The Myth of
Early-detection: Advances in Treatment--- not Screening Mammography--- Probably
Account for Decline in Breast Cancer Mortality Rates, 1963-2013
Part 1----Screening
Mammography/Early Detection
I
became skeptical of the early-detection hypothesis a few years ago when I heard
a cancer expert confess that no one knows “the natural history” of breast cancer,
i.e., that we don’t completely understand how breast cancers develop and
change. We’re not sure when they first occur, why the smallest that can be
detected has already metastasized or how some shrivel up and disappear on their
own. Apparently, for the past 50 years, we’ve naively assumed breast cancers
develop like everything else.
Hoping
it might help determine whether the early detection hypothesis had any
validity, I reviewed the history of screening mammography beginning with the
first screening trial 50 years ago. Below I’ve summarized the (1) trials and
studies, (2) mammography innovations and (3) the most recent research on the
risk of ionizing radiation I’ve reviewed. I’ve also inserted (in italics) my
experience as a breast cancer patient wherever I thought it might be useful.
The 1960s
The First Trial of Mammography Screening: The
Health Insurance Plan (HIP) Trial (1963-1986)1 was inspired by
Dr. Michael Shimkin, a researcher and bio-ethicist at the National Cancer
Institute. He was interested in a
brand-new mammography technology developed by Dr. Robert Egan, one of the “fathers
of modern mammography.” The HIP trial was the first randomized-controlled trial
of mammography screening. Unfortunately, it appears to have also been the last.
In its
simplest form, a randomized-control trial compares a study vs. a control group.
In the HIP trial, women in the study group were screened; those in the control
group weren’t. When the two groups are well-matched, any differences found between
them may be attributed to differences in how the study group and the control
group were treated.
The HIP
researchers’ access to the Health Insurance Plan of Greater New York2
enabled them to match each woman in the screening group to a woman in the
control group who was the same age, had the same number of children, had the
same family income, and lived and/or worked in the same Greater New York Area. The
two women weren’t twins or siblings, but they very well-matched neighbors. Approximately
30,000 women between the ages of 40 and 64 were assigned to each group.
Women
in the study group were screened by both mammography and clinical breast
examination (palpation). They were expected to attend an initial screening followed
by three yearly re-examinations.
The
women in the control group were left to follow their usual medical practices.
Since mammography for asymptomatic breast cancer was not a HIP benefit, it was
unlikely that any of them were screened.
In
spite of the researchers’ efforts, only 8,060 women completed all four
screenings. Ten thousand women in the study group didn’t attend the initial
screening (the refusers) and another 11,940 didn’t return for one or more of
the three yearly examinations.
The
refusers were an interesting group. They carried a greater burden of other
diseases and fewer of them than the other women in the control group developed
and died breast cancer.
The
HIP Trial Results: (1) Five percent fewer breast cancer deaths occurred among
the screened than the unscreened women. It’s unclear however, whether the lower
rate of breast cancer mortality was due to screening. It could have been due to
differences in treatment.
(2) Ten percent more of the screened than
unscreened women were treated with a radical mastectomy. The difference is large
enough to account for the observed difference in breast cancer deaths.
Apparently, the authors of the study didn’t notice.
(3) The evidence for early detection was meager: screened
women (1) were less likely to have traces of breast cancer cells in the lymph
nodes under their adjacent armpit and (2) 0.1% fewer had invasive breast
cancers. Plus, no data analyses comparing the mortality data for these
subgroups of women were done.
And, (4) ten years after having been first
screened, a greater proportion of 40-49 year-old women in the screening group
than the control group died of their breast cancer. Eighteen years after, most
of the 5% difference in mortality rate occurred in women whose breast cancers
were detected after the “women had passed their 50th birthday.”
What Can Be Concluded: (1)
Screening did not benefit women in their 40s, (2) many women saw no advantage
in being screened for breast cancer, and (3) the authors overlooked the
possible effect of treatment on mortality.
