Effectiveness of school food environment policies on children’s dietary behaviors: A systematic review and meta-analysis

Background School food environment policies may be a critical tool to promote healthy diets in children, yet their effectiveness remains unclear. Objective To systematically review and quantify the impact of school food environment policies on dietary habits, adiposity, and metabolic risk in children. Methods We systematically searched online databases for randomized or quasi-experimental interventions assessing effects of school food environment policies on children’s dietary habits, adiposity, or metabolic risk factors. Data were extracted independently and in duplicate, and pooled using inverse-variance random-effects meta-analysis. Habitual (within+outside school) dietary intakes were the primary outcome. Heterogeneity was explored using meta-regression and subgroup analysis. Funnel plots, Begg’s and Egger’s test evaluated potential publication bias. Results From 6,636 abstracts, 91 interventions (55 in US/Canada, 36 in Europe/New Zealand) were included, on direct provision of healthful foods/beverages (N = 39 studies), competitive food/beverage standards (N = 29), and school meal standards (N = 39) (some interventions assessed multiple policies). Direct provision policies, which largely targeted fruits and vegetables, increased consumption of fruits by 0.27 servings/d (n = 15 estimates (95%CI: 0.17, 0.36)) and combined fruits and vegetables by 0.28 servings/d (n = 16 (0.17, 0.40)); with a slight impact on vegetables (n = 11; 0.04 (0.01, 0.08)), and no effects on total calories (n = 6; -56 kcal/d (-174, 62)). In interventions targeting water, habitual intake was unchanged (n = 3; 0.33 glasses/d (-0.27, 0.93)). Competitive food/beverage standards reduced sugar-sweetened beverage intake by 0.18 servings/d (n = 3 (-0.31, -0.05)); and unhealthy snacks by 0.17 servings/d (n = 2 (-0.22, -0.13)), without effects on total calories (n = 5; -79 kcal/d (-179, 21)). School meal standards (mainly lunch) increased fruit intake (n = 2; 0.76 servings/d (0.37, 1.16)) and reduced total fat (-1.49%energy; n = 6 (-2.42, -0.57)), saturated fat (n = 4; -0.93%energy (-1.15, -0.70)) and sodium (n = 4; -170 mg/d (-242, -98)); but not total calories (n = 8; -38 kcal/d (-137, 62)). In 17 studies evaluating adiposity, significant decreases were generally not identified; few studies assessed metabolic factors (blood lipids/glucose/pressure), with mixed findings. Significant sources of heterogeneity or publication bias were not identified. Conclusions Specific school food environment policies can improve targeted dietary behaviors; effects on adiposity and metabolic risk require further investigation. These findings inform ongoing policy discussions and debates on best practices to improve childhood dietary habits and health.


Introduction
Diets of most children and adolescents (hereafter referred to as children) remain poor, with tremendous consequences for metabolic diseases, overweight and obesity, and other nutritionrelated illness [1][2][3][4]. Childhood is also a critical period to establish lifelong eating habits which influence future risk of obesity and cardiometabolic diseases [5][6][7]. Youth consume between one-third to one-half of meals at school, making this a crucial setting for interventions that alter the food environment [8]. Considering that almost all children obtain some years of schooling, and of diverse ethnic and socio-economic groups, health promotion efforts in schools could have a broader impact on eating behaviors and future disease risk. Promising school food environment policies include direct provision of healthful foods/ beverages such as fruits and vegetables (F&V), quality standards for competitive foods and beverages (foods and beverages sold outside of school meal programs), and quality standards (targets for foods, nutrients/energy) for school meals (lunch, breakfast) [8]. For example, in 2008, a US Fresh Fruit and Vegetable Program (FFVP) was expanded nationally for elementary schools with highest low-income enrolments to provide free F&V to students outside usual school meals [9]; and in 2007, a similar free school fruit programme was implemented in Norway to provide daily a free piece of fruit or vegetable to all secondary school students [10]. The Healthy, Hunger-Free Kids Act in 2010 [11] introduced Smart Snack Standards for competitive foods and beverages in schools receiving federal meal funding, including restriction of sugar-sweetened beverages (SSBs) to be fully implemented by 2014-15 [12]. In 2012, US National School Lunch and School Breakfast Programs nutrition standards were significantly updated to be more consistent with US Dietary Guidelines [13], and in 2015 the UK Department of Education mandated revised standards for all food served in schools [14].
Yet, effectiveness of these food environment policies for improving children's habitual dietary habits, adiposity, or metabolic risk is not well-established. Understanding these effects is critical to estimate benefits of existing programs as well as need for their expansion; and to elucidate potential harms from their elimination as suggested by potential new federal priorities in the US [15,16]. Prior studies have reviewed whether a range of school dietary interventions increase F&V consumption but often without focusing on environmental policies [17][18][19][20][21][22]; while other systematic reviews have been qualitative [23], assessed efficacy of competitive food/beverage standards informed mainly by cross-sectional studies [24], or focused on educational (rather than environmental) interventions [25]. Other reviews have grouped together highly varied programs, e.g., teacher training, child education, family components, labeling, pricing changes, behavioral techniques, and school gardens [26][27][28][29][30][31][32]. Thus, effectiveness of school food environment policies remain unclear, including potential differences for in-school vs. habitual (within and outside school) intakes. To address these gaps in knowledge, we systematically investigated and quantified the effects of school food environment interventions -carefully exploring sources of heterogeneity-, including provision of healthful foods/beverages, competitive food/beverage standards, and school meal standards, on habitual and inschool dietary consumption, adiposity, and metabolic risk factors in children. This investigation was performed as part of the Food-PRICE (Policy Review and Intervention Cost-Effectiveness) Project (www.food-price.org).

