Predicting the Probability of Lymph Node Involvement with Prostate Cancer Nomograms: Can We Trust the Prediction Models?

Abstract

Abstract Introduction Prediction of lymph node involvement (LNI) is of paramount importance for patients with prostate cancer (PCa) undergoing radical prostatectomy (RP). Multiple statistical models predicting LNI have been developed to support clinical decision-making regarding the need of extended pelvic lymph node dissection (ePLND). Our aim is to evaluate the prediction ability of the best-performing prediction tools for LNI in PCa in a Latin-American population. Methods Clinicopathological data of 830 patients with PCa who underwent RP and ePLND between 2007 and 2018 was obtained. Only data from patients who had ≥ 10 lymph nodes (LNs) harvested were included (n = 576 patients). Four prediction models were validated using this cohort: The Memorial Sloan Kettering Cancer Center (MSKCC) web calculator, Briganti v.2017, Yale formula and Partin tables v.2016. The performance of the prediction tools was assessed using the area under the receiver operating characteristic (ROC) curve (AUC). Results The median age was 61 years old (interquartile range [IQR] 56–66), the median Prostate specific antigen (PSA) was 6,81 ng/mL (IQR 4,8–10,1) and the median of LNs harvested was 17 (IQR 13–23), and LNI was identified in 53 patients (9.3%). Predictions from the 2017 Briganti nomogram AUC (0.85) and the Yale formula AUC (0.85) were the most accurate; MSKCC and 2016 Partin tables AUC were both 0,84. Conclusion There was no significant difference in the performance of the four validated prediction tools in a Latin-American population compared with the European or North American patients in whom these tools have been validated. Among the 4 models, the Briganti v.2017 and Yale formula yielded the best results, but the AUC overlapped with the other validated models.