A retrospective study of PD-L1 immunohistochemistry for hepatocellular carcinoma

Background Programmed cell death ligand 1 (PD-L1) expression plays an important role in selecting patients with hepatocellular carcinoma who will benefit from anti-programmed cell death 1 (PD-1)/PD-L1 monotherapy and directing those with lower levels to alternative treatments. Methods In total, 156 hepatocellular carcinoma tumors were stained with three PD-L1 immunohistochemistry assays (SP142, 28-8, and E1L3N). Two junior pathologists and one senior pathologist evaluated the pathological slides and recorded the percentages of tumor and immune cells stained at any intensity. Results Analytical comparisons demonstrated that the anti-PD-L1 assay SP142 is a significantly stronger reagent that stains a higher percent of both tumor and immune cells, while 28-8 and E1L3N behave similarly to each other. The correlation coefficients of the three assays ranged from 0.77 to 0.92 for tumor cells and from 0.66 to 0.75 for immune cells. Next, 1% and 5% tumor cell staining cutoffs were designated using various combinations of assays and cutoffs. Across all three reagents, 14 samples showed concordance above the 5% threshold, eight showed concordance within the range of 1% to 5% interval, and 93 showed concordance below the 1% threshold. The remaining 41 samples showed a combination of discordant results across all three reagents. High PD-L1 expression in tumor and immune cells tended to not be recurrent. Conclusions The anti-PD-L1 assays 28-8 and E1L3N were closely aligned with regard to staining tumor and immune cells, whereas SP142 showed higher percentages of staining for both cell types. All three assays indicated greater variability in immune cell staining than in tumor cells.