Introduction

The prevalence of obesity has increased in the world in the last decades ¹. The growing increase in obesity is influenced by contextual issues, since lifestyle, diet and physical activity practices are influenced by economic growth ^{2,3,4,5,6,7,8}. There is much evidence that family and contextual income influence obesity ^9,10,11. In Brazil, low-income women have a higher prevalence of obesity when compared to high-income women but the causal relationship is not well established ^12,13.

Interventions focused on people changing their eating behaviors to improve their nutritional status have had limited success ². Although the individual approach is necessary, it is not sufficient and changing the contextual determinants of healthy eating also needs to be considered.

Recently, the interest in the potential impact of the neighborhood on individual health and healthy diet components has increased. The food environment refers to physical, economic, political and sociocultural characteristics, opportunities and conditions that influence food behaviors ^14,15. The food environment can influence food acquisition practices and whether people make healthy food choices or not, because besides being influenced by the individuals’ preferences and knowledge about food, food behavior is influenced by the location of food stores and the price of available foods ^{15,16,17,18,19,20,21,22}. These characteristics, among other factors, may lead to the development of obesity and related conditions ^23,24.

Types of food stores have been considered proxies for healthy or unhealthy diets. In Brazil, supermarkets offer many different types of ultraprocessed foods, but, on the other side, they also offer healthy foods to consumers ^18,25. This is different from grocery stores which predominantly sell ultraprocessed foods and few healthy foods, fruits and vegetables ²⁶. A Brazilian study with children shows that the availability of natural foods was associated with an increase of these foods consumption and a decrease of ultraprocessed food consumption ²². In addition to the quality of food available in the neighborhood, the density of establishments is also a factor associated with nutritional status: in neighborhoods where there is a high diversity of food stores, residents tend to have a normal nutritional status ^27,28.

Neighborhood social disadvantage is another contextual characteristic related to obesity. Socially disadvantaged neighborhoods tend to have a lower availability of healthy foods when compared to more advantaged neighborhoods ^{29,30,31,32,33,34,35,36,37}. Menezes et al. ³⁵, when evaluated Brazilian adults found that average consumption of fruits and vegetables was higher in neighbourhoods with higher income and concentration of food stores. Furthermore, studies exploring the relationship between the food environment and obesity suggest that higher access to healthy food is associated with better diet quality and lower rates of obesity ^{26,29,38,39,40}.

Although the relationship between the food environment and obesity has been studied in developed countries ^41,42, few studies have focused on low and middle-income countries ^28,42,43. The objective of this study was to explore the relationship between the neighborhood food environment and obesity in women living in São Leopoldo, Rio Grande do Sul State, Brazil. We also assessed if relationships differed by income group. We hypothesized that living in a neighborhood with more supermarket and healthy food stores would protect residents from being obese, while living in areas with more grocery stores and convenience stores would increase the risk of being obese.

Materials and methods

This study was part of the larger project Living Conditions and Health of Adult Women: A Population-based Study in Vale do Rio dos Sinos - Evaluation after 10 Years, Rio Grande do Sul, Brazil. The present investigation had two stages of data collection. First, a cross-sectional population-based study of a representative sample of 1,096 women living in the urban area of São Leopoldo. Second, gathering information on the neighborhoods in which participants resided. The county of São Leopoldo is located in the Sinos River Valley, Porto Alegre Metropolitan Area, and in 2010, it had a population of 214,087 inhabitants ⁴⁴.

Results

From the 1,281 women in the full study sample, 153 refused to participate, 4 could not provide weight and height measurements, and 28 resided outside the buffer defined for their neighborhood. These were excluded from the analyses for a final analytical sample of 1,096 women. Despite exclusions, the analytical sample was representative, in terms of its demographic characteristics, to IBGE (Brazilian Institute of Geography and Statistics) data.

Table 1 shows socioeconomic, demographic and food purchase characteristics for the complete sample and according to obesity status. Women in the sample were white skin color (74.3%), married or in union (63.8%), aged 50 years or more (34.6%), had completed higher education (41.5%) and worked at the time of the interview (56.1%). In relation to food intake, they consumed on average 6.34 (SD: 2.90) fruits and vegetables per day, and drank sugar drinks 0.46 (SD: 0.40) times per day. In relation to food purchases, women reported accessing an average of 1.69 (SD: 0.70) types of food establishments, relying on their neighborhood to purchase, on average, 4.18 (SD: 3.52) food items out of the 8 enquired, and 59.7% of women reported making the majority of their purchases by car, motorcycle or bus (Table 1).

