Previous research on the influence of the food environment on weight status has often used impersonal measures of the food environment defined for residential neighborhoods, which ignore whether people actually use the food outlets near their residence. To assess whether supermarkets are relevant contexts for interventions, the present study explored between-residential neighborhood and between-supermarket variations in body mass index (BMI) and waist circumference (WC), and investigated associations between brands and characteristics of supermarkets and BMI or WC, after adjustment for individual and residential neighborhood characteristics.
Participants in the RECORD Cohort Study (Paris Region, France, 2007–2008) were surveyed on the supermarket (brand and exact location) where they conducted their food shopping. Overall, 7 131 participants shopped in 1 097 different supermarkets. Cross-classified multilevel linear models were estimated for BMI and WC.
Just 11.4% of participants shopped for food primarily within their residential neighborhood. After accounting for participants' residential neighborhood, people shopping in the same supermarket had a more comparable BMI and WC than participants shopping in different supermarkets. After adjustment for individual and residential neighborhood characteristics, participants shopping in specific supermarket brands, in hard discount supermarkets (especially if they had a low education), and in supermarkets whose catchment area comprised low educated residents had a higher BMI/WC.
Citation: Chaix B, Bean K, Daniel M, Zenk SN, Kestens Y, Charreire H, et al. (2012) Associations of Supermarket Characteristics with Weight Status and Body Fat: A Multilevel Analysis of Individuals within Supermarkets (RECORD Study). PLoS ONE 7(4): e32908. doi:10.1371/journal.pone.0032908
Editor: Sanja Stanojevic, Hospital for Sick Children, Canada
Received: June 28, 2011; Accepted: February 7, 2012; Published: April 4, 2012
Copyright: © 2012 Chaix et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work, as part of the RECORD project, was supported by the Institute for Public Health Research (Institut de Recherche en Santé Publique); the National Institute for Prevention and Health Education (Institut National de Prévention et d'Education pour la Santé) (Prevention Program 2007 074/07-DAS; 2010–2011 financial support; and 2011–2013 financial support); the National Institute of Public Health Surveillance (Institut de Veille Sanitaire) (Territory and Health Program); the French Ministries of Research and Health (Epidemiologic Cohorts Grant 2008); the National Health Insurance Office for Salaried Workers (Caisse Nationale d'Assurance Maladie des Travailleurs Salariés); the National Research Agency (Agence Nationale de la Recherche) (Health–Environment Program 2005 #00153 05); the Ile-de-France Regional Health Agency (Agence Régionale de Santé d'Île-de-France); the City of Paris (Ville de Paris); the Ile-de-France Regional Council (Conseil Régional d'Île-de-France, DIM SEnT and CODDIM); and the Ile-de-France Youth, Sports, and Social Cohesion Regional Direction (Direction Régionale de la Jeunesse et des Sports et de la Cohésion Sociale). The project was also supported by the Swedish Research Council (to JM), by the Canadian Institutes for Health Research (to YK), and by ANR-PNRA (to JMO). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
In recent literature, there has been a considerable interest in the influence that local food environments have on weight status . Because of the mixed evidence and inconsistent findings of previous literature on the food environment and weight status, there is growing debate on the strategies to measure the food environment , , . Previous research on this issue has overwhelmingly used measures of the food environment defined for the residential neighborhood , , , ,  or the geographic work environment  (e.g., the road network distance to the closest supermarket or the density of fast-food restaurants).
A major limitation of these measures, however, is that they do not reflect the personal food environment of individuals, as they do not account for whether participants effectively use or do not use food outlets located near their residence or workplace. Neighborhood-based measures of the local food environment may be meaningless for people who shop far from their residence. As suggested by studies documenting relationships between the type of store where people shop and dietary behavior , a necessary need, therefore, is to derive personal measures of the food environment through the identification of the food outlets people effectively use. Moreover, to assess whether supermarkets are appropriate contexts for developing preventive nutritional interventions, it is relevant to explore between-supermarket (in addition to between-residential neighborhood) variations in weight status and abdominal fat through multilevel analysis.
