Breath- and blood-based molecular assessment for gastroesophageal cancer

Abstract

Gastroesophageal cancer remains a significant health care problem characterized by diagnosis at advanced stages, rendering a significant proportion of patients ineligible for curative treatment. Conversely, early diagnosis and treatment of gastroesophageal cancer demonstrates markedly improved outcomes, with 5-year overall survival rates ranging from 83% to 96%. To date, gastroesophageal cancer surveillance primarily focuses on patients with Barrett’s esophagus, relying on invasive upper endoscopy. However, despite its high specificity, upper endoscopy demonstrates a suboptimal cancer yield and moderate missed rates, compromising its cost-effectiveness for screening in the general population. Moreover, screening strategies for gastroesophageal cancer are still lacking, in part owing to the invasiveness of current diagnostic methods.

To overcome these challenges, innovative minimally invasive technologies have emerged, including metabolomics of exhaled breath, and detection of circulating tumor DNA or proteomics in blood. Promising results from phase I diagnostic trials emphasize the potential of these approaches to advance early detection and monitoring of gastroesophageal cancer. As these methods further refine and consistently demonstrate adequate sensitivity and specificity, they are poised to challenge existing diagnostic modalities. This progress may prompt reconsideration of current one-size-fits-all paradigms, facilitating the evolution toward a patient-tailored management of gastroesophageal cancer.

In this article, we review the current minimally invasive breath- and blood-based diagnostic approaches for gastroesophageal cancer, elucidating the underlying biologic basis of identifiable biomarkers. As these approaches will become an integral part of future diagnostic paradigms, understanding the operational mechanisms, strengths, and limitations of these approaches is crucial for optimizing their implementation and interpretability in clinical practice.

Highlights

• •Despite dismal prognosis, there is currently no screening for gastroesophageal cancer.
• •Oncometabolites detected through breath or blood analysis show promise to function as onco-biomarkers.
• •Oncometabolites play a role in tumor microenvironment interactions.
• •Future trials are needed to validate these novel onco-biomarkers and facilitate screening trials on gastroesophageal cancer.

Introduction

Gastroesophageal cancer remains a global health care problem. Owing to its pauci-symptomatic nature in early-stage disease, gastroesophageal cancer is still predominantly diagnosed at advanced stages. This results in only one-third of the patients being eligible for treatment with curative intent.¹ Even despite curative interventions, serious complications related to the multimodality therapy and rapid recurrence contribute to the poor prognosis with 5-year overall survival rates of 20%-40% for late- and advanced-stage gastroesophageal cancer, respectively, causing 1.5 million deaths annually which resembles 15% of all cancer-related mortality.1, 2, 3, 4 Conversely, early diagnosis and treatment of gastroesophageal cancer yield significantly better outcomes, with 5-year overall survival rates ranging from 83% to 96%.^5,6 However, as symptoms suggestive of gastroesophageal cancer (such as dysphagia, odynophagia, and weight loss) exhibit low predictive value, the current approach to detection and diagnosis relies on invasive upper gastrointestinal endoscopy with biopsies.⁷ With an overall cancer yield of 2% for all annually-carried out endoscopies, this method lacks sufficient cost-effectiveness for the early detection of gastroesophageal cancer in the presence of upper gastrointestinal symptoms.⁸ Furthermore, the sensitivity of endoscopy in detecting dysplasia or cancer ranges from 51% to 90%, leading to a missed cancer rate (diagnosed within 1 year of endoscopy) of up to 25%.9, 10, 11 Therefore, there is a pressing need for improved, ideally non-invasive, diagnostic approaches that can facilitate a transition toward the detection of more early-stage gastroesophageal cancers, reserving endoscopy as a confirmatory test to obtain specimens for histopathologic evaluation.

The absence of proficient screening methods prompted researchers and clinicians to explore innovative minimally invasive approaches to identify diagnostic metabolic biomarkers of gastroesophageal cancer in exhaled breath and blood.^8,12, 13, 14, 15 Despite the potential of these biomarkers, their sometimes unknown biochemical origin or relationship to cancer has subjected these approaches to certain scrutiny, precluding the initiation of large well-designed reproducible real-life phase II and phase III diagnostic trials.¹⁶ Nevertheless, considering their close relation to the genetic alterations of cancer cells, these biomarkers have the potential to reevaluate the current diagnostic, treatment, and follow-up paradigms for gastroesophageal cancer. Therefore, this review aimed to synthesize the currently available minimally invasive diagnostic modalities for detecting metabolic biomarkers related to gastroesophageal cancer and the underlying biologic basis of identifiable biomarkers, aiming to enhance the general understanding and pave the way for future implementation in clinical practice.

