P392 Comprehensive video capsule endoscopy-based monitoring predicts short and long-term risk of disease flares in small bowel Crohn’s disease: A prospective cohort study

Abstract

Background: The optimal monitoring strategy for predicting clinical deterioration in Crohn's disease (CD) patients remains undefined Methods: This was a prospective observational cohort study of patients with stable CD involving the small‐bowel. Patients underwent magnetic resonance enterography (MRE) and ingested patency capsule. If patency was proved, patients were enrolled and underwent video capsule endoscopy (VCE) at baseline and every 6 months thereafter and bio‐markers assessment every three months. CD treatment was unchanged during follow‐up. Patients were followed for two years or until a clinical flare occurred, defined by a CDAI 70‐points rise. Results: Of 89 patients undergoing MRE and patency capsule, 28 were excluded (non‐patent small bowel n = 17, other causes n = 11). The included 61 patients swallowed a total of 231 VCEs over the 2‐year study without complications or retentions. Overall, 17/61 (28%) of patients flared during 24 months. No clinical or demographic baseline parameter predicted a future flare. Baseline CRP modestly discriminated between patients with subsequent flares or not (AUC = 0.73, p < 0.001, hazard ratio 6.7, 95% CI 2.6‐17.4). Baseline VCE Lewis score (LS) >350 for the worst SB tertile more strongly predicted future flares (AUC = 0.79, p < 0.0001, hazard ratio 10.7, 95% CI 3.8‐30.3) and a cumulative LS < 450 performed similarly. Mucosal inflammation in the second tertile of the small‐bowel correlated with future flares better than inflammation in either the first tertile or the distal third tertile. In contrast with VCE, baseline faecal calprotectin was not predictive (AUC = 0.63, p = 0.17, optimal cut‐off value 160μg/g) and MRE indices were only weakly predictive of future flare. However, on a time‐restricted analysis of all 425 serial measurements, the prediction‐accuracy of calprotectin progressively increased from AUC = 0.62 for 24‐months, through 0.67 for 12‐months, 0.76 for 6‐months and up to AUC = 0.81 for three‐months' flare prediction. Time‐restricted analysis for VCE also showed a stronger shortterm 6 months prediction accuracy (AUC = 0.83, p < 0.0001). Patient tolerability of intensive small‐bowel VCE monitoring was excellent. Conclusions: Calprotectin strongly predicts short‐term (3 months) risk of flares in patients with small‐bowel CD. In contrast, VCE accurately predicts both short‐term but also long‐term risk of disease exacerbation, and is safe and tolerable after small‐bowel patency has been established. A worst‐segment LS < 350 (or cumulative LS < 450) may be the clinically relevant target scores for small‐bowel mucosal healing. A randomised controlled trial has been launched to investigate proactive interventional strategy based on this predictive algorithm.