Development of a tissue-based risk prediction model combining histology and barrier biomarkers for clinical relapse in ulcerative colitis with endoscopic healing.

Background: Although endoscopic remission is a primary therapeutic target in ulcerative colitis (UC), a meaningful proportion of patients with endoscopic healing still experience clinical relapse (CR). Histologic activity and epithelial barrier dysfunction may help explain this residual risk. Objectives: To develop and internally assess an exploratory tissue-based risk prediction model combining histologic activity with epithelial barrier and immunoregulatory markers for relapse risk stratification in UC patients with endoscopic healing. Design: Multicenter prospective observational cohort study with internal model assessment. Methods: Consecutive UC patients with confirmed endoscopic remission (Mayo Endoscopic Subscore 0 or 1) were followed for 12 months. Baseline colonic biopsies were assessed using the Nancy Histological Index (NHI) and immunohistochemical quantification of vitamin D receptor (VDR), Claudin-2, and mucin-2 (MUC-2). A prespecified four-marker logistic regression model was developed using these tissue-based predictors. Model performance was described by receiver operating characteristic analysis, calibration assessment, 1000-sample bootstrap internal validation, and exploratory therapeutic subgroup analyses. Results: Among 145 patients, 36 (24.8%) experienced CR within 12 months. In the final four-marker model, higher Claudin-2 and NHI were associated with increased relapse risk, whereas higher VDR and MUC-2 expression were protective. The model showed high apparent discrimination in the development cohort (area under the curve (AUC) = 0.968, 95% confidence interval: 0.943-0.992). Bootstrap internal validation yielded an optimism-corrected AUC of 0.957, Brier score of 0.075, calibration intercept of -0.067, and calibration slope of 0.827. A data-derived Youden cut-off of 0.379 classified 39 patients as higher risk, of whom 32 relapsed, and 106 patients as lower risk, of whom 4 relapsed. These estimates reflect internal model performance and should be interpreted cautiously because of the limited number of events and absence of external validation. Conclusion: A tissue-based model integrating Claudin-2, VDR, MUC-2, and NHI may help identify UC patients with different relapse risks despite endoscopic healing. These findings are exploratory and hypothesis-generating, and external validation with standardized biomarker quantification is required before clinical implementation.