The prevalence of obesity in the sample was 33.1%. Table 1 shows the crude prevalence ratios for obesity with the characteristics of the sample. Compared to their respective reference category, a higher obesity prevalence was found in the higher age groups (PR = 1.95; 95%CI: 1.46-2.61), among those divorced, separated and widowed (PR = 1.66; 95%CI: 1.24-2.21), those with a lower schooling level (PR = 1.94; 95%CI: 1.30-2.89) and in the second tertile level of income (PR = 1.29; 95%CI: 1.04-1.61), as well as among women who were not working (PR = 1.19; 95%CI: 1.00-1.40). The frequency of consumption of fruits and vegetables and the number of stores accessed by women to purchase food were associated with obesity, with PRs of 1.03 (95%CI: 1.00-1.06; p = 0.013) and 1.15 (95%CI: 1.04-1.27; p = 0.005), respectively.

Concerning the food environment, 279 food stores were identified, of which 95 were supermarkets (34.1%), 95 grocery stores (34.1%), 17 (6.1%) convenience stores, 37 (13.3%) bakeries, 11 (3.9%) butchers, 19 (6.8%) fruit producers and 5 (1.8%) others (data not shown).

Table 2 shows the relationship between environmental-level variables (availability of four different types of food establishments and income) and obesity. The associations were not substantially different when using absolute compared to relative measures. Results for absolute measures are presented here, while those for relative measures can be found in the Supplementary Material (http://cadernos.ensp.fiocruz.br/site/public_site/arquivo/supl-e00144618_6899.pdf). In the crude analysis, women living in neighborhoods where supermarkets were available presented a 6% higher prevalence of obesity, when compared to those without this food establishment in their neighborhood (PR = 1.06; 95%CI: 1.01-1.10). After controlling for individual variables (age, marital status, family income, education, employment, consumption of fruits and vegetables and food purchasing behaviors), the association between supermarket exposure and obesity was maintained (PR = 1.05; 95%CI: 1.01-1.10). The direction of the association between availability of healthy establishments in the neighborhood and obesity was the same. After adjusting for demographic, socioeconomic, food intake and food purchase variables, obesity and healthy establishments were associated (PR = 1.02; 95%CI: 1.00-1.04). There was an inverse association between the mean income of the buffer and obesity, meaning that the higher the neighborhood-level income, the lower the prevalence of obesity, even after adjusting for potential confounders. A 36% lower prevalence of obesity was observed in women living in areas with a higher mean income, when compared to those living in more disadvantaged areas. (Table 2). Grocery stores and convenience stores were not associated with obesity in the multilevel Poisson regression analysis.

Table 3 shows the mean number of each type of food store for the full analytical sample and stratified by individual-level income tertile. The highest mean was found for healthy establishments (6.44; SD: 0.83), followed by supermarkets (2.98; SD: 5.41), grocery stores (2.57; SD: 2.89), and convenience stores (0.33; SD: 0.67). As income tertiles increased, there was an increase in the mean number of convenience stores.

Table 3 also shows results for the assessment of the interaction between type of food store and individual-level income (in tertiles). There was a significant interaction for the 1st tertile (lowest income) with supermarket and healthy stores, that low income women with access to these stores had a 7% and 2% higher obesity risk compared to those without access (PR = 1.07; 95%CI: 1.03-1.12 and PR = 1.02; 95%CI: 1.00-1.05, respectively). A significant interaction was also found among women in the 3rd income tertile but for supermarkets only (PR = 1.03; 95%CI: 1.00-1.07).

Discussion

We explored the association between the characteristics of the food environment and obesity in women living in a midsize Brazilian city, thus contributing to the literature on neighborhood effects on obesity in a middle-income country. We found that obesity was positively associated with the availability of supermarket and healthy food establishments in the neighborhood and inversely associated with the average income of the neighborhood.

This is consistent with the literature which shows that contextual characteristics are associated with obesity in men and women ¹, however, no studies have been found that highlight the relationship between food environment and obesity in women specifically. Importantly, in our study 76% of the women interviewed were responsible for purchasing food in their homes, and this reinforces the importance of studying the relationship between the food environment and obesity among women since they are responsible for making purchasing decisions in their households. The obesity prevalence of 33.1% among participating women, as well as obesity being positively associated with age and inversely related to family income and schooling are in line with previous research ^7,9,10,11,58. One of the possible explanations for the relationship between the neighborhood and obesity, especially in poor economic communities, is due to the high exposure to unhealthy foods and limited access to healthy foods ⁵⁹. Although economically disadvantaged populations have a good density of establishments ^60,61, the quality of the foods provided may inhibit a healthy eating behavior among the residents who depend on this trade ⁶². Evidence suggests that the neighborhood socioeconomic status, both in developing and developed countries, can influence the health of individuals and still affect many of the factors associated with the prevalence of obesity by environmental factors, such as access to healthy food stores, healthy restaurants, places for physical activity and parks, and price and availability of healthy food ^{29,38,42,58,63,64,65}.

