Randomized population trials and screening for lung cancer

Abstract

BACKGROUND At the current time, there is nearly universal agreement that screening for the early detection of lung carcinoma is not justified. This is based on the fact that, to the author's knowledge, no randomized population trial (RPT) to date has demonstrated a significant reduction in lung carcinoma mortality as a result of any screening intervention. METHODS To date, four RPTs, which have included a total of 37,724 male cigarette smokers, have been conducted. Studies at Memorial Sloan-Kettering Cancer Center and Johns Hopkins demonstrate no incremental effect by the addition of sputum cytology to annual chest X-ray (CXR) screening alone. However, CXR screening in all participants likely was responsible for stage and long term survival rates that were two- to three-fold higher than predicted based on contemporary statistics. Studies at the Mayo Clinic and in Czechoslovakia demonstrated significantly superior stage distribution and survival, but slightly inferior mortality in experimental populations that underwent periodic CXR screening. Such contradictory findings were made possible by a higher cumulative incidence rate of lung carcinoma in experimental populations. The constellation of improved survival, higher rate of incidence, and similar mortality led to the hypothesis that CXR screening leads to the overdiagnosis of lung carcinoma. RESULTS Abundant evidence based on epidemiologic, pathologic, and clinical considerations conclusively demonstrate that CXR screening does not lead to the significant overdiagnosis of lung carcinoma. Moreover, overdiagnosis is the only way to reconcile the results of existing RPTs with the conclusion that CXR screening is ineffective. The alternative conclusion is that significant stage, resectability, and long term survival advantages reflected the ability of CXR screening to improve cure rates. Population heterogeneity accounts for the failure of mortality to reflect screening efficacy accurately in these trials. There is direct evidence that population heterogeneity was responsible for the trend toward increased lung carcinoma mortality in the Czech study. Moreover, review of all RPTs demonstrates a consistent pattern in which population heterogeneity confounds the ability of mortality to represent an accurate measure of screening effectiveness. CONCLUSIONS Systematic analysis of RPTs supports two major conclusions: 1) an improvement in the cure rate rather than a reduction is cause specific mortality is the proper measure of screening effectiveness in the RPT setting and 2) CXR screening is associated with a two- to three-fold improvement in lung carcinoma cure rates. A paradigm shift is mandatory for the proper evaluation of conventional and newer screening modalities. Indeed, hundreds of thousands of lives would be saved annually on a global basis if CXR screening were offered to individuals at high risk for lung carcinoma. Cancer 2000;89:2399–421. © 2000 American Cancer Society.