Association Between Physical Activity and Indicators of Overweight/Obesity and Metabolically Unhealthy Obesity Risk in Children and Adolescents: A Systematic Review of Prospective Epidemiological Studies and Randomized Controlled Trials in Western Countries

ABSTRACT

This systematic review examined the etiologic association between physical activity (PA) and indicators of childhood overweight/obesity (OV/OB) and metabolically unhealthy obesity (MUO) risk. Original peer‐reviewed English reports published between January 01, 2013, and June 30, 2024, were retrieved from MEDLINE and Scopus. A total of 106 prospective epidemiological studies and randomized controlled trials (RCTs) conducted in Western countries among 2‐ to 19‐year‐olds with ≥12‐month follow‐up were eligible. Study quality was evaluated through risk of bias (RoB). Regarding OV/OB, epidemiological studies (n = 86, 57% high‐RoB) revealed a protective effect of certain accelerometry‐measured exposures: moderate‐to‐vigorous PA (n = 17/32), vigorous PA (n = 10/15), substituting lower‐intensity with higher‐intensity PA (n = 7/11), and longitudinally sustaining high PA (n = 3/3). Of the 16 RCTs (31% high‐RoB), 10 were successful and most utilized long‐term (≥ 1‐year) multicomponent PA interventions incorporating behavior change techniques. Random‐effects meta‐analysis of RCTs revealed a beneficial effect of PA on body mass index (BMI) (n = 5: −0.42 [95% CI: −0.64, −0.21] kg/m²) and BMI z‐score (n = 5: −0.09 [95% CI: −0.14, −0.04] SD) change; results were consistent in mixed‐effects linear regression models accounting for repeated measurements. PA had no impact on MUO indicators, albeit studies were scarce (n = 4). Promoting PA is a promising strategy against childhood OV/OB. Future research should focus on elucidating its role in MUO prevention.

Abbreviations

CI

confidence interval

OV/OB

overweight/obesity

MUO

metabolically unhealthy obesity

PA

physical activity

WHO

World Health Organization

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta‐Analyses

PECO

Population, Exposure, Comparator, Outcome

PICO

Population, Intervention, Comparator, Outcome

BMI

body mass index

RCT(s)

randomized controlled trial(s)

ROBINS‐E

Risk Of Bias In Nonrandomized Studies–of Exposures

RoB‐II

revised Cochrane Risk of Bias

1. Introduction

Childhood overweight/obesity (OV/OB) is a major global public health concern [1]. It is accompanied by a wide range of adverse psychosocial consequences, including poor cognitive functioning and academic performance, weight stigma, discrimination, anxiety, depression, and low quality of life [2, 3]. Most importantly, it often persists into adulthood, imposing serious short‐ and long‐term physical threats, mostly related, though not limited, to cardiometabolic health. In light of this, the concept of different OB phenotypes has emerged in the literature, and focus has been placed on metabolically unhealthy obesity (MUO), i.e., the combination of OB with poor metabolic health, such as impaired glucose metabolism, dyslipidemia, and/or hypertension, which leads to an increased risk of cardiovascular diseases, mortality, and disability [4, 5].

Childhood OV/OB is a multifactorial chronic disease, caused by a dynamic interplay between genetic predisposition and the environment, and results from a persistent imbalance between energy intake and energy expenditure in favor of the former [6]. Physical activity (PA) is the most modifiable component of energy expenditure and thus a crucial determinant of children's body weight status. In 2020, the World Health Organization (WHO) and other European scientific associations published updated guidelines on PA for children and adolescents, recommending a minimum of 60 min of moderate‐to‐vigorous PA per day for benefits in various health‐related outcomes, including adiposity and cardiometabolic health [7, 8]. However, evidence supporting these guidelines was drawn from the few available systematic reviews published up to 2019, most of which were of low methodological quality [7].

The latest data on the association between PA and the risk of childhood OV/OB or MUO, published from 2020 and on, are also characterized by a high degree of heterogeneity and present several limitations. Regarding childhood OV/OB, systematic reviews of epidemiological data usually include nonlongitudinal (cross‐sectional and case–control) studies, which cannot support etiologic associations, focus on certain age ranges not covering the whole spectrum of childhood, and/or emphasize specific PA exposures not covering the whole spectrum of PA behavior [9, 10, 11, 12, 13]. Accordingly, systematic reviews of interventional data include studies with a short follow‐up duration (usually ≤ 6 months), uncontrolled and/or nonrandomized clinical trials which present methodological drawbacks, and/or studies with multicomponent interventions (e.g., including a dietary component), which provide evidence for a healthy lifestyle in total and thus not specific to PA [14, 15, 16, 17, 18, 19]. Regarding childhood MUO, there are only a few systematic reviews of interventional studies available, which include short‐term clinical trials of various designs (uncontrolled and/or nonrandomized), focus on certain age ranges, emphasize either specific PA interventions (e.g., aerobic and resistance training) or multicomponent lifestyle interventions, and/or examine specific cardiometabolic indices as outcomes [20, 21, 22, 23, 24].

Childhood OV/OB is of particular importance in countries of the Western world, i.e., Europe, North America, and Oceania [25, 26]. For example, the World Obesity Atlas 2023 projects that by 2035, 17 million boys and 11 million girls aged 5–19 years will be living with OB in Europe [25]. Accordingly, in the USA, the prevalence of OB has increased by approximately 30% in 2‐ to 19‐year‐old children from 1998 to 2018 [26]. In Western countries, increasing urbanization has led to obesogenic environments, which hamper engagement in PA due to the lack of green areas, recreational spaces, and walkable neighborhoods [27], and facilitate the nutritional transition phenomenon, i.e., the abandonment of traditional dietary habits and the shift to processed energy‐dense foods, due to the globalization of food supply [28]. Given the above, targeted research on the etiology and prevention of childhood OV/OB in Western settings is crucial. However, most of the available systematic reviews on the association between PA and the risk of childhood OV/OB include studies conducted in both Western and non‐Western countries, which have substantial differences in sociocultural and environmental characteristics [9, 10, 11, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24]. This diversity in contexts contributes to a high degree of heterogeneity in study population background and hampers the generation of clear conclusions and implications for public health decision making.

In light of the existing knowledge and research gaps, in the present work, we aimed to systematically review evidence of the etiologic association between PA and indicators of childhood OV/OB and MUO risk in Western countries. We focused on data from long‐term studies with longitudinal designs of the highest methodological quality, which are able to indicate causal links/effects, covering the whole spectrum of PA‐related exposures/interventions and OV/OB‐ or MUO‐related outcomes throughout childhood and adolescence.

2. Materials and Methods

A systematic review of the scientific literature focusing on the association between PA and the risk of childhood OV/OB or MUO was performed according to the 2020 Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines [29, 30]. The PRISMA checklist is available in Table S1. The systematic review protocol was registered in the PROSPERO database on November 27, 2023 (identification number: CRD42023482948) and the methodology followed is described in detail in the following sections.

2.1. Research Questions

The research questions were formulated according to the “Population, Exposure/Intervention, Comparator, Outcome” (PE/ICO) model for epidemiological and interventional studies, respectively (Table 1) [31, 32]. The study population was defined as children 2–19 years of age; the lower age limit was chosen given that until the second year of life, there is limited autonomy for PA due to lack of advanced motor skills [33] and different approaches apply for the evaluation of body weight status [e.g., reference data for weight‐for‐length instead of body mass index (BMI) z‐score] [34], while the upper age limit is in line with the WHO definition of adolescence (the second decade of life, i.e., 10–19 years of age) [35]. We focused on epidemiological studies with PA‐related exposures and clinical trials utilizing interventions specifically aiming to promote PA, while the outcomes of interest were the presence and/or individual indicators of OV/OB and MUO.

TABLE 1.

Research questions.

	Prospective epidemiological studies	Randomized controlled trials
Population	Children and adolescents 2–19 years of age
Exposure	High/low level of PA	NA
Intervention	NA	Intervention to promote PA
Comparator	Low/high level of PA	Standard care or no intervention
Outcome	Presence (odds or incidence) of OV/OB and indicators of OV/OB (e.g., body mass index, waist circumference and body fat); Presence (odds or incidence) of MUO and indicators of MUO (e.g., blood pressure, blood lipids, glucose metabolism indices and OV/OB‐related metabolic comorbidities)

Abbreviations: MUO, metabolically unhealthy obesity; NA, not applicable; OV/OB, overweight/obesity; PA, physical activity.

2.2. Eligibility Criteria

The criteria based on which studies were assessed for eligibility for inclusion in the systematic review are presented in Table 2. Original peer‐reviewed journal articles, published in English within a decade before the registration of the systematic review in PROSPERO, i.e., from January 01, 2013, to June 30, 2024, were eligible. Studies with exposures/interventions and outcomes not relevant to the research questions were considered out of the scope of the systematic review. For the research question related to OV/OB, studies in general youth populations of mixed body weight status (mostly of healthy body weight) at baseline, which is the typical scenario in the majority of research focusing on the etiology and prevention of childhood OV/OB in community settings, were considered eligible, while those in children only with OV/OB at baseline were excluded. The opposite criteria were applied for the research question related to MUO (studies only in children with OV/OB at baseline were considered eligible, while those exploring the association/effect between/of PA and/or cardiometabolic outcomes in mixed body weight youth populations were excluded), to obtain and synthesize data relevant to the MUO phenotype, i.e., metabolic manifestations in the presence of OV/OB. Studies in children with noncommunicable diseases, intellectual/developmental disabilities, monogenic disorders, syndromic forms of OB, or receiving medical treatment for OB were additionally excluded. Regarding the study design, only prospective epidemiological studies and randomized controlled trials (RCTs), which provide robust evidence of etiologic associations [36], were included. A minimum follow‐up duration of 12 months was also set as a criterion for both study designs as adequate to observe meaningful long‐term associations/effects (in the case of RCTs, those with a ≥12‐month follow‐up from baseline were included regardless of the duration of the applied interventions). Regarding the geographical region, we focused on studies conducted in Europe, USA, Canada, and Oceania, which are characterized by a westernized way of living and share similar food and PA environments, while studies conducted in South America, Asia, or Africa were excluded.

