Model of outcomes of screening mammography: information to support informed choices
BMJ 2005; 330 doi: https://doi.org/10.1136/bmj.38398.469479.8F (Published 21 April 2005) Cite this as: BMJ 2005;330:936All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Barrat et al. compared screening compliers and refusers in their
incidence and mortality from breast cancer and their overall mortality.
They express surprise at the small difference in mortality between these
two groups. Indeed, it should be expected that the refusers have a much
higher mortality because of their higher background of health problems.
Therefore, it is likely that comparable groups (which these are not) would
make it even more difficult to detect a difference.
If even one of the world's flagships of mammography screening is
apparently needing magnifying glasses to detect the desired effects, then
the ordinary rest of us will face even greater difficulties in finding
support for women's hope of a mortality benefit through screening
mammography.
Furthermore, neither Barratt et al. nor the editorial by Taylor
consider the additional x-ray exposures required to localize screening
abnormalities for histology. Radiographically directed fine needle biopsy
requires an additional 7 to 9 x-ray exposures of the incriminated breast
during the procedure. There is no estimate as yet to take these additional
cancerogenic factors into account. They might obliterate any mortality
benefit teased out by screening as such, or even increase mortality by a
larger margin than we fear.
Given the uncontrolled nature of the study and the lack of
comparability between the two groups, one should be cautious in
attributing causality to any of the observed phenomena. Even the word
"reduction" may be misleading as it implies a directed activity. This
study does not allow such a conclusion.
Friederike M. Perl DRCOG
Consultant Surgeon Breast Centre,
Diakonie Klinikum,
Rosenbergstr. 38,
D.70173 Stuttgart/Germany
Competing interests:
None declared
Competing interests: No competing interests
All women would like accurate and balanced information to help them
decide whether or not to participate in screening programs. So the
question is: does the information presented by Barratt et al in this
article actually provide accurate or balanced information? We have our
doubts about this for the following reasons:
This article is based on assumptions about the “harms” associated
with false positive results from screening. Although we acknowledge these
“harms”, we would argue that the “harms” associated with false negative
results are even greater. False negatives may lead to death. Most women
would be prepared to cope with the problems of anxiety etc, if they
understood that there was a trade-off with the possibility of death.
So the next issue is how accurate are these “estimates”. We are not
in a position to argue this at the present time, the statisticians will
need to get to work to be able to determine the answer to this question.
But be reminded that they are estimates.
And the “estimates” are just a little rubbery. If the range of
“estimates” for overdetection ranges from 2-30% just how much value can be
placed on this concept. Indeed we would argue that overdetection is not
the issue; the main problem is the overwhelming desire by clinicians to
treat women, that is, overtreatment is more of a problem than
overdetection.
Even if these “estimates” are accurate, will this information
actually assist women in making their decisions. We would argue that
these figures are only of partial value to women. Although women will
take these estimates into account, they also base decisions on factors
such as their own personal value systems, their fears of developing breast
cancer, and their personal circumstances. “Estimates” such as this play a
limited role in their decision making.
Competing interests:
None declared
Competing interests: No competing interests
Barratt et al. report that for every 1000 women screened from the age
of 50 over 10 years, 2 fewer will die from breast cancer and 13 more will
be diagnosed with breast cancer (corresponding to 63% more cancers, which
mostly constitute overdiagnosis) (1). Similarly, we found that for every
woman who has her life prolonged, 5 healthy women, who would not have
received a breast cancer diagnosis in their lifetime if there had not been
screening, will be converted into cancer patients unnecessarily (2).
However, it is important to inform women that the frequently used
estimate of 25-30% for the survival benefit is uncertain and rather
optimistic (2,3). Barratt et al. even used 37% since they adjusted their
estimate of 25% for non-compliance. This procedure is doubtful since, as
noted by two of the investigators in the screening trials, women who
refuse to be screened have a worse prognosis, presumably because some of
them are afraid of having a suspicion of breast cancer confirmed (4). They
also have a much higher death rate from all causes (4).