Search strategy and selection criteria

References for this review were identified through searches of PubMed with the search terms ‘biomarkers’, ‘ctDNA’, ‘cell-free DNA’, ‘volatile organic compounds’, ‘gastroesophageal cancer’, ‘esophageal cancer’, ‘screening’, and ‘diagnosis’, from 1969 to 2024. Additional articles were also identified through searches of selected article’s reference lists and the authors’ own files. Only papers published in English were reviewed. The final reference list was generated on the basis of originality and relevance to the broad scope of this review.

Oncometabolites are generated in response to oxidative stress and altered protein functions related to genomic alterations in cancer cells (Figure 1)

Figure 1.

Oncometabolites are created through up-regulation of cytochrome P-450 (CYP450) and protein alterations in relation to genomic mutations. Oncometabolites can influence intracellular cell signaling pathways as well as the tumor immune microenvironment after excretion of the oncometabolites through transporters and exosome formation.

AKT, protein kinase B; AMPK, adenosie monophosphate activated protein kinase; EGFR, epidermal growth factor receptor; TCA, tricarboxylic acid; TKR, Tyrosine kinase receptor; VEGFR, vascular endothelial growth factor.

Illnesses, including cancer, can modify the biochemical composition of patients by either changing the concentration of metabolites or generating new metabolites, including organic compounds, proteins, and DNA/RNA remnants. One mechanism involved in the production of cancer-related metabolites (oncometabolites) is associated with increased oxidative stress and the subsequent overexpression of cytochrome P-450 enzymes.¹⁷ These enzymes play a crucial role in maintaining the balance of free radicals and reactive oxygen species (ROS) within the cell. ROS, originating from endogenous and exogenous (e.g. cigarette smoke, pollution, radiation) sources, can accumulate in the tissues, interact with DNA, fatty acids and proteins, and induce mutagenesis.¹⁸ The induced up-regulation of cytochrome P-450 enzymes consequently alters the clearance of affected proteins and fatty acids, resulting in the formation of various alcohol- or alkane-derived oncometabolites such as 1-propanol and pentanal.¹⁹ In this context, increased levels of arachidonic acid and linoleic acid have been identified as potential cancer biomarkers in lung cancer and have demonstrated involvement in cancer progression and proliferation through activating the AKT pathway.²⁰

Furthermore, in the early stages of cancer development, dysplastic cells undergo increased proliferation, reaching the limits of oxygen diffusion.^21,22 As a consequence, these cells transition to glycolysis to meet their energy and biosynthesis demands in response to the hypoxic state.^21,22 This metabolic shift is primarily regulated by increased activity of pyruvate kinase 2 (PKM2), leading to the production of lactic acid, which is excreted into the stromal tissue surrounding the cancer (tumor microenvironment).^21,23 Interestingly, this acidic environment may enhance tumor invasiveness by disrupting basement membranes and promoting immune evasion by reducing immune cell infiltration, differentiation, and function.^19,24

Oncogenic alterations to DNA induce changes in the number, composition, and function of intracellular proteins, impacting various metabolic pathways such as those involved in carbohydrates, fatty acids, and amino acids.²⁵ Consequently, these genomic alterations can generate distinct oncometabolic profiles that are secreted by the cancer cells. Given the significant variability in genomic profiles across different cancers, a characteristic used in cancer profiling,^26,27 the diverse metabolic profiles arising from these genetic alterations have also proven to serve as effective surrogates for diagnosing various cancer types.^8,28,29 For instance, mutations in genes related to isocitrate dehydrogenase, fumarate hydratase, or succinate dehydrogenase lead to elevated levels of 2-hydroxyglutarate, succinate, and fumarate, respectively.³⁰ Furthermore, several studies suggest that these metabolites may actively promote tumor growth by inhibiting epigenetic enzymes and suppressing DNA repair.²⁵ In addition to the effect of lactate in the tumor microenvironment, these findings suggest that cancer cells may deliberately produce specific oncometabolites. Additionally, when excreted, these altered proteins circulate in the bloodstream and may consequently function as oncometabolites themselves. Table 1 provides a summary of the different chemical classes of potential oncometabolic biomarkers.

