PD-L1 immunohistochemical expression in bladder urothelial cancer with SP263, SP142 and 22C3 antibodies: A comparative study

Programmed death ligand 1 (PD-L1) is currently the only biomarker used for the selection of patients with bladder urothelial cancer for immunotherapy. Several platforms, antibodies and scores are currently available for the evaluation of the expression of PD-L1 in immunohistochemistry (IHC). In this study three different antibodies (SP263, SP142 and 22C3) were compared to establish their performances and concordance rates. Twenty-four consecutive cases of surgically resected urothelial cancers of the bladder were enrolled. All cases were revised, and appropriate tumor areas were selected for IHC. Three commercially available PD-L1 antibodies were tested: 22C3 pharmDx with Dako Autostainer Link 48 (Dako, Carpinteria, Ca), and SP263 and SP142 with the Ventana BenchMark (Ventana Medical Systems, Tucson, AZ) platform. All slides were evaluated by an expert pathologist and both the tumor proportion score (TPS) and the combined positive score (CPS) were determined and compared at two different cut-off levels (>= 1 and >= 10). The SP263 and 22C3 clones produced more positive results with the CPS and TPS scores, respectively. The CPS score identified more positive cases than the TPS score, irrespectively of the clone or the cut-off used; the difference was statistically significant in both the SP263 and SP142 clones with the >= 1 cut-off. No statistically significant differences were found between the clones when the >= 1 cut-off was used, irrespectively of the score. At the contrary, a statistically significant difference (p = 0.024) and a trend to significance (p = 0.082) were respectively found for the TPS and CPS scores, when the SP22C3 and the SP142 clones were compared at a cut-off level of >= 10. The ICC test using CPS was 0.676 and 0.578 for the >= 1 and >= 10 cut-offs respectively, and 0.729 and 0.467 respectively for the same cut-offs using TPS. This suggests that the three antibodies under investigation cannot be used interchangeably, especially the 22C3 and SP142 clones which showed statistically significant difference when TPS was tested at a >= 10 cut-off.