Key points are not available for this paper at this time.
4573 Background: EV is used for treatment-refractory aUC with activity also noted in a frontline trial. Independent biomarkers associated with EV treatment outcomes are lacking. Methods: UNITE is a multi-site retrospective study of pts treated with EV and other agents. In this analysis we included pts treated with EV monotherapy and available next generation sequencing (NGS) data. Assessed biomarkers included genomic alterations (alts) in ≥5% of pts, tumor mutation burden (TMB) and PD-L1 expression. Pts were evaluable for observed response rate (ORR) if they had scans after ≥1 EV cycle. Biomarkers were evaluated individually in univariate analyses (UVA) and in separate multivariate analyses (MVA) which accounted for clinical variables. Progression-free survival (PFS) and overall survival (OS) from EV start were assessed using Kaplan Meier method and cox proportional hazard (cph) modeling. ORR was assessed using logistic regression. Results: Among 303 pts with NGS treated with EV monotherapy at 16 US sites, 207 had TMB and 146 had PD-L1 data. Median age was 70; 218 (72%) men, 262 (87%) White, 196 (66%) with pure urothelial histology, 212 (71%) with lower tract primary tumor, 204 (67%) had ≥2 prior therapies. ORR was 53% (140/264); mPFS was 6.0 mos, mOS was 13.1 mos. In UVA, longer OS was associated with higher Hgb, albumin and GFR (>30cc/min), lower neutrophil to lymphocyte ratio, older age, and absence of liver mets. MVA included these variables along with histology as covariates. In multivariate cph models, better outcomes were associated with alts in ERBB2, KDM6A and PIK3CA, while inferior outcomes with low TMB (<10 Mut/Mb) and high PD-L1 (CPS ≥ 10). Conclusions: MVA of this large, multi-site, retrospective cohort of aUC pts, identified several putative biomarkers independently associated with EV treatment outcomes. Longer OS was associated with ERBB2 and KDM6A alts, low PD-L1 and high TMB. Limitations include lack of central scan review, pt selection and confounding biases. These findings require external validation, while their underlying mechanisms merit translational study. Table: see text
Jindal et al. (Thu,) studied this question.