The
1970s
In 1972, the American Cancer
Society began its Breast Cancer Detection
Demonstration Project. The Project was supposedly based on the “success” of
the HIP trial, yet the Society rejected a randomized-controlled trial and decided
in favor of a demonstration project3.
It wanted to demonstrate: (1) that “early
detection cures cancer” and (2) that large numbers of women, particularly those
in their 30s and 40s, could be enticed to be screened. In addition, it needed to
raise enough money to continue to exist3.
The Project was a joint endeavor
of the Society and the National Cancer Institute. The Society set the agenda and the National
Cancer Institute contributed over $6 million a year. Twenty-nine detection
centers were created and the media was extensively used to attract 270,000
volunteers between the ages of 35 and 74. Chaos ensued.
In response to the
well-publicized breast cancer experiences of Happy Rockefeller and Betty Ford,
so many women in their 30s and 40s volunteered, they made up half of the women
in the Project. When news about the dangers of ionizing radiation appeared
during the Project’s early years, women refused to complete the required five
screening sessions. Later they reenrolled after hearing the Society had replaced
the more dangerous machines. And, when
the media reported that a pathologist reviewing pathology reports had
discovered that 66 women were misdiagnosed (perhaps the first known incidence
of over-diagnosis), women again stopped enrolling.
Although only asymptomatic women
were eligible, 6% had lumps in their breasts, 18% had had previous breast
surgery, and 20% had had at least one mammogram prior to entry--- as many as 118,800
of the 270,000 volunteers could have been ineligible. Furthermore, they did not represent women in
the general population. Most were upper middle-class.
Dr. Myles P. Cunningham,
President of the Society in 1997, put the best spin on it he could when he
claimed that “… the Project significantly advanced both the notion and the
science of population-based breast cancer screening.4” In fact, the
Project probably made it impossible to run a truly randomized control trial
ever again. Given its wide use of the media, researchers could no longer “assign”
women to a control group, i.e., to not be screened. The only options left were
to either use volunteers or study the effects of breast cancer screening in
already existing populations.
What Can Be Concluded: (1) Large numbers of women in their 30s and
40s could be enticed to be screened and (2) the Society succeeded in marketing
itself and mammography screening. It raised the money it needed to continue to
exist. Scientifically, its Project was a failure and probably restricted how
future screening mammography research could be done.
1980-2000
The Canadian National Breast Screening
Study-1 began In 19805. Only women in their 40s were eligible.
All 50,430 participants were volunteers.
I was
one.
We were told that we could be
randomly assigned to either the screening group or the control group. So, each of
us agreed to the possibility of not being screened. After agreeing we were randomly assigned to
either the screening or non-screening group: 25,214 to the screening group and 25,216,
to the non-screening group.
I was
assigned to the screening group.
As in the HIP trial (1) women
were screened by both mammography and a clinical breast examination; (2) women
in the control group were expected to continue their normal pattern of medical
care and (3) since mammography screening was not a benefit of Canada’s
universal health plan at the time, it was unlikely that women in the non-screening
group would be screened. However, 1799 women
(7.1%) in the screening group and 6655 (26.4%) in the non-screening group had
one or more mammograms outside of the trial.
The women in the non-screening
group who had mammograms outside the trial differed. They had a greater risk of
dying from breast cancer.
The Canadian Trial Results: (1) Screening
did not benefit women in their 40s. Ten years after their first mammogram, more
women in the screening groups than in the control group had died of breast
cancer. And, sixteen years after entry,
105 in the screening group and 108 in the non-screening group had died of
breast cancer.
(2) Again,
no mortality data were provided for possible evidence of early detection.
Seventy-one in situ breast cancers were detected in the screening group; 29, in
the control group.
And, (3) in the version of the
trial I read, no mention was made of how the women with breast cancer were
treated.
What Can Be Concluded: Again, (1) there was no reliable evidence that
screening benefits women in their 40s and (2) a data analysis that could have
shown a direct link between early detection and mortality was apparently not
done, but twenty years after the HIP trial (3) many women now saw an advantage
in being screened.
