Concordance between PD-L1 assays 28-8 and SP263 and their respective scoring algorithms in procured upper gastrointestinal adenocarcinoma samples

• Published in: Journal of clinical oncology Vol. 43; no. 4_suppl; p. 479
• Main Authors: Liu, Yihua, Sharpnack, Michael, Lum, Jenifer, Koralek, Daniel O., Das Thakur, Meghna
• Format: Journal Article
• Language: English
• Published: American Society of Clinical Oncology 01.02.2025
• ISSN: 0732-183X; 1527-7755
• DOI: 10.1200/JCO.2025.43.4_suppl.479

479Background: PD-L1 is being studied as a predictive immunohistochemical (IHC) biomarker for gastroesophageal junction (GEJ), gastric (GC), and esophageal (EAC) adenocarcinomas, and anti-PD-(L)1 therapies are used to treat them. Here, we compare the VENTANA PD-L1 (SP263) IHC assay and its tumor area positivity (TAP) algorithm with the clinically validated Agilent PD-L1 IHC 28-8 pharmDX combined positive score (CPS) algorithm in GEJ, GC, and EAC samples. The degree of concordance between SP263 TAP ≥ 5% and 28-8 CPS ≥ 5 is important for interpreting real-world settings and emerging clinical trial data. Methods: GEJ, GC, and EAC tumor samples procured from company-sponsored clinical trials (NCT02545504, NCT02862535, NCT02864381) and external vendors with informed consent were analyzed 4 ways: SP263 TAP, SP263 CPS, 28-8 TAP, and 28-8 CPS. Readouts were evaluated by a single pathologist and were analyzed with Pearson correlation coefficients. Results: 286 patient samples were analyzed (GEJ, n = 106; GC, n = 88; EAC, n = 92). The agreement among the SP263 TAP ≥ 5% and 28-8 CPS ≥ 5 prevalence is shown in the Table. The overall correlation coefficient between 28-8 CPS and SP263 TAP was high (R = 0.95; P < 2.2 × 10−16). In addition, when applying the same scoring algorithm to both assays (SP263 TAP vs 28-8 TAP or SP263 CPS vs 28-8 CPS), the correlation coefficients remained high (R = 0.97; P < 2.2 × 10−16). When both algorithms were applied to the same IHC assay (SP263 TAP vs SP263 CPS or 28-8 TAP vs 28-8 CPS), the correlation coefficients were also high (R = 0.98; P < 2.2 × 10−16). Conclusions: In this controlled experiment the PD-L1 assays 28-8 and SP263 concordance was high. The scoring algorithms for GEJ, GC, and EAC samples were observed as highly correlated; minor differences were likely driven by the distinct PD-L1 monoclonal antibody clones utilized in the IHC assays. These data suggest that the 2 assays are comparable for evaluating PD-L1 expression at the TAP ≥ 5% and CPS ≥ 5 cutoffs. Clinical trial information: NCT02545504, NCT02862535, NCT02864381. Concordance of SP263 TAP ≥ 5% with 28-8 CPS ≥ 5.Positive percentage agreement: 0.96Negative percentage agreement: 0.95Overall percentage agreement: 0.9528-8 CPS ≥ 5: Positive28-8 CPS ≥ 5: NegativeSP263 TAP ≥ 5%: Positive1169SP263 TAP ≥ 5%: Negative5156