Circulating Tumor DNA as a Biomarker in Pancreatic Cancer: Clinical Applications and Challenges

Abstract

Pancreatic cancer represents the most lethal malignancy among all gastrointestinal tumors. Recent medical advances, such as targeted therapy and immunotherapy, increasingly rely on tumor molecular profiling to tailor treatment of patients with advanced cancer. Given the risks associated with invasive biopsy, circulating cell-free DNA (cfDNA) has emerged as a cutting-edge approach for the detection and monitoring of cancer. The minimally invasive operation, coupled with the sensitive and timely detection of cancer across multiple clinical application, makes cfDNA a potential solution to transform precision oncology. Despite the advances, the expected widespread application of liquid-biopsy is still limited due to a series of substantial challenges in technique and clinical settings. In this article, we discuss technologies and methodologies in the detection of cfDNA. The opportunities to address substantial challenges, including achieving clinical meaningful detection sensitivities, enhancing of assay accessibility, evaluating the clinical sensitivity of circulating tumor DNA (ctDNA) burden in clinical settings, are highlighted to support the integration of liquid biopsies into cancer treatment. The most recent ctDNA-associated studies are summarized to provide a whole picture of the application of ctDNA in the adaptive management and surveillance of pancreatic cancers.

INTRODUCTION

Pancreatic cancer is currently the most lethal cancer, exhibiting the lowest 5-year survival rate among all gastrointestinal tumors. The survival rate is only 3% when pancreatic cancer is diagnosed at an advanced stage (1). Owing to the insidious and nonspecific symptoms of pancreatic cancer, more than 50% of patients with pancreatic cancer are diagnosed at an advanced stage. Surgical intervention remains the cornerstone of therapeutic management, whereas contemporary epidemiological data indicate that only 10%–20% of newly diagnosed patients meet the criteria for curative resection (2–4). Recent medical advancements in solid tumors, including pancreatic cancer, involve targeted therapy and immunotherapy applied in locally advanced or metastatic cases (5–11). These therapies require tumor molecular testing to identify the most suitable patients. Therefore, identifying novel biomarkers for pancreatic cancer holds immense potential to improve early detection, diagnosis, prognostication, and patient management (12,13). The application of molecular testing has allowed for targeted therapy in certain patients. Because the molecular testing has allowed for targeted therapy in certain patients, similar testing, particularly identification of novel biomarkers and less invasive methods of identifying these biomarkers, may ultimately lead to discovery in other aspects of cancer care. For instance, the progression in early detection may benefit from the novel methods of liquid biopsy.

Current clinical diagnostic approaches for pancreatic cancer encompass imaging modalities, serum biomarker assays represented by carbohydrate antigen (CA) 19-9, and histopathological examinations. However, these methods have demonstrated suboptimal efficacy in early-stage detection due to limitations in sensitivity and accuracy. Among these, CA19-9 remains the most widely used pancreatic cancer-associated tumor marker (14). Studies confirm its elevated levels in patients with pancreatic cancer, particularly in disease monitoring and prognostic evaluation (15–17). However, CA19-9 exhibits suboptimal specificity because it can also increase in benign pancreatic disorders, hepatobiliary diseases, and certain gastrointestinal malignancies, potentially leading to misdiagnosis (18). In addition, false-negative rates are notably higher in Lewis antigen-negative individuals (19). Therefore, reliable biomarkers for early diagnosis, prognosis evaluation, and treatment monitoring are still lacking.

In recent years, liquid biopsy has gained attention due to its lower invasiveness and ability to continuously monitor cancer progression (20,21). Liquid biopsy often refers to the analysis of tumor-derived materials, such as circulating tumor cells (CTCs), circulating tumor DNA (ctDNA), circulating tumor RNA (ctRNA), exosomes, and tumor-educated platelets (TEPs) from blood plasma (Figure 1). CTCs are released from tumor tissue; ctDNA is secreted from apoptotic or necrotic tumor cells; and exosomes are membrane-bound vesicles released from tumor cells. Compared with CTCs, ctRNA, exosomes, and TEP, ctDNA is extensively used as a biomarker used in clinical applications. The principal advantages of ctDNA are that it carries tumor-related genomic information, such as nucleosome footprint, gene variations, the methylation status, tumor mutation burden, and microsatellite instability, followed by its capacity to address intratumoral heterogeneity (22). Furthermore, the straightforward extraction protocols of ctDNA facilitate clinical implementation across the entire clinical workflow, including early diagnosis, therapeutic guidance, monitoring treatment, and prognostic stratification (23–25). Despite significant achievements, there is no systematic review comprehensively introducing recent developments in clinical applications of ctDNA in pancreatic cancer.

Figure 1.

The biology of liquid biopsy in patients with pancreatic cancer. Multiple types of biomolecules in peripheral blood, including ctDNA, cfRNA, exosomes, and CTCs, could serve as targets of liquid biopsies. The preanalytical and analytical methodologies in the testing of ctDNA are gaining maturity with the progression in technologies. The most classic detection assay of ctDNA focuses on the mutations in the most recent clinical settings and studies. Meanwhile, the potential to develop cfRNA, exosomes, and CTCs to serve as targets of liquid biopsy is still in preclinical explorations. The combination of the abovementioned novel biomarkers and the ctDNA assay might provide a novel solution to the challenges in the application of ctDNA. cfRNA, cell-free RNA; CTC, circulating tumor cell; ctDNA, circulating tumor DNA.

In this review, we describe the potential clinical application of ctDNA testing, from aiding cancer diagnosis to guiding treatment, detecting minimal residual disease (MRD), and dynamically monitoring treatment response (Figure 2).In addition, we highlight promising approaches, such as multiomics platforms integrated with artificial intelligence-driven analytical frameworks, which may be widely available in clinical applications in the future.

Figure 2.

