Food environment near schools and body weight — a systematic review of associations by race/ethnicity, gender, grade, and socioeconomic factors

Abstract

Previous research reported modest associations between food environments near schools and adiposity among children overall. The associations within socio-demographic subgroups have not been synthesized. This review assessed the evidence on the associations between food environments near schools and childhood obesity within different demographic and socioeconomic subgroups. PubMed and Scopus databases were searched to identify studies published in English between January 1, 1980 and April 25, 2019 examining presence of fast food outlets, convenience stores, supermarkets, and grocery stores near schools and measures of overweight/obesity by race/ethnicity, gender, grade, and income level. Twelve cross-sectional and two ecological studies were included. Fast food outlets were most commonly examined (n=12). The associations between fast food outlets near schools and obesity were generally positive among Latino, White, and African American students and across grade levels, though the strengths of evidence varied. The directions of association were mixed among Asian children. Three studies reported generally positive associations between convenience stores and obesity among Latino and African American children, and mixed associations among White and Asian children. Longitudinal studies are needed in addition to studies examining underlying mechanisms of the differential influence of food environments near schools within each subgroup.

Introduction

Childhood obesity has become a public health crisis around the world.¹ There is strong interest in addressing childhood obesity at the environmental level.² Given that children often spend a significant portion of their awake time around schools,³ accumulating evidence points to the role of the food environments near schools in shaping children’s dietary behaviors^4–6 and body weight.⁴ Previous research found that one trip to a corner store near schools contributed, on average, 360 kilocalories to schoolchildren’s dietary intake and these children most frequently bought energy-dense, low-nutritive foods at these stores.⁷

A prior systematic review found positive associations between fast food outlets (FFO) near schools and obesity among children.⁴ However, evidence on the associations by race/ethnicity, gender and income level have not been previously synthesized. Understanding how these associations differ among subgroups is important for several reasons. In many high income countries including the United States (U.S.), studies have shown heavy burden of childhood obesity amongst racial/ethnic minorities and socioeconomically disadvantaged populations.⁸ Cultural norms on diet are often present within close peer networks, grades, schools, and communities and may influence children’s dietary behaviors⁹ and this may depend on gender, race/ethnicity, income or grade. Developing effective strategies to reverse the current trends of obesity epidemics around the world requires assessment of risk factors specific to subgroups with high obesity prevalence. Hence, this review evaluated the existing evidence on the associations between the food environments near schools and obesity by demographic and socioeconomic subgroups.

METHODS

The current review included published articles that examined the association between food environments near schools and obesity by demographic and socioeconomic factors. The outcome of interest was weight status (body mass index (BMI)/overweight/obese). Initially, PubMed and Scopus databases were searched electronically to identify relevant articles published between 1980 and 2016. The PubMed search was subsequently updated on April 25, 2019. The references from the included articles were manually searched further to identify additional relevant articles. The PubMed search terms are included in the Appendix.

Studies were included in this review if they: 1) examined the associations between food environments near schools and weight status (self-reported or measured), 2) were published between January 1, 1980 and April 25, 2019, 3) were written in the English language, 4) studied children in elementary, middle, or high schools between ages six to eighteen, and 5) included subgroup analyses by race/ethnicity, gender, grade, or socioeconomic characteristics of schools or children. Studies were excluded if they examined food environments inside schools or residential food environments, or if weight status was not examined as an outcome.

Three reviewers conducted a systematic review of published articles in PubMed. One researcher searched on Scopus. Three reviewers also manually checked the references of selected articles to identify any relevant articles not found through the database searches and reviewed the final list of articles. The quality of the selected articles was checked against the modified version of the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies from the National Heart, Lung, and Blood Institute.¹⁰ The tool and the aggregated results of this assessment are included in the Appendix. The following data were extracted from each study by two reviewers: year of publication, study design, sample size, age/grade, study site location, measures of food environment, types of retail food outlets, weight status definitions, and effect estimates by race/ethnicity, gender, grade, and socioeconomic characteristics. Given the heterogeneity in study methods, settings, and research questions, meta-analyses were not conducted. The results are presented as a narrative review.

Results

Of the 3,027 articles identified in the initial searches, 14 articles met criteria for inclusion (Figure 1). Except for two ecological studies that used school-level overweight/obesity prevalence as the dependent variable,^11,12 all studies used cross-sectional data (Table 1).^6,13–23 Seven studies used U.S. samples; five were based in California, four of which used statewide data^12,15,16,19 and one study used county data from Los Angeles (LA)¹¹; one study used data from the Twin Cities in Minnesota,²⁰ and one study used U.S. national data.¹⁴ The remaining seven studies used data from Canada (Ontario),¹⁸ Korea (Seoul),¹³ Taiwan (national),^17,23 Finland (national),⁶ and England (one with national data,²¹ and the other from Norfolk County²²). Studies used a variety of data sources to assess exposures to food environments near schools, including direct observations,¹³ retail lists from commercial sources,^{6,11,12,14,18–20} and municipal or governmental databases.^{15–17,21–23} Twelve studies examined the influences of FFO near schools while eight studies assessed convenience stores. A smaller set of studies examined influences of grocery stores (n=3) and supermarkets (n=3). Ten studies used direct measures of height, weight, skinfolds, and/or bioelectrical impedance^{11–13,17–23} while four used self-reported height and weight.^6,14–16 A study that examined both BMI and fat mass index (FMI) and only showed FMI results was included as the results were reported to be similar.²²

Figure 1:

PRISMA Flow Diagram describing the study selection.

Table 1.

Characteristics of the studies included in this review.

