Detection of Drug-induced Interstitial Lung Disease Caused by Cancer Treatment Using Electronic Nose Exhaled Breath Analysis
Aim: To investigate whether eNose technology can differentiate between patients with cancer who have drug-induced interstitial lung disease, and those without drug-induced interstitial lung disease.
Take home message: ENose breath analysis shows promise as a non-invasive, rapid diagnostic tool for detecting drug-induced interstitial lung disease (DIILD) in cancer patients. Enose technology could be especially valuable for frequent monitoring of high-risk patients undergoing cancer treatment.
Introduction
The introduction of advanced cancer treatments, such as immune checkpoint inhibitors, tyrosine kinase inhibitors, and antibody-drug conjugates, has improved patient outcomes but has also increased the incidence of drug-induced interstitial lung disease (DIILD), a potentially life-threatening condition. Diagnosing DIILD remains challenging due to overlapping features with infections and malignancies, and current screening methods like CT scans involve radiation exposure and are often inconclusive. A non-invasive, rapid diagnostic tool is needed for early detection. This study investigates the use of eNose technology, which analyzes volatile organic compounds (VOCs) in exhaled breath, to differentiate patients with DIILD from those without.
Methods
This study included patients with confirmed cancer diagnosis and suspected DIILD between October 2021 and November 2023. A control group of similar cancer patients without DIILD was also recruited. Patients with pulmonary infections or recent alcohol intake were excluded. A multidisciplinary team of specialists determined the likelihood of DIILD for each patient. Breath analysis was conducted using the SpiroNose®, following a standardized maneuver. Data were processed and analyzed using sparse partial least squares discriminant analysis. The primary classification focused on distinguishing DIILD patients from controls, with additional sub-analyses to assess the influence of corticosteroid use.
Results
A total of 25 patients with suspected DIILD were included, with 20 control patients matched. After excluding low-quality data and cases with an alternative diagnosis, analysis showed that DIILD occurred at a median of 2.8 months after cancer treatment initiation. The eNose achieved an area under the curve (AUC) of 0.81, with a sensitivity of 0.75. In patients using corticosteroids, the AUC was 0.80, while in patients without corticosteroid use, the AUC improved to 0.87. These findings indicate a distinct breath profile for DIILD, independent of corticosteroid use.
Discussion
This proof-of-concept study is the first to use breath analysis for detecting DIILD, showing that eNose technology can effectively differentiate between patients with and without DIILD. Traditional biomarkers have not been validated for routine clinical use, and the complex nature of DIILD suggests that a single biomarker may not suffice. Instead, composite breath profiles analyzed by eNose provide a promising alternative. Despite the small sample size and patient heterogeneity, the eNose demonstrated high accuracy (AUC 0.81), suggesting robust performance even with varying cancer types and treatments. Future research should focus on validating eNose technology in larger, prospective studies and exploring its use for screening high-risk patients, potentially reducing the need for radiation-based diagnostics and enabling earlier treatment interventions.
