The review was conducted following recommended methods outlined by the Cochrane handbook (37). This is a secondary analysis of studies included in an existing Cochrane systematic review (32) of the effectiveness of school-based physical activity interventions for all children aged 6 to 18 years including studies published up to 1st June 2020. The review methods were registered prospectively via PROSPERO (CRD42022296549) and are reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for reporting (38).
Search
All studies included in the original review (32) that were identified via the implementation of a comprehensive search strategy were assessed for eligibility. The original review searched following electronic databases MEDLINE, EMBASE, PsycINFO, Cinahl, SportDiscus, BIOSIS, Sociological Abstracts and Cochrane Central Register of Controlled Trials (CENTRAL), using terms related to the condition (physical activity), context (school), population (school-aged children) and study design as were the reference lists of included studies. All included studies and their associated records identified in the original review were screened as per the updated eligibility criteria for this review (see below). Additional searches were also conducted to identify any other potentially eligible associated records for each study that may have reported results separately by gender. The authors of all studies included in the previous review were contacted via email to obtain any associated publications and unpublished data, including data reported separately for girls.
Eligibility criteria: Studies assessing the effects of school-based physical activity interventions in students aged 6 to 18 years were considered eligible for inclusion in the original review if they: employed a randomised controlled trial design, focused on the general population of school students, and included a measure of physical activity (including MVPA, sedentary time or the proportion of children meeting physical activity guidelines) measured via accelerometers or pedometers across the whole day or school day. All included studies from the original review were considered eligible for this secondary analysis (32) and were screened against the updated criteria for participants and outcome (described below).
Participant/population
The original review included studies in children aged 6 to 18 years who attended school, irrespective of their health status, gender or weight status. Studies were excluded if participants received a physical activity intervention as part of a treatment regimen for a specific illness or comorbidity. For this secondary analysis, only studies that reported intervention effects separately for female school students aged 6 to 18 years were eligible (32).
Interventions/exposures
The original review included any school-based physical activity intervention aimed at increasing physical activity and/or fitness in children and adolescents. School-based physical activity interventions included educational, counselling, health promotion or management strategies that focused on promoting physical activity and/or fitness. A range of interventions were included, such as a change to the school curriculum, teacher training regarding physical activity within the school curriculum, and educational materials for teachers, parents, and students. The primary focus for the interventions needed to target the school setting. Included studies also needed to be capable of being implemented by staff in local public health units or community settings, rather than physician or clinic based (32).
Comparator/control
Eligible comparison groups in the original review was no intervention, usual care, or another type of intervention which had to be the same in both intervention and comparator groups to establish fair comparisons.
Outcomes: To be eligible for the original review (32), studies had to report one or more of the following primary outcomes, presented as post-intervention measurement and standard deviation or confidence intervals, or as a change from baseline with standard deviations or confidence intervals: 1) MVPA, assessed via accelerometers or pedometers during school time or whole day or both or 2) Sedentary time, measured as time spent sedentary in total minutes or hours per day or per week, measured via accelerometers or pedometers. 3) Proportion of students meeting physical activity guidelines as reported or calculated by dividing the number of students engaged in 60min/day of MVPA by the total number of students allocated to either the intervention group or the control group. For this secondary analysis, only studies that report these outcomes separately for girls were eligible. Secondary outcomes included any adverse effects reported for girls.
Study selection process
Two review authors (NM and BM) independently assessed the eligibility of all full-text articles included in the original review against these eligibility criteria for this secondary data analysis. A third review author (AH) resolved any discrepancies.
Data extraction and management
The pre-piloted data extraction form used in the Cochrane review was adapted in this current review. Data extraction occurred independently by pairs of review authors (AS, BM, ERG, CG), with a fifth review author (NM) resolving any discrepancies. Information regarding study characteristics (design, participants, interventions, outcomes) from included studies were extracted from the original review, supplemented by additional information that was extracted specifically for female participants, including sample, outcome data and information to assess the risk of bias.