Author Year of publication	Location: Data source (Year of data collection)	Sample size; Age	Food environment	Subgroup analyses	Outcome1 (Measured (M) or Self-reported (SR))
Chiang et al. 2011	Taiwan: national, Nutrition and Health Survey (2001–2002)	2,283; Elementary school children (6–13 years old)	Fast food outlets Convenience stores	Gender	BMI z scores (M)
Chiang et al. 2017	Taiwan: national, Nutrition and Health Survey (2010)	1458; Grade 7–9	Fast food outlets Convenience stores	Gender	BMI z scores (M)
Currie et al. 20102	USA: California, Fitnessgram (1999, 2001–2007)	8,373 schools with ~3,000,000 9th graders	Fast food outlets	Race/ethnicity Gender Grade	Obese (M) [skinfold body fat >25% for boys, and >32% for girls]
Davis & Carpenter 2009	USA: California, California Healthy Kids Survey (2002–2005)	529,367; 5 –12th grades (mostly 7, 9, 11th)	Fast food outlets	Race/ethnicity	Overweight/obese (SR) [BMI≥85th / 95th percentile for age/sex]
Grier & Davis 2013	USA: California, California Healthy Kids Survey (2003–2005)	100,000; 5–12th grades (mostly 7, 9, 11th)	Fast food outlets	Race/ethnicity Socioeconomic advantage	BMI (SR)
Harrison et al. 2011	UK: Norfolk county, SPEEDY study (2007)	1,995; Year 5	Healthy food outlets (Supermarkets, greengrocers) Unhealthy food outlets (takeaways and convenience stores)	Gender	log FMI (M)
Langellier, 20122	USA: Los Angeles County, California, Fitnessgram (2008–2009)	1,694 schools with students in 5, 7, and 9th grades	Fast food outlets Corner stores (non-chain convenience stores, non-supermarket grocery stores, liquor stores)	Race/ethnicity	Overweight (M) [BMI≥ sex- and age-specific cutoffs defined by the Physical Fitness Testing program]
Larson et al. 2013	USA: Minneapolis/St. Paul Metropolitan Area (the Twin Cities), Minnesota, EAT study (2009–2010)	2,793 students; 6–12th grades	Fast food outlets Convenience stores	Gender	BMI z-scores (M)
Leatherdale et al. 2011	Canada: Ontario, Convenience sample of schools from PLAY-ON study (2007–2008)	1,207; 5–8th grades	Fast food outlets Grocery stores	Gender Grade	Overweight (M) [BMI≥85th percentile for age/sex]
Park et al. 2013	South Korea: Seoul, cross-sectional study (2011)	939; 4th – 9th grades	Fast food outlets Convenience stores Supermarkets	Gender	Overweight/obese (M) [BMI ≥85th /95th percentile for age/sex using 2007 Korean National Growth Charts]
Powell et al, 2007	USA: national, Monitoring the Future study (1997–2003)	73,079; 8th and 10th grades	Convenience stores Supermarkets Grocery stores	Race/ethnicity	BMI (SR)
Sánchez et al. 2012	USA: California, Fitnessgram (2007)	926,018; 5, 7, and 9th grades	Fast food outlets Convenience stores	Race/ethnicity Gender Grade	Overweight /Obese (M) [≥85th BMIz for age and sex]
Virtanen et al. 2015	Finland: national, Finish School Health Promotion study (2008–2009)	23,182; 8th and 9th grades	Fast food outlets & Grocery stores	Socioeconomic advantage	Overweight/obese (SR) [BMI≥25 kg/m2]
Williams et al. 2015	England: national, National Child Measurement Program (2010–2011)	8,211; Year 6	Fast food outlets Food stores (grocery stores, supermarkets, off-license stores, convenience stores, news agents)	Gender Grade	BMI z-scores (M)

M: direct measurement; SR: self-reported; BMI: body mass index; SPEEDY: Sports, Physical activity, and Eating behavior; Environmental Determinants in Young people; EAT: Eating and Activity in Teens; PLAY-ON: Play-Ontario.

Overweight or Obese (SR) means that the weight status categorization was based on self-reported height and weight.

^1.For BMI and BMI-based outcomes (i.e. overweight/obesity, BMIz), SR refers to height and weight used to derive these outcome measures.

^2.These studies used an ecological design; otherwise, all other studies were cross-sectional.

The availability of retail food outlets near schools was ascertained using a variety of metrics and spatial scales (Table 2 and 3). Nine studies used buffers around schools, based on either street network distances (n=4)^11,20–22 or straight-line, radial distances (n=5).^{13,17–19,23} The type of buffer was unclear for two studies.^12,15 Four studies used two different distances for network buffers: 800 meters (m)^21,22/0.5 miles (~800m)¹¹ or 1200m.²⁰ Availability of retail food outlets within radial buffers were measured at 500m,^13,17 0.5 mile¹⁹, or 1 kilometer (km)^18,23 distances. For the studies with unclear spatial scale,^12,15 food availability was examined within 0.1, 0.25, and/or 0.5 mile distance from schools. Three studies used distance as a continuous variable or categorized it based on pre-defined cut points and conceptualization of access: one study used Euclidian (straight-line) distance between the school and retail food outlet as a continuous variable,¹⁶ one categorized Euclidean distance (≤100m, 101–500m, >500m),⁶ and another used restaurant density per 10,000 residents within school zip codes.¹⁴

Table 2:

Associations between the 1) fast food outlets and 2) convenience stores near schools and weight, by race/ethnicity, gender, grade, or socioeconomic advantage.

