High-Intensity Interval Training versus Moderate-Intensity Continuous Training on Health Outcomes for Children and Adolescents: A Meta-analysis of Randomized Controlled Trials

s and title excluded during first screening (n=730) Figure 1: Flow diagram of literature search and trial selection process. 3 BioMed Research International


Introduction
Cardiorespiratory fitness (CRF) is an objective reproducible physiological response that is affected by physical activity habits, genetics, and disease status [1]. Low CRF in subjects is identified as a risk factor of cardiovascular morbidity and mortality [2]. Currently, the gold standard for CRF included maximal oxygen uptake, which is measured directly or indirectly by maximal graded cardiorespiratory test [3][4][5]. According to a study, high CRF during childhood and adolescence showed association with reduced risk of subsequent cardiovascular disease [6]. A study involving 25.4 million children and adolescents from 27 countries with CRF showed declination by 3.6% per decade [5]. Regular physical activity could improve CRF [7] and whether different types of physical activity yields differential effects on CRF and other cardiovascular risk factors in children and adolescents remains controversial.
Although the high-intensity interval training (HIIT) was completed within a shorter time, it increased aerobic fitness and mental health in children [8][9][10]. Children in schools require effective exercise programs to improve their physical fitness and spend shorter time to exercise in schools [11]. Recently, several studies have already compared the effects of HIIT versus moderate-intensity continuous training (MICT) on the outcomes of body mass index (BMI), hypertension, endothelial function, prediabetes, and type 2 diabetes in adults [12][13][14][15]. Moreover, the CRF, fat loss, and cardiometabolic health between HIIT and MICT in children and adolescents were compared, and various role of HIIT and MICT could explain by mitochondrial adaptations to short-term training [16,17]. Numerous studies have already been conducted to compare the effects of HIIT versus MICT in children and adolescents, which can enter into pooled analysis to reevaluate the effectiveness of HIIT versus MICT. Therefore, the current systematic review and meta-analysis was conducted on randomized controlled trials (RCTs) that compared the effects of HIIT versus MICT on the health outcomes of children and adolescents.

Materials and Methods
2.1. Data Sources, Search Strategy, and Selection Criteria. The Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) Statement was applied to guide the reporting of this systematic review [18]. Studies designed as RCT that compared the effects of HIIT versus MICT in children and adolescents were considered eligible in our study, and there is no restriction to publication language and status. The core search terms of (high intensity interval OR highintensity interval OR high intensity intermittent OR highintensity intermittent OR sprint interval OR HIIT OR HIIE) AND (children [MeSH] OR * adolescent [MeSH] OR boy OR girl OR youth [MeSH] OR kids OR student * ) AND (randomized controlled trials) were employed to search for potential trials from the PubMed, EmBase, and Cochrane Library electronic databases throughout December 2019. The reference lists from the retrieved studies were also reviewed manually to identify for studies that met the inclusion criteria.

Data
Collection and Quality Assessment. The data abstraction and quality evaluation from the retrieved studies were conducted by 2 authors independently, and any conflicts were resolved by group discussion. The data collected included the first author's surname, publication year, country, sample size, mean age, percentage male, intervention, control, follow-up duration, and reported outcomes. The Jadad scale was used to assess the quality of included trials, which is based on the following 5 subscales: randomization, concealment of treatment allocation, blinding, completeness of follow-up, and use of intention-to-treat analysis [19]. In this study, the trials that scored 4 or 5 were considered as high quality.
2.3. Statistical Analysis. The effects of HIIT versus MICT on health outcomes in children and adolescents were included as continuous data, and the weighted mean differences (WMDs) with 95% confidence intervals (CIs) were calculated before data pooling. The random-effects models were used to calculate the pooled effect estimates owing to it considering the underlying varies across included studies [20,21]. The heterogeneity across included trials were assessed using I 2 and Q statistics, and significant heterogeneity was defined as I 2 > 50:0% or P < 0:10 [22,23]. The stability of pooled conclusions was then assessed by a sensitivity analysis [24]. Subgroup analyses of investigated outcomes were conducted based on the mean age of subjects, and the differences between subgroups were assessed by interaction P test [25]. Publication biases for investigated outcomes were assessed by using the funnel plots, Egger, and Begg tests [26,27]. The inspection levels for pooled results are two-sided, and P < 0:05 was considered as statistically significant difference between HIIT and MICT. The data in this meta-analysis was analyzed by using the STATA software (version 12.0; Stata Corporation, College Station, TX, USA).

