Variation in Prostate Magnetic Resonance Imaging Performance: Data from the Prostate Biopsy Collaborative Group.

Background and objective: The quality and reporting of prostate magnetic resonance imaging (MRI) are operator dependent, leading to variations in estimates such as positive predictive value across sites. This impacts patient counseling, risk modeling, and risk calculators. This study assessed variation in Prostate Imaging Reporting and Data System (PI-RADS) score classification and subsequent probability of grade group (GG) ≥2 + prostate cancer.

Methods: Data from the Prostate Biopsy Collaborative Group, including multiple sites in North America, Europe, and Asia Pacific, were analyzed. Patients underwent multiparametric MRI (mpMRI) of the prostate followed by prostate biopsy during the years 2010-2023. Only those with MRI-targeted biopsy and PI-RADS score ≥3 were included. The risk of being assigned PI-RADS 4 or 5 and risk of GG ≥2 disease for these scores were estimated using logistic regression.

Key findings and limitations: The cohort included 7325 biopsies from 7320 unique patients from 13 sites. A two-fold variation in the probability of PI-RADS 4 or 5 assignment across sites persisted even after adjustment for patient risk (heterogeneity p < 0.001 for both). There were significant differences in the absolute risk of GG ≥2 disease for PI-RADS 4 and 5 (heterogeneity p < 0.001 for both), varying between 23% and 68% and between 49% and 87%, respectively. The use of prostate biopsy as a reference standard has limitations but reflects typical usage of mpMRI in clinical practice.

Conclusions and clinical implications: The probability of being assigned PI-RADS 4 or 5 and subsequent detection of GG ≥2 disease varies widely between institutions. This impacts counseling, risk stratification, and clinical practice, necessitating better standardization in the performance and interpretation of mpMRI.