We disagree that one can equate 2 fewer women who will die from
breast cancer with 2 fewer who will die from any cause (1). An effect of
screening on all-cause mortality has not been demonstrated (2,3); breast
cancer mortality is an unreliable outcome that is biased in favour of
screening (4); and the extra treatment due to overdiagnosis would be
expected to lead to excess mortality in the screened group (5).
1. Barratt A, Howard K, Irwing L, Salkeld G, Houssami N. Model of
outcomes of screening mammography: information to support informed
choices. BMJ, doi:10.1136/bmj.38398.469479.8F (published 8 March 2005)
2. Olsen O, Gøtzsche PC. Systematic review of screening for breast
cancer with mammography. 2001.
http://image.thelancet.com/lancet/extra/fullreport.pdf (accessed 8 Jan
2004).
3. 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(5 Part 1):347-60.
4. Gøtzsche PC. On the benefits and harms of screening for breast
cancer. Int J Epidemiol 2004;33:56-64.
5. Early Breast Cancer Trialists' Collaborative Group. Favourable and
unfavourable effects on long-term survival of radiotherapy for early
breast cancer: An overview of the randomised trials. Lancet 2000;355:1757-
70.
Competing interests:
None declared
Competing interests: No competing interests
Dear Editor,
The results of the interesting model study by Barratt et al. [1]
suggest that more breast cancer are diagnosed among screened than
unscreened women. For women aged 50 who have five biennial mammography
screenings, 40% higher cumulative 10-year cumulative incidence of breast
cancer was observed (and 67% higher if ductal carcinoma in-situ (DCIS) was
included). For women screened from other ages, the model predicted a
similar relative increase over ten year.
Two important underlying assumptions in mammography screening are i)
the sensitivity at the screening is relatively high (>75%) and ii) most
invasive tumours grow monotonously; i.e. spontaneous tumour regression is
uncommon.
From these assumptions it emerges that most of the difference in the
cumulative incidences between the screened and the unscreened women should
disappear if a prevalence screening of the previously unscreened women had
been performed at the end of the 10-year period.
However, a prevalence screening can only compensate for a part of
this difference. The reason for this is that the detection rate at a
prevalence screening repeatedly has been shown to be only about 3 times
higher than the background incidence at age 60 or 70 [2].
Moreover, the model by Barratt et al. predicts that if 1000 women
aged 40 have 15 biennial mammography screenings from age 40 to 69, the
cumulative difference between screened and unscreened for the three 10-
year periods is (17.6 – 13.2) + (28.1 – 19.8) + (32.5 – 23.9) = 21.3
invasive breast cancers [1]. This calculation leads to the conclusion that
the model predicts the detection rate at a prevalence screening of women
aged 70 to be twice that observed in screening programmes [3, 4].
This result suggests that at least one of the two assumptions above
should be modified. The prevalence screening detects most of the slow-
growing cancers, and the detection rate at the following screenings is
stable at a lower level. This indicates that it is true that the
sensitivity at the screening is relatively high.
We conclude that the other assumption is false. It is not true that
spontaneous tumour regression is uncommon.
1. Barrratt A, Howard K, Irwig L, Salkeld G, Houssami N. Model of
outcomes of screening: information to support informed choices. Br Med J
doi:10.1136/bmj.38398.469479.8F.
2. Zahl PH, Andersen JM, Maehlen, J. Spontaneous regression of
cancerous tumours detected by mammography screening. JAMA 2004; 292: 2579-
80.
3. Tabár L, Fagerberg G, Duffy SW, Day NE, Gad A, Gröntoft O. Update
of the Swedish Two-County program of mammographic screening for breast
cancer. Radiol Clin North Am 1992; 30: 187-210.
4. Smith-Bindman R, Chu PW, Miglioretti Dl, Sickles EA, Blanks R, et
al. Comparison of screening mammography in the United States and the
United Kingdom. JAMA 2003; 290: 2129-38.
