Clinical validation of ultra-sensitive WGS-based MRD detection in head and neck squamous cell carcinoma: Results from MONSTAR-SCREEN-3.

6066 Background: Circulating tumor DNA (ctDNA)-based molecular residual disease (MRD) detection has shown promise across various malignancies, yet limited data exist for Head and Neck Squamous Cell Carcinoma (HNSCC). The prospective, multicenter MONSTAR-SCREEN-3 study evaluates an ultra-sensitive whole-genome sequencing (WGS)-based MRD assay in patients with resectable solid tumors undergoing curative-intent therapy. Here, we report preliminary results from patients with resectable HNSCC enrolled in the definitive cohort (target n=1,100). Methods: Personalized ctDNA panels were generated using a WGS-based tumor-informed platform (Myriad Genetics), incorporating up to 1,000 tumor-specific variants identified through WGS of matched tumor tissue. Serial plasma samples were collected at baseline, 1 month post-surgery, quarterly during the first year, and biannually thereafter for up to two years. Results: As of November 2025, 44 patients with resectable HNSCC were enrolled; MRD results were available for 139 samples from 34 patients. Median age was 65 years (range: 39-87), with male predominance (65.7%). Clinical staging included Stage III (15.6%), IVA (75.0%), and IVB (9.4%). All patients underwent upfront radical surgery. Personalized panel creation succeeded in 100% of patients (34/34), identifying a median of 6,119 highly confident tumor-specific alterations per patient (range: 1,288-15,418) and yielding bespoke panels containing 717-1,000 alterations. Customized panels were created with 97.7% SNVs and 2.3% indels. The assay demonstrated 100% baseline ctDNA detection (34/34), with 5.9% detected at ultra-sensitive levels (tumor fraction <100 parts per million [ppm]; minimum detection: 83.3 ppm). MRD positivity at 1 month, 3 months, and 6 months was at ultrasensitive levels in 10/14 (71.4%), 6/16 (37.5%), and 4/5 (80%) patients, respectively. Among 10 patients who developed radiological recurrence, median lead time was 2.7 months (range: 0-4.6 months). MRD-positive patients at 1 month after surgery had poorer disease-free survival than MRD-negative patients (HR 18.9, 95% CI 2.4–148.9; log-rank P<0.0001). Nine of the 13 patients who were ctDNA-positive at 1 month experienced recurrence, whereas only 1 of the 20 ctDNA-negative patients recurred. Conclusions: The WGS-based personalized ctDNA assay achieved high technical feasibility in HNSCC, with comprehensive customized variant panels highlighting the critical importance of ultra-sensitive platforms. These results suggest potential clinical utility for recurrence surveillance in HNSCC. Updated results will be presented. Clinical trial information: UMIN000053975.