■ site map
June 2010 - December 2013
II - Mammography
2. Mammography: The neglected evidence
It is 16 years now since Lancet published a study by two Canadian researchers, attempting to give a rational, balanced account of what available research data says X-ray mammography does, and doesn't (Wright and Mueller, Screening mammography and public health policy: the need for perspective, 1995). The authors concluded that
the facts do not
support public funding for mammographic
The reason? As table below shows, they found that only
the first two, out of six
large randomized controlled trials at the time, have come up with
statistically significant reduction in mortality from breast
cancer in study populations screened by mammography: 23% and 31% in the New York and
one of the Swedish studies (Two-County),
respectively (for the relative mortality reduction, corresponding
absolute mortality reduction figures were much lower, 0.05% and
et al. Ten to fourteen year effects of breast cancer
screening on mortality
Note that Wright and Muller did not include Göteborg trial (started in 1982, ended 1990), possibly because complete data hasn't been published at the time (nor, for that matter, was as of 2009).
In the three trials following the New York and Two-County, mortality reduction in mammography groups was statistically insignificant, and in the last one the mammography group had higher mortality (though also statistically insignificant) than control.
To illustrate the risks and benefits, the authors made a simple calculation. The number of women needed to screen for 1 breast cancer death less ranged from 7086 to 63,264 in the five studies with mortality reduction, averaging 24,000. Combining it with the study that showed mortality increase for the mammography cohort, it comes to
1 breast cancer death less for every 32,000 women screened.
At the then-published rates, that one death less would come at the price of roughly 1,500 false positives, 4,000 false negatives, and $2 million spent on screening (note that the authors in their paper left out the Canadian study with its negative mortality effect, and used a rounded off number of screenings for 1 death less in 20,000 screened women, but it doesn't change the big picture).
In short, statistical benefit of mammography screening, as indicated by these six studies, is so small that
can be qualified as uncertain.
What is certain, are the harms caused by test's inaccuracy, its cost, and its radiation risk, which may be significant for highly sensitive women.
If so - and the evidence does support such conclusion - the huge amount of money spent on mammography and related procedures, could be used more efficiently (i.e. on some other forms of diagnostic test, and/or preventive treatments) to combat breast cancer.
Despite being fundamentally correct, the study made no impact. Nothing seemed to have the power to disturb the mammography doctrine. Instead, life-saving "benefits" of mammography screening were to be extended to women below age 50.
If Wright and Mueller stirred any change, it was instigating the move toward re-affirming values of mammography; a few more "significant" trial updates, trials and meta-studies that produced more of statistical benefits from mammography screening, seemingly removed any doubt in its great life-saving effect.
Good part of the reason was that mammography screening had already become firmly entrenched as the centerpiece of the official policy's weaponry against breast cancer, and its benefits were generally accepted among medical professionals as advertised by its promoters.
Official backing, professional acceptance, career stakes, pride and big profits, all gave to the mammography "benefits" life of their own. Any contrary, unfavorable evidence was ignored, sidelined, or labeled as "flawed" or "incompetent".
For instance, the Canadian study, which showed no benefits from mammographic screening vc. clinical breast examination was, believe or not, investigated by a hired private investigation company - and subsequently cleared - for tampering (Randomization in the Canadian National Breast Screening Study: a review for evidence of subversion, Bailar et al. 1997).
The study was given "stick and stone treatment" by the proponents of screening. Anything from unproven - and refuted, as mentioned above - accusation that the trial placed larger number of women with advanced cancers into its 40-49 year intervention group (Kopans; in fact, there was more clinically positive women in the control group), to alleging that the quality of its mammography was "indefensibly poor" (Kopans), while in fact it had documented above average efficacy for this age group, to numerous distortions and plain fallacies about the trial fed to media by the pro-screening side (for instance, that it had extraordinarily high rate of false negatives, or that it used ordinary X-ray units instead of mammography units).
