Concordance of PD-L1 combined positive score (CPS) analysis between primary gastric cancer and matched peritoneal metastasis.

e16068 Background: Peritoneum represents one of the most common metastatic sites of gastroesophageal adenocarcinoma (GEA) and is associated with poor prognosis. Programmed death-ligand 1 (PD-L1) expression is a recognized predictive biomarker of response to immune checkpoint inhibitors (ICIs) in the first-line setting. However, data from pivotal trials showed poorer clinical efficacy of ICIs in the sub-set of patients with peritoneal involvement, regardless of PD-L1 score magnitude. Here, we investigated the temporal and spatial heterogeneity of PD-L1 expression between primary tumor and matched peritoneal metastasis. Methods: PD-L1 expression according to CPS was determined by immunohistochemistry using VENTANA PD-L1 (SP263) assays on formalin-fixed paraffin-embedded (FFPE) tumor tissues derived from 22 patients treated at our institution for advanced GEA from September 2015 to August 2023. Concordance rate of PD-L1 expression between primary tumor and matched peritoneal metastasis using PD-L1 CPS cut-off of 1 and 5 was reported (spatial heterogeneity); additionally, concordance between PD-L1 value before and after the administration of systemic treatments was analyzed (temporal heterogeneity). Results: 12(54.5%)/5(22.7%) primary tumor samples and 12(54.5%)/1(4.5%) peritoneal metastases showed PD-L1 CPS ≥ 1/ ≥ 5, respectively. Concordance of PD-L1 CPS between primary and peritoneal specimens was 54.5% and 72.7% according to CPS cuts-off of 1 and 5, respectively, highlighting marked spatial heterogeneity. High concordance rate (94%) was reported between negative PD-L1 CPS primary tumor samples and matched peritoneal metastases, against the cut-off of CPS ≥ 5; in contrast, none of the positive primary tumor samples retained PD-L1 expression in the peritoneum. Evident temporal heterogeneity of PD-L1 expression was also observed, with 56% concordance against CPS ≥ 5 in the pre- versus post-treatment comparison. Conclusions: PD-L1 expression in GEA is characterized by spatial and temporal heterogeneity which is more prominent for lower CPS cut-offs values. This evidence could hamper its role as predictive biomarker for ICIs. The high intra-patient discordance rate of PD-L1 CPS expression between positive primary tumor samples and matched peritoneal metastases suggests that peritoneal specimens should not be recommended as the preferred source for PD-L1 assessment.