Just Published Canadian Study that Legitimately Questions the Utility of Annual Mammography — Will Not Change Minds — Due to Understandable Cancer Fears and “Costs Be Darned” Thinking

14 February 2014 (expanded 18 February 2014)

Citation

Anthony B Miller, Claus Wall, Cornelia J Baines, Ping Sun, Teresa To, and Steven A Narod, Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomised screening trial, British Medical Journal [BMJ] 348: g366, DOI: 10.1136/bmj.g366 (11 February 20-14)

This essay makes two main points

The just published Canadian study cautioning against annual mammograms for women from age 40 to 59 will not change minds, even though it should.

A simple Bayesian statistical example (essentially about the same subject) from the Sci Show on YouTube might alter screening behavior, provided thoughtful people pay attention to it.

When something is as deadly as breast cancer often is — calls to conserve health resources by limiting questionable screening processes understandably fall on deaf ears

This just published Canadian study’s statistically impressive findings essentially confirm the 2009 US Preventive Services Task Force recommendations to:

(a) avoid routine annual mammography breast cancer screening in women aged 40 to 49

and

(b) to reduce substitute biennial screenings for the previously recommended annual ones in women aged 50 to 74.

At the time, the USPSTF recommendations met with hostility from most of the medical establishment, as well as patients.

Note

I discussed the conflict — and the valid reasons for it — at length, here.

Although the newly published Canadian study basically proves the merit of the 2009 Task Force’s reasoning, it will meet the same fate

That so, despite the Canadian study’s statistical clarity regarding:

(a) the lack of merit to annual population mammography screening, as evaluated from a public health perspective

and

(b) the noticeable screening detriments attributable to unnecessary medical interventions attendant upon providing arguably too much mammography.

Why this essay?

The issue of breast mammography screening is a perfect confrontation between what we would like to think we control and the actuality of the little we know and can control. As such, it serves as an excellent overview of the conundrum that sets doctors and patients against public health’s arguably limited resources.

In other words, the issue sets legitimately frightened people against statistically-supported efforts at health care cost control.

My review’s structure

What follows presents some of the most important of the Canadian study’s findings.

I follow the findings with a brief discussion of why the study’s “proven” numbers will not persuade the unpersuadable. Meaning most of us.

And conclude by adding in a Baysian statistical analysis of why those who underestimate the numerical persuasiveness of the Canadian study are actually wrong.

The Canadian study — method

This was an admirable attempt to make sense of a difficult-to-probe subject.

The goal was:

To compare breast cancer incidence and mortality up to 25 years in women aged 40-59 who did or did not undergo mammography screening.

© 2014 Anthony B Miller, Claus Wall, Cornelia J Baines, Ping Sun, Teresa To, and Steven A Narod, Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomised screening trial, British Medical Journal [BMJ] 348: g366, DOI: 10.1136/bmj.g366 (11 February 20-14) (at Abstract)

The study began in 1980 and ended in 2005 — 89,835 women from 40 to 59 years old were followed for a mean period of 22 years. (Recall that “mean” means mathematical average.)

These nearly 90,000 women were randomly and blindly divided — meaning that the people who ran the study had no idea who went where — into an annual mammography (experimental) and no annual mammography (control) group.

The control group, however, did receive the usual standard of care for non-annual mammography screening, such as palpating breast exams and the usual follow-ups to finding suspicious lumps, including mammography, biopsy and cancer treatments.

Findings — two

First, annual mammography does not reduce the risk of dying from breast cancer, as compared to palpation and physician in-office exams with appropriate follow-ups.

Second, 22 percent of the patients who undergo annual mammograms receive unnecessary and potentially harmful medical interventions.

From the abstract:

Annual mammography in women aged 40-59 does not reduce mortality from breast cancer beyond that of physical examination or usual care when adjuvant therapy for breast cancer is freely available.

Overall, 22% (106/484) of screen detected invasive breast cancers were over-diagnosed, representing one over-diagnosed breast cancer for every 424 women who received mammography screening in the trial.

Details

The study makes clear that mammography has a slight advantage (compared to palpation) in detecting smaller tumors:

During the screening period the mean size of the cancers diagnosed in the mammography arm was 1.91 cm and in the control arm was 2.10 cm . . . .

