Cholangiocarcinoma: The era of liquid biopsy

Abstract

Cholangiocarcinoma (CCA) is a highly aggressive and heterogeneous malignancy arising from the epithelial cells of the biliary tract. The limitations of the current methods in the diagnosis of CCA highlight the urgent need for new, accurate tools for early cancer detection, better prognostication and patient monitoring. Liquid biopsy (LB) is a modern and non-invasive technique comprising a diverse group of methodologies aiming to detect tumour biomarkers from body fluids. These biomarkers include circulating tumour cells, cell-free DNA, circulating tumour DNA, RNA and extracellular vesicles. The aim of this review is to explore the current and potential future applications of LB in CCA management, with a focus on diagnosis, prognostication and monitoring. We examine both its significant potential and the inevitable limitations associated with this technology. We conclude that LB holds considerable promise, but further research is necessary to fully integrate it into precision oncology for CCA.

Key Words: Biliary tract cancer; Cholangiocarcinoma; Circulating tumour cells; Cell free DNA; Circulating tumour DNA; Circulating RNA; Biomarkers; Extracellular vesicles; Precision medicine

Core Tip: Cholangiocarcinoma (CCA) is an aggressive and heterogeneous malignancy that presents significant challenges in early detection, prognostication and monitoring using current diagnostic methods. Liquid biopsy (LB) emerges as a promising non-invasive technique, leveraging circulating tumour cells, cell-free DNA, circulating tumour DNA, RNA and extracellular vesicles as biomarkers. This review highlights the potential of LB to transform CCA management by improving early diagnosis, providing better prognostic insights and enabling dynamic patient monitoring. While LB shows substantial promise, critical limitations remain, necessitating further research to validate and optimise its integration into precision oncology for CCA.

INTRODUCTION

Cholangiocarcinoma (CCA) is a highly aggressive and heterogeneous malignancy arising from the epithelial cells of the biliary tract[1]. It represents the second most common primary liver malignancy after hepatocellular carcinoma (HCC), accounting for approximately 15% of all primary liver tumours and 3% of all gastrointestinal cancers[2]. CCA is anatomically classified into intrahepatic (iCCA), perihilar (pCCA), and distal (dCCA) subtypes, each presenting unique clinical, anatomical and molecular characteristics[3,4].

The incidence and mortality rates of CCA vary significantly across different regions worldwide, likely reflecting variations in environmental risk factors and genetic predispositions specific to each geographic area[5,6]. Several well-established risk factors are associated with CCA. Interestingly, some of these factors are shared across all three subtypes while others are subtype-specific. Nevertheless, up to 50% of CCA cases in the Western countries are diagnosed without any identifiable risk factor[5].

The pathogenesis of CCA is complex and multifactorial, involving a combination of genetic, environmental and inflammatory factors. Chronic inflammation, as found in conditions like primary sclerosing cholangitis (PSC) and parasitic infections, can lead to dysplasia and eventual malignant transformation[7,8]. Cholestasis, commonly found in chronic biliary obstruction, can also contribute to the pathogenesis of CCA. The prolonged exposure to toxic bile acids triggers oxidative stress, DNA damage and activation of pro-carcinogenic signalling pathways, such as mitogen-activated protein kinase, a pathway implicated in CCA carcinogenesis[9,10]. Recently, microbial factors, including gut microbiota, have been associated with the development of CCA[11].

CCA is characterised by molecular heterogeneity, presenting different genetic alterations between its subtypes. iCCA is commonly associated with mutations in the isocitrate dehydrogenase-1 (IDH1)/isocitrate dehydrogenase-2 (IDH2) genes and fibroblast growth factor receptor 2 (FGFR2) fusions. IDH1/IDH2 mutations, found in approximately 10%-22% of iCCA cases, alter cellular metabolism and are associated with less aggressive tumour behaviour[12]. FGFR2 gene fusions are present in approximately 10%-20% of iCCA cases, drive tumour growth and can be targeted by FGFR inhibitors[13]. KRAS mutations, TP53 and Erb-B2 receptor tyrosine kinase 2 amplification and overexpression are more common in extrahepatic CCA and are often associated with more aggressive disease and poorer outcomes[14,15]. Furthermore, distinct immune cell phenotypes have been associated with the risk of CCA development[16].

The updated 2023 practice guidance on PSC and CCA by the American Association for the Study of Liver Diseases effectively distinguished between de novo CCA and PSC-associated CCA, acknowledging that, due to their unique characteristics, they should be considered as distinct entities[17,18]. De novo CCA is typically diagnosed in older adults, usually during their seventh decade of life, and is more often encountered in Southeast Asia, due to parasitic infections, as previously mentioned. On the contrary, PSC-associated CCA often occurs in younger patients, typically in their fourth decade and is more common in Western countries where PSC has a higher prevalence, although the incidence of CCA appears to be increasing with advancing age at PSC diagnosis[1,19]. The lifetime incidence of CCA in PSC patients is estimated to be up to 20%, representing a 400-fold increased lifetime risk compared to the general population, with the majority of cases diagnosed within the first 24 months of PSC diagnosis[1,17,20]. De novo CCA tumours are characterised by heterogeneity, with less predictable patterns of genetic mutations, compared to PSC-associated CCA, where tumours often present distinct molecular characteristics, driven by chronic inflammation leading to oncogenesis[20].

Liquid biopsy (LB) is a modern and non-invasive technique comprising a diverse group of methodologies aiming to detect tumour biomarkers from body fluids such as blood, plasma, urine, bile and others[21]. Biomarkers detected using LB techniques include circulating tumour cells (CTCs), cell-free DNA (cfDNA), circulating tumour DNA (ctDNA), RNA and extracellular vesicles (EVs)[22].

CTCs are intact tumour cells that have detached from the primary tumour or metastatic sites and circulate in the bloodstream. They typically survive 1-2.5 hours in the bloodstream before being destroyed by the immune system[22]. The concentration of CTCs in blood is exceptionally low, typically ranging from 1 to 100 cells per millilitre. This makes the counting process challenging, as it generally requires enrichment of the target cells followed by their detection, isolation and release. This process is often time-consuming and costly[22,23].

cfDNA refers to DNA fragments, typically ranging in size from 140 to 170 base pairs, released from cells into the bloodstream[24]. It includes DNA from normal cells as well as from tumour cells in cancer patients, usually released through processes like apoptosis and necrosis. ctDNA is the subset of cfDNA that is of tumour origin[25]. ctDNA contains somatic variants, including tumour-specific mutations and epigenetic changes, such as DNA methylation patterns[22].

RNA molecules, such as microRNA (miRNA), messenger RNA (mRNA) and long non-coding RNA (lncRNA), can also circulate in the bloodstream, carrying genetic information regarding gene expression and regulation in tumours cells[26]. RNA in the bloodstream is less stable than cfDNA, with a half-life of only about 15 seconds. However, its stability is significantly improved when it is associated with proteins or encapsulated within EVs[27]. Specifically, miRNAs have been shown to play a critical role in cancer progression by regulating gene expression and impacting processes such as apoptosis, angiogenesis, metastasis and drug resistance. Depending on their function, miRNAs are categorised as oncogenic, tumour-suppressor or metastatic, with their dysregulation contributing to tumour development, progression and invasion through mechanisms including genetic, epigenetic and biogenesis alterations[28].

Finally, EVs, such as exosomes and microvesicles, are membrane-bound particles released by both normal and tumour cells, mediating communication between different cells types, including cancer cells and the surrounding environment, potentially influencing tumour invasion and metastasis[26]. EVs are small structures, approximately 100 nm in size, that can be found in nearly all body fluids. They contain a variety of biomolecules, including proteins, DNA and RNA, and their stability is preserved by a protective outer lipid membrane[22,27].

