Abstract: Objective To approach the sensitivity of sputum cytology and elucidate the influence of tumor location, histologic tumor type, tumor stage, specimens number and preparation method on the sensitivity. Methods From January 2005 to December 2005, there were 743 lung cancer patients investigated by sputum collection. Sputum was collected for three consecutive days, whenever possible. The first samples were processed by two different preparation methods (conventional preparation and ThinPrep). The second and the third samples were only processed by conventional preparation (CP). Results The overall sensitivity of the sputum was 34.7%. Sensitivity in central and peripheral carcinomas was 45.6% and 26.9%, respectively. Regarding tumor type, 46.3% squamous cell carcinoma, 31.1% small cell carcinoma and 27.0% adeno carcinomas could be diagnosed. Sputum sensitivity also showed relations to tumor stages. In stage Ⅰ~Ⅳ tumor the sensitivity was 21.6%, 31.9%, 42.6% and 41.3%, respectively. There were 560 patients whose first samples were processed both by ThinPrep (TP) and CP. The sensitivity was 30.4% and 19.3%, respectively. The sensitivity of CP was increased with the number of specimens. By examining three specimens, the sensitivity achieved 33.8%. Conclusion The sensitivity of sputum cytology is low, and it influenced by the histologic types, location or stage of the tumor. Using TP method and increasing specimen numbers can improve the sensitivity of sputum within certain extent.