Literature Search.
A total of 1,846 articles were identified by initial electronic search, and 1,055 articles of these were excluded owing to duplications. Next, 730 articles were further excluded because of irrelevant topics. Full-text evaluations were done for the remaining 61 studies, and 45 studies were excluded because of the following reasons: no appropriate control (n = 33), no desirable outcomes (n = 6), and no RCT design (n = 6). After this, a total of 16 RCTs were considered eligible and included in the final meta-analysis [28][29][30][31][32][33][34][35][36][37][38][39][40][41][42][43]. No new eligible trial was identified by manual searching of the reference lists of retrieved studies ( Figure 1). The baseline characteristics of the included trials are summarized in Table 1.

Study Characteristics.
A total of 543 children were included in the trials that are published from 2005 to 2019. The follow-up duration ranged from 3.0 to 24.0 weeks, and 13-94 children were included in each individual trial. Thirteen RCTs were conducted in Western countries or Africa, and the remaining 3 were conducted in Asia. The mean age of children and adolescents included in the trials ranged from 8.2 to 17.4 years, and 4 trials included only boys. The Jadad scale was used to assess the study quality, in which 5 trials had a score of 4, 7 trials scored 3, 3 trials scored 2, and the remaining 1 trial scored 1.
3.3. Peak VO 2 . The data regarding the effect of HIIT versus MICT on peak VO 2 were available in 10 trials. The peak VO 2 was significantly higher in the HIIT group (WMD: 2.68; 95% CI: 1.81 to 3.55; P < 0:001; Figure 2), and no heterogeneity was detected across the included trials (I 2 = 0:0% ; P = 0:460). The conclusion was robust and unchanged by sequential exclusion of each individual trial (Supplemental 1 and Supplemental 3). Subgroup analysis showed significant differences between HIIT and MICT in peak VO 2 in children ≥ 12:0 years (  Figure 3), and significant heterogeneity was detected among the included trials (I 2 = 83:7%; P < 0:001). Sensitivity analysis found a stable conclusion and was unaltered by sequential exclusion of individual trial (Supplemental 1 and Supplemental 3). Subgroup analysis suggested that HIIT was associated with lower HR max if the mean age of children was <12.0 years ( Table 2). No significant publication bias was observed for HR max (P value for Egger: 0.596; P value for Begg: 1.000; Supplemental 2).

Fat
Mass. The data regarding the effect of HIIT versus MICT on fat mass were available from 11 trials. There were no significant differences between HIIT and MICT for fat mass (WMD: -0.15; 95% CI: -1.85 to 1.55; P = 0:863; Figure 4), and similarly, no potential significant heterogeneity across the included trials was observed (I 2 = 48:7%; P = 0:034). The conclusion was robust and unaltered by sequential exclusion of individual trials (Supplemental 1 and Supplemental 3). The results of subgroup analyses were consistent with the overall analysis ( Table 2). Although the Begg test indicated no significant publication bias, the Egger test showed potentially significant publication bias for fat mass (P value for Egger: 0.019; P value for Begg: 0.350; Supplemental 2). The conclusion was changed when the trim and fill method was used to adjust potential publication bias (Supplemental 2).
3.6. Free Fat Mass. The data regarding the effect of HIIT versus MICT on free fat mass were available in 4 trials. HIIT showed no significant effect on free fat mass when compared with MICT (WMD: 0.38; 95% CI: -2.32 to 3.09; P = 0:781; Figure 5), and no evidence of heterogeneity was observed (I 2 = 0:0%; P = 0:920). The conclusion remained robust after reviewing the results of sensitivity analysis (Supplemental 1 and Supplemental 3). The conclusions of subgroup analyses were consistent with that of overall analysis (Table 2). There was no significant publication bias for free fat mass (P value for Egger: 0.345; P value for Begg: 0.308; Supplemental 2).