Methods
PRISMA recommendations were followed throughout all stages of this meta-analysis (Appendix A in S1 File) [33]. The objective, search strategy, and selection criteria were specified in advance (Appendix B in S1 File).

Primary exposures and outcomes
The primary intervention was school food environment policies targeting food/beverage availability across the school setting (e.g., classroom, cafeterias, vending machines, tuck shops) including direct provision (free, reduced-price, or full-price) of healthful foods or beverages outside of usual school meals (e.g., fresh F&V programs, water fountains, increased availability of healthy foods at vending machines), nutritional quality standards for competitive foods/ beverages, and nutritional quality standards for school meals (lunch, breakfast). The primary outcome was the change in habitual consumption of the targeted food, beverage, or nutrient, evaluated by reported intakes or objective sales/purchases data as a proxy for consumption. Secondary outcomes included changes in in-school meal nutrient content and intake (to compare and contrast to findings for habitual intake), total caloric intake, adiposity (body mass index (BMI), prevalence of overweight (!85 th -95 th percentile), obesity (!95 th percentile) or overweight/obesity combined); and metabolic measures (e.g., blood lipids, blood glucose, blood pressure).
PubMed for each final included article. Titles/abstracts were screened by one investigator; and for all potentially relevant articles, full-texts were retrieved.

Study selection
Full-text manuscripts were evaluated independently and in duplicate, with differences resolved by consensus or, if necessary, group discussion. Inclusion criteria were (a) all randomized or quasi-experimental interventions that (b) assessed the impact of school food environment policies in preschool, primary, or secondary schools on the outcomes of interest among generally healthy children age 2-18y; and (c) reported a quantitative change in the outcome (Appendix B in S1 File). We excluded cross-sectional, retrospective, case-control, modeling, methodology, and laboratory studies; reviews, commentaries, books, and studies for which full-text articles could not be retrieved. Studies were excluded if the policy focused on changes outside of food/ beverage availability (e.g., student education, food labeling, price changes), if the food/beverage environmental policy was a minor component (qualitatively, <30%, as judged by two independent reviewers) of a multi-component intervention, if intervention duration was <4 weeks, or if only knowledge or attitudes were evaluated as outcomes.

Data extraction
Data were extracted independently and in duplicate using standardized electronic templates (Microsoft Access, Office 2010). Extracted information included first author, publication year, study location, design, population (age, sex, race, sample size), intervention characteristics (components, targets, duration), outcome data including habitual (within and outside school) and in-school (e.g., lunch, breakfast, total in-school) intakes (definition, ascertainment methods, effect size, precision estimate), covariates, and for multi-component interventions, the relative contribution of the food environment policy component to the overall intervention (low: 30-59%, medium: 60-89%, high: !90%; qualitatively assessed independently and in duplicate). Missing data or definitions were resolved by direct author contact, where possible.
For outcomes evaluated at multiple time-points, we extracted the latest follow-up measure at end-intervention. Sustainability findings based on follow-up after end-intervention were also extracted when available and !4 weeks duration. Study quality was assessed independently and in duplicate based on study design, assessment of exposure, assessment of outcome, control for confounding, and evidence of selection bias (Table A in S1 File). Differences in data extraction and quality assessment between investigators were infrequent (concordance >95%) and resolved by consensus.

