Document resource
Background Immune checkpoint inhibitors (ICI) are the current standard of care for patients with recurrent/metastatic (R/M) head and neck squamous cell carcinoma (HNSCC), but only 20% of patients respond to treatment. Tissue-based biomarkers such as PD-L1 combined positive score (CPS) and tumor mutational burden (TMB) have limited success in predicting response. Recent work demonstrated the prognostic value of baseline peripheral blood biomarkers from routine laboratory tests, such as complete blood count (CBC) and comprehensive metabolic panel (CMP), in patients treated with ICI. 1 We tested the hypothesis that integrating more detailed immune profiles with clinical and laboratory data will improve HNSCC ICI prognostic models.Methods Patients with R/M HNSCC receiving ICI (n = 114) were recruited from Dana-Farber Cancer Institute, Dartmouth Cancer Center, and Rhode Island Hospital. Whole blood at baseline ICI treatment was used to extract DNA and measure methylation with the Illumina EPIC array. Methylation cytometry was used to quantify 12 immune cell subtypes. 2 A recursive partitioning algorithm called partDSA3 that uses ‘and’/’or’ logic was used to model complex interactions and identify immunotherapy prognostic groups with overall survival (OS) at 2-years as the outcome. We compared the base partDSA model with only clinical and demographic data to models that added routine CBC and CMP data, and a third model that added detailed immune profiles from methylation cytometry (figure 1). Risk groups from partDSA were evaluated with Kaplan-Meier, Cox proportional hazards models, and concordance index (C-index) values.Results The addition of immune-related variables to partDSA models improved stratification of HNSCC ICI outcome of OS at 2-years. The base model (HR: 3.2, 95% CI 1.8-5.5, C-index = 0.64) improved with the inclusion of CBC and CMP data (HR: 4.1, 95% CI: 2.5-6.9, C-index = 0.66) and improved further with the addition of detailed immune profile data from methylation cytometry (HR: 5.3, 3.0-9.4, C-index = 0.69). Differences in restricted mean survival time increased between the highest and lowest risk group from 220 days (base model) to 275 days (CBC+CMP model) and 293 days (methylation cytometry model). Methylation cytometry model features most predictive of 2-year OS included CD4 memory (%), CD4 naïve (%), sex, monocytes (%), red blood cells (RBC), and B memory cell count ( figure 2).Conclusions Incorporating methylation cytometry derived immune profiles with routinely collected clinical data improved prognostic stratification in R/M HNSCC, underscoring their clinical utility in assessing ICI response. Further validation studies are warranted in the R/M and neoadjuvant settings.References Yoo SK, Fitzgerald CW, Cho BA, Fitzgerald BG, Han C, Koh ES, et al. Prediction of checkpoint inhibitor immunotherapy efficacy for cancer using routine blood tests and clinical data. Nat Med. 2025 Mar;31(3):869–80.Salas LA, Zhang Z, Koestler DC, Butler RA, Hansen HM, Molinaro AM, et al. Enhanced cell deconvolution of peripheral blood using DNA methylation for high-resolution immune profiling. Nat Commun. 2022 Feb 9;13(1):761.Molinaro AM, Lostritto K, Van Der Laan M. partDSA : deletion/substitution/addition algorithm for partitioning the covariate space in prediction. Bioinformatics. 2010 May 15;26(10):1357–63.Ethics Approval This study was approved by the Institutional Review Board (IRB) at Dana-Farber Cancer Institute (Protocol #18-548) and all patients from Dana-Farber Cancer Institute, Dartmouth Cancer Center, and Rhode Island Hospital provided written informed consent to participate in this study.Abstract 1123 Figure 1Overview of study design and analysis (created with Biorender.com)Abstract 1123 Figure 2partDSA methylation cytometry model identifies key variables and cutoffs associated with 2-year overall survival (created with Biorender.com)