Authors: Xinmei Lin, Wenhan Shi, Zitong Zhou, Degang Li, Qiuyuan Yao, Yu Sun
Categories: Systematic Review, Early-onset gastric cancer, TCGA classification, Gene mutations, Immune biomarkers, Molecular features, Meta-analysis
Source: BMC Cancer
Authors: Xinmei Lin, Wenhan Shi, Zitong Zhou, Degang Li, Qiuyuan Yao, Yu Sun
Early-onset gastric cancer (EOGC), diagnosed before age 50, is characterized by distinct clinicopathological features, though its molecular landscape remains poorly defined.
A systematic literature search of PubMed, Embase, and Web of Science identified studies comparing molecular characteristics of EOGC and late-onset gastric cancer (LOGC). Meta-analyses assessed differences in The Cancer Genome Atlas (TCGA) molecular subtypes, gene mutations, therapeutic biomarkers, and serum tumor markers. Odds ratios (ORs) with 95% confidence intervals (CIs) were calculated; heterogeneity was assessed using the I^2^ statistic.
EOGC was associated with a higher prevalence of the genomically stable (GS) subtype (OR = 1.71, 95% CI: 1.37–2.12) and a lower prevalence of the chromosomal instability (CIN) subtype (OR = 0.62, 95% CI: 0.50–0.77). CDH1 mutations were more frequent in EOGC (OR = 3.44, 95% CI: 2.85–4.16), while HER2 expression (OR = 0.54, 95% CI: 0.43–0.67), dMMR/MSI-H status (OR = 0.25, 95% CI: 0.12–0.53), and p53 expression (OR = 0.56, 95% CI: 0.39–0.82) were significantly lower. Serum markers including CEA and CA19-9 were also less frequently elevated in EOGC.
EOGC represents a biologically distinct subset of gastric cancer with unique genomic and immunological features. These findings support age-specific diagnostic approaches and emphasize the value of multiomic strategies to uncover the mechanisms driving early-onset disease.
The online version contains supplementary material available at 10.1186/s12885-026-15567-5.
Gastric cancer (GC) remains a major global health burden, with approximately 968,000 new cases and over 660,000 deaths reported worldwide in 2022 alone [1]. Although advances in early detection, Helicobacter pylori eradication, and public health initiatives have contributed to a general decline in GC incidence and mortality, this downward trend has not been observed uniformly across all age groups [2]. Notably, recent epidemiological data suggest a rising incidence of GC among younger individuals, particularly those under the age of 50 [3]. This emerging subgroup—commonly referred to as early-onset gastric cancer (EOGC)—is increasingly recognized as a distinct clinical and molecular entity compared with late-onset gastric cancer (LOGC) [4].
The definition of EOGC varies across studies, with age cutoffs of 40, 45, or 50 years being most commonly used [5–8]. While GC in younger individuals often raises suspicion of hereditary cancer syndromes, only approximately 10% of EOGC cases can be attributed to germline mutations or familial predisposition, with the majority occurring sporadically [9]. Clinically, EOGC is more common in females and frequently presents aggressive pathological features, such as poor differentiation, extensive nodal involvement, predominant diffuse-type histology (particularly signet-ring cell carcinoma), and a more advanced TNM stage at diagnosis [10–12]. Collectively, these features contribute to a more aggressive disease course and poorer prognosis in younger patients.
In addition to clinicopathological characteristics, EOGC and LOGC also present distinct molecular profiles [13]. Recent studies suggest that EOGC is more frequently associated with CDH1 mutations and lower HER2 expression [8, 14]; however, these findings remain inconsistent, with several studies reporting contradictory results [15, 16]. With respect to the Cancer Genome Atlas (TCGA) classification, some authors have reported a higher prevalence of EBV-positive and genomically stable (GS) subtypes in EOGC [14, 17], whereas a study by Zhou et al. revealed no significant difference in the distribution of the EBV subtypes [18]. Similarly, discrepancies have been reported in the mutation status of ARID1A [8, 18], microsatellite instability (MSI) [10, 19], and serum tumor markers such as CA19-9 [20, 21]. Given the pivotal role of molecular biomarkers in guiding personalized treatment strategies, a comprehensive synthesis of current evidence is urgently needed to clarify these molecular distinctions and inform future research.
