Impact of 68Ga-PSMA PET/MRI on the Accuracy of MRI-Derived Grading Systems for Predicting Extraprostatic Extension in Prostate Cancer

Abstract

Objectives: Accurate preoperative staging and prediction of extraprostatic extension (EPE) are critical for optimal surgical planning in prostate cancer (PCa). This study evaluated the diagnostic accuracy of 68Ga-PSMA PET for EPE assessment, compared it with the standardized multiparametric MRI (mpMRI)-derived EPE-grading system, and examined whether integrating semi-quantitative PSMA PET parameters improves diagnostic performance using hybrid PET/MRI. Methods: This retrospective, single-center study included treatment-na & iuml;ve, biopsy-proven PCa patients who underwent 68Ga-PSMA-11 PET/MRI followed by radical prostatectomy. Diagnostic accuracy was assessed for clinical variables (PSA, ISUP grade), mpMRI features, mpMRI-derived EPE-grading system, visual PET findings, and semi-quantitative PET parameters (SUVmax, SUVmean, PSMA-tumor volume [PSMA-TV]). Optimal cut-offs were determined using the Youden index. Multivariate logistic regression and receiver operating characteristic (ROC) analyses were performed to compare the predictive value of clinical, mpMRI, or PET-derived variables, with histopathology as the reference standard. Results: Forty-five patients were included; EPE was histologically confirmed in 19 (42.2%). Predictors of EPE included capsular irregularity, neurovascular bundle asymmetry, curvilinear contact length >= 1.5 cm, seminal vesicle invasion, tumor size >= 14.25 mm, EPE grade >= 2, ISUP grade >= 3, overt EPE on PET, SUVmax >= 13.84, SUVmean >= 7.20, and PSMA-TV >= 1.40 cm3. The highest ROC performance (AUC = 0.890) was achieved by combining overt EPE on PET, SUVmax, and PSMA-TV. Incorporating PET parameters or tumor size into the EPE-grading system improved predictive accuracy. Conclusions: PSMA uptake in the primary tumor is an independent predictor of EPE. Integrating PSMA PET with mpMRI may provide additional information for preoperative EPE assessment.