Statistical analysis
Analyses were conducted using STATA14 (College Station, TX: StataCorp LP). For each policy, study-specific effect sizes were pooled using inverse-variance random-effects meta-analysis. For interventions with an external control group, we evaluated between-group continuous changes at follow-up, adjusted for baseline values and relevant covariates; for quasi-experimental studies with no control group, we evaluated within-group changes [34]. Statistical uncertainty (standard error, SE) was extracted or calculated based on other statistics (Appendix D in S1 File). For paired observations without reported covariance, we used a correlation of 0.5 for main analysis and 0.1 and 0.9 for sensitivity analyses [34]. In addition to continuous effect sizes, we extracted other relevant effect sizes (e.g., percentage meeting a cutpoint, odds ratio, ratio of the means, other relative changes) and their statistical uncertainty. Separate intervention arms or outcomes from the same study were included as separate estimates in the meta-analyses; subgroup findings from the same intervention arm or outcome (e.g., by sex, age) were first combined using study-specific meta-analysis.
We separately pooled findings for direct provision of healthful foods and beverages, competitive foods and beverage standards, and school meal standards. Effect sizes were standardized to consistent units: e.g., 80 g serving/d for F&V, 12-oz serving/d for SSBs, 8-oz serving (glass)/d for water, kcal/d for calories, % energy (E)/d, g/d or mg/d for nutrients, and kg/m 2 or z-score for BMI. Endpoints that could not be standardized (e.g., consumption expressed as a score, proportion of children consuming a given level) or separately meta-analyzed were included in qualitative assessment of the evidence. When multiple overlapping outcomes were reported (e.g., fruit with vs. without 100% juice), we extracted the outcome mostly closely aligned to a standardized definition, e.g. total fruits (fresh, raw, canned, or dried), excluding fruit juice; total vegetables, excluding white potatoes; and total SSBs (soda, energy drinks, sweetened teas, etc.). For studies reporting subcomponents of these definitions (e.g., separate subtypes of vegetables, of sweet snacks, F&V separately), we first summed these subtypes.

