Development and internal validation of a novel nomogram for predicting lymph node invasion for prostate cancer
Zhen Li1,2, Yixin Huang1,2, Diwei Zhao1,2
1Department of Urology, Sun Yat-sen University Cancer Center, Guangzhou, China.
Background:
Few studies have focused on the performance of Briganti 2012, Briganti 2017 and MSKCC nomograms in the Chinese population in assessing the risk of lymph node invasion(LNI) in prostate cancer(PCa) patients and identifying patients suitable for extended pelvic lymph node dissection(ePLND). We aimed to develop and validate a novel nomogram based on Chinese PCa patients treated with radical prostatectomy(RP) and ePLND for predicting LNI.
Methods:
We retrospectively retrieved clinical data of 631 patients with localized PCa receiving RP and ePLND at a Chinese single tertiary referral center. All patients had detailed biopsy information from experienced uropathologist. Multivariate logistic-regression analyses were performed to identify independent factors associated with LNI. The discrimination accuracy and net-benefit of models were quantified using the area under curve(AUC) and Decision curve analysis(DCA).The nonparametric bootstrapping were used to internal validation.
Results:
A total of 194(30.7%) patients had LNI. The median number of removed lymph nodes was 13(range, 11-18). In univariable analysis, preoperative prostate-specific antigen(PSA), clinical stage, biopsy Gleason grade group, maximum percentage of single core involvement with highest-grade PCa, percentage of positive cores, percentage of positive cores with highest-grade PCa and percentage of cores with clinically significant cancer on systematic biopsy differed significantly. The multivariable model that included preoperative PSA, clinical stage, biopsy Gleason grade group, maximum percentage of single core involvement with highest-grade PCa and percentage of cores with clinically significant cancer on systematic biopsy represented the basis for the novel nomogram. Based on a 12% cutoff, our results showed that 189(30%) patients could have avoided ePLND while only 9(4.8%) had LNI missing ePLND. Our proposed model achieved the highest AUC (proposed model vs Briganti 2012 vs Briganti 2017 vs MSKCC model: 0.83 vs 0.8 vs 0.8 vs 0.8, respectively) and highest net-benefit via DCA in the Chinese cohort compared with previous nomograms. In internal validation of proposed nomogram, all variables had a percent inclusion greater than 50%.
Conclusion:
We developed and validated a nomogram predicting the risk of LNI based on Chinese PCa patients, which demonstrated superior performance compared with previous nomograms.
