Dynamic changes in serum gelsolin combined with clinical parameters to construct a predictive model for mucosal healing in inflammatory bowel disease: A prospective cohort study.

Mucosal healing represents a critical therapeutic endpoint in inflammatory bowel disease (IBD), yet reliable noninvasive predictive biomarkers remain limited. Serum gelsolin (GSN) has emerged as a potential inflammatory biomarker, but its dynamic changes and predictive value for mucosal healing have not been systematically investigated. This study aimed to develop and internally validate a predictive model integrating dynamic GSN changes with clinical parameters for forecasting 6-month mucosal healing in IBD patients. We enrolled 298 IBD patients (182 Crohn disease, 116 ulcerative colitis) from a tertiary referral center and collected baseline and 6-month follow-up data including serum GSN levels, endoscopic scores, and clinical parameters. The primary endpoint was mucosal healing at 6 months (Mayo endoscopic score ≤1 or SES-CD <3). Linear mixed-effects models analyzed GSN longitudinal trajectories, identifying 3 change patterns through growth mixture modeling. Least absolute shrinkage and selection operator regression selected variables from 12 candidates, and random forest algorithms constructed prediction models. Time-dependent receiver operating characteristic curves evaluated predictive performance. Internal validation was performed using bootstrap resampling (1000 iterations). Propensity score matching compared GSN patterns between biologic responders and nonresponders. Restricted cubic splines examined GSN-mucosal healing relationships. Three distinct GSN change patterns emerged: rapid-increase (38.6%), slow-increase (42.3%), and nonresponse (19.1%). The integrated prediction model incorporating baseline GSN, ΔGSN, CRP/GSN ratio, and endoscopic scores demonstrated superior predictive performance (AUC 0.873, 95% CI 0.841 to 0.905) compared to models using clinical parameters alone (AUC 0.742, P <.001). Bootstrap internal validation confirmed model stability (optimism-corrected AUC 0.869 ± 0.018). decision curve analysis (DCA) revealed clinical net benefit across threshold probabilities of 10% to 60%. Among biologic-treated patients, rapid-increase GSN pattern was associated with higher mucosal healing rates (HR 3.24, 95% CI 2.15-4.89, P <.001). Dynamic GSN monitoring combined with clinical parameters shows promising predictive performance for mucosal healing in IBD under internal validation. This integrated approach may enable early identification of treatment responders and inform therapeutic decision-making, though external validation and cost-effectiveness analyses are needed before clinical implementation.