TABLE 2.

Eligibility criteria.

Parameter	Inclusion criteria	Exclusion criteria
Publication type	Original peer‐reviewed journal articles	Letters, editorials, reviews, study protocols, preprints
Publication time frame	From 01/01/2013 to 30/06/2024	Anything published till 31/12/2012
Publication language	English	Other than English
Study population	Humans 2–19 years of age	Animals Humans <2 and >19 years of age Humans with chronic noncommunicable diseases and intellectual/developmental disabilities Humans with monogenic disorders (e.g., melanocortin 4 receptor deficiency, leptin deficiency), with syndromic forms of obesity (e.g., Prader‐Willi syndrome, Alstrom syndrome), or on medication known to affect body weight (e.g., antidepressants, antiepileptics, antipsychotics, migraine management drugs, mood stabilizers, antimanic drugs, corticosteroids) Humans under medical treatment for OB (e.g., weight loss medication or bariatric/metabolic surgery)
Study design	Prospective epidemiological studies and randomized controlled trials	Cross‐sectional and case–control epidemiological studies, nonrandomized uncontrolled before‐after trials, controlled experiments, in vitro studies, in vivo animal studies, in silico studies
Study duration	≥12‐month follow‐up	<12‐month follow‐up
Study exposure	Related to PA	Not related to PA Multicomponent exposures (e.g., combination of multiple lifestyle habits in clusters/patterns) PA treated as confounder, moderator or mediator
Study intervention	Related to PA	Not related to PA Multicomponent interventions (e.g., targeting multiple lifestyle habits)
Study outcome	Related to OV/OB or MUO	Not related to OV/OB or MUO
Study region	Europe, USA, Canada, Oceania	Asia, Africa, South America
Other		Studies not reporting longitudinal associations between PA and childhood OV/OB or MUO For studies with OV/OB‐related outcomes: epidemiological studies longitudinally following children with established OV/OB at baseline and interventional studies utilizing PA interventions for the management of OV/OB For studies with MUO‐related outcomes: studies exploring the association/effect between/of PA and/or cardiometabolic indices in mixed body weight populations

Abbreviations: MUO, metabolically unhealthy obesity; OV/OB, overweight/obesity; PA, physical activity.

2.3. Search Strategy

The literature search was conducted by 1 researcher (MG) in MEDLINE (PubMed) and Scopus databases in July 2024. The search query was defined according to the PECO/PICO research questions and included keywords relevant to the target population (children and adolescents), the exposures/interventions of interest (related to PA), the outcomes of interest (related to OV/OB and MUO), and the desired study design (prospective epidemiological studies and RCTs). The exact search query and the applied filters for each database are presented in Table S2. A manual search of the reference lists of the included studies and previous systematic reviews on the topic was also performed by 1 researcher (MG) to enhance search comprehensiveness and mitigate the risk of missing relevant studies.

2.4. Study Selection

The results of the literature searches in MEDLINE (PubMed) and Scopus databases were merged in the Zotero software (https://www.zotero.org), duplicates were removed, and studies were screened for eligibility in 2 stages. Initially, a title and abstract screening was performed, and after exclusions, all remaining reports were full‐text reviewed. In both stages, reports were screened independently by 2 researchers (MG and IG) and any disagreements were discussed with a third senior researcher (GA) until a mutual decision was reached. The detailed process of study selection is presented in the PRISMA flowchart (Figure 1).

FIGURE 1.

PRISMA flow diagram. *A total of 132 reports were excluded for the reasons listed in the flowchart. The number of reasons exceeds the number of excluded reports, since some reports were excluded for multiple reasons. Abbreviations: MUO, metabolically unhealthy obesity; OV/OB, overweight/obesity; PA, physical activity; RCT, randomized controlled trial.

2.5. Risk of Bias (Quality) Assessment

Risk of bias was assessed by 1 researcher (AK or IG), with the second researcher randomly checking a sample (20%) of the eligible reports, and any disagreements were discussed with a third senior researcher (GA) until a mutual decision was reached. The ROBINS‐E (Risk Of Bias In Non‐randomized Studies–of Exposures) tool was used for prospective epidemiological studies [37], and the RoB‐II (revised Cochrane Risk of Bias) tool for individually randomized, parallel group clinical trials (adapted versions were used in case of cluster or crossover RCTs) [38].

2.6. Data Extraction and Synthesis

Data extraction was performed by 1 researcher (MG or IG), with the second researcher randomly checking a sample (20%) of the eligible reports. For each included report, information was extracted in relation to the study (year of publication, country, setting, duration), the population (sample size, age, sex distribution), the exposures (type, definition, assessment) or interventions (duration, delivery, groups, randomization, content), the outcomes (type, definition, assessment), the statistical analysis (analyzed sample, statistical models, covariates), and the study findings. In case of missing information or uncertainties, relevant information was sought from other published reports of the same study or via direct communication with the study investigators. Study findings are presented separately according to the main outcome category, i.e., OV/OB and MUO, and within each category separately according to study design, i.e., prospective epidemiological studies and RCTs. Within each study design, reports were considered for pooling if exposures/interventions were homogeneous. Prospective epidemiological studies were considered indicative of a protective effect of PA against OV/OB or MUO if a significant beneficial association between any PA exposure and any OV/OB or MUO‐related outcome was evident. Similarly, an RCT was considered successful if the PA intervention had a beneficial impact on any OV/OB or MUO‐related outcome, as indicated by a significant between‐group difference in favor of the intervention arm. A sensitivity analysis was also performed after excluding epidemiological studies with major methodological limitations (i.e., with high or very high risk of bias) to explore the consistency of findings.

2.6.1. Meta‐Analysis

A meta‐analysis of RCTs was conducted to explore the pooled effect of PA interventions on OV/OB‐related outcomes. The detailed methodology is presented in Appendix 1. In brief, random‐effects meta‐analysis was performed using the DerSimonian and Laird method. This analysis treated each study or follow‐up point as an independent observation, ignoring within‐study clustering. Pooled estimates were obtained using inverse‐variance weighting. Between‐study heterogeneity was quantified using τ², I ², and Cochran's Q statistic. To evaluate the stability of results, a leave‐one‐out meta‐analysis was performed by removing 1 study at a time and recalculating the pooled effect size. For outcomes with multiple follow‐up time points, weighted mixed‐effects linear regression models with a random intercept for study to account for clustering and repeated measures were additionally employed. Models were weighted by the inverse of the variance of each comparison. Follow‐up duration was included as a categorical fixed effect. Between‐study heterogeneity was assessed via τ² and pseudo‐I ². All analyses were conducted in Stata version 18.5 and statistical significance was set at p < 0.05.

3. Results

3.1. Study Selection

The detailed study selection process is presented in Figure 1. A total of 12,720 records (9863 from Scopus and 2857 from PubMed) were identified. After removing duplicates (2300), 10,420 unique records were screened for eligibility, of which 238 were full‐text reviewed. A total of 106 studies met the eligibility criteria and were included in the systematic review. Of those, 102 studies (86 prospective epidemiological studies and 16 RCTs) examined the relationship between PA and childhood OV/OB, and 4 studies (2 prospective epidemiological studies and 2 RCTs) examined the relationship between PA and childhood MUO.

3.2. PA and Indicators of Childhood OV/OB risk

3.2.1. Prospective Epidemiological Studies

3.2.1.1. General Characteristics

The 86 prospective epidemiological studies exploring the association between PA and indicators of childhood OV/OB risk are presented in detail in Table S3 [39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124]. Of those, 47 were conducted in Europe [40, 41, 42, 45, 48, 49, 50, 52, 54, 55, 56, 57, 58, 59, 60, 65, 66, 68, 70, 71, 72, 73, 74, 76, 77, 78, 80, 82, 83, 92, 93, 94, 95, 97, 98, 99, 100, 101, 104, 108, 109, 112, 113, 115, 121, 123, 124], 21 in the USA [39, 43, 46, 61, 64, 79, 81, 86, 88, 89, 90, 91, 102, 103, 105, 107, 111, 117, 118, 119, 122], 9 in Canada [44, 47, 53, 67, 69, 85, 87, 114, 120], 7 in Australia [51, 62, 63, 75, 84, 96, 106], 1 in New Zealand [116], and 1 both in Europe and the USA [110]. Of the 86 studies, 46 recruited participants from schools [39, 40, 42, 45, 46, 48, 49, 52, 54, 55, 56, 57, 58, 60, 61, 62, 64, 65, 67, 69, 70, 72, 73, 74, 75, 77, 78, 82, 85, 86, 87, 91, 93, 94, 100, 101, 108, 109, 111, 113, 114, 115, 117, 119, 120, 123], 30 from the general population [41, 43, 44, 50, 51, 53, 63, 68, 71, 80, 81, 83, 84, 88, 92, 95, 97, 98, 99, 103, 104, 105, 106, 107, 112, 116, 118, 121, 122, 124], 6 from various community settings [76, 79, 89, 90, 96, 102], 3 from primary care settings [47, 59, 66], and 1 from both schools and the general population [110]. The size of the baseline study samples ranged from 119 [102] to 21,357 individuals [39], and participants' baseline mean age ranged from 2.0 years [95] to 17.2 years [56]. Most studies (n = 76) included both male and female participants, with the percentage of females ranging from 44.7% [59] to 66.7% [56], 4 involved only females [49, 64, 81, 89], 2 involved only males [40, 115], and 4 did not report the population's sex distribution [44, 92, 100, 123]. The duration of follow‐up ranged from 1 year [45, 48, 59, 77, 100, 102, 122] to 14 years [88].