This study aims to conduct a systematic review and meta-analysis to elucidate the molecular differences between EOGC and LOGC. By comprehensively analyzing TCGA molecular subtypes, frequently mutated genes, targeted and immune-related biomarkers, and serum tumor markers, we sought to define and quantify the molecular characteristics that distinguish these two subtypes of gastric cancer. These findings are expected to provide a theoretical framework for the development of age-specific diagnostic tools, prognostic models, and precision therapeutic strategies tailored to the unique biology of EOGC.
This systematic review and meta-analysis was registered with PROSPERO (CRD420251011066) and conducted in accordance with the PRISMA guidelines [22].
A comprehensive and systematic literature search was conducted in PubMed, Embase, and Web of Science from their inception through February 24, 2025. The search strategy incorporated both Medical Subject Headings (MeSH terms) and relevant entry terms, including “Stomach Neoplasms,” “young onset,” and “early-onset gastric cancer.” The full search strategies for each database are detailed in Supplementary material 1. In addition, the reference lists of all included articles and relevant review papers were manually screened to identify additional eligible studies.
Inclusion criteria were as (1) original studies that directly compared molecular characteristics between EOGC and LOGC, with EOGC defined by an age threshold of 40, 45, or 50 years; (2) studies that reported data on at least one of the TCGA molecular subtypes; gene mutation frequencies (e.g., CDH1, TP53); targeted or immune-related biomarkers (e.g., HER2, dMMR/MSI-H); or serum tumor markers (e.g., CEA, CA19-9); and (3) studies involving pathologically confirmed tumor samples.
Exclusion criteria were as (1) case reports, letters, news articles, and reviews; (2) nonhuman studies or studies lacking definitive EOGC diagnosis; (3) studies with unclear or inconsistent definitions of EOGC, including those using age thresholds outside the 40–50 year range, which could introduce significant clinical heterogeneity; (4) studies without a clearly defined control group; (5) non-English publications; (6) studies with insufficient or invalid data for analysis; and (7) studies with duplicated or overlapping data.
Data extraction was independently performed by two authors (X.L. and W.S.). The extracted information included (1) the number of patients with early-onset and late-onset gastric cancer; (2) the distribution of TCGA molecular subtypes (EBV, MSI, CIN, and GS); (3) the frequencies of commonly mutated genes (e.g., CDH1, TP53, ARID1A, PIK3CA, and KRAS); (4) the expression levels of clinically relevant biomarkers (e.g., HER2, dMMR/MSI-H, PD-L1, COX-2, E-cadherin, and p53); (5) serum tumor markers (CEA and CA19-9); and (6) study-level characteristics, including first author, publication year, geographic region, study period, sample size, patient sex distribution, histological subtype (if available), molecular detection methods (e.g., IHC, PCR, NGS).
For studies presenting results only in graphical format, numerical values were extracted using GetData Graph Digitizer (version 2.24). For studies that reported only partial subgroup data (e.g., the number of diffuse-type cases in EOGC but not in LOGC), the missing values were inferred by subtraction from the total sample size when applicable. In addition, relevant data were also extracted from supplementary materials when not available in the main text.
Discrepancies in data extraction were resolved through discussion. Study quality was assessed using the Newcastle–Ottawa Scale (NOS), which evaluates methodological rigor across three selection (0–4 points), comparability (0–2 points), and outcome assessment (0–3 points). Only studies with an NOS score ≥ 6 were deemed eligible for inclusion. Quality assessments were independently performed by two authors (Y.S. and X.L.), with disagreements resolved by consensus.
Statistical analyses were conducted to compare the molecular features between EOGC and LOGC. For categorical variables, odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. Heterogeneity across studies was assessed via Cochran’s Q test and the I^2^ statistic. A random-effects model was applied when heterogeneity was significant (P < 0.10 or I^2^ > 50%); otherwise, a fixed-effects model was used. To evaluate the robustness of the pooled estimates, a leave-one-out sensitivity analysis was performed. Publication bias was assessed via funnel plots and Egger’s test. A two-tailed P value < 0.05 was considered statistically significant. All the statistical analyses were performed via Stata (version 12.0) and R (version 4.4.2).
