Arsenic Ingestion and Bladder Cancer Mortality—What Do the Dose-Response Relationships Suggest About Mechanism?

ABSTRACT

The Black–Foot Disease (BFD) endemic area of SW Taiwan has historically been the principal data source for assessing cancer risks from arsenic in drinking water in the United States, most recently in a 42–village ecological study. The data showed a discontinuity for bladder cancer risk at about 400 μg/L. A proposed explanation was that the arsenic–dependent bladder cancer risk was found only for those villages that were dependent on water from the artesian well aquifer (As > 350 μg/L and co–contamination with humic acids) and not for those villages receiving water from the shallow aquifer (As < 350 μg/L without humic acids). The humic acids were present from the algae that grew in the uncovered tanks holding the artesian water. The risk factors (slopes) developed from these subpopulations of the SW Taiwan study were applied to the data from an ecological study of median groundwater arsenic concentration and bladder cancer mortality in 133 U.S. counties dependent on groundwater to determine the slope most predictive of U.S. experience for bladder cancer mortality and arsenic ingestion (Lamm et al. 2004).The U.S. data excluded the SW Taiwan slope estimate derived from the artesian well–dependent subpopulation but were consistent with the slope estimate derived from the subpopulation using shallow aquifer water. Both the SW Taiwan data in the absence of high arsenic levels (< 350 μg/L) and humic acids and the U.S. 133–county data with As < 60 μg/L are consistent with no increased bladder cancer mortality risk from drinking water arsenic concentrations in the exposure range of observation. These analytic results are consistent with both co–carcinogenesis and high–exposure (hundreds of μ g/L As) dependence models of toxicological mode–of–action. These dose–response relationships should influence prioritization in the remediation of arsenic–contaminated drinking water supplies.

KeyWords:

• arsenic
• SW Taiwan
• bladder cancer
• humic substances
• cocarcinogenesis.

ACKNOWLEDGMENTS

An earlier version of this article was presented at the XIX Annual International Conference on Soils, Sediments, and Water held at the University of Massachusetts, Amherst, MA on October 20–24, 2003, and partial support was provided by the Wood Preservatives Science Council. The authors thank Drs. Arnold Engel, Daniel M. Byrd, and Timothy J. Jorgensen for their critical comments in the development of this article.

Notes

SMR = observed mortality rate/expected mortality rate.

To make this estimate of the start point, we found the X–intercept of the straight line that had the best fit with the data. Byrd et al. (1996) gave a more sophisticated analysis that used a genuine threshold model and found a start point of 103 μg/L.

The underlying data had been provided to us both by Dr. Louise Ryan of Harvard University and Dr. Andrew Schulman of the U.S. Environmental Protection Agency. We thank them both.

Chen et al. (1962) originally reported the ranges of arsenic concentrations as 0–150 μg/L for 14 shallow wells in the BFD–endemic region and 350–1,110 μg/L for 34 artesian wells in that region. (Note that these wells were not necessarily used by those included in the Wu et al. database.) Because this report is cited by Chen et al. (1985, 1986) and Wu et al. (1989), it is unclear why these more recent studies claim that the range for shallow wells is 0–300 μg/L instead of 0–150 μg/L. To account for possible misclassification due to this difference, we repeated our analyses using values of 150 and 250 μg/L when defining artesian dependent and non–artesian dependent villages. These different values did not appreciably affect the results.

Similar results are found using the alternative criteria described in footnote 4. Using the concentration of 150 μg/L to define artesian wells implies that there are 20 artesian well dependent villages, and that the slope estimate is +0.045 per μg/L (95% CI +0.009; +0.081). For the 22 non–artesian dependent villages, the slope estimate is +0.011 per μg/L (95% CI −0.026; +0.049). Using the concentration of 250 μg/L to define artesian wells implies that there are 19 artesian well–dependent villages for which the slope estimate is +0.054 per μg/L (95% CI +0.020; +0.087); for the 23 non–artesian dependent villages, the slope estimate is +0.001 per μg/L (95% CI −0.028; +0.029).

No confidence interval could be calculated for this slope estimate because the line that fit the data best had a negative slope and so (because of the non–negative constraint on the model) was forced to 0.

The estimated slope of the arsenic–SMR curve for male bladder cancer in all 42 villages is 0.032 per μg/L (95% CI, +0.017; +0.048). The estimated slope in the 14 artesian well–dependent villages is +0.074 per μg/L (95% CI +0.016; +0.133) and the estimated slope in the other 28 villages is +0.005 per μg/L (95% CI −0.020; +0.030). The predicted relationships were obtained by plotting the line = 0.967 + (Estimated Slope)× (1/3) × (As Median). The factor of 1/3 comes from the conversion factor used by EPA and the National Research Council to convert the daily mg As/kg body weight exposure in Taiwan to a comparable exposure for U.S. residents, and depends on assumptions about differences in average body weight and daily water consumption.