Does Tumor Volume Assessed by Cumulative Cancer Location Predict Grade Reclassification

RPE4 Does Tumor Volume Assessed by Cumulative Cancer Location Predict Grade Reclassification on Active Surveillance in the MRI Era? S. Fletcher, M. Mamawala, A. Holler, Z. Su, Y. Bhanji, C. de la Calle, C. Pavlovich Johns Hopkins Brady Urological Institute, Baltimore, MD, USA Introduction and Objective: Cumulative cancer location (CCLO) has previously been shown to be associated with grade reclassification (GR) during active surveillance (AS) in men without prostate MRI (Erickson et al, Eur Urol Onc 2018). Given the variability in interpreting metrics such as number of positive cores and maximum percentage of cancer in a core, we aimed to determine the ability of CCLO to predict GR in the MRI era. Methods: We identified patients enrolled in AS between 2011 and 2021 with Grade Group (GG) 1 disease who underwent prostate MRI. We developed an “MRI-CCLO” (mCCLO) score by summing the total number of uniquely involved sextants positive for cancer on both diagnostic and confirmatory biopsy, with an additional point for an MRI with at least one PI-RADS 2 lesion. Men were stratified into low (1-2) and high (3) mCCLO risk groups. The primary outcome was GR to GG2 on subsequent biopsies. Kaplan-Meier analysis was used to compare GR rates between mCCLO risk groups. Using multivariable analyses, we compared performance of a base model (age, confirmatory biopsy year, race, PSA density, and highest PI-RADS score) with the base model plus either mCCLO, number of positive cores, or maximum percentage of cancer in a core. Results: Among a total of 310 patients, the high mCCLO group had significantly higher rates of GR compared to the low mCCLO group (Figure). Each model had comparable discriminative ability (c-indices, 95% CI: mCCLO [0.68, 0.63-0.74], number of positive cores [0.68, 0.63-0.73], maximum percentage of cancer in a core [0.67, 0.61-0.72], base model [0.66, 0.60-0.72]). Conclusions: The mCCLO score is comparable to traditional biopsy metrics in predicting GR but may offer greater reproducibility and less variability in interpretation