In this comprehensive review, we explore the role of LB in the overall management of CCA (Figure 1).

Figure 1 The spectrum of liquid biopsy in cholangiocarcinoma: Types, biomarkers and applications. PCR: Polymerase chain reaction; NGS: Next-generation sequencing; CTCs: Circulating tumour cells; cfDNA: Cell-free DNA; ctDNA: Circulating tumour DNA; EVs: Extracellular vesicles; IDH: Isocitrate dehydrogenase; FGFR2: Fibroblast growth factor receptor 2.

Tissue sampling in CCA and biomarkers

Tissue sampling is an essential, though sometimes challenging process in the diagnosis and treatment planning of CCA, even though not always mandatory for patients eligible for resection or liver transplantation[29]. Several approaches are used to collect tissue samples, depending on the tumour’s anatomic location.

Endoscopic retrograde cholangiopancreatography (ERCP) is an invasive method commonly used for both diagnosis and symptom management in patients with pCCA and dCCA[7]. ERCP can be combined with conventional biliary brush cytology, fluorescence in situ hybridisation (FISH) and intraductal biopsies to confirm the presence of CCA[17,18]. Alternatively, percutaneous transhepatic cholangiography (PTC) is another invasive technique typically employed when ERCP is either not feasible or unsuccessful[17]. Endoscopic ultrasonography (EUS) is useful in the diagnosis and staging of pCCA and dCCA, since it provides a detailed assessment of the extrahepatic biliary tree and enables tissue acquisition via fine-needle aspiration (FNA)[8]. For iCCA, a computed tomography-guided or ultrasound-guided liver biopsy may be used to sample the liver mass directly[30]. A more recent advanced endoscopic technique, cholangioscopy using the SpyGlass SpyScope, enables direct endoscopic visualisation of the bile ducts and targeted biopsies of suspicious lesions[31].

As a final approach, if less invasive methods do not provide enough tissue to obtain a diagnosis, a laparoscopy or open surgery may be performed to directly visualise and sample the tumour[32].

Biomarkers play an increasingly important role in the diagnosis of CCA. Carbohydrate antigen 19-9 (CA19-9) is the most widely used serum biomarker for diagnosing CCA. Different cut-off values have been proposed for the diagnosis of de novo and PSC-associated CCA. For patients without PSC, the cut-off value is typically set at 100 U/mL, while for those with PSC, it is set at 129 U/mL, with a reported sensitivity of 79%, a specificity of 99% but a positive predictive value of only 57%[20]. CA 19-9 levels of 1000 U/mL or greater have been associated with unresectable disease[33].

Carcinoembryonic antigen (CEA) is another tumour marker, that can be elevated in CCA but is not specific to this cancer, as it is also elevated in other gastrointestinal cancers and benign diseases. CEA is often used in combination with CA19-9 to improve diagnostic accuracy[34]. CEA demonstrates lower diagnostic performance compared to CA19-9 for CCA, with levels exceeding 5 ng/mL showing a sensitivity of 33%-68% and a specificity of 82%-89%[20].

Diagnostic biomarkers (CA19-9, CEA) can assist in detecting CCA and sometimes help differentiate it from benign conditions, but their performance remains suboptimal due to important limitations, with the lack of specificity being an essential one. CA19-9 can be elevated in some non-malignant conditions, such as cholangitis, pancreatitis and biliary obstruction[35]. Moreover, these biomarkers are not sensitive enough to detect CCA in its early stages, which significantly affects prognosis. On the contrary, high levels of CA19-9 are associated with advanced, inoperable CCA[36]. As mentioned earlier, CCAs are highly heterogenic tumours, resulting in different biomarker profiles between CCA subtypes, but also between patients. In addition, approximately 7% of the population are Lewis antigen-negative and these individuals are unable to express CA19-9 completely[14].

The enhanced liver fibrosis (ELF) score is a non-invasive diagnostic tool used to assess liver fibrosis in PSC patients. It is derived from a combination of three serum biomarkers associated with extracellular matrix turnover: Hyaluronic acid, propeptide of type III procollagen and tissue inhibitor of metalloproteinases-1[37,38]. Recent studies have shown that the ELF score is significantly elevated in PSC patients diagnosed with CCA compared to PSC patients without CCA (P < 0.001)[37]. Additionally, a retrospective study comparing PSC patients with and without CCA, as well as patients with CCA in the absence of PSC, demonstrated that the ELF score was significantly higher in patients with CCA, regardless of PSC status, compared to those with PSC alone (P < 0.001)[39]. Among PSC patients, an ELF score ≥ 9.8 was identified as an independent predictor of CCA[39]. The utility of ELF in the early detection of CCA in patients with PSC requires further validation.

Current diagnostic techniques and their limitations

Although tissue sampling and serum biomarkers, combined with imaging studies, are considered the gold standard for diagnosing CCA, they have several limitations.

The anatomic location of CCA, especially for tumours located at the hilum or deeper in the liver, makes assessment challenging with methods such as conventional ERCP and EUS[40]. Additionally, these tumours are characterised by a highly desmoplastic nature and significant intratumoural heterogeneity, which makes adequate tissue sampling difficult. Due to the fibrotic nature of the tumour and the challenges in accessing deep areas, false negatives can occur, meaning cancer cells might not be present in the sample despite the underlying existence of CCA[41].

Cytology, although being a valuable tool in cancer diagnosis, has several limitations that can impact its accuracy and reliability. While cytology is a highly specific tool for diagnosing CCA (reaching 100%), its sensitivity is limited (5%-20%)[35]. To address the low sensitivity of conventional cytology, FISH is commonly used. FISH analysis detects chromosome abnormalities, such as amplifications or aneusomy (a marker for aneuploidy), that are typically present in tumour cells[33]. FISH analysis, when added to conventional cytology can increase sensitivity values from 15% to 38%-58%[1]. Nevertheless, the addition of FISH analysis in PSC patients reduces specificity to 70%, making it a suboptimal screening tool for this population[42,43].

Additionally, the invasive procedures, such as ERCP, EUS-FNA and PTC carry risks of complications (e.g., bile leakage, infection, bleeding) and may require multiple attempts. Finally, FNA may carry a risk of tumour seeding[44].

The limitations of imaging studies, tissue sampling techniques, cytology and available biomarkers in the diagnosis of CCA highlight the urgent need for new, reliable tools for early cancer detection, better prognostication and patient monitoring.

LB in CCA

Types: LB has evolved as an emerging tool in CCA diagnosis, using two types of body fluids: Blood/plasma and bile. As expected, blood-based biopsies have been more extensively studied[21]. Blood, as a fluid, offers several advantages, including easy sample collection, simple procedure, less pain and fewer risks compared to the traditional biopsies. Additionally, blood allows for serial measurements and multiple sampling, enabling intrasample comparison and providing real-time information on tumour dynamics and evolution[45]. Blood-based LBs are thought to offer a more comprehensive view of the overall tumour burden, as they capture information from a wider range of tumour sites compared to the localised sampling of a single tissue biopsy[46]. On the other hand, blood exhibits lower sensitivity in detecting small amount of tumour DNA and may miss localised tumours that do not shed DNA into the bloodstream[47].

