Urban violence as a predictor factor of obesity: longitudinal evidence from Sao Paulo, Brazil

Abstract

Background

Violence and obesity are global public health challenges that impose significant health burdens. However, the prospective association between urban violence and obesity remains insufficiently understood, especially in low- and middle-income countries. This study aimed to investigate if neighborhood crime-related violence is a possible predictor of obesity among adults residing in São Paulo, Brazil.

Methods

Data were from the ISA-Physical Activity and Environment cohort study. The sample comprised 815 adults without obesity at baseline, assessed in 2015 and 2021. Violent crime rates within 1,000-meter linear buffers were objectively measured and categorized into five pattens (persistent low violence, decrease, persistent intermediate violence, increase, persistent high violence). Log-binomial and linear regression models were used to evaluate the prospective associations of baseline rates and changes in urban violence with changes in BMI and incidence of obesity from 2015 to 2021.

Results

Over six years, the incidence of obesity was 14.8% (95%CI: 11.88–16.68), along with an average BMI increase of 1.08 kg/m² (95%CI: 0.87–1.28) among adults. Crime rates were associated with residential location, reflecting variations in urban violence across neighborhoods with different socioeconomic contexts. In the fully adjusted model, persistent intermediate violence (IR 2.74, 95%CI: 1.41–5.34), increase (IR 2.48, 95%CI: 1.23–4.98), and persistent high violence (IR 2.77, 95%CI: 1.45–5.26) in violence exposure were positively associated with obesity incidence. But for changes in BMI the relationship was close to statistical significance and was strongest in the persistent high violence group.

Conclusions

Our findings suggest that urban violence may play an important role in the causal pathway of obesity, and the importance of integrating social environmental factors into strategies for obesity prevention.

Supplementary information

The online version contains supplementary material available at 10.1186/s12889-025-24459-5.

Background

The World Health Organization defines violence as the intentional use of physical force or power that result in or could result in harm, death, psychological damage, impaired development, or deprivation [1]. Interpersonal violence includes crimes such as robbery and intentional homicides, and disproportionately affect low- and middle-income countries compared with high-income nations. For instance, in Japan, the homicide rate remained stable between 2015 and 2019 at approximately 0.3 per 100,000 inhabitants [2]whereas in the United States, it was approximately 5.0 per 100,000 inhabitants [3]. In contrast, Brazil’s homicide rate ranged from 28.9 per 100 000 inhabitants in 2015 [4] to 21.7 in 2019 [5]. According to United Nations data, between 2015 and 2018, Brazil ranked among the top 10 countries with the highest robbery rates [6] and among the top 15 for intentional homicide rates [2].

High rates of violence are a major public health concern. Beyond its direct impact on life expectancy by contributing to premature mortality [7]evidence shows that violence is associated with increased risks of injuries, mental health issues, adoption of harmful behaviors, and the development of non-communicable diseases such as cardiovascular diseases, cancer, and diabetes [8]. Moreover, exposure to violent environments can also negatively influence health-related behaviors such as physical activity and dietary patterns [9] contributing to an increased risk of obesity.

This relation between obesity and violence become even more relevant, considering that obesity has emerged as a global epidemic, affecting 16% of the adult population [10]. In middle-income countries such as Brazil, obesity has drastically increased in recent decades [11, 12] affecting 24.3% of the adult population [13]. The complexity of obesity underscores the need for a broader perspective to understand the factors contributing to its multifactorial etiology and to develop more effective prevention strategies in low- and middle-income countries, including its possible relationship with violence. Existing studies on behavior, obesity, and violence are predominantly in high-income countries [14].

One of the few studies conducted in a low- and middle-income country was by Jaime et al. (2011) [15] who performed an ecological analysis in São Paulo, Brazil. Although they did not find a statistically significant association between homicide rates and overweight in the adult population, the study observed an inverse correlation between overweight and access to leisure areas such as parks and public sports facilities. In contrast, a cross-sectional study in Philadelphia, USA, analyzed the relationship between exposure to fatal and non-fatal firearm crime rates and health outcomes and found a significant relationship between obesity and non-fatal firearm injuries [16].

