Background
The rise of obesity is an important public health concern due to its impact on nutrition-related chronic diseases such as diabetes mellitus, cardiovascular disease, and cancer, and its toll on disability and health care costs [
1,
2]. High body mass index (BMI) is the second most important risk factor for death and disease burden in Mexico [
3]. The combined prevalence of overweight and obesity (OWOB) in Mexican adults has increased considerably over the past two decades. For women between 20 and 49 years of age it rose from 34.5 % in 1988 to 71.4 % in 2006 and to 70.5 % in 2012. For men older than 20 years of age, OWOB increased from 60.7 % in 2000 to 66.7 % in 2006 and to approximately 70.0 % in 2012 [
4]. Mexico is now second, after the United States, among the OECD [
5] countries with the highest adult obesity prevalence, and has been classified among the top 20 countries with the highest prevalence of adult OWOB in the world [
6]. Although overall changes from 2006 to 2012 may not seem substantial, it is important to investigate if there have been significant changes in the distribution of OWOB among subpopulation groups in order to better understand the epidemic and who it is affecting most.
Socioeconomic indicators such as wealth, education, occupation, and marital status have been associated with OWOB or obesity in previous studies [
7‐
9]. These factors may affect bodyweight through influences on physical activity and diet [
10,
11]. In developed countries, higher socioeconomic status (SES) has been associated with decreased obesity risk, especially among women. In contrast, among developing countries, higher SES has been associated with an increased risk of obesity [
7,
12]. Sex-specific associations between socioeconomic indicators and OWOB have been extensively studied in high income countries, but fewer studies have investigated those associations in middle income countries [
13], with even fewer [
14] addressing time trends for these specific associations in upper-middle income countries such as Mexico.
Although socioeconomic indicators are correlated with each other (e.g. education and wealth), they may also be reflecting some specific relationships with OWOB that emerge when analyzed simultaneously. Therefore, it may be useful to estimate associations of each indicator when controlling for other indicators. For example, if two socioeconomic indicators are correlated positively with each other and also with OWOB but one of the indicators is omitted from the analysis, the expected value of the association of the included SES indicator with OWOB would be biased upwards. On the other hand, if the two SES indicators have opposite associations with OWOB and the association between the included SES indicator and OWOB is positive, the expected value of the estimated association would be biased downwards (given a positive correlation between SES covariates).
To the best of our knowledge, this is the first study addressing associations between multiple socioeconomic indicators, and OWOB for each sex in the Mexican adult population. Additionally, this is the first study estimating OWOB trends at different wealth levels with the most recent nationally representative surveys in Mexico. Knowing how associations have evolved during this study period may be helpful for determining whether the epidemic concentrates in certain characteristics of the population, and if Mexico then needs to refocus policies aimed towards reducing OWOB.
Discussion
The present study examined associations between multiple SES indicators and OWOB, and their evolution in the Mexican adult population from 2006 to 2012. Associations among women were generally consistent with the body of evidence on the relation between obesity and SES from highly developed countries. For low to middle wealth levels in women and for men in general, results were more concordant with what has been observed in low-middle developed countries [
7].
We found that prevalence of obesity among Mexican women increased from 2006 to 2012 at wealth values from the middle to the upper-middle part of the wealth distribution. It has been suggested that the burden of obesity shifts from higher to lower SES as per capita income increases in a given country [
13]; from 2006 to 2012 real per capita income increased about 13 % in Mexico [
23]. On the other hand, recent research has shown that in developing countries with relatively high per capita income and high-income inequality (Bolivia, Peru, Guatemala, Namibia, and Colombia), the prevalence of OWOB in adult women increased more rapidly in the wealthier groups [
24]. However, the authors interpreted these results conservatively; the availability of more data would clarify if this pattern persists. As of 2010, Mexico had a Gini index of 47.2 [
25], which would correspond to category of high income-inequality (Gini index from 42.2 to 74.3) reported in the aforementioned study [
24]. Although obesity in Mexican women significantly increased from 2006 to 2012 among wealth values located in the middle and upper-middle of the wealth distributions, the association between wealth and obesity remained negative for wealth values above the mean at both survey years.
In various SES categories, the proportion of subjects with no excess body weight (BMI < 25) decreased in men from 2006 to 2012. On the other hand, the proportion of women with no excess body weight did not significantly change. However the distribution of women with OWOB was concentrated at higher BMI values, which resulted in the shift of prevalence from overweight to obesity. This increases the risks of chronic diseases [
26‐
28].
