BACKGROUND Various definitions of biochemical failure (BF) have been used to predict cancer recurrence following prostate cryoablation. However to date, none of these definitions have been validated for this use. We have reviewed several definitions of BF to determine their accuracy in predicting biopsy‐proven local recurrence following prostate cryoablation. METHODS The Columbia University Urologic Oncology Database was queried for patients who underwent prostate cryoablation between 1994 and 2010, and who subsequently underwent surveillance biopsy due to clinical suspicion of prostate cancer recurrence. Serial postoperative prostate‐specific antigen (PSA) results were used to determine BF according to various definitions of BF. Biopsy results were used to determine local recurrence. Sensitivity, specificity, positive and negative predictive value, and receiver operating characteristic (ROC) curve area were calculated for each of the BF definitions. RESULTS A total of 110 patients met inclusion criteria for the study. These patients were treated with primary full‐gland (n = 38), primary focal (n = 24), or salvage cryoablation (n = 48). On surveillance biopsy, 66 patients (60%) were found to have locally recurrent prostate cancer. The most accurate BF definition overall was PSA nadir plus 2 ng/ml (Phoenix definition), with sensitivity, specificity, and ROC curve area of 68%, 59%, and 0.64, respectively. CONCLUSIONS Overall, the Phoenix definition best predicted local cancer recurrence following prostate cryoablation. These preliminary data may be useful for researchers evaluating the short‐term efficacy of cryoablation, and for urologists assessing their patients for potential cancer recurrence. Prostate 72:1802–1808, 2012. © 2012 Wiley Periodicals, Inc.