6056 Background: Highly sensitive minimal residual disease (MRD) assays using circulating tumor (ct)DNA have broad clinical potential and are impacting cancer care. Circulating human papillomavirus (HPV) DNA has emerged as a biomarker of occult MRD among patients (pts) with HPV+ oropharyngeal cancers, but most head and neck squamous cell carcinomas (HNSCC) are not virally driven. Despite multimodality therapy, nearly half of pts with locoregionally advanced HPV-negative HNSCC relapse. As ctDNA has the potential to identify MRD, predict outcomes, and guide treatment for these pts, we performed a retrospective cohort study to evaluate the performance of a custom-built, clinically and commercially available ctDNA assay for this population. Methods: Pts with newly diagnosed HPV-negative HNSCC (all mucosal sites) treated at a single academic institution with pre-treatment ctDNA testing (Signatera, Natera) performed during clinical care were identified. Signatera utilizes up to 16-plex PCR from matched tumor and blood to develop a personalized ctDNA assay. A subset of patients had additional ctDNA testing during follow-up. Study objectives were to understand baseline ctDNA detection in relation to disease characteristics, and to correlate ctDNA changes on- and post-treatment with disease status and survival. Results: Testing was performed in 92 of 115 (80%) pts with HPV-negative HNSCC (median age: 66, 69% male, 64% smokers); ctDNA testing could not be performed in 23 (20%) pts due to insufficient tumor tissue. Oral cavity (43%) and larynx (22%) were the most common subsites; with most having clinical T2-3 (54%) and N1-2 (51%) disease (AJCC 2017 8 th edition) treated with definitive surgery (46, 40%) and/or chemoradiation (59, 51%). Pre-treatment, 69/92 (75%) pts had detectable ctDNA (range: 0.03-4049.69 mean tumor molecules/mL). No baseline independent clinical features predicted pre-treatment ctDNA detectability or levels (multiple logistic/linear regression modeling), but levels varied by T and N stage in univariate analysis (both p<0.05; Kruskal-Wallis test). At a median follow-up of 5.2 months (range: 0.2-15.1), 47 pts (51%) had >1 test result (range: 1-7; 170 total samples). Of 47 pts, 17 (36%) had ctDNA detected after treatment. Disease-free survival was significantly worse for pts who were ctDNA positive vs. negative after treatment (HR 4.35, 95%CI: 1.68-11.21, p<0.01); 1-year overall survival was 83.7% vs. 100% for pts who were ctDNA positive vs. negative. Conclusions: Tumor-informed, personalized ctDNA testing is feasible among pts with non-virally mediated HNSCC. ctDNA positivity as an early indicator of MRD positivity post-treatment was associated with inferior survival, identifying a high-risk subgroup. Further research is warranted to understand whether ctDNA may be leveraged to guide therapy and improve outcomes for HNSCC.