In the RECORD Cohort Study , , , , , the supermarket where participants did most of their food shopping was assessed with its brand, exact location, and characteristics. The present study examined (i) whether participants who did most of their food shopping in the same supermarket had a more comparable body mass index (BMI) and waist circumference (WC) than participants who shopped in different supermarkets (after accounting for the residential neighborhood); and (ii) whether differences in BMI and WC between participants shopping in different supermarkets were explained by the distance from residence to the supermarket (which may influence food purchasing behavior ), the socioeconomic status (SES) of supermarket catchment area (possibly related to food preferences and purchasing power of consumers and types of foods available), the supermarket brand, the supermarket type, supermarket size, and by the quality of fruits and vegetables in the supermarket.
The study protocol was approved by the French Data Protection Authority. All the participants signed an informed consent to enter the study.
The RECORD Cohort Study (“Residential Environment and CORonary heart Disease”, www.record-study.org) includes 7 290 participants who were recruited between March 2007 and February 2008 , , , , . Participants were beneficiaries of the French National Health Insurance System for Salaried Workers, which offers a free medical examination every 5 years to all working and retired employees and their families (corresponding to 95% of the population of the Paris Ile-de-France region). Participants were recruited without a priori sampling (convenience sample) among the eligible persons attending these 2-hour-long preventive checkups conducted by the Centre d'Investigations Préventives et Cliniques in four of its health centers located in the Paris Ile-de-France region (Paris, Argenteuil, Trappes, and Mantes-la-Jolie). People seen at these preventive health checkups are not unhealthier than the average population, and attend these health centers either on a voluntary basis, through the referral of their family physician or workplace physician, following the advice of peers, etc. Eligibility criteria were as follows: age 30 to 79 years; ability to complete study questionnaires; and residence in one of the 10 (out of 20) administrative districts of Paris or 111 other municipalities of the metropolitan area selected a priori. These municipalities and districts were chosen so as to reflect disparities between advantaged and disadvantaged territories (with overrepresentation of disadvantaged municipalities) and disparities between urban and suburban areas. Among people presenting at the health centers who were eligible based on age and residence, 10.9% were not selected for participation because of linguistic or cognitive difficulties in filling out study questionnaires. Of the persons selected for participation, 83.6% accepted to participate and completed the entire data collection protocol.
Participants were geocoded with accuracy based on their residential address in 2007–2008. Research assistants rectified all incorrect or incomplete addresses with the participants by telephone. Extensive investigations with local Departments of Urbanism were conducted to complete the geocoding. Precise spatial coordinates and block group codes were identified for 100% of the participants.
Due to missing information on the supermarket (see below) or in anthropometric measures, the final sample available for analysis of BMI and WC comprised 7 072 and 6 926 participants, respectively.
Summary information on the scales of analysis and scales of measurement of the variables is provided in Table 1. Details on the anthropometric measures (body mass index and waist circumference) and individual sociodemographic variables analyzed in the present study are provided in Table 2.
Collective entities: residential neighborhoods and supermarkets.
Participants resided in 1 910 different census block group areas. The median number of residents in these local units was 2 425 (interquartile range: 2 083, 2 932) in the 2006 Population Census.
The participants were asked to report the brand and address of the supermarket where they did most of their food shopping (the survey question referred to supermarkets rather than food stores in general because, as shown with the data provided below, most people in France are shopping in chain supermarkets). During face-to-face contact, trained survey technicians sought to collect as much information as possible to identify these supermarkets. During the baseline survey, technicians did not, however, perform online searches to assist participants who were unable to identify their supermarket.
Each participant's primary supermarket could not always be straightforwardly identified. For example, Franprix is a brand of small supermarkets, and in Paris, there are often many Franprix very close to each other. Also, in certain cases, there were more than one supermarket of the same brand in the same street. In these cases and others, participants were systematically telephoned in the months subsequent to their health examination, in attempts to precisely identify the supermarket where they shopped (790 participants were so contacted). Technicians conducting these calls used Google Maps and the websites of the different supermarket brands to assist them in their searches. In the end, the official business identification code (SIRET) of each supermarket was retrieved.
Among the 7 290 participants, three participants who reported doing most of their shopping at the market (i.e., comprising numerous independent shopkeepers), 26 participants who did most of their food shopping via the internet, and 130 participants who did not make most of their shopping in a supermarket or could not provide enough information to identify the supermarket were removed from the study. For five participants who could not choose which one of two supermarkets was their main source of food, the largest supermarket was used for analysis. After the exclusions listed above, 7 131 participants were coded as conducting most of their food shopping in 1 097 distinct supermarkets.
