School-based obesity prevention interventions in Latin America: A systematic review

ABSTRACT OBJECTIVE To evaluate the implementation and effectiveness of school-based interventions to prevent obesity conducted in Latin America and provide suggestions for future prevention efforts in countries of the region. METHODS Articles published in English, Spanish, and Portuguese between 2000 and 2017 were searched in four online databases (Google Scholar, PubMed, LILACS, and REDALYC). Inclusion criteria were: studies targeting school-aged children and adolescents (6–18 years old), focusing on preventing obesity in a Latin American country using at least one school-based component, reporting at least one obesity-related outcome, comprising controlled or before-and-after design, and including information on intervention components and/or process. RESULTS Sixteen studies met the inclusion criteria. Most effective interventions (n = 3) had moderate quality and included multi-component school-based programs to promote health education and parental involvement focused on healthy eating and physical activity behaviors. These studies also presented a better study designs, few limitations for execution, and a minimum duration of six months. CONCLUSIONS Evidence-based prevention experiences are important guides for future strategies implemented in the region. Alongside gender differences, an adequate duration, and the combined use of quantitative and qualitative evaluation methods, evidence-based prevention should be considered to provide a clearer and deeper understanding of the true effects of school-based interventions.


INTRODUCTION
Childhood obesity is a global public health problem. The worldwide trend showed that, in 2013, 22.6% of girls and 23.8% of boys in developed countries were overweight or obese, whereas in developing countries it was 13% for both girls and boys 1 . Recently, a study on the trends from 1975 to 2016 estimated the global-age standardized prevalence of obesity in children and adolescents to be 5.6% among girls and 7.8% among boys, highlighting the sustained growth in developing countries 2 .
Latin America follows this trend: in 2016, the prevalence of overweight and obesity among 5 to 19 year-old Mexican children and adolescents ranged between 33% (both genders, aged 5-11 years) and 39.2% (girls aged 12-19 years; boys: 33.5%) 3 . Country-specific data from the Global Burden of Disease (GBD) 2015 Obesity Collaborators shows that in Chile the prevalence was 25.5% among girls and 36.4% among boys, and in Brazil 23.8% among girls and 27.3% boys (< 20 years old) 4 . As for Peru, GBD found 22.4% of girls and 19.2% of boys to be overweight or obese.
Environmental factors, lifestyle preferences, and cultural environments play key roles in the worldwide rising prevalence of obesity 5 . Numerous studies showed that sugary beverages and high-fat foods consumption added to low fruit and vegetable intake, decreased physical activity (PA), and increased sedentary behavior is positively associated with obesity [6][7][8] .
This evinces the importance in promoting healthier food choices and a more active lifestyle along with environment-related changes to improve healthy behavior in school-aged children and adolescents 9 . Schools play a pivotal role in promoting a healthy lifestyle among students. Reviews on school-based prevention programs for obesity, mostly in the United States and Europe, reported improvement in health-related behaviors and/or knowledge, as well as some positive impacts on body mass index (BMI) [9][10][11] .
Reviews on the extent and impact of interventions conducted in Latin America targeting obesity among school-aged children and adolescents are scarce. Most reviews are based on international efforts and include few studies conducted in Latin America, or approach PA interventions and report only PA-related outcomes 10,[12][13][14][15] . A review performed in 2013 reported significant outcomes in seven of the ten included studies, among which at least three presented an appropriate design and execution conducive to statistically significant changes in obesity-related outcomes 16 .
Considering that our systematic review seeks to assess how obesity prevention interventions targeting school-aged children and adolescents were implemented in different Latin American countries (i.e. implementation) and whether they were effective on obesity-related outcomes (i.e., effectiveness). Based on these findings, we provide suggestions for future prevention efforts implemented in Latin American countries.

