Sage Journals: Discover world-class research

Abstract

The purpose of this study was to examine the effects of exercise on cardiovascular disease (CVD) risk factors in children and adolescents with obesity. Randomized controlled trials (RCTs) of exercise ≥4 weeks in children and adolescents with obesity were included if one or more CVD risk factors were included as an outcome. Studies were retrieved by searching 7 electronic databases, cross-referencing, and expert review. Data were pooled using the inverse-variance heterogeneity (IVhet) model and strength of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) instrument. Thirty-nine studies representing 1548 participants (847 exercise, 701 control) met the inclusion criteria. Aerobic exercise improved 10 of 12 (83.3%) outcomes (P < .05 for all) while combined aerobic and strength training improved 5 of 8 (62.5%) outcomes (P < .05 for all). The strength of evidence ranged from “very low” to “moderate.” It was concluded that aerobic exercise, as well as combined aerobic and strength training, is associated with improvements in multiple CVD risk factors among children and adolescents with obesity. However, the generally low strength of evidence suggests a need for future well-designed and conducted RCTs on the effects of exercise, especially strength training, in children and adolescents with obesity.

Keywords

exercise meta-analysis obesity children adolescents

‘Cardiovascular disease (CVD) risk factors represent a major problem in youth with overweight and obesity, especially those with obesity.’

Overweight and obesity are major public health problems both in the United States and globally,^1,2 with obesity considered a disease since 2013.³ Between the years 1980 and 2013, the worldwide prevalence of children and adolescents 5 to 19 years of age with overweight and obesity increased from 16.9% to 23.8% in boys and from 16.2% to 22.6% in girls from developed countries.¹ For developing countries, increases were 8.1% to 12.9% in boys and 8.4% to 13.4% in girls.¹ Viewed from an absolute perspective, the World Health Organization (WHO) reported that 41 million children under the age of 5 and more than 340 million children and adolescents aged 5 to 19 had overweight or obesity in 2016.⁴ The prevalence of children and adolescents with overweight and obesity is especially problematic in the United States, increasing dramatically over approximately the past 5 decades. When compared to the years 1971 to 1974 and 2015 to 2016, the prevalence of US children and adolescents 2 to 19 years with overweight (body mass index [BMI] ≥ 85th to <95th percentile), obesity (BMI ≥ 95th percentile), and severe obesity (BMI ≥ 120th percentile) increased from 10.2% to 16.6%, 5.2% to 18.5%, and 1.0% to 5.6%, respectively.⁵ These increases were similar for both boys and girls.⁵ Finally, using National Health and Nutrition Examination Survey (NHANES) data for the years 1999 to 2016, a positive linear trend for all categories of overweight and obesity in children 2 to 19 years of age was observed, especially among adolescents.⁶ In addition, a significant increase in severe obesity among children 2 to 5 years of age was observed.⁶

Cardiovascular disease (CVD) risk factors represent a major problem in youth with overweight and obesity, especially those with obesity. For example, while the prevalence of high blood pressure (BP) in US youth decreased between 2001 and 2016, youth are still at risk, especially for those with obesity.⁷ Based on BMI percentile cutpoints, the prevalence of hypertension has been shown to be greater in youth with obesity compared to their healthy weight counterparts. Based on 2013 to 2016 data from the NHANES, the percentage of youth 12 to 19 years of age classified as having hypertension was reported to be 1.9% for those with a healthy weight, 1.9% for those who were overweight, and 9.4% for those who were obese, including those with severe obesity.⁷ When limited to those with severe obesity, the reported prevalence was 14.7%.⁷ Thus, when compared to healthy weight children and adolescents, those with obesity were 5 times more likely to have hypertension while those with severe obesity were 7 times more likely to be hypertense. Finally, when compared to normal weight children and adolescents 5 to 15 years of age, a systematic review with meta-analysis found increasing levels of resting systolic blood pressure (SBP) and diastolic blood pressure (DBP) in those with overweight or obesity, especially those with obesity.⁸ These findings suggest that youth with obesity are at the greatest risk of having elevated levels of resting SBP and DBP.

Abnormal cholesterol levels are also more prevalent in youth with overweight and obesity and also appear to be especially prominent in those with obesity. Based on 2011 to 2014 NHANES data in children and adolescents 6 to 19 years of age, the prevalence of total cholesterol (TC) ≥200 mg/dL was 6.8% for those who were normal weight, 6.9% for those who were overweight, and 11.6% for those classified as obese.⁹ For high-density lipoprotein cholesterol (HDL-C) <40 mg/dL, prevalence rates were 6.8%, 14.8%, and 33.2%, respectively, for healthy weight, overweight, and obesity,⁹ while non-HDL values (non-HDL-C) ≥145 mg/dL were 5.7%, 9.7%, and 16.7%.⁹ Collectively, the prevalence of either high TC, low HDL-C, or high non-HDL-C was 13.8% for those who were normal weight, 22.3% for those with overweight, and 43.3% for those with obesity.⁹ Thus, when compared to those with a healthy weight, those classified as overweight or obese were approximately 1.6 and 3.1 times more likely to have at least one abnormal cholesterol value. Another study, a systematic review with meta-analysis, found that when compared to normal weight children and adolescents, increased levels of TC, low-density lipoprotein cholesterol (LDL-C), and triglycerides (TG), as well as decreased levels of HDL-C were observed among those with overweight and obesity, especially among those with obesity.⁸ Thus, similar to resting SBP and DBP blood pressure, children with obesity seem to be at the greatest risk of having abnormal cholesterol levels.

Elevated levels of fasting glucose (FG) and insulin (FI) also appear to be problematic in youth with overweight and obesity, especially those with obesity. For example, a meta-analysis of 5- to 15-year-old children and adolescents found increasing levels of FG and FI among those with overweight and obesity, especially those with obesity.⁸ These findings in youth with obesity are similar to those observed for resting BP as well as lipids and lipoproteins.^7-9 Most notably, a systematic review concluded that children and adolescents with overweight and obesity are not only at an increased risk for premature morbidity during adulthood but also premature mortality.¹⁰

A cornerstone in the treatment of youth with overweight and obesity is exercise. However, and not surprisingly, both self-report and objective measures have shown that recommended levels of physical activity in children and adolescents with overweight and obesity are lacking. For example, using the US Youth Risk Behavior Surveillance System (YRBSS) data analysis tool and 2017 YRBSS data available at https://www.cdc.gov/healthyschools/data/yrbss.htm#null, the authors of this review estimated that 63.8% of high school females and 40.7% of high school males with overweight were not physically active at least 60 minutes per day on 5 or more of the past 7 days. For high schoolers with obesity, 70.7% of females and 52.0% of males were not physically active at least 60 minutes per day on 5 or more of the past 7 days (unpublished results). In addition, using NHANES data, Chung et al. examined accelerometer-based physical activity according to BMI percentiles derived from measured height and weight in 6- to 17-year-old males and females.¹¹ Overall, there was a greater decrease in meeting physical activity recommendations with (1) older age, (2) sex (females vs males), and (3) those with obesity when compared to those who were either at a healthy weight or overweight.¹¹ Finally, in a systematic review of 48 studies that objectively assessed habitual physical activity in children and adolescents, it was concluded that higher levels of habitual physical activity were protective against child and adolescent adiposity.¹² Collectively, these findings suggest that youth with obesity exhibit the lowest levels of physical activity.

There is evidence to suggest that CVD risk factors track into adulthood. For example, a systematic review with meta-analysis of 23 studies found a positive association between children with obesity and resting SBP, DBP, and TG, as well as an inverse association with HDL in adulthood.¹³ However, for those studies that adjusted for adult BMI, associations were reversed, suggesting that adult BMI may be a potential mediator.¹³ Another recent population-based cohort study in more than 1 million male adolescents found that low levels of cardiorespiratory fitness and obesity were associated with later risk of chronic disability due to CVD.¹⁴ The authors, as well as an accompanying editorial, suggested that preventive actions should begin at young ages and include the promotion of cardiorespiratory fitness and healthy body weight.^14,15 Finally, and most notably, obese youth have been shown to be at the greatest risk for premature morbidity and mortality throughout their lifetime.¹⁶ These as well as a number of other negative outcomes across the lifespan in youth with obesity have been discussed in detail elsewhere.¹⁷

Exercise has been deemed as important in the treatment of obesity in children and adolescents.¹⁸ However, despite numerous and conflicting results of randomized controlled trials (RCTs) focused on the effects of exercise (aerobic, strength training, or both) on CVD risk factors in youth with obesity,^19-58 a previous systematic review of previous systematic reviews with meta-analysis limited to RCTs of exercise,⁵⁹ and at least 17 original systematic reviews with meta-analysis limited to RCTs of exercise, have included both overweight and obese versus just obese youth.^60-73 One systematic review with meta-analysis of 9 randomized controlled exercise trials limited to obese children did report small standardized mean difference reductions, overall, in resting SBP and DBP⁷⁴ while another which also included 9 studies reported similarly small reductions in FG, FI, and percent body fat (BF%) as a result of aerobic exercise.⁷⁵ However, both meta-analyses were based on pooling models that have been shown to be less robust than currently available models.^76,77 Collectively, the above-mentioned factors are problematic because they prevent one from fully understanding the true magnitude of effect that exercise may have on CVD risk factors in those who are most vulnerable, that is, youth with obesity. Consequently, this inhibits clinicians, health care providers, and others from providing optimal support and treatment recommendations to this high-risk group. Thus, given (1) the prevalence of youth with obesity,^1,2,4,5 (2) the deleterious consequences of youth with obesity and CVD risk factors,^7-9,16 (3) the conflicting results of previous RCTs on exercise and CVD risk factors in youth with obesity,^19-58 and (4) the lack of focus of previous systematic reviews of previous meta-analyses as well as original systematic reviews with meta-analysis on exercise and multiple CVD risk factors in youth with obesity,^59-73 the purpose of this study was to conduct a systematic review and aggregate data meta-analysis of RCTs to determine the effects of exercise (aerobic, strength training, or both) on CVD risk factors in youth with obesity.

Methods

Overview

This study adhered to the guidelines from the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement for aggregate data meta-analyses of health care interventions.⁷⁸ In addition, the a priori protocol for this meta-analysis was registered in PROSPERO (#CRD42020179527)^79-81 but not published in a peer-reviewed journal.

Eligibility Criteria

The a priori inclusion criteria for this aggregate data meta-analysis were as follows: (1) participant-level RCTs with a comparative control group (no intervention, usual care, attention control, wait-list control, placebo), (2) exercise-focused intervention (aerobic, strength training, both), (3) studies lasting ≥4 weeks, (4) male and/or female children and adolescents 5 to 18 years of age, (5) participants considered obese, defined as a BMI ≥ 95th percentile or some other comparable criteria, (6) studies published in any language up to March 31, 2020, (7) data available for one or more of the following primary outcomes: measures of adiposity (BMI in kg/m², BMI-z-score, BMI percentile, fat mass [FM], BF%, waist circumference [WC]), blood pressure (resting SBP and DBP), lipids and lipoproteins (TC, HDL-C, LDL-C, TG), and markers of diabetes (FG, FI, glycated hemoglobin [HbA1c]). Studies were limited to RCTs because it is the only way to control for confounders that are not known or measured as well as the observation that non-RCTs tend to overestimate the effects of health care interventions.^82,83 For the purpose of this meta-analysis, exercise, aerobic exercise, and strength training were defined according to the 2018 Physical Activity Guidelines for Americans.¹⁸ Four weeks was chosen as the lower cut point for intervention length based on previous research demonstrating improvements in selected CVD risk factors over this period of time in 11-year olds.⁸⁴ There was no maximum cut point for intervention length. Participants were limited to obese children and adolescents, defined as either (1) ≥95th percentile based on BMI or BF%,^85-87 (2) a BMI >2 standard deviations above the WHO growth standard median (equivalent to a BMI of 30 kg/m² at 19 years of age) for ages 5 to 19,⁴ or (3) international BMI cut points according to age and sex for obesity in children ages 2 to 18.⁸⁸ It is important to note that while a BMI ≥25 kg/m² but <30 kg/m² is generally considered to represent overweight versus obesity in adults, many children and adolescents in this range and lower would be classified as obese, depending on age, sex, and country.⁸⁸ The focus was on obesity because it has been shown that this population is at the greatest risk for premature morbidity and mortality throughout their lifetime.^7-9,16 While some research has suggested that studies yielding statistically significant and positive results are more likely to be published in English-language versus non-English language journals,⁸⁹ other research has shown this to not be the case.^90,91 However, to be conservative, studies from both English and non-English-language articles were included, assuming there was an English-language abstract, with the latter translated into English by the second author using the freely available web-based Babelfish and Bing translators as well as the translation function in Google Documents. The primary outcomes chosen were based on their well-established association with CVD risk.¹⁷ Finally, multiple studies that included the same participants were handled by only including the originally published study.

Information Sources

Information sources included (1) the authors’ existing, large, and broad EndNote (version X8) database of randomized trials on the effects of exercise (aerobic, strength training, both) in overweight and obese children and adolescents,⁹² (2) conducting an updated search of 7 electronic databases, (3) cross-referencing from retrieved studies, and (4) expert review. A detailed description of the existing EndNote database has been previously described.^63,93 Briefly, this original database was derived by (1) searching 7 electronic databases, (2) cross-referencing from retrieved studies, and (3) expert review.^63,93 The database included published and unpublished (dissertations and theses) studies in any language, assuming an English language abstract was available, between January 1, 1973, and August 22, 2018. The year 1973 was chosen as the start date based on preliminary searching in which this was the earliest citation identified.⁹⁴ A post hoc decision was made to use the previously existing EndNote database for 2 reasons. First, this database was based on broader criteria focused on randomized trials as well as RCTs of exercise (aerobic, strength, training, concurrent aerobic and strength training) in both overweight and obese children and adolescents. Second, the investigative team did not want to duplicate effort.

In addition to the EndNote database, an updated search for studies available between August 23, 2018, and March 31, 2020, was conducted. This was accomplished by searching 7 electronic databases: (1) PubMed, (2) Web of Science, (3) Cochrane Central Register of Controlled Clinical Trials, (4) Cumulative Index to Nursing and Allied Health Literature, (5) SPORTDiscus, (6) Scopus, and (7) ProQuest Theses and Dissertations. Embase was not searched because Scopus has been reported to provide 100% coverage of Embase.⁹⁵ In addition, Embase was not readily available to the authors. Key words or forms of key words that were used and adapted to the unique characteristics of each database included the following: exercise, physical activity, physical fitness, children, adolescents, boys, girls, obesity, randomized controlled trials. For both the existing and new database searches, specific outcomes were not included in search strings because of concern about missing potentially eligible studies. All database searches, both existing and new, were conducted by the second author, with oversight from the first author. An example of the search strategy for one of the updated searches (PubMed) can be found in Supplementary File 1 (available online). For electronic database searches, the second author conducted separate searches in each database, imported them as separate files into Endnote X8, and then merged them into one large Endnote X8 file. The second author then removed duplicates both electronically and manually into another file, exclusive of duplicates. In addition to electronic database searches, cross-referencing from retrieved studies and previous reviews were performed as well as expert review by the third author (RP).

