|First author, number of included studies, year of publication, range of publication of primary studies in years||Study design and countries/populations covered||Exposure domain||Summary of results||ROBIS results|
|Urban–rural||Food||Physical activity||Social inequality||Other|
|Allender et al. , 9 studies, 2008, 1997–2007||
China, Russia, the Philippines, South Korea, South Africa, Cameroon, 1 global study
|X||Low-strength evidence of the following findings: a positive relationship between BMI/ cholesterol and food share of household expenditure and proportion of population in urban areas; increasing urbanization improves micronutrients but increases body weight, blood pressure and cholesterol; lifetime exposure to urban environment is correlated with BMI; lifetime exposure to urban environment is correlated with obesity and there is rapid change in diet, obesity and physical activity among LMIC.||Unclear|
|An et al. , 16 studies, 2018, 2008–2017||
Half longitudinal half cross-sectional|
US (9), Canada, Netherlands, Serbia, China, Italy, South Korea; adults study only in US, China, Italy and the Netherlands
Mixed associations, results vary by age sex and type of air pollutants. Associations between pollution and weight are probably not linear and mediated by health behaviours.|
Ratios of associations between exposure to respective pollutants and weight in adults by expected/unexpected/ non-significant:
|Angkurawaranon et al. , 45 studies, 2014, 1988–2013||
Malaysia, Laos, Vietnam, Thailand Indonesia, Timor-Leste, Philippines, Myanmar
Overall significant association between urbanicity and obesity. Pooled OR is 1.99 (95% CI: 1.64, 2.41) for South East Asia, in studies of only adults OR is 1.65 (95% CI: 1.36, 1.99) suggesting higher risks of obesity in urban areas.|
Heterogeneity in associations between and within countries, some can be attributed by economic status, age/sex, time of study, BMI classification used and populations studied.
35 studies have unclear risks, two studies have different response rate per area, the rest have low risks of bias.
|Black and Macinko , 36 studies, 2008, 1997–2006||
STUDY DESIGN not clearly defined|
US (17), Sweden (1), Netherlands (2, Eindhoven), Scotland (2), Canada (1), Australia (1), multicounty (1)
PA environment is more consistently associated to obesity than the food environment|
Neighbourhood SES is consistently negatively associated with weight outcomes after adjusting for personal SES.
Income inequality is positively associated to obesity but not on county level; racial composition studies have different results, but racial isolation is found to be associated with increased obesity prevalence.
|Casagrande et al. , 10 studies, 2009, 2002–2005||
STUDY DESIGN not clearly reported|
All US studies; on county, multi-site, state- or nationwide samples
|X||One study on the built environment and weight status, where higher BMI increased participants’ likelihood to report physical/ environmental barrier to exercising.||Unclear|
|Chandrabose et al. , 36 studies, 2019, 2007–2017||
35 observational and 1 natural experiment, all longitudinal|
US, Canada, Sweden, Australia, Finland, Germany, UK (Wales), Lithuania
Evidence for association between walkability, recreational facilities (except greenspace), urban sprawl and obesity- presented by weighted z-score with positive, statistically significant results.|
Studies on stayers tend to have higher percentage of significant associations compared to movers.
There was mismatch between perceived versus measured walkability in association with weight outcome.
Methodological issues include residential self-selection and possible mediation by physical activity.
|Cobb et al. , 71 studies, 2015, 2005–2014||
11 longitudinal, the rest cross-sectional|
US (64) and Canada (7)
Most of associations of food stores and weight are null, indices do have significant results more often but still dominated by nulls; quality of assessed studies sub-optimal- more than 30 studies have more than three errors.|
Ratios of associations between the following food stores with weight in adults by expected/unexpected/ non-significant:
Grocery store: 14/2/77
Fast food: 29/6/71
Relative healthiness food outlet index:16/1/20 (both adults and children)
Combined healthy outlets: 2/0/2
Combined unhealthy outlets: 0/0/2
|Durand et al. , 44 studies, 2011, 2002–2009||
Cross sectional (39) and longitudinal/ quasi-longitudinal (5)|
Sweden, US, Australia, UK, Belgium, Canada
Mostly non-significant associations between eight (over a total of ten) conceptual aspects of urban planning and weight. The other two aspects have no studies so far|
Associations in expected direction 0–33% of all studies, non-significant associations 66–100%.
|Feng et al. , 63 studies, 2009, 2001–2008||
61 cross-sectional, 4 longitudinal|
US (52), Australia (5), Canada (2), Denmark, Sweden, UK, European countries
Most consistent associations with sprawl index and land use mix, mixed results for fast food density.|
The overall ratio of expected/ unexpected/ non-significant associations is 40/2/38.
