Search in the Abstract Database

Search Abstracts 2016

* = Presenting author

P264 Meta-analytic Bayesian model for differentiating intestinal tuberculosis from Crohn’s disease utilising clinical, endoscopic, and cross-sectional imaging findings, and the interferon-gamma releasing assay

J. Limsrivilai1, 2, A. Shreiner2, P. D. Higgins*3

1Mahidol University, Bangkok, Thailand, 2University of Michigan, Internal Medicine - Gastroenterology, Ann Arbor, Michigan, United States, 3University of Michigan, Internal Medicine - Gastroenterology, Ann Arbor, Michigan, United States

Background

Distinguishing intestinal tuberculosis (ITB) from Crohn’s disease (CD) is problematic. Many studies have found clinical, laboratory, and imaging findings that help to differentiate these 2 diseases. Twenty-eight meta-analyses were performed to estimate the positive likelihood ratio (+LR) of each significant predictive finding and build a Bayesian model for differentiating ITB from CD.

Methods

A systematic literature search was conducted in MEDLINE, PubMed, and EMBASE from inception until September 2015. Studies differentiating ITB from CD by clinical characteristics, colonoscopy, pathology, computed tomography enterography (CTE), and laboratory testing were eligible. Random-effects models were used to combine estimates from all studies. The diagnostic odds ratio (OR) and 95% confidence interval (95% CI) for CD was calculated for each finding. The findings with low to moderate heterogeneity (I2 < 50%) and +LR ≥ 2 for either CD or ITB were included in the model.

Results

Thirty-six studies involving 3 646 patients (2 088 CD; 1 558 ITB) were included. The OR, 95%CI, I2, and +LR of each finding are shown in Table 1.

Table 1 Predictors of intestinal TB vs Crohn’s disease

The findings with high OR and lower bound of 95% CI > 1 favoured CD, whereas those with low OR and upper bound of 95% CI < 1 favoured ITB. After excluding the findings with high heterogeneity and +LR < 2, the model included the presence of perianal disease, extraintestinal manifestations, lung involvement, ascites, longitudinal ulcers, cobblestone appearance, mucosal bridging, transverse ulcers, patulous IC valve, fibro-fatty proliferation, short segmental involvement, intestinal wall stratification, and positive interferon-gamma releasing assay

Conclusion

This Bayesian model can incorporate the local pre-test probability and diagnostic LRs to produce estimates of the probability of TB and CD that are calibrated to local prevalence. We have developed and published a web app that allows use and testing of this model, which is needed for external validation. If validated, this model can guide treatment of patients with indeterminate ulceration of the intestine.