Study Screening

All studies were screened by the first 2 authors (GK and KK), independent of each other. The full report for each citation was retrieved for all titles and abstracts that appeared to meet the eligibility criteria as well as those where uncertainty existed. Multiple reports for the same study were handled by only including one of the studies. Blinding of the 2 screeners to journal titles, study authors, or institutions of the authors was not used during either the screening or data abstraction processes given previous research demonstrating neither a clinically nor statistically significant effect on results.⁹⁶ Major reasons for the exclusion of studies were based on PICOS criteria: (1) inappropriate population, (2) inappropriate intervention, (3) inappropriate comparison(s), (4) inappropriate outcome(s), (5) inappropriate study design and/or setting, (6) other.⁷⁸ Upon completion of initial screening, the first and second authors (GK and KK) met and reviewed their selections for agreement. Cohen’s kappa statistic (κ) was used to assess initial interselection agreement before resolving discrepancies.⁹⁷ Any discrepancies were resolved by discussion. If an agreement could not be reached, the third author (RP) provided a recommendation. Upon agreement of what studies to include, the second author then searched Web of Science to determine if any had been retracted. If so, they were excluded from the current study.⁹⁸ Based on the final number of studies to include, the overall precision of the searches was calculated by dividing the number of included studies by the total number of studies screened after removing duplicates.⁹⁹ The number needed-to-read (NNR) was then calculated as the reciprocal of the precision.⁹⁹ Both the existing and new database files were then merged into one overall file using version X8 of Endnote.⁹²

Data Abstraction

Microsoft Excel (version 2016) was used to develop a comprehensive electronic workbook with multiple worksheets that defined the coding process for each of the variables coded from each study. The workbook was created, pilot-tested, and finalized by the authors. The major groups of variables coded were (1) study characteristics (author, journal, year of publication, etc.), (2) participant characteristics (age, sex, height, body weight [BW], etc.), (3) intervention characteristics (length, frequency, intensity, mode, etc.), and (4) primary and secondary outcome data (sample sizes, baseline and postexercise means and standard deviations, method of assessment, etc.). Missing data for primary outcomes were requested from the corresponding author via electronic mail. To avoid data abstraction bias, the first 2 authors (GK and KK) used separate workbooks to independently code (dual coding) each item from every study to ensure accuracy and consistency. The authors then met to review their selections. Interrater agreement prior to resolving discrepancies was assessed using Cohen’s κ statistic.⁹⁷ Any disagreement in the items coded was discussed until mutual agreement was reached. If agreement could not be reached, the third author (RP) provided a recommendation. Based on Cohen’s κ statistic, the overall agreement rate prior to correcting disagreements was 0.96.

Outcomes and Prioritization

The primary a priori outcomes included measures of adiposity (BMI in kg/m², BMI-z-score, BMI percentile, FM, BF%, WC), blood pressure (resting SBP and DBP), lipids and lipoproteins (TC, HDL-C, LDL-C, TG), and markers of diabetes (FG, FI, HbA1c). Secondary outcomes included BW, fat-free mass (FFM), energy intake (EI), physical activity (PA) levels, relative maximum oxygen consumption (VO_2max in mL·kg·min⁻¹), and muscular strength. Post hoc, a decision was made to include the number of dropouts in the exercise and control groups as an outcome. Meta-analysis was conducted if there were at least 5 effect sizes for each outcome given that error estimation has been shown to deteriorate with less than 5 effect sizes for the inverse variance heterogeneity (IVhet) pooling model used in the current study.⁷⁶

Risk of Bias Assessment in Individual Studies

Risk of bias for each study was assessed using the recently revised Cochrane Risk of Bias instrument for RCTs (RoB 2).¹⁰⁰ Using one or more signaling questions, the RoB 2 instrument assesses risk of bias in 5 distinct domains: (1) bias arising from the randomization process, (2) bias due to deviations from intended interventions, (3) bias due to missing outcome data, (4) bias in measurement of the outcome, and (5) bias in selection of the reported result.¹⁰⁰ Based on signaling questions, each domain is assessed as either “low risk,” “high risk,” or “some concerns.”¹⁰⁰ Based on responses to each domain, the overall risk of bias for each study is then assessed as either “low risk,” “high risk,” or “some concerns.”¹⁰⁰ We chose to use this instrument over the various study quality instruments, including those focused on exercise intervention studies,^101,102 given the difficulty of the latter in differentiating between the quality of reporting and the quality in the conduct of a study.¹⁰⁰ Risk of bias was assessed by the first 2 authors (GK, KK), independent of each other. They then met and reviewed their selections for agreement. If agreement could not be reached, the third author (RP) rendered a decision. Using Cohen’s κ statistic, the overall agreement rate prior to correcting discrepant items was 0.89.

Data Synthesis

Baseline Characteristics

Differences in initial characteristics (age, height, BW, etc.) for each study were calculated by taking the difference in baseline values between the exercise and control groups and pooling their standard deviations. For studies that included multiple exercise intervention groups, the control group sample size was divided by the number of intervention groups so as to avoid overcounting.¹⁰³ The IVhet model, a model that has been shown to be more robust than the traditional random-effects model,^76,104,105 was then used to pool results. A 2-tailed z-α value ≤.05 and nonoverlapping 95% confidence intervals (CIs) were considered to represent significant between-group (exercise vs control) differences.

Calculation of Effect Sizes for Primary and Secondary Outcomes

With the exception of muscular strength and dropouts, effect sizes from each study and each outcome were calculated using the original metric. These were calculated as the change outcome difference in the exercise group minus the change outcome difference in the control group. Variances from each study were calculated from pooled change outcome standard deviations in the exercise and control groups. If change score standard deviations were not available, they were estimated from 95% CIs for change outcomes, 95% CIs for treatment effects, or pre and post standard deviation values according to procedures developed by Follmann et al.¹⁰⁶ Studies that reported data using metrics which could easily be converted between each other, for example, millimoles to milligrams per deciliter for TC, were converted into the same metric. For those studies that included multiple exercise intervention groups, the control group sample size was divided by the number of intervention groups so as to avoid overcounting.¹⁰³

Given the different metrics used to assess muscular strength across studies and the inability to convert between all of them, the a priori plan was to use the standardized mean difference effect size, adjusted for small sample bias (Hedge’s g) as the common metric.¹⁰⁷ However, since none of the exercise groups (aerobic, strength, combined aerobic and strength) included at least 5 effect sizes, this analysis was not conducted. For dropout comparisons in the exercise and control groups, the odds ratio (OR) was the metric of choice.

Pooling of Effect Sizes for Primary and Secondary Outcomes

Changes in all outcomes were pooled according to type of exercise (aerobic, strength, combined aerobic and strength) using the IVhet model, a quasi-likelihood model.^76,104,105 Two-tailed z-α values ≤0.05 and nonoverlapping 95% CIs were considered statistically significant, with a focus more on CIs.^108,109 Post hoc, a decision was made to also calculate 95% prediction intervals.^110,111 Prediction intervals may best be described as what result one might expect to find in an outcome if a new RCT was conducted.

Heterogeneity and inconsistency for pooled outcomes were estimated using the Q¹¹² and I² statistics,¹¹³ respectively. An α level <.10 for Q was considered to represent statistically significant heterogeneity.¹¹³ Higher values for I², a relative between-studies measure, represent the percentage of between-studies variance accounted for by true effects versus sampling error.¹¹³ I² values were categorized as very low (<25%), low (25% to <50%), moderate (50% to <75%), or large (≥75%).¹¹³ Absolute between-study heterogeneity was calculated using tau squared (τ²). In addition, influence analysis was conducted with each study deleted from the model once to examine the influence of each study on the overall results. Furthermore, cumulative meta-analysis, ranked by year, was conducted to examine the accumulation of results over time.¹¹⁴ Finally, while not part of the a priori protocol, a post hoc decision was made to conduct outlier analysis as a form of sensitivity analysis by excluding results for those effect sizes in which their 95% CI fell completely outside the pooled 95% CI.

Results for multiple intervention groups from the same study were based on group findings. For those studies that reported both per-protocol (PP) and intention-to-treat analyses (ITT), only ITT analyses were included since they usually result in more conservative estimates of effect.

Our a priori protocol planned for conducting simple meta-regression analysis. However, because of the collectively small number of effect sizes within each exercise modality (aerobic, strength training, or both), a post hoc decision was made to not conduct such analyses given the sample sizes suggested for continuous and categorical meta-regression results.^103,115

Because of concerns about missing possibly important findings that could be pursued in future RCTs, no adjustments were made for multiple testing.¹¹⁶ While some may consider this to be a “fishing expedition,” such analyses are considered important for providing investigators with potential direction for future RCTs, one of the very reasons for conducting a systematic review with meta-analysis.¹¹⁷

Meta-Biases

Small-study effects (publication bias, etc.) were assessed qualitatively using the Doi plot and quantitatively using the Luis Furuya-Kanamori (LFK index). The Doi plot has been suggested to be more intuitive than the funnel plot, and the LFK index more robust than the commonly used Egger’s regression-intercept test.^118,119 LFK values within ±1, greater than ±1 but within ±2, and greater than ±2 were considered to represent no, minor, and major asymmetry, respectively.¹¹⁸

Strength of Evidence

Strength of findings for each outcome was assessed using the most recent version of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) for meta-analysis tool.^120,121 Quality of evidence was assessed across the domains of risk of bias, consistency, directness, precision, and publication bias. The overall strength of evidence (certainty) was judged as either high (further research is very unlikely to change our confidence in the estimate of effect), moderate (further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate), low (further research is very likely to have an important impact on confidence in the estimate of effect and is likely to change the estimate), or very low (very uncertain about the estimate of effect).^120,121 Assessment was conducted using procedures similar to those for assessing risk of bias.

Software Used for Data Synthesis

All data were analyzed using (1) Stata (version 16.1, Stata/SE for Windows; Stata, 2020), (2) the admetan package within Stata, (3) Microsoft Excel for Office 365, (4) Meta XL (version 5.3, 2016), and (5) SSC-Stat (version 3.0, 2017, University of Reading, UK).

Results

Study Characteristics

Figure 1 contains a flow diagram that depicts the results of study screening. Of the 7629 studies screened, 39 representing 86 groups (47 exercise, 39 control) and 1548 participants (847 exercise, 701 control) met the criteria for inclusion.^19-34,36-58 Nine of the included studies were identified via the updated searches.^{22,26,27,37,41,44,49,51,57} The precision of the searches was 0.5% while the NNR was 196. The major reasons for the exclusion of studies were (1) inappropriate study design (44.3%), (2) inappropriate population (24.6%), (3) inappropriate intervention (14.8%), (4) inappropriate outcome (7.4%), (5) inappropriate comparison (1.0%), and (6) other factors such as not being an actual study (7.9%). A search of the 39 included studies on April 14, 2020, found that none had been retracted. A reference list of the 7590 excluded studies, including reasons for exclusion, is available upon reasonable request from the corresponding author. Missing primary outcome data were provided by all 3 studies for which data were requested.^20,22,37

Figure 1.

Flow diagram for the selection of studies.

A description of selected study-level characteristics is shown in Table 1 with additional information described below. Studies were published between 1997 and 2020 (X– ± SD = 2012 ± 5.0, median = 2013) in 28 different journals. The overall impact factor for the 26 journals in which impact data were available ranged from 0.02 to 12.64 (X– ± SD = 2.5 ± 2.8, median = 1.5). Thirty-five of 39 studies (89.7%) were published in the English language^{19-22,26-34,36-39,41-58} while 4 others were published in Chinese.^23-25,40 The studies took place in 18 different countries, 7 in South Korea,^{22,36,37,49-51,57} 5 in either China^23-25,40,52 or the United States,^{31,34,38,39,45} 4 in Tunisia,^21,28,46,47 3 in Iran,^30,44,58 2 each in either Australia,^54,55 Turkey,^33,48 or Iran^{30, 44} and 1 each in either Brazil,⁵³ Canada,¹⁹ France,²⁰ Germany,⁴³ Indonesia,²⁷ Singapore,⁵⁶ Sweden,³² Switzerland,²⁹ Taiwan,⁴¹ Thailand,²⁶ or the United Kingdom.⁴² Twenty-two (56.4%) of the studies reported funding associated with their study,^{19,21,26,28,29,31,32,34,36,38,39,41,43-47,50,53-56} with 7 reporting more than one source of funding.^{29,34,36,38,39,45,53} Funding was derived from government,^{21,28,29,31,32,34,38,39,41,45-47,53} university,^{19,26,29,34,36,43,44,50,56} and private sources.^{36,38,39,45,53-55} Thirteen studies (33.3%) provided sample size estimates for their primary outcome(s) of interest.^{22,29,34,38,41,44,45,49-51,53,55,57} For the 22 studies (56.4%) in which it could be clearly discerned,^{19,22,26,27,29,31,32,34,38-43,45-47,49,50,52,53,58} 12 used the PP approach to analyze their data,^{19,26,31,32,34,43,45-47,50,53,58} 5 used ITT,^{22,27,42,49,52} and 4 used both.^29,38,39,41 Another study reported that all participants initially enrolled completed the study.⁴⁰ Two studies reported matching subjects by either sex⁴¹ or sex and ethnicity.⁴⁵ Seven studies included more than one exercise group.^{24,26,38,39,44,46,47} Two studies included an attention-control group^37,51 while 4 wait-listed controls.^19,32,41,53 Another 2 studies reported using a randomized crossover design.^54,55 As shown in Table 1, and while noting that obesity cutpoints will vary by country,⁸⁸ the methods used for classifying obesity varied between the different studies.^19-34,36-58

Table 1.

General Characteristics of Studies.

