Harnessing Exhaled Breath for Lung Cancer Early Detection-Results From the ExPeL Study

Area of Science:

• Pulmonology
• Oncology
• Biochemistry

Background:

• Early lung cancer detection is crucial for improving patient outcomes and survival rates.
• Current diagnostic methods can be invasive or lack sufficient sensitivity for early-stage disease.
• There is a need for scalable, non-invasive tools to aid in early lung cancer detection and optimize referral pathways.

Purpose of the Study:

• To evaluate the Inflammacheck point-of-care device for non-invasive lung cancer detection using exhaled breath condensate (EBC).
• To assess the utility of EBC hydrogen peroxide (H₂O₂) and physiological parameters combined with machine learning in a real-world screening population.
• To compare machine learning model performance with untargeted LC-MS metabolomics for cancer discrimination.

Main Methods:

• The Exhaled Hydrogen Peroxide for Early Lung Cancer Detection (ExPeL) study enrolled participants from the UK Targeted Lung Health Check (TLHC) programme.
• Exhaled breath condensate (EBC) was collected using Inflammacheck, measuring H₂O₂, CO₂, humidity, temperature, and flow rate.
• Multivariate analyses (PCA, LDA, Mahalanobis) and supervised machine learning models (stacked and voting ensembles) were employed, alongside LC-MS metabolomics.

Main Results:

• The voting ensemble model achieved 85.7% accuracy, 80% sensitivity, 100% specificity, and a 0.90 ROC-AUC, with no false positives.
• Multivariate analysis demonstrated clear separation between lung cancer patients and controls.
• Untargeted metabolomics identified key metabolites with an AUC of 0.969 for cancer discrimination.

Conclusions:

• Inflammacheck provides an effective, rapid, and non-invasive breath test for distinguishing early-stage lung cancer.
• The findings are highly relevant for primary care and screening triage, addressing challenges of non-specific symptoms and low prevalence.
• This technology offers a promising tool to enhance early lung cancer detection strategies.