This article is more than 5 years old. Information may no longer be current.
CT screening for lung cancer: Wait for the data
The rush to use the technology is premature and may be harmful to public health and patient care.
Click Here to Manage Email Alerts
The American Cancer Society estimates that there will be 217,000 new cases of lung cancer in the United States in 2007, and there were more than 1.2 million cases worldwide in 2004.
Lung cancer is associated with more deaths in the United States than any other three cancers combined, with a five-year estimated survival rate of only 18%. The small improvements in survival from adjuvant therapy are just becoming felt.
|
Since the vast majority of lung cancer is found in a regionally or widely advanced state, finding earlier cancer should be a priority. Screening might be an option. Earlier detection, if possible, could improve outcomes, assuming that all lesions progress to death. We currently have technology that can detect smaller lesions, but are we ready to screen all high-risk individuals?
Fifty years ago, researchers evaluated the benefit of mass screening by X-ray. The study took place in Boston, with more than 500,000 X-rays in the general population. Although 43 cases were found, there were few early lung cancers.
In the early 1970s, the Philadelphia Pulmonary Neoplasm Project enrolled more than 6,200 high-risk men. These men were randomly assigned to yearly chest X-rays or usual care, when the X-rays would be performed if a patient became symptomatic. After more than 30 years of follow-up in this project, there was no survival benefit, despite the fact that they did find more cases of lung cancer in the group receiving chest X-rays.
More recently, a group at the Mayo Clinic evaluated the effects of screening with chest X-rays. After 20 years of follow-up, the researchers found 43 more cancers in the group that received the X-rays, but there was no difference in overall survival in this randomized trial.
The Mayo group concluded that annual chest X-rays do not improve outcomes because not all small lesions will inevitably progress to death from lung cancer. Some cases will be indolent, and this presents the problem of overdiagnosis.
Based on the data from these trials, experts concluded that chest X-ray is not an effective screening test. But what would be effective?
Effective screening
To be effective, a screening test must be both sensitive and specific. The screening test must be widely applicable to the screened population by being both low cost and widely available. The focus of the test should be on incident cases (those found after the initial screen), and the screened population must have improved survival when compared with a concurrent nonscreened population. Without the concurrent control, disease-specific mortality must be considered.
Does computed tomography provide an option for screening for lung cancer? Data from two recent studies of more than 5,000 CT scans each showed prevalence cases early and incidence cases evolving. The researchers concluded that screening works, but this may have been a premature conclusion.
High-resolution CT scanning can find nodules in nearly 25% of urban dwellers, but more than 95% of these lesions are benign. Although CT scanning has shown promise in observational studies, there is an estimated cost of more than $2 billion annually just for the screening tests, and we need to add the cost of follow-up scans and biopsies to determine the true cost.
In 2006, the I-ELCAT study by Henschke and colleagues (N Engl J Med. 2006;355:1763-1771) drew a lot of attention to CT as a screening tool. The researchers observed 30,000 patients in an observational, noncontrolled study and identified 492 cancers.
They claimed the screening was cost effective and that such screens should start now. Advocacy groups got wind of it and started clamoring for insurance coverage for all smokers and family members at risk. Legislation has been introduced to mandate coverage and advocates are calling for closure of randomized trials, but patience is called for.
Why wait?
Although this observational study has garnered a significant amount of attention, there are a number of pitfalls to any observational study: lead-time bias, overdiagnosis bias, wrong metrics, early prediction errors and the question of prevalence over incidence.
Lead-time bias happens when death occurs at the same biological time regardless of whether we find the cancer earlier. Earlier detection thus extends the measured survival time but not the actual mortality.
Overdiagnosis biases are caused by the basic assumption that all cases will proceed to lethality, but this may not be correct. Some cases are found incidentally at necropsy. Some cases grow slowly but are not the cause of symptoms or death, and these cases add apparent survival in uncontrolled studies.
Because of the lead-time and overdiagnosis bias, survival is the wrong metric for uncontrolled screening. Observational studies are thus usually inconclusive.
Evaluating I-ELCAT
The I-ELCAT study had 535 positive screens. Of these, 492 were cancers, which included 479 lung cancers, 13 other cancers and 43 false-positives.
Of the 479 lung cancers, 412 were stage I, including 405 prevalence cases and 74 incidence cases. Five cases were found between the scans.
Among the 412 stage I cases, eight never had any treatment, and all eight died.
The median follow-up of this study is only 3.3 years, and there are no cost-effectiveness data or analyses. Many of the conclusions are far too premature.
Even with a follow-up of only 3.3 years, Henschke and colleagues projected a 10-year survival of more than 90%. However, there is a huge lead-time bias here. Other trials for surgery with lesions smaller than 15 mm (at the same institution) showed an approximately 85% five-year survival and a 67% 10-year survival. The projection here is simply too premature.
The researchers also concluded that CT scanning could lead to a cure rate of 80% for lung cancer; however, this is unlikely. The overdiagnosis bias leads to us counting small lesions of unknown biological significance, and there is no scientific basis for this claim.
The researchers further asserted that CT screening for lung cancer is cost effective, but there is no information provided beyond the possible cost of a scan. There are no data on life years gained and no consideration of the cascade of secondary costs. If the screening failed to improve mortality, the costs of care would increase.
For this reason, the I-ELCAT study has been the subject of numerous contrary editorials and at least one contrary study.
Currently, there are two trials in the United States with more than 50,000 patients that are already completely enrolled. The monitoring will look at the data next year.
Until we see the randomized, controlled trial data, we should continue our current approach by counseling our patients to quit smoking. Above all we must protect the integrity of randomized, controlled trials.
For more information:
- Joseph Aisner, MD, Chief Medical Officer at the Cancer Institute of New Jersey, is Section Editor of Hem/Onc Today’s Lung Cancer section.