Head-to-head comparison of PI-RADS v2 and PI-RADS v1

To compare the reproducibility and diagnostic performance of PI-RADS version 2 (v2) and version 1 (v1) for the diagnosis of prostate cancer (PCa) on multiparametric MRI.

This IRB-approved retrospective study included 65 consecutive biopsy-naïve or biopsy-negative patients suspicious for PCa (mean age: 65 years, mean PSA: 10.8ng/ml) who were undergoing MR-guided biopsy after multiparametric 3T prostate MRI (T2w, DWI, DCE). Two independent readers (R1; R2) scored the prostate lesions according to the v2 score and the v1 sum score. Diagnostic measures (sensitivity, specificity, and area under the ROC-curve) were compared for all cases and stratified by location (transitional zone, TZ, peripheral zone, PZ). Inter-reader agreement was assessed by kappa statistics.

Inter reader agreement for v2 and v1 was substantial to almost perfect (kappa v2: 0.71, v1: 0.81). Overall, sensitivity between both readers and methods did not differ (p>0.05). Overall specificity was higher using v1 compared to v2 (R1: p=0.0078, R2: p=0.0313) In the TZ, v2 showed a higher AUC (0.81-0.84) compared to v1 (AUC 0.77-0.78). Here, the sensitivity of v2 (87.5-100%) was higher than that of v1 (75%) while v2 specificity (50%-56.3%) was lower than that of v1 (68.8-75%). In the PZ, AUCs were higher using v1 (AUC 0.82-0.83) compared to v2 (AUC 0.61-0.63). The specificity for v1 was higher (43.8-62.3%) than that for v2 (12.5-18.8%) while both v2 and v1 achieved 100% sensitivity.

PI-RADS v2 and v1 inter-reader agreement is excellent, but their diagnostic performance differs. While v2 appears to be the preferable method for the evaluation of TZ lesions, v1 performs better in the PZ.