Despite advancements in understanding the determinants of obesity, there remains a gap in knowledge regarding its relationship with neighborhood violence. The only systematic review specifically addressing this relationship highlighted that, out of six original studies, only two observed a statistical association between crime and obesity [14]. Importantly, all these studies were cross-sectional, which inherently limit the ability to capture temporality or causality. Therefore, in addition to the scarcity of studies in low- and middle-income countries where violence is more prevalent, there is also a lack of longitudinal evidence that can allow for the assessment of temporal relationships, help identify potential causal pathways, and enable the examination of changes in both exposures and outcomes over time. This design is essential to establish the mechanisms through which exposure to violent environments may influence complex diseases like obesity. This is particularly important in urban settings [14] where such diseases have physiological, psychological, social, and situational origins.

Given the high rates of exposure to violence and the rising prevalence of obesity in low- and middle-income countries, such as Brazil, this study aims to investigate the relationship between neighborhood crime-related violence and the incidence of obesity among adults residing in São Paulo, the largest city in Latin America.

Methods

Study design

This longitudinal study used data from two waves of the Health Survey (ISA): Physical Activity and Environment. Baseline data were collected through household interviews conducted between September 2014 and December 2015. The second wave of data collection took place from October 2020 to February 2021 using telephone interviews due to the COVID-19 pandemic. Interviewer training and skill development were conducted at all stages of data collection. Further technical details are available in previously published methodological papers [17].

Study location

This study was conducted in São Paulo, a city that accounts for approximately 10% of Brazil’s gross domestic product and is among the most populous cities globally. The city exhibits substantial social inequality across its 96 territories. For instance, disparities of up to 20 years in life expectancy, 327-fold differences in the number of households in slums, and 224-fold disparities in the availability of formal employment have been observed across districts [18]. Despite these, São Paulo ranked 28th in the National Municipal Human Development Index, with a relatively high score for Brazil [19].

Regarding criminality, there is up to 143-fold disparity in racial violence, 60-fold differences in femicide rates, and 16-fold differences in homicide rates across districts [18]. Notably, in 2000, São Paulo was the most violent capital in Brazil, with an intentional homicide rate of 51.05 per 100,000 inhabitants [20]. By 2015, it had become one of the least violent capitals, with a rate of 8.5 per 100,000 inhabitants [20, 21]. However, these rates remain high compared to other countries [2, 3].

Sample and sampling process

A stratified probabilistic sampling method was used at baseline, with a two-stage selection process involving census sectors and households. The baseline sample consisted of 3,410 individuals aged 18 years or older. In the second wave, 1,241 adults were evaluated, representing 36.4% of the baseline adult sample. Exclusion criteria for the second wave included relocation out of São Paulo, pregnancy, cognitive impairment that impeded questionnaire responses, recent accidents or surgeries preventing usual activities at the time of the interview, and permanent wheelchair use. A detailed flowchart of participant retention between the two waves is provided in a previously published article [22]. Overall, the analytical sample (n = 1,241) was slightly more represented by women, older and participants with more formal education than the original sample [22].

For this study, 277 individuals with obesity at baseline were excluded, along with 33 individuals with incomplete weight and height data and 116 individuals who had moved in the period. The final sample included 815 adults aged 18 or older who participated in both waves and did not have obesity at baseline.

Outcome variable

Two outcomes were assessed: change in Body Mass Index (BMI) and incidence of obesity. Calibration equations were applied to obtain more accurate estimates of self-reported weight and height values [23]. Age- and sex-specific obesity cut-off points were applied: +2 or more standard deviations from BMI for age and sex for individuals under 20 years [24] and BMI ≥ 30 kg/m² for those aged 20 or older [25].

Exposure to violence

Rates of violent crimes (intentional homicides, street robberies, vehicle thefts, and within public transportation robberies) per 1,000 inhabitants were calculated for 2015 and 2019 within 1,000-meter linear buffers around the participants’ residences. Detailed crime rate data collection procedures are outlined in Supplementary Material 1. Crime data were obtained from the University of São Paulo’s Center for Violence Studies (NEV-USP) and the Technology Transfer Coordination, which accesses official crime data from the São Paulo State Department of Public Safety.

