Predictive role of node-rads score in patients with prostate cancer candidates for radical prostatectomy with extended lymph node dissection: comparative analysis with validated nomograms

Background and objectives: The Reporting and Data System (RADS) have been used in the attempts to standardize the results of oncological scans in different scenarios, such as lymph nodes, adding configuration criteria to size determination. We analyze the predictive value of preoperative Node-RADS determination at imaging for pelvic lymph node (PLN) involvement in cases of prostate cancer (PC) considered for radical prostatectomy (RP) with extended lymph node dissection (eLND) and we compare it with validate predictive nomograms (MSKCC, Briganti and Gandaglia). Methods: 150 patients with a histological diagnosis of PC (high risk or intermediate with an estimated risk for pN+ higher than 5% using the Briganti or 7% using the Gandaglia nomogram) submitted for RP with an ePLND from 2018 and 2021 were retrospectively examined. Node-RADS determination was performed in all cases using the preoperative magnetic resonance (MR), performed by a radiologist blinded for pathologic results and compared with the MSKCC, Briganti 2012, Gandaglia 2017 and Gandaglia 2019 nomograms. Results: PLN involvement at final pathology (pN+) was found in 36/150 (24.0%) of cases and the mean percentage of positive LNs in pN+ cases was 15.90 ± 13.40. The mean number of PLNs removed at RP was similar (p = 0.188) between pN0 (23.9 ± 8.0) and pN+ (25.3 ± 8.0) cases. Considering a Node RADS 4-5 positive and a Node RADS 1-2 negative, the PPV was 100% and the NPV was 79.6%. A Node RADS score 4-5 showed a lower sensitivity (0.167 versus 0.972, 1.000, 0.971, 0.960 respectively), a higher specificity (1.000 versus 0.079, 0.096, 0.138, 0.186 respectively) and a similar AUC (0.583 versus 0.591, 0.581, 0.574, 0.597 respectively) when compared to MSKCC, Briganti 2012, Gandaglia 2017 and Gandaglia 2019 nomograms. Conclusions: Our evaluation suggests that Node RADS score, combining configuration criteria to size determination could improve specificity in terms of pathologic PLN prediction but a very low sensitivity has been also described.