Unification of favourable intermediate-, unfavourable intermediate-, and very high-risk stratification criteria for prostate cancer.

OBJECTIVE: To improve on the existing risk-stratification systems for prostate cancer.

PATIENTS AND METHODS: This was a retrospective investigation including 2 248 patients undergoing dose-escalated external beam radiotherapy (EBRT) at a single institution. We separated National Comprehensive Cancer Network (NCCN) intermediate-risk prostate cancer into 'favourable' and 'unfavourable' groups based on primary Gleason pattern, percentage of positive biopsy cores (PPBC), and number of NCCN intermediate-risk factors. Similarly, NCCN high-risk prostate cancer was stratified into 'standard' and 'very high-risk' groups based on primary Gleason pattern, PPBC, number of NCCN high-risk factors, and stage T3b-T4 disease. Patients with unfavourable-intermediate-risk (UIR) prostate cancer had significantly inferior prostate-specific antigen relapse-free survival (PSA-RFS, P < 0.001), distant metastasis-free survival (DMFS, P < 0.001), prostate cancer-specific mortality (PCSM, P < 0.001), and overall survival (OS, P < 0.001) compared with patients with favourable-intermediate-risk (FIR) prostate cancer. Similarly, patients with very high-risk (VHR) prostate cancer had significantly worse PSA-RFS (P < 0.001), DMFS (P < 0.001), and PCSM (P = 0.001) compared with patients with standard high-risk (SHR) prostate cancer. Moreover, patients with FIR and low-risk prostate cancer had similar outcomes, as did patients with UIR and SHR prostate cancer.

RESULTS: Consequently, we propose the following risk-stratification system: Group 1, low risk and FIR; Group 2, UIR and SHR; and Group 3, VHR. These groups have markedly different outcomes, with 8-year distant metastasis rates of 3%, 9%, and 29% (P < 0.001) for Groups 1, 2, and 3, respectively, and 8-year PCSM of 1%, 4%, and 13% (P < 0.001) after EBRT. This modified stratification system was significantly more accurate than the three-tiered NCCN system currently in clinical use for all outcomes.

CONCLUSION: Modifying the NCCN risk-stratification system to group FIR with low-risk patients and UIR with SHR patients, results in modestly improved prediction of outcomes, potentially allowing better personalisation of therapeutic recommendations.