Comparative analysis of programmed death ligand 1 expression in paired cytologic and histologic specimens of non–small cell lung cancer

Background: In advanced non–small cell lung cancer (NSCLC), cytologic specimens from transbronchial needle aspiration (TBNA) or transthoracic needle aspiration are often the only cancer tissue material available for the analysis of programmed death ligand 1 (PD-L1) expression. This study was aimed at assessing the concordance of PD-L1 expression in histologic and cytologic samples and at evaluating interobserver agreement on specimens in this setting. Methods: One hundred and thirty-eight specimens from 60 patients with NSCLC were analyzed. Histologic specimens were represented by endoscopic samples obtained with forceps (biopsies), whereas cytologic specimens were from TBNA and bronchial lavage (BL). PD-L1 expression was quantified with the immunohistochemistry (IHC)-based Ventana SP263 assay. For cytologic specimens, IHC was performed on cell block sections. Two independent pathologists who were blinded to the clinical data evaluated partial or complete membrane IHC staining. Concordance between 2 methods and between 2 pathologists was evaluated with normal and weighted Cohen's κ coefficients, overall agreement, and Bland-Altman plots. Results: PD-L1 expression was quantified in 138 specimens from 60 patients. Concordance between cytologic and histologic approaches was moderate (κ = 0.56; weighted κ = 0.55). Also, concordance in the biopsy-TBNA and biopsy-BL subgroups was moderate (κ = 0.43 and κ = 0.47, respectively), whereas interobserver agreement was substantial (weighted κ = 0.72). A Bland-Altman plot showed an underestimation in PD-L1 values from cytologic samples in comparison with histologic ones. Conclusions: The results demonstrate that in the absence of available histologic specimens, PD-L1 positivity in cytologic samples could be a reliable data for the oncologist to consider immune checkpoint inhibitor therapy. However, a comparison of cytologic and histologic samples has shown an underestimation of PD-L1 values in cytologic samples.