Competing interests:
None declared
Competing interests: No competing interests
Does screening mammography cause cancers?
The recent model of screening mammography by Alexandra Barratt (1) has some very interesting data. We could use their data to predict what might happen in the breast
of a woman from the age of 40 to 69 if not disturbed by the consequences of
mammographic screening. Let us wipe our slates clean and look at it
sans-prejudice.
The basis of screening is to
detect cancers before they became invasive or big enough to metastasise and
kill. So in theory, the total number of cancers should remain the same in the
screened and unscreened population, but the screened women would have smaller
ones. Or, perhaps the detection of ductal carcinoma in situ (DCIS) should
increase whilst as a corollary to that, invasive cancers should reduce in
incidence, if not in the first 10 years then at least over a 30 year period.
Looking at table 2 from the
original paper (1) and concentrating on the 3 columns from the ages of 40-69.
(The numbers for unscreened population between 70-79
are not available and hence are not included.)
To estimate the natural history of
breast cancer over 30 years, one just adds up the numbers to get Table 1, in
which one can compare what happens to unscreened breasts to those subjected to
two-yearly screening mammography
Table 1
Ages 40 to
69
Screened
Unscreened
Interval
29
29*
Invasive
49
28
DCIS
14
1
Total
92
58
Total diagnosed- excluding interval
cancers
63
29
*The
interval cancers appear in between the two-yearly mammograms so they are in
fact symptomatic cancers. Let us assume that they are not “caused” by the
screening process itself and ignore them at present.
Let us now interrogate the data with two obvious questions:
Does mammography “cause” cancers? And is mammography dangerous to life?
Thus we start off with 28 invasive
cancers and 1 DCIS that develops among 1000 women whose breasts are not
screened over 30 years. If these breasts are subjected to mammography, then
they are found to have 49 invasive cancers and 14 DCIS, over the same period.
That is 34 extra cancers. More than doubling of risk of
diagnosis of cancer. The unscreened women never seem to catch up with the
cancer diagnosis over 30 years, until they are 69.
Table 2 shows the same numbers for calculation of odds
ratios for testing alleged causality of invasive cancers, DCIS and total
cancers by mammography.
Table 2
Given Mammo-grams
Not Given Mammo-grams
Odds Ratio
95% Confidence intervals
No Invasive Cancer
951
972
Invasive cancer
49
28
1.79
(1.09 to 2.95)
p=0.015
No DCIS
986
999
DCIS
14
1
14.08
(1.96 to 289.75)
p=0.00076
No Cancer
937
971
All cancer
63
29
2.25
(1.41 to 3.62)
p=0.00029
Death
115.7
120.9
Remain Alive
884.3
879.1
1.05
(0.79 to 1.39)
p=0.72
So it appears that two-yearly mammography more than doubles
the risk of breast cancer diagnosis and increases the risk of DCIS diagnosis 14
times. Even if we include the interval cancers, the Odds ratio is 1.65 (95% CI
– 1.15 to 2.35, p=0.003). Fortunately, risk of remaining alive is not
statistically significantly altered (OR=1.05, 95% CI= 0.79-1.39, p=0.72).
So does mammographic screening
“cause” breast cancers? Yes, it either directlycauses cancers from ionising radiation or more likely
unearths cancers that would never surface in those 30 years if left to nature.
Is mammographic screening dangerous to life? Perhaps not,- there is no
statistically significant detrimental effect on life, suggesting that either
the cancers it induces or unearths are harmless or would have regressed anyway
(see Zahl et al above), or it detects other cancers
early enough to favourably influence their natural history, so negating the
harmful effect.
This unequivocal evidence from a
human experiment correlates well with
laboratory evidence that low dose radiation such as with mammography has
a carcinogenic effect in laboratory animals and cell cultures(2).
