Clinical, Endoscopic, and Histological Differentiations Between Crohn's Disease and Intestinal Tuberculosis

Abstract

The clinical, endoscopic, and histological features of Crohn's disease (CD) and intestinal tuberculosis mimic each other so much that it becomes difficult to differentiate between them. The aim was to find out clinical, endoscopic, and histological predictor features for differentiation between CD and intestinal tuberculosis. We recruited 106 patients, 53 each with CD and intestinal tuberculosis, in this study. The clinical, histological, and endoscopic features were subjected to univariate, bivariate, and multivariate analyses. On the basis of regression coefficients of the final multivariate logistic model, a score to discriminate between CD and intestinal tuberculosis was devised. For the validation of the score, the same model was tested on 20 new patients, each with CD and intestinal tuberculosis. On univariate analysis, although longer duration of disease, chronic diarrhea, blood in stool, perianal disease, extra-intestinal manifestations, involvement of left colon, skip lesions, aphthous ulcers, cobblestoning, longitudinal ulcers, focally enhanced colitis, and microgranulomas were significantly more common in CD, partial intestinal obstruction, constipation, presence of nodular lesions, higher number, and larger granulomas were significantly more common in intestinal tuberculosis. On multivariate analysis, blood in stool (odds ratio (OR) 0.1 (confidence interval (CI) 0.04-0.5)), weight loss (OR 9.8 (CI 2.2-43.9)), histologically focally enhanced colitis (OR 0.1 (CI 0.03-0.5)), and involvement of sigmoid colon (OR 0.07(0.01-0.3)) were independent predictors of intestinal tuberculosis. On the basis of regression coefficients of the final multivariate logistic model, a score that varied from 0.3 to 9.3 was devised. Higher score predicted more likelihood of intestinal tuberculosis. Once the cutoff was set at 5.1, then the sensitivity, specificity, and ability to correctly classify the two diseases were 83.0, 79.2, and 81.1%, respectively. Area under the curve for receiver-operating characteristic (ROC) to assess the ability of these features to discriminate between CD and intestinal tuberculosis was 0.9089. The area under ROC in the validation data set was 89.2% (95% CI 0.79-0.99). With a similar cutoff score of 5.1, sensitivity and specificity in the validation model were 90% (95% CI 66.9-98.2) and 60% (95% CI 36.4-80.0), respectively. Blood in stool, weight loss, focally enhanced colitis, and involvement of the sigmoid colon were the most important features in differentiating CD from intestinal tuberculosis.