Reproducibility and concordance of 4 clinically developed programmed death-ligand 1 (PD-L1) immunohistochemistry (IHC) assays in triple-negative breast cancer (TNBC)

Background Atezolizumab (an anti-PD-L1 antibody) has shown clinical activity in patients with metastatic TNBC who have PD-L1 expression on tumour-infiltrating immune cells (IC). We analysed the performance of 4 PD-L1 IHC assays for PD-L1 IC expression in TNBC.

Methods Thirty TNBC specimens were selected from a set of 107 based on PD-L1 IC expression per VENTANA SP142, to represent the distribution of PD-L1 IC-positivity in the pivotal atezolizumab studies. Serial histologic sections were stained with VENTANA SP142 and SP263, and DAKO 22C3 and 28-8, per manufacturer protocols. Slides were blinded for both assay and sample information and scored by trained readers at 7 sites for PD-L1 IC expression (% per tumour area), by online virtual microscopy.

Results Adjusted means of PD-L1 IC staining ranged from 3.7 % to 7.8 %; SP263 stained more IC than the other assays. Pairwise comparison of adjusted means showed small, non-significant differences (-1.2 % to 0.6 %) between SP142, 22C3 and 28-8, but a significant increase in PD-L1 staining for SP263 vs. the other assays (3.0 % to 4.2 %). Intra-class correlations for the assays showed moderate (0.460) to excellent (0.805) reader concordance.

Conclusions This first multicentre analytical PD-L1 assay comparison study in TNBC indicates good-to-high reproducibility and concordance of PD-L1 IC expression between the SP142, 22C3 and 28-8 assays, while higher expression was detected with SP263. Hence, SP142, 22C3 and 28-8 may be considered analytically interchangeable for PD-L1 IC testing.

Previously presented at ESMO 2019, 359P, Abstract Noske A et al. - Reused with permission.

Publication History

Article published online:
16 July 2020