Some Remarks and
Deliberations about Methodology, animated by the
cohort study “Breast cancer mortality in Copenhagen after
introduction of mammography screening” (BMJ 2005)
by Karl-Heinz Bayer M.D.
february in 2005
When you face mammography-screening without bias it becomes
clear that its methodology cannot solely be used to reduce
breast cancer mortality just as the EKG is not able to
solely reduce heart mortality rates. The question at issue
is whether one can detect more cancers and whether women
really derive a benefit from mammography screening more than
from regular mammograms or other examinations.
The possible screening benefit is highly overrated in
interested political and industrial circles, especially in
Germany, where the introduction of the national disease
management for breast cancer was justified by a non-proven
allegation of great numbers of wrong diagnoses in
mammography and ultrasounds.
The complexity of the problem is addressed by an article
about breast cancer prevention in Copenhagen 1)
and a very critical look into this article in the Ärzte
Zeitung of January 27th of 2005 2).
The rate of false positive diagnoses of breast cancers in
mammography screenings, according to all known studies, is
approximately 10 %. Unfortunately, this means that an equal
estimate of 10% false negatives also has to be granted
(allowed). On one hand, mammography screening creates
unjustified anxiesties in every tenth women examined, and on
the other hand, every tenth women with cancer is
unjustifiedly calmed and sent home without proper treatment.
As with all methods concerning relatively rare diseases –
breast cancer affects 0.4 % of women between the ages of 50
to 70 years of age – a statistically relevant reduction of
death rates is only possible whene there is a virtual 100 %
participation in screening. Every method concerning with
rare diseases reach their furthermost benefit statistically
only if the participation is next to 100. Voluntary
participation by 100 % of the women themselves is neither to
be expected, nor is it possible to be approved by political
These kinds of preventative measures would exceed 170
million Euro each year in Germany, following the
Ärztezeitung. Moreover, in any case, if a government was
willing to extend such funding and if all the women were to
participate in screening, the mortality rate would decrease
by 25 % at the most. Even the greatest possible use of
mammography could not prevent 75 % of those with breast
cancer dying from the disease. In absolute numbers, in
Copenhagen, 0.4 % of the 40,000 women examined represent
about 160 cases of newly detectet cancers. 40 of these women
could be saved, an average of 4 women each year. But 120
women or 12 each year would not be saved, even though they
were correctly diagnosed. And what is more relevant, 39,840
women needed to be screened to detect 4 additional cases of
breast cancer each year. The pros and contras of radiation
induced damage shall not even be mentioned here.
These considerations imply that based on numbers alone 25 %
of cancer mortality could be avoided through the use of
massive mammography screening alone. The past studies, most
from Scandinavian countries 3-9), are able
to create a doubt in this assumption. Maybe individual
behavior and regional anomalies are statistically as
important or even more important than participation in a
screening program 10,11). All further
conclusions drawn from the Copenhagen study include
unproven, that the follow-up cancer-treatment was successful
and that it was statistically equivalent in all cases. No
consideration was given to the idea that differing rates of
cancer mortality in different parts of the country might be
the result of superior or inferior treatment methods or
institutions, as well as where treatment takes place. Maybe
there also is a bias in the fundamental and individual
agreement or disagreement towards different treatments and
institutions. None of those parameters are discussed as part
of the reason for a possible and assumed mortality
There is evidence of biased calculation in the Copenhagen
study. The title “...after introducing mammography
screening” itself uncovers that an assumption was the
foundation of the reasoning.
The authors elaborated in their subsection "statistical
analysis" that it was impossible to adjust the data for
items such as local differences and time trends from the
effects of mammography as determinants for the reduction of
breast cancer rates. Therefore they felt compelled to use a
correction factor, expecting it to eliminate bias errors
arithmatically. This approach is not without pitfalls as it
is an attempt to calculate bias which cannot be
It is without doubt wrong and unscientific to describe the
risk reduction in fractions of percents and call it the "reduction
of the relative risk" 12). The authors
of the Copenhagen study stress that the numeral shift is
within the range of the statistical error, which is pitched
at the mark of 5 %, and they declare that the numerals are
also spread over a large range. But they seem to forget that
the range of errors does not decrease and change its
validity if only a so-called relative risk is determined. A
relative risk reduction of 25 % is basically (in reality)
only a risk reduction of 0.1 %. Undoubtedly 25 % sound
much better than 0.1 %, , but the risk of including errors
and biases exceeds the true and mathematically proven
benefit by about fifty times. These kinds of calculations
belong to the tricks of prestidigitators 13).
The Copenhagen study amplifies many biases and errors. One
of its chief problems is how the samples are selected. As a
whole, every third women was excluded from the study. The
result of having no information about a third of the sample
is that the likelihood of having more errors embedded in
the data increases. The results of the study being allowed
to be transferred to all women or at least to the whole
female Danish population can therefore not be supported. One
consequence may have been to exclude women in whom breast
cancer had already been detected, because it would be
obvious to screen them. Nonetheless, the number of these
women increases the share of those who certainly have no
benefit from mammography screening. So, in sum total, three
quarters of the women with cancers, although correctly
detected by the mammography screening, plus at least 10 % of
all false positive and false negative diagnosis, plus every
non-participant and plus every pre-diagnosed breast cancer
are comprising the group of the women not benefiting. Also
added to this group is the unidentified grey area of all
Open to critique also is the manner of building the cohorts.
It makes the study artificial.
