Abstract
Programmed death-ligand 1 (PD-L1) expression on tumor cells (TC) or tumor-infiltrating immune cells (IC) correlated in several studies with PD-L1/programmed death-1 (PD-1) checkpoint inhibitor efficacy. Since June 2018, a positive PD-L1 status is required for atezolizumab or pembrolizumab treatment of patients with advanced or metastasized urothelial bladder cancer, who are ineligible for cisplatin-containing therapy. We examined technical comparability and inter-reader agreement of four clinically developed PD-L1 assays in locally advanced disease. Archived, formalin-fixed, paraffin-embedded sections from 30 patients (73.3% cystectomies, 26.7% transurethral resections) were stained by PD-L1 immunohistochemistry using VENTANA SP142, VENTANA SP263, DAKO 22C3, and DAKO 28-8 at two sites per manufacturers' protocols and scored blinded at five sites for PD-L1 expression on IC (% per tumor area) and TC (%). Small, non-significant inter-assay differences were observed for IC. For TC, SP142 showed significantly lower staining percentages. Pairwise comparisons revealed - 0.3 to 1.6% differences in adjusted means between assays for IC, and for TC, - 10.5 to - 7.8% (SP142 versus others) and - 1.9 to 2.7% (other comparisons). Inter-reader and inter-assay agreement was moderate to high for both IC and TC. Allocation to binary cutoffs (1%, 5%, 10%) showed substantial to high Kappa agreement scores (0.440-0.923) for IC and TC between assays for each reader. This first multicenter study, with five independent readers blinded with respect to the assay used, suggests that all four currently clinically relevant assays are analytically similar for evaluation of PD-L1-stained IC and three (SP263, 22C3, and 28-8) for PD-L1-stained TC. Inter-observer agreement for trained readers in scoring of both IC and TC positivity was generally high.