Literature Search
Studies published from 2000 to 2017 were collected by searching four online databases: Google Scholar, PubMed, Literature in the Health Sciences in Latin America and the Caribbean (LILACS), and the Network of Scientific Journals of Latin America and the Caribbean, Spain, and Portugal (REDALYC). An update search was conducted in March 2019. Potentially eligible articles were also located by hand, by screening studies and articles reference list, across the 20 Latin American countries. Search strategy included Medical Subject Headings (MeSH) terms combined with text words based on the categories: population, intervention, outcomes, and type of studies addressed by the review. Search strategy is available at https://osf.io/yuz7e/. Publications written in English, Spanish, and Portuguese were covered. Search was conducted in three stages: first, one researcher conducted data search (RCCH); then identified studies were screened based on title and abstract; finally, studies were independently assessed in full-text considering the inclusion/exclusion criteria by two researchers (RCCH and EWN) ( Figure 1). In the event of disagreement or conflicts between the researchers, results were discussed based on full-text evaluation until reaching consensus.

Selection of Studies
Studies were included in this review if they met the following eligibility criteria: a) targeted school-aged children and/or adolescents (6-18 years old) of Latin American countries, b) addressed an intervention for primary prevention of obesity using at least one school-based component, c) reported obesity-related outcomes, d) comprised a controlled or before-and-after design, and e) included information on intervention components and/or process. Exclusion criteria were: a) studies that included children under six years old, categorized as preschool-aged, b) targeted overweight and/or obese children/adolescents as secondary prevention or treatment, c) addressed an intervention that was not conducted in a Latin American country, and d) reported solely dietary and/or PA outcomes.
Studies were not restricted regarding duration, follow-up period, risk of obesity, and intervention implementers.

Quality Assessment
Study quality was assessed using the standardized Effective Public Health Practice Project Quality Assessment Tool (EPHPP Tool). It comprises six components 17 : a) selection bias, b) study design, c) confounders, d) blinding, e) data collection methods (validity and reliability of tools), and f) withdrawals or drop-outs per group. Two researchers (RCCH and EWN) independently assessed the collected studies. Following EPHPP guidelines, each criterion was rated as good, fair, or poor, and each study received an overall score by the sum of the six ratings. In the event of any discrepancy between researchers, each study was reassessed collectively to reach a final decision.

Data Extraction
One researcher (RCCH) extracted details on each study intervention characteristics and summarized them in a table using a narrative synthesis. Collected data consisted of: first author name; publication year; country and specific setting (region-city); intervention focus; number of participants and age; target group (low socioeconomic status [SES]); intervention activities, duration and follow-up; theoretical basis, outcome measures, process and/or cost evaluation; and overall quality evaluation. Outcome measures were considered those related to obesity (prevalence of overweight and/or obesity, BMI, BMI z-score), dietary/nutrition behavior (e.g., fruit, vegetables, fast foods, and snacks intake), and PA behavior (e.g., PA time, exercise tolerance, and endurance). Effectiveness across studies was determined by effect size (ES). The following formula, also adopted by other studies 18 , was used to estimate Cohen's D ES: d = (x 1 -x 2 )/s, where x 1 is the mean of the intervention group (IG), x 2 the mean of the control group (CG), and s the pooled standard deviation. ES was considered trivial if < 0.2, small if equal to 0.2 or < 0.5, medium if equal to 0.5 or < 0.8, and large if equal to or higher than 0.8. For before-and-after study designs, ES was estimated based on the results of the last reported follow-up. ES values were tabulated for each outcome and based on intervention focus (whether dietary/nutrition, PA, or both).

Literature Search
Our initial search identified 8,273 publications on the databases and three by handsearching reference lists (Figure 1). After eliminating duplicates and screening studies titles/abstracts, 133 articles remained. By reading articles full-text and applying eligibility criteria, we found 71 articles unrelated to obesity prevention, 18 describing school-based programs not conducted in Latin America, 8 including children under 6 years old (categorized as preschoolers), 9 that did not report obesity-related outcomes, and 7 that were not based on school activities, all of which were excluded. Sixteen interventions (reported in 20 publications) met the inclusion criteria and were included in the systematic review. Several Latin America countries had very few or no school-based studies on neither of the four databases.

