Sir--I am concerned about some features of Ole Olsen and
Peter Gøtzsche's review (Oct 20, p 1340).
1 I refer
especially to their comments on systematic reviews of causes of
death, as applied by the investigators of the Health Insurance Plan
(HIP) trial,
2 and later adapted to the Canadian National
Breast Screening study (CNBSS)
3 and other trials.
4
I was one of the death reviewers for the HIP trial, initiated
the death review in the CNBSS, and I chair the Death Review
Committee for the Prostate Lung Colon and Ovary trial
(PLCO).
4
The HIP trial was done in an era when the size of breast
cancers was much larger than became usual in the subsequent two
decades in North America. I recall no instance in which the masking
of the allocation was compromised. The major difficulty for the
reviewers was not whether the patient died of cancer, nor whether
breast cancer had been diagnosed, but whether breast cancer was the
cause of death when the patient had been diagnosed with another
malignant disease. The decisions made were entirely masked, even
after consultation between the reviewers for initial disagreement.
Lumpectomy was not practised, although radiotherapy was used
frequently, especially for locally advanced disease. However, if an
unrecognised consequence of radiotherapy was the cause of death, and
was labelled as cardiovascular disease, such labelling would have
been applied without bias as to treatment assignment.
Olsen and Gøtzsche comment that equal numbers of patients, by
allocation, were reviewed, but that the numbers assigned to breast
cancer as a cause of death were fewer in the screening group, and
they cite this as evidence of bias. An alternative explanation is
more plausible. If screening was truly effective, then the numbers
of deaths from breast cancer in a reviewed series would be less in
the screened group. It seems not to have occurred to Olsen and
Gøtzsche that the equal numbers reviewed could have resulted from a
deliberate attempt to ensure that equivalent numbers of patients in
the two groups were reviewed at the same time to avoid bias.
I am also concerned by Olsen and Gøtzsche's comments on the
difference in numbers of women with breast cancer excluded from the
two groups. This difference arose because, when women in the
screened group attended for screening, previously diagnosed breast
cancers were identified, but this was not possible for the controls.
The 18-year follow-up, however, enabled all deaths from breast
cancer to be identified in the two groups; identification of the
date of diagnosis was then possible from hospital records. Patients
diagnosed before randomisation were excluded.
2 Although
this process did not eliminate the inequality in the numbers of
patients with previously diagnosed breast cancers still living at 18
years who were not excluded, the small difference in person-years of
observation is unlikely to have biased the results.
Therefore I believe the HIP trial results are valid, and the
trial should not be dismissed as flawed.
Anthony B Miller
Division of Clinical Epidemiology, Deutsches
Krebsforschungszentrum, D-69120 Heidelberg, Germany (e-mail:a.miller@dkfz-heidelberg.de)
1 Olsen O, Gøtzsche PC. Cochrane review on screening for
breast cancer with mammography.
Lancet 2001;
358: 1340-42. [
Text]
2 Shapiro S, Venet W, Strax P, Venet L. Periodic screening
for breast cancer: the Health Insurance Plan Project, 1963-1986, and
its sequelae. Baltimore: Johns Hopkins University Press, 1988.
3 Miller AB, Baines CJ, To T, Wall C. Canadian national
breast screening study, 1: breast cancer detection and death rates
among women age 40-49 years.
Can Med Assoc J
1992;
147: 1459-76. [
PubMed]
4 Miller AB, Yurgalevitch S, Weissfeld JL. Death review
process in the Prostate, Lung, Colorectal and Ovarian (PLCO) cancer
screening trials.
Control Clin
Trials 2000;
21: 400S-06S. [
PubMed]
Sir--Ole Olsen and Peter Gøtzsche's review
1
concurs with the findings of their earlier report
2 that
screening mammography programmes do not reduce breast-cancer
mortality and lead to more aggressive treatment. In his Oct 20
Commentary, Richard Horton
3 agrees with their conclusion
that screening mammography programmes are not therefore justified.
These conclusions could have a major effect on women's health care
and deserve careful scrutiny.
