Interpretation According to Clone-Specific PD-L1 Cutoffs Reveals Better Concordance in Muscle-Invasive Urothelial Carcinoma

Because immune checkpoint inhibitors have been approved for treating advanced urothelial carcinoma (UC), programmed death-ligand 1 (PD-L1) immunohistochemistry (IHC) assays have been widely used as companion or complementary diagnostic tests for predicting treatment outcomes. Because different clones, scoring algorithms, and cutoffs have been used for interpretation, this study investigated the variation, correlation, and concordance of four validated PD-L1 clones (SP142, SP263, 22C3, and 28-8) and proposed a practical solution for the harmonization of PD-L1 IHC. A tissue microarray, including 46 muscle-invasive UCs, was constructed for PD-L1 testing with the four clones. Tumor cell (TC) and immune cell (IC) expression was analyzed. SP142 had significantly low TC expression, whereas SP263, 22C3, and 28-8 exhibited a moderate correlation (rho ≥ 0.6), with almost perfect concordance (intraclass correlation coefficient > 0.8) in TC expression. Fair to moderate correlation and concordance were observed in IC expression in most pairwise comparisons of clones. Substantial concordance (kappa > 0.6) was noted when high PD-L1 expression was defined by applying clone-specific cutoffs to each clone. Our findings imply that a universal cutoff value is not feasible for UC; we propose that PD-L1 IHC assays for UC should be interpreted according to a clone-specific scoring algorithm and cutoff value.