Circulating tumor DNA in biliary tract cancers: A review of current applications

Abstract

Molecular profiling of biliary tract cancers (BTCs) has paved the way for a broader range of therapeutic options, leading to improved survival outcomes. Given the challenges of tissue evaluation in BTCs, circulating tumor DNA (ctDNA) has emerged as a promising non-invasive biomarker for genomic profiling. Bile has been proven to be a reliable ctDNA source, demonstrating higher concordance with tumor tissue than plasma. More importantly, ctDNA provides valuable insights into both clonal evolution and treatment response, including the detection of resistance mechanisms and mutation clearance, which are often associated with disease control. Although its role in recurrence monitoring remains investigational, early studies suggest that ctDNA detection may precede radiological recurrences. This review examines recent advancements in ctDNA analysis for patients with BTC, highlighting key developments, current clinical implications, and ongoing challenges. Large-scale prospective studies are needed to validate the clinical utility of ctDNA and to support its integration into BTC management.

Key Words: Biliary tract cancer; Circulating tumor DNA; Genomic profiling; Bile-derived circulating tumor DNA; Clonal evolution; Mutation clearance

Core Tip: Molecular profiling of biliary tract cancers has significantly transformed the management of these diseases by paving the way for a broader range of therapeutic options, leading to improved survival outcomes. Given the scarcity of tissue, circulating tumor DNA (ctDNA) has emerged as a promising non-invasive biomarker for genomic profiling, with high concordance with genomic alterations identified in tissue, particularly in therapy-naïve and metastasis-derived samples. Furthermore, ctDNA has a strong prognostic association and is an effective strategy for monitoring patients during systemic treatment.

INTRODUCTION

Biliary tract cancers (BTCs), including gallbladder cancer (GBC), intrahepatic cholangiocarcinoma (iCCA), and extrahepatic cholangiocarcinoma (eCCA), are aggressive malignancies with poor prognosis, with a median 5-year survival rate of approximately 2% in advanced stages[1]. Advancements in precision oncology have profoundly impacted patients’ clinical outcomes. By uncovering the molecular heterogeneity among BTC subtypes and identifying actionable genomic alterations, these advances have transformed the treatment landscape into metastatic setting, enabling the use of targeted therapies. However, obtaining adequate tissue samples for analysis remains challenging. Anatomical constraints and extensive desmoplastic stromal reactions frequently result in insufficient DNA quantity or quality for successful molecular characterization[2]. In this context, next-generation sequencing (NGS) has recently been integrated with liquid biopsy methods derived from peripheral blood, bile, and peritoneal fluid, including circulating free DNA (cfDNA), circulating tumor cells (CTCs), microRNAs, circulating proteins, and extracellular vesicles, thus expanding the opportunities for tumor genomic analysis.

Another advantage of liquid biopsy methods is their ability to capture intratumoral heterogeneity, which is inadequately assessed by single-site biopsies, allowing for the detection of tumor evolution and mechanisms of resistance[3]. The short half-life of ctDNA enables real-time tracking of genomic alterations, facilitating longitudinal disease monitoring, therapeutic response assessment, minimal residual disease (MRD) detection, and elucidation of resistance mechanisms to targeted therapies[4,5].

This systematic review summarizes the current applications of ctDNA in BTCs, addressing its role in molecular profiling, MRD assessment, treatment monitoring, and resistance mechanism detection.

cfDNA AND CELL-FREE TUMOR DNA

Circulating cfDNA, initially described in 1948[6], comprises short single- and double-stranded DNA fragments averaging approximately 167 base pairs[7,8]. In the bloodstream, cfDNA is protein-bound or encapsulated within extracellular vesicles, such as apoptotic bodies and microvesicles[9,10]. Although the precise mechanisms underlying cfDNA release remain partially understood, passive processes such as apoptosis, necrosis, and phagocytosis, along with active secretion by viable cells, have been implicated[11]. Under physiological conditions, cfDNA primarily originates from hematopoietic cells, vascular endothelial cells, and hepatocytes, maintaining a steady-state equilibrium between continuous generation and systemic clearance[12]. However, in patients with cancer, this balance is disrupted, resulting in elevated cfDNA levels derived mainly from tumor cells and adjacent stromal tissue due to increased cellular turnover and impaired clearance[13].

Circulating tumor DNA (ctDNA) is a fraction of cfDNA that predominantly originates from tumor cells and is characterized by genetic alterations mirroring those of tumor tissue, although normal cells can also contribute[14]. Typically, ctDNA is approximately 140 base pairs in length with a half-life of approximately two hours, ctDNA generally constitutes a minor fraction of the total cfDNA but is positively correlated with advanced clinical stage, greater tumor burden, and jaundice[5].

Common techniques for ctDNA analysis include digital PCR (dPCR), targeted NGS, whole-exome sequencing, and methylation profiling[15,16]. dPCR offers exceptional sensitivity and quantitative precision for detecting known hotspot mutations, making it ideal for monitoring MRD and tracking resistance variants in real time. However, its low multiplexing capacity and reliance on prior knowledge of specific mutations limit its use in broader genomic profiling[17]. Targeted NGS overcomes these limitations by enabling the simultaneous analysis of dozens to hundreds of cancer-related genes, capturing both point mutations and copy number variations with robust sensitivity and relatively fast turnaround times[18]. It also effectively captures tumor heterogeneity, albeit with a slightly increased false-positive rate[19].

