Impact of PD-L1 combined positive score (CPS) on clinical response to nivolumab in patients with advanced esophageal squamous cell carcinoma.

Abstract

e16045 Background: The ATTRACTION-3 trial showed that nivolumab (Nivo) significantly improved overall survival (OS) of patients (pts) with previously treated advanced esophageal squamous cell carcinoma (ESCC) regardless of tumor programmed cell death ligand-1 (PD-L1) expression assessed with tumor proportion score (TPS). On the other hand, pembrolizumab prolonged OS in advanced esophageal carcinoma pts with combined positive score (CPS) of ≥ 10 in the KEYNOTE-181 trial. Whether CPS can predict clinical outcome of Nivo in advanced ESCC pts remains unclear. Methods: We retrospectively evaluated advanced ESCC pts who received Nivo at a single institution between January 2014 and September 2020. The main eligibility criteria were as follows: refractory or intolerant to fluoropyrimidines and platinum, no prior anti-PD-1/PD-L1 antibody treatment, no comorbid malignancies, and available tissue specimens obtained before initiation of Nivo. PD-L1 immunostaining was performed using PD-L1 IHC 22C3 pharmDx assay, and the tumors were classified into three groups depending on CPS (≥ 10, 1–10, < 1). The adjusted hazard ratios (aHRs) for progression-free survival (PFS) and OS were calculated using a multivariate Cox model that contained variables with p values < 0.05 in the univariate analysis. Results: Among 69 pts, 50 were eligible (CPS ≥ 10/CPS 1–10/CPS < 1, 23/18/9). Patient characteristics were as follows (CPS ≥ 10/CPS 1–10/CPS < 1): age, ≥ 65, 70/67/67%; male, 83/67/78%; PS ≥ 1, 61/72/44%; second-line treatment, 65/61/56%; disease status, recurrent, 61/33/78%; prior esophagectomy, 65/33/44%; prior radiotherapy, 57/56/56%; prior taxane, 52/28/44%; the number of metastatic sites, ≥ 2, 48/61/89%; lymph node metastasis, 78/83/89%, lung metastasis, 26/22/56%; liver metastasis, 17/17/44%. Among 42 pts (84%) with disease progression after Nivo treatment, 24 pts received salvage-line chemotherapy. The adjustment factor for PFS was liver metastasis, while that for OS was not detected. The median PFS was 4.1 months (mo) in CPS ≥ 10 and 2.5 mo in CPS 1–10 (HR vs. CPS ≥ 10, 1.06; 95% CI, 0.51–2.17; p = 0.864; aHR, 1.15; 95% CI, 0.54–2.36; p = 0.713), and 1.4 mo in CPS < 1 (HR vs. CPS ≥ 10, 3,78; 95% CI, 1.53–8.92; p = 0.005; aHR, 1.91; 95% CI, 0.67–5.39; p = 0.223). The median OS was 12.3 mo in CPS ≥ 10, 10.2 mo in CPS 1–10 (HR vs. CPS ≥ 10, 1.72; 95% CI, 0.75–4.00; p = 0.201), and 9.0 mo in CPS < 1 (HR vs. CPS ≥ 10, 2.31; 95% CI, 0.92–5.69; p = 0.075). Objective response rate (CPS ≥ 10/CPS 1–10/CPS < 1) in 40 pts who had measurable lesions were 32/25/0%, respectively. Conclusions: The pts with CPS of < 1 did not respond to Nivo. Our study suggests that CPS is useful for predicting response to Nivo in pts with advanced ESCC.[Table: see text]