In a large study by Berchuck et al[48], 2068 serum cfDNA samples from 1671 patients with advanced CCA were collected and analysed. The study reported that genetic alterations in cfDNA were detected in 84% of patients. Concordance between cfDNA and tissue-based mutation detection was high for certain mutations, such as IDH1 (87%) and BRAF (100%), but notably lower for others, such as FGFR2 fusions (18%). The authors concluded that their findings strongly support the use of genomic profiling of cfDNA to guide clinical care for patients with advanced CCA[48].

Bile-based LBs have more recently emerged as an alternative to address the limitations of blood-based LBs. Bile, being in direct contact with the biliary system, is thought to provide more detailed information regarding tumour heterogeneity and exhibits higher specificity for detecting local cancers compared to blood biopsies[21]. Nevertheless, bile sampling requires an invasive procedure, typically ERCP or PTC, both of which carry risks of complications.

Gou et al[47], explored the potential role of bile as a LB source and compared it to plasma. They studied the somatic mutations found in tissue DNA, bile-cfDNA and plasma-cfDNA of 28 patients with CCA. They demonstrated that the overall concordance for mutations in bile was significantly higher than that in plasma (99.1% vs 78.3%, P < 0.0001), concluding that bile cfDNA was superior to plasma cfDNA in detecting tumour-related genomic alterations[47].

Driescher et al[49], also investigated bile as a potential LB source in pancreatobiliary cancers and compared it to plasma. They analysed 37 blood and 21 bile samples collected from CCA and pancreatic ductal adenocarcinoma patients and 16 blood and 21 bile samples from patients with non-malignant biliary obstructions. They found that, in the cancer group, 96.2% of pathogenic mutations identified by tissue sequencing could be detected in bile-cfDNA resulting in a sensitivity and specificity of 96.2% and 100%, respectively. In contrast, plasma-cfDNA had a detection rate of 31.6%. Additionally, when they compared KRAS allele frequency between bile and plasma, they observed that bile exhibited significantly higher allele frequencies than plasma in patients with localised disease (P = 0.0159)[49].

Methods: LB detects tumour biomarkers using a variety of methods. These methods can be divided into two categories, depending on the technology used: Polymerase chain reaction (PCR)-based approaches and next-generation sequencing (NGS)[50].

PCR-based methods focus on specific, known mutations, making them highly efficient for targeted analysis. These methods include quantitative real-time PCR (qRT-PCR), digital PCR (dPCR) and the mass-spectrometry-based method. They are relatively low cost and fast, making them ideal for clinical settings requiring rapid and specific mutation detection[51]. PCR-based methods can also demonstrate high sensitivity; quantitative PCR (qPCR) detects mutant allele fractions (MAF) that are greater than 10%, while newer methods such as dPCR and its variant beaming (beads, emulsion, amplification, and magnetics) can detect MAFs as low as 0.02%[51].

NGS is a high-throughput sequencing technology that provides comprehensive, large-scale information without being limited to predefined targets[52]. NGS can be broadly categorised into targeted and untargeted approaches. Targeted sequencing focuses on specific regions of the genome, such as known cancer-related genes and it is used for a specific panel of genes. Untargeted sequencing takes a broader approach, examining the entire genome [whole genome sequencing (WGS)], exome (whole exome sequencing) or transcriptome (RNA sequencing)[53]. This technology exhibits high sensitivity and specificity, detecting MAFs of less than 1%, reaching 0.1% when deep sequencing is applied[51]. It also requires small amounts of DNA and is suitable for whole-genome analysis in a single assay[54]. Nevertheless, these methods are more expensive and time-consuming due to their broad scope and high data output. Importantly, while NGS is accessible in high-income countries due to advanced healthcare infrastructure and reimbursement policies, its accessibility in resource-constrained settings remains a significant challenge due to high costs, lack of specialised expertise and limited diagnostic facilities[55].

Several studies have evaluated the cost-effectiveness of different molecular diagnostic approaches. In a Dutch study, Simons et al[56] explored the early cost-effectiveness of WGS for diagnosing advanced non-small-cell lung cancer compared to current clinical practices, which included FISH, immunohistochemistry and targeted NGS. They found WGS to be cost-effective only if priced at 2000 Euro per patient, identifying 2.7% more actionable patients than current practice[56]. Studies in the United States and Canada showed that NGS enhances targeted treatment selection in lung cancer patients, increasing life-years and productivity, with a minimal budget impact[57,58]. Conversely, a Brazilian study concluded that NGS for advanced lung adenocarcinoma while being precise was not cost-effective compared to PCR and FISH, in terms of quality-adjusted life years[59].

The available data on the cost-effectiveness of PCR and NGS in CCA are limited. The European Society for Medical Oncology Precision Medicine Working Group recently published recommendations for the use of NGS in patients with advanced cancer, taking into account cost-effectiveness and accessibility. Regarding CCA, the group concluded that, based on the number of alterations classified as level I, tumour NGS is recommended for patients with advanced CCA[60].

Liquid vs tissue biopsy in CCA

Despite the many advantages of LB, tissue biopsy remains the gold standard technique for diagnosis at present (Table 1). Nevertheless, numerous studies have evaluated the correlation between tumour tissue and LB genomic landscapes.

Table 1 Tissue biopsy remains the gold standard technique for diagnosis at present.

Liquid biopsy	Tissue biopsy
Advantages
Minimally invasive-easy sampling	Histological evaluation
Serial analysis-tumour evolution monitoring	Clinically validated and standardised
Ability to assess drug response-drug resistance	Available in most hospitals
Ability to use for postoperative follow-up/surveillance	Provides a direct sample of tumour tissue, ensuring accurate pathological analysis
Fast	Facilitates detailed genetic and molecular profiling of the sampled tissue
Comprehensive tumour profiling
Applicable to inaccessible tumours
Reveals tumour heterogeneity
Lower cost of sample isolations
Disadvantages
Not clinically validated	Invasive procedure
No histological evaluation	Not applicable in inaccessible tumours
Low availability	Localised tissue sampling
Low sensitivity (may miss low-abundance mutations or those not shed into the bloodstream	No capacity for tumour evolution monitoring, drug response-drug resistance monitoring, tumour heterogeneity assessment
Technology-dependent	Time intensive procedure
High costs (in case of NGS and WGS)	Higher cost of sample isolations

NGS: Next generation sequencing; WGS: Whole genome sequencing.

Han et al[61] attempted to determine whether ctDNA could replace tissue biopsy by studying the correlation among bile, plasma ctDNA and formalin-fixed paraffin-embedded (FFPE) tissue samples of CCA patients. For KRAS somatic mutation detection, they reported 80% concordance between paired bile ctDNA and FFPE samples, and 42.9% concordance between plasma ctDNA and FFPE samples. Additionally, transcriptomic sequencing of paired bile and FFPE samples revealed a strong positive correlation (r = 0.991, P < 0.001) in the expression of KRAS-associated oncogenes. These findings suggest that bile-based LB could serve as an alternative to tissue biopsy[61].

In another study, researchers analysed multiple gene mutations in tumour tissue and cfDNA from patients with pancreatobiliary carcinomas. Of the mutations detected by tissue biopsy, 90.3% [95% confidence interval (CI): 73.1%-97.5%] were also detected in cfDNA, showing a strong correlation (r = 0.93) and providing diagnostic accuracy of 97.7% for cfDNA[62]. Similarly, Chen et al[63] detected genomic alterations in blood-derived ctDNA using NGS and compared them to tissue genomic alterations. The results showed that the frequencies of single nucleotide variations in ctDNA were similar to those detected in tissue samples: 35.1% vs 40.4% for TP53 mutations and 20.1% vs 22.6% for KRAS mutations, suggesting that LB could be a viable alternative to tissue biopsy[63].

