Treatment-dependent and independent prognostic value of inflammation-based nutritional indices in immune checkpoint inhibitor-based therapy for esophageal cancer.

BACKGROUND: Phase III trials have established the efficacy of immune-based therapies for unresectable advanced or recurrent esophageal squamous cell carcinoma (ESCC). However, real-world evidence comparing regimens and clinically applicable prognostic biomarkers remains limited. METHODS: We retrospectively evaluated 245 patients with unresectable advanced/recurrent ESCC treated with first-line 5-fluorouracil plus cisplatin combined with nivolumab or pembrolizumab (CF + ICI; n = 114), first-line ipilimumab plus nivolumab (Ipi + Nivo; n = 46), or nivolumab monotherapy as second-line or later treatment (Nivo; n = 85). Efficacy, safety, and survival outcomes were assessed, focusing on inflammation-based nutritional indices and multivariate prognostic modelling. RESULTS: The objective response rates were 56%, 53%, and 27%, and the median overall survival was 26, 73, and 13 months in the CF + ICI, Ipi + Nivo, and Nivo groups, respectively. No treatment-related deaths occurred. In multivariate analyses, Glasgow Prognostic Score (GPS) ≥ 2 independently predicted poor survival across all regimens. High platelet-to-lymphocyte ratio (PLR) was independently associated with poor survival only in chemotherapy-free ICI regimens (Ipi + Nivo and Nivo). A pooled model showed a significant interaction between treatment regimen and PLR (p = 0.047). CONCLUSIONS: In real-world practice, GPS provides regimen-independent prognostic stratification, whereas PLR shows treatment-dependent prognostic relevance, supporting its potential utility for biomarker-guided and individualized treatment selection in ESCC.