Effects of physical activity on heart rate variability in children and adolescents: a systematic review and meta-analysis

The aim of the study was to investigate the effects of physical activity (PA) on heart rate variability (HRV) in children and adolescents. We conducted a research of Web of Science, PubMed, ScienceDirect, Springer-Link and EBSCO-host. The revised Newcastle-Ottawa Scale was used in an investigative analysis to assess bias risk. A total of 21 studies were included. Overall, me-dium-sized associations were found between PA and low frequency and high frequency in children and adolescents. High PA level had significantly higher standard deviation of RR intervals and root of the mean of the sum of the squares of differences between adjacent RR intervals in children and adolescents. The effects of PA on HRV were consistent in children and adolescents. Our systematic review and meta-analysis revealed medi-um-sized between PA and HRV in children and adolescents. Promoting children’s and adolescents’ participation in moderate-to-vigorous physical activity (MVPA) will increase parasympathetic nerve activity and decreased sympathetic nerve activity. Our findings support motivating children and adolescents to engage in more MVPA in their daily lives to improve autonomic nervous system function and promote cardiovascular safety.


introduction
Physical activity (PA) levels have been declining over the last few decades. About 25% of adults not get enough PA in the worldwide 1 . Strong evidence suggests that lower PA levels increases the risk of many adverse health conditions, such as cardiovascular diseases (CVD) 2 and lowers life expectancy 3 . Though CVD usually affects adults, it can begin in childhood 4 . Thus, the prevention of CVD and promotion of PA level should begin with children and adolescents 4,5 .
Heart rate variability (HRV) is a non-invasive, repeatable predictorof the cardiac autonomic nervous system function 6 and a major source of cardiovascular health 4 . Farah et al. 7 discovered no substantial association between children's PA and HRV using self-reported PA levels. However, the recall questionnaires should be with caution when used in children or adolescents 8 . Buchheit et al. 9 used an accelerometer to critically assess PA level, and found no significant correlation between the PA and HRV. Interestingly, in a previous study with larger sample size (3,395 adolescents) 10 , PA levels were significantly correlated with HRV and there were significant differences between PA levels. Furthermore, one systematic review 11 found a positive relationship between moderate-to-vigorous physical activity (MVPA) and HRV in children and adolescents (ages 5 to 18). Therefore, the relationship between PA levels and HRV is unclear due to the high heterogeneity of the previous studies.
Examining this relationship will assist in understanding how PA affects HRV in children and adolescents. Furthermore, HRV is regarded as a reliable factor that could demonstrate that PA reduces the risk of CVD 12 . Thus, the purpose of our studywas to conduct a systematic review and meta-analysis of the effects of PA for HRV in children and adolescents.

Methods
The review was conducted following the requirements of the international meta-analysis writing guidelines (the PRISMA statement for reporting systematic reviews and meta-analyses of studies) 13 .

identification of studies
A systematic search was conducted using Web of Science, PubMed, Science Direct, Spring-er-Link, EBSCO-host databases. Search terms included "ANS", "HRV", "PA", "child", "adolescent", all combined with "AND" (Figure 1). The retrospective approach was used for expanded retrieval to optimize the total literature. The retrieval period for data collection was between 2000/01-2019/10 ( Figure 1). Only papers written in peer-reviewed journals were considered. Reviews, conference sessions, and abstracts were not taken into account. Removed duplicates, studies were initially assessed by screening titles and abstracts.

inclusion and exclusion criteria
The inclusion criteria for relevant studies in the review were: (1) in non-intervention studies, the association between PA and HRV was studied, or the differences in HRV between PA groups were compared; (2) the participants were children and/or adolescents (3-18 years). Obese and overweight children/adolescents were also includes, but studies on other conditions (congenital heart disease, hypertension, metabolic syndrome, etc.) were excluded; (3) to evaluate the ANS function, the linear and non-linear for HRV analysed 6 . Linear indicators include time and frequency domain indicators. The time domain indicators are limited to standard deviation of RR intervals (SDNN) and square root of the mean of the sum of the squares of differences between adjacent RR intervals (RMSSD). The frequency domain indicators are limited to low frequency (LF), high frequency (HF) and LF/HF. Thenon-linear indicators are limited to the detrended fluctuation analysis (DFA1) and Poincaré plot (SD1). PA: objective assessment and subjective appraisal of PA studies using accelerometers, pedometers, global positioning system equipment, questionnaire surveys, and interviews. Total physical activity (TPA), vigorous physical activity (VPA), MVPA, light physical activity (LPA), and sedentary time (ST) for PA levels.