3.7.
Weight. The data regarding the effect of HIIT versus MICT on weight were available from 11 trials, and no significant differences between HIIT and MICT for weight were observed (WMD: -0.46; 95% CI: -2.29 to 1.37; P = 0:623; Figure 6). Moreover, unimportant heterogeneity was detected for weight (I 2 = 13:5%; P = 0:312). Sensitivity analysis revealed robust conclusion and showed nonsignificant difference by sequentially excluding each trial (Supplemental 1 and Supplemental 3). Subgroup analyses suggested that the conclusions were consistent with the overall analysis in all subsets (  Figure 7), and unimportant heterogeneity was detected (I 2 = 19:9%; P = 0:237). Sensitivity analysis suggested that the conclusion remained unchanged by sequential exclusion of each trial (Supplemental 1 and Supplemental 3). Moreover, the results of subgroup analyses were consistent with that of the overall analysis in all subsets (  Figure 9), and significant heterogeneity was detected (I 2 = 65:1%; P = 0:005). The results of sensitivity analysis suggested that the conclusion was robust and unaltered by sequentially excluding individual trial (Supplemental 1 and Supplemental 3). The results of subgroup analyses showed no significant differences between HIIT and MICT on SBP in all subsets (   Warm-up (stretching of the large muscle groups and cardiovascular exercises at 30% of peak watt for five minutes), a sprint interval block (10 minutes), continuous aerobic exercise (10 minutes), another sprint interval block (10 minutes), and cooling down (stretching of the large muscle groups and cardiovascular exercises at 30% of peak watt for 5 minutes)