However, large-scale studies comparing tissue and LB gene panels are needed to validate these findings.

APPLICATIONS OF LB IN CCA

Figure 2 summarises the key findings of major studies (based on their design, methodology and sample size) undertaken so far on LB in CCA.

Figure 2 Key findings of major studies on liquid biopsy in cholangiocarcinoma. CCA: Cholangiocarcinoma; CDO1: Cysteine dioxygenase type 1; CNRIP1: Cannabinoid receptor interacting protein 1; SEPT9: Septin 9; VIM: Vimentin; AUC: Area under the curve; Sens: Sensitivity; Sp: Specificity; PSC: Primary sclerosing cholangitis; NGS: Next generation sequencing; CTC: Circulating tumour cells; cfDNA: Cell-free DNA; CI: Confidence interval; HR: Hazard ratio; ctDNA: Circulating tumour DNA; DFS: Disease-free survival.

Diagnosis and molecular profiling

As mentioned, CCA diagnosis can be challenging due to several reasons, including the late presentation of symptoms, the anatomical location of tumours and the difficulty in differentiating CCA from benign conditions. LB offers a non-invasive approach to early detection, screening and differential diagnosis.

Early diagnosis: Wintachai et al[64] studied the plasma cfDNA levels of 62 patients diagnosed with CCA and compared them to cfDNA levels of 33 patients with benign biliary disease (BBD), including chronic cholecystitis, simple biliary cyst, and biliary cystadenoma, and 30 healthy controls. They found that the levels of cfDNA in CCA patients were significantly higher than in both normal controls (P < 0.0001) and BBD patients (P = 0.006). Additionally, in terms of diagnostic efficacy, higher cfDNA significantly differentiated CCA from healthy controls [88.71% sensitivity and 96.67% specificity, area under the curve (AUC) = 0.9715, P < 0.0001] and BBD (82.26% sensitivity and 57.58% specificity, AUC = 0.7229, P = 0.0004) with cut-off values of 0.2175 ng/μL, and 0.3388 ng/μL, respectively. Moreover, cfDNA outperformed the current commonly used serum tumour markers in distinguishing CCA from BBD (CEA: AUC = 0.5063 at the cut-off of 2.53 ng/mL; CA19-9: AUC = 0.5922, at the cut-off value of 39.90 U/mL). Based on these findings, the authors concluded that cfDNA could serve as a viable biomarker for diagnosing CCA[64]. Key limitations of the study include the small sample size and the fact that all patients were from a single institution.

In another study, plasma miRNAs levels of 25 patients with iCCA and 7 healthy controls were assessed using real time qPCR[65]. The researchers reported significant overexpression of miR-21 and miR-221 in patients with iCCA compared to healthy controls (P = 0.02 and P = 0.05, respectively). Moreover, circulating miR-21 expression demonstrated a high discriminatory ability between the two groups (AUC: 0.94). Based on these results, the study concluded that plasma miRNAs’ expression levels, particularly those of miR-21, could accurately differentiate patients with iCCA from healthy controls and potentially serve as supplements in diagnosis[65].

Lapitz et al[66], in 2020, attempted to characterise the transcriptomic profile of serum and urine EVs from patients with CCA, PSC, ulcerative colitis and healthy individuals. The majority of the identified transcripts were mRNAs, lncRNAs and miRNAs. The researchers reported differential RNA profiles in serum and urine EVs from CCA patients compared to both control groups (disease and healthy), as well as between CCA and PSC, with AUC values reaching 1.00 for certain mRNA markers. These markers include the ring finger and FYVE like domain containing E3 ubiquitin protein ligase, olfactory receptor family 4 subfamily F member 3 and the family with sequence similarity 107-member B. They concluded that these specific profiles could serve as potential diagnostic LB-based biomarkers for CCA[66].

Shigehara et al[67], in 2011, sought to confirm the presence of miRNAs in human bile and evaluate their potential as biomarkers for CCA. They collected bile samples from CCA and BBD patients and used cloning and PCR-based techniques to isolate miRNAs, confirming their existence in bile. Upon comparing different bile miRNAs, they identified that 335 out of 667 miRNAs were significantly more highly expressed in the CCA group than in the BBD group (P < 0.05) with 10 of these miRNAs (miR-9, miR-145, miR-105, miR-147b, let-7f-2, let-7i, miR-302c, miR-199a-3p, miR-222 and miR-942) showing significantly higher expression (P < 0.0005). In terms of diagnostic accuracy, with a specificity threshold set to 100%, certain miRNAs (miR-9 and miR-145) achieved a sensitivity of 88.9% and an AUC of 0.975. The researchers concluded that these findings suggest that bile miRNAs could be used as diagnostic biomarkers for CCA[67].

Li et al[68], also aimed to characterise bile EVs, including their miRNA content. In addition to confirming the presence of EVs and miRNAs in bile, they also demonstrated the stability of these particles and the potential of developing disease marker panels. Bile samples from 46 CCA and 50 BBD patients were collected and analysed to isolate EVs. Based on the different miRNA profiles between the two groups, the researchers developed a novel bile-based CCA diagnostic panel, which demonstrated a sensitivity of 67% and specificity of 96%, outperforming CA19-9 (sensitivity 58.9%) They concluded that their panel has potential clinical utility in diagnosing CCA[68].

Another study by Ikeda et al[69] assessed potential biomarkers for CCA diagnosis by performing proteomic analysis on bile samples from 10 CCA and 10 BBD. When comparing the two groups, they identified 166 CCA-specific proteins (P < 0.05) with one of them (Claudin-3) achieving, in terms of diagnostic accuracy, a sensitivity of 87.5%, a specificity of 87.5%, and an AUC of 0.945 (95%CI: 0.84-1), with a cut-off value of 37.61 pg/mL[69]. The small number of cases and single-centre design of the study represent important limitations.

Screening and surveillance in PSC patients: The applications of LB in early diagnosis can be further expanded into CCA screening in high-risk populations such as PSC patients and assist into developing cost-effective surveillance programmes that detect CCA before clinical or radiological abnormalities appear.

Voigtländer et al[70] studied miRNA patterns in serum and bile of patients with PSC, CCA and patients with PSC-associated CCA. They collected 40 serum and 52 bile samples from PSC patients, 31 serum and 19 bile samples of CCA patients, 19 bile samples of patients with PSC-associated CCA and 12 serum samples from healthy controls. Their analysis revealed distinct miRNA profiles between the groups. More specifically, bile MiR-412 (P = 0.001), miR-640 (P = 0.001), miR-1537 (P = 0.003) and miR-3189 (P = 0.001) were significantly different between patients with PSC and PSC-associated CCA, suggesting that these molecules could serve as diagnostic tools for early diagnosis of CCA in patients with PSC[70].

In another study, Vedeld et al[71] collected 344 bile samples from patients with sporadic and PSC-associated CCA using dPCR for DNA methylation analyses. They reported that samples from patients with PSC-associated CCA showed significantly higher methylation levels than those from patients with PSC for all identified biomarkers, including cysteine dioxygenase type 1, cannabinoid receptor interacting protein 1, septin 9 and vimentin (P < 0.0001), with AUCs, sensitivities, and specificities ranging from 0.77-0.87, 54%-76%, and 93%-98%, respectively. Notably, in the subgroup of PSC patients diagnosed with CCA within 12 months from bile sampling, their panel achieved a sensitivity of 100% and a specificity of 90% for detecting CCA in patients with underlying PSC, up to 12 months before CCA diagnosis. Based on their findings, the researchers concluded that their panel could serve as a valuable tool in screening and surveillance programmes of PSC patients[71]. Although the samples for this study were collected from three independent centres across three different countries, validation in larger prospective multicentre studies is still required.