Crime rates in 2015 and 2019 were categorized into tertiles, where the first tertile represented the lowest crime rates and the third tertile the highest. To analyze violence exposure changes, the following classification was used: persistent low violence, decrease, persistent intermediate violence, increase, and persistent high violence (see Table 1).

Table 1.

Change in urban violence exposure using terciles of violent crime rates in 2015 and 2019 within 1,000-metter buffers, São Paulo – Brazil

Violent crime rates (terciles)*			Change in urban violence (exposure variable)
2015		2019
1st tercile	→	1st tercile	Persistent low violence (both in less violent neighborhoods)
2nd or 3rd tercile	→	1st tercile	Decrease
3rd tercile	→	2nd tercile
2nd tercile	→	2nd tercile	Persistent intermediate violence
1st tercile	→	2nd tercile	Increase
1st or 2nd tercile	→	3rd tercile
3rd tercile	→	3rd tercile	Persistent high violence (both in more violent neighborhoods)

*1st tercile is the category with lower violent crimes rates and 3rd tercile is the category with the highest crimes rates.

Environmental perception and socioeconomic variables

At baseline, participants reported whether their neighborhood safety perception was safe, relatively safe, somewhat violent, or violent. This variable was dichotomized as “safe” or “violent”.

The Geographic Index of the Socioeconomic Context for Health and Social Studies (GeoSES) assessed the socioeconomic profile of the environment. GeoSES variable was weighted for the 1,000-meter linear buffers to provide a more accurate reflection of the neighborhood socioeconomic situation. It was categorized into quartiles, with the first quartile representing the lowest index values. Supplementary Material 2 shows this in more detail.

Individual social and demographic variables

Baseline social and demographic data were collected through interviewer-administered questionnaires. The covariates considered were sex (female, male), age (18–29, 30–39, 40–49, 50–59, 60 years or older), educational level (up to 4 years, 5–10 years, 11 or more years of schooling), ethnicity/color (White/Asian, Black/Brown/other), and marital status (living with or without a partner). Moreover, the categorization of ethnicity/color aligns with affirmative action policies for public university admissions supported by Brazilian legislation [26] which considers Brazil’s racial diversity and its history of social inequalities. The conceptual model is shown in Supplementary Material 3.

Statistical analyses

Descriptive analyses of absolute and relative frequencies were performed to assess the sample profile and evaluate BMI changes and obesity incidence. Crime rates were also analyzed. Bivariate analyses were applied to assess the relationship between crime rates and environmental variables. Log-binomial regression models with incidence ratios and linear regression models were used to evaluate the relationship between violent crime rates, obesity incidence and BMI changes, respectively. Three models were structured: Model 1, crude analysis; Model 2, adjusted analysis for individual variables (sex, age, education, ethnicity/color, and marital status); Model 3, analysis adjusted for individual and environmental variables (neighborhood safety perception, GeoSES weighted for the buffer). Wald tests were conducted to assess the significance of the relationship between the outcomes and exposure. Sensitivity analyses were performed to test for statistical interactions among models. Statistical significance was set at p < 0.05.

Crime rate calculations within the buffers were performed using Quantum Geographic Information System – QGIS Desktop 3.36.0, and statistical analyses were conducted using STATA version 16.1.

Results

The sample primarily consisted of females (59.9%), predominantly of White/Asian (57.3%), aged 60 years or older (33.6%), and with at least 11 years of study (28.7%). Regarding the characteristics of the residential environments of participants at baseline, most participants reported a perception of the neighborhood as safe. During the study period, the incidence of obesity was 14.8% (95%CI: 11.88–16.68%), with a mean BMI increase of 1.08 kg/m² (95%CI: 0.87–1.28) among adults.

In 2015, the median rate of violent crimes within the neighborhoods was 10.91 per 1,000 inhabitants. By 2019, this median rate had decreased to 4.95. Median rates of violent crimes across tertiles were 7.41, 10.95 and 16.46 per 1,000 inhabitants in 2015, and 3.25, 4.96 and 7.56 in 2019. Median rates of intentional homicides within the violent crimes rate were 0.08, 0.07 and 0.09 per 1,000 inhabitants in 2015, and 0.04, 0.05 and 0.06 in 2019. The highest median violent crime rate was observed in the third tertile of 2015.