There is difference in definition of “place” across studies, heterogeneity in metrics to measure the built environment and geographic range. There is added value in composite scores and indices compared to single measures in obesity studies; there is dominance of cross-sectional studies reduces causation or inference power.
|Ferdinand et al. , 169 studies, 2012, 1995–2010||
164 quantitative observational and 5 qualitative observational; no specification whether cross-sectional or longitudinal|
Most studies within US (69), 60 outside and 40 with unknown locations
173/194 significant associations between food and PA environment with health benefits, results are similar in Southern states compared to others.|
Characteristics associated with higher scientific vigour have fewer significant associations with health benefits.
|Fleischghacker et al. , 40 studies, 2011, 1998–2008||
Cross sectional except for Sturm and Datar (longitudinal, children)|
US, UK, Canada Australia, New Zealand
Six out of ten adult studies reported increased BMI with higher fast food access.|
Most studies do not have the same definition for fast food outlets or consistent sources of information or exposure methodologies.
Fast food access is also associated to increased percent of minorities and lower SES
|Fraser et al. , 33 studies, 2010, 2002–2009||
Cross-sectional and ecological, except for Sturm and Datar (longitudinal)|
US (12), Canada, UK, New Zealand (1), Australia (1)
Weak association between fast food access and obesity/ overweight. There are some associations between fast food and deprivation. Overall ratios of expected/ non-expected/ non-significant associations are 6/2/5; associations are significant especially when weights are self-reported.|
Highlighting inconsistencies in definition of fast food, various measures of access and geographical settings.
|Gamba et al. , 51 studies, 2015, 2004–2014||
40 cross-sectional, 7 longitudinal, 4 repeated cross-sectional|
32% of all associations are significant in expected direction. Measure of presence of food stores is most likely to identify significant findings.|
Community nutrition environment should include more food store and food store types. Definition of fast food restaurants is not consistent across studies and quality of food data sources is sometimes questionable.
|Giskes et al. , 28 studies, 2011, 2003–2009||
Cross sectional except for 1 experiment (consumption)|
US, UK (obesity), Australia, Netherlands, Japan, New Zealand
Studies with weight-related outcome are mostly about access to/ density or presence of food sources; associations with weight as outcomes are notably more consistent than with diet.|
Ratios of associations in adults by expected/unexpected/ non-significant for weight status is 12:3:11 and for diet is 7:0:13.
|Grasser et al. , 34 studies, 2013, 1997–2010||
33 cross-sectional, 1 prospective|
US, UK, Australia, Canada, EU
Walkability is more consistently associated with walking than weight outcomes; associations with weight outcomes were mixed and sometimes in the unexpected direction (especially for connectivity measure); some were significant in only a subpopulation (one city/ one gender etc.). Overall level of evidence was concluded to be low.|
In terms of quality: 8 good, 16 fair and 10 poor studies.
|Hernández et al , 18 studies, 2012, 1964–2010||
14 cross-sectional, 1 retrospective 3 prospective cohorts|
Chile, Iran, Senegal, Kenya, China, Papua New Guinea, Peru, Panama, Guatemala, Tanzania, Poland, Bangladesh, India, Indonesia
Higher BMI in migrants compared to rural, lower BMI in migrants compared to urban, higher obesity in migrants compared to rural, lower obesity in migrant compared to urban in almost all measures of weight.|
Meta-analyses show no differences between urban and rural areas. Studies are highly heterogeneous.
