To determine the effectiveness of school-based strategies for obesity prevention and control using methods of systematic review and meta-analysis.
Peer-reviewed studies published between 1966 and October 2004 were considered for review. Studies meeting eligibility criteria were published in English, targeted children aged 3–18 in a school setting, reported weight-related outcomes, included a control measurement and had at least a 6-month follow-up period. Studies employed interventions related to nutrition, physical activity, reduction in television viewing or combinations thereof. Weight related data were analyzed using RevMan software.
Sixty-four studies were considered for inclusion. Fourteen did not meet inclusion criteria; 29 were excluded due to poor methodological quality. Twenty-one papers describing 19 studies were included in the systematic review and 8 of these were included in the meta-analysis. Nutrition and physical activity interventions resulted in significant reductions in body weight compared with control ((standardized mean difference, SMD=−0.29, 95% confidence interval (CI)=−0.45 to −0.14), random effects model). Parental or family involvement of nutrition and physical activity interventions also induced weight reduction ((SMD=−0.20, 95%CI=−0.41 to 0.00), random effects model).
Combination nutrition and physical activity interventions are effective at achieving weight reduction in school settings. Several promising strategies for addressing obesity in the school setting are suggested, and warrant replication and further testing.
The prevalence of childhood overweight has tripled over the past two decades in the United States.1, 2 Applying the current definition of childhood overweight, the 95th percentile for age- and sex-adjusted body mass index (BMI), approximately 16% of American children 6–19 years of age are overweight;1, 2 this conservative definition likely underestimates the true prevalence of childhood obesity. The physical health of obese children is compromised in a manner similar to adults. Childhood obesity is associated with an increased risk of hyperinsulinemia,3 insulin-resistance,4 type 2 diabetes,5, 6 hypertension,7, 8 hypercholesterolemia,3 chronic inflammation,9 abnormalities in endothelial function,10 hyperandrogenemia,4 gallstones,11 hepatitis,12, 13 asthma, cancer14 and orthopedic problems.10 Obesity has also been shown to diminish children's quality of life severely14 and is associated with decreased self-esteem15, 16 and depressive symptoms.17 Obese children are subjected to teasing, discrimination and victimization and may be socially excluded.18
In response to these threats, there has been a surge in obesity research in recent years. Schools have been a popular setting for implementation of interventions, as they offer continuous, intensive contact with children during their formative years. School infrastructure and physical environment, policies, curricula and personnel have great potential to positively influence child health. Despite the apparent advantages of addressing obesity in schools, a relative lack of evidence of effectiveness has led some to question the wisdom of allocating scarce resources to school-based programs. This issue was the subject of debate among health professionals at the 2005 annual meeting of the North American Association for the Study of Obesity in Vancouver, BC.19
Several systematic reviews of childhood obesity programs have been published.20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35 With the exception of one recent review20 and an earlier review conducted by our team in collaboration with the Community Guide Branch of the Center for Disease Control and Prevention (CDC),25 this work is unique in that it focuses specifically on studies undertaken in schools. Previous reviews have focused solely on the prevention of obesity22 or weight loss interventions among already overweight children.23 Recognizing that the strategies to prevent unnecessary weight gain and to treat obesity in children overlap considerably, this review includes studies with both aims. Many authors have reported difficulty in evaluating findings across studies quantitatively. Although some difficulty was encountered due to heterogeneity of outcome measures and specific intervention components, sufficient data were available for meta-analysis.
The literature search strategy was developed by the research team and critiqued by an expert panel. The following search engines were used to locate overweight/obesity intervention studies published between 1966 and February 2000: MEDLINE, HealthStar, Psych Info and Embase. Review of these studies was conducted in collaboration with the CDC and Prevention's Guide to Community Preventive Services branch and has been published.25 Additional searches were conducted to retrieve studies published between February 2000 and October 2004 using Medline, Ovid, Cinahl and PsychInfo. The Cochrane Library was searched to identify systematic reviews to be used for manual bibliography searching. Other meta-analyses, review articles and articles written by prominent authors in the field of obesity were also reviewed for relevant citations.
