Prognostic significance of circulating tumor DNA in non-metastatic colorectal cancer

Abstract

Purpose

Circulating tumor DNA (ctDNA) is the small fragments of DNA released by tumor cells into the blood. By detecting the abnormal changes in ctDNA in the blood, it can assist in the early screening of cancer, monitoring of treatment efficacy, early warning of recurrence, and guidance for individualized treatment. Using a systematic review and meta-analysis approach, in this study, we assessed the significant prognostic value of ctDNA in non-metastatic colorectal cancer (CRC) patients.

Methods

Studies on the predictive value of ctDNA in CRC were considered. RevMan 5.3 was used to conduct meta-analyses and subgroup investigations classified by tumor population, two-year or three-year DFS, and tumor location. The analysis included 11 trials conducted with 605 individuals. Definitions of positive ctDNA: Detection of ≥ 1 tumor-informed mutation, or tumor methylation signature above threshold, or combined mutation and fragmentome analysis.

Results

Five studies investigated the associations between pre-operative ctDNA status and patient outcomes. Patients who had higher ctDNA levels had considerably lower DFS and OS, but these differences were not statistically significant. 10 studies investigated how post-operative ctDNA status affects patient outcomes. Patients with high levels of ctDNA had worse DFS and OS, regardless of the tumor population, two-year DFS, three-year DFS, or tumor location. Surgical resection or chemotherapy increased the rate of ctDNA+/–. The study was registered on PROSPERO (CRD42024575151).

Conclusion

Post-operative ctDNA positivity is related to a poor outcome. Surgical excision or chemotherapy influences the ctDNA shift status. Circulating tumor DNA can serve as a promising prognostic indicator for patients with non-metastatic colorectal cancer. However, more studies are needed to assess the potential of ctDNA status with survival outcomes in Non-metastatic colorectal cancer.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-025-04171-x.

Introduction

Colorectal cancer (CRC) is regarded as the third most commonly recognized malignancy globally and the second major cause of cancer-related death. Projections estimate that until 2040, about 3.2 million new cases of CRC and 1.6 million individuals worldwide will die from the disease [1]. Most early-stage CRCs (tumor-node-metastasis [TNM] stages I–III) can be cured by surgery; however, about 20% of CRC patients experience recurrence in the post-operative period [2–4]. Whether adjuvant chemotherapy can be performed to decrease the likelihood of recurrence primarily depends on surgery, clinicopathology, and patient tolerance [5, 6]. To predict tumor prognosis, current assessments include the instability of micros (MSI) state, TNM stage, the BRAF, KRAS, and NRAS mutation status, lymph node infiltration, tumor burden, and degree of tumor differentiation [7]. Although prognostic indicators are useful, they do not fully capture the heterogeneity of tumor biology. Additionally, abuse of chemotherapy causes intolerable medication toxicity, whereas insufficient treatment causes recurrence [5, 8]. These are important obstacles, and more precise biomarkers are needed for detecting early recurrence and survival.

The use of liquid biopsy for malignancies has increased because it is safe and has a degree of invasiveness, especially since the discovery of circulating tumor DNA (ctDNA). ctDNA is produced when cells from tumors are discharged into the circulatory system following necrosis, apoptosis, and secretion [9, 10]. Cancers that have spread to neighboring organs and tissues via the lymphatic system or arterial system, release ctDNA [11]. ctDNA can be adopted for assessing the dynamic characteristics of CRC [12, 13], evaluate the response to therapy, stratify options for therapy [14–16], and predict survival [17, 18]. The prognostic predictive ability of ctDNA in CRC has attracted considerable attention. ctDNA can serve as a dependable prognostic marker associated with adverse outcomes [19]. Positive detection of ctDNA indicates short DFS and overall survival (OS) in individuals with CRC treated via surgery, radiotherapy, chemotherapy, or targeted therapy [20–22]. However, some studies have shown no survival disparity between ctDNA+ and ctDNA− CRC patients [23].

