The Analysis of Volatile Organic Compound Profiles in the Breath as a Biomarker of Lung Cancer

Abstract

Rationale: Cancer cells have distinct metabolic properties that could lead to an accurate volatile metabolic signature in the urine of patients with lung cancer. We aim to determine the accuracy of colorimetric sensor array derived volatile organic compound profiles of urine outgas for the rapid and non-invasive identification of patients with lung cancer. Methods: 100 uL of urine was added to blotting paper then sealed in a vesicle containing a colorimetric sensor array in the gas space above the paper. The urine was added either as neat urine or after exposure to one of five additives designed to diversify the volatile species profile in the urine enclosure headspace. The sensor was imaged at the beginning of testing and every 3 minutes for a total of 4 hours of exposure. Lung cancer subjects were biopsy confirmed and had not received treatment. Control subjects had indeterminate lung nodules, proven to be benign, or had risk factors for the development of lung cancer. Subjects were recruited from clinics at the Cleveland Clinic. Sensor response features were first reduced using the entire dataset (to approximately 70 features), and then selected from a training set (70% of the cohort, approximately 20 features). Random forest models were built on the training set using the selected features. The models were validated on a testing set (the remaining 30%). This process was repeated 100 times with average results described. Results: 57 subjects with lung cancer and 24 control subjects were included in the analysis. 33 of the lung cancers were adenocarcinoma and 19 were squamous cell carcinoma. The validated C-statistics for models assessing cancer vs. control, adenocarcinoma vs. control, and squamous cell carcinoma vs. control, using the sensor responses to the urine headspace gas were 0.91, 0.91, and 0.98 respectively. Conclusions: Colorimetric sensor array derived volatile organic compound profiles of urine headspace gas can accurately distinguish a signature of people with lung cancer from those of relevant controls.