Our primary criticism is the notion that a screening
programme can be justified only if it reduces mortality. Mortality
is a very important outcome, and easily measured, and choosing an
outcome parameter for ease of measurement is a common bias among
clinical investigators. However, outcomes such as quality of life
are also important to patients and could justify screening
programmes. Unfortunately, such outcomes have been understudied in
the assessment of breast-cancer screening. Rather than making
assumptions about whether false-positive tests and earlier detection
of breast cancer improves or harms quality of life, objective
measurement is needed of how mammography screening programmes affect
quality of life, and whether certain types of patients will benefit
more than others.
Olsen and Gøtzsche conclude that screening mammography leads
to overly aggressive treatment. If that is true, it is a
condemnation of treatment practices, not of screening. Our
organisation is one of many encouraging better education of patients
and physicians to lower the prevalence of unnecessarily aggressive
treatment. The best solution to improving treatment is better
technology, education, and training, along with improved
communication between physicians and patients. These strategies
deserve more attention from the medical community before a decision
is made to eliminate screening programmes that lead to earlier
diagnoses. There is no need to throw out the baby with the bath
water.
We are also concerned with the meta-analysis itself. In six
of seven randomised trials (including the extended Malmö
trial
4) analysed by Olsen and Gøtzsche, a reduction in
breast-cancer mortality was associated with the use of screening
mammography. We recommend caution before rejecting these findings,
which incorporated data from hundreds of thousands of patients, on
the basis of one meta-analysis.
Meta-analysis requires many subjective decisions on the
selection and weighting of variables. Each decision can introduce
bias, no matter how systematic the decision-making process. A second
set of investigators carefully analysing the same data might reach
very different conclusions. For example, many investigators did not
agree with Olsen and Gøtzsche's decision to deem certain clinical
trials unreliable on the basis of small baseline differences in mean
age between screening and control groups.
5
Meta-analyses are an evolutionary improvement over the
traditional review article, but their limitations must be
acknowledged before they are used to formulate major health-policy
decisions. The attention garnered by Olsen and Gøtzsche illustrates
the need to assess each meta-analysis thoroughly, as well as to
reassess the proper role of meta-analyses.
*Jae Hong Lee, Diana Zuckerman
National Center for Policy Research for Women and Families, 1444
Eye Street NW, Suite 900, Washington, DC 20005, USA (e-mail:jl@center4policy.org)
1 Olsen O, Gøtzsche PC. Cochrane review on screening for
breast cancer with mammography.
Lancet 2001;
358: 1340-42. [
Text]
2 Gøtzsche PC, Olsen O. Is screening for breast cancer with
mammography justifiable?
Lancet 2000;
355: 129-34. [
Text]
3 Horton R. Screening mammography--an overview revisited.
Lancet 2001;
358: 1284-85. [
Text]
4 Andersson I, Janzon L. Reduced breast cancer mortality in
women under age 50: updated results from the Malmo Mammographic
Screening Program.
J Natl Cancer Inst
Monogr 1997;
22: 63-67. [
PubMed]
5 Hayes C, Fitzpatrick P, Daly L, Buttimer J. Screening
mammography re-evaluated.
Lancet 2000;
355: 749.
Sir--In their overview of breast-cancer screening, Ole Olsen
and Peter Gøtzsche
1 claim that the main reason for the
continuing controversy about their analysis of imbalances in
baseline variables is that their opponents keep referring to the
criticisms of their paper without referring to their reply. On the
contrary, their replies have contained nothing worth referring
to.
2,3
Olsen and Gøtzsche do not address the original criticism of
their study, which is the real problem. This inaction is a shame,
because some other points they make are reasonable and, in view of
the uncritical acceptance of the value of screening by politicians,
the case against is worth making.
We have previously pointed out to Olsen and Gøtzsche that
cluster-randomised trials will be expected to show greater
variability in balance at baseline than trials randomised at patient
level.
4 This variability is not a problem provided that
such trials are analysed properly. The Cochrane Collaboration's
favourite tool, RevMan, can analyse only single-centre
parallel-group trials with no covariates, and, therefore, their
researchers are apt to see any other sort of trial as problematic.
This problem should not be inflicted on a research community that is
not limited to primitive analysis.