Characteristics of the Intervention
After reviewing each study specificities, we identified three intervention domains, as summarized in Table 1, following the Health Promoting School Framework developed by the World Health Organization (WHO) 38 . The main domains entail school environment, curriculum, and partnership with families and/or the wider community. As most interventions adopted a school curriculum component, only those including at least two more domains were considered multi-component. Most interventions targeting diet alone (four out of five) were mainly educational, promoting classroom sessions encouraging healthy eating 21,24,30,31 , whereas the remaining one was a school breakfast program 36 . Among these, one was considered multi-component for embracing efforts for environmental change along with education activities for parental involvement 21 . The intervention focused on PA compared a curriculum of PA programming with a conventional physical education (PE) class 32,33 . Interventions targeting nutrition and PA (n = 10) implemented both educational and environmental activities 20,22,23,25,26 , and six of them were multi-component interventions that also included parental involvement 19,27,28,29,34,35,37

Obesity-related outcomes
a. Prevalence of overweight and/or obesity: few studies reported this outcome (n = 3) 21,35,36 . Two multi-component interventions significantly decreased the prevalence of overweight and/or obesity on -9.5 to -3.5 percentage points 21,35 (    b. Diet: five studies recorded positive effects on one or several dietary behaviors (two targeting diet-only and three diet and PA) 19,20,24,27,31 . However, we managed to estimate ES for only two of the interventions targeting both domains (Table 4). One intervention targeting diet alone reported a one-percentage point increase in the daily intake of two healthy food (orange juice and skim milk) among girls and a significant decrease in the consumption of hamburger and hotdog among both boys and girls 24 . As for interventions targeting both domains, one managed to sustain fruits intake in the IG (≥ 5 days/week), but reported a significant decrease in the CG (-10.3%). Whereas IG also showed an increasing water intake (+6.1%), in CG it decreased similarly to fruits (-10.6%) 19 . The two remaining studies reported a trivial and a small positive ES on similar behaviors (ranging from 0.11 to 0.29) 20,27 .