The cohorts are not really chosen at random. Copenhagen is
a) urban and b) as Denmark´s capital has a different than
the rural samples. The authors of the Copenhagen study were
confronted by those differences, which they described as
local and time trend differences. These were so severe that
they were forced to use a correction factor, but that is not
the solution to the problem. Another conundrum is indicated
because initially the cancer mortality rate in Copenhagen
was significantly higher than in the rest of Denmark. The
question arises as to whether this data is based on an
underlying, undetected error or whether it is the results of
medical care in the Danish capital being below average. If
the second theory is the correct then the reversal of the
mortality rates in Copenhagen and the other cohorts may only
show that Copenhagen reached the high standard of the rest
of the country by initiating mammography screening. And this
would also show that high standards in cancer prevention are
to be obtained by other means as well.
Thus, if many methods lead so similar results, it is fair to
assess their advantages and disadvantages and costs.
Mammography, used as a screening tool, is very expensive. In
order to show positive effects at all, participation must be
close to 100 %. However, the more women participate, the
greater the number of false positives and false negatives.
Before declaring the mammography screening to be the
gold-standard method, one has to consider other methods,
such are ultrasound and palpation. It is a fact that the
palpation of the mammary glands has caused the largest
advance in the fight against cancer; and it is true also,
that until today most of the breast cancers are detected by
Ultrasounds, especially as practiced by less experienced
health care professionals, results in twice as many errors
in diagnosis than mammography. But the statistical effects
are not as obvious as presumed. Let us assume that
mammography detects 90 % of all breast cancers and
ultrasound 80 %; and let us then translate these assumption
into the statistics underlying the Copenhagen study.
Mammography there has prevented 40 out of 160 cases of
death, and it has missed 4 more cases of undetected cancer.
If it were not mammography but ultrasound screening, the
statistics would show 32 cases prevented and 8 cases not
detected. Mammography is more effective, but the benefit is
very small and far beyond estimated benefits. The question
remains whether the high effort end expense of mammography
used as a method of screening is really justified, all the
more as all the calculations have to be adjusted by the
described statistical deficiencies.
What is the best advice for a woman between 50 and 70 years
First of all to accept that breast cancer is one of the
relatively rare diseases. But it is a cancer with best
conditions for detection. Every woman ought to be guided and
instructed in the methods of the palpation of the own
breast´s glands. Ultrasounds shall not be considered as a
second rate quality method, because it is unerring enough
for a screening method and has no negative collateral
effects like induced cancers. Mammography remains a
high-quality method, especially when used as indicated and
individually, but definitely not as a screening method.
In all cases of pre-existing suspicious results from
whatever kind of examination used, the 10 % errors of
mammography are less important in both directions. If the
result is false positive and more invasive examinations are
made, it is justified by the pre-existing suspicion. And if
it is false negative, the pre-existing suspicion demands
for closely coordinated follow-up examinations in any case.
Both aspects differ fundamentally to which we are accustomed
in a screening-area.
The reasonable use of all existing
methods combined in a logical system is supposed to be more
effective than the preference of one method especially when
it is used in screening.
Dr.med. Karlheinz Bayer
D-77740 Bad Peterstal
AH Olsen, Sisse H Njor I Vejborg, W Schwartz,
P Dalgaard et al. Breast cancer mortality in Copenhagen
after introduction of mammography screening: cohort study
BMJ 2005;330:220 (29 January),
Mammographie-Screening - die Propaganda mit der Angst
Ärztezeitung vom 27.1.2005
Nyström L, Rutqvist LE, Wall S, Lindgren A, Lindqvist M,
Ryden S, et al. Breast cancer screening with mammography:
overview of Swedish randomised trials.
Lancet 1993;341: 973-8
Andreasen AH, Andersen KW, Madsen M, Mouridsen H, Olesen KB,
Lynge E. Regional trends in breast cancer incidence and
mortality in Denmark prior to mammographic screening.
Br J Cancer 1994;70: 133-7
Blanks RG, Moss SM, McGahan CE, Quinn MJ, Babb PJ. Effect of
NHS breast screening programme on mortality from breast
cancer in England and Wales, 1990-8: comparison of
observed with predicted mortality.
Otto SJ, Fracheboud J, Looman CWN, Broeders MJM, Boer R,
Hendriks JHCL, et al. Initiation of population-based
mammography screening in Dutch municipalities and effect on
breast-cancer mortality: a systematic review.
Lancet 2003;361: 1411-17
Jonsson H, Nyström L, Törnberg S, Lenner P. Service
screening with mammography of women aged 50-69 years in
Sweden: effects on mortality from breast cancer.
J Med Screen 2001;8: 152-6
Duffy SW, Tabar L, Chen HH, Holmqvist M, Yen MF, Abdsalah S,
et al. The impact of organized mammography service screening
on breast carcinoma mortality in seven Swedish counties.
Cancer 2002;95: 458-69
Alexander F, Anderson TJ, Brown HK, Forrest APM, Hepburn W,
Kirkpatrick AE, et al. years of follow-up from the Edinburgh
randomised trial of breast-cancer screening.
Lancet 1999;353: 1903-8
Gøtzsche PC, Olsen O. Is screening for breast cancer with
Olsen O, Gøtzsche PC.
Cochrane review on screening for breast
cancer with mammography.
R Kürzl, Evidenzbasierte Mißverständnisse beim Mammakarzinom
Deutsches Ärzteblatt (Heft 36-2004)
W Krämer So lügt man mit Statistik
Campus Verlag 1992