Therefore, we performed a meta-analysis to consolidate the available evidence on the prognostic significance of ctDNA among those with non-metastatic CRC. By focusing on this subset of patients, we aimed to elucidate the role of ctDNA in predicting DFS and OS, which are pivotal for clinical decision-making and can guide more personalized treatment approaches. Furthermore, the findings that surgical interventions and chemotherapy influence ctDNA dynamics can provide insights into the interactions between treatment modalities and tumor genetic profiles, ultimately improving the management of CRC.

ctDNA detection refers to detecting the abnormal changes in small fragments of DNA released by tumor cells into the blood. This can assist in the early screening of cancer, monitoring of treatment efficacy, early warning of recurrence, and guiding individualized treatment. Defined key terms: ctDNA status: presence/absence of detectable tumor-derived DNA fragments in blood; high ctDNA level: concentration above study-specific thresholds, it has study-specific, ranging from 0.1 to 1.0% variant allele frequency for mutation-based assays; Across the 11 selected trials, ctDNA status was defined as Positive: Detection of ≥ 1 tumor-informed mutation (8 studies), or tumor methylation signature above threshold (2 studies), or combined mutation and fragmentome analysis (1 study).

Materials and methods

A systemic assessment was performed based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) standards [24]. The study was registered on PROSPERO (CRD42024575151). This study is a systematic review and meta-analysis of previously published data; therefore, no ethical approval or patient consent was required.

Search strategy

The PubMed, Embase, Cochrane Library, and Web of Science databases were searched for studies published before June 4, 2024. In addition to database searches, we manually screened reference lists from relevant review articles and eligible studies to identify additional publications. We also searched clinical trial registries (ClinicalTrials.gov, WHO ICTRP) for ongoing or unpublished studies. Conference proceedings from major oncology meetings (ASCO, ESMO, AACR) were reviewed to include potentially eligible abstracts. Furthermore, we consulted experts in the field to ensure no key studies were missed. All records identified through these “other sources” were subjected to the same eligibility criteria as database-derived records. The search was conducted using the following terms: ctDNA, liquid biopsy, CRC, survival, and prognosis. The section on Supplementary Material-1 describes the process. All abstracts identified during the search were evaluated, and relevant complete texts were examined. Two researchers individually managed the choices of studies, collection of data, and evaluation of quality. Discrepancies were settled through discussions with another researcher.

Outcomes

In this systematic review, we evaluated how ctDNA status at various times (baseline or before surgery, during therapy, following treatment, or following surgery) and its changes are correlated with DFS and OS. Subgroup analyses by geographic region (Asian vs. non-Asian) were performed based on documented biological, technical, and clinical practice differences that could modulate ctDNA’s prognostic utility.

Study selection

The criteria for including studies were established using the PICO framework, which consists of population, intervention or exposure, comparator, and outcome [25].

The results were included based on the following specific qualifying standards:

• (P) The population investigated consisted of individuals who were 18 years old or older with resectable CRC who were undergoing surgical treatment.

• (I) Participants provided liquid biopsy samples before and after surgery.

• (C) Two studies were conducted: (1) time points for fluid biopsy status (before and after surgery) as positive versus negative and (2) the transformation occurrence after surgery or chemotherapy in everyone as the transformation to positive versus shift to negative.

• ctDNA -/+ indicates that a patient transitioned from a negative pre-operative liquid biopsy to a positive postoperative biopsy.

• ctDNA +/- indicates a transition from a positive pre-operative liquid biopsy to a negative postoperative biopsy.

• (O) Outcomes assessed included OS, DFS, mortality, and recurrence rates.

• The exclusion criteria were as follows: (1) Patients lacking a histological confirmation of CRC. (2) Studies missing complete ctDNA data or outcome information. (3) Duplicates, reviews, editorials, ongoing research, or case studies.

Data extraction

All studies were cataloged in an Excel spreadsheet, and duplicates were removed. Ruiyang Li and Wenjian Liu independently reviewed titles, abstracts, and full texts based on the set criteria. Disagreements were resolved by consensus between Shuchao Wang and Baojun Liu. Additional references were manually selected from the bibliographies of the included studies. Key data, such as the name of the first author, date and nation of publication, participant count, TNM staging, study design, sample type, technique for detecting liquid biopsies, number of centers, follow-up time, and results, were independently extracted by the reviewers. The results are reported as a hazard ratio (HR), risk ratio (RR), or odds ratio (OR) with 95% confidence intervals. Only studies on resectable CRC patients who underwent surgical intervention were considered for meta-analysis. Studies that could not be stratified were excluded from the meta-analysis. The data were collated from the tables, charts, and text of the articles, including supplementary materials, when necessary.

