BACKGROUND: Whether metformin use may affect the risk of oral cancer required further investigation. METHODS: The reimbursement database of the National Health Insurance in Taiwan was used. Patients with type 2 diabetes mellitus at an onset age of 25-74 years during 1999-2005 and newly treated with either metformin (n = 288198, "ever users of metformin") or other antidiabetic drugs (n = 16263, "never users of metformin") were followed for at least 6 months for oral cancer until December 31, 2011. The treatment effect of metformin (for ever versus never users, and for tertiles of cumulative duration of therapy) was estimated by Cox regression adjusted for propensity score (PS) or incorporated with the inverse probability of treatment weighting (IPTW) using PS. RESULTS: The respective numbers of incident oral cancer in ever users and never users were 1273 (0.44%) and 119 (0.73%), with respective incidences of 92.7 and 163.6 per 100,000 person-years. The overall hazard ratios (95% confidence intervals) suggested a significantly lower risk [0.584 (0.483-0.707) for PS-adjusted model, and 0.562 (0.465-0.678) for IPTW model]. In tertile analyses, the PS-adjusted hazard ratios (95% confidence intervals) for the first (<21.5 months), second (21.5-45.9 months) and third (>45.9 months) tertile of cumulative duration were 1.403 (1.152-1.708), 0.557 (0.453-0.684) and 0.152 (0.119-0.194), respectively; and were 1.244 (1.024-1.511), 0.526 (0.429-0.645) and 0.138 (0.108-0.176), respectively, for IPTW. CONCLUSIONS: Metformin may significantly reduce the risk of oral cancer, especially when the cumulative duration is more than 21.5 months.