Author, Year	Outcome; Age/grade	Exposure within spatial scale	Measure of association	Subgroup	Effect size
1) Fast food outlets
Race/ethnicity
Currie et al, 2010	Obesity prevalence;	1+ vs 0 outlet within 0.1 miles (unspecified type of buffer)	β	White	2.8149  (SE: 1.0163); p<0.01
	Schools with 9th graders			Latino	2.0067(1.0135); p<0.05
				AA	−1.5417 (1.2056); p≥0.1
Davis & Carpenter, 2009	BMI; 5–12th grades	1+ vs 0 outlet within 0.5 miles (unspecified buffer type)	β	AA	0.20 (95%CI: 0.04, 0.36)
				Other minorities	(data not shown; mentioned no other racial/ethnic minorities had associations larger than baseline associations representing all students”)
Grier & Davis, 2013	BMI; 5–12th grades	Distance to nearest FFO. Euclidean.	β	White (non-urban, higher income schools)	−0.05 (95%CI: −0.10, 0.00); p<0.05
			Difference (Ref= white students at non-urban higher income schools)	Latino	−0.03 (interaction: 0.02 (95%CI: −0.05, 0.09)).
			Difference (Ref= white students at non-urban higher income schools)	AA	−0.07 (interaction: −0.02 (−0.14, 0.10))
			Difference (Ref= white students at non-urban higher income schools)	Asian	−0.04 (interaction: 0.01 (−0.08, 0.11))
Langellier, 2012	Overweight prevalence;	1+ vs 0 outlets within 0.5 miles network buffer.	Difference (Ref = majority Latino schools)	majority White	0.41 (interaction: 0.06 (SE: 1.27); p=0.97)
	Schools with students in 5, 7, 9th grades		β	majority Latino	0.35 (0.52); p=0.48
			Difference (Ref = majority Latino schools)	majority AA	3.39 (interaction: 3.04 (2.24); p=0.16)
			Difference (Ref = majority Latino schools)	majority Asian	−1.91 (interaction: : −2.26 (1.86); p=0.22)
			Difference (Ref = majority Latino schools)	No majority	−2.29 (interaction: −2.64 (1.06); p=0.01)
Sànchez et al., 2012	Overweight/ obese;	1+ vs 0 outlets within 0.5 miles radius.	PR	White	1.02 (95%CI: 1.00, 1.04)
	5, 7, 9th grades			Latino	1.02 (1.01, 1.03)
				AA	1.03 (1.00, 1.06)
				Asian	0.94 (0.91, 0.97)
Gender
Chiang, 2011	BMIz; Elementary schoolchildren (6–13 years old)	n of outlets within 500 meter radius of a school.	β	Boys	0.077; p<0.05
				Girls	0.032; p≥0.05
Chiang et al, 2017	BMIz; Grade 7–9	n of outlets within 1000 meter circular buffer.	β	Boys	−0.009; p≥0.05
				Girls	0.009; p≥0.05
Currie et al, 2010	Obesity prevalence;	1+ vs 0 outlet within 0.1 miles (unspecified buffer type).	β	Boys	1.3833 (SE: 0.8002); p<0.1
	Schools with 9th graders			Girls	1.9248 (1.0002); p<0.1
Larson et al, 2013	BMIz; 6–12th grades	1+ vs 0 outlet within 800m network buffer.	β	Boys	0.242(SE: 0.149); p 0.104
				Girls	−0.012 (0.139); p 0.932
Park et al, 2013	Obesity; 4–9th grades	Per unit change in groups (low, middle, high) of store density within a 500m radius.	OR	Boys	1.01 (95%CI: 0.99, 1.03); p≥0.05
				Girls	1.03 (1.01, 1.05); p<0.05
Sánchez et al, 2012	Overweight/ obese;	1+ vs 0 outlets within 0.5 miles radius.	PR	Boys	1.01 (95%CI: 1.00, 1.02)
	5, 7, 9th grades			Girls	1.02 (1.01, 1.04)
Williams et al, 2015	BMIz; Year 6	Store density (0 vs tertiles) within 800m street network buffer.	β (Ref: 0 stores)	Boys	Lowest density: 0.002 (95%CI: −0.096, 0.100); Middle density: −0.002 (−0.133, 0.129); High density: 0.101 (−0.052, 0.254)
				Girls	Lowest density: −0.001 (−0.110, 0.108); Middle density: −0.019 (−0.165, 0.126); Highest density: 0.018 (−0.152, 0.187)
Grade
Currie et al, 2010	Obesity prevalence	1+ vs 0 outlet within 0.1, 0.25, 0.5 miles (unspecified buffer type).	β	5th	0.1miles: 6.1332 (SE2.8280); p<0.05; 0.25 miles: −1.0562 (0.7568); p≥0.1; 0.5 miles 0.0418 (0.4985); p≥0.1
