2283-PUB: Metformin Treatment and Prostate Cancer Risk: Cox Regression Analysis with Time-Dependent Covariates of 150,000 Men with Incident Diabetes Mellitus

Abstract

Background: There is conflicting evidence regarding the association between metformin use and prostate cancer risk in diabetic men. Objective: We investigated the association of metformin treatment, as a time dependent exposure, with prostate cancer incidence, in a population-based cohort of T2D incident men, while accounting for major time related biases, time-dependent confounding by glucose levels, and for use of other diabetes medications. Methods: During 2002-2012, we followed a cohort of 145,617 incident diabetic men aged 21-87 years insured in the largest health maintenance organization in Israel. We used a discrete form of the weighted cumulative exposure to metformin to evaluate its association with prostate cancer incidence. This was implemented in a time-dependent covariate Cox model, adjusting for age, ethnic background, and socioeconomic status. Other diabetes medications were not found to be strongly associated with prostate cancer. To account for time- dependent confounding by glucose levels, we used Propensity Score weighting for receiving metformin at each time-point. Results: There were 1,592 prostate cancer events occurring over a total of 910,000 person-years. We found a positive association of prostate cancer with metformin exposure in the previous year (hazard ratio (HR) for 1 defined daily dose = 1.66; 95%CI 1.23, 2.24), but a negative association with metformin exposure in the previous 2-7 years (HR=0.57; 95%CI 0.32, 1.02). Conclusion: Our results do not support a clear causal effect of metformin treatment on the incidence of prostate cancer. The positive association with metformin given in the previous year could be due to reverse causation of the cancer raising glucose levels and causing increased prescription of metformin. The negative association with metformin given 2-7 years previously is of borderline statistical significance and requires further research. Disclosure R. Dankner: None. H. Murad: None. N. Agay: None. L. Olmer: None. L.S. Freedman: None.