Residential neighborhood- and supermarket-level contextual variables.
Three distinct characteristics were considered for residential neighborhood SES and SES of supermarket catchment area (hereafter referred to as supermarket neighborhood SES) (see Table 2). Using databases geocoded at the building level, these characteristics were determined within circular areas of different scales centered on the exact place of residence or on the exact supermarket location , . Moreover, to account for the fact that supermarket catchment areas greatly vary in size between supermarket brands, SES of supermarket catchment areas was also determined within circular areas with brand-specific radiuses (see Table 2).
Variables related to each participant's supermarket are described in Table 2: supermarket brand (Table 3 provides descriptive information on each brand), supermarket type (see Information S1 for additional information on supermarket types), whether one's supermarket is within one's residential neighborhood, the distance to the supermarket, whether one's supermarket is the closest available, supermarket size, and the quality of fruits/vegetables in one's supermarket.
After deriving descriptive statistics, we estimated multilevel linear/logistic models with individual education, household income, neighborhood education, and neighborhood income as explanatory factors, and the following variables as the outcomes: (i) one's primary supermarket located or not in one's administrative residential neighborhood; (ii) street network distance to one's primary supermarket; (iii) shopping or not at the closest available supermarket; (iv) shopping or not at the closest supermarket of a particular type (citymarket, small/large supermarket, hypermarket, hard discount supermarket); (v) shopping or not at the closest supermarket of a particular brand.
For BMI and WC separately, we estimated cross-classified multilevel linear models  with two distinct random effects at the neighborhood level and at the supermarket level (levels not nested within each other).
Models adjusted for age and sex were first estimated. A second series of models adjusted for individual sociodemographic variables assessed the spatial scale on which residential and supermarket neighborhood SES were most strongly associated with BMI and WC , . The different supermarket variables were then included in a third series of models adjusted for individual and residential neighborhood SES. At each step, only those variables that remained associated with the outcomes were retained. We assessed interactions between the coefficients for individual or residential neighborhood SES and supermarket characteristics.
All models were estimated with SAS 9.2. The Akaike Information Criterion (AIC) was used to compare the models: the lower the AIC, the better the overall combination of fit to the data and parsimony of the model.
Descriptive data on the study sample are provided in Information S2.
Descriptive data on participants' shopping behavior
Just 11.4% of participants reported shopping for food primarily within their residential census block group neighborhood. Even on a larger scale, no more than 64.2% of the participants did most of their food shopping in their municipality of residence. Regarding street network distance to the primary supermarket, as much as 50% and 31% of the participants had their primary food store located further than 1 km away and 2 km away, respectively, from their residence.
The proportion of participants who shopped in their administrative residential neighborhood tended to increase with socioeconomic status, especially with education of neighborhood residents (Information S3). On the opposite, descriptive data and regression modeling indicated that the average street network distance to the supermarket was larger for residents of high education neighborhoods (Information S3). The contradiction is only apparent: residents of high education neighborhoods more frequently shopped in their residential neighborhood (possibly due to more frequently residing in urban centers) but at the same time covered larger average distances to their primary supermarket (some of them, e.g., having holiday homes or working far from their residence).
Comparing the effective participants' primary supermarket to the available supermarkets nearby their residence, we found that 29.5% of the participants performed most of their food shopping at the closest available supermarket. As shown in Information S3, participants residing in neighborhoods with a high average education level were more likely to shop at the closest supermarket available from their residence.
Models for BMI and WC
The mean BMI of participants was 25.5 kg/m2 (interdecile range: 20.7, 30.7). Mean WC was 77.7 cm (interdecile range: 65, 94) among women and 89.0 cm (interdecile range: 76, 103) among men.
After adjustment for age and sex, models that included both residential neighborhood-level and supermarket-level random effects had a better combination of fit to the data and parsimony than models including only the residential neighborhood- or the supermarket-level random effect (with a 18.2 and 17.4 points lower AIC, respectively, for BMI and a 15.0 and 9.4 points lower AIC for WC).
In models including both residential neighborhood-level and supermarket-level random effects, there were variations in BMI and WC both between residential neighborhoods and between supermarkets (see Information S4). Even after accounting for each person's residential environment, participants who shopped in the same supermarket had a more comparable BMI or WC than those shopping in different supermarkets. Overall, 4.2% and 3.1% of the variations in BMI and WC were at the residential neighborhood level, and 2.5% and 2.2% of the variations were at the supermarket level.
