Performance and Clinical Utility of the Memorial Sloan Kettering Cancer Center Nomogram for Predicting Lymph Node Invasion in Patients Undergoing Radical Prostatectomy.

Introduction Pelvic lymph node dissection remains the most accurate method for nodal staging in patients with prostate cancer undergoing radical prostatectomy (RP), although its potential morbidity warrants careful patient selection. Preoperative nomograms may help estimate the probability of lymph node invasion (LNI) and identify patients for whom pelvic lymph node dissection could potentially be omitted. This study aimed to evaluate the discriminatory performance and potential clinical utility of the Memorial Sloan Kettering Cancer Center (MSKCC) pre-RP nomogram for predicting LNI in a contemporary, single-center cohort. Materials and Methods We conducted a retrospective observational study of 302 consecutive patients who underwent RP for prostate cancer between January 2020 and December 2023 at a tertiary urology center. Of these, 260 patients underwent pelvic lymph node dissection and were included in the cohort for evaluating nodal outcomes. Nomogram performance was assessed using receiver operating characteristic (ROC) curve analysis, calibration measures, and diagnostic performance at different risk thresholds. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the proportion of potentially avoidable lymph node dissections were evaluated, with particular emphasis on an 8% predicted-risk threshold. Results Among the 260 patients who underwent pelvic lymph node dissection, 34/260 (13.1%) had pathologically confirmed LNI. The nomogram demonstrated good discrimination for LNI, with an area under the ROC curve (AUC) of 0.859 (95% confidence interval (CI): 0.798-0.911). At an 8% threshold, sensitivity was 34/34 (100.0%), specificity was 93/226 (41.2%), PPV was 34/167 (20.4%), and NPV was 93/93 (100.0%). Overall, 93 of the 260 pelvic lymph node dissections (93/260, 35.8%) could potentially have been avoided without missing any of the 34 patients with LNI. Increasing the threshold to 9% would have increased the proportion of potentially avoidable dissections to 112/260 (43.1%), but at the cost of missing 1/34 (2.9%) patients with LNI. The mean predicted risk exceeded the observed prevalence of LNI (17.1% vs. 34/260 (13.1%)), suggesting a degree of risk overestimation. Conclusions The MSKCC nomogram showed good discriminatory performance for LNI in this cohort. An 8% predicted-risk threshold identified all 34 patients with pathologically confirmed LNI (34/34, 100.0%) while potentially avoiding 93/260 (35.8%) pelvic lymph node dissections. However, the model had limited specificity and demonstrated some degree of risk overestimation. Importantly, because pathological tumor stage was used in cases where clinical stage was unavailable, these results should not be considered a formal external validation of the preoperative nomogram. Further validation using exclusively preoperative variables is warranted before routine adoption of this threshold.