Arsenic Exposure and Diabetes Mellitus Risk

To the Editor: Chiu et al1 have presented the annual (and 3-year moving average) diabetes mellitus (DM) standardized mortality ratios (SMRs) for 1971 to 2000 for males and females from the Blackfoot disease (BFD) endemic area of southwest Taiwan. They have examined the temporal pattern of the data following the transfer by 1975 of the population from an artesian well water supply with a median arsenic level of 780 μg/L (ppb) to a reservoir-sourced water system with an arsenic concentration of less than 10 μg/L. The authors concluded from their ecologic analysis that the DM mortality rate showed a significant and causal decline for females but not for males. Our reanalysis of their data suggests that this conclusion is based on an artifactual data analysis and a too-facile interpretation. In Figure 1, we present a graphing of their 3-year moving average SMRs with data for both males and females on the same graph unlike their presentation of separate graphs. Our examination of this figure suggests that the DM SMR pattern for males and females differ considerably during the 1970s but not during the 1980s and 1990s. In Figure 2, we have included linear regressions (or time-series regressions) for the 1980–2000 data points and find that the DM SMRs for the males and for the females each show a similar significant negative slope of approximately 3% per year. Thus, we dispute the conclusion that there is a gender difference in temporal pattern (at least during the 1980s and the 1990s).Fig. 1.: Time trends for Diabetes Mellitus SMRs (3-year moving averages) for Males (Blue Diamonds) and Females (Red Squares) in Blackfoot Disease area (from Chiu et al., 2006).Fig. 2.: Linear regression analyses for 1980–1999 data applied to Time trends for Diabetes Mellitus SMRs (3-year moving averages) for Males (Blue Diamonds) and Females (Red Squares) in Blackfoot Disease area (from Chiu et al., 2006).During the 1970s, there is a similar pattern also for males and females; however, the pattern for the females seems to have come 3 years earlier. During the 1970s, both males and females showed a sharp drop in the DM SMR down to near 50 with a sharp rise to approximately 150. The relationship of this pattern to the change in water supply is uncertain, because the change was only complete in 1975. Certification of DM as the underlying cause of death is quite susceptible to selection and reporting biases, and analyses based on small numbers should be cautious. Over the 30-year observation period, the number of DM deaths expected annually has increased 10-fold, which suggests marked changes in either demography or diagnostic choice. The DM mortality data, themselves, during the 1970s are sparse with approximately two DM deaths per year for males and fewer than six per year for females. In contrast, during the 1990s, there was an average of 17 male and 26 female DM deaths per year, providing much greater stability to the analysis. Even if the decline in DM mortality was causally associated with the change in water supply, it is unclear that the causal agent is the arsenic in the water. Historically, the most prominent health characteristic of the BFD endemic area is BFD, a condition closely associated with high arsenic level artesian well water use2 for which no new cases appeared after the cessation of use of the artesian wells.3 However, BFD never appeared in areas in northeast Taiwan that had similar arsenic levels to those in the BFD area.4 A number of papers have cited fluorescent compounds (humic substances) in the artesian well water as a potential causative agent for BFD,5,6 and township stratification of the BFD area cancer mortality data revealed the suggestion of a separate nonarsenic carcinogenic risk factor related to the drinking water.7 Thus, the decline in diabetes mortality may be the consequence of withdrawal of a water-borne factor, but whether arsenic is that factor is not clear. Whether arsenic exposure and DM mortality are associated is not yet clear; the literature cited by Chiu et al1 is incomplete. They have cited a possible association at a Swedish smelter, but not the negative findings from either a U.S. copper smelter (observed/expected = 54/65)8 or a U.K. tin smelter (observed/expected = 0/3.1).9 Pooled together, they yield an SMR = 0.79 (95% confidence interval = 0.60–1.03). Clearly, this is not indicative of an increased DM mortality risk with at least arsenic inhalation exposure. Both of these cohorts had sufficient arsenic exposure to demonstrate a significant increased mortality from lung cancer. Further work is necessary to delineate the causal factors for DM in the BFD area.10 Steven H. Lamm, MD Consultants in Epidemiology and Occupational Health, LLC. (SHL, AE)