Clinical use cases for ctDNA testing in pancreatic cancer. The application of ctDNA testing is promising in multiple clinical settings. In MRD detection, the use of ctDNA is relatively ready for prime time. Studies have been conducted to detect MRD by ctDNA analyses in patients with various types of solid tumors. Many clinical trials in pancreatic cancer are ongoing. Studies have documented that ctDNA not only plays a role in serving as biomarkers of therapeutic efficacy but can also provide guidance in recurrence detection in serial ctDNA monitoring compared with the standard-of-care surveillance. Emerging data demonstrated the early screening, prognosis evaluation, and identification of targets for therapy. Although various challenges regarding screening and early detection still remain to be addressed, the ctDNA testing is potentially useful in multiple clinical practices. ctDNA, circulating tumor DNA; MRD, minimal residual disease.

METHODOLOGIES OF ctDNA DETECTION

The preanalytical process of ctDNA detection

Circulating cell-free DNA (cfDNA), first reported by Mandel and Metais in 1948, refers to the length of DNA fragments that are less than 200 base pairs (26). Because cfDNA is derived from various cells and rapidly cleared from circulation, it typically presents in body fluid for a short duration and at limited levels. Thus, optimizing experimental techniques for cfDNA isolation and analysis is critical.

The preanalytical phase of liquid biopsy approaches is critical to isolate the analytes of interest properly. Whole blood contains ctDNA, cell-free RNA, CTCs, TEPs, tumor-derived proteins and metabolites, and tumor-derived extracellular vesicles (27–29). Whether the analytes of interest can be extracted in sufficient quantity and quality depends on the blood processing method. The preanalytical phase of plasma-derived ctDNA analysis contains blood collection, cell-free plasma processing, and cfDNA extraction.

Several factors can affect the quality of the ctDNA in plasma during blood collection. The application of butterfly needles in Phlebotomy significantly reduces hemolysis compared with conventional intravenous catheters (30). Cellular breakage contaminates the ctDNA with high molecular weight genomic DNA, thereby increasing background DNA levels and impeding the detection of ctDNA (31,32). Therefore, hemolysis is a critical concern in sensitive ctDNA detection and ensuring accurate variant identification. Although the use of needles with larger gauges is associated with less hemolysis (31,33), evidence on whether needle gauge size affects ctDNA integrity is lacking. The correlation between biological factors and ctDNA integrity remains to be investigated, including psychosocial and physical stress, site of venipuncture, food consumption, and circadian rhythm (31,34,35).

Improper preservation and processing of blood samples can induce destruction of cells, thereby releasing high molecular weight genomic DNA into the cell-free plasma fraction and diluting the ctDNA even more. Anticoagulant EDTA can stabilize cfDNA and prevent contamination with germline DNA released from normal blood cells. A recent study reported that specialized cfDNA collection tubes with the stabilization reagent provide higher flexibility for sample processing, up to 14 days without affecting cfDNA detection (36). Blood collection tubes (BCTs) without preservative agents, such as EDTA, require almost immediate processing (within 2–4 hours) of blood sampling to prevent hemolysis. Thus, the use of BCTs that contain formaldehyde-free cell preservative agents, such as Streck or PAXgene, facilitates the processing procedure because they stabilize blood cells, avoiding the need for rapid processing. It is necessary to mix the blood properly with the preservative agents for optimal prevention of cell degeneration. The heparin BCTs are not appropriate for use in ctDNA testing due to the potential effect on polymerase chain reaction (PCR) applications (37). Several studies have summarized the use of different BCTs (31,38–41), drawing a conclusion that the ctDNA levels and quality were similar between preservation BCTs and EDTA when worked up within the recommended maximal processing times. The choice of BCTs is therefore often based on costs and the biobanking infrastructure.

Blood collection is a routine procedure in all medical centers. However, the processing of blood for ctDNA extraction and ctDNA testing is generally centralized, within a region, in a specialized department (in academic institutions) or at a dedicated facility. Consequently, blood samples must be transported either within the hospital or from outpatient clinics to a central laboratory. In many cases, the transportation time exceeds the maximum storage duration for BCTs without preservatives, necessitating the use of preservative-treated BCTs. During transport, BCTs should be handled carefully to prevent cellular degradation due to mechanical stress and sample agitation. A recent study confirmed that transport by pneumatic tube systems within an institution increases the release of cellular DNA into plasma collected in EDTA BCTs compared with transport by courier (42). By contrast, DNA levels in Streck BCTs remained unchanged, indicating that preservative BCTs offer better resistance to mechanical stress. Vacuum BCTs were initially made of glass but are now more commonly manufactured from plastic materials.

The fraction of ctDNA in total cfDNA can vary widely, from less than 0.1% to over 90% (43–46). Detecting ctDNA requires methods with a high analytical sensitivity and specificity. Several high-sensitivity approaches are available, including peptide nucleic acids (PNA)-based methods (47), quantitative PCR and droplet digital PCR (ddPCR), beads emulsion amplification and magnetics (48). However, these methods are inadequate for high-throughput screening. Next-generation sequencing (NGS) platforms offer advantages such as screening for unknown mutations and structural and copy-number variations. Targeted NGS has been developed for ctDNA testing, including cancer personalized profiling by deep sequencing (CAPP-Seq) (49), targeted error correction sequencing (50), and tagged amplicon-based sequencing (51). Whole-exome sequencing (WES) or whole-genome sequencing (WGS) with deep coverage can provide more comprehensive profiling of ctDNA (52,53). However, the clinical implementation of these methodologies in pancreatic cancer management remains constrained by high costs and the requirement for large-volume blood sampling.

In recent years, integrated detection platforms combining artificial intelligence algorithms, enzymatic signal amplification systems, and nanomaterials-based enrichment have been developed to effectively increase the content of ctDNA fragments, thereby facilitating the identification of target gene mutations within ctDNA and inferring the origin of tumors (54). Parallel to these developments, multiple analytical platforms exist for cfDNA fragmentation profiling and methylation mapping. Whole-genome bisulfite sequencing remains the benchmark methodology for genome-wide methylation profiling, enabling precise identification of differentially methylated regions characteristic of malignant transformation (55,56). Nevertheless, whole-genome bisulfite sequencing applications face intrinsic limitations, including DNA degradation-induced signal attenuation during bisulfite conversion and challenges in detecting low-abundance methylated alleles.

