Postoperative Circulating Tumor DNA Monitoring in Stage I-III Driver-Mutant Colon Adenocarcinoma

Background Colorectal cancer (CRC) ranks the 3rd most common malignancy worldwide and the 2nd leading cause of cancer death. In China, CRC incidence and mortality keep rapidly rising, with \~517,000 new cases and \~240,000 deaths annually. Radical resection is the curative standard for stage I-III CRC, yet postoperative recurrence reaches 20%-40% even after microscopically margin-negative (R0) resection. Traditional tumor-node-metastasis (TNM) staging and clinicopathological features fail to precisely stratify recurrence risk within the same stage, lacking capacity for early, real-time recurrence surveillance. Oncogenic driver mutations (KRAS, NRAS, BRAF, PIK3CA) dominate colorectal tumorigenesis and correlate with higher recurrence risk and worse prognosis, but static mutation status cannot reflect dynamic postoperative residual tumor burden. Circulating tumor DNA (ctDNA)-based molecular residual disease (MRD) enables non-invasive early detection of microscopic residual tumor, predicting recurrence months prior to radiological abnormalities, with strong prognostic value validated by multiple landmark trials. However, nearly all existing MRD cohorts lack stratification by driver mutation status; large-scale prospective multicenter data focusing exclusively on driver-mutant stage I-III colon adenocarcinoma are absent globally. This study fills the clinical gap by combining driver genotyping and longitudinal ctDNA-based MRD surveillance. Scientific Hypothesis In patients with driver-mutant stage I-III resectable colon adenocarcinoma, postoperative ctDNA-based MRD status acts as an independent prognostic biomarker for recurrence regardless of conventional clinicopathological factors. Dynamic serial MRD shifts can alert tumor relapse earlier than routine contrast-enhanced computed tomography (CT), and different driver mutation subtypes demonstrate heterogeneous MRD detectability and predictive efficacy. Persistent MRD positivity predicts high recurrence risk, while sustained MRD negativity indicates excellent prognosis and potential de-escalation of adjuvant chemotherapy. Study Design \& Procedures A total of 10 tertiary hospitals with colorectal surgery and molecular pathology capacity will consecutively enroll eligible patients within 1 year. All participants receive standardized perioperative and follow-up care per clinical guidelines without study-mandated intervention modification. Serial peripheral blood samples (20 mL per draw) are collected at fixed time points: pre-operation (T0, baseline), postoperative day 7±2 (T1), 1 month post-surgery before adjuvant chemotherapy (T2), every 3 months up to 12 months (T3-T6), every 6 months at 18/24 months (T7-T8), and 36 months post-operation (T9). ctDNA-based MRD testing is performed centrally by Amoy Diagnostics using a fixed 10-gene next-generation sequencing (NGS) panel covering all core CRC driver mutations, sequencing depth ≥100,000×. MRD positivity is defined as detection of any tumor-derived driver mutation in plasma at any postoperative time point. All patients receive standardized 3-year follow-up: physical exam + tumor markers every 3 months in year 1; every 6 months in year 2; annual follow-up in year 3. Contrast-enhanced computed tomography (CT) of the chest, abdomen, and pelvis is conducted every 6 months, colonoscopy at 12, 24 and 36 months postoperatively. All recurrence events are adjudicated by an blinded independent central review (BICR). Statistical Plan Primary analysis uses Kaplan-Meier curves and log-rank test to compare RFS between MRD-positive and MRD-negative groups; multivariate Cox proportional hazards regression adjusts for age, sex, TNM stage, differentiation grade, adjuvant chemotherapy regimen and driver mutation subtype to confirm MRD as independent prognostic factor. Secondary endpoints including OS, 1- and 2-year RFS rates, median lead time of MRD ahead of radiological recurrence are analyzed via survival statistics. All statistical tests are two-sided with α=0.05, analyzed using R 4.3.0 and SPSS software. Sample size calculation accounts for 20% dropout rate, targeting total enrollment of 392 patients to ensure statistical power of 0.95 for primary endpoint analysis.