Bedside analysis of volatile organic compounds in exhaled human breath as non-invasive diagnostic tool for colon diseases.

BACKGROUND: Measurements of volatile organic compounds (VOCs) in exhaled air represent a promising noninvasive biomarker modality for gastrointestinal diseases and has potential to enable point-of-care diagnostics. AIM: To evaluate the feasibility of differential ion mobility spectrometry (DMS) combined with machine learning to distinguish patients with colon diseases from healthy controls based on VOCs analysis in breath. METHODS: Breath samples were collected from 50 participants, including patients with inflammatory bowel disease (IBD), diverticulosis and healthy controls. Exhaled VOC profiles were acquired using a bedside differential mobility spectrometry (DMS) platform and subsequently analyzed with machine learning models. RESULTS: Based on machine-learning analysis there was moderate discrimination between patients with colon disease and healthy controls (best ROC-AUC 0.784). In exploratory disease-specific analyses the highest discriminative performance for ulcerative colitis (ROC-AUC 0.862) has been noted, whereas discrimination between Crohn's disease and diverticular disease was less accurate. In patient-level permutation testing it has been confirmed that the observed diagnostic performance was unlikely to have occurred by chance (permutation p = 0.005). CONCLUSION: DMS-based breath analysis combined with machine learning appears promising noninvasive diagnostic approach in patients with colon diseases.