© 2003 American Thoracic Society
Medicine on Lung Cancer ScreeningA Different ParadigmDepartment of Radiology Weill Medical College of Cornell University New York, New York In this issue of the Journal (pp. 1167–1173), McWilliams and coworkers (1) address the laudable goal of stratifying individuals by risk for lung cancer to optimize computer tomography (CT) screening. For lung cancer, such a test would need to be more sensitive and specific than CT, be easily and rapidly performed, be acceptable to potential screenees, and cost less than CT screening (typically about $300). McWilliams and coworkers report on their baseline screening results. They still need to demonstrate that sputum cytology will perform well for considerably smaller malignancies seen on annual repeat CT screening (2–3) because this determines the major value of screening. Baseline occurs only once for any screenee. Even if automated sputum cytometry were not useful on repeat screening, it may still be useful during baseline screening. Further analysis might even show it to be more useful when performed only in patients who have non-calcified nodules on CT. McWilliams and coworkers appear to have data to evaluate this option. McWilliams and coworkers enrolled 561 high-risk individuals and obtained a sputum sample that was analyzed by automated quantitative cytometry. In 544 patients, a baseline low-dose CT was done immediately after sputum collection. In another 17 patients, shown in retrospect to have normal sputum, the baseline low-dose CT was done some time later. The initial low-dose CTs on 322 screenees were done on a single-slice scanner with 7 mm collimation; the subsequent 229 scans were done later on a 4-slice scanner with 1.25 mm collimation. Details of the CT readings—by whom, how, blinded to other results, and so on—were not given nor were the location or size of the CT-diagnosed cancers. Validation of the automated readings, including comparison to the readings by the cytologists, was not provided in this paper. This information is critically important. McWilliams and coworkers found sputum atypia in 75% (423/561) of the individuals; a low-dose CT was positive in 46% (261/561). Henschke and coworkers (4) showed recently that the positive result at baseline using multi-slice, low-dose CT can be reduced to below 15% without any increase in false negatives. Further research on sputum cytometry may also achieve a reduction of its positive rate as well. Of the 561 individuals enrolled in the study, 378 also had autofluorescence bronchoscopy (LIFE). A total of 14 primary lung cancers were detected: 10 (71%) solely by CT and 4 (29%) solely by autofluorescence bronchoscopy (the latter 4 were squamous carcinomas, 3 in situ and 1 Stage IA). Of the 14 lung cancers, 13 (93%) had sputum atypia. McWilliams and coworkers suggest that overdiagnosis is an important issue. I agree. From the viewpoint of minimizing overdiagnosed cancers, I suggest that the three in situ squamous cancers identified solely by LIFE bronchoscopy are far more likely candidates for overdiagnosis than those of Stage IA or higher. The most important point about overdiagnosis is to ensure that such cases are not overtreated, although this might be of concern as two of the three in situ cancer cases underwent lobectomy. Even McWilliams and coworkers stated that the adenocarcinomas diagnosed by CT appeared to be more aggressive than those diagnosed by LIFE bronchoscopy. Thus, focusing only on the 11 cancers of Stage IA or higher, CT detected 10 of them, as did sputum atypia, resulting in the same sensitivity, except that all those with sputum atypia would require two tests (sputum cytology and subsequent CT) as compared with a single CT. McWilliams and coworkers (1) used an optimized sputum collection method by combined hypertonic saline induction and high-frequency chest-wall oscillation followed by postinduction collection to minimize unsatisfactory specimens. This process is time consuming for the screenee and technician, requiring at least 2 hours as opposed to the 20 seconds required for a CT scan. The reliability of obtaining a satisfactory CT is nearly 100%; the equivalent number was not provided for sputum collection and analysis. Obtaining sputum in former smokers is challenging, and relevant information on this issue was not provided in the paper. Neither were the cost of the sputum collection and subsequent automated cytometry listed. Conversely, CT screening, based on actual data, is highly cost-effective (5). McWilliams and coworkers have probed important questions about improving the early detection of lung cancer. Fundamental limitations of the study undermine their optimistic conclusions regarding the value of automated sputum cytometry relative to CT. A more rigorous study is required to determine the appropriate role for automated sputum cytometry in a cost-effective CT screening program for lung cancer. As demonstrated in the study of McWilliams and coworkers (1), surely cytomorphologic evaluation of the sputum should find some cancers at an earlier and more curable stage, and it should thus be lifesaving in at least some patients (based on simple deductive reasoning). It is unfortunate that research on the usefulness of sputum cytology in screening for lung cancer has been slow to advance. Slow progress is partly the consequence of nihilistic conclusions regarding its effectiveness as a result of earlier randomized trials evaluating the effectiveness of sputum cytology for lung cancer screening in the 1970s. The sputum analysis performed in those studies was limited when compared with today's techniques. I believe, however, that serious methodologic mistakes contributed to the negative results of those studies. Similar mistakes were identified in the randomized trials of mammography screening, leading to considerable confusion about its effectiveness (6). As has been pointed out (7), the misleading results of such randomized trials were mainly due to insufficient years of repeat screening and the failure to focus on the relevant time interval during which the benefit of screening can be identified in such trials. Such trials typically require at least 10 years of continuous screening and the effect is seen only when the analysis is focused on the deaths in the last 2 to 3 years of the trial (7). Investigators who are interested in conducting screening research, including research with cutting-edge biomarkers, need to recognize that their innovations will not gain acceptance until they are shown to be effective in carefully designed clinical studies. The traditional approach for such evaluation is a randomized trial in which screening is only performed for a short period because of the high cost of the trial and that such trials typically fail to show a benefit. Ultimately, those involved in assessment of sputum cytology, like any other type of screening, will have to address the evaluation of its effectiveness, as it needs to rest firmly on evidence as to its robustness, cost effectiveness, and reduction in lung cancer deaths. The most appropriate methodology for such evaluation still awaits serious and open scientific discussions. FOOTNOTES Conflict of Interest Statement: C.I.H. has no declared conflict of interest. REFERENCES
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
