Abstract
Introduction: Programmed death-ligand 1 (PD-L1) staining is used in clinical practice to guide the proper use of immune checkpoint inhibitors. This study aimed to investigate the accuracy of PD-L1 staining of non-small cell lung cancer (NSCLC) cytological cell block samples. Methods: Paired cytological cell block and surgical resection samples were consecutively collected from January 2016 to February 2017 in Shanghai Pulmonary Hospital, Tongji University. Two trial-validated PD-L1 assays (28-8 and SP142) were used to quantify PD-L1 expression. Results: A total of 112 pairs of specimens were collected, including 68(60.7%) adenocarcinomas and 28(25.0%) squamous cell carcinomas. Based on a tumor proportion score (TPS) cutoff of 1% for the 28-8 and SP142 assays, PD-L1 expression was positive in 78.6% and 58.9% of surgical samples respectively, while PD-L1 expression was positive in 67.9% and 25.0% of cytological cell block samples. Based on staining by each antibody, fair to substantial concordance of PD-L1 expression was observed for cytological cell block specimens as compared to surgical resection (𝛋 ranges from 0.377 to 0.686). However, as the tumor cells in the cell block specimen increased, the consistency of PD-L1 expression increased. The concordance of PD-L1 expression in cell blocks with abundant cellularity was nearly perfect with various cutoffs (28-8: tumor cells over 400; SP142: tumor cells over 500). Conclusion: Cytological cell block specimens may serve as a surrogate for PD-L1 staining in patients of NSCLC when more than 400-500 cancer cells were contained (over 400 cancer cells for 28-8, over 500 cancer cells for SP142).