BioMed Research International
Warming up (stretching of the large muscle groups and cardiovascular exercises at 30% of peak watt for five minutes), cycling (10 minutes), walking/running (10 minutes), stepping (10 minutes), and cooling down (stretching of the large muscle groups and cardiovascular exercises at 30% of peak watt for five minutes) 15 Figure 10), and no evidence of heterogeneity was seen (I 2 = 0:0%; P = 0:681). The conclusion was robust and unaffected by sequential exclusion of individual trial (Supplemental 1 and Supplemental 3) and was stratified by the mean age ( Table 2). No significant publication bias for DBP was observed (P value for Egger: 0.165; P value for Begg: 0.174; Supplemental 2).
3.12. Glycemia. The data regarding the effect of HIIT versus MICT on glycemia were available in 4 trials. No significant difference between the groups for glycemia was observed (WMD: -2.14; 95% CI: -7.62 to 3.35; P = 0:445; Figure 11), and potentially significant heterogeneity was detected (I 2 = 64:3%; P = 0:038). The results of sensitivity analysis (Supplemental 1 and Supplemental 3) and subgroup analysis were consistent with that of the overall analysis (Table 2). No significant publication bias for glycemia was observed (P value for Egger: 0.879; P value for Begg: 0.734; Supplemental 2).
3.13. Insulinemia. The data regarding the effect of HIIT versus MICT on insulinemia were available in 6 trials. The results showed no significant effect of HIIT on insulinemia when compared with MICT (WMD: -1.72; 95% CI: -4.13 to 0.69; P = 0:163; Figure 12), and no significant heterogeneity across the included trials was observed (I 2 = 24:5%; P = 0:250). The results of sensitivity analysis (Supplemental 1 and Supplemental 3) and subgroup analysis were consistent with that of the overall analysis (Table 2). There was no significant publication bias for insulinemia (P value for Egger: 0.359; P value for Begg: 0.707; Supplemental 2).
3.14. Total Cholesterol. The data regarding the effect of HIIT versus MICT on TC were available in 5 trials. There was no significant difference between HIIT and MICT for TC (WMD: -6.22; 95% CI: -12.87 to 0.42; P = 0:066; Figure 13), and no evidence of heterogeneity across included trials (I 2 = 0:0%; P = 0:584). The results of sensitivity analysis (Supplemental 1 and Supplemental 3) and subgroup analysis were consistent with that of the overall analysis (  Figure 14), but significant heterogeneity was observed (I 2 = 72:5%; P = 0:012). The results of sensitivity analysis indicated that HIIT might be associated with low HDL (Supplemental 1 and Supplemental 3). Moreover, subgroup analysis found no significant difference between groups on HDL in all subsets (Table 2). There was no significant publication  Figure 15), and moderate heterogeneity was detected across the included trials (I 2 = 32:0%; P = 0:221). The results of sensitivity analysis (Supplemental 1 and Supplemental 3) and subgroup analysis were consistent with those of the overall analysis ( 3.17. Triglyceride. The data regarding the effect of HIIT versus MICT on TG were available from 5 trials. The results showed no significant differences between HIIT and MICT for TG was detected (WMD: -6.03; 95% CI: -13.83 to 1.77; P = 0:130; Figure 16), and no evidence of heterogeneity among the included trials (I 2 = 0:0%; P = 0:900). The results of sensitivity analysis (Supplemental 1 and Supplemental 3) and subgroup analysis were consistent with that of the overall analysis (Table 2). There was no significant publication bias     Figure 19), and significant heterogeneity was detected among the included trials (I 2 = 82:5%; P = 0:003). The results of sensitivity analysis (Supplemental 1 and Supplemental 3) and subgroup analysis were consistent with that of the overall analysis (Table 2). No significant publication bias was detected for leptinemia (P value for Egger: 0.888; P value for Begg: 1.000; Supplemental 2).

