Abstract 4969: Investigating digital pathology tumor microenvironment features for immunotherapy outcome in patients with advanced non-small cell lung cancer

Abstract

Abstract Introduction The composition of the tumor microenvironment (TME) has been shown to influence response to immunotherapy (IO). In this study, we investigated digital pathology TME features of IO outcomes among patients with non-squamous, non-small cell lung cancer (NSCLC) within a real-world dataset. Methods From the nationwide de-identified Flatiron Health-Foundation Medicine NSCLC clinico-genomic database, a cohort of 50 EGFR/ALK-negative, non-squamous NSCLC patients treated with first-line pembrolizumab monotherapy (mono-IO) or pembrolizumab in combination with carboplatin/cisplatin and pemetrexed (chemo-IO) that had available whole slide images of H&E-stained tissue resection specimens were included. The de-identified data originated from approximately 280 US cancer clinics (~800 sites of care). AI-powered TME characterization models developed by PathAI (Boston, MA; commercially available as PathExplore™) were deployed on scanned H&E slides to extract a panel of cell- and tissue-level human interpretable features (HIFs). Cox proportional hazards regression was used to identify digital pathology features or HIFs associated with overall survival. Results 17 HIFs were associated with better survival and 19 HIFs were associated with worse survival among patients treated with mono-IO (p < 0.05). Lymphocyte features were among the HIFs associated with favorable survival, including density of lymphocyte cells in cancer epithelium (hazard ratio (HR) per unit standard deviation (SD) = 0.46 [0.24, 0.88], p-value = 0.02) and proportion of lymphocyte cells relative to total immune cells in cancer epithelium (HR per unit SD = 0.61 [0.39, 0.95], p-value = 0.03). Cell-level features associated with worse survival included density of fibroblast cells in stroma (HR per unit SD = 1.51 [1.01, 2.25], p-value = 0.05). These associations remained after adjusting for tumor mutational burden (TMB) (N = 29 low, 21 high) and PD-L1 status (N = 0 negative, 1 low-positive, 17 high-positive, 32 unknown). Additionally, similar associations were not observed among patients treated with chemo-IO. High lymphocyte density was associated with better survival compared to low lymphocyte density in mono-IO treated patients (adjusted HR per unit SD = 0.36 [0.18, 0.74], p-value = 0.01) but not in chemo-IO treated patients (adjusted HR per unit SD = 1.10 [0.76, 1.60], p-value = 0.6). Conclusion These results indicate that the composition of TME assessed via digital pathology may have utility in identifying NSCLC patients that will respond to first-line immune checkpoint inhibitors beyond the established IO biomarkers. Citation Format: Ericka M. Ebot, Kuei-Ting Chen, Mikayla Biggs, Douglas I. Lin, Julia A. Elvin, Garrett M. Frampton, James J. Pao. Investigating digital pathology tumor microenvironment features for immunotherapy outcome in patients with advanced non-small cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4969.