I believe Olsen and Gøtzsche's previous criticisms about
randomisation have been irrelevant and unfair. For example, the
Edinburgh randomised trial of breast cancer
5 allowed for
cluster randomisation, which RevMan cannot do. Two of that study's
coauthors were, however, statisticians well known for their
expertise in mixed-effect models. Whatever other features of Olsen
and Gøtzsche's criticism are valid, the way they have used baseline
tests of covariate balance on cluster-randomised trials is not.
Stephen Senn
Department of Statistical Science and Department of Epidemiology
and Public Health, University College London, London WC1 6BT, UK
(e-mail:stephens@public-health.ucl.ac.uk)
1 Olsen O, Gøtzsche PC. Cochrane review on screening for
breast cancer with mammography.
Lancet 2001;
358: 1340-42. [
Text]
2 Gøtzsche PC, Olsen O. Screening mammography re-evaluated.
Lancet 2000;
355: 752.
3 Gøtzsche PC, Olsen O. More on mammography.
Lancet
2000;
356: 1276.
4 Cates C, Senn S. Screening mammography re-evaluated.
Lancet 2000;
355: 750.
5 Alexander FE, Anderson TJ, Brown HK, et al. 14 years of
follow-up from the Edinburgh randomised trial of breast-cancer
screening.
Lancet 1999;
353: 1903-08. [
Text]
Sir--Predictably, Ole Olsen and Peter Gøtzsche's
conclusion
1 that mammographic screening is of uncertain
benefit and leads to greater use of more aggressive treatment and
other harms, has provoked adverse criticism from many who believe
that screening saves lives. Richard Horton, in his
Commentary,
2 and the alternative version of the report on
the journal's internet site
3 have exposed disagreement
among the editorial ranks of the people responsible for producing
Cochrane systematic reviews of evidence for breast cancer: this is a
problem of serious import.
The new elements in this latest review that confirm and
strengthen Olsen and Gøtzsche's previous findings, rest on their
statement that breast cancer mortality is a misleading outcome
measure. It is occasioned by their findings that mammographic
screening leads to greater use of more aggressive treatment,
increases the number of mastectomies by about 20%, and inflicts
important psychological distress on more than 10% of the healthy
population who attend screening programmes, evidenced in a cited
systematic review.
These arguments have been in the public domain for some
time,
4 but are largely dismissed as irrelevant by workers
zealously pursuing the goal of reduction of mortality from breast
cancer, without regard for the risks, limitations, and consequences,
to women invited for screening but not informed about the risk of
the procedure or of receiving a diagnosis of ductal carcinoma in
situ.
Olsen and Gøtzsche point out that mass screening involves the
entire population and may cause a screening-associated increase in
mortality, as well as other harms.
Attempts by the Cochrane Breast Cancer Review Group to
suppress these features, by insistence on changes to the review as a
condition of its publication in the
Cochrane Library, are a
blatant threat to the over-riding principle of freedom of appointed
investigators to place all legitimate findings in the public domain
without judgmental interference.
Such threats to integrity must not be allowed to go
unchallenged. Attempts to muzzle unpopular findings, if successful,
would undermine trust in research processes and methods to the harm
of the general welfare of society and progress in health care. The
gate-keeping function of medical journalism has again been put to
the test.
5 Preservation of freedom of speech is essential
so that careful reflection and proper discussion may follow, because
the implications for women and policy makers are
substantial.
2
Hazel Thornton
31 Regent Street, Rowhedge, Colchester, Essex CO5 7EA, UK
(e-mail:hazelcagct@aol.com)
1 Olsen O, Gøtzsche PC. Cochrane review on screening for
breast cancer with mammography.
Lancet 2001;
358: 1340-42. [
Text]
2 Horton R. Screening mammography--an overview revisited.
Lancet 2001;
358: 1284-85. [
Text]
3 Olsen O, Gøtzsche PC. Cochrane review on screening for
breast cancer with mammography.
http://image.thelancet.com/lancet/extra/fullreport.pdf
(accessed on Nov 21, 2001)
4 Thornton H. Randomised clinical trials: the patient's point
of view. In: Silverstein M, ed. Ductal carcinoma in situ of the
breast. Philadelphia: Lippincott, Williams and Wilkins, 1997.
