Trajectory identification of high-risk subgroups after colorectal cancer surgery in patients with minimal residual disease: a single-center longitudinal study

Introduction

Colorectal cancer (CRC) is one of the most prevalent malignant neoplasms of the digestive tract, ranking third in incidence and second in mortality among all malignant tumors (1). Surgical intervention remains the primary treatment modality for CRC. However, the literature suggests that approximately 30% to 50% of CRC patients develop distant organ metastases following radical surgery (2). Traditional monitoring techniques, such as imaging examinations, are constrained by radiation exposure and inherent delays (3,4), often resulting in missed surgical opportunities by the time metastases are detected, which significantly worsens the prognosis. Consequently, there is an urgent need for more precise detection methodologies to improve monitoring efficacy.

Minimal residual disease (MRD) denotes the presence of a small population of cancer cells that persist in the body following radical surgical intervention in CRC patients (5). Traditional imaging techniques are inadequate for the detection of MRD, necessitating the use of liquid biopsy methods. Initially, MRD assessment was applied to hematologic malignancies to evaluate the risk of recurrence and inform subsequent therapeutic strategies. Over time, its application has been extended to solid tumors, including lung, breast, prostate, and ovarian cancers, where it plays a significant role in guiding postoperative adjuvant therapy and predicting recurrence and metastasis (5-9). Furthermore, MRD status serves as both a prognostic biomarker (10) and a predictive tool for guiding adjuvant therapy (11) and decisions regarding metastatic treatment in CRC (12).

Earlier studies have demonstrated that circulating tumor DNA (ctDNA)-MRD detection in CRC can identify distant metastases approximately 3.3 months before traditional imaging methods (13). A large international prospective observational clinical trial (GALAXY) found that postoperative ctDNA-MRD positivity was strongly linked to shorter progression-free survival (PFS) and a higher risk of tumor recurrence, offering more guidance than postoperative pathological staging (3). In 2024, the American Society of Clinical Oncology (ASCO) revealed the 5-year follow-up results of the DYNAMIC study, indicating that the long-term survival outcomes for patients in the ctDNA-MRD guided management group were comparable to those in the standard management group (14). Previous studies have shown that ctDNA-MRD positivity is a crucial prognostic marker for stage II/III CRC, with a recurrence risk nearly 12 times greater than that of ctDNA-MRD negative patients. Extensive research indicates that ctDNA-MRD positive patients are at a higher risk of recurrence or distant metastasis, whereas ctDNA-MRD negative patients have a lower chance of postoperative recurrence and metastasis (3). Despite being classified as ctDNA-MRD negative, certain CRC patients continue to experience postoperative recurrence or distant metastasis, suggesting potential limitations in current ctDNA-MRD testing methodologies (15,16). For this high-risk subgroup, the existing ctDNA-MRD evaluation framework is inadequate, and their risk characteristics remain unidentified, underscoring the need for precise identification of individuals at high risk for postoperative distant metastasis in CRC.

This study seeks to examine the relationship between ctDNA-MRD and clinicopathological factors, as well as prognosis, in order to identify high-risk subgroups of CRC through a longitudinal analysis of ctDNA-MRD. We present this article in accordance with the TRIPOD reporting checklist (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0476/rc).

Methods

Study design and population

This retrospective cohort study, conducted at The First Affiliated Hospital of Kunming Medical University, focused on a single-center analysis. A total of (n=124) patients diagnosed with CRC and who underwent radical resection, postoperative gene panel testing, and ctDNA-MRD assessment between January 2017 and December 2024 were included. Comprehensive data on preoperative clinical baseline characteristics, laboratory parameters, and postoperative pathological findings were systematically collected. A detailed technical overview of the study workflow is provided in (Figure 1). This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of The First Affiliated Hospital of Kunming Medical University (No. 2025-L-27). Due to the retrospective nature of the study and the use of de-identified data, the requirement for informed consent was waived.

Figure 1 Study flowchart. FN, false negative; FP, false positive; MRD, minimal residual disease; TN, true negative; TP, true positive.

Inclusion and exclusion criteria

Patients were deemed eligible for inclusion in the study if they satisfied all of the following criteria: (I) they were aged 18 years or older; (II) they had a histopathologically confirmed diagnosis of primary colorectal adenocarcinoma; (III) they had undergone an R0 radical resection of the primary tumor; (IV) they had completed at least one postoperative ctDNA based MRD assay; and (V) they had comprehensive documentation of essential clinical, pathological, and follow-up data. The exclusion criteria were as follows: (I) the presence of another primary malignancy that had not been cured or had recurred within the previous five years; (II) the presence of synchronous or metachronous multiple primary cancers; (III) a confirmed diagnosis of hereditary or familial CRC syndromes [e.g., Lynch syndrome, familial adenomatous polyposis (FAP), MUTYH-associated polyposis (MAP), POLE/POLD1 syndromes] or cancer associated with inflammatory bowel disease (IBD); (IV) non-adenocarcinoma histologies [e.g., neuroendocrine tumors, gastrointestinal stromal tumors (GIST)]; and (V) failure to meet quality control metrics for ctDNA-MRD testing, including insufficient plasma volume, severe haemolysis, sequencing failure, or inadequate coverage depth.

