Tracking circulating tumor DNA through liquid biopsy after radiotherapy for dynamic risk monitoring in cancer patients.

6053 Background: Circulating tumor DNA (ctDNA) analysis provides a useful tool for non-invasive tumor burden assessment. However, the management of detectable ctDNA without clinical or radiographic evidence of disease during follow-up remains poorly studied and clinically challenge. Here, we explored whether monitoring ctDNA dynamics can differentiate phenotypes in these patients. Nasopharyngeal carcinoma (NPC) is an ideal model for the sensitive detection of cell-free Epstein-Barr virus DNA (cfEBV DNA). Methods: The study analyzed medical records of 14,611 non-metastatic NPC cases treated with definitive radiotherapy with or without chemotherapy from two cohorts: a discovery cohort (n = 4,485) and an independent validation cohort (n = 10,126). Cox regression and recursive partitioning analysis (RPA) were used for survival analyses and patient stratification. Results: In the discovery cohort, 23.1% of cases presented detectable cfEBV DNA during follow-up and experienced significantly worse survival compared to those with longitudinally undetectable cfEBV DNA (5-year DFS: 39.1% vs. 91.5%, p < 0.001). Pre-treatment cfEBV DNA level (≥2000 vs. < 2000 copies/ml) and the maximum level during follow-up (≥500 vs. < 500 copies/ml) were significantly associated with disease failure. Based on the longitudinal dynamic changes, patients with bounce of cfEBV DNA after its clearance had worse survival than those without bounce (HR DFS = 4.3 [3.1-6.0], p < 0.001), and patients with persistent cfEBV DNA had the worst survival (HR DFS = 11.5 [8.3–15.9], p < 0.001). RPA was then conducted incorporating other clinical prognostic factors and four distinct prognostic phenotypes were identified (Table). The model showed a significantly higher positive predictive value (96.6% vs. 60.5%) and similar negative predictive value (67.4% vs. 65.8%) compared to single-point cfEBV DNA assessments. Patients who had clearance of low cfEBV DNA burden had a favorable prognosis (5-year DFS, 83.4%). In contrast, most patients with persistent cfEBV DNA had disease failure (5-year DFS, 8.0%); this subgroup can be defined as molecular relapse. The results were consistent in the validation cohort. Notably, patients in the molecular relapse group benefited from early salvage therapy with improved DFS, whereas patients in other phenotypes responded poorly. Conclusions: Tracking longitudinal ctDNA dynamics after definitive treatment can enable more precise risk assessment and facilitate risk-adapted, individualized patient management. Prognostic phenotypes based on RPA and cfEBV DNA dynamics. HR DFS Group Discovery cohort Validation cohort Clearance of low cfEBV DNA burden (<500 copies/ml) Reference Clearance of high cfEBV DNA burden (≥500 copies/ml) 3.0 [1.9-4.7] 1.6 [1.1-2.2] Bounce after clearance 5.9 [4.1-8.7] 3.5 [2.5-4.7] Persistent cfEBV DNA 15.7 [10.8-22.7] 13.0 [9.6-17.7] p <0.001 p <0.001