A total of 22,091 records were identified through systematic searches of PubMed, Embase, and Web of Science, from which 4,424 duplicates were removed. After screening titles and abstracts on the basis of predefined eligibility criteria, 17,557 articles were excluded. An additional 88 full-text articles were excluded after detailed assessment. Ultimately, 21 studies met the inclusion criteria and were included in the final analysis [8, 10, 13–15, 17–21, 23–33]. A PRISMA flow diagram illustrating the study selection process is presented in Fig. 1.Fig. 1PRISMA flow diagram of study selection. Flowchart depicting the study selection process for the systematic review and meta-analysis
These studies were published between 2006 and 2024 across 10 countries and collectively included data from 5,593 patients diagnosed with EOGC. The key characteristics of the included studies are summarized in Table 1. Study quality, assessed using the NOS, ranged from 6–9, with a mean score of 7.52 indicating generally high methodological quality. The detailed results of the quality assessment are provided in Supplementary material 2.Table 1Characteristics of included studiesStudyPeriodCountryNumber of casesNumber of EOGCEarly-onset definition (y)Age categories (y)Molecular characteristicsNOS score Milne et al., 2006 [23]1993–2002Netherlands204113 ≤ 45 ≤ 45; > 45COX-2, E-Cadherin, p539 Park et al., 2009 [20]2000–2005Korea3362654 ≤ 45 ≤ 45; > 45CEA, CA19-97 Moelans et al., 2011 [24]/Netherlands199108 ≤ 45 ≤ 45; > 45HER29 Qiu et al., 2011 [25]1996–2006China1000294 < 50 < 50; ≥ 50CEA, CA19-98 Schildberg et al., 2013 [26]1995–2005Germany42356 < 50 < 50; ≥ 50COX-27 Schildberg et al., 2014 [32]1995–2005Germany42356 < 50 < 50; ≥ 50E-cadherin, p537 Gurzu et al., 2015 [19]2006–2013Romania19137 ≤ 45 ≤ 45; > 45HER2, dMMR/MSI-H, PD-L1, COX-2, E-Cadherin, p537 Cho et al., 2017 [8]/Korea224109 ≤ 45 ≤ 45; > 45TCGA classification, CDH1/TP53/ARID1A/PIK3CA/KRAS mutation8 Hakkaart et al., 2019 [27]2009–2013New Zealand9421 ≤ 45 ≤ 45; > 45CDH1 mutation6 Choi et al., 2020 [31]/Korea993110 < 45 < 45; ≥ 45CDH1 mutation8 Machlowska et al., 2020 [15]1993–2003Poland5335 ≤ 45 ≤ 45; > 45CDH1 mutation9 Moore et al., 2020 [17]1997–2016Israel7439 ≤ 45 ≤ 45; > 50EBV, HER2, dMMR/MSI-H8 Huang et al., 2021 [28]2009–2019China312154 ≤ 45 ≤ 45; > 50HER28 Machlowska et al., 2021 [33]1993–2003Netherlands, Finland and Poland18191 ≤ 45 ≤ 45; > 45p536 Setia et al., 2021 [13]/America105681 < 50 < 50; ≥ 50CDH1/TP53/ARID1A mutation6 Qu et al., 2022 [21]2000–2019China9406672 ≤ 40 ≤ 40; > 40CEA, CA19-97 Zhou et al., 2022 [18]/China1688308 ≤ 45 ≤ 45; > 45TCGA classification, CDH1/TP53/ARID1A/PIK3CA mutation9 Chu et al., 2023 [29]1994–2019China1959202 ≤ 39 < 40; ≥ 40CEA7 Han et al., 2024 [30]/China42328 ≤ 50 < 50; ≥ 50CDH1/TP53/ARID1A mutation6 Liu et al., 20242000–2022China18,8772206 ≤ 45 ≤ 45; > 45HER2, dMMR/MSI-H, CEA, CA19-9, HER28 Lumish et al., 2024 [14]2005–2018America1123219 < 50 < 50; ≥ 50TCGA classification, CDH1/TP53/ARID1A/PIK3CA/KRAS mutation, HER28
Data extracted from studies comparing TCGA molecular subtypes revealed significant differences between early-onset and late-onset gastric cancer. As no significant heterogeneity was observed for the CIN subtype (P = 0.4274, I^2^ = 0%) or the GS subtype (P = 0.4186, I^2^ = 0%), a fixed-effects model was applied. The GS subtype was significantly more prevalent in EOGC than in LOGC (Fig. 2; OR = 1.71, 95% CI: 1.37–2.12; P < 0.0001), whereas the CIN subtype was less common in the EOGC group (Fig. 2; OR = 0.62, 95% CI: 0.50–0.77; P < 0.0001). However, there were no statistically significant differences in the prevalence of the MSI and EBV subtypes between the two groups (Fig. 2).Fig. 2Distribution of TCGA molecular subtypes in EOGC vs LOGC. EOGC was associated with a significantly greater prevalence of the GS subtype and a lower prevalence of the CIN subtype. No significant differences were found in the EBV-positive or MSI subtypes
