Postoperative circulating tumor DNA detection predicts recurrence and guides adjuvant chemotherapy duration in high-risk stage III colorectal cancer

Abstract

BACKGROUND

Colorectal cancer (CRC) patients with high-risk stage III disease (T4 or N2) face a significant risk of postoperative recurrence despite radical surgery and adjuvant chemotherapy. The optimal duration of capecitabine plus oxaliplatin chemotherapy 3 months or 6 months remains debated, with no reliable biomarker to guide individualized decisions. Circulating tumor DNA (ctDNA), as a minimally invasive marker of minimal residual disease, has shown promise in early risk stratification, yet its role in guiding adjuvant chemotherapy duration in this population remains unclear.

AIM

To ascertain if postoperative ctDNA status (measured 4 weeks after surgery) predicts recurrence and guides decisions about the duration of adjuvant chemotherapy in patients with high-risk stage III CRC (those with T4 or N2 disease).

METHODS

This is a retrospective cohort study of 50 patients with high-risk stage III CRC who received radical surgery from January 2022 to September 2025 in Foshan First People’s Hospital. All patients received adjuvant capecitabine plus oxaliplatin chemotherapy for 3 months or 6 months postoperatively. ctDNA status was assessed by next-generation sequencing 4 weeks after surgery (prior to chemotherapy administration). Patients were subsequently divided into the ctDNA-negative group (n = 29) and ctDNA-positive group (n = 21). Survival outcomes were compared to evaluate the predictive significance of ctDNA status for prognosis and its role in guiding chemotherapy across these groups.

RESULTS

The median follow-up was 18 months (1-43 months). The recurrence rate in the ctDNA-positive group (47.62%, 10/21) was higher than that in the ctDNA-negative group (17.24%, 5/29; P = 0.021) by November 2025. Kaplan-Meier analysis showed a significant difference in recurrence-free survival (RFS) between the two groups (Log-rank P = 0.026). Univariate Cox regression analysis revealed that tumor stage [hazard ratio (HR) = 0.172, 95% confidence interval (CI): 0.035-0.860, P = 0.032], T stage (HR = 0.212, 95%CI: 0.044-1.015, P = 0.028) and postoperative ctDNA status (HR = 6.001; 95%CI: 1.203-29.927; P = 0.029) were independent risk factors for RFS. Multivariate analysis revealed tumor stage (HR = 0.241, 95%CI: 0.079-0.739, P = 0.013) and ctDNA positivity (HR = 4.304, 95%CI: 1.412-13.115, P = 0.010) as independent predictors of disease progression in this cohort. In ctDNA-positive (P > 0.05) or ctDNA-negative (P > 0.05) individuals, stratified analysis found no significant differences between 3-month and 6-month delivery of chemotherapy in recurrence rates or RFS. The recurrence rate of ctDNA-positive patients with 3-month chemotherapy was significantly higher than that of ctDNA-negative patients with 6-month chemotherapy (50.00%, 6/12 vs 13.33%, 2/15, Log-rank P = 0.038). KRAS, TP53 and APC were the most frequently detected variants among mutation spectrum analysis across all patients.

CONCLUSION

Postoperative ctDNA status is a potent independent predictor of recurrence in high-risk stage III CRC. Notably, ctDNA status constitutes additional reference information in regard to chemotherapy duration: Based on the residual risk of recurrence, ctDNA-positive patients should be treated for 6 months to decrease this risk; conversely, amongst those with undetectable ctDNA status post-therapy, these patients could consider de-escalating treatment at 3 month without compromising outcome or safety.

Key Words: Colorectal cancer; Circulating tumor DNA; Recurrence; Adjuvant chemotherapy; Prognosis

Core Tip: Postoperative circulating tumor DNA (ctDNA) provides early risk stratification in high-risk stage III colorectal cancer. In this retrospective cohort of 50 patients with T4 or N2 disease, ctDNA positivity 4 weeks after surgery was independently associated with poorer recurrence-free survival and earlier recurrence, with an approximately fourfold higher adjusted recurrence risk. Exploratory analyses suggested that ctDNA status may help inform adjuvant capecitabine plus oxaliplatin duration, although no significant within-group differences were observed between 3 months and 6 months of treatment. These findings support ctDNA as a promising biomarker requiring prospective validation.

