Comparison of three FDA-approved diagnostic immunohistochemistry assays of PD-L1 in triple-negative breast carcinoma

The Dako 28-8, Dako 22C3, and Ventana SP142 assays are among the approved programmed death ligand 1 (PD-L1) immunohistochemical companion/complementary diagnostics associated with cancer treatment. To address the concordance of these assays in triple-negative breast cancer (TNBC), we examined PD-L1 expression in 98 TNBC tumors and compared the positive rates using the three assays and three scoring methods: immune cell (IC), tumor cell (TC), and combined tumor cell and immune cell (TCIC) (an equivalent to combined positive score, or CPS). The positive rate for PD-L1 expression with a 1% cutoff was highest with 28-8, followed by the 22C3. These two assays demonstrated almost perfect or substantial agreement in all three scores. There was less agreement between SP142 and the other assays. Using the IC score or the TCIC score at a 1% cutoff (CPS 1), 4% of tumors were positive for PD-L1 with SP142 but negative with the other assays. Using SP142 with a 1% cutoff as a reference, the optimal cutoff for best agreement was at 1% for IC, 30% for TC, and 2% for TCIC (CPS 2) with the other two assays. A 2% cutoff for the 22C3 TCIC (CPS 2) yielded the best agreement with SP142 1% IC cutoff (kappa 0.65). Our study showed the lowest positive rate with SP142 among the three assays. However, the other two assays were not able to identify all tumors that would test positive with SP142 using IC or TCIC/CPS. It is unlikely to achieve high agreement between SP142 and the other two assays by changing the analytical cutoffs. (C) 2020 Elsevier Inc. All rights reserved.

Automated summary

The study compared the concordance of three PD-L1 assays in TNBC, finding that 28-8 and 22C3 had higher positive rates and good agreement in scoring methods, while SP142 showed lower concordance with the other assays. When using SP142 as a reference, optimal agreement was achieved by selecting specific cutoff values for IC, TC, and TCIC/CPS.