Comparison of four PD-L1 immunohistochemical assays in lung cancer

Introduction: Four different PD-L1 immunohistochemical assays are approved or in development as companion or complementary diagnostics to different immunotherapeutic agents in lung carcinoma. We sought to determine whether these assays are technically equivalent and whether one antibody can be used on an alternate staining platform. Methods: Serial sections of tissue microarrays constructed from 368 cases of resected lung cancer were stained for 22C3 and 28-8 on the Dako Link 48 platform, and for SP142 and SP263 on the Ventana Benchmark Ultra platform, strictly as per product insert. A protocol was developed to use the 22C3 antibody on the Ventana Benchmark Ultra platform. Results: Differences in mean TC and IC staining were observed between the four assays (p<0.001). Differences between 22C3 and 28-8 were not statistically significant. Concordance of TC scores was good (ICC=0.674), particularly when SP142 was excluded as an outlier (ICC=0.755). Highest concordance was seen between 22C3 and 28-8 (ICC=0.812). Concordance was poor for IC staining (ICC=0.212). When dichotomized according to clinically relevant cut-offs, pairwise comparisons showed poor to moderate concordance (k=0.196–0.578), with positive percent agreement ranging from 15.1% to 90.0%. 22C3 performed comparably on the Dako Link 48 platform and the alternate Ventana Benchmark Ultra platform (ICC=0.921, k=0.897). Conclusions: Concordance between the four PD-L1 immunohistochemical assays when performed and scored as intended show, apart from 28-8 and 22C3, they cannot be used interchangeably in clinical practice. A protocol was successfully developed to use 22C3 on an alternate platform, which may help to overcome some barriers to implementation.