Lung cancer is the leading cause of cancer-related mortality globally. Prior to the introduction of national lung cancer screening programs, nearly 70% of patients were diagnosed with advanced stage lung cancer, with an overall 5-year survival of 15% 1. Thus, early detection remains the most powerful strategy for improving lung cancer outcomes 1, 2. While low-dose computed tomography (LDCT) through screening programs has transformed early diagnosis in high-risk populations, challenges persist especially with overdiagnosis rates up to 19.7% and false-positive rates up to 96.4% 1, 2. Other barriers include infrastructure limitations, concerns surrounding radiation exposure, and limited patient uptake in real-world settings 3. Consequently, there has been increasing interest in the development and validation of cost-effective, accurate, and minimally-invasive approaches to risk-stratify LDCT findings 4. These include biomarker-based strategies on so-called liquid biopsies such as exhaled breath and blood 5. In particular, breath analysis represents a promising approach. Exhaled breath consists of non-volatile and volatile organic compounds (VOCs) 6, 7. Breath analysis utilising machine-based learning models has been explored to study the early detection of various malignancies such as breast, papillary thyroid carcinoma, lung, ovarian, colorectal, oesophageal, gastric, and hepatocellular carcinoma 6, 7. VOCs in malignancy can be classified into five groups comprising of aldehydes, ketones, alcohols, hydrocarbon, and aromatic compounds 6, 7. These compounds are detected in higher levels in malignancies and are a result of oxidative stress, inflammation, lipid peroxidation and raised glycolysis related to carcinogenesis 6, 7. In a recent study of 5047 patients who underwent screening for breast cancer, of whom 465 (9.21%) were newly diagnosed with breast cancer, a VOC-based diagnostic model (BreathBC) achieved an area under the curve (AUC) of 0.87 in external validation cohorts and further improved performance (AUC 0.94) on an expanded model, surpassing conventional breast imaging modalities such as mammography and ultrasound 8. Similarly, breathomics for colorectal cancer achieved an AUC of 0.89 which was superior to serum biomarkers 9. Whilst Gas Chromatography–Mass Spectrometry (GC–MS) is the common technique utilised in breathomics, other techniques utilized to assess VOCs include the application of an electric nose (eNose), selected ion flow tube mass spectrometry (SIFT-MS), high-pressure photon ionization-time-of-flight mass spectrometry (HPPI-TOFMS), and low-pressure photoionization mass spectrometry (LPP-MS) (Table 1) 6, 7, 10-12. With the advent of precision oncological management, recent studies have compared both VOC and blood-related genomic changes including EGFR, PIK3CA, ERBB2, BRAF and KRAS mutations through PCR and compared them with next-generation sequencing (NGS) tissue samples 11. However, challenges persist where studies have been constrained by relatively small sample sizes, lack of validation, and inconsistencies demonstrated by VOC biomarkers 10. It is important to recognise that currently VOC biomarkers lack disease specificity and may reflect broader biological processes rather than lung cancer alone 6, 10, 11. Furthermore, several factors may influence breath analysis including age, sex, hypoxia, gut microbiome, air pollution, smoking, food consumption, and medications 6. PTR-TOF-MS is emerging as a promising breathomics technique due to its high sensitivity and rapid response time, making it ideal for real-time online monitoring of trace and rapidly changing VOCs in human exhaled breath 11. In a recent publication in Respirology, Huang and colleagues present the largest validated cross-sectional breathomics study to date using proton transfer reaction–time-of-flight mass spectrometry (PTR-TOF-MS) combined with a machine learning approach to analyse exhaled breath samples for early lung cancer detection to distinguish lung cancer from healthy controls 12. This study of 4515 patients demonstrated robust diagnostic performance with sensitivity 95%, specificity 98%, and accuracy 98% when differentiating lung cancer from healthy controls 12. In addition, the model demonstrated sensitivity and specificity of 97% and 98% respectively, as well as accuracy of 98% when discriminating early-stage (stage one) lung cancer from benign pulmonary nodules 12. This highlights the potential of breathomics as an adjunctive tool for refining the assessment of indeterminate pulmonary nodules identified on LDCT. In this era of lung cancer screening, these findings are tantalizing. However, challenges exist pertaining to clinical implementation. PTR-TOF-MS instruments are sophisticated and resource-intensive, raising questions about scalability and cost-effectiveness 11. While breathomics is unlikely to replace established imaging-based screening strategies, it may have a potential role in composite risk stratification approached with integration of radiomics and liquid biopsy-based biomarkers 13. In summary, exhaled breath analysis holds considerable promise as a novel non-invasive diagnostic approach for lung cancer. The large and methodologically rigorous study by Huang and colleagues represents an important step toward the development of a clinically applicable point-of-care breath test. Despite the molecular heterogeneity of lung cancer, LDCT remains the cornerstone of screening, with tissue diagnosis as the gold standard and blood-based biomarkers to guide management and monitor treatment response. Within this evolving landscape, breathomics may emerge as a valuable adjunct in the early detection and risk stratification of lung cancer. The authors have nothing to report. T.L. is an Editorial Board member of Respirology and a co-author of this article. She was excluded from all editorial decision-making related to the acceptance of this article for publication. C.L.L. has no conflicts of interest to declare.
Leong et al. (Tue,) studied this question.