Risk of bias

The quality and biases of the included studies were independently evaluated by Ruiyang Li and Wenjian Liu by adopting the Quality in Prognosis Studies (QUIPS) tool [26]. Any discrepancies were resolved by Shuchao Wang and Baojun Liu. Publication bias for any quantitative evaluation was examined with a funnel plot.

Data analysis

The meaning of I² is to quantify the degree of heterogeneity among studies, that is, to assess the extent to which the results of each included study are inconsistent. It represents the percentage of heterogeneity variation among studies in relation to the total variation. Low heterogeneity: I² < 50%; High heterogeneity: I² >50%.

We pre-specified four heterogeneity sources for investigation: Technical: Stratified analyses by ctDNA detection method (ddPCR [n = 1] vs. NGS [n = 6] vs. PCR [n = 3] vs. dPCR [n = 1]). Temporal: Subgrouped sampling phases (pre-op, post-op ≤ 4 weeks, post-op > 4 weeks, post preoperative neoadjuvant chemoradiotherapy, postoperative adjuvant chemotherapy). Tumor location: colon cancer, rectal cancer, colorectal cancer. Strongest Signal: Postoperative ctDNA detection within 4 weeks showed the highest hazard ratio (HR = 14.34, 95%CI:5.02–40.97), indicating early post-surgical ctDNA positivity is a critical marker for recurrence risk. Most Robust Results: Late post-op (> 4 weeks) ctDNA (HR = 7.88, I² = 28%) and post-neoadjuvant ctDNA (HR = 6.91, I²=0%) demonstrated low heterogeneity, supporting high reliability. Notable Uncertainty: Preoperative ctDNA showed a non-significant trend (HR = 2.03, p = 0.16) with high heterogeneity (I²=77%), suggesting limited current evidence for pre-surgical use (Supplementary material, Table S1).

The RevMan 5.3 software was employed to carry out the meta-analysis. Heterogeneity was evaluated using I² statistics, with I² > 50% or P < 0.1 indicating substantial heterogeneity. Based on the heterogeneity among the studies, either a fixed-effect or a random-effect model was chosen, with the random-effect model being applied when significant heterogeneity was present. The subgroup assessment was performed according to two-year DFS or three-year DFS, ctDNA sampling time, tumor location, and Asian or non-Asian population. Using a random effects model, pooled HRs were computed to evaluate heterogeneity across studies. All P values presented are two-tailed, and the results were considered to be statistically significant at P < 0.05. The outcomes were presented as forest plots, and funnel plots were used to assess publication bias.

Results

Search results

Following the removal of duplicates, 605 articles were initially selected using the established search method. Finally, 11 studies involving 1945 patients were included. The research process selected according to the PRISMA criteria is shown in Fig. 1.

Fig. 1.

Flow diagram of the cochrane systematic review

Relevant studies and patient features

Among the included studies, 11 were prospective studies [27–36] and one was a retrospective cohort study [37]. The studies were performed in Norway [27], Australia [28, 30], China [29, 33], Korea [37], Denmark [31, 36], the UK [32], Netherlands [34], and Spain [35]. In all studies, blood samples were collected pre-operatively and postoperatively, before/after chemotherapy, and, in some cases, longitudinally after surgery or adjuvant chemotherapy. All studies were published from 2018 to 2024, showing the interest in this biomarker. The general features of the included articles are presented in Table 1. Figure 2 displays the findings of the risk of bias evaluation (Supplementary material, Figure S1).

Table 1.

