Oncological outcomes after radical prostatectomy of localized prostate cancer: stratified by magnetic resonance imaging and risk classification.

Background: We investigated whether combining T2-weighted magnetic resonance imaging (MRI) findings and clinical risk categories improves upon established prognostic indicators of oncological outcomes in prostate cancer.

Methods: Patients who underwent radical prostatectomy, but not preoperative hormone therapy, radiotherapy, or chemotherapy, for localized prostate cancer at Seoul National University Bundang Hospital from October 2007 to April 2016 were included. MRIs were classified according to the Prostate Imaging-Reporting and Data System (PI-RADS). Patients were divided into the following five groups: 1, no focal suspicious lesion; 2, organ-confined suspicious lesion PI-RADS ≤3; 3, organ-confined suspicious lesion PI-RADS 4 or 5; 4, suspicious lesion with extraprostatic extension (EPE), no seminal vesicle invasion (SVI); 5, suspicious lesion with EPE and SVI. Risk classified according to the National Comprehensive Cancer Network (NCCN) and MRI findings were combined to analyze survival curves for biochemical recurrence (BCR)-free and metastasis-free survival. The area under a time-dependent receiver operating characteristic was analyzed for event prediction after 5 years.

Results: We analyzed 1,290 patients. In multivariate Cox regression models, PI-RADS ≥4 (hazard ratio [HR] 2.33, P < 0.001), EPE (HR 1.46, P = 0.027), SVI (HR 5.03, P < 0.001) and NCCN high-risk (HR 2.33, 95% CI 1.66-3.26, P < 0.001) were associated with BCR. For metastasis, EPE (HR 2.33, P = 0.047), SVI (HR 13.08, P < 0.001) and NCCN high-risk (HR 2.78, P = 0.026) were independent risk factors. Depending on MRI group, BCR-free survival significantly decreased in NCCN intermediate-risk (P = 0.001) and high-risk (P < 0.001) groups, and metastasis-free survival decreased in the intermediate-risk group (P = 0.39) and significantly decreased in the high-risk (P < 0.001) group. Adding MRI group to NCCN risk classification significantly improved the predictive accuracy for BCR in comparison with NCCN risk classification alone (P = 0.042), but not for metastasis (P = 0.012).

Conclusion: Combining prostate MRI with NCCN risk classification improves the prediction value of BCR following radical prostatectomy for localized prostate cancer.