Analysis of key gene mutations revealed significant differences between EOGC and LOGC, particularly for CDH1. Pooled analysis demonstrated that CDH1 mutations were significantly more prevalent in EOGC (OR = 3.44, 95% CI: 2.85–4.16; P < 0.0001), although substantial heterogeneity was observed across studies (I^2^ = 78%, P < 0.0001), prompting the use of a random-effects model (Fig. 3).Fig. 3High-frequency gene mutations in EOGC vs LOGC. CDH1 mutations were significantly more common in EOGC. No significant differences were observed for TP53, ARID1A, PIK3CA, or KRAS
To explore the source of this heterogeneity, a sensitivity analysis was conducted by sequentially excluding individual studies. Removal of two studies—Choi et al. [31] and Hakkaart et al. [27]—substantially reduced heterogeneity (I^2^ = 19%, P = 0.2903), while the association between EOGC and CDH1 mutation remained statistically significant (OR = 2.76, 95% CI: 2.24–3.42; P < 0.0001; Supplementary Fig. 3). These two studies likely contributed to the observed heterogeneity due to methodological and clinical differences. Hakkaart et al. included only germline testing (whereas most other studies assessed somatic mutations or both), potentially underestimating the overall CDH1 mutation frequency [27]. In contrast, Choi et al. included a large EOGC cohort composed entirely of diffuse-type tumors, which may have led to an inflated mutation rate in that group [31] (Table 2).Table 2CDH1 Mutation Rates by Sex and Histology in EOGC and LOGCStudyEOGC CDH1 Mutation Rate (%)LOGC CDH1 Mutation Rate (%)Female EOGC (%)Male EOGC (%)Diffuse/Mixed EOGC (%)Intestinal EOGC (%)Diffuse/Mixed LOGC (%)Notes Cho et al., 2017 [8]36/109 (42.2%)20/115 (17.4%)31/59 (52.5%)15/50 (30.0%)36/109 (42.2%)NA20/115 (17.4%)Somatic + germline; all diffuse-type Hakkaart et al., 2019 [27]14/21 (66.7%)3/73 (4.1%)NANA14/20 (70%)NA30/60 (50.0%)Germline only; small Māori cohort; potential heterogeneity source Choi et al., 2020 [31]50/110 (45.5%)96/883 (10.9%)NANA50/50 (100.0%)0/< 60 (~ 0%)73/80 (91.3%)Somatic only; diffuse-type–enriched EOGC; potential heterogeneity source Machlowska et al., 2020 [15]31/35 (88.6%)18/18 (100.0%)NANA31/35 (88.6%)NANASomatic only Setia et al., 2021 [13]18/81 (22.2%)111/975 (11.4%)10/30 (33.3%)8/51 (15.7%)10/24 (41.7%)1/23 (4.3%)NASomatic onlyNA Not available in the original study
Exploratory data from selected studies suggest that the elevated frequency of CDH1 mutations in EOGC may be partly explained by specific clinicopathological features (Table 2). Female patients in the EOGC group showed a higher CDH1 mutation rate than males, and most mutated tumors were of diffuse or mixed histology. Intestinal-type tumors, by contrast, were rarely affected. When analysis was restricted to diffuse-type tumors, EOGC still demonstrated a higher mutation frequency than LOGC with the same histology. This difference may primarily reflect the higher proportion of diffuse-type tumors in EOGC, with a possible additional contribution from sex-related factors. However, due to limited stratified reporting across studies, a formal subgroup meta-analysis by sex or histological subtype could not be conducted.
In contrast, no statistically significant differences were observed between EOGC and LOGC in the mutation frequencies of TP53 (OR = 0.86, 95% CI: 0.73–1.02), ARID1A (OR = 0.92, 95% CI: 0.73–1.16), PIK3CA (OR = 0.84, 95% CI: 0.60–1.16), or KRAS (OR = 0.90, 95% CI: 0.57–1.42), with heterogeneity ranging from low to moderate (I^2^ = 23–72%) across comparisons (Fig. 3).
