Middle small-bowel segment Lewis score may predict long-term outcomes among patients with quiescent Crohn’s disease

BACKGROUND: Video capsule endoscopy (VCE) has been proven to accurately diagnose small-bowel inflammation and predict flares among patients with quiescent Crohn’s disease (CD). However, data regarding its predictive role in this population over an extended follow-up are scarce. OBJECTIVES: To predict clinical exacerbation and to assess the yield of Lewis score in identifying CD patients with future clinical exacerbation during an extended follow-up (>24 months). DESIGN: A post hoc analysis study. METHODS: Adult patients with quiescent small-bowel CD who were followed with VCE, inflammatory biomarkers and magnetic resonance enterography in a prospective study (between 2013 and 2018). We extracted extended clinical data (up to April 2022). The primary composite outcome (i.e. clinical exacerbation) was defined as intestinal surgery, endoscopic dilation, CD-related admission, corticosteroid administration, or biological/immunomodulator treatment change during follow-up. RESULTS: Of the 61 patients in the study [median age 29 (24–37) years, male 57.4%, biologic treatment 46.7%], 18 patients met the primary outcome during an extended follow-up [median 58.0 (34.5–93.0) months]. On univariable analysis, complicated [hazard ratio (HR) 7.348, p = 0.002] and stricturing disease phenotype (HR 5.305, p = 0.001) were associated with higher risk for clinical exacerbation during follow-up. A baseline VCE middle small-bowel segment Lewis score (midLS) ⩾ 135 identified patients with future exacerbation [AUC (area under the curve) 0.767, 95% confidence interval (CI) 0.633–0.902, p = 0.001, HR 6.317, 93% negative predictive value], whereas the AUC of the conventional Lewis score was 0.734 (95% CI: 0.589–0.879, p = 0.004). Sensitivity analysis restricted to patients with either complicated (n = 34) or stricturing (n = 26) disease phenotype revealed that midLS still predicted clinical exacerbation during follow-up (AUC 0.747/0.753, respectively), in these patients. CONCLUSION: MidLS predicts treatment failure in quiescent CD patients (median follow-up of 5 years) independently of disease phenotype.