To be eligible for inclusion, studies needed to: be published in English; target children aged 3–18 in a school setting; report commonly used weight-related outcomes (BMI, body weight, etc.); include a control measurement (either with pre/post-measures or using control group(s); and follow participants for at least 6 months from the beginning of the intervention. Studies included in the review aimed to prevent unnecessary weight gain or manage weight, and employed interventions related to nutrition, physical activity, reduction in television (TV) viewing or combinations thereof. The following definitions were used to place studies into categories on the basis of the mode of intervention: ‘nutrition interventions’ aimed to modify dietary intake of the students with no explicit focus on physical activity or TV; ‘physical activity’ (PA) interventions aimed to increase PA with no focus on dietary intake or TV; ‘TV reduction’ interventions focused only on reducing this sedentary activity without focus on changing dietary intake or increasing PA; and ‘combination’ interventions include both nutrition and PA elements and may also include reduction in TV viewing.
Data were extracted from each article by two reviewers independently, using a standardized protocol developed by the CDC.36 The CDC's Community Guide ‘data abstraction form’ contains two sections eliciting information about study description and results, followed by a third section to assess methodological quality. The quality scoring system takes into account suitability of study design and quantifies threats to validity.36 This score is used as the basis for excluding studies of poor quality from the review. The two reviewers compared quality scores and data extracted; the principal investigator was consulted when differing responses could not be reconciled.
The combined searches identified 64 papers on the basis of screening of titles and abstracts. Upon review of full text, 14 did not meet inclusion criteria, whereas data abstraction of the remaining articles resulted in exclusion of 29 with poor methodological quality, the majority from the initial review (see flow diagram in Figure 1). Twenty-one papers were included in the review, describing 19 studies (data from two studies were published twice, but included in the analysis once). No studies published before 1980 met quality criteria. Four of the included studies were published before 1996, five were published between 1996 and 2000 and 10 were published between 2000 and February 2004.
Of the 19 studies (21 papers) meeting criteria for inclusion, 14 were randomized controlled studies and five were non-randomized controlled studies. Six of the studies were ‘treatment’, and 13 were strictly primary prevention. These 19 studies intervened with a total of over 13 029 (sample size not reported in one study) students; 13 took place in elementary schools, 3 took place in middle schools and 3 targeted high school students. Ten took place in the United States, whereas nine were conducted overseas. One study targeted girls only.
All studies incorporated multiple strategies to impact on the weight status of student populations. Common strategies included: parent or family member participation;37, 38, 39, 40, 41, 42, 43, 44, 45 changing the physical environment;43, 44, 46, 47, 48 nutrition/PA-related classroom instruction;37, 38, 39, 40, 41, 43, 44, 46, 49 incorporation of nutrition/PA/TV reduction lessons across curricula;50 use of participatory, skill building activities;37, 38, 39, 40, 41, 43, 44, 49 dissemination of educational print materials;37, 38, 39, 40 use of student competitions;44 introduction of PA in addition to PE;40, 41, 43 modification to duration, frequency or intensity of PE activities;46, 49 focus on games, dance or other non-traditional PA;51 provision of new PE equipment or funds to support PE;52 teacher training;37, 38, 39 and student training in behavioral techniques such as self-monitoring.38, 39, 52, 53 Detailed descriptions of each study can be found in Table 1.
Eight of the 19 studies provided adequate data for inclusion in the meta-analyses. All the studies included in the meta-analysis were combination interventions (nutrition plus physical activity).37, 38, 39, 40, 41, 43, 44, 46, 49 Subgroup analyses were also performed by gender for each intervention type where applicable.
