Biomarkers for Gastric Cancer and Premalignant Gastric Conditions

Introduction

Gastric cancer (GC) remains a major global public health threat, ranking fifth in cancer incidence and mortality worldwide¹. Population-based endoscopic screening programs in high-incidence East Asian countries (Japan, South Korea) have reduced GC-specific mortality by 40% in recent decades, primarily by detection of early, curable-stage GC^2,3. In the U.S., GC is typically diagnosed at advanced stages when symptoms prompt diagnostic workup. Five-year survival for advanced stage GC is poor (regional: 37%, distant: 7.5%), and the dismal overall survival of 38% in the US reflects a missed opportunity⁴.

Upper endoscopy is the mainstay for GC screening, allowing simultaneous biopsy for risk stratification and resection of advanced neoplasia. However, it is resource-intensive, poses logistical barriers, and carries procedural risk. While population-wide screening may be appropriate for high GC incidence countries, targeted approaches are crucial in low-intermediate incidence countries such as the U.S. The recent American Gastroenterological Association (AGA) Clinical Practice Update suggests targeted endoscopic screening for higher-risk groups (e.g., select non-White racial/ethnic groups, chronic HP plus additional risk factors)⁵, and the American College of Gastroenterology (ACG) guidelines suggest consideration of screening on an individualized basis (e.g., first-degree family history of GC, first generation immigrants high-incidence countries)⁶. Currently, evidence for efficacy of endoscopic GC screening in high-risk U.S. subgroups is lacking^5,6, and risk stratification tools are imperfect to inform endoscopic screening/surveillance decisions. Precision methods to identify high-risk individuals would maximize the benefit-to-harm/cost ratio.

Biomarkers applied to serum or endoscopic assessment may improve risk stratification in lower-incidence countries, or higher-incidence/lower-resource settings. Biomarkers, “biological molecules that signal a normal or abnormal biological process”, include proteins (e.g., enzymes, hormones, antigens) and cancer-related genetic signatures (e.g., mutations, amplifications, translocations, methylation) found in bodily fluid/tissue⁷. Ideal biomarkers are highly sensitive and specific, cost-effective, and reproducible. Biomarkers may be applied across the cancer continuum, from assessing risk to treatment response; this article focuses biomarkers for detection of gastric premalignant conditions (GPMC) and gastric adenocarcinoma (GC) (Figure 1), addressing the following:

1. Can biomarkers be used for GC risk stratification to inform endoscopic screening or surveillance?

2. Can biomarkers accurately detect advanced GPMC and/or early GC?

Figure 1. Selected biomarker targets for gastric cancer risk stratification and early detection.

This landscape is expanding, and multi-analyte tests which utilize several of these targets are being developed, potentially alongside advanced modeling algorithms, to improve test accuracy.

While GC are subclassified by clinicopathologic features, most studies lack stratified test characteristics by subsite (cardia vs. non-cardia), or histology (adenocarcinoma, intestinal vs. diffuse). Accordingly, we examine biomarkers with this broader lens (Table 1).

Table 1.

