Performance characteristics of a tissue-agnostic genome-wide methylome enrichment MRD assay for head and neck malignancies.

3009 Background: Plasma cell-free DNA (cfDNA) tests have emerged as a promising approach for cancer management. cfDNA methylome approaches are well-suited for molecular residual disease (MRD) detection. Here we present data using a tissue-agnostic, genome-wide methylome enrichment platform based on cell-free methylated DNA immunoprecipitation and high throughput sequencing (cfMEDIP-seq) in head and neck cancer (HNC) to predict relapse for purposes of guiding adjuvant therapy after completion of curative-intent treatment and to detect early relapse. Methods: The cohort is comprised of biobanked samples from individuals diagnosed with stage I-IVB human papillomavirus (HPV)-negative and HPV-positive HNC with longitudinal data collection and sampling. The full cohort includes 325 unique patients with 1,155 samples. Samples were split into distinct sets to train and test a classifier consisting of differentially methylated regions. Blood collection time points include at diagnosis, and approximately 3 (landmark), 12 and 24 months after curative intent treatment. 5-10 ng of cfDNA isolated from each plasma sample was used for cfMEDIP-seq. MRD signals were quantified from average normalized counts across informative methylated regions and binarized into a positive (above the threshold) and negative groups. Recurrence-free survival (RFS) was compared for patients who tested positive to those who tested negative at 3 months post-curative treatment (i.e., landmark timepoint) and longitudinally. Results: A total of 196 post-treatment samples from 80 unique patients stage I (35%), II (15%), III (24%), IV (26%) were analyzed and correlated with recurrence, in this interim training result. At the landmark timepoint, patients who tested positive showed significantly worse RFS than those who tested negative (Hazard ratio (HR) 9.69; 95% CI, 4.39-21.4, P<0.001). Incorporating serial longitudinal samples, recurrence-free survival was worse in patients who tested positive compared to those that tested negative (HR 14.52; 95% CI, 5.78-36.46, P<0.001). Conclusions: Interim analysis demonstrates that MRD detection with a tissue-agnostic, genome-wide methylome enrichment platform in HNC patients after curative intent treatment correlates strongly with RFS with hazard ratios consistent with tumor-informed assays previously described. Updated analyses from the cohort will be presented at the meeting.