The arsenicosis-endemic areas of Zimapán and Lagunera were chosen for this study because tens of thousands of local residents are currently exposed to a wide range of iAs concentrations in drinking water and because relatively detailed information exists about historical levels of iAs in the local drinking water supplies. Unlike most previous studies, we used multiple biomarkers to classify diabetes and to characterize exposure to iAs. In spite of a relatively small number of subjects recruited for this study we were able to show that the exposure to iAs was positively associated with all three diabetes indicators. Our findings contradict results of a recent study that found no significant association of iAs exposure with diabetes in another arsenicosis area located in Bangladesh [30
]. This study used only glycated hemoglobin (HbA1c) and glucosuria to classify diabetic individuals. Although the HbA1c level in blood is now an accepted measure for diagnosis of diabetes in the US, it is unclear whether it was also validated for the Bangladeshi population. Glucosuria is not a reliable indicator of diabetes [31
]. Thus, using the validated diabetes indicators may be essential for linking iAs exposure to risk of diabetes. Notably, our data on the negative associations between iAs exposure and FPI and HOMA-IR suggest that the mechanisms of iAs-induced diabetes differ from those underlying type-2 diabetes, which is typically characterized by insulin resistance (i.e., increased HOMA-IR) and hyperinsulinemia [32
In our study, FPG and 2HBG were significantly associated with iAs concentration in drinking water, but not with tAs or sum of iAs metabolites in urine. Samples of water provided for iAs analysis represented the types and sources of drinking water that were used by study subjects on a daily basis for an extensive period of time. Thus, it is plausible that iAs concentrations in these samples are more representative of the current exposures to iAs than are tAs levels in spot urines, which could be affected by changes in drinking water sources or consumption during days or even hours before urine collection. Our results show that diabetes was not associated with cumulative exposure to iAs over the last fifteen years. Historical records show that levels of iAs in drinking water supplies in both Zimapán and Lagunera areas have changed in recent years as a result of interventions by local governments. These changes may explain the low correlation between estimated cumulative exposure and the current concentrations of iAs in drinking water. In addition, the use of bottled water for drinking and cooking has become common, particularly in Zimapán, which historically had the higher levels of iAs in the municipal water supplies.
Previous epidemiologic studies used MAs/iAs and DMAs/MAs ratios in urine to evaluate the efficiency of iAs methylation and detoxification by subjects exposed to environmental iAs. Some of these studies reported a negative association between iAs exposure and the DMAs/MAs ratio or %DMA in urine, suggesting saturation or inhibition of the methylation pathway for iAs at high exposure levels [33
]. Other reports have linked low DMAs/MAs ratio (or high MAs/DMAs ratio) and high %MAs (or low %DMAs) to an increased susceptibility to diseases associated with iAs exposure, including skin lesions [34
]. However, our data suggest that with increasing iAs exposure the efficiency of iAs methylation increases as indicated by higher DMAs/MAs and lower MAs/iAs ratios in urine.
An important goal of the present study was to examine associations between the prevalence of diabetes and urinary concentrations of MAsIII
. All urine samples collected in Lagunera and analyzed in UJED by the advanced HG-CT-AAS system contained both MAsIII
. The failure to detect DMAsIII
and, particularly MAsIII
in a large number of urines from Zimapán was likely due to a lower sensitivity of the conventional HG-CT-AAS system used in Cinvestav-IPN. Notably, DMAsIII
in urine was associated positively with FBG and 2HBG but negatively with FPI. These results suggest that individuals with high concentrations of DMAsIII
in urine are at increased risk of developing diabetes. DMAsIII
is a potent inhibitor of insulin-stimulated glucose uptake by adipocytes [8
also inhibits glucose-stimulated insulin secretion by isolated murine pancreatic islets (Styblo and Douillet, unpublished data). Thus, the mechanism of the diabetogenic effects of iAs exposure may involve inhibition by DMAsIII
of insulin-dependent glucose uptake and metabolism in peripheral tissues and/or inhibition by DMAsIII
of insulin production by pancreatic β-cells. Taken together these two mechanisms would produce fasting hyperglycemia and impaired glucose tolerance, as well as decreased FPI and HOMA-IR values, i.e., symptoms that are consistent with results of our study.
While this study had notable strengths, the interpretation of the results may be affected by several limitations. The study population was predominantly female because men in the area often emigrate to work. Although we adjusted for sex in the analysis, the ability to extend the findings to populations with a more balanced sex ratio may be limited to the extent that women and men may differ in susceptibility to As-related disease. As noted above, iAs exposures in the study area have changed in recent years as a result of modifications to municipal water systems and the use of bottled water. The negative correlation between current and cumulative exposures suggests that areas with historically high iAs levels were targeted for interventions to reduce exposure. Our data (not shown) suggest that individuals who had been told by a doctor that they had diabetes were not more likely to use bottled water, however. The recent changes in exposure add to the challenges of determining whether measurements in drinking water versus urine or current versus cumulative exposure are the best measure of biologically-relevant exposure to iAs. Although we were able to estimate exposure from 1993 onward, 64% of the subjects had lived in the study area before 1993 and had longer exposure histories than we could estimate. The sparse environmental data from the 1980s suggest that iAs levels in municipal water supplies may have been higher in that era than in later years. No measurements were available for years before 1978, but it is possible that iAs exposures were lower before the construction of municipal water systems served by deep wells began in the 1960s. In summary, there are plausible mechanisms for errors resulting in both under- and over-estimation of cumulative iAs exposure, but the magnitude and direction of bias in exposure-disease associations as a result of any such errors cannot be assessed without additional data.
The American Diabetes Association has recommended that FPG, OGTT or HbA1c be used to test for diabetes or to assess risk of future diabetes in asymptomatic patients [37
]. However, there is a potential for misclassification of diabetes status due to error in the field methods we used to measure these indicators. Any such error is likely to be random with respect to arsenic exposure, and so most likely to attenuate associations between diabetes status and exposure. We also lacked reliable indicators of social class and were consequently unable to adjust for it in the analysis. Both Zimapan and Lagunera are rural, low income areas of Mexico. Recent data suggest that type-2 diabetes is less prevalent among Mexican residents with low socioeconomic status [38
]. Nevertheless, the potential magnitude of confounding by social class is difficult to assess, because of the limited information about the extent to which it is a risk factor for diabetes in rural Mexican populations.
Finally, the cross-sectional design has well known limitations. In the present study, these relate primarily to the inability to measure the changes in exposure discussed above. Because we did not know when the onset of diabetes occurred, cumulative exposures may also have been overestimated for subjects who had the disease for a number of years. Such overestimation would tend to understate the association of exposure with disease. Cross-sectional studies can be susceptible to selection bias if exposed individuals withdraw from the population because of exposure or its early effects. However, exposure to iAs in drinking water was not recognized as a health concern until recently, so it is unlikely to have prompted more highly exposed individuals to move out of the study area. In addition, individuals with prevalent and new-onset diabetes may differ in the way they metabolize arsenic.