5 Davidoff F, DeeAnglis CD, Jaffrey M, et al. Sponsorship,
authorship, and accountability.
Lancet 2001;
358: 854-56. [
Text]
Sir--Ole Olsen and Peter Gøtzsche's full report on
www.thelancet.com
1 is riddled with misrepresentation,
inconsistency in the treatment of the randomised trials, and errors
of method and fact, some examples of which follow.
They cite inconsistencies of reporting for the Two-County
Trial. However, the individual trials and the Swedish overview
inevitably differ, since the latter reclassified causes of death and
redefined the study groups. This independent review of trial results
is exactly the sort called for by Olsen and Gøtzsche and Richard
Horton in his accompanying Commentary,
2 but was
overlooked. Other alleged inconsistencies cited by Olsen and
Gøtzsche are due to the numbers of deaths increasing with longer
follow-up; competent meta-analysts would be aware of such change and
differences between the primary studies and secondary researchers'
reports.
3 The Two-County Trial is further misrepresented
as two separate trials and as having failed to publish vital
information. It was, in fact, one trial. We have made public more
material on its design, conduct, and results than any of the other
trials.
In their attempts to justify the baseline imbalance in
palpable node-positive tumours in the Canadian National Breast
Screening Study,
4 Olsen and Gøtzsche note that study
group patients were generally treated in centres with more thorough
axillary-node dissection, but did not deem this difference a source
of bias. However, they wrongly cite this treatment difference as a
potential source in the other trials. In the Two-County study,
standard treatment according to stage of disease was applied to
study and control groups.
5
Olsen and Gøtzsche quote a significant excess of excluded
previous cancers in the people aged 60-69 years in the study group
of the Kopparberg part of the Two-County Trial, but do not seem to
have adjusted their significance test for the multiple testing that
they must have done to search for such chance heterogeneities.
The above do not inspire confidence in Olsen and Gøtzsche's
judgment of study quality or for the notion that all-cause mortality
is an appropriate endpoint. As if deaths from road-traffic accidents
or hip fractures were in some way indicative of the effect of
breast-cancer screening. We believe that Olsen and Gøtzsche
referring readers to a website for details of why they were right
all along, is a clever and convenient way of subverting the peer
review process and shielding their work from proper scientific
scrutiny. The full report is not the official Cochrane review,
despite its misleading title. Olsen and Gotzsche's review is of poor
quality, and, therefore, unreliable. Given this poor quality,
Horton's uncritical commentary is at best ill considered. Nothing in
the report or commentary disproves the finding of millions of
person-years of experimental research, and dozens of previous expert
reviews, that mammographic screening significantly reduces mortality
from breast cancer.
*Stephen W Duffy, Laszlo Tabar, Robert A Smith
*Imperial Cancer Research Fund, PO Box 123, London WC2A 3PX, UK;
Central Hospital, Falun, Sweden; and American Cancer Society,
Atlanta, GA, USA
1 Olsen O, Gøtzsche PC. Cochrane review on screening for
breast cancer with mammography.
http://image.thelancet.com/lancet/extra/fullreport.pdf
(accessed on Nov 21, 2001).
2 Nyström L, Larsson LG. Breast cancer screening with
mammography.
Lancet
1993;
341: 1531-32. [
PubMed]
3 Horton R. Screening mammography--an overview revisited.
Lancet
2001;
358: 1284-85. [
Text]
4 Miller AB, To T, Baines CJ, Wall C. The Canadian National
Breast Screening Study: update on breast cancer mortality.
Monogr Natl Cancer
Inst 1997;
22: 51-53. [
PubMed]
5 Holmberg LH, Tabar L, Adami HO, Bergström R. Survival in
breast cancer diagnosed between mammographic screening examinations.
Lancet 1986;
2: 27-30. [
PubMed]
Sir--I wonder why reports such as that of Ole Olsen and Peter
Gøtzsche
1 raise such a lot of angry criticism. It would
be rather too pessimistic and cynical to believe that the criticism
stems only from self-interest of radiologists, surgeons, managers,
and so on, whose daily bread depends on the continuation of
mammographic screening programmes. I think that such reports shake
the very basis of intuitive logic and are, therefore, not palatable.
But isn't the world frequently counterintuitive?
