Socioeconomic determinants of overweight and obesity among Mexican children and adolescents: Systematic review and meta-analysis

2.1 Search strategy

A sensitive search was developed to include index terms, free-text words, abbreviations, and synonyms to combine the key concepts for this review (Appendix 1). The databases searched included EMBASE, MEDLINE, LILACS, CINAHL, Global Health Library, ERIC, PsycINFO, ScienceDirect, Scopus, AGRICOLA, and SciELO Citation Index. Whenever possible, searches were done in Spanish to capture relevant references. In addition, the search engine Google Scholar and the COMO project database were used. The COMO project database currently includes over 1200 references relevant to childhood and adolescent obesity in Mexico. These references have been collected from indexed, non-indexed, and gray literature sources since 2020 and encompass evidence dating back to the early 1980s.¹⁵ In addition, reference lists of included papers were examined for additional publications. This review considered full-text papers and abstracts in English, Spanish, or Portuguese from studies published from 1995 onward, and searches were done in June 2024.

2.2 Selection criteria

2.3 Data selection

Two reviewers (MA-M, CFM-G) screened titles, abstracts, and relevant full texts. One reviewer (MA-M) extracted data from the studies and checked 10% by a second reviewer for consistency (CFM-G). A third author (Y.Y.G.-G.) was contacted in case of any disagreement. A data extraction form was created following the PECOS framework, which included relevant data from the included studies, such as population characteristics (e.g., sample size, target population, mean age, sex distribution), study design; setting characteristics (e.g., city, Mexican state, recruitment location); exposure (any SES indicator relevant to the household or environment of the included participant), outcome (e.g., BMI and any other anthropometric or adiposity measurement considered).

2.4 Data synthesis

The data extracted from the included studies were synthesized narratively, and the main characteristics were tabulated. As has been reported previously,^{16, 24} the nutritional status of young people might vary depending on the age of the references used to categorize BMI. Usually, four BMI categories are used when classifying children and adolescents, including “underweight,” “normal weight,” “at risk of overweight,” and “overweight.” Some other references use “underweight,” “normal weight,” “overweight,” and “obesity.” For the synthesis purposes of the current work, the categories “at risk of overweight” and “overweight” were unified. Moreover, the categories “overweight” and “obesity” will consider children and adolescents within the two highest BMI categories, regardless of the anthropometric reference used across studies.

Relevant SES indicators of participants, including measurements related to wealth, income, education, occupational prestige, and subjective social status, relevant to the children, their parents, household, or living/studying area, were categorized into the following main groups: wealth, income, living environment (urban vs. rural), education, ethnicity, and parental occupation or employment. The results section of this review outlines how each SES indicator or category was standardized and grouped for analytical purposes.

2.5 Risk of bias assessment

The JBI (formerly known as Joanna Briggs Institute) critical appraisal tool for cross-sectional studies was used to assess the quality of the included studies.²⁵ This tool assessed the methodological quality of the included studies by evaluating how well each study addressed potential biases. Within the evaluation, eight critical items were considered: explicit definition of inclusion and exclusion criteria; study participants and setting details; identification of confounding factors; strategies to deal with confounding factors; SES measurements; standard criteria used for measurement of the condition; outcome measurement validity and reliability; and statistical analysis appropriateness. Articles were not included or excluded based on their quality, but the appraisal results were considered in the synthesis process.

2.6 Data analysis

Whenever possible, the Odds Ratio (OR) for overweight and obesity were calculated. If raw data to estimate ORs was unavailable, but an unadjusted OR was provided, this was considered in the analysis. To showcase those participants with higher BMIs and increase the sample size in meta-analyses, as a primary analysis, participants with overweight and obesity were pooled and compared with data on lower BMIs (normal weight); if the data allowed it, the “underweight category” was excluded from the analysis. As a secondary analysis, meta-analyses of specific BMI (overweight or obesity) and/or SES categories were conducted, and results are provided in Appendix 2. Considering that all the papers included were observational studies, a DerSimonian and Laird method was used to construct a random-effects model to account for the heterogeneity within and between studies.²⁶ All results were reported with OR and 95% Confidence Intervals (CI), and the main results are presented in forest plots. The analysis was performed with R statistical software using the library “meta” and “metafor”.

3.1 Measurements of wealth

Twenty-eight studies^{27, 31, 32, 37-41, 43, 45, 48, 51, 54, 58-60, 62, 63, 65-68, 71-73, 75, 76, 78, 84} used household wealth measurements or neighborhood marginalization as proxies for SES. Most of these studies focused on household wealth based on housing characteristics and assets, except for six studies^{27, 45, 67, 68, 71, 84} which considered the degree of marginalization (including wealth indicators) of the area of residency or the location of the school that participants attended. All 28 studies, except for one⁴⁸, studied wealth and marginalization and its association with overweight or obesity.

