Previous studies on metformin use and gastrointestinal (GI) cancer risk have yielded inconclusive results on metformin’s chemoprotective effects. We aimed to evaluate GI cancer risk in users of metformin in The Netherlands using a time-varying apprach in a large population-based database. A cohort study was performed using the NCR-PHARMO database. Patients using >/=1 non-insulin anti-diabetic drug (NIAD) during 1998-2011 were included (N=57,621). Exposure to NIADs was modelled time-varyingly. Coxregression analysis estimated hazard ratios (HRs) of GI cancers in current metformin users versus current users of other NIADs. Covariables included age, sex, drugs known to impact cancer risk, history of hospitalization, and starting year of follow-up.A sensitivity analysis was performed, applying a new-user design. Current use of metformin was not associated with a decreased risk of GI cancer (HR 0.97, 95% CI 0.82-1.15), or specific GI cancer sites. The sensitivity analysis yielded comparable results. No decreasing trends were observed with increasing cumulative dose of metformin (HR 1.05, 95% CI 0.85-1.28, HR 0.89, 95% CI 0.73-1.10, HR 0.96, 95% CI 0.77-1.19 for dose tertiles low [<405 g], medium [405-999 g], and high [>/=999 g]). In contrast, an increased risk of pancreatic cancer was found in current users of metformin plus insulin (HR 4.90, 95% CI 2.64-9.10). In conclusion, no decreased risk of GI cancer was found in current metformin users compared to current users of other NIADs. Variations in the exposure definition of metformin use may be one of the explanations of previously found reduced cancer risks in metformin users.