Development and validation of a non-invasive biomarker-based model to identify endoscopic recurrences of Crohn’s disease

BACKGROUND: Endoscopic recurrence is common in postoperative patients with Crohn’s disease (CD). Monitoring endoscopic recurrence is important for selecting an appropriate treatment to prevent the development of postoperative disease. The aim of this study was to develop and validate a diagnostic model to identify endoscopic recurrence. METHODS: This was a retrospective cohort study recruiting postoperative CD patients who underwent endoscopy at the First Affiliated Hospital, Sun Yat-sen University from January 2016 to June 2020. Endoscopic recurrence was defined as Rutgeerts score > i1. Thirty non-invasive biomarkers, including C-reactive protein, erythrocyte sedimentation rate, vitamin D, complete blood count, and biochemical blood indices, were used as candidate predictors to build a multivariate logistic regression diagnostic model. The predictive ability of the diagnostic models was assessed by receiving the area under the characteristic curve (AUC) and calibration plots, and internal validation was performed by the bootstrap method. RESULTS: Two hundred and nineteen eligible patients were included in this study, and 135 (61.6%) patients had a postoperative endoscopic recurrence. The final diagnostic model included eight biomarkers with an AUC (95% confidence interval (CI)) of 0.796 (0.737–0.855) to identify endoscopic recurrence. The AUC, sensitivity, and specificity of this diagnostic model were 0.781 (0.780–0.782), 0.647 (0.643–0.651) and 0.811 (0.807–0.815), respectively, by internal validation. In addition, the diagnostic model exhibited good calibrability with calibration slope, calibration-in-the-large (‘mean calibration’) and Brier scores of 1.00, 0.00, and 0.175, respectively. CONCLUSION: This non-invasive biomarker-based diagnostic model has an excellent ability to identify endoscopic recurrence in patients with CD. Application of the model to clinical practice to monitor postoperative patients may be helpful for patient management.