Effectiveness of school‐based interventions to prevent obesity among children aged 4 to 12 years old in middle‐income countries: A systematic review and meta‐analysis

Economic and accompanying nutrition transition in middle‐income countries is resulting in rapidly increasing childhood obesity prevalence, exceeding acceleration rates in the West. Previous school‐based obesity prevention reviews have mainly included studies from high‐income countries. This review aimed to summarize the evidence from randomized controlled trials evaluating the effectiveness of school‐based interventions in preventing childhood obesity in middle‐income countries. Six electronic databases were searched: MEDLINE, EMBASE, CINAHL Plus, LILACS, IBECS and WPRIM. Eligibility criteria included middle‐income country setting, randomized/cluster‐randomized controlled trials, children aged 4–12 years and school‐based interventions targeting dietary intake and/or physical activity. Twenty‐one cluster‐randomized controlled trials, conducted in Asia (n = 10), South America (n = 4), North America (n = 4) and the Middle East (n = 3), were included. Fifteen studies reported a significant intervention effect on at least one adiposity‐related outcome. Characteristics of effective interventions included combined diet and PA interventions, school teacher‐delivery, duration of >8 months, parental involvement, education sessions and school food modifications. The risk of bias in these trials was mixed. The pooled estimate of the odds ratio for obesity in intervention versus control schools (nine studies) was 0.77; 95% CI, 0.63 to 0.94; p = 0.009. In conclusion, there is some evidence to support school‐based interventions in preventing childhood obesity in middle‐income countries.


| INTRODUCTION
Childhood obesity is an international health concern affecting all socio-economic groups and ethnicities, and the worldwide prevalence has risen dramatically over the last four decades from 4% in 1975 to over 18% in 2016. 1 If current trends continue, researchers have predicted that childhood obesity will surpass moderate and severe underweight by 2022. 2 Middle-income countries (MICs) are of particular interest at the current time because these countries are undergoing rapid economic transition. 3 Alongside this MICs are experiencing a 'nutrition transition': a corresponding shift in energy intake and expenditure behaviours. In contrast with high-income countries (HICs), where economic growth and accompanying nutritional changes have occurred over a long period of time, the transition in MICs is occurring at an accelerated rate. Consequently, the prevalence of obesity in adults and children in MICs has rapidly increased in the last two decades, whereas the increase previously seen in HICs occurred at a slower rate, with a plateau in prevalence now being observed in these countries. [4][5][6][7][8][9][10] There are differing contextual factors to take into account when considering the rapid rise in childhood obesity in MICs. The relationship between socio-economic status and childhood obesity in MICs is positive, with increasing wealth associated with higher prevalence, whereas in HICs, the higher childhood obesity prevalence is associated with deprivation. 11,12 Also in MICs, rapid urbanization, industrial modernization, adoption of new technologies and the increasing influence of mass media contribute to the increased access to and consumption of energy-dense foods as well as reductions in physical activity (PA). 6,13 The approach to noncommunicable disease prevention also differs in these settings, with lower prioritization of health promotion initiatives to address noncommunicable disease risk factors by governments, and less focus on health promotion policies in schools. 14 These particular contextual challenges and the current positioning of MICs within the obesity epidemic curve make them an important focus for childhood obesity prevention at the present time. 9,15,16 In 2011, the World Health Organization (WHO) adopted the target of halting the rise in global obesity by 2025. 17,18 Since this time, there has been much activity relating to the development and evaluation of obesity prevention approaches, with a large majority targeting children. [19][20][21][22] Schools are frequently the focal point of obesity prevention efforts because most children spend a substantial proportion of their day in school. Furthermore, schools offer a natural setting for education, with the potential to reach parents as well as children, and teachers are often seen as role models. [22][23][24] Hence, school-based interventions have the ability to reach many children and families through a contained environment that provides sustained contact.
A Cochrane review published in 2019 presented some evidence to suggest that combined diet and PA interventions in schools resulted in lower body mass index (BMI) z-score in children aged between 6 and 12 years (mean difference of −0.05, 95% CI, −0.10 to −0.01; meta-analysis of 20 randomized controlled trials (RCTs), low certainty of evidence). 25 Other systematic reviews have reported similar findings. [26][27][28][29][30][31] However, in general, there has been substantial heterogeneity among the studies included in these reviews, and the majority have been conducted in HICs, so the applicability of the findings to MICs setting is limited. The most recent systematic review of school-based studies focusing on low-and middle-income countries was undertaken by Verstraeten et al, 32 before most programmes prompted by the WHO targets commenced. The authors concluded that school-based interventions have the potential to prevent obesity but acknowledged that a number of methodological limitations prohibited firm conclusions being drawn. Other systematic reviews linked to MICs have addressed one region only. 33, 34 Feng et al 33  interventions. 34 Again, the small volume of trials, methodological issues and heterogeneity of interventions limited firm conclusions being drawn from this review.
As yet there has been no comprehensive review of the effectiveness of school-based interventions for the prevention of childhood obesity across all MICs. Moreover, methodological rigour of trials evaluating school-based interventions in MICs has improved since the Verstraeten review in 2012, and so it is the appropriate time for synthesis of this more robust evidence. Therefore, the aim of this systematic review is to summarize the evidence from RCTs evaluating the effectiveness of school-based interventions in preventing childhood obesity in MICs and thus inform future school-based interventions to prevent childhood obesity in these countries. 35 2 | METHODS

