Introduction
Upper gastrointestinal adenocarcinomas, comprising esophageal adenocarcinoma (EAC) and gastric cancer (GC), remain among the leading causes of cancer-related mortality worldwide despite substantial advances in endoscopic imaging, surgical techniques, and systemic therapies.1 The poor prognosis associated with both malignancies is largely attributable to delayed diagnosis, as early-stage disease is frequently asymptomatic, and many patients present with locally advanced or metastatic cancer. Five-year survival ranges from 49% to 77% when EAC and GC are detected at an early stage but declines dramatically to 5–8% following regional or distant spread, highlighting the importance of effective strategies for early detection and risk stratification.2,3
Both cancers develop through well-characterized precursor lesions that provide a window of opportunity for intervention. EAC typically arises through the intestinal metaplasia–dysplasia–carcinoma sequence, whereas intestinal-type GC develops through the Correa cascade, progressing from chronic Helicobacter pylori–associated gastritis to atrophic gastritis, gastric intestinal metaplasia (GIM), dysplasia, and invasive adenocarcinoma.4 Current surveillance strategies rely primarily on endoscopic evaluation with local biopsy and histopathologic assessment; however, these approaches are invasive, resource-intensive, susceptible to sampling error, and associated with considerable interobserver variability, limiting their effectiveness for population-based screening and longitudinal surveillance.5-7 These limitations have stimulated intense interest in molecular biomarkers capable of identifying high-risk lesions before the development of invasive malignancy.
Among the molecular alterations implicated in upper gastrointestinal carcinogenesis, aberrant DNA methylation has emerged as one of the earliest and most reproducible epigenetic events. Promoter CpG island hypermethylation can silence tumor suppressor genes involved in cell-cycle regulation, apoptosis, DNA repair, epithelial differentiation, and Wnt signaling, thereby contributing to neoplastic progression while remaining detectable in tissue, cytology specimens, and circulating cell-free DNA (cfDNA).8,9 Unlike many genetic mutations that occur later during tumor evolution, methylation changes frequently precede histopathologic metaplasia or dysplasia, making them attractive candidates for early detection, risk prediction, and minimally invasive surveillance. Advances in high-throughput methylation profiling, liquid biopsy technologies, and artificial intelligence have accelerated the discovery of increasingly accurate methylation signatures for both EAC and GC.10,11
Several recent reviews have comprehensively summarized DNA methylation biomarkers in GC and Barrett’s-associated neoplasia.12,13 However, these reviews primarily evaluate EAC and GC as separate disease entities and focus predominantly on cataloging candidate biomarkers. Comparatively little attention has been given to the shared epigenetic mechanisms that underlie upper gastrointestinal carcinogenesis, the overlap between methylation biomarkers identified across both organ systems, the comparative strengths and limitations of emerging methylation detection technologies, and the practical challenges that currently limit clinical implementation. Furthermore, rapid advances in minimally invasive sampling platforms, cfDNA methylation profiling, and multicancer early detection (MCED) assays have substantially expanded the translational landscape since many earlier reviews were published, warranting an updated synthesis of the field.
Accordingly, this review provides a comprehensive and comparative evaluation of DNA methylation biomarkers across Barrett’s esophagus (BE), EAC, GIM, and gastric adenocarcinoma. Rather than considering these malignancies independently, we examine their shared and disease-specific epigenetic mechanisms, compare emerging tissue-based, nonendoscopic, and liquid biopsy methylation platforms, critically evaluate barriers to clinical translation—including assay standardization, regulatory approval, and implementation—and discuss future directions for integrating methylation biomarkers into precision screening and surveillance strategies for upper gastrointestinal adenocarcinomas.
EAC and BE
Definitions and clinical context
Esophageal cancer comprises two major histologic subtypes: EAC and esophageal squamous cell carcinoma. Since the 1980s, the incidence of EAC has increased approximately six-fold in North America, representing one of the most rapidly rising malignancies in Western countries.14 More recent epidemiologic data suggest that although incidence continues to remain substantially elevated compared with historical levels, growth rates have begun to plateau in several Western populations, likely reflecting improvements in reflux management and changes in obesity trends. The 5-year survival rate for EAC is only 18–22% because most patients present at an advanced stage.15 The single major risk factor for the development of EAC is BE, in which intestinal-like glandular epithelium replaces the normal squamous mucosa of the esophagus.16,17 BE is a consequence of chronic gastroesophageal reflux disease (GERD).17,18 Roughly 15–40% of adults in Western countries have GERD, while 8–20% of adults with GERD develop BE.19,20 Patients with BE have a substantially increased risk of EAC compared with the general population; large population-based cohorts have reported annual EAC progression rates as low as approximately 0.12%, with risk estimates varying according to dysplasia status and cohort definition.16-18,21 Current guidelines generally recommend endoscopic surveillance every 3–5 years for patients with BE in order to identify dysplasia or early cancer.5,18 However, the limited cost-effectiveness and inconsistent patient compliance associated with these guidelines have resulted in a substantial number of EAC cases being diagnosed at advanced stages. Consequently, only a small fraction (10–15.9%) of EAC patients undergo regular BE surveillance.22
Current surveillance paradigm for BE
Currently, surveillance of individuals with BE relies on esophagogastroduodenoscopy (EGD) with systematic biopsy, which aims to detect dysplasia or early carcinoma before symptom onset. The American Society for Gastrointestinal Endoscopy guidelines recommend that patients with nondysplastic BE undergo surveillance EGD at regular intervals—typically every 3–5 years in the absence of dysplasia—using a systematic biopsy sampling protocol (Seattle protocol) with four-quadrant biopsies every 1–2 cm along the Barrett’s segment.5,23,24 More frequent surveillance is advised for those with confirmed low- or high-grade dysplasia (HGD), and targeted biopsies of any visible mucosal irregularities are also obtained during each procedure. Endoscopic assessment and histologic grading of dysplasia drive clinical decision-making, with treatment options such as endoscopic resection or radiofrequency ablation being effective in reducing progression for high-grade lesions.24,25 However, this surveillance paradigm is limited by sampling error and interobserver variability, resulting in suboptimal risk stratification,26 and many patients with a BE diagnosis follow divergent disease courses.27 There is an urgent need for clinically reliable biomarkers that can accurately stratify BE patients by their risk of progression to EAC. Despite advances in endoscopic imaging, population-based screening for BE remains limited by the invasive nature, cost, and relatively low diagnostic yield of conventional upper endoscopy. As highlighted by Lao-Sirieix and Fitzgerald, these limitations have driven the development of minimally invasive screening strategies capable of identifying BE in larger at-risk populations while maintaining acceptable diagnostic accuracy and patient acceptability.28
The Cytosponge is a nonendoscopic cell-collection device consisting of a swallowable capsule containing a compressed mesh sponge attached to a string; after the capsule dissolves in the stomach, the expanded sponge is withdrawn to collect esophageal cells for analysis.29-33 When combined with trefoil factor 3 (TFF3) immunohistochemistry, the Cytosponge-TFF3 test demonstrates high sensitivity (91.5%) for detecting BE, with a landmark randomized controlled trial showing a 10-fold increase in Barrett’s detection compared with usual care in primary care patients on acid-suppressant therapy.5,34,35 The device exhibits excellent safety, tolerability, and patient acceptability. Clinical translation of nonendoscopic sampling devices accelerated following the BEST2 trial, in which Cytosponge coupled with TFF3 immunohistochemistry demonstrated high sensitivity and specificity for detecting BE in a multicenter case-control cohort. This landmark prospective study established the feasibility of combining minimally invasive cell collection with molecular biomarker assessment and has served as the foundation for subsequent incorporation of DNA methylation biomarkers into Cytosponge-based diagnostic platforms.36 EsophaCap has been evaluated in clinical studies combining cytology and MUC2 immunohistochemistry, demonstrating reasonable sensitivity and high specificity for detecting intestinal metaplasia, dysplasia, and early EAC when compared with standard biopsies.37 When combined with panels of methylated DNA biomarkers, EsophaCap has demonstrated the ability to collect sufficient cytologic material to distinguish BE and early neoplastic changes with high sensitivity and specificity in prospective validation studies.38-40 Another class of devices, such as balloon-based esophageal samplers, similarly retrieve esophageal cells for analysis of methylated DNA markers like those used in commercially available assays, showing strong clinical utility and high rates of successful sample collection in real-world cohorts.41 These technologies illustrate how cytology-based, nonendoscopic tools may augment traditional EGD surveillance, particularly in settings where endoscopy access is limited or for prescreening at the point of care.