Direct provision of healthful foods and beverages
Interventions providing healthful foods/beverages were mainly in classrooms ("direct" provision) or via increased availability in cafeterias, tuck shops or vending machines ("indirect" provision) ( Table 1). F&V were most common.  Table B in S1 File). Effects were similar in randomized vs. quasi-experimental studies (Table 2, Figure A in S1 File). Effects appeared potentially higher with free provision [10,37,38,45,50,58, The impact of school food policies on diet: A meta-analysis (Continued) The impact of school food policies on diet: A meta-analysis  59] after direct provision was removed; no significant effect was seen (-0.18 (-0.51, 0.15)). Vegetables. Pooling 3 randomized and 8 quasi-experimental interventions with average duration 13.4 months, habitual vegetable intake was slightly increased (0.04 servings/d (0.01, 0.08)) (Fig 2). In 7 interventions providing free vegetables, effects appeared higher, although this heterogeneity was not statistically significant (P = 0.22) (Table 2, Figure B in S1 File). Findings were similar stratified by other study characteristics and in 3 studies [10,38,45] assessing in-school intake ( Table B in S1 File).
Combined fruits and vegetables. Sixteen studies (6 randomized, 10 quasi-experimental) assessed combined F&V intake, with average duration 15.4 months (11 of these studies also separately evaluated fruits or vegetables, above). Combined intake increased by 0.28 servings/d (n = 16 (0.17, 0.40)) (Fig 2). Findings were not significantly different in randomized vs. quasiexperimental studies or by other population or intervention characteristics (  We included all interventional studies including randomized controlled trials (RCTs) or quasi-experimental designs with (QED) or without an external control group (QED, no C) that assessed the impact of school food environment policy on dietary intake, adiposity, or metabolic outcomes in children. Specific interventions were represented by more than 1 study if different outcomes (e.g., intake vs content, school vs habitual) were reported. b School food environment policy interventions included the direct provision of healthful foods and beverages (DP), competitive food and beverage standards (CFS), and/or school meal standards (SMS). c Multi-component strategies were included only if the food environment policy was a major component, judged qualitatively to be at least 30% of the overall intervention. The relative contribution of the food environment policy component to the overall intervention was qualitatively assessed by each reviewer, independently and in duplicate, based on the number, types, and intensity of additional intervention components, as low (30 to <60%), medium (60 to <90%), and high (!90%).
Single-component strategies received 100%. d Additional intervention components in multi-component strategies included education (nutrition curricula) (Edu), promotion/ marketing (Mrk), family/ parent outreach (Fml), point-of-purchase labeling (Lbl), behavioral techniques (Bhv), other environmental change (Env), and economic incentive (Ecn). e Intervention duration (in months) was estimated from the end of data collection and start date of the intervention as reported. Periods that schools are closed (e.g., summer, holidays) were not taken into account in such estimations. f Quality assessment was performed by review of study design, assessment of exposure, assessment of outcome, control of confounding, and evidence of bias. Each of the 5 quality criteria was evaluated and scored on an integer scale (0 or 1, with 1 being better) and summed; quality scores from 0 to 3 were considered lower quality and 4 to 5 higher quality. g Additionally or exclusively [39,115] reported sustainability effects (i.e., change in reported outcome after the end of the intervention). Of these, 3 studies [37,50,59] within the same strategy (DP) could be meta-analyzed for changes in total fruit intake. One study, which published findings separately 1 yr, 3 yrs and 7 yrs after the intervention was not included in pooled analyses, as the reported outcome was fruit and vegetable intake combined [39,40,42] The impact of school food policies on diet: A meta-analysis in S1 File). In 6 studies assessing in-school consumption [10,38,40,41,45,52], combined F&V intake increased by 0.38 servings/d (n = 6 (0.23, 0.53)) ( Table B in S1 File). Total calories. Habitual caloric intake was reported in 6 studies [38,58,61,[73][74][75]], yet wasn't a target of direct provision in any of these. Pooling studies, no significant effect on habitual caloric intake was identified (-56 kcal/d; -174, 62) ( Table B in S1 File). Only 1 study reported school caloric intake [56], which was unchanged.
Other endpoints. Three studies [46,68,72] evaluated odds of consuming F&V [72] or varying percentage changes in F&V intakes, reported dichotomously [46,68]; these outcomes were generally not significantly improved. One study reported only sustainability data for F&V intake after end-intervention, finding sustained benefits for both in-school intake at 3 years [42] and habitual F&V intake at 3 [42] and 7 years [39] although this weakened over time. A few interventions provided low-fat/low-calorie items [49,51,69], or milk [57]. No significant improvements were found in consumption of low-fat items. A milk provision study in Iran aimed to increase students' weight, which was achieved.

Competitive food and beverage standards
Competitive food/beverage policies generally targeted SSBs and unhealthy snacks (Table 1). Strategies included product-specific restrictions; standards on nutrients, calories, or portion sizes; or both. All were performed prior to implementation of US national Smart Snacks guidelines in 2014.

Heterogeneity sources a Fruits, servings (80 g)/d Vegetables, servings (80 g)/d Combined fruits & vegetables, servings (80 g)/d N (n) b Mean (95% CI) c N (n) Mean (95% CI) c N (n) Mean (95% CI) c
Reduced/ Full 6 (6) 0.15 (0.02, 0.27) 4 (4) -0.01 (-0.12, 0.09) 6 (6) 0.07 (-0.05, 0.20) a Results are presented for selected heterogeneity sources (common across the three strategies of school food environment policies identified -Tables C and D in S1 Filewith the exception of "Cost of provision", specific to this strategy only) for the outcomes with the largest numbers of estimates. For all other outcomes not presented, no significant heterogeneity sources were identified. None of the identified differences by subgroups were statistically significant by meta-regression (P-heterogeneity>0.05 each). b Number of estimates (n, values in parentheses) can be higher than number of studies (N) included in the meta-analyses if multiple intervention groups or multiple comparisons were available from the same study c Study-specific effect sizes were pooled using stratified inverse-variance weighted random-effect models (metan command in STATA). Effect sizes correspond to mean changes standardized across studies to consistent units; and precision estimates to 95% confidence intervals (CIs).  Table 1). e In addition to the dietary targets, specific interventions also targeted non-dietary targets, such as physical activity and smoking.
f School food environment policy strategies included direct provision of healthful foods, quality standards for competitive foods/ beverages, and quality standards for school meals. The impact of school food policies on diet: A meta-analysis intake did not identify a significant effect (n = 5; -0.02 servings/d (-0.04, 0.01)). No significant heterogeneity sources were identified (Table C and Figure D in S1 File). Unhealthy snacks. Two interventions assessed habitual intake, which decreased by 0.17 servings/d (n = 3 (-0.22, -0.13)) (Fig 3). Four studies with 5 separate intervention arms assessed

