Discussion
This systematic review and network meta-analysis summarised data from 51 cluster randomised controlled trials comparing different nutritional interventions in school settings on multiple anthropometric and quality of diet outcomes in children and adolescents. For the primary outcomes, we found that nutritional interventions had little or no effect on body mass index, body mass index z score, body weight, body fat, or waist circumference compared with a control group. The results for specific nutritional interventions suggested that multicomponent interventions might reduce the prevalence and incidence of overweight compared with a control group; a multicomponent intervention was also found to be possibly more effective in reducing body mass index and likely more effective in reducing body fat than nutrition friendly school initiatives. For reducing body mass index z score, we found moderate certainty evidence that nutrition education and literacy is likely more effective than a control group. Likewise, moderate certainty evidence suggested that nutrition friendly school initiatives are likely more effective than a control group and multicomponent interventions in reducing waist circumference.
Findings for the secondary outcomes suggested that nutritional interventions were more effective than a control group for intake of fruit and vegetables, alone and combined, with no difference between groups for intake of fat and sugar sweetened beverages. For specific interventions, nutrition education and literacy and multicomponent interventions may be more effective than a control group for improving combined intake of fruit and vegetables. Findings with a low certainty of evidence further suggested that nutrition education and literacy and multicomponent interventions may be more effective than a control group for improving intake of fruit and that nutrition education and literacy may result in a larger increase in intake of vegetables than a control group.
Comparison with other studies
This is the first network meta-analysis on the effects of different nutritional interventions in the school setting on anthropometric and quality of diet outcomes in children and adolescents. A recently published network meta-analysis focused on nutrition, physical activity, and lifestyle interventions for the treatment of childhood obesity rather than prevention of obesity in school settings.123 The meta-analysis reported that nutritional interventions and comprehensive approaches with parental involvement were superior to no intervention in reducing anthropometric measures (ie, body mass index, body mass index z score, percentage body fat, or percentage overweight), which partly matches our results. Involving parents can have a positive effect on the nutritional environment (eg, in school canteens).124 125 Parents can bring about changes by influencing school management in parent-teacher conferences. On the other hand, a link between parental obesity and weight gain in their children exists,126 and comments by parents (even if well intentioned) about children’s efforts towards healthy weight management can be counterproductive.127–129 These concerns should be taken into account when implementing measures involving parents.
Consistent with our findings, a meta-analysis35 reported a small reduction in body mass index in adolescents after interventions in a school setting compared with a control group. A meta-analysis23 of nutritional interventions in a school setting for improving the eating habits of primary school children found that, compared with a control, experiential learning strategies (ie, school garden, cooking and food preparation activities), cross curricular approaches (ie, learning experiences delivered in ≥2 learning areas or subjects), and approaches based on the curriculum (ie, nutrition education programmes), were associated with medium to large effects for improving intake of fruit and vegetables, supporting our findings that nutritional interventions may be more effective than a control group for increasing both combined and separate intake of fruit and vegetables.
In another meta-analysis,24 all types of nutritional interventions in school settings were estimated to improve children’s daily intake of fruit and vegetables by an average of 0.25-0.33 portions (corresponding to a daily increase of 20-30 g) compared with a control group. Multicomponent programmes were found to be more likely to result in greater improvements in intake of fruit and vegetables than single component programmes, which also agrees with our results. Also, a meta-analysis30 on the effect of school food policies on dietary habits and obesity in children reported that direct provision of food and beverages increased daily intakes of fruit and vegetables (combined and separate) compared with a control group. Nutritional quality standards for school meals were also found to increase intake of fruit and reduce intake of total fat.In contrast with our results, no improvements were seen in the prevalence of overweight and obesity combined, overweight, body mass index, or body mass index z score compared with a control group. Similar to our study, conflicting qualitative findings were found for intake of sugar sweetened beverages. In contrast, another meta-analysis130 of food environment interventions in a school setting reported a small reduction in body mass index z score and small increases in intake of fruit, but no differences in intake of vegetables. Although we also found only a small decrease in body mass index z score, our findings showed larger increases in intake of fruit and vegetables.
Clinical and research implications
Excessive weight gain at an early age is associated with physiological and psychological problems in the subsequent course of childhood and adolescence, and has a considerable financial burden on the public health system. Moreover, childhood obesity increases the risk of non-communicable diseases, such as cardiovascular diseases in adulthood.7 8 11 Preventing or reducing overweight and obesity in children and adolescents is therefore critical for decreasing the risk of cardiovascular disease and the risk of developing non-communicable diseases. Also, research has shown an inverse relation between higher intake of fruit and vegetables and adiposity among children who are overweight,131 and a harmful association between unhealthy diets rich in sugar sweetened beverages and fat and the risk of overweight and obesity.132 Schools are important settings for shaping and promoting lifelong healthy eating habits in children and adolescents, and can provide important opportunities for prevention of overweight and obesity through health and nutrition programmes in the school setting.