General characteristics of the included articles

Study	Year	Country	Stage TNM	Tumor location	Study design	No. of patients	Detection methods	Sample type	Endpoint	Median patient age(years)	Median follow-up(months)	No.of centres	Sampling times
Kristin B	2023	Norway	I–III	Colon cancer	Prospective	50	ddPCR	Plasma	DFS	65(18–85)	52.8(12–67.2)	Single Centre	Preoperative, within 1 month post surgery, 3 months; 6 months and every 6 months thereafter
Jeanne Tie	2019	Australia	III	Colon cancer	Prospective	96	PCR	Plasma	DFS	64 (26-82)	28.9(11.6-46.4)	Multicenter	4 to 10weeks after surgery completion of treatment (within 6 weeks of the final cycle of chemotherapy)
Yaqi Wang	2021	China	II–III	Rectal cancer	Prospective	119	NGS	Plasma	DFS	57	21.5(1.2 – 30.8)	Multicenter	Before nCRT, at the 15th and the 25th fractions of nCRT, 0 to 1 days before surgery, 5 to 12 days after surgery
Jeanne Tie	2018	Australia	II–III	Rectal cancer	Prospective	159	PCR	Plasma	DFS	62(28–86)	24(1–55)	Multicenter	Pretreatment, post chemoradiotherapy, 4–10 weeks after surgery
Seung-Bum Ryoo	2023	Korea	II–III	Colorectal cancer	Retrospective	98	PCR	Plasma	DFS	67(31–87)	36.3(16.7–52.1)	Multicenter	Postoperative 3-week post-operative 1-year
Tenna Vesterman Henriksen	2022	Denmark	III	Colorectal cancer	Prospective	160	NGS	Plasma	DFS、OS	63 (27–80)	35(13–36 )	Multicenter	Diagnosis, postoperative, during adjuvant therapy, routine follow-up
Susanna Slater	2024	UK	I–III	Colorectal cancer	Prospective	214	NGS	Plasma	DFS	67 (30–88)	30.3(3.0–56.4)	Multicenter	Pre- and post-operatively, every 3 months for the year 1, every 6 months for years 2 and 3, annually for years 4 and 5 until discharge or recurrence
Jian Yang	2023	China	I–III	Colorectal cancer	Prospective	75	NGS	Plasma	DFS	67 (57–71)	39.1(33.8−41.3)	Multicenter	Pre-operative, post-operative day 7
Lisa S.M. Hofste	2023	Netherlands	I–III	Rectal cancer	Prospective	51	NGS	Plasma	DFS	66 (48–84)	13 (1–50)	Multicenter	Before treatment, 6-8 weeks after neoadjuvant therapy, 1-2 weeks after surgery
Joana Vidal	2021	Spain	II–III	Rectal cancer	Prospective	72	NGS	Plasma	DFS、OS	60 (33–75)	38(2.3–51.5)	Multicenter	At baseline; after TNT within 48 h before surgery
Tenna Vesterman Henriksen	2023	Denmark	II–III	Colorectal cancer	Prospective	851	dPCR	Plasma	DFS、OS	71 (64–77)	26(12–36 )	Multicenter	Before operation, after operation, every 3–4 months for up to 36 months after operation

Fig. 2.

Individual assessment of the risk of bias for each article was performed with the Cochrane risk bias assessment tool

Effect of surgery or chemotherapy on the dynamics of ctdna liquid biopsy

Three trials with 199 participants assessed the ctDNA transition from positive to negative (ctDNA+/−) or negative to positive (ctDNA−/+) before and after surgery. The funnel plot revealed no publication bias (Supplementary material, Figure S2A). The comparison revealed a greater proportion of patients with ctDNA+/− (RR = 0.14, 95% CI: 0.04–0.44, p = 0.0008; Fig. 3A).

Fig. 3.

A A forest plot of publications analyzed for ctDNA shifts following surgery in patients with non-metastatic CRC. B A forest plot of DFS comparing positive and negative ctDNA statuses before surgery in individuals with non-metastatic CRC. C A forest plot of DFS comparing positive and negative ctDNA statuses after surgery in individuals with non-metastatic CRC. D A forest plot of DFS comparing positive and negative ctDNA statuses after surgery in patients with locally advanced CRC

Three trials with 998 patients analyzed the ctDNA transition from positive to negative (ctDNA+/−) or negative to positive (ctDNA−/+) before and after chemotherapy. The funnel plot revealed no publication bias (Supplementary material, Figure S2B). This contrast revealed a greater proportion of individuals with ctDNA+/− (RR = 0.14, 95% CI: 0.08–0.24, p < 0.00001; Supplementary material, Figure S2C).

Survival analysis

Analysis of survival based on ctdna status before surgery

Two studies were considered for OS analysis before surgery. The funnel plot revealed no publication bias (Supplementary material, Figure S3B). The patients who were ctDNA + before surgery had poorer OS than those who were ctDNA− (HR = 3.49, 95% CI: 0.11–107.28, p = 0.47; Supplementary material, Figure S3A), although this difference was not significant.

Five studies investigated DFS before surgery based on ctDNA fluid biopsy status. There existed no publication bias (see Supplementary Material, Figure S3C). Patients who were ctDNA + before surgery demonstrated shorter DFS than those with ctDNA− (HR = 2.03, 95% CI 0.76–5.39, p = 0.16; Fig. 3B); nevertheless, the difference in DFS between the groups was not significant.

