DE-PASS Best Evidence Statement (BESt): A Systematic Review and Meta-analysis on the Effectiveness of Trials on Device-Measured Physical Activity and Sedentary Behaviour and Their Determinants in Children Aged 5–12 Years

Physical inactivity contributes to the rising global obesity crisis and the accompanying risks for non-communicable diseases [1]. Despite the well-known positive effects of physical activity (PA) and the detrimental effects of sedentary behaviour (SB) on physical and mental health, the majority of children do not meet the PA levels recommended by the World Health Organization (WHO)—at least a daily average of 60 min moderate-to-vigorous PA (MVPA)[2‐5]. Additionally, starting from the age of 7 years, children become less active as they grow into adolescents, which may have a negative impact on their growth and maturation [6, 7]. The WHO recommends children limit the time spent in SB, particularly recreational screen time [3]. According to the guiding objectives of the European Workplan on Sport 2021–2024, there is an urgent need to put forth evidence-based policies to promote participation in sports and health-enhancing PA among children to combat the low levels of PA and high levels of SB [8, 9]. However, the factors behind children’s PA/SB engagement are still not sufficiently understood and translated well enough to inform policy makers [10, 11]. To inform policy and implementation of interventions in the real world, a first step is to identify modifiable determinants of PA/SB and prioritize the most effective ones to target [8, 12, 13]. Furthermore, since modifiable determinants may be linked to specific settings (e.g. home, school etc.), the settings of interventions must be considered when assessing their effect [8, 14]. Previous reviews identifying determinants and correlates of PA/SB highlighted issues related to overall low methodological quality in the measurement of outcomes, methodological variability relating to high levels of heterogeneity among studies, and the reliance mostly on self-report measures of PA/SB [15‐18]. By nature, PA/SB are diverse behaviours and are influenced by many types of context dependent determinants [8, 10‐12].

A large body of research has explored the correlates of PA/SB in children, mainly examining their associations in cross-sectional designs [16‐19]. However, to identify modifiable determinants of PA/SB for real-world interventions and policy, the causality between identified determinants and PA/SB, and the modifiability of the determinants need to be examined [8, 10, 11, 20, 21]. Prospective experimental designs, including randomized controlled trials (RCTs) and controlled trials (CTs), can test for causal associations between PA/SB and their determinants. Specifically, RCTs and CTs allow verification of the causal associations by testing manipulated exposures (determinants) and evaluating the effects on PA/SB while controlling for lack of exposure [11, 22]. As such, RCTs and CTs allow both testing of the level of modifiability of determinants and provision of robust evidence for causality between manipulation of determinants and change in PA/SB [8, 22]. The goal of the current review was to contribute to a Best-Evidence Statement (BESt) aimed at informing public policies and interventions based on high-quality evidence regarding the key modifiable determinants of PA/SB in specific settings that should be prioritized and targeted [23].

The settings in which interventions are conducted have received much attention in recent years [14, 24, 25]. To what extent a determinant can be modified to achieve change in PA/SB is highly dependent on the setting of the intervention. Settings refer to social and geographical contexts, which can either be adapted to facilitate change in PA/SB, acting as a modifiable determinant itself, or facilitate modifiability of other determinants to achieve change in PA/SB [14]. A settings-based approach can therefore help identify which determinants underpin PA/SB and unravel how multiple determinants may interact to influence PA/SB [24]. Determinants of PA/SB can be further contextualized from a social-ecological perspective [26]. Using the social-ecological model as a classification framework of determinants can help policymakers target them within their respective domains and settings (e.g., through school/organization) and thereby increase the likelihood of effective interventions. As such, the current review addresses the calls for a settings-based approach to identifying (modifiable) determinants of PA/SB to inform policy, both on intra-/interpersonal levels and societal/policy levels [8, 14, 21, 27].

Moreover, the methods of measurement of PA/SB (e.g. device-based and self-report measures) need to be considered. Different methods of measurement of PA/SB may not yield comparable results, which adds to methodological variability and inconsistency in results [28‐30]. For example, both device-based and self-report measures of PA have been found reliable and valid in children, but they do not measure the same aspects of PA [28, 31]. Studies comparing self-report and device-based methods for the measurement of PA in children have shown that self-report measures overestimate PA levels, particularly at higher intensities [30‐32]. Therefore, as device-based measures involve continuous accelerometer-based PA/SB tracking, they may be particularly useful in children as they do not rely on recall [30]. Additionally, focussing only on device-based measurement of PA/SB can help mitigate methodological variability, which can contribute to inconsistency in results.

