Pre- and Post-treatment Serum Cytokine Expression Signatures and Survival Outcomes in Patients with Head and Neck Squamous Cell Cancer on Anti-PD-1 Therapy

M.S.I.S. Saif,K. Sehgal,Z. Zhang,K. Shirai,R. Butler,J. Wiencke,G. Ramush,M.K. Lee,A.M. Molinaro,J. Miller,H. Stolrow,L.A. Salas,D.C. Koestler,K.T. Kelsey,B.C. Christensen,R.I. Haddad
DOI: https://doi.org/10.1016/j.ijrobp.2024.01.198
IF: 8.013
2024-04-01
International Journal of Radiation Oncology*Biology*Physics
Abstract:Purpose/Objective(s) Clinically relevant biomarkers for anti-programmed death (PD)-1 therapy in recurrent/metastatic (R/M) head and neck squamous cell carcinoma (HNSCC) are limited to PD ligand-1 combined proportion score (PD-L1 CPS) on tumor biopsies. Cytokines are soluble proteins that act as a bridge between innate immunity, inflammation, and cancer. Given the interest in circulating biomarkers, this study sought to evaluate pre- and post-treatment serum cytokine expression signatures in R/M HNSCC treated with anti-PD-1 therapy. Materials/Methods In this ongoing, prospective multi-center study, 33 paired pre- and post-treatment (cycle 2 day 1) blood samples were analyzed for 92 cytokine biomarkers (Olink Target 96 Immuno-Oncology) in patients with R/M HNSCC on anti-PD-1-based therapy. Associations between pre-treatment, post-treatment, and change of pre- to post-treatment cytokine levels, overall survival (OS), and progression-free survival (PFS) were assessed using Kaplan-Meier curves, multivariable Cox proportional hazards (adjusted for age, sex, smoking, PD-L1 CPS, ECOG performance status, and platelet count), and unsupervised clustering analyses. Results There was a statistically significant association between pre-treatment cytokine expression signature, and both immune composition (e.g., relative fractions of circulating immune cell subtypes) and platelet count. Comparing pre- and post-treatment samples, expression levels of most cytokines exhibited a statistically significant increase (p <0.05). In pre-treatment samples, several cytokines exhibited nominally statistically significant associations (p < 0.05) with both OS and PFS in multivariable model. Similar results were seen in analyses of serum cytokines with PFS in post-treatment samples. Machine learning analysis using Elastic Net models identified a panel of cytokines (CD40.L, EGF, ANGPT1, MCP.4, CXCL1, IL10) that predicted survival with high prognostic performance (Cross Validated C-index of 0.75) in pre-treatment samples. Notably, a different panel of cytokines (ANGPT1, IL10, FGF2, TNFSF14, Gal.1, HGF, PTN, VEGFA, CCL20) predicted survival with even higher prognostic performance (Cross Validated C-index of 0.90) in post-treatment samples. Unsupervised clustering analysis based on the change in pre- versus post-treatment cytokine expression identified two subgroups which differed in their survival trajectories (log-rank p =0.09 and 0.01 for OS and PFS, respectively), where the subgroup with smaller changes in serum cytokine levels was associated with higher survival. Conclusion Our results highlight the potential prognostic and predictive value of serum cytokine expression signatures in R/M HNSCC treated with anti-PD-1 therapy and underscore the importance of longitudinal on-treatment assessments in peripheral blood.
oncology,radiology, nuclear medicine & medical imaging
What problem does this paper attempt to address?