Ultra-sensitive pan-cancer molecular residual disease assessment using whole-genome sequencing-based personalized ctDNA panel: Initial results from the MONSTAR-SCREEN-3 project.

3007 Background: While circulating tumor DNA (ctDNA) demonstrates promise as a molecular residual disease (MRD) biomarker, its clinical implementation has been primarily limited to tumors with favorable ctDNA shedding characteristics. We are evaluating an ultra-sensitive whole-genome sequencing (WGS)-based MRD assay in the MONSTAR-SCREEN-3 study to establish a comprehensive pan-cancer MRD platform inclusive of traditionally low-shedding tumors. Methods: MONSTAR-SCREEN-3, a prospective multicenter study targeting 1,100 patients with solid tumors undergoing curative-intent treatment in the definitive cohort, utilizes personalized panels constructed via Precise MRD (Myriad Genetics). These panels incorporate up to 1,000 tumor-specific alterations identified through WGS of matched tumor tissue, including both short variants and insertion-deletions. Serial plasma samples were collected at baseline, post-neoadjuvant chemotherapy (when applicable), 1-month post-surgery, quarterly in year 1, and biannually thereafter up to 2 years. The assay performance was evaluated across multiple cancer types for ctDNA detection and recurrence monitoring. Results: As of December 2024, 114 patients across 15 cancer types were enrolled, including colorectal (n = 33), gastric (n = 22), head and neck (n = 13), renal cell (n = 10), esophageal (n = 8), and pancreatic (n = 7) cancers. Treatment strategies included upfront surgery (n = 76) and neoadjuvant chemotherapy (n = 38). The median follow-up time was 2.4 months (range, 0.5–7.7). WGS analysis identified a median of 6,089 panel-eligible alterations per patient (range: 214-14,112), with high variants counts observed in a deficient mismatch-repair colorectal cancer, enabling comprehensive personalized panel design. Customized panel creation was successful in 69/71 patients (97.2%) across 8 cancer types, with two pancreatic cancer cases deferred to surgical specimens due to insufficient variants in FNA samples. The assay demonstrated 100% baseline sensitivity (41/41), detecting tumor fractions ranging from < 0.001% to 45.2% across all cancer types, including traditionally low-shedding tumors. Post-operative 1-month MRD assessment revealed 35.7% positivity (10/28), with tumor fractions ranging from < 0.001% to 0.27%. Two MRD-positive patients developed radiological recurrence with lead times of 2.5 and 3 months before conventional imaging detection. Conclusions: These interim results demonstrate successful pan-cancer implementation of WGS-based personalized ctDNA detection, achieving universal baseline sensitivity and ultra-sensitive MRD detection across tumor types, including those traditionally challenging to assess. Updated molecular and clinical outcome data will be presented. Clinical trial information: UMIN000053975 .