Adjuvant radiotherapy after radical prostatectomy (RP) with adverse pathological features (APFs) has a significant decrease biochemical failure (BF) and local recurrence. Three published randomized trials showed overtreatment 40-60% and these patients did expose to unnecessary radiation toxicities. Thus, we developed a new prognostic model to predict the probability of BF after RP to select more appropriate patients for adjuvant treatment. DESIGN: Multivariable Cox regression of a retrospective cohort of postoperative prostate cancer patients. SETTING: One university hospital. PATIENTS: 292 consecutive prostate cancer patients in 2005 to 2009 after RP with pT2-3N0M0 and at least one of APFs (extracapsular extension, seminal vesicle involvement, close or positive margin) without any adjuvant radiotherapy or hormonal therapy. OUTCOME: BF defined as rising of PSA ≥ 0.2 ng/ml with a second confirmatory level (apart at least 2 week) or start any salvage treatment. The median follow up time was 82 months and 99 of 292 (34%) patients who underwent RP alone had BF with median time to BF 24 months. Preoperative PSA risk group, postoperative PSA level, perineural invasion, sum of Gleason score were strongest predictor of BF. Other predictors included in the final model were extracapsular extension, seminal vesicle invasion and surgical margin status. Discrimination of new model was satisfactory, with C statistics 0.78 (95% CI: 0.74-0.83), compared with C statistics from the conventional three standard APFs 0.60 (95% CI: 0.55-0.66). Calibration plot was performed by dividing patients into 4 risk groups and comparing the survival curves between predicted BF-free survival probability and observed outcome from Kaplan-Meier method. These two curves were fitted by visual analysis. A simple chart was constructed to easily provide 5-year probability of BF from sum of risk score from all predictors and web-based calculator is also available. If criteria were set to have 10% of 5 year probability of BF, this prediction rule can decrease overtreatment by 12%. This new prediction model can predict 5-year probability of BF in localized prostate cancer patients with APFs after treated with RP. In addition, this new model can assist in decision making between patient and doctor to tailor the individually appropriate treatment. Cautiously, internal and external validation should be performed in the near future.