Singhi et al[72] developed a 28-gene NGS panel (BiliSeq) using bile samples from patients with bile duct strictures, focusing on the most commonly mutated genes associated with CCA. Among patients with PSC, this panel achieved a sensitivity of 83% and a specificity of 100% for diagnosing at least high-grade biliary dysplasia, outperforming both serum CA19-9 (67% sensitivity and 84% specificity) and pathological evaluation following biliary brushings or biopsies (8% sensitivity and 100% specificity). These findings led the researchers to conclude that their panel, combined with the current diagnostic methods, could improve the detection of malignant strictures in PSC patients[72]. Although a major strength of the study is the prospective evaluation and analysis of a large number of bile duct specimens, the availability of follow-up diagnostic pathology for only 66% of patients constitutes a significant limitation.

Lapitz et al[73], in 2023, studied a large number of potential protein biomarkers in serum EVs of patients with PSC, patients with concomitant PSC-associated CCA, PSC patients who developed CCA during follow-up and patients with CCA from non-PSC aetiology. They found that serum levels of C-reactive protein (CRP), fibrinogen, fibrinogen-like protein 1 (FRIL) and polymeric immunoglobulin receptor (PIGR) were significantly increased in PSC patients who developed CCA compared to PSC without malignancy, providing differential AUC values up to 0.828. Additionally, increased levels of CRP/fibrinogen/FRIL were associated with a 44-fold increased risk of CCA development, with positive and negative predictive values of 91.7% and 80%, respectively. Interestingly, in terms of predictive value, CRP/fibrinogen/FRIL and CRP/fibrinogen/FRIL/PIGR significantly outperformed common markers of liver injury and cholestasis (alanine aminotransferase: AUC 0.910 vs 0.514, aspartate aminotransferase: AUC 0.910 vs 0.579, total bilirubin: AUC 0.910 vs 0.725, GGT: AUC 0.910 vs 0.510 and ALP: AUC 0.910 vs 0.618) as well as CA19-9 (AUC 0.987 vs 0.650). Based on the above findings, the study concluded that the aforementioned LB-based protein biomarkers demonstrated predictive capacity for PSC-associated CCA development before clinical and radiological evidence of malignancy[73].

Differential diagnosis: Distinguishing CCA from other non-malignant biliary conditions can be challenging due to overlapping clinical features and imaging findings. LB offers a novel approach that aids in differential diagnosis.

Wasenang et al[74] quantified the methylation of two markers, namely opioid binding protein/cell adhesion molecule-like (OPCML) and homeoboxD 9 (HOXD9) in the serum cfDNA of CCA and BBD patients. The combined marker of OPCML and HOXD9 achieved a sensitivity of 62.50%, a specificity of 100%, a positive predictive value of 100% and a negative predictive value of 72.72%, suggesting that these markers could be useful for the differential diagnosis of CCA and BBD[74].

Arechederra et al[75], assessed the potential of bile cfDNA, using NGS analysis, to distinguish benign from malignant biliary strictures. They used a prospective cohort of 68 patients, classified as benign, indeterminate or malignant based on pathology. Their NGS assay, “Bilemut”, achieved a sensitivity of 96.4% and a specificity of 69.2% in diagnosing malignancy. They concluded that it could assist with early differential diagnosis of biliary strictures and improve malignancy detection[75]. Furthermore, the researchers proposed an algorithm for managing patients with biliary strictures, aiming to enhance presurgical diagnosis and prevent unnecessary surgeries in cases of benign aetiology.

Another study also explored the potential of bile EVs to distinguish between malignant and non-malignant congenital biliary dilatation (CBD) strictures. The researchers reported that the median concentration of EVs was significantly higher in bile samples from patients with malignant vs benign strictures (P < 0.0001) and they also established a cut-off value of 9.46 × 1014 nanoparticles/L in bile, which could differentiate malignant from non-malignant CBD strictures with 100% accuracy[76]. The limitations of this study included the small sample size and the technical constraints of the current nanoparticle tracking analysis technology, such as the inability of light scattering to distinguish between EVs and other minor nanosized structures.

Han et al[77] explored the application of bile-based LB into distinguishing CCA from BBD. They collected bile samples from 106 patients with obstructive biliary disease and analysed them to screen for miRNAs using qRT-PCR. CA19-9 and CEA levels were also assessed for comparison. The results highlighted that the expression levels of miR-30d-5p and of miR-92a-3p were significantly upregulated in the CCA group compared to the BBD group (P < 0.001). The reported AUCs for bile miR-30d-5p, miR-92a-3p, serum CA19-9 and CEA were 0.730, 0.652, 0.675 and 0.603, respectively. Among the studied markers, bile miR-30d-5p demonstrated the best diagnostic performance, with a sensitivity of 81.1% and a specificity of 60.5%[77]. Significant limitations of this study include its single-institution design, the limited number of participants, the lack of external validation, and the absence of comparisons with healthy individuals.

Additionally, LB has a potential role in differentiating liver malignancies, such as CCA and HCC. Distinguishing those entities can be challenging due to overlapping features, particularly in cases of iCCA. Traditional imaging and serum markers are often insufficient to clearly distinguish between these tumours, and a tumour biopsy is often required, in particular when systemic treatment is considered.

In a study by Urban et al[78], serum EVs from patients with CCA and patients with HCC were isolated and analysed. They reported that a combination of two tumour-related antigens (AnnV + CD44v6 +), which were diagnostic of CCA, demonstrated a sensitivity of 91% and a specificity of 69% for differentiating CCA from HCC. Additionally, when a combined approach incorporating EV levels with serum alpha-fetoprotein was evaluated, it provided perfect differentiation between CCA and HCC, with sensitivity, specificity, positive and negative predictive value all reaching 100%[78].

Another study also explored the potential application of LB in distinguishing iCCA from HCC. The researchers analysed the serum metabolome of patients with iCCA and patients with HCC, and developed an algorithm based on the concentrations of four variables (glycine, aspartic acid, sphingomyelins 42:3 and sphingomyelins 43:2) that achieved an AUC of 0.890, a sensitivity of 75% and a specificity of 90% for CCA diagnosis. Based on their findings, they concluded that these specific metabolite profiles could assist in differentiating iCCA from HCC[79]. The study results are based on a small sample size, nevertheless an important strength is the external validation in an independent cohort of patients.

Prognostication and risk evaluation

Prognostic stratification in CCA is essential for optimizing patient outcomes, as the identification of “good” and “bad” prognostic biomarkers such as genetic mutations or specific protein expressions can impact on the number of cases eligible for various specific therapeutic interventions and facilitate more successful treatments. This personalised approach enables more tailored therapeutic strategies and improved survival rates in patients with CCA.

In a multicentre randomised phase 2 study evaluating the use of cediranib or placebo in combination with cisplatin and gemcitabine chemotherapy for patients with advanced biliary tract cancer, researchers also analysed CTCs levels in the participants. Blood samples were collected from 95 patients prior to the initiation of chemotherapy. The analysis revealed that patients with no detectable CTCs in their blood samples had the longest overall survival [median 18.1 months (95%CI: 12.1-24.9)]. In contrast, the presence of one CTC [median 10.3 months (95%CI: 4.7-14.4)] or two or more CTCs [median 8.7 months (95%CI: 7.1-12.6)] was significantly associated with poorer overall survival outcomes[80].

