P079 Predicting the risk of acute severe colitis (ASC) at diagnosis of Ulcerative Colitis (UC): external validation
M. Cesarini*1, 2, G. Collins3, A. Ronnblom4, A. Santos5, 6, D. Sjoberg4, M. Parkes7, S. Keshav2, S. Travis2
1Sapienza, University of Rome, Department of Internal Medicine and Medical Specialties, Rome, Italy, Rome, Italy, 2John Radcliffe Hospital, Translational Gastroenterology Unit, Oxford, United Kingdom, 3University of Oxford, Centre for Statistics in Medicine, Oxford, United Kingdom, 4Uppsala University, Department of Medical Sciences, Uppsala, Sweden, 5Hospital Amato Lusitano, Gastroenterology Unit, Castelo Branco, Portugal, 6Cambridge University hospitals, Gastroenterology Unit, Cambridge, United Kingdom, 7University of Cambridge, Gastroenterology Unit, Cambridge, United Kingdom
A simple prognostic index applied at diagnosis of UC appears to predict the likelihood of patients developing ASC and therefore those at high risk of colectomy. This study evaluated the index in two independent cohorts of patients
An episode of ASC was defined by hospital admission with bloody diarrhoea >6/d and 1 or more Truelove & Witts' criteria. The index, calculated by the sum of 1 point each for extensive disease, CRP >10mg/L and haemoglobin <12.1g/dL (F) or < 13.8g/dL (M) at diagnosis to give a total score of 0/3 to 3/3, was applied retrospectively to two external validation cohorts (patients diagnosed and followed up for >3y in Cambridge (UK) and Uppsala (Sweden), excluding presentation with ASC within a month of diagnosis). Performance characteristics of the prognostic index were verified by discrimination and calibration.
Table 1 shows the characteristics of development and validation cohorts. Patients in Oxford (n = 111) and Cambridge (n = 96) had much higher rates of ASC during a median 3 years follow up (31% and 26% respectively) than patients in Uppsala (n = 296, 6% ASC). Nevertheless, the mean predicted risk of ASC using the index was similar in the three cohorts: 73%, 72% and 72% (Oxford, Cambridge, Uppsala). Of those who scored 3/3 at diagnosis, 8/11 (O: 73%), 18/18 (C: 100%) and 13/14 (U: 93%) subsequently developed ASC. The ability of the index to discriminate (c-index, where 1.0 = perfect discrimination) was 0.81 (Oxford), 0.95 (Cambridge), 0.97 (Uppsala).
The distribution of the predicted risks (from score = 0/3 to 3/3) was similar across each of the cohorts, indicating good calibration and the ability of the model to identify those at low or high risk of developing ASC. A nomogram allows individual risk to be estimated.
Despite geographic and demographic differences among three cohorts, this simple index has been verified as a reliable tool to predict ASC within 3 years from diagnosis. Patients with a score of 3/3 at diagnosis may merit early immunomodulator therapy.