A Modified Robust Nomogram for Predicting the Probability of Pelvic Lymph Node Invasion in Localized Prostate Cancer.

Abstract

Improved predictive models for the risk of pelvic lymph node invasion (LNI) in localized prostate cancer (PCa) is important in decision making of pelvic lymph node dissection (PLND) or whole pelvic RT (WPRT). We aimed to establish a nomogram with improved robustness (Ruijin model) based on pre-treatment information. Continuous localized PCa patients with detailed prostate tumor biopsy information, treated with laparoscopic radical prostatectomy and PLND between 2013 and 2022 in single institution were retrospectively reviewed. A multivariable logistic regression model was fitted and represented the basis for a coefficient-based nomogram of predicting LNI. Comparisons between the Ruijin model and the Roach formula were conducted using the receiver operating characteristic-derived area under the curve (AUC), calibration plot, and decision-curve analyses (DCAs). In total, 624 patients with median age (69.5 year) were included in this analysis. The median number of pelvis lymph nodes removed was 6 (range: 2-30). LNI was found in 35 (5.6%) patients. In the multivariable logistic regression model, total prostate-specific antigen (OR = 1.008, P = 0.025), percentage of cores with the highest-grade PCa (OR = 14.822, P<0.001), clinical stage (III vs. I-II: OR = 7.733, P = 0.008), and biopsy Gleason Grading Group (G3 vs G1-2: OR = 3.152, P = 0.082; G4 vs G1-2: OR = 3.065, P = 0.083; G5 vs G1-2: OR = 5.262, P = 0.008;) were included and formed the basis for the nomogram. The predictive accuracy of Ruijin nomogram in our cohort was 87.7%. Using a cutoff of 4% based on Ruijin nomogram, 395 (63%) PLND would be spared and LNI would be missed in only two (0.5%) patients. The sensitivity, specificity, and negative predictive value associated with the 4% cutoff were 94.3%, 66.7%, and 99.5%, respectively. As compared with the Roach formula, the Ruijin model showed higher AUC (87.7% vs 80.9%, Z = -2.013, P = 0.044), better calibration characteristics, and a higher net benefit at DCA. We developed a novel nomogram for predicting the LNI in localized PCa patients with detailed biopsy information. PLND or WPRT could be avoided in patients with a risk of LNI <4%, so as to spare more than 60% of unnecessary pelvic nodal treatment with a cost of missing only 0.5% LNIs.