Table 1.

The different biochemical classes of potential biomarkers related to cancer

Chemical class	Origin	Examples
Fatty acids	•Cellular metabolism	Pentanoic acid Hexanoic acid Arachidonic acid Linoleic acid
Alkanes	•Peroxidation of polyunsaturated fatty acids •Reactive oxygen specimens	Pentane 3-methylhexane 1-heptene 2,4-dimethyl-1-heptene
Alcohols	•Hydroxylases of hydrocarbonmetabolism •Cytochrome P-450	Methanol 1-propanol 2-methyl-propanol
Aldehydes	•Dehydrogenases of alcohols •Reduction of hydroperoxide in cytochrome P-450 (fatty acid peroxidation)	Butanal Heptanal Decanal
Ketones	•Peroxidation of polyunsaturated fatty acids •Amino acid metabolism	Hydroxybutyrate Acetoacetate Propanone
Esters	•Loss-of-function enzymatic mutations •Hydrolysis causes acid formation	Butyl acetate 2-hydroxyglutarate Succinate Fumarate
Nitriles	•Cytochrome P-450 mono-oxygenase •Exogeneous	Acetonitrile Acrylonitrile
Aromatics	•Exogeneous and cause peroxidative damage	Trimethyl-benzene 4-isopropyl-toluene Methyl phenol Ethyl phenol
Sulfides	•Desulfhydration of cysteine	Hydrogen sulfide Carbon disulfide
Proteins	•Transcripts of RNA •Release by transporters or exosomes •Surface proteins on exosomes	Tumor growth factor beta Pepsinogen CD54 Immunoglobulin G Creatinine
Nucleic acids	•DNA breakage or RNA transcripts •Exosomal release	Mutated cell-free DNA or RNA

Non-exhaustive list. Composed of different sources.^{8,18,20,54,55,64,65,67,76,77}

Excretion of oncometabolites is based on concentration equilibria and exosome formation, enabling their detection through breath- and blood-based approaches (Figure 2)

Figure 2.

Collection of the oncometabolites through breath analysis and blood analysis. Breath analysis can be carried out through either electronic nose technology, using semi-selective sensors, or liquid chromatography/mass spectrometry. Blood analysis can be carried out through liquid chromatography/mass spectrometry for protein and metabolite detection or DNA or RNA sequencing for circulating tumor DNA.

The excretion of oncometabolites is based on the concentration equilibrium between tissue and blood, as well as blood to air, with each oncometabolite possessing a unique partition coefficient that determines its likelihood of being retained in the blood or excreted in the alveoli or kidneys.^19,31,32 Depending on their solubility and volatility, oncometabolites are present at varying concentrations and can be more easily detected through either breath- or blood-based analyses.^33,34 In regard to the breath analysis, the exchange of oncometabolites even extends beyond the alveoli, with studies revealing that metabolites with low solubility in the blood are primarily exchanged in the alveoli, while those with high solubility are exchanged in the respiratory and even smaller conducting bronchioles.³¹ In addition to transmembranous excretion, metabolites, DNA, and RNA are also excreted through exosome formation via the endocytic pathway. While initially considered waste products of cellular metabolism, recent investigations suggest that tumor-derived exosomes play a crucial role in cell signaling, altering the tumor microenvironment and enhancing the metastatic potential of cancer cells.^14,35, 36, 37, 38 These exosomes may contribute to the creation of a pre-metastatic niche by attracting neutrophils, which then form neutrophil extracellular traps for circulating cancer cells.36, 37, 38

