Current dietary intakes, demographics, and BMI in children

Data on mean dietary intakes, demographics, and BMI and corresponding uncertainty in children ages 5–18 years were obtained from the two most recent cycles of the National Health and Nutrition Examination Survey (NHANES) in 2009–10 and 2011–12 (N = 4,165), preceding widespread implementation of the Smart Snacks Standards or FFVP [20,21]. The FFVP was implemented in 2008 in the lowest income elementary schools, but not other elementary schools or middle or high schools, so dietary changes in this subset of schools may already be partially captured in NHANES 2009–11. The HHFK Act was passed in 2010, but not fully implemented until 2016, so it is unlikely that its effects would greatly alter our baseline SSB intake estimates from NHANES.

The NHANES 24-hour recall protocol and data collection methods are thoroughly documented elsewhere [22]. We used survey weights that accounted for the sample design and probabilities of selection to obtain nationally representative data for children. Data from two 24-hour recalls were averaged for each participant and dietary exposure, further accounting for within-person variation in determining distributions. The US Department of Agriculture Food Patterns Equivalents Database for years 2009–10 and 2011–12 was used to disaggregate multi-component foods into specific food groups, including fruits, vegetables, and SSBs [23,24]. The definitions used for these dietary exposures have been recently published [25]. NHANES demographic data were used to stratify dietary intakes by age (5–10 years [elementary school], 11–14 years [middle school], 15–18 years [high school]), race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, other), and sex. BMI was obtained and similarly stratified by age, race/ethnicity, and sex, as directly measured in NHANES by trained personnel using standardized methods.

Effects of F&V provision and SSB restriction on dietary intakes in children

The effects of school policies for F&V provision and SSB restriction on the consumption of fruits, vegetables, and SSBs in children were obtained from a meta-analysis of intervention studies, including 6 randomized trials, 8 quasi-experimental studies, and 1 regression discontinuity study of F&V provision programs, and 3 quasi-experimental trials of SSB restriction programs [6]. We focused on the effect sizes of each intervention on total food consumption, not only in-school intakes, using 24-hour recalls and food frequency questionnaires to account for any compensatory dietary changes outside of school. The identified trials were generally 1–2 years in duration, and therefore reflected the 1–2 year effects of these policies.

Etiologic effects of dietary changes on BMI in children

We conservatively assumed no effects of F&V consumption on childhood BMI in our main analysis, based on insufficient evidence linking F&V intake to BMI in childhood [26,27]. In sensitivity analysis, we used pooled estimates from 3 large cohort studies in adults to estimate the relationship between changes in fruit and vegetable intake and changes in BMI in children [28]. For SSBs, the change in BMI per serving change in daily SSB intake was obtained from a randomized controlled trial in children [29], after confirming similar findings in a second randomized trial [30] as well as long-term observational studies (S1 File, S1 Table) [31].

Current dietary intakes, demographics, and BMI in adults

Data on mean dietary intakes, demographics, and BMI and corresponding uncertainty in adults ages 25 and older were obtained from NHANES 2009–2010 and 2011–2012 (N = 8,510) using similar methods as described for children [20,21]. Dietary intakes and BMI were stratified by age (25–34, 35–44, 45–54, 55–64, 65–74, 75+ years), race/ethnicity, and sex.

Effects of childhood dietary changes sustained into adulthood

We estimated the effects of these policies on long-term changes in dietary intake from a systematic review of cohort studies that correlated within-individual dietary intakes from childhood to adulthood [3]. The average correlation was approximately 0.35 or higher for most dietary habits [3]. Within-individual correlations of dietary intakes were comparable over various durations of follow-up, and for different nutrients and food categories. Data on the sustained absolute differences in these dietary intakes were not reported. In our main analysis, we used these findings to assume that 35% of the dietary changes associated with school policies in childhood would be sustained into adulthood; and varied this assumption from 25% to 50% in sensitivity analyses.