Whole-exome sequencing, on the other hand, provides comprehensive genomic coverage but is generally restricted to exploratory or research settings because of its lower sensitivity for low-frequency variants, higher cost, and longer processing time[20]. Methylation-based ctDNA assays, which detect cancer-specific hypermethylation patterns, show promise for early detection and tissue-of-origin identification by capturing epigenomic alterations that can precede mutational events[16]. Additionally, RNA-based anchored multiplex PCR techniques complement DNA-based NGS by enhancing the detection of gene fusions and resolving challenges related to large intronic regions and structural variants[21].

A comprehensive genomic analysis of 1671 patients with BTC identified an average of 4.6 genomic alterations per patient, comprising short variants, copy number alterations, and gene rearrangements. The most frequently altered genes were tumor protein p53 (TP53, 40.0%), cyclin-dependent kinase inhibitor 2A (CDKN2A, 29.0%), KRAS (22.6%), CDKN2B (19.7%), AT-rich interactive domain-containing protein 1A (ARID1A, 16.0%), mothers against decapentaplegic homolog 4 (11.7%), isocitrate dehydrogenase 1 (IDH1, 10.2%), and BRCA1-associated protein 1 (BAP1, 10.2%)[22,23]. Studies have consistently demonstrated high concordance rates (70%-92%) between ctDNA and tissue biopsies in diverse tumors, including BTC[24], for which a 66.67% overall concordance rate was reported between cfDNA and matched tissue DNA, achieving notably high sensitivity and specificity for mutations in TP53, epidermal growth factor receptor (EGFR), and MET genes[25]. By contrast, a lower concordance (18%) was observed for fibroblast growth factor receptor 2 (FGFR2) fusions because of the numerous fusion partners described in iCCA[26]. The timing of ctDNA sampling also significantly impacts mutation detection, with IDH1 mutations showing optimal concordance (100%) when sampled before the initiation of systemic therapy[23]. Qualitative analyses have identified tumor-specific genomic alterations in cfDNA, including mutations in KRAS and p53, microsatellite instability, loss of heterozygosity, aberrant methylation, and tumor-specific messenger RNA (mRNA), supporting its use as a cancer biomarker[27].

BILE ctDNA

Due to the anatomical challenges of obtaining BTC biopsies, bile fluid emerges as an ideal source for liquid biopsy, outperforming plasma ctDNA, with higher sensitivity and a strong concordance rate (87.5%)[28]. The sensitivities range from 83.3% to 94.7%, with specificities nearing 100%[29]. Bile is rich in proteins, mRNA, and microRNA, providing a deeper understanding of the underlying cancer biology[30]. Notably, KRAS mutations are detected more frequently in bile ctDNA than in plasma (48% vs 18.8%)[31]. Although the mutation detection rate in bile does not vary with disease stage, its presence is associated with poorer survival, suggesting its potential as a prognostic biomarker.

Despite its promise, the use of bile in molecular diagnostics remains limited by the lack of standardized protocols for sample collection, processing, and storage, which influence biomarker stability and the reliability of tests. A recent study partially addressed this gap by showing that ctDNA and protein biomarkers in bile remain stable for up to seven hours at room temperature, highlighting bile as a practical diagnostic method that can be used in settings where immediate processing is not possible[32].

In addition to bile, other non-blood sources of ctDNA emerge as promising alternatives for molecular profiling, particularly in cancers that shed low amounts of ctDNA into the bloodstream. Although several studies have investigated cfDNA from these alternative fluids, the effectiveness of detection varies according to tumor histology and disease stage[33]. While preliminary findings support their utility for diagnosis, prognosis, and treatment monitoring, clinical implementation remains limited due to the lack of standardized protocols and validation in larger cohorts. As a result, despite encouraging preliminary findings, the clinical use of cfDNA from non-blood fluids remains investigational.

MULTICANCER EARLY DETECTION

Surgical resection remains the sole curative therapeutic option for patients with BTC. However, most patients present with advanced disease due to the absence of specific clinical symptoms, laboratory biomarkers, or imaging modalities with good sensitivity and specificity for early diagnosis[34-36].

Exploratory studies have consistently demonstrated elevated cfDNA levels in patients with BTC, enabling high sensitivity and specificity in distinguishing these cases from healthy controls and cholecystitis[37,38]. ctDNA and CTC analyses serve as valuable diagnostic tools, particularly for patients presenting with radiographic suspicion, allowing for more accurate identification and classification[39]. The PREVAIL trial reported 100% sensitivity and 75% specificity for tumor detection using ctDNA[40]. Mishra et al[24] further supported these findings, identifying significant expression changes in a panel of five long non-coding RNAs in serum samples from patients with BTC, reinforcing their diagnostic and prognostic potential.

However, despite promising results for early cancer detection, its clinical utility remains limited due to low ctDNA shedding, poor sensitivity for alterations such as FGFR2 fusions, and a high risk of false-negative and false-positive results[41]. Technical barriers such as DNA degradation, limited detection of copy number variations and rearrangements, and lack of spatial specificity further complicate its reliability[42].

To overcome some of these limitations, Liu et al[43] demonstrated that enriching shorter cfDNA fragments (90–150 base pairs) through a single-stranded DNA library significantly improved the sensitivity for detecting low-frequency variants. However, the study was limited by a small cohort, necessitating further validation, particularly for tumors with low ctDNA shedding, such as brain tumors or brain-only progression.

