126 Development of a visual scoring method to determine inflamed, excluded or desert immune phenotype in non-small cell lung cancer, colorectal cancer, and urothelial cancer

<h3>Background</h3> The induction of an anti-tumor immune response is an essential element of effective checkpoint immunotherapy and requires recruitment of functional immune cell populations to the tumor microenvironment to interact directly with tumor cells. Thus, characterization of spatial distributions of inflammatory cells relative to tumor cells can be used as a biomarker to identify patients likely to respond to immunotherapy.¹ A reproducible scoring method is needed to capture the heterogeneity of an immune cell infiltrated tumor as well as provide the inflammatory status of that tumor. Here, a visual scoring approach to identify desert, excluded, and inflamed immune phenotypes in non-small cell lung cancer (NSCLC), colorectal cancer (CRC), and urothelial cancer (UC) samples is described with the aim to create a reproducible method for pathologists to identify significant patterns of tumor inflammation. <h3>Methods</h3> Commercial FFPE NSCLC, CRC and UC resection samples were stained by H&amp;E and dual pan-CK/CD8 immunohistochemistry. Stained slides were scanned and whole slide images were analyzed using an in-house developed Visiopharm algorithm to determine the number and density of CD8 cells in the tumor stroma and tumor cell nests. In a first iteration of the method, pathologists estimated the percentage of each compartment that had no, or a low, moderate and high density of CD8 cells. Results were reported in four scoring bins for each of the two compartments. The method was simplified to report the percent of tumor stroma and tumor cell nests with high density CD8 cell infiltration. For both methods, the desert, excluded and inflamed immune phenotypes are determined by comparison of the proportion of tumor stroma and tumor cell nests with high-density CD8 cell infiltration. <h3>Results</h3> The density of CD8+ cells in each tumor type covered a dynamic range from 8 to 3500 cells/mm² in the tumor area. Visual scoring of CD8+ cells was performed in two tumor compartments, tumor associated stroma and tumor cell nests. The results of intra- and inter-pathologist scoring and immune phenotyping of NSCLC (n = 84), CRC (n = 82) and UC (n = 71) samples will be presented for eight category and four category methods with a concordance assessment of desert, excluded and inflamed phenotypes. <h3>Conclusions</h3> A simplified four category method can improve robustness and pathologist concordance for both CD8 cell density scores as well as resultant immune phenotypes. We conclude that the presented approach is a reproducible method to identify desert, excluded and inflamed immune phenotypes in the tested indications. <h3>Reference</h3> Li X, Eastham J, Giltnane JM, Zou W, Zijlstra A, Tabatsky E, Banchereau R, Chang CW, Nabet BY, Patil NS, Molinero L, Chui S, Harryman M, Lau S, Rangell L, Waumans Y, Kockx M, Orlova D, Koeppen H. Automated tumor immunophenotyping predicts clinical benefit from anti-PD-L1 immunotherapy. <i>J Pathol</i> 2024 Jun;<b>263</b>(2):190-202. <h3>Ethics Approval</h3> This study included commercially procured biobank specimens that were collected and stored under the Compendium Biobank FAGG (fagg-afmps.be) guidelines and policies.