The independent and joint associations of serum selenium and vitamin A (retinol) and E (alpha tocopherol) concentrations with the risk of death from cancer were studied in 51 case-control pairs--that is, 51 patients with cancer, each paired with a control matched for age, sex, and smoking. Case-control pairs came from a random sample of some 12000 people aged 30-64 years resident in two provinces of eastern Finland who were followed up for four years. Patients who died of cancer during the follow up period had a 12% lower mean serum selenium concentration (p = 0.015) than the controls. The difference persisted when deaths from cancer in the first follow up year were excluded. The adjusted risk of fatal cancer was 5.8-fold (95% confidence interval 1.2-29.0) among subjects in the lowest tertile of selenium concentrations compared with those with higher values. Subjects with both low selenium and low alpha tocopherol concentrations in serum had an 11.4-fold adjusted risk. Among smoking men with cancer serum retinol concentrations were 26% lower than in smoking controls (p = 0.002). These data suggest that dietary selenium deficiency is associated with an increased risk of fatal cancer, that low vitamin E intake may enhance this effect, and that decreased vitamin or provitamin A intake contributes to the risk of lung cancer among smoking men with a low selenium intake.