Study groups and definitions

Patients were categorized into four distinct cohorts based on their postoperative ctDNA-MRD status and long-term clinical outcomes. The experimental group comprised individuals who were ctDNA-MRD negative postoperatively but exhibited documented recurrence or metastasis. The negative control group consisted of patients who were ctDNA-MRD negative postoperatively and did not experience recurrence or metastasis. The positive control group included those who were ctDNA-MRD positive postoperatively and subsequently developed recurrence or metastasis. Lastly, the blank control group encompassed patients who were ctDNA-MRD positive postoperatively but did not show recurrence or metastasis.

Follow-up and study endpoints

The follow-up period terminated on December 31, 2025, or at the time of patient mortality, whichever occurred earlier. The follow-up duration ranged from a minimum of 12 months to a maximum of 85 months, with a median duration of 31.5 months and a mean duration of 30.1 months.

Observational indicators

Clinical and pathological variables

A comprehensive analysis was conducted on (n=124) clinicopathological parameters, encompassing ctDNA-MRD, sex, age, body mass index (BMI), family history of cancer, ethnicity, blood type, history of chemotherapy, preoperative white blood cell count (WBC), neutrophil-lymphocyte ratio (NLR), monocyte-to-lymphocyte ratio (MLR), neutrophil-monocyte-to-lymphocyte ratio (NMLR), systemic inflammatory response index (SIRI), systemic immune-inflammation index (SII), total protein (TP), serum globulin (GLB), serum albumin (ALB), cholinesterase (ChE), blood urea nitrogen (BUN), creatinine (Cr), uric acid (UA), prothrombin time (PT), fibrinogen (Fib), thrombin time (TT), activated partial thromboplastin time (APTT), alpha-fetoprotein (AFP), carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (CA199), carbohydrate antigen 125 (CA125), carbohydrate antigen 15-3 (CA153), neuron-specific enolase (NSE), MRD status, tumor location, tumor morphology, tumor size, tumor type, differentiation grade, pathological T stage (pT stage), pathological N stage (pN stage), Union for International Cancer Control (UICC) stage, vascular invasion, perineural invasion, and mismatch repair (MMR) protein expression.

Inflammatory and immune index calculations

The following indices were calculated from preoperative complete blood count data: the NLR was determined by dividing the absolute neutrophil count by the absolute lymphocyte count; the MLR was calculated as the absolute monocyte count divided by the absolute lymphocyte count; the NMLR was obtained by dividing the sum of the absolute neutrophil and monocyte counts by the absolute lymphocyte count; and the SIRI was derived by multiplying the platelet count by the absolute neutrophil count and subsequently dividing by the absolute lymphocyte count.

Cut-off value determination

Optimal cut-off values for the continuous variables, including age, WBC, NLR, MLR, NMLR, SIRI, SII, PLR, TR, ALB, GLB, ChE, BUN, Cr, UA, PT, Fib, TT, APTT, and tumor size, were established using X-tile software, resulting in thresholds of 64yearS, 5.1, 3.8, 0.3, 4.1, 0.8, 813.4, 69.5g/L, 39.8g/L, 33.6g/L, 6.2U/L, 3.5mmol/L, 80.1mol/L, 363.7umol/L, 12.7s, 4.9g/L, 19s, 33.9s, 0.5s, and 3cm, respectively. The variables AFP, CEA, CA125, CA153, CA199, and NSE were dichotomized based on institutional reference thresholds from The First Affiliated Hospital of Kunming Medical University, specifically 7ng/mL, 5ng/mL, 24U/mL, 24U/mL, 30U/mL, and 10U/mL, respectively. BMI categories were defined according to international weight classification standards: underw eight (<18.5 kg/m²), normal weight (18.6-23.9 kg/m²), overweight (24-27.9 kg/m²), and obese (>28 kg/m²) (17).

MRD longitudinal trajectory variables

For each ctDNA-MRD assay, the recorded data points included the date of blood collection, the postoperative day of collection, ctDNA-MRD status (categorized as positive or negative), and oncentration, which encompassed mutation count, total variant allele fraction (VAF), and maximum VAF. Additionally, a comprehensive variant list was documented, detailing the gene, locus, functional annotation, and VAF. Serial measurements from the same patient were associated using a unique sample identifier.

