Mexican Americans are at higher risk for obesity and diabetes than the general US population (1-5). The economic and social ramifications of these epidemics are substantial (6,7). In 2006, more than 20 million Americans were estimated to have type 2 diabetes, and nearly one-third of the cases were undiagnosed (8). By 2050, the number of US patients with diagnosed diabetes is projected to rise to 39 million (9). Race and ethnicity, encompassing genetics, culture, economics, and elements of the environment, are underlying risk factors for diabetes (9). In the United States, racial/ethnic minority populations, including Mexican Americans, have a higher risk of diabetes, but this tendency is confounded by socioeconomic status (SES) (10,11). Similarly, even though health in the United States, including diabetes control, improved in general from 1999 through 2006, this did not hold true in socioeconomically disadvantaged racial/ethnic minority populations, particularly before the age of eligibility for Medicare (≥65 years) (10,12).

Racial disparities measured in years of potential life lost from major diseases, including diabetes, are well documented (13). The influx of immigrants from Mexico has been in the forefront of a national debate over the past decade or more. What is often lost in this debate is the health status of Mexican Americans in the United States, and particularly on the United States-Mexico border where they are often the dominant ethnic group. Brownsville (Cameron County, Texas; 2006 population: 172,437 [14]) provides a unique environment to examine health disparities in Mexican Americans in the United States. This community provides an advance view of health outcomes due to demographic changes that are likely to spread to many parts of the country. In 2004, we began to recruit a cohort, now numbering 2,000, drawn from randomly selected households on the basis of 2000 Census tract data (15), with the purpose of establishing a "Framingham-like" cohort in this population. The Framingham cohort originated in 1948 as the Framingham Heart Study, consisting of a random sample of 5,209 adults from Framingham, Massachusetts, aged 30 to 62 years (16). This cohort has grown to serve numerous studies and supply risk factor data for common diseases. The primary objective of our study was to see what effect minor economic advantages might have on the prevalence of obesity and diabetes in a Mexican American population.

Participation in the cohort in Brownsville did not significantly differ by SES stratum (Figure 1 and Table 1). The only substantial difference between the 2 socioeconomic strata was the proportion of participants who completed high school. Less than a fourth of the participants reported having private health insurance. A few (5%) were on Medicaid and a few others (5%) reported having Medicaid or Medicare combined with private managed care.

More than half of the participants in both SES strata had BMI scores in the obese range (≥30 kg/m²) (Table 2). Slightly more participants in the higher SES stratum were morbidly obese (BMI ≥40 kg/m²), but the difference was not significant. Self-reported diabetes differed only slightly between strata, but we found significant differences between strata for undiagnosed diabetes. Nearly 1 in 10 of all participants in the lower SES stratum were informed through this study that they have diabetes. In both strata, mean HbA1c levels among participants with diabetes were 9.4%. In logistic regression models allowing for potential confounding variables, undiagnosed diabetes continued to be the only significant outcome associated with socioeconomic strata (Table 2).

The only significant difference in biological measurements by SES strata when using weighted data was for fasting blood insulin (Table 3). Results stratified by sex show significant differences between SES strata in women for fasting insulin and insulin resistance. Still, mean insulin resistance for both sexes was well above the upper limit of the normal range (3.15).

When we ranked these participants by annual household income, we found the mean household income of the fourth quartile to be $20,532 and of the first quartile to be $8,160. When we compared these quartiles, we found no differences in frequency of obesity or in mean BMI. However, consistent with what we observed using census tract quartiles, we found a significantly higher mean fasting blood glucose of 123.2 mg/dL (95% CI, 114.5-131.9 mg/dL; n = 204) in the first quartile compared with 108.0 mg/dL (95% CI, 101.3-114.5 mg/dL; n = 202) in the fourth quartile. We also observed a marked difference in prevalence of diabetes (diagnosed and undiagnosed), 21% (43/204) in the first quartile compared with 10% (20/202) in the fourth quartile (χ² = 6.596, P < .01). We found no difference in mean insulin or insulin resistance levels between the 2 quartiles.

