PD-L1 Tumour-Cell/Immune-Cell Phenotype, Early Nodal Status and Serum Biomarkers in Locally Advanced Head and Neck Squamous Cell Carcinoma Receiving Neoadjuvant Therapy: A Retrospective Cohort Study With Exploratory Survival Analysis.

PD-L1 tumour-cell (TC) and immune-cell (IC) expression are scored separately in head and neck squamous cell carcinoma (HNSCC) but are often collapsed into a combined positive score, potentially obscuring distinct patterns. In this retrospective, single-centre cohort of 61 patients with locally advanced, hypopharynx-predominant HNSCC receiving neoadjuvant chemoimmunotherapy (n = 46) or chemotherapy alone (n = 15), we scored PD-L1 TC% and IC% separately on pretreatment biopsies, defined four TC/IC phenotypes, classified patients by combined PD-L1/early lymph node metastasis (ELNM) profile and related these to serum biomarkers, neoadjuvant response and survival. TC% and IC% were not correlated (Spearman ρ = -0.13, p = 0.40), supporting their biological independence; among 46 patients with both scores the phenotypes were cold (TC-/IC-, n = 10), TC-intrinsic (n = 10), IC-dominant (n = 12) and hot (TC+/IC+, n = 14). The overall objective response rate after two neoadjuvant cycles was 91.4% (53/58 evaluable), with no progressive disease and did not differ significantly across PD-L1/ELNM profiles (p = 0.15) or between chemoimmunotherapy and chemotherapy alone (95% vs. 79%; p = 0.085). Serum IL-6 showed a non-significant inverse association with IC% (ρ = -0.33, p = 0.10) and was numerically lowest in the hot phenotype; serum tumour markers did not differ by ELNM status. Over a median follow-up of 24.0 months there were three deaths and nine progression events; in exploratory, underpowered analyses neither progression-free nor overall survival differed by PD-L1/ELNM profile (PFS log-rank p = 0.40) or treatment modality (p = 0.64), and the double-positive group did not show the favourable survival suggested by uncontrolled comparisons. These data indicate that PD-L1 TC% and IC% capture independent information and that the TC/IC phenotype distribution is heterogeneous; with few events the survival analysis is hypothesis-generating only, and adequately powered prospective studies with longer follow-up are required.