Comprehensive assessment of PD-L1 immunohistochemistry on paired tissue and cytology specimens from non-small cell lung cancer

Objectives: PD-L1 staining assessed by immunohistochemistry (IHC) is a predictive biomarker used to select advanced stage non-small cell lung carcinoma (NSCLC) patients who are likely to respond to PD-1/PD-L1 inhibitors. Cytology specimens represent a significant percentage of the diagnostic samples and additional data are required to show that they provide reliable PD-L1 results when compared to tissue specimens. We aimed to compare PD-L1 staining obtained from patient-matched tissue and cytology specimens. We also want to assess the feasibility of PD-L1 testing on cell blocks with two assays by evaluating the intra- and inter-observer agreement and the level of difficulty for determining the percentage of stained tumor cells (TPS).

Materials and methods: Forty-six patients with NSCLC were selected. Each patient provided a surgical specimen and a cytology sample (cell block) and/or a biopsy at diagnosis. PD-L1 staining using Agilent PD-L1 IHC 28-8 pharmDx and VENTANA PD-L1 (SP263) assays was evaluated by four pathologists using the TPS. Sixty slides were rescored to document intra-observer agreement. Pathologists were asked to score the level of difficulty for evaluating PD-L1 TPS for each slide. Fleiss's and Cohen's kappas (k) were used to assess the agreement between paired specimens as well as intra- and inter-observer agreement.

Results: The concordance in PD-L1 TPS between cell blocks and surgical specimens (k varying from 0.56 to 0.82) or biopsies (k from 0.43 to 0.81) was moderate to substantial, depending on the cut-off. On cell blocks, inter-observer agreement was substantial (k from 0.74 to 0.82) and intra-observer agreement was almost perfect (k from 0.85 to 0.93). The perceived difficulty of PD-L1 evaluation of cell blocks was not different from surgical specimens but more difficult than biopsy samples.

Conclusion: PD-L1 TPS was concordant between cell blocks and tissue specimens, mainly at 10, 25 and 50 % cut-offs. PD-L1 evaluation on cell blocks was feasible and reproducible between different observers and assays.