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 approach in a large population-based database. A cohort study was performed using the NCR-PHARMO database. Patients using ≥1 non-insulin antidiabetic drug (NIAD) during 1998 to 2011 were included ( = 57,621). Exposure to NIADs was modeled time-varyingly. Cox regression analysis estimated 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% confidence interval (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 with 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. .

<p>Supplementary Table 2. Hazard ratios of GI cancer subsites by sex</p>

<p>Supplementary Table 1. Hazard ratios of GI cancer overall by sex</p>

<p>Supplementary Table 1. Hazard ratios of GI cancer overall by sex</p>

<p>Supplementary Table 2. Hazard ratios of GI cancer subsites by sex</p>