3.2.1.2. Risk of Bias

According to the ROBINS‐E tool, of the 86 prospective epidemiological studies, 19 (22%) were characterized as very high risk of bias [43, 49, 52, 53, 54, 56, 69, 75, 82, 84, 85, 87, 92, 93, 95, 110, 111, 115, 117], 30 (35%) as high risk of bias [44, 45, 47, 50, 55, 57, 59, 60, 61, 62, 67, 71, 73, 74, 76, 78, 79, 80, 88, 89, 91, 94, 96, 98, 101, 105, 109, 112, 114, 120], 30 (35%) as raising some concerns [39, 40, 41, 42, 51, 58, 63, 64, 65, 66, 68, 70, 72, 77, 81, 86, 90, 97, 99, 100, 103, 106, 107, 113, 116, 118, 119, 121, 122, 124], and 7 (8%) as low risk of bias (Table S4) [46, 48, 83, 102, 104, 108, 123]. Risk of bias tended to be higher in domain 1 (confounding), which was evaluated as very high or high risk in 35 studies and as raising some concerns in 36 studies, and domain 5 (missing data), which was evaluated as very high or high risk in 19 studies and as raising some concerns in 36 studies.

3.2.1.3. Exposures

A total of 50 prospective epidemiological studies measured PA objectively via accelerometry (Figure 2a) [41, 42, 43, 45, 48, 49, 50, 54, 55, 57, 58, 59, 60, 61, 62, 64, 67, 68, 70, 72, 73, 75, 76, 77, 79, 80, 82, 83, 86, 88, 90, 92, 94, 96, 97, 98, 99, 100, 102, 104, 105, 108, 113, 114, 115, 116, 119, 120, 121, 123]. Accelerometer wear time ranged from 3 days [105] to 14 days [94], while the definition of nonwear time, valid measurements, and different PA intensities varied among studies. In 34 studies, the exposure was time spent in PA of various intensities (baseline, change from baseline to follow‐up or longitudinal), usually expressed as minutes/day [41, 42, 43, 45, 48, 49, 50, 54, 57, 62, 64, 67, 73, 75, 77, 79, 80, 82, 83, 86, 88, 92, 94, 97, 98, 99, 102, 104, 105, 108, 114, 120, 121, 123]. Among these, 3 studies also assessed PA energy expenditure [73, 82, 102], and 1 study additionally evaluated participants' PA level as exposures [102]. In 10 studies, participants' compliance with PA thresholds was used as exposure [48, 59, 70, 86, 90, 96, 99, 114, 115, 119]. The most common thresholds were 60 min/day [48, 59, 70, 96, 114, 115], or those defined based on the distribution of moderate‐to‐vigorous PA (e.g., tertiles or quartiles) [86, 99, 114, 119]. In 11 studies, the isotemporal substitution of PA was used as exposure [43, 45, 55, 58, 60, 61, 72, 77, 100, 116, 123], of which 8 studies evaluated the reallocation of time between different PA variables [43, 45, 55, 58, 60, 61, 77, 100] (from sedentary time to PA of various intensities [43, 45, 55, 58, 60, 61, 77, 100], from light or moderate PA to vigorous PA [77, 100], and from light PA to moderate‐to‐vigorous PA [61]), and 5 studies assessed the isometric log‐ratio of PA, i.e., increasing a PA behavior while decreasing the rest of PA behaviors [55, 72, 100, 116, 123]. In 3 studies, longitudinal moderate‐to‐vigorous PA data were used to identify PA trajectories [68, 76, 113].

FIGURE 2.

Distribution of exposures in prospective epidemiological studies in which (a) PA was measured objectively through accelerometery (n = 50) and (b) PA was assessed subjectively through children's or parents' reports (n = 39). *The sum of the number of studies with each exposure exceeds the number of total studies, since some studies had multiple exposures. **Based on the combination of time spent in PA and sedentary activities. Abbreviation: PA, physical activity.

A total of 39 prospective epidemiological studies evaluated PA subjectively through children's or parents' reports (Figure 2b) [39, 40, 44, 46, 47, 49, 51, 52, 53, 54, 56, 63, 64, 65, 66, 69, 71, 74, 78, 81, 84, 85, 87, 89, 91, 93, 95, 101, 103, 106, 107, 109, 110, 111, 112, 117, 118, 122, 124]. In 14 studies, the exposure was participation in organized PA (sports) [51, 56, 63, 64, 65, 71, 85, 89, 91, 93, 106, 110, 111, 117] (2 used thresholds [65, 93]). In 11 studies, the exposure focused on extracurricular/leisure‐time activities [44, 47, 49, 66, 69, 74, 84, 85, 101, 103, 109] [2 used thresholds [66, 101]], of which 3 used longitudinal data to identify trajectories [74, 103, 107]. In 7 studies, the exposure was total PA [52, 78, 81, 87, 106, 118, 124] (3 used thresholds [52, 118, 124]), of which 1 used longitudinal data to identify trajectories [118]. In 4 studies, the exposure was school commuting (characterized as active [e.g., walking, cycling] versus passive [e.g., car, bus]) [39, 54, 56, 112], of which 1 used longitudinal data to identify trajectories [112]. In 3 studies, the exposure was outdoor play time (e.g., in a playground) [46, 65, 95] (2 used thresholds [65, 95]). In 3 studies, the total PA profile based on the combination of time spent in PA and sedentary activities was used as exposure [40, 53, 107], of which 1 used longitudinal data to identify trajectories [107]. Only 1 study evaluated the reallocation of time from sedentary activities to PA [122].

3.2.1.4. Outcomes

Study outcomes varied among studies (Tables 3 and 4). Indicators of general OV/OB, i.e., BMI [42, 44, 45, 49, 54, 56, 58, 59, 61, 69, 71, 74, 75, 76, 77, 78, 79, 80, 83, 85, 86, 87, 94, 95, 99, 100, 102, 110], BMI z‐score [39, 41, 48, 50, 58, 59, 60, 66, 71, 84, 92, 93, 96, 98, 104, 105, 111, 112, 114, 116, 117, 118, 122, 124], BMI percentile [64, 80, 113], BMI/BMI z‐score trajectories [51, 65, 91, 107, 120], follow‐up odds of OV/OB [40, 41, 46, 52, 53, 59, 63, 93, 99, 101, 107, 108, 113, 115], and incidence of OV/OB [40, 66, 90, 93, 103, 109, 114], as well as indicators of abdominal OV/OB, i.e., waist circumference [42, 44, 47, 54, 58, 60, 69, 70, 74, 75, 84, 96, 104, 108, 114, 123], waist circumference z‐score [41, 42, 48, 58, 66, 73], waist circumference percentile [113], waist‐to‐height ratio [45, 61], follow‐up odds of abdominal OB [41], and incidence of abdominal OB [66], were assessed in most studies. Skinfold thickness was measured in 6 studies [45, 48, 54, 69, 81, 87] and used to assess body composition in 3 studies [45, 81, 87]. Several studies utilized body composition analysis methods to estimate fat mass [42, 43, 49, 50, 55, 57, 60, 61, 62, 64, 67, 68, 71, 72, 73, 74, 77, 80, 82, 83, 84, 86, 88, 89, 95, 97, 100, 102, 104, 106, 108, 112, 116, 119, 121, 124], fat‐free/lean mass [42, 72, 74, 77, 83, 89, 100, 102, 104, 116, 121, 124], or fat mass components (visceral adipose tissue [43, 55, 88, 124], subcutaneous adipose tissue [43, 124], total abdominal adipose tissue [43], and trunk fat mass [57, 60]). The most common methods were bioimpedance and dual‐energy x‐ray absorptiometry, while air‐displacement plethysmography [72, 77, 100] and magnetic resonance imaging [43, 124] were also applied.

TABLE 3.

Overview of prospective epidemiological studies exploring the association between objectively measured PA and indicators of childhood OV/OB risk.