The maps in Fig. 1 illustrate the distribution of changes in exposure to violence within the sample, along with a map of the weighted quartiles of the GeoSES index. The highest concentration of participants with persistent low violence of exposure occurred in the northern part of the map, while the decrease group was more concentrated in the west. The persistent intermediate violence group was more widely dispersed; the increase group was concentrated in the south, and the persistent high violence group was concentrated in the eastern and southern parts of the map. The fourth quartile of the GeoSES, representing the best index values, was concentrated in the central-west area, whereas the first quartile, indicating the lowest values, was concentrated in peripheral areas. A statistically significant relationship was observed between crime rates over time and the local socioeconomic environment, indicating differences in urban violence according to residential location (Fig. 2).

Fig. 1.

Distribution of neighborhood violence exposure for 2015–2019 (A), and weighted GeoSES (B) within 1000 m buffers from the 815 participant’s residence. São Paulo, Brazil

Fig. 2.

Percentages by GeoSES for violent crimes (A–B) and change in violence (C) in 2015–2019

Analyses of obesity incidence and violent crime exposure revealed significant differences for persistent intermediate violence, increase, and persistent high violence in violence exposure compared to the persistent low violence category. All these groups showed a positive association with obesity incidence. Importantly, individuals in the persistent high violence category of exposure to violent crime had a 2.77-fold increased risk of developing obesity (Table 2). Regarding BMI changes and violent crime exposure, a statistically significant association was observed for the persistent high violence category in Models 1 (95%CI 0.07–1.24) and 2 (95%CI 0.04–1.17), with this relationship approaching significance in Model 3 (95%CI −0.01–1.16). Although not statistically significant, the relationship followed a pattern similar to that observed for obesity incidence and urban violence, with residents in more violent environments showing an increase in BMI (Table 2).

Table 2.

Incidence ratios and coefficients for the association between violent crime exposure, obesity incidence and change in BMI for adults of the “Health Survey of Sao Paulo: Physical Activity and Environment” cohort residing in São Paulo – Brazil

	Obesity incidence			Change in Body Mass Index (BMI)
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
	IR (95%CI)	IR (95%CI)	IR (95%CI)	IR (95%CI)	IR (95%CI)	IR (95%CI)
Change in urban violence
Persistent low violence	Ref	Ref	Ref	Ref	Ref	Ref
Decrease	1.47 (0.71–3.02)	1.59 (0.76–3.32)	1.78 (0.83–3.79)	−0.03 (−0.70-0.64)	0.02 (−0.63-0.68)	0.06 (−0.60-0.72)
Persistent intermediate violence	2.51* (1.34–4.69)	2.49* (1.31–4.74)	2.74* (1.41–5.34)	0.59 (−0.06-1.23)	0.44 (−0.19-1.06)	0.40 (−0.23-1.03)
Increase	2.33* (1.22–4.44)	2.27* (1.16–4.43)	2.48* (1.23–4.98)	0.45 (−0.22-1.13)	0.23 (−0.43-0.88)	0.23 (−0.45-0.90)
Persistent high violence	2.42* (1.33–4.40)	2.53* (1.37–4.68)	2.77* (1.45–5.26)	0.66 (0.07–1.24)	0.60 (0.04–1.17)	0.57 (−0.01-1.16)
Ref. Reference category, IR Incidence ratio, Coef. Coefficients, Bold = p-value < 0.05; * = Wald test p-value < 0.05; Model 1 = crude analysis; Model 2 = adjusted for sex, age, ethnicity/color, education, marital status; Model 3 = adjusted for Model 2 + perception of neighborhood safety and weighted GeoSES. Missing values: 7 for ethnicity/color, 5 for neighborhood safety perception

Sensitivity analyses showed no significant interaction between adjustment variables and changes in exposure to violence.