INTRODUCTION

Colorectal cancer (CRC) is the third most common type of cancer and the second leading cause of cancer-related death globally, with 1.9 million new cases and more than 900000 deaths estimated each year[1,2]. The management of high-risk stage III CRC in the postoperative setting is a therapeutic conundrum that raises one critical issue: How to balance the effectiveness of adjuvant chemotherapy with therapy-related toxicities? While oxaliplatin-based adjuvant chemotherapy has increased cure rates by 10%-20%, approximately 20%-30% of individuals derive no benefit and nearly half are incapable of completing treatment due to life-threatening toxicities[3]. This therapeutic dilemma of over-treating some and potentially under-treating others demonstrates the urgent need for biomarkers to identify patients who can benefit from intense treatment vs those who can safely undergo de-escalation.

The International Duration Evaluation of Adjuvant collaborative trial attempted to tackle this issue through risk stratification based on pathological factors: Low-risk patients (T1-3N1) would get 3 months of capecitabine plus oxaliplatin (CapOX), while high-risk patients (T4 or N2) should receive 6 months[4-6]. Measuring only tumor-node-metastasis-based stratification remains limited; however, it cannot identify which high-risk patients will experience recurrence despite finishing standard treatment and which can safely de-escalate therapy[6]. These critical deficits underscore the call for dynamic, molecular biomarkers that reflect residual disease burden after surgery and impact decisions surrounding adjuvant chemotherapy duration.

Postoperative circulating tumor DNA (ctDNA) represents a promising solution to this clinical problem[7-10]. As a sensitive marker of minimal residual disease (MRD), ctDNA at 4 weeks post-surgery may identify high-risk patients for recurrence prior to clinically detectable disease[8-10]. Recent landmark studies have illustrated the strength of ctDNA prognostication: CtDNA direct management resulted in 50% less unnecessary chemotherapy for stage II patients without compromising outcomes, as seen in the DYNAMIC trial[11,12]; meanwhile, the CIRCULATE-Japan GALAXY study confirmed that ctDNA-positive stage II-III patients significantly benefit from adjuvant chemotherapy[13,14]. However, the utility of ctDNA status in decisions on chemotherapy duration in high-risk stage III patients remains underexplored.

We propose that postoperative ctDNA status can differentiate high-risk patients into prognostically-informative groups, guiding less vs more intensive chemotherapy treatment decisions: CtDNA-positive patients who need 6-month regimens and bear the toxicity, while ctDNA-negative patients may de-escalate to a 3-month regimen. This study tests this hypothesis by determining whether postoperative ctDNA detection predicts recurrence and informs the duration of adjuvant chemotherapy in patients with high-risk stage III CRC.

MATERIALS AND METHODS

Study subjects

A retrospective review of high-risk stage III CRC patients from January 2022 to September 2025 was performed in Foshan First People’s Hospital.

Inclusion criteria: (1) Pathologically confirmed CRC meeting criteria for high-risk disease stage III (T4 or N2 or both) (n = 68 screened, 50 met this criterion); (2) Received radical surgery (All eligible patients received radical operation); (3) Age > 18 years (none excluded for age), and a Karnofsky Performance Status score > 70%; and (4) Distant metastases were excluded by preoperative computed tomography (CT) and magnetic resonance imaging (MRI) of chest, upper abdomen and pelvis with liver MRI to confirm absence of metastatic disease (3 patients were excluded for metastatic disease). A total of 50 patients met all inclusion criteria and were enrolled.

The major exclusion criteria were: (1) Those with a past or concurrent history of other malignant tumors; (2) Preoperative neoadjuvant therapy or conversion therapy; (3) Pregnant or lactating women, patients in terminal status; (4) Postoperative expected survival time < 6 months; and/or (5) Complicated with severe cardiac, pulmonary, hepatic or renal insufficiency.