A typical breath or blood sample contains between 3000 and 6000 different metabolites at concentrations as low as parts-per-trillion by volume or potentially even lower in early-stage disease, given the direct relationship between oncometabolite concentrations and tumor volume.^39,40 Consequently, the detection of these oncometabolites in early-stage tumors necessitates highly sensitive technologies. Primarily, quantitative methods based on gas chromatography and mass spectrometry have been utilized to identify metabolites and construct disease-specific reference libraries.³⁴ Technological advancements, including proton transfer reaction, selected ion flow, or liquid chromatography/mass spectrometry, have improved detection rates to parts-per-billion per volume with the potential for instant read-out.⁴⁰ However, these techniques are time-consuming, require expensive laboratory equipment and space, and need the expertise of specially trained personnel.³⁴ Furthermore, they may not detect trace levels of oncometabolites present in in situ carcinoma or precancerous states, complicating implementation into clinical practice. Moreover, these technologies currently require the transport of oncometabolites from the collection site to the analysis site using tethered bags, which can result in the loss of volatile compounds.¹⁹ These limitations may contribute to an increased false-negative rate, as the technologies may lack the sensitivity to detect trace levels of oncometabolites. Technologies, such as chill-stored carbon cartridges or instant read-out devices, aim to mitigate compound loss, thereby enhancing the clinical applicability of these methods.^41,42

In this context, qualitative approaches have been introduced for exhaled breath analysis using electronic nose devices (E-nose), leveraging the principles of combinatorial selectivity analogous to the olfactory system. These E-noses discern patterns of oncometabolites through semi-selective sensor arrays, comprising nanomaterial conductors, electroacoustic sensors, colorimetric, or thermodynamic sensors.^28,42, 43, 44 These sensors exhibit distinct interactions with a selected variety of oncometabolites, generating a multidimensional signal that discriminates not only between different cancers but also benign diseases.^45,46 E-nose devices offer the advantage of demonstrating a higher discriminatory capacity compared with the classical quantitative approaches.³⁴

Finally, it is important to note that, to date, both chromatography/spectrometry and E-nose technology require participants to adhere to fasting protocols before testing, preventing dilution by exogenous components. However, development of ultrasensitive instant read-out devices capable of measuring quantities of components rather than concentrations may mitigate this fasting requirement, thereby enhancing patient convenience.⁴² Furthermore, the continued miniaturization of these devices will support their clinical implementation beyond hospital settings, enabling their use in health care-vulnerable areas.

Despite lower sensitivity, tumor-agnostic assays are required to detect circulating tumor DNA and RNA in the setting of screening

Following tumor cell death, remnants of DNA and RNA (ctDNA/ctRNA) are released into the tumor microenvironment and bloodstream, serving as detectable biomarkers in blood samples.⁴⁷ However, only a minute fraction—∼0.014%—of a tumor cell’s DNA is typically shed into the bloodstream,⁴⁷ a process influenced by factors such as tumor size, cell turnover, and number of detectable mutations via diagnostic assays.^47,48 Currently, there is no universally standardized ctDNA assay, and detection methods rely on either a tumor-informed or tumor-agnostic approach. In a tumor-informed assay, DNA is extracted from the primary tumor. A second sample is then compared with this sample, potentially improving sensitivity by leveraging the knowledge of somatic DNA alterations. However, this approach lacks the ability to screen for cancer via blood analysis, as no comparative tumor sample is available. Conversely, the tumor-agnostic approach avoids the need for prior tissue acquisition. Consequently, these DNA assays are more readily available for use in screening programs and demonstrate improved cost-effectiveness but may translate into lower sensitivity.^15,49 It is important to note that, to date, there has been no direct comparison between the different ctDNA tumor-agnostic assays, which complicates the interpretation and comparison of various trials. Unlike the tumor-specific approach, which employs droplet digital or real-time fluorescence quantitative PCR or next-generation sequencing, the tumor-agnostic approach utilizes whole-genome or whole-exome sequencing for accurate tumor origin assessment.⁵⁰ Consequently, detecting early-stage tumors through ctDNA remains challenging. Nevertheless, technological advancements, such as assessing methylation patterns or utilizing phased variant enrichment and detection sequencing, hold promise in improving the diagnostic precision, potentially rendering them more suitable for early-stage tumor detection in screening programs.^49,51 Moreover, ctDNA enables the identification of potentially targetable genomic alterations. In addition, a recent study conducted at our university investigated the diagnostic value of cell-free DNA fragmentation profiling.⁵² This profiling leverages the distinct functioning of nuclease-dependent nucleosomes. In particular, cell-free DNA fragments derived from these nucleosomes typically consist of 147 base pairs, intricately wound around histone octamer cores. Given that nucleases demonstrate a preference for cleaving DNA at specific sites depending on the cell of origin, these cell-free DNA fragmentation assays hold promise for accurately determining the cancer’s origin in a tumor-agnostic manner.⁵² However, before widespread adoption, this technique requires further validation in the clinical settings. Moreover, its application in gastroesophageal cancer has not yet been established. Despite these potential major benefits, the cost associated with detecting ctDNA or cell-free DNA—ranging from €168/$199 to €7638/$9124 per sample—should be taken into consideration.⁵³ This variation in cost is related to the assay sensitivity, detection methodology, and laboratory volume.⁵³

