To examine an association between selenium level and gastric cancer (GC) risk, a hospital-based case-control study was conducted in Cali, Colombia. Selenium concentrations in toenails were compared between 142 GC patients and 244 controls selected from hospitalized non-cancer patients. GC risk was lowest in the lowest quartile of selenium level and highest in the second highest quartile (age-, sex-, hospital-, and sampling-season-adjusted odds ratio [OR]: 5.9, 95% confidence interval: 2.8, 12.4). This association was not modified by either tumor location or Lauren's histological type. The magnitude of ORs was not affected by other diets that were significantly associated with GC risk. Since selenium levels were relatively high in cases and in controls, our results indicate that an inverse association between selenium level and GC risk may exist only among populations with low selenium levels.