Individual socioeconomic variables associated with BMI or WC are reported in Information S5. As previously reported , after individual-level adjustment, residential neighborhood education was more strongly associated with BMI and WC than were neighborhood income or neighborhood dwelling values. BMI and WC increased with decreasing neighborhood education. As shown in Information S6, model fit was the best when residential neighborhood education was assessed within 500 m radius circular areas.
Similarly, after adjustment for individual characteristics, the relationship between supermarket neighborhood SES and BMI or WC was stronger when assessed with education than with income or dwelling values. As shown in Information S6, model fit was the best when supermarket neighborhood education was assessed within 5 000 m radius circular areas centered on the supermarket, i.e., on a much broader scale than residential neighborhood education. Supermarket neighborhood education assessed within brand-specific radius circular areas led to a slightly better fit to the data only in the model for BMI. Accordingly, supermarket neighborhood education was assessed in circular areas with brand-specific radiuses in the model for BMI and in 5 000 m radius circular areas in the model for WC.
There was a strong correlation between residential neighborhood education and supermarket neighborhood education. Nonetheless, when residential neighborhood education and supermarket neighborhood education were introduced together into the models, the two variables remained associated with either BMI or WC (Table 4).
After full adjustment for individual, residential neighborhood, and supermarket characteristics, participants who did most of their food shopping far from home had a slightly larger WC and tended to have a slightly higher BMI (Table 4). After adjustment, the quality of fruits and vegetables in the supermarket and supermarket size showed no association with BMI or WC.
After controlling for individual and residential neighborhood SES and distance to the supermarket, and using the Monoprix brand (expensive citymarkets located in city centers) as the referent, participants shopping in certain supermarket brands, especially in hypermarkets such as CORA or in hard discount supermarkets such as Ed or Lidl, had a greater BMI and WC (Table 3).
Considering supermarket type instead of supermarket brands, people shopping in small/large supermarkets, hypermarkets, and more particularly in hard discount supermarkets had a greater BMI and WC compared to those shopping in citymarkets, while participants shopping in organic shops had a markedly lower BMI and WC (Table 4).
There was a strong interaction between the coefficients for individual education and shopping in a hard discount supermarket. Models for BMI and WC re-estimated with a variable combining individual education and shopping in a hard discount store (Figure 1) indicated that the association between shopping in a hard discount store and BMI or WC was markedly stronger for lower education levels.
The interactions were estimated from cross-classified multilevel linear models for body mass index (BMI) (part A) and waist circumference (WC) (part B) adjusted for individual sociodemographic characteristics, residential neighborhood education, supermarket type, distance to the supermarket, and supermarket neighborhood education (interactions assessed with a variable combining categories of the two variables). RECORD Cohort Study, Paris Metropolitan Area, 2007–2008.
As detailed in Information S7, relationships between supermarket type and BMI/WC were re-estimated after propensity score matching, i.e., among participants shopping in different types of supermarkets matched according to their probability to shop in a specific supermarket type (probability estimated in function of a number of individual socioeconomic characteristics). In these analyses, as shown in Table 5, participants shopping in small/large supermarkets and hypermarkets tended to have a higher BMI/WC than those shopping in citymarkets. However, relationships between shopping in hard discount supermarkets and anthropometry were lost in these propensity score matched analyses, probably due to the sharp decrease in sample size after matching (which decrease resulted from a limited exchangeability of participants shopping in citymarkets and hard discount supermarkets according to their individual socioeconomic profile).
Supplementary analyses also indicated that all of the associations between supermarket characteristics and WC disappeared after adjustment for BMI .
The results indicate that a greater distance from one's home to one's primary supermarket, a lower SES of supermarket catchment area, and shopping in specific supermarket brands or supermarket types (especially in hard discount supermarkets among low educated individuals) were associated with a greater BMI and WC. This research supports the utility of deriving personal measures of the food environment by geocoding supermarkets where people do most of their food shopping and the relevance of multilevel studies of individuals nested within food stores for targeting nutritional interventions to specific supermarkets showing higher obesity prevalence .
