PD-L1 expression in urothelial bladder cancer varies more among specimen types than between companion assays

Urothelial bladder cancer (UBC) patients ineligible to platinum-based chemotherapy can be treated with immune-checkpoint inhibitors (ICI) in Programmed Death Ligand 1 (PD-L1) positive cases. Although concordance exists between different PD-L1 assays, little is known on PD-L1 expression variability in matched UBC samples. We compared PD-L1 expression in whole slides of matched transurethral resections (TURBT), radical cystectomies (RC), and lymph node metastasis (LN). Immunohistochemistry using the VENTANA PD-L1 (SP263) assay was performed on 115 patients and scored positive if expression occurred in ≥25% immune cells (IC), ≥25% tumour cells (TC), or both. PD-L1 was positive in 42.7% TURBT, 39.8% RC, and 27.3% LN specimens. Concordance was moderate (κ=0.52; P<0.001) between TURBT and RC, and fair between LN and TURBT (κ=0.31; P=0.048) or RC (κ=0.25; P=0.075). Comparison with the VENTANA PD-L1 (SP142) assay which had been performed previously on the same cohort showed moderate to substantial inter-assay agreement (κ=0.42–0.66). Although TC staining is not part of the SP142 scoring algorithm, discordant PD-L1 assay outcome could be attributed to SP263 TC≥25% staining in only 41% of cases. These results show that PD-L1 expression variability between matched specimens is higher than that between individual assays. Optimal specimen determination for PD-L1 testing needs to be addressed in future studies. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s00428-021-03094-6.