According to the Global Cancer Observatory 2022, esophageal cancer (EC) is the 11th most common cancer and the 7th leading cause of cancer deaths globally, resulting in about 510,000 new cases and about 450,000 deaths (1).EC mainly consists of esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC). EC patients are asymptomatic or atypical in the early stages. Because of their very aggressive biology it often leads to a poor prognosis due to significant progression of the disease by the time of diagnosis. In most countries and regions, the 5-year survival rate after EC diagnosis is between 10% and 30% (2). Currently, the treatment of EC is based on a staged combination model, in which patients with locally advanced disease are usually treated radically with a multimodal approach including surgery, e.g., neoadjuvant therapy in combination with surgery, and palliatively with advanced (metastatic or disseminated) and recurrent disease (3). Although the treatment modalities of EC have improved significantly, the clinical management of EC is still unable to accurately and timely capture the recurrence risk of patients under the current clinical technology, resulting in no significant improvement in their prognosis. There is an urgent need to develop a new biomarker to predict the risk of EC patients at different stages of treatment and to implement individualized treatment so that timely intervention can be made to improve overall survival outcomes.

In recent years, more and more researchers have begun to focus on applying liquid biopsy in oncology, among which circulating tumor DNA (ctDNA) has attracted much attention (4). ctDNA is released into the bloodstream during apoptosis or necrosis of tumors and through circulating tumor cells (5–8). Compared with clinically available serum tumor biomarkers (e.g., CEA, SCC), ctDNA has higher sensitivity and specificity (8, 9). ctDNA has a very short half-life of approximately 2 hours, which allows it to reflect the real-time tumor load of a tumor patient more accurately and to be used as a dynamic biomarker for tracking the presence of tumors (10, 11). Some studies have reported the application of ctDNA in the early diagnosis of tumors, identification of tiny residual lesions, response to treatment, and survival prognosis, respectively (12–15). Most studies focused on the relationship between ctDNA and survival prognosis in EC. There are three systematic reviews and meta-analyses on the relationship between ctDNA and EC prognosis (16–18), the most comprehensive of which is from Zhang and Jin et al. (18). They conducted a complete subgroup analysis of the prognosis of ctDNA in patients with EC in terms of different ethnicities, histological subtypes, anatomical locations, ctDNA testing methods, testing times, and different treatment modalities. Although previous studies have confirmed the prognostic value of ctDNA in EC, a core issue that has been overlooked is that the vast majority of research has pooled ctDNA data collected at different time points for analysis, failing to adequately account for the inherent dynamic evolution of ctDNA levels over time and in response to treatment. This ‘static’ analytical approach fails to capture the full value of ctDNA as a biomarker that reflects tumor burden and treatment response in real time. This study aims to systematically evaluate the prognostic value of ctDNA at three critical clinical time points (baseline, after neoadjuvant therapy, and follow-up) using a “dynamic” framework. It further compares the differences in clinical recurrence detection timing between ctDNA and traditional imaging techniques, seeking to provide precise evidence for the application of ctDNA at specific clinical decision points.

The present systematic review and meta-analysis were conducted by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) ( Supplementary Table S1 ) and the Assessing Methodological Quality in Systematic Reviews (AMSTAR) guidelines (19, 20). The study protocol was registered on the International Prospective Register of Systematic Reviews (PROSPERO) (ID: CRD42024612909).

This study differs from previous systematic reviews and meta-analyses in that it systematically elucidates the prognostic value of ctDNA in EC from a dynamic temporal perspective. By employing univariate and multivariate statistical methods combined with tumor-informed and non-informed analysis strategies, we provide a more comprehensive, time-specific basis for risk stratification and efficacy assessment.

