Introduction: External beam radiotherapy is a treatment option for clinically localised prostate cancer; however, some 15% of patients will undergo treatment failure within 5 years. The objective was to compare the Cancer of the Prostate Risk Assessment (CAPRA) score (based on the clinical-pathological findings) and the sub-types of minimal residual disease (MRD) (based on the biological properties of the cancer cells) risk classifications to predict biochemical failure (BF) after external beam radiotherapy. Methods and Patients: Clinical-pathological findings were obtained from the prostate biopsy to determine the CAPRA score and used to define low-, intermediate-and high-risk patients. Blood and bone marrow were obtained 3 months after radiotherapy; circulating prostate cells (CPCs) and micro-metastasis were detected using immunocytochemistry with anti-prostate specific antigen. CPCs and micro-metastasis were classified as positive if at least one cell was detected in the sample. Three subgroups were formed Group A (MRD negative), Group B (micro-metastasis positive, CPC negative) and Group C (CPC positive) Patients were followed up for 10 years or until biochemical failure. Biochemical failure free survival (BFFS) curves were constructed using Kaplan-Meier (observed), a flexible parameter model (predicted survival) and the restricted mean survival time (RMST) was calculated for each subgroup. Results: 309 men participated with a median follow-up of 8 years. The risk of biochemical failure increased proportionally with increasing CAPRA score, hazard ratio 1.18 for intermediate and 1.69 for high risk patients. After 10 years, the percentage BFFS and RMST to failure were 74%, 49%, 16% and 9, 7 and 6 years, respectively. The agreement between observed and predicted BFFS was acceptable (Harrell´s C 0.62). The BFFS curves for MRD were different and not proportional, survival curves for men MRD negative and only micro-metastasis were similar up to 5 years, and then there was increasing failure in the micro-metastasis only group. After 10 years, the percentage BFFS and RMST to failure were 95%, 59%, 28% and 10, 9 and 6 years, respectively. The CAPRA score failed to distinguish between Groups A and B, one third of high risk Group C had low risk CAPRA scores. The agreement between observed and predicted BFFS was very good (Harrell´s C 0.92). Minimal residual disease hazard ratios were Group B 1.84 and Group C 4.51. Research