Tumor-informed and tumor-naïve approaches

ctDNA-based MRD assessment involves primarily 2 approaches: tumor-informed and tumor-agnostic methods, distinguished by whether the approach relies on previous tumor sequencing information. The choice between these 2 strategies depends on various factors, including research objectives, tumor tissue availability, desired sensitivity, and cost considerations (57).

Tumor-informed assays seem more sensitive than tumor-agnostic assays in many contexts, although sufficient quantity and quality of tumor tissue may not be available in all cases (e.g., from lung cancer biopsies). Among tumor-informed methods, several established ctDNA assays use WES to derive patient-specific probe data from tissue (58). By contrast, broader WGS is used by newer technologies such as the NeXT Personal ultrasensitive ctDNA assay (Personalis), which aims to enhance sensitivity for MRD detection (59).

Key platforms using tumor-informed methods for MRD detection include Signatera, RaDaR, and ArcherDX PCM, all of which use amplicon-based targeted NGS, with limits of detection ranging from 0.001% to 0.02% (60,61). Despite these advantages, WES-based platforms may exhibit uneven coverage in challenging genomic regions, potentially leading to the omission of clinically significant variants. To overcome these limitations, WGS-based tumor-informed platforms offer broader genomic coverage and use advanced computational methods to increase sensitivity (62,63). Finally, hybrid capture-based platforms leverage phased variants to achieve a sensitivity for tumor fractions less than 0.0001% (64,65).

Tumor-naïve methods involve blood-based assays that do not require previous tumor sequencing. Instead, they use predefined panels of recurrent cancer-associated genomic or epigenomic alterations (66,67). These universal panels make tumor-agnostic platforms broadly applicable, offering faster turnaround times and lower costs. However, the lack of individualization may reduce sensitivity, as patient-specific mutations unique to heterogeneous tumors might be missed. Furthermore, their broader genomic coverage can increase background noise, necessitating advanced bioinformatics tools to increase specificity and accuracy (65,68).

Tumor-naïve platforms use either amplicon-based or hybrid capture-based methods. Amplicon-based platforms offer comparable sensitivity for various applications. Hybrid capture-based platforms provide a broader genomic perspective, enabling simultaneous analysis of several genomic regions. A prominent example is Guardant Reveal, a tumor-agnostic ctDNA assay for MRD detection and recurrence monitoring, which has demonstrated clinical validity in colorectal cancer and is under investigation in other solid tumors. This assay integrates genomic and epigenomic alterations to increase sensitivity. Similarly, the U.S. Food and Drug Administration-approved FoundationOne Liquid CDx achieves a limit of detection of 0.37%–0.9% mutant allele fraction (MAF) and is used in non-small cell lung cancer, colorectal cancer, and breast cancer. The AVENIO ctDNA assays use CAPP-Seq technology and a 197-gene panel optimized for non-small cell lung cancer and colorectal cancer, achieving a limit of detection of 0.5%–1% MAF (64).

Emerging tumor-naïve technologies integrate advanced methods such as fragmentomics and methylation analysis to further improve sensitivity and specificity. GRAIL Galleri and OverC extend tumor-agnostic applications to early cancer detection, focusing on using methylation analysis to classify the original tissue of the ctDNA and monitor recurrence. Both platforms have shown promising utility in both MRD and recurrence detection, complementing tumor-informed strategies (65).

Without relying on molecular profiling of the tumor, tumor-naive liquid biopsies can bypass some challenges associated with tumor-informed assays (especially when tumor tissue is limited), although they are overall less sensitive. Most mutation-based tumor-naive approaches use NGS with hybrid capture of fixed genomic segments, performing detection in plasma alone or in both plasma and matched white blood cell samples (49,50,69). Although most tumor-naive liquid biopsy assays are based on mutation, ctDNA assays that assess differentially methylated regions—either alone or in combination with mutations—have demonstrated favorable analytical performance and clinical utility (70). Although initially developed for early detection, tumor-naive whole-genome approaches that leverage cfDNA fragmentomics to calculate and track the tumor fraction are attracting increasing attention (71,72).

CLINICAL APPLICATIONS OF CTDNA IN PANCREATIC CANCER

ctDNA testing for detection of minimal residue disease

Determining MRD after surgical resection poses a clinical challenge, as 60% of recurrence or death within 3 years, even with adjuvant therapy (73). Currently, computed tomography (CT) and magnetic resonance imaging (MRI) are used for postsurgery surveillance of patients with pancreatic cancer. However, disease recurrence detection is often delayed due to uncertain recurrent locations and small tumor sizes. Therefore, longitudinal ctDNA assessment might aid in detecting MRD and predicting the risk of recurrence.

The sensitivity of ctDNA testing depends on the level of tumor DNA released into the blood and optimal sample collection timing. Liquid biopsy approaches have been developed to identify low ctDNA levels, even less than 0.01% of total cfDNA (74,75). Recent studies demonstrated that ctDNA testing has the ability to detect MRD and tumor relapse after surgery in various cancer types, including pancreatic cancer (76–78). Interestingly, some studies showed that ctDNA testing could detect resistance mutations or disease progression before CT imaging (77,79,80).

In a prospective study of ctDNA assessment for MRD, 116 patients with resected pancreatic cancer underwent tumor-informed ctDNA assays. Detectable postoperative ctDNA was strongly associated with inferior recurrence-free survival (RFS), with hazard ratios of 8.2 (95% confidence interval [CI] 3.4–19.9) and adjusted hazard ratio of 5.6 (95% CI 2.5–12.9) (81). Similarly, in the largest retrospective study, Botta et al reviewed tumor-informed ctDNA levels from 231 patients with stages 1–3 pancreatic ductal adenocarcinoma who underwent curative-intent resection. MRD, as defined by ctDNA detection within the first 12 weeks postsurgery, was associated with pathologic stage; 22.2% of stage 1 patients had detectable ctDNA compared with 19.4% of stage 2 and 50% of stage 3 patients (81). Across all stages, patients with undetectable postoperatively ctDNA had significantly improved disease-free survival (DFS) compared with those with detectable MRD (33.3 months vs 6.4 months).