				7th	miles: 1.2712 (1.1135); p≥0.1; 0.25 miles: −1.5916 (1.1223); p≥0.1;0.5 miles: 0.6946 (0.6353); p≥0.1
				9th	0.1 miles: 6.3337 (2.8750); p<0.05; 0.25 miles: −1.7947 (1.2095); p≥0.1; 0.5 miles: −0.8311 (1.0871); p≥0.1
Leatherdale1, 2011	Overweight/ obese	n of outlets within 1km radius: 0 and 8 outlets are shown here.2	OR	5th	~1.0 and ~4.75
				6th	~0.75 and ~3.25
				7th	~0.75 and ~3.2
				8th	~0.70 and ~2.8
Sánchez et al. 2012	Overweight/ obese	1+ vs 0 outlets within 0.5 miles radius.	PR	5th	1.03 (95%CI: 1.01, 1.04)
				7th	1.02 (1.00, 1.04)
				9th	1.01 (0.99, 1.03)
Income Levels
Grier & Davis, 2013	BMI; 5–12th grades	Distance to nearest FFO. Euclidean.	Difference (Ref=white students at non-urban higher income schools)	Low income	−0.04 (interaction 0.01 (95%CI: −0.04, 0.07); p≥0.05)
Virtanen et al, 20152	Overweight/ obese; 8, 9th grades	Categorized distance to nearest FFO or grocery stores (<100m, 100–500m, >500m). Euclidean.	PR (ref>500m)	Low-SES	101–500m: 1.18 (95%CI: 0.98, 1.42) ≤100m: 1.22 (1.00, 1.48)
			PR (ref>500m)	Intermediate-SES	101–500m: 0.84 (0.70, 1.02) ≤100m: 0.90 (0.71, 1.12)
			PR (ref>500m)	High-SES	101–500m: 0.98 (0.87, 1.11) ≤100m: 0.79 (0.62, 1.01)
2) Convenience stores
Race/ethnicity
Langellier, 2012	Overweight prevalence;	1+ vs 0 stores within 0.5 miles network.	Difference (Ref = majority Latino schools)	majority White	−1.34 (interaction −2.97 (SE: 1.29); p=0.02)
	Schools with students in 5, 7, 9th grades		β	majority Latino	1.63 (SE: 0.61); p=0.007
			Difference (Ref = Majority Latino schools)	majority AA	2.13 (interaction 0.50 (2.35); p=0.83)
			Difference (Ref = Majority Latino schools)	majority Asian	−0.09 (interaction −1.72 (1.88); p=0.36)
			Difference (Ref = Majority Latino schools)	No majority	1.42 (interaction −3.05 (1.08); p=0.005)
Powell et al, 2007	BMI; 8th, 10th grades	n of stores per 10,000 people within each schools’ zip code	β	White	0.0162 (SE: 0.0121); p≥0.1
				Latino	0.0988 (0.0515); p<0.05
				AA	0.0476 (0.0341); p≥0.1
Sànchez et al, 2012	Overweight/ obese	1+ vs 0 stores within 0.5 miles radius..	PR	White	1.01 (95% CI: 1.00, 1.02)
	5, 7, 9th grades			Latino	1.01 (1.00, 1.01)
				AA	1.02 (1.00, 1.03)
				Asian	0.99 (0.97, 1.00)
Gender
Chiang et al, 2011	BMIz; Elementary schoolchildren (6–13 years old)	n of stores within 500 meter circular buffer	β	Boys	0.013;p≥0.05
				Girls	0.015; p≥0.05
Chiang et al, 2017	BMIz; Grade 7–9	n of stores within 1000 meter circular buffer	β	Boys	−0.005; p≥0.05
				Girls	0.001; p≥0.05
Harrison et al, 2011	log FMI; Year 5	Tertiles of weighted sum of the distance to unhealthy food outlets within 6 km	β (Ref: no access)	Girls	Commute on foot or bicycle Middle access:−0.1203 (95%CI: −0.014, −0.254) Best: 0.144 (0.009, −0.280) Commute by car, bus, or train Middle: 0.006 (−0.098, 0.110) Best: 0.012 (−0.107, 0.130)
				Boys	(data not shown; mentioned "no significant relationships were seen between FMI and access to the food outlets or facilities measured here among boys".)
Larson et al, 2013	BMIz; 6–12th grades	1+ vs 0 outlet within 800m network buffer.	β	Boys	−0.262(SE: 0.143); p=0.066
				Girls	−0.153(0.129); p=0.238
Park et al, 2013	Obesity; 4–9th grades	Per unit change in groups (low, middle, high) of store density within a 500m radius	OR	Boys	0.98 (95%CI: 0.94, 1.02)
				Girls	0.97 (0.93, 1.02)
Sánchez et al, 2012	Overweight/ obese	1+ vs 0 store within 0.5 miles radius.	PR	Boys	1.01 (95%CI: 1.00, 1.01)
	5, 7, 9th grades			Girls	1.01 (1.01, 1.02)
Grade
Sánchez et al, 2012	Overweight/ obese	1+ vs 0 outlets, 0.5 miles radius.	PR	5th	1.01 (95%CI: 1.00, 1.02)
				7th	1.01 (1.00, 1.02)