| Protocol and registration
The protocol for this systematic review was registered on the International Prospective Register of Systematic Reviews (registration number CRD42019128058) and has been created according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. 36 2.2 | Eligibility criteria

| Population
The participants of interest were primary school children aged 4 to 12 years. Studies comprising populations of a wider age range were eligible for inclusion where the majority (>75%) of participants fulfilled the specified age range. Due to the variation in school systems and age of school entry between countries, the specified age range was selected to allow inclusion of all children in primary, or equivalent, education. Participants were children who reflected the weight status profile within the general child population. Studies were excluded if the target population was children with excess weight only as these were considered to be trials of obesity treatment rather than prevention.
Studies in which the target population was predominantly formed by children with co-morbidities or critical illnesses were also excluded.

| Intervention
Interventions targeting dietary intake and/or PA were included, irrespective of other intervention components. Eligible interventions were delivered through primary schools or equivalent settings but may have included components based within the family or community.
No restrictions were placed on the intervention duration or the personnel delivering the intervention. Interventions that had been specifically designed for the treatment of childhood obesity were excluded.

| Comparator
Studies were included where the control group received either no intervention, usual school-based activities relating to health education or an intervention with no specific diet or PA content.

| Outcomes
The primary outcomes comprised BMI, BMI z-score, skinfold thickness, body fat percentage, waist circumference, waist: height ratio, prevalence of overweight, prevalence of obesity and other measures of obesity based on weight and height. The secondary outcomes considered were measures relating to PA, dietary intake and quality of life.

| Study design and setting
All relevant RCTs or cluster RCTs conducted in MICs were eligible for inclusion. MICs were defined according to the World Bank classification of countries by income level 2018-2019. 37 Studies reported as full-texts only were included; published and unpublished studies (identified through published protocols) were eligible, but the latter were included where results were made available upon contact with the authors within a 2-week timeframe.

| Search strategy
The following electronic databases were searched: MEDLINE (OVID

| Study selection
Literature search results were imported into the referencing software, Mendeley Desktop, to aid deduplication and organization of references for screening. Titles and abstracts of the studies were screened by JS. A random sample of 5% of the studies, generated via the random number function in Microsoft Excel, was independently screened by another review author (CH or MP). Full-texts of potentially relevant studies were assessed for inclusion independently by two reviewers (see Data S1). Middle-income setting was assessed at the abstract screening stage. If the study setting was unclear from the abstract, the study was included in the full-text assessment stage. Disagreements at the abstract and full-text screening stages were discussed within the author team, and consensus was reached in determining eligible studies. The PRISMA flow diagram is shown in

| Data extraction
Data were extracted into data extraction forms, which had been designed and piloted a priori, shown in Data S1. The data of interest comprised location, study design, setting, duration, number of trial arms, number of participating schools, recruitment process, randomization method, participant characteristics, intervention (including personnel, duration and parental involvement), control, primary and secondary outcome data at baseline and follow-up time points, acknowledgement of adverse outcomes, response rate, attrition, method to deal with losses to follow-up, method of statistical analysis and a summary of study findings. This was carried out by JS for all studies, and each form was verified by one other reviewer (CH or MP). Missing data were sought, where possible, by email contact with study authors. Any discrepancies were identified and resolved through discussion within the author team. For the non-English studies retrieved, translators with the relevant language skills provided assistance with the data extraction.