Study	Country	Population				FU	Exposures						Outcomes									RoB*	Results
		BL N	BL age	Sex	Setting		TPA	LPA	MPA	VPA	MVPA	iso PA	BMI	OV/OB	WC	AO	WHtR	SF	FM	FFM	FM components
Wiersma, 2020 [41]	Netherlands	1135	C	F + M	GP	5 y		BL	BL	BL	BL		FU z, Δ z	FU OD	FU z, Δ z	FU OD						SC	+
Aars, 2020 [42]	Norway	1039	A	F + M	S	2 y		BL			BL		Δ		Δ				Δ	Δ		SC	+
Kracht, 2023 [43]	USA	342	C	F + M	GP	2 y		BL	BL	BL		BL							FU		FU	VH	+
Reisberg, 2020 [45]	Estonia	256	C	F + M	S	1 y		BL	BL	BL	BL	BL	FU, Δ				FU, Δ		FU, Δ	FU, Δ		H	+
Skrede, 2021 [48]	Norway	1129	C	F + M	S	1 y	BL			BL	BL, BL thr		FU z		FU z			FU z				L	+
Le, 2023 [49]	Denmark, Estonia, Norway, Portugal	NR	C	F + M	S	6 y		BL	BL	BL			FU									VH	−
Tanaka, 2018 [50]	UK	510	C	F + M	GP	2 y		BL, Δ	BL, Δ	BL, Δ	BL, Δ		Δ z						Δ			H	+
Camiletti‐Moirón, 2020 [54]	Spain	2225	A	F + M	S	2 y		BL			BL		FU		FU			FU				VH	−
Rubin, 2022 [55]	Czech Republic	311	C	F + M	S	5 y						Δ							FU, %Δ		FU, %Δ	H	+
Marques, 2016 [57]	Portugal	510	C	F + M	S	2 y					BL								FU		FU	H	+
Dalene, 2017 [58]	Norway	1306	C	F + M	S	6 y						BL	FU, FU z		FU, FU z							SC	−
Berglind, 2018 [59]	Sweden	965	C	F + M	PC	1 y					BL thr		FU, Δ, FU z, Δ z	FU OD								H	−
Sardinha, 2017 [60]	Portugal	1042	C	F + M	S	2 y						BL	FU z,		FU				FU		FU	H	+
McAlister, 2021 [61]	USA	202	C	F + M	S	30 mo						Δ	FU				FU		FU			H	−
Telford, 2019 [62]	Australia	556	A	F + M	S	4 y	Δ				Δ								Δ			H	−
Cohen, 2014 [64]	USA	303	A	F	S	2 y			Δ	Δ			Δ p						Δ			SC	+
Henderson, 2016 [67]	Canada	630	C	F + M	S	2 y					BL								FU			H	+
Kwon, 2022 [68]	UK	3533	C	F + M	GP	6 y					tr								FU			SC	+
Leppanen, 2022 [70]	Finland	512	C	F + M	S	2 y					BL thr				FU							SC	−
Migueles, 2023 [72]	Sweden	315	C	F + M	S	5 y						BL							FU	FU		SC	−
Vaisto, 2019 [73]	Finland	512	C	F + M	S	2 y	Δ PAE	Δ		Δ	Δ				Δ z				Δ			H	+
Contardo Ayala 2018 [75]	Australia	528	A	F + M	S	2 y		BL, Δ			BL, Δ		Δ		Δ							VH	−
Aira, 2023 [76]	Finland	583	A	F + M	Co	4 y					tr		FU, Δ									H	+
Leppanen 2017 [77]	Sweden	315	C	F + M	S	1 y			BL	BL	BL	BL	FU, Δ						FU, Δ	FU, Δ		SC	+
Mitchell, 2013 [79]	USA	1091	C	F + M	CO	6 y					Δ		Δ									H	+
Janssen, 2019 [80]	UK	502	C	F + M	GP	8 y					Δ		Δ, Δ p						Δ			H	+
Collings, 2016 [82]	UK	930	A	F + M	S	2.5 y	BL PAE	BL	BL	BL									%Δ			VH	−
Griffiths, 2016 [83]	UK	13,681	C	F + M	GP	4 y	BL				BL		FU						FU	FU		L	+
Dowda, 2017 [86]	USA	1083	C	F + M	S	2 y					BL, BL thr		FU						FU			SC	+
Janz, 2017 [88]	USA	339	C	F + M	GP	14 y		14‐y			14‐y								14‐y		8‐y	H	+
Heerman, 2019 [90]	USA	605	C	F + M	CO	3 y					BL thr			IN								SC	−
Steinsbekk, 2015 [92]	Norway	797	C	NR	GP	2 y	BL						Δ z									VH	−
Aadland, 2023 [94]	Norway	365	C	F + M	S	4 y		BL	BL	BL			Δ									H	−
Hinkley, 2020 [96]	Australia	471	C	F + M	CO	6 y	BL thr				BL thr		FU z		FU							H	−
Zahl‐Thanem, 2022 [97]	Norway	797	C	F + M	GP	8 y	BL				BL								FU			SC	+
Remmers, 2014 [98]	Netherlands	334	C	F + M	GP	4 y	2‐y	2‐y			2‐y		Δ z									H	+
Hamer, 2018 [99]	UK	6497	C	F + M	GP	7 y				BL, BL thr	BL		FU	FU OD								SC	+
Migueles, 2022 [100]	Sweden	315	C	NR	S	1 y						BL	FU						FU	FU		SC	+
Butte, 2016 [102]	USA	119	C	F + M	CO	1 y	BL, BL PAE, BL PAL				BL		Δ						Δ	Δ		L	+
Stamatakis, 2015 [104]	UK	6469	C	F + M	GP	4 y					BL		FU z		FU				FU	FU		L	+
Martinez, 2021 [105]	USA	323	C	F + M	GP	2 y					BL		FU z									H	−
van Sluijs, 2016 [108]	UK	2064	C	F + M	S	4 y			BL, Δ	BL, Δ				FU OD	FU, Δ				FU, Δ			L	−
Sprengeler, 2021 [113]	Belgium, Cyprus, Estonia, Germany, Hungary, Italy, Spain, Sweden	16,229	C	F + M	S	6 y					tr		FU p	FU OD	FU p							SC	+
Carson, 2014 [114]	Canada	841	C	F + M	S	2 y		BL thr	BL thr	BL thr	BL thr		FU z	IN	FU							H	+
Latt, 2015 [115]	Estonia	313	C	M	S	2 y				BL thr	BL thr			FU OD								VH	+
Taylor, 2018 [116]	New Zealand	802	A	F + M	GP	4 y						BL	FU z						FU	FU		SC	−
Dowda, 2024 [119]	USA	1083	C	F + M	S	2 y					2‐y thr								2‐y			SC	+
Sohi, 2024 [120]	Canada	630	C	F + M	S	7 y					BL		tr z									H	+
Agbaje, 2023 [121]	UK	7159	C	F + M	GP	4 y		BL			BL								FU	FU		SC	−
Delisle Nyström, 2023 [123]	Sweden	315	C	F + M	S	5 y						BL			FU							L	+

Note: The results of each study are categorized as “+” for a beneficial association between any PA exposure and any OV/OB‐related outcome or as “−” for nonsignificant results.

^*Based on the ROBINS‐E (Risk Of Bias In Nonrandomized Studies–of Exposures) tool (https://www.riskofbias.info/welcome/robins‐e‐tool). A detailed overview of risk of bias assessment is available in Table S4.

Abbreviations: A, adolescents (≥12 years); AO, abdominal obesity; BL, baseline; BMI, body mass index; C, children (<12 years); CO, community; F, females; FM, fat mass; FFM, fat‐free mass; FU, follow‐up; GP, general population; H, high; IN, incidence; iso, isotemporal substitution; L, low; LPA, light physical activity; M, males; mo, month(s); MPA, moderate physical activity; MVPA, moderate‐to‐vigorous physical activity; NR, not reported; OD, odds; OV/OB, overweight/obesity; p, percentile; PA, physical activity; PAE, physical activity expenditure; PAL, physical activity level; PC, primary care; RoB, risk of bias; S, school; SC, some concerns; SF, skinfold thickness; thr, threshold; TPA, total physical activity; tr, trajectories; VH, very high; VPA, vigorous physical activity; WC, waist circumference; z, z‐score; WHtR, waist‐to‐height ratio; y, year(s); Δ, change; %, percentage.

TABLE 4.

Overview of prospective epidemiological studies exploring the association between subjectively assessed PA and indicators of childhood OV/OB risk.

Study	Country	Population				FU	Exposures						Outcomes								RoB*	Results
		BL N	BL age	Sex	Setting		Organized PA (sports)	Extracurricular/leisure‐time PA	School commuting	Outdoor play	PA profile	TPA	BMI	OV/OB	WC	AO	SF	FM	FFM	FM components
Mendoza, 2014 [39]	USA	21,357	NR	F + M	S	5 y			BL				FU z								SC	+
O'Neill, 2017 [40]	Ireland	8570	C	M	S	4 y					BL			FU OD, IN							SC	+
Kosak, 2022 [44]	Canada	2837	C	F + M	GP	2 y		BL					FU		FU						H	−
Arcan, 2013 [46]	USA	454	C	F + M	S	15 mo				Δ				FU OD							L	−
Potter, 2018 [47]	Canada	2114	C	F + M	PC	3 y		BL, Δ							FU						H	−
Le, 2023 [49]	Finland	396	C	F	S	7 y		BL					FU					FU			VH	−
Zulfiqar, 2019 [51]	Australia	4606	C	F + M	GP	8 y	BL						tr								SC	+
Oellingrath, 2017 [52]	Norway	865	A	F + M	S	3 y						BL thr		FU OD							VH	−
Smithers, 2014 [53]	Canada	4386	C	F + M	GP	2 y					BL			FU OD							VH	−
Camiletti‐Moirón, 2020 [54]	Spain	2225	A	F + M	S	2 y			BL				FU		FU		FU				VH	−
Deforche, 2015 [56]	Belgium	2726	A	F + M	S	1.5 y	Δ		Δ				Δ								VH	+
Zulfiqar, 2019 [63]	Australia	4386	C	F + M	GP	6 y	BL							FU OD							SC	−
Cohen, 2014 [64]	USA	303	A	F	S	2 y	Δ						Δ p					Δ			SC	−
Koning, 2016 [65]	Netherlands	4072	C	F + M	S	6 y	BL thr			BL thr			tr z								SC	−
Bawaked, 2020 [66]	Spain	1480	C	F + M	PC	3 y		BL thr					FU z	IN	FU z	IN					SC	−
Belanger, 2018 [69]	Canada	1294	A	F + M	S	5 y		5‐y					FU		FU		FU				VH	−
Basterfield, 2015 [71]	UK	585	C	F + M	GP	3 y	BL, %Δ						FU, %Δ, FU z, %Δ z					FU, %Δ			H	+
Aars, 2019 [74]	Norway	1039	A	F + M	S	2 y		BL, tr					Δ		Δ			Δ	Δ		H	+
de Souza, 2015 [78]	Portugal	6894	A	F + M	S	2 y						2‐y	2‐y								H	−
Narla, 2019 [81]	USA	2879	C	F	GP	10 y						BL						FU			SC	+
Vella, 2018 [84]	Australia	3956	A	F + M	GP	2 y		BL					FU z		FU			FU			VH	−
Cairney, 2017 [85]	Canada	1999	C	F + M	S	3 y	BL	BL					FU								VH	+
Barnett, 2013 [87]	Canada	1195	A	F + M	S	3 y						BL, Δ	FU					FU			VH	+
		951	A	F + M								BL, Δ	FU					FU				+
Day, 2015 [89]	USA	211	C	F	CO	2 y	BL, Δ											FU	FU		H	+
Seo, 2015 [91]	USA	5309	C	F + M	S	18 mo	18‐mo						tr								H	+
Bel‐Serrat, 2019 [93]	Ireland	2755	C	F + M	S	3 y	BL thr						Δ z	FU OD, IN							VH	+
Saldanha‐Gomes, 2017 [95]	France	1266	C	F + M	GP	3 y				BL thr			FU					FU			VH	+
De Coen, 2014 [101]	Belgium	621	C	F + M	S	30 mo		BL thr						FU OD							H	−
Kornides, 2018 [103]	USA	14,490	A	F + M	GP	7 y		tr						IN							SC	+
Vella, 2019 [106]	Australia	3994	C	F + M	GP	4 y	BL					BL						FU			SC	−
Zarrett, 2014 [107]	USA	1482	A	F + M	GP	4 y					tr		tr	FU OD							SC	+
Hebert, 2017 [109]	Denmark	1197	C	F + M	S	59 wk		53‐wk						IN							H	+
de la Rie, 2023 [110]	Netherlands	6690	C	F + M	GP	4 y	BL						FU								VH	−
	Germany	2349	C	F + M	S		BL						FU									−
	UK	13,847	C	F + M	GP		BL						FU									−
	USA	16,660	C	F + M	S		BL						FU									−
Jackson, 2017 [111]	USA	NR	C	F + M	S	9 y	BL						FU z								VH	−
Falconer, 2015 [112]	UK	NR	A	F + M	GP	5.5 y			tr				FU z					FU			H	+
Loren, 2022 [117]	USA	NR	C	F + M	S	2 y	BL						FU z								VH	−
Fung, 2023 [118]	USA	10,574	C	F + M	GP	2 y						tr thr	Δ z								SC	−
Zink, 2024 [122]	USA	11,876	C	F + M	GP	1 y						BL iso	FU z								SC	+
Wu, 2023 [124]	Netherlands	NR	C	F + M	GP	7 y						BL, BL thr	FU z					FU	FU	FU	SC	+

Note: The results of each study are categorized as “+” for a beneficial association between any PA exposure and any OV/OB‐related outcome or as “−” for nonsignificant results.