The combinatorial presence of breath-based biomarkers demonstrates high sensitivity and specificity for gastroesophageal cancer and the precancerous Barrett’s esophagus

In contrast to blood-based testing, it is crucial to highlight that research on breath analysis for diagnosing gastroesophageal cancer is still relatively scarce, often due to the false assumption that the tumor must be in direct contact with the airways in order to be detectable. However, building on the initial investigation of volatile oncometabolites, called volatile organic compounds (VOCs), in the urine and gastric headspace content using selected ion flow tube mass spectrometry,^54,55 Kumar and colleagues pioneered the detection of VOCs associated with gastroesophageal cancer in the exhaled breath.⁵⁵ They were able to identify a selection of VOCs significantly increased in gastroesophageal cancer compared with healthy controls, which were predominately derived from the phenol (phenol, methyl phenol, and ethyl phenol) and aldehyde pathways (hexanoic acid).⁵⁵ These results subsequently led to investigating differences between the VOC profiles of esophageal adenocarcinoma, Barrett’s esophagus and normal upper gastrointestinal tissue.⁸ In this study, a set of 12 VOCs, including fatty acids, phenols and aldehydes, successfully distinguish gastroesophageal cancer from non-cancer patients, including Barrett’s metaplasia, with a sensitivity and specificity of 88% and 83%.⁸ Notably, the results of this study showed no influence of tumor stage on diagnostic performance,⁸ while in a smaller series from Zou and colleagues early-stage gastroesophageal cancer could be differentiated from advanced-stage cancers based on the concentration of VOCs.⁴⁰ Given the multifactorial nature of biomarker excretion, which can be influenced by tumor volume, activity, and genomic profile, further research, exploring the correlation between these features and VOC concentration, is warranted. Nevertheless, the studies from Kumar and Zou suggest that VOCs have the potential to detect early-stage tumors and could be used for early detection in screening programs.^8,40

In the context of Barrett’s esophagus—the only recognized precursor of esophageal adenocarcinoma—no discernable difference in VOCs was detected between normal upper gastrointestinal tracts and Barrett’s metaplasia, using selected ion flow mass spectrometry.⁸ While identifying such differences could aid in recognizing individuals at risk of developing esophageal cancer and guide further follow-up, it is essential to note that metaplasia is still considered a benign cellular transition phase.⁹ Consequently, it is plausible that other VOCs, not currently captured or identified with existing technologies, could differentiate Barrett’s metaplasia from benign upper gastrointestinal tissue. Supporting this hypothesis, the combinatorial selectivity and pattern-recognition platform of the E-nose demonstrated the ability to distinguish more subtle differences in VOC profiles between Barrett’s esophagus and normal gastroesophageal tissue.^45,56 Two proof-of-concept studies using E-nose in the setting of Barrett’s esophagus demonstrated a sensitivity and specificity of 82%-91% and 74%-80%, respectively.^45,56 The presence of long-segment Barrett’s esophagus improved the discriminatory ability of E-nose technology, suggesting an influence of the disease burden on the concentrations of VOCs. Importantly, E-nose performance appeared independent of clinical conditions such as the intake of proton pump inhibitors, the presence of hiatal hernia, or gastroesophageal reflux disease.⁴⁵ While the study by Chan and colleagues had a small number of patients diagnosed with intramucosal carcinoma (4.5%),⁵⁶ E-nose technology has not yet been validated in the context of in situ carcinoma or the full spectrum of Barrett’s dysplasia. Therefore, our center is currently enrolling patients in a phase IIa diagnostic trial comprising individuals with Barrett’s esophagus, dysplasia, and early-stage esophageal cancer to address this knowledge gap.

