PI-RADS v2 and ADC values: is there room for improvement?

PURPOSE: To determine the diagnostic accuracy of ADC values in combination with PI-RADS v2 for the diagnosis of clinically significant prostate cancer (CS-PCa) compared to PI-RADS v2 alone. MATERIALS AND METHODS: This retrospective study included 155 men whom underwent 3-Tesla prostate MRI and subsequent MR/US fusion biopsies at a single non-academic center from 11/2014 to 3/2016. All scans were performed with a surface coil and included T2, diffusion-weighted, and dynamic contrast-enhanced sequences. Suspicious findings were classified using Prostate Imaging Reporting and Data System (PI-RADS) v2 and targeted using MR/US fusion biopsies. Mixed-effect logistic regression analyses were used to determine the ability of PIRADS v2 alone and combined with ADC values to predict CS-PCa. As ADC categories are more practical in clinical situations than numeric values, an additional model with ADC categories of ≤ 800 and > 800 was performed. RESULTS: A total of 243 suspicious lesions were included, 69 of which were CS-PCa, 34 were Gleason score 3+3 PCa, and 140 were negative. The overall PIRADS v2 score, ADC values, and ADC categories are independent statistically significant predictors of CS-PCa (p < 0.001). However, the area under the ROC of PIRADS v2 alone and PIRADS v2 with ADC categories are significantly different in both peripheral and transition zone lesions (p = 0.026 and p = 0.03, respectively) Further analysis of the ROC curves also shows that the main benefit of utilizing ADC values or categories is better discrimination of PI-RADS v2 4 lesions. CONCLUSION: ADC values and categories help to diagnose CS-PCa when lesions are assigned a PI-RADS v2 score of 4.

Evaluating the performance of PI-RADS v2 in the non-academic setting.

PURPOSE: To evaluate the utility of PI-RADS v2 to diagnose clinically significant prostate cancer (CS-PCa) with magnetic resonance ultrasound (MR/US) fusion-guided prostate biopsies in the non-academic setting. MATERIALS/METHODS: Retrospective analysis of men whom underwent prostate multiparametric MRI and subsequent MR/US fusion biopsies at a single non-academic center from 11/2014 to 3/2016. Prostate MRIs were performed on a 3-Tesla scanner with a surface body coil. The Prostate Imaging Reporting and Data System (PI-RADS) v2 scoring algorithm was utilized and MR/US fusion biopsies were performed in selected cases. Mixed effect logistic regression analyses and receiver-operating characteristic (ROC) curves were performed on PI-RADS v2 alone and combined with PSA density (PSAD) to predict CS-PCa. RESULTS: 170 patients underwent prostate MRI with 282 PI-RADS lesions. MR/US fusion diagnosed 71 CS-PCa, 33 Gleason score 3+3, and 168 negative. PI-RADS v2 score is a statistically significant predictor of CS-PCa (P < 0.001). For each one-point increase in the overall PI-RADS v2 score, the odds of having CS-PCa increases by 4.2 (95% CI 2.2-8.3). The area under the ROC curve for PI-RADS v2 is 0.69 (95% CI 0.63-0.76) and for PI-RADS v2 + PSAD is 0.76 (95% CI 0.69-0.82), statistically higher than PI-RADS v2 alone (P < 0.001). The rate of CS-PCa was about twice higher in men with high PSAD (≥0.15) compared to men with low PSAD (<0.15) when a PI-RADS 4 or 5 lesion was detected (P = 0.005). CONCLUSION: PI-RADS v2 is a strong predictor of CS-PCa in the non-academic setting and can be further strengthened when utilized with PSA density.