Over-diagnosis in breast cancer screeningBMJ 2006; 332 doi: https://doi.org/10.1136/bmj.38768.401030.7C (Published 23 March 2006) Cite this as: BMJ 2006;332:691
All rapid responses
Møller and Davies (M & D) draw attention to the importance of
overdiagnosis in breast cancer screening , but misrepresent both
benefits and harms. They extrapolate the effect seen in the screening
trials to the remaining lifespan of the invited women and use lifetime-
breast cancer mortality (after age 50) in their calculation of benefit
. However, an undiminished effect on mortality after screening has
ceased is very unlikely. M & D also disregard that breast cancer
mortality increases with age, and their estimate of overdiagnosis needs to
be corrected for opportunistic screening in the control group .
Overdiagnosis is defined as cancers that would never have caused
symptoms, nor been detected, in the absence of screening . With
reference to the updated Malmö trial , M & D suggest that the level
of overdiagnosis is 10%. However, the serious flaws in this estimate,
which has been diluted by 15 years without screening, need to be taken
into account [2,5]. It is possible to infer the number of undetected
cancers that exist at the end of the trial in the control group because
they are subsequently diagnosed as “catch-up” cancers (calculated as the
difference between number of cancers identified in the screened group and
the control group during the next 15 years) . Thus, it is possible to
adjust for the lead-time bias caused by screening. Adding these cancers to
the control group, the level of overdiagnosis when the trial ended becomes
18% . However, it is also necessary to adjust for the finding that 24%
of the women in the control group had undergone screening at least once
during the trial . In the screening trials from Stockholm, Göteborg and
Two-County, the control group was invited to screening when the trials
ended, and this had the effect that the number of cancers became very much
the same as in the screened group (relative risks 0.98-0.99 [7-9]). Thus,
a single screening led to a similar level of overdiagnosis as repeated
screenings. This means that we can estimate the level of overdiagnosis, x,
in the Malmö trial from:
(1+x)/(1+0.24x) = 1.18, which gives an overdiagnosis of 25%.
In the Malmö trial, as originally reported, there were 588 cancers in
the screened group and 447 in the control group, i.e. a difference of 141
cancers, corresponding to a crude overdiagnosis of 32% . The adjusted
calculation gives 588 - 588/1.25 = 118 cases of overdiagnosis (25%). There
were 3 fewer deaths from breast cancer in the screened group than in the
control group . Thus, after nine years of screening in the Malmö trial,
39 healthy women had been diagnosed and treated for breast cancer
unnecessarily for every woman who had her life prolonged because of
screening, i.e. a harm/benefit ratio of 39.
It is more informative, of course, to estimate the harm/benefit ratio
from the results of all the screening trials. There are many problems with
the quality of the other screening trials [10,11], and the often quoted
estimate of a 30% reduction in breast cancer mortality is not tenable. In
their review of the randomised trials, The US Preventive Services Task
Force reported a breast cancer mortality reduction of 16% . We used
15% as our estimate which gives 2000 as the number needed to screen to
prolong one life . By their inappropriate extrapolations, M & D
estimated the number needed to screen at only 250, which is 4 times more
optimistic than what has ever been reported in any screening trial [10,12]
or screening program.
We believe that our estimate of 30% overdiagnosis is also reasonable
, in particular considering that observational studies have reported
higher estimates [12-14]. This gives a harm/benefit ratio of 10 after 10
years of screening , which is 5 times bigger than the harm/benefit
ratio of 2 suggested by M & D.
Finally, M & D claim that when screening ceases, the breast
cancer incidence rate returns to a lower than expected rate. The IARC
report on screening noted that “No population into which breast screening
has been introduced has yet been reported to show an unequivocal return of
incidence rates to the baseline expected from pre-screening trends” .
Indeed, the incidence has remained elevated in Sweden in the older age
groups, long after screening ceased , and persistent increases in the
population have also been observed in other countries that have had
screening for many years, e.g. in the USA and the UK [13,15].
