Incremental Diagnostic Value of the Serum Pepsinogen Ratio for Non-Invasive Prediction of Histological Gastric Atrophy: An Internally Validated Clinical–Infectious Model

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Background: Histological gastric atrophy represents a critical step in the cascade leading to intestinal-type gastric cancer. Although upper gastrointestinal endoscopy with systematic biopsy remains the reference standard for detecting atrophic changes, its invasiveness, cost and limited suitability for repeated population-level assessment highlight the need for reliable non-invasive biomarkers.The serum pepsinogen I-to-II ratio (PGR) reflects the functional integrity of gastric glands and may decrease following loss of corpus glandular tissue. However, the clinical value of PGR should not be assessed solely through comparisons between histological groups. Its incremental contribution beyond routinely available demographic and infectious variables, particularly Helicobacter pylori and Epstein–Barr virus status, remains insufficiently defined.Aim: To evaluate the diagnostic performance of serum PGR for histologically confirmed gastric atrophy and to determine whether incorporation of PGR improves a clinical–infectious prediction model.Methods: This prospective observational study included 120 consecutive patients who underwent upper gastrointestinal endoscopy with targeted gastric biopsy sampling.Histological assessment included gastric atrophy, intestinal metaplasia and epithelial dysplasia. The severity of gastritis was staged according to the Operative Link on Gastritis Assessment system. Serum pepsinogen I and II concentrations were measured, and PGR was calculated as PGI divided by PGII.H. pylori and EBV status were assessed by polymerase chain reaction. Demographic variables and major comorbidities were also recorded.PGR distributions were compared between histological groups using the Kruskal–Wallis test. Associations with age and OLGA stage were evaluated using Spearman correlation. Mann–Whitney U testing was used to compare PGR according to infectious status.Histological atrophy versus non-atrophic histology was used as the binary outcome. Univariable logistic regression was followed by a prespecified multivariable model incorporating age, sex, H. pylori status, EBV status and PGR. Model discrimination was evaluated using receiver operating characteristic analysis. Bootstrap resampling was used to estimate confidence intervals and optimism-corrected discrimination. The optimal PGR threshold was determined using the Youden index. Model accuracy was additionally evaluated using the Brier score.Results: Histological examination identified gastric atrophy in 32 patients (26.7%), intestinal metaplasia in 53 patients (44.2%) and epithelial dysplasia in 35 patients (29.2%).PGR differed significantly across the three histological groups (Kruskal–Wallis H = 9.77; p = 0.0076). Median PGR was lowest in patients with gastric atrophy: atrophy: 4.87, interquartile range 4.16–5.42; intestinal metaplasia: 6.53, interquartile range 4.50–8.35; epithelial dysplasia: 6.27, interquartile range 5.24–7.41.PGR was not significantly correlated with age (Spearman ρ = 0.086; p = 0.349) or OLGA stage (ρ = −0.007; p = 0.939).Patients with H. pylori positivity had significantly lower PGR values than H. pylori-negative patients: median 5.15 versus 6.98 (p = 0.00014). Conversely, EBV-positive patients had higher PGR values than EBV-negative patients: median 6.53 versus 5.42 (p = 0.0014). VCA IgG status was not associated with PGR (p = 0.954).In univariable logistic regression, each one-unit increase in PGR was associated with a 19.6% reduction in the odds of gastric atrophy: OR = 0.80; 95% CI 0.67–0.97; p = 0.025.In the prespecified multivariable model, PGR remained independently associated with histological atrophy: adjusted OR = 0.77 per unit increase; 95% CI 0.62–0.96; p = 0.020.EBV positivity was also independently associated with atrophy: adjusted OR = 3.07; 95% CI 1.20–7.87; p = 0.020.Age, sex and H. pylori status were not independently associated with atrophy after adjustment.PGR alone showed moderate discrimination for gastric atrophy, with an AUC of 0.684 (bootstrap 95% CI 0.573–0.788). The clinical–infectious model without PGR had an AUC of 0.644 (95% CI 0.566–0.787). Addition of PGR increased the apparent AUC to 0.732 (95% CI 0.648–0.871).The optimal data-derived PGR threshold was ≤5.56, providing: sensitivity: 81.3%; specificity: 68.2%; positive predictive value: 48.1%; negative predictive value: 90.9%.The combined model had a Brier score of 0.171. After bootstrap correction for model optimism, the estimated AUC was 0.669, indicating moderate internally validated discrimination.Conclusion: Serum PGR is independently associated with histologically confirmed gastric atrophy and provides additional diagnostic information beyond demographic and infectious variables.A PGR threshold of ≤5.56 demonstrated high sensitivity and a high negative predictive value, suggesting that PGR may be particularly useful as a non-invasive rule-out marker for gastric atrophy. Integration of PGR into a clinical–infectious model improved apparent discrimination compared with clinical and infectious variables alone.However, bootstrap correction demonstrated that the model’s internally validated performance was more modest than its apparent performance. These findings support PGR as a promising component of non-invasive gastric risk stratification, while emphasizing the need for external validation before implementation as a stand-alone screening or endoscopy-triage tool.