The irony is that this Canadian
randomized clinical trial actually raised quality level not only
with respect to all other breast cancer studies from that time,
but also with respect to most trials after it as well. Its
deputy director illustrated all absurdity of the continuous
dismissal of this study by X-ray mammography advocates as being
of "low quality" and "flawed" (Baines, Rational and
irrational issues in breast cancer screening, 2011).
Not a single one of these vital indicators of trial quality favors trial B; it is trial A that is superior in all these aspects. Yet it is trial B - the Two County Trial - that pro-mammography camp up to this day presents as a high quality and reliable, while dismissing trial A - the Canadian trial - as "low quality" and "flawed". The reason can be seen at the very bottom: the Two County trial is the only major trial after the very first (HIP) that produced statistically significant mortality reduction in the screened population.
On the other hand, the Canadian trial is the only one that came up
with near-zero overall life-saving benefit from screening.
It is this single aspect that placed these two trials at the center of controversy and determined which one will be honored and promoted, and which one denigrated and dismissed.
The above study design data are not complete, and that is more of a disadvantage for the Canadian trial, since the insight into many relevant aspects provided by, and its transparency, are unmatched, not only by the Two County trial, but also by any of the breast cancer trials before and after that.
Meanwhile, the publicity based on poorly designed, biased studies establishing "clear benefit" from mammography screening, was used to further promote the practice. The results from Two-County trial were, for instance, the deciding factor in the introduction of organized screening in Netherlands (1990), and among major factors prompting its introduction in the U.K.
While it took decades to unveil poor quality of the Two County trial, clear indications of flaws in the other major mammography-supporting trial, the New York trial, were noted and brushed aside at the time it came up with its results. Specifically, number of breast cancers diagnosed was nearly identical in the screened and control group - frankly, with those averaged in the screened group being actually larger, which is
logically incompatible with the study conclusion of
In fact, the main factor producing such outcome seems to be the exclusion of far more women with prior breast cancer from the treatment (screened) group,
directly lowering its mortality rate vs. that in the control group.
The only thing that changed in the practice of promoting
mammography screening at any cost is that the main reason for
doing so switched from the belief that it "has to be
beneficial", to merely protecting interests of the screening
The magnitude of facts distortion by die-hard supporters
of standard mammography screening nowadays is well illustrated by this
online article by a U.S. radiologist (The Journal of Lancaster
General Hospital, PA). Here's the excerpt describing the Canadian
trial (which consisted of two separate branches, for 40-49y and
50-59y age group):
"...Importantly, none of those large Canadian studies was a truly
randomized controlled trial, since patients were assigned to the
screening or no-screening groups only after a physical examination
of the breasts. Since women with palpable masses were selectively
assigned to the screening cohort, it contained 4X as many advanced
cancers as the non-screened group. This difference inevitably
altered the mortality statistics, which compromised the parameter by
which the effectiveness of screening mammography was assessed.
Errors in certification of the cause of death were also raised by
the authors in the first study, a fact that was even noted by the
authors of the paper. Compliance among the women in these studies
was also a problem, with nearly 20% of those assigned to regular
screening not getting these examinations over a 4-5 year period, and
many women assigned to the no screening cohort getting screening
mammograms outside the study. Questions have also been raised about
the quality of mammographic examination, the skills of the
interpreting physicians, design study, randomization, contamination
of controls, follow-up, and ascertainment, thus further eroding the
conclusions reached in this study..." (Controversies in
screening mammography, N.K. Tanna, 2010)
"...Importantly, none of those large Canadian studies was a truly randomized controlled trial, since patients were assigned to the screening or no-screening groups only after a physical examination of the breasts. Since women with palpable masses were selectively assigned to the screening cohort, it contained 4X as many advanced cancers as the non-screened group. This difference inevitably altered the mortality statistics, which compromised the parameter by which the effectiveness of screening mammography was assessed. Errors in certification of the cause of death were also raised by the authors in the first study, a fact that was even noted by the authors of the paper. Compliance among the women in these studies was also a problem, with nearly 20% of those assigned to regular screening not getting these examinations over a 4-5 year period, and many women assigned to the no screening cohort getting screening mammograms outside the study. Questions have also been raised about the quality of mammographic examination, the skills of the interpreting physicians, design study, randomization, contamination of controls, follow-up, and ascertainment, thus further eroding the conclusions reached in this study..." (Controversies in screening mammography, N.K. Tanna, 2010)
Statements that neither branch of the Canadian trial was "truly randomized", and that its screening cohort "contained 4X as many advanced cancers" are plain false. As mentioned above, because of accusations by the mighty pro-screening camp, study's randomization was investigated more than a decade prior to this article, and no evidence of tampering was found.