Data were analyzed using RevMan software version 4.2.54 For clusters of articles reporting the same outcomes and shared intervention types (nutrition, PA or combination), meta-analysis was performed. Data entered in RevMan included the following: pre/post-changes in outcome measures for the intervention and the control groups, the corresponding number of participants for each treatment group in the trial and standard deviation (σ). Where studies38, 39, 51, 55, 56 reported mean changes in outcome measures from baseline (pre) to end point (post) for each intervention group (that is, ΔI or ΔC), data were entered in RevMan with their corresponding number of participants and σ's. For trials40, 41, 43, 44, 46 that reported both pre- and post-intervention values, changes were computed by subtracting the pre- from the post-measurements. In case of missing σ's, the σ's were estimated by imputing σ's of other studies using the same outcome measures.57
The standardized mean differences (SMD) were computed with their 95% confidence interval (CI). We used SMD (that is, effect size) because the trials included in the analysis assessed weight loss in different ways (e.g., BMI, weight and ponderosity). Using SMD transformed the results of trials into a uniform scale so that they could be pooled. The SMD gives size of the treatment effect in each trial relative to the variability observed in that trial. Positive SMD represents weight gain and a negative SMD represents weight loss.
To assess the net effect of different interventions on weight loss, we computed pooled SMD and 95% CIs for clusters of trials with similar interventions. Weight reduction for each intervention type (nutrition, PA, combination) compared with placebo was considered statistically significant when the 95% CI around the pooled SMD did not include zero. Comparisons across interventions were made also using the 95% CIs around the pooled SMD. When the 95% CIs of pooled SMD of one intervention was not included within the 95% CI of pooled SMD of another intervention (that is, non-overlap), we considered the two interventions significantly different at P<0.05.
The Q test was used to test for heterogeneity.58 The Q test is based on the χ2 distribution and provides a measure of the sum of the squared differences between the results observed and the results expected in each trial, under the assumption that each trial estimates the same treatment effect. To obtain the SMD with 95% CI for treatment and control, we used a random effects model when heterogeneity (χ2) was present (P<0.05); otherwise, the fixed-effects model was used. The I2 test was also computed. This statistic provided the degree of inconsistency of the results of the trials. The value is expressed as a percentage of the total variation across trials that are attributed to heterogeneity rather than chance.
Combination interventions, the single nutrition intervention and TV reduction were equally effective. All showed significant (P<0.05) reduction of body weight in children. The pooled effect sizes of the combination, nutrition interventions and TV reduction were ((SMD=−0.29, 95%CI=−0.45 to −0.14), random-effects model) (see Figure 2); (SMD=−0.39, 95% CI=−0.56 to −0.23) and (SMD=−0.35, 95% CI=−0.63 to −0.06), respectively. PA intervention did not show body weight reduction (SMD=1.87, 95% CI=1.31–2.42).
When including only studies with a parent/family component in the analysis, the combination intervention improved body weight in school children ((SMD=−0.20, 95% CI=−0.41–0.00), random effects model) (see Figure 3). The magnitude of the SMD for the combination intervention decreased ((SMD=−0.16, 95% CI=−0.320–0.00), random effects model) (see Figure 4) when including only studies with both a family component and an environmental component.
The TV reduction intervention resulted in significant reduction in body weight (SMD=−0.35, 95% CI=−0.63 to −0.06). This intervention effect (SMD=−0.35, 95% CI=−0.63 to −0.06) did not differ from the effect of nutrition (SMD=−0.39, 95% CI=−0.56 to −0.23), or combination interventions ((SMD=−0.29, 95% CI=−0.45 to −0.14), random-effects model) (see Figure 2).
