Comparison of Two Methods to Analyse Exhaled Breath for Accurate Asthma Detection in School‐Aged Children

Asthma is the most common chronic disease in childhood, yet its diagnosis remains challenging due to the nonspecific nature of asthma-like symptoms and the absence of a reliable stand-alone diagnostic test [1, 2]. A promising alternative involves analysing the chemical composition of exhaled breath using mass spectrometry (MS) [3, 4]. This study evaluated the diagnostic accuracy of exhaled breath analysis using Selected Ion Flow Tube Mass Spectrometry (SIFT-MS) and Gas Chromatography Time-of-Flight Mass Spectrometry (GC-TOF-MS) in school-aged children. A cross-sectional proof-of-concept study design was conducted. Asthma was diagnosed by a paediatric pulmonologist based on the ERS definition of asthma (presence of asthma symptoms combined with a reversible airway obstruction and/or bronchial hyperresponsiveness) [2]. The study was approved by the local ethics committee (2018-0545); written informed consent was obtained from parents. Breath samples were collected using standardised procedures. Each sampling bag was connected directly to the heated inlet of a VOICE200 Ultra instrument (Syft Technologies, Christchurch, New Zealand) for SIFT-MS analysis. The breath remaining in the sampling bag was transferred into a thermal desorption tube to store the volatile compounds (VOCs) until GC-TOF-MS analysis. Detailed information about the breath collection and analysis can be found in Supporting Information A. To differentiate between healthy and asthmatic individuals, data from SIFT-MS and GC-TOF-MS were analysed separately using a 5-step computational procedure to build and test the two classification models. The datasets were randomly divided into training and internal test sets. Details are described in Supporting Information B. Complete data were available for 119 participants, of which 51 participants were diagnosed with asthma. As expected, children with asthma had a significantly higher percentage of inhalation allergies, eczema and first-degree relatives with asthma [2]. Half of the asthmatic patients were well controlled according to the Global Initiative for Asthma (GINA) score, with almost 90% using moderate doses of daily inhalation corticosteroids. Still, 40% did experience an exacerbation in the 3 months preceding the measurement. Details are described in Supporting Information C. After pre-processing, the final dataset consisted of 377 and 302 variables for SIFT-MS and GC-TOF-MS, respectively. The evaluation of the discriminative power of the different variables measured by both techniques was performed by building two distinct classification models. This resulted in two groups of 51 (SIFT-MS) and 18 (GC-TOF-MS) most discriminatory variables. The results of the classification models (see Figure 1) showed that both SIFT-MS and GC-TOF-MS yielded significant discriminatory power, with SIFT-MS offering the advantage of real-time analysis and high user-friendliness towards non-expert staff. The diagnostic accuracy of SIFT-MS was decent (sensitivity and specificity of both 70%, and an area under the receiver operating characteristic curve (AUROC) of 0.81 (CI 0.72–0.88)) and comparable to the diagnostic ability of GC-TOF-MS (sensitivity and specificity of 78% and 72% respectively and an AUROC of 0.78 (CI 0.70–0.86)). The combination of both methods provided the highest diagnostic accuracy, i.e., sensitivity of 84% and specificity of 87% with an AUROC of 0.97 (CI 0.93–0.99). The evaluation of any effect of possible confounders, such as the usage of inhaled corticosteroids or level of asthma control, was performed using PCA models without observing clear grouping. However, given that nearly 90% of asthma patients used inhaled corticosteroids, drug metabolites may have contributed to some discriminative features. In conclusion, our study demonstrates the potential of breath analysis by SIFT-MS as a valuable user-friendly tool for asthma diagnosis. The combination of both techniques offers an improved diagnostic methodology for asthma. Future research should address the further external validation of the diagnostic models proposed in this work on larger cohorts of patients, in an attempt to move closer to the integration of breath analysis in clinical practice. We gratefully acknowledge the invaluable support of our research nurse, Karen Groot, the financial assistance from Health~Holland (LSHM17071), and we extend our heartfelt thanks to the children who participated in this study. The authors declare no conflicts of interest. The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions. Data S1: all70084-sup-0001-Supinfo1.docx. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.