Background There is evidence that Pseudomonas aeruginosa (Pa) produces volatile organic compounds (VOCs) such as hydrogen cyanide (HCN) and 2-aminoacetophenone (2-AA). VOCs in exhaled breath are therefore proposed as potential biomarkers of infection. We hypothesised that selective ion-flow mass spectrometry (SIFT-MS) breath analysis might allow discrimination of CF patients with (CF + Pa) and without Pa (CF-Pa).

Methods 79 adults (31 CF + Pa, 22 CF-Pa and 26 healthy controls) provided starved, single tidal exhalation breath samples into Nalophan^TM bags. Quantification of 15 VOCs was performed within two hours on SIFT-MS. All results are presented as (median parts-per-billion by volume [IQR]).

Results 2-AA was significantly higher in CF + Pa than CF-Pa (5.0 [3.4 7.1] vs. 1.3 [0.0 3.2], p <0.01). However, there was significant overlap and median co-efficient of variation was 35.41%; clinical utility is therefore questionable.

Dimethyl disulphide was also significantly higher in CF + Pa (95.2 [41.3 211.2 vs. 35.5 [22.1 79.8], p < 0.01). When combined with 2-AA, area under ROC curve was 0.867.

Counter to our sputum results, there was no difference in HCN between CF + Pa and CF-Pa (8.1 [5.0 11.9] vs. 6.9 [4.4 11.0], n/s) or between all CF patients and healthy controls (7.8 [4.9 11.5] vs. 7.0 [4.6 11.5], n/s).

Our early in vitro data showed decreased butanol above Pa cultures, suggesting consumption. This was replicated in breath with lower levels in CF + Pa vs. CF-Pa (37.4 [24.3 87.6] vs. 91.7 [46.9 143.7], p < 0.05).

Of VOCs likely to be of host origin, isoprene was increased in CF vs. controls (108.0 [83.4 195.5] vs. 69.6 [46.9 89], p < 0.01) with no difference between CF + Pa vs. CF-Pa. Acetone was reduced in CF (269.9 [161.9 356.4] vs. 324.9 [236.7 598.9], p < 0.01).

Conclusions 2-AA is a potential biomarker of Pa infection but clinical applicability is uncertain. Dimethyl disulphide and butanol also show promise. Mouth-exhaled HCN assessed by SIFT-MS does not appear to fulfil its promise as a Pa biomarker. Other VOCs assessed were either similar between Pa groups or different between healthy controls and CF, but unable to differentiate between Pa status. This study provides proof-of-concept for the development of a non-invasive tool with which to screen for lower airway bacterial infection in CF though a clinically applicable test remains some way off.