Genomic alterations are key drivers of EC development and metastasis (46). Studies indicate that the most frequently mutated genes in ESCC include TP53, NFE2L2, and MLL2, while EAC predominantly involves TP53, CDKN2A, and ARID1A (47). Among these, TP53 exhibits the highest mutation rate in both subtypes (47–49) and correlates with poorer survival outcomes (50). Since Hsieh et al. (51) first reported the association between plasma cell-free DNA and EC recurrence and survival, numerous prospective and retrospective studies have focused on the prognostic value of ctDNA in EC. This systematic review and meta-analysis primarily assessed the clinical significance of ctDNA monitoring at three time points: baseline, after neoadjuvant therapy, and during follow-up.

Baseline ctDNA positivity is often associated with greater tumor burden and more advanced clinical staging (25–27, 33, 35, 43), suggesting its potential as a biomarker reflecting EC tumor burden and assessing prognosis. Although some included studies did not find a significant association with prognosis (25, 26, 32, 35, 40), others reported a clear correlation (27, 28, 33, 36, 43). The pooled results of this meta-analysis indicate that baseline ctDNA positivity is a significant risk factor for recurrence and mortality in EC patients, consistent with reports in other cancer types such as breast and lung cancer (15, 52, 53). Further analysis revealed that ctDNA positivity detected after neoadjuvant therapy and during follow-up was also significantly associated with higher risks of recurrence and mortality. Notably, the combined hazard ratio (HR) for recurrence or death following ctDNA positivity after neoadjuvant therapy was approximately 1–2 times that of baseline status, suggesting superior predictive efficacy. The HR for ctDNA positivity during follow-up further increased to 1–2 times that after neoadjuvant therapy, indicating that its prognostic value progressively escalates over time, peaking during follow-up. This precisely quantified pattern of risk evolution, systematically elucidated for the first time in the EC field through meta-analysis, highlights the significant advantage of dynamic monitoring over single-time-point detection. Notably, this pattern of risk change highly correlates with the dynamic clearance pattern of ctDNA. Multiple studies consistently demonstrate that a decrease or clearance of ctDNA levels during or after treatment is significantly associated with improved patient survival, with those maintaining persistent negativity or achieving clearance exhibiting better prognosis (25, 26, 32). Therefore, our discovery of this dynamic pattern represents a significant addition to previous review conclusions and provides a solid foundation for the clinical translation of ctDNA.

More importantly, viewed from the perspective of tumor ecology and evolution (54), these dynamic patterns reflect the adaptive evolution of tumor clonal populations under therapeutic pressure. Persistent ctDNA positivity after neoadjuvant therapy may indicate the selection and amplification of adaptive resistance subclones, while its clearance may reflect effective suppression of the dominant clone by treatment. This understanding extends beyond the static interpretation framework of traditional molecular biomarkers, situating ctDNA dynamics within the ecological system of tumor-microenvironment-therapy interactions. It provides a novel theoretical perspective for deepening our understanding of treatment response and resistance mechanisms.

Based on these time-specific risk quantification evidence and theoretical understanding, persistent ctDNA positivity after neoadjuvant therapy should be regarded as a strong indicator of tumor resistance to treatment and extremely high recurrence risk. Regardless of imaging findings, these patients should be considered for intensified adjuvant therapy (e.g., immunotherapy, extended chemotherapy, or targeted therapy) to eradicate residual disease. For ctDNA positivity emerging during follow-up, immediate initiation of more frequent imaging (e.g., PET-CT) is warranted, along with exploring preemptive treatment before clinical recurrence. This study provides evidence-based support for incorporating ctDNA-guided temporal stratification into the comprehensive management of EC patients, advancing from prognostic assessment to personalized intervention. However, the primary challenge to clinical translation remains the absence of standardized detection thresholds and unified reporting standards. Future research should focus on establishing clinically actionable criteria for ctDNA positivity at various time points. Prospective intervention trials—such as randomized controlled trials evaluating intensified treatment strategies for ctDNA-positive patients—are needed to validate the actual survival benefit. Ongoing clinical trials (e.g., NCT06914011, NCT05965479) provide an initial framework for ctDNA-guided therapy.