If we believe Olsen and Gøtzsche, failure of screening to
save lives is a sad fact that we might need to accept, and a
plausible explanation may make it easier. Absence of benefit from
screening may actually point to a very fundamental hidden biological
clue. Judah Folkman
2 has elegantly shown that tumours
cannot grow beyond 200 µm without stimulating their own blood
supply. Thus, many primary tumours can exist in a state of dynamic
equilibrium and suppress the growth of their own secondaries by
antiangiogenic paracrine secretions.
3 Removal of the
primary can provoke local angiogenesis from surgical trauma and
release the secondaries from this inhibition and stimulate their
growth.
The paracrine secretion may be proportional to the tumour
size until a critical level, when it can no longer continue to
inhibit the secondary growth effectively. This critical size may be
different for individual tumours and could well be higher than the
clinical threshold.
There is, therefore, no meaning to the term early cancer. We
must realise that by age 60 years, every one of us harbours at least
one of prostate, breast, thyroid, or lung cancer in a subclinical
(dormant) state, which is generally harmless.
3-5
Effective treatment of cancer and reduction in mortality from
cancer in general will be possible only if we follow a strategy
based on these principles. Once a primary cancer can no longer
effectively inhibit its secondaries, it could be safely removed and
its antiangiogenic effect replaced and supplemented so that the
secondaries remain suppressed. Removal of a primary cancer at an
early stage when it is effectively suppressing its secondaries will
only remove the inhibition of angiogensis that will nulify any
benefit from reduction of further metastasis. This mechanism could
be the elusive answer to the puzzle of the disappointments of
screening. We should stop talking about early cancer and start
thinking about dormant cancer and inappropriate early surgery.
Perhaps the only way to cure some cancers is to learn ways to live
comfortably with them.
Jayant S Vaidya
Department of Surgery, University College London, London W1W 7EJ,
UK
1 Olsen O, Gøtzsche PC. Cochrane review on screening for
breast cancer with mammography.
Lancet 2001;
358: 1340-42. [
Text]
2 Folkman J. Angiogenesis in cancer, vascular, rheumatoid and
other disease.
Nat
Med 1995;
1: 27-31. [
PubMed]
3 Whitmore WF Jr. The natural history of prostate cancer.
Cancer 1973;
32: 1104-12. [
PubMed]
4 Nielsen M, Thomsen JL, Primdahl S, Dyreborg U, Andersen JA.
Breast cancer and atypia among young and middle-aged women: a study
of 110 medicolegal autopsies.
Br J
Cancer 1987;
56: 814-19. [
PubMed]
5 Vaidya JS, Baum M. Low-dose spiral computed tomography for
lung-cancer screening.
Lancet 1998;
352: 236.
Sir--We have followed with interest the debate provoked by
Ole Olsen and Peter Gøtzsche's work
1 on the usefulness of
breast-screening programmes. Clearly, one of the most potent issues
raised by their findings is what patients should be told about
breast screening.
In an era of emphasis on partnership with patients and
informed consent, underscored by the Kennedy principles,
2
we would expect that the current uncertainty about the risks and
benefits of participation would be fully explained to women.
The UK National Breast Screening Programme has issued a new
version of the leaflet given to women invited for screening.
Entitled
Breast screening: the facts, the leaflet, according
to the National Health Service Cancer Screening Programme
website,
3 purports to ensure that women are told what
screening can and cannot achieve so that they can "make a genuinely
informed choice about why they are attending for screening".
Social science teaches us that multiple readings are
available for any text, and it is perhaps possible that one reading
of this leaflet would indeed fully inform a decision about taking
part in screening. Our interpretation, however, is as follows. The
leaflet introduces what breast screening is with no hint of the
disadvantages. It treats women's need for screening as self-evident.
It draws on a persuasive discourse of "catch it early" without
providing a full account of the possible risks of attempting early
detection.
The leaflet suggests that the breast-screening programme has
not been extended to women younger than 50 years because "it has not
been proven to reduce the number of deaths from breast cancer", but
makes no mention of the controversy about whether screening reduces
deaths even in the ages currently screened.