Due to the complexity of PA/SB interventions, few reviews on the determinants (or correlates) of PA/SB in children have conducted meta-analyses, leaving a gap concerning the effectiveness of interventions on determinants of PA/SB [15, 33]. Additionally, previous meta-analyses have mainly focussed on one specific setting (e.g. school or family) and, to the authors’ knowledge, none to date have provided a comprehensive overview on the effectiveness of interventions across settings [33, 34]. Finally, the current review used Bayesian meta-analysis, which provides nuanced conclusions and realistic measures of heterogeneity and adjusts for publication bias [35]. Therefore, the current systematic review and meta-analysis aimed to identify modifiable determinants of device-measured PA and SB targeted in available intervention studies with RCT and CT designs in children and early adolescents (5–12 years old) and quantify the effects of the interventions within their respective settings on the determinants of PA/SB, and the outcomes PA and SB. More specifically, the aim was three-fold: (1) identify determinants of PA and SB targeted in existing interventions, and quantify the effects of the interventions (2) on the identified determinants and (3) on PA and SB.

The current systematic review and meta-analysis was part of a series of reviews investigating the effect of interventions on PA/SB and modifiable determinants of PA/SB, with a common search strategy. The methods for the review process were outlined in a pre-published protocol [23]. The review was pre-registered in the International Prospective Register of Systematic Reviews (PROSPERO; 12/10/2021; CRD42021282874). The reporting in the current systematic review and meta-analysis was guided by the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) [36]. The main outcome measures were device-based PA/SB [3] and modifiable determinants of PA/SB. Based on the WHO definitions, PA was defined as “Any bodily movement produced by skeletal muscles that requires energy expenditure” and SB was defined as “Any waking behaviour characterized by an energy expenditure of 1.5 METs or lower while sitting, reclining or lying” [3]. Modifiable determinants were identified by the context of each intervention, in which factors targeted for manipulation were hypothesized to have an effect on PA/SB [23].

A literature search was conducted in MEDLINE (Ovid), PsycINFO (EBSCO), Web of Science, SPORTDiscus and Cochrane Central Register of Controlled Trials (CENTRAL) on 12/09/2021 (see Table 1 for the search strategy). The search results were filtered for studies published from 2010, when the first global WHO PA guidelines were published, until September 2021, when the initial search was completed. An updated search was conducted on 12/07/2023 to identify studies published after September 2021 following the procedures outlined by Bramer and Bain [37]. The included studies were selected based on study design (RCTs and CTs only), measurement method for PA/SB (device-based only) and population (children and adolescents).

The screening, data extraction and risk of bias assessment were completed by a group of 26 reviewers, who were experienced researchers. The records resulting from the search were screened initially by one member of the review team to remove duplicates and any grey literature. The screening of each record by title and abstract and by full-text was completed in Covidence [38] by two blinded reviewers, with one additional reviewer to resolve any conflicts. Each reviewer screened approximately 2000 records at the title and abstract stage and ca 70 records at the full-text stage. A pre-piloted decision tree based on the inclusion and exclusion criteria was used for the screening to promote consistency among reviewers. Studies were included if they comprised a sample of children and/or early adolescents with a mean age within the range of 5.00–12.99 years (henceforth termed children), an RCT or CT study design, a device-based measure of PA and/or SB, modifiable determinants, measures of the outcomes included at least at two timepoints (pre- and post-intervention), and a control group. Studies were excluded if they included non-clinical populations, i.e. participants with diagnosed medical conditions known to affect the ability to engage in PA and/or patients undergoing treatment on all levels of care (e.g. studies including patients with cancer or individuals with anterior cruciate ligament injury or studies where the intervention takes place in a clinical setting). Studies were also excluded if they included participants with disabilities (i.e. impairments, activity limitations and participation restrictions, denoting the negative aspects of the interaction between an individual and that individual’s contextual factors). Additionally, studies published in languages other than English for which translation could not be obtained were excluded. Data extraction and risk of bias assessment were also completed in Covidence [38] by two independent reviewers, each completing approximately six records. Reviewers discussed any conflicts to resolve them, and a third reviewer cross-checked and resolved any remaining conflicts in both data extraction and risk of bias assessment. The overall characteristics of the included studies and numerical data for the outcomes for all time points were extracted for use in meta-analyses. Authors were contacted to supplement additional or unreported data. The extracted data included intervention description (design, intervention content, control activity, location), sample description (sample size, grouping, sex, age), outcome measures (targeted determinants, type of PA/SB measures, instruments used to measure outcomes), time frames (duration of intervention, time between measures and follow-up) and quantitative outcome data (measures of central tendency and variance for each outcome measure).