In subgroup analysis by gender, the PA study showed better statistically significant weight reduction in girls (SMD=−0.38, 95% CI=−0.74 to −0.02). The weight reduction in girls (SMD=−0.38, 95% CI=−0.74 to −0.02) did not differ significantly when compared to boys (SMD=0.14, 95% CI=−0.17–0.44). In studies employing combination interventions, girls ((SMD=−0.53, 95% CI=−1.37–0.30), random-effects model) (see Figure 5) showed non-significant improvement in weight reduction compared with boys (SMD=−0.22, 95% CI=−0.32 to −0.12), fixed-effects model) (see Figure 6). Girls in the PA intervention (SMD=−0.38, 95% CI=−0.74 to −0.02) showed non-significantly better improvement compared to girls in trials that used combination interventions ((SMD=−0.53, 95% CI=−1.37–0.30), random-effects model) (see Figure 5). However, in the combination interventions, boys ((SMD=−0.22, 95% CI=−0.32 to −0.12), fixed-effects model) (see Figure 6) showed non-significantly better improvement compared with boys in the PA intervention trial (SMD=0.14, 95% CI=−0.17 to 0.44).
This review is an update of the research team's previous work published in 200525 that concluded that there were not enough studies of adequate quality available to determine the effectiveness of school-based interventions, using the CDC's Community Guide to Preventive Services criteria. Although this review update employed Community Guide methods of data abstraction (with quality assessment) for consistency, we did not rely on Community Guide criteria for recommending interventions. However, the addition of several high-quality studies published since the initial review would likely have allowed for formal recommendations. The additional studies provided sufficient data to take a meta-analytic approach.
The robustness of these findings is limited, as there is a high degree of heterogeneity as measured by χ2 or I2. The consistency of a meta-analysis depends on the similarity of magnitude of the treatment effects of the trials included in the analysis. To critically appraise these findings, the presence of any underlying potential source of heterogeneity has to be explored.59 Typically, heterogeneity is associated with reporting bias, differences in the intensity or duration of interventions, the underlying risk, the effect size and irregularities of data.
This meta-analysis helps to address uncertainty about the utility of school-based interventions to combat the childhood obesity epidemic. Although behavioral programs directed at weight control in schools may eventually prove to be less critical than the establishment of school policies and environmental modifications that promote health behavior change (without requiring significant motivation on the part of the student), most have in fact demonstrated success.57 In these analyses, we observed a pooled effect size of −0.29 in clusters of studies using combination nutrition and physical activity interventions. This effect size corresponds to 21.3% non-overlap between the intervention and the control conditions according to Cohen's interpretation.60
Results of the analyses of studies using TV reduction, nutrition alone and physical activity alone should be interpreted cautiously, as only one study was available in each category. Nevertheless, these results indicate that the major contributing factor to the success of combination nutrition and physical activity interventions may be the nutrition component. The extent to which physical activity interventions contributed to weight reduction was minimal. However, it is important to consider that lack of evidence of effectiveness is not the same as evidence of ineffectiveness. Too few studies were available to adequately evaluate these strategies. Heterogeneity of outcome measures is compounded by heterogeneity of intervention methods and, to a lesser extent, populations, thus rendering the aggregation of data and findings challenging.
Ideally schools should be provided with information on the exact elements that make an obesity prevention or control program effective. Given that intervention components vary considerably across studies, it is not possible to do so definitively; however, several salient considerations emerge from this review. An assessment limited to the findings of formal meta-analysis suggests that the following commonly used program components may be valuable: parent involvement (as discussed above),37, 38, 39, 40, 41, 43, 44 classroom (or after-school) instruction on improving dietary intake or increasing PA,37, 38, 39, 40, 41, 43, 44, 46, 49 participatory/hands-on, skill building student activities,37, 38, 39, 40, 41, 43, 44, 49 the provision of print materials,37, 38, 39, 40 teacher training for program implementation,37, 38, 39 student competitions,44 improvements to the nutritional environment (school cafeteria offerings, etc.),43, 46 implementation of PA programs in addition to routine PE,40, 41, 43 modifications to duration, frequency or intensity of existing PE,46, 49 use of non-competitive PA,49 training in behavioral techniques (including self monitoring, goal setting, etc.) or coping skills (decreasing irrational thoughts, improving self-talk, etc.)38, 39, 41, 49 and program tailoring for cultural relevance.41, 43 Future studies should examine strategies for the optimal blending of these intervention approaches.