AA: African American; PR: prevalence ratio; OR: odds ratio. SE: Standard Errors.

¹Leatherdale et al showed the association between relative odds ratio for students being overweight in terms of grades and the number of fast-food retailers near schools (0 to 8 outlets). We estimated the relative ORs from the graphs and noted the minimum and maximum relative ORs (which corresponded to 0 and 8 outlets for each grade); the relative ORs for 1–7 outlets all fell between these values for each grade in an increasing pattern.

²Virtanen et al combined fast food outlets and grocery stores.

³In the published manuscript, this was reported as 0.120 but from the 95%CI; in the present review, it was assumed that this was a typo for −0.120. The authors were contacted to verify the validity of this assumption but a response could not be obtained.

Table 3:

Associations between the 1) supermarkets and 2) grocery stores near schools and weight, by race/ethnicity, gender, or grade.

Author, Year	Outcome; Age/grade	Exposure within spatial scale	Measure of association	Subgroup	Effect size
1) Supermarkets
Race/ethnicity
Powell et al1, 2007	BMI; 8th, 10th grades	n of outlets per 10,000 people within a schools’ zip code.	β	White	−0.0959 (SE 0.0349); p<0.01
				Latino	−0.0898 (0.0478); p<0.1
				AA	−0.3187 (0.0987); p<0.01
Gender
Harrison et al, 2011	log FMI; Year 5	Tertiles of weighted sum of the distance to healthy food outlets within 6 km.	β (Ref: no access)	Girls	Commute on foot or bicycle Middle access: −0.019 (95%CI:−0.116, −0.078); Best: −0.089 (−0.183, −0.006) Commute by car, bus, or train Middle: −0.035 (−0.120, 0.050) Best: 0.021 (−0.068, 0.110)
				Boys	(data not shown; mentioned “no significant relationships were seen between FMI and access to the food outlets or facilities measured here among boys”.)
Park et al, 2013	Obesity	Per unit change in groups (low, middle, high) of store density within 500m radius.	OR	Boys	1.05 (95%CI: 0.97, 1.13)
	4–9th grades			Girls	1.04 (0.96, 1.12)
2) Grocery stores
Race/ethnicity
Powell et al, 2007	BMI	n of outlets per 10,000 people within a school’s zip code.	β	White	0.0105 (SE: 0.0083); p≥0.1
	8th, 10th grades			Latino	0.0401 (0.0354); p≥0.1
				AA	0.0026 (0.0238); p≥0.1
Gender
Williams et. al, 20152	BMIz; Year 6	Store density (0 vs tertiles) within 800m street network buffer.	β (Ref: 0 stores)	Boys	Lowest density: 0.034 (95%CI: −0.070, 0.138); Middle density: 0.056 (−0.077, 0.189); High density: −0.053 (−0.212, 0.106)
				Girls	Lowest density: 0.113 (0.000, 0.227); Middle density: 0.106 (−0.043, 0.254); High density: 0.070 (−0.110, 0.250)
Grade
Leatherdale et al3, 2011	Overweight/obese	n of outlets within 1km radius (0 and 7 outlets shown here).	OR	5th	~1.0 and ~4.25
				6th	~0.70 and ~3.00
				7th	~0.70 and ~2.75
				8th	~0.60 and ~2.50

AA: African American; BMI: body mass index; OR: odds ratio.

¹Those are results from chain supermarkets; for non-chain supermarkets, the estimates were: White: β: 0.0494 (SE: 0.0324); AA: −0.0721 (0.1608); Latino: −0.0578 (0.1613), with p-values≥0.1 for these three estimates.

²Williams et al combined supermarkets, corner stores, convenience stores, and grocery stores.

³Leatherdale et al showed the association between relative odds ratio for students being overweight in terms of grades and the number of grocery stores near schools (0 to 7 stores). We estimated the relative ORs from the graphs and noted the minimum and maximum relative ORs (which corresponded to 0 and 7 stores for each grade); the relative ORs for 1–6 outlets all fell between these values for each grade in an increasing pattern.

Subgroup results of the associations between retail food outlets and weight status

Findings are presented for FFO and convenience stores (Table 2) and for supermarkets and grocery stores (Table 3). Within each outlet type, results are described by race/ethnicity, gender, grade, or child or school-level socioeconomic advantage.

1). Fast Food Outlets

Five studies reported results on availability of FFO near schools and weight status separately for racial or ethnic groups,^{11,12,15,16,19} seven studies by gender,^{12,13,17,19–21,23} three studies by grade,^12,18,19 and two studies by socioeconomic advantage.^6,16

Associations by race or ethnicity

Except for one study using county data from LA, all studies examining the associations by race/ethnicity used statewide data from California. There was more consistent evidence of positive associations among Latino children. There was weak evidence of negative associations between availability of FFO and weight status among Asian children.

Among Latino children, three studies reported clear evidence of positive associations between greater availability of FFO and weight status.^12,16,19 The school-level comparison in Los Angeles County did not find strong evidence for positive association between FFO availability within 0.5 miles of schools and overweight prevalence.¹¹

Among White children, results were similar to Latino children; two studies showed weak to moderate of evidence for positive associations between availability of FFO and weight status^12,19 and one study a negative association between distance to FFO and weight status.¹⁶ The ecological study in Los Angeles County found no strong evidence of association between FFO availability within 0.5 miles of schools and overweight prevalence in majority White schools (main effect of distance: 0.35; SE: 0.52; p=0.48; interaction with race/ethnicity 0.06; 1.27; p=0.97).¹¹

For African American children, two studies showed weak to moderate evidence for positive associations between FFO availability and greater body weight.^15,19 One study found that distance from schools to the nearest FFO was negatively associated with BMI, although there was no clear evidence of effect modification by African American race/ethnicity (main effect of distance: −0.05; 95%CI: −0.10 to 0.00; interaction:−0.02; −0.14 to 0.10).¹⁶ Two other studies using California statewide data and Los Angeles County school-level data found no clear evidence for a positive association between the presence of a FFO (within 0.1 or 0.5 miles of schools respectively) and obesity.^11,12

Among Asian children, studies showed mixed evidence for associations between availability of FFO near schools and weight status. One study reported that greater distance from a school to the nearest FFO was associated with lower BMIz scores in non-urban, high income neighborhoods (main effect of distance: −0.05; 95%CI: −0.10 to 0.00; interaction with Asian race/ethnicity: 0.01; −0.08 to 0.11).¹⁶ Two studies found either a negative association¹⁹ or no clear evidence of association¹¹ between availability of FFO and overweight/obesity prevalence.

Associations by gender

There was some variation in the evidence of association between FFO availability and weight status in the gender-stratified analyses.

