Breath test shows promise in detecting lung cancer
Click Here to Manage Email Alerts
An exhaled breath test that uses high-pressure photon ionization time-of-flight mass spectrometry accurately identified patients with lung cancer, according to recent data in JAMA Network Open.
“Exhaled breath holds promising clinical application in lung cancer screening,” Shushi Meng, MD, of the department of thoracic surgery at the Peking University People’s Hospital in Beijing, and colleagues wrote.
Previous data have shown that low-dose computed tomography (LDCT) screening in high-risk populations can reduce mortality from lung cancer by 20%, the researchers reported. However, they noted there are several disadvantages to LDCT, including radiation exposure, high costs and a high false-positive rate. Exhaled breath tests, the researchers added, could potentially improve lung cancer screening.
“High-pressure photon ionization time-of-flight mass spectrometry (HPPI-TOFMS) is a promising tool for breath testing, because it is highly sensitive, does not require pretreatment of exhaled breath, and holds great tolerance for humidity,” Meng and colleagues wrote.
The researchers conducted a case-control diagnostic study to investigate whether a breath test that combines HPPI-TOFMS with a support vector machine (SVM) algorithm could distinguish 139 patients with lung cancer from 289 healthy controls. Most patients with lung cancer (n = 126) had early-stage disease.
Meng and colleagues collected exhaled breath samples from March 1, 2019, to Sept. 1, 2019. The breath collection process lasted about 1 minute for each participant, and no adverse events were reported, the researchers wrote. They randomly assigned samples from 381 participants to a discovery data set, which was further broken down into a training set of 286 samples — obtained from 90 patients with lung cancer and 196 healthy controls — and a test set of 95 samples — obtained from 30 patients with lung cancer and 65 healthy controls.
The SVM algorithm was used to establish the lung cancer detection model based on data from the training set. The test set was used to evaluate the accuracy of the detection model.
After 500 iterations of fourfold cross validation, the researchers determined that the detection model performed with 92.97% sensitivity, 96.68% specificity and 95.51% accuracy in the test set.
Forty-seven samples were then assigned to a blinded validation data set. In this data set, the model performed with 100% sensitivity, 92.86% specificity and 95.74% accuracy, according to the researchers. The area under the receiver-operating characteristic curve was 0.9586, and the model yielded a positive predictive value of 90.48% and a negative predictive value of 100%.
Meng and colleagues highlighted several limitations of the study, including the small size of the validation data set. They said additional studies are needed to verify the findings.
“In this study, we demonstrated that the HPPI-TOFMS breath test is feasible in clinical practice,” they wrote. “This diagnostic study’s results suggest encouraging findings that breath testing may be a reliable approach to lung cancer detection and HPPI-TOFMS may provide fast and precise detection of exhaled breath.”
Perspective
Back to Top
In this study, Meng and colleagues attempt to use a newer mass spectrometry technology combined with machine learning to assess ability to distinguish lung cancer from healthy controls in a screening population.
This study reports higher sensitivity, specificity and accuracy compared to prior studies on volatile organic compounds (VOCs) in its ability to detect cancer from noncancer patients.
Studies on VOCs have always been limited by lack of standardization, inability to control patient-environmental factors for appropriate breath collection and their reproducibility. This study suggests a newer technique that perhaps will help overcome some of those limitations. It adds to the growing literature supporting further research on VOCs for early detection and may suggest that machine learning algorithms possibly improve our assessment models. However, it stays limited by small numbers and no external validation.
Collapse
Read more about
Click Here to Manage Email Alerts