Summary of biomarkers for gastric cancer risk-stratification and early detection

Main Category	Target	Name/Components	Clinical Applications	Pros	Cons
Biomarkers for GC risk stratification
H. pylori-related Biomarkers	H. pylori infection	H. pylori IgG (ELISA)	Diagnosis of H. pylori infection. GC risk stratification	Convenient, inexpensive; sensitivity 85%, specificity of 79% in pooled estimates; high NPV (97.4%)	Low PPV in low-prevalence settings. Can't distinguish past vs. current infection
	H. pylori infection	Stool antigen test	Diagnosis of H. pylori infection	High sensitivity (86%), specificity (92%)	Requires stool sample, less convenient than serology
	H. pylori infection	Urea breath test	Diagnosis of H. pylori infection	High sensitivity (96%), specificity (93%)	Requires specialized equipment, less accessible in some settings
	H. pylori virulence proteins	CagA, VacA, UreA	Enhanced GC risk stratification	Stronger GC association (OR ~9.2); identifies high-risk phenotypes	Limited validation. Not commercially available.
	H. pylori virulence factors	Multiplex Serology (VacA, GroEl, HcpC, HP1564)	Enhanced GC risk stratification	High sensitivity (100%), specificity (90%) for active infection	Limited validation. Not commercially available.
	H. pylori immunoproteome	Ggt, HsIU, NapA, CagA	GC and GPMC detection	Moderate performance (AUC 0.73) for GC vs. controls	Limited validation. Not commercially available.
	H. pylori outer membrane proteins	Omp, HP0305	GPMC detection	High odds (OR 7.43) for GPMC; AUC 0.75 with additional clinical variables	Limited validation. Not commercially available.
Biomarkers related to GPMC or AIG	Gastric mucosal atrophy	Pepsinogen (PGI, PGII, PGI/PGII ratio)	Detection of AG, and GPMC risk stratification	Modest performance (AUC 0.76 for GC, 0.85 for AG); widely studied	Low sensitivity in Western populations; high variability in cut-offs
	H. pylori + PG	ABC(D) method	Stratifies GC risk into 4 groups	Used in Japan; enhanced differentiation	Less predictive in U.S.
	PG, gastrin-17, H. pylori IgG	GASTROPanel	Enhanced risk stratification for AG	High specificity (0.94), good agreement with endoscopic biopsy (91%)	Limited sensitivity (0.70); mostly validated in Asia/Europe
	Parietal cell	Anti-Parietal Cell Antibody (PCA)	AIG diagnosis	High sensitivity (70–80%) for AIG	Low specificity due to overlap with H. pylori inflammation
	Parietal cell	Anti-Intrinsic Factor Antibody (IFA)	AIG diagnosis	High specificity (>95%) for AIG	Low sensitivity (30–60%)
	GIM	Expression signature of 26 genes	GC risk stratification	Identifies advanced GIM (OLGIM III or IV); enhanced GC risk stratification	Research setting; not yet clinical
Genetic and epigenetic markers	Hereditary GC	Germline pathogenic variants (APC, ATM, BRCA1, BRCA2, CDH1, etc.)	Risk stratification for familial or hereditary GC	Identifies high-risk individuals. Enhanced risk stratification based on H. pylori status.	Accounts for <3% of GC cases; requires genetic testing
	GC susceptibility	Polygenic Risk Scores (PRS)	Risk stratification	Identifies individuals with mutations in low-penetrance genes.	Poor to moderate discrimination (AUC 0.56–0.78); not ready for clinical use
	Epigenetic changes	DNA methylation (miR-124a-3, EMX1, NKX6-1) or histone modification	GC risk stratification	Up to 3-fold increased GC risk with high methylation	Requires validation in diverse populations; not commercially available
Other serologic markers	Epstein-Barr virus (EBV)	Anti-BALF2, anti-LF2, anti-BORF2	GC subtype classification (EBV-associated GC)	Good discrimination (AUC 0.81–0.85) for EBV-associated GC; useful in prediction of response to therapy	Not useful for GC detection due to ubiquity of EBV, role in prognostication
	GC/GPMC	Trefoil Factors (TFF)	GC and GPMC detection	Improved PPV (0.828) and NPV (0.878) with PG	Limited validation.
	Atrophic gastritis	HE-4 with PG	AG detection	Improved sensitivity (85.2%) over PG alone	Reduced specificity; limited validation
Biomarkers for early detection of GC
Liquid biopsy	MicroRNA (miRNA)	miR-421	Early GC detection	High AUC (0.92) for stage I-II GC	Validated mainly in Chinese cohorts; requires broader validation
	miRNA panel	GASTROClear® (a 12-miRNA panel)	Early GC detection	Sensitivity (0.71–10.87); cost-effective in Singapore ($44,531/QALY); FDA breakthrough designation	Moderate specificity in some cohorts.
	Circulating methylated DNA of Reprimo	RPRM cfDNA	Detection of early GC / high-grade dysplasia	Combined with PG improves sensitivity (78%)	Validation needed; moderate specificity
	Extracellular lncRNA	GClnc1	Early GC detection	AUC 0.94 for stage I/II GC; distinguishes GC from GIM	Investigational; no clinical use yet

