AR analysis. However, the values for HF square milliseconds were significantly greater with FFT, as compared with AR analysis in standing position (P < .05). Moreover, Bland & Altman method highlighted a large discrepancy between the results of FFT and AR analysis for all heart rate variability indices in the 3 conditions. Therefore, parametric and nonparametric spectral analyses could not be considered as interchangeable at rest in healthy subjects even if they give same qualitative results.