Determinants of variability of five programmed death ligand-1 immunohistochemistry assays in non-small cell lung cancer samples

Abstract

Programmed death ligand-1 (PD-L1) expression as determined by immunohistochemistry (IHC) is potentially predictive of clinical outcome. The aim of this study was to assess the concordance of reported PD-L1 IHC assays and investigate factors influencing variability. Consecutive sections from 20 non-small cell lung cancers (NSCLCs) comprising resection, core biopsy, cytology and pleural fluid samples underwent IHC with 5 different antibody/autostainer combinations: 22C3/Link48, 28-8/BOND-MAX, E1L3N/BOND-MAX, SP142/BenchMark and SP263/BenchMark. PD-L1 RNA levels were assessed using RNAscope. The frequency of positive cases using scoring thresholds from clinical trials was 72%, 33%, 61%, 56%, and 33% for the 5 IHC protocols respectively, and 33% for RNAscope. Pairwise agreement on the classification of cases as positive or negative for PD-L1 expression ranged from 61%-94%. On a continuous scale, the lowest correlation was between 28-8/BOND-MAX and SP142/BenchMark (R2=0.25) and highest was between 22C3/Link48 and E1L3N/BOND-MAX (R2=0.71). When cases were ordered according to tumor cell (TC)%, a similar ranking of cases across IHC protocols could be observed, albeit with different quanta and limits of detection. Single-slide OPAL 7-color fluorescence IHC analysis revealed a high degree of co-localization of staining from the 5 PD-L1 antibodies. Using SP142 antibody in a BOND-MAX protocol led to increased TC% quanta, while retaining a similar ranking of samples according to TC%. The results of this study highlight tumor PD-L1 status can vary significantly according to IHC protocol. Protocol-dependent staining intensities and nominated thresholds for positivity contribute to this variability, while the antibody used appears to be less of a factor.

Response to Editorial: The PD-L1 Immunohistochemistry Biomarker: Two Steps Forward, One Step Back?

“…[3][4][5][6] Most studies in this field have found that 10% to 20% of cases will be classified differently as above or below clinical cutoffs, depending on the assay used. 3,4,[7][8][9] We agree that our data are "not greatly different from those in the Blueprint Phase 1 study." 1(para.5) However, we could not interpret these data and conclude, as other authors have, that the SP263, 22C3, and 28-8 assays are interchangeable.…”

External quality assessment demonstrates that PD‐L1 22C3 and SP263 assays are systematically different

“…Taking the CAP guidelines into account this does not support the conclusion from the authors of ‘analytical evidence for interchangeability of the 22C3, 28‐8 and SP263 assays’. Remarkably, other studies reported that SP263 had higher PD‐L1 scores than 22C3 in several cases or low correlation between these two assays , while Kim and colleagues show more positive cases in 22C3 than in SP263 . Interestingly, when looking at individual cases around the threshold of 1% PD‐L1 positivity, comparison of 22C3 with SP263 shows in some cases higher staining for SP263 and in other cases higher for 22C3 .…”

Patterns of TIGIT Expression in Lymphatic Tissue, Inflammation, and Cancer

“…Details on the used antibodies, antibody retrieval procedures, and OPAL dyes are given in Table 1. The experimental procedure was performed according to the manufacturer's instructions [23]. Slides were initially boiled in a microwave (15 minutes at 100°C) for antigen retrieval.…”