Prebiopsy multiparametric MRI and PI-RADS version 2.0 for differentiating histologically benign prostate disease from prostate cancer in biopsies: A retrospective single-center comparison

Abstract

PurposeTo investigate the diagnostic performance of Prostate Imaging-Reporting and Data System version 2.0 (PI-RADSv2.0) for differentiating clinically significant prostate cancer (csPCa) from benign prostate disease on prebiopsy multiparametric MRI stratified by total prostate specific antigen (PSA) concentration. Materials and methods150 patients who had prebiopsy mpMRI, serum PSA concentration and subsequent biopsy were retrospectively analyzed. Patients were stratified by PSA concentration (Group1 ≥ 10 ng/mL; Group2 4.0–<10 ng/mL). MRI findings were assessed using PI-RADSv2.0 by two blinded radiologists. Lesions were graded histopathologically using the International Society of Urological Pathology (ISUP) score. Diagnostic performance of PI-RADSv2.0 was evaluated and compared to PSA and PSA Density (PSAD). The performance of the radiologists was compared including inter-observer agreement for PI-RADSv2.0. The correlation between imaging and histopathological biopsy results was analyzed. ResultsThe differences in total PSA, free/total PSA ratio and PSAD between benign (n = 78) and malignant (n = 72) groups were significant (p < 0.05). The PI-RADSv2.0 scores of the radiologists were strongly correlated (r = 0.912, p < 0.001) with excellent agreement, κ = 0.97 (95%CI: 0.90–1.03; p < 0.005). Receiver operating characteristics curve analysis showed significantly high predictive power for PI-RADSv2.0, total PSA and PSAD alone. Comparison of age, prostate volume, PSAD, free/total PSA ratio and total PSA values between ISUP1 and ISUP ≥ 2 cases revealed significantly increased PSAD (p < 0.001) and total PSA (p = 0.001) in the ISUP ≥ 2 group. ConclusionPI-RADSv2.0 had high diagnostic accuracy in both PSA groups. PI-RADSv2.0, PSAD and total PSA alone had significant high predictive power to detect csPCa. However, the combination of PI-RADSv2.0 and PSAD or total PSA for each reader showed no statistically significant improvement when compared to PI-RADSv2.0 alone.