Computerized tomography-based predictive model for differentiation of Crohn's disease from intestinal tuberculosis.

Abstract

Intestinal tuberculosis (ITB) and Crohn's disease (CD) have clinical, radiological, endoscopic, and histological resemblance. There is paucity of literature regarding differentiation of CD and ITB based on radiology using computed tomography (CT). The present study was designed to compare CT features of ITB and CD and develop a predictive model to differentiate ITB and CD. Patients with ITB and CD, who underwent CT enteroclysis/CT enterography/CT abdomen before starting treatment, were recruited. Specific findings were noted by a radiologist who was blinded to the diagnosis. A predictive model was developed based on the features which were significantly different in these diseases. Fifty-four patients with CD and 50 patients with ITB were compared. On univariate analysis, left colonic involvement, ileocecal involvement, long-segment involvement, comb sign, presence of skip lesions, involvement of ≥3 segments and ≥1-cm sized lymph nodes were significantly different between CD and ITB. On multivariate analysis, ileocecal involvement, long-segment involvement and the presence of lymph node ≥1 cm were statistically significant. Based upon the latter three variables, a risk score (with values ranging from 0 to 3) was generated, with scores 0 and 1 having specificity of 100 % and 87 %, respectively, and positive predictive values (PPV) of 100 % and 76 %, respectively, for ITB and scores 2 and 3 having specificity of 68 % and 90 %, respectively, and PPV of 63 % and 80 %, respectively, for CD. A predictive model based on the presence of long-segment involvement, ileocecal involvement and lymph nodes sized ≥1 cm on CT could differentiate ITB and CD with good specificity and PPV.