Discussion

The aim of this study was to investigate the associations between neighborhood violence and incidence of obesity over 6 years among adult residents of the largest city in Latin America. Overall, findings of this study indicated that, on average, BMI increased 1.08 kg/m² from 2015 to 2021, with 14.8% of adults transitioning to obesity during this period. Furthermore, individuals residing in neighborhoods with the highest rates of violence within a 1,000-meter radius of their residences from 2015 to 2019 were 2.77 time more likely to develop obesity than those individuals who were residing in neighborhoods with the lowest rates of violence. This is one of the first prospective studies conducted in a middle-income country to suggest that violence rates may play an important role in the causal pathway of obesity in adults. These findings highlight the importance of considering violence in the design of public health initiatives aimed at addressing the obesity epidemic.

The observed obesity incidence in this study was higher than the 5.2% reported over an average 5-year period among British adults [27]; however, this value was lower than the 38.1% reported over an average 8-year period among adults in Tehran [28]. This incidence reflects the continued rise of obesity rates in Brazil. The most recent trend report from the Ministry of Health shows that adult obesity in Brazil has nearly doubled over 15 years, increasing from 11.8% in 2006 to 22.4% in 2021 [12]. In contrast, high-income countries have experienced a more gradual rise in obesity, as observed between 2009 and 2019 in the United States (7%), the United Kingdom (4%), and France (12%) [29]. However, the increase in low- and middle-income countries has been more pronounced, as observed between 2000 and 2020 in Argentina and Bolivia, with approximate 68% and 95%, respectively [30]. Therefore, controlling this disease is critical in low- and middle-income countries like Brazil.

In 2015, Brazil recorded one of the highest robbery rates globally, with 7.6 per 1,000 inhabitants. Similarly, high rates were reported in 2018 at 7.0 per 1,000 [7]. This study observed a higher median rate of violent crimes in 2015, which declined by 2019. Regarding homicide rates in 2015, Brazil also ranked high rates globally with 0.28 per 1,000 inhabitants [2]. Nevertheless, the homicide rates observed among the most violent tertiles in this study were comparatively lower. Despite decreases in violent crime and homicide rates, a sense of insecurity remains. For instance, in 2018, a study conducted in São Paulo reported that 53.7% of the residents believed that neighborhood violence had increased over the past 2 years. Although São Paulo records the lowest rate of intentional violent deaths among capital cities that year [31] 82.2% of respondents feared that a close relative could be murdered. Additionally, 3% reported avoiding walking on foot, 46.2% refrained from visiting certain neighborhoods or streets, and 51.4% avoided going out at night [32].

This study revealed that individuals who continued to live in violent areas or experienced an increase in neighborhood violence had a higher risk of developing obesity. Aligning, a study using 14 waves from the Household Income Labour Dynamics Australia (HILDA) survey to evaluate the relationship between violent crime, obesity, and BMI utilized postal codes to access crime rates corresponding to neighborhoods in major cities or towns in rural Australia. For each standard deviation increase in crime rate, there was a 0.1 standard deviation increase in BMI and a 2.1% increase in obesity prevalence. This association was mediated by physical activity, as residents of more violent areas were less likely to engage in physical activity, thereby increasing the risk of obesity [33].

Two other studies, conducted in the United States, also found a relationship between crime and obesity. In a cross-sectional study of adults in New York City, Stolzenberg et al. (2019) [34] examined the association between violent crime rate across 34 areas designated by the local health department and obesity. They observed a relationship between violent crime rates and obesity when considering interactions with race/ethnicity, highlighting that Hispanic and Black adults were more likely to have obesity when exposed to violent crimes. Similar findings were reported in an ecological study by Singleton et al. (2023) [35] who analyzed the violent crime rate and obesity prevalence across 798 census tracts in Chicago, stratified by the predominant race/ethnicity in each tract.

In Brazil, two studies were identified for comparison: one cross-sectional and the other ecological. Mendes et al. (2013) [36] using data from the Surveillance of Risk and Protective Factors for Chronic Diseases by Telephone Survey (VIGITEL) in Belo Horizonte, Brazil, measured urban violence using the homicide rate in the census tracts and observed a significant prevalence ratio of 1.45 for overweight. However, details about the treatment and use of homicide data were limited because this analysis was not the primary focus. Jaime et al. (2011) [15] conducted an ecological study across 31 districts of São Paulo and found no statistically significant association between the homicide rate and overweight prevalence, suggesting that other environmental factors may have a more prominent impact on overweight prevalence. These findings may not accurately reflect individuals’ actual exposure to the social environment and urban violence, as data were collected across large, variably sized areas, encompassing remarkable socioeconomic and demographic variation.