Two key biomarkers relevant to targeted and immune therapies were evaluated across six studies. Heterogeneity was low for both HER2 (I^2^ = 14%, P = 0.3239) and dMMR/MSI-H (I^2^ = 12%, P = 0.3211), allowing the use of fixed-effects models. Meta-analysis showed that HER2 expression was significantly lower in EOGC compared with LOGC (Fig. 4; OR = 0.54, 95% CI: 0.43–0.67; P < 0.0001). Similarly, dMMR/MSI-H status was also less frequent in EOGC (Fig. 4; OR = 0.25, 95% CI: 0.12–0.53; P = 0.0003).Fig. 4Expression of HER2 and immune biomarkers in EOGC vs LOGC. Compared with LOGC, EOGC had lower HER2 and dMMR/MSI-H expression
Three additional molecular markers—E-cadherin, p53, and COX-2—were analyzed across five studies. Among them, p53 expression showed low heterogeneity (I^2^ = 17%) and was significantly lower in EOGC compared to LOGC (OR = 0.56, 95% CI: 0.39–0.82; P = 0.0027), as determined using a fixed-effects model (Fig. 5). In contrast, E-cadherin and COX-2 demonstrated substantial heterogeneity (I^2^ = 94% and 92%, respectively). Random-effects models were applied, and no significant differences were observed between EOGC and LOGC for either E-cadherin (OR = 0.13, 95% CI: 0.01–1.61; P = 0.1126) or COX-2 (OR = 0.29, 95% CI: 0.05–1.55; P = 0.1477) (Fig. 5).Fig. 5Expression of additional molecular markers in EOGC vs LOGC. p53 expression was significantly lower in EOGC. No significant differences were observed for E-cadherin or COX-2 expression between the two groups
Four studies reported comparisons of serum tumor markers between EOGC and LOGC patients. For CEA, the pooled analysis showed significantly lower levels in EOGC than in LOGC (Fig. 6; OR = 0.43, 95% CI: 0.36–0.51; P < 0.0001). Heterogeneity was low (I^2^ = 0%, P = 0.3925), and a fixed-effects model was applied. Similarly, analysis of CA19-9 from three studies demonstrated a significantly lower frequency of elevated levels in EOGC compared to LOGC (Fig. 6; OR = 0.72, 95% CI: 0.60–0.88; P = 0.0009), also with no significant heterogeneity (I^2^ = 0%, P = 0.5422).Fig. 6Serum tumor marker levels in EOGC vs LOGC. EOGC patients had significantly lower rates of elevated CEA and CA19-9 compared to LOGC patients
A leave-one-out sensitivity analysis was conducted to evaluate the robustness of the pooled estimates. All the results remained stable, except for CA19-9 (Supplementary material 4). The results from the funnel plot and Egger's test collectively indicated no significant publication bias (Supplementary material 5).
EOGC has garnered increasing research interest due to its rising incidence and distinctive clinicopathological features [4, 5, 34]. However, its molecular underpinnings remain incompletely understood. Prior studies have reported inconsistent molecular profiles, often limited by small sample sizes, geographic variability, and methodological heterogeneity. These limitations have hindered the establishment of robust molecular classifications and age-tailored therapeutic strategies. To address these gaps, we conducted a comprehensive meta-analysis comparing EOGC and LOGC across multiple molecular domains—including TCGA subtype, common gene mutations, therapeutic biomarkers (e.g., HER2, MSI), and serum-related markers. By quantitatively synthesizing available evidence, our findings clarify the molecular distinctiveness of EOGC and underscore its clinical relevance for precision medicine, particularly in the context of targeted and immune-based therapies.
In our analysis based on TCGA molecular subtypes, EOGC exhibited a significantly higher proportion of the GS subtype and a lower proportion of the CIN subtype compared to LOGC. This distribution aligns with previous observations, such as those reported by Setia et al., who found that the GS subtype was enriched in diffuse-type and signet-ring cell carcinomas—histological features more commonly seen in younger patients. Conversely, the CIN subtype is more frequently associated with intestinal-type tumors, which tend to occur in older individuals [13]. Crucially, GS classification does not imply genomic quiescence, but rather reflects a distinct oncogenic mechanism involving impaired cell adhesion, stromal activation, and epithelial–mesenchymal transition (EMT) [35]. These tumors often harbor CDH1 and RHOA mutations, which disrupt epithelial polarity and drive infiltrative growth, without the widespread chromosomal instability characteristic of CIN-type tumors [35]. Our meta-analysis found no significant difference in the prevalence of the EBV subtype between EOGC and LOGC. This contrasts with earlier studies suggesting higher EBV positivity in younger patients [17], likely attributable to population heterogeneity or methodological differences.