This study was approved by the Hospital Ethics Committee, approval No. Lun Shen Yan[2022] No. 57, and conformed to the principles outlined in the Declaration of Helsinki. A total of 50 patients were eventually included, comprising 27 males and 23 females (mean age 63.28 ± 10.53 years). All patients underwent radical surgery and postoperatively received 3 or 6 months of adjuvant CapOX chemotherapy. Adjuvant chemotherapy duration was not randomized, but instead were separated according to case-by-case clinical decision-making based on Eastern Cooperative Oncology Group performance status, baseline comorbidities, early treatment-related toxicities, especially peripheral neuropathy, and patient preference after multidisciplinary discussion. Out of the 50 patients, 27 received chemotherapy for 3 months and 23 received chemotherapy for 6 months.

Data collection

Patient baseline data included age, sex, medical history, tumor marker levels [such as carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9], tumor location (including ampullary region carcinoid tumors), tumor size (in cm), pathological staging based on the current American Joint Committee on Cancer staging system, differentiation grade and lymph node metastasis status, and adjuvant treatment information. ctDNA detection was performed from preoperative blood samples collected from patients 4 weeks after their operation and prior to beginning adjuvant chemotherapy. These plasma samples were isolated within a timespan of 2 hours from collection time and stored at -80 °C until further application. Genetic testing was conducted using next-generation sequencing (NGS) technology.

ctDNA detection methods

ctDNA was detected using the Zhenxin’an MRD Panel (Shanghai Tongshu Biotechnology Co., Ltd.) based on NGS technology following the manufacturer’s standard operating procedures. This panel covers 36 gastrointestinal tumor-associated genes. Plasma cell-free DNA was extracted, amplified, and sequenced for target regions. Variants were considered ctDNA-positive if they had an allele frequency ≥ 0.2%, consistent with thresholds adopted in prior published studies of NGS-based ctDNA detection in CRC[7,8]. Detection was performed using standard procedures for quality control, library preparation, sequencing and bioinformatics analysis. Patients with one or more tumor-specific genetic variants detected in the plasma were classified as ctDNA-positive, and all other samples were classified as ctDNA-negative. Importantly, ctDNA was assessed at a single time-point (4 weeks post-surgery, before starting chemotherapy). This study did not perform serial measurements to quantify ctDNA during/after chemotherapy.

Follow-up and outcome measures

All patients were followed up routinely, the postoperative follow-up took place at 3-month intervals. Follow-up consisted of tumor markers, imaging scans (CT, MRI or positron emission tomography-CT) and colonoscopic assessment to define recurrence as accurately as possible. Mean follow-up time was until November 2025. The primary endpoint of this study was recurrence-free survival (RFS), which was defined as time from surgical resection to first evidence of local recurrence or distant metastasis or death.

Statistical analysis

Statistical analyses were conducted with SPSS 26.0 software (IBM Corp., Armonk, NY, United States). Categorical variables were expressed as frequencies (percentages), and intergroup comparisons were performed using a χ² test. The Cox proportional hazards regression model was used to compare the association between ctDNA status and RFS. The Kaplan-Meier method was used to create survival curves, and differences in survival outcomes between groups were compared using Log-rank test. P < 0.05 two-tailed was considered statistically significant.

RESULTS

Characterization of ctDNA mutational spectrum

Plasma samples were obtained to detect ctDNA in 50 enrolled high-risk stage III CRC participants 4 weeks after surgery. ctDNA genetic variants were positive in 21 (42%) cases. KRAS (42.3%), TP53 (23.8%) and APC (19.0%) were the most frequently mutated genes among all 50 enrolled patients. Seven other mutated genes were identified, including NRAS, PIK3CA, BRAF, ERBB2, RET and NTRK. The distribution of identified mutations is summarized in Figure 1.

