Exhaled Volatile Organic Compounds as Non-Invasive Biomarkers for the Diagnosis, Phenotyping, and Treatment Response Prediction in Asthma

Technical Summary: Exhaled Volatile Organic Compounds as Non-Invasive Biomarkers in Asthma

Problem Statement
Current asthma management faces a critical bottleneck: existing clinical assessment tools (such as pulmonary function tests, FeNO, and blood eosinophil counts) are insufficient for comprehensive disease evaluation. These traditional markers often rely on patient cooperation, fail to capture early metabolic alterations preceding functional impairment, and predominantly reflect Type 2 inflammatory pathways. Consequently, they inadequately represent the full spectrum of asthma phenotypes (including non-Type 2 endotypes), struggle to distinguish disease control status accurately, and have limited capacity to predict individual treatment responses. There is an unmet clinical need for non-invasive, multidimensional biomarkers capable of addressing diagnosis, phenotyping, treatment selection, and longitudinal monitoring simultaneously.

Methodology
This study employed a single-center, prospective, observational cohort design conducted between June and December 2025 at the First Affiliated Hospital of Fourth Military Medical University.

• Cohort: The study recruited 147 asthma patients (APs) and 50 healthy controls (HCs). After quality control and exclusion criteria (e.g., COPD, incomplete sampling, acute exacerbations), the final analysis included 114 APs and 41 HCs.
• Stratification: APs were stratified by:
  • Control Status: Controlled (ACG, ACT ≥ 20) vs. Uncontrolled (AUCG, ACT < 20).
  • Phenotype: Type 2 (T2G) vs. Non-Type 2 (NT2G), defined by FeNO, eosinophils, and IgE levels.
  • Treatment: Inhaled Corticosteroid (ICS) combination vs. Biologic plus ICS (Bio-ICS).
  • Longitudinal: A subset of 11 patients provided paired pre- and post-treatment samples (3 months of therapy).
• Data Acquisition: Exhaled breath samples were collected in FEP bags and analyzed using Gas Chromatography-Mass Spectrometry (GC-MS) with Solid-Phase Microextraction (SPME).
• Data Processing & Analysis:
  • Preprocessing: Included baseline correction, peak alignment, noise reduction, and removal of ambient air interference. Peaks were matched against the NIST2020 library (quality > 80%).
  • Statistical Modeling: Univariate analysis (Mann-Whitney U test with FDR correction) and multivariate analysis (Orthogonal Partial Least Squares Discriminant Analysis - OPLS-DA) were used to identify significant VOCs.
  • Machine Learning: Random Forest algorithms were utilized for classification modeling, with performance evaluated via the Area Under the Receiver Operating Characteristic Curve (AUC).
  • Correlation: Pearson correlation analysis assessed relationships between VOCs and clinical parameters (lung function, FeNO, IgE, EOS).

Key Contributions

1. Holistic VOC Profiling: The study systematically evaluated exhaled VOCs across the entire clinical spectrum of asthma (diagnosis, control status, phenotyping, and treatment response) within a single cohort, addressing a gap in integrated literature.
2. Multimodal Integration: It investigated the synergistic effect of combining VOC profiles with traditional biomarkers (FeNO, IgE, EOS) to enhance diagnostic and phenotypic classification accuracy.
3. Phenotype Differentiation: The study provided specific VOC signatures distinguishing Type 2 from Non-Type 2 asthma, a distinction critical for targeted therapy but difficult to achieve with current standard markers alone.
4. Mechanistic Correlation: It established correlations between specific VOCs and clinical inflammatory markers/lung function, suggesting biological plausibility for these compounds as disease biomarkers.

Results

• Diagnosis (APs vs. HCs): Ten VOCs were identified as discriminatory. Eight were downregulated in APs (e.g., 4-Methyldecane, 2-Methylbenzaldehyde), and two were upregulated (1-Iodododecane, Nonadecane). The machine learning model achieved an AUC of 0.819 (95% CI: 0.751–0.887).
• Disease Control (ACG vs. AUCG): Nine VOCs differentiated controlled from uncontrolled asthma. The model yielded an AUC of 0.737 (95% CI: 0.641–0.833).
• Treatment Response: Four VOCs (including Dodecane) showed significant downregulation in patients responding to Bio-ICS therapy compared to ICS-C therapy. The predictive AUC was 0.673 (95% CI: 0.465–0.881), indicating modest discriminatory capacity.
• Phenotyping (T2G vs. NT2G): Four VOCs distinguished phenotypes (3 upregulated in NT2G, 1 in T2G). While VOCs alone yielded an AUC of 0.686, the combination of VOCs + FeNO + IgE + EOS significantly improved the AUC to 0.959 (95% CI: 0.892–1.000), compared to 0.909 for clinical markers alone.
• Correlations: Specific VOCs showed significant correlations with clinical parameters. For instance, Phenol correlated positively with IgE and EOS, while Decane and Nonadecane showed negative correlations with inflammatory markers. 2,4-Dimethylundecane correlated positively with EOS and FeNO.

Significance and Claims
The authors claim that this study establishes a foundation for using exhaled VOCs as non-invasive metabolic fingerprints for the multidimensional assessment of asthma. The primary significance lies in demonstrating that:

1. VOC profiles can effectively differentiate asthmatic patients from healthy controls and distinguish between controlled and uncontrolled disease states.
2. Distinct VOC variations exist between patients undergoing different therapeutic regimens (ICS vs. Biologics), suggesting potential for monitoring treatment efficacy.
3. Integrating VOCs with traditional biomarkers creates a robust framework for precision phenotyping, significantly outperforming traditional markers alone in distinguishing Type 2 from Non-Type 2 asthma.

The paper concludes that while limitations exist (notably sample size and the modest AUC for treatment response prediction), the findings advocate for breath analysis as a readily implementable, non-invasive technique to identify airway inflammatory phenotypes and assess treatment eligibility, potentially revolutionizing personalized asthma management through real-time metabolic monitoring.