In 2008, Navas-Acien et al1 published a report suggesting that exposure to inorganic arsenic was associated with type 2 diabetes mellitus in a population with relatively low exposure. Although the report was based on only 93 cases and cross-sectional data, it appeared in a prominent journal and received much press. The biomarkers of exposure, while relatively sophisticated, were hobbled due to high detection limits for several arsenic species and by the fact they reflected only recent exposure. The data were from the 2003–2004 National Health and Nutrition Examination Survey (NHANES), meaning they were generalizable to the United States, and also easily accessible by others. In this issue of Epidemiology, Steinmaus et al2 present their analysis of the same data, in which they report no association. The Editors invited Navas-Acien et al to provide a defense of their previous work, which the authors did by augmenting their analysis with newer NHANES data.3 Interpretation of these conflicting results may affect the direction of future research. It's worth examining the technical issues behind the debate.
The difference between the 2 analyses hinges on how best to estimate exposure to inorganic arsenic, given the available biomarkers in urine (mainly total arsenic, dimethyl arsenic, and arsenobetaine). The 2 analytic approaches reflect different views about how ingested arsenic contributes to various arsenic species in urine. Steinmaus et al2 assume that urinary arsenite, arsenate, methylarsonate, and dimethylarsinate (DMA) originate primarily from ingested inorganic arsenic (see Fig. 1 in the response from Navas-Acien at al.3). Thus their estimate of inorganic arsenic exposure is total urine arsenic minus the arsenic metabolites that come from seafood (primarily arsenobetaine, and a minor constitutent, arsenocholine). Navas-Acien et al,4 however, note that some DMA comes from seafood. To assume that subtracting arsenobetaine (and arsenocholine) from total arsenic gives a good measure of inorganic arsenic ignores the contribution of seafood to DMA. On that basis, Navas-Acien et al's measure of inorganic arsenic as total arsenic with adjustment of results for markers of seafood intake has merit.
Nonetheless, the original report of Navas-Acien et al1 had weaknesses inviting criticism. Their original findings were heavily dependent on arsenobetaine having an inverse relation with odds of diabetes—a curious finding. Fish intake is not known to prevent diabetes.5,6 The inverse relation means the patients with diabetes were eating less fish. (They may also have changed their diet in other ways because of their diagnosis, but more on this point later.) In the results of Navas-Acien et al, an association of diabetes with total arsenic was not apparent until after they adjusted for arsenobetaine (and Hg). The median arsenobetaine levels in diabetes were about half those in nondiabetics (P = 0.03; Table 1 [in the article by Navas-Acien et al]).1 Statistical significance aside, the categorical model results in their Table 3 show that the strongest odds ratios were between diabetes and arsenobetaine rather than total arsenic. The correlation of ln(arsenobetaine) with ln(total arsenic), on a creatinine basis, was 0.46. All other aspects of diet aside, I suppose it's reasonable to ask what is the association of diabetes with total arsenic. But the particular constellation of cross-sectional data and relationships among key variables gives one the impression that the adjusted results ought to be viewed with skepticism. To defend against such skepticism, the authors showed that, after restricting the analysis to those with arsenobetaine levels below the median, the odds ratio between total arsenic (80th vs. 20th percentile) and diabetes was 2.4 (95% CI = 0.7–8.4). That wide confidence interval also provided shaky ground for their original premise.
Now that data from the 2005–2006 NHANES can be brought to bear, there is better statistical power to examine the association of inorganic arsenic with diabetes. In the response by Navas-Acien et al3 (Table 1), after restriction of the analysis to those with undetectable arsenobetaine, the odds ratio between total arsenic (80th vs. 20th percentile) and diabetes was 2.6 (95% CI = 1.1–6.0). While the confidence interval is still wide, the evidence for an association appears stronger.
As noted above, Steinmaus et al contended that in human urine, and base their analyses on this assumption. This assumption may be true, but a recent human feeding experiment4 (cited by Steinmaus et al), as well as other data,7,8 suggest that seafood intake can increase the amount of DMA in urine. Seafood can contain DMA or other compounds such as arsenosugars that are metabolized to DMA.4,9 The potential contribution of seafood to urinary DMA is apparent in those NHANES 2003–2004 subjects who ate seafood in the past 24 hours: the correlation of ln(arsenobetaine) and ln(DMA), on a creatinine basis, was 0.50. Arsenobetaine and DMA from seafood are present in urine more than 24 hours after ingestion.4 Thus, the strategy of excluding those eating seafood in the past 24 hours (used by both groups1,2) would have been only partly successful in eliminating seafood-derived DMA. As discussed above, arsenobetaine was associated with decreased odds of disease. Suppose that seafood-derived DMA was similarly associated. In that case, even if only a small amount of urinary DMA was from seafood, ignoring the seafood origin of this DMA could mask an association of diabetes with inorganic arsenic (estimated as total minus arsenobetaine). Estimating inorganic arsenic as total minus arsenobetaine might serve well in populations with high inorganic arsenic exposure and low seafood intake. But in the NHANES analyses discussed, more than 10% of subjects who were otherwise eligible were excluded because of fish or seafood consumption in the past 24 hours. With the increase in seafood in the US diet,10 and the many unknowns about the fate of some ingested organic arsenic species,4 the total-minus-arsenobetaine approach raises concerns, especially where seafood is a confounder.
