Measurement Of Spatial And Antibody-Based Pd-L1 Heterogeneity In Non-Small Cell Lung Cancer.

9040 Background: Assessment of PD-L1 is not yet standardized and is based largely on “assay performance” rather than protein measurement; with each assay requiring its own specialized platform and antibody. Here we focus on PD-L1 measurement using five monoclonal antibody clones (SP142, E1L3N, 9A11, SP263 and 28-8). We analyzed concordance between antibodies and then determined the heterogeneity between readers and tissue samples. Methods: We designed a PD-L1 index tissue microarray including 30 NSCLC cases with full dynamic range of PD-L1 expression, tissue and cell line controls. The concordance between the antibodies was assessed using diaminobenzidine (DAB) and quantitative immunofluorescence (QIF). Additionally, the correlation among 5 pathologists was assessed on 35 cases of resected NSCLC using SP142 DAB-stained samples on 3 blocks from the same tumor from each case. Results: Correlations between the antibodies ranged predominantly from 0.8-0.9 for tumor cores and 0.83-0.97 (R2) for cell line cores. The 28-8 antibody was excluded due to lack of reproducibility. Among the 5 pathologists a 93.2% intra-class correlation coefficient (ICC) was seen using the percent of tumor cells with membrane or cytoplasmic signal from DAB staining while the ICC was only 19.5% for immune cell staining. The ICC for tumor staining between the 3 blocks of each cases was 93.7% while the ICC between blocks for immune cell staining was 77.7%. A mixed effects model for the QIF assessment of PD-L1 tumor expression showed that the variance between fields of view accounted for 88% of total variance while the variance between blocks only represents 12%. Conclusions: Detection of PD-L1 by a range of antibodies shows high level concordance, suggesting that differences seen in expression is not attributable to the antibodies used. Heterogeneity studies showed that the variance seen is largely between fields of view within the area of the tumor block and that the block to block variance is low (12%). This data suggests that assays that measuring PD-L1 using different antibodies from different regions of patient tumors are likely to be concordant, while demonstrating the microscopic (field to field) appearance of heterogeneity.