Purpose We analyzed our experience treating localized prostate cancer to determine the impact of androgen deprivation (AD) on clinical outcome if administered at the time of isolated biochemical failure (BF) vs. after clinical failure (clinical failure), and the associated impact of various BF definitions. Methods A total of 1,201 patients with stage T1–T3N0M0 prostate cancer were treated with external beam radiotherapy (EBRT) to a median dose of 66.6 Gy. Early AD was defined as administration of AD after BF, without evidence of clinical failure. Delayed AD was defined as administration of AD after clinical failure. Multiple BF definitions were tested for capacity to predict subsequent clinical failure. For each BF definition, outcome was compared for BF patients receiving early AD vs. no or delayed AD. Results Five-year clinical failure (from date of BF) was 60% for patients who experienced a prostate-specific antigen rise to ≥3 ng/mL above nadir. For these patients, early AD was associated with decreased 5-year local failure (4% vs. 33%), distant metastasis (13% vs. 44%), cause-specific death (9% vs. 24%), and death due to any cause (32% vs. 48%), despite poorer prognostic factors in patients receiving early AD. On multivariate analysis, early AD remained independently significant for each of these end points. Conclusion The efficacy of AD after BF varies depending on the BF definition. When an optimal BF definition is applied, early AD decreases distant metastasis and improves survival. Prostate-specific antigen elevation to ≥2 or ≥3 ng/mL above nadir seems optimal in establishing clinically significant BF and the timing of AD intervention.