A basic prediction model with clinical parameters to diagnose lung cancer

Introduction: Lung cancer remains the leading cause of cancer-related death worldwide. There is an increasing demand for innovative, noninvasive diagnostic tools to detect lung cancer early to improve survival and decrease invasive diagnostics. A prognostic model based on readily available clinical parameters can serve as a base to combine with non-invasive diagnostic tools to improve the diagnosis of lung cancer. Methods: In a multinational, multi-centre study with 302 subjects diagnosed with non-small cell lung cancer (NSCLC) and 429 healthy subjects clinical parameters and results from breath analysis with the Aeonose™ were collected. The clinical parameters were analysed in a multivariate logistic regression to develop a prediction model for NSCLC. The diagnostic accuracy of this model is presented as Area under the Curve (AUC). Results: Confirmed NSCLC patients (68.3 (8.5) years; 58.3% male) were compared with controls without NSCLC (63.1 (8.7) years; 52.2% male) yielding an AUC of 0.73 (95% CI 0.70–0.77). By choosing an appropriate threshold value in the ROC-diagram of the multivariate model, we observed a sensitivity of 92.3%, a specificity of 32.7%, and a positive and negative predictive value of 48.9% and 85.9% respectively Conclusion: A model based on easily available clinical parameters obtained in a large multinational multicentre study can be used to detect the presence or absence of lung cancer. This basic model can be extended with results of other diagnostic tools, such as liquid biopsies and exhaled-breath analyses to further improve the detection of lung cancer in the future.