Study	Year	Country	N	Age (years) (X– ± SD) or range	Sex	Obesity criteria^a
^bAlberga et al.¹⁹	2013	Canada	AE + ST = 12CON = 7	AE + ST = 10.0 ± 1.0CON = 10.0 ± 2.0	MF	≥95th percentile
Andre and Beguier²⁰	2015	France	AE = 8CON = 8	AE, CON 14.4 ± 1.5	MF	BMI = 37.7 kg/m.²
Ben Ounis et al.²¹	2010	Tunisia	AE = 16CON = 16	AE = 13.4 ± 0.4CON = 13.2 ± 0.6	MF	BMI >97th percentile
Bharath et al.²²	2018	South Korea	AE + ST = 20CON = 20	AE + ST = 14.6 ± 4.5CON = 14.8 ± 4.5	F	BMI ≥30 kg/m²
Cao et al.²³	2012	China	AE = 20CON = 20	AE = NACON = NA	M	BMI ≥ 25 kg/m²
Chen et al.²⁴	2015	China	AE =15ST = 15AE + ST = 15CON = 15	AE = 14.1 ± 3.1ST = 13.9 ± 2.2AE + ST = 14.2 ± 3.8CON = 14.4 ± 3.2	NA	BMI ≥ 25 kg/m²
Cheng et al.²⁵	2012	China	AE + ST = 30CON = 30	AE + ST, CON = 13.0-14.0	M	BMI ≥ 25 kg/m²
Chuensiri et al.²⁶	2018	Thailand	AE (HIIT) = 16AE (SHIIT) = 16CON = 16	AE (HIIT) = 11.0 ± 1.0AE (SHIIT) = 11.1 ± 0.8CON = 10.6 ± 1.0	M	BMI ≥ 2 SD based on WHO criteria
Djaafar et al.²⁷	2019	Indonesia	AE = 49CON = 49	AE = 11.0 ± 0.7CON = 11.1 ± 0.8	MF	Body fat >30% in AE and CON groups
Elloumi et al.²⁸	2011	Tunisia	AE = 7CON = 8	AE = 13.1 ± 1.0CON = 13.2 ± 0.2	M	BMI >97th percentile
Farpour-Lambert et al.²⁹	2009	Switzerland	AE + ST = 22CON = 22	AE + ST = 9.1 ± 1.4CON = 8.8 ± 1.6	MF	BMI >97th percentile
Fazelifar et al.³⁰	2013	Iran	AE + ST = 12CON = 12	AE + ST, CON = 11-13	M	BMI >28 kg/m²
Gutin et al.³¹	1997	USA	AE = 17CON = 18	AE = 9.6 ± 0.8CON = 9.5 ± 1.3	MF	Body fat >43% in AE and CON groups
Hagstromer et al.³²	2009	Sweden	AE + ST = 16CON = 15	AE + ST = 13.7 ± 2.0CON = 13.6 ± 2.2	MF	Referred to obesity treatment clinic; BMI >31 kg/m² in AE and CON groups
Karacabey³³	2009	Turkey	AE = 20CON = 20	AE = 11.8 ± 0.5CON = 11.2 ± 0.2	M	BMI ≥ 30 kg/m²
Kelly et al.³⁴	2015	USA	ST = 13CON = 13	ST = 15.2 ± 0.9CON = 15.5 ± 0.9	M	≥95th percentile based on CDC BMI charts
Kim et al.³⁶	2007	South Korea	AE = 14CON = 12	AE = 17 ± 0.4CON = 17.0 ± 0.4	M	Mean BMI = 29.5 kg/m²
Kim et al.³⁷	2020	South Korea	AE =24CON = 24	AE = 15.0 ± 1.0CON = 15.0 ± 1.0	F	BMI >2 SD based on WHO criteria
Lee et al.³⁸	2012	USA	AE = 16ST = 16CON = 13	AE =15.2 ± 0.9ST = 14.6 ± 1.5CON = 14.8 ± 1.4	M	≥95th percentile based on CDC BMI charts
Lee et al.³⁹	2013	USA	AE = 16ST = 16CON = 12	AE = 14.6 ± 1.9ST = 14.8 ± 1.9CON = 15.0 ± 2.2	F	≥95th percentile based on CDC BMI charts
Li et al.⁴⁰	2014	China	AE = 20CON = 20	AE = 15.4 ± 2.6CON = 14.6 ± 3.5	M	BMI ≥ 25 kg/m²
Liu et al.⁴¹	2018	Taiwan	AE = 40CON = 40	AE = 13.9 ± 0.8CON = 14.2 ± 0.7	MF	BMI > 24 kg/m² for Taiwan
McNarry et al.⁴²	2015	UK	AE = 15CON = 11	AE, CON = 9.3 ± 0.9	MF	BMI ≥ 95th percentile
Meyer et al.⁴³	2006	Germany	AE = 33CON = 34	AE = 13.7 ± 2.1CON = 14.1 ± 2.4	MF	>97th percentile for German pediatric population
Moslehi et al.⁴⁴	2019	Iran	AE (IE) = 10AE (OE) = 10CON = 10	AE (IE) = 10.9 ± 0.5AE (OE) = 11.1 ± 0.5CON = 10.9 ± 0.5	M	BMI 29-31 kg/m²
Owens et al.⁴⁵	1999	USA	AE = 35CON = 39	AE = 9.5 ± 1.2CON = 9.4 ± 1.3	MF	Triceps skinfold >85th percentile for gender, age, and ethnicity (mean percent fat >44% in AE and CON groups)
Racil et al.⁴⁶	2013	Tunisia	AE (MIIT) = 11AE (HIIT) = 11CON = 12	AE (MIIT) = 16.3 ± 0.5AE (HIIT) = 15.6 ± 0.7CON = 15.9 ± 1.2	F	BMI >97th percentile according to French standards
Racil et al.⁴⁷	2016	Tunisia	AE (IT) = 23AE + ST (IT + PLYO) = 26CON = 19	AE(IT) = 16.6 ± 0.9AE + ST (IT + PLYO) = 16.5 ± 1.2CON = 16.9 ± 1.0	F	Obese based on CDC growth charts
Saygin and Ozturk⁴⁸	2011	Turkey	AE = 20CON = 19	AE, CON = 10.0-12.0	F	Reported as obese but no cutpoints provided (body fat >31% in AE and CON groups)
Son et al.⁴⁹	2017	South Korea	AE + ST = 20CON = 20	AE = 15.0 ± 0.5CON = 15.0 ± 0.5	F	BMI ≥ 30 kg/m², ie, 95th percentile for age and sex, and abdominal obesity (waist circumference >80 cm)
Song et al.⁵⁰	2012	South Korea	AE = 12CON = 10	AE = 12.7 ± 0.7CON = 12.6 ± 0.6	M	Body fat >30% for AE and CON groups
Sung et al.⁵¹	2019	South Korea	AE = 20CON = 20	AE = 15.0 ± 1.0CON = 15.0 ± 1.0	F	Abdominal obesity (waist > 80 cm) and BMI ≥95th percentile for age and sex
Tan et al.⁵²	2010	China	AE = 30CON = 30	AE = 9.4 ± 0.5CON = 9.5 ± 0.5	MF	Body mass >20% of standard body mass for height of Chinese children
Vasconcellos et al.⁵³	2016	Brazil	AE = 10CON = 10	AE = 14.1 ± 1.3CON = 14.8 ± 1.5	MF	BMI >2 SD above age- and sex-specific WHO reference medians
Watts et al.⁵⁴	2004	Australia	AE + ST = 19CON = 19	AE+ST,CON = 14.3 ± 1.5	MF	All children reported as obese, but no definition given
Watts et al.⁵⁵	2004	Australia	AE = 14CON = 14	AE, CON = 8.9 ± 1.6	MF	All children reported as obese based on age- and sex-specific cutpoints from a previous study
Wong et al.⁵⁶	2008	Singapore	AE + ST = 12CON = 12	AE + ST = 13.8 ± 1.1CON = 14.3 ± 1.5	M	All children reported as obese, but no definition given
Wong et al.⁵⁷	2018	South Korea	AE + ST =15CON = 15	AE + ST = 15.2 ± 4.7CON = 15.3 ± 4.3	F	BMI ≥ 95th percentile for age and sex and abdominal obesity; waist to height ratio of >0.5
Zehsaz et al.⁵⁸	2017	Iran	AE = 16CON = 16	AE = 10.8 ± 0.9CON = 10.3 ± 0.9	M	BMI >25 kg/m²

Abbreviations: USA, United States of America; UK, United Kingdom; N, number of participants; AE, aerobic exercise, ST, strength training; AE + ST = aerobic exercise + strength training; CON, control; X –± SD, mean ± standard deviation; HIIT, high-intensity interval training; SHIIT, supra-high-intensity interval training; IE, indoor exercise; OE, outdoor exercise; MIIT, moderate-intensity interval training; IT, interval training; IT + PLYO, interval training + plyometrics; M, male; F, female; NA, not available; SD, standard deviation; WHO, World Health Organization.

Cutpoints for obesity vary by country.

Considered to be combined aerobic and strength training because each strength training session preceded by 20 minutes of aerobic exercise at 60% to 70% of maximum heart rate.

Participant Characteristics

Baseline characteristics for participants are shown in Table 2. As can be seen, WC and FI levels were greater for exercise versus control groups. No other significant differences were observed. For those studies in which it could be delineated, 15 (39.5%) included both males and females,^{19-21,27,29,31,32,41-43,45,52-55} 14 (36.8%) were limited to males,^{23,25,26,28,30,33,34,36,38,40,44,50,56,58} and 9 (23.7%) to females.^{22,37,39,46-49,51,57} For those studies reporting data, race/ethnicity, as reported by the authors, included Blacks, Whites, Hispanics, Asians, as well as multiple groups.^{22-25,31,34,36-41,45,49-52,55-57} Studies included participants with hyperlipidemia,^21,29 hypertension or prehypertension,^{21,29,37,49,51} hyperinsulinemia,^37,49,57 and insulin resistance.²⁹ None of the studies reported any type of drug use by participants at the start of the trial that could impair exercise or weight loss. In addition, none of the studies reported that participants were taking any type of hormone replacement therapy, smoked cigarettes, consumed alcohol, had type 1 or type 2 diabetes, heart disease, metabolic syndrome, cancer, or asthma. As defined by the authors, 12 studies specifically reported that participants were sedentary prior to study initiation.^{22,26,29,30,33,37-39,43,49,51,57} One study reported that participants in the exercise group decreased their normal daily physical activity levels during the intervention,³² while 2 other studies reported a decrease in physical activity in the control groups.^29,45 No serious adverse events were reported by any of the studies.

Table 2.

Baseline Characteristics of Participants^a.

	Exercise			Control
Variable	#S/G/P	X– ± SD	Mdn	#S/G/P	X– ± SD	Mdn	Dif (P)
Age (years)	35/43/770	13.2 ± 2.3	13.9	35/35/629	13.1 ± 2.5	14.2	.57
Height (cm)	30/36/651	157.6 ± 9.7	161.2	30/30/543	156.4 ± 10.6	159	.06
BW (kg)	36/44/754	75.8 ± 14.3	76.4	36/36/613	74.6 ± 15.3	75.5	.90
BMI (kg/m²)	36/42/746	29.7 ± 3.3	29.3	36/36/642	29.6 ± 3.3	29.3	.91
BMI z-score	5/7/142	2.8 ± 0.4	2.9	5/5/103	2.8 ± 0.4	2.8	.54
FM (kg)	17/21/340	31.0 ± 9.4	29.3	17/17/268	30.8 ± 8.4	30.8	.62
BF (%)	30/38/692	36.3 ± 6.8	35.7	30/30/557	36.4 ± 6.4	35.4	.32
WC (cm)	19/24/404	93.7 ± 8.4	92.6	19/19/309	92.0 ± 8.5	87.5	.04*
FFM (kg)	18/23/393	48.5 ± 9.8	50.9	18/18/300	47.3 ± 10.7	49.2	.77
SBP (mm Hg)	17/18/317	124.6 ± 7.3	125.0	17/17/301	123.6 ± 6.8	122.8	.26
DBP (mm Hg)	10/18/306	73.3 ± 7.7	75.5	17/17/289	72.9 ± 9.1	72.9	.47
TC (mg/dL)^b	10/12/183	157.0 ± 18.3	161.0	10/10/178	160.7 ± 21.2	164.2	.88
HDL-C (mg/dL)^b	14/16/274	44.1 ± 5.3	43.3	14/14/248	44.0 ± 5.6	45.2	.80
LDL-C (mg/dL)^b	12/14/236	99.5 ± 9.9	104.1	12/12/210	98.8 ± 13.1	98.1	.88
TG (mg/dL)^c	13/15/254	114.0 ± 25.2	118.4	13/13/228	109.3 ± 23.1	111.6	.07
FG (mg/dL)^d	18/22/380	89.6 ± 6.1	90.7	18/18/301	89.6 ± 6.9	73.9	.42
FI (µU/mL)	12/16/287	24.6 ± 9.6	21.6	12/13/209	22.3 ± 8.8	19.8	<.001*
HbA1c (%)	ID^e	ID	ID	ID	ID	ID	ID
VO_2max (mL·kg·min⁻¹)	17/25/35	30.3 ± 4.7	30.4	17/17/261	29.8 ± 6.0	29.2	.71
EI (kcal)	ID	ID	ID	ID	ID	ID	ID

Abbreviations: #S/G/P, number of Studies, Groups, and Participants, with descriptive analysis based on number of groups; X –± SD, mean ± standard deviation; Mdn, median; Dif (P), alpha-value for differences between exercise and control groups; BW, body weight; BMI, body mass index; FM, fat mass; BF (%), percent body fat; WC, waist circumference; FFM, fat-free mass; SBP, resting systolic blood pressure; DBP, resting diastolic blood pressure; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; TG, triglycerides; FG, fasting glucose; FI, fasting insulin; HbA1c, glycated hemoglobin; VO_2max, maximum oxygen consumption; EI, energy intake.

Baseline data for muscular strength and physical activity not reported because of the lack of a common metric across studies.

To convert TC, HDL-C, and LDL-C to millimoles per liter, divide by 38.67.

To convert TG to millimoles per liter, divide by 88.57.

To convert FG to millimoles per liter, divide by 18.02.

ID, insufficient data (N < 5).

Statistically significant (P < .05).

For the 16 studies (41.0%) that reported adequate data,^{22,26,27,29,34,38-41,43,45-47,49,52,53} the dropout rate ranged from 0% to 37.5% in both the exercise and control groups (exercise, X– ± SD, 10.1 ± 11.5%, control, 12.9 ± 13.6%). No significant difference was observed in dropouts between the exercise and control groups (OR, 0.82; 95% CI, 0.52 to 1.29, P = .39). Reasons given by participants for dropping out of the study included personal and family reasons, lack of interest, transportation issues, intensity of the exercise intervention, unhappiness with being assigned to the control group, and pregnancy. Reasons given by investigators for dropping participants from the study included lack of compliance to the intervention assigned, not completing follow-up testing as well as participants dieting or taking metformin during the study.

