Optimized PD-L1 scoring of gastric cancer

BACKGROUND: PD-1/PD-L1-Immunotherapy has been approved for gastric carcinoma. PD-L1 assessment by immunohistochemistry is the principle biomarker. Are biopsies able to map the actual PD-L1 status of the entire tumor? METHODS: Whole tumor slides of 56 gastric carcinoma were analyzed to determine the distribution of PD-L1 positive cells in the entire tumor areas. Tissue micro arrays with four cores of the tumor surface, which represents the endoscopically accessible biopsy zone, were built from the same tumors. The PD-L1 CPS value was determined separately for each core. Preoperative diagnostic biopsies were available for 22 of the tumors. PD-L1 prevalence, sensitivity and specificity were analyzed using the whole tumor slides as reference scores. Molecular subtyping was performed and related to the PD-L1 status. RESULTS: 27.3% of cases were PD-L1 negative (CPS < 1), 43.6% showed low PD-L1 expression (CPS ≥ 1 to < 5), 12.7% moderate (CPS ≥ 5 to < 10) and 16.4% strong expression (CPS ≥ 10). The biopsies showed best test characteristics if four surface biopsies were analyzed combined, i.e., the CPS was calculated across all four biopsies. The prevalence showed a distribution similar to the resection specimens, sensitivity was 0.73 and specificity 1.0. Using fewer surface biopsies decreased sensitivity and specificity and caused false-negative classifications. Compared to the TMAs, the preoperative biopsies showed reduced sensitivity (0.412). CONCLUSIONS: This is the first comprehensive study to optimize PD-L1 assessment in gastric cancer using endoscopically available tissue. The obtained PD-L1 prevalence is consistent with data of current clinical studies. Calculation of the test characteristics shows that surface biopsies can be indicative of the true PD-L1 status based on the resection specimen. However, an adequate number of biopsies is required. In this study, n = 4 biopsies yielded best results. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s10120-021-01195-4.