, based on data from a prospective study of approximately 800 individuals.
Lung cancer remains a major cause of death in patients with chronic obstructive pulmonary disease (COPD), but many cases may go undetected in the early stage because of lack of screening and lack of validated predictive biomarkers, wrote Rianne de Vries, PhD, of the University of Amsterdam, and colleagues.
Accurate, noninvasive tests to screen patients with COPD for lung cancer are needed, and molecular profiling of exhaled breath using electronic nose (eNose) technology has shown potential as a method of early detection by identifying patterns of exhaled volatile organic compounds (VOCs), they said.
In a study published in the journal Chest, the researchers reviewed data from 682 adults with COPD and 211 with lung cancer who were enrolled in BreathCloud, a multicenter, observational study of healthy controls and individuals with suspected or confirmed diagnosis of asthma, COPD, or lung cancer.
Patients’ breath profiles were collected at study enrollment, between May 2017 and November 2018, using a metal oxide semiconductor eNose (SpiroNose).
Data from the eNose included the highest sensor peak normalized to the most stable sensor and the ratio between sensor peak and breath hold point. These variables were combined into four principal components (PCs) that captured 78.4% of variance in the dataset, and training and validation sets were constructed for all subjects. The researchers calculated a receiver operating characteristic (ROC) curve, including the area under the curve (AUC).
All patients were treated with standard clinical care and were monitored for development of clinically diagnosed lung cancer for 2 years, confirmed via CT imaging. The mean age of the patients was 64 years, and demographics at baseline were similar for patients with and without lung cancer.
After exclusion of 116 patients with both COPD and lung cancer, the analysis showed an accuracy of 90% and a ROC-AUC of 0.95.
Within 2 years of study enrollment, 37 patients with COPD (5.4%) developed lung cancer. In training sets and validation sets, the principal components one, two, and three were significantly different in patients with COPD who developed lung cancer and those who did not, (P = .002, P < .001, P < .001, respectively). The ROC-AUCs of the testing and validation sets were 0.89 and 0.86, respectively.
“Interestingly, the VOC pattern associated with early development of lung cancer in COPD did not match to the pattern related to lung cancer stages, as the former was mainly captured by PC2 and the latter by PC3,” the researchers wrote in their discussion. “This suggests that early identification of upcoming clinically manifest lung cancer in patients with COPD by eNose is not driven by VOCs that are predominantly associated with a particular stage of the disease,” they said.
The findings were limited by several factors including the lack of CT scanning at baseline because of the real-world design, so the presence of any baseline tumors was unknown, although none of the COPD patients showed symptoms indicative of lung cancer at baseline, the researchers noted.
However, the results suggest that eNose technology can identify lung cancer-specific VOC patterns early in cancer development in COPD patients, which provides a possible opportunity for early intervention, they concluded.
The study received no outside funding. De Vries disclosed personal fees and a substantial interest in the start-up company Breathomix.
A version of this article first appeared on Medscape.com.