Exercise Program Characteristics

Exercise program characteristics are shown in Table 1 and Table 3. Twenty-seven studies included one or more aerobic exercise groups,^{20,21,23,24,26-28,31,33,36-48,50-53,55} 4 included a strength training group,^24,34,38,39 while 12 included a combined aerobic and strength training group.^{19,22,24,25,29,30,32,47,49,54,56,57} Three studies included one or more aerobic interval training groups.26,46,47 Training modalities included walking, jogging, cycling, swimming, various games, jump rope, elliptical exercise, and strength training.^19-34,36-58 Based on American College of Sports Medicine cutpoints for the different methods of relative exercise intensity reported in those studies in which data were available,¹²² 19 included one or more moderate intensity exercise groups,^{19,24,29-32,37-40,48-51,54-58} 12 included one or more vigorous intensity exercise groups,^{21-23,33,38,39,41,44-46,52,53} while 3 included one or more maximal or near-maximal, that is, interval exercise groups.^26,46,47 With the exception of one study in which exercise was unsupervised,³⁴ all others were supervised.^19-33,36-58 For those studies that included a strength training group and in which data were provided, the number of sets ranged from 1 to 4,^{19,24,29,30,34,38,39,47,57} the number of repetitions per exercise from 5 to 25,^{19,22,24,29,30,34,38,39,47,56,57} and the number of exercises from 3 to 13.^{19,22,30,38,39,47,56} The reported rest period between exercises ranged from 30 to 180 seconds.^{24,30,39,47,57} Resistance for the strength training interventions included free weights, machine weights, elastic bands, medicine balls, and one’s own body weight (pushups, etc.). Three studies reported using a circuit training protocol.^54,57,58

Table 3.

Training Program Characteristics^a.

Variable	#S/G/P^b	X– ± SD	Mdn	Range^c
Length (weeks)	39/47/847	11.1 ± 3.3	12	6-24
Frequency (days/week)	38/46/839	3.2 ± 1.0	3	1-5
Duration (min/session)	36/45/820	42.8 ± 20.2	40	3-90^d
Compliance (%)^e	8/10/173	94.4 ± 6.1	97	83-100
Minutes per week^f	36/43/805	134.1 ± 74.3	130	8-360^g
Minutes per week (adj)^h	8/10/173	171.8 ± 70.7	163	103-360

Descriptive data based on number of groups.

#S/G/P, number of studies, groups, and participants in which data were available for.

Bewteen-group ranges.

Duration as low as 3 minutes because interval training performed.

Compliance (%), percentage of exercise sessions attended.

Minutes per week, calculated as the product of frequency * duration.

Minimum of 8 minutes because interval training-type activity performed.

Minutes per week, adjusted for compliance.

Risk of Bias Assessment

Overall risk of bias based on the Cochrane Risk of bias assessment instrument (v 2.0) is shown in Figure 2, while study-level results are shown in Supplementary File 2. As can be seen, the overall risk of bias for most studies (76.9%) was high.

Figure 2.

Risk of bias results (pooled).

Data Synthesis

Changes in Primary Outcomes

Adiposity outcomes

As can be seen in Table 4 and Supplementary File 3, significant reductions in BMI in kg/m² were observed for aerobic exercise (P < .001) but not combined aerobic and strength exercise (P = .22). Compared to a mean baseline value of 29.2 kg/m², changes were equivalent to reductions of 6.5% in the aerobic exercise groups. Significant heterogeneity and a large amount of inconsistency were observed (Supplementary File 3). In addition, major asymmetry suggestive of small-study effects was found (Supplementary File 4). The 95% PI was −4.1 to 0.3 kg/m² for aerobic exercise and −2.1 to 1.3 kg/m² for combined aerobic and strength training. Changes in aerobic exercise groups remained statistically significant across all deletions, ranging from −1.9 to −1.7 kg/m². When 3 outlier results from 2 studies were deleted from the aerobic group,^26,33 results remained statistically significant ( $\bar{X}$ , 95% CI, −1.9, −2.4 to −1.4 kg/m²). Cumulative meta-analysis, ranked by year, revealed that results for all aerobic exercise groups have been significant since the year 2010, although the magnitude of reduction has decreased by 12.9% since that time (Supplementary File 5).

Table 4.

Changes in Primary and Secondary Outcomes.

	Aerobic			Strength			Aerobic + Strength
Outcomes	S/G/P (#)	$\bar{X}$ (95% CI)	I² (%)	S/G/P (#)	$\bar{X}$ (95% CI)	I² (%)	S/G/P (#)	$\bar{X}$ (95% CI)	I² (%)
Primary
BMI (kg/m²)	24/26/891	−1.9 (−2.7, −1.1)*	87	ID	ID	ID	11/11/370	−0.4 (−0.9, 0.2)	72
BMI z-score	ID	ID	ID	ID	ID	ID	ID	ID	ID
BMI percentile	ID	ID	ID	ID	ID	ID	ID	ID	ID
FM (kg)	10/10/305	−3.5 (−5.3, −1.7)*	85	ID	ID	ID	7/7/212	−2.6 (−5.2, 0.1)	94
BF (%)	20/23/778	−2.6 (−3.5, −1.8)*	81	ID	ID	ID	11/11/378	−2.0 (−3.3, −0.7)*	83
WC (cm)	14/16/456	−3.2 (−4.7, −1.6)*	77	ID	ID	ID	5/5/184	−3.6 (−5.1, −2.0)*	74
SBP (mm Hg)	11/12/400	−7.5 (−10.6, −4.4)*	88	ID	ID	ID	6/6/235	−5.9 (−10.0, −1.7)*	87
DBP (mm Hg)	10/11/333	−4.1 (−6.9, −1.2)*	94	ID	ID	ID	6/6/235	−2.9 (−5.3, −0.4)*	60
TC (mg/dL)^a	7/9/216	−7.9 (−16.0, 0.2)	81	ID	ID	ID	ID	ID	ID
HDL-C (mg/dL)^a	10/12/355	5.3 (2.0, 8.7)*	94	ID	ID	ID	ID	ID	ID
LDL-C (mg/dL)^a	9/11/323	−17.5 (−27.9, −7.1)*	93	ID	ID	ID	ID	ID	ID
TG (mg/dL)^b	9/11/315	−13.2 (−24.1, −2.4)*	93	ID	ID	ID	ID	ID	ID
FG (mg/dL)^c	10/11/297	−5.1 (−10.8, 0.5)	93	ID	ID	ID	8/8/312	−5.9 (−11.4, −0.4)*	91
FI (µU/mL)	8/10/304	−4.4 (−5.9, −2.9)*	84	ID	ID	ID	5/5/190	−6.2 (−15.2, 2.9)	96
HbA1c (%)	ID	ID	ID	ID	ID	ID	ID	ID	ID
Secondary
BW (kg)	23/26/735	−2.6 (−3.7, −1.5)*	57	ID	ID	ID	12/12/406	−0.8 (−2.3, 0.8)	78
FFM (kg)	11/12/354	0.5 (−0.3, 1.3)	46	ID	ID	ID	8/8/268	0.4 (−1.8, 2.5)	91
EI (kcal)	ID	ID	ID	ID	ID	ID	ID	ID	ID
Physical activity levels (g)	ID	ID	ID	ID	ID	ID	ID	ID	ID
VO_2max (mL·kg·min⁻¹)	12/15/3334	2.7 (0.9, 4.4)*	95	ID	ID	ID	7/7/255	3.4 (1.4, 5.4)*	86
Muscular strength (g)	ID	ID	ID	ID	ID	ID	ID	ID	ID

Abbreviations: #S/G/P, number of Studies, Groups, and Participants, with analysis based on number of groups; $\bar{X}$ (95% CI), mean, 95% confidence interval; I², I-squared; ID, insufficient data for pooling (<5 groups, i.e., effect sizes); BMI, body mass index; FM, fat mass; BF, body fat; WC, waist circumference; SBP, resting systolic blood pressure; DBP, resting diastolic blood pressure; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; TG, triglycerides; FG, fasting glucose; FI, fasting insulin; FFM, fat-free mass; EI, energy intake; (g), Hedge’s standardized effect size, adjusted for small-sample bias; VO_2max, maximum oxygen consumption.

To convert TC, HDL-C, and LDL-C to millimoles per liter, divide by 38.67.

To convert TG to millimoles per liter, divide by 88.57.

To convert fasting glucose to millimoles per liter, divide by 18.02.

Statistically significant (P < .05, and nonoverlapping 95% CI, boldfaced).

For FM (kg), statistically significant reductions were observed for the aerobic exercise groups (P < .001) with a trend for significant reductions in the combined aerobic and strength training groups (P = .06, Table 4, Supplementary File 6). Compared to a mean aerobic group baseline value of 30.4 ± 9.4 kg, changes were equivalent to reductions of approximately 11.5%. Significant heterogeneity and a large amount of inconsistency were observed (Supplementary File 6). No asymmetry was observed (Supplementary File 7). The 95% PI was −8.9 to 2.0 kg for aerobic exercise and −11.5 to 6.4 kg for combined aerobic and strength training. With each result deleted from the model once, changes remained statistically significant across all deletions, ranging from −4.0 to −2.4 kg, including when one outlier was deleted.⁴⁰ Cumulative meta-analysis, ranked by year, revealed that results for all aerobic exercise groups have been significant since inception of the first study in 1999, with more prominent reductions since 2014 (Supplementary File 8).

When results for BF% were pooled, statistically significant reductions were observed for both the aerobic (P < .001) and combined aerobic and strength training (P = .004) groups (Table 4, Supplementary File 9). Compared to mean baseline values of 35.2 ± 6.6% for aerobic exercise and 38.0 ± 5.3% for combined aerobic and strength training, relative reductions were equivalent to approximately 7.5% and 5.2%, respectively. The 95% PI for aerobic and combined aerobic and strength training was −5.4% to 0.1% and −6.1% to 2.2%, respectively. Significant heterogeneity and a large amount of inconsistency were observed for both interventions (Supplementary File 9). No asymmetry was observed for either aerobic (Supplementary File 10) or combined aerobic and strength training (Supplementary File 11). With each result deleted from the model once, changes remained statistically significant across all deletions for both aerobic (range, −2.8% to −2.5%) and combined aerobic and strength training (range, −2.3% to −1.7%), including when one outlier was deleted for the latter.²⁴ Results also remained significant when 2 outlier results were deleted for aerobic exercise groups ( $\bar{X}$ , 95% CI, −2.5%, −3.0% to −1.9%.^23,40 Cumulative meta-analysis, ranked by year, revealed that results for aerobic exercise groups have been significant since publication of the first included study in 1999, with results stable since approximately 2016 (Supplementary File 12). For combined aerobic and strength exercise, results have been consistently significant since the year 2016 (Supplementary File 13).

Statistically significant reductions in WC were observed for both aerobic and combined aerobic and strength training (P < .001 for both, Table 4, Supplementary File 14). Compared to mean baseline values of 92.3 ± 7.5 cm for aerobic exercise and 90.6 ±4.8 cm for combined aerobic and strength training, relative reductions were equivalent to approximately 3.4% and 3.9%, respectively. Significant heterogeneity was observed for both aerobic (P < .001) and combined aerobic and strength training (P = .004) while a large and moderate amount of inconsistency was observed for aerobic and combined aerobic and strength training, respectively (Supplementary File 14). No asymmetry was observed for aerobic exercise (Supplementary File 15) while minor asymmetry was observed for combined aerobic and strength training (Supplementary File 16). The 95% PI was −7.0 to 0.6 cm for aerobic exercise and −8.6 to 1.5 cm for combined aerobic and strength training. With each result deleted from the model once, changes remained statistically significant across all deletions for both aerobic (range, −3.6. to −2.8 cm) and combined aerobic and strength training (range, −3.9 to −3.2 cm), including when one outlier was deleted for the aerobic exercise groups.²¹ Cumulative meta-analysis, ranked by year, revealed that WC reductions for aerobic exercise have been significant since the year 2010 (Supplementary File 17) and since 2017 for combined aerobic and strength training (Supplementary File 18).

BP outcomes

When results for changes in resting SBP were pooled, statistically significant reductions were observed for both aerobic (P < .001) and combined aerobic and strength training groups (P = .005, Table 4, Supplementary File 19). Compared to mean baseline values of 126.0 ± 6.3 mm Hg for aerobic exercise and 118.1 ± 9.9 mm Hg for combined aerobic and strength training, relative reductions were equivalent to approximately 5.9% and 5.0%, respectively. Significant heterogeneity and a large amount of inconsistency were observed for both interventions (Supplementary File 19). Minor asymmetry suggestive of small-study effects was observed for aerobic exercise (Supplementary File 20) while no asymmetry was observed for combined aerobic and strength training (Supplementary File 21). The 95% PI for aerobic and combined aerobic and strength training was −15.5 to 0.5 mm Hg and −18.7 to 7.0 mm Hg, respectively. With each result deleted from the model once, changes in resting SBP remained statistically significant across all deletions for both aerobic (range, −8.0 to −6.9 mm Hg) and combined aerobic and strength training (range, −6.9 to −4.2 mm Hg), including when 2 outliers were deleted for aerobic exercise groups^36,55 and one for combined aerobic and strength exercise.³² Cumulative meta-analysis, ranked by year, revealed that results for aerobic exercise groups have been significant since the year 2012 (Supplementary File 22) and since 2017 for combined aerobic and strength exercise (Supplementary File 23).

For resting DBP, statistically significant reductions were observed for both aerobic (P = .01) and combined aerobic and strength training (P = .02) groups (Table 4, Supplementary File 24). Compared to mean baseline values of 75.5 ± 6.9 mm Hg for aerobic exercise and 70.4 ± 8.8 mm Hg for combined aerobic and strength training, relative reductions were equivalent to approximately 5.4% and 4.1%, respectively. Significant heterogeneity was observed for both aerobic and combined aerobic and strength exercise while a large and moderate amount of inconsistency was observed for aerobic and combined aerobic and strength exercise, respectively (Supplementary File 24). Minor asymmetry suggestive of small-study effects was observed for aerobic exercise (Supplementary File 25) while major asymmetry was observed for combined aerobic and strength training (Supplementary File 26). The 95% PI was −12.0 to 3.8 mm Hg for aerobic exercise and −9.2 to 3.5 mm Hg for combined aerobic and strength training. With each result deleted from the model once, changes in resting DBP remained statistically significant across all deletions for both aerobic range, -3.0 to – 4.9 mm Hg) and combined aerobic and strength training (range, −3.1 to −1.6 mm Hg). Results also remained significant when 2 outliers for aerobic exercise were deleted from the model ( $\bar{X}$ , 95% CI, −3.2, −4.6 to −1.8 mm Hg).^36,55 Cumulative meta-analysis, ranked by year, revealed that results for aerobic exercise groups have been significant since the year 2014 but with a trend for smaller reductions since that time (Supplementary File 27), while reductions for combined aerobic and strength exercise have been significant since 2012, with the same general trend (Supplementary File 28).