We further analyzed high-frequency mutated genes and found that CDH1 mutations were more prevalent in EOGC, with prior studies reporting rates of 22.2%–42.2% in EOGC versus 11.4%–17.4% in LOGC [8, 13]. CDH1 encodes E-cadherin, a calcium-dependent adhesion protein essential for maintaining epithelial architecture and preventing tumor invasion [36]. In sex-stratified reports, two studies observed a higher prevalence of CDH1 mutations in female than in male EOGC patients, although such data were limited [8, 13]; moreover, most CDH1-mutated tumors were of diffuse or mixed histology, whereas intestinal-type tumors were rarely affected [13, 31]. This pattern suggests that both histology and sex may contribute to the apparent age-associated enrichment of CDH1 alterations in EOGC. Mechanistically, CDH1 loss disrupts adherens junctions and weakens epithelial cohesion, promoting the discohesive infiltrative growth typical of diffuse-type gastric cancer [37, 38]. E-cadherin dysfunction may also activate Wnt signaling via β-catenin nuclear translocation, further supporting invasion [39, 40]. This sex difference raises the possibility of biological modifiers, as estrogen/ERα signaling has been linked to diffuse-type tumor growth and gastric cancer invasiveness, although the causal relationship remains incompletely defined [41, 42]. In addition, CDH1 alterations have been linked to adverse outcomes in young Cho et al. reported shorter survival in early-onset diffuse gastric cancer with CDH1 alterations (hazard ratio 3.4) [8], and Setia et al. similarly observed worse overall survival in EOGC with CDH1 mutations (median 9 vs 22 months) [13]. Thus, CDH1 status may represent an age-relevant prognostic marker and underscores the aggressive biology of EOGC. Clinically, young patients with diffuse-type or signet-ring cell gastric cancer, especially when there is a suggestive family history, should undergo evaluation for hereditary diffuse gastric cancer, including germline CDH1 testing. For carriers of pathogenic or likely pathogenic germline CDH1 variants, NCCN guidance supports management at experienced centers using a multidisciplinary approach, with individualized discussions of risk-reducing gastrectomy versus endoscopic surveillance [43]. While CDH1 has clear implications for hereditary-risk evaluation and prognostication in young patients, treatment-specific outcome data by CDH1 status in EOGC remain sparse.
For targeted- and immunotherapy-related biomarkers, EOGC exhibits significantly lower HER2 expression and dMMR/MSI-H prevalence. HER2, a transmembrane receptor tyrosine kinase, is typically localized to the apical surface of polarized, gland-forming epithelial cells [44]. In gastric cancer, HER2 overexpression is more frequently observed in intestinal-type tumors, which retain cohesive glandular architecture favorable for HER2 membrane localization [45]. In contrast, the non-cohesive growth pattern and lack of gland formation characteristic of diffuse-type tumors—prevalent in EOGC—create an unfavorable structural context for HER2 expression [10]. Frequent CDH1 loss or mutation may further modulate HER2 signaling by destabilizing cell–cell junctions and impairing receptor clustering [46]. Likewise, the reduced MSI-H frequency in EOGC may be primarily attributed to its histological background, as microsatellite instability is more common in intestinal-type cancers, while diffuse-type tumors are typically microsatellite stable [47]. Furthermore, MLH1 methylation, a key epigenetic driver of sporadic MSI-H, is infrequently observed in EOGC, consistent with its lower MSI-H frequency [48, 49]. Notably, despite lower MSI-H rates, PD-L1 expression appears higher in EOGC [17, 50], potentially due to EBV-related tumor biology in younger patients. EBV-positive gastric cancers have been shown to exhibit higher PD-L1 expression [51], and age-related variations in the tumor immune microenvironment may also contribute to PD-L1 regulation independently of MSI status. From a treatment-selection perspective, the lower HER2 positivity and dMMR/MSI-H prevalence in EOGC may reduce the proportion of patients eligible for HER2-targeted therapy or MSI-H–directed immunotherapy, while EBV/PD-L1 status may still aid immunotherapy stratification.