Apart from pancreatic cancer, ctDNA also plays an important role in monitoring MRD in other gastrointestinal cancers, such as colorectal cancer.

ALTAIR Study was a randomized, double-blind, phase III trial, part of the CIRCULATE-Japan research program, aiming to explore the clinical significance of using trifluridine/tipiracil as preventive therapy in patients with colorectal cancer who were ctDNA-positive after curative resection. Results presented at the 2025 European Society for Medical Oncology Gastrointestinal Cancers Congress suggested that for patients with ctDNA-positive colorectal cancer after curative surgery, trifluridine/tipiracil as preemptive treatment showed a positive trend toward benefit compared with placebo for extending median DFS and improving ctDNA clearance rate, although statistical significance was not reached (82).

DYNAMIC Study is a series of studies, including DYNAMIC-II and DYNAMIC-III. DYNAMIC-II was the first to demonstrate that personalized escalation/de-escalation adjuvant treatment decisions could be made for stage II colon cancer postsurgery based on MRD status, with MRD-negative patients undergoing observation only and treatment administered solely to MRD-positive patients (83). DYNAMIC-III is a multicenter, randomized, phase II/III clinical trial enrolling patients with stage III colon cancer who had undergone surgical resection and were suitable for adjuvant chemotherapy. Patients were randomized 1:1 to either a ctDNA-guided group or a standard management group (84).

Results from the escalation group of the DYNAMIC-III study were presented at the 2025 American Society of Clinical Oncology Annual Meeting. In the ctDNA-guided (escalation) group, patients with a positive ctDNA result detected by a tumor-informed assay at 5–6 weeks postsurgery initiated an intensified adjuvant chemotherapy regimen. The results confirmed the prognostic significance of detectable ctDNA and, for the first time, showed that recurrence risk increased significantly with ctDNA burden. However, treatment intensification did not improve RFS (85). Results from the de-escalation group of the DYNAMIC-III study were presented at the 2025 European Society for Medical Oncology (ESMO) Congress. In this group, patients received de-escalated adjuvant chemotherapy guided by ctDNA. The results suggested that a ctDNA-guided de-escalation strategy for adjuvant chemotherapy is feasible, significantly reducing oxaliplatin exposure (from 88.6% to 34.8%) and decreasing treatment-related hospitalizations and serious adverse events. However, within the overall ctDNA-negative population, this strategy failed to demonstrate noninferiority in the 3-year RFS rate (86).

The PEGASUS Study aimed to evaluate the feasibility of using liquid biopsy to guide clinical management after surgery and adjuvant therapy in patients with high-risk stage II and stage III colon cancer. The study stratified patients by ctDNA testing and applied different intensified treatments, thereby altering postoperative treatment choices. Results reported at the 2023 ESMO Congress showed that for high-risk stage II and III colon cancer, dynamic escalation/de-escalation interventions based on MRD status after curative surgery could be applied to guide postoperative clinical management by reducing unnecessary toxicity and improving response to standard chemotherapy, and introduced the novel concept of plasma-conversion therapy (87). (Ref) Final results presented at the 2025 ESMO Annual Meeting suggested that a ctDNA-guided treatment strategy holds promise for individualizing adjuvant therapy based on molecular risk, yet did not meet expectations (88).

Hence, longitudinal ctDNA monitoring facilitates early detection of molecular recurrence, enabling timely therapeutic intervention that may suppress metastatic dissemination and tumor cell proliferation.

ctDNA testing for therapeutic response

Serial molecular testing is critical for evaluating therapeutic response in neoadjuvant therapy and adjuvant settings. Patients often favor minimally invasive ctDNA testing over repeated invasive tumor biopsies. ctDNA testing provides comparable efficacy in predicting therapeutic outcomes. The release of ctDNA into the bloodstream is triggered by tumor cell death, making dynamic fluctuations of ctDNA concentrations a potential biomarker for early treatment response assessment. The longitudinal monitoring of ctDNA has been shown to have therapeutic efficacy and detect disease progression approximately 5 months earlier than conventional radiological imaging or CA19-9 biomarker evaluation (89,90). Declining ctDNA levels correlate strongly with favorable treatment responses, whereas rising levels are indicative of progressive disease (49,74,77,79,91). Notably, sustained undetectable ctDNA levels across serial assessments may signal a complete pathologic response.

ctDNA has emerged as a critical dynamic biomarker for monitoring therapeutic efficacy during neoadjuvant treatment. Groot et al (25) reported a markedly reduced probability of ctDNA detected in patients undergoing neoadjuvant chemotherapy compared with those without preoperative chemotherapy (21% vs 69%; P < 0.001). In a cohort of 38 patients with locally advanced pancreatic cancer treated with neoadjuvant FOLFIRINOX followed by chemoradiotherapy, preoperative ctDNA-negative status (16 patients) was associated with a significantly higher rate of R0 resection and node-negative pathology (88%) compared with ctDNA-positive patients (22 patients; 50%; the Fisher exact P = 0.036) (92). Similarly, Du et al highlighted the utility of serial ctDNA monitoring in predicting tumor response to preoperative immunochemotherapy, demonstrating that patients with a >50% reduction in maximum variant allele frequency between baseline and first evaluation exhibited prolonged survival, enhanced treatment response rates, and increased surgical resectability postneoadjuvant therapy (7). These findings align with broader evidence supporting the predictive value of ctDNA testing for immunotherapy outcomes in non-small cell lung cancer (93) and gastric cancer (94). These findings establish ctDNA as a minimally invasive biomarker that mirrors real-time tumor dynamics, capturing critical transitions such as disease progression or regression, thereby refining therapeutic personalization in oncology.

