Intercenter reproducibility of software-based fusion biopsies for grade group prediction when targeting suspicious MRI lesions

Abstract

PurposeTo assess the intercenter reproducibility of software-based fusion targeted biopsy (TB) for grade-group assessment and pretherapeutic evaluation of highly suspicious MRI lesions. Patients and methodsIn this study, were included 380 consecutive patients who underwent radical prostatectomy (RP) after prostate cancer diagnosis and a prebiopsy MRI showing Prostate Imaging-Reporting and Data System (PIRADS) score 4 or 5 lesions. All patients underwent systematic biopsies (SB) combined with software-based fusion TB in the 2 centers. Biopsies were only performed by expert urologists or radiologists in a contemporary time frame. The primary endpoint was the center difference of concordance/upgrading rates between biopsy and RP specimens. ResultsPathological features on biopsy and RP specimens were significantly different among centers with more unfavourable disease in center 1. The rate of TB upgrading was 33.6% in center 1 vs. 35.4% (P = 0.860) in center 2. Grading concordance was also comparable among centers (50.0% vs. 47.1%) as well as the SB upgrading rate. Regression analysis did not find any baseline characteristics (Age, prostate-specific antigen, MRI lesions, center) predictive for TB upgrading. These findings were achieved by using fewer TB per lesion in center 1 (2.3 vs. 5.0, P < 0.001), at the expense of more SB cores (14.4 vs. 8.5, P < 0.001). The influence of MRI characteristics (lesion size and number, PIRADS score) on upgrading rates was consistent among centers. ConclusionsSoftware-based fusion TB technique leads to comparable outcomes in terms of grade group prediction accuracy in PIRADS 4 to 5 lesions, insignificant between centers, in spite of different non imaging-based aggressiveness features.