Significant differences in protein and serum biomarker expression were identified between EOGC and LOGC. p53 expression was lower in EOGC despite similar TP53 mutation frequencies. This discrepancy may reflect differences in mutation types or tumor histology. Truncating mutations observed in EOGC can result in unstable p53 protein that is undetectable by immunohistochemistry, while diffuse-type gastric cancers common in younger patients frequently lack p53 nuclear accumulation [33, 52]. Additionally, Helicobacter pylori infection, more prevalent in intestinal-type tumors, has been shown to promote p53 degradation via USF1-mediated pathways [53, 54]. COX-2, previously reported to be downregulated in younger patients, showed no significant age-based difference in our meta-analysis after adjusting for cohort overlap and incorporating broader international data [23, 55]. Regarding serum markers, both CEA and CA19-9 were significantly less frequently elevated in EOGC. This likely reflects underlying tumor biology, as diffuse and poorly differentiated tumors—common in younger patients—tend to secrete lower levels of these markers. Although some heterogeneity in CA19-9 data was noted due to sample size variation, the overall trend remained consistent and was supported by large retrospective cohorts [10].
CLDN18 exhibits distinct molecular alterations in EOGC. A meta-analysis of five studies on CLDN18–ARHGAP fusion genes found these alterations to be more frequent in younger patients, females, and diffuse-type tumors [56]. CLDN18–ARHGAP fusions disrupt tight junctions via Rho GTPase signaling, promoting EMT and tumor invasiveness. Their presence is associated with lymph node metastasis, advanced TNM stage, and poor overall survival, suggesting independent prognostic significance [56]. In addition, CLDN18.2, a splice variant of CLDN18 and a therapeutic target in gastric cancer, has been investigated in relation to age. Liu et al. reported CLDN18.2 positivity in 17% of EOGC and 11.7% of LOGC, without a statistically significant difference [10]. Future studies should report CLDN18.2 expression together with histology and treatment exposure to determine whether CLDN18-directed strategies are similarly applicable in EOGC.
DNA methylation plays a key role in gastric cancer and may differ between EOGC and LOGC. MLH1 promoter hypermethylation, a major cause of microsatellite instability in sporadic GC, occurs more frequently in older patients, suggesting an age-related pattern [49]. In contrast, BRCA1 and EIF4E hypermethylation has been observed in early-onset cases, potentially contributing to DNA repair and translational dysregulation, though comparative data remain limited [37, 57, 58]. A more systematic, age-stratified epigenetic profiling effort is needed to determine whether these methylation differences represent distinct etiologic pathways or simply reflect histology and cohort composition.
In summary, our meta-analysis integrates the available evidence and demonstrates consistent molecular differences between EOGC and LOGC. At the same time, EOGC is heterogeneous, and the observed EOGC–LOGC contrasts should be considered in light of the marked imbalance in histologic subtype distribution between age groups. Because several molecular features assessed here are closely associated with histology, the observed differences may partly reflect tumor composition alongside age. This limitation also reflects incomplete reporting in the primary literature, as many studies did not provide within-histology comparisons, such as analyses restricted to diffuse-type or intestinal-type tumors, or histology-adjusted estimates. Larger, well-characterized cohorts with standardized molecular testing, histology-stratified analyses, and multivariable adjustment are needed to better disentangle age- and histology-related signals and strengthen clinical relevance.
This meta-analysis highlights the distinct molecular landscape of EOGC, characterized by a higher prevalence of CDH1 mutations and the GS subtype, alongside significantly lower rates of HER2 expression, MSI-H status, p53 expression, and elevated serum tumor markers (CEA and CA19-9). These differences underscore the need for age-tailored diagnostic and therapeutic strategies. Molecular profiling—including CDH1 testing—should be considered in the diagnostic workup of younger patients. Furthermore, emerging therapeutic targets such as CLDN18–ARHGAP fusion genes, which are more common in EOGC and associated with aggressive clinicopathological features, may offer new opportunities for personalized treatment in this distinct patient population.
Supplementary Material 1.
Supplementary Material 2.
Supplementary Material 3.
Supplementary Material 4.
Supplementary Material 5.
Supplementary Material 6.
Supplementary Material 7.