Lipid and lipoprotein outcomes

As can be seen in Table 4 and Supplementary File 29, reductions in TC were in the direction of benefit but not statistically significant for aerobic exercise interventions (P = .05 with overlapping 95% CIs). Compared to a mean baseline value of 155.6 mg/dL, overall changes were equivalent to reductions of 5.1%. Statistically significant heterogeneity and a large amount of inconsistency were observed (Supplementary File 29). Minor asymmetry suggestive of small-study effects was found (Supplementary File 30). The 95% PI was −31.1 to 15.2 mg/dL. With each study deleted from the model once, changes remained nonsignificant except when 2 studies were deleted separately.^36,55 Results also remained nonsignificant when the one outlier was deleted.⁵³ Cumulative meta-analysis, ranked by year, revealed that results for aerobic exercise have remained nonsignificant since inception of the first study in 2004 but with a trend toward significance in more recent years (Supplementary File 31).

Limited to aerobic exercise, statistically significant increases in HDL-C were observed (P = .002, Table 4 and Supplementary File 32). Compared to a mean baseline value of 43.3 ± 5.7 mg/dL, overall changes were equivalent to relative increases of 12.2%. Statistically significant heterogeneity and a large amount of inconsistency were observed (Supplementary File 32). Minor asymmetry suggestive of small-study effects was found (Supplementary File 33). The 95% PI was −4.5 to 15.1 mg/dL. With each study deleted from the model once, changes remained significant, ranging from 4.3 to 6.0 mg/dL. When 2 outliers were deleted from the model simultaneously,^26,48 results remained significant ( $\bar{X}$ , 95% CI, 4.9, 1.7 to 8.2 mg/dL). Cumulative meta-analysis, ranked by year, revealed that results for aerobic exercise and increases in HDL-C in mg/dL have remained significant since the year 2010, with a general trend toward smaller increases since that time (Supplementary File 34).

For LDL-C, statistically significant reductions (P = .001) limited to aerobic exercise groups were found (Table 4 and Supplementary file 35). Compared to a mean baseline value of 100.0 ± 9.2 mg/dL, overall changes were equivalent to relative decreases of 12.2%. Statistically significant heterogeneity and a large amount of inconsistency were observed (Supplementary File 35). Major asymmetry suggestive of small-study effects was also observed (Supplementary File 36). The 95% PI was −46.0 to 11.0 mg/dL. With each study deleted from the model once, changes remained significant, ranging from −19.9 to −10.3 mg/dL. When 2 outliers were deleted from the model simultaneously,^36,55 results remained significant ( $\bar{X}$ , 95% CI, −19.2, −28.3 to −10.1 mg/dL). Cumulative meta-analysis, ranked by year, revealed that decreases in LDL-C in mg/dL for aerobic exercise have remained significant since the year 2011, with a general trend toward smaller overall decreases since that time (Supplementary File 37).

Statistically significant reductions in TG (P = .02) were found for aerobic exercise groups (Table 4 and Supplementary file 38). Compared to a mean baseline value of 124.1 ± 18.1 mg/dL, relative decreases of 10.7% were observed. Statistically significant heterogeneity and a large amount of inconsistency were present (Supplementary File 38). Major asymmetry suggestive of small-study effects was also found (Supplementary File 39). The 95% PI was −42.6 to 16.1 mg/dL. With each study deleted from the model once, changes remained significant, ranging from −16.7 to −9.5 mg/dL, except when the study by Ben Ounis et al.²¹ was deleted from the model ( $\bar{X}$ , 95% CI, −9.5, −21.3 to 2.2 mg/dL). No outliers were detected. Cumulative meta-analysis, ranked by year, revealed that decreases in TG associated with aerobic exercise have remained significant the year 2017 (Supplementary File 40).

Diabetes-related outcomes

As can be seen in Table 4 and Supplementary File 41, overall results for FG were in the direction of benefit but not statistically significant for aerobic exercise (P = .07) while results were statistically significant (P = .04) for combined aerobic and strength training groups. Compared to mean baseline values of 88.7 ± 4.9 mg/dL for aerobic exercise and 89.2 ± 8.3 mg/dL for combined aerobic and strength training, relative reductions were equivalent to approximately 5.8% and 6.6%, respectively. Significant heterogeneity and a large amount of inconsistency were observed for both aerobic and combined aerobic and strength training groups (Supplementary File 41). Major asymmetry suggestive of small-study effects was observed for both aerobic (Supplementary File 42) and combined aerobic and strength training (Supplementary File 43). The 95% PI was −17.0 to 6.7 mg/dL for aerobic exercise and −23.2 to 11.4 mg/dL for combined aerobic and strength training. With each result deleted from the model once, changes in FG remained nonsignificant for the aerobic exercise groups except when one outlier study was deleted from the model³⁶ ( $\bar{X}$ , 95% CI, −5.6, −4.6 to −0.9 mg/dL). When 3 outlier results were deleted for combined aerobic and strength training,^22,29,57 reductions remained statistically significant ( $\bar{X}$ , 95% CI, −6.4, −10.3 to −2.4 mg/dL). Cumulative meta-analysis, ranked by year, revealed that results have remained nonsignificant for aerobic exercise groups since the year 2007 but with a trend for greater reductions in more recent years (Supplementary File 44). For combined aerobic and strength exercise, there has also been a trend for greater reductions in FG in more recent years, with results only statistically significant since the year 2018 (Supplementary File 45).

Statistically significant reductions in FI were observed for aerobic (P < .001) but not combined aerobic and strength (P = .18) groups (Table 4 and Supplementary File 46). Compared to mean baseline values of 23.0 ± 9.2 µU/mL for aerobic exercise and 20.4 ± 7.2 µU/mL for combined aerobic and strength training, relative reductions were equivalent to approximately 19.2% and 30.2%, respectively. Significant heterogeneity and a large amount of inconsistency were observed for both aerobic and combined aerobic and strength training groups (Supplementary File 46). Major asymmetry suggestive of small-study effects was observed for both aerobic (Supplementary File 47) and combined aerobic and strength exercise (Supplementary File 48). The 95% PI for aerobic and combined aerobic and strength training was −8.4 to −0.4 µU/mL and −29.6 to 17.3 µU/mL, respectively. With each result deleted from the model once, changes remained significant for aerobic exercise across all deletions (range, −4.7 to −3.8 µU/mL), including when one outlier study was deleted from the model ( $\bar{X}$ , 95% CI, −4.7, −6.0 to −3.4 µU/mL).⁴³ For combined aerobic and strength training, results remained nonsignificant across all deletions, including when one outlier was deleted ( $\bar{X}$ , 95% CI, −6.6, −16.2 to 3.1 µU/mL).²⁹ Cumulative meta-analysis, ranked by year, revealed that results have been statistically significant for aerobic exercise groups since the year 2007, with a trend for greater reductions in more recent years (Supplementary File 49). For combined aerobic and strength exercise, results were statistically significant in the year 2017 but nonsignificant before and after that time (Supplementary File 50).

For primary outcomes, insufficient data were available (N < 5) to assess BMI z-score and BMI percentile in the aerobic exercise groups, all outcomes in the strength training groups, and BMI z-score, BMI percentile, TC, HDL-C, LDL-C, and TG in the combined aerobic and strength training groups (Table 4).

Changes in Secondary Outcomes

BW

When results for BW were pooled, statistically significant reductions were observed for aerobic (P < .001) but not combined aerobic and strength training (P = .34) groups (Table 4 and Supplementary File 51). Compared to mean baseline values of 74.5 ± 14.5 kg for aerobic exercise and 75.7 ± 12.2 kg for combined aerobic and strength training, relative reductions were equivalent to approximately 3.5% and 1.0%, respectively. Significant heterogeneity and a moderate amount of inconsistency were observed for aerobic exercise while statistically significant and a large amount of inconsistency was observed for combined aerobic and strength training (Supplementary File 51). No asymmetry suggestive of small-study effects was observed for aerobic exercise (Supplementary File 52) while major asymmetry was observed for combined aerobic and strength exercise (Supplementary File 53). The 95% PI was −5.9 to 0.6 kg for aerobic exercise and −5.4 to 3.9 kg for combined aerobic and strength training. With each result deleted from the model once, changes in aerobic exercise groups remained statistically significant across all deletions (range, −3.4 to −2.5 kg), including when one outlier was deleted from the model ( $\bar{X}$ , 95% CI, −2.5, −3.5 to −1.6 kg).²⁴ For combined aerobic and strength training, results remained nonsignificant across all deletions (range, −1.1% to −0.6%), including when 3 outliers were deleted simultaneously ( $\bar{X}$ , 95% CI, −0.3, −1.4 to 0.7 kg).^22,24,57 Cumulative meta-analysis, ranked by year, revealed that results for aerobic exercise groups have been significant since the year 2012 (Supplementary File 54), while results for combined aerobic and strength exercise have been nonsignificant since inclusion of the first study in 2004 (Supplementary File 55).

FFM

No statistically significant differences were found for FFM for either the aerobic (P = .20) or combined aerobic and strength training (P = .74) groups (Table 4 and Supplementary File 56). Compared to mean baseline values of 47.1 ± 9.2 kg for aerobic exercise and 46.7 ± 10.6 kg for combined aerobic and strength training, relative increases were approximately 1.1% and 0.8%, respectively. Significant heterogeneity and low inconsistency were observed for aerobic exercise while statistically significant and a large amount of inconsistency was observed for combined aerobic and strength training (Supplementary File 56). Minor asymmetry suggestive of small-study effects was observed for aerobic exercise (Supplementary File 57), while major asymmetry was found for combined aerobic and strength training (Supplementary File 58). The 95% PI for aerobic and combined aerobic and strength training was −1.4 to 2.4 kg and −4.7 to 5.4 kg, respectively. With each result deleted from the model once, changes in aerobic exercise groups remained nonsignificant across all deletions except when the study of Zehasz et al.⁵⁸ was deleted from the model ( $\bar{X}$ , 95% CI, 0.7, 0.1 to 1.4 kg). For combined aerobic and strength training, results remained nonsignificant across all deletions, including when one outlier study was deleted ( $\bar{X}$ , 95% CI, 0.1, −0.8 to 1.1 kg).²⁵ Cumulative meta-analysis, ranked by year, revealed that results for aerobic exercise groups have been nonsignificant since the year 2012 (Supplementary File 59), while results for combined aerobic and strength training have been nonsignificant since inclusion of the first study in 2004 (Supplementary File 60).

Maximum oxygen consumption

Statistically significant increases in VO_2max in mL·kg·min⁻¹ were found for both aerobic (P = .002) and combined aerobic and strength (P = .001) exercise (Table 4 and Supplementary File 61). Compared to mean baseline values of 30.4 ± 5.1 mL·kg·min⁻¹ for aerobic exercise and 31.0 ± 4.5 mL·kg·min⁻¹ for combined aerobic and strength training, relative increases were equivalent to approximately 8.9% and 11.0%, respectively. Significant heterogeneity and a large amount of inconsistency were observed for both aerobic and combined aerobic and strength training groups (Supplementary File 61). Major asymmetry suggestive of small-study effects was observed for aerobic exercise (Supplementary File 62) while no asymmetry was observed for combined aerobic and strength training (Supplementary File 63). The 95% PI was −2.3 to 7.7 mL·kg·min⁻¹ for aerobic exercise and −1.6 to 8.4 mL·kg·min⁻¹ for combined aerobic and strength training. With each result deleted from the model once, increases in aerobic exercise groups remained statistically significant across all deletions (range, 2.2 to 3.0 mL·kg·min⁻¹), including when 2 outliers^38,48 were deleted from the model ( $\bar{X}$ , 95% CI, 2.0, 1.0 to 4.4 mL·kg·min⁻¹). For combined aerobic and strength training, increases also remained significant across all deletions (range, 2.9 to 4.1 mL·kg·min⁻¹), including when one outlier²⁵ was deleted ( $\bar{X}$ , 95% CI, 2.9, 1.9 to 3.9 mL·kg·min⁻¹). Cumulative meta-analysis, ranked by year, revealed that results for aerobic exercise groups have remained significant since the year 2013 (Supplementary File 64) while results for combined aerobic and strength exercise have been significant since 2012 (Supplementary File 65).

For secondary outcomes, insufficient data were available (N < 5) for analyzing EI, physical activity levels, and muscular strength in the aerobic exercise groups, all secondary outcomes in the strength training groups, and EI, physical activity levels, and muscular strength in the combined aerobic and strength training groups (Table 4).

Strength of Evidence (GRADE)

Strength of evidence for primary and secondary outcomes partitioned by intervention type (aerobic or combined aerobic and strength training) and outcome category (body composition, blood pressure, lipids and lipoproteins, markers of diabetes, aerobic fitness) is shown in Supplementary Files 66 to 74. For those outcomes that could be pooled (N ≥ 5), the strength of evidence for body composition outcomes (BMI, FM, BF%, WC, BW, FFM) ranged from “very low” to “moderate” for both aerobic (Supplementary File 66) and combined aerobic and strength training (Supplementary File 67). For resting SBP and DBP, the strength of evidence was either “very low” or “low” for aerobic exercise (Supplementary File 68) and “low” for combined aerobic and strength training groups (Supplementary File 69). For lipid and lipoprotein outcomes (TC, HDL-C, LDL-C, TG), the strength of evidence was considered either “very low” or “low” for aerobic exercise groups (Supplementary File 70). The strength of evidence for markers of diabetes (FG and FI) were all considered “very low” for both aerobic (Supplementary File 71) and combined aerobic and strength training groups (Supplementary file 72). For aerobic fitness (VO_2max in mL·kg·min⁻¹), the strength of evidence was limited to “very low” and “low,” respectively, for aerobic (Supplementary File 73) and combined aerobic and strength training (Supplementary File 74).

Discussion

Overall Findings

With a focus on point estimates and 95% CI, the overall findings of the current systematic review with meta-analysis suggest that either aerobic or combined aerobic and strength exercise may improve selected CVD risk factors in children and adolescents with obesity. For aerobic exercise, this includes statistically significant improvements in body composition (BMI in kg/m², FM, BF%, WC, BW), resting SBP and DBP, lipids and lipoproteins (HDL-C, LDL-C, TG), FI, and aerobic fitness (VO_2max in mL·kg·min⁻¹). While in the direction of benefit, no statistically significant improvements were observed for TC, FG, or FFM. The observed findings were reinforced by the lack of change in results for BMI in kg/m², FM, BF%, WC, BW, SBP, DBP, HDL-C, LDL-C, FI, and VO_2max in mL·kg·min⁻¹, but not TG, when influence and outlier analysis was conducted as well as the accumulation of results over time. However, for TC, FFM, and FG, results became statistically significant when either influence or outlier analysis was conducted. Assuming that an improvement of 5% from baseline values is clinically relevant for all outcomes and not just adiposity results,¹²³ the current findings suggest clinically important improvements for all statistically significant outcomes except WC, BW, and FFM.

