Abstract
Objective: Prognosis of patients with lung cancer (LC) largely depends on early diagnosis. Previous studies reported successful detection of LC in exhaled breath samples using sniffer dogs. However, questions remain concerning the discrimination potential of these approaches regarding LC and chronic inflammatory lung conditions, such as COPD.
Methods: In a prospective study we tested exhaled breath samples of patients with LC (group A), patients with COPD (group B) and healthy individuals (group C). Four sniffer dogs were trained to identify LC in human exhaled breath samples with each breath sample used only once. For analysing the diagnostic accuracy of LC detection, each dog performed the following tests in a standardized fashion: I) group A (n=10) vs group C (n=40), II) group A (n=5) vs group B+C (n=20), III) group A (n=10) vs group B (n=40).
Results: Sensitivity and specificity were in test I) LC vs healthy individuals 50% and 88% resp., in test II) LC vs mixed collective of COPD patients and healthy individuals 95% and 95% resp., and in test III) LC vs COPD 80% and 95% resp. Overall test sensitivity was 90% (CI 0,78-0,97), specificity 72% (CI 0,51-0,88), PPV 86% (CI 0,74-0,94) and NPV 78% (CI 0,56-0,93), inter-rater variability k=0,436 given the experimental conditions. Tests were analyzed for confounders (study population, smoking habits, nutrition, medication).
Conclusions: Exhaled breath analysis is a promising approach towards future non-invasive LC screening methods. By using sniffer dogs as a “detection device”, our results set a benchmark for the identification of LC in exhaled breath samples and the discrimination of LC and COPD.
- © 2011 ERS