| Quality appraisal
The quality of the included studies was assessed independently and in duplicate using the Cochrane Collaboration's Tool for Assessing Risk of Bias, which was modified for use with cluster RCTs. 38,39 Any disagreements were resolved through discussion between the two review authors, and where necessary, consultation within the entire author team. Due to the nature of the intervention, performance bias was not considered because it is not possible to blind study personnel or participants. The additional risk of bias assessments for cluster RCTs included the following: recruitment bias, analysis bias, baseline imbalance and loss of clusters. A judgement of either 'low risk' , 'high risk' or 'unclear risk' was assigned for each specific domain along with a support for the judgement. The results of the risk of bias assessments were exported to Review Manager 5.3 software to create a visual representation. 40

| Synthesis of results
Trial outcomes expressed as mean differences, risk ratios and odds ratios were extracted. Study findings have been presented by outcome measures and grouped based on the following intervention characteristics to provide a narrative synthesis: type of intervention, F I G U R E 1 A PRISMA flow diagram outlining the identification, screening, eligibility and inclusion of studies in this review. Only studies for which estimates of effect were adjusted for clustering and reported in the appropriate format were eligible for inclusion in the meta-analyses duration, personnel delivering the intervention and parental involvement. These specific intervention characteristics were evaluated because previous reviews indicate that these are important variables influencing the effectiveness of school-based interventions. 23,41,42 When it was deemed appropriate, based on the assessment of the clinical and methodological homogeneity of the studies, a metaanalysis was conducted to support the narrative synthesis. Metaanalyses of adiposity-related outcomes that had been reported in an appropriate format by five or more studies were carried out using the Review Manager 5.3 software. 40 Only studies for which estimates of effect were adjusted for clustering were included in the meta-analyses. We organized the forest plots by intervention type with the following subgroups: diet only, physical activity only and combined. We used these groupings to look at the relative effectiveness of these types of interventions to generate further hypotheses. A random effects model was applied because of the assumed clinical and methodological heterogeneity among the studies. 43 We used odds ratios as the statistical output for dichotomous outcomes, with 95% confidence intervals (CIs). For continuous data, we used mean differences with 95% CIs. We tested for statistical homogeneity of pooled estimates of effectiveness using the Chi 2 test and the I 2 statistic, for which a statistically significant (p value ≤ 0.1) value of the Chi 2 test together with I 2 value of at least 50% indicates heterogeneity.

| Search results
The systematic search strategy identified 3014 articles. After the removal of duplicates, 2321 unique titles and abstracts were screened. A total of 21 studies (presented in 34 articles) were found to be eligible for inclusion in this review (see Figure 1).  Two of the 34 papers reporting two of the included studies required translation. 71,74

| Description of studies
The characteristics of the included studies are presented in Table 1. All of the included studies were cluster-RCTs. The majority of the studies were conducted in Asia (n = 10), followed by South America (n = 4), North America (n = 4) and the Middle East (n = 3). Papers reporting the included studies were published between 1998 and 2019 (71.4% of studies published from 2012 onwards). The total number of participants per study ranged from 221 to 10 091, and the mean age of participants varied from 4.5 to 11.5 years. Fifteen of the 21 studies targeted populations in which the mean age of participants was between 9 and 12 years. Three of the studies from Brazil targeted low socio-economic areas, whereas the remaining 18 studies either did not specifically address socio-economic status or did not report this information. 68,69,77 Most studies (n = 19) used a control group that received routine school health education or no intervention. The remaining two studies either did not state the control or delivered general health sessions to participants in the control group. 53,77 Three studies evaluated diet-only interventions, three studies evaluated PA-only interventions and 14 studies evaluated interventions involving diet and PA components (referred to as combined interventions for the purpose of this review), see Table 1. The remaining study had three trial arms investigating one PA-only and two separate diet-only interventions. 55 The duration of the interventions ranged from 3 to 36 months, whereas the interval between baseline and follow-up measures varied from 3.5 to 72 months. The personnel delivering the intervention included school staff (n = 12), external deliverers (individuals from the research team or health professionals; n = 6) or a mixture of both (n = 2). One study did not specify who delivered the intervention. 72 Only nine of the 21 studies addressed adverse intervention effects. [44][45][46][47]50,55,58,72 All of the diet-only studies incorporated nutrition education sessions and parental involvement. 68,72,77 One diet-only study included modifications to the school food provision. 72 Two of the three PAonly studies provided additional PA sessions ranging from 100 to 105 min per week, which were integrated into the existing school curriculum. 45,67 The remaining PA-only study comprised monthly education sessions, changes to the school environment and an additional home setting component with parental involvement. 58 The key characteristics of the combined intervention studies were education sessions (n = 13), modifications to the school food provision (n = 11) and parental involvement (n = 13), outlined in Table 1. Four of the 14 combined intervention studies included daily PA sessions and four studies included an additional home setting component.