That environments where children and adolescents spend time offer good opportunities to promote healting eating habits, has been confirmed in a meta-analysis133 that synthesised data on intervention strategies to promote healthy meals in restaurants and canteens. The most prominent improvements for intake of healthy food groups were found in studies in children.133 The school environment is therefore a suitable setting for implementing these strategies. Also, the availability of healthy items in school canteens is associated with an increased willingness of children and adolescents to buy these food groups.134 According to the WHO Global Nutrition Policy Report, 142 of 160 countries (89%) implemented healthy diet and nutrition programmes in 2016-17, although implementation has generally declined in recent years.135 Comprehensive or multicomponent nutritional interventions were also rarely implemented.135 The Commission on Ending Childhood Obesity strongly recommends implementing comprehensive programmes that promote the intake of healthy foods and reduce the intake of unhealthy foods, create healthy school environments, and promote health and nutrition literacy in school aged children and adolescents.136 Our findings support these recommendations, because we showed that the effects on anthropometric and quality of diet outcomes differed across single and multicomponent nutritional interventions. Although beneficial effects were seen with some single component interventions (eg, nutrition friendly school initiatives), nutrition education and literacy as well as multicomponent interventions mostly ranked highest.
Factors that contribute to childhood overweight and obesity are complex and multifaceted and require a whole system approach, targeting multiple stakeholders and environments to drive behavioural change (eg, nutrition education at the individual, family, community, and school levels). Thus developing multicomponent interventions is essential and should ideally involve multidisciplinary teams with participation of all relevant stakeholders (eg, parents, schools, and municipalities), including experts in nutrition education and didactics.
Strengths and limitations
Our systematic review and network meta-analysis has several strengths and limitations. The strengths include the comprehensive and rigorous literature searches in multiple electronic databases and trial registries, a priori published protocol, network meta-analysis methodology incorporating direct and indirect evidence to compare and rank interventions that have not been previously compared, detailed risk of bias assessment with the new risk of bias 2 tool, extensive subgroup and sensitivity analyses, and the GRADE framework for assessing the certainty of the evidence.
Our results were limited by the exclusion of interventions that combined both nutrition and physical activity or other non-nutritional components, and which were implemented in non-school settings (eg, after school settings). Findings on intake of fruit and vegetables (separately and combined) were limited by large statistical heterogeneity. Also, in nearly all of the networks, interventions were compared with controls, resulting in little or no direct comparative evidence for the different nutritional interventions. This limitation also prevented an assessment of inconsistency because of the lack of both direct and indirect evidence for pairs of nutritional interventions. Also, reporting of outcomes was inconsistent across studies; for example, anthropometric outcomes, such as body mass index, were reported less often and some studies did not report combined intake of fruit and vegetables. Hence selective reporting of outcomes cannot be excluded and might have influenced our analyses. These factors contributed to the small number of trials for many comparisons and might explain why pairwise meta-analyses for most primary outcomes showed little to no effects, with wide confidence intervals. Similarly, most studies did not report baseline data for the prevalence of obesity, restricting the interpretation of findings. However, because the prevalence of obesity was relatively low (<22%), we did not assume that the other studies included >30% of children with obesity.
Most cluster randomised controlled trials differed in study length (range 1-68 months) and lacked longitudinal follow-up data, limiting interpretations of the long term effects of different nutritional interventions. Future research should include well designed (cluster) randomised controlled trials assessing the long term effects of nutritional interventions in the school setting with more rigorous reporting of study characteristics and findings. The studies included in the meta-analyses used different instruments to assess dietary intake outcomes, including 24 hour dietary recalls, food diaries, and food frequency questionnaires on one or multiple days, which might explain some of the observed heterogeneity. Subgroup analyses on sex, socioeconomic status, and migration background could not be conducted in our network meta-analysis, and children and adolescents with differences in these characteristics might respond differently to nutritional interventions; we did not consider subgroup analyses of other factors that might have influenced the results, such as school year, baseline overweight status, and different definitions of overweight and obesity. Hence future research efforts should investigate interactions between nutritional interventions and sex, socioeconomic status, background, and school year.
Finally, many of the interventions (ie, multicomponent interventions) included in our review involved multiple nutritional components which are likely to have had synergistic effects but could also not have been similar enough (in components, content, or extent of implementation) across studies to be combined into one (multicomponent intervention) group. Component network meta-analysis was not possible in this review, however, and only limited conclusions can be drawn about the effects of the individual nutritional components and their combined effect in multicomponent interventions. This problem is not limited to nutritional interventions. Comparable difficulties in assessing the effectiveness of individual strategies in multicomponent interventions were reported137 in a systematic review of studies investigating the promotion of physical activity during school recess in children and adolescents.
Conclusion
Nutritional interventions in school settings showed beneficial effects on reducing the risk of overweight and on increasing the combined and separate consumption of fruit and vegetables. Future studies should distinguish between the effects of individual strategies within multicomponent interventions so that synergies can be better recognised and implemented in holistic measures. Future studies should also include process evaluations and cost effectiveness analyses of interventions, which could be of interest to policy makers in countries where resources are scarce.138 139 The results of our network meta-analysis could be of interest to public health authorities and policy makers worldwide in developing and implementing effective, evidence based nutritional intervention strategies in school settings.