Discussion
This systematic review and meta-analysis was included RCTs that compared the effectiveness of HIIT and MICT on the health outcomes of children and adolescents. The findings of this meta-analysis found that HIIT could significantly improve peak VO 2 when compared with MICT, and HIIT showed association with lower HDL and high HbA1c levels. Our results found that children in the HIIT group had high peak VO 2 and HbA1c than those in the MICT group when the mean age of children was ≥12.0 years, and the HR max in children in the HIIT group was significantly lower than those in the MICT group if the mean age of children was <12.0 years. However, no significant differences were observed between HIIT and MICT on HR max , fat mass, free fat mass, weight, BMI, WC, SBP, DBP, glycemia, insulinemia, TC, HDL, LDL, TG, HOMA-IR, HbA1c, and leptinemia.
Several systematic reviews and meta-analyses have investigated the effectiveness of HIIT. The meta-analysis conducted by Bacon et al. including 334 subjects aged <45.0 years from 37 studies revealed the association of HIIT with high peak VO 2 [44]. A meta-analysis conducted by Costigan et al. suggested that HIIT was regarded as a feasible and timeefficient approach to improve CRF and body composition in adolescents [45]. Milanović   10 BioMed Research International underwent endurance training [46]. A meta-analysis conducted by García-Hermoso et al. included 9 studies and suggested that HIIT is considered as an effective and timeefficient approach to improve blood pressure and aerobic capacity levels than other forms of training in overweight and obese adolescents [47]. Maillard et al. conducted a meta-analysis by including 39 studies and found that HIIT could reduce fat-mass deposits in normal weight and overweight/obese adults [12]. Thivel et al. reported that HIIT significantly improved the maximal oxygen uptake and reduced body mass, body fat, SBP, DBP, and HOMA-IR in overweight and obese children and adolescents [48]. A metaanalysis conducted by Depiazzi et al. found that aquatic HIIT significantly improved the aerobic performance and lower limb strength in a nonathletic population [49]. A metaanalysis conducted by Cao et al. found that HIIT versus MICT showed significant improvement in CRF in children and adolescents [16]. However, the comprehensive health outcomes between HIIT and MICT in children and adolescents were not reported in prior studies. Therefore, the current meta-analysis was conducted to compare the effectiveness of HIIT with MICT in children and adolescents. The current study revealed that HIIT significantly improved the peak VO 2 than MICT, and the significant difference between the groups was mainly observed if the mean age of children was ≥12.0 years. Although the pooled results were consistent with those in the previous meta-analysis study, irrespective for adults or children [16,46], they applied standard mean difference as an effect estimate, and no stratification was done by the mean age of the subjects. The potential mechanism for this could be that HIIT versus MICT displayed a greater increase in mitochondrial content, including citrate synthase maximal activity, type II fiber activation, adenosine monophosphate activated protein kinase activity, and mass-specific oxygen flux [50,51]. Also, HIIT played an important role on central adaptation, including       13 BioMed Research International maximal stroke volume, cardiac output, and blood volume [52][53][54]. Finally, the improvement of peak VO 2 in the HIIT group was more evident in children aged 12.0 years or more, which could be explained by the intensity and regularity of training.
In this study, no significant differences between HIIT and MICT for HR max , fat mass, free fat mass, weight, BMI, WC, SBP, DBP, glycemia, insulinemia, TC, HDL, LDL, TG, HOMA-IR, HbA1c, and leptinemia were observed, which was not consistent with that of the previous findings. However, the results of HDL and HbA1c between the groups were not robust. The potential reason for this could be that smaller number of included trials reported these parameters, and the power was not enough to obtain a stable result between the HIIT and the MICT groups. Moreover, HbA1c was increased in the HIIT group if the mean age of children was ≥12.0 years. This result showed correlation with high energy intake after training in older children. Interestingly, HIIT showed association with low HR max than MICT if the children were < 12.0 years age. However, this result was based on 2 included tri-als, and so large-scale RCT should be conducted to verify the result.
There are several strengths in this study that should be highlighted: (1) this study provided the comprehensive health results between HIIT and MICT in children and adolescents; (2) the analysis of this study was based on RCTs, which included high evidence level results; (3) the analysis was based on inclusion of large number of trials, and the results were robust than any individual trial, and (4) stratified analyses for investigated outcomes according to the mean age of subjects were also conducted. Although above strengths, the limitations of this study should be acknowledged: (1) a smaller number of children were included in individual study, and the deviation was large that could affect the robustness of pooled conclusions; (2) various training types and intensities were included, which could affect the net effects between HIIT and MICT; (3) the status of subjects varied, and the place of training showed significant correlation with health outcomes; (4) several outcomes stratified by the mean age were restricted owing to only 1 study included; (5) whether the effectiveness between HIIT and 14 BioMed Research International MICT are differing based on weight status and sexual maturation were not conducted owing to mostly studies did not reported data stratified by weight status and sexual maturation; (6) the data abstracted was based on pooled results in each trial, restricting us in conducting a more detailed analysis; and (7) this analysis was based on published RCTs, and so publication bias was considered inevitable.

Conclusions
In summary, a high peak VO 2 in the HIIT group was observed than in the MICT group in children and adolescents. Sensitivity analyses suggested that the levels of HDL and HbA1c might differ between the HIIT and the MICT groups. Finally, the effects of HIIT versus MICT on peak VO 2 , HR max , and HbA1c might differ based on the mean age of the subjects. The long-term effects of HIIT versus MICT in children and adolescents require assessment in further large-scale RCTs.

Conflicts of Interest
The authors declare that there is no conflict of interest regarding the publication of this paper.

Authors' Contributions
Jun Yin and Zhixiong Zhou contributed equally to this work.

Acknowledgments
This work was supported by the support program for Highlevel Teacher Team Development of Beijing Municipal Institutions (IDHT20170515) and Serving National Special Needs in Doctoral Talents Development Program-Performance Training and Health Promotion for Adolescents.

