Non-Invasive Biomarker Monitoring in UC on Biologics

Design: Single-centre, prospective, longitudinal observational cohort study at University Hospitals Coventry and Warwickshire NHS Trust (UHCW), UK. No intervention is assigned; participants continue standard-care maintenance intravenous biologic therapy. Population: Adults (≥18 years) with ulcerative colitis (UC), confirmed by clinical, endoscopic and histological criteria per ECCO guidelines, established on maintenance intravenous infliximab or vedolizumab and attending the UHCW medical day unit. Patients in the induction phase are not eligible. Target enrolment is 99 participants. Procedures: Participants are assessed at four consecutive routine infusion visits (every 4, 6 or 8 weeks; follow-up 12-24 weeks). At each visit, pre-infusion blood is taken for biologic trough concentration, anti-drug antibodies, C-reactive protein (CRP), full blood count, cytokines (IL-6, IL-10, IL-23) and DNA methylation. Stool and urine are collected at home before the visit. Stool is used for faecal calprotectin (FCP), VOC analysis and whole-genome shotgun metagenomic sequencing; urine is used for VOC analysis. Research samples are stored at -80 °C at the Arden Tissue Bank, UHCW, and VOC analysis is performed at the University of Warwick. Classification: Trough concentrations are classified as sub-therapeutic at ≤1 µg/mL for infliximab and \<12.7 µg/mL for vedolizumab, according to the reporting laboratories' thresholds, fixed before any VOC analysis. Remission is defined as FCP \<250 µg/g and CRP ≤10 mg/L; relapse as FCP ≥250 µg/g and/or CRP \>10 mg/L, or corticosteroid rescue, biologic dose escalation, or hospitalisation for UC flare (STRIDE-II). Analysis: The primary analysis uses one baseline sample per participant. A pre-specified set of supervised machine-learning classifiers will be evaluated using nested cross-validation stratified by trough status and biologic agent, with feature selection and tuning confined to training folds. The primary outcome is cross-validated AUROC (95% CI); the hypothesis is supported if the lower 95% CI bound exceeds 0.5, with AUROC ≥0.70 considered acceptable. The urinary VOC analysis is tested in fixed sequence after the primary analysis. False discovery rate correction is applied to multi-VOC association analyses. Repeated-visit analyses use participant-grouped cross-validation. Sample size: 77 participants (23 sub-therapeutic, 54 therapeutic) provide 80% power at two-sided α=0.05 to detect AUROC 0.70 versus 0.50 (Hanley-McNeil), assuming 30% sub-therapeutic. Allowing 20% attrition, 99 participants will be recruited.