Figure 1 Circulating tumor DNA mutation spectrum in high-risk stage III colorectal cancer patients (all 50 enrolled patients). The bar chart illustrates the mutation frequencies of key oncogenes and tumor suppressor genes detected in circulating tumor DNA among 50 enrolled high-risk stage III colorectal cancer patients. KRAS exhibited the highest mutation frequency (42.3%), followed by TP53 (23.8%), APC (19.0%), and NRAS (14.3%). PIK3CA and BRAF showed intermediate frequencies (9.5% and 9.3%, respectively), while ERBB2, RET, and NTRK were the least frequently mutated genes (4.0% each). These findings indicate that KRAS and TP53 mutations are the predominant somatic alterations identified in the circulating tumor DNA of this patient cohort.

Clinico-pathological features associated with postoperative ctDNA status

The positive rate of post-operation ctDNA was 42% (21/50). In addition, correlation analysis of ctDNA status with the clinicopathological characteristics indicated that only tumor recurrence rate (P = 0.021) was significantly correlated with ctDNA status. There was no difference between the two groups with respect to age, sex, tumor site, differentiation degree, tumor stage (IIIb/IIIc), T stage (T3/T4), N stage (N1/N2), perineural invasion and vascular invasion or CEA level (all P > 0.05). These results are summarized in Table 1.

Table 1 Relationship between postoperative circulating tumor DNA status and clinicopathological features of high-risk stage III colorectal cancer.

Clinicopathological features	n	Postoperative ctDNA		χ2	P value
		Negative (n = 29)	Positive (n = 21)
Age
< 60 years	20	9	11	2.313	0.128
≥ 60 years	30	20	10
Sex
Male	27	15	12	0.144	0.704
Female	23	14	9
Tumor location
Rectum	16	11	5	1.116	0.291
Colon	34	18	16
Tumor differentiation
Moderate/well-differentiated	30	17	13	0.055	0.815
Poorly differentiated/mucinous carcinoma	20	12	8
Tumor stage
IIIb	29	20	9	3.408	0.065
IIIc	21	9	12
T stage
T3	26	17	9	1.213	0.271
T4	24	12	12
N stage
N1	17	11	6	0.475	0.490
N2	33	18	15
Perineural invasion
Yes	31	19	12	0.363	0.547
No	19	10	9
Vascular invasion
Yes	25	12	13	2.053	0.152
No	25	17	8
CEA (ng/mL)
< 5	23	12	11	0.539	0.441
≥ 5	27	17	10
Recurrence
Yes	15	5	10	5.352	0.021
No	35	24	11

ctDNA: Circulating tumor DNA; CEA: Carcinoembryonic antigen.

Recurrence-free survival analysis

In univariate Cox proportional hazards regression analysis, tumor stage (IIIb vs IIIc), T stage (T3 vs T4) and postoperative ctDNA status were independent risk factors for CRC recurrence in high-risk stage III CRC patients (all P < 0.05). In contrast, age, sex, tumor location, differentiation grade, N stage, perineural invasion, vascular invasion and number of harvested lymph nodes and adjuvant chemotherapy duration and CEA level did not significantly correlate with RFS (all P > 0.05). Univariate associations were identified, and statistically significant variables were incorporated into a multivariable Cox proportional hazards regression model. Results confirmed that tumor stage [hazard ratio (HR) = 0.241, 95% confidence interval (CI): 0.079-0.739, P = 0.013] and postoperative ctDNA positivity (HR = 4.304, 95%CI: 1.412-13.115 P = 0.010) were independent prognostic factors for postoperative RFS in this cohort (Table 2).

Table 2 Univariable and multivariable Cox proportional hazards regression analysis of risk factors for recurrence-free survival in patients with high-risk stage III colorectal cancer.