School policies and programs going forward should be informed by the evidence available to date, but not limited by it. Consistent patterns not yet assessable by meta-analysis have emerged. Weaving nutrition, physical activity or TV reduction lessons into the standard curriculum, so that key messages are reinforced in multiple contexts, appears useful.50 Programs appear to have been enhanced when physical activity was not merely encouraged, but made a routine part of the school day.45, 51, 53 Very few studies focusing on TV reduction were available at the time this review was conducted52, 55 and only one was amenable to computation of effect size.55 This strategy remains appealing despite the relative paucity of data and deserves further attention.
Television viewing likely imposes four interrelated adverse effects related to weight control in children. First, time spent viewing TV is sedentary, and thus conducive to energy imbalance. Second, it has long been hypothesized that TV displaces other, more active leisure pursuits from a child's day,58 and thus it represents both the addition of sedentary time and the subtraction of active time. Third, snacking while watching TV is common. In a recent study of the affects of food advertising during children's television programs, 90% of 3- to 8-year-olds ate while watching TV, and the majority of foods consumed were snack foods of low nutritional value.59 This and other studies provide evidence that TV exposes children to frequent advertisements for fast food, and high sugar, high salt, high-fat ‘junk’ food of low nutritional value. These advertisements directly affect eating habits in the short term—a majority of children ask their parents for foods seen on TV59 and potentially in the long term, as food-marketing strategies focus on creating lifelong customers.61 For these reasons, reducing TV viewing should have favorable effects on energy balance.
Interventions that modify school policies and the physical environment in ways that support improved dietary practices and regular physical activity, but do not provide behavioral programs, are absent from this review. Few policy or environmental interventions were candidates for inclusion simply because those published at the time of the review did not report measurement of weight-related outcomes. These types of strategies to combat childhood obesity are gaining political support and are expected to make a significant impact. Adoption of such approaches, with concurrent evaluation, is warranted.
The conclusions drawn from this meta-analysis are consistent with those reported in previous systematic reviews. We found that combination interventions (nutrition and PA) with a parent or family component produced significant weight reduction. This result is supported by a recent meta-analysis of seven intervention studies on weight loss for overweight children.23 The authors reported a significant weight loss effect of diet, physical activity and parental involvement in all seven studies. Another review found that 17 of 25 studies resulted in statistically significant reduction in BMI.20 On the other hand, results from interventions aimed at obesity prevention have not been as promising. A recent meta-analytic review of 64 obesity prevention programs (46 trials) for children and adolescents showed that 79% of programs did not produce statistically reliable weight-gain-prevention effects.60 Another systematic review examining interventions for childhood obesity prevention also reported that the majority of studies reviewed did not have a significant impact.22 These results suggest that the demonstration of weight loss among overweight children may be more easily attainable than the demonstration of the prevention of weight gain. As noted above, this review combined intervention and prevention studies, as the strategies used are substantially the same in both cases.
This study does not provide definitive guidance toward the optimal school-based strategies for obesity prevention and control. It does, however, provide hopeful evidence that progress toward such guidance is being made. No single intervention, in school or elsewhere, is likely to be sufficient to reverse the childhood obesity trend. Epidemic obesity has been created by a veritable flood tide of obesigenic factors in our society, from fast food to labor-saving technology. In schools, the increasing availability of highly palatable foods of questionable nutritional properties has occurred in tandem with reduced physical activity, in part because of federal legislation,62, 63 reflecting the adverse influences on both sides of the energy balance equation that have played out in society at large. Only when an array of strategies commensurate with and opposite to the prevailing obesigenic influences have been aligned can we reasonably hope to see epidemic obesity begin to subside. To turn the tide of epidemic obesity we will likely need to combine many strategies in schools, communities, clinics, worksites and households. This article offers guidance for the important contribution schools can and should make to that systematic effort.
About this article
Evaluation of a student participatory, low-intensity program to improve school wellness environment and students’ eating and activity behaviors
International Journal of Behavioral Nutrition and Physical Activity (2016)