For combined aerobic and strength training, statistically significant improvements were observed for selected body composition outcomes (BF%, WC), resting SBP and DBP, FG, and aerobic fitness (VO_2max in mL·kg·min⁻¹). While in the direction of benefit, no significant differences were observed for BMI in kg/m², FM, BW, FFM, or FI. These findings were reinforced by the lack of change in statistically significant results when influence and outlier analyses were conducted as well as the accumulation of findings over time. Assuming again that an improvement of 5% from baseline values is clinically relevant, the current findings suggest clinically relevant improvements for all statistically significant outcomes except WC and SBP.

While the current findings are encouraging, they should be considered with respect to (1) the overall high risk of bias for more than 75% of the studies, (2) the uncertainty in the evidence, as assessed by GRADE, and (3) the overlapping 95% PI for all outcomes except FI in the aerobic exercise groups. The generally large amount of heterogeneity and inconsistency observed, both of which are reflected in risk of bias and GRADE assessments, may be thought to derive from interindividual differences in response to exercise. However, recent research by the investigative team in which true interindividual response differences were appropriately quantified suggests that this may not be the case.^124,125 However, this previous work included children with either overweight or obesity versus obesity only, and were limited to BMI in kg/m², FM, and BF%.^124,125

Implications for Research

Reporting and Conduct of Future Research

The reporting of future RCTs addressing the effects of exercise in children and adolescents with obesity could be improved. For example, appropriately reporting the elements contained in the Cochrane Risk of Bias assessment instrument for RCTs would be helpful.¹⁰⁰ Along those lines, only 7 studies (17.9%)^{22,27,29,38,39,41,53} included a Consolidated Statement of Reporting Trials (CONSORT) figure that depicted the flow of participants through the trial,¹²⁶ while less than half (33.3%) reported power estimates for their primary outcome(s) of interest.^{22,29,34,38,41,44,45,49-51,53,55,57} Second, better reporting of the elements of RCT exercise interventions is suggested. For example, only 8 of 39 studies (20.5%) reported adequate data on compliance of children and adolescents to the exercise interventions.^{19,21,29,31,38,39,45,56} Adherence to the Consensus on Exercise Reporting Template (CERT) could go a long way in addressing these elements when planning future RCTs.¹²⁷ Third, it is suggested that future studies report the EI and energy expenditure of participants as well as the total physical activity levels of participants beyond the actual exercise intervention itself. The reporting of EI and expenditure of participants is important given the need to account for the EI-energy expenditure spectrum and any potential effect on the outcomes examined. Accounting for the total physical activity levels of both intervention and control participants is important given the potential for exercise compensation. For example, when compared to a control group, Hagstromer et al. found that adolescents with obesity who were assigned to the exercise group had a greater reduction in their total daily physical activity.³²

Future Research Directions

There are several suggestions for future RCTs of exercise intervention studies in children and adolescents with obesity. First, based on the investigative team’s inclusion criteria, there was insufficient data (N < 5) to examine any of the outcomes for strength training interventions. Given these findings, it appears that there is a need for additional well-designed RCTs focused on strength training in children and adolescents with obesity. Second, given the lack of data for all intervention types with respect to BMI z-score, BMI percentile, HbA1c, EI, and total physical activity, future RCTs should consider including these outcomes in their studies given their potential impact on CVD risk factors in children and adolescents with obesity. Third, given a lack of data, future studies that focus on combined aerobic and strength training interventions should also consider the assessment of lipids and lipoproteins (TC, HDL-C, LDL-C, TG) in this population. Finally, given the increased interest in precision medicine,^128-130 including precision exercise¹³¹ and precision obesity,^132,133 future RCTs should appropriately quantify any true interindividual differences in response to exercise among children and adolescents with obesity before examining for potential moderators and mediators, methods of which have been reported elsewhere.^134,135 Briefly, this includes a focus on the change outcome SDs of both the intervention and control groups as the unit of analysis, including a comparison against a minimally important threshold. For example, in 2 recent interindividual response meta-analyses by the authors, little evidence was found to suggest that true interindividual response differences exist with respect to aerobic exercise on FM, BF%, and BMI in children and adolescents with overweight and obesity, once random variation was accounted for.^124,125 However, whether interindividual differences exist for other outcomes, for example, resting SBP and DBP, as well as when limited to children and adolescents with obesity, is not known.

Implications for Practice

While the results of the current meta-analysis suggest that aerobic and combined aerobic and strength training may improve multiple CVD risk factors, the strength of this evidence is generally low. The former notwithstanding, it would seem reasonable to suggest that clinicians consider the treatment approach to managing obesity in adolescents recently suggested by Cardel et al.,¹³⁶ much of which could also be applied to children with obesity. This approach includes (1) classifying weight status using BMI, (2) conducting a detailed history, physical examination, and laboratory assessment, (3) discussing the diagnosis with the child/adolescent and caregiver(s), and (4) recommending behavioral interventions that include, but are not limited to, dietary modifications and at least 60 minutes of physical activity and less than 2 hours of nonacademic sedentary activities per day.¹³⁶ If BMI targets are not reached through lifestyle changes in adolescents, long-term use of a Food and Drug Administration–approved drug such as orlistat in those ≥12 years of age may be given, and as a last option, possible surgical intervention. Specific to physical activity, of which exercise is nested in, it is suggested that clinicians and other health care professionals counsel children and adolescents with obesity to strive toward meeting or exceeding the recommendations set forth by the 2018 Physical Activity Guidelines for Americans.¹⁸ For preschool children 3 to 5 years of age, this includes being physically active throughout the day with adult caregivers encouraging active play.¹⁸ For children and adolescents 6 years of age and older, a minimum of 60 minutes of moderate to vigorous physical activity is recommended.¹⁸ The majority of the 60 minutes should consist of aerobic activity, for example, bicycling, of which 3 days per week should be vigorous intensity activity. Included within the 60 minutes or more should be muscle strengthening and bone strengthening activity at least 3 days per week.¹⁸ Inclusive in the authors’ recommendation is interval training types of activities.^137,138

Implications for Policy

From the authors’ perspective, the results of this systematic review with aggregate data meta-analysis have application to policy. Specifically, it is suggested that third-party payment for exercise programming be considered. For example, while the Early and Periodic Screening, Diagnostic and Treatment (EPSDT) benefit offers comprehensive and preventive health care services, including health education, for children <21 years of age who are enrolled in Medicaid, no benefit that specifically targets proper payment for exercise programming currently exists.¹³⁹ As a result, this probably has a deleterious impact on obtaining quality exercise programming in this vulnerable population. Consequently, this fails to meet the United States Preventative Services Task Force (USPSTF) recommendations for the treatment of children and adolescents with overweight and obesity.¹²³ Given the lack of training that clinicians and other health care personnel receive in exercise prescription,^140-142 this goal may best be addressed by having the clinician or other relevant health care personnel refer the child or adolescent with obesity to a certified exercise programming professional, the latter of whom is then appropriately compensated for her/his services. Unfortunately, in both the public and private sector, a lack of appropriate reimbursement continues to be a challenge for obesity and treatment services in children and adolescents with obesity.¹⁴³

Strengths and Limitations

From the authors’ perspective, the major strengths of the current systematic review with meta-analysis include (1) the focus on studies in children and adolescents with obesity, (2) the multiple CVD risk factors examined, (3) use of the most robust methods currently available for conducting a systematic review with aggregate data meta-analysis,^{76,100,104,105,118-121} including the calculation of 95% PI for estimating what result one might expect if they conducted their own RCT,¹¹⁰ (4) suggestions for researchers in the conduct and reporting of future RCTs on exercise and CVD risk factors in children and adolescents with obesity, and (5) suggestions for clinicians and other health care personnel, as well as policymakers, for making more informed decisions regarding exercise in the treatment of children and adolescents with obesity.

In addition to strengths, there are several potential limitations that should be mentioned. First, given the collective lack of data available, a post hoc decision was made to not conduct any type of meta-regression analysis. As a result, potential covariates that might have affected the overall results for each outcome within each type of intervention were not conducted. However, meta-regression within the context of an aggregate data meta-analysis does not support causal inferences because covariates are not randomly assigned in studies.¹⁴⁴ Therefore, any potential associations observed would have needed to be tested in original RCTs.¹⁴⁴ Second, given the lack of studies available, we were unable to examine the effects of strength training-only interventions when limited to children and adolescents with obesity. Third, while many statistical tests were conducted, no adjustments were made for such. As a result, some findings could have been nothing more than the play of chance. However, common to most aggregate data meta-analyses, no statistical adjustments were made for multiple testing because of concerns about missing possibly important findings that could be tested in original RCTs.¹¹⁶ Fourth, since this was an aggregate data meta-analysis, the potential for ecological fallacy, specifically Simpson’s Paradox, exists.¹⁴⁵ Consequently, the possibility exists that the findings observed at the group level in the current study would not be applicable at the individual level of the child or adolescent. While an individual participant data (IPD) meta-analysis would avoid this limitation, such an approach has to be considered with respect to (1) the inability to obtain IPD from original trial investigators,^146-151 (2) the loss of power and increased risk of bias with fewer eligible studies, (3) the substantially increased costs (up to 8 times greater) associated with conducting an IPD meta-analysis,^151,152 and (4) the general lack of differences in overall results observed between aggregate and IPD meta-analyses.^152-157 Finally, like any meta-analysis, the limitations of the included studies carry over to the current meta-analysis.

Conclusions

The results of the current systematic review with meta-analysis of RCTs suggest that exercise, particularly aerobic exercise, as well as combined aerobic and strength training, is associated with improvements in multiple CVD risk factors among children and adolescents with obesity. However, the generally low strength of evidence suggests a need for future well-designed and conducted RCTs on the effects of exercise, especially strength training, in children and adolescents.

CME/CE Article Quiz

American College of Lifestyle Medicine (ACLM) members can earn FREE CME/CE credit by reading this approved CME/CE article and successfully completing the online CME/CE activity. Non-members can earn CME/CE for $40 per article. Visitlifestylemedicine.org to join the ACLM.

Instructions.

AJLM CME/CE Articles and Quizzes are offered online only through the American College of Lifestyle Medicine and are accessible at lifestylemedicine.org/store. ACLM Members can enroll in the activity, complete the quiz, and earn this CME/CE for free. Non-members will be charged $40 per article.

A Passing score of 80% or higher is required in order to be awarded the CME/CE credit.

Supplemental Material

sj-pdf-1-ajl-10.1177_1559827620988839 – Supplemental material for Exercise and Cardiovascular Disease Risk Factors in Children and Adolescents With Obesity: A Systematic Review With Meta-Analysis of Randomized Controlled Trials

Supplemental material, sj-pdf-1-ajl-10.1177_1559827620988839 for Exercise and Cardiovascular Disease Risk Factors in Children and Adolescents With Obesity: A Systematic Review With Meta-Analysis of Randomized Controlled Trials by George A. Kelley, Kristi S. Kelley and Russell R. Pate in American Journal of Lifestyle Medicine

Supplemental Material

sj-pdf-2-ajl-10.1177_1559827620988839 – Supplemental material for Exercise and Cardiovascular Disease Risk Factors in Children and Adolescents With Obesity: A Systematic Review With Meta-Analysis of Randomized Controlled Trials

Supplemental material, sj-pdf-2-ajl-10.1177_1559827620988839 for Exercise and Cardiovascular Disease Risk Factors in Children and Adolescents With Obesity: A Systematic Review With Meta-Analysis of Randomized Controlled Trials by George A. Kelley, Kristi S. Kelley and Russell R. Pate in American Journal of Lifestyle Medicine

Supplemental Material

sj-pdf-3-ajl-10.1177_1559827620988839 – Supplemental material for Exercise and Cardiovascular Disease Risk Factors in Children and Adolescents With Obesity: A Systematic Review With Meta-Analysis of Randomized Controlled Trials

Supplemental material, sj-pdf-3-ajl-10.1177_1559827620988839 for Exercise and Cardiovascular Disease Risk Factors in Children and Adolescents With Obesity: A Systematic Review With Meta-Analysis of Randomized Controlled Trials by George A. Kelley, Kristi S. Kelley and Russell R. Pate in American Journal of Lifestyle Medicine

Supplemental Material

sj-pdf-4-ajl-10.1177_1559827620988839 – Supplemental material for Exercise and Cardiovascular Disease Risk Factors in Children and Adolescents With Obesity: A Systematic Review With Meta-Analysis of Randomized Controlled Trials

Supplemental material, sj-pdf-4-ajl-10.1177_1559827620988839 for Exercise and Cardiovascular Disease Risk Factors in Children and Adolescents With Obesity: A Systematic Review With Meta-Analysis of Randomized Controlled Trials by George A. Kelley, Kristi S. Kelley and Russell R. Pate in American Journal of Lifestyle Medicine

Footnotes

Authors’ Note

Data are available from the corresponding author on reasonable request.

Author Contributions

GAK was responsible for the conception and design, acquisition of data, analysis and interpretation of data, drafting the initial manuscript, and revising it critically for important intellectual content. KSK was responsible for the conception and design, acquisition of data, and reviewing all drafts of the manuscript. RRP was responsible for the conception and design, interpretation of data, and reviewing all drafts of the manuscript. All authors read and approved the final manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by the American Heart Association, Grant #17GRNT33630158 (GA Kelley, Principal Investigator). The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the American Heart Association.

Ethical Approval

Not applicable, because this article does not contain any studies with human or animal subjects.

Informed Consent

Not applicable, because this article does not contain any studies with human or animal subjects.

Trial Registration

Not applicable, because this article does not contain any clinical trials.

ORCID iDs

George A. Kelley

Kristi S. Kelley

Supplemental Material

Supplemental material for this article is available online.

References

Fleming

Robinson

, et al. Global, regional, and national prevalence of overweight and obesity in children and adults during 1980-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet. 2014;384:766-781.

NCD Risk Factor Collaboration. Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: a pooled analysis of 2416 population-based measurement studies in 128.9 million children, adolescents, and adults. Lancet. 2017;390:2627-2642.

Kyle

Dhurandhar

Allison

DB.

Regarding obesity as a disease: evolving policies and their implications. Endocrinol Metab Clin North Am. 2016;45:511-520.

World Health Organization. Obesity and overweight. Published April 1, 2020. Accessed January 5, 2021. http://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight

Fryar

Carroll

Ogden

. Prevalence of overweight, obesity, among children and adolescents: United States, 1963-1965 through 2011-2012. Accessed January 5, 2021. https://www.cdc.gov/nchs/data/hestat/obesity_child_15_16/obesity_child_15_16.htm

Skinner

Ravanbakht

Skelton

Perrin

Armstrong

SC.