DNA extraction

Tumour DNA was extracted from baseline tissues using TIANamp Genomic DNA Kit (TIANGEN). Buffy coat DNA was isolated with TGuide S32 automated system (TIANGEN) and magnetic‑bead kit‑T5C. Plasma cfDNA was extracted from 3.5–4 mL plasma using MagMAX Cell‑Free DNA Kit (Thermo Fisher). DNA concentration and quality were assessed by Qubit dsDNA HS Assay and Agilent 2100 BioAnalyzer. For each sample, 30–300 ng genomic DNA was sheared to –200 bp (Covaris LE220) for library prep.

Library prep & capture

Libraries were constructed with KAPA Hyper PCR‑free Kit, amplified with KAPA Library Amplification Kit, purified with AMPure XP beads, and ligated with UMI adapters. After quantification and size validation, target regions (–2.4 Mb, 769 cancer genes) were captured using HyperCap Enrichment Kit (Roche). The panel, designed and validated by Genecast (CAP‑accredited), included genes from TCGA, COSMIC, OncoKB, and internal data. Captured libraries were amplified, purified, and sequenced on NovaSeq 6000 (150‑bp paired‑end).

Tumour single nucleotide variant/insertion and deletion (SNV/InDel) calling

Reads were trimmed (Trimmomatic v0.36), aligned to hg19 (BWA v0.7.17), and duplicates marked (Picard v2.23.0). SNVs/InDels were called with VarDict v1.5.1 and complex mutations with FreeBayes v1.2.0. Variants were filtered for quality, strand bias, and excluded if overlapping low‑complexity/segmental duplication regions or in‑house error lists.

Somatic filtering

Germline/clonal hematopoiesis (CH) variants were removed if: (I) peripheral blood lymphocytes (PBL) VAF ≥5%; (II) PBL VAF <5% but >1/5 of tumour VAF; or (III) gnomAD MAF ≥2%. Remaining variants required ≥5 supporting reads and VAF thresholds of 4% (SNVs) and 5% (InDels).

Plasma ctDNA monitoring (tumour‑informed)

Only tissue‑detected SNVs/InDels were tracked in plasma. Candidates were tested against an internal background library (>1,000 plasma + matched tissue/PBL) to remove artefacts. After germline/CH filtering, remaining variants were fitted to an inverse‑gamma distribution. Zero‑inflation was addressed via Monte Carlo simulation, and each VAF was tested by binomial test (alternate/total reads). Variant‑level positivity: P<0.05; sample‑level combined P<0.01 defined a positive plasma. The calculation formula is: mean VAF = sum of positive VAFs/total trackable mutations. ctDNA concentration (hGE, molecules/mL) was calculated.

Plasma copy number variation (CNV) monitoring

CNVs in tissue and plasma were analysed by CNVkit v0.9.2 (paired PBL reference). Thresholds: gain >4.0 (tissue)/>3.0 (plasma); loss <1.0 (tissue)/<1.2 (plasma). Plasma CNV was reported only if the same gene‑level alteration was present in the matched tissue.

Plasma fusion monitoring

Fusions were identified by FACTERA v1.4.4 and FusionMap; only canonical driver fusions involving anaplastic lymphoma kinase (ALK), c-ros oncogene 1 receptor tyrosine kinase (ROS1), or rearranged during transfection (RET) kinase domains were included. Plasma was fusion‑positive if ≥1 read perfectly matched the tissue‑derived fusion reference and spanned the breakpoint.

Annotation

Variants were annotated with ClinVar, COSMIC, OncoK. To reduce sparsity, multiple variants per gene were merged into a binary variable (1 if any pathogenic/likely pathogenic/tumour‑associated).

Statistical analysis

All statistical analyses were conducted utilizing SPSS version 22.0 (IBM Corp., Armonk, NY, USA), while graphical representations were created with GraphPad Prism version 8.0 (GraphPad Software, San Diego, CA, USA). Mutation landscapes were visualised with ComplexHeatmap. Post hoc pairwise comparisons among the four subgroups were adjusted for multiple testing using the Bonferroni correction. Cumulative survival probabilities were estimated via the Kaplan-Meier (KM) method, with survival differences evaluated using the log-rank test. Both univariable and multivariable survival analyses were carried out using Cox proportional hazards regression models. Longitudinal MRD data were analyzed using R (version 4.1.2) and Python (version 3.10+). To address repeated measures and time-dependent effects, generalized estimating equations (GEE), mixed-effects models, and time-dependent survival assessments were implemented in R and Python. The Benjamini-Hochberg procedure was employed to manage the false discovery rate in the context of multiple comparisons. A two-tailed P value of less than 0.05 was regarded as indicative of statistical significance.