The leaflet does not describe the psychological distress
associated with false-positive tests, or the consequences of
findings such as ductal carcinoma in situ, for which the best
treatment is still unknown. The leaflet states that "Around half the
cancers that are found at screening are still small enough to be
removed from the breast. This means that the whole breast does not
have to be removed"; it does not say that in clinically detected
cancers, the breast conservation rate is even higher, at about 75%.
In fact, one of the findings of Olsen and Gøtzsche's review is a
paradoxical increase in mastectomy rates for screen-detected
disease. No explanation is made of the possible consequences for
personal insurance, including insurance of participants'
children.
4 Finally, the leaflet gives no indication of
the overall chance of benefit.
We suggest that this view of the benefits of breast screening
is partial and potentially misleading, and that it represents a
prioritising of political and professional agendas over those of
patients.
5 We believe it is time we had more honest and
responsible partnerships with patients supported by resources that
truly reflect current evidence.
*Mary Dixon-Woods, Michael Baum, Jennifer J Kurinczuk
*Department of Epidemiology and Public Health, University of
Leicester, Leicester LE1 6TP, UK; and Cancer Research
Campaign/University College London Cancer Trials Centre, London
(e-mail:md11@le.ac.uk)
1 Olsen O, Gøtzsche P. Cochrane review on screening for
breast cancer with mammography
Lancet 2001;
358: 1340-42.
[
Text]
2 The inquiry into the management of care of children
receiving complex heart surgery at the Bristol Royal Infirmary.
http://www.bristol-inquiry.org.uk/
(accessed Nov 21, 2001)
3 Breast screening: an informed choice.
http://www.cancerscreening.nhs.uk/breastscreen/publications/ia-02.html
(accessed Nov 21, 2001).
4 Thornton H. Consequences of breast screening.
Lancet. 2000;
356: 1033.
5 Dixon-Woods M. Writing wrongs? An analysis of published
discourses about the use of patient information leaflets.
Soc Sci
Med 2001;
52: 1417-32. [
PubMed]
Sir--Ole Olson and Peter Gøtzsche
1 confirm their
previous conclusions that there is no reliable evidence that
screening for breast cancer reduces mortality.
2
Undoubtedly, this paper will raise much more heated debate.
A few months ago, we tried to find out the effect of their
first publication on clinical practice. We sent a questionnaire to
all Belgian gynaecologists and to a randomly selected sample of
gynaecologists in four other countries (in one mailing with prepaid
envelopes for responses to guarantee anonymity). We asked physicians
about their screening attitudes and about whether they had modified
their attitudes since the publication of Olsen and Gøtzsche's
report.
In some countries (UK, Denmark, and Sweden) gynaecologists
are not responsible for calling patients for mammography screening.
The survey results are presented in the table. The proportion of
physicians who recommend systematic screening to women was very high
except in Denmark. Moreover, more than 90% of the physicians who
answered the survey had not changed their attitude, whether or not
they had read or heard about Olsen and Gøtzsche's first publication.
|
Denmark |
Sweden |
UK |
France |
Belgium |
|
(n=264) |
(n=383) |
(n=500) |
(n=593) |
(n=1150) |
| Response rate (%) |
31 |
36 |
18 |
51 |
44 |
| Proportion physicians recommending |
37 |
95 |
87 |
98 |
95 |
| mammography |
| Proportion physicians not changed |
90 |
98 |
93 |
99 |
98 |
| screening attitude (%) |
| Rate of recommendation of mammography
screening after age 50 years and change in attitude to
screening after reading Olsen and Gøtzsche's
report |
Do our results mean that physicians don't care about their
patients? We don't think so. We purport that, unless harm from
screening, in terms of mortality is proven, physicians may prefer to
screen their individual patients, even in the absence of proof of
reduced mortality.
*S Rozenberg, H Ham, F Liebens
Departments of *Obstetrics and Gynaecology, and Nuclear Medicine,
Hospital St Pierre, Free University of Brussels, 1000 Brussels,
Belgium
1 Olsen O, Gøtzsche PC. Cochrane review on screening for
breast cancer with mammography
Lancet 2001;
358: 1340-42. [
Text]
2 Gøtzsche PC, Olsen O. Is screening for breast cancer with
mammography justifiable?