Prevalence of obesity and severe obesity in US children, 1999-2016. Pediatrics. 2018;141:e20173459.

Jackson

Zhang

Wiltz

, et al. Hypertension among youths—United States, 2001-2016. MMWR Morb Mortal Wkly Rep. 2018;67:758-762.

Friedemann

Heneghan

Mahtani

Thompson

Perera

Ward

AM.

Cardiovascular disease risk in healthy children and its association with body mass index: systematic review and meta-analysis. BMJ. 2012;345:e4759.

Nguyen

Kit

Carroll

Abnormal cholesterol among children and adolescents in the United States, 2011-2014. NCHS Data Brief. 2015;(228):1-8.

10.

Reilly

Kelly

Long-term impact of overweight and obesity in childhood and adolescence on morbidity and premature mortality in adulthood: systematic review. Int J Obes (Lond). 2011;35:891-898.

11.

Chung

Skinner

Steiner

Perrin

EM.

Physical activity and BMI in a nationally representative sample of children and adolescents. Clin Pediatr (Phila). 2012;51:122-129.

12.

Jiménez-Pavón

Kelly

Reilly

. Associations between objectively measured habitual physical activity and adiposity in children and adolescents: systematic review. Int J Pediatr Obes. 2010;5:3-18.

13.

Umer

Kelley

Cottrell

Giacobbi

Jr Innes

Lilly

CL.

Childhood obesity and adult cardiovascular disease risk factors: a systematic review with meta-analysis. BMC Public Health. 2017;17:683.

14.

Henriksson

Tynelius

, et al. Cardiorespiratory fitness, muscular strength, and obesity in adolescence and later chronic disability due to cardiovascular disease: a cohort study of 1 million men. Eur Heart J. 2020;41:1503-1510.

15.

Chiesa

Charakida

Deanfield

JE.

Adolescent health and future cardiovascular disability: it’s never too early to think about prevention. Eur Heart J. 2020;41:1511-1513.

16.

Dietz

WH.

Health consequences of obesity in youth: childhood predictors of adult disease. Pediatrics. 1998;101(3 pt 2):518-525.

17.

Kelly

Barlow

Rao

, et al. Severe obesity in children and adolescents: identification, associated health risks, and treatment approaches: a scientific statement from the American Heart Association. Circulation. 2013;128:1689-1712.

18.

US Department of Health and Human Services. 2018 Physical Activity Guidelines Advisory Committee Scientific Report. US Department of Health and Human Services; 2018.

19.

Alberga

Farnesi

Lafleche

Legault

Komorowski

The effects of resistance exercise training on body composition and strength in obese prepubertal children. Phys Sportsmed. 2013;41:103-109.

20.

Andre

Beguier

Using motivational interviewing as a supplement to physical activity program in obese adolescents: a RCT study. Eat Weight Disord. 2015;20:519-523.

21.

Ounis

Elloumi

Makni

, et al. Exercise improves the ApoB/ApoA-I ratio, a marker of the metabolic syndrome in obese children. Acta Paediatr. 2010;99:1679-1685.

22.

Bharath

Choi

Cho

, et al. Combined resistance and aerobic exercise training reduces insulin resistance and central adiposity in adolescent girls who are obese: randomized clinical trial. Eur J Appl Physiol. 2018;118:1653-1660.

23.

Cao

Zhao

Effects of aerobic interval training on cardiac function and physical fitness in obese teenagers [in Chinese]. J Jilin Univ Med. 2012;38:586-589. http://onlinelibrary.wiley.com/o/cochrane/clcentral/articles/209/CN-00904209/frame.html

24.

Chen

Cao

Zhao

Effects of different exercise modes on body composition, inflammatory factors, and exercise capacity of obese teenagers [in Chinese]. J Jilin Univ Med. 2015;41:1070-1075. http://onlinelibrary.wiley.com/o/cochrane/clcentral/articles/798/CN-01177798/frame.html

25.

Cheng

Peng

Zhu

Qin

YS.

Effects of eight weeks exercise prescription intervention on aerobic capacity, body composition, blood lipid and C-reactive protein in obese adolescents [in Chinese]. J Jilin Univ Med. 2012;38:745-749.

26.

Chuensiri

Suksom

Tanaka

Effects of high-intensity intermittent training on vascular function in obese preadolescent boys. Child Obes. 2018;14:41-49.

27.

Djaafar

Hadisaputro

Widjanarko

, et al. The effects of physical fitness gymnastics (SKJ) 2012 towards body mass index, body fat percentage, and physical fitness in obese children. Int J Pharm Res. 2019;11:250-254.

28.

Elloumi

Makni

Ounis

, et al. Six-minute walking test and the assessment of cardiorespiratory responses during weight-loss programmes in obese children. Physiother Res Int. 2011;16:32-42.

29.

Farpour-Lambert

Aggoun

Marchand

Martin

Herrmann

Beghetti

Physical activity reduces systemic blood pressure and improves early markers of atherosclerosis in pre-pubertal obese children. J Am Coll Cardiol. 2009;54:2396-2406.

30.

Fazelifar

Ebrahim

Sarkisian

Effect of concurrent training and detraining on anti-inflammatory biomarker and physical fitness levels in obese children. Rev Bras Med Esporte. 2013;19:349-354.

31.

Gutin

Owens

Slavens

Riggs

Treiber

Effect of physical training on heart-period variability in obese children. J Pediatr. 1997;130:938-943.

32.

Hagstromer

Elmberg

Marild

Sjöström

Participation in organized weekly physical exercise in obese adolescents reduced daily physical activity. Acta Paediatr. 2009;98:352-354.

33.

Karacabey

The effect of exercise on leptin, insulin, cortisol and lipid profiles in obese children. J Int Med Res. 2009;37:1472-1478.

34.

Kelly

Loza

Lin

, et al. The effect of a home-based strength training program on type 2 diabetes risk in obese Latino boys. J Pediatr Endocrinol Metab. 2015;28:315-322.

35.

Kelly

Holmberg

Schroeder

, et al. Effect of home-based strength training program on IGF-I, IGFBP-1 and IGFBP-3 in obese Latino boys participating in a 16-week randomized controlled trial. J Pediatr Endocrinol Metab. 2019;32:1121-1129.

36.

Kim

, et al. Improved insulin sensitivity and adiponectin level after exercise training in obese Korean youth. Obesity (Silver Spring). 2007;15:3023-3030.

37.

Kim

Son

Headid

Pekas

Noble

Park

SY.

The effects of a 12-week jump rope exercise program on body composition, insulin sensitivity, and academic self-efficacy in obese adolescent girls. J Pediatr Endocrinol Metab. 2020;33:129-137.

38.

Lee

Bacha

Hannon

Kuk

Boesch

Arslanian

Effects of aerobic versus resistance exercise without caloric restriction on abdominal fat, intrahepatic lipid, and insulin sensitivity in obese adolescent boys: a randomized, controlled trial. Diabetes. 2012;61:2787-2795.

39.

Lee

Deldin

White

, et al. Aerobic exercise but not resistance exercise reduces intrahepatic lipid content and visceral fat and improves insulin sensitivity in obese adolescent girls: a randomized controlled trial. Am J Physiol Endocrinol Metab. 2013;305:E1222-E1229.

40.

G, X

Wang

Lyu

Zhao

LJ.

Influence of aerobic exercise in heart rate variability in obese adolescents [in Chinese]. J Jilin Univ Med. 2014;40:1093-1097.

41.

Liu

Alderman

Song

Chen

Hung

Chang

YK.

A randomized controlled trial of coordination exercise on cognitive function in obese adolescents. Psychol Sport Exerc. 2018;34:29-38.

42.

McNarry

Lambrick

Westrupp

Faulkner

The influence of a six-week, high-intensity games intervention on the pulmonary oxygen uptake kinetics in prepubertal obese and normal-weight children. Appl Physiol Nutr Metab. 2015;40:1012-1018.

43.

Meyer

Kundt

Lenschow

Schuff-Werner

Kienast

Improvement of early vascular changes and cardiovascular risk factors in obese children after a six-month exercise program. J Am Coll Cardiol. 2006;48:1865-1870.

44.

Moslehi

Khalvati

Playing in form of outdoor aerobic exercise is more effective than indoor treadmill exercise on serum Orexin-A and weight loss in obese adolescent boys. Obes Med. 2019;15:100104.

45.

Owens

Gutin

Allison

, et al. Effect of physical training on total and visceral fat in obese children. Med Sci Sports Exerc. 1999;31:143-148.

46.

Racil

Ounis

Hammouda

, et al. Effects of high vs moderate exercise intensity during interval training on lipids and adiponectin levels in obese young females. Eur J Appl Physiol. 2013;113:2531-2540.

47.

Racil

Zouhal

Elmontassar

, et al. Plyometric exercise combined with high-intensity interval training improves metabolic abnormalities in young obese females more so than interval training alone. Appl Physiol Nutr Metab. 2016;41:103-109.

48.

Saygin

Ozturk

MA.

The effect of twelve week aerobic exercise programme on health related physical fitness components and blood lipids in obese girls. Afr J Pharm Pharmacol. 2011;5:1441-1445.

49.

Son

Sung

Bharath

Choi

Park

SY.

Combined exercise training reduces blood pressure, arterial stiffness, and insulin resistance in obese prehypertensive adolescent girls. Clin Exp Hypertens. 2017;39:546-552.

50.

Song

Stebbins

Kim

Kang

Chai

JH.

Effects of 12 weeks of aerobic exercise on body composition and vascular compliance in obese boys. J Sports Med Phys Fitness. 2012;52:522-529.

51.

Sung

Pekas

Scott

Son

Park

SY.

The effects of a 12-week jump rope exercise program on abdominal adiposity, vasoactive substances, inflammation, and vascular function in adolescent girls with prehypertension. Eur J Appl Physiol. 2019;119:577-585.

52.

Tan

Yang

Wang

Physical training of 9- to 10-year-old children with obesity to lactate threshold intensity. Pediatr Exerc Sci. 2010;22:477-485.

53.

Vasconcellos

Seabra

Cunha

, et al. Health markers in obese adolescents improved by a 12-week recreational soccer program: a randomised controlled trial. J Sports Sci. 2016;34:564-575.

54.

Watts

Beye

Siafarikas

, et al. Exercise training normalizes vascular dysfunction and improves central adiposity in obese adolescents. J Am Coll Cardiol. 2004;43:1823-1827.

55.

Watts

Beye

Siafarikas

, et al. Effects of exercise training on vascular function in obese children. J Pediatr. 2004;144:620-625.

56.

Wong

PCH

Chia

MYH

Tsou

IYY

, et al. Effects of a 12-week exercise training programme on aerobic fitness, body composition, blood lipids and C-reactive protein in adolescents with obesity. Ann Acad Med Singap. 2008;37:286-293.

57.

Wong

Sanchez-Gonzalez

Son

Kwak

Park

SY.

The effects of a 12-week combined exercise training program on arterial stiffness, vasoactive substances, inflammatory markers, metabolic profile, and body composition in obese adolescent girls. Pediatr Exerc Sci. 2018;30:480-486.

58.

Zehsaz

Farhangi

Ghahramani

Exercise training lowers serum chemerin concentration in obese children. Sci Sports. 2017;32:39-45.

59.

Kelley

KS.

Effects of exercise in the treatment of overweight and obese children and adolescents: a systematic review of meta-analyses. J Obes. 2013;2013:783103.

60.

Kelley

Pate

Exercise and BMI z-score in overweight and obese children and adolescents: a systematic review and network meta-analysis of randomized trials. J Evid Based Med. 2017;10:108-128.

61.

Kelley

Pate

RR.

Effects of exercise on BMI z-score in overweight and obese children and adolescents: a systematic review with meta-analysis. BMC Pediatr. 2014;14:225.

62.

Kelley

Pate

RR.

Exercise and BMI in overweight and obese children and adolescents: a systematic review with trial sequential meta-analysis. Biomed Res Int. 2015;2015:704539.

63.

Kelley

Pate

RR.

Exercise and adiposity in overweight and obese children and adolescents: a systematic review with network meta-analysis of randomised trials. BMJ Open. 2019;9:e031220.

64.

Atlantis

Barnes

Singh

MAF

. Efficacy of exercise for treating overweight in children and adolescents: a systematic review. Int J Obes (Lond). 2006;30:1027-1040.

65.

Dias

Green

Ingul

Pavey

Coombes

JS.

Exercise and vascular function in child obesity: a meta-analysis. Pediatrics. 2015;136:e648-e659.

66.

Garcia-Hermoso

Gonzalez-Ruiz

Triana-Reina

Olloquequi

Ramirez-Velez

Effects of exercise on carotid arterial wall thickness in obese pediatric populations: a meta-analysis of randomized controlled trials. Child Obes. 2017;13:138-145.

67.

Garcia-Hermoso

Ceballos-Ceballos

RJM

Poblete-Aro

Hackney

Mota

Ramirez-Velez

Exercise, adipokines and pediatric obesity: a meta-analysis of randomized controlled trials. Int J Obes (Lond). 2017;41:475-482.

68.

Garcia-Hermoso

Cerrillo-Urbina

Herrera-Valenzuela

Cristi-Montero

Saavedra

Martinez-Vizcaino

Is high-intensity interval training more effective on improving cardiometabolic risk and aerobic capacity than other forms of exercise in overweight and obese youth? A meta-analysis. Obes Rev. 2016;17:531-540.

69.

Garcia-Hermoso

Sanchez-Lopez

Martinez-Vizcaino

Effects of aerobic plus resistance exercise on body composition related variables in pediatric obesity: a systematic review and meta-analysis of randomized controlled trials. Pediatr Exerc Sci. 2015;27:431-440.

70.

Han

Liu

Zhao

, et al. Does physical activity-based intervention improve systemic proinflammatory cytokine levels in overweight or obese children and adolescents? Insights from a meta-analysis of randomized control trials. Obes Facts. 2019;12:653-668.

71.

Garnett

Baur

, et al. Impact of dietary and exercise interventions on weight change and metabolic outcomes in obese children and adolescents: a systematic review and meta-analysis of randomized trials. JAMA Pediatr. 2013;167:759-768.

72.

Marson

Delevatti

Prado

Netto

Kruel

LF.

Effects of aerobic, resistance, and combined exercise training on insulin resistance markers in overweight or obese children and adolescents: a systematic review and meta-analysis. Prev Med. 2016;93:211-218.

73.

Stoner

Rowlands

Morrison

, et al. Efficacy of exercise intervention for weight loss in overweight and obese adolescents: meta-analysis and implications. Sports Med. 2016;46:1737-1751.

74.