I. Biomarkers for Gastric Cancer Risk Stratification

Helicobacter pylori (HP) is a Group I (definitive) carcinogen for gastric adenocarcinoma (attributable risk 80–90%) and mucosa-associated lymphoid tissue lymphoma^8,9. Additional modifiable risk factors include diet (e.g., high salt, preserved foods), tobacco and heavy alcohol use. Variation in exposures alongside host genetic diversity underlie the wide intra-/inter-country differences in GC incidence/mortality¹⁰. In the U.S., immigrant and racial/ethnic minority groups disproportionately bear GC risk and are the focus of targeted endoscopic screening⁵. However, race, ethnicity, and immigration status are social constructs, are not genetically or biologically defined. Better characterizing the exposure profile (‘exposome’) or preneoplasia (e.g., molecular profiling) using biomarkers may improve risk assessment beyond surrogate markers.

Helicobacter pylori-related biomarkers

Although Helicobacter pylori (HP) colonizes half the global population, <1–3% of infected individuals develop GC^11,12. In some, chronic HP-induced gastric inflammation progresses to gastric mucosal atrophy, intestinal metaplasia (GIM), and dysplasia – precursors in the histologic progression to intestinal-type GC (Correa cascade)¹³. HP has evolved sophisticated virulence mechanisms. The best-characterized are CagA, an oncoprotein that induces epithelial cell morphologic changes (aid motility/invasion) and disrupts apoptosis, and VacA, which impairs immune cell function¹⁴. The HP immunoproteome of over 1,500 proteins, offers many targets for non-invasive diagnosis or risk stratification¹⁵. Identifying higher-risk HP phenotypes and defining present or past HP infection informs GPMC and GC risk. As the vast majority of HP-colonized individuals will never develop GC, precision approaches that can identify high-risk individuals will improve cost-effectiveness, antibiotic stewardship, and minimize unnecessary treatment¹⁶.

HP IgG (ELISA), is widely available and inexpensive, with pooled sensitivity of 85%, and specificity of 79% for current/prior HP infection¹⁷. Positive HP IgG (ELISA) is consistently associated with increased GC risk (pooled odds ratio [OR] 3.01 for risk of GC, 95% confidence interval [CI] 2.22–4.07)¹⁸. This may underestimate true risk; those with advanced GPMC but cleared infection may test negative^19,20. More sensitive tests (immunoblot including CagA, VacA, UreA or multiplex serology) generally show a stronger association (pooled OR 9.22, CI 3.12–27.21) compared to ELISA¹⁸. While IgG has a high negative predictive value (NPV, 97.4%), its low positive predictive value (PPV, 22.8%, with further age-related declines) limits use in low-prevalence clinical settings, including many populations in the U.S.^19,20, and non-invasive diagnosis of HP relies upon stool antigen (sensitivity 86%, specificity 92%) and urea breath testing (UBT, sensitivity 96%, specificity 93%)^21,22. Importantly, serology performance varies between commercially available assays and populations, highlighting the need for local validation²³. Assays that discriminate current and prior HP infection would be valuable but are not clinically available. IgM, given rapid decline after infection clearance, was evaluated for this purpose, but has poor sensitivity (6.8%)²⁰.

Multiplex serology can identify various targets (HP status, GC, GPMC). Multiplex serology combining ≥2 of 4 HP virulence factors (Vac A, GroEl, HcpC, HP1564) can identify active HP infection with specificity of 90% and sensitivity of 100%, compared to UBT²⁴. This cohort was predominantly African American, and broader validation is required. A Chilean cohort study (N= 50 GC, 50 controls) found a four-protein combination demonstrated moderate discrimination between GC and controls (Ggt, HsIU, NapA, CagA, among 1,527 HP analyzed proteins; area under the curve [AUC] of 0.73)¹⁵. The humoral response to HP can also identify GPMC; in a high-incidence Chinese cohort (Linqu County, Shandong Province, N= 1,402) seropositivity to both Omp and HP0305 was associated with a 7-fold higher odds if GPMC on endoscopy (OR 7.43, CI 5.59–9.88)²⁵. A classification model including Omp, HP0305, (plus age, smoking, HP seropositivity) demonstrated moderate discrimination of GPMC from controls (AUC 0.75, CI 0.73–0.78)²⁵. Similarly, in a Chilean cohort (N= 50 GIM, 50 controls), a model including 11 IgG antibodies to HP proteins demonstrated an AUC of 0.81 (CI 0.75–0.87) for discriminating GIM from controls, though it did not report GIM severity²⁶.