From those studies measuring household wealth, most^{31, 32, 37, 38, 40, 41, 43, 51, 54, 58-60, 72, 73, 75, 76, 78} used a component analysis, including different household characteristics, housing quality materials on the structure (e.g., floor, walls and roof), public services (e.g., public sanitary sewer system, public water network or electricity availability) and assets ownership (e.g., motor vehicle, television, fridges). Two studies^{48, 65} measured only household assets, and in two other studies^{63, 71} it was unclear how wealth was recorded. Six studies evaluated the degree of overcrowding (i.e., the number of people per room, which differs from the total number of family members, synthesized below in “Family structure and size”).^{38, 39, 62, 65-67} Three studies^{27, 45, 68} used the degree of marginalization provided by the National Population Council, which considers housing services, security perception in the neighborhood, education, and monetary income of the areas where the household is located. Three studies considered wealth or characteristics of neighborhoods or municipalities for the analysis^{45, 67, 84}. One study⁶⁸ considered the type of household, the perception of security, and the distance to the closest park to indicate wealth. One abstract⁷¹ reported using “school social class” as a variable, but how this was classified was unclear.

Overall, most studies found that overweight and obesity were more prevalent among children and adolescents from wealthier households, characterized by better household conditions and higher possession of goods.^{37, 39, 40, 43, 45, 46, 51, 54, 58-60, 62, 63, 67, 72, 73, 75, 76, 78, 79} Only one study⁶⁸ that included school-aged children recruited from a public family clinic found that children who lived in a neighborhood with a high or very high degree of marginalization had a higher prevalence of overweight than those who lived medium to a very low degree of marginalization neighborhoods (p = 0.022). One study⁴¹ that included national data from children 0–23 months reported an insignificant pattern of overweight prevalence based on wealth. However, this same study reported that from 2012 to 2020, the prevalence of overweight and obesity almost doubled in individuals with lower wealth.

Ten studies^{37, 41, 59, 60, 62, 65, 66, 72, 73, 75, 76} that reported wealth measurements at a household level were meta-analyzed. The meta-analysis revealed that children and adolescents from wealthier households (with better characteristics and structure, classified as “medium” or “high” SES) were significantly more likely to have either overweight or obesity compared to those classified as having a low SES (OR 1.43, 95% CI 1.19, 1.72, Figure 2). This likelihood remained significant for children when accounting only for obesity (OR 1.47, 95% CI 1.08, 2.01, Appendix 3, Figure S1).

Also, when considering only those children with the highest wealth (i.e., the best household characteristics and structure, and removing those deemed to have a “medium” SES) compared to those with the lowest wealth categories (i.e., worst household characteristics and structure), the likelihood became higher, showing that those with the best household characteristics were significantly more likely to have either overweight or obesity (OR 1.65, 95% CI 1.46, 1.86, Appendix 3, Figure S2), a likelihood which was even higher when considering obesity only (OR 2.62, 95% CI 1.63, 4.19, Appendix 3, Figure S3).

These results remained significant even when analyzing them in light of the methods used to recall the measurement. For those studies measuring wealth through household goods, services, and structure, those children considered to live with better services/structure were likelier to have either overweight or obesity (OR 1.42, 95% CI 1.17, 1.74, Appendix 3, Figure S4), or obesity only (OR 1.69, 95% CI 1.22, 2.35, Appendix 3, Figure S5). [Correction added on 29 April 2025, after first online publication: The previous sentence has been corrected in this version.] For those studies using household overcrowding as a measurement of household wealth, a similar trend was found with those children living in not-overcrowded households (as a proxy of better wealth) with a higher likelihood of having overweight or obesity (OR 1.59, 95% CI 1.12, 2.27, Appendix 3, Figure S6).

3.2 Living setting (urban vs rural areas)

Twenty-seven studies considered the living environment, based on the size of the communities where participants lived, a proxy of SES.^{27, 31-34, 36, 37, 39-41, 43-45, 48, 49, 51, 54, 56-58, 60, 67, 69, 73-78, 82, 85}. Generally, studies considered rural areas to be those with less than 2500 inhabitants. (Appendix 2) Overall, most of the studies showed higher prevalences of overweight and obesity in children living in urban settings. However, two studies^{33, 34, 36} found no statistically different prevalence of overweight or obesity among those children living in rural or urban areas. One of these studies, presented in two references, included only Indigenous children of Tarahumara origin.^{33, 34}

Fourteen studies^{33, 37, 41, 44, 45, 69, 73, 75-77, 82} that provided sufficient data were meta-analyzed. Results show that participants living in urban areas are significantly more likely to have overweight or obesity compared to those living in rural areas (OR 1.41, 95% CI 1.20, 1.66, Figure 3). This likelihood was higher and remained significant when considering obesity only (OR 1.64, 95% CI 1.39, 1.93, Appendix 3, Figure S7).

