INTRODUCTION: To identify the most significant cut-off of tumor volume (TV) for prediction of clinical failure (CF) among high-risk prostate cancer (hPCa) patients. METHODS: Within a multi-institutional cohort, 262 patients treated with radical prostatectomy (RP) for hPCa were identified. CF was defined as local recurrence or distant metastases. A time dependent ROC curve was used to evaluate the area under the curve (AUC) using TV as single marker to predict clinical failure at 10 years. We searched for the TV cut off value with the highest combined sensitivity and specificity predicting CF. Three multivariable Cox regression analyses (MVA) tested the predictors of CF after RP. Predictors of the model 1 were pre-operative PSA, pathologic stage (PT), pathologic Gleason sum (GS), surgical margin status, and lymph node invasion. Predictors of the models 2 and 3 were the same of model 1 plus TV as a continuous or dichotomous variable using the defined cutoff, respectively. Validation (leave-one-out-cross-validation-LOOCV) of each model was performed. RESULTS: Overall, 46 (17.6%) patients experienced CF. The TV value was 6.29 ml. In MVA of models 2 and 3, PT and GS remained independent predictors of CF. Moreover, in model 2 TV (HR:1.07,) and in model 3 TV >6.29 ml (HR:2.99,) were independently associated with CF. In LOOCV, the C-index of models 1-3 were 65.53%, 71.75%, and 70.26%, respectively. CONCLUSIONS: TV is an independent predictor of CF in hPCa patients. Patients with a TV exceeding the cut-off of 6.29 ml are more likely to develop CF. Prostate 77:3-9, 2017. © 2016 Wiley Periodicals, Inc. © 2016 Wiley Periodicals, Inc.