As for "4X as many advanced cancers" in the screened group, it could be - unintentional? - mix up related to the fact that the screened group in the Canadian study had more node-positive cancers with four or more nodes involved detected. The actual numbers are 290 vs. 237 after five years, and 592 vs. 552 at 13-year follow-up for the screened population and controls, respectively. Nothing remotely close to "4X as many advanced cancers".
"Errors in certification" probably refers to the data presented in the trial paper where, due to confidentiality limitations imposed by hospitals, an external panel blind to the screening status could not investigate and confirm cause of death in all cases (for the 40-49 age group, it reviewed 67% and 77% certificates for screened and control group, respectively, with the cause of death for the rest taken as reported on death certificate); this, of course, is not an "error", rather a cut above the Two County trial, in which reviewers weren't blinded to the status, and study authors were directly involved in determining the cause of death.
Compliance was actually better in the Canadian trial than in the Two County; "contamination" by women in the control group receiving occasional screening or diagnostic mammograms (about 20% averaged for both trial branches) is somewhat higher than the all-trial average (as published), but still within the range. Malmo study, for instance, had 25% contamination, and some studies (HIP, Stockholm, Edinburg) did not report it at all. The effect in the Canadian trial was estimated to be minor by the authors.
The unreferenced list of flaws at the end of the excerpt is a fairly good summary of other baseless objections to the quality of the Canadian study, serving as the pretext for its dismissal.
Also very typically, article author has
no single objection
to the design or execution of any of the trials that did produce more tangible - even in most cases still statistically insignificant - mortality reduction in the screened population. Swedish trials, collectively, are declared as "some of the best" (no specific reasons given), showing "nearly 30-45% decrease in mortality in the screened groups".
But wait, it gets even better: adding "recent updates, still unpublished", from author's conversation with Dr. Tabar (lead author of the Two County trial), "may show a nearly 65% reduction in mortality from breast cancer" due to screening X-ray mammography.
You may think that biased ramblings of an obscured U.S. radiologist don't matter much in the big picture, but think twice. It is such radiologists and the myths of screening mammography imposed as the official view that
many a woman depend on for information and guidance.
These ramblings are merely reflection of the official view supported by hard core advocates of the standard mammography screening practice, still dominant in the "official" medicine.
So, that was how it all started with the X-ray mammographic screening, and how it kept going: by assuming benefit, embracing all that seemingly supported it, while disregarding evidence to the contrary. The reality is that the X-ray screening mammography has become a major part of breast cancer public policy in some of the most developed countries, including U.S. It also became the backbone of quite a few academic careers, accepted practice for doctors, comforting, trusted presence for millions of women, and a multi-billion dollar business.
But all this time it was standing on shaky ground.
In 1999, a small epidemiological study in Sweden - one of the most active countries in studying the effects of mammography screening - came up with a news outright shocking for most, although not entirely unexpected for a few. It found that in the ten-year period from 1986 to 1995, breast cancer mortality in over 600,000 Swedish women undergoing mammography screening declined only 0.8% vs. those that did not use the screening - much less than what the 11% officially accepted estimate was projecting (Sjönell and Ståhle, 1999).
This little study set in motion chain of events.