Development and validation of a novel Endoscopic ulcer Activity Score for Evaluation of Crohn's Disease (EASE-CD) using data from two randomised controlled trials

Abstract

Background: Endoscopy is essential for measuring luminal disease activity in Crohn's disease. Existing indices for evaluating endoscopic Crohn's disease activity are only modestly correlated with clinical disease activity, contain items with ambiguous definitions and poor interobserver reliability, and do not have validated thresholds for defining clinically relevant changes. To address these issues, we conducted a multiphase study to develop and validate a novel endoscopic index for Crohn's disease.

Methods: Paired baseline and post-induction ileocolonoscopy videos from a phase 3b adalimumab trial (NCT00348283) and phase 2 risankizumab trial (NCT02031276) were assessed by multiple central readers using candidate endoscopic items with face validity derived using modified RAND/UCLA appropriateness methods. The four readers were masked to all clinical information, including treatment assignment and patient symptom scores. A total of 112 and 99 paired ileocolonoscopy videos with treatment assignment were available from the adalimumab trial and risankizumab trial, respectively; after the four readers assessed the 112 videos, two readers then reassessed 44 post-treatment videos to assess inter-rater reliability in index development. Items with acceptable inter-rater reliability (with an intraclass correlation coefficient [ICC] of ≥0·41) and responsiveness (win probability [WinP], the probability that a patient receiving active treatment has a better endoscopy score than a patient receiving placebo, of ≥0·56) were included as covariables in regression for model building with internal validation with bootstrapping. External validation was conducted using the paired videos from the risankizumab trial.

Findings: Ten endoscopic items met ICC and WinP thresholds and were considered for novel index development. They included items from the SES-CD (presence and size of ulcers, extent of ulcerated surface, and extent of affected surface) and CDEIS (percentage of surface involved by Crohn's disease, percentage of ulcerated surface, deep ulceration, and superficial ulceration) and exploratory items (presence of Crohn's disease-related lesions, Crohn's disease-related lesion severity, and Crohn's disease-related lesion involvement). The final model, Endoscopic ulcer Activity Score for Evaluating CD (EASE-CD), included the presence and size of ulcerations measured on an ordinal scale of 0 (none), 1 (ulcers <5 mm), 2 (ulcers between 5 mm and 20 mm), and 3 (ulcers >20 mm); the presence or absence of deep ulcerations measured dichotomously, with 0 for absent and 1 for present; and the proportion of ulcerated surface area, expressed as a continuous proportion from 0-1, with each item evaluated and averaged across visualised bowel segments. The total score ranges from 0-100 with higher scores indicating greater endoscopic activity. The r² and optimism adjusted r² of EASE-CD in internal validation were 0·608 and 0·554, and the index was well calibrated with a slope of 0·983. In external validation, the adjusted r² was 0·565 and the calibration slope was 0·997. EASE-CD also demonstrated substantial inter-rater reliability (ICC 0·798 [95% CI 0·711-0·836]) and a large degree of responsiveness (WinP 0·769 [95% CI 0·658-0·851], p<0·001) in external validation. A 10-point reduction in EASE-CD had 82·4% specificity (95% CI 78·6-86·0) and 69·9% sensitivity (63·3-76·4) for endoscopic response, defined by at least 50% reduction in SES-CD or CDEIS. Ulcer-free endoscopic remission results in EASE-CD score of 0.

Interpretation: EASE-CD is a novel, internally and externally validated continuous measure of endoscopic ulcer activity in Crohn's disease that is reliable and responsive to treatment.

Funding: None.