The mechanism and influencing factors of missed diagnosis risk in new gastric cancer screening scoring systems for low-risk populations: A review

Suboptimal Variable Selection and Weighting: Models are predominantly developed from high-risk population data, often overlooking risk factors specific to or more prevalent in low-risk cohorts (e.g., specific genetic backgrounds, atypical lifestyles). Insufficient weight is assigned to indicators of occult risk (e.g., extensive intestinal metaplasia)

Data and Sample Bias: Training datasets are often derived from high-incidence regions or enriched with high-risk individuals, leading to underrepresentation of low-risk populations. This compromises the model’s generalizability and discriminative power for low-risk groups

Inflexible Risk Score Thresholds: Thresholds are often set high to maximize specificity or cost-effectiveness, inadvertently reducing sensitivity and misclassifying individuals with “occult high-risk” features into the low-risk category

Pathological and Molecular Heterogeneity: Gastric cancers arising in low-risk individuals may belong to well-differentiated, slow-growing subtypes (e.g., some intestinal-type) or possess distinct molecular phenotypes (e.g., HER2-negative, microsatellite stable), which behave differently from the typical high-risk GC models

Insidious Early Progression Mechanisms: Some tumors exhibit “skip” progression or early peritoneal micrometastasis, disseminating while the primary lesion remains minuscule and endoscopically inconspicuous

Immunoediting and Immune Evasion: Early-stage tumor cells can sculpt an immunosuppressive microenvironment (e.g., via PD-L1 upregulation, Treg recruitment) to achieve “immune stealth,” evading immune surveillance and enabling prolonged latency before clinical presentation

Limitations of Conventional Serological Biomarkers: Serum pepsinogen (PG) levels may remain normal in cases of atrophy confined to the gastric antrum. Gastrin-17 (G-17) levels are influenced by gastric acid status and the location of atrophy, resulting in overall low sensitivity in low-risk populations

Immuturity of Novel Biomarkers: Although promising, novel biomarkers like ctDNA, miRNAs, and metabolomic profiles lack validation in large-scale screening studies of low-risk populations. Challenges remain in standardization, cost, and assay robustness

Endoscopic Technique Variability and Blind Spots: Early gastric cancer presents with subtle endoscopic features (e.g., slight redness, discoloration, irregular mucosal patterns) easily missed by less-experienced endoscopists or those using non-high-definition equipment. Anatomical blind spots (e.g., high gastric body, posterior wall of cardia) are common

Pathological Diagnostic Limitations: Biopsy sampling is inherently random and may miss focal lesions or areas of dysplasia. Pathological interpretation has inherent subjectivity, potentially leading to undergrading of low-grade lesions

Provider Awareness and Training Gaps: Insufficient vigilance regarding the potential risk in low-risk populations, unfamiliarity with the endoscopic spectrum of early GC, and suboptimal management of follow-up adherence contribute to missed opportunities

Inadequate Screening Intervals and Follow-up Mechanisms: Excessively long screening intervals (e.g., 5–10 years) for low-risk groups, coupled with a lack of systematic re-assessment and recall systems, lead to interval cancer development

Lack of Personalized Screening Strategies: A “one-size-fits-all” approach regarding starting age and modality fails to account for individual genetic risk and dynamic risk changes (e.g., new H. pylori infection)

Ineffective Multidisciplinary Collaboration (MDT): Insufficient information sharing and a lack of regular case discussions between gastroenterology, pathology, and radiology departments can lead to diagnostic errors in complex or borderline cases

Incorporation of multi-omics biomarkers (e.g., genomic, proteomic)

Implementation of dynamic, adjustable risk thresholds

Integration of longitudinal data for evolving risk assessment

Combining high-definition endoscopy with AI-based image analysis

Supplementing with liquid biopsy (e.g., ctDNA, miRNA) for non-invasive monitoring

Utilizing radiomics from CT/MRI for complementary information

Defining individualized screening initiation age and intervals based on comprehensive risk profiles

Tailoring screening modalities (e.g., endoscopy vs. non-invasive tests) to individual risk

Establishing integrated workflows among gastroenterology, pathology, radiology, and oncology

Utilizing AI platforms to facilitate data sharing and consensus decision-making

Mandatory standardized training and certification for endoscopists in early cancer detection

Widespread adoption of AI-based quality control systems to reduce procedural variability and oversight