Results

Baseline clinical characteristics of the study cohort

A total of 124 patients diagnosed with CRC satisfied the predefined inclusion and exclusion criteria for this study. Among these patients, 36 individuals (29.03%) were identified as positive for postoperative ctDNA-MRD, whereas 88 individuals (70.97%) were negative. The cohort consisted of 68 males and 56 females, resulting in a male-to-female ratio of 1.21. The age distribution indicated that 56 patients were younger than 60 years, while 68 patients were aged 60 years or older. In terms of BMI, 12 patients had a BMI of less than 18.5 kg/m², 60 patients had a BMI ranging from 18.6 to 23.9 kg/m², and 52 patients had a BMI of 24 kg/m² or higher. With respect to CEA levels, 65 patients (52.42%) exhibited CEA levels below 5 ng/mL, and 59 patients (47.58%) had CEA levels of 5 ng/mL or greater. Tumor location was categorized as follows: right-sided colon cancer in 33 cases (26.61%), left-sided colon cancer in 35 cases (28.23%), and rectal cancer in 56 cases (45.16%). Postoperative pathological staging indicated that 52 patients were classified with stage II disease, and 72 patients were classified with stage III disease. MMR protein status was proficient (pMMR) in 109 cases and deficient (dMMR) in 15 cases. Comprehensive baseline characteristics are detailed in (Table 1).

Association of clinicopathological features and ctDNA-MRD status with progression-free and overall survival (OS)

The analysis of survival follow-up data from the cohort of 124 patients identified several factors significantly associated with PFS in univariate analysis, including ctDNA-MRD status, history of chemotherapy, thrombin time, AFP levels, tumor size, pT stage, pN stage, UICC stage, and vascular invasion (all P<0.05). Subsequent multivariable Cox regression analysis indicated that ctDNA-MRD negativity and AFP levels below 7 ng/mL were independent protective factors for PFS. Conversely, pN1 stage, and pN2 stage, were identified as independent risk factors associated with poorer PFS outcomes (refer to Table 2, Table S1 and Figure 2). In relation to OS, univariate analysis revealed significant associations with MRD status, NLR, ALB, ChE, BUN, Fib, AFP, CEA, CA125, CA153, CA199, pN stage, and UICC stage, with all associations achieving statistical significance (P<0.05). Subsequent multivariable Cox proportional hazards modeling identified ctDNA-MRD negativity, NLR <3.8, AFP <7 ng/mL, CEA <5 ng/mL, and CA125 <24 U/mL as independent protective factors for OS. In contrast, ALB <39.8 g/L, pN1 stage, and pN2 stage emerged as independent prognostic risk factors associated with diminished OS (refer to Table 2, Table S1 and Figure 3).

Figure 2 Survival analysis of clinical, hematological, and pathological factors and progression-free survival. (A) KM survival analysis of MRD status; (B) KM analysis of thrombin time; (C) KM analysis of alpha-fetoprotein; (D) KM analysis of pN stage; (E) KM analysis of tumor size. AFP, alpha-fetoprotein; KM, Kaplan-Meier; MRD, minimal residual disease; N, node; PFS, progression-free survival; pN, pathology node; TT, thrombin time.

Figure 3 Survival analysis of clinical, hematological, and pathological factors and overall survival. (A) KM survival analysis of MRD status; (B) KM analysis of NLR; (C) KM analysis of albumin; (D) KM analysis of alpha-fetoprotein; (E) KM analysis of cholinesterase; (F) KM analysis of carcinoembryonic antigen; (G) KM analysis of carbohydrate antigen 125; (H) KM analysis of carbohydrate antigen 153; (I) KM analysis of pN stage. AFP, alpha-fetoprotein; ALB, albumin; CA125, carbohydrate antigen 125; CA153, carbohydrate antigen153; CEA, carcinoembryonic antigen; ChE, cholinesterase; KM, Kaplan-Meier; MRD, minimal residual disease; N, node; NLR, neutrophil-lymphocyte ratio; OS, overall survival; pN, pathology node.

Association between ctDNA-MRD longitudinal trajectory and clinicopathological factors

Our findings reveal that 14.77% (13/88) of patients with negative MRD nonetheless developed distant metastasis, a recurrence pattern consistent with the metastatic risk observed in previously diagnosed stage II–III CRC (15). To investigate the underlying factors contributing to recurrence and metastasis in ctDNA-MRD negative individuals, we incorporated a cohort of patients who underwent two or more ctDNA-MRD assessments (n=40). Subsequently, we conducted a subgroup stratification analysis based on their prognostic outcomes to elucidate their distinct prognostic characteristics. In this cohort of patients, univariate analysis revealed significant correlations among BMI, chemotherapy history, MLR, pT stage, UICC stage, and vascular invasion across subgroups (P<0.05) (Table 3 and Table S2). However, after conducting post hoc pairwise comparisons with Bonferroni correction, only BMI remained significantly different between the false negative (FN) group and the true negative (TN) group. No other clinicopathological factors exhibited significant associations across the subgroups (Tables 4-6).