Strengths and limitations
Strengths of the present study include (i) the time-consuming and meticulous geocoding of participants' supermarkets, (ii) the assessment of abdominal fat beyond general weight status, (iii) the cross-classified multilevel regression models  estimated to simultaneously partition variance in BMI and WC between residential neighborhoods and supermarkets, (iv) the propensity score matching analyses performed to examine whether supermarket effects are separable from individual socioeconomic effects, and (v) the investigation of spatial scale of predictors that revealed that supermarket neighborhood education was related to BMI/WC on a much broader scale than residential neighborhood education.
Study limitations include (i) the cross-sectional nature of the data that did not allow us to investigate whether frequenting specific supermarkets is related to weight gain over time, (ii) the inability to control for food preferences that condition supermarket choice and therefore likely confound the relationships of interest (discussed below), (iii) the limited amount of information available on each supermarket , e.g., the absence of data on the proportion of shelf space devoted to each type of food  as more direct exposures than those measured in the present study, and (iv) the absence of information on participants' food purchasing behavior as a mediator of the supermarket–body weight relationship.
Main study findings
Most previous studies of the relationship between the food environment and weight status have relied on impersonal measures of the food environment related to residential or geographic work environments assumed to reflect where individuals obtain food , , , , , . However, the present data remarkably indicate that as much as 88.6% and 35.8% of participants did not shop in their own neighborhood or municipality, respectively. Therefore, neighborhood-based measures of the food environment may in many circumstances provide incomplete picture of individuals' food environments. Moreover, our data suggest that a focus on the residential neighborhood may differentially affect the quality of measurement of the food environment in the different socioeconomic groups, thereby potentially biasing the assessment of the role of the food environment in socioeconomic disparities in obesity risk and poor health. As a complementary strategy, there is a need to develop personal measures of the food environment based on participants' frequenting of given supermarkets or food providers. Such a strategy enabled (i) accounting for food shopping supermarkets, in addition to residential neighborhoods, as a component of group-level variability in multilevel analyses, and (ii) deriving supermarket-level variables to assess in relation to BMI and WC.
Longer distances between one's home and primary supermarket tended to be related to greater BMI and WC, as also reported previously . People shopping far from their residence may less frequently go to their supermarket and may accordingly rely on less fresh products. It is not clear, however, whether this process may contribute to the relationship between distance to the supermarket and weight status.
Regarding supermarket characteristics, independent associations were documented between, on the one hand, supermarket brand, supermarket type, and educational level of supermarket catchment area and, on the other hand, BMI and WC. Even if the model was adjusted for individual and neighborhood variables that may condition supermarket choice, associations reported here reflect a mixture of causal effect and residual confounding that are intrinsically difficult or almost impossible to disentangle (see Figure 2). Of course, our discussion of different causal scenarios does not imply that we draw any causal conclusion from our observational data.
Causal effects of supermarket brand, type, and SES (in support of which our observational data do not provide solid evidence) may stem from the differential availability of healthy foods such as fruits/vegetables or fish, from the availability of low-cost energy-dense foods in hard discount supermarkets, or from the differential advertisement, showcasing, or nutritional labeling of these products in the different supermarkets, all of which may constrain or influence individual purchasing behavior , .
We reviewed the published evidence on the nutritional quality of low priced or hard discount food products as relevant to correctly interpret the associations documented in our study between shopping in hard discount supermarkets and BMI/WC. A study conducted in France reported that 36% of surveyed consumers thought that low priced products are of lower quality than the branded ones . In coherence with this perception, a French study published more than 15 years ago indicated that branded products were more regularly of high quality (but not always better) in terms of nutrient content, food security, and taste indicators than low cost ones . A more recent study indicated that, within a given food category, the lowest-priced foods were not different from the equivalent branded products in terms of overall energy or total fat content . Another French study published in 2009 reported no systematic difference between low priced and branded products in terms of nutrient content, raw materials, microbiological analysis, or taste . However, in the aforementioned French study, a weak relationship suggested that the overall quality of ingredients increased with the price of foods (within a given food category) . Moreover, basic nutritional information and dietary recommendations were less often provided on low priced foods than on branded products , , . Overall, the published information is scarce and provides only mixed evidence in support of the idea that hard discount supermarkets are obesogenic nutritional environments .
