PD-L1 immunohistochemistry in clinical diagnostics: Inter-pathologist variability is as high as assay variability.

Abstract

e20637 Background: Several checkpoint inhibitors targeting the PD1 axis are now approved for the treatment of non-small cell lung cancer (NSCLC). The best predictor for therapy response is PD-L1 expression on tumor cells assessed by immunohistochemistry (IHC). However, the PD-L1 assays have been developed independently for each available drugs. In this study, we evaluated the different assays to determine the comparability between the assays and the concordance between pathologists under conditions reflecting the clinical situation. Methods: We tested four IHC assays (28-8, 22C3, SP263 and SP142) and one in-house protocol (28-8 on the Ventana system) on a tissue microarray (TMA) including 55 resected tissue specimens of NSCLC patients. The TMAs were independently annotated by seven pathologists with regard to the percentage of tumor cell membrane staining using a six-grade scale ( < 1%; 1-≤5%; 5%-≤10%; 10-≤25%; 25-≤50%; ≥50%). Results: The staining scores were comparable for the three assays with clones 28-8, 22C3 and SP263 as well as the in-house 28-8 protocol (κ-value: 0.75-0.91). The use of the SP142 clone resulted in a lower percentage of positive tumor cells and positive cases (κ: 0.44-0.63). When a reference score was defined based on the majority of scores, the proportion of positive cases (≥1%) were 38% for 28-8, 29% for 22C3, 42% for SP263, 37% for 28-8 in-house and 16% for SP142%. With ≥1% cut-off 3-8 of the 55 cases (5-15%) were differently classified as positive or negative when comparing any two assays excluding SP142. Analysing the scoring of the seven pathologists a clear inter-rater variability (κ: 0.71-0.96) was observed independent from the assay. With a ≥1% cut-off, 20% of cases were annotated differently by at least one pathologist. Conclusions: The concordance between the three antibody clones 28-8, 22C3 and SP263 was acceptable and was in the similar range as inter-rater variability of different pathologists. The results suggest that assessment of tumor cell staining for Nivolumab, Pembrolizumab and Durvalumab might be performed with only one of these assays. The assessment of the pathologist presents an intrinsic source of error that should be considered especially at low PD-L1 scores.