Magnetic resonance enterography to predict disabling disease in newly diagnosed Crohn's disease: the METRIC-EF multivariable prediction model, multicentre diagnostic inception cohort study.

Background: The ability to predict whether patients with a new diagnosis of Crohn's disease will develop disabling disease is an unmet clinical need. Magnetic resonance enterography is a first-line investigation for Crohn's disease, but its role in prognostication is unknown. Objective(s): To improve prediction of disabling Crohn's disease within 5 years of diagnosis by developing and internally evaluating a multivariable prediction model comprising clinical predictors and adding magnetic resonance enterography scores (Magnetic resonance Enterography Global Score, Simplified Magnetic Resonance Index of Activity and Lémann Index). To estimate the healthcare costs incurred within 5 years of Crohn's disease diagnosis and to explore factors driving costs. Design: A multicentre diagnostic inception cohort. Setting: Nine National Health Service hospitals. Participants: Aged ≥ 16 years with newly diagnosed Crohn's disease. Main outcome measures: Comparative predictive ability of prognostic models, including magnetic resonance enterography scores (Magnetic resonance Enterography Global Score, Simplified Magnetic Resonance Index of Activity and Lémann Index) versus a model based on clinical predictors alone for the development of modified Beaugerie disabling Crohn's disease within 5 years of diagnosis. Statistical analysis: We censored development of modified Beaugerie disabling disease ≤ 90 days from diagnosis, and utilised time-to-event models using Royston-Parmar flexible parametric models. Risk group definitions were prespecified; for risk group definition 1, the high-risk patients were the top 40% with the greatest predicted risk, and the high-risk patients had an absolute risk ≥ 10% for risk group definition 2. The absolute risk cut-off was calculated by sorting patients by predicted risk and using the risk of the eighth (10% of 81) patient who developed modified Beaugerie disabling disease. Results: We studied 194 patients, median age 29, interquartile range 22-44 years. Within 5 years from diagnosis, 42% (81/194) developed modified Beaugerie disabling disease. There was a univariable association between initial need for steroid therapy and developing modified Beaugerie disabling disease [hazard ratio 2.11 (95% confidence interval 1.36 to 3.26)]. Using risk group definition 1, the baseline clinical model had 49% (95% confidence interval 39 to 60) sensitivity and 66% (95% confidence interval 57 to 74) specificity for predicting the development of modified Beaugerie disabling disease. There was no difference in sensitivity and specificity between models incorporating Magnetic resonance Enterography Global Score, Simplified Magnetic Resonance Index of Activity and Lémann Index compared to the baseline clinical model. Using risk group definition 2, the model, including magnetic resonance enterography predictors, had 86% (95% confidence interval 77 to 92) sensitivity and 35% (95% confidence interval 27 to 45) specificity for predicting the development of modified Beaugerie disabling disease. There was no difference in sensitivity between the clinical model and models incorporating Magnetic resonance Enterography Global Score, Simplified Magnetic Resonance Index of Activity and Lémann Index, but specificity was significantly lower for models incorporating Magnetic resonance Enterography Global Score [29% (95% confidence interval 22 to 38)] and Lémann Index [29% (95% confidence interval 22 to 38)]. The mean total 5-year per-patient cost of health care was £24,267 (standard deviation £33,108). Mean 5-year costs were £29,763 (standard deviation £38,278) compared to £20,327 (standard deviation £28,368) for those with and without disabling disease, respectively. The largest contributor to costs was biologic use. Age under 40 years, presence of perianal disease and presence of severe endoscopic disease were associated with higher costs. Limitations: Liège and Montreal criteria for disabling disease could not be studied due to an insufficient event rate. Conclusions: Addition of magnetic resonance enterography scores to a multivariable model comprising existing clinical predictors did not improve prediction of modified Beaugerie disabling disease. Healthcare costs were increased in those aged under 40 years and patients with perianal and severe endoscopic disease. Future work: Testing the predictive ability of magnetic resonance enterography against alternative definitions for disabling Crohn's disease. Trial registration: This trial is registered as ISRCTN76899103. Funding: This award was funded by the National Institute for Health and Care Research (NIHR) Health Technology Assessment programme (NIHR award ref: 15/59/17) and is published in full in