As illustrated in Figure 2, the most significant source of bias for the associations between supermarket characteristics and BMI or WC likely stems from confounding from characteristics that are independent determinants of both supermarket choice and nutritional status. There are at least three major confounding mechanisms possibly involved in the reported associations. The first is related to neighborhood SES, the second one to individual/household financial capacities (low income populations are constrained to rely on energy-dense, nutrient-poor foods to obtain daily calories at an affordable cost ), and the third one is related to food preferences.
Regarding the first bias, residential neighborhood SES influences which supermarket brands are available in one's local vicinity (SES is an important criterion in geomarketing location strategies ). Moreover, neighborhood SES is associated with other exposures influencing weight status (e.g., opportunities for physical activity) . Therefore, neighborhood SES may be a major source of confounding in the associations between supermarket characteristics and weight status or abdominal fat .
As shown in Figure 2, confounding by individual financial capacities and food preferences is attributable to the fact that individual income and dietary preferences not only influence supermarket choice, but also shape the food purchasing behavior of typical customers in each supermarket (i.e., typical customers of supermarket A would have a different purchasing behavior than typical customers of supermarket B if they were constrained to shop in supermarket B).
We addressed the biases related to neighborhood SES and financial capacities by controlling for residential neighborhood education and testing adjustment for household income. However, even though the models were adjusted for individual education, we did not control for dietary preferences, which may confound the estimated associations.
Moreover, it is important to keep in mind that propensity score matching analyses revealed that it is difficult to disentangle the supermarket type – BMI/WC associations from the effects of individual socioeconomic factors. For example, there was evidence that adjustment for individual socioeconomic variables of the relationships between shopping in a hard discount supermarket and BMI/WC was to some extent based on excessive models' extrapolations.
In order to address these issues, future studies should combine information on the types and brands of participants' primary supermarkets, on the food products sold in these supermarkets, on individual food purchasing behavior and consumption, and on anthropometric indices.
Whether partly attributable to confounding or not, the results of the study have important implications in suggesting that the supermarket where most of the food shopping is done could be targeted by public health interventions aimed at promoting adequate nutritional practices . While numerous studies proposed to consider residential neighborhoods to identify high-priority areas for interventions, far fewer studies have suggested supermarkets as the basic unit for targeting interventions, as implied by the multilevel analysis of individuals nested within supermarkets reported here. Implementing interventions focused on food purchasing behavior in specific supermarkets may be an efficient strategy because supermarkets are the very place where dietary preferences are concretely materialized and translated into a definite set of purchased foods. The present study that considered various supermarket characteristics may prove useful in identifying supermarkets, based on their brand and catchment area characteristics, where nutritional interventions may be particularly beneficial.
Additional information on supermarket type and supermarket brands.
Descriptive information on the study sample.
Information on participants' shopping behavior.
Between-neighborhood and between-supermarket variations in BMI and WC from models adjusted for age and gender.
Models with individual and residential neighborhood variables associated with BMI and WC.
Identification of the residential and supermarket neighborhood socioeconomic variables most strongly associated with BMI or WC and assessment of the optimal spatial scale of measurement of the variables.
Relationships between supermarket type and body mass index (BMI) or waist circumference (WC) estimated through propensity score matching.
We particularly thank Alfred Spira, head of IReSP, for his advice and support. We are grateful to INPES (and Pierre Arwidson) for its continued support since the beginning of the study. We also thank Mélani Alberto from ARS-IdF and Nathalie Catajar and Muriel Hirt from DRJSCS-IdF for their support in our project. We are grateful to Insee, the French National Institute of Statistics and Economic Studies, which provided support for the geocoding of the RECORD participants and allowed us to access to relevant geographical data (with special thanks to Pascale Breuil and Jean-Luc Lipatz). We thank Geoconcept for allowing us to access to the Universal Geocoder software. Regarding the data used in the present analysis, we are grateful to Paris-Notaries and to the National Geographic Institute. We also thank CNAM-TS and the Caisse Primaire d'Assurance Maladie de Paris (CPAM-P, France) for helping make this study possible.
Conceived and designed the experiments: BC FT KB BP. Analyzed the data: BC. Contributed reagents/materials/analysis tools: KB MD SNZ YK HC CL FT NK CW JMO CS JM BP. Wrote the paper: BC.
- 1. Leal C, Chaix B (2011) The influence of geographic life environments on cardiometabolic risk factors: a systematic review, a methodological assessment and a research agenda. Obes Rev 12: 217–230.
