Meta-Analytic Bayesian Model For Differentiating Intestinal Tuberculosis from Crohn's Disease.

Area of Science:

• Gastroenterology
• Infectious Diseases
• Radiology

Background:

• Distinguishing intestinal tuberculosis (ITB) from Crohn's disease (CD) presents a significant clinical challenge.
• Various clinical, endoscopic, imaging, and laboratory findings have been reported to aid differentiation, but their predictive power requires quantification.
• A need exists for a comprehensive model to accurately estimate the probability of ITB versus CD.

Purpose of the Study:

• To estimate the predictive power of various findings in differentiating ITB from CD.
• To construct a comprehensive Bayesian prediction model for estimating the probability of ITB versus CD.
• To provide a tool for clinical application in differentiating these two conditions.

Main Methods:

• A systematic literature search was performed across MEDLINE, PUBMED, and EMBASE up to September 2015.
• Fifty-five meta-analyses were conducted to determine the odds ratio for each predictive finding.
• A Bayesian prediction model was developed, incorporating significant findings and local pretest probability.

Main Results:

• Thirty-eight studies involving 2,117 CD and 1,589 ITB patients were analyzed.
• Specific findings significantly favored either CD (e.g., male gender, perianal disease, cobblestone appearance) or ITB (e.g., fever, night sweats, ascites, positive interferon-γ release assay).
• The validated model demonstrated high diagnostic performance for ITB with 90.9% sensitivity, 92.6% specificity, and 91.8% accuracy.

Conclusions:

• A Bayesian model integrating meta-analytic results provides a robust method for estimating the probability of ITB versus CD.
• The model is calibrated to local disease prevalence, enhancing its clinical applicability.
• A publicly available web application facilitates the use of this model in patient management.