Current blood-based biomarker tests demonstrate high specificity for the detection of gastroesophageal cancer

The use of ctDNA in the context of early-stage gastroesophageal cancer is still not well established. One of the main challenges remains the low proportion (0.07%) of ctDNA among the circulating DNA in a blood sample, related to the low DNA shedding (0.014%) into the bloodstream upon cell death in localized cancer.^47,57 This low abundance of ctDNA necessitates the use of highly sensitive ctDNA assays to diagnose esophageal cancer, especially in the early stage. To overcome this low-abundance issue, DNA methylation, considered an early carcinogenic event that often occurs in promoter regions of oncogenes, has been explored as a marker to identify distinctive methylation patterns for distinguishing and confirming cancer.^48,51 In this context, two pan-cancer trials from Klein and colleagues¹⁵ and Liu and colleagues⁴⁸ demonstrated a sensitivity for early-stage gastroesophageal cancer of 13%-17% and 17%, respectively, while the specificity, using this method, was 99% in both trials. Of note, for patients detected in more advanced gastroesophageal cancer (stage III and IV), the sensitivity increased to >80%, suggesting the performance is still primarily dependent on the resolution of the DNA methylation assays. These observations are reiterated specifically for squamous-cell carcinoma in a study by Iwaya and colleagues,⁵⁸ which observed ctDNA positivity in one in seven patients (14%) with stage I disease and 23 in 27 patients (85%) with stage II disease or higher. In a study by Qiao and colleagues⁵⁹ sensitivity for stage 0-II squamous-cell carcinoma was 59%. Interestingly, a longitudinal study by Chen and colleagues demonstrated that ctDNA detection could precede the clinical diagnosis by up to 4 years.⁶⁰ However, in the same study, the detection of gastric cancer as well as early-stage gastroesophageal cancer was the most challenging with a wide range of prediction scores.⁶⁰ Nevertheless, this observation supports the use of minimally invasive diagnostics to prompt more intensive follow-up in order to detect these cancers early. Moreover, in Barrett’s esophagus the fraction of allelic loss has shown to incrementally increase with the transition from metaplasia to high-grade dysplasia, suggesting its potential use as a diagnostic biomarker for early detection of esophageal cancer.⁶¹ Importantly, the presence of clonal hematopoiesis has been associated with false-positive results on ctDNA analysis. Since somatic mutations gradually accumulate with increasing age, the majority of mutations identified in these typically older esophageal cancer patients may stem from clonal hematopoiesis.⁶² While exclusion of the hematopoietic variants may reduce the detection rate of ctDNA, integrating the ctDNA analysis with that of matched white blood cells can facilitate the differentiation of tumor-related mutations, thereby preserving the overall detection rate of these assays.

Despite the challenges faced by the tumor-agnostic approaches to detect early-stage tumors, valuable insights can be gained by examining the performance of tumor-specific assays in the context of minimal residual disease. In this scenario, where limited tumor cells are present, resulting in minimal circulating material, tumor-specific approaches exhibit a sensitivity of ∼70%, a significantly higher performance compared with current tumor-agnostic approaches, while maintaining a specificity of 99%.^12,57 Conversely, a longitudinal study including 245 samples in 97 patients from the study by Ococks and colleagues assessing minimal residual disease through a tumor-agnostic panel by Roche on a limited panel of 77 genes demonstrated a sensitivity of 36% with a 97% specificity.⁶³ This suggests the sequencing panel needs to be sufficiently comprehensive.

The detection of oncometabolites in blood has been feasible for several decades. Serum biomarkers, such as carcinoembryonic antigen, P53 antibody, cytokeratin fragment antigen 21-1, squamous-cell carcinoma antigen, protein kinase D1, matrix metalloproteinase 2, tissue inhibitor of metalloproteinases-2, and serum macrophage colony-stimulating factor, have demonstrated to be increased in the presence of esophageal cancer. However, these markers are also linked to other cancer types or inflammatory processes, leading to an overall moderate sensitivity and specificity.^64,65 To address this issue of moderate sensitivity and specificity, several investigators explored combinatorial change in circulating oncometabolites. Similar to the breath analysis, Bhatt and colleagues⁶⁶ investigated the headspace of plasma using ion flow tube mass spectrometry, revealing significant reductions in concentration of acetonitrile, acrylonitrile, carbon disulfide, isoprene, 1-heptene, 3-methylhexane, [E]-2-nonene, hydrogen sulfide, and triethylamine in gastroesophageal cancer compared with patients with reflux disease. Logistic regression identified acrylonitrile and carbon disulfide as the independent predictors of esophageal cancer, and a model based on these two components demonstrated an area under the curve of 0.83.⁶⁶ Moreover, Zhang and colleagues were able to identify 12 differentiating oncometabolites (lactate, valine, leucine, methionine, carnitine, tyrosine, tryptophan, myristic acid, margaric acid, linolenic acid, linoleic acid, and pyroglutamic acid) for esophageal cancer in the serum of patients using liquid chromatography/mass spectrometry. Using principal component analysis, they constructed a prediction model with an area under the curve of 0.82.⁶⁷