Validation of these tests in target populations is critical, as HP virulence factors linked to GC vary by region²⁷. In Europe, only CagA and GroEL are associated with GC risk, whereas in East Asia, 5 additional proteins are implicated (HP0305, HyuA, NapA, Omp, VacA)²⁸. This heterogeneity likely reflects different exposures, HP serotypes, and interplay with host genetics¹⁶. Though not yet commercially available, serologic HP markers show potential for GC risk stratification.

Biomarkers related to gastric premalignant changes

As distinct steps in GC development, GPMC are promising biomarker targets. Serum pepsinogens (PGI and PGII) reflect gastric mucosal status; a reduced PGI:PGII ratio indicates progression of HP-associated gastritis from antrum to the corpus. “Standard” cutoff values predictive of GC risk (PGI ≤70 ng/mL and ≤3 PGI:PGII ratio) were developed in East Asia and validated in other settings such as Portugal²⁹. Optimal threshold values may vary by population. A meta-analysis of 31 studies (1,520 GC, 2,265 atrophic gastritis, mostly in Asia) found modest diagnostic accuracy of PG for GC (sensitivity 0.69, CI 0.60–0.76; specificity 0.73, CI 0.62–0.82; AUC 0.76, CI 0.72–0.80) and for atrophic gastritis (sensitivity 0.69, CI 0.55–0.80, specificity 0.88, 0.77–0.94; AUC 0.85, CI 0.82–0.88)³⁰. Limitations include high between-study heterogeneity, potential for publication bias, and limited generalizability beyond East Asia. There are limited U.S. studies, in one study using pre-diagnostic serum samples from the Prostate, Lung, Colorectal and Ovarian Cancer (PLCO) Screening Trial (N= 105 GC, 209 controls), PG was associated with GC risk (OR 10.6, CI 4.3–26.2), after adjusting for HP status, family history, smoking³¹. However, sensitivity was low (0.34–0.44) with high specificity (0.92–0.94) using standard cutoffs. Alternative thresholds (PG ratio ≤ 5.2, HP IgG ≥ 17.5 U/mL) improved sensitivity (0.88) but reduced specificity (0.51)^31,32. Standard cutoffs again showed low sensitivity for GPMC detection (15–47%) but high specificity (78–95%), and addition of HP IgG/gastrin-17 offered no significant improvement (AUC<0.7)³³. This cohort (N= 135) was high-risk (largely foreign-born, Hispanic or East Asian, 12% with GC family history), highlighting need for improved tools in diverse populations³³.

Efforts to further improve PG-based risk stratification include the ”ABC(D) method”, which classifies GC risk based on HP IgG positivity and PG (using standard cutoffs). In Japan, GC risk increased progressively across groups, compared to Group A (HP-/PG-, i.e., no exposure to HP, no atrophy): Group B (HP+/PG-) had 9.8-fold, Group C (HP+/PG+) had 19.6-fold, and Group D (HP-/PG+) 120.4-fold risk of GC over 8 years³⁴. In the U.S.-based PLCO cohort, only PG+ individuals (Groups C and D) demonstrated elevated GC risk compared to Group A, though with a notably lower effect estimate (OR 9.3, CI 3.4–25.2)³².

GastroPanel^® (Biohit Oyj, Helsinki, Finland) is a commercial ELISA panel of PGI, PGII, gastrin-17, HP IgG for dyspepsia evaluation/atrophic gastritis. A meta-analysis of 27 studies (in Asia and Europe, N= 8,654) demonstrated a sensitivity of 0.70 (CI 0.64–0.76) and specificity of 0.94 (CI 0.90–0.95) for atrophic gastritis³⁵. In a prospective study (United Kingdom, N= 324, 21.6% HP positive), GastroPanel results also agreed with the updated Sydney System in 91% (CI 0.88–0.94), supporting its potential for non-invasive risk stratification³⁶.

