The association between serum selenium concentration and the risk of cancer was studied in 1110 men aged 55 to 74 years in two rural areas of Finland. The men were followed-up prospectively for 9 years and there were 109 new cases of cancer, with the cases of the first follow-up year excluded. The serum selenium concentrations were adjusted for age, area, smoking, serum cholesterol, and alcohol intake. The patients had a slightly lower adjusted mean serum selenium than the subjects without cancer at the end of the follow-up (+/- standard error of mean) 53.9 +/- 1.5 and 55.3 +/- 0.5 micrograms/l, respectively. The relative risks of cancer were essentially the same when these were calculated in the tertiles of the serum selenium distribution. Thirty-seven men had a history of cancer at baseline or had cancer diagnosed during the first follow-up year and their adjusted mean serum selenium was 49.4 +/- 2.6 micrograms/l, which was significantly lower (P less than 0.05) than that of the subjects without cancer during the follow-up.