Abstract :
[en] Exhaled breath contains hundreds of volatile organic compounds (VOCs) that reflect ongoing metabolic and inflammatory processes, making breath analysis an attractive non-invasive approach in respiratory medicine. Despite three decades of research, no VOC-based test has entered routine clinical practice. This thesis addresses that gap by treating analytical performance and clinical feasibility as co-equal requirements rather than sequential concerns, following the translational pathway a breath biomarker must travel: how the sample is collected, what it adds to clinical assessment, whether it can inform a treatment decision before it is taken, and how it could eventually be measured in a clinical laboratory. All work was built upon an established thermal desorption comprehensive two-dimensional gas chromatography mass spectrometry (TD-GC×GC-MS) platform.
The three offline sampling devices in widest use were first compared within the peppermint benchmarking protocol. No device proved universally optimal. Tedlar® bags showed the lowest overall variability (mean pooled RSD 54%) but sensitivity to exogenous contamination; ReCIVA® offered superior selectivity and reduced background at the cost of detecting fewer compounds, greater complexity and higher cost; BioVOC-2® provided operational simplicity but limited sensitivity. Focus groups with clinicians, researchers and participants established practical requirements for clinical deployment, yielding use-case-dependent guidance and extending the peppermint reference database.
VOC profiling was then evaluated as a supplementary clinical parameter in two contrasting settings. In 79 patients with uncontrolled moderate-to-severe asthma randomised to a 12-week submaximal aerobic training programme, clear improvements in symptom control, quality of life and exercise tolerance occurred without detectable modulation of airway or systemic inflammation, including VOCs. In a bicentric cohort of 42 systemic sclerosis patients, a nine-compound signature discriminated those with interstitial lung disease with an AUC of 0.82, comparable to DLco (0.83).
An eosinophil-derived 17-VOC signature measured immediately before the first anti-IL-5/Rα injection predicted clinical response with AUROCs of 0.95 and 0.98 in two independent cohorts analysed through entirely separate workflows, outperforming FeNO and blood eosinophil count. Comparable accuracy was retained by a five-compound subset of reactive aldehyde species.
Finally, SICRIT coupled to high-resolution mass spectrometry was evaluated for the direct measurement of 16 clinically relevant aldehydes. Linearity was good for 13 compounds (R² ≥ 0.99), with limits of detection between 1.33 and 25.13 ppbv and low intra-day variability. Protonated species emerged as the preferred monitoring targets, ionisation competition from a Tedlar® contaminant was identified and mitigated, and cyclic ion mobility permitted tentative isomer separation at a prohibitive sensitivity cost.
Taken together, this work shows that a single breath sample and a single workflow can address clinically diverse questions that conventional parameters answer only partially, and that the obstacles to implementation are identifiable and addressable. It does not deliver a ready-to-use test, but contributes comparative data for standardisation, a refined molecular target, and an analytical route toward measurement compatible with clinical timescales.