The
2000s
Computer-aided
detection (CAD) --- pattern recognition software that marks and brings
suspicious-looking regions of mammograms to radiologists’ attention---was
introduced in about 2004.6 CAD‘s design is based on the
early-detection hypothesis.
Detecting anything that could possibly be a
breast cancer as soon as possible becomes of utmost importance if it’s true
that women will die when their breast cancers are not detected as early as
possible. CAD marks everything that
might be suspicious. As a result, it produces an average of 167 to 500 false
marks per 1000 mammograms.
Radiologists
using CAD find more breast cancers (many of which are minuscule) and make more
false alarms. That’s fine if the early detection hypothesis is true; it’s not
fine if it’s false. False positives and over-diagnoses ---diagnoses of breast
cancer that would not have been diagnosed in woman’s lifetime if they had not
been screened----reduce the quality of women’s lives.
The anxiety and fear associated
with false positives are not short-lived7. Some women suffer from sleep disturbances and
other symptoms of anxiety for as long as three years. And, over-diagnosis, in addition to
subjecting women to unnecessary treatment, leads to life-long recurring
episodes of fear that the breast cancer will recur and metastasize, to say
nothing about the depression that often accompanies breast cancer diagnoses.
In 2009 the United States
Preventative Services Task Force reviewed a large and varied set of trials and
studies, assessed their reliability and validity and recommended (1) women in
their 40s not initiate regular screening and (2) 50-74 year-old women be
screened biennially and not annually. Their new recommendations attempt to
minimize the harms of over-diagnosis and false positives while retaining the
purported benefits of screening.
What
Can Be Concluded: In the 2000s, (1) a powerful imaging adjunct was
developed that increased the risk of false positives and over-diagnosis and (2)
the United States Preventative Services Task Force revised its guideline to
minimize the potential harms of mammography.
2010-2012
As far as I know, the Task Force
did not take into account the possible harm of ionizing radiation. In 2010, an
article containing estimates of the number of
mammography-induced breast cancers and deaths due to repeated exposure of screening
mammography’s low levels of ionizing radiation appeared in Radiology8. (An
average glandular dose of 4.7 mGY per screening was assumed).
The estimated risk is 0.91 per
1000 women for women screened annually from age 40 to 80. That’s high, almost 1
per 1000 women. If that were the risk for a drug, even if it were effective and
saved women’s lives, it would be shunned by many patients and physicians.
However if women wait until
they’re 50 to initiate screening; the risk is reduced to 0.40 per 1000 women. And
the risk of dying from a mammography-induced breast cancer is reduced from 0.25
to 0.12. That could mean that during their life-span, five of the 25,214
women in their 40s who were screened in the Canadian trial will develop a
mammography-induced breast cancer, one or two of whom might die. Thus, while
screening may have saved the lives of three women, it also may have endangered
the lives of one or two others.*
In 2011, the Canadian Task Force on Preventive
Care suggested that “No” was a legitimate response to mammography screening for
women who (1)didn’t place a high value on the small chance that screening might
save their lives and (2) were concerned about false-positive readings and
over-diagnoses9.
Professor
Peter C. Goestzache, Director of the Nordic Cochrane Center, Rigshospitalet,
Copenhagen, citing the results of Danish and Swedish mammography studies, suggested
that “It may be most wise to avoid screening altogether, at any age”9, 10.
In one
large area of Denmark, screening was available for 17 years. In another, it
wasn’t. The decline in mortality from breast cancer was greater (by 1%) in the
area where screening was not available,
suggesting that some breast cancers disappear. The authors estimated that a third
of the women who had been screened were over-diagnosed.
And, in
fact, untreated cancers do sometimes simply disappear: When my bone metastases were detected, CT and bone scans showed that
some of the lesions had completely healed even though I had had no treatment during
the ten years between when my breast cancer was first detected in 1985 and its recurrence
discovered in 1995.
A
group of Swedish women who had been screened for a period of six years were
compared to a group who were not screened during the first four years of the
same six-year period. The accumulated
number of breast cancers was lower in the group of women who were not screened for the initial four years
---yet more evidence that screening leads to over-diagnoses.
