Background: Combined androgen deprivation therapy (ADT) and radiation therapy (RT) is a frequently used localized prostate cancer (PC) treatment. Testosterone recovery (TR) after combined ADT-RT is not well-characterized. We studied TR in men who received RT and either short-term (ST) ADT or long-term (LT) ADT with LHRH agonists. Methods: We identified consecutive localized PC patients treated with ADT-RT at the Durham VA Medical Center (DVAMC) from 1/2011-10/2016. All patients had a documented baseline testosterone (T) level. Individual patient records were reviewed. TR was defined as time from last ADT injection to T normalization ( > 240 ng/dL). The Kaplan-Meier method was used to estimate time to TR. Cox proportional hazards models were generated to identify TR predictors with a nomogram built based on a parsimonious multivariate model. Results: 252 patients were identified. Median follow-up was 26.7 months. Median age was 65. Prior to treatment, 69% had a normal baseline T. 67% were treated with STADT, median duration 6 months. 33% were treated with LTADT, median duration 18 months. Median time for TR was 22.6 months for all patients (19.5 months for STADT and 25.6 months for LTADT). At 1 and 2 years post ADT, estimated TR was 13% and 53% (17% and 57% for STADT and 3% and 42% for LTADT). 2-year biochemical control was 99.2% and 97.6% for STADT and LTADT, respectively; 98.9% and 98.6% for those with and without TR, respectively. On multivariate analysis, higher pre-treatment T (HR = 1.004 95% CI 1.003-1.006, p < 0.001), use of STADT (HR = 2.48 95% CI 1.45-4.25, p = 0.001), and lower BMI (HR = 0.95 95% CI 0.91-0.98, p = 0.001) were associated with shorter time to TR. White race was a negative TR predictor (HR = 0.65 95% CI 0.43-0.9992, p = 0.049). Age, smoking, and Charlson Comorbidity Index were not significant independent TR predictors. A nomogram was generated to predict probability of TR at 1, 2, and 3 years. Conclusions: In this VA population of localized PC patients treated from 2011-2016, TR following the use of ADT-RT was variable. Using pre-treatment T levels, ADT duration, BMI, and race, a predictive nomogram can estimate the likelihood of TR.