Use of genomic dedifferentiation signature to predict clinical outcomes with doublet vs triplet therapy in the phase III COSMIC-313 trial.

Abstract

523 Background: The Genomic Dedifferentiation Signature (GDS) has previously been shown to differentiate clinical outcomes to immune checkpoint inhibitor (ICI)-based combinations and anti-VEGFR monotherapy in patients with metastatic renal cell carcinoma (mRCC; Salgia et al. Cancer Cell. 2025). COSMIC-313 is a randomized phase III trial evaluating cabozantinib-nivolumab-ipilimumab (C/N/I) vs nivolumab-ipilimumab (N/I) in mRCC. Here, we evaluated the prognostic and predictive utility of the GDS in COSMIC-313. Methods: Bulk transcriptomic profiling (RNAseq) was obtained for 398 patients (pts) enrolled in COSMIC-313 and used to assign GDS scores by single sample gene set enrichment analysis. Pts were classified as GDS high or GDS low based on median GDS expression. Clinicopathologic variables and objective response rates (ORR) were compared using Wilcoxon Rank-Sum and Fisher’s Exact Tests. Progression-free survival (PFS) and overall survival (OS) analyses employed Kaplan-Meier estimates and Cox proportional hazard models. Results: In the ITT population (both trial arms), GDS scores were significantly enriched in pts with IMDC poor vs intermediate risk (p = 0.0097) and in those with vs without sarcomatoid features (p = 2.2x10 -5 ). ORR did not significantly differ between GDS high and GDS low pts treated with C/N/I (48% vs. 58%; p = 0.21), though a non-significant improvement in ORR amongst GDS high vs. GDS low pts was seen with N/I (50% vs. 35%; p = 0.057). Among GDS low pts, C/N/I yielded a significantly higher ORR than N/I (58% vs. 35%; p = 0.0019), with no difference between arms in GDS high pts (48% vs. 50%; p = 0.98). PFS was not significantly different between GDS high and GDS low , pts in the ITT, C/N/I, and N/I cohorts, with HRs of 1.0 (95% CI: 0.81–1.3, p = 0.75), 1.2 (95% CI: 0.81–1.7, p = 0.4), and 0.88 (95% CI: 0.62–1.3, p = 0.48), respectively. OS was significantly worse in GDS high pts in the ITT population (HR: 1.3, 95% CI: 1.0–1.8, p = 0.043), with non-significant differences in the C/N/I and N/I arms (HRs: 1.4 and 1.2, p = 0.14 and 0.36, respectively). Conclusions: Front-line management of mRCC currently relies on ICI-ICI and VEGFR-ICI combinations, though no biomarker yet exists to determine which patients differentially benefit from each strategy. These results not only reinforce the negative prognostic capacity of the GDS, but also support the potential predictive utility of the GDS, suggesting that GDS low patients disproportionately benefit from the addition of VEGFR-directed agents to ICI-based therapy for the first-line treatment of mRCC.