Effect of competitive food and beverage standards in schools on sugar-sweetened beverage and unhealthy snack intake in children.
Intakes represent habitual or total in-school consumption, except for 1 study that assessed in-school lunch intake. Solid squares represent study specific continuous changes in reported intakes; and lines, 95% confidence intervals (Cis). Vertical line represents pooled effect size (ES); and open diamond, corresponding 95% CI. Multi-component strategies were included only if food environment policy was a major component, judged qualitatively to be at least 30% of the overall intervention. The relative contribution of the food environment policy component to the overall intervention was qualitatively assessed as low (30 to <60%), medium (60 to <90%), and high (!90%). a A single estimate was obtained by summing separately reported outcomes (n = 2) that their total aligned to the single optimal definition (i.e., sweet snacks). SSBs, sugar-sweetened beverages; RCT, randomized controlled trial; QED, quasi-experimental intervention with external control group; QED, no C, quasi-experimental intervention without external control group; CA, Canada; UK, United Kingdom; US, United States of America. https://doi.org/10.1371/journal.pone.0194555.g003 The impact of school food policies on diet: A meta-analysis in-school intake, which decreased by 0.05 servings/d (n = 9; -0.08, -0.02) (Fig 3). No significant heterogeneity sources were identified (Table C and Figure D in S1 File).
Other targeted dietary factors. Other targeted diet factors included total fat [88,90] and saturated fat [86,88] intake; habitual and in-school lunch total fat intake decreased (n = 3), but not in-school lunch saturated fat intake (n = 2) ( Table B in S1 File).
Adiposity and metabolic measures. Several studies assessed the prevalence or odds ratios of childhood overweight (n = 6 and n = 6 estimates, respectively), obesity (n = 10, n = 8), or overweight/obesity (n = 5, n = 2) ( Figures G and H in S1 File), as well as BMI (n = 6; Figure I in S1 File) and BMI z-score (n = 5; Figure J in S1 File). Durations ranged from 2.3 to 69 months (mean 31.5). Competitive food/beverage standards did not significantly reduce any of these measures ( Table B in S1 File), although the central effect estimate often tended to be slightly and nonsignificantly lower. Prevalence of overweight/obesity was nonsignificantly higher across 5 studies evaluating this outcome (n = 5; 0.24%; -0.54, 1.02), largely driven (70.51% of the weighted estimate) by 1 quasi-experimental study [84] that compared changes in rates among schoolchildren in California (n =~600,000). Only 2 studies evaluated effects on metabolic risk factors and could not be pooled [67,91]; individually, these found significant improvements in various risk factors assessed [67] other than fasting glucose [91].