| Quality appraisal
The results of the methodological quality assessment are shown in baseline imbalance (n = 17), loss of clusters (n = 16), analysis bias (n = 12), random sequence generation (n = 11) and incomplete outcome data (n = 10). 'Unclear risk of bias' was the most common assessment for the remaining four domains: allocation concealment (n = 17), selective reporting (n = 17), blinding of outcome assessment (n = 11) and recruitment bias (n = 11). The key methodological limitations encountered (as indicated by a 'high risk of bias' assessment) included the following: unblinded outcome assessment (n = 5), discrepancies in the reporting of prespecified outcome data (n = 4), absence of reporting on losses to follow-up (n = 3), inappropriate data analysis method (n = 3) and loss of clusters (n = 3). Seven included studies were assessed as 'low risk of bias' for five or more domains. 46,47,50,58,68,75,77 Two of the more recent trials had predominantly low risk judgements and appear to be higher quality than the Other: Public schools and children from families of low socio-economic levels.

Method of randomization:
Not specified.
The PAAPPAS intervention combined primary prevention at school with an additional home intervention targeting children with excess weight.
The primary prevention at school consisted of two monthly 30-min health education games with the main message of reducing consumption of sodas, cookies and sugar-sweetened beverages.
Water consumption was promoted, culinary classes took place along with PA games relating to the Olympic Games.
Routine school health activities.

Method of randomization:
Not specified. One classroom lesson every fortnight was provided based on a text designed by the research team on childhood obesity prevention. An additional intervention was provided to children with obesity or overweight and those failing physical education, who had 20-min after school lessons from Mondays to Thursdays. A nutritional lecture was given to parents once per semester by a researcher.
Routine school health activities.

Secondary setting component:
No.
Length of follow-up (mo):  School grade:

Method of randomization:
Random number table. The intervention consisted of six 40-min lectures, and the Happy 10 programme involving 10-min PA sessions twice daily. The school food provision was evaluated and modified accordingly. Two lectures were also provided to parents who were sent bulletins on nutrition.
No intervention.

Secondary setting component:
No.
Length of follow-up (mo): Trial arms: School grade:

Method of randomization:
Computer random number generator.
The intervention targeted school-level policies and implemented health education activities. Four 40-min health education lessons were delivered every 2 weeks. At least three 45-min physical education classes were provided each week of moderate to vigorous intensity. Students were encouraged to exercise at home and were provided with instruction manuals. The school lunches were evaluated and modified for improvement. As part of school policies, children were told not to drink sugar-sweetened beverages or eat unhealthy food in school.
Routine school health activities.

Secondary setting component:
No.
Length of follow-up (mo): 12 Primary outcomes: BMI, BMI z-score, prevalence of OW, and prevalence of O.
Secondary outcomes: Median number of days doing more than 1 h of MVPA per week, median number of days per week for consumption of vegetables, fruits and meat.

Secondary setting component:
No.

Method of randomization:
Not specified.

Method of randomization:
Not specified.
The Nutrition on the Go programme involved 6 nutrition and PA workshops. Daily PA occurred prior the start of classes and a puppet theatre was performed was once a week by the 5th grade students to younger children.
The school shop sold fruits, vegetables and water. Parents were provided with recipe calendars.
Not stated.

Secondary setting component:
No.
Length of follow-up (mo): Trial arms: School grade: Other: Semi-rural populations.

Method of randomization:
Not specified.
The CATCH Programme consisted of education on health habits, PA sessions and modifications to the school food provision. The implementation of the intervention was supervised by the researchers.
No intervention.

Secondary setting component:
No.
Length of follow-up (mo):  School grade:

Method of randomization:
Coin tossing and drawing of lots.
The intervention consisted of five 40-min education sessions that were provided at 1-month intervals. Fresh fruits were sold daily at the school restaurant and two additional BMI measurements were performed for participants.
Two 1-h education sessions were provided to families who were also sent a healthy nutrition brochure.
Routine school health activities.

Personnel:
Research team members.

Secondary setting component:
No.

Secondary setting component:
No.

Method of randomization:
Random number table.
The nutrition education intervention consisted of six 40-min lectures. The school food provision was evaluated and modified accordingly. Two lectures were also provided to parents who were sent bulletins on nutrition.
No intervention.