We agree that overdiagnosis must be recognised and estimated. It is
the most important harm of mammography screening and it is not clear
whether the benefit outweighs the harms. It is indefensible that women are
not being told that screening will turn many of them into cancer patients
1. Møller H, Davies E. Commentary: Over-diagnosis in breast cancer
screening. BMJ 2006;332:691-2.
2. Gøtzsche PC. Overdiagnosis in the Malmö mammography screening
trial was considerably underestimated. BMJ 2006;332:727.
3. Vainio H, Bianchini F. IARC Handbooks of Cancer Prevention. Vol 7:
Breast Cancer Screening. IARCPress; Lyon 2002.
4. Zackrisson S, Andersson I, Janzon L, Manjer J, Garne JP. Rate of
over-diagnosis of breast cancer 15 years after end of Malmö mammographic
screening trial: follow-up study. BMJ 2006;332:689-92.
5. Welch HG. How much overdiagnosis?
6. Andersson I, Aspegren K, Janzon L et al. Mammographic screening
and mortality from breast cancer: the Malmo mammographic screening trial.
7. Frisell J, Eklund G, Hellstrom L, Lidbrink E, Rutqvist LE, Somell
A. Randomized study of mammography screening-preliminary report on
mortality in the Stockholm trial. Breast Cancer Res Treat 1991;18:49-56.
8. Bjurstam N, Björneld L, Duffy SW, Prevost TC. Author Reply. Cancer
9. Tabar L, Fagerberg G, Day NE, Duffy SW, Kitchin RM. Breast cancer
treatment and natural history: new insights from results of screening.
10. Olsen O, Gøtzsche PC. Cochrane review on screening for breast
cancer with mammography. Lancet 2001;358:1340-2.
http://image.thelancet.com/extras/fullreport.pdf (accessed 8 Feb 2006).
11. Humphrey LL, Helfand M, Chan BK, Woolf SH. Breast cancer
screening: a summary of the evidence for the U.S. Preventive Services Task
Force. Ann Intern Med 2002;137:347-60.
12. Jørgensen KJ, Gøtzsche PC. Content of invitations for publicly
funded screening mammography. BMJ 2006; 332;538-541.
13. Gøtzsche PC. On the benefits and harms of screening for breast
cancer. Int J Epidemiol 2004;33:56-64.
14. Zahl PH, Strand BH, Mæhlen J. Incidence of breast cancer in
Norway and Sweden during introduction of nationwide screening: prospective
cohort study. BMJ 2004;328:921-4.
15. Advisory Committee on Breast Cancer Screening. Screening for
breast cancer in England: past and future. Sheffield: NHS Cancer Screening
Programmes. NHSBSP Publication No 61, February 2006. ISBN 1 84463 026 9.
Competing interests: No competing interests
The arguments and counter-arguments about screening for breast cancer
have been difficult to follow. There have been many people, including some
who have written letters published in this issue of the BMJ, arguing that
women must be given the correct information. Up until now, that
information has often involved arcane statistical arguments that few women
will have been able to understand. In their commentary, Møller and Davies
state the problem simply (albeit, inevitably, simplistically) and clearly:
"In a population where the lifetime risk of breast cancer is 8% and
the lifetime risk of dying from breast cancer from age 50 onwards is 2.5%,
screening 250 women may prevent about one death from breast cancer.
Screening would, however, also lead to the over-diagnosis of two cases."
So there it is: how do we decide whether the risk-benefit is "worth
it"? I am not a woman, but do I want a 1 in 250 chance of being saved? Or
a 2 in 250 chance of treatment that I don't need, of being told I am ill?
This is a cruel and excellent example of many decisions that we
pretend that patients - or anybody - can make. They are intangibles. There
is a new operation; it is better than the old one because the recovery is
faster. But it has a higher risk of (rare) disaster. At what level of risk
do we decide that the new operation is in all senses better? What of the
patients badly harmed by the new operation? How do we balance 100 (or 200,
or 300?) patients who used less morphine and were back at work within 2
weeks against 1 (or 2, or 3) patients permanently disabled?
One thing is certain: the risks must be fully stated, but that won't
make life, or medicine, easier.
Competing interests: No competing interests