We hypothesized that being exposed to a supermarket would reduce the prevalence of obesity, since women would have greater diversity in choice and thus greater opportunities for buying healthy foods, but our findings did not support this hypothesis. In our study supermarket availability was positively associated with obesity, whereas the majority of previous research, which was conducted in different social contexts, for it occurred in developing countries, has reported an inverse relationship ^18,25,29. However, two previous studies reported results similar to ours. In one study, the relationship between obesity and living in a “food desert” (i.e., areas without healthy foods available), in adults in Detroit (United States), Budzynska et al. ⁶⁶, found that people who shopped at supermarkets had higher BMIs compared to those who bought groceries from other food stores. Similarly, Chaparro et al. ⁶⁷ found that in Los Angeles (United States), residents who had a supermarket in their neighborhood had twice the risk of being obese when compared to those without supermarket access (OR = 2.02; 95%CI: 1.44-2.83), however, the neighborhood density of unhealthy food outlets did not modify this association. One possible explanation for this association is that supermarkets sell different types of foods, both healthy and unhealthy, and that the proportion of the latter to the former is often higher ^67,68. In addition ultraprocessed foods of low nutritional quality are usually less expensive than fresh healthy foods, so although both can be bought in supermarkets, cheaper foods might be more popular ^36,68,69,70.

We also found that obesity was positively associated with the access to a healthy food establishment. This relationship needs to be studied in more depth, but one avenue of research might be to understand if the price of healthy food is one of the possible factors that explains this result. There is evidence suggesting that healthier diets tend to be more expensive and that when food prices increase, the quality of diet decreases ^68,71,72. Making healthy foods cheaper is one way to encourage people to buy these foods, and this is only possible through monetary policies. A promising way to improve food intake is through food taxation ⁷³. This is already evident in the scientific literature, demonstrating that both subsidies and taxes influence food purchasing behaviors, and consequently, food intake, but none of the evidence is from Brazil specifically ^74,75.

In this study we used the NEMS score to classify food establishments. In the literature different measurement methods can be used to evaluate the food environment, both at the community level, through the location and density of food retailers, as well as at the consumer level, to assess in-store food availability, variety, price, and quality and to qualify stores as healthy or unhealthy ⁷⁶. In one study on adults from Philadelphia (United States), that has used the same method to describe the food environment, people were more likely to shop at food stores closest to home if the store had higher NEM-S scores, and to travel for shopping if stores had lower scores ⁶². Many studies, however, classify the environment into healthy or unhealthy based on commercial classifications only, such by categorizing “supermarkets” as “healthy stores” ^25,77. This classification presents limitations, and one is that stores usually offers both healthy and unhealthy foods ³⁶. A combination of methodologies as we used here may enable a more comprehensive assessment of the relationship between the environment and obesity ³⁰.

The results of the present study should be interpreted in light of methodological limitations. First, the cross-sectional design does not allow to establish causal relationships and reverse causality cannot be excluded. For example, an unfavorable food environment could lead to a behavioral change that would influence obesity, but, on the other hand, it is possible that the food environment would have adapted to the demand of obese customers. Considering the current Brazilian food policy, a second limitation is the classification of establishments according to the NEMS classification which considers the nutritional composition of foods but not the level of industrial processing, which is being recognized as an important determinant of health ^78,79. The third limitation is our focus on the residential neighborhood since women can buy food near their workplace or travel to shop in different areas. This is especially important considering that participants reported buying only half of the staple items in establishments located in what they considered to be their neighborhood. Finally, the timeframe for assessing food intake (i.e., 3 months prior to interview), differs from the 12-month period often used in other studies. Nonetheless, the timeframe chosen was consistent with timing of food environment.

Worldwide, changes in the political landscape related to nutrition are occurring, with an increasing number of countries taking regulatory action to improve the food environment to promote healthy and quality foods and Brazil is no different. Important public policies have been proposed, including some that acknowledge that the environment can benefit the population’s health by reinforcing the contextual approach to obesity prevention and promoting healthy eating ^80,81,82,83.

This paper adds to the literature from elsewhere that suggests residential neighborhood may be an important contextual factor for obesity but that understanding it properly, and within country-specific settings, is likely to be an important part of ensuring that obesity prevention strategies are effective.