^*Based on the ROBINS‐E (Risk Of Bias In Nonrandomized Studies ‐ of Exposures) tool (https://www.riskofbias.info/welcome/robins‐e‐tool). A detailed description of risk of bias assessment is available in Table S4.

Abbreviations: A, adolescents (≥12 years); AO, abdominal obesity; BL, baseline; BMI, body mass index; C, children (<12 years); CO, community; F, females; FM, fat mass; FFM, fat‐free mass; FU, follow‐up; GP, general population; H, high; iso, isotemporal substitution; IN, incidence; L, low; M, males; mo, month(s); NR, not reported; OD, odds; OV/OB, overweight/obesity; p, percentile; PA, physical activity; PC, primary care; RoB, risk of bias; S, school; SC, some concerns; SF, skinfold thickness; thr, threshold; TPA, total physical activity; tr, trajectories; VH, very high; WC, waist circumference; wk., week(s); y, year(s); z, z‐score; Δ, change; %, percentage.

3.2.1.5. Synthesis

Of the 50 prospective epidemiological studies in which PA was measured objectively through accelerometry, 32 (64%) were indicative of a beneficial effect of PA on OV/OB‐related outcomes (Table 3) [41, 42, 43, 45, 48, 50, 55, 57, 60, 64, 67, 68, 73, 76, 77, 79, 80, 83, 86, 88, 97, 98, 99, 100, 102, 104, 113, 114, 115, 119, 120, 123]. Among studies evaluating different PA intensities, a relatively consistent pattern was observed in favor of PA of higher intensity. Specifically, a beneficial effect was observed for vigorous (10/15 studies [41, 43, 48, 50, 64, 73, 77, 99, 114, 115]) and moderate‐to‐vigorous PA (17/32 studies [48, 50, 57, 67, 73, 77, 79, 80, 83, 86, 88, 97, 98, 102, 104, 119, 120]), while evidence was limited for light (2/15 studies [42, 45]) and moderate PA (3/11 studies [45, 50, 114]). Most isotemporal substitution studies (7/11) revealed a beneficial effect of reallocating time from lower‐intensity to higher‐intensity activities on OV/OB‐related outcomes; specifically, this was evident for substituting sedentary time with PA of various intensities in 5 studies [43, 45, 55, 60, 100] and increasing vigorous PA relatively to lower‐intensity PA in 5 studies [55, 77, 100, 123]. The 3 studies with PA trajectories as exposure also revealed a beneficial impact of longitudinally sustaining a high PA level [68, 76, 113]. Of note, 5/32 (16%) of studies showing a beneficial effect of objectively measured PA on OV/OB‐related outcomes were characterized as low risk of bias [48, 83, 102, 104, 123], as compared to 1/18 (6%) of studies with nonsignificant results [108]. Findings were confirmed when data synthesis was restricted among the 23 studies with low risk and some concerns in risk of bias. Specifically, a beneficial effect was evident for light PA in 1/3 studies [42], moderate PA in 0/4 studies, moderate‐to‐vigorous PA in 8/14 studies [48, 77, 83, 86, 97, 102, 104, 119], vigorous PA in 5/6 studies [41, 48, 64, 77, 99], reallocating time from lower‐intensity to higher‐intensity activities in 3/6 studies [77, 100, 123], and longitudinally sustaining a high PA level in 2/2 studies [68, 113]. No other pattern of associations/effects between PA‐related exposures and OV/OB‐related outcomes was evident in relation to population characteristics (e.g., age and sex) and study outcomes.

Of the 39 prospective epidemiological studies in which PA was assessed subjectively, 19 (49%) were indicative of a beneficial effect of PA on OV/OB‐related outcomes (Table 4) [39, 40, 51, 56, 71, 74, 81, 85, 87, 89, 91, 93, 95, 103, 107, 109, 112, 122, 124]. Synthesis by exposure did not reveal a consistent pattern of associations/effects. Specifically, a beneficial effect of subjectively assessed PA on OV/OB‐related outcomes was observed in 7/14 studies with organized PA (sports) as exposure [51, 56, 71, 85, 89, 91, 93], in 3/11 studies with extracurricular/leisure‐time PA as exposure [74, 103, 109], in 3/7 studies with total PA as exposure [81, 87, 124], in 2/4 studies with school commuting as exposure [39, 112], in 1/3 studies with outdoor play as exposure [95], and in 2/3 studies with PA profiles as exposure [40, 107]. The sole isotemporal substitution study also revealed a beneficial effect of reallocating self‐reported time from sedentary activities to PA on OV/OB risk, but only among female participants [122]. The risk of bias was high or very high, indicative of a lower methodological quality, in 11/19 (58%) of studies showing a beneficial effect of subjectively assessed PA on OV/OB‐related outcomes [56, 71, 74, 85, 87, 89, 91, 93, 95, 109, 112], and in 13/20 (65%) of studies with nonsignificant results [44, 47, 49, 52, 53, 54, 69, 78, 84, 101, 110, 111, 117]. In sensitivity analysis focusing on the remaining 15 studies characterized as low risk or raising some concerns in terms of risk of bias, a beneficial effect on OV/OB‐related outcomes was evident for organized PA (sports) in 1/5 studies [51], total PA in 2/4 studies [81, 124], extracurricular/leisure‐time PA in 1/2 studies [103], outdoor play in 0/2 studies, PA profiles in 2/2 studies [40, 107], school commuting in 1/1 study [39], and reallocating time from sedentary activities to PA in 1/1 study [122]. No other pattern of associations/effects between PA‐related exposures and OV/OB‐related outcomes was evident in relation to population characteristics (e.g., age and sex) and study outcomes.

3.2.2. Randomized Controlled Trials

3.2.2.1. General Characteristics

The 16 RCTs exploring the effect of PA interventions on indicators of childhood OV/OB risk are presented in detail in Table S5 [125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140]. Of those, 10 were conducted in Europe [125, 127, 128, 129, 131, 132, 135, 136, 137, 140], 5 in Australia [126, 130, 133, 134, 139], and 1 in New Zealand [138]. All were parallel cluster RCTs, of which 1 had a 2 × 2 factorial design [139] and 1 had a pragmatic design [128]. The size of the baseline study samples ranged from 361 [126] to 2280 individuals [128], and participants' baseline mean age ranged from 4.7 [140] to 14.1 years [134]. In 14 RCTs the sample included both male and female participants, with the percentage of females ranging from 47.5% [128] to 57.0% [139], while 1 involved only females [129] and 1 involved only males [126]. The study follow‐up ranged from 1 [128, 134, 137] to 6 years [125]. All 16 RCTs were conducted in schools (1 in kindergartens [140], 7 in primary schools [127, 131, 133, 135, 136, 138, 139], 1 in junior and primary schools [128], and 7 in secondary schools [125, 126, 129, 130, 132, 134, 137]). Most of the RCTs (15/16) had a single intervention/control group, while 1 had 4 arms (3 intervention groups and 1 control group) [139]. In all RCTs, control participants followed the standard school curriculum [in 4 RCTs they received the intervention after study completion [126, 130, 134, 139]). Randomization in study groups was school‐based in 14 RCTs [125, 126, 127, 128, 129, 130, 132, 133, 134, 136, 137, 138, 139, 140] and class‐based in 2 RCTs [131, 135].

3.2.2.2. Risk of Bias

According to the RoB‐II tool, of the 16 RCTs, 5 (31%) were characterized as high risk of bias [125, 127, 131, 132, 136], 5 (31%) as raising some concerns [129, 130, 135, 138, 140], and 6 (38%) as low risk of bias (Table S6) [126, 128, 133, 134, 137, 139]. Risk of bias was mostly evident in domain 2 (deviations from intended interventions), which was judged as high risk in 3 RCTs and as raising some concerns in another 3 RCTs, and in domain 5 (selection of the reported result), which was judged as high risk in 1 RCT and as raising some concerns in another 7 RCTs.