Etiologic effects of dietary changes on disease outcomes

The associations of dietary changes with disease-specific mortality in adults were obtained as previously described from a systematic review of meta-analyses of prospective cohorts and randomized trials [32]. This included effect estimates for consumption of fruits, vegetables, and SSBs in relation to CHD, stroke, and type 2 diabetes (S2 File). Data on disease-specific deaths in US adults by age, sex, and race/ethnicity were obtained from the National Center for Health Statistics for 2012, including deaths from coronary heart disease, ischemic stroke, hemorrhagic stroke, unidentified and other non-ischemic/hemorrhagic stroke, and diabetes mellitus [33].

Data analysis

A CRA modeling framework was utilized as previously described [32], including probabilistic sensitivity analysis using 1000 Monte Carlo simulations to jointly incorporate the uncertainty of multiple inputs including baseline dietary intakes and BMI, effects of the school policies on dietary changes, effects of dietary changes on BMI in childhood, and effects of dietary changes on CMD mortality later in life.

F&V provision and SSB restriction policies and diet and BMI in children

We estimated that F&V provision in schools would increase habitual fruit intake by 17.1% in elementary school students (95% uncertainty interval (UI): 10.8, 23.6%), 22.2% in middle school students (95% UI: 14.1, 31.5%), and 25.0% in high school students (95% UI: 15.4, 37.7%) over 1–2 years (Table 1; Fig 1). This policy was estimated to have a much smaller effect on vegetable intake in children. SSB restriction in schools would change habitual SSB intake by an estimated -26.5% (95% UI: 6.4, 46.4), 19.2% (95% UI: 4.5, 34.5%), and 14.5% (95% UI: 3.4, 25.2%) among elementary, middle, and high school students, respectively. By child gender, we estimated that F&V provision would lead to a 21.3% (95% UI: 13.5, 30.1%) and a 19.4% (95% UI: 12.2, 27.2%) increase in fruit intake in males and females, respectively, while vegetable intake would increase by 3–4% in males and females (S3 Table). SSB restriction would lead to a slightly larger reduction in SSB intake in males than females (20.5% [95% UI: 5.3, 36.1%] in males versus 18.5% [95% UI: 4.5, 32.5%] in females). Stratified by race, we estimated increases in fruit intake would range from 18.1% (95% UI: 11.4, 25.1%) among Hispanics to 27.4% (95% UI: 17.3, 37.6%) among non-Hispanic blacks to, while decreases in SSB intake would be as large as 26.8% (95%UI: 6.3, 48.5%) among children in the other race category (S4 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Current and estimated changes in dietary intakes associated with national school policies on F&V provision and SSB restriction among US children by age.

Current intakes are based on NHANES 2009–10 and 2011–12 (N = 4,165 children age 5–18 years), where bars represent the mean and error bars, the 95% confidence intervals. Estimates for dietary intake with policies are based on a comparative risk assessment framework incorporating policy effects from intervention studies, where bars represent the median values from 1,000 Monte Carlo simulations and error bars, the 95% uncertainty intervals.

https://doi.org/10.1371/journal.pone.0200378.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Select model inputs and estimated changes in child diet and child BMI associated with F&V provision and SSB restriction in US schools¹.

https://doi.org/10.1371/journal.pone.0200378.t001

A national school policy to restrict SSBs was estimated to change childhood BMI by -0.7% (95% UI: -1.2, -0.2%), -0.5% (95% UI: -1.0, -0.2%), and -0.5% (95% UI: -0.8, -0.1%) over 1–2 years for children in elementary, middle, and high school, respectively (Table 1). In sensitivity analyses including potential effects of F&V on BMI, a national school policy for F&V provision would lead to an approximate 0.1% additional reduction in BMI, largely related to increased fruit intake (S5 Table).

