Detection of lung cancer by automated sputum cytometry

Introduction: Biomarkers may prove to be valuable tools to manage those at risk of lung cancer. Sputum analysis using DNA cytometry has shown promise, but an automated, objective sputum analysis test has yet to be developed. This study evaluated the performance characteristics of the LungSign test for lung cancer and compared them to conventional cytology. Methods: A multicenter validation trial was conducted in which sputum specimens were prospectively collected from subjects suspected of having lung cancer during diagnostic workup. Specimens were placed on slides, DNA stained using Feulgen thionin, and analyzed using an automated cytometry-based scoring system. Smears were also prepared from the sputum specimens, stained by the Papanicolaou procedure, and analyzed using conventional cytology. LungSign scores and conventional cytology results were compared with the subject diagnoses. Results: A total of 1235 high-risk subjects were enrolled at nine clinical sites. Of 1123 subjects included for analysis, 370 were found to have lung cancer-a 33% prevalence. The a priori selected LungSign score threshold detected 40% of all lung cancers and 35% of stage I lung cancers with 91% specificity. Test performance was statistically equivalent across cancer stages, histologic types, and localizations for 330 analyzable lung cancer subjects. LungSign receiver operating characteristic area under the curve measure for the test was 0.692. Conventional cytology detected 16% of lung cancers with 99% specificity. Conclusions: DNA cytometry of sputum using the LungSign test detects stage I lung cancer and may provide a new tool to manage high-risk individuals.