selection of studies and data extraction
Titles, abstracts and full texts were screened by CH, and XJ, with disagreements discussed between these two authors. The authors compared their screening results, and if the inclusion results were contradictory, they conferred the decision  with the third author (XF). If there was a problem such as missing data in the included literature, they contacted the first and/or corresponding authors through e-mail. If the literature was still not available, remove it. The information in the literature includes author, purpose, design, sample characteristics, recruitment procedures, participant exclusion and inclusion criteria, measurement of results, description of confounding factors and processing methods, and statistical methods and results (content) of leading indicators, used in this analysis. The literature is in Table 1.

study quality and risk of bias assessment
Two authors (CH and XJ) independently assessed risk of bias using the Newcastle-Ottawa Scale (NOS) document quality evaluation scale (revised edition) assessing risk of bias 14,15 . It consists of seven assessment contents divided into four categories: selection bias, design bias, statistical bias, and result bias. In the NOS scale, seven criteria were rated as either "high" (score = 3) or "low" (score = 0) for each included study, and the highest score is 21 points (see the Chart 1).

statistical analysis
The original or standardized regression coefficient (β) or the coefficient of determination (R 2 ) and standard error (SE) were extracted. The correlation coefficient (r) and standard deviation (SD) are extracted and summarized for correlation analysis. Used random-effects meta-analysis to derive a pooled estimate of the association between PA and HRV (SDNN, RMSSD, LF, HF). Applied Fisher's z transformation to correlation coefficients to calculate the relevant statistics (variance, standard error, confidence intervals) before converting to the correlation to report the summary effect size (ES). Using an inverse variance weighting procedure for independent effect sizes to improve overall precision 16 . Independent sample (k) as the unit of analysis. Pearson's r was the effect size metric selected to report results. Effect sizes used to Cohen's criteria for small (> 0.20), moderate (> 0.40), and large (> 0.80) aid the interpretation of results. Rosenthal's classic fail-safe N was used to examine publication bias. Analyses is carried out using comprehensive meta-analysis (version-2) 17 .
Depending on whether the units are consistent, selected the standardized mean differences (SMD) or mean difference (MD) for processing. Calculate the ES and 95% confidence interval (CI) of the outcome indicator. The small number of studies included (n = 4), Utilized Random Effects Models and Bayesian-classical analysis 16 . Analyses is carried out using Stata 16.0 software. The significance level was set at 0.05.

search results
The initial search yielded a total of 3031studies (EBSCO host = 248, Science Direct = 21, Springer Link = 1,245, PubMed = 1,169, Web of Science = 141). The reviewers excluded 2918 redundant documents in Endnote, removed 47 irrelevant documents after reading the title and abstract, 36 studies after reading the full text, and finally included 21 studies ( Figure 1).

Bias category score Comment content Evaluation criteria (0 to 3 points) Selection bias 3 points selection
This domain contained one subdomain regarding the source of the population. Low risk of bias was considered when random sampling was used and a high risk of bias was considered when a convenience sample was used without explanation of the recruitment procedures undertaken. Design bias 3 points sample size Low risk of bias was considered when the study provided an appropriate power analysis for sample size calculation, and when the study controlled for important confounders using appropriate statistical methods.

points confounders
Statistical bias 3 points statistical approach Low risk of bias was considered when an appropriate statistical approach was used, and the authors properly described how missing cases were handled. Studies not mentioning missing cases were considered to have a low risk of bias.

points missing data
Result bias 3 points PA test Low risk of bias was considered when the exposures were objectively measured and sufficient details provided to enable the measurement to be replicated by the reader. In addition, the appropriate subdomain for the measurement of HRV was duplicated. In this case, a low risk of bias was considered when studies presented sufficient details of the procedures taken before and during the measurement of HRV according to published guidelines [11].

points HRV test
Source: Authors.  Note: †Fisher's Z was used to calculate effect size statistics, k number of effect sizes, r effect size, SE standard error, s 2 variance, 95% CI 95% confidence interval, Z the test of the null hypothesis, Q total Q-value used to determine heterogeneity, I 2 the percentage of total variation across studies that is due to heterogeneity rather than chance, fail-safe N: the potential for publication bias to have influenced the results of a meta-analysis. Fail-safe N is the number of additional studies that would be needed to increase the P-value for the meta-analysis to above 0.05. The Same as below.