Analysis of survival based on the ctdna status after surgery

Two studies assessed OS based on the postoperative ctDNA status. The funnel plot (Supplementary information, Figure S4B) revealed no publication bias. The postoperative prognosis of ctDNA + patients was poorer than that of ctDNA− patients (HR = 2.63, 95% CI: 1.65–4.17, p < 0.0001; Supplementary material, Figure S4A).

Ten studies assessed DFS based on the postoperative ctDNA status. There was no publication bias (Supplementary Material, Figure S4C). Patients who were ctDNA + following surgical treatment had a shorter DFS than those with ctDNA− (HR = 9.84, 95% CI: 6.53–14.82, p < 0.00001; Fig. 3C).

There was no publication bias in six studies analyzing DFS in locally advanced CRC patients based on the postoperative ctDNA status (Supplementary material, Figure S4E). Patients who were ctDNA + following surgery exhibited shorter DFS than those with ctDNA− (HR = 10.03, 95% CI: 7.03–14.33; p < 0.00001; Fig. 3D).

Two studies examined DFS using longitudinal ctDNA data monitoring in CRC patients. No publication bias was recorded (Supplementary Material, Figure S4F). Serial ctDNA detection was related to an increased risk of recurrence (HR = 31.58, 95% CI: 21.22–46.99, p < 0.00001; Supplementary material, Figure S4D). Among them mean value and 95%CI of DFS and OS in each of the main group (Supplementary material, Table S2).

Survival analysis based on the ctdna status following neoadjuvant chemoradiotherapy

Four studies investigated DFS according to the ctDNA status after neoadjuvant chemoradiotherapy. No publication bias was recorded, in accordance with the funnel plot (Supplementary material, Figure S5). Patients who were ctDNA + following neoadjuvant chemoradiotherapy had lower DFS than those who were ctDNA− (HR = 6.91, 95% CI: 3.98–12, p < 0.00001; Fig. 4A).

Fig. 4.

A A forest plot of DFS comparing positive and negative ctDNA statuses after neoadjuvant chemoradiotherapy in patients with non-metastatic CRC. B A forest plot of publications analyzing the comparison of DFS between positive versus negative ctDNA status after postoperative adjuvant chemotherapy in patients with non-metastatic CRC. C A forest plot of recurrence rates comparing positive versus negative ctDNA status after surgery in patients with non-metastatic CRC. D A forest plot of recurrence rates comparing positive versus negative ctDNA status after postoperative adjuvant chemotherapy in patients with non-metastatic CRC

Survival analysis based on the ctdna status following postoperative adjuvant chemotherapy

Three studies assessed the relationship between DFS and the ctDNA status after postoperative adjuvant chemotherapy. No publication bias was recorded (Supplementary Material, Figure S6). Patients who were ctDNA + following adjuvant chemotherapy had lower DFS than those who were ctDNA− (HR = 17.08, 95% CI: 5.21–55.96, p < 0.00001; Fig. 4B).

Recurrence risk analysis

Recurrence risk assessment based on the ctdna status following surgery

The recurrence rate was assessed in 903 participants from eight studies. No publication bias was found (Supplementary material, Figure S7A). Patients with ctDNA− following surgery had a lower rate of recurrence than those with ctDNA+ (RR = 5.5, 95% CI: 4.12–7.33, p < 0.00001; Fig. 4C).

Four studies analyzed the relationship between recurrence risk and postoperative ctDNA status in locally advanced CRC patients. No publication bias was observed (Supplementary material, Figure S7B). Patients who were ctDNA + after surgery exhibited a greater risk of recurrence than those who were ctDNA− (HR = 5.75, 95% CI: 3.34–9.89, p < 0.00001; Supplementary material, Figure S7C).

Recurrence risk analysis based on the ctdna status after postoperative adjuvant chemotherapy

The recurrence rate was assessed in 568 participants from four investigations. No publication bias was observed (Supplementary material, Figure S8). Patients who were ctDNA + after postoperative adjuvant chemotherapy had a greater recurrence risk than those who were ctDNA− (RR = 4.77, 95% CI: 2.56–8.86, p < 0.00001; Fig. 4D).

