Limited reproducibility of PI-RADS v2.1 upgrading rule subcategorization in MRI-visible peripheral zone lesions: A lesion-level targeted-biopsy study.

PURPOSE: To evaluate inter-reader reproducibility of PI-RADS v2.1 upgrading rule subcategorization in prelocalized MRI-visible peripheral zone (PZ) lesions and assess its association with lesion-level clinically significant prostate cancer (csPCa) yield at MRI-targeted biopsy.

METHODS

This retrospective single-center lesion-level study included 310 men with 375 MRI-visible PZ lesions who underwent 3-T multiparametric MRI followed by MRI-TRUS fusion-targeted biopsy. Three radiologists independently rescored prelocalized lesions using PI-RADS v2.1. PI-RADS 4 lesions were classified as upgraded when assigned DWI score 3 with positive dynamic contrast-enhanced (DCE) imaging, or non-upgraded when assigned DWI score 4 regardless of DCE status. csPCa was defined as ISUP grade group ≥ 2. Inter-reader agreement was assessed, and generalized estimating equations were used to account for within-patient clustering.

RESULTS

csPCa was present in 176/375 lesions (47 %). Inter-reader agreement was moderate-to-substantial for overall PI-RADS categorization, but only fair-to-moderate for PI-RADS 4 subcategorization (Gwet's AC1 range, 0.29-0.57). Across readers, csPCa yield ranged from 39 % to 68 % for upgraded PI-RADS 4 lesions and from 63 % to 81 % for non-upgraded lesions. The magnitude and significance of this subgroup contrast varied by reader.

CONCLUSION

In this multireader, lesion-level, targeted-biopsy cohort, subcategorization of the PI-RADS 4 upgrading rule showed limited reproducibility. Although nonupgraded lesions generally showed higher csPCa yield, subgroup-specific estimates were reader-dependent, supporting cautious interpretation of this split.