Abstract Background: Epidemiologic studies have suggested that metformin use is associated with cancer death reduction. The results for pancreatic ductal adenocarcinoma (PDAC) are inconsistent. Retrospective studies of cancer survival may potentially have unintended biases; due to complexities of diabetes management, metformin use exposure is commonly categorized as simply “yes/no.” Nevertheless, in part based on epidemiologic studies, there are at least 20 open cancer clinical trials that include metformin in a treatment arm. We aimed to study metformin use and survival of patients with PDAC, addressing potential inherent biases. Methods: A retrospective cohort study for the years 2000-2011 was conducted using the patient database of the Mayo Clinic Specialized Programs of Research Excellence (SPORE) in Pancreatic Cancer. 1,360 PDAC patients (59% male; mean age 67) with diabetes were studied; overall survival was the primary outcome. Initiation and duration of metformin use was carefully abstracted from medical records. Patients were categorized into 5 groups based on metformin initiation: (A) Never used, n = 908 (reference group); (B) Metformin started >1 year prior to PDAC diagnosis, n = 84; (C) Metformin started within 1 year prior to PDAC diagnosis, n = 212; (D) Metformin started <30 days post-PDAC diagnosis, n = 104; (E) Metformin started >30 days post PDAC diagnosis, n = 34. Hazard ratios (HR) and 95% confidence intervals (C.I.) for survival differences between groups were computed using Cox proportional hazards models, adjusting for age, sex, disease stage, body mass index, and diagnosis year group. Results: Median survival of patients in Groups B-E was 292 days compared to 308 for Group A; this was not significantly different. However, among 413 resectable patients, metformin users appeared to survive longer than non-users (782 vs. 612 days, p = 0.07). Survival in each group was 308, 245, 249, 243 and 818 days, respectively. Because Group E had already survived >30 days, a bias existed. With Group E excluded, there was no survival difference among the other 4 groups (p = 0.18). Compared to Group A, HR and 95% C.I. for metformin use were 1.08 (0.85-1.37), 0.99 (0.85-1.17), 1.04 (0.83-1.31) and 0.49 (0.33-0.74), for Group B-E, respectively. Conclusions: Our thorough analysis suggests that metformin use does not improve survival of patients with PDAC. The benefit of metformin was observed only in patients who started metformin post-PDAC diagnosis; however, there is an inherent survival bias in this group of patients. Studies of medication exposure and cancer survival warrant very careful and detailed data collection. Researchers should exercise caution when initiating clinical trials based on retrospective epidemiologic studies. Citation Format: Roongruedee Chaiteerakij, David B. Zhen, Patrick A. Burch, Kari G. Chaffee, William R. Bamlet, Ann L. Oberg, Lewis R. Roberts, Gloria M. Petersen. Metformin use does not increase survival of pancreatic cancer patients: A cautionary lesson. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr LB-183. doi:10.1158/1538-7445.AM2015-LB-183