- 2. McKinnon RA, Reedy J, Morrissette MA, Lytle LA, Yaroch AL (2009) Measures of the food environment: a compilation of the literature, 1990–2007. Am J Prev Med 36: S124–133.
- 3. Glanz K (2009) Measuring food environments: a historical perspective. Am J Prev Med 36: S93–98.
- 4. Lytle LA (2009) Measuring the food environment: state of the science. Am J Prev Med 36: S134–144.
- 5. Mehta NK, Chang VW (2008) Weight status and restaurant availability a multilevel analysis. Am J Prev Med 34: 127–133.
- 6. Li F, Harmer P, Cardinal BJ, Bosworth M, Johnson-Shelton D, et al. (2009) Built environment and 1-year change in weight and waist circumference in middle-aged and older adults: Portland Neighborhood Environment and Health Study. Am J Epidemiol 169: 401–408.
- 7. Morland K, Diez Roux AV, Wing S (2006) Supermarkets, other food stores, and obesity: the atherosclerosis risk in communities study. Am J Prev Med 30: 333–339.
- 8. Powell LM, Auld MC, Chaloupka FJ, O'Malley PM, Johnston LD (2007) Associations between access to food stores and adolescent body mass index. Am J Prev Med 33: S301–307.
- 9. Wang MC, Kim S, Gonzalez AA, MacLeod KE, Winkleby MA (2007) Socioeconomic and food-related physical characteristics of the neighbourhood environment are associated with body mass index. J Epidemiol Community Health 61: 491–498.
- 10. Jeffery RW, Baxter J, McGuire M, Linde J (2006) Are fast food restaurants an environmental risk factor for obesity? Int J Behav Nutr Phys Act 3: 2.
- 11. Zenk SN, Schulz AJ, Hollis-Neely T, Campbell RT, Holmes N, et al. (2005) Fruit and vegetable intake in African Americans. Income and store characteristics. Am J Prev Med 29: 1–9.
- 12. Chaix B, Bean K, Leal C, Thomas F, Havard S, et al. (2010) Individual/neighborhood social factors and blood pressure in the RECORD Cohort Study: which risk factors explain the associations? Hypertension 55: 769–775.
- 13. Chaix B, Billaudeau N, Thomas F, Havard S, Evans D, et al. (2011) Neighborhood effects on health: correcting bias from neighborhood effects on participation. Epidemiology 22: 18–26.
- 14. Leal C, Bean K, Thomas F, Chaix B (2011) Are associations between neighborhood socioeconomic characteristics and body mass index or waist circumference based on model extrapolations? Epidemiology 22: 694–703.
- 15. Havard S, Reich BJ, Bean K, Chaix B (2011) Social inequalities in residential exposure to road traffic noise: An environmental justice analysis based on the RECORD Cohort Study. Occup Environ Med 68: 366–374.
- 16. Chaix B, Kestens Y, Bean K, Leal C, Karusisi N, et al. (2011) Cohort profile: residential and non-residential environments, individual activity spaces, and cardiovascular risk factors and diseases: the RECORD Cohort Study. Int J Epidemiol in press.
- 17. Inagami S, Cohen DA, Finch BK, Asch SM (2006) You are where you shop: grocery store locations, weight, and neighborhoods. Am J Prev Med 31: 10–17.
- 18. Thomas F, Bean K, Pannier B, Oppert JM, Guize L, et al. (2005) Cardiovascular mortality in overweight subjects: the key role of associated risk factors. Hypertension 46: 654–659.
- 19. Benetos A, Thomas F, Pannier B, Bean K, Jego B, et al. (2008) All-cause and cardiovascular mortality using the different definitions of metabolic syndrome. Am J Cardiol 102: 188–191.
- 20. Human Development Report 2007/ (2008) New York, NY 2008. Available: http://hdr.undp.org/en/media/HDR_20072008_EN_Complete.pdf. Accessed 2011 May 7: United Nations Development Programme.
- 21. Beckman A, Hakansson A, Rastam L, Lithman T, Merlo J (2006) The role country of birth plays in receiving disability pensions in relation to patterns of health care utilisation and socioeconomic differences: a multilevel analysis of Malmo, Sweden. BMC Public Health 6: 71.
