Observer variations in short period spectral analysis of heart rate variability

In spectral analysis of heart rate variability (HRV), the ratio between pow er in the low-frequency (LF) and power in the high-frequency (HF) regions i s widely used to assess the balance between cardiovagal and sympathetic act ivity (LF/HF ratio). We studied the variation between three observers in th e analysis of seven healthy volunteers and eight patients with renovascular hypertension under three conditions (rest, nitroprusside and phenylephrine infusion), in order to determine whether manual selection of the most repr esentative and steady periods increases the reliability of the method. For each condition a period of 6 min of assumed steady-state ECG activity was m arked for further analysis. LF/HF ratios were calculated for 3-min periods, using three analysis methods: 'all manual', 'fixed frequency' and 'automat ic'. In 'all manual', the observer selected the optimal 3-min period as wel l as die LF and HF regions. In 'fixed frequency', the manually selected opt imal 3-min periods were used with fixed frequency regions for LF (0.04-0.15 Hz) and HF (0.15-0.40 Hz). In 'automatic', the fixed frequency regions (0. 04-0.15 and 0.15-0.40 Hz) were used and the periods were systematically sel ected regardless of HR data. The variance in this 'automatic' analysis is c onsidered to reflect intrasubject variability, For all analysis methods, th e mean LF/HF ratio was highest in the nitroprusside condition, in which the largest interobserver as well as intrasubject variations were found. Calcu lated per method and per patient condition, the standard deviation between observers was always >25% of the corresponding mean LF/HF ratio. The smalle st interobserver variations were found for the 'fixed frequency' method. Th e data showed that it is advantageous to select the 3-min ECG periods but n ot to select the frequency regions. Selection of the latter led to an incre ase in interobserver variation. The results of this study give a realistic impression of the intrasubject and interobserver variation to be expected w hen measuring the LF/HF ratio. This variation is considerable. (C) 2000 Els evier Science B.V. All rights reserved.