3.3 Parental education

Twenty-four studies^{32, 35, 36, 38-40, 42, 46-48, 51, 53-55, 58, 59, 62, 64, 67, 70, 72, 78, 80, 81} considered parental education as a proxy of SES. However, four studies included this characteristic in a comprehensive measurement of SES, and no direct association with the weight status of children was provided^{35, 38, 39, 67}. From the rest, 11 studies^{32, 40, 46, 51, 55, 58, 59, 64, 70, 72, 81} included maternal education, and the remaining studies included the education level of both parents.^{36, 42, 47, 48, 53, 54, 62, 78} One study examined maternal education by years of schooling, while paternal education was assessed based on overall literacy levels.⁵⁴

Most of the studies reported higher prevalences of childhood and adolescent overweight or obesity among households with better-educated parents compared to those with the least-educated parents.^{32, 40, 47, 48, 51, 54, 55, 58, 62, 70, 72} However, a few studies also reported that children of mothers who were less educated or had lower literacy were likely to have overweight or obesity.^{46, 53, 79} One study found no difference in the nutritional status of offspring based on maternal education.⁵⁹

One study⁵¹ reported a high correlation between maternal education and overall SES (wealth estimation). Ullman et al⁷⁸ used data from a national sample and reported that higher paternal education was associated with a lower prevalence of obesity. Nonetheless, the association between maternal education and obesity was positive but not always significant (as it largely depended on other SES variables included in the adjusted analysis). Also, a study conducted among Mayan children⁴² showed that compared to the normal weight mother and child dyads, with each year increase of maternal education, there was a significant decrease in the odds of overweight and obesity in mothers and children. Conversely, with each year's increase in the father's education, there was also an increase in the odds of having overweight or obesity among fathers and children.

Ten studies^{40, 48, 51, 54, 58, 59, 62, 64, 72, 81} that provided sufficient data on maternal education were meta-analyzed. Children and adolescents with well-educated mothers (secondary school or higher) were significantly more likely to have overweight or obesity than those with the least educated mothers (primary/elementary studies or less) (OR 1.44, 95% CI 1.14, 1.82, Figure 4). This likelihood was even higher and remained significant when comparing participants with the most educated mothers (college studies or more) compared to those with the least educated mothers (primary/elementary studies or less) (OR 1.94, 95% CI 1.79, 2.10, Appendix 3, Figure S8). Due to insufficient data, meta-analyses of paternal education were not possible.

3.4 Family structure and size

Eighteen studies^{30, 35, 38-40, 42, 46, 56-58, 60, 62, 64, 68, 72, 78, 79, 81} accounted for family size or structure as a proxy of SES. From these, three studies^{38, 57, 78} accounted for this characteristic as a covariate within their analysis, and no direct association with overweight and obesity was presented. Twelve studies^{30, 35, 39, 40, 42, 46, 56, 58, 60, 62, 64, 68} accounted for the number of people conforming to the household as a proxy of SES. Seven studies^{35, 39, 42, 58, 62, 64, 68} found that participants with the highest prevalences of overweight or obesity were found among smaller families, while two^{30, 56} reported higher prevalences of overweight or obesity among participants from larger families. Two other studies found^{40, 46} no difference in the prevalence of overweight or obesity according to the number of people in the house. However, one study found a significant likelihood of overweight and obesity only when accounting for a higher number of siblings but not with a higher number of people conforming to the family.⁶⁴

Along with the number of family members, nine studies^{40, 46, 56, 58, 62, 68, 72, 79, 81} also considered parents' marital status or maternal partner status. Three studies^{40, 46, 72} found that the partner status, marital status, or father's absence of mothers were not predictors of overweight or obesity among participants. One⁵⁶ reported a significant difference in the prevalence of obesity depending on the composition of the family, with a higher percentage of participants with obesity from families with no fathers. One abstract⁷⁹ found that children from unmarried couples were more likely to have obesity than those from married couples. Two studies^{58, 62} found that children with single mothers were less likely to have overweight or obesity compared to those with both parents at home. One study⁸¹ found a higher prevalence of absence in the home of one or both biological parents among children with overweight or obesity. Two studies also considered adolescent participants and considered the marital status of these adolescents as a predictor of overweight and obesity among female participants,^{38, 60} and one of these⁶⁰ found that those who cohabit with a partner report a higher prevalence of overweight but not obesity.

