Effects of "real life" prostate MRI inter-observer variability on total needle samples and indication for biopsy.

Abstract Purpose Prostate multiparametric magnetic resonance imaging (mpMRI) improves diagnosis of clinically significant cancer and reduces over-detection of nonsignificant cancer. Disagreement in the interpretation of mpMRI readings is well-known, with a reported discrepancy rate of 10% to 42%. We report the clinical repercussions of this variability on prostate biopsy candidates. Materials and Methods Medical records of patients referred from 11 medical centers for MR-guided prostate biopsy (MRGpB) between October, 2017 and January, 2019 were retrospectively analyzed. Patients with at least one prostate imaging reporting and data system (PI-RADS) 3 or greater prostate lesion were selected, and the mpMRI studies (all read by others) were reviewed by our prostate mpMRI reader. Outcomes included changes in PI-RADS score and the subsequent effect on total needle samples and indication for biopsy. Results Eighty-two patients with 128 lesions were suitable for analysis (mean age 66.5 ± 7.1 years, mean PSA 6.8 ± 8.5 ng/ml). Nine (11%) patients had suspicious rectal exams (T2a). Following our prostate mpMRI reader's imaging revisions, the PI-RADS score was downgraded in 66 (52%) lesions, upgraded in 15 (12%), and unchanged in 47 (37%), leaving a total of 84 suspected lesions (kappa = 0.17). Biopsy was deferred in 22 (27%) patients, and an estimated 136.4 (34.4%) samples were avoided (P = 0.0001 for both). There was a trend toward prostate size to correlate with imaging revision and abortion of biopsy (P = 0.06) while enrollment in active surveillance correlated with proof from such outcome (P = 0.007). Conclusion These data suggest that high interobserver disagreement in prostate mpMRIs from diverse institutes significantly affects prostate biopsy practice. The clinical consequences of this discord are significant.