Comparison of breath VOCs collected from controls, healthy smokers, smokers with COPD, and ex-smokers with COPD at two sites

Abstract

In this ongoing two-center collaboration study (Marburg, Hannover, German Center for Lung Research) we aim to include sufficiently high numbers of patients and controls (total 240, >70% completed) to identify specific COPD related VOC biomarker signals that are independent of the influence of study site.

Tenax adsorption tubes were conditioned in Hannover and shipped to Marburg. Breath collection was performed using identical sampling devices (inhalation via A2 filter and exhalation (flow control) into a stainless steel reservoir), but different mouth pieces. 3L of breath were continuously drawn onto 2 adsorption tubes. Currently data for 53 commonly observed substances is available. Of these 13 VOCs were excluded due to significant differences between study sites.

Including data of 72 samples from healthy non-smokers and COPD ex-smokers from both sites the stepwise inclusion of variables resulted in a model with 4 VOCs that discriminated 83.0% of subjects correctly. In the cross-validation process 77.8% of subjects were correctly identified. Including the 41 healthy smokers and smokers with COPD the model correctly identified only 68.3% of the subjects and the cross-validation value was 61.0%.

Excluding site specific VOCs we were able to identify a VOC pattern of 4 substances that appear to be related to COPD. The model to discriminate COPD patients from controls showed a better performance in the non-smoking groups. Despite lower subject numbers in the actively smoking groups, this result emphasizes that it is important to carefully control for active smoking when searching for COPD related VOC patterns.