Predicting Endoscopic Improvement in Ulcerative Colitis Using the Ulcerative Colitis Severity Index.

Introduction: We developed and internally validated a prognostic scoring index for ulcerative colitis (UC) patients that includes baseline patient-reported outcomes (PROs), biomarkers, endoscopy, and histology for achieving 1-year endoscopic improvement (EI).

Methods: This post hoc analysis included 644 patients treated with ustekinumab induction therapy. Data were randomly split to obtain a 70% training and 30% testing cohort. Multivariate analyses assessed baseline variables and those with P < .05 were assigned weights based on their relative prognostic value from logistic regression modeling for predicting 1-year EI (Mayo endoscopic score ≤1). A cutoff was obtained by calculating the maximum Youden index and validated in the testing cohort.

Results: Prior biologic failure, albumin <40 g/L, C-reactive protein >5 mg/L, Mayo stool frequency subscore, endoscopic erosions/ulcerations, and chronic histologic structural/architectural changes demonstrated significant associations with 1-year EI and were included in the final model. The Ulcerative Colitis Severity Index (UCSI) had acceptable discriminative ability for 1-year EI in the training (area under the curve [AUC], 0.78; 95% confidence interval, 0.70-0.86) and testing cohort (AUC, 0.76; 95% CI, 0.68-0.85). Compared with the UCSI, the Mayo Clinic score demonstrated poor accuracy (AUC, 0.49; 95% CI, 0.40-0.58) for predicting 1-year EI (P = .0006). The UCSI predicted 1-year endoscopic healing (Mayo endoscopic score = 0), clinical remission (total Mayo Clinic score ≤2 and no subscore >1), partial Mayo score remission <2, and 2-item Patient-Reported Outcome score (Mayo stool frequency and rectal bleeding subscore = 0) with significantly greater accuracy compared with the Mayo Clinic score.

Discussion: The UCSI is an internally validated prognostic scoring tool that accurately predicts 1-year EI at baseline among moderate-to-severe UC patients initiating therapy. Further validation with additional datasets is needed.