Comparison of 22C3 PD-L1 Expression between Surgically Resected Specimens and Paired Tissue Microarrays in Non–Small Cell Lung Cancer

The extent to which intratumoral heterogeneity of programmed death ligand 1 (PD-L1) expression causes discordance of PD-L1 expression between paired samples remains unclear. Here, PD-L1 status was compared between whole sections from NSCLCs and the corresponding tissue microarrays (TMAs) serving as surrogate biopsy specimens.PD-L1 expression was evaluated by 22C3 immunohistochemistry assay on 190 archival surgical specimens and matched to the TMA results. PD-L1 expression was determined by the tumor proportion score (TPS) and classified as TPS lower than 1%, TPS of 1% to 49%, and TPS of 50% or higher. Agreement statistics were used.The percentage of PD-L1 expression on tumor cells differed greatly between individual TMAs and matched surgical specimens. When PD-L1 TPS was adopted, a total of 36 of 190 discordance cases (18.9%) were observed, with a κ-value of 0.630 between paired samples. The TMAs underestimated or overestimated PD-L1 status in 19 of 36 (52.8%) and 17 of 36 (47.2%) of the matched surgical specimens, respectively (p = 0.118). The discordance rate was much lower in cases with a PD-L1 TPS lower than 1% compared with in cases with a TPS of 1% to 49% and TPS of 50% or higher (18.4% versus 56.7% and 43.3%, p0.001). When a TPS of 50% or higher was used as the cutoff, the discordance rate of PD-L1 TPS less than 50% was further reduced to 7.5%. Such discrepancies were due mainly to intratumoral heterogeneity of PD-L1 expression and nonsignificant association with clinicopathological features.PD-L1 expression in TMAs correlates moderately well with that in the corresponding surgical specimens, indicating that evaluating PD-L1 expression in diagnostic biopsy specimens could be misleading in defining sensitivity to pembrolizumab treatment yet may be reliable as a way to exclude patients with a PD-L1 TPS less than 50% from first-line pembrolizumab treatment.