ctDNA-MRD longitudinal trajectory stratifies postoperative recurrence risk

To further elucidate the factors contributing to subgroup differences, a longitudinal analysis of ctDNA-MRD trajectories was conducted within a cohort of 40 CRC patients. Sequencing results identified recurrent alterations in canonical CRC driver genes and pathways (Tables S3,S4). The genes most frequently mutated were APC and TP53, followed by additional recurrent mutations in genes associated with chromatin regulation and DNA repair (Figure 4A; Table S3). At the level of signaling pathways, there was notable enrichment in the Wnt/β-catenin signaling pathway, chromatin remodeling/epigenetic regulation, p53/cell cycle control, and DNA damage repair pathways, consistent with known CRC tumor biology (Figure 4B; Table S4). The integration of postoperative ctDNA-MRD results (Tables S5,S6) revealed that patients who eventually experienced recurrence exhibited a higher baseline ctDNA-MRD breadth and burden compared to those who did not recur, with the baseline breadth showing a more pronounced difference between the post-treatment (TP) and treatment-naive (TN) time points (Figure 4C; Table S7). Additionally, baseline ctDNA-MRD burden was negatively correlated with the time to recurrence (Figure 4D). Importantly, longitudinal trajectories of ctDNA-MRD offered more precise risk stratification compared to cross-sectional analyses. The tumor-informed VAF displayed distinct patterns among different groups at various postoperative intervals (Figure 5A; Table S5). In patients who experienced relapse, the ctDNA-MRD signal was frequently persistently detectable and exhibited an increasing trend during follow-up, whereas patients who did not relapse generally showed absent or declining signals. Stratification based on patient outcomes, contrasting true positive (TP)/FN (relapsed) with false positive (FP)/TN (non-relapsed) cases, highlighted the clinical significance of dynamic ctDNA-MRD monitoring (Figure 5B,5C).

Figure 4 Gene mutation detection and baseline status in subgroups of colorectal cancer. (A) Gene mutation information in subgroups; (B) enriched biological pathways in subgroups; (C) baseline gene mutation status in subgroups; (D) mutation burden and recurrence. AF, allele fraction; FN, false negative; FP, false positive; TN, true negative; TP, true positive.

Figure 5 Longitudinal trajectory analysis of ctDNA-MRD in colorectal cancer. (A) Longitudinal ctDNA-MRD trajectories in patients with serial MRD monitoring. (B) Trajectories in patients who developed recurrence/metastasis. (C) Trajectories in patients without recurrence/metastasis. The x-axis represents days after surgery, with surgery defined as the reference date. Each horizontal line represents the follow-up interval for one patient. Each colored circle represents one plasma ctDNA-MRD test at the corresponding follow-up time point, and circle size is proportional to the tumor-informed VAF/AF signal at that time point; very small circles indicate undetectable or near-zero signal. Black “X” symbols indicate clinically or radiologically confirmed recurrence/metastasis. Group definitions were standardized as follows: TP, MRD-positive with recurrence/metastasis; FN, MRD-negative with recurrence/metastasis; FP, MRD-positive without recurrence/metastasis; TN, MRD-negative without recurrence/metastasis. AF, allele fraction; ctDNA, circulating tumor DNA; FN, false negative; FP, false positive; MRD, minimal residual disease; TN, true negative; TP, true positive; VAF, variant allele frequency.

Variant allele fractions and detection breadth across the four subgroups

In the cohort of 40 patients with CRC, the mean VAF value in the ctDNA-MRD positive group was significantly elevated compared to the ctDNA-MRD negative group (P=0.04), suggesting an association between ctDNA-MRD positive status and increased ctDNA intensity (refer to Figure 6A and Table S8). In terms of detection breadth, TP cases generally demonstrated a broader detection range compared to TN cases. Notably, FN cases, despite eventual recurrence, exhibited limited detection breadth, which is consistent with a restricted traceable signal and/or borderline detectability. However, no significant differences in detection breadth were identified among the four subgroups (see Figure 6B). Among the subset of 23 patients who underwent two or more postoperative plasma ctDNA-MRD monitoring time points, the total VAF value in the recurrence group was higher than that in the non-recurrence group, although this difference did not achieve statistical significance (refer to Figure 6C). The TP cases are characterized by consistently elevated ctDNA-MRD levels with extensive traceability. TN cases exhibit persistently absent or low ctDNA-MRD levels. FP cases show detectable but non-progressing ctDNA-MRD, which may indicate transient shedding, treatment effects, or a non-progressive residual signal. FN cases display low breadth and intensity of ctDNA-MRD despite poor clinical outcomes. These findings indicate that cross-sectional ctDNA-MRD negativity can still occur under unfavorable kinetic conditions when the traceable signal is limited or remains near the detection threshold (see Figure 7).