Methylation markers in EAC as molecular signatures of malignancy
Epigenetic dysregulation, particularly aberrant DNA methylation, is now firmly established as a central molecular hallmark of EAC and a critical driver of Barrett’s-associated neoplastic progression. CpG islands are CpG-rich genomic regions, typically located within gene promoters, whose methylation suppresses gene transcription and represents one of the principal mechanisms of epigenetic gene silencing. Early candidate-gene studies demonstrated progressive promoter hypermethylation of tumor suppressor genes such as CDKN2A and APC, establishing epigenetic silencing as a stable and early event in the transition from BE to invasive carcinoma (Table 1).8,10,40,42-47 Subsequent genome-wide methylation profiling revealed widespread epigenetic remodeling across the Barrett’s–dysplasia–EAC sequence, identifying distinct methylation signatures that differentiate normal squamous epithelium, nondysplastic BE, HGD, and EAC.10,43-46 These studies demonstrated that global CpG methylation alterations accumulate progressively and define molecular subclasses of disease.
| Approach | Key findings | Clinical relevance | Notes | References |
|---|
| Genome-wide methylation profiling | Distinct signatures across normal → BE → dysplasia → EAC | Disease classification | Defines molecular subtypes |
43,45,46
|
| Integrated methylome studies | Epigenetic subclasses of BE/EAC | Risk stratification | Links to transcriptome/genome | 10 |
| CpG island methylation accumulation | Progressive methylation changes | Early carcinogenesis | Stepwise progression model | 8,44 |
| Expanded methylome datasets | Novel biomarker discovery | Early detection potential | Enables nonendoscopic testing | 40,47 |
More recent large-scale discovery efforts have refined these observations by identifying restricted panels of CpG loci consistently hypermethylated in HGD and EAC, mapping to genes involved in transcriptional regulation, epithelial differentiation, and Wnt signaling pathways—including ZNF682, ZNF132, CDO1, and SFRP1.10,40,48 These loci form the basis of multimarker methylation panels that demonstrate strong diagnostic accuracy for distinguishing advanced neoplasia from nondysplastic BE. Importantly, ongoing translational work has extended these findings beyond tissue-based diagnostics. Analytical validation studies and multicenter case-control trials have demonstrated that methylated DNA marker panels retain high sensitivity and specificity when applied to minimally invasive, nonendoscopic sampling platforms such as EsophaCap and EsoCheck.38,49,50 In parallel, more studies continue to expand methylome-wide datasets, identifying additional biomarkers with potential utility for early detection and nonendoscopic screening of BE and EAC.40,47
Collectively, these advances position DNA methylation not only as a mechanistic hallmark of EAC progression but also as a clinically actionable biomarker platform capable of enabling nonendoscopic detection, risk stratification, and potentially surveillance optimization in BE. Whether aberrant DNA methylation functions primarily as a driver of carcinogenesis or as a biomarker of malignant transformation remains an area of active investigation. Functional studies demonstrate that promoter hypermethylation of tumor suppressor genes, including CDKN2A, APC, RUNX3, and secreted frizzled-related protein (SFRP) family members, can directly suppress transcription and alter cell-cycle regulation, apoptosis, and Wnt signaling, supporting a mechanistic contribution to neoplastic progression. However, many methylation alterations likely reflect broader epigenomic instability accompanying chronic inflammation and metaplastic evolution. Consequently, individual methylation events may serve both as biologically relevant contributors to carcinogenesis and as clinically useful biomarkers of disease progression.
Diagnostic methylation biomarkers for BE
Aberrant DNA methylation is an early and pervasive event in the BE–dysplasia–EAC sequence, making it an attractive platform for molecular diagnosis. Genome-wide profiling studies have demonstrated that nondysplastic BE already harbors extensive epigenetic reprogramming distinct from normal squamous epithelium, with additional methylation changes accumulating in dysplasia and invasive carcinoma (Table 2).38,43,44,51-53 One study tested CCNA1 DNA methylation as a BE biomarker in cytology brushings of the distal esophagus from 173 individuals with or without BE. CCNA1 DNA methylation demonstrated an area under the curve (AUC) of 0.95 for discriminating BE-related metaplasia and neoplasia cases from normal individuals. When VIM and CCNA1 were combined, the resulting two-biomarker panel was 95% sensitive and 91% specific. In balloon samples from 86 individuals, tests of CCNA1 plus VIM DNA methylation detected BE metaplasia with 90.3% sensitivity and 91.7% specificity.52 Wang et al.38 also studied five methylation biomarkers (p16, HPP1, NELL1, TAC1, and AKAP12), comparing BE with controls. The AUC was 0.894, with a sensitivity of 94.4% and specificity of 62.2% in the training set and a sensitivity of 78.6% and specificity of 92.8% in the test set.38 Another 5-marker methylated DNA panel, including ZNF682, VAV3, NDRG4, BMP3, and ZNF568, was tested in 199 patients in the training group and 89 patients in the test group. The sensitivity of the 5-marker panel for BE diagnosis was 93% at 90% specificity in the training set and 93% at 93% specificity in the test set. AUCs were 0.96 and 0.97 in the training and test sets, respectively.53 These methylation signatures retain performance when applied to minimally invasive, nonendoscopic platforms such as EsophaCap, EsoCheck, and capsule sponge devices, supporting their feasibility for large-scale screening and surveillance applications (Table 3).31,32,37-39,41,49,52,53 Collectively, these studies establish methylation panels as robust molecular tools for diagnosing BE beyond the limits of histopathology alone.
| Biomarker / Panel | Sample type | Clinical use | Performance | Key notes | References |
|---|
| CCNA1 | Cytology brushings | BE detection | AUC 0.95 | Strong single biomarker | 52 |
| CCNA1 + VIM | Cytology / balloon | BE detection | Sens 90–95%, Spec ~91% | Validated in nonendoscopic samples | 52 |
| p16, HPP1, NELL1, TAC1, AKAP12 | Cytology (EsophaCap) | BE diagnosis | AUC 0.894; Sens 78.6–94.4%, Spec 62.2–92.8% | Multimarker panel | 38 |
| ZNF682, VAV3, NDRG4, BMP3, ZNF568 | Cytology | BE detection | Sens 93%, Spec 90–93%; AUC 0.96–0.97 | Highly validated panel | 53 |
| Platform | Biomarker(s) | Clinical use | Performance | Key notes | References |
|---|
| Cytosponge-TFF3* | TFF3 (IHC) ± methylation panels | BE detection | Sens ~91.5% | 10-fold increase in detection vs. usual care | 31,32 |
| EsophaCap | Cytology + methylation markers | BE/EAC detection | High sensitivity & specificity | U.S.-developed capsule device | 37-39 |
| EsoCheck | EsoCheck + Methylated DNA panel | BE/EAC detection | High diagnostic accuracy | Balloon-based sampling | 41,49 |
Methylation biomarkers for predicting high-risk BE progression
Importantly, many of the methylation alterations observed in EAC emerge earlier along the BE–dysplasia–adenocarcinoma sequence, indicating that epigenetic reprogramming precedes overt histologic malignancy.45 Recently, the role of epigenetic changes in the pathogenesis of BE and esophageal cancer was extensively studied (Table 4).10,45,51,54-62 Multiple methylation biomarkers have been identified that can discriminate between high-risk and low-risk BE, including APC/p16,59MGMT,55,58PKP1,51TIMP3/TERT,54RUNX3/HPP1,58 and OR3A4.60 Targeted genetic and epigenetic profiling has further confirmed frequent APC, CDKN2A, MGMT, and TIMP3 promoter hypermethylation in both nondysplastic BE and EAC.63 Dilworth et al.60 identified 44 methylation markers that may be able to discriminate between nondysplastic BE that either progresses to adenocarcinoma or remains static. Hypomethylation of tumor suppressor OR3A4 (probe cg09890332) was validated in a separate cohort of samples. For OR3A4, median methylation was 67.8% in progressors versus 96.7% in nonprogressors. Using receiver operating characteristic curve modeling, the AUC of this model was 0.70, and adjustment for a prevalence of 0.7% using a threshold of 58% demonstrated a sensitivity of 33.3% and specificity of 78.6%.60 Another notable example of a translational effort is Previse’s Esopredict test, a methylation-based prognostic assay developed to risk-stratify patients with BE according to their likelihood of progression to HGD or EAC within 5 years. Esopredict quantifies methylation at four loci—p16 (CDKN2A), RUNX3, HPP1 (TMEFF2), and FBN1—and incorporates patient age into a locked risk algorithm that generates a continuous risk score and categorical risk designation.61,62 In a multicenter clinical validation study of 240 patients with BE, the average 5-year risk of progression to HGD or EAC was 21.5% among patients classified as high risk, compared with 1.85% among those classified as low risk. Importantly, the approximately four-fold figure reported in the study refers to the high-risk group’s risk relative to the overall 5-year progression prevalence in the BE population, rather than to the low-risk group. Compared with patients in the low-risk category, those in the high-risk category were 11.6 times more likely to progress within 5 years.62 These findings support the potential utility of methylation-based risk stratification for distinguishing patients with substantially different risks of neoplastic progression, although prospective studies are needed to determine whether incorporating Esopredict into surveillance strategies improves clinical outcomes.
