| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorials |
1 From the Department of Radiology, Dartmouth-Hitchcock Medical Center, One Medical Center Dr, Lebanon, NH 03756 (W.C.B.); Department of Radiology, University of California, Los Angeles, California (D.R.A.); and Division of Cancer Prevention, National Cancer Institute, Bethesda, Maryland (C.D.B.). Received May 22, 2006; final version accepted May 25. Address correspondence to W.C.B. (e-mail: william.black{at}hitchcock.org).
Screening for lung cancer with computed tomography (CT) has been a highly debated topic in medicine for several years. Much of the controversy pertains to the choice of study designs that should be used to determine the benefits, harms, and costs of such screening (1–4). In a recent editorial (3), Dr Gur urges the radiology community to initiate a large field trial of CT screening for lung cancer. Although he does not explicitly define what he means by a large field trial, he refers to the Breast Cancer Detection Demonstration Project (BCDDP), which was a single-arm cohort study of screening in more than 280 000 women (5). Dr Gur suggests "two annual screening examinations followed by repeat screening every 2 or 3 years thereafter" in high-risk adults 55–70 years of age. Assuming an average age of 62 years at enrollment and life expectancy of 15 years, the average number of CT screenings per participant would be about eight. We disagree with Dr Gur's recommendation and take this opportunity to explain why we feel a large field trial would be inappropriate.
First, a large field trial would not answer the fundamental question of whether CT screening reduces lung cancer mortality. Because such a study would lack an internal control group, the observed lung cancer mortality would have to be compared with that of some external control group, which would almost certainly introduce selection bias. If the control group had a lower risk of developing lung cancer or better treatment for the disease than did the study group, then the comparison would be biased against CT screening. Conversely, if the control group had a higher risk or poorer treatment, then the comparison would be biased in favor of CT screening.
For example, the average annual age-adjusted lung cancer mortality rate in Kentucky is 80.0 deaths per 100 000 persons, while the rate in Utah is only 24.8 deaths per 100 000 persons (6). (This difference can be largely explained by the different smoking rates among adults, 32.2% in Kentucky versus 17.1% in Utah [7].) If researchers for Kentucky were to perform a state-wide field trial of CT screening for lung cancer and compared their lung cancer mortality rate with that of the entire United States, 55.1 deaths per 100 000 persons (6), they might falsely conclude that the intervention was very harmful. Conversely, if researchers for Utah were to do the same, they might falsely conclude that the same intervention was very effective. Even if researchers for one of these states (or for any other geographic region) were to use their own population from an earlier period of time as the control group, the comparison could be biased by temporal changes in lung cancer mortality. Smoking rates and probably many other factors related to lung cancer mortality change over time, and because some of these factors may be unknown or difficult to measure, analytic adjustments for these temporal changes are not reliable. Similarly, without an internal control group, there would be no reliable way to determine the effects of detecting non–lung cancer abnormalities.
Comparing lung cancer survival statistics from the field trial with those from less intensively screened populations, such as those composing the Surveillance Epidemiology and End Results database, would be even more problematic than comparing mortality statistics. As pointed out in the opposing editorial by Reich (4), the comparison of lung cancer survival would be strongly biased in favor of screening because of self-selection, lead-time, length, and overdiagnosis biases. The only study design that effectively avoids all these biases is a randomized clinical trial, such as the National Lung Screening Trial (NLST) (8,9).
Second, a large field trial would be inappropriate because CT screening could be harmful to a large number of asymptomatic individuals because of false-positive screening results and overdiagnosis (10). Because the participants would be at high risk to die of other smoking-related conditions, principally ischemic heart disease and chronic obstructive pulmonary disease, and because CT can depict very small lung cancers that often have very long doubling times (11,12), overdiagnosis could be a major problem.