There are some hypotheses regarding mediation mechanisms that possibly explain the relationship between violent regions and increased obesity risk that were not assessed in this study. One possible mediation is the limited opportunity to adopt healthy behaviors, including physical activity, healthy eating, and access to public spaces such as squares, parks, bike lanes, public clubs, and establishments offering healthy foods that promote health living. Neighborhood crime can lead to a lack of investment in, or deterioration of, public infrastructure, commercial establishments, and environmental aesthetics [37, 38]. Consequently, violent areas may experience cycles of low maintenance and inadequate facilities essential for promoting these behaviors. This, in turn, may encourage the persistence of local danger and crime, deterring residents from accessing public spaces and businesses. Supporting this hypothesis, studies have demonstrated that more violent environments often have fewer suitable facilities for physical activity and a heightened sense of insecurity for outdoor activities [39, 40].

In a study on changes in the availability and distribution of open public spaces in São Paulo, Teixeira et al. (2022) [41] emphasized the inequality in their distribution across the city, revealing that wealthier areas have more of these spaces. The presence and maintenance of open public spaces promote physical activity during leisure time [42] a behavior associated with obesity prevention [23, 40]. An Australian study observed that participating in physical activity mediated the relationship between crime and obesity [33]. Similarly, Richardson et al. (2017) [43] conducted a longitudinal United States study and found that increased neighborhood crime was associated with higher BMI, a relationship mediated by physical activity, primarily among a predominantly black adult sample in lower socioeconomic neighborhoods. Additionally, a longitudinal study of Australian adults revealed a positive association between sedentary behavior and environmental vandalism, suggesting that the sedentary behavior-crime relationship becomes significant when neighborhood maintenance and aesthetics are low [44].

Another mediation hypothesis suggests that violent environments may impact food establishment accessibility and quality. Violent areas often have lower availability of more healthy food options [45] leading to environments characterized as “food deserts” or “food swamps” [46]. These environments promote increased consumption of ultra-processed foods, which are already associated with obesity and other chronic diseases [47]. Recent research has shown that only a few favelas in São Paulo, more socially vulnerable places, have formally registered food establishments [48]. Thus, further investigation is needed to explore the relationships between urban violence, socioeconomic segregation, access to healthy food, dietary quality, and obesity.

Crime is closely related to socioeconomic and racial segregation. Socially vulnerable populations are more likely to be exposed to urban violence [38, 49]. A Brazilian study using data from 152 cities found a significant relationship between economic segregation and homicide, showing a 50% increase in homicide rates per standard deviation increase in the segregation index [50]. Another national survey in Brazil reported a higher prevalence of interpersonal violence among Black/Brown adults with lower educational levels and income [51]. This social vulnerability extends beyond economic aspects to encompass racial/ethnic dimensions, given the historical link between racism and socioeconomic and health inequalities [37, 49] – particularly in Brazil, a highly mixed-race country. A study in Chicago, USA, found that violent crime rates were associated with obesity in communities predominantly composed of Black and Hispanic individuals, an association not observed in predominantly White communities [35]. These inequalities were observed in our study, where relationships were found between violence variables and the GeoSES index, which includes socioeconomic and health components, thereby underscoring its importance in the relationship between neighborhood violence and obesity.

Another mechanism that may explain the observed relationship between obesity and urban violence is continuous exposure to a violent environment. This exposure signals a breakdown in social order for residents, creating a perceived sense of potential danger that leads to stress, insecurity, and fear [33, 40]. Chronic exposure to violent environments may trigger either physical or biological stress responses [9]. Physical responses include reduced participation in social and community activities [9]while biological responses are related to allostatic load, causing cumulative physical wear and tear on the body. Allostatic load involves the continuous adaptation of the body to a hostile and stressful environment. This adaptation stimulates physiological mechanisms that release stress-responsive hormones such as glucocorticoids, which may disrupt inflammatory cytokines levels [52]. Elevated glucocorticoid release can influence food intake by increasing appetite and preference for high-calorie foods, raising insulin levels that can promote greater body fat accumulation [52]. However, further evidence is required to elucidate this potential explanatory mechanism [43].