In 2012, a study by Drs. A.
Bleyer (Oregon Health and Science University in Portland) and H. Gilbert Welch
(Geisel School of Medicine at Dartmouth College, Hanover, NH) appeared in the New England Journal of Medicine showing
that over the past 30 years, screening had: (1) failed to reduce the number of
women walking around with late-stage breast cancer, something which should not have
happened if early detection had worked as hypothesized and (2) led to 1.2
million women being over-diagnosed.
However, EUROSCREEN, a cooperative group made
up of experts who evaluated population-based screening programs in nine
countries (France, Italy, Germany, the Netherlands, the United Kingdom, Norway,
Sweden, Denmark, and Spain), estimated that the number (7-9 per 1000 women) of
lives saved by breast cancer screening programs exceeded the risks of over-diagnoses
(4 per 1000 women) and false positives (200 per 1000 women)11.
The
EUROSCREEN numbers however, are not relevant for the U.S. They’re based on
biennial screening programs for women 50-69 years old, a much more efficient
screening strategy than the one commonly used in the U.S.
Furthermore,
the estimated number of lives saved is probably too high. The EUROSCREEN estimates
combined the reduction in mortality rate observed in case-control studies (48%)
with the reduction in incidence based mortality rates (38%). Case-control studies are notoriously biased
and tend to be overly positive.
What
Can Be Concluded: In the past three
years, evidence of the harms of mammography screening---mammography-induced
breast cancers, false alarms and over-diagnoses---has been rapidly
accumulating.
Conclusion
Over
the past fifty years, screening research has overlooked the contribution of
treatment to the reduction in rate of breast cancer mortality, even though,
without effective treatments, screening and the early detection of breast
cancers would be useless.
Early
detection may have been important when treatments were mostly ineffective. For
example when mastectomy was the only treatment option (see the HIP trial
results above). It might still be important for breast cancers for which the
new targeted therapies are not effective. We don’t know.
We
do know however, that women in their 40s are still being urged to be screened
even though they do not benefit from screening and are particularly vulnerable
to ionizing radiation. Those data could not be clearer. And other, more recent,
data indicate that large numbers of women of all ages have been needlessly exposed
to the harms of false positives and over-diagnosis.
Despite the consistent lack of evidence during the last fifty years, the hugely successful and persistent marketing
of the myth behind mammography screening has drained our resources and distracted
us from doing what we should have been doing----decoding the natural history of
breast cancers. And that, I fear, may be its greatest harm.
References
* Since the 25,214 women were
screened just five times between the ages of 40 and 49, the number of women who
might develop a mammography-induced breast cancer is 6.3 (25,214 x 0.25). And
the number who might die of that breast cancer is 1.575 women (6.3 x 0.25).
1Shapiro, S., Venet,
H., Strax, P. & Venet, L., 1988, the
Health Insurance Plan Project and Its Sequelae, 1963-1986. Baltimore: The
Johns Hopkins University Press.
2The American Journal of Public Health, Vol. 35, June, 1945, pp. 643-644.
3Lerner, B.H., March,
2001, To See Today with the Eyes of
Tomorrow: a History of Screening Mammography, A Background Paper for the
Institute of Medicine report: “Mammography and Beyond: Developing Technologies
for the Early Detection of Breast Cancer.”
4Cunningham, M.P. 1997,
The Breast Cancer Detection Demonstration Project 25 Years Later, CA---A Cancer Journal for Clinicians, 47, 131-132.
6Zivian, M.T. &
Gershater, R., 2008, The Accuracy of Diagnostic Radiology, in M. A. Hayat (ed.)
Cancer Imagining Instrumentation and Applications, Vol. 2, Elsevier Academic
Press, pp.109-118.
9Time to stop
mammography screening, Canadian Medical
Association Journal (2011), 183
(17), pp. 1957-58)
10British Medical Journal, October, 2012.
11Journal of Medical Screening, 2012, 9: 5-13.