Age had a marked influence on the effect of SES. Rates of diabetes were significantly higher for people aged 55 to 64 years in the lower SES stratum compared with those in the higher SES stratum (Figure 2). There was a decrease in self-reported diabetes and an increase in undiagnosed diabetes, which was also age-dependent.

In a largely uninsured racial/ethnic minority population on the urban United States-Mexico border, SES was associated with undiagnosed diabetes and development of diabetes. Despite the narrow range of income in this population, people in the lower SES stratum were significantly more likely to have undiagnosed diabetes and the risk of diabetes increased markedly with age. This adverse effect of lower SES stratum on health became starker when we compared the top and bottom quartiles of self-reported household incomes. However, we found no differences between strata in prevalence of obesity.

We found that women in the lower SES stratum were more likely to have elevated insulin resistance and fasting insulin levels compared with women in the higher SES stratum, but we did not observe similar differences for men, consistent with published observations (21,22). Mexican American women in the lower SES stratum may have less access to health care than do their counterparts in the higher SES stratum, or perhaps lifestyle differences between strata contribute to how women care for their own health.

National data may underestimate the prevalence of obesity and diabetes in border communities (1). Rates of diabetes in these communities were 1.4 times higher than those reported nationally for Mexican Americans aged 20 years or older (1) and nearly twice as high as in all Americans (aged 40-59 years) (1,3,23). Undiagnosed diabetes was also twice as high in our border population as nationally. Similarly, our rates of obesity were more than one and a half times the rate in Mexican Americans nationally (aged 40-59 years), and one-tenth of our women were morbidly obese (3). These excess rates offer a glimpse of the potential future health of people of Mexican origin residing in other parts of the United States.

Our study sample is not yet eligible for Medicare (age ≥65 years). In both SES strata, nearly 80% of participants reported that they had no medical coverage. These data highlight the need for prevention and intervention aimed at people aged 35 or older but younger than 65, who are most at risk and least likely to be able to afford health care. This population is at particular risk for expensive complications such as renal failure and amputations. Complications of diabetes reduce ability to work, exacerbating the socioeconomic vulnerability of this age group.

There were several limitations to our study. Despite intensive efforts, more women than men participated in the study; this was not unexpected in a community where most men are on hourly wages. We could not assess those randomly selected people who chose not to participate. Exclusion of the second and fourth SES strata weakens our estimates of prevalence. Our data raise issues of acculturation and immigration experience, but we did not collect data to distinguish those factors and SES. Though the population is largely Mexican in origin, participants in this study had lived in Brownsville an average of 16 years, making them long-term residents of the United States. Nevertheless, more than 70% elected to answer the questionnaires in Spanish. Residents in Brownsville access Mexican culture freely through media outlets and imported products in local retail and grocery stores; those who can also move freely across the border. In many ways, Brownsville more closely resembles its Mexican neighboring city, Matamoros, than its closest major US city, Corpus Christi. The process of acculturation to Brownsville may differ from that observed in the general US population. We plan to gather precise acculturation data in our future studies with this cohort to determine patterns in this community.

The strengths of our study are considerable in that we have recruited the first exclusively Mexican American cohort in a border city with extreme health disparities. With the support of our clinical research unit funded by the National Institutes of Health, we have been able to recruit and process participants in a standardized fashion, by using staff expert in good clinical practices and by using well-monitored data management and analysis processes. This enables us to report reliable data that can be used to plan for appropriate prevention and equitable health care for an underserved population.

The overall picture of health in this Mexican American population is poor. The rate of diabetes in Hispanics/Latinos is nearly twice that in whites, occurring at an earlier age and with higher rates of complications and death (2). In Mexican Americans on the United States-Mexico border, published rates of diabetes in rural communities are more than twice those in non-Hispanic whites (24). We show the same elevated risk in urban border populations. Furthermore, we demonstrate high rates of diabetes, particularly undiagnosed diabetes, in the poorest of our poor who have little or no access to health care through insurance. This presents a major economic burden on the community and on the health care system because it delays diagnosis and treatment until the disease is far more complex and costly to address. Intervention in the form of community-based participatory research will likely be essential to achieve sustainable solutions (25). A focus on early diagnosis and prevention of diabetes will be key to success in border health.