Background: To determine the rate and time of testosterone (T) recovery to normal level in patients (pts) with prostate cancer treated with radiotherapy plus 6, 18 or 36 months of androgen deprivation therapy (ADT) and considered cured from their disease. Methods: We randomized 1230 pts with prostate cancer, into two phase III trials: 600 with intermediate risk and 630 with high-risk. We selected those considered cured to avoid subsequent T variations due to reintroduction of ADT for recurrence. We excluded the following pts: no ADT at all (126) or not receiving exactly 6, 18 or 36 months of ADT (69), survival less than one year (21), no T measured at baseline or during follow-up (75), biochemical failure (195) or evidence of metastatic/recurrent disease (137).T recovery rate was compared between baseline normal/abnormal T (values below biochemical normal range) and by ADT duration with Chi-square test or Fisher's exact test. A multivariable logistic regression model to predict the probability of recovering normal T was performed by including normal/abnormal T at baseline, age, Zubrod, comorbidities and ADT duration. A second model was performed by replacing ADT duration with baseline PSA, Gleason score and stage. The median time to T recovery was calculated only on pts who recovered normal T. Results: Results are reported with a median follow-up of 14 years. 607 pts fit the criteria and are available for analysis: 309 pts in the 6 months ADT schedule, 185 in the 18 and 113 in the 36. Overall, 76.7%, 54.6% and 45.1% pts recovered normal T on the 6, 18 or 36 months schedule, respectively (p<0.001). The median time to T recovery was 1.5, 3.1, and 5.1 years for the 6, 18 or 36 months schedule, respectively (p<0.001). 79.7% presented with a normal T at baseline while 20.3% had an abnormal T level. By splitting pts between a normal vs. abnormal presenting T level, the T recovery rate was as follows: 82.1%, 63.3%, and 50% for the normal T cohort, compared to 53.4%, 28.3% and 21.1% for the abnormal T cohort at 6, 18 or 36 months, respectively. There was a significant difference in the overall recovery rate (p<0.001) between normal vs. abnormal T level and at all ADT duration lengths between the two cohorts. In multivariable model, baseline normal T was a strong predictor of T recovery. Older age, diabetes, longer ADT, higher clinical stage, higher PSA and higher Gleason score reduced significantly the chance for T recovery. In pts recovering T post-ADT, except for the 6 months duration (p=0.01), the median time for T recovery was not significantly different between normal or abnormal T at baseline in 18 and 36 months cohorts. Conclusions: Older age, longer ADT and poor disease features are associated with a lower T recovery. Even after adjusting for several variables and ADT duration, a higher T recovery post-ADT is significantly associated with a normal T at baseline.