ctDNA demonstrates detectable levels in 75%–99% of metastatic pancreatic cancers and exhibits considerable potential as a clinical management tool for both localized and advanced disease stages (95–97). Wei et al investigated the utility of ctDNA for monitoring therapeutic response in a prospective cohort of 17 patients with pancreatic cancer treated with the FOLFIRINOX regimen. Among the 12 patients achieving clinical response, 11 demonstrated a significant reduction in the MAF of ctDNA. Conversely, all 5 patients with confirmed chemotherapy resistance exhibited an increase in ctDNA MAF during disease progression (98). Serial ctDNA quantification further serves as a precision biomarker for evaluating radiation therapy efficacy. A multicenter analysis of 138 patients with metastatic gastrointestinal cancer (29% pancreatic cancer) identified 101 cases with trackable ctDNA mutations. Longitudinal monitoring revealed that a reduction of ctDNA MAF at 4 weeks strongly predicted radiographic response, with median ctDNA decreases of 98.0% in patients achieving partial response compared with 49.0% in those with progressive disease (P < 0.001) (99). These findings highlighted the utility of ctDNA testing applied in guiding personalized therapeutic strategies, including chemotherapy intensification/de-escalation and radiation field optimization.

ctDNA testing for the evaluation of prognosis

ctDNA serves as a critical prognostic biomarker in pancreatic cancer, providing insights into the tendency for tumor metastasis, recurrence risk, and patient survival outcomes. Substantial evidence supports the prognostic distinction between ctDNA-detectable and nondetectable cases, with most studies using PCR-based methods to detect mutant KRAS, a driver mutation present in 90%–93% of pancreatic malignancies (100–102). Chen et al established a significant correlation between KRAS-mutant ctDNA and clinical outcomes. In patients with nonelevated CA19-9 levels, ctDNA detection rates reached 93.7% and 86.4% for predicting progression and overall survival (OS), respectively; meanwhile, KRAS testing could also accurately predict treatment response in 80% of cases (103). ddPCR-based analyses further corroborated these findings. A recent study reported markedly poor prognosis in patients with KRAS-mutated ctDNA, showing a median OS of 170 days compared with 489 days in mutation-negative counterparts. Notably, tissue-based KRAS mutations failed to exhibit comparable prognostic significance (24). Another study investigated postoperative outcomes through analyzing 2 KRAS mutations (G12D and G12V) in ctDNA, revealing significantly reduced DFS in patients harboring both mutations (104). Strikingly, Guo et al specifically assessed the prognostic impact of the KRAS G12D variant in PDAC (n = 26), indicating dramatically shorter median OS (12.1 vs 24.9 months, P < 0.001) and RFS (6.3 vs 17.4 months, P < 0.001) in G12D-positive patients vs negative patients (105).

ctDNA analysis is also reflective of prognosis in patients undergoing neoadjuvant or adjuvant therapy (7,106). In a cohort of 38 patients with locally advanced pancreatic cancer receiving neoadjuvant FOLFIRINOX followed by chemoradiotherapy, preoperative ctDNA-negative status (n = 16) correlated with significantly higher rates of R0 resection and node-negative pathology (88%) compared with ctDNA-positive patients (n = 22, 50%; P = 0.036) (92). Further evidence highlighted the prognostic relevance of serial ctDNA monitoring (106). Among 35 patients with postoperative ctDNA assessment, detectable postoperative ctDNA (n = 13) was associated with markedly reduced median RFS (5.4 months vs 17.1 months in ctDNA-negative patients). At a median follow-up of 38.4 months, 12/22 ctDNA-negative patients remained recurrence-free, whereas all ctDNA-positive patients (13/13) experienced recurrence. Notably, patients achieving ctDNA clearance after therapy (preoperative-positive to postoperative-negative conversion) demonstrated improved outcomes, with a median RFS of 12.2 months (106). The efficacy and safety results of a phase II trial evaluating tislelizumab combined with hypofractionated radiotherapy plus nab-paclitaxel/gemcitabine in patients with BRPC/LAPC showed encouraging clinical activity with a manageable safety profile. Dynamic biomarker exploration revealed that baseline interleukin 6 level (>5 pg/mL) predicted better PFS. Moreover, ctDNA status and clearance demonstrated superior survival (107).

Hence, ctDNA serves as a predictive biomarker for therapeutic efficacy and prognostic stratification in cancer management.

ctDNA testing in early diagnosis

Currently, the main diagnostic methods for pancreatic cancer are endoscopic ultrasound-fine needle aspiration, MRI, and CT. However, due to its insidious and nonspecific symptoms, most patients are diagnosed only when the tumor obstructs the bile duct or invades surrounding nerves (108,109). Invasive biopsy remains the gold standard but carries high risks. Therefore, noninvasive or minimally invasive approaches such as ctDNA testing are attractive alternatives, potentially replacing tissue biopsy in specific situations (such as insufficient tumor tissue), improving treatment success and survival rates, and reducing side effects and costs.

The KRAS gene has garnered significant attention regarding ctDNA mutations because it is highly mutated in pancreatic cancer (101,102). In a study of 26 patients with cancer using DNA-based NGS, KRAS, TP53, APC, FBXW7, and SMAD4 mutations were found in 90% of matched tumor biopsies, with a diagnostic accuracy of 97.7%, average sensitivity of 92.3%, and specificity of 100% across all 5 genes (110). Concordance rates between tumor sequencing and ctDNA mutation detection vary but are generally high; for example, 1 study showed 86% for TP53 mutations and 81% for KRAS mutations (96). Thus, ctDNA is particularly useful for identifying cancer mutation profiles in patients with insufficient tumor tissue and can detect uncommon but highly actionable mutations such as KRAS G12C (5,6).