Figure 6 Analysis of AF values and detection breadth among the four subgroups. (A) Analysis of average AF values among the four subgroups; (B) analysis of detection breadth among the four subgroups; (C) analysis of total AF values among the four subgroups in the 23-patient population. AF, allele fraction; FN, false negative; FP, false positive; MRD, minimal residual disease; TN, true negative; TP, true positive.

Figure 7 Manifestations of the longitudinal trajectory status of MRD among the four subgroups. ctDNA, circulating tumor DNA; FN, false negative; FP, false positive; MRD, minimal residual disease; TN, true negative; TP, true positive.

Discussion

Recurrence and metastasis are key factors for having poor outcomes after CRC surgery, with conventional diagnostics failing to detect metastasis early enough for timely treatment (18). Detecting ctDNA-MRD allows for identifying microscopic residual disease not visible through standard imaging, offering a new approach for postoperative risk assessment and monitoring (19). This study found that: (I) ctDNA-MRD status is significantly linked to PFS and OS, serving as an independent prognostic marker; (II) Patients with metastasis consistently showed detectable and increasing ctDNA-MRD signals, while those without metastasis had absent or decreasing signals. This suggests that tracking ctDNA-MRD over time provides more valuable insights than a single MRD measurement.

In this study, the ctDNA-MRD positivity rate for stage II–III CRC was 29.03% (36/124), with higher rates in these stages due to the increased residual risk (19). However, variability across studies frequently emerges from several factor: (I) technological differences, including platforms [polymerase chain reaction (PCR), next-generation sequencing (NGS), methylation] and positivity thresholds, which affect comparability (20-22); and (II) real-world selection bias, where high costs and retrospective follow-up may exclude non-compliant patients, leading to biased cohorts. Consequently, comparisons of ctDNA-MRD positivity rates across different studies should be conducted with caution.

ctDNA-MRD is a key prognostic factor for stage II–III CRC, aligning with previous research (23). ctDNA-MRD, made up of DNA fragments from tumor cells, indicates circulating tumor burden and potential residual disease (3). Postoperative ctDNA-MRD positivity means tumor signals remain at the molecular level, suggesting a molecular cure might not be achieved even after R0 resection, increasing recurrence risk. Therefore, ctDNA-MRD is valuable for postoperative risk assessment, follow-up planning, and personalized treatment strategies.

The optimal timing for initiating adjuvant chemotherapy is generally acknowledged to be two weeks following surgery, with ctDNA-MRD status serving as a valuable tool for guiding treatment decisions. Concurrently, residual ctDNA-MRD present in the circulation may be further reduced through adjuvant chemotherapy. To facilitate a dynamic and precise assessment of the CRC patient cohort, blood samples were collected at multiple predefined intervals throughout this study for longitudinal ctDNA-MRD analysis. This approach to dynamic surveillance enables the provision of intensified postoperative therapy for high-risk patients, while allowing for the de-escalation of treatment intensity in low-risk individuals. Consequently, this strategy mitigates chemotherapy-related toxicity and reduces the frequency of ctDNA-MRD testing, thereby avoiding unnecessary costs associated with repeated assays.

In addition, analysis of the longitudinal ctDNA-MRD trajectory revealed a significant increase in the mean VAF within the recurrence group (P=0.047), accompanied by a sustained upward trend. This observation suggests that reliance solely on radiographically detectable lesions for recurrence detection is insufficient. Instead, two consecutive elevations in VAF should be regarded as an early indicator of biological recurrence, warranting earlier clinical attention and optimization of intervention timing. In the clinical application of MRD monitoring, it is essential to consider both the assay turnaround time and the associated economic costs. In this study, the interval between the first postoperative test and the initiation of chemotherapy was deemed adequate, thereby preventing treatment delays.

To enable more frequent dynamic monitoring, a risk-adapted strategy is advised, prioritizing resource allocation towards high-risk patients specifically those with pN2 stage or initially MRD-positive results for intensified follow-up. Conversely, for low-risk patients with consistently negative results, the frequency of testing may be decreased. This stratified management approach not only aligns with health economic principles but also enhances the predictive utility of the ctDNA-MRD trajectory in forecasting recurrence.

This study confirmed classic CRC driver mutations, specifically in the (APC and TP53) genes (24), and demonstrated that patients experiencing recurrence exhibited a broader spectrum of variant detection and elevated ctDNA levels post-surgery, negatively correlating with baseline burden and recurrence time. The breadth of ctDNA-MRD reflects tumor heterogeneity, with a broader range indicating stronger adaptability and recurrence potential. Clinically, the breadth and depth of the first post-surgery ctDNA-MRD are crucial quantitative indicators for predicting the growth and invasiveness of residual tumors.

