A nomogram to predict pathologic T2 stage in candidates to robot-assisted radical prostatectomy with iT3 prostate cancer on preoperative multiparametric MRI: results from a multi-institutional collaboration

In candidates to robot-assisted radical prostatectomy (RARP) for locally advanced (iT3) prostate cancer on preoperative MRI, the performance of MRI for local staging is demonstrably suboptimal, and currently no prediction tools that might help surgeons in preoperative planning are available. We analyzed data of 685 patients with iT3 prostate cancer (PCa) who received RARP at five participating institutions between 2012 and 2020. Multivariable logistic regression model investigated predictors of pT2 disease among variables available before surgery (i.e: preoperative PSA, biopsy ISUP group, clinical T stage on digital rectal examination-DRE, prostate volume on MRI, PIRADS score of index lesion, seminal vesicles invasion on MRI, location suspicious for T3 disease on MRI). Coefficients from such model were used to build a nomogram to predict organ-confined (i.e. pT2) disease on final pathology. Internal validation was performed using the leave-one-out cross-validation. Median (interquartile range) preoperative PSA was 7.5 (5.2, 11.9) ng/ml, and 280 (41%) and 216 (32%) had biopsy ISUP group 4-5 disease and palpable disease on DRE, respectively. Preoperative MRI was suspicious for iT3 disease on the mid-posterior part of the gland in 485 (71%) men, and 527 (77%) men had a PIRADS 5 lesion. After surgery, a total of 192 (28%) patients had organ-confined disease (i.e. pT2). All variables fitted into the model and were considered to build the nomogram. After internal validation, the AUC was 73% (95% confidence interval: 69%, 77%). Awaiting external validation, we provided data that is relevant to optimize surgical strategy in men diagnosed with iT3 PCa who are scheduled for RARP.