Utility of pre- and post-pembrolizumab (Pembro) Vesical Imaging-Reporting and Data System (VIRADS) to predict the pathological response in muscle-invasive urothelial bladder cancer (MIBC): An analysis of the PURE-01 cohort.

552 Background: The possibility to predict the pathologic complete response (pT0) or downstaging (pT≤1) after neoadjuvant therapy may have profound impact on the management of MIBC and orient next-generation bladder-sparing trials. The VIRADS is a standardized reporting system that uses mpMRI parameters to predict the probability of MIBC. No studies have analyzed the ability of VIRADS to predict the pT0 or pT≤1 response post-immunotherapy (IO). Methods: In PURE-01 patients (pts) were staged with bladder multiparametric magnetic resonance imaging (mpMRI: T2-weighted imaging, diffusion-weighted imaging, dynamic contrast enhancement) before and after treatment (3 cycles of pembro) prior to radical cystectomy (RC). All mpMRI scans were centrally reviewed. Logistic regression models analyzed pre- and post-pembro VIRADS against pT≤1 (primary endpoint) and pT0 (secondary endpoint). VIRADS scores were dichotomized between 0-3 and 4-5. Covariates included cT-stage, age, gender, PD-L1 combined positive score (CPS) and tumor mutational burden (TMB). Results: In total, 58 pts were had centrally-reviewed MRI scans (N=116 mpMRI), treated between 02/17 and 08/18. Demographic characteristics and outcomes were generally similar between the all-treated and VIRADS-evaluable populations of PURE-01. Median age was 65 years, 52 (89%) had pure/predominant urothelial carcinoma (UC) histology, 25 (43.1%) had cT3-4N0 MIBC. Pre-pembro: 8 pts (13.8%) had no measurable disease (VIRADS=0), 20 (34.5%) a VIRADS 1-3 score, and 30 (51.7%) had a VIRADS 4-5 score. Six pts (10.3%) had a downstage from VIRADS 4-5 to VIRADS 0-3 post-pembro. Both pre-pembro and post-pembro VIRADS 0-3 scores were significantly associated with pT≤1 endpoint on multivariable analyses (MVA): the strongest effect was seen with post-pembro VIRADS 0-3 against pT≤1 response (OR: 30.2, 95%CI: 6.2-223.2, p<0.0001). The AUC of this model was 0.92. Regarding pre-pembro VIRADS 0-3: OR: 4.35, 95%CI: 1.1-19.7, p=0.04; AUC: 0.83. CPS was another significant variable for pT≤1 endpoint only in MVA using pre-pembro VIRADS 0-3 (OR: 1.02, 95%CI: 1-1.05, p=0.02). Post-pembro VIRADS 0-3 was also associated with pT0 on MVA (OR: 4.59, 95%CI: 1.2-19.6, p=0.02; AUC: 0.78), whereas pre-pembro VIRADS was not (p=0.13). Mean apparent diffusion coefficients (ADC) were similar between pre- and post-pembro lesions (0.86 vs 0.87). Conclusions: To our knowledge, this is the first evidence establishing the predictivity of the VIRADS score towards the pathological downstaging, both in the pre and post IO settings. Post-pembro VIRADS, along with the combination of pre-pembro VIRADS and CPS, emerged as the strongest features based on which selecting pts for bladder-sparing strategies. Clinical trial information: NCT02736266 .