Our analysis at a single baseline showed no significant differences in mean VAF values across four groups, with some overlap between FP and TP, and FN and TN groups. This suggests intensity alone cannot distinguish outcomes. However, longitudinal data provided better insights: the TP group often had increasing signals, while the TN group remained negative. The FP group showed limited signal growth, possibly due to temporary shedding, treatment effects, or low-level residual disease. The FN group had low intensity but still recurred, likely due to low tumor burden, minimal ctDNA-MRD shedding, specific metastatic sites, or missed sampling. Overall, these findings underscore the importance of trajectory morphology and dynamic information in interpreting ctDNA-MRD positive or negative outcomes and lay the groundwork for optimizing follow-up strategies.

The TP group showed continuous ctDNA-MRD growth and mutation tracking, with conventional chemotherapy often ineffective. The TN group consistently showed no ctDNA-MRD, suggesting that might avoid chemotherapy. The FP group had low or limited ctDNA-MRD growth, providing more insight than a binary state and reducing unnecessary chemotherapy. Despite negative ctDNA-MRD, the FN group faced recurrence and metastasis, possibly due to low DNA shedding in certain tumors or limitations in current sequencing technology for detecting low-frequency signals. In the future, integrating multi-omics ctDNA-MRD strategies based on epigenetics (such as DNA methylation, fragmentation characteristics) or increasing sequencing depth will be key to overcoming the FN population.

Limitations

There are several limitations in this study due to its single-center, retrospective design. There is selection bias in ctDNA-MRD testing and follow-up adherence, and variability in blood collection intervals may affect longitudinal indicators. The matched cohort sample size is small, especially for FN and FP phenotypes, making some comparisons statistically insignificant and exploratory. Limited events for OS may cause unstable hazard ratio estimates with wide confidence intervals. Additionally, the ctDNA-MRD detection platform and threshold are specific to the center, limiting cross-study comparability. Future research should involve prospective, multi-center cohort studies to validate and standardize the findings.

Conclusions

In CRC, the clinical application of ctDNA-MRD has evolved from a simple binary evaluation to a more complex analysis. By evaluating the breadth, intensity, and velocity of ctDNA-MRD over time, we can accurately trace the tumor’s microscopic evolution in the patient. Clinical trials in the future should incorporate ctDNA-MRD for choosing patients and evaluate changes in ctDNA-MRD dynamics as an alternative endpoint for determining the efficacy of novel therapies.

Acknowledgments

We thank The First Affiliated Hospital of Kunming Medical University. We also acknowledge the contributions of the research team, data analysts, and medical staff in patient care and data collection. We would like to thank Tawfik A. H. Alburiahi for his help in polishing our paper.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0476/rc

Data Sharing Statement: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0476/dss

Peer Review File: Available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0476/prf