Subgroup analysis of postoperative DFS

Subgroup analysis according to tumor population

In the 10 trials that analyzed DFS based on the postoperative ctDNA status, patients with a ctDNA + status after surgery had lower DFS than those with ctDNA− status in the Asian population (HR: 10.64, 95% CI 4.61–24.59, p < 0.00001) and the non-Asian population (HR: 10.93, 95% CI: 6.66–17.94, p < 0.00001) (Fig. 5A). No publication bias was recorded (Supplementary material, Figure S9A).

Fig. 5.

A A forest plot of publications exploring the comparison of postoperative DFS between patients with positive versus negative ctDNA status according to the tumor population in non-metastatic CRC patients (Asian population or non-Asian population). B A forest plot of publications analyzing the comparison of two-year DFS and three-year DFS between patients with positive versus negative ctDNA status in non-metastatic CRC

Subgroup analysis according to two-year or three-year DFS

The two-year DFS and three-year DFS rates were assessed in 686 participants from five trials. Those who were ctDNA + following surgery had a shorter two-year DFS (HR: 5.09, 95% CI: 2.13–12.18, p = 0.0003) and three-year DFS (HR: 7.35, 95% CI: 4.47–12.10, p < 0.00001) than patients with ctDNA− (Fig. 5B). No publication bias was recorded (Supplementary material, Figure S9B). As shown in Fig. 5B, patients with ctDNA + versus ctDNA− had a poorer three-year DFS than two-year DFS.

Subgroup analysis according to tumor location

Based on tumor location and ctDNA status, DFS was analyzed in 10 articles. The funnel plot revealed no publication bias (Supplementary material, Figure S9C). The results revealed a decrease in DFS among colon cancer patients (HR = 25.46, 95% CI 1.27–511, p = 0.03), rectal cancer patients (HR = 10.61, 95% CI 5.21–21.62, p < 0.00001), and CRC patients (HR = 11.35, 95% CI: 6.31–20.42, p < 0.00001) who were ctDNA + following surgery compared to patients who were ctDNA− (Supplementary material, Figure S9D).

Discussion

Main findings and comparison with prior studies

This systematic review and meta-analysis revealed that non-metastatic CRC patients with ctDNA-positive status after surgery showed significantly lower DFS (HR = 9.84, 95% CI: 6.53–14.82, p < 0.00001) and OS (HR = 2.63, 95% CI: 1.65–4.17, p < 0.0001). An evaluation of the subgroups revealed that irrespective of the Asian population or non-Asian population, two-year DFS or three-year DFS, tumor location, that increased levels of ctDNA indicated a worse postoperative outcome. The observed regional differences may reflect variations in tumor biology, assay methodologies, or treatment approaches, though our study was not powered to definitively establish these relationships. These findings suggested that individuals who are positive for ctDNA after surgery have a poorer prognosis than ctDNA-negative individuals. According to these findings, ctDNA assessment can promising indicate individuals at increased risk of recurrent illnesses and those who can benefit from adjuvant systemic treatments. Moreover, this can prevent patients from receiving unnecessary or harmful treatments. Thus, ctDNA analysis could be applied as a prediction marker. Postoperative ctDNA detection is associated with a significant probability of recurrence [38–41]. ctDNA results are used to may guide patient treatment and examine the clinical value of ctDNA-guided techniques for CRC therapy and surveillance after surgery, though its monitoring utility requires further standardization.

In our analysis, pre-operative ctDNA positivity showed a non-significant trend toward worse DFS (HR = 2.03, 95% CI: 0.76–5.39, p = 0.16) and OS (HR = 3.49, 95% CI: 0.11–107.28, p = 0.47). The wide confidence intervals, particularly for OS, reflect substantial uncertainty in these estimates, likely due to limited sample size, heterogeneity in sample timing (e.g., proximity to surgery), and variability in ctDNA detection methods. While these exploratory findings suggest a potential association between pre-operative ctDNA and adverse outcomes, the lack of statistical significance precludes definitive conclusions. Future studies with standardized protocols and larger cohorts are needed to clarify whether pre-operative ctDNA has utility in guiding neoadjuvant therapy decisions. Similarly, in a study on surgically treatable and unresectable stage IV CRC, Parikh et al. reported no significant connection between pre-operative ctDNA and OS [18]. Our findings also revealed that after neoadjuvant chemoradiotherapy, patients who were ctDNA + had a lower DFS following surgery than patients who were ctDNA+. (HR = 6.91, 95% CI 3.98–12, p < 0.00001).

