Clinicopathologic and genomic correlates of tumor-infiltrating immune cells and immunotherapy efficacy in NSCLC.

Abstract

9121 Background: Tumor-infiltrating immune cells and PD-L1 expression are associated with improved clinical outcomes in patients (pts) with NSCLC treated with immune checkpoint inhibitors (ICIs). However, as tumor-infiltrating immune cells are not a well-established biomarker for NSCLC, further data are needed to integrate and identify clinicopathological and genomic factors that influence the tumor microenvironment. Methods: We collected clinicopathologic and genomic data from pts with NSCLC who underwent multiplexed immunofluorescence. Uniform Manifold Approximation and Projection (UMAP) was used to identify distinct immunophenotypic clusters according to the number of intratumoral PD-1+ immune cells (ICs), CD8+, and Foxp3+ T cells, as well as PD-L1 on tumor and immune cells. An unbiased recursive partitioning (URP) algorithm was used to investigate an optimal cluster with respect to objective response rate (ORR) in the subset of pts treated with ICIs. Results: Among 304 pts, UMAP identified 5 clusters: PD-L1-high with high vs low CD8+ and PD-1+ ICs (clusters A & B, respectively); PD-L1-low with high vs low CD8+ and PD-1+ ICs (clusters C & D respectively); PD-L1-low and moderate levels of CD8+ and PD-1+ ICs (cluster E). Clinicopathological characteristics of the clusters shown in Table. URP analysis identified immune rich clusters A and C as optimal responders to ICIs. From the start of ICIs, we observed a significantly higher ORR (53.3% vs 4.3%; P<0.001), a significantly longer median progression-free survival (mPFS 25.6 vs 3.7 months; HR: 0.12 [95% CI: 0.05-0.32]; P<0.001), and longer median overall survival (mOS 45.1 vs 22.3 months; HR: 0.25 [95% CI: 0.1-0.68]; P=0.006) in clusters A + C (N=15) vs other clusters (N=23). After adjusting for other variables such as performance status, histology, presence of oncogenic driver mutation, and line of treatment, clusters A + C were significantly associated with improved mPFS (HR: 0.08 [95% CI: 0.03-0.24], P<0.001) and mOS (HR: 0.11 [95% CI: 0.03-0.40], P<0.001). Conclusions: Incorporation of multiplex immunofluorescence may improve prediction of response and resistance to immunotherapy in NSCLC.[Table: see text]