Supplementary Materials
Supplemental 1 Figure S1: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on peak VO 2 . Figure S2: sensitivity analysis of highintensity interval training versus moderate-intensity continuous training on peak heart rate. Figure S3: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on fat mass. Figure S4: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on free fat mass. Figure S5: sensitivity analysis of high-intensity interval training versus moderateintensity continuous training on weight. Figure S6: sensitivity analysis of high-intensity interval training versus moderateintensity continuous training on body mass index. Figure S7: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on waist circumference. Figure S8: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on systolic blood pressure. Figure S9: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on diastolic blood pressure. Figure S10: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on glycemia. Figure  S11: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on insulinemia. Figure S12: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on total cholesterol. Figure S13: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on high density lipoprotein. Figure S14: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on low density lipoprotein. Figure S15: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on triglycerides. Figure S16: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on HOMA-IR. Figure S17: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on HbA1c. Figure S18: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on leptinemia. Supplemental 2 Figure S1: funnel plot of highintensity interval training versus moderate-intensity continuous training on peak oxygen uptake (VO 2 ). Figure S2: funnel plot of high-intensity interval training versus moderateintensity continuous training on peak heart rate. Figure S3: funnel plot of high-intensity interval training versus moderate-intensity continuous training on fat mass. Figure  S4: results of Trim and fill method for high-intensity interval training versus moderate-intensity continuous training on fat mass. Figure S5: funnel plot of high-intensity interval training versus moderate-intensity continuous training on free fat mass. Figure S6: funnel plot of high-intensity interval training versus moderate-intensity continuous training on weight. Figure S7: funnel plot of high-intensity interval training versus moderate-intensity continuous training on body mass index. Figure S8: Funnel plot of high-intensity interval training versus moderate-intensity continuous training on waist circumference. Figure S9: funnel plot of high-intensity interval training versus moderate-intensity continuous training on systolic blood pressure. Figure S10: funnel plot of highintensity interval training versus moderate-intensity continuous training on diastolic blood pressure. Figure S11: funnel plot of high-intensity interval training versus moderateintensity continuous training on glycemia. Figure S12: funnel plot of high-intensity interval training versus moderateintensity continuous training on insulinemia. Figure S13: funnel plot of high-intensity interval training versus moderateintensity continuous training on total cholesterol. Figure S14: funnel plot of high-intensity interval training versus moderate-intensity continuous training on high density lipoprotein. Figure S15: funnel plot of high-intensity interval training versus moderate-intensity continuous training on low density lipoprotein. Figure S16: funnel plot of highintensity interval training versus moderate-intensity continuous training on triglycerides. Figure S17: funnel plot of highintensity interval training versus moderate-intensity continuous training on HOMA-IR. Figure S18: funnel plot of highintensity interval training versus moderate-intensity continuous training on HbA1c. Figure S19: funnel plot of highintensity interval training versus moderate-intensity continuous training on leptinemia. Supplemental 3 Table S1: 15 BioMed Research International sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on peak VO 2 . Table  S2: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on peak heart rate. Table S3: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on fat mass. Table S4: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on free fat mass. Table S5: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on weight. Table S6: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on body mass index. Table S7: sensitivity analysis of highintensity interval training versus moderate-intensity continuous training on waist circumference. Table S8: sensitivity analysis of high-intensity interval training versus moderateintensity continuous training on systolic blood pressure. Table  S9: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on diastolic blood pressure. Table S10: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on glycemia. Table S11: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on insulinemia. Table S12: sensitivity analysis of highintensity interval training versus moderate-intensity continuous training on total cholesterol. Table S13: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on high density lipoprotein. Table S14: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on low density lipoprotein. Table S15: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on triglycerides. Table S16: sensitivity analysis of high-intensity interval training versus moderate-intensity continuous training on HOMA-IR. Table S17: sensitivity analysis of highintensity interval training versus moderate-intensity continuous training on HbA1c.