A study investigating alterations in N-glycosylation of circulating immunoglobulins in 496 patients revealed that fucosylation of IgG is associated with the development of esophageal cancer, while mannosylation is linked to the absence of esophageal cancer.⁶⁴ Similar to the potential carcinogenic role of certain generated oncometabolites, the authors propose that these glycosylation patterns may either enhance or inhibit antibody-dependent cellular cytotoxicity toward the cancer cell.⁶⁴ However, the question as to whether the tumor promotes alterations in glycosylation as an immune-evasive strategy or if the changes in glycosylation facilitate carcinogenesis remains elusive.

Studies investigating surface proteins on exosomes have successfully differentiated various gastroesophageal cancer-specific proteins, including CD36, TENM2, MUC1, DSC1, ITGA1, ANXA1, CDH13, ESAM, TIMP2, LAG3, and NES.⁶⁸ Moreover, overexpression of CD54 on exosomes has been identified as a predictor of gastroesophageal cancer proliferation, migration, and invasion.⁶⁹ This observation further emphasizes the active role of exosomes in creating a cancer-promoting environment, both locally and distantly. Although conducted on tumor tissue, a study by Liu and colleagues demonstrated loss of esophagus-specific proteins such as KRT4, KLK13, KRT78, SPINK5, SPINK7, and CRNN, along with an increase in protein phosphorylation.⁷⁰ These changes in the cancer surrounding stromal tissue can result not only in concomitant alterations in the composition of blood biomarkers, detectable through the use of liquid chromatography/mass spectrometry, but also suggest a potential functional role of these proteins and metabolites in the tumor microenvironment.

Table 2 summarizes the current modalities used for the detection of esophageal cancer and Barrett’s esophagus in different mediums, along with their performance.

Table 2.

Overview of the different minimally invasive modalities for detection of gastroesophageal oncometabolites

	Technology	Oncometabolites	Diagnostic accuracy	Refs
Gastroesophageal cancer
Breath	SIFT-MS	Pentanoic acid Hexanoic acid Phenol Methyl phenol Ethyl phenol Butanal Pentanal Hexanal Heptanal Octanal Nonanal Decanal	Multivariable logistic regression Sensitivity: 98% Specificity: 92% AUC: 0.97	8
	PTR-MS	Not further specified: m/z 136 m/z 34 m/z 63 m/z 27 m/z 95 m/z 107 m/z 45	Stepwise discriminant analysis Sensitivity: 86% Specificity: 90% AUC: 0.94	40
	SIFT-MS	Butyric acid Pentanoic acid Hexanoic acid Butanal Decanal	Multivariable logistic regression Sensitivity: 80% Specificity: 81% AUC: 0.85	13
Urine	SIFT-MS	Acetaldehyde Acetone Methanol Hexanoic acid Acetic acid Hydrogen sulfide Phenol	Sum concentrations Sensitivity: NA Specificity: NA AUC: 0.90	55
Gastric content	SIFT-MS	Acetone Formaldehyde Acetaldehyde Hexanoic acid Hydrogen sulfide Hydrogen cyanide Methyl phenol	Sum concentrations Sensitivity: NA Specificity: NA AUC: 0.90	54
Plasma	Whole-genome sequencing	DNA methylation profiles	Deep-learning interface Sensitivity: 13%-17% Specificity: 99%	15,48
	SIFT-MS	Acetonitrile Acrylonitrile Carbon disulfide Isoprene 1-heptene 3-methylhexane 2-nonene Hydrogen sulfide triethylamine	Multivariable logistic regression Sensitivity: 84% Specificity: 90% AUC: 0.83	66
	LC-MS	Lactate Valine Leucine Methionine Carnitine Tyrosine Tryptophan Myristic acid Margaric acid Linolenic acid Linoleic acid Pyroglutamic acid	Principal component analysis Sensitivity: 77% Specificity: 86% AUC: 0.82	67
	Nano-flow cytometry and PCR	CD54 CD36 TENM2 MUC1 DSC1 ITGA1 ANXA1 CDH13 ESAM TIMP2 LAG3 NES	AUC: 0.985	65,68
Barrett’s esophagus
Breath	E-nose	No individual metabolites detected	Artificial neural network: Sensitivity: 90% Specificity: 74% AUC: 0.91	45
	E-nose	No individual metabolites detected	Artificial neural network: Sensitivity: 82% Specificity: 80% AUC: 0.79	56
Plasma	Whole-genome sequencing	Loss of heterozygosity	NA