Parameter	Univariable analysis		Multivariate analysis
	HR (95%CI)	P value	HR (95%CI)	P value
Age (≥ 60 years vs <60 years)	0.613 (0.142-2.459)	0.512	-	-
Sex (male vs female)	0.257 (0.062-1.055)	0.059	-	-
Tumor location (rectum vs colon)	0.238 (0.044-1.273)	0.093	-	-
Tumor differentiation (moderate/well-differentiated vs poorly differentiated/mucinous carcinoma)	0.899 (0.222-3.631)	0.881	-	-
Tumor stage (IIIb vs IIIc)	0.172 (0.035-0.860)	0.032	0.241 (0.079-0.739)	0.013
T stage (T3 vs T4)	0.212 (0.044-1.015)	0.028	0.561 (0.193-1.632)	0.289
N stage (N1 vs N2)	0.212 (0.044-1.015)	0.052	-	-
Perineural invasion (no vs yes)	2.580 (0.529-12.573)	0.241	-	-
Vascular invasion (no vs yes)	1.277 (0.263-6.207)	0.761	-	-
Number of harvested lymph nodes (≥ 12 vs < 12)	1.678 (0.366-7.681)	0.505	-	-
Adjuvant chemotherapy duration (3 months vs 6 months)	2.852 (0.527-15.434)	0.224	-	-
CEA (ng/mL) (< 5 vs ≥ 5)	0.526 (0.116-2.381)	0.405	-	-
Postoperative ctDNA status (positive vs negative)	6.001 (1.203-29.927)	0.029	4.304 (1.412-13.115)	0.010

ctDNA: Circulating tumor DNA; HR: Hazard ratio; CI: Confidence interval; CEA: Carcinoembryonic antigen.

Total survival based on ctDNA status

Based on the ctDNA-positive group, recurrence was observed in 47.62% (10/21) of patients, while the rate was 17.24% (5/29) in the ctDNA-negative group, with a significant difference between groups (χ² = 5.352, P = 0.021) at follow-up cut-off date (November 2025). Kaplan-Meier survival analysis showed that RFS was shorter in ctDNA-positive patients compared with ctDNA-negative patients (Log-rank χ² = 4.918, P = 0.026) (Figure 2A).

Figure 2 Kaplan-Meier analysis of recurrence-free survival according to postoperative circulating tumor DNA status and adjuvant chemotherapy duration. A: Kaplan-Meier curves for recurrence-free survival (RFS) stratified by postoperative circulating tumor DNA (ctDNA) status. The curves represent ctDNA-positive and ctDNA-negative patients (Log-rank P = 0.026); B: Kaplan-Meier curves for RFS in ctDNA-positive patients stratified by chemotherapy duration. The curves represent the 3-month and 6-month capecitabine plus oxaliplatin (CapOX) chemotherapy subgroups (Log-rank P = 0.689); C: Kaplan-Meier curves for RFS in ctDNA-negative patients stratified by chemotherapy duration. The legend represents the 3-month and 6-month CapOX chemotherapy subgroups (Log-rank P = 0.512); D: Kaplan-Meier curves for RFS comparing ctDNA-positive patients receiving 3 months of CapOX chemotherapy and ctDNA-negative patients receiving 6 months of CapOX chemotherapy (Log-rank P = 0.038). ctDNA: Circulating tumor DNA; CapoX: Capecitabine plus oxaliplatin.

Stratification of chemotherapy duration by ctDNA status

To assess the usefulness of ctDNA status to personalize chemotherapy duration, stratified analyses were conducted in each subgroup based on ctDNA status according to chemotherapy durations (3 months vs 6 months). For ctDNA-positive patients: 6-month chemotherapy group, n = 9 (44.44%, 4/9) vs the 3-month chemotherapy group, n = 12 (50.00%, 6/12). There was no difference in recurrence rate (χ² = 0.125, P = 0.724) or RFS curves (Log-rank P = 0.689) between these two subgroups (Figure 2B). For ctDNA-negative patients: Recurrence rates were 13.33% (2/15) and 21.43% (3/14) in the 6-month chemotherapy and 3-month chemotherapy group, respectively. No major differences in recurrence rate (χ² = 0.457, P = 0.499) or RFS curves (Log-rank P = 0.512; Figure 2C) were found between these two subgroups.

