Biomarkers in Gastric Cancer: Blood Biomarkers, Liquid Biopsy, Artificial Intelligence, and Risk-Stratified Care

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Background: Gastric cancer is often diagnosed in advanced stages for patients in countries without organized screening processes. Non-invasive biomarkers have been proposed to risk-stratify and identify those who should undergo endoscopy. Methods: We searched PubMed/MEDLINE, Scopus, and Google Scholar for studies published between January 2006 and May 2026 reporting diagnostic accuracy or health–economic outcomes for gastric cancer detection in asymptomatic adults or patients with precursor mucosal lesions. Twenty-five studies met eligibility criteria and were synthesized narratively, with reporting informed by PRISMA-DTA and Ferrari guidance. Results: Systemic inflammatory indices derived from routine blood counts are inexpensive but non-specific, with individual areas under the receiver operating characteristic curve (AUC) between 0.65 and 0.83. Serological markers of mucosal atrophy have a clearer biological basis and carry the review’s only large Western validation dataset but preferentially detect corpus-predominant disease. Liquid biopsy platforms report the highest accuracy, up to an AUC of 0.968, though sensitivity falls to 62.4% for precancerous lesions and to 44% for stage I disease on some platforms. Magnifying narrow-band imaging reported the highest accuracy among the endoscopic modalities reviewed, with an AUC of 0.97. In low-incidence Western settings, one-time serum pepsinogen screening with endoscopy reserved for positive results is cost-effective at USD 4913 per QALY gained. Conclusions: Reported accuracy is encouraging across all four classes of gastric cancer markers but derives mainly from retrospective East Asian case–control cohorts and declines at the earliest stages of disease. The evidence supports using these markers to triage toward endoscopy rather than to diagnose directly. Validation in Western populations remains a gap.