Barbato et al[44] developed a genetic model for GBC based on specific genetic alterations observed during tumor development, aiming to predict the risk of GBC and inform preemptive gallbladder removal decisions. An artificial intelligence-based diagnostic tool was proposed to distinguish GBC from other gallbladder disorders. GBCseeker integrates cfDNA genetic signatures, radiomic features, and clinical data from 301 patients with gallbladder-occupying lesions, achieving high accuracy in distinguishing malignant from benign conditions and reducing surgeons' diagnostic errors by 56.24%[45]. Phallen et al[46] developed targeted error correction sequencing (TEC-Seq), a sensitive and specific NGS-based approach for detecting low-frequency genomic alterations in cfDNA. TEC-Seq identified ctDNA alterations in up to 67% of patients with stage I cancers. Similarly, the Cancer SEEK assay achieved 83% accuracy in predicting tissue origin, with detection rates ranging from 59% to 71%, depending on the cancer type[47]. Although the low ctDNA concentration in early disease stages often causes the mutant allele fraction to fall below the detection limits of current methods[48], these recent technologies have already enabled ctDNA detection in stage I tumors of certain cancers.

Further research demonstrated that the bile-derived cfDNA levels of syncytin-1 and solute carrier family 7 member 11 were significantly higher in patients with CCA than in those with gallstones (3.06 vs 1.32, P < 0.001; 2.39 vs 1.30, P < 0.001), correlating significantly with disease burden. Combining these markers with carbohydrate antigen 19-9 (CA19-9) substantially improved diagnostic accuracy (area under the curve: 0.927), suggesting their potential as novel early diagnostic biomarkers[49].

To date, the value of ctDNA for early diagnostic purposes is used only for adjunctive purposes combined with conventional evaluations for confirmation and localization of disease, as summarized in Table 1. It cannot be used as a substitute for standard screening.

Table 1 Studies evaluating circulating tumor DNA for early detection.

Ref.	Title	Clinical relevance
Phallen et al[46], 2017	Direct detection of early-stage cancers using circulating tumor DNA	Investigates ctDNA levels in patients with early-stage solid tumors; higher levels are associated with recurrence and worse overall survival
Liu et al[37], 2024	Cell-free DNA in plasma reveals genomic similarity between biliary tract inflammatory lesion and biliary tract cancer	Compares molecular profiles of biliary tract cancer and inflammatory bile diseases to aid differential diagnosis
Mencel et al[40], 2022	Liquid biopsy for diagnosis in patients with suspected pancreatic and biliary tract cancers: PREVAIL ctDNA pilot trial	Evaluates the utility of ctDNA in confirming cancer diagnosis in patients with radiologically suspected pancreatic or biliary malignancies
Yang et al[45], 2023	Multimodal integration of liquid biopsy and radiology for the noninvasive diagnosis of gallbladder cancer and benign disorders	Develops a ctDNA-based tool for the preoperative diagnosis of gallbladder cancer, supporting surgical decision-making
He et al[49], 2025	Bile-derived cfDNA of syncytin-1 and SLC7A11 as a potential molecular marker for early diagnosis of cholangiocarcinoma	Explores the role of syncytin-1 and SLC7A11 expression in cholangiocarcinoma pathogenesis
Konczalla et al[39], 2020	Clinical significance of circulating tumor cells in gastrointestinal carcinomas	Reviews the potential of CTCs for cancer screening and patient stratification
Mishra et al[24], 2024	Diagnostic utility of next-generation sequencing in circulating free DNA and a comparison with matched tissue in gallbladder carcinoma	Analyzes genomic alterations in cfDNA from gallbladder cancer to improve diagnostic precision and inform therapeutic strategies

cfDNA: Cell free DNA; CTC: Circulating tumor cells; ctDNA: Circulating tumor DNA; SLC7A11: Solute carrier family 7 member 11.

ctDNA FOR ACTIONABLE FINDINGS

Targetable alterations are identified in up to 43%-68% of patients with BTC, most frequently FGFR2 fusions, IDH1, and BRAF V600E mutations[23]. Despite being grouped under the same umbrella as GBC, iCCA and eCCA differ not only in their anatomical locations but also in their molecular characteristics and etiological backgrounds[50]. Several studies have identified clusters of BTC-associated molecular aberrations with unique genomic, epigenetic, and clinical features[51]. These mutations have subtype-specific prevalence. iCCA display higher rates of IDH1, BAP1, and polybromo-1 mutations, and FGFR2 fusions. eCCA, on the other hand, more frequently harbor KRAS, CDKN2A, human epidermal growth factor receptor 2 (HER2), and BRCA1 mutations. GBCs often present with homologous recombination repair deficiencies and HER2/neu overexpression or amplification. Moreover, iCCA and GBCs tend to exhibit higher rates of positive predictive biomarkers for immune checkpoint inhibition, such as programmed cell death ligand 1 expression, high microsatellite instability, and high tumor mutational burden, than eCCA[52].

FGFR2 fusions occur in approximately 10% of iCCA cases, typically associated with a better prognosis and a relatively indolent clinical course[22]. FGFR inhibitors, including futibatinib and pemigatinib, have been approved for use in patients with previously treated CCAs with FGFR2 fusions or rearrangements. In a pivotal trial of pemigatinib, patients with FGFR2 fusions or rearrangements achieved a median progression-free survival (mPFS) of 6.9 months and a median overall survival (mOS) of 21.1 months, with an objective response rate of 35.5%, highlighting its clinical benefit in this molecularly defined subgroup[53].