Secondary setting component:
No.
Length of follow-up (mo):  School grade: 3rd and 4th Other:

Method of randomization:
Not specified.
The Happy 10 Programme consisted of two daily 10-min PA sessions of moderate to vigorous intensity conducted during classroom breaks. The programme was incorporated into the curriculum and did not replace any regular activities.

Secondary setting component:
No.

Method of randomization:
Random number  School grade:

nd
Other: Private kindergartens.

Method of randomization:
Not specified.
The intervention consisted of a 15-min walk before the morning class and a 20-min aerobic dance session in the afternoon. These activities occurred three times a week and did not interfere with routine school activities.
Routine school health activities.

7.5
Personnel: Research team members.

Secondary setting component:
No.
Length of follow-up (mo): rest of the field. 47,58 Confidence in the results from these trials is higher than those which have likely sources of bias. Outcomes have only been included where a statistical comparison between the intervention and control group has been provided.  Table 3.

| BMI z-score
Five of the nine studies which included the outcome of BMI z-score reported a significant decrease in this outcome in the intervention group compared with the control group. 45,47,55,58,73 Three of the five studies reporting positive effects evaluated combined diet and PA interventions. Other characteristics of the five studies reporting significant intervention effects included the following: parental involvement (n = 4), delivery by school teachers (n = 4), intervention duration of at least 12 months (n = 3) and daily PA sessions (n = 3). All four studies with parental involvement provided at least two health education sessions for parents. Two studies did not report significant effects on BMI z-score. 46,55 The remaining two studies did not statistically compare outcomes across study groups. 53,65

| Body fat percentage
Only two of the seven studies reporting body fat percentage found a significant intervention effect. 45 and education sessions (n = 3). Six studies did not report significant effects on prevalence of obesity. 46,55,60,72,77 The characteristics of these studies were similar to those with significant intervention effects: delivery by school teachers (n = 4), combined diet and PA intervention (n = 3) and an intervention duration of at least 12 months (n = 2). The remaining four studies did not statistically compare outcomes across study groups. 65,[67][68][69] Nine of the 14 studies, with a total of 19 762 participants, reported data in the appropriate format to be pooled in a meta-analysis, shown in Figure 4. 44 However, substantial heterogeneity was detected among the studies (I 2 = 78%; p < 0.0001).

| Prevalence of overweight
Only two of the 11 studies with data on prevalence of overweight reported significant intervention effects. 44,73 Both of these studies evaluated combined diet and PA interventions, delivered over a duration of at least 6 months and provided at least two education sessions for parents. Five studies did not report significant effects on prevalence of overweight. 55,60,72,77 The characteristics of the studies reporting nonsignificant effects were similar to those with significant intervention effects: intervention duration of at least 6 months (n = 5), parental involvement (n = 5), delivery by school teachers (n = 3) and a combined diet and PA intervention (n = 2). The remaining four studies did not statistically compare outcomes across study groups. 53,65,68,69 Six of the 11 studies, involving a total of 7723 participants, reported data in the appropriate format to be pooled into a meta-analysis, shown in Figure 5. 44,60,68,69,73,77 Although there was a lower prevalence of overweight in the intervention group compared with the control group, the difference was not significant (odds ratio = 0.89; 95% CI, 0.74 to 1.09; p = 0.26). Substantial levels of heterogeneity were observed for this comparison (I 2 = 65%; p = 0.01).

| Dietary outcomes
Thirteen studies assessed outcomes relating to dietary behaviour, of which nine found significant intervention effects on at least one outcome. 46,50,62,68,[70][71][72]75,77 Six of these nine studies evaluated com-  46 This study reported effects on dietary outcomes. Six of the nine studies that reported positive effects on dietary behaviour also reported significant intervention effects on at least one adiposity-related outcome. 46,47,68,70,71,77 One study did not report significant intervention effects on dietary outcomes. 69 The remaining three studies did not statistically compare outcomes across study groups. 60,62,73 3.4.9 | Physical activity outcomes Ten of the included studies reported outcomes relating to PA behaviour, of which six found significant intervention effects on at least one outcome. 46 behaviour also reported significant intervention effects on at least one adiposity-related outcome. 46,58,70 One study did not report significant intervention effects on physical activity outcomes. 47 The remaining three studies did not statistically compare outcomes across study groups. 60,62,73
Total no. of clusters analysed:

