ORLANDO A 2-minute breath test to detect volatile
organic compounds (VOCs) successfully predicted lung cancer in patients with
stage I-IV disease, according to a study reported by Michael Phillips, MD, at
the American Society of Clinical Oncology 41st Annual Meeting (abstract 9510).
"The breath test may potentially become a useful adjunct to lung cancer
detection," said Dr. Phillips, clinical professor of medicine, New York Medical
Prior studies using gas chromatography and mass spectroscopy
have identified VOCs associated with lung cancer. Dr. Phillips had previously
identified oxidative stress products (alkanes and methylated alkanes) as being
sensitive and specific biomarkers for lung cancer.
Lung Cancer Development
Lung cancer development is thought to require a high-risk
genotype and a high-risk exposure to a toxin, Dr. Phillips said. Combined, they
induce enzyme production that results in a high-risk phenotype. These induced
enzymes metabolize the toxins into carcinogens, resulting in lung cancer.
Normal oxidative stress generates VOCs, which are excreted in the breath.
"Our hypothesis [see Figure] is that these compounds undergo
rapid accelerated clearance by the induced enzymes, thereby resulting in
detectable changes in the breath VOCs," he said.
Dr. Phillips and his colleagues sought to determine whether
the breath VOCs could be clinically useful biomarkers for lung cancer. They
enrolled 213 smokers older than age 60 in the control group and 238 patients
with suspected primary lung cancer, untreated and in their first episode with
The control group received a spiral CT scan of the chest and
a 2-minute breath test. The suspected lung cancer patients underwent biopsy and
the same breath test. Breath samples were analyzed with gas chromatography and
mass spectroscopy. Patients with no evidence of disease after CT scan or biopsy
(the controls) and patients with documented primary lung cancer underwent a 2
to 1 split into a training set (to construct a predictive model based on the
breath VOCs) and a prediction set (to test the model).
The team ended up with 212 controls and 195 patients with
primary lung cancer, primarily non-small-cell lung cancer. Patient
characteristics (sex, age, smoking history) were similar in the cancer group
and the controls.
Dr. Phillips used "fuzzy logic" for a multivariate analysis
of the data to identify biomarkers of lung cancer. "Despite its name, fuzzy
logic is a rigorous and precise technique," he said.
In the training set, fuzzy functions generated typicality
matrixes for the controls and the lung cancer patients. Those data were placed
in the prediction set along with new breath samples. Using the typicality
matrixes, the investigators determined two values, one for lung cancer and one
for no disease.
The fuzzy logic identified 28 VOC markers of lung cancer.
Ten were alkanes and methylalkanes (biomarkers of oxidative stress); 20 were
alkane derivatives (ie, downstream metabolites of these products of oxidative
stress). Seventeen of the compounds showed reduced abundance in lung cancer
patients. The researchers entered the breath VOCs into the predictive model.
"Breath VOCs predicted lung cancer in the prediction set
with a sensitivity of 90.6%, a specificity of 82.7%, and an area under the
curve (AUC) of 0.91," Dr. Phillips said. The team also split the datasets using
a leave-one-out method. This method produced similar results, with sensitivity
of 81.5%, specificity of 87.3%, and identical AUC (0.91). The predictive curves
were comparable for all stages of cancer, Dr. Phillips said, "although we did
see the best predictive value in early-stage cancer." The predictions were not
affected by tobacco smoking.
Dr. Phillips noted that the breath test was also positive
for lung cancer in other groups, including those with cancer metastatic to the
lung, postoperative lung cancer, mesothelioma, and suspicious imaging with a
negative biopsy. He explained that excision of the lung cancer does not affect
extrapulmonary sites, which are the sources of most of the oxidative stress in
Dr. Phillips calculated that if the breath test were
administered to 100,000 smokers older than age 60, assuming a 2% prevalence of
lung cancer, about 14% of the tests would be positive and 86% negative. The
positive predictive value would be 11.6% and the negative predictive value
99.6%. The high negative predictive value is important, he noted, because the
test identifies the majority of patient who do not have lung cancer.
"This means that if a breath test is positive, it would make
sense to do chest imaging, because the positive predictive value will greatly
improve and thereby improve the detection of lung cancer," Dr. Phillips said.
"However, if the breath test is negative, it probably would not make sense to
do imaging, because the negative predictive value is unlikely to improve much
more than 99.6%." Thus, he concluded, use of the breath test could lower costs
and decrease patients’ exposure to radiation by reducing the need for CT
screening in smokers.