To the best of our knowledge, this is the first cohort study from a middle-income country to use direct crime data to examine the incidence of obesity and BMI changes. This study provides robust evidence to elucidate the relationship between obesity and violence, given that systematic reviews indicate a scarcity of longitudinal and convergent evidence regarding the determinants of obesity [14, 53]. In addition to obtaining direct measurements of violence using data from official and reliable sources, this study considers people’s perceptions of security. Assessing neighborhood exposure through buffer zones rather than census tracts provided a more accurate representation of individuals’ exposure to violence in their immediate surroundings. Importantly, the longitudinal design, combined with the analysis of incidence and changes in exposure, also strengthened the results, while the statistical approach allowed the direct calculation of the incidence ratio and, therefore, the risk of developing obesity.

This study has several limitations. First, the study period overlapped with the COVID-19 pandemic, which may have introduced residual confounding in the observed associations, particularly regarding obesity incidence, as health related behaviors were likely affected by pandemic-related restrictions and stressors, which could not be fully accounted for in the analyses. Secondly, violence data may be underestimated, as they include only reported crimes. Third, urban violence data may undergo adjustments before being openly released, potentially affecting comparisons over time. Fourth, this study also revealed differences in loss to follow-up across data collection points that could have introduced some attrition bias in our analytical sample, demanding caution in interpreting the results. However, the study’s sampling process minimized selection bias and enhanced sample representativeness, and all statistical models were adjusted for the variables where differences were observed. Finally, our study focus was manly epidemiological in nature, to examine the potential role of urban violence in the causal pathway of obesity, rather than to generalize findings to the entire adult population of São Paulo.

Conclusions

The incidence of obesity during the study period was substantial, affecting approximately 15% of adults in the city of São Paulo. This finding confirms that current public policies have not yet succeeded in stabilizing obesity rates. Importantly, our study provides evidence that exposure to urban violence is associated with a higher risk of obesity. This suggests that only promoting healthy habits may not be adequate to mitigate the risk of obesity in violent contexts. Consequently, our findings support the necessity of considering the social environment when designing obesity prevention and health promotion strategies and actions, such as urban planning and community safety initiatives.

Finally, it remains essential to conduct further longitudinal studies to investigate the mediating mechanisms underlying the relationship between urban violence and obesity. They will aid in refining the focus of interventions and public policies, enhancing their effectiveness in addressing obesity.

Supplementary information

Below is the link to the electronic supplementary material.

Author contributions

BMO participated in the study’s conceptualization and design at all stages, data analysis and critical interpretation, and manuscript drafting and is responsible for all aspects of the work, ensuring accuracy and integrity. AAF contributed to the conceptualization and design of the study in all phases. MBN analyzed and interpreted data related to geoprocessing, violence rates, and the social environment. ITP participated in the geoprocessing of participant data and violence rates. GIM contributed to the discussion and interpretation of the statistical analyses. All authors were involved in the critical review of intellectual content at all stages and have read and approved the final manuscript.

Funding

The ISA-Physical Activity and Environment study was part of a thematic project funded by the São Paulo Research Foundation (FAPESP, grant number 2017/17049-3), along with an academic scholarship for BMO (FAPESP, grant number 2021/03277-0). AAF holds a research productivity scholarship funded by The National Council for Scientific and Technological Development (CNPq; grant number 309301/2020-3). Research, Innovation and Dissemination Centers was funded by the São Paulo Research Foundation (FAPESP, grant number 2013/07923-7).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki and it was approved by the Ethics Committees of the School of Arts, Sciences and Humanities (reference number: 10396919.0.0000.5390) of the University of São Paulo, and by the Department of Health of São Paulo City (reference number: 32344014.3.3001.0086).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.