Lancet 2000;
355: 129-33. [
Text]
Author's reply
Sir--Breast-cancer mortality is unreliable and biased in
favour of screening. Masked endpoint committees are not the solution
to this issue. First, they cannot include cardiac deaths due to
radiotherapy. Second, to work in an unbiased way, death certificates
and other important documents must have been completed and patients
and documents selected for review without awareness of status, and
it should not be possible to break the masking during any of these
processes, including review of causes of death. These three
assumptions seem close to impossible to fulfil. The decrease in
breast-cancer mortality with screening in the Two-County study when
the endpoint committee did not know status was similar to that when
cause of death was assessed openly (and where we found bias in the
classification process). Therefore, our findings that masked
endpoint committees make biased assessments are supported.
Anthony Miller is not correct that a deliberate attempt was
made in the New York study to select for masked review an equal
number of cases and controls; they were selected according to
prespecified criteria. Miller writes that the major difficulty was
to assess cause of death when the patients had another cancer in
addition to breast cancer. We noted that all-cancer mortality was
similar in the groups invited to screening and in the control groups
in the trials we reviewed (relative risk 1·02 [95% CI 0·95-1·10] for
the two best trials and 1·00 [0·91-1·10] for the Two-County trial).
If screening reduced breast cancer mortality by 30%, as the Swedish
trialists have claimed, relative risk for all-cancer mortality
should be only 0·95 (assuming that breast-cancer mortality
constitutes 18% of all-cancer mortality in women older than 40
years, as in Denmark).
Jae Hong Lee and Diana Zuckerman claim that overly aggressive
treatment is a condemnation of treatment practices, not of
screening. Overtreatment cannot be avoided, however. In the UK, for
example, 29% of the women who had carcinoma in situ were treated by
mastectomy compared with 28% of those with invasive cancer in
1999-2000.
1 The surgeons were blamed for not having
treated even more women with carcinoma in situ by mastectomy. Thus,
increased quality assurance would only increase this harm (carcinoma
in situ is a frequent diagnosis among screened women that is rarely
detected without screening, but it develops into invasive cancer in
only a minority of cases).
Stephen Duffy and others had the chance of answering the
crucial questions we raised in our review, such as whether the
Kopparberg study was randomised, and if so, how. They did not.
Laszlo Tabar tries to blame his co-workers for the inconsistencies
we identified, but he has been a researcher on several of the
inconsistent papers, for example those that reported fewer deaths
from breast cancer with increasing follow-up than those identified
initially by the masked endpoint committee. We have also noted that
Tabar's main report
2 is inconsistent with the trial
protocol and a later thesis. This finding reinforces our conclusion
that the published data are of poor quality and very probably
flawed. Contrary to their assertion, the data we present on
The
Lancet website have been extensively peer reviewed.
Miller and Stephen Senn argue that we have treated the New
York and Edinburgh studies unfairly. We disagree, and note how these
studies are classified is not important for our results: there is no
evidence that screening decreases mortality. Contrary to what Senn
asserts, our previous criticisms of the cluster randomised trials is
not relevant for this finding. We furthermore doubt that
retrospective exclusion of women after 18 years of follow-up, as in
the New York study, is reliable.
In the absence of good arguments, some commentators try to
persuade by big numbers--hundreds of thousands of patients, millions
of person-years, and dozens of previous expert reviews. By contrast,
Richard Horton notes in his Commentary that our results require
careful reflection and discussion. We look forward to this
discussion. And as long as we have not been proved wrong, our
conclusions stand: there is no evidence of benefit, but evidence of
harm.
Peter C Gøtzsche
Nordic Cochrane Centre, Rigshospitalet Department 7112, 2100
Copenhagen, Denmark (e-mail:pcg@cochrane.dk)
1 NHS cancer screening programmes.
http://www.cancerscreening.nhs.uk/breastscreen/publications.html
(accessed Dec 12, 2001).
2 Tabar L, Fagerberg CJ, Gad A, et al. Reduction in mortality
from breast cancer after mass screening with mammography: randomised
trial from the Breast Cancer Screening Working Group of the Swedish
National Board of Health and Welfare.
Lancet
1985;
1: 829-32. [
PubMed]
