Immune checkpoint blockade induced shifts in cytokine expression patterns in peripheral blood of head and neck cancer patients are linked to outcome

Background: Immune-checkpoint blockade (ICB) of programmed-death-1 (PD-1) with pembrolizumab or nivolumab is approved for treating recurrent/metastatic (R/M) head and neck squamous cell carcinoma (HNSCC). NadiHN and ADRISK are phase IIB trials investigating in locally advanced (LA) HNSCC having low or high risk of recurrence the potential benefits from adding nivolumab to post-operative radiotherapy or pembrolizumab to cisplatin-based radio-chemotherapy.Methods: Along five randomized controlled ICB trials including NadiHN and ADRISK, blood samples were taken before and after starting ICB in n=25 patients. Concentrations of vascular endothelial growth factor A (VEGF), CCL2 (MCP-1), interleukin-6 (IL-6), IL-8, interferon-gamma (IFN-gamma), and CXCL10 (IP-10) pre- and post-ICB in EDTA-anticoagulated plasma and serum were compared. We used receiver operating characteristic (ROC) curves to identify optimal cutoff for defining subgroups before analyzing overall survival (OS) applying Kaplan-Meier plots and multivariate Cox regression.Results: We detected huge heterogeneity between cytokine patterns in pre-and post-ICB plasma and serum. We observed high correlation between concentrations of some cytokines. Despite absent systematic OS differences after ICB with pembrolizumab or nivolumab or between LA-HNSCC versus R/M HNSCC patients, we noticed improved outcome of patients having lower IFN-gamma concentrations pre- and post-ICB and following ICB reduced concentrations of VEGF, IL-6, and IL-8 but not MCP-1. Contrarily, increases in IL-6, IL-8, and VEGF levels correlated with impaired outcome. Multivariate Cox regression revealed five independent OS predictors among cytokines; using natural logarithms of their hazard ratios to estimate an individual's risk of dying, three cytokine-expression pattern (CEP)-risk groups with no death within mean (95% confidence interval) follow-up of 29.2 (22.1-36.2) months and median OS of 11.3 (8.8-13.8) and 2.9 (0.4-5.4) months were found.Conclusion: Whereas individual pre- or post-ICB cytokine concentrations in serum or plasma alone failed to predict the survivor group, CEP-risk groups may support the identification of individual patients with long-lasting benefit from ICB.