Plasma selenium and glutathione peroxidase in erythrocytes were analyzed in a case-control study encompassing 441 cases with breast cancer and 191 controls with benign breast disease. No difference in mean serum selenium level between cases and controls on supplementary selenium intake was seen. If only individuals without supplementary intake, 278 cases and 135 controls, were considered a preventive effect was found increasing with selenium level. This finding was significant among women 50 years old or more with Mantel-Haenszel odds ratio = 0.16 for individuals with serum selenium > 1.21 mumol/L. Also for subjects with serum selenium in the range 1.00-1.21 mumol/L a significant preventive effect was seen with odds ratio = 0.38. For women under 50 years of age a nonsignificant preventive effect was seen. Glutathione peroxidase in erythrocytes did not correlate well with serum selenium and was not a marker for the risk of breast cancer.