To comprehensively assess the value of ctDNA testing in early diagnosis, researchers are exploring combinations with other biomarkers. Cohen et al (111) conducted a case-control study involving 221 patients with resectable pancreatic cancer and 182 healthy individuals, demonstrating that genetic alteration could be detected with elevated protein markers. The combination of ctDNA and protein markers performed better than individual markers in screening. Another study on 68 patients with solid pancreatic tumors (58 malignant, 10 benign) achieved a sensitivity of 78% and specificity of 91% for the diagnosis of pancreatic cancer using a combination analysis involving CA19-9, ctDNA, and CTCs (112). Furthermore, a registered clinical study (NCT03334708) aims to develop blood-based biomarkers, including ctDNA, for early diagnosis and treatment response assessment in pancreatic cancer.

As an alternative diagnostic test, ctDNA hypermethylation patterns have been described for several genes. For instance, ADAMTS1 and BNC1 methylation status in cfDNA demonstrated excellent diagnostic performance for early-stage pancreatic cancer (sensitivity: 97.4%, specificity: 91.6%, area under the curve: 0.95) (113). In another study, HOXD8 and POU4F1 hypermethylation showed a sensitivity of 56.8% in a large cohort of 372 patients with metastatic pancreatic cancer (114). Singh et al reported that SPARC and NPTX2 hypermethylation can distinguish pancreatic cancer from healthy controls and patients with chronic pancreatitis (115). Alternatively, ZNF154 has also been reported to differentiate patients with cancer, including those with pancreatic, colorectal, and liver malignancies, from healthy controls (116); however, such testing inherits intrinsic limitations to specificity. Other studies have combined metrics to improve test performance. For example, Fujimoto et al reported increased sensitivity from 50.9% to 85.5% for a RUNX3-based assay when combined with CA19-9 measurement (117). Although further optimization may be required for clinical applications, methylation markers offer a promising and noninvasive diagnostic strategy for pancreatic cancer.

CHALLENGES AND PERSPECTIVES

The application of ctDNA testing in the clinical management of pancreatic cancer embodies the principles of precision oncology. Liquid biopsies, which enable noninvasive acquisition of tumor-derived biological material, provide comprehensive molecular profiles of tumor dynamics. Furthermore, these genomic insights enhance our understanding of tumorigenic pathways and metastatic evolution, empowering clinicians to tailor therapeutic strategies based on individual molecular signatures. Notably, ctDNA profiling demonstrates particular utility in patients with inadequate tumor tissue for conventional biopsy, overcoming sampling limitations while identifying rare but clinically actionable targets, such as the KRAS G12C mutation (5,6).

Although blood-based ctDNA analysis has shown significant potential for molecular diagnostics across various cancers, its clinical application has been hindered. The primary reasons for this include substantial variability in preanalytical workflows and a lack of standardization for analytical applications. Improper handling during blood collection and processing can lead to contamination of ctDNA by genomic DNA, diluting the ctDNA and generating uninformative variants that complicate molecular interpretation. Quantitative and qualitative assessment of ctDNA, using patient-matched reference samples, is necessary to determine its suitability for molecular analysis. The variety of commercially available products and the absence of universally applicable reference standards inherently contribute to differences in analytical procedures across laboratories, which, in the absence of collaboration, severely affect interlaboratory comparability and quality assurance. Moreover, these variations hinder the generation of evidence regarding the clinical utility of liquid biopsy assays. Initial collaborative efforts among (international) consortia and associations have developed guidelines, paving the way for harmonization; however, these guidelines lack stringent quality standards and evaluation protocols. It is essential to establish guidelines that define quality criteria for key steps across the workflow, including preanalytical (e.g., selection of BCTs, plasma processing, and ctDNA extraction methods), analytical (e.g., choice of detection method, use of [patient-matched] reference samples, and analysis of results), and postanalytical (e.g., variant interpretation and reporting) phases. Interinstitutional collaboration through external quality assessment can help develop standardized guidelines and standard operating procedures, which are necessary for the clinical implementation and reimbursement of ctDNA-based liquid biopsy testing.

The application of ctDNA analysis for MRD detection is gaining clinical traction across solid tumors, including pancreatic cancer (106,118,119). Emerging evidence has prompted the initiation of prospective clinical trials evaluating the prognostic and therapeutic utility of postoperative ctDNA monitoring in resected pancreatic cancer. A pivotal case is the multicenter DYNAMIC Pancreas trial (CTRA U1111-1209-6200), which randomizes patients after curative resection into 2 arms: one receiving standard adjuvant chemotherapy without knowledge of postoperative ctDNA status, and the other using a ctDNA-guided therapeutic strategy. In the intervention arm, adjuvant therapy is dynamically adjusted based on serial ctDNA quantification and mutational profile trends. This paradigm aims to address unmet needs in precision postresection management by stratifying patients for treatment intensification or de-escalation.

ctDNA has become the most important biomarker in solid tumors for the next 5–10 years. Previous research has clearly established its significant value in prognostic assessment and predicting the efficacy of adjuvant therapy. The ALTAIR study, published in 2025, further indicates that ctDNA can be used to predict the efficacy of neoadjuvant therapy. This study demonstrated that dynamic monitoring of ctDNA can accurately predict the efficacy of chemoradiotherapy and recurrence risk in anal squamous cell carcinoma. Molecular-level recurrence signaled by ctDNA occurred significantly earlier than clinical/radiological recurrence, highlighting its potential as a highly effective predictive marker.

Notably, however, despite considerable interest in prospective interventional studies guided by ctDNA, studies such as ALTAIR, DYNAMIC-III, and PEGASUS all failed to achieve their expected outcomes, making this a year of setbacks for the field.

Among these, the reasons why the DYNAMIC-III study failed to meet its primary endpoints warrant in-depth analysis. The failure can likely be attributed to 2 main factors. On one hand, there were methodological limitations in MRD detection. The study, initiated in 2017, used a tumor-naive MRD detection strategy rather than a tumor-informed assay. Data from another study presented by the research team at the 2024 American Society of Clinical Oncology Annual Meeting showed that compared with newer methods based on WES, this older detection method had significant shortcomings in both sensitivity and specificity. The research team subsequently abandoned the original method in the follow-up DYNAMIC-IV study, adopting a new strategy instead, which indirectly confirms the inadequacy of the initial methodology. On the other hand, there were flaws in the treatment strategy design. The study formulated interventions based solely on the binary positive/negative status of MRD, overlooking the critical dimension of quantitative ctDNA levels. This one-size-fits-all approach failed to provide refined stratification based on the depth of a patient's tumor burden, potentially leading to unnecessarily intensified treatment of some patients with low-level positivity.

