Preoperative MRI-based predictive model for biochemical recurrence following radical prostatectomy.

Abstract

Purpose: To determine the biochemical recurrence (BCR)-related pelvic anatomic characteristics before radical prostatectomy (RP) and to establish a new predictive model for BCR-free survival (BCRFS).

Methods: The study involved 170 patients who underwent RP between January 2015 and December 2022. Kaplan-Meier plots were applied to estimate survival probabilities. Multivariate Cox regression models were employed to identify predictors for BCRFS, which were subsequently incorporated into an MRI-based nomogram to visualize the model. The Harrell's concordance index (C-index) was employed to evaluate the discrimination, and compared with a basic model without incorporating pelvic anatomy. Time-dependent receiver operating characteristic (ROC) curves and decision curve analysis (DCA) were applied to identify the advantage of the new predictive model. Three risk categories were created.

Results: Multifactorial analysis revealed that age, capsule contact length (CCL), tumor's distance to the proximal membranous urethra (UD), urethral width, and annual surgery volume were independent risk factors for BCR (all p < 0.05). The established predictive model yielded a C-index of 0.850 that was superior to aforementioned basic model with C-index of 0.771 (p < 0.001). Our new model with an area under the ROC curve (AUC) of 0.893 revealed better predictive ability in BCRFS than basic model with the AUC of 0.823 (p = 0.01), and DCA demonstrated that our model generated more net benefits.

Conclusion: UD and urethral width are independent predictors of BCRFS. Our new model exhibits superior predictive accuracy with respect to BCRFS relative to the basic model. Apart from tumor features, pelvic anatomical features should also be considered before the treatment decision making of PCa patients.