Automated detection of genetic abnormalities combined with cytology in sputum is a sensitive predictor of lung cancer.

Detection of lung cancer by sputum cytology has low sensitivity but is noninvasive and, if improved, could be a powerful tool for early lung cancer detection. To evaluate whether the accuracy of diagnosing lung cancer by evaluating sputa for cytologic atypia and genetic abnormalities is greater than that of conventional cytology alone, automated scoring of genetic abnormalities for 3p22.1 and 10q22.3 (SP-A) by fluorescence in situ hybridization (FISH) and conventional cytology was done on sputa from 35 subjects with lung cancer, 25 high-risk smokers, and 6 healthy control subjects. Multivariate analysis was performed to select variables that most accurately predicted lung cancer. A model of probability for the presence of lung cancer was derived for each subject. Cells exfoliated from patients with lung cancer contained genetic aberrations and cytologic atypias at significantly higher levels than in those from control subjects. When combined with cytologic atypia, a model of risk for lung cancer was derived that had 74% sensitivity and 82% specificity to predict the presence of lung cancer, whereas conventional cytology achieved only 37% sensitivity and 87% specificity. For diagnosing lung cancer in sputum, a combination of molecular and cytologic variables was superior to using conventional cytology alone.