In the ctDNA-negative patients completing 6-month chemotherapy, RFS was statistically significantly lower than that of ctDNA-positive patients receiving 3-month chemotherapy (13.33%, n = 2/15 vs 50.00% (n = 6/12); χ² = 4.356; P = 0.038)(Figure 2D).

ctDNA status-based time to recurrence analysis

The dynamics of timing analysis of recurrence were also different in ctDNA-positive and ctDNA-negative patients. Median time to recurrence was 11.0 months (interquartile range: 7.5-15.0 months) for the 10 ctDNA-positive patients. While one patient experienced recurrence at > 3 years, most (70.0%, n = 7/10) recurred within the first year following surgery. In contrast, the median time to recurrence in the five ctDNA-negative patients was significantly longer at 18.0 months (interquartile range: 14.0-22.0 months), with only a minority of them experiencing recurrence < 12 months after surgery (1/5, 20.0%) (P = 0.041). These time-associated recurrence patterns suggest that postoperative ctDNA status not only predicts recurrence but also provides critical information on when such a recurrence is likely to occur, as ctDNA-positive patients are at much higher risk for early relapse, possibly in need of extended chemotherapy and intensive follow-up procedures (Figure 3A and B).

Figure 3 Time to recurrence analysis by circulating tumor DNA status and chemotherapy duration. Box plots showing the distribution of time to recurrence. A: Comparison between circulating tumor DNA (ctDNA)-positive (n = 10) and ctDNA-negative (n = 5) patients. ctDNA-positive patients had significantly shorter median time to recurrence (11.0 months vs 18.0 months, P = 0.041, Mann-Whitney U test); B: Comparison between 3-month (n = 6) and 6-month (n = 4) chemotherapy durations in ctDNA-positive patients. No significant difference was observed (9.5 months vs 14.0 months, P = 0.167). Box plots display the median (horizontal line), mean (diamond), interquartile range (box), and range (whiskers). ctDNA: Circulating tumor DNA; NS: Not significant.

DISCUSSION

ctDNA consists of cell-free DNA fragments shed by cancer cells into body fluids, and its profile closely matches that of the primary tumor tissue. ctDNA can be released into circulation by as few as 1 million tumor cells and can be detected through polymerase chain reaction[15] or NGS[16]. Plasma levels of tumor-derived cell-free DNA are a sensitive and important biomarker for evaluating risk of tumor resurgence following surgery and can be clinically significant for malignancy recurrence prediction, guiding adjuvant chemotherapeutic strategies, monitoring therapeutic response, and predicting treatment failure through the identification of recurrent and metastatic lesions in at-risk patients[17].

In this retrospective cohort study of 50 high-risk stage III CRC patients, RFS was higher in ctDNA-negative patients compared with ctDNA-positive patients (82.76%, 24/29 vs 52.38%, 11/21; χ² = 5.352, P = 0.021). Consistent with this result, Kaplan-Meier survival analysis indicated that ctDNA-positive patients had shorter RFS compared with ctDNA-negative patients (Log-rank χ² = 4.918, P = 0.026). Univariate and multivariable Cox proportional hazards regression analyses confirmed that ctDNA positivity was an independent predictor of postoperative recurrence (HR = 6.001, 95%CI: 1.203-29.927, P = 0.029 for univariable analysis; HR = 4.304, 95%CI: 1.412-13.115, P = 0.010 for multivariable analysis) and lower RFS (Table 2).

ctDNA is a key biomarker for monitoring tumor heterogeneity and genomic alterations, detecting postoperative MRD, and predicting recurrence in CRC[7-10]. Moreover, mutation spectrum analysis in this cohort showed that KRAS, TP53 and APC were the most frequently observed variants across patients. Recurrence patients had significantly higher frequencies of KRAS and TP53 mutations, reaffirming earlier findings that these mutations are associated with aggressive tumor biology and chemoresistance[18,19]. Several studies have further validated that plasma ctDNA status after curative resection is correlated with poor RFS and increased recurrence risk: CtDNA-positive patients show significantly increased rates of recurrence (1.55- to 18-fold higher recurrence risk compared to ctDNA negative patients) and significantly shorter time to recurrence[16]. Postoperative ctDNA status is more prognostic for RFS than any single or combined clinicopathological risk factors among a wide spectrum of clinicopathological variables associated with RFS[14]. Likewise, consistent with our results, the RFS of ctDNA-negative patients is approximately 36.7% higher than those that are ctDNA-positive (52.38% vs 82.76%). The presence of ctDNA led to a four-fold increase in hazard ratio for recurrence, and thus lower RFS, according to Cox regression after proper adjustment of several confounding factors[8]. The results of this study indicate that postoperative ctDNA status in patients with high-risk stage III CRC can lead to finer stratification of RFS, particularly because ctDNA positivity is consistently associated with an increased recurrence risk[17].