Two studies from California found evidence of positive associations between FFO near schools and weight status for both girls and boys.^12,19 One study in Taiwan saw stronger evidence of positive association between FFO availability and weight status among boys in elementary schools while the study in Korea showed stronger evidence for girls.^13,17 Studies from Minnesota and England as well as another study from Taiwan focusing on junior high school students did not see strong evidence of association between FFO and weight status for either gender.^20,21,23

Associations by school grade

Across children in all school grades, there were generally positive associations. Among children in the 5th grade in California, FFO availability within 0.1 miles was associated with a higher obesity prevalence (β: 6.13 percentage points; p<0.05) but there was no clear evidence of association at longer distances (0.25 o 0.5 miles).¹² Among children in the 7^th grade in California, a study showed weak evidence of association between availability of FFO and higher prevalence of overweight/obesity (Prevalence Ratio (PR): 1.02; 95%CI: 1.00 to 1.04)¹⁹ while another study found no clear evidence of association.¹² One study in California showed evidence of positive association¹² while another California study did not find clear evidence of association among students in the 9^th grade.¹⁹ A study from Ontario, Canada reported a “significant” interaction effect with school grades. The study graphically showed the steepest increase in odds ratios for having obesity with the increasing number of FFO within 1km radius of schools among students in Grade 5, in comparison to Grade 6–8, though the estimates from hypothesis testing were not reported.¹⁸

Associations by socioeconomic advantage

Two studies examined associations between distance to or availability of FFO near schools and weight status according to socioeconomic advantage. One study in Finland found some evidence of greater prevalence ratios of overweight/obesity (BMI≥25 kg/m²) associated with shorter distance to the closest FFO or grocery store only among students from socioeconomically disadvantaged backgrounds.⁶ The same study saw weak evidence of negative associations between closer distance from school to a FFO or grocery store and overweight/obesity among students with socioeconomic advantage.⁶ A study in California did not find strong evidence of differences in association between distance to FFO and BMIz scores among children in low vs high income-level schools (interaction with income level: 0.01(95% CI: −0.04 to 0.07); p≥0.05).¹⁶

2). Convenience Stores

Eight studies investigated convenience store availability near schools and weight status, reporting results separately by race/ethnicity,^11,14,19 gender, ^{13,17,19,20,22,23} and/or grade.¹⁹

Associations by race or ethnicity

Three studies found some evidence of positive associations between convenience stores and weight status among Latino and African American children, and more mixed results among White and Asian children.

Three studies using data from Los Angeles County, California statewide and national data showed positive associations between convenience store availability near schools and overweight/obesity among Latino students.^11,14,19

Among White children, the ecological study using Los Angeles County data reported a negative association between convenience store availability near schools and overweight/obesity¹¹ while a study using statewide California data found weak evidence of a positive association.¹⁹ In a U.S. national study analyzing zip-code level densities of convenience stores, there was no strong evidence of association between convenience store availability near schools and BMI among White children.¹⁴

Among African American children, Los Angeles County and statewide California studies showed some evidence of positive associations between convenience store availability near schools and overweight/obesity.^11,19 In a U.S. national study, there was no clear evidence that convenience store availability near schools measured at the zip-code level was associated with higher BMI (β: 0.05; SE: 0.03; p≥0.1).¹⁴

Among Asian children, one study showed weak evidence that convenience store availability near schools was associated with lower overweight/obesity prevalence ratios (PR: 0.99; 95% CI: 0.97 to 1.00).¹⁹ Another study from Los Angeles County showed positive association (main effect of convenience store availability: 1.63; SE: 0.61; p<0.01; interaction by race/ethnicity: −1.72; 1.88; p = 0.36).¹¹

Associations by gender

Six studies reported results by gender. A study based on California data found convenience store availability was associated with greater prevalence of overweight/obesity (PR: 1.01; 95%CI: 1.01 to 1.02) and weaker evidence for boys (1.01; 1.00 to 1.01).¹⁹ Five studies from Minnesota, Korea, Taiwan, and England found no clear evidence of association between convenience store availability and weight status in girls or boys.^{13,17,20,22,23}

Associations by school grade

A California study reported weak evidence that greater availability of convenience stores within a 0.5 mile radius of a school was associated with higher prevalence of overweight/obesity among 5th grade (PR: 1.01; 95% CI: 1.00 to 1.02) and 7^th grade students (PR: 1.01; 95% CI: 1.00 to 1.02); there was no clear evidence of association among students in the 9^th grade (PR: 1.00, 95% CI: 0.99, 1.01).¹⁹

3). Supermarkets

Three studies examined associations between supermarkets near schools and weight status (Table 3). ^13,14,22

Associations by race or ethnicity

Chain supermarket density per 10,000 capita within school’s zip codes was associated with lower BMI among White, Latino and African American students in the U.S.¹⁴

Associations by gender

Better access to supermarkets near schools was negatively associated with FMI only among girls who took active transport (i.e. on foot or bicycle) to school in Norfolk, England but not among boys or girls who took cars, trains, or buses to school.²² A study from Seoul, Korea, found no clear evidence of association between supermarket density within 500 m of school’s neighborhoods and odds of obesity for either boys or girls.¹³

4). Grocery Stores

Three studies examined the associations of grocery stores near school on weight status (Table 3). One study investigated the relationship between grocery stores near schools and student weight status by race/ethnicity,¹⁴ one study by gender,²¹ and one study by school grade.¹⁸