Data analysis and synthesis
Where possible random-effects meta-analyses were undertaken to estimate a pooled effect size for the primary outcomes (minutes of MVPA across the whole/school day, minutes of sedentary time across the whole/school day, and percent meeting PA guidelines). As all measurements were reported in consistent units (i.e., minutes of physical activity), the pooled mean difference for MVPA and sedentary behaviour was calculated using the inverse variance method. For outcomes where too few studies (2 or less) were included to enable meta-analysis, estimates were summarised separately by individual study. Between-group differences at follow-up were prioritised for synthesis as the main effect of interest, however change from baseline was included if between-group differences at follow-up were not reported. As we included only randomised trials and standardised effect sizes were not calculated, it was appropriate to pool both follow-up and change scores (39). Both cluster and individual level randomised trials were pooled in the analysis. We applied a design effect to the standard errors as recommended in the Cochrane Handbook for cluster-level trials that did not appropriately adjust for clustering in their analysis. Where the intra-class correlation coefficient (ICC) for the design effect could not be obtained from the original study (i.e., from the sample size calculation or results section), an ICC value of 0.01 was used, consistent with the methods used in the original review (32). For studies that reported multiple intervention arms, where feasible we combined the intervention effect into one composite effect estimate to allow inclusion in meta-analyses, as recommended in the Cochrane Handbook (37) following recommended formula (40, 41).
Subgroup analysis and investigation of heterogeneity: Statistical heterogeneity was explored using I-squared statistics from the random effects meta-analysis. Heterogeneity was also explored in two pre-specified sub-groups: 1) age – primary (children aged 6 to 12 years) and secondary (aged 12 to 18 years), as physical activity levels are shown to decline as children age, particularly in girls (42); and 2) intervention type (classified according to original review, a) Multi-component whole-school interventions that included a range of whole-school strategies, such as changes to the school environment or curriculum. b) Interventions focused on general school time physical activity promotion, providing opportunities to increase physical activity within school time, such as active academic lessons. c) Interventions that looked at enhancing PE whereby the frequency, duration, or intensity of regular PE classes was increased. d) Other interventions provided opportunities for increasing physical activity during before or after school activities, such as walking or dance groups). As the original review found some differences by intervention type (32), we explored whether such differences persist for female students alone. Sub-group analyses were only undertaken if there was a minimum of two studies per sub-group. Differences in effect estimates across sub-groups were explored using statistical comparisons based on the Q-test.
Missing data
Where possible missing data were estimated using other data reported in the original study (e.g., 95% CIs used to estimate missing standard deviations).
Risk of Bias: Consistent with the original review, we used the Cochrane risk of bias tool version 1.0 (RoB 1) to determine the quality of each study. Risk of bias assessments reported in the initial systematic review for the following categories were extracted and reported: sequence generation, allocation sequence concealment, blinding of participants and personnel and other biases. The following outcome-specific domains were re-assessed for this secondary data analysis based on extracted outcomes: blinding of outcome assessment, incomplete outcome data, and selective outcome reporting. Two review authors (NM & ERG) independently assessed the risks of bias in the included trials and summarised them in tabular form, for one study where NM was involved two review authors not involved (ERG and AS) conducted the RoB assessment. A third review author (AH) was consulted to resolve any discrepancies between review authors. Single studies were defined as having a low risk of bias when all but one domain was assessed at low risk of bias or two domains if one was “blinding of participants and personnel” due to the difficulty in doing this well in school-based physical activity interventions. Single studies were defined as being at high risk of bias when at least three domains were assessed as having an unclear or high risk of bias, or four domains when one was “blinding of participants and personnel.” (32)
Sensitivity analysis
Robustness of estimated results were assessed via sensitivity analyses by removing studies classified as high risk of bias.
Quality Assessment of the evidence
The Grading of Recommendations, Assessment, Development and Evaluations (GRADE) approach was used to assess the overall certainty of evidence for each primary outcome measure. GRADE takes into account issues related to both internal and external validity to state how confident we are in the effect estimates presented. Two review authors (NM and AH) independently rated the certainty of evidence for each outcome, including the RoB, inconsistency, indirectness, imprecision, and publication bias. Differences were resolved by discussion and checked by a third review author (RKH). For each outcome, we rated evidence certainty as very low, low, moderate, or high based on the GRADE domains as described in the Cochrane Handbook (43). All studies started at a high level of certainty, as they were randomised controlled trials (RCT’s) and were then downgraded based on the five GRADE domains.