Prediction of immune checkpoint inhibitor (ICI) response in patients with urothelial cancer (UC) by a novel RNA-based ICI response signature (ICI-PRS).

e16502 Background: Immune checkpoint inhibition (ICI) shows great promise, including approval of pembrolizumab, nivolumab and avelumab in urothelial cancer (UC). However, the overall percentage of UC patients with clinical benefit is relatively small, despite biomarker use to select patients (e.g., PD-1 expression and TMB). Recently, several accelerated approvals (atezolizumab and durvalumab) were subsequently withdrawn following failure to show improved survival compared to standard of care (SOC) treatment in Phase 3 +/- biomarker use. To address the need for better ICI biomarkers, we developed an RNA-based predictive response signature (ICI-PRS) with the primary objective of being prognostic for ICI-treated patients, but not for SOC treatment (Shepherd et al, 2024 ASCO GU). ICI-PRS uses a novel ICI-specific survival signature with tumor-derived and immune associated features to evaluate the likelihood of responding and having longer survival with ICI therapy independent of non-ICI therapy. Herein we report the initial clinical validation of the ICI-PRS using independent datasets of patients with UC treated with ICI. Methods: RNA expression and clinical data were retrospectively evaluated in two independent cohorts of patients with UC treated with ICI: Bladder Cancer Advocacy Network (BCAN) UC-GENOME cohort (dbGaP phs003066.v1.p1 and cBioPortal (n=105) [PMID36333289]) and BACI cohort (Gene Expression Omnibus (GEO;GSE176307 (n=84 for patients treated with >1 cycle) [PMID: 34294892])). ICI-PRS was applied and an optimal cutoff value to designate binary response prediction (complete response + partial response (PR/CR) vs. stable disease + progressive disease (SD/PD)) was derived with Youdin’s method using the receiver operator characteristic (ROC) curve of the original IMVigor210 training set. Results: Samples predicted to be ICI-PRS positive show consistent associations with overall survival (OS) when receiving ICI treatment. A lower hazard ratio (HR) was found for both BACI [HR (95% CI): 0.455 (0.24-0.86); and BCAN 0.540 (0.30-0.98)]. Response to ICI therapy was also associated with ICI-PRS scores, as patients achieving either CR or PR had significantly lower ICI-scores than those with SD or PD in both BACI (p=0.001) and BCAN (p=0.012). Conclusions: Herein we describe initial clinical validation of ICI-PRS in two independent cohorts of UC patients treated with ICI. It provides ICI-treatment specific prognostic and predictive value for patients with UC and may have application alone or in combination with other signatures when determining treatment sequencing or rational combination strategies. These results support further evaluation of the ICI-PRS and potential development as a diagnostic test for selecting UC patients most likely to benefit from ICI therapy either alone or in combination with non-ICI therapy.