118P Artificial intelligence (AI)-powered spatial analysis of tumor-infiltrating lymphocytes (TILs) as a predictive biomarker for anti-PD-1 in advanced biliary tract cancer (BTC)

Recently, anti-PD-1/L1 plus gemcitabine and cisplatin (GemCis) has demonstrated significant survival benefits in randomized phase 3 trials for advanced BTC. However, no biomarker has been established for anti-PD-1 in BTC. Here, we assessed TILs using artificial intelligence (AI)-powered spatial analysis in advanced BTC treated with anti-PD-1. An AI-powered whole-slide image (WSI) analyzer (Lunit SCOPE IO, Lunit, Seoul, Korea) was used to segment tumor epithelium and stroma, and identification of intratumor TIL (iTIL) and stromal TIL (sTIL). H&E-stained WSI from pre-treatment samples was acquired from Asan Medical Center and Yonsei Cancer Center (n = 354), and a total of 337 samples (95.2%) after quality control were used for the final analysis. Immune phenotypes (IP) were defined as follows: inamed as high iTIL and sTIL; immune-excluded as low iTIL and high sTIL; immune-desert as low TIL overall. Among them, 29 patients (pts) were available for multi-color flow cytometry analysis (FACS) using peripheral blood mononuclear cells (PBMC), collected at baseline and C1D8. All patients were treated with anti-PD-1 monotherapy, and 188 pts (55.8%) were treated as 2nd line. In overall pts, anti-PD-1 showed median overall survival (OS) of 5.7 months (mo), median progression-free survival (PFS) of 2.0 mo, and objective response rates (ORR) of 10.1%. IP were classified as inflamed in 40 pts (11.9%), immune-excluded in 167 (49.6%), and immune-desert in 130 (38.6%). The inflamed group showed better OS (12.5 vs. 5.1 mo, P<0.001), PFS (5.0 vs. 2.0 mo, P=0.001), and ORR (27.5% vs. 7.7%, P<0.001) than non-inflamed groups. In the FACS, the inflamed group showed significantly higher proportion of baseline CD8+Tem and lower baseline CD8+Tnaive than non-inflamed groups. At C1D8 of anti-PD-1, CD69+CD8+T, GZMB+CD8+T, and PRF1+CD8+T cells were increased, while PD1+CD8+T cells was decreased in PBMC of the inflamed group. Immune phenotype classified by AI-powered spatial TIL analysis was effective to predict the efficacy outcomes of advanced BTC pts treated with anti-PD-1 therapy. Further evaluation is required in the setting of anti-PD-1/L1 plus GemCis.