In colorectal and breast cancers, tumor-informed ctDNA testing demonstrated superior performance compared to non-tumor-informed assay, exhibiting higher sensitivity (52, 55, 56). This study found that the prognostic value of ctDNA derived from both testing methods was generally consistent across three time points: baseline, after neoadjuvant therapy, and follow-up. Notably, at baseline and after neoadjuvant therapy stages, the hazard ratios (HRs) for recurrence and mortality risk derived from the tumor-informed assay tended to be numerically higher than those from the non-tumor-informed assay, suggesting potential advantages during periods of high tumor burden due to greater sensitivity. However, given the limited research data and high uncertainty surrounding the non-tumor-informed assay, the actual effect of the tumor-informed assay may be overestimated. The conclusion presented here is only a preliminary hypothesis: a tumor-informed assay may be more sensitive in identifying high-risk recurrence and mortality in EC patients. Currently, systematic comparative studies in the EC field are lacking, and further research is needed to clarify the differences and clinical applicability of these two strategies.

During the tumor-free interval following treatment, when imaging fails to detect lesions, serial liquid biopsies can be used for early prediction of recurrence risk. This study found that in EC, the emergence of ctDNA positivity occurred on average 4.53 months earlier than imaging detection. This trend has been consistently reported across multiple tumor types, including breast cancer and colorectal cancer (52, 57, 58). Current clinical guidelines still regard imaging as the gold standard for recurrence diagnosis and treatment decisions. It is important to note that intervention based solely on ctDNA positivity carries risks, such as false positives due to clonal hematopoiesis or technical errors, which may lead to overtreatment. Therefore, after detecting ctDNA positivity, it is recommended to promptly combine imaging examinations for careful evaluation to ensure robust clinical decision-making and patient safety. Furthermore, the current early warning value of ctDNA remains based on limited studies and requires further validation across diverse populations and healthcare settings. Future large-scale, rigorously designed multicenter prospective studies are necessary to confirm these findings.

This study has several limitations. First, outcome measures were grouped into two broad categories—”recurrence-related (including PFS, DFS, RFS, and FFP)” and “mortality-related (including OS, DSS, and CSD)”—resulting in a certain lack of outcome consistency. Second, criteria for defining ctDNA positivity varied across studies, and the absence of standardized testing protocols reduced comparability of results. Third, the study did not perform stratified analyses by histological subtype (EAC vs. ESCC) or geographic region (Asia vs. Western countries). Given the significant molecular and clinical differences between these subtypes, along with regional heterogeneity in clinical practice and epidemiological backgrounds, this limitation restricts the generalizability of findings and the interpretability of subgroup analyses. Additionally, substantial heterogeneity existed in baseline OS analyses. Although random-effects models and sensitivity analyses were employed to ensure robustness, interpretation of results may still be affected. At the same time, the studies included primarily utilized univariate HRs, which may introduce residual confounding bias. This heterogeneity and bias primarily stem from clinical and methodological variations. Moreover, the included studies were evenly split between prospective and retrospective designs. Although quality assessments controlled for bias, further high-quality prospective studies are needed for validation. Finally, despite comprehensive search strategies, publication bias may still exist in some analyses. Given these limitations, future research should focus on: ① Establishing unified standards for ctDNA detection and reporting to enhance comparability of results; ② Conducting large-scale, multicenter, prospective studies targeting specific histological subtypes and regional populations, and integrating multidimensional data for individual patient data (IPD)-based meta-analyses to elucidate interactions among factors and more accurately assess the prognostic value of ctDNA; ③ Collectively enhancing the reliability and clinical applicability of relevant evidence through the aforementioned measures.

Throughout the entire treatment course, positive ctDNA detection consistently correlated with poor prognosis in EC patients. Notably, its prognostic value increased over time, continuously strengthening from baseline to follow-up stages. This study supports ctDNA as an effective dynamic prognostic biomarker and preliminarily explores its potential feasibility for integration into the treatment management workflow for EC patients.measures