Biomarkers in autoimmune gastritis

Autoimmune gastritis (AIG) involves inflammatory destruction of the oxyntic mucosa, replacement by atrophic metaplastic mucosa. Resulting achlorhydria, hypergastrinemia, and enterochromaffin-like cell hyperplasia may progress to type 1 gastric neuroendocrine tumors – the primary neoplastic pathway for AIG. Epidemiologic data suggest an additional association between AIG and GC (adenocarcinoma). Pernicious anemia (a late stage manifestation) confers a 7-fold risk of GC³⁷, and meta-analysis of 8 observational studies demonstrated a pooled relative risk of 11.05 (CI 6.39–19.11) for GC among patients with metaplastic AIG³⁸. Limitations include heterogeneity in AIG severity and incomplete/absent consideration of HP as a confounder. In one important study that included HP exposure status, HP-naïve subjects showed no association between AIG and GC risk; limitations included the small cohort size (N= 211) and follow up duration (median 7.5 years)³⁹. Additionally, AIG and HP-related metaplasia are indistinguishable in murine and human data, with similar histopathologic features, immunofluorescence staining for key markers (GKN3, MUC6, GF), and transcriptomic signatures, with ANPEP/CD13 identified as a shared GC-related marker⁴⁰.

Anti-parietal cell antibody (PCA) and anti-intrinsic factor antibodies (IFA), widely used to diagnose AIG, could serve as GC biomarkers in select groups, such as women. In Finnish cohorts, PCA was associated with GC risk among HP seronegative women (OR 5.52, CI 3.16–9.64), but not among men or HP seropositive women⁴¹. PCA is more sensitive (70%–80% sensitivity) for diagnosing AIG but less specific compared to IFA⁴². PCA may be positive in HP gastritis, other autoimmune conditions, and falsely negative for advanced AIG (near complete loss of parietal cells)⁴³. In contrast, IFA, appearing in more advanced AIG, is less sensitive (30%–60% sensitivity) but highly specific (>95% specificity)^41,42. In a prospective study (Italy, N= 100, validation cohort), combining PCA and IFA with B12, chromogranin A, and hemoglobin achieved a reasonable accuracy for AIG diagnosis (sensitivity 85.7%, specificity 83.7%), compared to histopathology⁴⁴. Additionally, advanced-stage (AIG stage 3) patients had elevated gastrin-17, even as PCA positivity did not change with stage (N= 282)⁴⁵. Unlike PG, IFA or PCA levels are not directly linked to GC risk^44,45.

European and U.S. guidelines suggest endoscopic GC screening for individuals newly diagnosed with pernicious anemia^46,47, though more data are needed to guide surveillance recommendations⁶. Given mechanistic and epidemiologic overlaps, biomarkers for gastric mucosal atrophy/inflammation (e.g., low PGI and PG ratio) may be applicable in patients with AIG⁴⁸. However, as AIG is corpus-dominant, cutoff values differ from HP-associated gastric atrophy⁴⁸.

Tissue-based Biomarkers for Gastric Intestinal Metaplasia

GIM, characterized by replacement of gastric epithelial cells with mature and immature epithelium of intestinal lineage, is associated with 1.6% GC risk at 10 years (agnostic of clinical and histologic risk stratification)⁴⁹. In the U.S., GIM has lower inter-observer variability and more widespread pathologic reporting compared to atrophic gastritis⁴⁹. Higher-risk GIM includes anatomically extensive (corpus extension) and incomplete-type GIM (extensive GIM: RR 2.1, CI 1.0–4.4, incomplete GIM: RR 3.3, CI 2.0–5.6)⁴⁹. However, the need to rely on endoscopic and pathologic data poses logistical and cost challenges, and more refined, less invasive stratification methods are necessary.