/A N/A N/A N/A N/A N/A N/A N/A N/A Adolescent N/A N/A N/A N/A N/A N/A N/A N/A N/A
Source: Authors.
High PA level significantly increases cardiac autonomic nervous control in children and adolescent. Five studies 10,25,26,30,33 have reported significant differences between the high and low PA levels. Three studies 10 10,25,26 reported LF/HF for reflects the sympathetic-vagal balance. A subgroup analysis showed that LF/HF was non-significant differences in children (MD = 0.33; 95%CI = -0.31, 0.97).

Discussion
The findings of our review show that: (a) PA and HRV were significantly positive correlated, PA can effectively improve cardiac autonomic function; (b) VPA and MVPA can improve HRV, while LPA and ST may have no effect on improving HRV; (c) The differences in the physiological characteristics of the participants (age, sex) and the testing tools (PA, HRV) may affect the results.

The relationship between PA and HRv
Our results show that there is a positive relationship between PA and HRV in children and adolescents. Our findings are consistent with previous studies which investigated different age groups, including young adults 35 and seniors 36 . There are three main ways to increase the ANS function: increased parasympathetic nerve activity, decreased sympathetic nerve activity and the role of the vagus nerve on sympathetic-parasympathetic. Our study shows a significantly positive correlated between PA and vagus nerve and parasympathetic nerve activity RMSSD (r = 0.22; 95%CI = 0.13, 0.30), HF (r = 0.46; 95%CI = 0.39, 0.53). Meanwhile, RMSSD was significantly higher in participants with high PA than the low PA levels (MD = 28.08; 95%CI = 17.50, 38.66). However, no statistical differences were found in HF (MD = 602.78; 95%CI = 0.47, 1,205.08). This is most likely due to the natural log transformation distribution distorting the short-term linear HRV metrics 10,33 . According to the effects of the sympathetic nerve activity, PA was significantly positive correlated with the LF (r = 0.58; 95%CI = 0.52, 0.64). The LF decreased significantly as compared to the population with low PA levels (MD = 309.96; 95%C I = 96.24, 523.68). This means the positive relationship between children's and adolescents' PA and HRV.
According to the current findings, MVPA has a significant effect on improving HRV and the impact of VPA on HRV is weak. Four of the 21 studies investigated the relationship between VPA and parasympathetic nerves 9,22,24,27 . Only one study 22 found a negative correlation between VPA and the HF (r = -0.66, P < 0.05) using a pedometer to distinguish PA levels. However, this study did not control for confounding factors and the pedometer test's validity for children and adolescents is low 8 . In addition, while combining MPA and VPA (MVPA), there is a significantly positive relationship between MVPA and HRV. Gutin et al. 21 and Radtke et al. 23 used subjective question-naire distinguish PA level, and it was discovered that MVPA was unrelated to HRV. However, the recall questionnaires should be with caution when used in children or adolescents 8 . Four studies 11,24,27,29 used accelerometers to distinguish PA leveland discovered a strongpositive association between MVPA and HRV. The current findings show that MVPA could be a significant factor for increasing HRV and improving ANS function.

Confounding factors and test tools
Our research aims to investigate the effect of PA levels on HRV in 3-18 years. The interpretation of the results is hampered by sample size, statistical methods, regional, outcome assessments and confounding factors. There is a positive relationship between PA level and HRV for participants aged 3-18 years remains controversial.

Confounding factors
The standard deviationof participants' age is very small (< 2) in the included study,except for Subramanian's study 25 . It means that the age range of included study is not large. Although half of the studies reported confusion about the age factor, only one 11 study controlled for age variables and discovered a significant link between 3-6 year old children's PA and HRV (β = -0.7, P < 0.01). There was also a link in the studies between the ages 6.1-13 y (β = 0.17, P < 0.05) 29 and 13.1-18 years (β = 0.18, P < 0.05) 27 . However, the difference inregression coefficient suggested that the influence of age in the results may be slight. Although biological maturity increases with age in children and adolescents. There was no research investigated the effect of biological maturity on the relationship between PA and HRV. It is difficult to conduct subgroup analyses based on maturity. Our findings show that reporting HRV in childhood alone is similar to combining childhood and adolescence, as describe in Table 2 and Table 3. Therefore, we speculated that maturity may have a minor impact on the relationship between PA and HRV.
The effect of body weight status on the observed relationship is weak. Two studies 20,33 found a significant positive correlation between PA and HRV in overweight children, although one of study has a higher risk of bias. In comparison, the lower risk of bias study with Farah et al. 7 also found a significant correlation. However, no significant correlation was found in a study of the risk of the same bias 21 . Therefore, we hypothesize that obesity will disrupt the normal maturation of cardiac autonomic function, while the weight shift induced by PA will encourage cardiac autonomic function improvement 37 .
Concerning the effect of sex on relationships, two studies 28,29 discovered a significant positive correlationin boys. Boys and girls aged 8-9 years were surveyed in studies, suggesting that the correlation between PA and HRV differs by sex 28,29 . Furthermore, boys have more MVPA than girls, which may explain the reproductive consequences of sex differences 29 .

