Predictors of ISUP upgrading from biopsy to radical prostatectomy: development and internal validation of a preoperative model in a single-center cohort

Background: Biopsy grade may underestimate tumor aggressiveness; a preoperative estimate of upgrading risk could improve counseling and planning. Methods: Retrospective single-center cohort of men with biopsy ISUP 1-3 undergoing radical prostatectomy (2023-2025) with pre-biopsy mpMRI and systematic biopsy. A multivariable logistic regression model using PSA density, number of positive cores, PI-RADS category, clinical T stage (>=cT2b vs <=cT2a), and biopsy ISUP was developed. Discrimination (AUC), calibration (intercept/slope), decision curve analysis, and two prespecified operating points (rule-in specificity >=0.90; rule-out sensitivity >=0.90) were evaluated with 2,000-bootstrap internal validation. Results: Among 166 patients, upgrading occurred in 77 (46.4%). The model achieved an apparent AUC of 0.740; optimism-corrected AUC was 0.712 after 2,000 bootstrap resamples. Calibration was excellent (intercept ~0; slope 1.00). The rule-in cutoff was 0.626 (specificity 0.90; sensitivity 0.416; PPV 0.800), and the rule-out cutoff was 0.255 (sensitivity 0.90; specificity 0.281; NPV 0.781). Conclusions: A parsimonious preoperative model integrating PSA density, biopsy tumor burden, PI-RADS, clinical stage, and biopsy ISUP provides moderate discrimination, strong calibration, and interpretable operating points to support risk stratification for upgrading in biopsy ISUP 1-3 patients.