PD-L1 Testing in Gastric Cancer by the Combined Positive Score of the 22C3 PharmDx and SP263 Assay with Clinically Relevant Cut-offs

Purpose We provide a comparison between 22C3 pharmDx and SP263 assay, for evaluating programmed death ligand 1 (PD-L1) expression in advanced gastric cancer (GC) patients. Materials and Methods The PD-L1 immunohistochemistry by 22C3 pharmDx and SP263 assays was performed in the center of the tumor (CT) and invasive margin (IM) in 379 GC tissues using tissue microarrays and interpreted as combined positive score (CPS) and tumor proportion score (TPS). Of the total samples, 55 samples were independently reviewed by five pathologists. Results The two assays showed a high correlation in both the CPS and TPS. At a CPS >= 1 cut-off, 219 (57.8%) and 231 (60.9%) GCs were positive for PD-L1 with the 22C3 and SP263 assays, and at >= 10 cut-off, 37 (9.8%) and 36 (9.5%) GCs were positive, respectively. The overall percent agreement (OPA) was greater than 90% with CPS >= 1 and >= 10 cut-offs, and TPS >= 1% and >= 10% cut-offs. There was higher OPA between the two assays with a CPS cut-off >= 10 (99.2%) than >= 1 (94.7%). The percent agreement between the CT and IM was higher with a CPS cut-off >= 10 (92.9%) than >= 1 (77.6%). Patient with positive expression at CPS >= 5 cut-off had a significantly better outcomes in both assays. Interobserver variability among five pathologists was higher than the assay variability. Conclusion Two assays for PD-L1 expression in GC showed high agreement. These results provide guidance for selecting eligible patients with GC for pembrolizumab treatment.