LAUSR

BACKGROUND Recently, the Food and Drug Administration (FDA)/European Medicines Agency (EMA) restricted first-line use of atezolizumab and pembrolizumab in patients with metastasised urothelial carcinoma by defining distinct programmed cell death ligand-1 cut-offs. We analysed the diagnostic performance of all FDA/EMA-approved programmed cell death ligand-1 assays with emphasis on new restrictions for first-line treatment with atezolizumab and pembrolizumab. PATIENTS AND METHODS Two hundred fifty-one urothelial carcinomas were analysed on tissue microarrays with four cores of each tumour. Stains were performed in certified laboratories on Ventana Benchmark Ultra and Dako Link 48 autostainers. Stains were read on an assay-by-assay basis by two trained pathologists. Overall percentage agreement (OPA) was calculated across the preset cut-offs. Positive percentage agreement (PPA) and negative percentage agreement (NPA) were calculated across different scoring algorithms. Venn diagrams were constructed to illustrate discordance according to the recent FDA/EMA guidelines. RESULTS The Dako 28-8, 22c3 and the Ventana SP263 assays showed high interassay correlation (r-range 0.83-0.91). Interassay correlation between the Ventana SP142 and the three other assays was moderate (r-range 0.66-0.75). OPA of 93.3% was achieved between the Dako 28-8, 22c3 and Ventana SP263 assays. OPA including the SP142 was 84.1%. Pooled PPA and NPA of different scoring algorithms was 89.4% and 95.3% for the Dako 28-8, 22c3 and the SP263 assays, respectively. With the SP142 assay, pooled PPA was 59.1%. The SP142 assay identifies fewer eligible patients for first-line treatment with atezolizumab/pembrolizumab. CONCLUSION Dako 28-8, 22c3 and SP263 assays show interchangeable performance. The SP142 assay shows divergent staining results. Interassay variability leads to different detection rates of eligible patients for first-line treatment with atezolizumab and pembrolizumab.