Five studies^{39, 40, 58, 62, 68} considered cohabitation with other family members (usually grandmothers) as a potential factor contributing to overweight or obesity. One⁵⁸ described that children from parents or single mothers who cohabit with relatives were more likely to have overweight or obesity. One study⁶² found that children from nuclear families had a higher prevalence of overweight or obesity compared to those from other familial compositions, including extended families. However, one study found that cohabiting with grandparents does not differentiate the prevalence of overweight or obesity among participants.⁴⁰ The rest did not report differences between cohabitation status and nutritional status.

3.5 Ethnic origin

Sixteen studies^{32, 33, 36, 37, 40-42, 46, 48, 54, 58, 60, 61, 67, 76, 82} considered the ethnic origin of participants as a proxy of SES. However, five^{33, 42, 47, 67, 82} recruited participants from specific ethnic groups but did not study the role of ethnicity in participants' overweight or obesity status. From those studies that recruited non-Indigenous and Indigenous participants, seven studies^{32, 36, 37, 41, 54, 60, 76} recalled Indigenous status if at least one woman in the household or household head spoke an indigenous language. One study⁴⁰ captured ethnicity by asking mothers if they self-identified as Indigenous. One more⁴⁶ used two variables to account for ethnicity, whether an indigenous language was spoken in the household or for the type of community (greater or not than 50% of the community is indigenous). One study⁶¹ recruited participants in Yucatan and recorded ethnicity through the last name of participants (having two Mayan surnames, one Mayan surname, or a non-Mayan surname). Three studies^{27, 50, 67} also considered children from Indigenous schools, and two reported a lower prevalence of overweight and obesity for those participants attending Indigenous schools compared to general public and private schools.^{27, 50}

Generally, the prevalence of overweight and obesity among non-Indigenous participants was higher when compared to their counterparts.^{37, 40, 54, 58, 76} However, some studies only found higher prevalences in specific age or sex groups (e.g., women 11–19 years only, but not younger populations).^{32, 60, 76} Some others did report no differences in the prevalence of overweight and obesity according to the ethnic origin of participants.^{36, 46}

A meta-analysis was conducted, including six studies^{37, 41, 48, 54, 60, 76} that categorized their participants as Indigenous or non-Indigenous. The results show that non-Indigenous participants are significantly more likely to have overweight or obesity compared to Indigenous children (OR 1.55, 95% CI 1.22, 1.96, Figure 5).

This likelihood was higher when considering only obesity (OR 2.22, 95% CI 1.93, 2.56, Appendix 3, Figure S9).

3.6 Income or monetary measurements

Thirteen studies^{27, 35, 36, 39, 52, 53, 55, 61, 62, 67, 68, 73, 80} included household income or monetary measurements as a proxy of SES. However, half of these studies included income as part of a compound measurement or as a covariate within their analysis, and no direct association between income and overweight or obesity was presented.^{27, 35, 36, 39, 62, 67, 73} From the other half, income or monetary assets were measured differently. For example, three studies^{55, 61, 80} accounted for the total family income. One⁵² accounted for two indicator variables for medium and high-income inequality: GDP per capita and schooling for each federal entity. Another study⁵³ included monthly family income, spending on food per month (percentage of minimum wage), and expenses in food per capita per month (percentage of salary minimum). One study⁶⁸ accounted for the perception of financial issues.

One study found that the prevalence of overweight or obesity at high inequality levels was higher than the low or medium inequality level participants. However, this was only significant for boys.⁵² Gonzalez-Rico et al⁵³ found that food expenditure per capita per month (percentage of monthly salary) was significantly associated with obesity. However, this was included as a covariate of a model measuring the association of family dysfunction with obesity. Mendez et al⁶¹ reported that the odds of having overweight or obesity were significantly lower among children from higher-income families. Ramirez-Serrano et al⁶⁸ found that those children living with families that perceived having economic issues were significantly more likely to have overweight or obesity. On the contrary, Jimenez-Cruz et al⁵⁵ reported that children living in households with a monthly income >600 US dollars were more likely to have obesity. Due to the differences in the data and measuring methods of income or monetary measurements, meta-analyses were not possible.

