The association of serum selenium with the subsequent risk of death from cancer was investigated in a case-control study that was nested in a prospective nine-year follow-up study in the Netherlands. In 1975-1978, 10,532 persons in the Dutch town of Zoetermeer who were aged five years or more participated in a medical survey. Serum samples were collected and stored at -20 C. For the 82 persons who died of cancer since the baseline examination, 164 cohort members still alive by the end of 1983 were selected as controls and matched for age, sex, and smoking. Cancer deaths that occurred in the first year of follow-up were excluded, leaving 69 cases for statistical analyses. The mean serum selenium level of 116.7 +/- 4.0 micrograms/liter among male cancer deaths (n = 40) was significantly different (p = 0.04) from that in the control subjects (126.4 +/- 3.1 micrograms/liter). In females, selenium levels were similar among cases and controls. The adjusted risk of death from cancer for men in the lowest quintile of serum selenium (below 100.8 micrograms/liter) was more than twice that of subjects with higher levels (relative risk = 2.7,90% confidence interval = 1.2-6.2). These data support recent findings of an increased cancer risk associated with low serum selenium levels in men but not in women.