The other, more plausible
hypothesis (see Zahl et al above), to explain this
phenomenon is that the extra cancers are also present in unscreened women, but
they either don’t progress or regress spontaneously. However there is some
evidence to the contrary. A study of untreated DCIS cases in the Nurses health
study (3), in which 13 DCIS cases were wrongly labelled as benign. Six of these
developed invasive cancers and 4 more were diagnosed as DCIS suggesting that
probably half of DCIS progresses to invasive cancer. In the classic
study by David Page and colleagues(4), 28 women with
DCIS were treated with biopsy only. Eleven of these women developed invasive
carcinoma in a 30 year follow up. Applying this to the table 1, Of the 28
cancers in the unscreened group, 6 would have resulted from the 13 cases of
undetected DCIS. This gives us an estimate that 22of the 49 cancers detected by screening become
clinically apparent- a proportion similar to DCIS! (NB Many of these women
being screened and having biopsies and this may confound the issue).
The proportion of screen detected invasive cancers that may have
progressed to clinically evident cancers is not known. Let us now estimate how
many cancers would need to be diagnosed during screening to account for these 22
cases, for different probabilities of progression to clinical cancers. Then the
number of cancers caused by screening can be calculated by subtracting this
number from the 49 cancers that are actually detected by screening. This is
shown in Table 3
Table 3 shows the results of 4 possible proportions of
invasive cancers detected at screening that may progress to clinical cancers
over 30 years. Please read this table, column by column
Percent
of screen detected invasive cancers that we believe would have naturally
progressed to clinical cancer
45%
(similar to DCIS)
60%
80%
100%
Number of
cancers needed to be diagnosed during screening to account for the 22
cases diagnosed in the unscreened population
49
36
27
22
Number of
cancers really caused by mammographic screening per 1000 women over 30 years
0
13
22
27
So one has to either believe that more than half of invasive
cancers detected by screening would never have become clinically evident, at
least until the age of 69, (column 1 of table 3), or, if a larger proportion (60%, 80% or
100%) of these invasive cancers progress to clinical tumours, then screening mammography actually causes extra
cancers(13, 22 or 27, respectively), as in columns 2, 3 or 4 of Table 3.
Neither of these hypotheses is easily testable. Women should
be made aware that mammographic screening would double the chance of their
being diagnosed with cancer, although it is unlikely that this will adversely
affect their chance of remaining alive at the age of 69. The relevance of this may be more sinister in
women with family history of breast cancer because tissues of women with BRCA1
or BRCA 2 gene mutations are even more sensitive to low dose of radiation (5).
Yours sincerely,
Jayant S Vaidya
References
1. Alexandra Barratt, Kirsten
Howard, Les Irwig, Glenn Salkeld,
NehmatHoussami
Model of outcomes of screening mammography: information to
support informed choices 2005; 330 (7497): 936.
2. Frankenberg D,
Kelnhofer K, Bar K, Frankenberg-Schwager M.
Enhanced neoplastic transformation
by mammography X rays relative to 200 kVp X rays:
indication for a strong dependence on photon energy of the RBE(M)
for various end points. Radiat Res. 2002 Jan;157(1):99-105.
3. Collins LC, Tamimi RM, Baer HJ,
Connolly JL, Colditz GA, Schnitt
SJ Cancer. Outcome of patients with ductal carcinoma in situ untreated after
diagnostic biopsy.
2005 May 1;103(9):1778-84.
4. Sanders ME, Schuyler PA, Dupont
WD, Page DL The natural history of low-grade ductal carcinoma in situ of the
breast in women treated by biopsy only revealed over 30 years of long-term
follow-up. Cancer. 2005 May 9; [Epub ahead of print]
5. Vaidya JS, Baum M. Benefits and risks
of screening mammography in women with BRCA1 and BRCA2 mutations.JAMA. 1997 Jul 23-30;278(4):290
Competing interests:
None declared
Competing interests: The recent model of screening mammography by Alexandra