Estimated cardiometabolic deaths averted among adults

If such school policies had been implemented when current US adults were children, F&V provision would increase fruit intake in adults by 19.1% (95% UI: 12.3, 25.7), vegetable intake by 2.2% (95% UI: 0.5, 4.0%), and decrease SSB intake by 24.2% (95% UI: 5.1, 43.4%) (Table 2). F&V provision and SSB restriction were estimated to jointly prevent 22,383 CMD deaths/year (95% UI: 18735, 25930), or 3.2% of total annual CMD deaths (95% UI: 2.7, 3.7%) (Table 3). Among individual dietary targets, reductions in SSBs were estimated to have the greatest impact, with 14,132 deaths averted/year (95% UI: 10453, 17365) (Table 3, Fig 2). Increases in fruit intake associated with F&V provision were estimated to avert 7,457 CMD deaths per year (95% UI: 6615, 8428), while changes in vegetable intake had a small effect (983 CMD deaths averted/year; 95% UI: 790, 1217). For subtypes of CMD, CHD mortality would be most reduced by SSB restriction (3.0% reduction; 95% UI: 2.0, 3.9%), while stroke mortality would be most reduced by provision of fruit (3.4% reduction; 95% UI: 3.0, 3.8%). Diabetes mortality was only influenced by SSB restriction, with a 3.7% annual mortality reduction (95% UI: 2.8, 4.5%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Estimated annual cardiometabolic deaths averted from implementation of national school policies on F&V provision and SSB restriction, separately and jointly, in US elementary, middle, and high schools.

Bars represent the median values from 1,000 Monte Carlo simulations in a comparative risk assessment framework; and error bars, the 95% uncertainty intervals. Health effects are estimated for the current US adult population if exposed to these school environment policies during their childhood in elementary, middle, and high school. Panel (A) assumes that 25% of the dietary changes achieved in childhood are sustained into adulthood; panel (B), that 35% of dietary changes are sustained; and panel (C) that 50% of dietary changes are sustained.

https://doi.org/10.1371/journal.pone.0200378.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Select model inputs for modeling CMD outcomes in adults.

https://doi.org/10.1371/journal.pone.0200378.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Estimated cardiometabolic deaths averted associated with F&V provision and SSB restriction in US schools¹.

https://doi.org/10.1371/journal.pone.0200378.t003

In sensitivity analyses assuming that either 25% or 50% of the dietary changes associated with these school policies were sustained into adulthood, an estimated 17,390 (95% UI: 14978, 19975) and 26,210 (95% UI: 21741, 30818) CMD deaths would be averted per year, respectively (Table 3, Fig 2).

Plausibility and coherence of findings

Relative effects by age and race reflect current intake patterns in US schoolchildren. For example, fruit intake is lower in high school than elementary school students, while SSB intake is nearly twice as high in high school compared to elementary school students [20,21]. The small effect on habitual vegetable consumption of typical school policies to provide fresh F&V is consistent with prior analyses [36, 37]. This may reflect both a stronger student preference for fruit as well as the relative ease and practicality for schools of serving fruit outside of usual school meals, compared with vegetables.

We identified a modest effect on childhood BMI of school restrictions on SSBs over 1–2 years, consistent with prior empiric experiences [38–40]. By 2009–12, many US schools had already reduced in-school SSB intakes due to educational or other policies, which would limit the estimated magnitude of implementing national school SSB restrictions. In addition, the majority of SSB consumption by children occurs outside of school, especially at home and also at restaurants and retail outlets [41]. Furthermore, the observed intervention effects incorporate the potential for partial compensation by increased SSB consumption outside of school. For all these reasons, our school-based intervention would be expected to have a modest impact on habitual SSB intake. However, even a small change in childhood BMI is relevant across the US population, and translation of even a portion of this lower SSB intake into adulthood diets led to substantial numbers of fewer CHD and diabetes deaths.