School meal standards
Policies on school meal (mainly lunch) standards (foods, nutrients/energy) generally targeted F&V, dietary fats, and sodium (Table 1). Five studies evaluated implementation of the 2012 US school lunch guidelines, while all studies were performed prior to the implementation of the revised 2015 UK school meal standards.
Dietary fats. Most studies specified target levels for dietary fats, which were generally consistent across studies; these ranged from 30-35%E/lunch for total fat and 10-11%E/lunch for saturated fat. Six studies assessed habitual total fat, which decreased by 1.49%E (-2.42, -0.57) (Fig 4). In g/d, the reduction in habitual fat intake was greater (~6 g/d total fat) in magnitude to achieved reductions in in-school meal content and intake (~3-4 g/d total fat) (Figure K in S1 File). Standards also reduced habitual saturated fat (n = 4; -0.93%E (-1.15, -0.70)), in-school lunch saturated fat (n = 9; -2.75%E (-4.39, -1.11)), and in-school meal (lunch or breakfast) The impact of school food policies on diet: A meta-analysis saturated fat (n = 10; -2.46%E (-4.04, -0.89)) ( Table B in S1 File). Again, absolute magnitudes of reduction were marginally higher for habitual intake (~3 g/d) compared with in-school meal intake and content (~1-2 g/d) (Figure L in S1 File). No significant heterogeneity sources were identified, with the exception of study region for in-school total fat intake (P = 0.042); larger reduction was observed for studies in Europe/New Zealand compared to US/Canada (Table D and Figures E and F in S1 File).
Other endpoints. Two studies reported dichotomous changes in healthy food (e.g., fruit, vegetable, bread, milk, cereal [98,112], reported as % meeting a threshold) that could not be meta-analyzed due to varying cutpoints. Results were conflicting, with a reduced overall healthy food score and a higher intake of healthy items at breakfast [112]. One study [115] reported only sustainability data, evaluating total calories, total fat, and saturated fat in lunches 5 years after school meal standards were removed, finding further decreases in %E from total and saturated fat, but increased caloric content.

Publication bias
Visual inspection of funnel plots provided little evidence for publication bias (Figures O-Q in S1 File). Begg's or Eggers test did not identify statistical evidence for publication bias.