Garcia-Hermoso

Saavedra

Escalante

Effects of exercise on resting blood pressure in obese children: a meta-analysis of randomized controlled trials. Obes Rev. 2013;14:919-928.

75.

Garcia-Hermoso

Saavedra

Escalante

Sanchez-Lopez

Martinez-Vizcaino

Aerobic exercise reduces insulin resistance markers in obese youth: a meta-analysis of randomized controlled trials. Eur J Endocrinol. 2014;171:R163-R171.

76.

Doi

Barendregt

Khan

Thalib

Williams

. Advances in the meta-analysis of heterogeneous clinical trials I: the inverse variance heterogeneity model. Contemp Clin Trials. 2015;45(pt A):130-138.

77.

Doi

Barendregt

Khan

Thalib

Williams

. Advances in the meta-analysis of heterogeneous clinical trials II: the quality effects model. Contemp Clin Trials. 2015;45(pt A):123-129.

78.

Liberati

Altman

Tetzlaff

, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. Ann Intern Med. 2009;151:W65-W94.

79.

Booth

Clarke

Dooley

, et al. The nuts and bolts of PROSPERO: an international prospective register of systematic reviews. Syst Rev. 2012;1:2.

80.

Page

Shamseer

Tricco

AC.

Registration of systematic reviews in PROSPERO: 30,000 records and counting. Syst Rev. 2018;7:32.

81.

Sideri

Papageorgiou

Eliades

Registration in the international prospective register of systematic reviews (PROSPERO) of systematic review protocols was associated with increased review quality. J Clin Epidemiol. 2018;100:103-110.

82.

Sacks

Chalmers

Smith

Randomized versus historical controls for clinical trials. Am J Med. 1982;72:233-240.

83.

Schulz

Chalmers

Hayes

Altman

DG.

Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials. JAMA. 1995;273:408-412.

84.

Jago

Jonker

Missaghian

Baranowski

Effect of 4 weeks of Pilates on the body composition of young girls. Prev Med. 2006;42:177-180.

85.

Kuczmarski

Ogden

Grummer-Strawn

, et al. CDC Growth Charts: United States Advance Data From Vital and Health Statistics, No. 314. National Center for Health Statistics; 2000.

86.

Kromeyer-Hauschild

Zellner

Trends in overweight and obesity and changes in the distribution of body mass index in schoolchildren of Jena, East Germany. Eur J Clin Nutr. 2007;61:404-411.

87.

Laurson

Eisenmann

Welk

GJ.

Body fat percentile curves for US children and adolescents. Am J Prev Med. 2011;41(4 suppl 2):S87-S92.

88.

Cole

Bellizzi

Flegal

Dietz

WH.

Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ. 2000;320:1240-1243.

89.

Egger

Zellwegerzahner

Schneider

Junker

Lengeler

Antes

Language bias in randomised controlled trials published in English and German. Lancet. 1997;350:326-329.

90.

Moher

Pham

Klassen

, et al. What contributions do languages other than English make on the results of meta-analyses? J Clin Epidemiol. 2000;53:964-972.

91.

Nussbaumer-Streit

Klerings

Dobrescu

, et al. Excluding non-English publications from evidence-syntheses did not change conclusions: a meta-epidemiological study. J Clin Epidemiol. 2020;118:42-54.

92.

EndNote [computer program]. Version X8. New York, NY: Thomson Reuters; 2016.

93.

Kelley

Pate

RR.

Exercise and adiposity in overweight and obese children and adolescents: protocol for a systematic review and network meta-analysis of randomised trials. BMJ Open. 2017;7:e019512.

94.

Harris

Hallbauer

ES.

Self-directed weight control through dieting and exercise. Behav Res Ther. 1973;11:523-529.

95.

Burnham

JF.

Scopus database: a review. Biomed Digit Libr. 2006;3:1.

96.

Berlin

JA.

Does blinding of readers affect the results of meta-analyses?

Lancet. 1997;350:185-186.

97.

Cohen

Weighted kappa: nominal scale agreement with provision for scaled disagreement or partial credit. Psychol Bull. 1968;70:213-220.

98.

Fanelli

Moher

What difference do retractions make? An estimate of the epistemic impact of retractions on recent meta-analyses. bioRxiv. Published online August 13, 2019. doi:10.1101/734137

99.

Lee

Dobbins

DeCorby

Mcrae

Tirilis

Husson

An optimal search filter for retrieving systematic reviews and meta-analyses. BMC Med Res Methodol. 2012;12:51.

100.

Sterne

JAC

Savovic

Page

, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ. 2019;366:l4898.

101.

Smart

Waldron

Ismail

, et al. Validation of a new tool for the assessment of study quality and reporting in exercise training studies: TESTEX. Int J Evid Based Healthc. 2015;13:9-18.

102.

Maher

Sherrington

Herbert

Moseley

Elkins

Reliability of the PEDro scale for rating quality of randomized controlled trials. Phys Ther. 2003;83:713-721.

103.

Higgins

JPT

Thomas

Chandler

, et al. Cochrane Handbook for Systematic Reviews of Interventions. Version 6.1. Cochrane; 2020.

104.

Doi

SAR

Furuya-Kanamori

Thalib

Barendregt

. Meta-analysis in evidence-based healthcare: a paradigm shift away from random effects is overdue. Int J Evid Based Healthc. 2017;15:152-160.

105.

Doi

SAR

Furuya-Kanamori

. Selecting the best meta-analytic estimator for evidence-based practice: a simulation study. Int J Evid Based Healthc. 2020;18:86-94.

106.

Follmann

Elliot

Suh

Cutler

Variance imputation for overviews of clinical trials with continuous response. J Clin Epidemiol. 1992;45:769-773.

107.

Hedges

LV.

Distribution theory for Glass’s estimator of effect size and related estimators. J Educ Stat. 1981;6:107-108.

108.

Amrhein

Greenland

McShane

Scientists rise up against statistical significance. Nature. 2019;567:305-307.

109.

Wasserstein

Schirm

Lazar

NA.

Moving to a world beyond “P < 0.05.”

Am Stat. 2019;73(suppl 1):1-19.

110.

IntHout

Ioannidis

Rovers

Goeman

JJ.

Plea for routinely presenting prediction intervals in meta-analysis. BMJ Open. 2016;6:e010247.

111.

Higgins

Thompson

Spiegelhalter

DJ.

A re-evaluation of random-effects meta-analysis. J R Stat Soc Series A. 2009;172:137-159.

112.

Cochran

WG.

The combination of estimates from different experiments. Biometrics. 1954;10:101-129.

113.

Higgins

JPT

Thompson

Deeks

Altman

. Measuring inconsistency in meta-analyses. BMJ. 2003;327:557-560.

114.

Lau

Schmid

Chalmers

TC.

Cumulative meta-analysis of clinical trials builds evidence for exemplary medical care: the Potsdam International Consultation on Meta-Analysis. J Clin Epidemiol. 1995;48:45-57.

115.

Gartlehner

Grant

, et al. Conducting quantitative synthesis when comparing medical interventions: AHRQ and the Effective Health Care Program. J Clin Epidemiol. 2011;64:1187-1197.

116.

Rothman

KJ.

No adjustments are needed for multiple comparisons. Epidemiology. 1990;1:43-46.

117.

Sacks

Berrier

Reitman

Ancona-Berk

Chalmers

TC.

Meta-analysis of randomized controlled trials. N Engl J Med. 1987;316:450-455.

118.

Furuya-Kanamori

Barendregt

Doi

SAR

. A new improved graphical and quantitative method for detecting bias in meta-analysis. Int J Evid Based Healthc. 2018;16:195-203.

119.

Furuya-Kanamori

Lin

, et al. P value-driven methods were underpowered to detect publication bias: analysis of Cochrane review meta-analyses. J Clin Epidemiol. 2020;118:86-92.

120.

Schunemann

Brozek

Guyatt

Oxman

GRADE Handbook for Grading Quality of Evidence and Strength of Recommendation. GRADE Working Group; 2013.

121.

GRADEpro Guideline Development Tool [computer software]. McMaster University (developed by Evidence Prime, Inc); 2015.

122.

Garber

Blissmer

Deschenes

, et al. Quantity and quality of exercise for developing and maintaining cardiorespiratory, musculoskeletal, and neuromotor fitness in apparently healthy adults: guidance for prescribing exercise. Med Sci Sports Exerc. 2011;43:1334-1359.

123.

Grossman

Bibbins-Domingo

Curry

, et al. Screening for obesity in children and adolescents: US Preventive Services Task Force Recommendation Statement. JAMA. 2017;317:2417-2426.

124.

Kelley

Pate

RR.

Inter-individual differences in body mass index were not observed as a result of aerobic exercise in children and adolescents with overweight and obesity. Pediatr Obes. 2021;16:e12692.

125.

Kelley

Pate

RR.

Are there inter-individual differences in fat mass and percent body fat as a result of aerobic exercise training in overweight and obese children and adolescents? A meta-analytic perspective. Child Obes. 2020;16:301-306.

126.

Schulz

Altman

Moher

; CONSORT Group. CONSORT 2010 statement: updated guidelines for reporting parallel group randomised trials. BMJ. 2010;340:c332.

127.

Slade

Dionne

Underwood

, et al. Consensus on Exercise Reporting Template (CERT): modified Delphi study. Phys Ther. 2016;96:1514-1524.

128.

Collins

Varmus

A new initiative on precision medicine. N Engl J Med. 2015;372:793-795.

129.

Joyner

MJ.

Precision medicine, cardiovascular disease and hunting elephants. Prog Cardiovasc Dis. 2016;58:651-660.

130.

Joyner

Paneth

Seven questions for personalized medicine. JAMA. 2015;314:999-1000.

131.

Ross

Goodpaster

Koch

, et al. Precision exercise medicine: understanding exercise response variability. Br J Sports Med. 2019;53:1141-1153.

132.

Bomberg

Ryder

Brundage

, et al. Precision medicine in adult and pediatric obesity: a clinical perspective. Ther Adv Endocrinol Metab. 2019;10:2042018819863022.

133.

Kelly

Marcus

Yanovski

Osganian

SK.

Working toward precision medicine approaches to treat severe obesity in adolescents: report of an NIH workshop. Int J Obes (Lond). 2018;42:1834-1844.

134.

Atkinson

Batterham

AM.

True and false interindividual differences in the physiological response to an intervention. Exp Physiol. 2015;100:577-588.

135.

Hopkins

WG.

Individual responses made easy. J Appl Physiol (1985). 2015;118:1444-1446.

136.

Cardel

Jastreboff

Kelly

AS.

Treatment of adolescent obesity in 2020. JAMA. 2019;322:1707-1708.

137.

Thivel

Masurier

Baquet

, et al. High-intensity interval training in overweight and obese children and adolescents: systematic review and meta-analysis. J Sports Med Phys Fitness. 2019;59:310-324.

138.

Cao

Quan

Zhuang

Effect of high-intensity interval training versus moderate-intensity continuous training on cardiorespiratory fitness in children and adolescents: a meta-analysis. Int J Environ Res Public Health. 2019;16:1533.

139.

Centers for Medicare & Medicaid Services. Early and periodic screening, diagnostic, and treatment. Accessed June 20, 2020. https://www.medicaid.gov/medicaid/benefits/early-and-periodic-screening-diagnostic-and-treatment/index.html

140.

Mealy

Richardson

Miller

Smith

Juvancic-Heltzel

JA.

Exercise is Medicine®: knowledge and awareness among exercise science and medical school students. Int J Exerc Sci. 2019;12:505-514.

141.

Dirks-Naylor

Griffiths

Gibson

Luu

JA.

The prevalence of exercise prescription-related course offerings in United States pharmacy school curricula: Exercise is Medicine. Adv Physiol Educ. 2016;40:319-322.

142.

Cardinal

Park

Kim

Cardinal

MK.

If exercise is medicine, where is exercise in medicine? Review of US Medical education curricula for physical activity-related content. J Phys Act Health. 2015;12:1336-1343.

143.

Slusser

Staten

Stephens

, et al. Payment for obesity services: examples and recommendations for stage 3 comprehensive multidisciplinary intervention programs for children and adolescents. Pediatrics. 2011;128(suppl 2):S78-S85.

144.

Littell

Corcoran

Pillai

Systematic Reviews and Meta-Analysis. Oxford University Press; 2008.

145.

Rücker

Schumacher

Simpson’s paradox visualized: the example of the rosiglitazone meta-analysis. BMC Med Res Methodol. 2008;8:34.

146.

Kelley

KS.

Retrieval of individual participant data for exercise meta-analyses may not be worth the time and effort. Biomed Res Int. 2016;2016:5059041.

147.

Polanin

JR.

Efforts to retrieve individual participant data sets for use in a meta-analysis result in moderate data sharing but many data sets remain missing. J Clin Epidemiol. 2018;98:157-159.

148.

Duval

Vazquez

Baker

Jacobs

Jr ; CODA Study Group. The collaborative study of obesity and diabetes in adults (CODA) project: meta-analysis design and description of participating studies. Obes Rev. 2007;8:263-276.

149.

Nevitt

Marson

Davie

Reynolds

Williams

Smith

CT.

Exploring changes over time and characteristics associated with data retrieval across individual participant data meta-analyses: systematic review. BMJ. 2017;357:j1390.

150.

Villain

Dechartres

Boyer

Ravaud

Feasibility of individual patient data meta-analyses in orthopaedic surgery. BMC Med. 2015;13:131.

151.

Lyman

Kuderer

NM.

The strengths and limitations of meta-analyses based on aggregate data. BMC Med Res Methodol. 2005;5:14.

152.

Steinberg

Smith

Stroup

, et al. Comparison of effect size estimates from a meta-analysis of summary data from published studies and from a meta-analysis using individual patient data for ovarian cancer studies. Am J Epidemiol. 1997;145:917-925.

153.

Olkin

Sampson

Comparison of meta-analysis versus analysis of variance of individual patient data. Biometrics. 1998;54:317-322.

154.

Mathew

Nordstrom

On the equivalence of meta-analysis using literature and using individual patient data. Biometrics. 1999;55:1221-1223.

155.

Smith

Marcucci

Nolan

, et al. Individual participant data meta-analyses compared with meta-analyses based on aggregate data. Cochrane Database Syst Rev. 2016;9:MR000007.

156.

Tierney

Fisher

Burdett

Stewart

Parmar

MKB

. Comparison of aggregate and individual participant data approaches to meta-analysis of randomised trials: an observational study. PLoS Med. 2020;17:e1003019.

157.

Huang

Tang

Tam

WWS

, et al. Comparing the overall result and interaction in aggregate data meta-analysis and individual patient data meta-analysis. Medicine (Baltimore). 2016;95:e3312.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.09 MB

2.59 MB

0.15 MB

0.17 MB