Liu et al[81] assessed the potential value of miR-21 as a prognostic factor in CCA patients by analysing serum samples of CCA and BBD patients but also healthy controls using qPCR methods. The findings of the study indicated that, beyond its potential use as a diagnostic marker, high miR-21 expression was significantly correlated with clinical stage, invasion depth, lymph vessel infiltration, metastasis status, differentiation status and poor survival in CCA patients (P < 0.05 for all)[81]. In a similar line of research, another group from China investigated the prognostic role of different miRNAs in CCA. Their study demonstrated that miR-26a levels were significantly reduced in the postoperative setting (post liver resection) compared to preoperative levels (P < 0.001) among patients subjected to potentially curative treatment surgeries. Additionally, high preoperative levels of this marker were significantly associated with clinical stage, lymph vessel infiltration, distant metastasis, differentiation status and poor survival (in terms of overall survival and progression-free survival) in CCA patients (P < 0.05 for all)[82]. A significant limitation of utilizing miR-26a as a prognostic marker for CCA is its low specificity, as it is already recognised as a marker for other cancers, including HCC, ovarian cancer and renal cell carcinoma. Consequently, it remains uncertain whether its expression is specifically linked to CCA or represents a broader phenomenon associated with cancer cell progression.

Cheng et al[83] studied the prognostic value of serum miR-106a and found that downregulation of miR-106a in serum samples was linked to a poor prognosis in CCA patients. Specifically, decreased levels of miR-106a (defined as below the cut-off value of 1.00) were predictors of the presence of lymph node metastasis [hazard ratio (HR) = 18.3, 95%CI: 5.9-56.4, P < 0.01] and shorter overall survival (HR = 5.1, 95%CI: 2.2-11.8, P < 0.01)[83]. In another study, the exosomal miR-200 family was evaluated as potential prognostic biomarker. Blood samples of 36 CCA patients and 12 healthy controls were collected and 20 exosomal miRNAs were analysed. Among these, miR-200c-3p showed significant prognostic value, as its expression was positively associated with tumour stage, demonstrating marked upregulation in patients with advanced disease (stage III–IV, P < 0.05)[83]. Nevertheless, these findings need to be validated through large-scale prospective studies.

Loosen et al[84,85] published two prospective studies investigating prognostic factors for CCA. These studies included 94 and 107 patients, respectively, who underwent resection of CCA. Blood samples were collected from participants both prior to surgery and 6-7 days following tumour resection. In the first study, they examined osteopontin, an EV glyco-phosphoprotein, and found that postoperative osteopontin levels were significantly higher in CCA patients who died during the follow-up period compared to those who survived. Using a cut-off value of 320 ng/mL, the analysis demonstrated a more pronounced difference in long-term survival, favouring patients with lower postoperative osteopontin levels (P = 0.004)[84]. In the second study, the researchers analysed serum concentrations of a four-miRNA panel (miR-122, miR-192, miR-29b, and miR-155). They found that while preoperative miRNA serum levels were not suitable for predicting patient prognosis, a postoperative reduction in miR-122 was associated with a favourable prognosis, with a HR of 2.949 (95%CI: 1.147-7.584) (P = 0.025). In both studies, the researchers concluded that the biomarkers investigated represent promising prognostic indicators for patients with resectable CCA[85].

Other LB-based biomarkers have also demonstrated potential for prognostic stratification. Csoma et al[86] found a strong correlation between the estimated tumour volume (as determined by preoperative imaging) and cfDNA levels (r = 0.9326, P < 0.0001)[86]. Furthermore, a study from China analysed bile cfDNA from CCA patients using whole-genome sequencing to investigate correlations between genetic alterations and prognosis. The researchers reported that specific genetic mutations in bile, such as those involving homologous recombination, were associated with poorer overall survival (P = 0.0049)[87]. This study once again features a low number of cases, all originating from a single institution.

A prospective study by Yang et al[88] analysed blood CTCs levels in 88 patients with CCA and evaluated their association with tumour extent and overall survival. Elevated CTCs levels were linked to poorer survival outcomes. Patients with CTCs ≥ 2 had a median survival of 5 months compared to 27 months for those with CTCs < 2 (P < 0.01), with a HR in multivariate analysis HR = 2.5; 95%CI: 1.1-5.4; P = 0.02. Similarly, patients with CTCs ≥ 5 had a median survival of 5 months vs 20 months for CTCs < 5 (P < 0.01, HR = 4.1, 95%CI: 1.4-10.8, P = 0.01), underscoring CTCs levels as independent prognostic markers. Additionally, higher levels of CTCs were significantly correlated with tumour extent, in terms of larger tumour size, multinodular disease, bilobar involvement (P < 0.01), and extrahepatic metastasis (P = 0.02)[88]. This single-centre study has limited generalisability due to its modest sample size and heterogeneity in patient characteristics, including CCA subtypes, tumour extent and treatment approaches. While this diversity allowed exploration of various patient and tumour factors influencing outcomes, it also constrained statistical robustness. The lack of serial blood draws prevented analysis of trends in CTCs and their predictive value for overall survival.

In a study conducted in Germany, researchers assessed the potential of CTCs as a preoperative marker for detecting occult metastases. The findings demonstrated that CTCs detection was significantly associated with worse overall survival (HR = 3.59, 95%CI: 1.79-7.1, P = 0.04) and with the presence of metastases identified only intraoperatively or shortly after surgery (P = 0.49). The researchers concluded that preoperative CTCs detection could provide prognostic value in predicting existing metastases, which could help improve patient stratification and reduce the frequency of open-close laparotomies[89]. Nevertheless, this study has important limitations. Besides the small sample size, the cohort included patients with various types of CCA, which are associated with distinct mortality risks, a fact that may have influenced survival outcomes. Furthermore, the heterogeneity of patients regarding their lymph node status, distant metastases and treatment modalities may also have affected the statistical analysis.

Additionally, an ongoing clinical trial seeks to develop a novel LB assay (based on exo-miRNAs in preoperative blood samples), that can accurately detect lymph node metastasis before treatment in iCCA patients[90].

Monitoring

Monitoring in CCA is crucial for guiding clinical treatment decisions, as it allows for the assessment of treatment response and the timely detection of relapse or minimal residual disease (MRD). Regular monitoring also helps identify resistance to therapies, prompting adjustments to treatment plans. Early detection of significant changes can lead to effective interventions to prevent disease progression.

Yang et al[91] assessed the potential of cfDNA to predict responses to immune checkpoint inhibitor (ICI)-based treatment in advanced hepatobiliary cancers. They analysed copy number variations (CNVs) in plasma cfDNA from 187 patients divided into three cohorts based on their treatment regimens: Combination therapy with a programmed cell death protein 1 inhibitor and lenvatinib, ICI-based therapy, and non-ICI therapy. Using these data, they developed a CNV risk score model based on the copy numbers of eight genes commonly associated with hepatobiliary malignancies. Their findings indicated that 53% of patients with low CNV risk scores achieved durable clinical benefit, while 88% of patients with high CNV risk scores showed no benefit from combination therapy (P = 0.004), suggesting that cfDNA could be useful for assessing treatment response[91]. Limitations of this study are its retrospective design, single-centre setting, and the relative heterogeneity of the cohorts with respect to cancer types and ICI-based treatment regimens.

