The hypothesis that low selenium may in some circumstances be a risk factor for lung cancer was investigated in a case-control study nested within a longitudinal study. Serum samples from 9,101 cancer-free individuals were collected and stored at -20 degrees C by the Finnish Mobile Clinic in 1968-1971 and 1973-1976. During follow-up until the end of 1991, 95 cases of lung cancer were diagnosed. Selenium concentrations were determined from the serum samples of the cases and 190 controls, individually matched for sex, age, and place of residence. Mean levels of serum selenium in cases and controls were 53.2 microg/liter and 57.8 microg/liter, respectively. The relative risk of lung cancer between the highest and lowest tertiles of serum selenium, adjusted for smoking, serum alpha-tocopherol, serum cholesterol, serum copper, serum orosomucoid, and body mass index (kg/m2), was 0.41 (95% confidence interval (CI) 0.17-0.94). The association was stronger at lower levels (<5.9 mg/liter) of alpha-tocopherol (relative risk=0.24, 95% CI 0.07-0.85). The association was also pronounced among current smokers and at higher levels of serum orosomucoid and serum copper. The relative risk for smokers who were twice ranked in higher selenium tertiles, at an interval of 4-7 years, in comparison with smokers who remained in the lowest tertile was 0.16 (95% CI 0.04-0.74). In accordance with the hypothesis, the findings suggest that very low selenium status may contribute to the risk of lung cancer.