The authors examined the hypothesis that relatively high levels of transferrin saturation increase the risk of cancer. They studied a cohort of prepaid health plan members whose transferrin saturation levels were measured during the period 1969-1971 and who were followed for cancer through 1990. After the exclusion of 10 percent of the subjects who received treatment for one or more of six chronic conditions or who were pregnant when the measurement was made and persons who contributed less than 5 years of follow-up, the authors were left with 38,538 persons who were followed for an average period of 17.7 years. In women, a positive association was observed between transferrin saturation and risk of stomach carcinoma (> or = 34.5% compared with < or = 20.3%: relative risk (RR) = 3.5, 95% confidence interval (CI) 0.98-12). In men, transferrin saturation was inversely associated with risk of colon and rectal carcinoma (> or = 40.7% compared with < or = 26.0%: colon, RR = 0.62, 95% CI 0.35-1.1; rectum, RR = 0.30, 95% CI 0.08-1.1) and with non-Hodgkin's lymphoma (32.1-40.6% compared with < or = 26.0%: RR = 0.31, 95% CI 0.11-0.88; no cases observed with transferrin saturation > or = 40.7%). The authors did not find evidence that the risk of epithelial cancer (all sites combined) was related to transferrin saturation level or to iron deficiency (< or = 15%) or overload (> or = 60%).