| Biomarker / Panel | Clinical use | Key findings | Performance | Notes | References |
|---|
| APC / p16 | BE progression risk | Early tumor suppressor methylation | Associated with progression | Foundational markers | 42,59 |
| MGMT | BE progression | Promoter methylation | Risk association | DNA repair pathway | 55,58 |
| PKP1 | BE progression | Aberrant methylation | — | Cell adhesion gene | 51 |
| TIMP3 / TERT | BE progression | Distinguishes high-risk BE | — | Telomere biology | 54 |
| RUNX3 / HPP1 | BE progression | Risk stratification | — | Frequently replicated | 58 |
| OR3A4 | BE progression | Hypomethylation in progressors | AUC 0.70; Sens 33%, Spec 79% | Limited standalone utility | 60 |
| Esopredict (p16, RUNX3, HPP1, FBN1 + age) | 5-year progression risk | High vs. low risk: 21.5% vs. 1.85% | Strong risk stratification | Clinically validated prognostic assay | 61,62 |
Summary: Emerging genomic and epigenomic data support a model in which epigenetic instability—particularly aberrant DNA methylation—arises early in Barrett’s metaplasia and persists throughout disease evolution, functioning as both a biomarker and a mechanistic driver of malignant transformation. While diagnostic methylation panels for detecting EAC and HGD compared with normal tissue or BE now demonstrate reproducibly high accuracy, biomarkers predicting progression from BE to EAC remain more challenging to develop, reflecting the biological heterogeneity of BE and the relatively low annual progression rate.
Gastric cancer biomarker review
Definitions and clinical context
Gastric adenocarcinoma is the fifth most common cancer worldwide, and demonstrates striking geographic variation in incidence and mortality.64-66 The highest age-standardized incidence rates—often exceeding 30 per 100,000 persons—are observed in Eastern Asia, Eastern Europe, and parts of South America. The disease shows a clear male predominance and a median age at diagnosis of 68 years in the United States. Globally, nearly 85% of tumors arise in the distal stomach (noncardia), whereas in North America and Western Europe, 30–40% originate in the cardia or gastroesophageal junction. Significant racial and ethnic disparities persist, with Asian, Hispanic, Black, and American Indian populations experiencing substantially higher incidence rates of noncardia gastric cancer compared with non-Hispanic White individuals. These epidemiologic patterns underscore the combined influence of environmental exposures, infection prevalence, and host susceptibility in disease distribution.64-66
H. pylori infection remains the dominant etiologic factor, accounting for an estimated 75–89% of noncardia gastric cancers. Classified as a World Health Organization Group 1 carcinogen, H. pylori infects approximately 43.9% of adults globally (about 17.6% in the United States), yet only 1–3% of infected individuals ultimately develop malignancy.67-69 Carcinogenic progression is driven by bacterial virulence factors—particularly CagA and VacA—host genetic susceptibility, and environmental influences. The Correa cascade describes the well-established sequence from chronic active gastritis to atrophic gastritis, intestinal metaplasia, dysplasia, and ultimately invasive adenocarcinoma.67,68 Emerging data further highlight synergistic interactions between H. pylori infection and germline pathogenic variants in homologous recombination genes such as BRCA1, BRCA2, ATM, and PALB2; infected carriers may face a markedly elevated lifetime risk compared with uninfected carriers.68 Although approximately 10% of cases demonstrate familial aggregation, recognized hereditary cancer syndromes account for only 1–3%, including hereditary diffuse gastric cancer due to CDH1 mutations and Lynch syndrome. Additional risk factors include pernicious anemia, prior gastric ulcer disease, advanced age, lower socioeconomic status, and immigration from high-incidence regions, collectively reflecting the multifactorial pathogenesis of gastric adenocarcinoma.64-66
Current screening and prevention
Standard clinical surveillance for GC relies primarily on endoscopic evaluation with systematic biopsy in high-risk populations, combined with serologic screening when appropriate. Guidelines for gastric cancer screening vary among countries and high-risk populations. The fundamental difference between US and Asian guidelines lies in their population-based screening strategies. Japan and South Korea implement universal screening programs for all adults starting at age 40–50 years with biennial endoscopy or radiography, with Japan starting screening at age 50 and South Korea at age 40, while the US employs a risk-stratified approach targeting only high-risk individuals.3,70-75 The American Gastroenterological Association (AGA) Clinical Practice Update recommends high-quality endoscopy with systematic biopsies for individuals at elevated risk, such as those with extensive GIM, family history of GC, or first-generation immigrants from high-incidence regions, with intervals typically ranging from 3–5 years depending on histologic findings.3
H. pylori testing and eradication are critical components of both primary and secondary prevention strategies, given its role as the dominant etiologic factor for GC, particularly in East Asian populations, where prevalence rates remain high and H. pylori infection accounts for the majority of sporadic GC cases.3,67-69,73,76,77 Epidemiologic studies indicate that, although eradication programs and declining infection rates have begun to reduce GC incidence in younger cohorts, the overall disease burden in Asian countries such as Japan, Korea, and China remains substantially higher than in Western populations.68,73,77,78 Therefore, surveillance strategies in these regions emphasize endoscopic screening combined with risk-stratified H. pylori management to detect premalignant lesions early and prevent progression to carcinoma.3
Traditional biomarkers for screening
Traditional serologic biomarkers for gastric cancer screening primarily include serum pepsinogen I (PGI), pepsinogen II (PGII), the PGI/PGII ratio (PGR), carcinoembryonic antigen (CEA), CA19-9, CA72-4, gastrin-17 (G-17), and H. pylori IgG antibodies, which are used to detect atrophic gastritis and stratify gastric cancer risk, particularly in high-incidence Asian populations (Table 5).69,70,73,76,79-89 Reduced PGI (≤70 μg/L) and PGR (≤3.0) reflect corpus atrophy and form the basis of pepsinogen screening; pooled Japanese data demonstrate sensitivities of 59–77% and specificities of 73–89% for gastric cancer detection, while meta-analyses report a summary sensitivity of approximately 0.69 and specificity of 0.73 (AUC ~0.76) for cancer and higher accuracy for atrophic gastritis (AUC ~0.85).81,82 Pepsinogen testing performs best for noncardia gastric cancer and was associated with an 8–11-fold increased cancer risk in pepsinogen-positive individuals in a US PLCO-based study, suggesting potential value even in Western populations.90,91 Gastrin-17 reflects antral function and shows a J-shaped association with cancer risk, with both low and high levels associated with increased risk; when combined with PGI and PGII, diagnostic performance improves substantially (AUC up to 0.93 in some studies).76,80,83H. pylori serology identifies current or past infection—the dominant etiologic factor for noncardia cancer—and is most effective when combined with pepsinogen testing in the ABC method, which stratifies individuals into stepwise risk groups and confers a 6–8-fold higher cancer risk among those with serologic atrophy.68,69,73,78,84,85 Expanded multimarker “serological biopsy” panels incorporating PGI, PGII, PGR, G-17, and H. pylori antibodies further improve discrimination (C-statistic ~0.80) for precancerous lesions and future cancer risk.86 However, these biomarkers have important limitations, including variability by ethnicity, tumor subtype, medication use, and infection status, and they are not routinely recommended for population screening in low-incidence regions such as the United States due to inconsistent validation and limited availability.3,69 CEA, CA19-9, and CA72-4 have limited diagnostic utility for gastric cancer. When used individually at standard cutoff values, CEA demonstrates 20–28% sensitivity, CA19-9 shows 27–39% sensitivity, and CA72-4 achieves 28–43% sensitivity.87-89 While combination testing improves sensitivity to 48–75%, this remains inadequate for screening purposes, and these markers are most frequently elevated in advanced, incurable disease rather than early-stage cancer. These markers have greater value for prognosis and monitoring recurrence than for initial diagnosis.87-89
