Development and External Validation of a Multiparametric Magnetic Resonance Imaging and International Society of Urological Pathology Based Add-On Prediction Tool to Identify Prostate Cancer Candidates for Pelvic Lymph Node Dissection.

We sought to expand current prediction tools for lymph node invasion in patients with prostate cancer using current state-of-the-art available tumor information, including multiparametric magnetic resonance imaging based tumor stage and detailed biopsy information. We selected patients with prostate cancer for study who had available registered information on ISUP (International Society of Urological Pathology) based biopsy grading and multiparametric magnetic resonance imaging, and who had undergone radical prostatectomy with extended pelvic lymph node dissection. We developed a lymph node invasion prediction tool in 420 patients and externally validated it in 187. A concordance index was estimated to quantify the discriminative performance of the model. In the development cohort a median of 21 lymph nodes were removed per patient and 71 patients (16.9%) were diagnosed with lymph node invasion. Statistically significant predictors of lymph node invasion were the initial prostate specific antigen value, multiparametric magnetic resonance imaging based T stage, maximum tumor length in 1 core in mm and ISUP grade group corresponding to the maximum tumor involvement in 1 core. The predictive accuracy of this lymph node invasion prediction tool was 79.7% after fivefold internal cross validation and 72.5% after external validation. We report a contemporary, externally validated prediction tool for lymph node invasion in patients with prostate cancer. This prediction tool is a response to the paradigm shift from systematic to targeted biopsies by incorporating additional core specific biopsy information instead of the percent of positive cores. This new tool will also overcome stage migration, which is a potential risk when multiparametric magnetic resonance imaging information is used in digital rectal examination based nomograms.