Objectives: This study aims to find out whether volatile organic compounds (VOCs) from exhaled breath differ significantly between patients with schizophrenia and healthy controls and whether it might be possible to create an algorithm that can predict the likelihood of suffering from schizophrenia.
Methods: To test this theory, a group of patients with clinically diagnosed acute schizophrenia as well as a healthy comparison group has been investigated, which have given breath samples during awakening response right after awakening, after 30 min and after 60 min. The VOCs were measured using Proton-Transfer-Reaction Mass Spectrometry.
Results: By applying bootstrap with mixed model analysis (n = 1000), we detected 10 signatures (m/z 39, 40, 59, 60, 69, 70, 74, 85, 88 and 90) showing reduced concentration in patients with schizophrenia compared to healthy controls. These could safely discriminate patients and controls and were not influenced by smoking. Logistic regression forward method achieved an area under the receiver operating characteristic curve (AUC) of 0.91 and an accuracy of 82% and a machine learning approach with bartMachine an AUC of 0.96 and an accuracy of 91%.
Conclusion: Breath gas analysis is easy to apply, well tolerated and seems to be a promising candidate for further studies on diagnostic and predictive clinical utility.
Keywords: Schizophrenia; breath gas; classification; clinical utility; volatile organic compounds.