Diagnostic performance of PI-RADS and PSA density for detecting clinically significant prostate cancer

Multiparametric magnetic resonance imaging (mpMRI), interpreted using the Prostate Imaging Reporting and Data System (PI-RADS), is increasingly used to improve prostate cancer detection and reduce unnecessary biopsies. However, its diagnostic accuracy compared with histopathological confirmation remains variable across institutions. This study aimed to evaluate the correlation between mpMRI findings and TRUS-guided prostate biopsy results in detecting clinically significant prostate cancer (csPCa) in our patient cohort.

This retrospective diagnostic accuracy study included 100 biopsy records (86 unique patients) who underwent mpMRI followed by TRUS-guided biopsy. mpMRI findings were scored using PI-RADS v2, while biopsy histopathology served as the reference standard. Clinically significant prostate cancer (csPCa) was defined as ISUP Grade Group ≥ 2. Diagnostic performance was assessed for two interpretive rules: (1) Baseline rule: PI-RADS ≥ 4 as positive; (2) Combined rule: PIRADS ≥ 4 or PI-RADS = 3 with PSA density (PSAD) > 0.15 ng/mL/mL. Sensitivity, specificity, predictive values, and area under the RoC curve (AUC) were calculated. Hierarchical logistic regression assessed the independent contribution of PSAD and clinical covariates.

Malignant cases showed higher PSA (median 10.0 ng/mL vs 7.0 ng/mL) and PSAD (0.32 vs 0.13 ng/mL/mL) and smaller prostate volumes (36.5 mL vs 61.0 mL) compared with benign cases. csPCa detection increased with rising PI-RADS category (3.7% for PI-RADS 3, 56.9 % for PI-RADS 4-5). At the patient level, the Baseline rule achieved sensitivity = 86.7% and specificity = 66.1%, while the Combined rule increased sensitivity to 90.0 % with specificity = 55.4%. The ordinal PI-RADS score demonstrated excellent discrimination (AUC = 0.826, 95% CI 0.713-0.924). In logistic regression, adding PSAD improved model AUC from 0.836 to 0.888 (p < 0.001), and inclusion of age and prostate volume further increased AUC to 0.900 (p = 0.044). within PI-RADS 3 lesions, the optimal PSAD threshold (youden index) was 0.163 ng/mL/mL, yielding 100% sensitivity and 76% specificity. Postbiopsy complications were within expected ranges, with mild hematuria (29%), minor rectal bleeding (23%), and UTI (7%) being most common.

mpMRI findings strongly correlated with histopathological outcomes from TRUS-guided biopsy. Incorporating PSA density significantly enhanced the diagnostic accuracy for csPCa, particularly in equivocal PI-RADS 3 cases. Combining mpMRI and PSAD can refine patient selection for biopsy and improve early detection of clinically significant prostate cancer.