Same health profile observed for other cardiovascular conditions.
|Holsten , 7 studies, 2009, 2004–2006||
6 studies were cross-sectional, 1 was ecologic|
All US except for Simmons (Australia)
Inconclusive results due to inconsistent correlations and various methods used.|
Most studies focus only on fast food restaurants and associations are only significant in one subset of populations.
|Kondo et al. , 68 studies, 2018, 1991–2017||
Experimental studies between (14) or within (21) subjects. 20 longitudinal, 3 case–control crossover, 9 quasi-experimental|
US (27), UK (13), Netherlands (5), Canada (5), Japan (4), Australia (3), Lithuania (3), Denmark (2), Germany (2), Finland, Italy and Spain (1 each)
Studies focusing on adult population found no association between BMI and green space exposure.|
There is a general lack of consistency in defining urban nature and measurement of its exposure hinders study quality and generalizability.
|Lachowycz et al. , 60 studies, 2011, 2002–2009||
Australia, Canada, England, Europe, New Zealand, Portugal, Sweden, Netherlands, US
Three studies with obesity-related outcome finds association between higher green space and lower adverse obesity-related outcome.|
Most studies found some sort of evidence for a relationship between greenspace and weight or reported mixed results across subgroups, according to measure of greenspace.
Ratios of associations in adults by positive/ mixed/ negative/ non-significant is 3:6:1:4.
|Larson et al. , 54 studies, 2009, 1986–2008||
Study type not mentioned in review, however discussion mentioned that they are mainly cross-sectional|
Higher access to fresh supermarkets and limited access to corner stores are negatively associated with obesity while restaurant availability has mixed results. In general, more access to full restaurants & limited access to fast food restaurant is associated to less obesity.|
There is inequality in access to supermarkets, corner stores and restaurants (lower SES—more fast food, higher SES—healthier restaurants).
|Leal and Chaix , 131 studies, 2011, 1985–2009||
14 longitudinal (9 on weight), the rest cross sectional|
US (86), Sweden, UK, Canada, Netherlands, Germany & Czech Republic, France, Italy, Lithuania, Portugal, Slovakia, Switzerland, Australia, Japan, New Zealand, Spain
Most consistent association is between low SES and increased weight; low urbanized associated with high weight; high supermarket low convenience store low fast food associated with low weight; high street connectivity, high density of intersections and services associated with lower obesity; criminality and insecurity associated with high weight.|
High traffic noise associated with high triglyceride level.
|Lovasi et al. , 45 studies, 2009, 1995–2009||
Study type not mentioned in review, either individually or in statistics summary|
Targeted groups disadvantaged in terms of access to food stores, fast food outlets, places to exercise, aesthetics and safety. Strongest support for importance of food stores, exercise facilities and safety.|
The built environment might affect the high SES group more than the low SES groups due to low exposure.
Built environment interventions are more effective if targeted to the disadvantaged; especially those that help reduce disparities.
|Mackenbach et al. , 92 studies, 2014 2003–2013||
6 longitudinal, 84 cross-sectional, 2 both|
US (74), Canada, Australia, New Zealand, UK, Belgium, Portugal, EU, France, Denmark
Heterogeneity in metrics used and findings, except for urban sprawl and land use mix with clear associations with obesity within North America.|
Stratification results show remaining heterogeneity within continents, mode of measurement, methodological quality. Overall weak associations between environment and weight status.
29 strong, 53 moderate and 8 weak primary studies methodologically. Reporting was moderate or strong in quality.
|Malambo et al. , 18 studies, 2016, 2005–2015||
17 cross-sectional and 1 longitudinal (study on stroke)|
US, New Zealand, Australia, China, Sweden, Canada
|X||X||BMI is lower in walkable neighbourhoods with recreationally dense neighbourhood. BMI is higher in places with high densities of fast food restaurants not supermarkets (both American studies in somewhat older populations).||High|
|McCormack et al. , 55 studies, 2019, 1998–2017||
Cross-sectional (36), prospective and retrospective cohort, longitudinal, case–control, case-crossover, time series, quasi-experimental|
Canadian provinces Ontario, BC, Quebec, Alberta, Nova Scotia
Consistent associations between aggregate built environment score, greenness, land use, food environment and weight status.|
Somewhat less consistent associations for population/ dwelling density and mostly non-significant for route characteristics.
