Effectiveness of Metformin in Preventing Colorectal Cancer Among Japanese Patients With Type 2 Diabetes: A Target Trial Emulation.
Observational studies have repeatedly reported lower cancer incidence among metformin users than nonusers, but findings are inconsistent and often affected by issues such as immortal time, unclear comparators, and use of total cancer as a composite outcome, which precludes adequate adjustment for site-specific confounding. These limitations can be mitigated by explicitly emulating a target trial with a suitable active comparator.
We emulated a target trial using a Japanese claims database (April 2014-March 2024) to assess whether metformin reduces colorectal cancer (CRC) risk compared with dipeptidyl peptidase-4 inhibitors (DPP-4is) in patients with type 2 diabetes. DPP-4is served as an active comparator because they are widely used as first-line alternatives in Japan and have no clear evidence of affecting CRC risk. We estimated the observational analogue of the per-protocol effect using pooled logistic regression with inverse probability weighting to adjust for baseline and time-varying confounders.
Among 26 273 metformin users and 108 299 DPP-4i users, the 5-year risk of CRC was 1.55% (95% confidence interval, 1.16 to 2.04) versus 1.26% (1.13 to 1.41), yielding a risk difference of 0.29% (-0.12 to 0.79) and a risk ratio of 1.23 (0.91 to 1.66).
Metformin use did not reduce the 5-year risk of CRC compared with DPP-4is. This finding is consistent with meta-analyses of randomized trials and contrasts with earlier observational reports of substantial benefit, which were likely inflated by methodological flaws. Explicitly emulating a target trial minimized design-related biases and provided estimates that support causal interpretation.
We emulated a target trial using a Japanese claims database (April 2014-March 2024) to assess whether metformin reduces colorectal cancer (CRC) risk compared with dipeptidyl peptidase-4 inhibitors (DPP-4is) in patients with type 2 diabetes. DPP-4is served as an active comparator because they are widely used as first-line alternatives in Japan and have no clear evidence of affecting CRC risk. We estimated the observational analogue of the per-protocol effect using pooled logistic regression with inverse probability weighting to adjust for baseline and time-varying confounders.
Among 26 273 metformin users and 108 299 DPP-4i users, the 5-year risk of CRC was 1.55% (95% confidence interval, 1.16 to 2.04) versus 1.26% (1.13 to 1.41), yielding a risk difference of 0.29% (-0.12 to 0.79) and a risk ratio of 1.23 (0.91 to 1.66).
Metformin use did not reduce the 5-year risk of CRC compared with DPP-4is. This finding is consistent with meta-analyses of randomized trials and contrasts with earlier observational reports of substantial benefit, which were likely inflated by methodological flaws. Explicitly emulating a target trial minimized design-related biases and provided estimates that support causal interpretation.