Based on previous reports, it is uncertain whether serum cholesterol levels are inversely related to colon cancer risk. In this study, serum cholesterol levels were measured in 7926 Japanese-American men who were followed for over 20 years. Two hundred thirty-one incident cases of colon cancer and 97 cases of rectal cancer were identified. An increase in serum cholesterol levels was associated with a decrease in risk for colon cancer (P value for trend = .01) but not for rectal cancer. This association appeared stronger as the site of cancer moved from the sigmoid colon to the cecum. The data were further analyzed by interval from examination to diagnosis. The inverse association was present for colon cancer cases diagnosed within 10 years of examination (P value for trend less than .01), especially for cecum-ascending colon cancer cases (P less than .01). A similar inverse pattern was found for cecum-ascending colon cancer cases diagnosed after 10 years, but the association was not statistically significant. The results suggest that the preclinical effects of undiagnosed colon cancer contributed to the inverse association, but these effects do not entirely explain why the relationship with hypocholesterolemia was stronger in men who were subsequently diagnosed with right-sided colon cancer.