Associations by race/ethnicity

Powell et al reported no clear evidence of association between grocery store density and BMI among White (β: 0.01 kg/m²; SE: 0.01; p≥0.1), Latino (0.04; 0.04; p≥0.1), or African American students (0.003; 0.02; p≥0.1) using a national U.S. sample.¹⁴

Associations by gender

A study from England did not find clear patterns of association between density of food stores near schools and BMIz in boys or girls.²¹

Associations by school grade

A study from Ontario, Canada found that the number of grocery stores within 1 km of the students’ school was associated with greater odds of being overweight among students in the 5^th grade than 6–8^th grades, though the study only graphically presented this finding based on the interaction effects and the estimates from hypothesis testing were not available in the publication.¹⁸

Quality of Evidence

The quality of the evidence were similar among the included studies (Appendix). Many studies used nationally/regionally representative or surveillance data. All studies used cross-sectional or correlational data. Many studies used the latest food environment data that were measured during the same time periods when the outcomes were measured. This may be a reasonable choice in food environment studies as the availability of retail food outlets typically do not change rapidly and it is plausible that students have been exposed to these retail food outlets for some time before their weight status were measured. Two studies clearly described the use of prior exposure data: one study used food environment data from one year before¹² and the other specifically excluded establishments that opened in the year when the outcomes were measured.¹⁹

Discussion

The findings from this systematic review revealed that only a small number of studies have reported results specifically for subgroups, and all of those studies used cross-sectional or correlational data. We observed more consistent evidence for positive associations of obesogenic food environments near schools and negative associations of healthier types of retail food outlets with weight status among Latino students. For other racial/ethnic groups, the magnitudes, directions of the effects, and the strengths of evidence were less consistent. The included studies – many from California, which has large populations of Latino children - often had greater sample sizes of this group than children of other race/ethnicities. The wider exposure distributions among Latino students may have resulted in greater power to detect associations in this group.

The limited body of evidence suggests consistent associations between obesogenic food environments near schools and weight status across all grade levels, with some studies reporting greater associations in younger grades. However, there is a need for future studies in more diverse settings. A recent ecological study from New York found that FFO were positively associated only among middle/high school students (7^th and 10^th school grades combined) while greater density of farmer’s markets within the school districts were negatively associated with obesity rates only among younger students (pre-kindergarten/kindergarten, 2^nd and 4^th combined).²⁴

We found that only two studies reported associations separately by relative socioeconomic advantage. These studies found some evidence of greater positive associations between food environments near schools and obesity among children and school-neighborhoods at greater socioeconomic disadvantage. Studies have shown higher concentrations of FFO and convenience stores near schools in socioeconomically disadvantaged neighborhoods than schools in affluent neighborhoods.^6,16,25 Additional research is needed to examine any differential effects of food environments near schools on childhood obesity across socioeconomic strata.²⁶

Grier and Davis tested for interactions among race/ethnicity, income, and urbanicity in examining associations between food environments near schools and weight status. They found stronger associations among African American and Latino students in lower income urban schools compared with white students in higher income non-urban schools.¹⁶ Prior research found differences in availability of unhealthy retail food outlets near schools by urbanicity in California.²⁶ Given these findings, consideration to interactions with race/ethnicity, income, and urbanicity would be helpful in understanding the potentially interrelated nature of these dimensions.

Multiple factors may explain the differences in results across studies reviewed here, including various spatial scales used to characterize exposure to retail food outlets, geographic locations of the samples, and differences in the characteristics of the study populations. We observed variation in the classification of retail food outlet exposures, which prevents direct comparisons of results across studies. For example, convenience stores, the second most-studied exposure, were differentially categorized as corner stores, liquor stores, chain stores, and non-chain stores. Additionally, a range of spatial scales were used to delineate food environment exposures near schools, including network, radial buffers and zip-code-level density, along with varying distances from schools ranging from 0.1 miles to 5 km. Study locations included countries from Europe, Asia, and North America. Although it would be interesting to explore cross-national and cultural differences in the impact of food environments on children’s consumption behaviors by subgroups, there is currently insufficient evidence to characterize and explain these differences even within single countries. Additionally, all studies included in this review were from high income countries. Future research is needed to examine associations in lower income countries, given that the burden of childhood obesity has been rising in some of these countries. Another important factor to consider for future studies is means of transportation during commute (e.g., by parental cars, walking, bicycle) as these could modify the association between food environments near schools and weight status and perhaps explain some of the differences among subgroups examined in this review.

Previous reviews reported modest overall associations between the food environments near schools and weight status.^4,14 The present review contributes to the body of the literature concerning obesity influences of the food environments near schools by summarizing the evidence reported separately for racial/ethnic minority populations and socioeconomically disadvantaged subgroups —who are at higher risk of obesity— as well as by gender and grade levels. The limitations of this review include its focus on studies written in English and published in peer-reviewed journals, and the exclusion of government reports and other types of publications that may have reported associations for specific sociodemographic subgroups. In this review, we observed negative associations between grocery stores/supermarkets and weight status in some subgroups, but there were much fewer studies examining such associations. Therefore, the evidence for associations on these specific food outlets in subgroups remains unclear. Four studies used self-reported weight status, which may have reduced accuracy. Due to heterogeneous study settings and differences in methods used (e.g., measures of outcomes and spatial scales), we did not perform meta-analyses to obtain overall estimates of the association within each subgroup. Although two large search databases were used and manual search in the relevant papers were performed, the possibility of missing relevant articles cannot be fully eliminated.