This further prompts reflection on the clinical dilemmas facing the MRD field: Which ctDNA detection technology can truly represent MRD? How should its criteria be established? Which technological path is superior? Currently, ctDNA detection strategies are mainly divided into tumor-informed and tumor-naive approaches. Overall, personalized tumor-informed technology has become the mainstream direction for MRD detection in colorectal cancer. Simultaneously, in MRD-guided intervention decisions, the MRD status during the window period is currently the core basis for formulating treatment strategies.

The future direction for MRD lies in quantitative analysis. Future MRD assessment must rely on quantitative calculations to reliably measure tumor burden, thereby providing a basis for clinical intervention. Current escalation interventions based on MRD primarily involve extending the duration of adjuvant chemotherapy—existing research suggests one possible mechanism for prolonged survival with extended adjuvant chemotherapy is the clearance of positive ctDNA (MRD). However, even after completing current standard adjuvant chemotherapy, it remains difficult to completely clear all positive ctDNA. Therefore, for patients with residual positivity, combining other treatment strategies should be considered, and choices should be personalized based on patient molecular characteristics. For example, immunotherapy could be combined for microsatellite instability-high patients, or bevacizumab could be considered for patients with RAS mutations.

ctDNA testing is being studied for additional novel clinical applications, including multicancer screening and early detection (70,75,120). However, implementation in routine clinical practice remains challenging due to the inherently low ctDNA concentrations in peripheral blood. The CancerSEEK assay, a multimodal platform combining multiplex PCR-based ctDNA mutation profiling across 16 cancer-associated genes with protein biomarker analysis (CA19-9 and carcinoembryonic antigen), demonstrated 69%–98% sensitivity for 5 malignancies (liver, gastric, pancreatic, esophageal, and colorectal cancers), with at >99% specificity (75). Emerging methodologies beyond WGS show enhanced ctDNA detection capabilities. Whole-genome bisulfite sequencing of cfDNA showed better ability than WGS-based classifiers in inferring the tumor tissue of origin. identifying cancer-specific methylation patterns and predicting tumor tissue of origin (54). Despite these advances, current multicancer early detection tests using cfDNA methylation signatures exhibit modest overall sensitivity (51.5%; 95% CI 49.6–53.3), with markedly reduced performance in stage I malignancies (16.8%; 95% CI 14.5–19.5) (121). Thus, although ctDNA assays show promise for early detection of cancers, insufficient sensitivity for small early-stage tumors with minimal ctDNA shedding is a significant barrier (122). Furthermore, recent studies highlighted the diagnostic potential of nonplasma biofluids (stool, bile, urine, peritoneal fluid), which frequently demonstrate elevated ctDNA concentrations compared with peripheral blood, suggesting their utility as complementary liquid biopsy sources (123–125).

Evaluation of multiple liquid biopsy biomarkers has the potential to enhance the accuracy of MRD detection, as different classes of biomarkers reflect distinct aspects of MRD biology. In principle, multiomic assessment within a single blood sample is feasible, albeit requiring separation of plasma components (cfDNA, extracellular vesicles, proteins, and metabolites) from cellular components (CTCs, immune cells, circulating cancer-associated fibroblasts, and endothelial cells) (27). Technological advances have enabled automated sorting of these diverse analytes (126), although some require specific preanalytical conditions (127).

To our knowledge, multiomic assessment of liquid biopsy samples has only been performed in a limited number of studies and is not yet routinely integrated into clinical decision-making. As noted above, ctDNA analysis can incorporate a combination of genetic and epigenetic features (128–132), and fragmentomics approaches (133–135). These technologies can improve MRD detection performance but come with increased assay complexity and cost.

In patients with cancer, only a small fraction of cfDNA (typically 0.01%–5%) is ctDNA shed by tumor cells into the bloodstream (136). Ultrasensitive targeted methods, such as ddPCR, beads emulsion amplification and magnetics, or real-time PCR assays, enable rapid, cost-effective, and highly sensitive detection of predefined cancer-associated mutations with high sensitivity.

NGS-based methods include both targeted and untargeted approaches and are distinguished by their capacity for massively parallel sequencing of millions of DNA sequences. Targeted NGS methods such as tagged amplicon-based sequencing, safe-sequencing system, and CAPP-Seq allow simultaneous detection of multiple rare mutations in ctDNA. Despite their high analytical sensitivity, targeted methods are limited to mutations in a predefined set of genes. By contrast, untargeted approaches such as WGS or WES offer the opportunity to detect novel, clinically relevant genetic alterations without previous knowledge of the primary tumor. However, the clinical application of untargeted methods is often limited by generally lower sensitivity, higher sample input requirements, and increased costs (137–139).

Furthermore, the appropriate clinical trial design for ctDNA and other liquid biopsies as a biomarker poses a challenge due to the complex balance between trial feasibility of patient inclusions and costs, and the need for meaningful outcomes. The current challenges include concerns in the validation of liquid biopsies as a stratification marker, the validation of liquid biopsy dynamics as a surrogate end point in current clinical trials, and randomizing patients with a positive assay, and the incorporation of current stratification based on clinicopathological factors. A series of ongoing clinical trials are summarized in Table 1.

Table 1.

