Development and internal validation of a novel predictive model to guide an individualized risk assessment in prostate cancer patients

Abstract

Introduction and objectives

To provide a risk-adapted strategy to manage prostate cancer (PCa) patients eligible for curative surgery by developing an individualized risk calculator to predict oncologic outcomes.

Materials and methods

Data of consecutive patients treated with robot-assisted radical prostatectomy (RARP) between March 2020 and June 2023 at a single tertiary referral center were prospectively collected and analyzed. Multivariate analysis using Cox proportional hazards model were performed to explore predictors of 3-year biochemical failure (BCF). Both preoperative and postoperative models explored, with key variables including tumor-related features and surgical delay. Based on the significant variables identified, two nomograms were developed to estimate the risk of 3-year BCF. The area under the receiving operator characteristics (ROC) curves (AUC) was used to quantify predictive discrimination. Internal validation using bootstrapping techniques was performed to assess the model's accuracy and calibration.

Results

Overall, 2017 patients were enrolled. At the multivariable analysis for preoperative model, cT stage, cN stage, ISUP grade on prostate biopsy, PIRADS of the index lesion on prostate MRI and surgical delay were significant predictive factors of 3-year BCF. At the multivariable analysis for postoperative predictive model, pT stage, pN stage, ISUP grade on final histopathological examination, surgical margins and surgical delay were significant predictive factors of 3-year BCF. The preoperative and postoperative model showed a ROC AUC of 60.7 % and 71.9 %, respectively. The final nomograms for both preoperative and postoperative models were built. Both models underwent internal validation using bootstrapping with 1000 repetitions.

Conclusions

To optimize the timing of surgery in PCa patients based on individual risk profile, we finally designed and internally validated two nomograms, which serve complementary roles. The preoperative nomogram offers early, albeit less precise, risk stratification to guide initial treatment planning, while the postoperative nomogram refines BCF predictions using definitive pathological data.