Discussion
This systematic review and meta-analysis is the first, to our knowledge, to determine quantitative effects of school food environment policies on children's habitual dietary intakes in interventional studies. Direct provision policies increased fruit intake by 0.27 servings/d and vegetable intake by 0.04 servings/d, but not water intake. Competitive food/beverage standards reduced SSBs by 0.18 servings/d and unhealthy snacks by 0.17 servings/d. School meal standards increased fruit intake by 0.76 servings/d, reduced total fat intake by~1.5% energy and saturated fat intake by~1% energy, and reduced sodium by 170 mg/d. All of these policies influenced dietary composition, without altering total calories. Measures of adiposity were generally unchanged; and few studies assessed metabolic factors, with mixed findings.
We separately evaluated in-school vs. habitual intakes to determine effects on children's overall nutritional habits, given potential for compensatory changes outside of school. For example, restricting SSBs or unhealthy snacks at school could lead to increased consumption after school or at home. Such compensation is suggested in some cases; for instance, school meal standards significantly reduced meal calorie content, but not in-school meal calorie intake or habitual calorie intake. Conversely, reductions were similar for in-school vs. habitual sodium intake, suggesting that sodium reduction at school does not lead to meaningful compensation elsewhere. For some policy outcomes, e.g. for competitive food standards and SSBs and snacks, the pooled findings from interventions evaluating in-school effects were smaller than those evaluating habitual intakes. These were generally different studies, suggesting possibly other differences in the types of studies evaluating in-school intakes. Overall, our results support the importance of schools as a setting to improve overall dietary habits of children within and outside school.
Our findings suggest efficacy of a range of food environment policies, including direct provision, competitive food/beverage standards, and school meal standards. The results for both direct provision and school meal standards suggest greater efficacy for fruit intake, compared with vegetables; consistent with greater palatability of many fruits and generally less need for preparation or cooking. Water intake was unchanged in the limited studies that assessed this outcome, likely further due to difficulties in assessing fluid intake and measurement error. Our findings further highlight key gaps for many other dietary targets, such as other healthier foods (e.g., legumes, whole grains, fish, yogurt) or less healthy foods (e.g., processed meats) or other nutrients of concern (e.g., calcium, vitamin D, potassium, unsaturated fats, fiber). Given updated Dietary Guidelines for Americans that focus on healthier foods, overall diet patterns, and specific nutrients of concern [127], future studies are needed to assess how school food environment policies impact these priorities.
Evidence on the health impact of policies targeting the school food environment is especially relevant and timely given the potentially evolving priorities of the new federal US administration. Congress did not reauthorize the Healthy, Hunger-Free Kids Act (HHFKA) as scheduled in Sept 2015, so the future of Smart Snack Standards, now covering 99% of public and 83% of private schools [12], remains uncertain. Further, current policy debates include a focus on weakening or eliminating national school lunch standards [15,16]. A recent analysis indicated that in-school selections have improved with the new lunch standards [128]. Our findings build upon and expand this prior work by demonstrating changes in actual habitual intake, further supporting efficacy of meal and snack standards and informing ongoing debates. Similarly, the current national FFVP only applies to elementary schools with high proportions of low-income students [9], about 4 million students across the US [129]. Our investigation supports efficacy of such direct provision programs, which should be considered for a broader range of elementary, middle, and high schools. Finally, while identified dietary improvements were meaningful at a population level, these will not fully address the suboptimal diets of most children. Thus, our results confirm a need for multiple programmatic and policy interventions, including within and outside schools, to improve children's diets.
While several dietary benefits were confirmed, changes in adiposity metrics were generally not significant. This may be because such policies improve dietary quality or composition (more relevant for general and metabolic health) but not dietary quantity (more relevant for obesity, at least in the short-to intermediate-term). Because dietary composition influences numerous pathways for health and well-being, the absence of a documented effect on obesity does not preclude efficacy of these interventions. Few studies evaluated metabolic risk factors, for which improvements may be more readily detected compared with adiposity. Also, establishing lifelong healthier dietary habits may have benefits decades later, during adulthood. Our findings provide quantitative summaries of how school food environment policies affect specific dietary targets, allowing modeling of potential effects on childhood obesity and future diets and disease risk in adulthood.
Prior reviews of a more varied range of school interventions identified effects of similar magnitude for total F&V consumption [18,19,21]. A previous systematic review on competitive foods/beverages was qualitative, and included mostly cross-sectional studies in the US alone [24]. Similarly, another systematic review on school food environment was also qualitative, excluded direct provision studies, and grouped together various heterogenous interventions [23]. Importantly, most prior reviews did not specifically evaluate potential effects of school food environment interventions on dietary intakes, and have grouped together highly varied programs potentially leading to biased inferences [26][27][28][29][30][31][32]. Our findings extend these results by specifically evaluating school food environment policies and quantifying their effects on dietary intakes, as well as separately evaluating direct provision, competitive food/beverage standards, and school meal standards with careful consideration of potential heterogeneity. We also looked for sustainability: while few studies were identified, the results suggested that dietary improvements are difficult to sustain if school food environment policies are cancelled.
Our evaluation has several strengths. Evidence was based on interventions, most of which were randomized, increasing reliance in validity of results. We evaluated changes in diet, adiposity, and metabolic risk factors, providing a more coherent and comprehensive picture of the evidence. We focused on habitual (within and outside school) dietary intakes, rather than in-school intake alone. A systematic search of multiple databases made it less likely that major relevant reports were missed. Standardized methods and analytic techniques and duplicate full text reviews and data extractions reduced errors and bias. Standardization of interventions and outcomes facilitated quantitative pooling. We explored multiple factors for potential modifying effects.
Potential limitations should be considered. Educational systems and schools vary within and across nations, which could contribute to unmeasured heterogeneity. Intensity or success of policy implementation could modify results, but these are difficult to quantify; e.g., due to varying professional education or technical assistance for food service directors; differences in how schools prepare, offer, sell, serve, or purchase food; and policy nutritional guidelines. Most studies did not report details by socioeconomic indicators, which could modify efficacy of some programs. Costs and cost-effectiveness were generally not reported. Several studies included other intervention components that might contribute to impact. Some studies were judged to have lower quality scores, that could weaken or bias results. Evaluation of heterogeneity and publication bias is dependent on total numbers of studies, limiting statistical power for some endpoints. Most studies were from high-income Western countries, highlighting the need for research in lower-income nations.
In conclusion, this systematic review and meta-analysis demonstrates that specific school food environment policy interventions can improve targeted dietary behaviors. These findings inform ongoing policy discussions and debates on best practices to improve childhood dietary habits and health.
Supporting information S1 File. Supplementary material. Appendix A. PRISMA Checklist. Appendix B. Study protocol. Appendix C. Search query for PubMed/ Medline. Appendix D. Statistical Analysis. Table A. Quality Assessment Criteria. Table B. Meta-analyses of randomized and quasi-experimental interventions evaluating school food environment policies and dietary habits or adiposity in children. Table C. Prespecified sources of heterogeneity explored among interventions evaluating the effect of competitive food and beverage standards in schools on dietary intakes or adiposity in children. Table D. Prespecified sources of heterogeneity explored among interventions evaluating the effect of school meal standards on dietary intakes or meal contents in children.