Introduction: We aim to estimate the positive predictive value (PPV) of Prostate Imaging Reporting & Data System (PI-RADS) score to identify any prostate cancer (PCa) and clinically significant prostate cancer (csPCa). Further, we sought to analyze cases of PI-RADS 5 discordance with biopsy pathology to develop a quality improvement protocol.
Methods: We retrospectively reviewed the Duke Prostate Cancer (PCa) registry for men who had a 3T multiparametric magnetic resonance imaging (mpMRI) of the prostate and underwent fusion biopsy by a single surgeon from 11/2015 – 3/2021. PI-RADS version 2-2.1 scores for all mpMRI lesions were compared to corresponding targeted biopsy results. The PPV of PI-RADS lesions was calculated for all PCa and csPCa, defined as Gleason Grade Group (GGG) > 2. The subset of discordant cases (PI-RADS 5 lesions with benign or GGG 1 pathology) were re-reviewed by a radiologist with expertise in prostate mpMRI to determine opportunities for quality improvement.
Results: A total of 524 targeted lesions were identified in 392 patients undergoing fusion biopsy of which a PI-RADS score of 5, 4, and 3 was assigned to 121 (23.1%), 310 (59.2%), and 90 (17.2%) lesions, respectively. The PPV of PI-RADS 5, 4, and 3 for all PCa and csPCa was 0.80, 0.55, 0.24, and 0.63, 0.33, and 0.09, respectively. Detection of csPCa was significantly higher for PI-RADS 5 lesions (62.8%) versus PI-RADS 4 lesions (33.2%) (p < 0.001). Still, PIRADS 5, discordant biopsy results were identified for 45 (37.2%) lesions. On mpMRI re-review, 18 (40.0%) lesions were considered “resolved” discordances as their suspicion was downgraded because either the lesion was > 1.5 cm alone without further abnormality (n=13), low signal central zone was mistaken for PCa (n=2), or the mpMRI was generally misread (n=3). The remaining 27 discordant cases had the level of suspicion confirmed on re-review and were classified as “unresolved.” After excluding “resolved” discordant cases, the PPV of PI-RADS 5 lesions improved to 0.74.
Conclusions: Although PIRADS 5 lesions are considered high risk for csPCa, a diagnostic dilemma occurs when targeted biopsy returns benign or low grade cancer. In our cohort, a proportion of discordances were "resolved" by mpMRI re-review. Further work is needed to determine the value of re-biopsy in cases of “unresolved” discordance. We found re-review of PI-RADS 5 discordance events to be a source of potential quality improvement as the review of these cases allows for the identification of those who may benefit from re-biopsy versus close surveillance.