Modified versions of Cochrane’s tools for risk of bias tools were used—Risk of Bias in Non-randomised Studies of Interventions (ROBINS-I) for CTs [39] and Cochrane Risk of Bias Tool for Randomised Trials (RoB2) for RCTs [40]. The modified risk of bias tools included an additional domain each to assess the risk of bias in the measurement of the determinant(s). The quality of the produced evidence in each meta-analysis was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach in GRADEpro [41]. The GRADE assessment considers the results of the meta-analysis, their inconsistency, indirectness and imprecision, and the overall risk of bias of the included studies and yields a score on the quality of evidence provided by the meta-analysis.

The aims of the current systematic review and meta-analysis were to identify modifiable determinants of device-measured PA and SB targeted in available intervention studies with RCT and CT designs in children and early adolescents (5–12 years old) and quantify the effects of the interventions within their respective settings on the determinants of PA/SB, and the outcomes PA and SB. Overall, the results showed little effect of the interventions on PA, SB, and the modifiable determinants of PA/SB. The effects on the determinants were largely trivial-to-small with a few exceptions of moderate-to-large effects with moderate level of evidence for the observed effects, such as parental PA modelling in the family/home setting. The effects of the interventions on PA/SB were trivial-to-moderate, with the exception of moderate effect on the part-day measure of SB in school-based RCTs, with moderate level of evidence for the effect. Studies in the family/home setting showed relatively lower risk of bias compared to the other settings and a higher overall quality of evidence as indicated by GRADE. Although the interventions had little effects on the determinants of PA/SB and the PA and SB outcomes, the ones observed were found in RCTs but not in CTs, indicating that the higher level of methodological rigor associated with RCTs may have contributed to the outcomes.

In the school setting, the interventions showed mostly trivial effects on the determinants of PA and SB, and the PA and SB outcomes, except for the part-day measure of SB, which showed moderate evidence for the presence of an effect of ES = 0.58 and a moderate quality of evidence. The results in the school setting echoed previous reviews on the effects of school-based interventions on PA and SB, where a reduction in SB was observed but no increase in PA (MVPA, more specifically) [33, 92]. The school setting is particularly important for the promotion of healthy lifestyles as it is the setting in which children spend the most time outside of the family/home setting, in which they learn, socialize and have the opportunity to be physically active [93]. Several interventions in the school setting targeted the provision of PA opportunities and specific activities during school time (whether in class or during recess/breaks), or targeted teaching methods to include more PA elements in class (e.g., standing in class). An effect was found on part-day measured SB, but not any of the other PA/SB measures. The selection of whole-day measures of PA/SB in the current review was to enable comparison across studies, but it may be more important to consider outcome measures that correspond to the exposure within the interventions, such as a school-based intervention with a measure of PA/SB during school hours rather than a school-based intervention with a PA/SB measure that represents the whole day [10, 94].

In the family/home setting, the results showed moderate evidence for an effect of ES = 0.69 and a high quality of evidence for parental PA modelling. Only trivial evidence was found for the presence of any effects on PA and SB. The family/home setting showed consistently higher quality of evidence compared to the other settings. A previous meta-analysis on family-based interventions showed small significant effects of the interventions on PA [34]. A potential success of interventions conducted in the family/home setting alludes to an important role played by the immediate social environment, which largely involves parents and family members [26, 34]. Furthermore, it is plausible that interventions conducted in the family/home setting, particularly those including parents, could have a transferable effect to other settings [95]. For example, the determinants co-PA and parental PA modelling show promise and the recommended inclusion of parents in PA interventions in children echoes the results in previous literature [95]. The observed effects in the family/home setting in the current review are worth considering in future interventions. Finally, the community setting, and the combined settings showed no evidence for the presence of any effects on the determinants of PA or SB, likely due to a low number of included studies.

Regarding the theoretical underpinnings and contents of the interventions, a small number of theories were used as the bases for most of the interventions in the included studies. Social cognitive theory was the most cited theory (50% of included studies), of which self-efficacy and intention are central determinants of PA/SB [96]. Three studies did not mention a theoretical basis for their interventions, some of which were nevertheless still consistent with some of the well-established theories, targeting for example self-efficacy, perceived athletic competence, social support and PA outcome expectancies [67, 81, 85]. Four studies that did not mention a theoretical basis targeted motor competence and muscular/aerobic fitness, aiming to improve health outcomes along with increased PA/reduced SB [64, 72, 73, 97]. It has been argued that consistent small effects observed in the literature may reflect a failure of interventions to capture a wider array of determinants, and combinations of theories may lend a more comprehensive approach to interventions aiming to change PA/SB [8, 20]. It has been suggested that multilevel interventions based on the ecological approaches may provide a more comprehensive approach [26]. Ecological approaches, such as the social ecological model, emphasize the interaction between the individual and their social and physical environments and highlight that external factors can influence individual tendencies to engage in PA/SB, such as built environment, provision of safe outdoor areas, and provision/improvement of organized PA [26, 98, 99]. Three interventions in the included review were based on ecological theories, including ecological systems theory by Carson et al. [59] and socioecological model by Cohen et al. [62, 63]. As such, multilevel interventions based on ecological models, targeting determinants at individual and social and physical environment levels, warrant further exploration for future interventions. Emerging research in recent years has deviated from the largely social-cognitive approach with a focus on the individual in PA behaviour regulation (automaticity and habit) expanding on and challenging the current approach [100‐103]. Three of the included studies [82, 86, 88] touched on such concepts and for a better understanding of them, further research is warranted.

