Search in the Abstract Database

Abstracts Search 2014

* = Presenting author

P241. Development of a novel prediction model for natural course in Crohn's disease: A Korean multicenter, retrospective cohort study

Y. Park1, J.H. Cheon2, W.H. Kim2, B.D. Ye3, Y.S. Kim4, D.S. Han5, J.S. Kim6, Y.H. Kim7, 1Yonsei University College of Medicine, Internal Medicine, Seoul, Korea, Republic of, 2Yonsei University College of Medicine, Department of Internal Medicine and Institute of Gastroenterology, Seoul, South Korea, 3University of Ulsan College of Medicine, Asan Medical Center, Gastroenterology, Seoul, South Korea, 4Inje University College of Medicine, Internal Medicine, Seoul, Korea, Republic of, 5Hanyang University Guri Hospital, Gastroenterology, Guri, Korea, Republic of, 6Seoul National University College of Medicine, Internal Medicine, Seoul, Korea, Republic of, 7Sungkyunkwan University School of Medicine, Internal Medicine, Seoul, Korea, Republic of

Background

The natural course of Crohn's disease (CD), with continuing relapses and remissions, leads to irreversible intestinal damage. Early adoption of immunomodulators and/or biologics has been proposed in order to prevent bowel damage in patients with CD at high risk of disease progression. However, appropriate selection of patients who might be potential candidates for early intervention remains unanswered. The aim of our study was to identify factors predictive of a natural course and prognosis at diagnosis, and to develop and validate a novel prediction model that can guide patient selection in Korean patients with CD.

Methods

The prognostic model was derived from a retrospective, multi-center longitudinal cohort. The cohort consisted of 1,338 patients from 29 centers diagnosed as having CD between 1982 and 2010. The entire population was randomly split into training and validation sets. The Korean Crohn's Disease Prediction (KCDP) model was developed from the training data set using the Cox proportional hazards model including risk predictors identified from univariate and multivariate analysis, and validated using the separate validation data set.

Results

During the follow-up period of 6,435 patient-years (median 4.6 years), 545 patients (40.7%) underwent surgery. Independent predicting factors at diagnosis significantly associated with surgery in the training set were the presence of granuloma (OR 1.29 [95% CI 1.04–1.58]), jejunal involvement (OR 1.38 [95% CI 1.07–1.78]), and the presence of perianal disease (OR 1.74 [95% CI 1.39–2.18]), stenosis (OR 1.40 [95% CI 1.04–1.88]), and perforation (OR 3.42 [95% CI 2.41–4.84]). These predictors were included in the final model to estimate individual probabilities of surgery at 5 years. The final model was as follows: Risk of OP at 5 years =1–0.664·exp(prognostic index), where prognostic index = −0.002(age − 27.6) + 0.364(jejunum involvement − 0.15) + 0.247(granuloma − 0.32) + 0.649(perianal − 0.43) − 0.722(initial behavior − 0.79). This prognosis of CD could be differentiated in the model using the validation set. The discrimination using the KCDP model in the Korean cohort was acceptable with Herral's c-index of 0.687 at 5 years.

Conclusion

This was the first validated surgery risk prediction model for Korean patients with CD and provides accurate individualized estimates for the probability of surgery using clinical parameters collected at diagnosis. This model may guide appropriate patient selection for early intensive treatment in CD.