PD-L1 score: Tool effectiveness in non-small cell lung cancer.

Abstract

e20594 Background: Programmed death (and ligand)-1 (PD-(L)1) inhibitors are standard of care for treatment of non-small cell lung cancer (NSCLC). There are four FDA approved PD1 immunohistochemistry (IHC) assays used to measure expression on tumor (TC) and immune cells. Prior studies note discordance in expression across assays with unclear clinical significance. It remains poorly understood whether PD-L1 scoring assays adequately predict response to immunotherapy (IO). Methods: We conducted a retrospective study to investigate discordance in PD-L1 IHC scores among patients with NSCLC. Pathology samples were obtained from patients diagnosed at TJUH between 2019-2022. All samples were stained for PD-L1 IHC using two FDA approved assays, Ventana SP142 and Ventana SP263. PD-L1 expression was categorized as < 1%, 1-49% and > 50%. Patients with assay results in different expression categories were discordant, while patients with assay expression in the same categories were concordant. Patients with concordant PD-L1 scores served as a comparator group. The influence of PD-L1 discordance on treatment was determined by treating physician documentation indicating assay results impacted therapy choice. Descriptive statistics were used to describe demographic and clinical characteristics. Kaplan-Meier curves were used to analyze overall survival (OS) and progression free survival (PFS). Results: Of 133 patients, 107 had discordant assays and 26 were concordant. There were no significant differences between groups with respect to race, stage, or histologic subtype. Median age at diagnosis was 66 vs 70 years (control vs discordant, p = 0.04). Discordant patients were largely female (56%) and concordant patients were largely male (65%, p = 0.011). Most patients presented with stage IV disease (64% discordant vs 56% concordant). Targetable mutations were more frequent in the discordant group (30% vs 3.8%, p = 0.005), predominately EGFR (17.8%). Among discordant patients Ventana SP142 assay scores were all < 50%, while Ventana SP263 assay scores were > 50%. There was no significant difference in IO use or toxicity between cohorts. Discordance in PD-L1 score influenced treatment choice in 26% of patients (p = 0.002). This remained significant after excluding patients eligible for frontline targeted therapy (30.7%, p = 0.001). In both groups, treatment choice remained largely guideline directed (72% concordant vs 67% discordant p = 0.8). Discordance did not significantly impact PFS or OS. Conclusions: Findings confirm higher rates of PD-L1 discordance in patients with targetable mutations, particularly EGFR. Among discordant patients Ventana SP142 scores were all less than 50%, consistent with prior studies showing reduced and discordant TC staining with this assay due to lower sensitivity. While discordance influenced frontline therapy choice, there was no significant impact on clinical outcomes, suggesting PD-L1 is not an optimal predictive marker.