|Papas et al. , 20 studies, 2007, 2002–2006||
18 cross-sectional investigations, three of which were ecologic studies. Two longitudinal studies|
US, Australia, and Europe
Food environment is less well studied than PA environment, associations differ by age groups or races. 17/20 studies found statistically significant associations between aspects of the built environment and weight.|
Concerns include inconsistency of measurements of the built environment across studies, the cross-sectional design of most investigations, and the focus on aspects of either diet or physical activity but not both.
|Patterson et al. , 10 studies, 2019, 2008–2017||
Longitudinal, controlled trials or natural/ quasi experiments|
US (6), UK and China
Meta-analysis of five studies show initiating public transport use was associated with 0.30 units BMI reduction 95% CI (0.14, 0.47).|
Distance from the residential address to the nearest bus route had mixed associations with adiposity.
|Renalds et al. , 23 studies, 2010, 2005–2008||
Mostly cross-sectional, 1 experiment, 1 retrospective review of longitudinal study|
Countries not listed but seem mostly US
Mixed land use alone is not sufficient for BMI research, more investigation into specific land use (type of business within a residential neighbourhood) is needed.|
Urbanity, low land use mix, crime, low street connectivity, automobile dependency increases overweight in adults.
|Schüle and Bolte , 33 studies, 2015, 2005–2013||
Most cross-sectional except for one British study|
Belgium, Australia, Canada, US, Sweden, Germany, UK
Nine studies looked at adult obesity and/or BMI. Eight out of 11 studies had significant associations between neighbourhood SES (NSES) and BMI/ overweight/ obesity: four high SES—low BMI and four low SES—low BMI; seven studies found significant association between built environment and BMI independent of NSES.|
Two studies found associations between NSES and individual characteristics (sex/ race); six studies found interaction between BE and individual SES (ISES, in other words, associations are only significant for some subgroups: women, whites, car owners etc.)
Differences between NSES and ISES should be considered: it is not clear how NSES, built environment and sex interacted with one another.
|Sugiyama et al. , 41 studies, 2014, 2010—2013||
34/41 cross-sectional, 6 prospective, 1 both|
US (23), Canada, UK, Australia, France, Brazil, Egypt, Belgium & Nigeria
Composite environmental indices have significant associations with weight related outcome (sprawl index, pedestrian + public transport + residential density) unless it is walkability. Among the walkability components, land use mix is most consistently correlated to weight outcomes.|
Utilitarian destinations also have significant associations. Leisure-activity related attributes do not contribute to obesity/ weight. Transport-related physical activity should be a target for future studies.
Ratios of associations in adults by expected: unexpected: non-significant in cross-sectional studies is 56/6/86 and in prospective studies is 9/1/10.
|Tseng et al. , 17 studies, 2018, 2008—2018||
17 natural experiments|
UK, Australia but mostly US (12)
Low level of evidence for weight/ BMI change in intervention groups.|
Four out of nine interventions in PA has significant BMI reduction (ratio reduction/ inconsistent/ no difference is 4/3/2). None of the five food interventions shows BMI reduction.
The mixed environment studies (both food and PA) also have low level of evidence. Studies with significant results show associations with low clinical significance. Most studies have high risk of biases.
|Wilkins et al. , 113 studies, 2019, 2005–2018||
87 cross-sectional, 26 longitudinal|
US, UK, Australia, Germany, New Zealand
Overall, null associations dominated other associations. Fast food outlets associations more positive when defined more narrowly, measured in proximity rather than presence, and in low SES groups.|
Percentage of associations in adults by expected/unexpected/ non-significant for each outlet is:
Fast food: 20.8/4.2/75.0 (404 associations in total)
Convenience stores: 10.9/8.5/79.6
Supermarkets/groceries store: 6.6/12.9/80.5
Diverse methods used to measure food environment in five different aspects: exposure data source, extraction method, methods and definitions of food outlets, geocoding method and retail food environment metrics. Measurement methods are not well reported.