Notably, ctDNA levels may fluctuate following treatment [42]. Therefore, we also assessed whether ctDNA levels change after surgery or chemotherapy in CRC patients and found through a meta-analysis that post-treatment ctDNA levels are reduced or cleared in CRC patients. These findings suggested that ctDNA may be serve as a dynamic marker to monitor treatment response. This is particularly relevant for tailoring adjuvant therapy, where the presence of ctDNA after primary treatment can be used to identify patients who may benefit from more aggressive or extended treatment protocols. This finding was demonstrated across multiple studies in non-CRC patients [43, 44]. However, we did not find improvement in survival among patients who exhibited a drop or clearance in ctDNA levels before or after treatment. One such factor is the limited number of studies that provide precise information on fluid biopsy outcome changes in each individual. Additional factors, including advanced age, surgical difficulties, and adjuvant chemotherapy, may also affect DFS, whereas additional help therapy following recurrence may affect OS.

Patients who were positive for ctDNA following treatment with adjuvant chemotherapy had a poorer DFS than those who were negative for ctDNA (HR = 17.08, 95% CI: 5.21–55.96; p < 0.00001). Post-adjuvant chemotherapy ctDNA status can be used to estimate the possibility of recurrence and the necessity for more closely monitored observation [45, 46]. Additionally, serial ctDNA detection is a prognostic indicator of recurrence (HR = 31.58, 95% CI: 21.22–46.99, p < 0.00001). Although tumor-informed assays require a longer time frame for reversal, they can yield sensitive outcomes and effective longitudinal surveillance. In a study conducted with 138 patients with advanced digestive tract cancer, Parikh et al. reported that serial ctDNA tracking can predict how patients respond to systemic treatment [41, 47].

Biological plausibility

Biomarkers, more specifically, ctDNA, constitute a new technique used in the diagnosis, prognosis, and therapy of cancer [48–51]. Recently, research on ctDNA and CRC has increased. Faulkner et al. reported in their meta-analysis that positive ctDNA detection after surgery has a significant detrimental effect on DFS in patients with CRC [52]. This association was also reported by Chidharla et al. [53]. In the quantitative study, both meta-analyses included patients at various tumor stages, such as metastatic patients, as well as patients receiving chemotherapy and surgical therapy for curative and palliative purposes.

There was a lack of agreement among the studies concerning the optimal timing for ctDNA sampling. Three studies assessed ctDNA levels after surgery and adjuvant chemotherapy, revealing. They found that ctDNA measurements taken after chemotherapy predicted outcomes more accurately [28, 38, 54]. Delaying postponement of the start of adjuvant chemotherapy for more than eight weeks can lead to poorer long-term results [55], highlighting the importance of timing when incorporating postsurgical ctDNA into treatment plans. ctDNA analysis should be conducted once it has been cleared from circulation and after the primary tumor has been removed. Chen et al. studied ctDNA clearance following lung cancer surgery by taking serial measurements immediately after the operation. They reported that ctDNA levels continue to decrease until the third day postoperative, and ctDNA detected beyond this period is more accurately correlated with prognosis [56].

Limitations

While our meta-analysis provides important insights, several limitations warrant consideration: Methodological Limitations: The relatively small number of included studies (n = 11) limits the statistical power for some subgroup analyses and increases susceptibility to type II errors. Significant inter-study heterogeneity was observed, reflecting variations in study designs, patient populations, and follow-up protocols. Potential Biases: Publication bias is plausible, as small negative studies might be underrepresented. Language bias may exist as we included only English-language publications, potentially excluding relevant studies from non-English speaking regions. Most included studies were from academic centers, possibly limiting generalizability to community practice settings. Clinical Confounders: Important prognostic factors like BMI, comorbidities (particularly cardiovascular diseases), and detailed treatment adherence data were rarely reported, preventing adjustment for these potential confounders. Age distribution varied significantly across studies, yet age-adjusted analyses were unavailable in most cases. Tumor molecular characteristics (e.g., MSI status, RAS mutations) that may influence ctDNA shedding were not consistently reported. Temporal Considerations: The seven years-long inclusion period (2018–2024) encompasses evolving ctDNA technologies and changing treatment standards. Variability in ctDNA sampling timepoints (pre-treatment, post-op, during surveillance) complicates cross-study comparisons.Technical Variability: First, Substantial differences in ctDNA detection methods across studies (including tumor-informed vs. tumor-agnostic approaches, varying sensitivity thresholds from 0.01% to 0.1% VAF) may affect result comparability. Pre-analytical factors (blood collection tubes, processing delays) were inconsistently reported, potentially introducing technical noise. The evidence for the absolute specificity of many methylation markers is still under investigation. Potential co-methylation in non-malignant conditions (e.g., clonal hematopoiesis, benign colorectal polyps, or inflammation) could theoretically lead to false-positive results. Secondly, the included studies utilized a variety of detection platforms (e.g., PCR-based, NGS-based) and, most importantly, lacked a standardized, universally accepted threshold for defining ctDNA positivity. This heterogeneity in methodological approaches is a significant source of the statistical heterogeneity observed in our meta-analysis (as indicated by the high I² values) and complicates the direct comparison and pooling of results. It underscores that our pooled estimate should be interpreted as an average effect across different technologies and definitions. Therefore, our results must be interpreted with caution. Future efforts towards standardizing assay techniques, validating marker specificity, and establishing clinically validated cut-off values are paramount for the reliable translation of ctDNA analysis into routine clinical practice.

