[The spectral analysis of heart rate variability. The methodological aspects]

Abstract

Objectives: A study is made of the influence of methodology on the analysis of cardiac cycle variability in terms of frequency and time domains.

Methods: Twenty-five individuals were divided into three groups: 1) non-smokers (n = 10); 2) smokers (n = 8); and 3) non-smokers without guided respiratory frequency (n = 7). An analysis was made of RR variability during 5 min intervals in time domain (standard deviation, variation coefficient, and difference between maximum and minimum RR), and frequency domain (spectral analysis, Fast Fourier Transform algorithm using 5 types of data windows).

Results: On comparing the results obtained in the 25 individuals with the 5 data windows, significant differences were observed (ANOVA; p < 0.001) in maximum and total amplitudes of the spectrum (in absolute terms) in the low frequency band (0.04-0.15 Hz). No significant differences were found between the normalized values or between the low/high (0.15-0.40 Hz) frequency ratio: W1 = 1.63 +/- 0.30; W2 = 1.62 +/- 0.29; W3 = 1.65 +/- 0.31; W4 = 1.52 +/- 0.26, and W5 = 1.55 +/- 0.27 (X +/- SE). On averaging the RR intervals each 5 cycles, significant differences were encountered for great part of the parameters studied. No significant differences were encountered for great part of the parameters studied. No significant differences were noted between the non-smokers as regards the use or non-use of a metronome (low/high frequency ratio in group 1 = 1.85 +/- 0.39 vs group 3 = 1.78 +/- 0.43; NS). The linear regressions between the variability parameters in terms of time and frequency domains (absolute values), and age (group 1) were significant, with greater regression coefficients on using the linear transformation of an exponential model. On comparing groups 1 and 2, a non statistically significant trend towards smaller maximum and total amplitudes (absolute values) was noted in group 2 for both frequency bands.

Conclusions: 1) The type of data window (Fourier analysis) introduces significant modifications of some parameters expressed in absolute values, but not between normalized values; 2) signal averaging greatly modifies the information obtained; 3) the decrease in variability parameters observed with age fits better to an exponential model; 4) no differences were detected on guiding or not respiratory frequency, and 5) smoking appears to induce a tendency towards reducing cardiac cycle variability.