Immune Checkpoint Inhibitor Therapy for Advanced, Unresectable Esophageal Squamous Cell Carcinoma: A Series of Patient-level Meta-analyses From Phase III Trials.

[AIM] This study reconstructed patient-level data to provide updated evidence on survival outcomes across PD-L1 expression subgroups.

[MATERIALS AND METHODS] A systematic search was conducted on March 2, 2025, using PubMed, Embase, Web of Science, Cochrane, and Scopus for RCTs comparing ICIs to chemotherapy in advanced ESCC. Kaplan-Meier curves were reconstructed using a time-to-event algorithm for all patients and for PD-L1 subgroups. KMSubtraction was used to retrieve unpublished survival data. A patient-level meta-analysis was performed using a stratified Cox model.

[RESULTS] Thirteen phase III trials (6,672 patients) were included. In the first-line setting, patients with TPS <1% (from CheckMate-648 and ESCORT-1st) showed no overall survival (OS) benefit (HR: 0.88, p = 0.187), whereas those with TPS ≥1% from the same trials showed a significant benefit (HR: 0.61, p < 0.001). For a combined positive score (CPS) ≥10, pooled data from ASTRUM-007, KEYNOTE-590, ORIENT-15, and GEMSTONE-304 showed significant OS benefit (HR: 0.60, p < 0.001). CPS <10 (KEYNOTE-590, ORIENT-15, and GEMSTONE-304) also showed benefit (HR: 0.81, p = 0.02), but no survival benefit was seen in CPS 1-9 (ASTRUM-007 and GEMSTONE-304) (HR: 0.82, p = 0.117).

[CONCLUSION] PD-L1 appears to be a useful biomarker in advanced ESCC. Patients with PD-L1 levels of 10 or higher consistently showed survival benefits. For PD-L1 <10, treatment decisions may be less clear; further research on combining PD-L1 with other biomarkers is needed. Patients with PD-L1 <1 showed minimal benefit and should be informed about the limited efficacy of ICIs.