3.7 Parental occupation or employment

Twelve studies accounted for parental employment as a proxy of SES.^{38-40, 53, 56, 57, 62, 64, 72, 80} Three studies included this variable as a covariate within their analysis as part of an index or compound measurement, and no direct association between parental occupation and overweight or obesity was presented.^{38, 53, 56} The categorization of this factor varied widely across studies, from those reporting parents having or not having jobs to those such as Campos et al⁴⁰ that conceptualized maternal employment as having a paid job (including the hours worked in a week) but also accounted for whether the job was formal (including a tax contributory social protection system) and if it was a full- or part-time job.

Overall, six studies reported no statistical differences in the prevalence of overweight or obesity among children from working mothers compared to those with unemployed mothers.^{39, 40, 62, 72, 80, 81} Manila et al⁵⁷ studied two cohorts of children in Oaxaca in 1972 and 2000. This study categorized SES by using only the working status of parents. They found that the prevalence of overweight and obesity did not differ among SES groups in 1972 and 2000; however, the prevalences were significantly higher within each SES group in 2000 compared to 1972.

A meta-analysis was conducted, including five studies^{40, 58, 62, 72, 81} that evaluated maternal employment and overweight and obesity prevalence. We found that participants with employed mothers (any type of employment) are more likely to have overweight or obesity, compared to those with unemployed mothers (OR 1.39, 95% CI 1.30, 1.48, Figure 6). Due to insufficient data, the analysis only for obesity considering different types of employment (e.g., full-time vs part-time, or formal vs. informal), or considering fathers' employment was not possible.

3.8 Type of school

Eight studies^{27-29, 50, 59, 70, 83, 84} considered the type of school participants attended as a proxy of SES. Two study papers considered participants from different types of schools, including public (without a financial fee), private (with an economic fee), Indigenous (public schools located in rural communities with Indigenous populations monolingual and bilingual), and schools from the National Council of Educational Development (public schools located in small rural areas that also benefit the migrant community and have a maximum of 29 children each).^{27, 50} One paper also considered the type of public school (offering full-time education, morning or afternoon shift).²⁷ Five studies^{27-29, 50, 84} reported a higher prevalence of overweight and obesity among private schools. Nevertheless, two studies^{59, 70} showed a higher prevalence of overweight and obesity among children from public schools. One study found no differences in the prevalence of overweight and obesity, considering the different school types.⁸³

A meta-analysis was conducted, including six studies^{27-29, 59, 83, 84} that provided information about public and private schools and the prevalence of overweight and obesity. We found no significantly higher likelihood of overweight and obesity among participants attending private schools than those attending public schools. (OR 1.10, 95% CI 0.82, 1.46, Figure 7).

According to one study,²⁷ those participants assisting with an afternoon shift schooling were less likely to have overweight or obesity than those attending morning shifts or full-time schooling. Moreover, one longitudinal study³⁸ found that adolescents who finished only primary school are more likely to transition to obesity compared to adolescents who continue or finish secondary school.

3.9 Other factors identified as SES proxies in Mexico

As part of the ENSANUT survey, health service affiliation was considered a potential SES attribute.^{60, 75, 76} No statistical differences in the prevalence of overweight or obesity were reported for these characteristics.⁶⁰ Also, ENSANUT records whether participants are food aid program recipients; however, no statistical differences in the prevalence of overweight or obesity were reported according to their participation in food programs.^{43, 60, 75, 76} Other studies that recorded food aid as an SES characteristic found similar results, such as Fernald et al⁴⁶ which conducted a community survey and reported whether a federal breakfast program benefited the community. Such a study found no significantly higher likelihood of overweight or obesity among children in communities that received such programs. Morales-Ruan et al⁶³ reported in an abstract that no association existed between food aid and overweight or obesity. Two studies^{38, 60} recorded child/adolescent work as a potential SES indicator from participants, but no results were provided in relation to their nutritional status.

3.10 Risk of bias

Most of the studies covered all the items evaluated through the JBI tool. However, abstracts were the ones falling short of reporting key methodological details. From the items evaluated, the criteria for inclusion in the sample were clearly defined for most of the included papers, except for 11 papers (mainly abstracts),^{33, 39, 48, 50, 52, 63, 70, 77, 79, 80, 85} where this was unclear. Likewise, some details were ambiguous or not reported when describing the setting in 14 studies.^{27, 33-35, 42, 49, 50, 52, 53, 56, 63, 71, 77, 79, 80, 85} The methods to collect or categorize SES data within 11 studies were unclear.^{45, 48, 50, 52, 53, 62, 63, 70, 71, 79, 80, 85} Moreover, weight measurements to estimate overweight and obesity were mainly recollected using validated and reliable methods across all papers. Appendix 4 provides a complete overview of the evaluation of the risk of bias.