Metformin use and the risk of incident breast cancer: a systematic review and meta-analysis accounting for time-related biases, and why imeglimin cannot yet be evaluated

Background: Observational studies have suggested that metformin lowers breast cancer risk, but many were susceptible to immortal time and time-window bias. Imeglimin, a structurally related glucose-lowering drug approved in Japan in 2021, has been proposed as a candidate for cancer prevention. Methods: We searched PubMed (1 November 2026) and the reference lists of earlier meta-analyses for cohort and case-control studies that reported breast cancer-specific estimates for metformin use and incident breast cancer. One report per source population was pooled with a DerSimonian-Laird random-effects model. Studies were classified by whether they addressed time-related biases, using a new-user design, time-varying exposure or target trial emulation. For imeglimin, we applied the same criteria in a systematic search of PubMed and ClinicalTrials.gov, assessed whether a meta-analysis was feasible, and reviewed non-clinical and structural data. Results: The search identified 368 PubMed records and 28 reports from other sources. Twenty-two reports from 15 source populations were eligible, and 12 independent studies were pooled. The pooled relative risk (RR) was 0.89 (95% CI 0.74-1.07; I-squared = 90.5%). In nine studies that addressed time-related biases, the RR was 1.00 (0.94-1.08; I-squared = 0%). In three that did not, or were unclear, it was 0.69 (0.56-0.84); the between-group difference was Q = 12.78, p = 0.0004. The difference was not significant when the one unclear study was reclassified (Q = 1.15, p = 0.28). For imeglimin, the search returned 218 PubMed records and 10 registered studies, but no clinical study reported breast cancer or any cancer outcome. No meta-analysis was possible. The available evidence consisted of one in vitro study in breast cancer cells, one docking study, and models of other cancers at concentrations at least about 40-fold above clinical plasma levels. A comparative cohort would need about 631 breast cancer events to detect a hazard ratio of 0.80. Conclusions: In studies that addressed time-related biases, metformin use was not associated with incident breast cancer. A meta-analysis of imeglimin and breast cancer is not yet possible, because no clinical study exists and exposure time since approval is short. Its structural relation to metformin does not by itself predict a preventive effect. The next step should be an adequately powered new-user, active-comparator cohort study. Registration: PROSPERO CRD420261301586 (submitted 6 February 2026; registered 1 November 2026). Deviations from the protocol are reported in the Methods.