Background Healthcare associated infections, including ventilator associated pneumonia, are difficult to diagnose and treat, and are associated with significant morbidity, mortality and cost. We aimed to demonstrate proof of concept that breath volatile profiles were associated with the presence of clinically relevant pathogens in the lower respiratory tract.

Methods Patients with sterile brain injury requiring intubation and ventilation on the intensive care unit were eligible for inclusion. Serial clinical and breath data were obtained three times a week, from admission up to a maximum of 10 days. Bronchial lavage for semiquantitative culture was collected immediately prior to breath sampling. Breath samples were collected in triplicate for off-line analysis by thermal-desorption/gas chromatography/time-of-flight mass spectrometry. Breath data were recorded as retention time/mass ion pairs, and analysed (pathogen present vs absent) by ANOVA-mean centre principal component analysis.

Results Samples were collected from 46 patients (mean (SD) age 49 (19) years; 27 male). The dominant factors affecting breath sample analysis were the individual breath profile and duration of intubation. When these were taken into account, clear separation was seen between breath profiles at each time point by the presence/absence of pathogens. Loadings plots identified consistent metabolite peaks contributing to this separation at each time point.

Conclusions Breath volatile analysis is able to classify breath profiles of patients with and without significant pathogen load in the lower respiratory tract. If validated in independent cohorts, these findings could lead to development of rapid non-invasive point-of-care surveillance systems and diagnostics for lower respiratory tract infection in the intensive care unit.