While numerous laboratory investigations have shown that selenium may have anticarcinogenic activity, the epidemiological data have been inconsistent. In this report, meta-analysis was used to quantitatively summarize the existing epidemiological evidence on selenium and lung cancer and identify sources of heterogeneity among studies. When all studies were combined, the summary relative risk (RR) for subjects with higher selenium exposures was 0.74 [95% confidence interval (CI) 0.57-0.97]. In subgroup analyses based on the average selenium level in the study population, the summary RR for areas where selenium levels were low was 0.72 (95% CI 0.45-1.16), while the RR for areas where selenium levels were higher was 0.86 (95% CI 0.61-1.22). In both studies in high selenium areas where RRs were markedly below 1.0, protective effects were only found when subjects in the lowest category of selenium exposure were used as referents. No clear protective effects were seen when highly exposed subjects were compared with those in the middle exposure categories. The summary RR was lower in studies assessing selenium exposure using toenails (RR 0.46, 95% CI 0.24-0.87) than in studies using serum selenium (RR 0.80, 95% CI 0.58-1.10) or studies assessing dietary intake (RR 1.00, 95% CI 0.77-1.30). Overall, these results suggest that selenium may have some protective effect against lung cancer in populations where average selenium levels are low. The evidence for these findings is greater in studies of toenail selenium than in studies involving other measures of exposure.