In a case-referent study on the possible role of selenium in human mammary carcinogenesis, serum selenium was found to be 79 +/- 12 micrograms/l in 66 cases and 81 +/- 12 micrograms/l in 93 referents. An internal trend in serum selenium was observed among cases (TNM stage I 81 +/- 11 micrograms/l and TNM stage II 76 +/- 13 micrograms selenium/l), indicating disease-mediated changes. The evaluation of selenium as a risk indicator in human breast cancer was therefore restricted to TNM stage I patients (n = 36). Multiple logistic regression analyses including variables associated with selenium levels revealed no association between selenium levels and breast cancer risk.