| Biomarker | Biological basis | Clinical use | Performance | Key notes | References |
|---|
| Pepsinogen I (PGI) | Corpus gland function | Detect atrophic gastritis | Sens 59–77%, Spec 73–89% | Low in corpus atrophy | 81,82 |
| Pepsinogen II (PGII) | Gastric inflammation | Used with PGI | — | Elevated in inflammation | 81,82 |
| PGI/PGII Ratio (PGR) | Gastric atrophy index | GC risk stratification | AUC ~0.76 (GC), ~0.85 (atrophy) | ≤3 suggests atrophy | 81,82 |
| Gastrin-17 (G-17) | Antral function | GC risk | AUC up to 0.93 (combined) | J-shaped association | 76,80,83 |
| H. pylori IgG | Infection status | Risk stratification | — | Used in ABC method | 69,73,84,85 |
| CEA | Tumor antigen | Prognosis/monitoring | Sens 20–28% | Poor early detection | 88,89 |
| CA19-9 | Tumor marker | Prognosis | Sens 27–39% | Limited screening utility | 87,88 |
| CA72-4 | Tumor marker | Prognosis | Sens 28–43% | Elevated in advanced disease | 88,89 |
| Multimarker panels | Combined markers | Risk prediction | C-stat ~0.80 | Improved discrimination | 86 |
DNA methylation biomarkers for gastric cancer
Although traditional models of gastric carcinogenesis emphasize the multistep Correa cascade—from chronic gastritis to atrophic gastritis, intestinal metaplasia, dysplasia, and carcinoma—recent genomic and epigenomic studies indicate that epigenetic alterations, particularly aberrant DNA methylation, arise early and persist throughout disease evolution.12,13,92-95 Genome-wide and targeted methylation profiling demonstrates that premalignant gastric lesions, including intestinal metaplasia, already harbor extensive DNA methylation changes distinct from normal gastric mucosa, supporting a model in which epigenetic dysregulation is a foundational driver of malignant transformation rather than a late event.96
cfDNA released from tumor cells into the bloodstream provides a noninvasive source for detecting epigenetic alterations associated with cancer. DNA methylation changes occur early in gastric carcinogenesis and are often more frequent than somatic mutations, making them attractive biomarkers for early cancer detection. Recent studies have demonstrated that cfDNA methylation profiling can detect gastric cancer with reasonable accuracy (Table 6).13,95-108
| Biomarker | Sample type | Clinical use | Performance | Key notes | References |
|---|
| SEPT9 | Plasma cfDNA | GC detection | Sens 48–61%, Spec 86–87% | Not GC-specific | 99,100 |
| SFRP2 | Plasma/serum | GC detection | Sens up to ~70% | Wnt signaling gene | 102,103 |
| SFRP2 + RPRM (Reprimo) | Plasma/serum | GC detection | Sens 57–79%, Spec >90% (combo) | Cell cycle regulator | 105 |
| RNF180 | Plasma | GC detection | — | Tumor suppressor | 101,104 |
| PCDH10 | Serum | GC detection | Detected in 12.9% of GC sera; 0% in controls | Low serum detection despite frequent tissue methylation | 105 |
| APC | Serum | GC detection | Detected in 6.25% of GC sera; 0% in controls | Low serum detection; limited standalone sensitivity | 105 |
| ELMO1 + ZNF569 + C13orf18 | Plasma | GC detection | Sens 86% (95% CI 71–95%), Spec 95% | Three-marker plasma methylated DNA panel | 106 |
GHR +
GLRB ± GATM | Plasma (dPCR) | Early GC detection | Sens 83–87%, Spec ~90% | Strong early detection | 107 |
| FGFR2 | Blood leukocyte DNA | GC vs. IM | Sens 85%, Spec 80% | Nontumor DNA source | 108 |
In gastric cancer, plasma methylated SEPT9 has demonstrated sensitivities ranging from approximately 48–61% with specificities around 86–87%, depending on assay algorithms. However, SEPT9 methylation is not specific to gastric cancer, as it is also frequently hypermethylated in colorectal, pancreatic, and esophageal cancers, limiting its diagnostic specificity for gastric malignancy.99,100,109 Methylation of SFRP2, RNF180, RPRM, APC, and PCDH10 has been detected in serum or plasma of gastric cancer patients and is associated with tumor presence and progression.101-105 A dual-gene panel of SFRP2 and RPRM demonstrated sensitivities up to 57–79% with specificities exceeding 90% in distinguishing gastric cancer from controls.105 In a separate plasma pilot study, a three-marker methylated DNA panel comprising ELMO1, ZNF569, and C13orf18 detected 86% of gastric adenocarcinomas (95% confidence interval (CI), 71–95%) at 95% specificity.106 More recently, a plasma-based digital polymerase chain reaction (PCR) assay targeting methylated GHR and GLRB achieved 83.3% sensitivity and 90% specificity, and adding GATM as a third marker increased sensitivity to 86.7% overall and 81.6% for stage I disease while maintaining high specificity.107 FGFR2 methylation in blood leukocytes demonstrated 85% sensitivity and 80% specificity for distinguishing gastric cancer from intestinal metaplasia patients.108
Comprehensive genome-scale methylation analysis using methods such as methylated CpG tandem amplification and sequencing (MCTA-Seq) has identified extensive panels of methylation biomarkers (Table 7).109-114 One study identified 153 cfDNA methylation biomarkers, including DOCK10, CABIN1, and KCNQ5, achieving sensitivities of 44%, 59%, 78%, and 100% for stages I, II, III, and IV, respectively, at 92% specificity.110 Another genome-wide approach identified 21 differentially methylated regions, genomic regions demonstrating statistically significant methylation differences between normal and diseased tissues, that enabled gastric cancer detection with 93.9% sensitivity and 95.2% specificity in the discovery set, and 88.4% sensitivity and 94.2% specificity in validation.111 MCED approaches based on cfDNA methylation sequencing have also demonstrated the ability to identify tumor-specific methylation signatures and classify cancer types with high accuracy.11 A meta-analysis of 32 studies including 4,172 patients found that blood-based DNA methylation testing had an overall sensitivity of 57% and specificity of 97% for gastric cancer detection.114 Plasma-based testing showed superior sensitivity (71%) compared with serum-based testing (50%), though with lower specificity (89% vs. 98%). Multigene panels achieved 76% sensitivity at 85% specificity, significantly outperforming single markers. For early-stage (TNM I+II) disease, methylation testing achieved 55% sensitivity at 96% specificity.114 The GutSeer assay, combining DNA methylation and fragmentomics across 1,656 markers, demonstrated 65.3% sensitivity for gastric cancer specifically, with overall gastrointestinal cancer detection sensitivity of 82.8% at 95.8% specificity.112
| Panel / Approach | Markers | Sample | Clinical use | Performance | Key notes | References |
|---|
| MCTA-Seq–derived 153-marker cfDNA panel | DOCK10, CABIN1, KCNQ5 | Plasma | GC detection by stage | Sens 44% (I), 59% (II), 78% (III), 100% (IV); Spec 92% | Genome-scale cfDNA methylation analysis | 110 |
| 21 DMR panel | Differential regions | Plasma | GC detection | Sens 88–94%, Spec 94–95% | High accuracy | 111 |
| GutSeer assay | 1,656 markers | Plasma | Multicancer detection | GC Sens 65.3% | Includes fragmentomics | 112 |
| Meta-analysis panels | Multigene | Blood | GC detection | Sens 76%, Spec 85% | Superior to single markers | 114 |
Methylation DNA biomarkers for high-risk GIM
GIM is a key precancerous lesion in the Correa cascade leading to gastric adenocarcinoma. However, only a subset of intestinal metaplasia lesions progress to cancer, creating a critical need for molecular biomarkers that stratify high-risk patients. Recent advances in genomics, epigenomics, and transcriptomics have identified several candidate biomarkers reflecting early malignant transformation.
CpG island hypermethylation and genomic DNA hypomethylation are found not only in gastric cancers but also in associated premalignant lesions. Aberrant DNA methylation represents one of the earliest molecular alterations in gastric carcinogenesis, occurring even before histologically detectable neoplastic changes. H. pylori infection induces aberrant CpG island hypermethylation in gastric mucosa. The timing of methylation varies by gene. THBS1 and TIMP3 show marked increases in hypermethylation frequency from chronic gastritis to intestinal metaplasia, while MLH1 and p16 methylation increases more prominently from metaplasia to cancer (Table 8).13,96,112,115-117 This temporal pattern suggests that different methylation events contribute to distinct stages of the carcinogenic cascade.
