Despite the potential of exhaled-breath analysis of volatile organic compounds to diagnose lung cancer, clinical implementation has not been realized partly due to the lack of validation studies.

This study addressed two questions: 1) Can we simultaneously train and validate a prediction model to distinguish non-small cell lung cancer (NSCLC) patients from non-lung cancer subjects based on exhaled-breath patterns? 2) Does addition of clinical variables to exhaled-breath data improve the diagnosis of lung cancer?

In this multicentre study, subjects with NSCLC and control subjects, performed a measurement of 5 minutes of breathing in the Aeonose™. A training cohort was used for developing a prediction model based on breath data, whereas a blinded cohort was used for validation. Multivariable logistic regression analysis was performed including breath data and clinical variables, where the formula and cut-off value for the probability of lung cancer were applied on the validation data.

376 Subjects formed the training, and 199 subjects formed the validation set. The full training model, including clinical parameters and breath data showed, at a cut-off probability of 16% for lung cancer, a sensitivity of 95%, specificity of 51%, a negative predictive value (NPV) of 94% with an area under the receiver operating characteristic curve (AUC) of 0.87. Performance of the prediction model on the validation cohort showed corresponding results with a sensitivity of 95%, specificity of 49%, NPV of 94%, and an AUC of 0.86.

Combining exhaled-breath data and clinical variables in a multicentre, multi-device validation study can adequately distinguish lung cancer patients from subjects without lung cancer in a non-invasive manner. This study paves the way to implement exhaled-breath analysis in the daily practice of diagnosing lung cancer.

The Netherlands Trial Register, NL7025.