Introduction: Mechanical ventilation on the Intensive Care Unit (ICU) is a life-saving intervention but can be associated with the development of ventilator-associated pneumonia (VAP). Diagnosing VAP remains challenging and many patients receive antimicrobial treatment unnecessarily, contributing to the well-known global problem of antimicrobial resistance. Breath analysis offers a novel diagnostic perspective in critically ill mechanically ventilated patients. Two breath sampling methods have already been developed and shall be investigated in patients suspected of VAP on ICU. Thus proof will be sought that these volatile compounds can distinguish between the presence and absence of VAP.Methods: A ventilator circuit was constructed to model the patient setting on ICU. Two methods of breath sampling were investigated: 1) a semi-invasive method using a suction catheter inserted via the endotracheal tube; and 2) a non-invasive method that sampled in the external ventilator connection circuit. A quality control (QC) mix was inserted in the ventilator circuit. Samples were collected on steel adsorbent tubes, which were analysed using Thermal Desorption - Gas Chromatography - Mass Spectrometry (TD - GC - MS). On three ICUs in Greater Manchester a proof-of-concept study was set up to test both breath sampling techniques in critically ill patients suspected of VAP.Results: Both breath sampling methods resulted in good chromatograms reflecting signals deriving from the compounds of the QC mix. The compounds with relatively large molecular weight were not sampled as well using the non-invasive compared to the semi-invasive method. The semi-invasive method resulted in contaminating signals on the chromatograms. The clinical study suffers from a low recruitment rate, which appears to be due to a reduction in the prevalence of VAP on the participating ICUs. Discussion: A realistic model of human mechanical ventilation was established. A QC mix mimicking compounds present in exhaled breath was successfully inserted in the system and sampled using non-invasive and semi-invasive breath sampling methods. Standard curve linearity and detection limits of the system were determined. The clinical study is currently recruiting and data is being collected.