Postoperative ctDNA status also has valuable implications for tailoring adjuvant chemotherapy strategies to maximize RFS. In patients with stage II colon cancer, a randomized ctDNA-guided strategy reduced the use of adjuvant chemotherapy without compromising RFS, supporting the potential value of ctDNA for treatment de-escalation[11,12]. In addition, postoperative adjuvant chemotherapy was associated with improved disease-free survival among patients with ctDNA-positive high-risk stage II or III CRC, whereas no clear benefit was observed among ctDNA-negative patients[13]. However, whether postoperative ctDNA can reliably guide the duration of adjuvant chemotherapy in high-risk stage III CRC remains uncertain and requires prospective validation.

Stage III CRC patients at a high-risk clinical stage (T4 or N2 stage) have poor prognosis and are clinically recommended for 6 months of adjuvant chemotherapy with the CapOX regimen[20,21]. Nonetheless, long-term chemotherapy administration is linked to particular toxicities; oxaliplatin-related neurotoxicity can occur within 3 months and is irreversible[21,22]. The 6-month chemotherapy group had a significantly higher incidence of ≥ grade 3 neurotoxicity than the 3-month chemotherapy group, and the total toxic burden for those receiving the 6-month regimen is approximately five times that of those receiving the 3-month regimen[22].

Clinically, nearly half of the patients fail to complete the full 6-month treatment course due to intolerable toxicities[20], which may compromise RFS. Nevertheless, multiple studies have confirmed that the RFS of stage III patients with high-risk pathological factors is correlated with chemotherapy duration, and insufficient treatment often results in lower RFS[6,21,22]. In the present study, among ctDNA-positive patients, those who received 3-month short-course chemotherapy had the lowest RFS (50.00%), which was significantly lower than the RFS of ctDNA-negative patients who completed 6-month chemotherapy (86.67%, P = 0.038). Therefore, our findings suggest that postoperative ctDNA status may provide valuable reference information when considering adjuvant chemotherapy strategies in patients with high-risk stage III CRC. ctDNA-positive patients appeared to have poorer outcomes when receiving shorter chemotherapy duration, while ctDNA-negative patients maintained relatively favorable RFS regardless of chemotherapy duration. However, these observations should be interpreted cautiously, and definitive treatment recommendations require validation in larger prospective studies.

Limitations

This study has multiple limitations that need to be highlighted. First, it is a single-center study with limited number of cases (n = 50), which limits the applicability of the RFS. Secondly, the follow-up period for this study (median: 18 months) is short, and long-term RFS of 3 years or 5 years has yet to be evaluated. Third, this study did not include data on chemotherapy-associated toxicities, and therefore, further analysis of the risk-benefit balance (toxicity vs RFS) based on different durations of chemotherapy should be performed.

CONCLUSION

In the case of high-risk stage III CRC, postoperative ctDNA status is significantly associated with RFS prognosis; a positive status indicates lower RFS during follow-up and worse outcomes for patients. Additionally, ctDNA status may represent a helpful reference for postoperative adjuvant chemotherapy to inform RFS; a full course of adjuvant treatment in ctDNA-positive patients is recommended, while ctDNA-negative patients may benefit from treatment de-escalation. However, this biomarker-driven multi-trial intervention approach to improve RFS still needs prospective validation in large-scale and multi-center individual clinical trials.