In this collaborative study by the Russian Society of Clinical Oncology and the Russian Society of Pathology, we assessed the concordance among three validated, commercially available PD-L1 immunohistochemistry assays for patients with urothelial cancer. Tumors from 100 urothelial cancer patients were stained with the antibody clones 22C3 (Agilent), SP142 (Ventana Medical Systems), and SP263 (Ventana Medical Systems), which are used in clinical trials of second-line therapy with checkpoint inhibitors. Four trained pathologists independently evaluated the percentages of tumor cells (TC) and tumor-infiltrating immune cells (IC) that were stained at any intensity by each of the antibodies. The test-specific cutoffs for the proportions of stained cells in a positive sample were pre-specified as TC + IC ≥ 10% or TC ≥ 10% for 22C3, IC ≥ 5% for SP142, and TC ≥ 25% or IC ≥ 25% for SP263. Three hundred immunohistochemistry slides were scored. The percentages of PD-L1 staining in the three assays without using any cutoff were higher in the IC than in the TC (55% versus 24% for 22C3, 45% versus 8% for SP142, and 72% versus 27% for SP263, respectively). The Pearson correlation coefficients for anti-PD-L1 staining in the IC were 0.5, 0.69, and 0.85 with 22C3/SP142, 22C3/SP263, and SP142/SP263, respectively. The Pearson correlation coefficients for PD-L1 staining in the TC were 0.93, 0.99, and 0.91 for the same pairs. Among the patients who were negative for PD-L1 staining by one test, 91-100% were also negative by the other tests. Among the patients who were positive by one test, 43-100% were also positive by the other tests. Our data indicate that repeated testing can be avoided as a patient with urothelial cancer who is classified as negative for PD-L1 expression by one of the three single tests using the corresponding cutoff rule is highly likely (91-100%) to be classified as negative by either of the other tests.
Keywords: Concordance analysis; Immunohistochemistry diagnostic assays; PD-L1 expression.