After tracking smokers for 20 years, a large study indicates that screening for lung
cancer with chest X-rays does not save lives. The new report reinforces original conclusions
from the Mayo Lung Project, published in the mid-1980s, that X-rays at frequent intervals do
not decrease the death rate from lung cancer. The analysis also points to a potential problem
for any type of lung imaging: the detection of tumors that are not life threatening.
The study by Pamela Marcus, Ph.D., and colleagues from the National Cancer Institute (NCI),
Bethesda, Md., and the Mayo Clinic, Rochester, Minn., appears in the Aug. 16, 2000,
Journal of the National Cancer Institute*.
In their analysis, Marcus and colleagues present compelling evidence that a substantial
number of tumors uncovered between 1971 and 1983 in the 9,211 participating men turned out
never to cause serious illness or death. In the absence of screening, these tumors would not
have been found. Such over-diagnosis can lead to unnecessary worry or, more seriously, to
expensive and risky biopsies or surgery.
Marcus' findings arrive in the middle of a debate over a newer screening technology, spiral
computed tomography (CT) scans, and could slow enthusiasm for the scans until they are
properly studied, she said. "A significant reduction in death rates is the gold standard
for any cancer screening test. Our follow-up of the Mayo Lung Project shows that an intense
regimen of chest X-rays in the 1970s and 1980s did not meet this standard. Likewise, until
spiral CT scans are proven to save lives, they should not be recommended as a cancer screening
test. The benefits of any screening test must outweigh the harm."
In the Mayo Lung Project, men were split into two groups: half received free chest X-rays
and sputum tests three times yearly for six years; half received the Mayo's standard 1970
recommendation to receive the same tests annually. At the end of 1996, with an average
follow-up of more than 20 years, the number of deaths from lung cancer was statistically
identical for each group: 337 men in the screening arm had died from the disease, compared
with 303 in the so-called "usual care" arm. If effective, the screening would have
significantly reduced the number of deaths in the screening arm. Instead, the mortality rates
were indistinguishable (4.4 per 1,000 person-years in the screening arm vs. 3.9 in the usual
care arm – not statistically significant).
After publication in 1986, the Mayo Lung Project generated controversy among the medical
community. Some researchers argued that the study did not include enough volunteers to
detect a small benefit for X-ray screening. As designed with the knowledge and resources
available, the study had the statistical power to detect a 50 percent or larger
reduction in mortality.
Though such a large decrease was not seen, the possibility of X-ray screening achieving
a smaller benefit – perhaps a 10 percent to 20 percent reduction in death rates – still
lingers. A much larger NCI-funded study, the Prostate, Lung, Colon, and Ovarian (PLCO) Cancer
Screening Trial, is expected to answer this question by 2015. (Unlike the Mayo Lung
Project, PLCO includes women.)
But another criticism of the Mayo Lung Project remained after publication in 1986: it did
not track participants long enough to show a true benefit. Some lung tumors grow slowly, and
the short follow-up time after screening (three years, on average) may have been inadequate
to see a true mortality reduction, asserted the critics.
To test whether they were right, Marcus and her team matched medical files from the
project with death records from the Mayo Clinic and the National Death Index, housed at the
National Center for Health Statistics. If no death certificate was found for a participant, the
researchers assumed he was still alive. If the participant was reported dead, the researchers
recorded the cause and date. This process led to the primary finding and to the conclusion
that chest X-rays may lead to over-diagnosis of lung cancer.
While the men in the screening arm did not experience a mortality benefit, they did have
longer survival times, measured from diagnosis to death. Intuitively, this sounds like
screening worked. However, according to the authors' analysis, something else happened: chest
X-rays detected tumors that did not lead to death.
"We now have good evidence that lung cancer lesions with limited clinical relevance
exist," said Marcus. "These lesions would never be diagnosed in the absence of screening, and
would not cause death. The Mayo Lung Project picked them up with chest X-rays. Spiral CT is
much more sensitive and will probably pick up even more of them." She added that high
mortality rates, as seen with lung cancer, do not imply that all lung cancers are
lethal. They simply imply that clinically diagnosed cancers are lethal.
Other cancers – prostate and breast, for instance – also appear in forms (prostate
intraepithelial neoplasia and ductal carcinoma in situ of the breast) that only sometimes
progress to full-blown cancer. These conditions, virtually unknown before the advent of
widespread screening, account for a substantial portion of all diagnosed prostate and breast
tumors. How – or if – to treat them remains an open question.
Consequences of over-diagnosis of lung cancer include psychological stress and
unnecessary biopsies or surgery. Biopsies are potentially risky procedures that remove
a small amount of tissue, either through a scope fed down the windpipe (bronchoscopy) or
with a needle through the rib cage (CT-directed needle biopsy). Possible complications
from biopsies include bleeding, infection, and pain and discomfort. Depending on the size
and location of the lesion, chest surgery (thoracotomy) to obtain a larger biopsy may be
recommended. Thoracotomies are major surgical procedures that remove substantial amounts
of tissue; the procedure can damage nerves in the chest and may lead to chronic pain.
Lung cancer will be diagnosed in an estimated 164,100 people and claim 156,900 lives
in the United States this year.
This research was supported by funds from NCI's Division of Cancer Prevention.
Co-authors include Richard M. Fagerstrom, Ph.D., and Philip C. Prorok, Ph.D., from NCI; and
Erik J. Bergstralh, M.S., William F. Taylor, Ph.D., David E. Williams, M.D., and
Robert Fontana, M.D., from the Mayo Clinic.
* The article is titled, "Lung Cancer Mortality in the Mayo Lung Project: Impact of
Extended Follow-up," and appears in the Journal of the National Cancer Institute, Vol. 92,
No. 16, Aug. 16, 2000.