W1908 Diagnostic Potential of Volatile Organic Compounds in Inflammatory Bowel Disease

Abstract

Introduction: Crohn's disease (CD) and ulcerative colitis (UC) are diagnosed using invasive investigations which patients find unpleasant. Non-invasive modalities to asses the clinical course of the disease are yet to be adopted. Volatile organic compounds (VOCs) are chemicals which emit in the faeces normally, and understanding changes in their concentration and pattern could provide information about various bowel diseases. Objective: We aim to investigate VOCs as faecal biomarkers in diagnosing and predicting the clinical course of inflammatory bowel disease (IBD). Method: We prospectively studied a group of 205 individuals, 75 patients with active IBD (CD=46, UC=29), 70 with disease in remission (CD =35, UC =35) and 60 healthy controls. Diagnosis was confirmed histologically and disease activity was established using clinical scoring and laboratory indices. Fresh samples were collected from all patients and equal quantity was aliquoted in 10mls vial. For analysis, samples were heated at 60°C for an hour. VOCs were extracted using SPME (solid phase micro-extraction) fibre injected into the headspace for 10 minutes and were analysed by gas chromatography / mass spectrometer. VOCs extracted by this method were identified using NIST library. Results: VOCs extracted in all 5 groups were compared to identify the combination that provided the best discrimination. Univariate analysis with the active CD, UC and the normal controls selected 35 VOCs out of 225 as statistical significant at the 5% level. These were used in a forward stepwise discriminant analysis to construct a predictive model which had reasonable discriminatory power but some degree of overlap between the groups. Two further discriminant models were created using these 35 VOCs comparing CD with normal and UC with normal both of which achieved excellent discrimination. On cross validation using 80:20 split of data arbitrarily on 8 occasions and averaging the results, this approach appears to have excellent stability, the average area under the curve for the CD model is 0.95 test sets and 0.80 for the validation sets, and the UC models give 0.94 and 0.86 respectively. Conclusion: This preliminary data demonstrates the potential of pattern analysis of faecal VOC in patients with IBD may provide a non-invasive method in the diagnosis of IBD.