Intervention:
Diet and PA.
Total no. of clusters analysed:
Total no. of clusters analysed:

| Strengths and limitations
This is the first systematic review of school-based interventions to prevent childhood obesity that focuses specifically on MICs. The unique situation of these countries in terms of their rapid nutrition transition and consequent steep rise in childhood obesity prevalence highlights the importance of achieving an understanding of the types of interventions that are likely to work in these settings.
A comprehensive search strategy was developed and included approaches for identifying unpublished studies. In addition, studies were not excluded based on publication language, which was a common methodological limitation of previous reviews. 31,32,34 Eight of the 21 studies had not previously been included in a systematic review, and two included studies had not been published at the time the review was undertaken. 46,47,58,[69][70][71][72][73] The quality of the reporting and risk of bias was improved in two trials published more recently and included in this review ( Figure 2). 47,58 The low quality of evidence has been a common limitation in previous reviews. 31,32 A further strength is that additional obesity-related outcomes such as body fat percentage and waist circumference were evaluated in addition to BMI, providing a more comprehensive picture of intervention effectiveness.
There are also some limitations. The search strategy was restricted to international electronic databases, which may not have identified studies published in local or national journals. Nevertheless, our search strategy included a comprehensive approach to identifying international literature and also utilized the knowledge of experts in the field. Abstract and title screening of articles was only carried out in duplicate for a 5% random sample; however, there was a high level of agreement between reviewers for this sample. Despite our attempts, we were unable to contact some study authors to gain further information and so were unable to include two potentially eligible studies. 78,79 Due to the design and reporting of many studies, we were unable to include them in our meta-analyses. In addition, the high heterogeneity observed in the three meta-analyses undertaken needs to be considered when interpreting the findings. Consistent with the limitations described in previous reviews, outcome measurements for some studies were restricted to BMI, which can result in misleading results because BMI may be insensitive to changes in adiposity for children and is an age-and sex-dependent variable. 26,62,80 Also, many studies used self-report measures for PA and dietary behaviour, which may have low validity. sonnel. An important consideration in these settings is ensuring that intervention approaches are contextually appropriate. 46 For example, the nutritional environment in schools in MICs poses a challenge as often street-food vendors will intermingle with privately owned school food shops. 53,62,63 A further example is the strong emphasis on academic performance found in some settings such China. Obesity prevention efforts need to complement and synergize with this academic focus. 50,51 Intervention programmes must be tailored to take account of these different social, environmental and educational contexts.

| Recommendations for future research
In this review, we have acknowledged that few of the included studies reported adverse outcomes, including the prevalence of children with underweight, at follow-up. In MICs, undernutrition commonly coexists alongside obesity. Therefore, there is a potential for harm from obesity prevention programmes if they cause weight loss in children who are already living with underweight. 3,81 Consequently, future studies should ensure that underweight and other potential adverse outcomes, such as bullying or effect on quality of life (only assessed in one included study), are evaluated. 47 Although, some of the more recent studies had higher methodological quality and more comprehensive reporting, there is still a large variation across all the included studies. Thus, a further recommendation is for more rigorous F I G U R E 3 A forest plot presenting data on the effects of school-based obesity prevention interventions on BMI with 95% confidence intervals of individual studies, grouped by intervention type. The pooled estimates with 95% confidence intervals were calculated using a random effects model and inverse variance methods. The unit of measurement on the horizontal axis is weight (kg)/height (cm) 2 . Studies have only been included where data were adjusted for clustering and reported in the appropriate format. SD, standard deviation; IV, inverse variance; CI, confidence interval F I G U R E 4 A forest plot presenting data on the effects of school-based obesity prevention interventions on the prevalence of obesity with 95% confidence intervals of individual studies, grouped by intervention type. Studies have used different definitions of obesity, either using the WHO or International Task Force classifications. The pooled estimates with 95% confidence intervals were calculated using a random effects model and inverse variance methods. Studies have only been included where data were adjusted for clustering and reported in the appropriate format. SD, standard deviation; IV, inverse variance

| CONCLUSION
This review presents some evidence to support the effectiveness of school-based interventions in preventing childhood obesity in MICs. F I G U R E 5 A forest plot presenting data on the effects of school-based obesity prevention interventions on the prevalence of overweight with 95% confidence intervals of individual studies, grouped by intervention type. Studies have used different definitions of overweight, either using the WHO and or International Task Force classifications. The pooled estimates with 95% confidence intervals were calculated using a random effects model and inverse variance methods. Studies have only been included where data were adjusted for clustering and reported in the appropriate format. SD, standard deviation; IV, inverse variance; CI, confidence interval