Objectives: To investigate the reliability of heart rate variability (HRV) measures at rest and during light exercise in children.
Methods: Short term (five minute) HRV was assessed in 12 children (11–12 years of age). HRV measures were collected at rest with the children supine, breathing at 12 breaths/min, and during exercise on a cycle ergometer while exercising at 25% of peak oxygen uptake. Both resting and exercise data were collected twice from each child.
Results: Intraclass correlation coefficients were low to moderate for most measures with wide confidence intervals for each variable in both resting and exercise conditions. Random variation (typical error) within repeated measurements ranged from 31% to 187%.
Conclusions: These preliminary findings suggest that HRV measures are unreliable at rest and during light exercise in children aged 11–12 years. Tighter control of extraneous influences is recommended.
Statistics from Altmetric.com
- HRV, heart rate variability
- RMSSD, square root of the mean of the sum of the squares of the differences between adjacent R-R intervals
- pNN50, the proportion of pairs of adjacent intervals differing by more than 50 ms
- SDNN, standard deviation of all the R-R intervals
- LF, low frequency power
- HF, high frequency power
- TP, total power
- ICC, intraclass correlation coefficient
Variations in heart rate period arise through the balance of sympathetic and parasympathetic nervous modulation of heart rate. Through differences in the physiological response of the myocardium to the neurotransmitters acetylcholine and noradrenaline (norepinephrine), a greater variation in heart rate period arises when the parasympathetic nervous system predominates than with predominant sympathetic modulation.1,2 Analysis of this natural variation in heart rate period (heart rate variability (HRV)) is considered a valuable tool in providing a non-invasive window to sympathovagal regulation of heart rate.3
Time and frequency domain statistical procedures are used to analyse HRV data,4,5 with the following measures being calculated from short term—less than five minute—heart rate recordings: in the time domain, square root of the mean of the sum of the squares of the differences between adjacent R-R intervals (RMSSD), the proportion of pairs of adjacent intervals differing by more than 50 ms (pNN50), and standard deviation of all the R-R intervals (SDNN); in the frequency domain, low frequency power (LF), high frequency power (HF), and total power (TP).
In adults, the HRV response to exercise has been previously documented.2 In addition, the relation of HRV to peak oxygen uptake (peak V̇o2) and physical activity levels and the effects of exercise training on HRV are widely understood,3,6,7 yet their use with children has largely focused on the neonate,4 and data on the reliability of HRV are restricted to adult subjects.
In healthy adult subjects, there exists a wide interindividual variation in HRV measures at rest. Between subject variance is reported to be in the range 12–15% for mean R-R interval, 41–155% for LF, and 70–163% for HF, although the cause(s) of such a wide variation is unclear.8,9
Although group mean reliability data may indicate no significant difference between HRV measures, on an individual basis significant day to day variation exists for each person: mean (SD) variation of 23.5 (14.6)% for pNN50 and 10.7 (6.8)% for SDNN.12 There appear to be no published data on the reliability of short term resting HRV measures in healthy children.
HRV measures during exercise have been investigated previously in adults. Data are sparse, but indicate that exercise HRV measurements are reliable. Tulppo et al13 reported the limits of agreement between repeated measures14 during cycling at intensities of 50 and 75 W. The range of differences for HF expressed as a coefficient of variation was 4.4%. The reliability of HRV measures in children during exercise is not known and must be established to evaluate the use of this technique with young people.
The purpose of this study was therefore to assess the reliability of HRV measures in children during rest and light exercise.
Twelve children (seven girls, five boys) volunteered for the study. Written informed consent was obtained from the participants and their legal guardians. The institutional ethics committee granted approval for the study. All children were healthy, and none were taking any prescription medications. Participants were fully habituated to equipment, protocols, and experimenters. Stature was measured using a Holtain stadiometer (Holtain, Crymych, Dyfed, UK), and body mass using Avery beam balance scales (Avery, Birmingham, UK). The children were asked to abstain from consuming caffeinated beverages and excessive physical activity in the 24 hours preceding data collection. HRV measures were collected at the same time of day.
Resting HRV data were collected on two separate days using a Polar Vantage telemetry system (Polar, Kemple, Finland), with the children in a supine position in a designated quiet room for 10 minutes. Breathing was paced at 12 breaths/min by means of a computer generated image that increased and decreased in size in rhythm with the required breathing rate, supplemented by an audible “beep” signal.