DISCUSSION
Among the 16 studies that met inclusion criteria and were included in our systematic review, 10 (60%) achieved significant positive effects on at least one of their reported outcomes. Of these, three interventions focused on diet, one on PA, and six on both PA and diet. Within the three diet-only interventions, two reported an improvement in dietary behaviors, but failed in positively impacting an obesity-related outcome 24,31 , and one reduced both of its reported obesity-related outcomes (prevalence of overweight and obesity) by reducing BMI z-score in IG 21 . The PA-only intervention managed to improve PA performance within the overall sample for most tests, but this effect had no significantly impact on any obesity-related outcome 33 . Of the six diet/PA-focused interventions, three managed to improve a single behavioral outcome 19,20,22 , one positively affected both diet and PA, and two impacted both behavioral and obesity-related outcomes (one PA+ BMI; one PA+ BMI+ obesity prevalence) 35,37 .
Considering these results, we identified evidence-based effectiveness in three obesity prevention interventions conducted in Latin America that targeted promoting healthy diet and PA by associating environmental, educational, and parental involvement activities (multi-component), corroborating other international reviews 12,39,40 . Although these interventions ES often range from trivial to small, they have a significant capability of providing benefits if scaled to a greater level. The results of two interventions implemented in Chile and one in Mexico were strengthened based on these studies, showing the best methodological quality (moderate) among our sample 27,35,37 . However, this data also stress the weakness within this research field in Latin America -as already reported by prior studies 41,42 -regarding the amount and methodological quality of publications. Our systematic review found studies conducted in only 5 of the 20 countries in the region (25% representation) to meet inclusion criteria, and most of them presented a poor methodological quality.
Parental involvement has been considered of key influence for improving children's lifestyle behaviors and preventing obesity. Our findings emphasize the importance of their role within the school setting, corroborating other reviews worldwide 10,43 , which may also be explained by the culture of family tradition in Latin America 10 .
The potential influence of age in intervention effects is an ongoing debate. While some studies argue that programs targeting older students tend to achieve better outcomes 11,44 , other meta-analysis found younger children (elementary school students, aged 4-9 years) to experience greater BMI effects than middle (10-13 years) and high school students (14 years or older) 45 . However, regional and national disparities in education systems should be regarded when comparing and interpreting results based on age. In Latin America, the education system is characterized by three basic levels: pre-school, primary school, and high school, which, in most countries, entails students aged approximately 3-5 years, 6-13 years, and 13-18 years 46,47 , respectively. In our review, thirteen studies targeted primary school students, ranging from 7.6 to 10.9 years old (an average of 9.09 years). We suggest further studies to approach both children and adolescents, enabling a deeper understanding regarding age impact on intervention outcomes.
The greater improvement on boys' physical condition after interventions, reported in two studies, were previously discussed 48,49 . Considering these findings, it seems that boys tend to engage in more vigorous activity than girls, who tend to be less active 9 . One study found a greater improvement in dietary intake among girls 24 . Such differences may be explained by the intervention focus, which did not address PA specifically, but rather educational components grounded on social learning theory, to which Kropski et al. (2008) suggest girls might respond better. Yet, further research are required for a better discernment of gender differences within PA and dietary intake behaviors.
Intervention duration and its association with effectiveness are still an ongoing debate, as well as an agreed differentiation cut-off. Bautista-Castano et al. (2004) found that interventions lasting between 6 months and one year are more effective (triceps skin-fold and BMI anthropometrics) than shorter and longer-term interventions 50 . Another meta-analysis found short-term interventions (0 to 12 weeks) to have negative effects on BMI, whereas longer interventions (13 weeks or more) are associated with small, significant, and positive BMI effects 45 . The duration of the three effective interventions identified in our systematic review ranged from 6 to 11 months, within the cutoff suggested by the aforementioned reviews. Besides duration, follow-up period may also be important for identifying intervention sustained benefits. However, these data are often disregarded in the literature, which might be justified by the difficulty in assessing effects of groups no longer under the intervention arm.
Around 40% of the studies reported evaluation process mostly regarding the overall percentage of individuals targeted by the intervention (students, parents, and teachers) who participated or attended nutrition or PA sessions, that is: the percentages of adherence or attendance. Yet, various researchers argue the need of a throughout evaluation of the implementation process to better contextualize and assess the program true effect 51,52 . Such need is even sharper in complex or multi-component interventions 53 that are not solely based on education. Considering that, we recommend further study to employ both qualitative and quantitative approaches by a mixed-method design, to better plan, correct, and evaluate interventions and their affecting factors.
Our systematic review pose some limitations. First, comparing and interpreting effect sizes from heterogeneous studies targeting children and adolescents is always challenging, and only a few studies provided the appropriate pediatric measures of obesity-related outcomes (BMI z-scores). Second, many methodological deficiencies identified by the quality assessment instrument owed to lack of information, so we urge studies to provide more detailed information regarding the adopted methods. Third, we considered intervention effects regardless of the evaluation data process, because these results were not thoughtfully reported. As our focus was obesity-related outcomes, we might have potentially excluded studies addressing nutrition and PA behaviors as primary outcomes that also contributed to obesity prevention, not only by weight loss.

CONCLUSIONS
We found evidence of the effectiveness of three school-based interventions for preventing obesity among school-aged children and adolescents in Latin America. These interventions were characterized as moderate quality, included the multi-components of health education, school environment, and parental involvement focusing on healthy eating and PA behaviors, and had better study designs, few execution limitations, and a 6-month minimum duration. Future efforts on preventing obesity in Latin American countries should consider evidence-based preventions experiences, such as those identified in our review, as guides. They should also consider gender differences, appropriate duration, and mixed-method evaluation designs combining both quantitative and qualitative approaches, as their association could provide a clearer and deeper understanding of the school-based interventions true effect.