Metabolic alterations in pneumonia assessed by breath analysis: an exploratory paired longitudinal study comparing acute infection and post-recovery states

BACKGROUND: The diagnosis and monitoring of pneumonia remain a challenge due to its diverse etiology, overlapping clinical symptoms with other respiratory infections, and limitations in current diagnostic technologies for timely and accurate detection. We evaluated whether real-time secondary electrospray ionization-high‑resolution mass spectrometry (SESI‑HRMS) breathomics can capture pneumonia‑associated breath-profile changes in a paired longitudinal cohort. Methods: In this single-center longitudinal study at University Hospital Zurich, Switzerland, adults with clinically confirmed pneumonia (n = 47, median age: 63) were enrolled between 2020 and 2023 sampled during acute infection and again after recovery (median interval 85 days). After pre-processing, 3,635 mass‑spectral features were retained across ion modes, and 1,498 could be linked to at least one Human Metabolome Database (HMDB) entry by accurate‑mass matching (±1 ppm). Breath profiles were assessed using a subject‑grouped, nested cross‑validated partial least squares discriminate analysis (PLS‑DA) and paired differential analysis. Results: PLS-DA showed framework moderate exploratory internal discrimination between pneumonia and post‑pneumonia measurements with moderate performance (median balanced accuracy 0.72; median AUROC 0.80; Q² 0.24; two latent variables), with a permutation p-value of 0.112. Paired differential analysis identified 104 recovery‑associated features (58 higher and 46 lower during pneumonia) meeting |mean log2 fold change| ≥ 1 and FDR q ≤ 0.05, with putative accurate-mass matches to 239 HMDB accessions. Conclusion: Real‑time breathomics captures reproducible changes in exhaled breath mass-spectral profiles from pneumonia to convalescence and reveals breath-profile heterogeneity not fully reflected by standard clinical indices. These results should be interpreted as exploratory and hypothesis-generating, motivating multicenter validation with tandem mass spectrometry (MS/MS)‑supported compound identification, clinically relevant control groups, and longitudinal outcome linkage.