Ongoing Clinical trials for ctDNA in pancreatic cancer

Study ID	Study type	Study population	Enrollment (estimated)	MRD detection method	Primary outcome	Official title
NCT03334708	Observational	Advanced pancreatic cancer, operable pancreatic cancer, and other control cohorts	700	Tumor-informed ctDNA assay	Change in biomarkers to determine sensitivity and specificity of the assay to diagnose early stage pancreatic cancer	Development of biomarkers for the early detection, surveillance, and monitoring of pancreatic ductal adenocarcinoma
NCT07080021	Observational	High-risk resectable, borderline resectable, or locally advanced stage pancreatic cancer	119	Tumor-informed ctDNA assay	The correlation between ctDNA-MRD status at serial monitoring points during neoadjuvant therapy and therapeutic efficacy (R0 resection)	A prospective observational cohort study on longitudinal monitoring of ctDNA MRD in neoadjuvant therapy for pancreatic cancer
NCT05479708	Observational	Stage I–III pancreatic adenocarcinoma	150	Tumor-informed ctDNA assay	Death and overall survival (time frame: 3 yr)	Postoperative ctDNA-based minimal residual disease detection for resected pancreatic adenocarcinoma: a prospective observational cohort study
NCT06151691	Observational	Stage I/II pancreatic ductal adenocarcinoma	51	Multiomics (methylation, fragments, and CNV)	DFS (time frame: 1.5 yr)	Explorations of cell-free DNA multiomics technology in detection of minimal residual disease and disease prognosis after surgery in early pancreatic ductal adenocarcinoma: a single-center, prospective, observational case study
NCT07054879	Observational	Pancreatic ductal adenocarcinoma underwent R0 radical resection	70	Tumor-informed ctDNA assay	2-yr DFS rate	A prospective single-center observational study: ctDNA-guided maintenance therapy for postoperative pancreatic cancer
NCT06043921	Observational	Clinical stage III/IV pancreatic ductal adenocarcinoma (T1-3N2M0, T4 anyNM0, anyTanyNM1)	150	Tumor-informed ctDNA assay	Unresectable pancreatic cancer: rate of concordance of KRAS mutations between tumor tissue and blood sample Resectable pancreatic cancer: success rate of WES assays and selections of personalized genes using tumor tissue specimens obtained by EUS-FNA/FNB	Multicenter study of circulating tumor DNA in patients with pancreatic cancer using a personalized panel
NCT05802407	Interventional	Pancreatic cancer who had undergone curative-intent surgery (R0 or R1 resection)	100	Tumor-naïve ctDNA assay	DFS (time frame: 3 yr)	The value of molecular residual disease monitoring based on ctDNA in resected pancreatic cancer
NCT06966440	Interventional	Stage I–III pancreatic ductal adenocarcinoma underwent R0 resection	856	Tumor-naïve ctDNA assay	DFS (time frame: 3 yr)	Assessing the impact of ctDNA MRD-guided adjuvant therapy on outcomes in resectable pancreatic cancer
NCT05802394	Interventional	Borderline resectable or locally advanced pancreatic cancer	100	Tumor-naïve ctDNA assay	Resection rate	The value of molecular residual disease monitoring based on ctDNA in borderline resectable or locally advanced pancreatic cancer
NCT06867146	Interventional	Stage I–III pancreatic cancer, who are scheduled to undergo curative surgery	60	Tumor-naïve ctDNA assay	DFS (time frame: 1.5 yr)	The role of molecular residual disease in the treatment strategy and prognosis prediction of pancreatic cancer patients undergoing adjuvant therapy

CNV, copy number variant; ctDNA, circulating tumor DNA; DFS, disease-free survival; EUS, endoscopic ultrasound; FNA, fine needle aspiration; FNB, fine needle biopsy; MRD, minimal residual disease.

In summary, ctDNA-based liquid biopsy is becoming an indispensable tool for tumor diagnosis and therapeutic monitoring. Notably, its clinical utility is particularly promising for improving the management of pancreatic cancer, a malignancy plagued by late-stage detection and limited diagnostic options. Furthermore, the convergence of technological advancements in sensitivity and the completion of prospective large-scale clinical trials are required for validating multiple tumor-derived biomarkers. These developments underscore the transformative potential of ctDNA testing as a cornerstone of precision medicine in oncology.

CONFLICTS OF INTEREST

Guarantor of the article: Guoliang Yao, PhD.

Specific author contributions: G.Y.: supervision, writing original draft. Y.Z.: writing original draft, methodology. H.L.: project administration, visualization. L.T.: formal analysis. Y.M.: writing-review and editing. S.W.: formal analysis. C.L.: conceptualization. Y.F.: visualization. D.C.: conceptualization, supervision. All authors reviewed the manuscript.

Financial support: None to report.

Potential competing interests: None to report.

Consent for publication: This manuscript has been read and approved by all the authors to publish and is not submitted or under consideration for publication elsewhere.

Ethics declaration: This study was approved by the Medical Research Ethics Committee, the First Affiliated Hospital of Henan University of Science and Technology. The protocol adhered to the ethical principles of the Declaration of Helsinki and Good Clinical Practice Guidelines.

ABBREVIATIONS:

ACT

adjuvant chemotherapy

AI

artificial intelligence

BCTs

Blood collection tubes

BEAMing

beads emulsion amplification and magnetics

CA

carbohydrate antigen

CAPP-Seq

cancer personalized profiling by deep sequencing

cfDNA

cell-free DNA

CT

computed tomography

CTCs

circulating tumor cells

ctDNA

circulating tumor DNA

ctRNA

circulating tumor RNA

ddPCR

droplet digital PCR

ESMO

European Society for Medical Oncology

EUSFNA

endoscopic ultrasound-fine needle aspiration

EVs

extracellular vesicles

FTD/TPI

trifluridine/tipiracil

gDNA

genomic DNA

HMW

high molecular weight

MAF

mutant allele frequency

MRD

minimal residue disease

MRI

magnetic resonance imaging

NGS

next-generation sequencing

PDAC

pancreatic ductal adenocarcinoma

qPCR

quantitative polymerase chain reaction

RFS

recurrence-free survival

TAm-Seq

tagged amplicon-based sequencing

TEC-Seq

targeted error correction sequencing

TEPs

tumor-educated platelets

WES

whole-exome sequencing

WGBS

whole-genome bisulfite sequencing

WGS

whole-genome sequencing