Objective: A retrospective cohort study, using a population-based reimbursement database, was conducted for investigating the relationship between diabetes and colon cancer and assessing whether metformin had a protective effect. Methods: Overall, 493 704 men and 502 139 women, covered by the National Health Insurance, without colon cancer were followed from 2003 to 2005. Cox regression evaluated the adjusted relative risk (RR), considering confounders and detection examinations. Results: Even though diabetes patients had a significantly higher probability of receiving examinations that could lead to the detection of colon cancer, they had a significantly higher risk (24%) of this cancer after adjustment. Metformin users had a significantly lower risk (27%) of colon cancer. While comparing patients with diabetes for !1, 1-3, and R3 years to nondiabetes individuals, the adjusted RR (95% confidence interval) was 1.308 (1.020-1.679), 1.087 (0.900-1.313), and 1.185 (1.055-1.330) respectively. The higher risk among those with diabetes for !1 year suggested a possible reverse causality or a link with prediabetes. However, diabetes still might play some role in the development of colon cancer in those with diabetes for R3 years. The duration of metformin use showed an inverse trend, with a significant RR of 0.643 (0.490-0.845) in users for R3 years, when compared with nonusers. In addition, metformin may reduce colon cancer risk associated with chronic obstructive pulmonary disease (a surrogate for smoking). Conclusions: Following adjustment for potential detection bias and other covariates, diabetes remains a significant risk factor for colon cancer. Metformin may protect against colon cancer.