| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
Washington, DC
Dr. Colice is Director, Pulmonary, Critical Care, and Respiratory Services, Washington Hospital Center, and Professor of Medicine, The George Washington University School of Medicine.
Correspondence to: Gene L. Colice, MD, FCCP, Washington Hospital Center, 110 Irving St NW, Washington, DC 20010; e-mail Gene.Colice{at}Medstar.net
Pinsky et al1 suggest in their article appearing in this issue of CHEST (see page 688) that chest radiographs (CXRs) performed as a screening test for lung cancer may also provide important prognostic information about premature death from respiratory and cardiovascular disease. These investigators took advantage of the huge database obtained as part of the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (> 70,000 subjects had CXRs performed annually for 3 years in this trial) to calculate the hazard ratios for all-cause mortality, respiratory mortality, and cardiovascular mortality based on CXR interpretations of abnormalities that were not suspicious for cancer. Approximately 35% of all CXRs were found to have findings that were abnormal but not suspicious for cancer, with granulomas, scarring/pulmonary fibrosis, and cardiac abnormalities being frequently described. Significantly increased hazard ratios for cardiovascular mortality were found for cardiac abnormalities and pleural fluid; COPD/emphysema and scarring/pulmonary fibrosis were associated with significant increases in respiratory mortality. The results are intriguing, but the reader should be cautious about using CXRs, or extrapolating these results to CT scans, for more ambitious screening for three reasons.
First, the implicit rationale behind this analysis is that radiographic screening for lung cancer will prove effective. Although this has not been shown for CXRs,2 there is considerable hope that lung cancer screening with CT scanning will provide a survival advantage.3 Surveys4 of Americans have revealed enthusiasm for cancer screening in general, even if screening would generate unnecessary tests through false-positive results. Surprisingly, most surveyed American adults would want to be screened for cancer even though effective treatment might not be available. However, pursuing a national policy of using CT scans as lung cancer screening tests has enormous financial implications. It is agreed that CT scans can detect lung cancers at smaller sizes and earlier stages than CXRs, but it is still unclear whether early detection with CT scanning will translate into a survival advantage.56 Until the results of the National Lung Screening Trial,7 a 50,000-patient study that is designed to specifically address whether annual CT scan screening in a high-risk population will reduce lung cancer mortality, are available, clinicians should understand that CT screening for lung cancer is not recommended. Taking the next step and using CT scanning to screen for other conditions predictive of early death is similarly not reasonable.
Second, there is an important difference between hypothesis-generating analyses and hypothesis-confirming studies. The allure and availability of the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial database of CXR interpretations made it easy to use sophisticated analyses to look for associations between types of abnormalities found on CXRs and early death. However, findings of a significant association do not constitute adequate proof of a causal relationship. The analyses performed by Pinsky and colleagues1 should be viewed as preliminary and sufficient to generate hypotheses, but they require confirmation by adequately designed confirmatory studies. Ideally, a prospective study examining the relationship between various abnormalities found on CXRs and cardiovascular/respiratory death would also provide information about the pathophysiology involved and the possible benefits of early intervention.
Third, applying a test to large populations requires a clear understanding of the performance characteristics of the test. For CXRs, there are two aspects of performance that should be considered: technical features related to obtaining the image; and the interpretation of the image by the radiologist. Determining the performance characteristics of a test typically involves comparing the test result with a "gold standard." This information is not provided for the CXR interpretations obtained in this study (the "gold standard" presumably being clinical evaluation), which leaves the reader uncertain as to the clinical relevance of the CXR interpretation. More disconcerting is the information provided on intrasubject and interrater agreements on interpretations. The CXRs were interpreted without knowledge or review of prior CXRs by > 300 radiologists. Perhaps not surprising for those of us who review CXRs routinely, a subjective interpretation process, such as reading CXRs, commonly results in disagreement among radiologists and disagreement by the same radiologist on rereading the CXR. This aspect of the approach to both CXR and CT scan screening for disease needs further careful consideration.
In summary, Pinsky et al1 have made interesting, preliminary observations based on a retrospective analysis. Others will need to confirm these findings in well-designed prospective trials. Whether CXR abnormalities would truly predict early cardiovascular/respiratory death and, more importantly, would enable early interventions is a reasonable hypothesis to test in future well-designed prospective trials. However, before pursuing this question, please let us ensure that we are getting what we pay for with lung cancer screening: improved survival through early detection.
Footnotes
The author has served as a consultant or speaker for Adanis, Kos, Teva, SanofiAventis, Pfizer, GlaxoSmithKline, and Almorall.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
