OBJECTIVES: For localized prostate cancer, both primary androgen deprivation therapy (PADT) and surgery is frequently used, without any comparative evidence of survival benefit between them.  This retrospective study aimed to examine the predictors of receiving PADT versus radical prostatectomy (RP) as a monotherapy among clinically localized prostate cancer patients. METHODS: Male veterans with localized prostate cancer were identified from the Veterans Affairs Veterans Integrated Service Network (VISN) 16 (01/2004-12/2008). Eligible patients had no other cancer diagnosis before the first prostate cancer diagnosis and used PADT or RP as monotherapy within 6 month after first diagnosis. A proxy variable (preference) was created to account for potential physician influence on treatment choice in each of the ten medical centers of VISN 16. Predictors fitted in a logistic regression model included age, race, marital status, non-VA insurance, cancer stage, baseline prostate specific antigen (PSA), Charlson comorbidity index (CCI), alcohol and tobacco use, calendar year at diagnosis, and preference. RESULTS: The analysis sample included 878 patients, 523 in PADT cohort and 355 in RP cohort. The age [Mean (SD)] in PADT was 71.4 (8.92) years old, 60.3 (5.91) in RP, p<0.001. The significant predictors of receiving treatment of PADT over RP included being 65 years old or over (OR=9.152, 95% CI: 5.688, 14.725), higher baseline PSA (OR=1.184, 95% CI=1.133, 1,237), higher CCI (CCI=1: OR=2.246, 95% CI=1.062, 4.753; CCI>=2: OR=7.328, 95% CI=2.599, 20.665) and preference of PADT (OR=4.603, 95% CI=3.068, 6.906). Other explanatory variables including race, marital status, cancer stage, year at diagnosis were not significant (all p-values>0.05). CONCLUSIONS: Among localized prostate cancer patients in VA, age, baseline PSA and CCI were strong predictors of treatment selection of PADT versus RP. We did not observed a significant treatment selection patterns change over time as expected.