Environmental selenium and cancer: risk or protection?

One of the joys of epidemiology is the discovery of a new association that has not been previously reported in the literature. In this issue of Cancer Epidemiology, Biomarkers and Prevention (1), Vinceti et a!. purport to have observed just such an association between malignant melanoma and exposure to selenium in drinking water from a municipal water supply in Reggio Emilia, a municipality in northern Italy. Unfortunately, the authors ignore the basic issue of the biological plausibility of the association, and either omit or misinterpret a number of articles to support their hypothesis. Thus, this study must be interpreted with extreme caution and considered only a chance association, unless additional information is uncovered that would change our basic understanding of the nutritional essentiality of selenium. The stated motivation for conducting the study was “a moderate imprecisely measured increased risk of melanoma” in the Nurses Health Cohort and “an increase in mortality from cancer of certain sites, including melanoma” in Reggio Emilia. The 63 nurses in the Nurses Health Cohort with a diagnosis of malignant melanoma had a nonsignificant (P = 0.10) 5% higher toenail selenium level than controls (0.886 versus 0.844 i.tg/g; Ref. 2). Our previous review of this study identified a number of methodological issues limiting its interpretability; in addition, no adjustment was made for the multiplicity of end points tested, which affects the interpretation of any site-specific P values (3). Regarding the stated excess mortality from melanoma in Reggio Emilia, surprisingly the excess is limited to males in the referenced abstract, and consists of only one case in an earlier publication of cancer mortality rates in Reggio Emilia (4). Furthermore, the authors’ literature review overlooks an earlier case-control study of selenium and melanoma in Germany, a region with a selenium status similar to that of Italy (5). This study with 101 cases observed a significant inverse association with serum selenium levels. The association was the strongest with low selenium levels for lentigo maligna melanoma and superficial spreading melanoma, tumors with typically small initial tumor mass. Although we have previously discussed the limitations of case-control studies of selenium and cancer (6), these issues are less applicable to malig-