3.2.2.3. Interventions

The PA interventions differed among RCTs (Table 5). Intervention duration ranged from 10 weeks [134] to 4 years [125, 133, 135]. The PA intervention was delivered by ordinary classroom and/or physical education teachers supported by researchers in 10 RCTs [128, 130, 131, 132, 133, 135, 136, 137, 139, 140], or by various combinations of researchers, school staff, community stakeholders, policymakers, parents, and students in 6 RCTs [125, 126, 127, 129, 134, 138]. A total of 7 RCTs utilized digital tools (e.g., websites, smartphone applications) in intervention delivery [126, 128, 129, 130, 132, 134, 137]. Most RCTs (11/16) applied PA interventions targeting various combinations of the school curriculum, the school environment, and/or the family environment [125, 126, 127, 128, 130, 132, 134, 136, 137, 139, 140], with the exception of 3 RCTs solely focusing on a supervised exercise program [131, 133, 135] and 2 RCTs solely focusing on the school environment [129, 138]. Interventions targeting the school curriculum included supervised exercise added to regular physical education [125, 126, 128, 130, 131, 133, 134, 135, 136, 137, 140] and educational activities (e.g., lectures, workshops) [125, 126, 127, 128, 130, 132, 134, 136, 137, 139, 140]. Interventions targeting the school social/organizational environment introduced new opportunities for PA, such as sporting events, PA breaks during lessons, and PA during recess [125, 126, 127, 129, 130, 132, 134, 136, 138, 139]. Of note, 6 RCTs provided teachers with support through consultation or materials [128, 129, 132, 134, 137, 140], 7 RCTs involved teachers' training [126, 127, 129, 130, 134, 137, 140], and 4 RCTs involved students in intervention delivery [126, 127, 129, 134]. Interventions targeting the school physical/built environment focused on the provision of fitness equipment to schools [126, 130, 132, 134, 138, 139], the improvement of the outdoor school area [127, 136, 138, 139], or the design of new school areas for PA (e.g., playgrounds) [127, 138]. Interventions targeting the family environment aimed to create a supportive context for PA at home via in‐person meetings with parents, websites, or newsletters [125, 126, 128, 130, 132, 136, 139, 140].

TABLE 5.

Overview of RCTs exploring the effect of PA interventions on indicators of childhood OV/OB risk.

Study	Country	Population			Intervention										RoB*	Results
		BL Ν	Education level	Sex	Duration	Theoretical framework	Content					Teachers' training	Students as deliverers	Use of digital tools
							School curriculum		School environment		Family environment
							Supervised exercise	Educational activities	Social/organizational	Physical/built
Simon, 2014 [125]	France	954	SEC	F + M	48 mo	✓	✓	✓	✓		✓				H	+
Lubans, 2016 [126]	Australia	361	SEC	M	20 wk	✓	✓	✓	✓	✓	✓	✓	✓	✓	L	−
Christiansen, 2013 [127]	Denmark	1348	PRI	F + M	1 y	✓		✓	✓	✓		✓	✓		H	−
Breheny, 2020 [128]	UK	2280	PRE & PRI	F + M	12 mo	✓	✓	✓			✓			✓	L	+
Harrington, 2018 [129]	UK	1752	SEC	F	14 mo	✓			✓			✓	✓	✓	SC	−
Hollis, 2016 [130]	Australia	1150	SEC	F + M	19–24 mo	✓	✓	✓	✓	✓	✓	✓		✓	SC	+
Sacchetti, 2013 [131]	Italy	497	PRI	F + M	2 y		✓								H	+
Isensee, 2018 [132]	Germany	1296	SEC	F + M	3 mo	✓		✓	✓	✓	✓			✓	H	+
Daly, 2016 [133]	Australia	727	PRI	F + M	4 y		✓								L	+
Kennedy, 2018 [134]	Australia	607	SEC	F + M	10 wk	✓	✓	✓	✓	✓		✓	✓	✓	L	+
Müller, 2016 [135]	Germany	366	PRI	F + M	4 y		✓								SC	−
Meyer, 2014 [136]	Switzerland	502	PRI	F + M	9 mo	✓	✓	✓	✓	✓	✓				H	+
Ten Hoor, 2018 [137]	Netherlands	695	SEC	F + M	1 y	✓	✓	✓				✓		✓	L	+
Farmer, 2018 [138]	New Zealand	840	PRI	F + M	1 y				✓	✓					SC	−
Salmon, 2023 [139]	Australia	593	PRI	F + M	30 mo	✓		✓	✓	✓	✓	✓			L	+
Roth, 2015 [140]	Germany	709	PRE	F + M	11 mo		✓	✓			✓	✓			SC	−

Note: The characteristics of the intervention implemented in each study are marked with the “✓” symbol. The results of each study are categorized as “+” for a beneficial impact of the PA intervention on any kind of OV/OB‐related outcome or as “−” for nonsignificant results.

^*Based on the RoB‐II (revised Cochrane Risk of Bias) tool for cluster randomized trials (https://www.riskofbias.info/welcome/rob‐2‐0‐tool/rob‐2‐for‐cluster‐randomized‐trials). A detailed overview of risk of bias assessment is available in Table S6.

Abbreviations: BL, baseline; F, females; H, high; L, low; mo, month(s); M, males; OV/OB; overweight/obesity; PA, physical activity; PE; physical education; PRE, preschool education (kindergartens, preschools, junior schools); PRI, primary education (primary/elementary schools); RCT, randomized controlled clinical trial; RoB, risk of bias; SC, some concerns; SEC, secondary education (secondary/middle/high schools); wk., week(s); y, year(s).

Of the 16 RCTs, 6 (38%) implemented PA interventions developed a priori on the basis of a theoretical framework for behavior change [126, 127, 130, 134, 136, 139]. These included the social cognitive theory [126, 130, 134, 139], the socioecological framework/model [127, 130, 136, 139], the self‐determination theory [126, 134], and the behavioral choice theory [139]. In 5 RCTs (31%), individual behavioral techniques linked to a theoretical framework were utilized (e.g., self‐monitoring, goal setting, social support, problem solving, reinforcement, and counseling based on motivational interviewing) [125, 128, 129, 132, 137], while in the remaining 5 RCTs (31%), the PA intervention applied was not based on or included components of a behavior change theory [131, 133, 135, 138, 140].

3.2.2.4. Outcomes

Study outcomes varied among RCTs. Indicators of general OV/OB, i.e., body weight [130, 133, 137, 138], BMI [125, 126, 130, 131, 134, 138], BMI z‐score [125, 126, 128, 129, 130, 134, 136, 138, 139], BMI percentile [132, 135, 140], follow‐up odds of OV/OB [131], and incidence of OV/ΟΒ [125], were assessed in most RCTs. Indicators of abdominal OV/OB, i.e., waist circumference [126, 127, 138, 139], waist circumference z‐score [136], and waist‐to‐height ratio [132, 138], were measured in 6 RCTs. The sum of 4 skinfolds (triceps, biceps, subscapular, and suprailiac) was measured as an index of adiposity in 2 RCTs [136, 140]. A total of 7 RCTs utilized body composition analysis methods to estimate fat mass [125, 128, 129, 132, 133, 137] or fat‐free/lean mass [133, 135, 137]. The main assessment method was bioimpedance, whereas dual‐energy x‐ray absorptiometry [133] and deuterium dilution [137] were applied in 2 RCTs.

3.2.2.5. Synthesis

The qualitative synthesis of RCTs yielded mixed results in terms of the effects of PA interventions on OV/OB‐related outcomes (Table 5). Of the 16 RCTs, 10 (62.5%) revealed a beneficial impact of PA interventions on any kind of OV/OB‐related outcome [125, 128, 130, 131, 132, 133, 134, 136, 137, 139]. Most of the successful RCTs shared common characteristics, i.e., utilized multicomponent PA interventions targeting any combination of the school curriculum, the school environment, and the family environment (7/10) [125, 128, 130, 132, 134, 136, 139], had an intervention duration of ≥ 1 year (7/10) [125, 128, 130, 131, 133, 137, 139], and were based on or included techniques of a theoretical framework for behavior change (8/10) (Figure 3a) [125, 128, 130, 132, 134, 136, 137, 139]; however, these characteristics were also evident in some nonsuccessful RCTs (Figure 3b). Of the 10 successful RCTs, 4 (40%) had all 3 of the abovementioned characteristics [125, 128, 130, 139], as compared to 1/6 (17%) of nonsuccessful ones [127]. The use of digital tools was evident in 5/10 (50%) of successful RCTs [128, 130, 132, 134, 137] and 2/6 (33%) of nonsuccessful ones [126, 129]. Moreover, 5/10 (50%) of successful RCTs were characterized as low risk of bias [128, 133, 134, 137, 139], as compared to 1/6 (17%) of nonsuccessful ones [126]. Data synthesis did not reveal a differentiating effectiveness of PA interventions in the prevention of childhood OV/OB by country, population characteristics (e.g., age and sex), intervention delivery mode, or study outcomes.

FIGURE 3.

Distribution of various study and intervention characteristics in (a) successful RCTs (n = 10) and (b) nonsuccessful RCTs (n = 6). *Based on the RoB‐II (revised Cochrane Risk of Bias) tool for cluster randomized trials (https://www.riskofbias.info/welcome/rob‐2‐0‐tool/rob‐2‐for‐cluster‐randomized‐trials). A detailed overview of risk of bias assessment is available in Table S6. **Intervention targeting any combination of the school curriculum, the school environment, and the family environment. Abbrevation: RCT, randomized controlled trial.

3.2.2.6. Meta‐Analysis

3.2.2.6.1. Change in BMI

Random‐effects meta‐analysis of 5 RCTs [125, 126, 130, 131, 134] (2 with multiple follow‐ups [125, 130]) revealed a significant difference between intervention and control groups in change in BMI (−0.42 [95% CI: −0.64, −0.21] kg/m²) (Figure 4a). Heterogeneity between studies was not present (I ² = 0.00%; Q = 1.15, p = 0.99). Results from a leave‐one‐out sensitivity analysis were constant across most iterations, but removing the study of Sacchetti et al. [131] attenuated the pooled effect (−0.26 [95% CI: −0.64, 0.11] kg/m²) (Figure S1). Results remained similar when accounting for the repeated follow‐up measurements in mixed‐effects linear regression models (Figure 4b). Intervention groups exhibited a more favorable BMI change compared to control groups in studies with a follow‐up duration of 1 year, with a mean difference of −0.27 (95% CI: −0.41, −0.13) kg/m². There was no significant change in effect for studies with 1–2 years of follow‐up (p = 0.53). In studies with > 2 years of follow‐up, the intervention effect appeared smaller, with the between‐group difference in BMI change reduced by 0.033 kg/m², suggesting a potential attenuation over time. The mixed‐effects linear regression models yielded higher heterogeneity estimates (τ² = 0.021; pseudo‐I ² = 99%) compared to the conventional random‐effects meta‐analysis, which may reflect real differences in intervention effects across time points or study settings that were masked in the simpler model. Excluding the study of Sacchetti et al. [131] slightly reduced heterogeneity (τ² = 0.009; I ² = 97%) and attenuated the overall effect (−0.21 [95% CI: −0.32, −0.10] kg/m²). The pattern of follow‐up duration effects remained consistent in both time points.