Capturing tumour heterogeneity may be particularly important for evaluating resistance to treatment. Additionally, the ability to conduct serial cfDNA analyses can facilitate studies of tumour evolution. Two studies have identified genetic mechanisms of acquired resistance to FGFR inhibition in patients with FGFR2 fusion-positive CCA, highlighting the role of LB in tracking resistance. In the first study, LB was used to confirm FGFR inhibitor resistance in patients with FGFR2 fusion-positive CCA through recurrent FGFR2 point mutations detected in cfDNA and validated by tumour biopsy[92]. The second study demonstrated that strategic sequencing of FGFR inhibitors, informed by serial biopsies and ctDNA, may extend the effectiveness of FGFR inhibition in FGFR2 fusion-positive iCCA. Specifically, TAS-120, an irreversible pan-FGFR inhibitor, showed efficacy in four patients with FGFR2 fusion-positive CCA who had developed resistance to previous FGFR inhibitors. Together, these findings suggest that monitoring resistance mechanisms through LB and strategic inhibitor sequencing can improve FGFR-targeted treatment outcomes[93].

While conventional imaging remains the standard for assessing tumour response to treatment, LB is increasingly being integrated as an additional tool for monitoring therapeutic efficacy. Although data are limited regarding CCA, ctDNA has been evaluated as a prognostic biomarker to predict treatment response in patients with metastatic colorectal cancer, lung cancer and breast cancer[94-97]. One key advantage of LB is its ability to detect early signs of treatment resistance before it becomes apparent through clinical symptoms or conventional imaging. By monitoring dynamic changes in tumour-derived markers in real time, LB can provide critical insights into emerging resistance mechanisms. While data on CCA remain scarce due to its rarity, the early detection of treatment resistance using LB has been successfully demonstrated in lung cancer studies[98,99].

Finally, LB also holds potential for predicting clinical outcomes and identifying treatment failure before radiological progression. Additionally, it can aid in detecting relapse and the presence of MRD that may be undetectable by high-resolution imaging technologies[100,101]. The potential of LB to predict recurrence and MRD has been explored in several malignancies. Wang et al[102] investigated the role of ctDNA in detecting recurrence earlier than conventional surveillance in 58 patients with resected colorectal cancer. The study found that ctDNA was detectable before radiographic evidence of recurrence in all patients who relapsed. Notably, no relapses occurred in patients with negative ctDNA. While these findings are preliminary, LB biomarkers demonstrate significant potential for cancer monitoring and could become integral components of surveillance programs in the near future[102]. In a prospective study by Garcia-Murillas et al[103], plasma ctDNA was analysed in 55 early breast cancer patients to evaluate its utility for monitoring MRD. The study demonstrated that detecting ctDNA after surgery strongly predicted metastatic relapse. A single postsurgical time point yielded a HR of 25.1 (95%CI: 4.08-130.5, P < 0.0001), while serial analyses provided an HR of 12.0 (95%CI: 3.36-43.07, P < 0.0001). These results highlight the high accuracy of ctDNA in predicting relapse[103].

Regarding CCA, a study examining the potential role of ctDNA in detecting residual disease after potentially curative resection included 11 patients with pancreatobiliary malignancies. Although statistical significance was not reached and the sample size was very small, the authors observed a trend indicating that patients with detectable postoperative ctDNA had an increased risk of relapse. This suggests ctDNA could serve as a marker for MRD and may aid in identifying patients at higher relapse risk post-surgery[104]. In a second multicentre, randomised phase II study evaluating the role of ctDNA in predicting recurrence, plasma samples from 89 patients who underwent surgery for extrahepatic CCA were prospectively collected before and after chemotherapy initiation (at 12 and 24 weeks)[105]. The results demonstrated that ctDNA positivity was significantly associated with worse disease-free survival across all subgroups (all P < 0.029). Furthermore, patients with serially negative ctDNA tests exhibited longer disease-free survival compared to those with sustained ctDNA positivity or those who converted to ctDNA positive status (both P < 0.001). Based on these findings, the researchers concluded that ctDNA status and dynamics could aid in recurrence prediction and serve as monitoring tool for post-surgery surveillance in CCA patients. Furthermore, an ongoing clinical trial from the Mayo clinic (LB monitoring of CCA for treatment response and prognostic outcomes) explores the potential value of a new LB-based test approach to detect measurable residual disease or early recurrence/progression in patients with CCA[106].

DISCUSSION

LB is a promising modern technique with diverse applications in various fields of oncology, including diagnosis, prognostication, monitoring and overall management of cancer patients. Among its advantages, the non-invasive nature of the test with its capacity for serial analysis distinguish it among current methods. Unlike tissue biopsies, which carry procedural risks each time a new sample is taken, LB reduces these risks by using a simple blood test to access ctDNA[107]. This advantage is especially beneficial for patients who have insufficient tumour tissue to biopsy or whose tumours are inaccessible, such as in several cases of CCA, making traditional biopsies challenging or unfeasible[108]. In such cases, LB could act as an alternative option, providing a feasible and safe way to gather essential molecular information without the need for repeated invasive procedures to obtain tumour tissue. Even in the case of bile-based LB, a great proportion of CCA patients undergo biliary drainage as part of their treatment, so additional invasive procedures are not required. The current progress in the field holds promise that LB may acquire an even more prominent relevant role in the future.

LB enables comprehensive tumour profiling, capturing the tumour’s overall genetic landscape and providing a broader reflection of cancer heterogeneity[109]. Heterogenous tumours often exhibit distinct mutational profiles across different regions. ctDNA reflects a mix of DNA from all tumour regions, capturing a more complete genetic profile. This allows LB to identify both dominant and subclonal mutations. This capability is particularly valuable for CCA, which exhibits a high degree of intratumoural heterogeneity that is challenging to capture with a single tissue biopsy from the centre of the lesion[108,110]. ctDNA also includes contributions from metastatic lesions, providing a more holistic view of tumour heterogeneity across primary and secondary sites. This is particularly important for understanding metastatic drivers or resistance mechanisms not present in the primary tumour[110]. Different LB platforms vary in their ability to identify diverse genetic alterations. NGS and particularly deep sequencing NGS is highly effective in detecting low-frequency variants, which is essential for addressing the complexity of heterogeneous tumours. Methylation-based assays, on the other hand, focus on detecting epigenetic changes in ctDNA. These assays can reveal tumour-specific regulatory alterations and help identify subclonal populations with unique methylation patterns[111]. LB facilitates early diagnosis by detecting biomarkers even when tumours are undetectable via imaging, offering, additionally, opportunities for better surveillance and monitoring[101]. Moreover, LB biomarkers show promise as prognostic tools for risk-stratification and survival improvement. LB also supports real-time tracking of tumour evolution, treatment response and resistance mutations, aiding in personalised treatment planning[112].

Although LB holds tremendous potential to revolutionise the management of CCA and cancer patients in general, it has certain limitations. Firstly, the low concentrations of LB-based biomarkers compared to other blood components present significant challenges not only in isolating and analysing these biomarkers but also in establishing cell lines and xenograft models[50]. Furthermore, another challenge to overcome regarding the biomarker detection, is the low abundance of ctDNA in the bloodstream, especially in early-stage cancers or in patients with slow-growing tumours. This limited concentration of ctDNA often makes it difficult to differentiate tumour-derived DNA from the background cfDNA released by normal cells, resulting in reduced sensitivity[109]. Efforts are ongoing to develop more robust assays and computational tools to overcome these challenges and fully realise the potential of ctDNA-based diagnostics[113,114].

