Tumor-naïve multimodal cfDNA MRD assay to predict recurrence in a prospective cohort of patients undergoing curative-intent lung cancer resection.

8016 Background: Tumor-informed circulating tumor DNA (ctDNA) assays can achieve high sensitivity for minimal residual disease (MRD) detection after lung cancer resection but require tumor tissue and individualized assay design, limiting scalability and timeliness in routine practice. Tumor-naïve MRD approaches offer a practical alternative for postoperative surveillance; however, their sensitivity remains constrained. Integrating complementary cfDNA features beyond somatic mutations may enhance tumor-naïve MRD detection and postoperative risk stratification. Methods: A prospective cohort of 212 patients with resectable stage I–IV lung cancer, predominantly consisting of stage I disease (146 of 212), was enrolled. Postoperative plasma samples were collected at a landmark timepoint (~1 week after surgery) and longitudinally every 3–6 months for up to 3 years. MRD was assessed using ShieldingUltra, a tumor-naïve multimodal cfDNA assay integrating somatic mutations, copy number variations (CNVs), and fragmentomic features based on ultra-deep UMI-based sequencing of an integrated panel covering >2000 cancer-related genes. MRD positivity was determined from integrated multimodal cfDNA signals. Survival outcomes were evaluated using Kaplan–Meier analysis and Cox proportional hazards models, with subgroup analyses by disease stage, histology, and postoperative adjuvant therapy. Results: At the postsurgical landmark timepoint, MRD positivity was strongly associated with an increased risk of recurrence (hazard ratio [HR], 9.72; log-rank p=2.06×10⁻⁸). Longitudinal MRD monitoring further improved risk stratification, with MRD-positive patients exhibiting a markedly higher recurrence risk (HR, 16.64; log-rank p=1.66×10⁻⁹) while maintaining high specificity (~90%). Among the 27 patients who developed radiographically confirmed recurrence, MRD was detected prior to imaging in the majority of cases (21 of 27), providing a median lead time of 264 days. The prognostic value of MRD status was consistently observed across disease stages, histologic subtypes, and postoperative adjuvant treatment strategies, with particularly robust prognostic discrimination observed in patients with stage I disease. Multivariable analyses confirmed MRD positivity as an independent predictor of recurrence in both landmark and longitudinal settings. Conclusions: Tumor-naïve multimodal MRD assessment integrating mutation, CNV, and fragmentomic cfDNA features enables sensitive and clinically informative detection of residual disease after lung cancer surgery. Early and longitudinal MRD monitoring provides robust prognostic stratification and meaningful lead time over imaging, supporting its clinical utility for postoperative surveillance in resectable lung cancer.