FIGURE 4.

Meta‐analysis of RCTs with change in BMI as outcome. (a) Forest plot for the mean difference in BMI change from baseline to follow‐up between intervention and control groups. The vertical grey line represents the null association. Each follow‐up point of a study was treated as an independent observation: Hollis, 2016 (1): 12 mo; Hollis, 2016 (2): 24 mo; Simon, 2014 (1): 48 mo; Simon, 2014 (2): 24 mo; Simon, 2014 (3): 76 mo. (b) Difference in BMI change from baseline to follow‐up between intervention and control groups by follow‐up duration. Abbreviations: BMI, body mass index; CI, confidence interval; diff, difference; mo: month; RCT, randomized controlled trial.

3.2.2.6.2. Change in BMI Z‐Score

Random‐effects meta‐analysis of 5 RCTs [125, 126, 130, 134, 139] (3 with multiple follow‐ups [125, 130, 139]) revealed a significant difference between intervention and control groups in change in BMI z‐score (−0.09 [95% CI: −0.14, −0.04] SD) (Figure 5a). Heterogeneity between studies was not present (I ² = 0.00%; Q = 2.83, p = 0.94), and results remained consistent in a leave‐one‐out sensitivity analysis (Figure S2). Results remained similar when accounting for the repeated follow‐up measurements in mixed‐effects linear regression models (Figure 5b). Intervention groups exhibited a more favorable BMI z‐score change compared to control groups in studies with a follow‐up duration of 1 year, with a mean difference of −0.062 (95% CI: −0.113, −0.011) SD. The difference at 1–2 years was slightly higher (0.030 [95% CI: 0.029, 0.031] SD, p < 0.001), indicating a modest attenuation of effect, while no meaningful difference was observed for > 2 years of follow‐up (0.001 [95% CI: −0.076, 0.078] SD, p = 0.971). Variance (τ²) was 0.0021, suggesting about 0.05 SD of true difference in BMI z‐score change between studies, and pseudo‐I ² was approximately 79%, indicating moderate heterogeneity and suggesting that differences in intervention effects across time points or study settings could be masked in the simpler model.

FIGURE 5.

Meta‐analysis of RCTs with change in BMI z‐score as outcome. (a) Forest plot for the mean difference in BMI z‐score change from baseline to follow‐up between intervention and control groups. The vertical grey line represents the null association. Each follow‐up point of a study was treated as an independent observation: Salmon, 2023 (1): 18 mo; Salmon, 2023 (2): 30 mo; Simon, 2014 (1): 24 mo; Simon, 2014 (2): 48 mo; Simon, 2014 (3): 76 mo; Hollis, 2016 (1): 12 mo; Hollis, 2016 (2): 24 mo. (b) Difference in BMI z‐score change from baseline to follow‐up between intervention and control groups by follow‐up duration. Abbreviations: BMI, body mass index; CI, confidence interval; diff, difference; mo: month; RCT, randomized controlled trial.

3.2.2.6.3. Other Outcomes

Random‐effects meta‐analysis of 3 RCTs [132, 135, 140] with change in BMI percentile as outcome (Figure S3) and 3 RCTs [128, 129, 138] (1 with multiple follow‐ups [138]) with follow‐up BMI z‐score values as outcome (Figure S4) revealed no significant differences between intervention and control groups, while heterogeneity between studies was very low in both cases.

3.3. PA and Indicators of Childhood MUO Risk

3.3.1. Prospective Epidemiological Studies

Only 2 prospective epidemiological studies exploring the association between PA and the risk or indicators of childhood MUO were identified and included in the systematic review [141, 142]. The study of Roberge et al. [141] was conducted in Canada among 69 individuals with metabolically healthy OB, evaluated the incidence of MUO as outcome, and was characterized as high risk of bias. The study of Remmel et al. [142] was conducted in Estonia among 25 males with OV, evaluated the change in several individual cardiometabolic markers as an outcome, and was characterized as very high risk of bias. In both studies, PA was measured objectively through accelerometry, and participants were followed for 2 years. The synthesis of the 2 studies revealed no significant associations between PA and the risk or indicators of childhood MUO [141, 142]. More details can be found in Appendix 2 and Tables S7 and S8.

3.3.2. Randomized Controlled Trials

Only 2 RCTs exploring the effect of PA interventions on the risk or indicators of childhood MUO were identified and included in the systematic review [143, 144]. The study of Jones et al. [143] was conducted in Australia among 37 participants with or at risk of OV/OB who were recruited from community settings and followed for 12 months. The study of Davis et al. [144] was conducted in the USA among 175 participants with OV/OB who were recruited from primary schools and followed for 20 months. Both RCTs had a parallel design, included 2 arms (a single intervention/control group) in which randomization was participant‐based with an allocation ratio of 1:1, compared an intensive PA intervention based on or including a behavior change framework or techniques with a control intervention (less intensive or sedentary activity program), evaluated anthropometric, body composition, and cardiometabolic indices as outcomes, and were characterized as raising some concerns in terms of risk of bias [143, 144]. The synthesis of the 2 RCTs did not reveal a significant impact of PA interventions on indicators of childhood MUO [143, 144]. More details can be found in Appendix 2 and Tables S9 and S10.

4. Discussion

4.1. Summary of Findings

This systematic review aimed to identify and synthesize evidence of the association between PA and indicators of OV/OB and MUO risk among children and adolescents 2–19 years of age, analyzing data from long‐term (≥12 months of follow‐up) longitudinal studies conducted in Western countries. Although the available epidemiological literature is heterogeneous and characterized by methodological limitations, there is evidence of a beneficial effect of specific PA exposures, measured objectively through accelerometry, i.e., engaging in moderate‐to‐vigorous PA, substituting sedentary time with PA, and longitudinally sustaining a high PA level, on childhood OV/OB‐related outcomes. The qualitative synthesis of data from the available RCTs yielded mixed results in terms of the effectiveness of PA interventions in childhood OV/OB prevention, but most successful RCTs seem to share common features, namely a long‐term (≥1‐year) multicomponent PA intervention incorporating behavior change techniques. The meta‐analysis of a subgroup of RCTs sharing common outcomes revealed a significant but slight beneficial impact of PA interventions on some OV/OB‐related outcomes, including change in BMI and change in BMI z‐score, but not on others (change in BMI percentile and follow‐up BMI z‐score values). Data on the association between PA and indicators of childhood MUO risk were scarce and insufficient to yield robust conclusions.

4.2. PA and Indicators of Childhood OV/OB Risk

The 86 prospective epidemiological studies included in the present systematic review differed significantly in terms of exposures and outcomes, while just over half (57%) were judged as high or very high risk of bias, most commonly due to insufficient control for confounders and missing data. Although studies were mixed in terms of the effects of PA on OV/OB‐related outcomes, our synthesis revealed the protective role of specific PA exposures, discussed in detail below, and is supportive of the importance of PA in childhood OV/OB prevention.

A differential link between different PA intensities, measured objectively trough accelerometery, and risk/indicators of childhood OV/OB was observed, according to which engaging in moderate‐to‐vigorous PA seems to have a stronger beneficial effect compared to light‐to‐moderate PA. Although the energy equilibrium during childhood is positive to support normal growth, a great imbalance in favor of energy intake can lead to excess body weight [6]. It may be that only PA of higher intensity can prevent such an imbalance, through inducing a higher PA‐related energy expenditure and a more beneficial body composition, i.e., higher fat‐free/lean mass, which can lead to a higher basal metabolic rate. Moreover, PA is closely and bidirectionally related to physical fitness, defined as “the ability to carry out daily tasks with vigor and alertness, without undue fatigue and with ample energy to enjoy leisure‐time pursuits and to meet unforeseen emergencies” [145]. It is possible that children engaging in PA of higher intensity develop a higher level of physical fitness and are in turn more capable of maintaining a physically active lifestyle with benefits for weight status. PA of higher intensity can also beneficially influence children's affective states, leading to acute postexercise experiences of positive feelings, emotions, or mood that can reinforce long‐term engagement in PA [146], and appetite regulation, leading to anorexigenic effects [147], which may also moderate the observed protective effects of moderate‐to‐vigorous PA on body weight status. Another explanation may be that PA of higher intensity (e.g., active play and sports), compared to that of lower intensity, is more appealing for children and thus more easily sustained in the long‐term.

Research utilizing isotemporal PA substitution models, based on accelerometry, although limited in volume, also revealed a beneficial impact of reallocating time from activities of lower intensity to those of higher intensity on indicators of childhood OV/OB risk. Our findings are in line with previous research showcasing the benefits of the isotemporal substitution of sedentary time with PA on mortality, cardiometabolic health, adiposity, mental health, and fitness, in both youth and adult populations [148, 149, 150]. It is noteworthy that given the 24‐h day time constraint, PA and sedentariness are 2 “competing” behaviors, and as such, an increase in one leads to a decrease in the other. However, “inactivity” is not identical to “sedentariness”, meaning that a physically active child (e.g., engaging in 60 min/day of moderate‐to‐vigorous PA) can exhibit a high level of sedentariness (e.g., recreational screen time) at the same time. The available epidemiological research is also in support of this notion, showing that PA cannot entirely diminish the health risks associated with sedentary behavior, and that individuals characterized as both “inactive” and “sedentary” are at greatest risk of poor health [151]. Therefore, increasing time spent in PA at the expense of sedentariness might represent the optimal approach for public health strategies to prevent childhood OV/OB.