The overall quality of evidence in the current review was assessed as low, with the exception of the family/home setting, which showed an especially high quality of evidence. The main components contributing to the low quality of evidence were the imprecision of studies, which denotes a wide variance among effect sizes (large credible intervals), i.e. low confidence in the results generated. Large credible intervals can be attributed to small sample sizes within studies and a small number of studies in a meta-analysis [104]. The small number of studies included in the meta-analyses may have been the main reason for the low confidence in the generated evidence.

In response to the quest of the European workplan on sport for evidence-based sport policies, the goal of the current review was to summarize and evaluate the available evidence on PA/SB and the modifiable determinants of PA/SB in children to inform a BESt to be used by policy makers and future researchers alike [9]. A strength of the current review was the use of the Bayesian approach in the meta-analyses, which provides a nuanced interpretation of the results, and particularly RoBMA, which yields publication bias-corrected effect sizes. Meta-analyses based on the frequentist statistical approach were also conducted for the current review and have been reported in the Supplementary file 1—Table S1. It is important to note that the results of RoBMA and the frequentist meta-analyses are not comparable due to the difference in inference methods and the publication bias adjustment in RoBMA. Another strength of the current review was the attention to the quality of the produced evidence, which is key for the interpretation of the results and translation of the evidence to policy (e.g. BESt). A limitation in the current review was related to the exclusive focus on modifiable determinants limited the breadth of the current review, as additional studies may have been included that target non-modifiable determinants. However, since the current review had the goal to provide evidence for future interventions and policy, determinants were sought that were modifiable and thus possible to manipulate in interventions. Furthermore, studies including clinical populations were excluded from the current review. Among the included studies, there were studies including participants with overweight and/or obesity (the whole sample or parts of the sample) that did not explicitly disclose any medical conditions. However, it could be that participants with overweight/obesity may have unknown medical conditions that affect their ability to engage in PA. Additionally, participants with overweight/obesity may be more motivated to take part in PA/SB interventions and, as a result, may respond particularly well to the intervention [105]. Such participant characteristics must be taken into account when interpreting the results of the meta-analyses. Moreover, the current review targeted RCTs and CTs as having robust designs that produce a high level of evidence and can provide indication of causality due to repeated measures of the determinants and the PA and SB outcomes. However, due to the controlled nature of RCTs and CTs, their application in the real world may be difficult and their chosen methods may not be ecologically valid. Therefore, the application of PA and SB interventions would require accurate translation to policy, adaptation of methods to target environments and theory-informed methods for implementation in their respective settings [106]. Finally, the studies with both PA and SB as outcomes did not distinguish between determinants of PA or SB, but rather aimed to manipulate their targeted determinants and measured PA and SB simultaneously. As such, it was not possible to make any distinctions between determinants of PA and SB in the current review.

Taken together, the results of the current systematic review and meta-analysis suggest that the included interventions have not been effective in changing PA/SB or the determinants of PA/SB overall and the few changes shown were not sustained over time. Parental PA modelling in the family/home setting indicated a moderate effect and a moderate level of evidence, which may suggest that parental involvement may be key to increasing PA/reducing SB in children. The part-day measure of SB in the school setting also indicated a moderate effect and a moderate level of evidence, which may suggest that an outcome measure (school-based measure) that corresponds to the exposure (school-based intervention) may be an important consideration. Despite the lack of effects overall, it seems that the family/home setting shows promise with consistently higher quality of evidence compared with the school and community settings and the combinations of settings, which is an important finding. The current review also helps uncover the breadth of methodological variability and aspects of PA research in children which need to be addressed in future research, relating to the selection of both determinants and behaviour change techniques. The results of the current review shed light on the continued need for refined theory-based interventions and the continued need to deepen our understanding of how determinants underpin PA/SB and provide a relevant basis for evidence-based policies and interventions. Gaps were identified that relate to determinants on the physical environmental level, the integration of determinants based on an ecological approach and the emergence of research on automaticity and habit of PA behaviour on the individual level that warrant further investigation.