diagnosis by screening has a diminishing impact on
mortality.20 This may now be the case for breast cancer, where
adjuvant therapy can substantially reduce mortality from
early disease whether it is diagnosed clinically or by
mammographic screening.16,19,20,21,22,23,24
The analytic techniques that are used to evaluate
population-based mammographic screening programmes are
also controversial. The advice of WHO-IARC is that
observational studies, especially case-control studies, should
not be used; they were used extensively in a review of
European mammographic screening programmes.3,25 This
WHO advice is important, and echoes earlier cautions
published in the BMJ: "that confounding and selection bias
often distort the findings from observational studies and that
there is a danger that meta-analyses of observational studies
produce very precise but equally spurious results".26
It is increasingly apparent that over-diagnosis of in situ and
invasive breast cancers, which would never have troubled
women in their lifetimes, is substantial and has been
estimated in a meta-analysis as being 52% of women
diagnosed by screening27, although published estimates of
over-diagnosis vary widely.27,28,29,30 Over-diagnosis is a
common consequence of screening asymptomatic people and
finding biological forms of a disease that will never cause
clinical disease. This leads to unnecessary investigations and
unnecessary treatment, so the ratio of mortality reduction
(benefit) to over-diagnosis (harm) consequent upon
mammographic screening for breast cancer in some
countries may now be unfavourable. Evaluations of
mammographic screening programmes that report on overdiagnosis must make
an estimate of the lead time from
screening. It has been reported that modelling generally
overestimates lead time and hence underestimates overdiagnosis as compared to clinical measures of lead time.30
Therefore estimates of over-diagnosis that come from
modelling studies should be viewed with caution.
Today, the dilemma in countries with mammographic
screening programmes is whether the, at most, modest
benefits of mammographic screening are now outweighed by
the harms of over-diagnosis.31 In particular, one of the
advertised benefits of mammographic screening has been
that it detects small breast cancers which can be treated by
partial mastectomy and radiotherapy and so preserves the
breast.20 Unfortunately, it is now apparent that external beam
radiotherapy, which is the usual mode of administering
radiotherapy to the breast after breast-conserving surgery,
can later cause cancer of the lung and/or accelerated
atherosclerosis of the left anterior descending artery
predisposing to myocardial infarction, since both the lungs
and heart lie just a few centimetres beneath the breasts.
Therefore, more women treated with radiotherapy in this
way may die from these complications than avoid death from
breast cancers detected by screening.20,32 It is important to
emphasize that studies of mammographic screening, which
do not measure all-cause mortality as well as breast cancerspecific mortality,
will not detect these lung cancer and
myocardial infarct deaths and therefore give a spurious
picture of the benefits of screening.32
The balance of benefits and harms for mammographic
screening is critical in low- and middle-income countries,
where the incidence of breast cancer is lower than in highincome countries and where breast cancer is more likely
to be
a premenopausal disease, with most women diagnosed with
breast cancer at less than 50 years of age.
1,33 Since
mammographic screening is more sensitive in
postmenopausal women, who have less dense breasts than
premenopausal women, this combination of lower incidence
and mainly premenopausal breast cancer means that,
compared with high-income countries, many more women will
have to be screened in low- and middle-income countries to
detect one breast cancer.1 For example, it has been calculated
that in a low- or middle-income country where the incidence
of breast cancer is half that of the United States "for women
aged 40 to 49 years, more than 3,800 women would need to
be invited for screening at the cost of more than 41,000 visits
for mammography and more than 4,000 false-positive
diagnoses, in order to prevent one death from breast cancer
during 11 years of follow up".
1This should give pause for thought
to governments contemplating introducing mammographic
screening in low- and middle-income countries.
In the United States, where the incidence of breast cancer
is one of the highest worldwide34, the cost effectiveness
of mammography has been estimated to be $US
30,000-100,000 per quality adjusted life year gained.1 Even
though a mammogram may be cheaper to deliver in a lowand middle-income country,
because of the sensitivity and
incidence issues described above, when compared with a
high-income country, the cost-effectivenss of mammography
in a low- and middle-income country may be worse.
Finally, there has only been one mammographic screening
RCT reported for a non-European-derived population which
was carried out in Singapore, where a low participation rate
amongst Chinese women resulted in selection bias in favour
of women of higher socioeconomic status and education in
the screened group and no mortality outcomes were
reported.35
The debate about the balance of benefits and harms of
mammographic screening has important implications for all
DISEASE-SPECIFIC CANCER CONTROL
100 CANCER CONTROL 2014