NHANES data indicate that, compared with nondiabetics, the urine of patients with diabetes has less arsenobetaine1 and more bisphenol A,11 and their blood has higher levels of several persistent organic pollutants.12 Perhaps we have a lot to learn about how diabetes affects eating habits, and about the pharmacokinetics of contaminants. Could patients with diabetes be drinking more local water and fewer sugary beverages bottled with water relatively low in arsenic? Could creatinine correction of urine concentrations not work as hoped in patients with diabetes?13
In a 2006 review of arsenic and diabetes, Navas-Acien et al14 said of the epidemiologic data that “…methodologic problems limit the interpretation of the association…” and “…prospective epidemiologic studies measuring arsenic biomarkers and appropriately assessing diabetes should be a research priority.” The recent analyses of the NHANES data have not moved the field forward much. One hopes that future prospective studies, incorporating improved bioassays, can bring this question closer to resolution.
ABOUT THE AUTHOR
MATTHEW LONGNECKER is a Senior Investigator in the Epidemiology Branch, National Institute of Environmental Health Sciences, National Institutes of Health. He has studied associations between levels of environmental contaminants in biologic samples and health for over 15 years. His research interests include environmental risk factors for type 2 and gestational diabetes mellitus.
1. Navas-Acien A, Silbergeld EK, Pastor-Barriuso R, Guallar E. Arsenic exposure and prevalence of type 2 diabetes in US adults. JAMA
2. Steinmaus C, Yuan Y, Liaw J, Smith AH. Low-level population exposure to inorganic arsenic in the United States and diabetes mellitus: a re-analysis. Epidemiology
3. Navas-Acien A, Silbergeld EK, Pastor-Barriuso R, Guallar E. Arsenic exposure and prevalence of type 2 diabetes in US adults: updated findings from the National Health Nutrition and Examination Survey, 2003–2006. Epidemiology
4. Lai VW, Sun Y, Ting E, Cullen WR, Reimer KJ. Arsenic speciation in human urine: are we all the same? Toxicol Appl Pharm
5. Fedor D, Kelley DS. Prevention of insulin resistance by n-3 polyunsaturated fatty acids. Curr Opin Clin Nutr Metab Care
6. Schulze MB, Hu FB. Primary prevention of diabetes: what can be done and how much can be prevented? Annu Rev Public Health
7. Raml R, Goessler W, Traar P, Ochi T, Francesconi KA. Novel thioarsenic metabolites in human urine after ingestion of an arsenosugar, 2’,3’-dihydroxypropyl 5-deoxy-5-dimethylarsinoyl-beta-D-riboside. Chem Res Toxicol
8. Francesconi KA, Tanggaar R, McKenzie CJ, Goessler W. Arsenic metabolites in human urine after ingestion of an arsenosugar. Clin Chem
9. Devesa V, Súñer MA, Algora S, Vélez D, Montoro R, Jalón M, Urieta I, Macho ML. Organoarsenical species contents in cooked seafood. J Agric Food Chem
10. Committee on Nutrient Relationships in Seafood: Selections to Balance Benefits and Risks, Malden C. Nesheim and Ann L. Yaktine, eds. Seafood Choices: Balancing Benefits and Risks
. Washington, DC: National Academies Press; 2007.
11. Lang IA, Galloway TS, Scarlett A, et al. Association of urinary bisphenol A concentration with medical disorders and laboratory abnormalities in adults. JAMA
12. Everett CJ, Frithsen IL, Diaz VA, Koopman RJ, Simpson WM Jr, Mainous AG III. Association of a polychlorinated dibenzo-p-dioxin, a polychlorinated biphenyl, and DDT with diabetes in the 1999–2002 National Health and Nutrition Examination Survey. Environ Res
13. Barr DB, Wilder LC, Caudill SP, Gonzalez AJ, Needham LL, Pirkle JL. Urinary creatinine concentrations in the U.S. population: implications for urinary biologic monitoring measurements. Environ Health Perspect
14. Navas-Acien A, Silbergeld EK, Streeter RA, Clark JM, Burke TA, Guallar E. Arsenic exposure and type 2 diabetes: a systematic review of the experimental and epidemiological evidence. Environ Health Perspect