Molecular, genomic, and microbiome profiling of GIM may improve GC risk stratification. For instance, specific GIM mutational burdens (chromosomal alterations, shortened telomeres) were associated with dysplasia or GC, while epigenomic patterns resembling normal mucosa showed regression in the Gastric Cancer Epidemiology Program (GCEP), a large Asian biobank^50,51. Further evaluation is needed for Western populations. Active HP infection alone may induce molecular changes, thus mucosal signatures independently associated with GC risk may not be identified in GCEP, which included individuals with active HP. The GAstric Precancerous Conditions Study (GAPS), in a multiethnic HP-negative cohort, identified a signature of 26 genes (including CDH17, KRAS) for advanced GIM (operative link for gastritis assessment staging system; OLGIM III-IV), which localized to metaplastic glands and intestinal-type GC, but not normal mucosa⁵². Further studies will refine molecular and genomic risk models.

Genetic and Epigenetic Alterations

10% of GCs demonstrate familial clustering, but <3% are attributed to hereditary cancer syndromes from germline pathogenic variants (PV; APC, ATM, BRCA1, BRCA2, CDH1, MLH1, MSH2, MSH6, PALB2)⁵³. Monoallelic PV in homologous recombination genes (BRCA1, BRCA2, PALB2, ATM) are more common in GC patients, and a strong synergistic increase in GC risk is seen in co-carriers (PV plus HP)⁵³. This supports a two-hit hypothesis; HP activates signaling pathways (via CagA) resulting in double-strand DNA breaks in the target epithelial cell, which, in carriers of homologous recombination PV, are repaired by error-prone mechanisms (e.g., mutagenic nonhomologous, or microhomology-mediated end-joining pathways), resulting in gross chromosomal rearrangements that confer greater risk GC⁵³.

Genome-wide association studies (GWAS) have identified lower-penetrance genes which may be targets for risk stratification. For instance, ANKRD50 and PTGER4 are consistently associated with GC, in Chinese and European populations^54,55. Polygenic risk scores (PRS), derived from such loci, can identify individuals missed by existing clinical risk prediction methods, such as in breast and prostate cancer⁵⁶. In a Chinese cohort, genetic predisposition to GC (high PRS) modified HP-associated GC risk and eradication reduced risk only among high PRS-individuals, highlighting the interplay between host genetic factors and HP¹⁶. However, GC-susceptibility PRS have shown insufficient discrimination to date (AUC 0.56 – 0.78, East Asian and European cohorts), and like other biomarkers, need to be evaluated in the target population⁵⁷.

Finally, epigenetic changes may modulate carcinogenic pathways. DNA methylation burden is elevated in HP-induced gastritis, increases with advanced GPMC, and may affect key cancer-related pathways (e.g., WNT, p53)^58,59. In a Japanese cohort, the methylation burden of select genes (miR-124a-3, EMX1, NKX6–1) was associated with GC risk; individuals with the highest methylation burden had 3-fold GC risk (HR 3.0, CI 1.58–5.72) compared to low-burden individuals^60,61.

Additional biomarkers for GC risk stratification

Epstein-Barr Virus (EBV)-associated GC is a discrete subtype that constitutes 10% of incident GC, with distinct genomic, histopathological, and clinical characteristics⁶². It is associated with greater immune cell infiltration and favorable outcomes compared to EBV-negative GC, independent of HP co-infection⁶³. EBV serology (including anti-BALF2, anti-LF2, anti-BORF2) demonstrates good discrimination (AUC 0.81–0.85) of individuals with EBV-associated GC from EBV-negative GC controls, but the ubiquity of EBV limits its utility for early GC detection⁶⁴. However, given elevated PD-L1 expression in EBV-associated GC, it may predict treatment response to immune checkpoint inhibitors⁶⁵.

Trefoil factors (TFF), secreted by gastrointestinal goblet cells, are elevated in GC and GPMC; combining PG and TFFs improved GC detection (PPV 0.828, NPV 0.878), over PG alone (PPV 0.608, NPV 0.874)⁶⁶. Adding HE-4 to PG increased sensitivity for atrophic gastritis (85.2%, CI 72.9–93.4%) compared to PG alone (sensitivity 55.6%. 41.4–69.1), at the cost of specificity⁶⁷. Finally, GC risk was associated with lower body iron stores in the European Prospective Investigation into Cancer and Nutrition (EPIC) study, possibly due to achlorhydria from severe atrophic gastritis limiting iron uptake⁶⁸. These markers show limited promise to date.

