OBJECTIVES: To evaluate hemoglobin A1c (HbA1c) reduction in exenatide and liraglutide in a veteran population. METHODS: Retrospective cohort study investigating exenatide and liraglutide use in a veteran population over a 24-month follow-up period. Patients were included if they were >/=18 years, eligible for veterans benefits, and initiated exenatide or liraglutide at the Veterans Health Administration (VHA). Patients were excluded if they were prescribed both medications during the follow-up period or crossed over into the other group. Clinical data were extracted from the VHA Corporate Data Warehouse. HbA1c reduction was evaluated at 12 months and 24 months. Main dependent variable was HbA1c reduction from baseline at 24 months. Multiple linear regression evaluated reduction in HbA1c controlling for potential confounders. Statistical significance was set at P=.05. RESULTS: A total of 1318 patients were identified receiving exenatide (N=1181) and liraglutide (N=137) based on inclusion and exclusion criteria. A majority of the exenatide patients were male (N=1102, 93%); average age was 59.8 years,  average Charlson comorbidity index (CCI) was 1.53, and average BMI was 37.53 kg/m^2. Similarly, a majority of the liraglutide patients were male (N=126, 92%); average age was 62.93 years; average CCI was 1.51; and average BMI was 34.68 kg/m^2. At 12 months HbA1c reduction was significant within the exenatide (P<0.0001) and liraglutide (P<0.0001) groups. We observed similar findings at the 24-month follow up. In the multiple linear regression, exenatide was associated with an HbA1c reduction of 0.093% (95% Confidence Interval (CI): -0.55, 0.74) compared to liraglutide while controlling for confounders (R-square = 0.37). CONCLUSIONS: These results have important implications regarding formulary decisions. Other factors such as dosing regimen, adherence, and side effect profile may be important attributes for formulary preference. Moreover, evaluating clinical outcomes (e.g., stroke and mortality) may reveal important differences. Generalizability may be limited to a non-veteran population.