Search in the Abstract Database

Abstracts Search 2014

* = Presenting author

DOP049. Predicting the individal risk of acute severe colitis at diagnosis

M. Cesarini1, G. Collins2, L.M. Wang3, N. Fernandopulle1, C. Corte1, J. Brooks1, S. Keshav1, S. Travis1, 1University of Oxford, Translational Gastroenterology Unit, Oxford, United Kingdom, 2University of Oxford, Centre for Statistics in Medicine, Oxford, United Kingdom, 3University of Oxford, Department of Cellular Pathology, Oxford, United Kingdom

Background

Acute Severe Colitis (ASC) occurs in 24.8% of patients with ulcerative colitis (UC) and can be life-threatening (overall mortality 1.2%). Furthermore, the colectomy rate increases from 3.4% to 39.8% in patients with at least one admission for ASC, making it the commonest reason for surgery. Risk factors to predict ASC have never been studied systematically. Our aim was to use disease characteristics at diagnosis to predict the risk of developing ASC within 3 years.

Methods

Clinical and biochemical data for patients diagnosed with UC in Oxford from 2007 to 2010 were analysed retrospectively using multivariable logistic regression modeling. We analysed age, gender, extent of disease (Montreal E1, E2, or E3), C-reactive protein (CRP, mg/L), haemoglobin (g/dL), endoscopic appearance (mild, moderate, severe as judged by the endoscopist), and histopathology (mild, moderate, or severe, Truelove & Richards' index). Performance of the model was assessed by discrimination and calibration. Internal validation was performed by bootstrapping and a simplified nomogram developed for clinical application.

Results

111 patients met inclusion criteria (median follow-up 46m, range 36–60). 34/111 (31%) were admitted with ASC at least once within 3 years of diagnosis (median 14m, range 1–29). Patients with ASC at initial presentation were excluded. The final prediction model included three factors: extent of disease, CRP and haemoglobin. The model had good discrimination (c-index = 0.77). Bootstrapping for internal validation showed good calibration. A binary score was assigned as follows: 0 for extent E1 or E2, 1 for E3, 0 for CRP ≤ 10 mg/L, 1 for CRP > 10 and 0 for normal haemoglobin, 1 for low (13.5 g/dL for men, or 12.1 g/dL for women). At diagnosis, 39 patients scored 0 points, 32 scored 1, 29 scored 2, and 11 scored 3. The percent developing ASC in the follow-up period, predicted risk and interquartile range based on modeling are shown in the table.

Table: Simplified model for prediction of ASC
Prognostic Score 123
n39322911
Number with ASC (%)4 (10%)6 (19%)16 (55%)8 (73%)
Predicted Risk
 Mean11%27%46%73%
 IQR8 to 13%18 to 28%32 to 52%61 to 82%

Conclusion

This model predicts the likelihood of ASC within 3 years of the diagnosis. Prospective validation of the score is needed.