Head-to-head Comparison of Three Commonly Used Preoperative Tools for Prediction of Lymph Node Invasion at Radical Prostatectomy

• Published in: Urology (Ridgewood, N.J.) N.J.), 2011-12, Vol.78 (6), p.1363-1367
• Main Authors: Abdollah, Firas, Schmitges, Jan, Sun, Maxine, Thuret, Rodolphe, Djahangirian, Orchidee, Tian, Zhe, Shariat, Shahrokh F, Briganti, Alberto, Perrotte, Paul, Montorsi, Francesco, Karakiewicz, Pierre I
• Format: Article
• Language: English
• Subjects: Adenocarcinoma - pathology; Adenocarcinoma - surgery; Adult; Aged; Aged, 80 and over; Area Under Curve; Biological and medical sciences; Decision Support Techniques; Humans; Lymph Node Excision; Lymph Nodes - pathology; Lymph Nodes - surgery; Lymphatic Metastasis; Male; Medical sciences; Middle Aged; Neoplasm Staging; Nephrology. Urinary tract diseases; Nomograms; Predictive Value of Tests; Prostate-Specific Antigen - blood; Prostatectomy; Prostatic Neoplasms - pathology; Prostatic Neoplasms - surgery; Urology
• ISSN: 0090-4295; 1527-9995
• DOI: 10.1016/j.urology.2011.07.1423

Objective A formal validation and head-to-head comparison of the National Comprehensive Cancer Network (NCCN) practice guideline lymph node invasion (LNI) nomogram, Partin tables, and D'Amico risk-classification was conducted for prediction of LNI at radical prostatectomy (RP). Methods We focused on 20,877 patients treated with RP and pelvic lymph node dissection (PLND) between 2004 and 2006 within the Surveillance, Epidemiology and End Results database. The discrimination of the 3 tools in predicting histologically confirmed LNI was quantified using the area under the curve (AUC). Calibration plots were used to graphically depict the performance characteristics of the examined tools. In addition, we relied on decision curve analyses to compare the 3 models directly in a head-to-head fashion. Results Overall, 2.5% of patients had LNI. The NCCN LNI nomogram (AUC 82%) outperformed the Partin tables (73%) and the D'Amico risk-classification (75%) for prediction of LNI. Calibration plots revealed that all 3 tools overestimated the risk of LNI. Partin tables showed the highest net-benefit for probability threshold range between 1% and 4%. Conversely, the NCCN LNI nomogram showed the highest net-benefit for the remaining threshold probabilities. Conclusion The NCCN LNI nomogram had the highest discrimination accuracy. However, using the decision curve analysis, the Partin tables demonstrated the highest net benefit when a threshold probability of LNI is