Biomarker analysis from the phase 3 CheckMate 9ER trial of nivolumab + cabozantinib v sunitinib for advanced renal cell carcinoma (aRCC).

Abstract

608 Background: Prior studies investigated determinants of anti–PD-L1 ± anti-angiogenic response in aRCC. The mechanisms underlying response to first-line anti–PD-1 + VEGF-targeted therapy (anti-VEGF) remain largely unknown. In this exploratory post hoc analysis we conducted an analysis of anti–PD-1 (N) + anti-VEGF (C) v S by assessing potential predictive biomarkers for efficacy in CheckMate 9ER. Analyzed biomarkers included CD8 T-cell frequency (% of total tumor cells, CD8%), CD8 topology, PD-L1 immunohistochemistry, gene set enrichment analysis (GSEA), and 7 gene expression signatures (GES). Methods: Progression-free survival (PFS) and overall survival (OS) were evaluated by tumor PD-L1 expression (< 1% or ≥ 1%), CD8% (low, medium, high by tertiles), and CD8 topology phenotype (desert, excluded, inflamed), and assessed for association using Kaplan–Meier (KM) methods with log-rank test (PD-L1 and CD8), and Cox proportional hazard (Cox PH) models (CD8) (to avoid arbitrary or biased categorization of continuous-valued predictor variables, a potential limitation of KM analyses). Pre-ranked GSEA was performed to assess enrichment for hallmark gene sets using all genes ranked by interaction effect estimates derived using Cox PH regression. Seven GES (Angio, Myeloid, Teff, TIS, IFN-γ, EMT8, Javelin), including several previously found to be predictive of anti-PD-L1 ± anti-VEGF outcomes, were assessed for association with PFS within treatment arms. Results: At 44 mo median follow-up, median PFS and OS were improved with N+C v S regardless of PD-L1 status. PD-L1 < 1% v ≥ 1% was associated with longer median PFS in the S arm only ( P = 0.00011). In KM analyses, higher CD8% was associated with improvements in PFS with N+C, but not S. Of the 410 patients (pts) in the CD8 topology analysis, the predominant CD8 phenotype was inflamed (44.6%), then desert (38.5%) and excluded (16.8%). CD8 topology supported an association between the inflamed phenotype and improved survival outcomes with N+C v S (PFS, P < 0.0001; OS, P = 0.0016). However, these associations were not confirmed in Cox PH models. Common hallmark gene sets with positive enrichment (with false discovery rate < 0.05) in genes associated with longer PFS and OS with N+C v S included oxidative phosphorylation, hypoxia, adipogenesis and P53 pathway. Pts receiving N+C had longer median PFS in the high IFN-γ GES tertile ( P = 0.041) and shorter median PFS in the high EMT8 tertile ( P= 0.013). However, all 7 GES tested, including IFN-γ and EMT8, were not predictive for N+C outcomes in Cox PH models. Conclusions: In this exploratory post hoc analysis, biomarkers previously found to be predictive of anti-PD-L1 ± anti-VEGF outcomes, including established GES, were not predictive of efficacy with anti-PD-1 + anti-VEGF (N+C) using Cox PH models. This suggests that key determinants of response to anti–PD-1 v anti–PD-L1 therapies may differ. Clinical trial information: NCT03141177 .