- 22. Mujahid MS, Diez Roux AV, Morenoff JD, Raghunathan T (2007) Assessing the measurement properties of neighborhood scales: from psychometrics to ecometrics. Am J Epidemiol 165: 858–867.
- 23. Chaix B, Lindstrom M, Rosvall M, Merlo J (2008) Neighbourhood social interactions and risk of acute myocardial infarction. J Epidemiol Community Health 62: 62–68.
- 24. Chaix B (2009) Geographic Life Environments and Coronary Heart Disease: A Literature Review, Theoretical Contributions, Methodological Updates, and a Research Agenda. Annu Rev Public Health 30: 81–105.
- 25. Chaix B, Merlo J, Chauvin P (2005) Comparison of a spatial approach with the multilevel approach for investigating place effects on health: the example of healthcare utilisation in France. J Epidemiol Community Health 59: 517–526.
- 26. Chaix B, Merlo J, Subramanian SV, Lynch J, Chauvin P (2005) Comparison of a spatial perspective with the multilevel analytic approach in neighborhood studies: the case of mental and behavioral disorders due to psychoactive substance use in Malmö, Sweden, 2001. Am J Epidemiol 162: 171–182.
- 27. Muntaner C, Li Y, Xue X, Thompson T, O'Campo P, et al. (2006) County level socioeconomic position, work organization and depression disorder: a repeated measures cross-classified multilevel analysis of low-income nursing home workers. Health Place 12: 688–700.
- 28. Rosmond R, Lapidus L, Bjorntorp P (1996) The influence of occupational and social factors on obesity and body fat distribution in middle-aged men. Int J Obes Relat Metab Disord 20: 599–607.
- 29. Gittelsohn J, Song HJ, Suratkar S, Kumar MB, Henry EG, et al. (2010) An urban food store intervention positively affects food-related psychosocial variables and food behaviors. Health Educ Behav 37: 390–402.
- 30. Glanz K, Sallis JF, Saelens BE, Frank LD (2007) Nutrition Environment Measures Survey in stores (NEMS-S): development and evaluation. Am J Prev Med 32: 282–289.
- 31. Farley TA, Rice J, Bodor JN, Cohen DA, Bluthenthal RN, et al. (2009) Measuring the food environment: shelf space of fruits, vegetables, and snack foods in stores. J Urban Health 86: 672–682.
- 32. Glanz K, Sallis JF, Saelens BE, Frank LD (2005) Healthy nutrition environments: concepts and measures. Am J Health Promot 19: 330–333, ii.
- 33. Franco M, Diez Roux AV, Glass TA, Caballero B, Brancati FL (2008) Neighborhood characteristics and availability of healthy foods in Baltimore. Am J Prev Med 35: 561–567.
- 34. Moati P, Ranvier M (2005) Faut-il avoir peur du hard discount? CREDOC, consommation et modes de vie 188: 1–4.
- 35. Michels S (1995) Qualités nutritionnelles comparées des produits alimentaires de marque et de premier prix. La Lettre de l'IFN 35: 1–8.
- 36. Darmon N, Caillavet F, Joly C, Maillot M, Drewnowski A (2009) Low-cost foods: how do they compare with their brand name equivalents? A French study. Public Health Nutr 12: 808–815.
- 37. (2009) Produits bas prix, quelle qualité? Que Choisir 470:
- 38. CNA (2010) Qualité nutritionnelle, consommation et distribution des produits alimentaires “hard discount” et “premiers prix”: état des lieux et perspectives. Avis n°67. Available: http://cna-alimentation.fr/index.php?option=com_docman&Itemid=28. Accessed 2012 February 8.
- 39. Oqali (2010) Rapport d'activité 2009. Available: http://www.oqali.fr/oqali/publications_oqali/rapports_d_activite. Accessed 2012 February 8.
- 40. Drewnowski A, Monsivais P, Maillot M, Darmon N (2007) Low-energy-density diets are associated with higher diet quality and higher diet costs in French adults. J Am Diet Assoc 107: 1028–1032.
- 41. Diez Roux AV (2003) Residential environments and cardiovascular risk. J Urban Health 80: 569–589.
- 42. Chaix B, Leal C, Evans D (2010) Neighborhood-level Confounding in Epidemiologic Studies: Unavoidable Challenges, Uncertain Solutions. Epidemiology 21: 124–127.