gms | German Medical Science

Background: Metformin (MET) has shown protective effects on carcinogenesis and cancer proliferation in preclinical studies but observational studies assessing the impact of MET use on cancer risk reported conflicting findings. This matched cohort study assessed how incident use of metformin impacts on cancer risk while attempting to address the methodological shortcomings of previous studies.

Materials and Methods: Data from a disease management programme (DMP) for patients with type 2 diabetes mellitus was linked to a population-based cancer registry. The cohort was assembled from treatment-naive patients who were enrolled between 2003 and 2009 and who had no prior cancer diagnosis. MET initiators were matched to a comparator group of up to three untreated patients based on time since DMP enrolment, diabetes duration at baseline, sex, age and BMI categories. Remaining imbalances were redressed by including propensity score weights in the estimation of hazard ratios (HR) for any cancer and the most frequent cancer entities in a Cox proportional hazards model. Follow-up time was analyzed in an intention-to-treat (ITT) approach, and an as treated (AT) approach, censoring for a new antidiabetic therapy start or therapy discontinuation. Time-dependent effects on cancer risk were evaluated by conducting sensitivity analyses around induction and latency time periods and modelling hazard ratios over time by using restricted cubic (natural) splines.

Results: In ITT-analysis, there were 1,287 first invasive cancer cases identified over 64,648 person-years of follow-up. Initiating MET was related to a significant risk reduction for all cancer sites (HR = 0.85, 95% CI [0.75-0.97]). Time-dependent analyses revealed, however, that cancer risk was reduced in particular in the first months after starting on MET and increased to comparator levels again thereafter. Cancer-specific analyses produced similar findings but were all statistically non-significant. AT analyses (862 cases, 40,347 person-years) essentially confirmed these results with even more pronounced time effects.

Conclusion: This study found indication of a time-dependent course of cancer risks after metformin initiation, which might help to explain previous contradictory study results in this field. Limited sample size precluded detailed analyses for single cancer entities.