| Stage | Biomarkers | Methylation pattern | Key insight | References |
|---|
| Chronic gastritis → IM | THBS1, TIMP3 | Increasing hypermethylation | Early carcinogenic event | 115 |
| IM → Dysplasia/GC | MLH1, p16 | Increased methylation later | Late-stage progression | 115 |
| Intestinal metaplasia (IM) | RASSF1A, APC | Elevated methylation | IM resembles GC epigenetically | 117 |
| Incomplete IM subtype | Intergenic hypermethylation | Cancer-like profile | High-risk precursor | 116 |
| H. pylori-associated mucosa | Multiple CpG loci | Infection-driven methylation | Early driver of carcinogenesis | 96,115 |
| Advanced GC | Global CpG alterations | Widespread dysregulation | Defines tumor biology | 13,112 |
Recent genomic profiling reveals that incomplete (colonic-type) intestinal metaplasia exhibits distinct methylation patterns compared with complete intestinal metaplasia. Incomplete intestinal metaplasia demonstrates extensive intergenic hypermethylation resembling native antral mucosa, while complete intestinal metaplasia displays promoter hypermethylation of tumor suppressor genes with a more fully intestinalized epigenetic profile. Incomplete intestinal metaplasia represents a phenotypically unstable, epigenetically deregulated state with molecular resemblance to gastric cancer, establishing it as a true precursor lesion.116
RIMS1 has emerged as one of the most robustly validated single-gene methylation markers for gastric cancer risk stratification (Table 9).117-121 A prospective multicenter study of 1,624 patients after H. pylori eradication demonstrated that those in the highest quartile of RIMS1 methylation had a gastric cancer incidence rate of 972.8 per 100,000 person-years compared with 127.1 in the lowest quartile (adjusted hazard ratio (HR) 5.7, 95% CI 1.3–25.5). A methylation cutoff of 25.7% identified a “super-high-risk” population requiring more frequent screening than currently recommended.118
| Biomarker | Clinical context | Key findings | Performance | Notes | References |
|---|
| RIMS1 | Post–H. pylori eradication | High methylation → high GC risk | HR 5.7 | Strong prospective evidence | 118 |
| miR-124a-3 methylation | Metachronous GC risk | Increased recurrence risk | HR 2.30 | Validated cohort | 119 |
| EMX1 | GC progression | Risk association | — | Often combined | 119 |
| NKX6-1 | GC progression | Combined risk model | HR ~3.1 | Multimarker model | 120 |
| RPRM | IM → GC progression | Progressive hypermethylation | — | Early transformation marker | 117 |
| ZNF793 | IM progression | Hypermethylation | — | Early epigenetic shift | 117 |
| IGF2 (blood DNA) | Preclinical detection | Elevated ≥5 years prior | OR 12.5 | Strong early signal | 121 |
| N33 (blood DNA) | Early detection | Elevated in progressors | — | Noninvasive biomarker | 121 |
A prospective cohort study of 782 patients after endoscopic resection of early gastric cancer demonstrated that the highest quartile of miR-124a-3 methylation had a significant multivariate-adjusted HR of 2.30 (95% CI 1.03–5.10) for developing metachronous gastric cancers. Similar trends were observed for EMX1 and NKX6-1.119 In a Caucasian population, MIR124-3 hypermethylation was significantly associated with development of metachronous gastric lesions (adjusted HR 2.31, 95% CI 1.03–5.17), particularly in females and H. pylori-negative patients. NKX6-1 was found to be hypermethylated in patients with synchronous lesions. A molecular-based methylation model incorporating both genes was associated with a threefold increased risk for metachronous gastric lesion development (adjusted HR 3.10, 95% CI 1.07–8.95).120 In a 16-year Colombian cohort study, methylation levels of AMPH, PCDH10, RSPO2, SORCS3, and ZNF610 predicted progression of gastric lesions independent of H. pylori infection duration, baseline diagnosis, or inflammatory scores.122 Although these biomarkers demonstrated statistically significant associations with progression in the original longitudinal cohort, independent external validation remains limited, and their clinical utility has yet to be established. Blood-based markers also show potential. Hypermethylated IGF2 and N33 in blood leukocyte DNA were elevated at least 5 years before clinical gastric cancer diagnosis, with markedly increased frequency in intestinal metaplasia patients who progressed to cancer (odds ratio 12.52 for IGF2 in intestinal metaplasia progressors).121 Comprehensive epigenomic profiling of 138 intestinal metaplasia cases from a 10-year prospective study revealed that intestinal metaplasia patients with shortened telomeres and chromosomal alterations combined with hypermethylation signatures were associated with subsequent dysplasia or gastric cancer, while those with normal-like epigenomic patterns showed association with regression.123
Longitudinal studies confirm that intestinal metaplasia partially recapitulates patterns of aberrant methylation of intestinal-type gastric cancer, independently of H. pylori status. Genes such as RASSF1A and APC consistently showed increased methylation in intestinal metaplasia with respect to earlier precursor lesions. Importantly, H. pylori-dependent methylation in intestinal metaplasia suggests that eradication in late stages of precursor lesions may not prevent epigenome reprogramming toward a cancer signature.117
Clinical implementation considerations
Current AGA guidelines identify high-risk intestinal metaplasia patients based on incomplete (vs. complete) histology (3.3-fold relative risk), extensive (corpus-extended) vs. limited gastric intestinal metaplasia (2.1-fold relative risk), and family history of gastric cancer (4.5-fold relative risk).3 Methylation biomarkers provide an additional molecular layer for risk stratification that may complement these histologic and clinical factors. Several challenges remain for clinical implementation. Most reported methylation markers are in early discovery stages and have been evaluated primarily in tissue samples. Noninvasive blood-based assays lack standardized sampling protocols, preanalytical procedures, and multicenter validation. Many markers exhibit insufficient sensitivity for early-stage gastric cancer detection, and most are pan-cancer markers rather than gastric cancer-specific.13
Emerging prediction markers
Beyond DNA methylation, emerging work also supports combining methylation profiles with other biomarkers (e.g., microbiome indices) to enhance predictive performance, especially in H. pylori-negative GC cases, underscoring the complexity and heterogeneity of epigenomic influences in gastric tumorigenesis.124 Other nucleic acid-based biomarkers are under investigation for early GC detection. Blood microRNA signatures, including miR-124-3p and miR-125a-3p, have demonstrated high predictive performance for early detection and prognosis.94 Epigenetic alterations also intersect with other layers of gene regulation, including histone modifications and noncoding RNA dysregulation, which collectively shape transcriptional networks that drive proliferation, invasion, immune evasion, and resistance to therapy.92 Integrative epigenomic profiling continues to refine molecular taxonomy of GC, revealing subtypes with distinct prognostic and therapeutic implications and creating opportunities for precision risk stratification beyond traditional histopathology.13,92 When comparing performance, targeted methylation panels generally provide higher specificity, whereas microRNA and transcriptomic signatures may enhance sensitivity and enable multilayered risk stratification when combined with methylation markers.94,106,125
Shared epigenetic mechanisms across upper gastrointestinal adenocarcinomas
Although EAC and GC arise from distinct epithelial lineages and exhibit important differences in epidemiology, environmental risk factors, and clinical management, both malignancies develop through remarkably similar epigenetic mechanisms. Chronic inflammation induced by GERD in BE and H. pylori infection in the stomach initiates progressive epigenetic remodeling that accompanies the transition from metaplasia to dysplasia and ultimately invasive adenocarcinoma. These observations suggest that aberrant DNA methylation represents a common molecular hallmark of upper gastrointestinal carcinogenesis, while organ-specific methylation signatures reflect differences in tissue lineage, inflammatory microenvironment, and selective pressures during tumor evolution.
A defining feature shared by both diseases is the early establishment of widespread promoter CpG island hypermethylation. Genome-wide methylation profiling has demonstrated that extensive methylation abnormalities are already present in nondysplastic BE and GIM, well before the development of invasive carcinoma.8,10,13,43,117 These findings support a model in which epigenetic dysregulation is not merely a late consequence of malignant transformation but rather an early event that creates a permissive environment for neoplastic progression.
Several methylation targets appear consistently across both EAC and GC, suggesting that common biological pathways are disrupted despite differences in tissue origin. Promoter hypermethylation of CDKN2A (p16), APC, RUNX3, and members of the SFRP family has been repeatedly identified in Barrett’s-associated neoplasia as well as gastric carcinogenesis.62,92-95,112,126-130 These genes regulate fundamental cellular processes, including cell-cycle control, apoptosis, epithelial differentiation, and Wnt/β-catenin signaling, indicating that convergent epigenetic silencing of tumor suppressor pathways represents a shared mechanism underlying upper gastrointestinal adenocarcinoma development. Similarly, methylation of RASSF1A, TIMP3, and additional genes involved in extracellular matrix remodeling and genomic stability has been reported across multiple stages of both disease pathways, further supporting common mechanisms of inflammation-associated epigenetic reprogramming.131
Among these shared pathways, dysregulation of canonical Wnt signaling deserves particular attention. Aberrant methylation of APC, SFRP1, and SFRP2 results in loss of negative regulation of Wnt signaling, promoting epithelial proliferation, stem-cell expansion, and resistance to apoptosis. Because activation of the Wnt pathway is observed in both Barrett’s-associated carcinogenesis and gastric carcinogenesis, these methylation events likely represent fundamental drivers of upper gastrointestinal tumor initiation rather than organ-specific phenomena.130 Likewise, epigenetic silencing of RUNX3, an important mediator of transforming growth factor-β signaling and immune regulation, has been implicated in both diseases, suggesting that impaired differentiation and altered immune surveillance may represent another common mechanism of malignant transformation.36,58,105,127,132
Despite these similarities, important lineage-specific methylation patterns distinguish EAC from GC and likely reflect differences in epithelial origin and disease biology. Barrett’s-associated neoplasia demonstrates recurrent methylation of genes including ZNF682, ZNF132, VAV3, NDRG4, BMP3, PKP1, and OR3A4, many of which have shown promising diagnostic or prognostic performance in nonendoscopic sampling platforms such as Cytosponge, EsophaCap, and EsoCheck.133,134 In contrast, gastric adenocarcinoma exhibits frequent methylation of RNF180, RPRM, RIMS1, GHR, GLRB, FGFR2, and PCDH10, several of which have demonstrated encouraging performance in plasma cell-free DNA assays for early cancer detection.104,135-137 These organ-specific methylation signatures likely reflect differences in developmental lineage, exposure to distinct inflammatory stimuli, and divergent selective pressures imposed by reflux-associated versus H. pylori-associated carcinogenesis.
