Short-term heart rate variability (HRV) is used to monitor autonomic function. HRV is a generic term which describes a number of distinct measures. The reliability (test-retest repeatability) of these measures is rarely reported (Sinnreich et al. 1998; Marks & Lightfoot, 1999). Comparisons are difficult because of the variation in the number of test conditions, sampling epochs, test-retest durations and statistical analyses. A range of instruments is now available to record HRV and the purpose of this study is to assess the agreement between such systems and their individual reliability.
Subjects were 30 (20 male and 10 female) volunteers (median age 27.5, range 19-59 years). They gave informed consent. The three HRV instruments were: VariaCardio TF5 (Advanced Medical Diagnostics Group Ltd, UK), CardioTens (Meditech Ltd, UK) and a specialised HRV module within a CardioPerfect ECG analysis system (Carido Control Ltd, MN, USA.). Simultaneous, 5 min ECG recordings were made under three conditions: subject supine, subject standing, subject supine with paced (0.2 Hz) breathing. This recording protocol was repeated 1 week later at the same time of day with identical pre-test conditions. Agreement between instruments was assessed using repeated measures ANOVA and intraclass correlation coefficient 3.1 (R). Reliability was assessed using coefficient of variation (CV).
The consistent HRV measurement across each machine were mean normal-to-normal beat interval (NN), the route mean square of the standard deviation of NN intervals (RMSSD), the standard deviation of the NN intervals (SDNN), high and low frequency spectral power in raw units (HF and LF) and in normalised units (HFnu and LFnu). A high level of agreement existed in all HRV measurements under all conditions between the three instruments (median R = 0.95, range 0.79-0.99). A small but statistically significant (5.6 ms, P < 0.05) difference was found between the NN intervals recorded by the instruments. These were translated into significantly (P < 0.05) higher values for SDNN intervals calculated by the CardioTens. The reliability (CV %) was found to be poor for the TF5 (median 26.5, range 1.1-189.4 %) the CardioTens (median 29.6, range 6.2-238.5) and the CardioPerfect (median 35.5, range 7.2-90.4). Similar CVs were commonly produced by all three instruments, for given measurements, made under given conditions, although this was not consistent.
The results demonstrate that data manipulation is similar in these three instruments. The variability in R-R interval length and SDNN was found to be due to slight variation in the number of R-R intervals included in each instrument’s sampling epoch, a function of their data collection protocols. Similarities between all remaining measures demonstrate data interpretation to be so similar between instruments they may be used interchangeably. This will enable data collected between laboratories to be compared.
The results also demonstrate low levels and wide ranges of reliability for HRV measures due mainly to biological variation. This demonstrates the importance of assessing and reporting reliability of HRV parameters under specific experimental conditions. This will provide essential information on which to base any intervention.