Determination of peak V̇o2 was performed on a different day from the collection of the resting HRV data, using an electronically braked cycle ergometer (Lode Excalibur Sport, Groningen, the Netherlands). During exercise, gas exchange variables were measured and displayed on line using a calibrated EX670 mass spectrometer (Morgan Medical Ltd, Kent, UK). Peak V̇o2 was determined using a ramp test, using increments of 10 W/min, to voluntary exhaustion. Participants pedalled at a cadence of 70 (5) rpm, being actively encouraged throughout. Peak V̇o2 was taken as the highest recorded 10 second stationary average value during the maximal exercise test.
On separate, subsequent visits, participants completed two constant work rate exercise tests. All participants completed a six minute bout of light exercise of unloaded cycling, which represented an intensity of about 25% peak V̇o2. A pedal cadence of 70 (5) rpm was maintained throughout the test. Each test was completed on separate days, but at the same time of day.
HRV data were firstly cleaned using the Polar Precision Performance software set at moderate filtering level. The cleaned data were then analysed with HRV Analysis Software (Biomedical Signal Analysis Group, University of Kuopio, Finland). A continuous five minute HR recording period was selected from the resting HR data and the exercise steady state HR data. The following time domain statistics were generated: RMSSD, NN50, and pNN50. Frequency domain analysis used an autoregressive statistical method. Optimal autoregressive model order was calculated using the forward-backward linear least squares method,15 which indicated a fixed model order of 16. To remove the influence of a large low frequency baseline trend component, detrending of the data was performed using a smoothness priors based method.16 Frequency domain statistics are presented in both ms2 and normalised units (NU).4
Paired t tests and intraclass correlation coefficients (ICCs) were calculated for all HRV variables. Typical error was also calculated.17 Statistical significance was set at p⩽0.05.
Data were collected from 12 children, mean (SD) age 12.8 (0.3) years, mean (SD) body mass 46.6 (10.1) kg, and mean (SD) stature 1.56 (0.06) m. Group mean peak oxygen uptake was 41 (6) ml/kg/min. Constant load exercise was performed at an intensity of 25 (4)% peak V̇o2.
At rest and during light exercise there were no significant differences between the group mean values between trial 1 and 2 for any time domain or frequency domain HRV variable (table 1). None of the ICCs for resting measures were significant (p⩾0.05; table 2), however, during light exercise ICCs for RMSSD, NN50, pNN50, and HF were all significant (p⩽0.05). Calculation of typical error indicated a random variation in the results of 31–187% (table 3).
HRV analysis is a novel research tool in exercise and sports medicine,4 yet before this work can be extended to children it is imperative to establish the reliability of such measures within that population. Determination of reliability is a composite measure that provides information on the stability, accuracy, and size of measurement error of a dependent variable. By minimising measurement error and controlling for influencing factors, the stability and accuracy of the dependent variable can be better determined. The benefits of ensuring reliability are that the resultant data can be interpreted in a meaningful way, the measurement tool/procedure can be replicated with confidence by others, and the experimental conditions used have suitably attenuated the effects of any confounding variables.18
Although there were no significant differences between the group means between trial 1 and 2 for any HRV variable, the standard deviations of these mean scores were large. The ICCs were lower than those reported for adults.5,10,11 In addition, most of the ICCs were non-significant or had associated wide 95% confidence intervals (table 2). Hopkins17 argues that calculating typical error gives a more informative reliability statistic than simply calculating differences and ICCs. Typical error represents how much of the change in scores is due to random variation—this random variation may arise through both day to day variation (biological variation) and through measurement error. The typical error statistic for every HRV variable indicated the presence of a sizeable random variation (31–187%) (table 3), and it is this randomness of the resultant HRV data that may help to explain the large standard deviations, low ICCs, and wide 95% confidence intervals observed with these data.
Take home message
These preliminary data suggest that the reliability of heart rate variability measures at rest and during light exercise in children is weak. Further research is required to determine whether reliability can be improved with tighter control of extraneous variables.
Although attempting to control for the potentially confounding effects of drugs, previous activity, and caffeinated beverages on HRV in these children, differences in preparations before measurement may have occurred, leading to the significant day to day variations seen. For example, it is well documented that children are poor at recalling previous physical activity, and their perception of exercise intensity is more inconsistent than that of adults.19 Consequently, it is difficult to guarantee that all the children avoided excessive exercise on the days before data collection. In addition, although breathing rate was held constant during resting HRV measures, tidal volume, which also influences HRV in children,20 was not. These factors, combined with the small sample size, may have colluded to produce the wide variation seen.
These data suggest that the reliability of HRV measures in children is unconvincing, therefore researchers and clinicians should be cautious in using and interpreting them. Whether such reliability could be improved with tighter control of potentially confounding variables warrants further investigation.
The results were analysed with HRV Analysis Software 1.1 for windows developed by The Biomedical Signal Analysis Group, Department of Applied Physics, University of Kuopio, Finland.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.