IDH1 inhibitors, such as ivosidenib, have shown efficacy in IDH1-mutant cholangiocarcinoma’s previously treated with systemic therapies. In the ClarIDHy trial, ivosidenib achieved a mOS of 10.3 months compared with 7.5 months with placebo (or 5.1 months when adjusted for crossover) and a previously reported mPFS of 2.7 months vs 1.4 months with placebo[54]. IDH1-mutant tumors also demonstrate unique mRNA expression profiles, copy number alterations, and notably increased methylation of the ARID1A promoter, resulting in decreased ARID1A expression[55].

Alterations in the ERBB/HER family of proteins are especially relevant in eCCA and GBC, with frequencies ranging from 6% to 20%[56]. Roa et al[56] observed that HER2 overexpression in GBC correlates with worse survival outcomes, whereas Javle et al[57] suggested that ERBB2 (also known as HER2) amplification predicts better treatment response compared to ERBB2 mutations. The HERIZON-BTC-01 trial demonstrated the promising efficacy of zanidatamab in previously treated patients with HER2-positive BTC, achieving a disease control rate of 65%, mPFS of 6.7 months, and a 1-year OS rate of 79.1%[58]. Patients with HER2 immunohistochemistry (IHC) scores of 3+ had significantly longer PFS than those with IHC 2+ confirmed by fluorescence in situ hybridization or NGS (8 months vs 1.4 months; P = 0.02)[59]. These findings were supported by the DESTINY-PanTumor02 trial, which reported an overall response rate of 22% [95% confidence interval (CI): 10.6-37.6; P = 0.05] in the overall BTC cohort and 56.3% (95%CI: 29.9-80.2; P = 0.05) in patients with HER2 IHC 3+ expression when using trastuzumab deruxtecan[60]. Combination therapies targeting HER2 have also demonstrated promising efficacy[61-63]. Another commonly identified mutation is KRAS. It is identified in approximately 15.6% of CCA cases, with common variants including G12D (37.0%), G12V (24.0%), and Q61H (8.2%). Patients with KRAS mutations have reduced overall and recurrence-free survival. Elevated cfDNA levels of KRAS G12/G13 mutations combined with high CA19-9 Levels correlate strongly with poor prognosis, particularly in CCA[64]. Although KRAS mutations are well-known activators of the RAS-RAF-MEK-ERK signaling pathway, clinical trials have not yet demonstrated significant clinical benefits[65].

Kawakami et al[66] identified TP53 as the most mutated gene across various cancers, including BTC, with a mutation rate of 90%, linking it to poor survival outcomes, though clinical trials targeting TP53 remain lacking. Other relevant genomic alterations in BTC include EGFR mutations (about 15% in GBC), which are linked to gemcitabine sensitivity and poor prognosis[67], BRAF mutations (variably reported between 0% and 33% in GBC)[68], and MET overexpression (ranging from 5% to 74% in BTC overall, with a 4.1% mutation rate)[69]. Although less commonly observed, NTRK fusions, BRAF V600E mutations, and RET fusions are identifiable in a subset of patients with BTC, providing opportunities for tumor-agnostic targeted therapies[70].

CTDNA FOR MONITORING MRD ASSESSMENT

MRD reflects the presence of persistent cancer cells that are undetectable by imaging or clinical examination. Due to the limitations of radiological methods in detecting microscopic disease, minimally invasive approaches, such as cfDNA estimation, have emerged to monitor treatment response and disease progression. ctDNA positivity at any time point strongly predicts recurrence, often preceding radiographic detection by months. Sensitivity and specificity can be refined by adjusting positivity thresholds, with ctDNA decline from baseline to the first evaluation correlating with improved survival. Defining both ctDNA presence and dynamic changes, typically based on fold changes or arbitrary thresholds, is crucial[71]. Standardization efforts, such as ctDNA-response evaluation criteria in solid tumors (RECIST), aim to formalize assessments, although confirmatory testing remains crucial to distinguish true progression from biological variability[72,73].

ctDNA-based MRD detection is well established across multiple malignancies, including non-small cell lung cancer (NSCLC), breast cancer, melanoma, head and neck cancer, urothelial cancer, and colorectal cancer[74]. Studies have consistently linked cfDNA levels to survival outcomes: Decreasing levels predict better treatment response and improved survival, while persistently high levels indicate poor prognosis, lack of response, systemic spread, and progression[75].

In BTC, ctDNA research remains limited but promising. A small prospective study by King et al[76] (12 patients’ post-resection) found that all recurrences were ctDNA-positive before imaging detection. Another study of 24 patients post-R0/R1 resection linked a vimentin-positive CTC and cfDNA fragment score to recurrence risk (75% sensitivity, 87.5% specificity)[77]. The largest prospective study in resected eCCA (89 patients, 254 plasma samples, median follow-up 52.8 months) conducted by Yoo et al[78] and a retrospective study (56 patients with BTC, median follow-up 12.8 months) confirmed ctDNA’s prognostic superiority over carcinoembryonic Antigen and CA19-9. Serial ctDNA negativity predicted significantly longer disease free survival (DFS) [hazard ratio (HR) = 6.7; P < 0.001], independent of chemotherapy regimen, with persistence or emergence of ctDNA correlating with poor outcomes (HR = 5.8; P < 0.001)[79]. ctDNA dynamics (clearance vs non-clearance) remained prognostic for DFS and OS, independent of chemotherapy (P-interaction: DFS = 0.746, OS = 0.7314).