Future directions

Future studies should standardize ctDNA detection methods and define optimal time points for ctDNA sampling in the clinical workflow of CRC management. Additionally, assessing the genetic landscape of ctDNA can reveal specific mutations that are prognostically relevant and can be targeted therapeutically.

To summarize, this meta-analysis supports the prognostic value of ctDNA in non-metastatic CRC, particularly in the postoperative context. With the advancement in the field of liquid biopsy, integrating ctDNA analysis into clinical practice can revolutionize personalized treatment approaches, which in turn can improve the management and outcomes for CRC patients. Further studies are needed to refine the utility of ctDNA and expand its application across different stages of CRC.

Conclusions

To conclude, in the current meta-analysis, we evaluated the effect of ctDNA status on non-metastatic CRC. To summarize, the detection of ctDNA following surgery for CRC indicates a poor prognosis. We found that postoperative ctDNA detection is associated with significantly decreased DFS and OS, highlighting the potential of ctDNA as a valuable biomarker for monitoring disease progression and patient outcomes. Conversely, presurgical ctDNA levels indicated poorer outcomes, although the results were not statistically significant, suggesting that the prognostic utility of pre-operative ctDNA might be less pronounced than that of postoperative ctDNA levels. The analysis also revealed that surgical resection and chemotherapy influence ctDNA dynamics, potentially leading to a shift in ctDNA status, which can affect patient monitoring and treatment strategies. Despite promising insights, larger, well-designed studies are needed to further investigate how shifts in ctDNA status correlate with changes in DFS and OS. Such findings can help refine the use of ctDNA in clinical practice, improving patient management and treatment strategies. Additionally, the decrease in DFS observed in patients with persistent ctDNA positivity following neoadjuvant chemoradiotherapy highlights the need for continuous monitoring and intensified follow-up of these high-risk patients.

Despite this extensive body of research, there is still no consensus on various logistical factors, such as ctDNA sampling time and ctDNA test methods (dPCR, ddPCR, and next-generation sequencing (NGS)). Thus, further studies are required to refine our understanding of ctDNA dynamics and its integration into clinical practice. Larger, well-designed studies are needed to determine the full potential of ctDNA in guiding therapeutic decisions and improving patient management in non-metastatic CRC.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Ruiyang Li: Conceptualization, Methodology, Formal analysis, Investigation, Resources, Writing – original draft. Shuchao Wang: Project administration, Investigation. Jie Xu: Conceptualization. Heng Zhang: Methodology. Yibo Yin: Supervision. Min Zhao: Validation. Xue Han: Visualization. Baojun Liu: Writing – review & editing, Funding acquisition. Wenjian Liu: Writing – review & editing.

Funding

This work was supported by Shandong Province Traditional Chinese Medicine Science and Technology Project (grant numbers: M20243703); Medical and Health Science and Technology Development Project of Shandong Province (grant numbers: 2019WS404); Shandong Province Traditional Chinese Medicine Science and Technology Development Project (grant numbers: 2019–0358); Tai ‘an city science and technology innovation development project (grant numbers: 2022NS218).

Data availability

All data supporting the findings of this study are available within the article and Supplementary Files.

Declarations

Ethics approval and consent to participate

This study is a systematic review and meta-analysis of aggregated data from previously published studies; therefore, ethical approval and patient consent were not required.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.