An additional shared feature of both malignancies is the phenomenon of epigenetic field cancerization. Chronic inflammatory injury induces widespread methylation alterations throughout histologically normal-appearing mucosa, creating an epigenetically altered field from which dysplastic lesions subsequently emerge. In BE, methylation abnormalities extend beyond areas of visible dysplasia and are detectable throughout the Barrett’s segment, supporting the concept of diffuse molecular field defects. Similarly, H. pylori infection produces persistent methylation changes throughout the gastric mucosa that frequently remain even after bacterial eradication, potentially explaining the continued risk of gastric cancer in patients with established intestinal metaplasia. This persistence of epigenetic alterations provides a biological rationale for long-term surveillance despite apparent histologic regression.
These common mechanisms have important implications for biomarker development. Rather than viewing BE and GIM as entirely separate diseases, their shared epigenetic architecture suggests that similar biomarker discovery strategies may be applicable across upper gastrointestinal adenocarcinomas. Genome-wide methylation profiling, machine learning-based biomarker selection, and minimally invasive sampling approaches have already demonstrated success in both disease settings. Furthermore, emerging MCED platforms, including those developed through the Circulating Cell-free Genome Atlas (CCGA) program and evaluated prospectively in the PATHFINDER study, leverage methylation signatures across multiple tumor types to simultaneously detect cancer and predict tissue of origin.11 Although current MCED assays demonstrate lower sensitivity for localized upper gastrointestinal cancers than for some other malignancies, continued refinement of organ-specific methylation classifiers may substantially improve their performance for EAC and GC.
Nevertheless, important challenges remain before a unified methylation-based approach to upper gastrointestinal cancer detection can be realized. Significant heterogeneity exists across published studies with respect to patient selection, specimen type (tissue, cytology, plasma, serum, or cell-free DNA), laboratory platform, methylation assay, reference standards, and statistical modeling. Consequently, direct comparison of reported sensitivities, specificities, and AUC values across studies should be interpreted cautiously, as differences often reflect methodological variation rather than true biomarker performance. Standardization of assay platforms, multicenter prospective validation, and integration of methylation biomarkers with established clinical risk factors and histopathologic staging systems will be essential before these assays can be incorporated into routine clinical practice.
Collectively, these observations support a conceptual framework in which DNA methylation represents a shared biological hallmark of upper gastrointestinal adenocarcinoma while preserving sufficient tissue-specific information to distinguish organ of origin and disease stage. Future biomarker development should therefore emphasize both the common molecular pathways linking reflux-associated and H. pylori-associated carcinogenesis and the lineage-specific methylation signatures that enable accurate diagnosis, risk stratification, and precision surveillance.
Clinical translation of DNA methylation biomarkers: Barriers to implementation and the path toward clinical adoption
The development of DNA methylation biomarkers for BE, EAC, and GC has progressed rapidly over the past decade, with several assays demonstrating excellent diagnostic accuracy in retrospective and prospective validation studies. Nevertheless, despite encouraging analytical performance, no methylation-based assay has achieved widespread adoption as the standard of care for screening, surveillance, or risk stratification in upper gastrointestinal malignancies. This disconnect highlights a fundamental challenge in translational biomarker research: demonstrating diagnostic accuracy alone is insufficient for clinical implementation. Successful translation requires robust evidence of analytical validity, clinical validity, clinical utility, cost-effectiveness, regulatory approval, reimbursement, and integration into existing healthcare systems.138-143
Esopredict represents one of the potential methylation assays developed for BE surveillance, incorporating multiple methylation biomarkers to identify patients at increased risk of dysplasia or progression to EAC. Early validation studies have reported encouraging discrimination between low-risk and high-risk Barrett’s lesions, suggesting that molecular risk stratification may complement conventional histopathologic assessment. The clinical validation study demonstrated substantial separation of progression risk across assay categories, but prospective evidence that assay-guided management improves outcomes remains limited.62 The current evidence remains insufficient to support replacement of established surveillance strategies based on high-definition endoscopy with systematic biopsy protocols. Most validation cohorts remain relatively small, enriched for high-risk populations, and originate from specialized tertiary referral centers, limiting generalizability to community practice. Large prospective multicenter studies demonstrating improved patient outcomes compared with current surveillance strategies are still lacking.
The major obstacle is the lack of prospective evidence demonstrating clinical utility.139,140 Most published studies focus on diagnostic performance metrics, including sensitivity, specificity, and AUC, rather than patient-centered outcomes. Regulatory agencies, clinical guideline committees, and healthcare payers increasingly require evidence that introducing a new biomarker changes clinical management, reduces cancer incidence or mortality, decreases unnecessary endoscopic procedures, or improves cost-effectiveness.141 Demonstrating statistical superiority over existing biomarkers is therefore insufficient if the assay does not clearly improve clinical decision-making. Consequently, randomized or pragmatic prospective studies comparing methylation-guided surveillance with current guideline-directed management represent an essential next step for clinical translation.142,143
Cost-effectiveness also remains a critical consideration. Current surveillance for BE relies on periodic upper endoscopy with systematic biopsy following the Seattle protocol, an approach that is invasive, resource-intensive, and associated with substantial interobserver variability in histopathologic interpretation.7,144 Although methylation assays have the potential to reduce unnecessary endoscopic procedures by identifying patients at genuinely elevated risk of progression, widespread implementation will depend on demonstrating that molecular testing provides meaningful economic value relative to existing surveillance strategies.25 Formal health economic analyses evaluating laboratory costs, downstream diagnostic procedures, treatment expenditures, and quality-adjusted life years remain limited. Without clear evidence of cost-effectiveness, reimbursement by public and private insurers is likely to remain inconsistent, substantially slowing adoption into routine clinical practice.
An equally important limitation is the absence of standardized analytical methodology across published methylation studies. Differences in specimen acquisition, DNA extraction protocols, bisulfite conversion efficiency, methylation detection platforms, bioinformatic processing, and statistical modeling contribute substantially to variability in reported diagnostic performance. Identical biomarkers frequently demonstrate different sensitivities and specificities depending on whether they are measured using methylation-specific PCR (MSP), pyrosequencing, droplet digital PCR (ddPCR), targeted bisulfite sequencing, or next-generation sequencing (NGS). Without standardized laboratory protocols and external quality assurance programs, reproducibility between institutions remains uncertain, limiting confidence in routine clinical implementation.9
Regulatory approval represents an additional hurdle. Before a methylation assay can be incorporated into evidence-based clinical guidelines, it must satisfy regulatory standards for analytical validity, reproducibility, and clinical performance. Incorporation of the biomarkers into professional society guidelines, including those issued by U.S. organizations such as the American College of Gastroenterology, AGA, American Society for Gastrointestinal Endoscopy, and European organizations such as European Society of Gastrointestinal Endoscopy, requires consistent evidence from multiple independent prospective studies demonstrating reproducible benefit across diverse patient populations. At present, methylation biomarkers are recognized as promising investigational tools but are not recommended as replacements for standard endoscopic surveillance.5,23,145
Implementation within routine pathology and gastroenterology practice presents further practical challenges. Adoption of molecular diagnostics requires laboratory infrastructure capable of performing standardized methylation analyses, specialized personnel trained in molecular pathology and bioinformatics, validated quality assurance protocols, and efficient workflows for integrating molecular results into electronic health records and clinical decision-making. These requirements may be readily achievable in academic medical centers but remain substantially more difficult in community hospitals and lower-resource healthcare systems, where most BE surveillance is performed. Ensuring equitable access to molecular diagnostics therefore represents an important consideration as these technologies mature.145
The emergence of MCED assays further illustrates both the promise and complexity of clinical translation. Large prospective studies, including the CCGA program and the PATHFINDER trial, have demonstrated that genome-wide methylation signatures can detect multiple malignancies simultaneously while accurately predicting tissue of origin.11,146 However, the sensitivity of current MCED platforms for early-stage EAC and GC remains lower than for several other solid tumors, underscoring the need for disease-specific methylation classifiers and continued refinement of upper gastrointestinal-focused biomarker panels. Rather than replacing organ-specific surveillance, these technologies may ultimately function as complementary screening tools that identify individuals requiring targeted endoscopic evaluation.