Optimizing ctDNA-based MRD monitoring requires precise timing of blood sampling. Early sampling enables timely interventions but risks false negatives due to post-surgical cfDNA elevations, supporting a ≥ 2-week recovery period[80]. Landmark analysis at a fixed time point (e.g., 4 weeks post-surgery) aids early adjuvant therapy decisions and cost reduction, whereas multiple sampling improves detection sensitivity. Evidence suggests that ctDNA may persist for up to 4 weeks post-surgery, indicating that delayed sampling may enhance accuracy[81].

Despite sequencing advancements that have improved ctDNA assay sensitivity and specificity, challenges remain, particularly regarding the negative predictive value. The low concentration of ctDNA in circulation and its rapid degradation by liver macrophages and nucleases pose significant limitations[82]. To overcome this, Martin-Alonso et al[83] introduced two novel priming strategies using liposomal nanoparticles to reduce cfDNA degradation, increasing cfDNA levels up to tenfold when administered 1-2 hours before blood collection. These approaches may enhance ctDNA detectability, particularly in BTC, where cfDNA degradation poses a key challenge.

While sample size constraints persist, the correlations between cfDNA and recurrence-free survival warrant further investigation[76]. In BTC, tumor-informed (requiring prior tumor sequencing) and tumor-agnostic (profiling ctDNA without tumor reference) approaches impact test feasibility, but their comparative efficacy remains unclear. Further studies are needed to refine ctDNA-guided monitoring and enhance its clinical utility in BTC. Representative studies are outlined in Table 2.

Table 2 Studies evaluating circulating tumor DNA for minimal residual disease.

Ref.	Title	Clinical relevance
Alcaide et al[73], 2020	Evaluating the quantity, quality and size distribution of cell-free DNA by multiplex droplet digital PCR	Presents a novel droplet digital PCR assay to identify suboptimal samples and aberrant cfDNA size distributions, the latter typically associated with high ctDNA levels
King et al[76], 2023	Prospectivelongitudinal tumor-informed ctDNA in resectable biliary tract cancers	Assesses the utility of ctDNA levels in evaluating response in the absence of radiographically visible disease. ctDNA showed a higher detection rate than CA 19-9 prior to resection
Park et al[77], 2024	Ultrashort cell-free DNA fragments and vimentin-positive circulating tumor cells for predicting early recurrence in patients with biliary tract cancer	Investigates the effectiveness of cell-free DNA and circulating tumor cells in predicting early recurrence after curative surgery and adjuvant therapy in patients with BTC
Yoo et al[78], 2024	Circulating tumor DNA status and dynamics predict recurrence in patients with resected extrahepatic cholangiocarcinoma	Evaluates superiority of ctDNA over conventional biomarkers in predicting recurrence and informing adjuvant chemotherapy in resected extrahepatic cholangiocarcinoma
Yu et al[79], 2025	Detecting early recurrence with circulating tumor DNA in stage I-III biliary tract cancer after curative resection	Evaluates serial ctDNA testing for surveillance after curative resection in early-stage BTC. Identified recurrence in 93.8% of cases, with a median lead time of 3.7 months

BTC: Biliary tract cancer; cfDNA: Cell free DNA; ctDNA: Circulating tumor DNA.

CTDNA FOR MONITORING METASTATIC DISEASE

The RECIST guidelines are widely used to assess tumor progression and guide treatment decisions. Despite their widespread validation, they rely on imaging that detects only gross changes, often missing early tumor responses or progression and complicating interpretation due to pseudo-progression[84]. In this context, ctDNA analysis has emerged as a promising complementary tool, offering real-time and potentially more accurate insights into disease status.

Liquid biopsy provides a more comprehensive view of tumor heterogeneity than conventional tissue biopsies, enabling dynamic longitudinal monitoring through serial ctDNA assessments[85]. Notably, early ctDNA responses (as soon as day 21 of the first treatment cycle) can predict subsequent imaging responses, offering a critical lead time for assessing disease progression, rapidly identifying non-responders, and allowing timely treatment adjustments[85]. This early insight is particularly valuable for allowing ineffective treatments and their associated toxicities to be discontinued sooner[86].

In a phase 2 trial, ctDNA was monitored every three cycles in 5 patients with solid tumor on ICI therapy, correlating with tumor status and showing predictive value at baseline and post-treatment. Most patients who continued treatment beyond ctDNA progression experienced rapid disease progression, and ctDNA dynamics correlated with OS, independent of RECIST response[87]. Recent studies have highlighted ctDNA clearance as a potential marker for progression risk in patients on immunotherapy for ≥ 1 year, aiding personalized strategies—discontinuation for those with undetectable ctDNA or treatment escalation for those with detectable levels[88].

Data on ctDNA dynamics and resistance mutations in tyrosine kinase inhibitor-treated patients with BTC are limited. However, a retrospective analysis of the ClarIDHy trial in IDH1-mutant cholangiocarcinoma demonstrated that IDH1 mutation clearance in plasma ctDNA was correlated with improved disease control, highlighting its potential as a biomarker for monitoring treatment response[89]. Similar findings from studies on NSCLC and metastatic breast cancer reinforce the clinical utility of ctDNA in detecting resistance mutations (e.g., ERBB2, TP53, and phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit alpha) and predicting treatment responses, including responses to immune checkpoint inhibitors[90,91].