Very few studies have evaluated the long-term effects of childhood interventions on CMD outcomes later in life. Two cohorts have examined the long-term associations of dietary habits in childhood with cardiovascular risk and outcomes in young adulthood [42–45]. For example, the Cardiovascular Risk in Young Finns Study followed children ages 9–18 years for up to 27 years and found that children with higher vegetable consumption had a 14% lower risk of metabolic syndrome during adulthood, especially due to lower blood pressure and trigylcerides. The association persisted after adjustment for adulthood vegetable consumption, suggesting benefits of lifelong dietary habits (not captured in our analysis) [42]. Similar findings were reported in this cohort for childhood F&V consumption or generally healthy childhood diets combined with other healthy behaviors in relation to arterial pulse wave velocity and stiffness in adulthood [43,44]. A second long-term cohort following British children from 1937–39 found higher childhood intake of vegetables, but not fruit or other foods, was linked to lower risk of subsequent stroke [45]. Our investigation builds upon and extends these findings by pooling and incorporating the evidence from multiple studies to assess the potential impact of specific school dietary policies on CMD outcomes in the US.

Policy implications

Evidence on the health impact of national policies targeting the school food environment is especially relevant and timely given the potentially evolving priorities of the new federal administration. Congress did not reauthorize the HHFKA as scheduled in Sept 2015, so the future of the Smart Snack Standards, which currently cover US students in 99% of public schools and 83% of private schools [46], remains uncertain. Our findings suggest health benefits of these standards during both childhood and later adulthood, supporting a need for reauthorization. Our results also suggest that most childhood SSB consumption will not be eliminated by school policies alone, making programmatic interventions to reduce SSB availability and intake in other venues important for children.

Currently, the national FFVP is available only for elementary schools with high proportions of low-income students [18], covering about 4 million students across 50 states and the territories [47]. Our investigation suggests meaningful health benefits of extending this program nationwide to elementary, middle, and high schools across the US. Indeed, our results indicate that the combination of F&V provision and SSB restriction has complementary health benefits.

Strengths and limitations

Our analysis included nationally representative data on dietary intakes, BMI, demographics, and CMD mortality, increasing generalizability to the US population. Our estimates for the impact of these school policies on diet, as well as the etiologic effects of diet on CMD, were obtained from systematic reviews and meta-analyses, increasing validity of these estimates. Model inputs were stratified by age, sex, and race wherever relevant, accounting for differences in prevalent diets, BMI, and cardiometabolic mortality. We utilized conservative assumptions whenever possible, such as no significant effects of F&V intake on childhood BMI, reducing the likelihood we overestimated health benefits. We modeled and estimated effects in both children and adults, providing a clearer picture of overall population effects. Uncertainty was incorporated using Monte Carlo simulations; and sensitivity analyses tested the robustness of our results to changes in inputs and assumption.

Potential limitations should be considered. National estimates of dietary intakes were based on self-report, which could introduce error. To minimize this, we averaged two 24-hour recalls per person and accounted for within-person variation. The school interventions used to estimate the impact of these programs were of relatively brief duration (~1–2 years); and longer-term consecutive programs could have larger effects, particularly if there are lag-effects or acceleration of effects over time. The intervention studies varied in target populations, protocols, and implementation, and therefore are best considered as representing an average population effect of school policies for F&V provision and SSB restriction rather than being applied to any one child. For instance, additional technical assistance and resource support in implementing these policies could increase their impact. Although stochastic and parameter uncertainty were included in our UI estimates, we recognize the additional unquantifiable uncertainty, such as in the sustainability of these intervention effects. Direct evidence was not available on sustained impacts of these school policies on adult dietary habits, requiring estimation from other studies tracking diets from childhood into adulthood. For example, food habits from a school-based intervention may track differently into adulthood than a family-based change. To address this uncertainty, we conducted a sensitivity analysis on the sustained effects of these interventions on diet.

Conclusions

These findings suggest that national school policies on F&V provision and SSB restriction in US elementary, middle, and high schools would modestly but meaningfully improve dietary habits and BMI in children and CMD mortality later in life. Our findings have implications for the public, school officials, health care providers, and policymakers aiming to improve the health and nutrition of children and reduce the burden of CMD in adults.