Future implementation will likely require integration of methylation biomarkers into multivariable clinical prediction models rather than reliance on molecular markers alone. Combining methylation profiles with established clinical risk factors—including age, sex, obesity, GERD duration, smoking history, Barrett’s segment length, histopathologic findings, and established risk stratification systems such as the Operative Link on Gastric Intestinal Metaplasia Assessment (OLGIM)—may improve predictive performance while facilitating individualized surveillance intervals. Such integrated models are more likely to gain acceptance within evidence-based clinical guidelines because they complement, rather than replace, existing standards of care.
Ultimately, the greatest challenge facing methylation-based assays is not whether they demonstrate promising diagnostic accuracy, but whether they can consistently improve patient outcomes within real-world healthcare systems. Future research should therefore prioritize multicenter prospective validation, standardized laboratory methodologies, cost-effectiveness analyses, regulatory evaluation, and implementation science alongside continued biomarker discovery. Addressing these barriers will determine whether methylation-based diagnostics evolve from promising investigational tools to routinely incorporated components of precision surveillance for upper gastrointestinal adenocarcinomas.
Technical platforms for DNA methylation analysis: Advantages, limitations, and clinical translation
The rapid expansion of DNA methylation biomarkers for EAC, BE, gastric adenocarcinoma, and GIM has been accompanied by significant advances in methylation detection technologies. However, differences in assay platforms represent a major source of heterogeneity across published studies and remain a significant barrier to clinical implementation. Variations in analytical sensitivity, quantitative accuracy, genomic coverage, DNA input requirements, cost, and laboratory standardization complicate direct comparisons between studies and contribute to the limited reproducibility of many candidate biomarkers. Consequently, understanding the strengths and limitations of individual methylation platforms is essential when interpreting published performance metrics and evaluating their suitability for clinical translation.9
Bisulfite-based DNA methylation analysis
Sodium bisulfite conversion remains the cornerstone of most DNA methylation assays. Bisulfite treatment selectively converts unmethylated cytosine residues to uracil while preserving methylated cytosines, allowing subsequent discrimination between methylated and unmethylated CpG sites by PCR amplification or sequencing.147 Many conventional targeted methylation assays rely on bisulfite-converted DNA. Therefore, the efficiency of this conversion step critically influences assay accuracy and reproducibility. Incomplete conversion may lead to false-positive methylation calls, whereas DNA degradation during treatment can reduce assay sensitivity, particularly when analyzing low-abundance cfDNA samples.9
Pyrosequencing
Pyrosequencing is one of the most widely used quantitative methods for validating candidate methylation biomarkers. Following bisulfite conversion and PCR amplification, pyrosequencing quantitatively measures methylation at individual CpG sites through sequencing-by-synthesis, providing highly reproducible estimates of methylation percentages.148 Compared with MSP, pyrosequencing offers superior quantitative accuracy and is particularly useful for validating methylation gradients across the metaplasia–dysplasia–carcinoma sequence.
Numerous studies investigating CDKN2A, RUNX3, APC, and SFRP family methylation in BE and gastric carcinogenesis have employed pyrosequencing because of its high analytical precision and relatively straightforward implementation.43,122 However, pyrosequencing is limited to predefined genomic regions, requires moderate DNA input, and is less suitable for large-scale genome-wide discovery studies.148
MSP and quantitative MethyLight assays
MSP and its quantitative derivative, MethyLight PCR, remain among the most commonly used techniques in translational methylation studies because of their simplicity, low cost, and excellent analytical sensitivity.8,149 These assays use methylation-specific primers following bisulfite conversion to selectively amplify methylated DNA, making them particularly useful for detecting low-frequency methylated alleles in plasma, cytology specimens, or nonendoscopic sampling devices.
Many early studies evaluating VIM, CCNA1, VAV3, ZNF682, RNF180, and RPRM employed MSP- or MethyLight-based approaches because of their robust performance using limited DNA input.52,150 Nevertheless, these assays are generally limited to a small number of candidate loci and may exhibit interlaboratory variability due to differences in primer design, amplification conditions, and threshold determination. Consequently, although MSP remains valuable for targeted clinical assays, its limited multiplexing capacity restricts its utility for comprehensive biomarker discovery.9
ddPCR
ddPCR has emerged as one of the most promising technologies for clinical methylation testing, particularly for liquid biopsy applications. In ddPCR, DNA samples are partitioned into thousands of nanoliter-sized droplets, allowing absolute quantification of methylated DNA molecules without requiring standard calibration curves.151 This partitioning substantially improves analytical sensitivity and enables reliable detection of rare methylated alleles present in circulating tumor DNA.
Because tumor-derived methylated DNA may be present at very low abundance within circulating cfDNA in patients with early-stage gastrointestinal cancers, ddPCR offers important advantages over conventional quantitative PCR.152 Recent investigations evaluating plasma methylation biomarkers for EAC and GC have demonstrated that digital PCR platforms improve analytical sensitivity and quantitative reproducibility while requiring only minimal DNA input.107 Its relatively rapid turnaround time and compatibility with Clinical Laboratory Improvement Amendments (CLIA)-certified laboratories further support its potential for future clinical implementation. However, ddPCR remains restricted to targeted genomic regions and is therefore better suited for biomarker validation than initial discovery.
Targeted next-generation bisulfite sequencing
Targeted NGS combines bisulfite conversion with massively parallel sequencing, allowing simultaneous quantitative assessment of hundreds to thousands of CpG loci within predefined genomic panels. Compared with single-gene assays, targeted NGS provides substantially greater multiplexing capacity while maintaining high analytical sensitivity.
Recent methylation panels developed for BE and gastric cancer increasingly utilize targeted NGS to evaluate multiple biomarkers simultaneously, thereby improving diagnostic accuracy through machine learning-based classification algorithms rather than reliance on individual genes.10,52 Targeted sequencing also permits assessment of methylation density, regional methylation patterns, and differentially methylated regions, all of which may outperform single CpG measurements for cancer detection. Nevertheless, higher costs, increased bioinformatic complexity, and the need for specialized laboratory infrastructure currently limit widespread clinical adoption.
Genome-wide DNA methylation arrays
Genome-wide methylation arrays, including the Illumina HumanMethylation450 BeadChip and the more recent Infinium MethylationEPIC array, have transformed biomarker discovery by enabling interrogation of more than 450,000 and 850,000 CpG sites, respectively.153 These platforms provide comprehensive epigenomic profiling across promoters, enhancers, CpG islands, shores, shelves, and open sea regions, facilitating unbiased identification of novel methylation biomarkers.
Numerous genome-wide studies have used these arrays to identify candidate biomarkers associated with BE progression, GIM, and early gastric cancer.10,43,117 These approaches have substantially expanded understanding of upper gastrointestinal epigenetic evolution and enabled the development of multimarker diagnostic panels. However, methylation arrays remain research tools rather than routine clinical assays because they require relatively high-quality DNA, sophisticated bioinformatic analysis, and extensive external validation before translation into clinical diagnostics.9
Cell-free methylated DNA immunoprecipitation sequencing (cfMeDIP-seq)
cfMeDIP-seq represents an emerging technology that enriches methylated cfDNA fragments using antibodies against 5-methylcytosine prior to NGS. Unlike bisulfite-based methods, cfMeDIP-seq preserves DNA integrity and requires substantially lower DNA input, making it particularly attractive for liquid biopsy applications where circulating tumor DNA concentrations are extremely limited.154
Recent studies have demonstrated that cfMeDIP-seq can identify highly cancer-specific methylation signatures while preserving sufficient tissue-specific information for tumor localization.154 These characteristics have generated considerable interest in integrating cfMeDIP-seq into MCED assays, where methylation patterns can simultaneously detect malignancy and predict tissue of origin.11 Although early results are encouraging, prospective validation in upper gastrointestinal cancers remains limited, and standardization of laboratory protocols will be necessary before routine clinical implementation.
Whole-genome bisulfite sequencing (WGBS)
WGBS provides single-base resolution of DNA methylation across the entire genome and is considered the gold standard for comprehensive methylome characterization.155 WGBS enables identification of novel differentially methylated regions, enhancer methylation, and non-CpG methylation patterns that may not be captured by targeted panels or methylation arrays.
