Background. Many volatile organic compounds (VOCs), principally alkanes and
benzene derivatives, have been identified in breath from patients with lun
g cancer. We investigated whether a combination of VOCs could identify such
patients.
Methods. We collected breath samples from 108 patients with an abnormal che
st radiograph who were scheduled for bronchoscopy. The samples were collect
ed with a portable apparatus, then essayed by gas chromatography and mass s
pectroscopy. The alveolar gradient of each breath VOC, the difference betwe
en the amount in breath and in air, was calculated. Forward stepwise discri
minant analysis was used to identify VOCs that discriminated between patien
ts with and without lung cancer.
Findings. Lung cancer was confirmed histologically in 60 patients. A combin
ation of 22 breath VOCs, predominantly alkanes, alkane derivatives, and ben
zene derivatives, discriminated between patients with and without lung canc
er, regardless of stage (all p < 0.0003). For stage 1 lung cancer, the 22 V
OCs had 100% sensitivity and 81.3% specificity. Cross-validation of the com
bination correctly predicted the diagnosis in 71.7% patients with lung canc
er and 66.7% of those without lung cancer.
Interpretation. In patients with an abnormal chest radiograph, a combinatio
n of 22 VOCs in breath samples distinguished between patients with and with
out lung cancer. Prospective studies are needed to confirm the usefulness o
f breath VOCS for detecting lung cancer in the general population.