We have shown that a colorimetric sensor array is capable of identifying the unique chemical signature of the breath of individuals with lung cancer with moderate accuracy.
The results were not affected by the subjects' demographics, smoking history or stage of cancer. The results were similar for a small group of individuals with indeterminate lung nodules.
The earliest study to analyse the pattern of VOCs in the exhaled breath as a diagnostic test was published in 1985. In this study, 12 individuals with advanced stage non‐small cell lung cancer and 17 healthy controls provided breath samples for analysis using GC–MS. Forty‐nine VOCs were found to have differences in peak occurrence or concentration. From these 49, 22 peaks were used to develop a discriminate function. Only 7 of the 22 were needed to discriminate fully between the two groups studied.8
Two recent studies using GC–MS to analyse breath constituents in lung cancer have been published. The first study included 67 individuals with lung cancer, 41 healthy volunteers, 15 with metastases to the lung and 91 without cancer (who were undergoing a diagnostic bronchoscopy). A model using nine compounds to discriminate lung cancer from healthy controls had a sensitivity of 85.1% and a specificity of 80.5%. When lung cancer samples were compared with those from patients with negative bronchoscopy results, the specificity was 37.4%. When the comparison was with those with metastases to the lung, the sensitivity was 66.7%.13
The second study compared concentrations of 13 VOCs from the breath of patients with early‐stage non‐small cell lung cancer (n
36) with asymptomatic smokers (n
35), control non‐smokers (n
50) and subjects with mild to moderate COPD (n
25). There was an overall accuracy of 82% with a sensitivity and specificity of 72.2% and 93.6%, respectively, for lung cancer.14
The major benefits of GC–MS analysis are its sensitivity and ability to identify the chemical differences that make the breath of patients with lung cancer unique. The downside is that the expense of the system and the expertise needed for interpretation render the technology difficult to use as a point‐of‐care test. Gaseous chemical sensing devices are easier to use and less expensive. The ability of two of these systems to detect lung cancer has been reported previously.
In the first report, the authors used a quartz microbalance sensor system. The study population included 35 individuals with lung cancer, nine with previously resected lung cancer and 18 healthy controls. A model developed from this system was reported to show 100% correct classification of lung cancer and 94% classification of controls (although the figure in the paper suggests a small amount of overlap between the groups). The model was not validated.18
The second system reported was a carbon polymer sensor system. This study was performed in two phases—a training phase and a validation phase. In the training phase, the study population included 14 patients with lung cancer, 27 controls with other lung conditions and 20 healthy controls. Analysis suggested that the sensor output from the patients with lung cancer was distinguishable from healthy controls, whereas the output from other disease groups was not. The validation group included 14 patients with lung cancer, 32 controls with other lung conditions and 30 healthy volunteers. The accuracy of the model for detecting lung cancer was 85% (sensitivity of 71.4% and specificity of 91.9%).19
This study differs from previous reports in the type of gaseous chemical‐sensing device used, the breath collection method, the type of analysis and the study population. Our results seem to be more modest than those reported in the study of the quartz microbalance system. However, that study only reported the accuracy of the model building phase and did not report data on validation of that model or include controls with disease. Our results were also lower than reported for the carbon polymer sensor system. This may represent a true difference in the ability of these sensor systems to detect the discriminatory patterns of VOCs, or it may represent differences in accuracy related to the breath collection methods (ie, real‐time sampling of tidal breathing in this study vs collection of a forced exhalation into a sampling bag in the prior report) or in the populations studied (ie, more patients with lung cancer in the model building phase and a greater portion of controls with disease in the model building and validation phases of this study). This report is the only one to test the model on small indeterminate lung nodules.
Gaseous chemical‐sensing devices have been criticised for their lack of ability to identify the specific chemical compounds in the breath, as well as their lack of sensitivity to detect all of the potentially important VOCs. However, the goal of these devices is not to identify the breath constituents but to detect patterns of VOCs that serve as bio‐signatures of lung cancer. The current proof of principle study supports the promise that these systems can be developed into a useful clinical test. Future identification of the specific chemical differences in the breath by other technologies could be applied to the refinement of current, or development of new, sensor systems. The most accurate sensor system published to date is the ability of dogs to distinguish the breath of patients with lung cancer from that of healthy controls. In the double‐blind phase of the study, the dogs had an accuracy of 99%.20
This highlights the fact that pattern recognition in the absence of specific identification has the potential to produce results accurate enough to be clinically useful.
In summary, a colorimetric sensor array can detect the unique pattern of VOCs in the breath of patients with lung cancer with moderate accuracy. Further work may clarify the nature of the distinct breath constituents. This would help to guide refinement of the sensor array and breath collection system to maximise the diagnostic accuracy of the test. Ultimately, this line of investigation could lead to an inexpensive, non‐invasive screening or diagnostic test for lung cancer.