Contrary to the perception that OWOB or obesity is shifting to low SES groups, we found that OWOB increased among men with a high education level, and obesity increased among women with relatively high wealth index, as previously noted. Additional research is required to properly identify the determinants of such increases. One possibility could be higher accessibility to and consumption of ultra-processed foods. Retail sales per-capita of ultra-processed drinks and food products increased 29.2 % from 2000 to 2013 in Mexico, and exposure to ultra-processed food products has been linked to urbanization and a higher per-capita income, among other factors [
29]. In regard to formal education and OWOB prevalence among men, our results underscore the necessity to strengthen nutritional and health education within school programs from early grades on up to the highest grades. Prevention at early ages in life should be reinforced. The high prevalence of OWOB in Mexico is justification for public health policy targeted to all population groups, with a focus on subpopulations that are at higher risk of OWOB. In Mexico, public policy is aligning toward OWOB prevention through a combination of interventions in multiple sectors [
29,
30].
We found that OWOB prevalence was higher at the married/cohabitating category compared to the single category, with a greater difference for men. Single individuals may have more awareness of body shape and assign it a greater value. The difference between male and female may reflect the fact that women are more aware of body size than men. Even when married, women may still tend to pursue thinness, which is a culturally reinforced value particularly in developed countries [
12,
31].
Other factors previously shown to modify weight include accessibility to foods with high caloric density, availability of opportunities to engage in physical activity, and lack of public awareness of overweight health hazards [
32,
33]. Individuals with a high SES have more available resources to modify their diet and physical activity and can therefore more easily regulate their weight. It is possible that women become more culturally connected to western body size values as SES increases.
Most studies relating SES to OWOB or obesity are limited to cross sectional samples, even fewer studies have assessed sex-specific change in the prevalence of OWOB or obesity at categories of SES indicators. Most recent studies on women from low and middle-income countries have focused on either wealth or education as a SES measure or analyzed them separately [
24,
34]. We used a multivariate approach for estimating associations between various SES indicators, which allowed us to avoid potential biases caused by omitting any of the available SES indicators. Therefore, coefficients our estimates reflect associations attributable to the variable in question when all other covariates are held constant. It is possible that wealth is a mediator between education and OWOB. We did not estimate this potential mediation, but our additional analyses indicated that part of the total relationship between education and OWOB may be mediated by wealth (Additional file
8).
Although each indicator attempts to measure SES, they may be reflecting different aspects of development. The distinction may be especially relevant when a country is at a transitional stage of its development; it could be that as countries develop, these associations converge. Knowing whether individual SES indicators converge or diverge in their associations with excess body weight highlights the importance of a multivariate approach to this analysis. Under isolation from other SES indicators, this approach can identify specific socioeconomic categories more closely related to excess body weight.
Our approach for assessing change in OWOB or obesity at given wealth levels was different from other studies; we could apply the exact same definition of wealth in both years since the wealth items were identical in both surveys. For more time-distant surveys, definitions of some of the items and their meaning as indicators of wealth may change. Under such circumstances using quintiles or other distributional categorization would be preferable.
The present study is based on the 2006 and 2012 NHNS surveys, which are representative of the Mexican population. The most recent studies relating SES to obesity in the Mexican adult population were limited either to the 2006 or to the 2012 data [
35,
36], or focused solely on trends in women and educational categories [
20].
Limitations of our study should be noted. We did not include dietary intake, physical activity, smoking or parity in the analyses. Both dietary intake and physical activity have been recognized as mediator variables in previous literature [
10,
11,
31], and as such their omission would result in the estimation of total associations of SES with excess body weight. That is, the calculated associations may incorporate or absorb the pathway through these mediators.
Interactions for area of residence were significant only for the wealth-OWOB association among men. Graphical analysis showed that wealth ranges have lower levels concentrated in rural areas. This may be driving the interaction significant since area of residence groups do not sufficiently overlap at low wealth levels. On the other hand, standard errors were much larger for the rural area, which also complicates detection of actual changes between 2006 and 2012.
Failure to reject the null of no interactions from our joint tests for area of residence does not imply that such interactions do not exist. Sample sizes of the SES categories within rural areas were relatively small and therefore could result in less precise estimates. Furthermore, given the correlation between the included SES indicators, standard errors are expected to be greater than a situation in which the covariates are not correlated.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
ADQ conceived the study, analyzed data and participated in writing; ALL participated in writing and interpretation of data. Both authors read and approved the final manuscript.