CTC measurements offer insights into tumor evolution under therapeutic selection and help identify resistance mechanisms (e.g., secondary mutations)[92]. CTC counts change before anatomical imaging, making them effective for early response assessment. Present in patients with or without primary tumors and prevalent in recurrent cases. Their presence reflects the tumor burden at all stages and aids in tracking immunophenotypic and genetic changes.

CTDNA FOR DETECTING MECHANISMS OF RESISTANCE

Clonal evolution and resistance mechanisms emerge from dynamic genomic changes that occur during treatment. Targeted panel NGS for ctDNA detection enables real-time monitoring, showing that over 60% of patients acquire new driver mutations as the disease progresses[93]. This approach also reveals tumor heterogeneity, which is a key factor in treatment resistance. Although ctDNA analysis may not fully capture rapid tumor evolution, serial sampling allows for the detection of emerging subclones. The increasing number of alterations over time reflects growing tumor complexity and resistance, as evidenced by a significantly higher mutation burden in later ctDNA samples than in the initial samples. In the BTC scenario, the higher frequency of mutated genes was TP53, KRAS, and FGFR2, which are commonly associated with cancer aggressiveness and resistance to anticancer therapy[23].

Despite the remarkable efficacy of targeted therapies in inhibiting key molecular pathways, cancer cells develop resistance through on-target, off-target, and phenotypic adaptations[94]. On-target alterations have been detected via ctDNA in liquid biopsies of patients treated with FGFR inhibitors, demonstrating their potential as biomarkers for monitoring resistance. Goyal et al[95] identified the acquired V564F mutation in progressing patients through serial cfDNA monitoring, showing strong concordance between tissue and ctDNA samples.

Notably, patients exhibited up to nine mutations, and autopsy data from 12 metastases confirmed that all retained the FGFR fusion but harbored distinct mutations. This suggests that resistance arises from convergent mutational events rather than FGFR fusion-negative clone selection. In FGFR2 fusion-positive BTC, FGFR kinase domain mutations are known to cause acquired resistance to FGFR inhibitors[95]. TAS-120, an irreversible pan-FGFR inhibitor, was effective in four FGFR2 fusion-positive CCA patients who had developed resistance to prior FGFR inhibitors, selected through serial biopsies, ctDNA analysis, and patient-derived tumor models[96].

Methylated homeobox A9 (meth-HOXA9), a tumor-specific prognostic and predictive biomarker, was evaluated by Andersen et al[97] to assess its clinical impact in patients with advanced BTC receiving last-line therapy. The study found that increased meth-HOXA9 Levels were negatively associated with survival, suggesting its potential role in guiding early discontinuation of ineffective treatment.

Further supporting the utility of liquid biopsy, initial studies provided proof of concept by associating high pretreatment variant allele frequency (VAF) in ctDNA with poor prognosis and reduced response duration to systemic therapies[23]. Another trial showed that high pretreatment VAF in cfDNA has predictive value in advanced BTCs, with a shorter mPFS (2.6 months vs 7.6 months) and OS (7.7 months vs 19.9 months) compared to those with low VAF (HR = 2.1, 95%CI: 1.1–4.0; P = 0.030)[98]. Previous studies have highlighted the role of miRNAs in chemoresistance[99]). Meng et al[100] demonstrated their involvement in gemcitabine sensitivity, showing that miRNA inhibition significantly increased cytotoxicity and apoptotic effects in CCA cell lines while also impacting cell proliferation and differentiation. A high-throughput screening of 997 Locked nucleic acid miRNA inhibitors in six CCA cell lines treated with cisplatin and gemcitabine identified miR-1249 inhibition as a key factor in enhancing chemotherapy sensitivity across all tested cells[101]. miR-1249 was found to be upregulated in CD133+ cells from BTC stem cell niches and chemoresistant CCLP cells. Its knockout impairs CD133+ subclone expansion, reduces cancer stem cell marker expression, and increases chemosensitivity[102]. Furthermore, miR-1249 overexpression was detected in 41% of human BTC cases, suggesting its potential as both a biomarker of chemoresistance and therapeutic target.

The development of resistance mechanisms hampers accurate treatment response assessment and may misguide therapeutic decision-making. Effectively tracking clonal evolution and resistance dynamics through ctDNA analysis remains a challenge, underscoring the need for continued advancements to enhance clinical reliability.

FUTURE PESPECTIVES

The role of ctDNA in precision oncology continues to expand. However, several challenges must be addressed for their full integration into clinical practice. One key question is whether performing both blood and tumor NGS at baseline provides superior insight for treatment decision-making and patient selection in clinical trials compared with using either modality alone. Given the intratumor heterogeneity observed in BTC, comprehensive genomic characterization through ctDNA analysis remains critical. Comparative analyses of plasma and biliary ctDNA are warranted to clarify the best source for analysis and their better suitability throughout the course of the disease. Two independent studies have leveraged ctDNA-based NGS and machine learning to define the broader genomic and biological landscape of non-CRC gastrointestinal cancers, highlighting their potential for refining molecular profiling strategies[103].