II. Biomarkers for Early Detection of Gastric Cancer

Liquid Biopsies are novel tests targeting tumor components (e.g., proteins, circulating tumor cells, cell-free DNA/RNA) shed into biologic fluid such as blood)^69,70. As tumor necrosis releases these markers, liquid biopsies are more sensitive/specific in advanced cancer. Thus, for early detection, test characteristics must be assessed in a cohort with high proportion of early disease⁷⁰. Standalone protein biomarkers (e.g., CEA, CA19–9, AFP, CA-125) perform poorly for GC, but are used in multi-analyte tests discussed below⁷¹.

MicroRNA (miRNA) are released by tumor cells through non-necrotic mechanisms (e.g., secretion via exosomes) and demonstrate promise for early GC⁷². miR-421 (linked to cell cycle regulation) demonstrated an impressive AUC of 0.92 in a meta-analysis of studies conducted in China (stage I-II: 27.6–58.9%)⁷³. GASTROClear^® (MiRXES), a 12-serum miRNA panel for stage I-II GC is commercially available in Singapore, and has sensitivity of 0.86 and specificity of 0.85 (AUC 0.91, validation cohort)⁷⁴. In individuals referred for endoscopy (standard clinical indications), GASTROClear showed a similar sensitivity of 0.87, but low specificity of 0.68 (AUC 0.85)⁷⁴. In a Japanese cohort (58% stage I GC) sensitivity was 0.71 and specificity was 0.75 (AUC 0.8), improving to AUC 0.88 (sensitivity 0.71, specificity 0.87), with the addition of PG and HP serology⁷⁵. It appears cost-effective for mass screening for Singaporean Chinese males (50–75 years; number needed to screen= 489)⁷⁴. In the U.S., GASTROClear has Food and Drug Administration (FDA) Breakthrough Device Designation (expedites the approval process for devices that demonstrate more effective treatment/diagnosis for important conditions)^76,77. Similarly, a four-miRNA panel showed sensitivity of 0.996 and specificity of 0.953 for early GC (AUC 0.998) in Japanese subjects, with ongoing prospective studies⁷⁸.

Combination assays targeting circulating RNA, miRNA, and DNA methylation markers are under development^79–81. Circulating methylated cell-free DNA of Reprimo (a tumor suppressor gene, RPRM cfDNA) showed sensitivity of 65% (CI 53.5 – 75.3%) and specificity 75.9 (CI 73.2 – 78.5%) to distinguish GC from normal/GPMC. RPRM cfDNA positivity had a 4-fold risk of high grade dysplasia or early GC (OR 3.9, CI 1.53 – 9.93), addition PGI:II ratio achieved a sensitivity of 78% (CI 54.4 – 94.0%) and specificity 63.0% (CI 52.3 – 72.9%) for detecting high grade dysplasia/early GC in Chilean subjects⁸². Additionally, extracellular vesicle-derived long noncoding RNA (GClnc1) identified stage I-II GC with AUC of 0.94 (CI, 0.91–0.97), and distinguished early-stage GC from GPMC in cases where other markers (CEA, CA72–4, CA19–9) were negative (AUC 0.89; CI 0.86 – 0.92)⁸¹.

Challenges and Future Directions

HP-associated intestinal-type GC has been the focus for GC biomarker development, facilitated by its higher incidence and greater availability of epidemiologic, clinical and mechanistic data. Even as the intestinal subtype accounts for most GC (74%), diffuse-type GC (16%) and other epithelial cancers (10%) still constitute significant proportions of incident cases⁸³. Few studies evaluate biomarkers by histology or anatomic subsite, likely due to insufficient statistical power. It is unclear whether these biomarkers can reliably be applied to all GC, or whether data aggregation results in blind spots for detection.