In the recently completed Mayo Clinic study of 1520 high-risk adults, one or more noncalcified nodules were detected in 1118 (74%) participants during the five annual CT screenings (13). Although most of these participants were evaluated with chest CT alone (without contrast material enhancement), many were evaluated with invasive diagnostic procedures, including 13 participants who underwent 15 surgeries for benign disease. (Surgery for benign disease is not necessarily benign [14].) In addition, screening depicted potentially resectable lung cancer in 55 participants. Fifty-three of these participants were known to have had surgical resection, but the treatments for two participants remain unknown (S.J. Swensen and A.O. Bungum, oral communication, March 28, 2006). Of these potentially resectable lung cancers, 11 (20%) were stage I bronchioloalveolar cell carcinomas, which tend to have very long doubling times and are of questionable clinical significance (11,12,15).
Extrapolating from the Mayo Clinic results, we estimate that CT screening in the proposed field trial would depict noncalcified nodules in more than 200 000 participants and lead to more than 10 000 surgeries for potentially resectable lung cancer and more than 3000 surgeries for benign disease. Although it is certainly possible that these harms would be outweighed by a reduction in lung cancer mortality, there is no strong evidence to date that this would be the case. Results of the landmark Mayo Lung Project (16), a randomized clinical trial of lung cancer screening with chest radiography, did not demonstrate even a trend in the right direction. Furthermore, in contrast to the proposed field trial, the BCDDP was begun after a 33% breast cancer mortality reduction had already been demonstrated in results of the Health Insurance Plan of Greater New York study (17)—the first randomized clinical trial of any cancer screening intervention. The BCDDP was intended primarily to demonstrate the feasibility of this effective screening on a larger scale in a broader population of women and clinics.
Third, a large field trial would cost a large amount of money and consume a lot of health care resources. With 53 000 participants, three rounds of screening, and equal allocation to screening with CT and chest radiography, the NLST is projected to cost $200 million (18). The CT arm is projected to cost substantially more than the chest radiographic arm because CT scans are much more expensive than chest radiographs. The projected cost of the NLST does not include the costs of diagnostic evaluations for screening abnormalities or treatment for lung cancer and other conditions detected with screening. (These costs are assumed to be covered by the participant's insurance, the participant, or financial assistance from the trial site or nearby organization.) Given more than five times the number of participants, more than twice the number of screening rounds, and CT screening for all participants, the field trial would cost at least 10 times as much as will the NLST—more than $2 billion! And this estimate does not include the costs of diagnostic procedures to evaluate CT abnormalities or treatments for lung cancer and other conditions detected with screening.
We believe that observational studies can provide useful insights into screening for lung cancer and other diseases, particularly in initial exploratory investigations. However, such studies should not require an intervention in 280 000 subjects. In fact, a number of useful, appropriately sized studies have already been conducted. The Early Lung Cancer Action Project (19) and the Mayo CT study (13) are single-arm prospective cohort studies with 1000 and 1520 participants, respectively. Results of each study have provided information on the feasibility of CT screening for lung cancer, including the prevalence and incidence of lung cancer in populations eligible for screening and the accuracy of CT screening. However, neither study has a control group that allows for an unbiased comparison of lung cancer mortality, which is necessary to determine the effectiveness of CT screening.
In summary, we oppose a large field trial of CT screening for lung cancer. Such a study could seriously harm tens of thousands of previously asymptomatic individuals, consume billions of dollars, and still fail to answer the fundamental question—does CT screening reduce lung cancer mortality. The NLST is on track to answer this fundamental question, as well as other important questions about the benefits, harms, and costs of CT screening for lung cancer.
 |
FOOTNOTES |
|---|
D.R.A. has given expert testimony related to the subject of this article.
See also the editorial by Gur in this issue.
|
References |
|---|
 TOP References |
|---|
Related Article
This article has been cited by other articles:
|
|
 |
|
  D. F. Yankelevitz, W. C. Black, D. R. Aberle, C. D. Berg, and on behalf of the Executive Committee of the Nation Will the National Lung Screening Trial Be Able to Demonstrate a Mortality Reduction? Radiology, February 1, 2008; 246(2): 653 - 654. [Full Text] [PDF]
|
|
|
|
|
|
D. Gur Large Field Trial for Lung Cancer Screening: Putting the Wrong Cart before the Horse?--Commentary Radiology, May 1, 2007; 243(2): 317 - 318. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