Funding: This work was supported by the projects of Wu Jieping Medical Foundation (No. KYFY-KJ01-202504-0225) and the Science and Technology Projects of Yunnan Universities Serving Key Industries (No. FWCY-BSPY2025076).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jgo.amegroups.com/article/view/10.21037/jgo-2026-0476/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of The First Affiliated Hospital of Kunming Medical University (No. 2025-L-27). Due to the retrospective nature of the study and the use of de-identified data, the requirement for informed consent was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Siegel RL, Kratzer TB, Giaquinto AN, et al. Cancer statistics, 2025. CA Cancer J Clin 2025;75:10-45. [Crossref] [PubMed]
2. Robinson TP, Kaiser K, Lark M, et al. NCCN guideline concordance in colon and rectal cancer patients within a comprehensive health system. Am J Surg 2025;240:116114. [Crossref] [PubMed]
3. Kotani D, Oki E, Nakamura Y, et al. Molecular residual disease and efficacy of adjuvant chemotherapy in patients with colorectal cancer. Nat Med 2023;29:127-34. [Crossref] [PubMed]
4. Boute TC, Swartjes H, Greuter MJE, et al. Cumulative Incidence, Risk Factors, and Overall Survival of Disease Recurrence after Curative Resection of Stage II-III Colorectal Cancer: A Population-based Study. Cancer Res Commun 2024;4:607-16. [Crossref] [PubMed]
5. Dong A, Jahedi Zargar M, Mirzazadeh A, et al. Circulating biomarkers in leukemia. Clin Chim Acta 2026;584:120865. [Crossref] [PubMed]
6. Moding EJ, Shahrokh Esfahani M, Jin C, et al. Integrating ctDNA Analysis and Radiomics for Dynamic Risk Assessment in Localized Lung Cancer. Cancer Discov 2025;15:1609-29. [Crossref] [PubMed]
7. Venetis K, Pepe F, Pescia C, et al. ESR1 mutations in HR+/HER2-metastatic breast cancer: Enhancing the accuracy of ctDNA testing. Cancer Treat Rev 2023;121:102642. [Crossref] [PubMed]
8. Lewis AR, Valle JW, McNamara MG. Pancreatic cancer: Are "liquid biopsies" ready for prime-time? World J Gastroenterol 2016;22:7175-85. [Crossref] [PubMed]
9. Taliento C, Morciano G, Nero C, et al. Circulating tumor DNA as a biomarker for predicting progression-free survival and overall survival in patients with epithelial ovarian cancer: a systematic review and meta-analysis. Int J Gynecol Cancer 2024;34:906-18. [Crossref] [PubMed]
10. Negro S, Pulvirenti A, Trento C, et al. Circulating Tumor DNA as a Real-Time Biomarker for Minimal Residual Disease and Recurrence Prediction in Stage II Colorectal Cancer: A Systematic Review and Meta-Analysis. Int J Mol Sci 2025;26:2486. [Crossref] [PubMed]
11. Masfarré L, Vidal J, Fernández-Rodríguez C, et al. ctDNA to Guide Adjuvant Therapy in Localized Colorectal Cancer (CRC). Cancers (Basel) 2021;13:2869. [Crossref] [PubMed]
12. Parikh AR, Chee BH, Tsai J, et al. Minimal Residual Disease using a Plasma-Only Circulating Tumor DNA Assay to Predict Recurrence of Metastatic Colorectal Cancer Following Curative Intent Treatment. Clin Cancer Res 2024;30:2964-73. [Crossref] [PubMed]
13. Mo S, Ye L, Wang D, et al. Early Detection of Molecular Residual Disease and Risk Stratification for Stage I to III Colorectal Cancer via Circulating Tumor DNA Methylation. JAMA Oncol 2023;9:770-8. [Crossref] [PubMed]
14. Tie J, Wang Y, Lo SN, et al. Circulating tumor DNA analysis guiding adjuvant therapy in stage II colon cancer: 5-year outcomes of the randomized DYNAMIC trial. Nat Med 2025;31:1509-18. [Crossref] [PubMed]
15. Malla M, Loree JM, Kasi PM, et al. Using Circulating Tumor DNA in Colorectal Cancer: Current and Evolving Practices. J Clin Oncol 2022;40:2846-57. [Crossref] [PubMed]
16. Emiloju OE, Storandt M, Zemla T, et al. Tumor-Informed Circulating Tumor DNA for Minimal Residual Disease Detection in the Management of Colorectal Cancer. JCO Precis Oncol 2024;8:e2300127. [Crossref] [PubMed]
17. Zhang Y, Han F, Qu J. Global, regional, and national burden of early-onset colorectal cancer attributable to high BMI (1990-2021). Front Med (Lausanne) 2025;12:1631392. [Crossref] [PubMed]
18. Henriksen TV, Tarazona N, Frydendahl A, et al. Circulating Tumor DNA in Stage III Colorectal Cancer, beyond Minimal Residual Disease Detection, toward Assessment of Adjuvant Therapy Efficacy and Clinical Behavior of Recurrences. Clin Cancer Res 2022;28:507-17. [Crossref] [PubMed]
19. Nakamura Y, Kaneva K, Lo C, et al. A Tumor-Naïve ctDNA Assay Detects Minimal Residual Disease in Resected Stage II or III Colorectal Cancer and Predicts Recurrence: Subset Analysis from the GALAXY Study in CIRCULATE-Japan. Clin Cancer Res 2025;31:328-38. [Crossref] [PubMed]
20. Pantel K, Alix-Panabières C. Liquid biopsy and minimal residual disease - latest advances and implications for cure. Nat Rev Clin Oncol 2019;16:409-24. [Crossref] [PubMed]
21. Sahin IH, Yanes R, Saridogan T, et al. The Role of Circulating Tumor DNA for Management of Patients With Rectal Cancer: Challenges and Opportunities. Cancer J 2024;30:290-6. [Crossref] [PubMed]
22. Luo H, Zhao Q, Wei W, et al. Circulating tumor DNA methylation profiles enable early diagnosis, prognosis prediction, and screening for colorectal cancer. Sci Transl Med 2020;12:eaax7533. [Crossref] [PubMed]
23. Henriksen TV, Demuth C, Frydendahl A, et al. Unraveling the potential clinical utility of circulating tumor DNA detection in colorectal cancer-evaluation in a nationwide Danish cohort. Ann Oncol 2024;35:229-39. [Crossref] [PubMed]
24. Jin Y, He X, Wu Y. Gastrointestinal cancer: molecular pathogenesis and targeted therapy. Mol Biomed 2025;6:136. [Crossref] [PubMed]