Exhaled volatile organic compounds analyzed via GC-IMS as non-invasive biomarkers for severity stratification and prognosis prediction in critical illness.
TL;DR
While single-modal VOC profiling has limited prognostic performance due to the heterogeneity of critical illness outcomes, the distinct differential signatures identified here provide a solid foundation for future research.
Abstract
Background
Exhaled volatile organic compound (VOC) profiling is a promising non-invasive tool for disease diagnosis, but its clinical value in general critical illness remains underexplored. This study investigated whether exhaled breath VOC signatures can distinguish critically ill patients from healthy controls, stratify disease severity, and predict short-term prognosis in a heterogeneous intensive care unit (ICU) population.
Methods
We conducted a prospective observational study enrolling 278 consecutive ICU patients and 145 healthy controls. Exhaled VOC profiles were collected via gas chromatography-ion mobility spectrometry (GC-IMS). Differential VOC analysis was performed, and an orthogonal partial least squares-discriminant analysis (OPLS-DA) model was constructed to evaluate diagnostic accuracy, stratification capacity, and prognostic performance.
Results
A total of 862 exhaled breath samples were analyzed, identifying 28 of 44 detected VOC signals. Eight of these compounds showed significant abundance alterations in ICU patients relative to controls. The OPLS-DA model achieved an area under the receiver operating characteristic curve (AUC) of 0.97 for distinguishing ICU patients from healthy individuals. VOC profiles also differed significantly between patients stratified by high vs. low APACHE Ⅱ scores. For predicting 30-day all-cause mortality, the model yielded a moderate AUC of 0.73, with 3 specific VOCs (2-Butanone-3-hydroxy-D, ethyl acetate-M, and dimethyl sulfide) showing significant abundance differences between non-survivors and patients with clinical improvement.
Conclusions
Exhaled VOC profiling effectively identifies critical illness and stratifies disease severity. While single-modal VOC profiling has limited prognostic performance due to the heterogeneity of critical illness outcomes, the distinct differential signatures identified here provide a solid foundation for future research.