Despite its unparalleled genomic coverage, WGBS remains prohibitively expensive for routine clinical use because of high sequencing costs, extensive computational requirements, and large DNA input needs. Consequently, WGBS currently serves primarily as a discovery platform for identifying candidate biomarkers that are subsequently validated using more clinically practical targeted assays.155
Comparative considerations for clinical translation
Although remarkable progress has been made in methylation assay development, no single platform is optimal for every clinical application. Genome-wide technologies, including EPIC arrays and WGBS, excel in biomarker discovery but are currently impractical for routine diagnostics because of their cost and analytical complexity. In contrast, pyrosequencing, ddPCR, and targeted bisulfite sequencing provide highly reproducible quantitative assessment of predefined biomarkers and are better suited for clinical validation and eventual implementation.9
Importantly, differences in assay methodology represent one of the principal reasons reported diagnostic performance varies considerably across published studies. Factors including DNA source (fresh tissue, formalin-fixed paraffin-embedded tissue, cytology specimens, plasma, serum, or cfDNA), bisulfite conversion efficiency, sequencing depth, normalization strategies, and statistical classification algorithms all influence measured sensitivity, specificity, and AUC. Therefore, direct comparison of biomarker performance across studies should be interpreted cautiously, particularly in the absence of standardized laboratory protocols or prospective multicenter validation.139,140
Future clinical implementation will likely rely on targeted, multiplexed methylation assays capable of simultaneously evaluating multiple biomarkers from minimally invasive specimens such as plasma, Cytosponge, EsoCheck, or gastric brushing specimens. Integration of these assays with machine learning algorithms, clinicopathologic risk factors, and established surveillance strategies may substantially improve early detection and personalized risk stratification for patients with BE, GIM, and upper gastrointestinal adenocarcinomas.11 Continued standardization of analytical methods, external validation across diverse populations, and adherence to reporting guidelines such as Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) will be essential for translating promising methylation biomarkers from discovery into routine clinical practice.139
Limitations
Several limitations should be considered when interpreting the findings presented in this review. First, the conclusions drawn from the existing literature are inherently limited by substantial heterogeneity among published studies. Considerable variability exists in study design, patient demographics, disease stage, specimen type (tissue, cytology, plasma, or cfDNA), methylation detection platforms, gene panels, and statistical methodologies. Consequently, reported measures of diagnostic performance, including sensitivity, specificity, and AUC, should not be compared directly across studies, as differences in patient selection, reference standards, and laboratory methods preclude meaningful quantitative comparisons without formal meta-analytic approaches.9,140
Second, although numerous methylation biomarkers have demonstrated promising diagnostic and prognostic performance, relatively few have undergone external validation in large, prospective, multicenter cohorts. Most available studies remain retrospective or case-control in design and evaluate biomarker performance under controlled research conditions rather than real-world clinical settings. Furthermore, many investigations report analytical performance without assessing whether methylation-guided testing improves patient-centered outcomes, such as earlier cancer detection, reduced mortality, optimized surveillance intervals, or decreased healthcare costs. As a result, the clinical utility of many candidate biomarkers remains to be established despite encouraging preliminary findings.139,142,143
Third, while this review emphasizes shared epigenetic mechanisms across Barrett’s-associated EAC and gastric carcinogenesis, important biological differences exist between these malignancies with respect to their etiologic factors, tissue microenvironments, molecular evolution, and clinical management. The comparative framework presented herein is intended to highlight common principles of DNA methylation biology and opportunities for translational biomarker development rather than imply that these diseases are biologically identical. Similarly, although several methylation markers are shared across both disease pathways, their functional significance, temporal acquisition, and diagnostic performance may differ substantially depending on tissue context and disease stage.10,13
Finally, as a narrative review, this manuscript does not employ a formal systematic review methodology or quantitative meta-analysis. Although every effort was made to include landmark studies, recent advances, and clinically relevant investigations, publication bias and selective reporting within the existing literature remain unavoidable limitations. In addition, the rapid evolution of methylation profiling technologies, liquid biopsy platforms, and MCED assays means that new biomarkers and validation studies are likely to emerge following publication. Consequently, the conclusions presented should be interpreted as a contemporary synthesis of a rapidly evolving field rather than definitive evidence supporting immediate clinical implementation.
Future directions and clinical translation of DNA methylation biomarkers in upper gastrointestinal adenocarcinoma
DNA methylation has emerged as one of the most promising molecular approaches for improving the early detection, risk stratification, and surveillance of Barrett’s-associated EAC and gastric adenocarcinoma. However, despite the growing number of candidate biomarkers with excellent reported diagnostic performance, relatively few have progressed beyond the discovery or early validation stages. Future research should therefore prioritize rigorous clinical validation rather than continued identification of novel methylation targets. A major challenge remains the substantial methodological heterogeneity among published studies, including differences in specimen types, methylation detection platforms, laboratory protocols, and bioinformatic analyses, which complicate direct comparison of biomarker performance and limit reproducibility.9 Standardization of analytical methods and adherence to reporting frameworks such as REMARK will be essential for facilitating multicenter validation and eventual clinical implementation.139
Future biomarker development is likely to shift from single-gene assays toward multiplexed methylation signatures integrated with machine learning algorithms and complementary molecular datasets. Rather than relying on individual methylation events, emerging diagnostic models increasingly combine panels of biomarkers to improve diagnostic accuracy and tissue-of-origin prediction. Shared methylation events involving CDKN2A (p16), APC, RUNX3, and members of the SFRP family have been consistently implicated in both Barrett’s-associated neoplasia and gastric carcinogenesis, reflecting common epigenetic mechanisms associated with chronic inflammation, Wnt signaling dysregulation, cell-cycle disruption, and progressive metaplasia.8,10,13,43 In contrast, biomarkers such as VIM, CCNA1, VAV3, and ZNF682 appear to exhibit greater specificity for BE and EAC, whereas RNF180, RPRM, BARHL2, PCDH10, SOX17, and ZIC1 have demonstrated particular promise in GIM and gastric adenocarcinoma.13,52,106,122 Integrating these shared and lineage-specific methylation signatures with established clinicopathologic risk factors—including Barrett’s segment length, dysplasia grade, H. pylori infection status, and histologic staging systems such as OLGIM—may improve personalized risk prediction models capable of more accurately identifying patients who would benefit from intensified surveillance.6,10,156 Continued investigation is also needed to determine whether these epigenetic alterations function primarily as causal drivers of neoplastic progression or as molecular biomarkers reflecting chronic inflammatory injury and metaplastic evolution, as accumulating evidence suggests that both mechanisms likely contribute to upper gastrointestinal carcinogenesis.8,10,13,43
Minimally invasive sampling strategies and liquid biopsy technologies offer significant opportunities to expand molecular screening beyond conventional endoscopy. Devices such as Cytosponge, EsoCheck, and EsophaCap, together with plasma-derived cfDNA, have demonstrated the feasibility of combining nonendoscopic sampling with methylation-based diagnostics.36,49,52,157,158 More recently, genome-wide cfDNA methylation profiling has enabled the development of MCED assays capable of identifying multiple malignancies while predicting tissue of origin, as demonstrated by the CCGA and PATHFINDER studies.11,146 Although current performance for early-stage upper gastrointestinal cancers remains suboptimal, continued refinement of tissue-specific methylation classifiers and targeted biomarker panels may substantially enhance the role of liquid biopsy in early detection, surveillance, and treatment monitoring.
Successful clinical translation, however, will require considerably more than high diagnostic accuracy. Before methylation biomarkers can be incorporated into routine practice, prospective multicenter studies must demonstrate analytical validity, clinical validity, clinical utility, and cost-effectiveness while establishing meaningful improvements over existing surveillance strategies.8,139,143 Integration into evidence-based clinical guidelines will additionally require standardized laboratory workflows, regulatory approval, reimbursement, and implementation across diverse healthcare settings. Rather than replacing established endoscopic and histopathologic evaluation, methylation biomarkers will most likely serve as complementary tools that improve patient selection for screening, personalize surveillance intervals, and reduce unnecessary procedures through more precise molecular risk stratification.5,6
Conclusions
Barrett’s-associated EAC and GC represent biologically distinct malignancies that nevertheless share fundamental epigenetic mechanisms characterized by progressive DNA methylation during chronic inflammation and metaplastic transformation. Advances in high-throughput methylation profiling, liquid biopsy technologies, and computational risk modeling have accelerated the field toward precision prevention and early detection. Several of these tools have already crossed into clinical or near-clinical use: Cytosponge-TFF3 coupled with biomarker panels has been prospectively evaluated in real-world UK surveillance pathways, the EsoCheck/EsoGuard methylated VIM and CCNA1 assay is available as a CLIA laboratory-developed test, and Esopredict has become the first epigenetic prognostic assay clinically validated to risk-stratify patients with BE, while blood-based MCED assays incorporating cfDNA methylation are undergoing prospective evaluation. Continued collaboration among molecular biologists, gastroenterologists, pathologists, bioinformaticians, and implementation scientists will be critical to standardize assays, validate biomarkers across diverse populations, and integrate molecular diagnostics into clinical care. Addressing these translational challenges will ultimately determine whether DNA methylation biomarkers evolve from promising investigational tools to evidence-based components of routine management for patients at risk of upper gastrointestinal adenocarcinomas.
Declarations
Funding
No specific funding was received for the preparation of this review.
Conflict of interest
The authors declare that they have no competing interests.
Author contributions
Conceptualization (MH, ZZ), literature search (MH), data curation (MH), investigation (MH), original draft preparation (MH), review and editing (MH, ZZ), supervision (ZZ), validation (ZZ), and project administration (ZZ). Both authors have approved the final version and publication of the manuscript.