Objective. To examine the association between metformin use and cancer stage at diagnosis among elderly men with preexisting diabetes mellitus and incident prostate cancer. Methods. This study used a population-based observational cohort of elderly men (≥66 years) with preexisting diabetes and incident prostate cancer between 2008 and 2009 (N = 2,652). Cancer stage at diagnosis (localized versus advanced) was based on the American Joint Cancer Committee classification. Metformin use and other independent variables were measured during the one year before cancer diagnosis. Logistic regressions with inverse probability treatment weights were used to control for the observed selection bias. Results. A significantly lower percentage of metformin users were diagnosed with advanced prostate cancer as compared to nonusers (4.7% versus 6.7%, p < 0.03). After adjusting for the observed selection bias and other independent variables, metformin use was associated with a 32% reduction in the risk of advanced prostate cancer (adjusted odds ratio, AOR: 0.68, 95% confidence interval, CI: 0.48, 0.97). Conclusions. This is the first epidemiological study to support the role of metformin in reducing the risk of advanced prostate cancer. Randomized clinical trials are needed to confirm the causal link between metformin use and prostate cancer diagnosis stage.