In the few available studies with PA trajectories as exposure, long‐term engagement in objectively measured PA also emerged as protective against the risk of childhood OV/OB [68, 76, 113]. Behavior, including PA‐related behavior, is determined by several individual, interpersonal, and environmental factors, and presents a high degree of variability over time [152, 153]. Rather than focusing on a single time point assessment of PA, the analysis of PA trajectories is a more robust approach that allows the evaluation of PA patterns (stability or change) over time and their dynamic relationship with health/disease [154, 155]. This is important in light of evidence that PA declines from childhood to adolescence [156], mainly due to an increase in sedentariness, driven by higher recreational screen time and prolonged sitting for academic reasons [157], and puberty, the onset of which is characterized by changes in children's body composition and self‐perception [158]. Thus, findings of studies linking past PA level with future OV/OB might be hampered by this decline in PA. In any case, sustaining a high PA level over the course of childhood might be crucial for maintaining a healthy body weight and should be encouraged in the context of a physically active lifestyle that can be tacked into adulthood.

The 16 RCTs included in the present systematic review were heterogeneous in terms of interventions and outcomes, while more than half (62%) were characterized as raising concerns or high risk of bias, mostly due to deviations from intended interventions and selection of the reported results. Of the 16 RCTs, 10 (62.5%) revealed a positive impact of PA on any kind of OV/OB‐related outcome, while the findings of the remaining 6 (37.5%) were nonsignificant. Most of the successful RCTs shared common intervention features. First, they utilized multicomponent PA interventions targeting various combinations of the school curriculum, the school environment and/or the family environment. Interventions targeting multiple settings/stakeholders might be more successful in preventing childhood OV/OB. The school setting is considered a crucial environment, given the significant amount of time that children spend at school, the mandatory physical education classes and the opportunities for PA during recess [159]. The family environment has also been identified as an important determinant of children's body weight status due to the significant role of parents/caregivers in shaping children's lifestyle behaviors [160]. Second, they had an intervention duration of ≥1 year. Although the optimal duration of interventions for the prevention of childhood OV/OB remains vague, the American Academy of Pediatrics is in favor of intensive interventions with ≥26 h of face‐to‐face family‐based counseling over ≥3–12 months for the treatment of childhood OV/OB [34]. Third, they were based on behavior change theories/techniques, in line with data showcasing the importance of behavior change in the prevention and management of childhood OV/OB [161, 162]. Definite conclusions cannot be drawn, since the aforementioned characteristics were also present, albeit to a lower extent, in nonsuccessful RCTs. However, the present findings are in accordance with previously published data regarding the effectiveness of school‐based interventions in promoting healthy lifestyle behaviors in Europe, according to which the most successful ones lasted for 1 school year, incorporated parental involvement, and were designed based on a theoretical model for behavior change [163].

Meta‐analysis of a subgroup of RCTs with common outcomes also revealed a significant but slight beneficial effect of PA interventions on change in BMI and BMI z‐score. This effect was generally greater in shorter compared to longer durations of follow‐up, suggesting a gradual attenuation of intervention effects and challenges in sustaining healthy behavior change [164, 165]. Our findings are in line with previous meta‐analyses of RCTs focusing on childhood OV/OB prevention, showing that PA interventions, compared with control interventions, can lead to significant but only slight reductions in BMI and BMI z‐score [17, 18, 19]. Intervention effects on other endpoints, such as change in BMI percentile and follow‐up BMI z‐score values, were nonsignificant, albeit the number of meta‐analyzed RCTs was small in both cases.

4.3. PA and Indicators of Childhood MUO Risk

Although the cardiometabolic benefits of PA are well‐established [166], longitudinal studies examining the etiological link between PA and childhood MUO in Western countries are scarce. The 4 studies (2 prospective epidemiological and 2 RCTs) included in the present systematic review were characterized by methodological limitations, had small sample sizes and do not support a significant association between PA and MUO‐related outcomes. Of note, out of the 4 included studies relevant to MUO, only 1 evaluated the MUO phenotype as an outcome, defined as the combination of OB with at least 1 metabolic risk factor (e.g., dyslipidemia, hyperglycemia or hypertension) [141], while in the rest 3 studies the association/effect of PA with/on individual cardiometabolic indices was explored in youth populations with OV/OB [142, 143, 144]. The paucity of research data, combined with the heterogeneity in MUO‐related outcomes, precludes a clear data synthesis and hampers the generation of robust conclusions. Given the significant health burden of excessive body weight that can persist through adulthood, leading to increased morbidity [167], future research should focus on OB‐related cardiometabolic disorders and their lifestyle determinants in childhood. In this context, high‐quality longitudinal studies with adequate samples, valid methodologies for PA assessment, and the use of a harmonized definition of MUO are needed to elucidate the role of PA in the etiology and prevention of childhood MUO and guide decision making for clinical practice and public health.

4.4. Strengths and Limitations

In the present work, the latest evidence of the association between PA and indicators of childhood OV/OB and MUO risk was synthesized using a systematic review approach. The research questions were carefully formulated based on the PECO/PICO models, while the methods and findings of the systematic review were comprehensively and transparently reported according to the PRISMA guidelines. In an effort to explore etiologic associations between PA and childhood OV/OB or MUO, focus was placed only on studies with a longitudinal design. However, some limitations of the present work must also be acknowledged. Because the search did not include gray literature or non‐English reports and was performed only in 2 databases, i.e., MEDLINE (PubMed) and Scopus), the possibility of missing eligible studies cannot be excluded; however, the selected databases provide comprehensive coverage of biomedical and public health literature, while an additional manual search of the reference lists of included studies and previously published systematic reviews was conducted to partially mitigate the risk of limited study retrieval. In addition, the vast majority of studies included in the present systematic review were of prospective epidemiological design, which is superior to the cross‐sectional one due to its longitudinal nature but still prone to reverse causation, meaning that a negative association between PA and OV/OB related outcomes might be distorted by the fact that the presence of OV/OB can hamper engagement in PA. Moreover, the focus of the present work was on studies conducted in Western countries, a fact that limits the generalizability of its findings in different sociocultural settings. Most importantly, the findings of the present systematic review are limited by the scientific quality of the included studies, most of which were of medium/high risk of bias, and the large heterogeneity in terms of exposures/interventions related to PA and outcomes related to OV/OB. Last but not least, the present work was unable to produce robust conclusions regarding the causal link between PA and childhood MUO due to the paucity and heterogeneity of the available data, and highlights this topic as an understudied field that should be addressed by future research.

5. Conclusions

The available data from studies with a longitudinal design are in favor of the beneficial effect of PA on indicators of OV/OB risk among children and adolescents. Engagement in moderate‐to‐vigorous PA, substituting sedentary time with PA, and longitudinally sustaining a high PA level over the course of childhood emerged as key objectively measured exposures showing beneficial associations with OV/OB‐related outcomes in epidemiological studies. Preventive strategies for childhood OV/OB should aim at facilitating the adoption of a physically active lifestyle in multiple settings through long‐term multicomponent PA interventions based on behavior change. Given the growing burden of the MUO phenotype among youths but the paucity of relevant research data, future studies should prioritize this research gap to elucidate the role of PA in childhood MUO and inform effective and targeted prevention strategies.

Author Contributions

Michael Georgoulis and Meropi D. Kontogianni: Conceptualization. Michael Georgoulis, Georgios Saltaouras, Eirini Bathrellou, Vasiliki Bountziouka, George Dimitrakopoulos and Meropi D. Kontogianni: methodology. Michael Georgoulis, Ismini Grapsa, Alexandra Karachaliou, Giannis Arnaoutis, and Vasiliki Bountziouka: formal analysis. Michael Georgoulis, Ismini Grapsa, Alexandra Karachaliou, Giannis Arnaoutis, Georgios Saltaouras, and Eirini Bathrellou: investigation. Michael Georgoulis, Ismini Grapsa, and Giannis Arnaoutis: writing—original draft. Alexandra Karachaliou, Georgios Saltaouras, Eirini Bathrellou, Vasiliki Bountziouka, Mary Yannakoulia, George Dimitrakopoulos, and Meropi D. Kontogianni: writing—review and editing. George Dimitrakopoulos and Meropi D. Kontogianni: supervision. All authors have read and agreed to the final version of the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: PRISMA 2020 checklist.

Table S2: Literature search query.

Table S3: Characteristics of prospective epidemiological studies exploring the association between PA and indicators of childhood OV/OB risk.

Table S4: Risk of bias in prospective epidemiological studies exploring the association between PA and indicators of childhood OV/OB risk.

Table S5: Characteristics of RCTs exploring the effect of PA interventions on indicators of childhood OV/OB risk.

Table S6: Risk of bias in RCTs exploring the effect of PA interventions on indicators of childhood OV/OB risk.

Figure S1: Leave‐one‐out sensitivity analysis for the mean difference in BMI change from baseline to follow‐up between intervention and control groups.

Figure S2: Leave‐one‐out sensitivity analysis for the mean difference in BMI z‐score change from baseline to follow‐up between intervention and control groups.

Figure S3: Forest plot for the mean difference in BMI percentile change from baseline to follow‐up between intervention and control groups.

Figure S4: Forest plot for the mean difference in follow‐up BMI z‐score values between intervention and control groups.

Table S7: Characteristics of prospective epidemiological studies exploring the association between PA and indicators of childhood MUO risk.

Table S8: Risk of bias in prospective epidemiological studies exploring the association between PA and indicators of childhood MUO risk.

Table S9: Characteristics of RCTs exploring the effect of PA interventions on indicators of childhood MUO risk.

Table S10: Risk of bias in RCTs exploring the effect of PA interventions on indicators of childhood MUO risk.

Georgoulis M., Grapsa I., Arnaoutis G., et al., “Association Between Physical Activity and Indicators of Overweight/Obesity and Metabolically Unhealthy Obesity Risk in Children and Adolescents: A Systematic Review of Prospective Epidemiological Studies and Randomized Controlled Trials in Western Countries,” Obesity Reviews 27, no. 1 (2026): e70014, 10.1111/obr.70014.

Data Availability Statement

The authors have nothing to report.