Test tools
HRV -the first guideline for HRV measurement standards, physiological interpretation, and clinical use was published in 1996 6 . Our review included studies that were all published after 2000, adopting the HRV method and the measurement method obtained is generally well reported. The majority of studies provide linear HRV metrics indicators, which are described in Table 1. Only two studies have 11,23 reported on the currently popular non-linear indicator. Furthermore, using different devices (wearlink31, 810s, Rs810cs) to record HRV or different algorithms for analysis when comparing results will introduce possible deviations.When compared to ECG 38 , these Polar recorders have been recorded to be reliable and useful tools, particularly in such application scenarios. They can use Polar to collect and edit HRV data if appropriate 38 . Furthermore, the method of deciding HRV presents some challenges. The majority of the research used daytime test times and five-minute durations, but Herzig et al. used deep sleep to prevent environmental interference 11 . However, some tests for 2 minutes 18 , 6 minutes 26 , 5 hours 22 and 24 hours 24 . There were some studies using 256 RRi 21,27 , 1000 RRi 19 and short-term HRV. Although short-term HRV results are easy to measure, they can be difficult to interpret 39 , resulting in mistakes.
PA -self-reporting accounts for nearly half of the study's data, which has a number of advantages, convenience, efficiency, burden, and low cost 40 . The questionnaire, on the other hand, is inaccurate, particularly when evaluating the PA of young children 40,41 . While activity monitors, such as accelerometers, are an alternative or complementary tool for assessing PA 40 , but there is currently no agreement on which specifications should be used to collect and process data 41 . Published research, on the other hand, defined the criteria for collecting and processing accelerometer data in order to determine PA 42 . Among them are wearing position, sampling frequency, filter, epoch length, non-wearing time, significant day (week) time, sedentary behaviour, and cut-off value. Despite the fact that many studies failed to reveal significant methodological problems, seven studies 9,22,24,[27][28][29]31 include the accelerometer cut-off values (Table 1). Therefore, we recommend that researchers report detailed methodologies to improve the quality and reproductivity of future research.

Limitations
One limitation of our systematic review and meta-analysis is that it is cross-sectional in design. Therefore, cause and effect relations cannot be deduced. Furthermore, confounding factors such as gender, age, and BF% may interfere with the effect of PA on HRV. In addition, due to the relatively small number of studies on this issue, adolescents cannot be analysed separately when grouped by age. However, we present the results of the analysis for the children group and use the separated ages (children and adolescents) as the benchmark. The results show that there was no difference in analysis between combined (children) and separated ages (children and adolescents).It was also in the studies between the ages of 6.1-13 years (β = 0.17) 29 and 13.1-18 years (β = 0.18) 27 . However, the difference inregression coefficient suggested that the influence of age on the results may be slight. Therefore, we speculate that the effects of PA on HRV were consistent in children and adolescents' further studies are needed to determine whether age affects the relationship between PA and HRV.

Conclusion
The findings of our study revealed that there was a medium-sized association between PA and HRV in children and adolescents. Promoting children's and adolescents' participation in moderate-to-vigorous physical activity (MVPA) will increase parasympathetic nerve activity and decreased sympathetic nerve activity. The confounding factors and the testing tools will influence the relationship between PA and HRV. Our findings support motivating children and adolescents to engage in more MVPA in their daily lives to improve autonomic function and promote cardiovascular safety. In the future, there is still needed for high-quality cross-sectional studies in children and adolescents, more stringent control factors, more unified testing tool, greater doses of PA and more diverse cardiovascular health outcomes.