First North American validation and head-to-head comparison of four preoperative nomograms for prediction of lymph node invasion before radical prostatectomy.

ABSTRACT

OBJECTIVES: To perform a head-to-head comparison of four nomograms; namely, the Cagiannos, the 2012-Briganti, the Godoy and the online-Memorial Sloan Kettering Cancer Center (MSKCC), for prediction of lymph node invasion (LNI) in a North American population. PATIENTS AND METHODS: A total of 19 775 patients with clinically localized prostate cancer (PCa) who had undergone radical prostatectomy and pelvic lymph node dissection (PLND) were identified within the Surveillance Epidemiology and End Results (SEER) database. All four nomograms were tested using Heagerty's concordance index (C-index), calibration plots and decision curve analysis (DCA). In addition, we examined specific nomogram-derived thresholds to compare the number of avoided PLNDs and missed LNI-positive cases. RESULTS: All nomograms were found to have highly comparable C-index values the Cagiannos, 78.6%; the Godoy, 78.2%; the 2012-Briganti, 79.8%; and the MSKCC, 79.9%. The Cagiannos nomogram showed the best calibration, followed by the 2012-Briganti, the Godoy and the online-MSKCC. In DCA, the 2012-Briganti and the Cagiannos, in that order, provided the best results, followed by the Godoy and the online-MSKCC models. For each nomogram, the threshold associated with ≤10% missed LNI cases avoided 8 693 (46.6%), 8 652 (46.4%), 8 461 (45.4%) and 8 590 (46.1%) PLNDs, respectively, with the use of the Cagiannos (2.6% threshold), the online-MSKCC (4.3% threshold), the Godoy (3.6% threshold) and the 2012-Briganti (4.6% threshold) nomograms. CONCLUSION: The Cagiannos and the 2012-Briganti nomograms exhibited the best calibrations and DCA results. Conversely, C-index values and ability to avoid unnecessary PLNDs were virtually the same for all four nomograms examined.