For HP-focused biomarkers, heterogeneity of populations and prevalent HP strains in different regions present a challenge. High GC incidence regions (East Asia, South America) have generally dominated in biomarker research, but HP prevalence/genomes demonstrate differences even between subgroups within multi-racial/ethnic countries.^84,85. For instance, Cag-A positivity was double in Black vs. non-Hispanic White (NHW) individuals in the U.S. (82% vs. 36%, p <.0001)²⁷. Thus, applying available test characteristics directly to a diverse low-GC incidence population requires additional validation. Finally, adequate representation of high-GC risk groups in studies is essential, especially as these demographics are underrepresented in trials⁸⁶. 70% of cancer deaths result from cancer types with no screening recommendations⁸⁷, spurring development of multicancer detection (MCD) tests, which aim to screen for multiple cancers non-invasively. MCDs apply liquid biopsy targets and advanced machine learning algorithms to identify multiple cancers simultaneously. While an exciting field, GC-specific data are limited. The PATHFINDER study was a proof-of-concept trial with 6,662 participants (median age 63 years, 63.6% female, 91.7% NHW)^88,89. The Galleri^® test (GRAIL, Menlo Park, CA) evaluates serum methylation patterns using a computational algorithm to predict the cancer signal origin (i.e., tissue origin) for 50 cancer types; while targets included GC, no cases occurred. Similarly, a multi-analyte blood test (cfDNA plus 8 protein biomarkers including CEA, CA 19–9), was evaluated in a large prospective study of women aged 65 −75 years (N= 10,006, DETECT-A study, Exact Sciences Corporation, Madison, WI); while 3 patients developed GC, these were missed by the test^90,91. Multiple MCDs that include GC as a target are being developed⁹², and even as GC-specific data are pending, these studies highlight the important downstream considerations for MCD screening (overdiagnosis, false positives, follow-up testing, adverse events)^89,90.

Currently, no MCDs carry FDA approval, but some are commercially available under the Clinical Laboratory Improvement Amendments (CLIA) program, which prioritizes reliability and accuracy of laboratory tests. In response to reports of erroneous test results, patient harm, and rapid growth of this market, FDA has extended oversight to in vitro diagnostic products in July 2024, with premarket compliance required in 2027⁹³. Without FDA approval, MCDs are generally not covered by insurance and may incur high out-of-pocket costs for the MCD and follow-up diagnostic procedures. Finally, important public health questions remain; the overall effectiveness of biomarkers/MCDs, intervals for effective screening, cost burden, and downstream impacts need to be carefully evaluated⁹⁴.

Summary

Gastric cancer has major global public health impacts, but few countries have organized screening/prevention efforts. While endoscopy is the current gold standard for detection and risk stratification, biomarker development is underway to provide non-/minimally-invasive options with benefits of convenience, safety, and lower cost. Biomarkers target various aspects GC carcinogenesis, including HP, mucosal atrophy, genetic or epigenetic changes. To date, none have demonstrated standalone test characteristics sufficient for implementation, though combination panels of multiple biomarkers and advanced modeling have improved accuracy. Further assessment in target populations and evaluation of downstream effects are needed to refine these promising tools prior to implementation.

Key Points:

• Gastric cancer (GC) is a major cause of preventable cancer death worldwide. Organized screening efforts are limited to select countries, endoscopy has barriers to implementation

• Biomarkers are promising screening tools, offering convenience, are minimally/non-invasive, and improving accuracy

• Biomarkers can target various aspects of GC carcinogenesis, and for screening purposes, may identify 1) high risk patients needing endoscopic screening and 2) presence of advanced gastric neoplasia or early GC

• Multi-target biomarker panels are in development and show potential for future use

• Priorities for translation of GC biomarkers into clinical use include validation studies in intended populations, refinement and innovation improve accuracy and reproducibility, cost-effectiveness, and accessibility

Synopsis:

Gastric cancer (GC) is a major cause of preventable cancer death worldwide. Endoscopic screening is the gold standard for early detection, but due to logistical and cost barriers, organized efforts are limited to select high-incidence countries. Biomarkers applied to GC prevention may provide an adjunctive strategy, by targeting multiple aspects of GC carcinogenesis including H. pylori infection, premalignant changes, and genetic aberrancies. While no single biomarker is sufficient, combination panels show potential. We summarize evidence on biomarkers for risk stratification of gastric premalignant conditions and GC early detection, outlining knowledge gaps and future directions to advance prevention efforts.