Validation of models predicting lymph node involvement probability in patients with prostate cancer

Area of Science:

• Urology
• Oncology
• Medical Statistics

Background:

• Accurate prediction of lymph node involvement is crucial for prostate cancer management.
• Several models exist to predict lymph node involvement, but external validation in diverse populations is necessary.

Purpose of the Study:

• To externally validate established prostate cancer lymph node involvement prediction models in a Japanese cohort.
• To compare the performance of different models in predicting lymph node metastasis.

Main Methods:

• Retrospective analysis of 331 patients undergoing robotic-assisted radical prostatectomy with extended pelvic lymph node dissection.
• External validation of multiple prediction models including the Memorial Sloan Kettering Cancer Center web calculator, Yale formula, Briganti nomogram, and Partin table.
• Model performance assessed using receiver operating characteristic curves (AUC), calibration plots, and decision curve analyses.

Main Results:

• Lymph node involvement was identified in 18.4% of patients.
• The Memorial Sloan Kettering Cancer Center web calculator achieved the highest AUC (0.78), followed closely by the Yale formula (0.77) and Briganti nomogram (0.76).
• The Memorial Sloan Kettering Cancer Center web calculator and Briganti nomogram demonstrated better calibration and higher net benefit for predicting low-risk lymph node involvement.

Conclusions:

• Established prostate cancer lymph node prediction models show varying accuracy when validated in a Japanese cohort.
• The Memorial Sloan Kettering Cancer Center web calculator and the 2017 Briganti nomogram are recommended as the most accurate models for predicting lymph node involvement in this population.