Survival and mortality in a randomized study of lung cancer detection.

In a randomized prospective study of lung cancer detection in a high-risk population of over 6000 heavy smokers semiannual screening by X-ray and sputum cytology was compared to screening at a 3-year interval. The comparison of Kaplan-Meier estimates of survival curves done without and with correcting for lead-time bias disclosed a rather important impact of lead-time bias on survival comparisons. On the contrary, controlling for possible length bias had no obvious effect on the shape of survival curves. The evaluation of mortality from lung cancer, being used as a basic criterion, indicated no traceable benefit from semiannual screening. The higher incidence of lung cancer in the frequently screened group was paralleled by a higher mortality. It is concluded that currently available screening techniques will not solve the problem of lung cancer mortality in smokers. The results underline the importance of primary prevention for lung cancer.