The retail food environment and its association with body mass index in Mexico

The retail food environment—comprising the types, locations, and accessibility of food outlets—has been recognized as a major determinant of health and obesity. Mexico has seen rapid rises in obesity during the past three decades, coinciding with structural changes in its food system and retail landscape, including influences following NAFTA such as the entry of global fast-food chains and greater availability of low-cost ultra-processed foods. Individual-level factors alone cannot explain the rapid population-wide increase in BMI, suggesting environmental drivers play a key role. Prior research in Mexico indicates excessive access and exposure to calorie-dense foods and beverages, with healthy options relatively more expensive. This study aims to analyze whether densities of different retail food outlets in urban Mexico are associated with adult BMI, and to identify which aspects of the retail environment might be most obesogenic to inform targeted policy interventions.

Most retail food environment studies have focused on high-income countries, with fewer studies in low- and middle-income countries and very few testing associations with obesity. Mexico’s retail food context has been influenced by the U.S., including greater exposure to energy-dense, nutrient-poor foods and larger portion sizes. Evidence from a Mexican city suggested excessive access to calorie-dense foods and beverages. Meanwhile, consumption of sugar-sweetened beverages is high in Mexico, prompting a national SSB tax, but there are no national retail regulations to improve healthy food availability. These strands of literature motivated examination of retail outlet densities in relation to BMI at the national level in Mexico.

Design: Secondary analysis of cross-sectional, population-based data linking individual BMI and covariates with neighborhood-level retail food outlet densities in urban Mexico. Population and data sources: Adult participants (N = 22,219) from urban areas (≥2,500 inhabitants) in the 2012 National Health and Nutrition Survey (ENSANUT) with measured height and weight. Exclusions: pregnant women, participants <18 years, missing anthropometrics, and outlier BMI values (<15 or >58 kg/m²). Retail data: Food outlet geolocations and characteristics from the 2014 INEGI economic census, which involved ground-truthing and digital georeferencing. Geography: Neighborhoods proxied by census tract areas (CTAs). There were 55,427 urban CTAs (mean geometric area 0.59 km², range 0.009–5.20 km² in one description; elsewhere urban CTAs mean 0.46 ± 0.56 km², range 0.0002–36.41 km²). Participants were geocoded to CTA centroids (exact addresses unavailable). Food outlet classification: Using INEGI classifications refined by outlet characteristics (product types, menus, and where needed, website information). Categories: convenience stores (mainly SSBs and snacks), fast-food outlets (e.g., pizza, hamburgers, hotdogs, fried chicken, including informal/mobile carts), restaurants (à la carte, sit-down, healthy options included), supermarkets (mega-supermarkets and grocery stores offering broader options including fruits/vegetables), and fruit and vegetable stores. No in-store assessments were conducted. Exposure measures: Outlet density calculated as outlet count per CTA divided by CTA area (km²). Densities computed for each outlet category. Outcomes and covariates: Primary outcome BMI (kg/m²). Covariates included age, gender, household socioeconomic position (SEP; quintiles), physical activity (inactive, moderately active, active; subsample n=10,587), car ownership (owns/does not own; subsample n=8,635), region (north, center, metropolitan, south), neighborhood deprivation (low/high), urbanicity (urban/metropolitan; rural excluded), participation in food assistance programs (yes/no), and health insurance (covered/not covered). Statistical analysis: Multilevel linear regression models with random effects to account for clustering at state (Models A and C) or CTA (Model B); Model D used linear regression to account for selection bias. Models informed by directed acyclic graphs. Multicollinearity assessed via variance inflation factors (<4.0). Survey design, clustering, stratification, finite-population corrections, and weights were accounted for in descriptive and regression analyses. Bonferroni corrections applied. Models: Model A included age, gender, SEP with state as second level; Model B added physical activity, car ownership, deprivation, urbanicity, food assistance, health insurance with CTA as second level; Model C included Model A plus deprivation and urbanicity with state as second level; Model D included Model A plus deprivation, urbanicity, food assistance, and health insurance (selection bias) via linear regression. A mutually adjusted model (state random effect) included all food outlet types together. Sensitivity analyses: Re-ran analyses using waist circumference as outcome (smaller sample); tested interactions between SEP and outlet density; used two-level multinomial logistic regression for SEP and retail environment associations.