AUC, area under the curve; E-nose, electronic nose; LC-MS; liquid chromatography/mass spectrometry; PTR-MS, proton transfer reaction mass spectrometry; SIFT-MS, selected ion flow tube mass spectrometry.

Future investigations should focus on unraveling the pathophysiologic role and genomic pathways associated with these oncometabolites as well as developing a minimally invasive diagnostic platform

While these recent studies demonstrate high diagnostic performance and predictive value, it will be crucial to conduct external validation of these results through multicenter collaborations including diverse demographic patient populations. Furthermore, leveraging the current advancements in artificial intelligence, these collaborations can facilitate the identification of universal oncometabolic biomarkers or profiles,^71,72 thereby supporting necessary paradigm shifts in early detection, treatment assessment, and follow-up. Such paradigm shifts will be critical to address important shortcomings in the treatment of patients with gastroesophageal cancer and improve the cost-effectiveness of modern health care systems, particularly in light of escalating health care costs associated with cancer treatments.⁷³ Anticipating the use of these technologies in settings with varying disease prevalences also necessitates the adoption of tailored metrics to evaluate their performance and clinical applicability. In the context of modern screening, the primary goal of diagnostic tests should be to minimize false-negative results while maintaining low false-positive rates. Traditional metrics such as sensitivity, specificity, and accuracy remain valuable tools; however, additional metrics like the F1 score may provide further insights into performance of diagnostic tests. The F1 score [$F1=\frac{2\times TruePositives}{2\times TruePositives+FalsePositives+FalseNegatives}$] combines precision (the ratio of correctly classified positive samples to all samples classified as positive) and recall (the rate of positive samples correctly classified), offering a balanced evaluation of true-positive detection.⁷² This approach reduces the influence of the more abundant true-negative cases, making it less dependent on disease prevalence. Such metrics are particularly valuable in the setting of high gastroesophageal cancer prevalence, such as in Asian populations or for monitoring disease recurrence and progression. Similarly, alternative measures like Matthew’s correlation coefficient, calibration curve analysis, and decision-curve analysis account for disease prevalence and should play a more prominent role in the future evaluation of diagnostic tools.⁷⁴ These advanced metrics enable a more nuanced understanding of diagnostic performance, ensuring applicability across diverse clinical settings.

In addition, future research should focus on elucidating the pathophysiologic and genomic pathways associated with the development of these oncometabolites. These insights will enhance our understanding of the function of these oncometabolites in cancer biology, cell signaling, and their interaction with the tumor microenvironment.⁷⁵ Moreover, the potential synergistic effects of the genomic, proteomic, and breathonomic analyses can contribute to the development of a pan-cancer minimally invasive diagnostic platform. Such a platform can have several promising applications, including screening of the general population. Subsequently, more invasive tests (i.e. endoscopy, imaging, biopsy) can be used for confirmation and tissue acquisition to support further molecular and genomic analysis.

Conclusion

Minimally invasive diagnostic approaches, such as detecting oncometabolites or circulating tumor DNA in exhaled breath or blood, are poised to become essential components of contemporary cancer assessment and care. These cutting-edge technologies pave the way for patient-tailored treatment strategies that can improve both quality of life and cancer-related survival. A comprehensive understanding of these techniques in combination with modern analytical tests enables clinicians to choose the most suitable approach for specific indications and interpret results with greater precision.