Head to Head Comparison of Nomograms Predicting Probability of Lymph Node Invasion of Prostate Cancer in Patients Undergoing Extended Pelvic Lymph Node Dissection

Abstract

To validate the Briganti nomogram and compare it with 2 current lymph node invasion (LNI) nomograms (the Cagiannos nomogram and the updated 2007 Partin tables). The Briganti nomogram predicts the probability of LNI in patients undergoing extended pelvic lymph node dissection (EPLND) during radical prostatectomy for prostate cancer. Irrespective of the risk of LNI, 173 consecutive patients were treated for localized prostate cancer with radical laparoscopic prostatectomy and EPLND. The area under the receiver operating characteristics curve was used to estimate the predictive accuracy of the nomograms, and calibration plots were used for comparisons between the predicted and observed probabilities of LNI. The median number of nodes removed was 15 (range 10-34). Of the 173 patients, 12 (6.9%) had LNI. The Briganti nomogram achieved a receiver operating characteristic curve of 0.88 versus 0.83 with the Cagiannos nomogram and 0.84 with the 2007 Partin tables. The difference in predictive accuracy was not statistically significant (P < .2). The Briganti nomogram showed only minor departures from the ideal predictions in the low-risk range and the Cagiannos nomogram showed major departures from the ideal predictions for the entire risk range. The Briganti nomogram provides highly accurate predictions of the risk of LNI after EPLND. Its performance tended to be increased without being significantly better. The other tools also performed reasonably well but underestimate the true risk of LNI. We recommend the use of these tools to identify patients at low risk of LNI for whom EPLND can be safely spared.