Despite these advances, technological and biological challenges continue to affect the efficacy and applicability of ctDNA analysis. Issues such as sample quality, low ctDNA shedding in certain tumor types, and the inherent stochastic nature of ctDNA release must be addressed[103]. Future research should prioritize the development of ctDNA enrichment strategies and explore their integration with multi-omics platforms, including proteomics and metabolomics, for increased sensitivity and specificity.

Another emerging area of interest is ctDNA methylation, which has shown promise as an additional biomarker. Studies suggest that changes in methylation frequency correlate with disease progression and therapeutic response, offering a potential tool for refining response criteria in clinical trials[104]. However, establishing standardized thresholds, such as the maximum VAF and percent change in VAF, remains crucial. Additionally, the timing of ctDNA assessments must be carefully considered, given the varying ctDNA clearance rates and drug pharmacokinetics, which directly impact the treatment duration and response evaluation. Studies should aim to define these optimal time points for ctDNA sampling across BTC subtypes and disease stages.

A critical challenge in ctDNA implementation is the lack of standardized protocols and harmonization across platforms. Variability in sample processing, sequencing technologies, and bioinformatics pipelines contribute to inconsistencies in the results, raising concerns about reproducibility[105]. Establishing universal quality control measures and standardizing analytical workflows are essential for integrating ctDNA analysis into routine clinical practice. Standardization across commercial testing platforms is also lacking, leading to variability in sensitivity, mutation calling, and report formats, which limit cross-platform comparability and mutual recognition[106-113] (Table 3).

Table 3 Studies evaluating circulating tumor DNA in advanced settings.

Ref.	Title	Clinical relevance
Zhang et al[108], 2020	Prognostic and predictive impact of circulating tumor DNA in patients with advanced cancers treated with immune checkpoint blockade	Analyzed ctDNA from multiple tumor types in durvalumab trials, showing that higher pretreatment VAF is associated with worse overall survival, supporting its role as a prognostic not predictive biomarker
Gouda et al[85], 2022.	Longitudinal monitoring of circulating tumor DNA to predict treatment outcomes in advanced cancers	Evaluation of longitudinal ctDNA to predict early dynamic changes with advanced solid tumors
Bratman et al[87], 2020.	Personalized circulating tumor DNA analysis as a predictive biomarker in solid tumor patients treated with pembrolizumab	Examines ctDNA in advanced solid tumor patients treated with pembrolizumab, showing that baseline ctDNA and its kinetics during treatment correlate with clinical outcomes
Aguado et al[89], 2020	IDH1 mutation detection in plasma ctDNA and association with clinical response in patients with advanced intrahepatic cholangiocarcinoma from the phase 3 ClarIDHy study	Extends analysis from the ClarIDHy trial, demonstrating 92% concordance between plasma and tissue for mIDH1-R132 detection in iCCA, supporting its use in liquid biopsy for patient selection when tumor tissue is limited
Andersen et al[97], 2022	The clinical impact of methylated homeobox A9 ctDNA in patients with non-resectable biliary tract cancer treated with erlotinib and bevacizumab	Demonstrates that rising plasma meth-HOXA9 Levels after one treatment cycle are significantly associated with worse survival in late-stage BTC
Uson Junior et al[98], 2022	Cell-free tumor DNA dominant clone allele frequency is associated with poor outcomes in advanced biliary cancers treated with platinum-based chemotherapy	Demonstrates that higher dominant clone allele frequency in pretreatment ctDNA is associated with significantly worse progression-free and overall survival in metastatic BTC patients receiving platinum-based chemotherapy
Meng et al[100], 2006	Involvement of human micro-RNA in growth and response to chemotherapy in human cholangiocarcinoma cell lines	Identifies dysregulated miRNAs in cholangiocarcinoma and shows that targeting miR-21, miR-141, and miR-200b alters tumor growth and gemcitabine sensitivity
Berchuck et al[23], 2022	The clinical landscape of cell-free DNA alterations in 1671 patients with advanced biliary tract cancer	Demonstrates that ctDNA sequencing in cholangiocarcinoma shows high concordance with tumor tissue, reflects tumor burden, and can track mutational evolution during chemotherapy
Ettrich et al[27], 2019	Genotyping of circulating tumor DNA in cholangiocarcinoma reveals diagnostic and prognostic information	Demonstrates that ctDNA sequencing in CCA enables noninvasive monitoring of tumor mutations, with high concordance to tissue, correlation with tumor burden and PFS, and dynamic changes during chemotherapy
Okamura et al[109], 2021.	Comprehensive genomic landscape and precision therapeutic approach in biliary tract cancers	Shows that genomic profiling via ctDNA and/or tissue-DNA is feasible in BTC, with higher concordance between ctDNA and metastatic tissue, and that matched targeted therapies based on profiling improve PFS and disease control rates
Weinberg et al[52], 2019	Molecular profiling of biliary cancers reveals distinct molecular alterations and potential therapeutic targets	Comprehensively profiles BTCs, revealing distinct molecular alterations by subtype and supporting the use of site-specific molecular profiling to guide therapy and clinical trial design
Nakamura et al[110], 2015	Genomic spectra of biliary tract cancer	Molecularly characterizes BTCs, identifying subtype-specific genomic alterations and mutational signatures, with nearly 40% harboring actionable targets and a hypermutated subgroup potentially responsive to immunotherapy
Jusakul et al[51], 2017	Whole-genome and epigenomic landscapes of etiologically distinct subtypes of cholangiocarcinoma	Performs integrative genomic and epigenomic analysis of CCAs, revealing distinct molecular subtypes
Farshidfar et al[55], 2017	Integrative genomic analysis of cholangiocarcinoma identifies distinct IDH-mutant molecular profiles	Integrative TCGA analysis of predominantly intrahepatic CCA identifies an IDH-mutant subtype with distinct epigenetic and metabolic features, revealing molecular heterogeneity with potential therapeutic relevance
Thongyoo et al[64], 2025	KRAS mutations in cholangiocarcinoma: Prevalence, prognostic value, and KRAS G12/G13 detection in cell-free DNA	Assesses KRAS G12/G13 mutations in cfDNA as non-invasive prognostic biomarkers in CCA; elevated MAF combined with CA19-9 Levels predicts poorer survival
Leone et al[111], 2006	Somatic mutations of epidermal growth factor receptor in bile duct and gallbladder carcinoma	Identifies somatic EGFR mutations and pathway activation in biliary tract cancers
Tannapfel et al[68], 2003	Mutations of the BRAF gene in cholangiocarcinoma but not in hepatocellular carcinoma	Identifies BRAF and KRAS mutations in cholangiocarcinoma; supports ctDNA-based profiling of MAPK pathway alterations to stratify patients for targeted therapies
Toyota et al[101], 2015	Mechanism of gemcitabine-induced suppression of human cholangiocellular carcinoma cell growth	Exploration of microRNAs and angiogenic molecules as biomarkers of sensitivity or resistance to gemcitabine in CCC; potential targets for overcoming chemoresistance
Carotenuto et al[102], 2020	Modulation of biliary cancer chemo‐resistance through microRNA‐mediated rewiring of the expansion of CD133+ cells	Targeting microRNA 1249 to overcome chemoresistance to gemcitabine–cisplatin in BTC
Gerlinger et al[93], 2012.	Intratumor heterogeneity and branched evolution revealed by multiregion sequencing	Identification of Highlights the limitations of single-site, reinforcing ctDNA as a non-invasive method to capture spatial and temporal tumor heterogeneity for molecular profiling and treatment monitoring
Deshpande et al[112], 2011	Mutational profiling reveals PIK3CA mutations in gallbladder carcinoma	Highlights ctDNA for subtype differentiation and therapeutic decision-making by identification of subtype-specific mutations
Kawakami et al[66], 2021	Stepwise correlation of TP53 mutations from pancreaticobiliary maljunction to gallbladder carcinoma: A retrospective study	Supports the use of TP53 mutations as a biomarker for early detection and risk stratification in GBC patients, potentially detectable via ctDNA
Kim et al[67], 2015	Molecular subgroup analysis of clinical outcomes in a phase 3 study of gemcitabine and oxaliplatin with or without erlotinib in advanced biliary tract cancer	Highlights the potential utility of ctDNA for assessing KRAS and PIK3CA mutational status as predictive biomarkers to guide anti-EGFR therapy in BTC
Saetta et al[113], 2004	Mutational analysis of BRAF in gallbladder carcinomas in association with KRAS and p53 mutations and microsatellite instability	Supports the inclusion of BRAF exon 15 (V600-equivalent codon 599) in ctDNA panels for gallbladder cancer, enabling detection of mutually exclusive RAS/RAF alterations with potential diagnostic and therapeutic relevance