Additionally, developing LB models can be time-consuming and costly, which may limit their application in clinical practice[50]. Although the potential of LB to complement traditional tissue sampling has been evaluated in multiple studies, it cannot at this stage be considered a complete replacement for tissue biopsy[115]. Despite its promise, concerns remain about the sensitivity of LB in detecting mutations, limiting its effectiveness as a standalone diagnostic tool. Furthermore, some mutations are predominantly identified through tissue biopsies rather than LB[116]. These factors underscore the current need for a hybrid diagnostic approach that integrates both ctDNA analysis and tissue sampling.

In the field of CCA, the current pool of evidence regarding LB applications is derived predominantly from small, single-institution studies. Prospective, multicentre and well-organised studies are notably lacking. Additionally, the available studies are heterogeneous, with considerable variation in the biomarkers investigated and methodologies employed. Many studies focus on narrow scientific aspects, limiting their broader applicability. This fragmentation has limited the emergence of robust, universally accepted findings from larger-scale investigations, underscoring the urgent need for more comprehensive and standardised research in this area. Given the variety of the available ctDNA assays, there is a critical need for more comprehensive cross-platform comparisons to created standardised and universal models[45]. Additionally, each individual LB platform requires independent validation through multicentre, collaborative studies, particularly for rare cancers like CCA[41]. In this regard, enhancing the clinical utility of LB will also require conducting studies with large, independent cohorts to validate promising biomarker candidates for CCA and to develop new methodologies for biomarker identification that can support routine patient care[36].

Standardising LB procedures is essential to ensure reproducibility, reliability and comparability of results across studies and clinical settings[117]. Standardisation is particularly needed for factors influencing the assay’s performance and reporting, including pre-analytical variables, analytical considerations and laboratory assay reporting. During the pre-analytical phase, variables such as sample collection (e.g., type of collection tubes and volume of fluid), processing time, centrifugation protocols and plasma storage conditions must be clearly defined[118]. In the analytical phase, it is essential to use validated kits for biomarker isolation, reference standards to quantify biomarker concentration and integrity and clearly specify thresholds for sensitivity and limit of detection for each technology (e.g., PCR-based or NGS). Additionally, implementing quality control checkpoints ensures consistency and accuracy across assays[119]. For the reporting phase, adopting a standardised framework is critical. This framework should include clinical and demographic information, details of pre-analytical and analytical conditions, mutational data with variant allele frequency thresholds and the sensitivity and specificity metrics of the assay. Reports should also highlight assay limitations and provide interpretative guidance on the clinical relevance of each detected alteration[119].

Analytical validity studies, which assess how accurately and reliably a test measures a biomarker, are crucial to support the adoption of new technologies as the standard of care[120]. In the context of CCA, various detection platforms and methodologies for LB are under investigation. However, there is currently no consensus on the most effective method for clinical application in LB and further analytical validation is necessary to establish reliability and accuracy. Clinical validity studies, which evaluate how well a test predicts clinical outcomes, are equally important[121]. In CCA, ctDNA has been explored as a prognostic biomarker. Nevertheless, comprehensive studies demonstrating diagnostic accuracy across diverse clinical scenarios in CCA remain limited, underscoring the need for further research to establish robust clinical validity. Finally, clinical utility studies, which assess the impact of a test on patient outcomes, are vital to demonstrate the tangible benefits of new technologies[121]. While LB shows promise in guiding treatment decisions and monitoring therapeutic efficacy across various cancers, evidence specific to CCA is still emerging. The integration of LB into clinical practice for CCA holds the potential to enable earlier interventions and more personalised treatment strategies. However, larger and more comprehensive studies are required to clearly establish its impact on patient outcomes in CCA.

LB has exciting potential for future applications in managing CCA, especially when paired with advancements in artificial intelligence (AI) and machine learning (ML). AI/ML algorithms can process vast amounts of complex datasets, integrating multiple types of biological data, including genomics, epigenomics, proteomics and others, enhancing diagnostic precision, enabling better predictions of patient outcomes and allowing for more accurate and personalised treatment strategies[109,122,123]. In an attempt to compare plasma and bile cfDNA, researchers used a multigene panel for plasma LB (LiquidPlex) to analyse bile cfDNA samples from 24 biliary tract cancer patients, including 17 cases of CCA. A ML approach was employed to classify tumour-derived genetic variants. The analysis revealed that, across all biliary tract cancer cases, bile cfDNA demonstrated a higher detection rate for cancer driver mutations (54%) compared to plasma cfDNA (17%)[124]. The integration of AI/ML significantly improved the classification and detection of tumour-derived variants, reinforcing the utility of these advanced computational approaches in precision oncology. ML techniques applied to broad proteomic profiling of serum samples have shown promise in identifying biomarkers for disease presence, severity and cirrhosis in patients with PSC[125]. These advancements suggest a potential future application of LB in CCA, where similar proteomic and ML approaches could be expanded to identify PSC patients at risk of developing CCA or enhance early detection and monitor disease progression in PSC-associated CCA.

Recently, the first multi-omics database, cfOmics, was introduced, integrating comprehensive LB data, including NGS-based multi-omics, as well as proteome and metabolome data derived from mass spectrometry, across 69 disease conditions[126]. The combination of LB and radiomics offers exciting possibilities for improving the care of patients with CCA[127]. LB provides molecular insights by detecting tumour-derived markers in the blood, while radiomics extracts detailed imaging features that reveal the tumour’s shape, texture and progression[128]. This synergy could improve early diagnosis, prognostication and monitoring, and may also uncover new treatment options and allow for highly personalised care tailored to each patient’s unique cancer profile[129,130]. The integration of conventional imaging techniques with ctDNA analysis holds the potential to improve the assessment of prognostication and treatment response in cancer patients. In the literature, studies evaluating this integration are limited and have primarily focused on its application in developing prognostic models to predict survival[131]. Integrating LB with conventional imaging can improve follow-up strategies by providing a more comprehensive view of tumour status, enabling earlier intervention and optimising treatment decisions.

Additionally, LB could be used to guide the selection of targeted treatment agents, by identifying specific mutations or biomarkers. This information would allow the selection of personalised therapies, moving towards the implementation of precision oncology[132,133]. Furthermore, recent LB-based transcriptomic profiling has identified novel subgroups within iCCA, which may allow for a more refined classification in the future, facilitating the development of targeted treatments for distinct molecular subtypes[134,135]. Finally, emerging technologies enhancing the sensitivity and specificity of LB but also the integration of imaging techniques could provide a more comprehensive view of CCA by combining molecular data with anatomical insights, improving diagnosis, prognostication and ongoing monitoring[127,136].

Altogether, these advances could transform the use of LB in CCA, paving the way for personalised, real-time disease management. While currently not universally accessible and affordable, the reduced procedural burden and potential for repeated testing make LB an economical choice for longitudinal monitoring, potentially offsetting costs by reducing the need for multiple invasive biopsies. As research continues to validate its utility, LB may soon become a standard component of future guidelines and algorithms, offering a non-invasive, dynamic and comprehensive approach to managing CCA at various stages.

CONCLUSION

In conclusion, LB holds significant promise for transforming the management of CCA by providing a non-invasive means to diagnosis, prognosis and treatment monitoring. As a fundamental component of personalised medicine, LB offers rapidly available genetic biomarkers that could significantly aid diagnosis, streamline and improve treatment decisions, optimise therapeutic monitoring and provide early insights into relapse or disease progression. Given the genetic complexity and heterogeneity of CCA, advancing our understanding of these tumours is essential to enable early diagnosis, identify patients most likely to benefit from specific therapies and develop more effective targeted treatments. However, further research is essential to fully realise this potential, moving towards a comprehensive, multimodal approach that integrates LB into precision oncology for CCA.