Prevalence and Heterogeneity of PD-L1 Expression by 22C3 Assay in Routine Population-Based and Reflexive Clinical Testing in Lung Cancer

Programmed death-ligand 1 (PD-L1) is used as a biomarker for anti–programmed cell death protein-1 (PD-1) or anti-PD-L1 immunotherapies in NSCLC. We report here the results of population-based PD-L1 testing using the 22C3 IHC pharmDx Assay (Agilent Technologies) in a large Canadian regional reference pathology laboratory. Testing was conducted reflexively on biopsies and resections for NSCLC during an 8-month period. Tumor proportion score (TPS) cutoffs for low and high expression were 1% and 50%, respectively. Altogether, 2031 PD-L1 tests were performed on specimens from 1795 patients, with 107 inconclusive results (5.3%). Excluding cases with inconclusive/missing data, proportions for the remaining 1713 patients were 41.6% for TPS less than 1%, 28.6% for TPS 1% to 49%, and 29.8% for TPS greater than or equal to 50%. Higher PD-L1 expression rates were noted in EGFR wild-type versus mutant tumors (p < 0.001), squamous versus adenocarcinoma (p < 0.001), and metastatic versus primary tumors (p < 0.001). PD-L1 among 103 patients with paired biopsy and resection specimens revealed moderate concordance (κ = 0.67). A total of 52% (25 of 48) of biopsies with TPS less than 1% had TPS greater than 1% in resection, whereas 84.6% (22 of 26) of biopsies with TPS greater than or equal to 50% were concordant in resected tumors. Discordance rates between biopsy and resection were 71.4% for biopsies with less than 8 mm2 total area, compared with 33.3% for biopsies with greater than or equal to 8 mm2 area (p < 0.026). Concordance among 27 patients with paired primary lung and metastatic tumor biopsies revealed only weak concordance (κ = 0.48). Intratumoral heterogeneity of PD-L1 expression may result in misclassification of PD-L1 status in a substantial proportion of PD-L1–negative small biopsy samples. Biopsy of metastatic site may increase proportion of patients with high PD-L1 expression.