BTC: Biliary tract cancer; CCA: Cholangiocarcinoma; CCC: Cholangiocellular carcinoma; ctDNA: Circulating tumor DNA; EGFR: Epidermal growth factor receptor; GBC: Gallbladder carcinoma; HOXA9; Homeobox A9; iCCA: Intrahepatic cholangiocarcinoma; MAF: Mutant allele frequency; MAPK: Mitogen-activated protein kinase; mIDH1; Mutated form of the isocitrate dehydrogenase 1 enzyme; PIK3CA; Phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha; PFS: Progression-free survival; TCGA: The Cancer Genome Atlas; TP53: Tumor protein p53; VAF: Variant allele frequency.

In addition to analytical performance, several practical barriers continue to hinder the widespread clinical adoption of ctDNA in BTC. Turnaround time remains a critical issue, as delays in resulting delivery may compromise timely therapeutic decisions in aggressive disease settings. Moreover, interpreting ctDNA results requires specialized training, and many clinicians may lack confidence or familiarity with integrating molecular data into routine care.

Additionally, cost and accessibility are major barriers. High-throughput sequencing technologies and complex bioinformatics tools can be expensive, limiting their adoption, particularly in resource-constrained settings. Reimbursement policies further complicate access, with inconsistent insurance coverage and high out-of-pocket costs acting as deterrents to access. To maximize the impact of ctDNA in oncology, efforts must focus on developing cost-effective platforms and improving accessibility to ensure its widespread clinical utility.

CONCLUSION

ctDNA is a valuable non-invasive biomarker in BTC, playing a crucial role in molecular profiling, prognosis, personalized treatment selection, and real-time disease monitoring. Its dynamic nature facilitates treatment decisions and improves patient management by identifying actionable genomic alterations relevant to the targeted therapies. Despite current limitations, such as small sample sizes, standardization issues, and cost-effectiveness challenges, ongoing research and large-scale prospective trials are expected to further validate the clinical utility of ctDNA and enhance its integration into precision oncology for BTC.