Forty-three studies (36,579 children) were included in the review: 23 cluster randomised controlled trials (RCTs), 15 RCTs, two quasi-RCTs and three non-RCTs.
The pooled estimate of BMI change was -0.17kg/m² (95% CI -0.26 to -0.08; 43 studies, 60 comparisons). Significant statistical heterogeneity was found (Ι²=93.4%).
When stratified by gender, a significant reduction in mean BMI change was found for girls (-0.28kg/m², 95% CI -0.50 to -0.06) but not for boys. When intervention type was considered, a significant reduction in change in BMI was found for physical activity alone (-0.13kg/m², 95% CI -0.22 to -0.04) and in combination with nutrition (-0.17kg/m², 95% CI -0.29 to -0.06). Interventions delivered to all pupils produced a change in BMI of -0.16kg/m² (95% CI -0.25 to -0.06; 54 comparisons). Interventions delivered only to overweight/obese pupils produced a change in BMI of -0.35kg/m² (95% CI -0.58 to -0.12; six comparisons). Statistical heterogeneity was not reported for these results.
None of study characteristics assessed were found to be significant predictors of overall effect and no significant relationship between length of follow-up and effect size was found. The earliest studies were found to have the greatest effect size but the pooled estimate had been stable since 2007.
The authors reported that there was no evidence of publication bias.