Implications

Inferences about the causal effect of the food environments near schools on body weight remain unclear, given the cross-sectional and correlational nature of the designs in the studies included in this review. Despite this caveat, findings from the present review suggest that policy makers need to be aware that some subgroups, such as Latino children and children in younger grades, may be particularly vulnerable to influences of the food environments near schools.

Furthermore, a previous study in California suggested that the availability of FFO increased between 2000 and 2010 among majority African American, majority Latino, and majority Asian schools in the least affluent neighborhoods.²⁶ Exposure to marketing and advertisements for calorie-dense food and beverages may also increase sociodemographic disparities in childhood obesity. Powell et al. showed that Latino and African American children were more likely to be exposed to fast food marketing.²⁷ A combination of increased exposures to both advertisements and to retail food outlets near schools may work in unison to influence diet and obesity among children from more vulnerable subgroups, through its effects on purchasing and consumption behaviors.

Researchers and policy makers around the world have called for regulation on the availability of junk food and sweetened beverages near schools.^28–30 For example, the U.S. National Academies of Sciences recommended that local governments enact policies to limit the number of fast food establishments near schools, and to create incentives that encourage retail food outlets to reduce unhealthy food marketing and increase offerings of healthier food options.³¹ These strategies could be enhanced by additional evidence on which subgroups of children are more likely to be exposed to and influenced by food environments near schools. Future studies with longitudinal designs and subgroup analyses will be able to strengthen the evidence base for how differential environmental regulations could effectively reduce obesity among high-risk subgroups. Analyses of aggregated data may mask environmental inequalities, potentially diminishing our ability to plan effective interventions for children at higher risk of obesity. Additionally, non-U.S. studies included in this review did not provide information on the effect modification by racial/ethnic groups. Further evidence on this topic from other countries can enrich our understanding of complex relationships among cultural, socioeconomic, and demographic characteristics. Finally, there is a need for more studies to assess how students from diverse backgrounds respond to different food environments by examining their purchasing patterns at these outlets.

Conclusions

A limited number of studies using ecological or cross-sectional data has reported variable associations between the food environments near schools and obesity separately by race/ethnicity, gender, grade and socioeconomic advantage. The studies in this review found generally positive associations between fast food outlets near schools and obesity among Latino, White, and African American children as well as children in all grade levels, with potentially large effects for children in younger grades. Additional research is needed to better understand causal mechanisms of the effects of the food environments near schools on weight status among subgroups, especially those at higher risk of childhood obesity such as children from racial/ethnic minority and socioeconomically disadvantaged populations.

Appendix

1).

PubMed Search Terms by Hedge Terms

Hedge Terms	Food Environment	Childhood Obesity
	Keywords [All Fields]	Keywords [All Fields]	MeSH
Search Terms	“neighbourhood” “neighborhood” “food environment “nutrition environment” “food availability” “health food” “grocery” “supermarket” “Convenience store” “restaurant” “takeaway” “food access” “food store” “fast food” “mobile vendor”	“obesity” “overweight” “body mass index” BMI	“obesity” “overweight” “body mass index”

Full search terms were:((((((((((((“neighbourhood”[All Fields] OR “neighborhood”[All Fields]) OR “food environment”[All Fields]) OR “nutrition environment”[All Fields]) OR “food availability”[All Fields]) OR “health food”[All Fields]) OR “grocery”[All Fields]) OR “supermarket”[All Fields]) OR “Convenience store”[All Fields]) OR “restaurant”[All Fields]) OR “takeaway”[All Fields]) OR “food access”[All Fields]) OR “food store”[All Fields] OR “fast food”[All Fields] OR “mobile vendor”[All Fields]) AND ((((“obesity”[MeSH Terms] OR “obesity”[All Fields]) OR (“overweight”[MeSH Terms] OR “overweight”[All Fields])) OR (”body mass index”[MeSH Terms] OR “body mass index”[All Fields])) OR BMI[All Fields]) AND ((“1980/01/01”[PDAT] : “2019/04/25”[PDAT]) AND “humans”[MeSH Terms] AND English[lang])

2).

Summary Results of the Modified National Heart, Lung, and Blood Institute Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies

	Yes (n)	No (n)	Other1 (n)
1. Was the research question or objective in this paper clearly stated?	14	0	0
2. Was the study population clearly specified and defined?	14	0	0
3. Was the participation rate of eligible persons (or schools in ecological studies) at least 50%?	12	0	2
4. Were all the subjects selected or recruited from the same or similar populations (including the same time period)? Were inclusion and exclusion criteria for being in the study prespecified and applied uniformly to all participants?	14	0	0
5. Was a sample size justification, power description, or variance and effect estimates provided?	13	1	0
6. For the analyses in this paper, were the exposure(s) of interest measured within a year of the outcome(s) being measured?1	12	0	2
7. Was the timeframe sufficient so that one could reasonably expect to see an association between exposure and outcome if it existed?1	14	0	0
8. For exposures that can vary in amount or level, did the study examine different levels of the exposure as related to the outcome (e.g., categories of exposure, or exposure measured as continuous variable)?	14	0	0
9. Were the exposure measures (independent variables) clearly defined, valid, reliable, and implemented consistently across all study participants?	14	0	0
10. Was the exposure(s) assessed more than once over time?	2	10	2
11. Were the outcome measures (dependent variables) clearly defined, valid, reliable, and implemented consistently across all study participants?	12	2	0
12. Were the outcome assessors blinded to the exposure status of participants?	0	14	0
13. Was loss to follow-up after baseline 20% or less?	0	0	14
14. Were key potential confounding variables measured and adjusted statistically for their impact on the relationship	14	0	0

^1.Cannot determine; Not applicable; or Not reported