- Descriptive environment: Convenience stores had the highest density per CTA (mean (SD) 50.0 (36.9)/km²), followed by restaurants (23.22 (29.3)), fruit and vegetable stores (5.90 (22.7)), fast-food outlets (4.85 (6.4)), and supermarkets (0.35 (1.2)). Many CTAs lacked healthier outlets: 42% (n=10,145) had no fruit/vegetable store and 88% (n=21,209) had no supermarket, whereas only 0.5% (n=119) had no convenience store. Metropolitan areas showed the highest concentrations, with up to 105 convenience stores per CTA and densities up to 438 per km². - Associations with BMI: Higher convenience store density was consistently associated with higher BMI: • Model A: β = 0.003 kg/m² per unit density increase (95% CI: 0.0006, 0.005; p=0.011). • Model C: β = 0.003 (95% CI: 0.0009, 0.005; p=0.006). • Model D: β = 0.003 (95% CI: 0.0001, 0.005; p=0.041). • Mutually adjusted model (all outlet types): β = 0.003 (95% CI: 0.001, 0.006; p=0.006). Model A had the highest ICC (0.11; SE 0.02; 95% CI: 0.07, 0.17). A 10% increase in convenience store density in a high-density CTA corresponds to an estimated 0.13 kg weight increase for an adult 1.60 m tall; per the abstract, a 10% increase corresponds to ~0.34 kg for a 1.60 m adult. - Metropolitan areas: Showed the strongest associations with BMI (β = 0.01, 95% CI: 0.004–0.01; p<0.001). A 10% store density increase in metropolitan areas would represent approximately a 1 kg weight increase for a 1.60 m adult. - SEP and demographics: Lower SEP groups were more likely to have obesity than higher SEP groups; inactivity or moderate activity was associated with higher obesity risk compared to active individuals; women and those aged 45–54 had higher obesity prevalence (descriptive).

Findings indicate that greater availability of convenience stores—outlets assumed to mainly sell sugar-sweetened beverages, snacks, and ultra-processed foods—is associated with higher mean BMI among Mexican adults. This supports the hypothesis that aspects of the retail food environment contribute to obesity risk, particularly in metropolitan areas where outlet densities are highest. The results remain after adjustment for sociodemographic factors, neighborhood deprivation, urbanicity, and after accounting for potential selection bias, suggesting robustness. Given the widespread presence of convenience stores and relative scarcity of supermarkets and fruit/vegetable stores in many CTAs, the retail food environment may promote exposure to energy-dense, nutrient-poor options. These findings underscore the potential importance of environmental and policy approaches to improve the healthfulness of local food environments in Mexico, complementing individual-level interventions.

This national-level study in urban Mexico finds that higher convenience store density is associated with higher adult BMI, with the strongest effects in metropolitan areas. The ubiquity of convenience stores and limited access to outlets retailing healthier options such as fruits and vegetables may increase obesity risk. The study contributes novel evidence from a middle-income country context and suggests that policies targeting the retail food environment—such as improving healthy food availability and limiting exposure to ultra-processed options—may help reduce obesity risk. Future research could leverage longitudinal designs, finer-grained exposure measures, and direct in-store assessments to better establish causal pathways and evaluate specific policy interventions.

- Cross-sectional design limits causal inference despite use of DAGs to guide modeling. - Participant geocoding used CTA centroids due to data protection; exact residential addresses and proximity measures were unavailable, potentially introducing exposure misclassification. - Retail outlet classification relied on INEGI data and, where needed, outlet websites; no in-store assessments were conducted to verify product offerings. - Physical activity and car ownership were available only for subsamples, reducing power and potentially introducing residual confounding. - Waist circumference analyses had a smaller sample size. - Although ground-truthing was conducted for nine geographic samples, comprehensive national validation of outlet data was not performed.