In the mammography arm, 30.6% of cancers (n=204) were [lymph] node positive and 68.2% (n=454) were palpable.

In the control arm, 32.4% of the cancers (n=170) were node positive . . . and all were palpable.

Overall, 454 palpable cancers were detected in the mammography arm and 524 in the control arm, whereas similar proportions of palpable cancers were identified as node positive.

On average, palpable cancers were larger than cancers that were detected only by mammography (2.1 cm v 1.4 cm . . . and were more likely to be node positive (34.7% v 16.5% . . . .

The 25 year survival was 77.1% for women with tumours of less than 2 cm, compared with 54.7% for tumours greater than 2 cm . . . .

BUT — according to the research team, smaller tumor size detection advantage of annual mammography did not noticeably carry over into extended survival times for that group.

The authors’ Figure 2 graph — under the Results - Breast cancer mortality section — shows that the plots for 25 year survival for both arms of the study overlap.

This means that for these close to 90,000 study subjects, there was NO long-term survival benefit to being in the annual mammography arm of the study.

Was there a benefit to annual mammography regarding shorter-than-25 year survival?

No.

Figure 3 of the paper plots the probability of survival for patients in both arms of the study against a timeline running from Year 0 to Year 25. Again, the plots overlap with an apparently statistically meaningless slight survival disadvantage to annual mammography for cancer diagnoses made in approximately Years 8 to 11 of the enrollment period.

If there was no survival advantage to annual mammography — was there (at least) no disadvantage to providing it?

Again, no.

Annual mammography has a noticeable disadvantage in addition to increased health care costs:

At the end of the screening period, an excess of 142 breast cancer cases occurred in the mammography arm compared with control arm (666 v 524) . . . .

Fifteen years after enrolment, the excess became constant at 106 cancers. This excess represents 22% of all screen detected invasive cancers—that is, one over-diagnosed breast cancer for every 424 women who received mammography screening in the trial.

“Overdiagnosed” is a challenging choice of words for non-medical readers

What the team appears to mean is that either:

(a) interventions were made in regard to “lesions” (abnormalities) that were not cancerous or harmful — usually called false positives

(b) interventions were made for lesions that were cancerous, but not serious or progressive enough to warrant the scope of the intervention in light of the patient’s overall health.

The significance of the overdiagnosis distinction is that we do not want to intrude into the body, unless the intervention is both necessary and likely to be effective.

Pertinent to this dilemma, consider the three categories of “true positives” (cancers that are real). These are:

(a) Cancers detectable by screening that are curable after clinical diagnosis [palpation, for example].

(b) Cancers detectable by screening that are incurable after clinical diagnosis but curable after [mammography] screen detection.

It is category B that we seek to maximize through screening, none of those cases in categories A or C benefit from [mammography] screening, though they cannot be specifically identified at the time.

Medical intrusions themselves carry risks of infection, maiming, and cosmetic damage — as well as unnecessary psychological distress, the magnitude of which many or most patients appear to underestimate, until it happens to them.

Then there’s this bombshell

Medicine still presumes that arguably unnecessary fiddling with a tumor can:

(a) spread or accelerate the spread of its malignant cells

and

(b) at the very least, induce the body’s tissue repair process, which potentially itself promotes more rapid growth of the tumor.

Most obvious general objection to the study — technological advance

The most obvious objection to the Canadian data is that it is reviewing the effects of comparatively old technology. No study, in these technology-accelerating times, can overcome that caveat.

My own view, however, is that nothing in breast cancer mammography or breast cancer’s pathological identification has advanced so definitively as to obviate the gist of the study’s findings. Medicine, generally, is nowhere near as advanced as researchers, physicians, and patients wish it to be.

Note

I discussed that depressing reality from the too often hyped “omics” perspective, here.

What are women to do?

Talk to your doctor about the Canadian study. Press her (or him) to find out if she has open-mindedly considered the study’s implications.

That said, the situation essentially remains as it was when I wrote the following (last year):

Medical providers are in a difficult position.

They want to do what is best for the individual patient, not for the health system generally. And most providers would take it personally, if a patient died because the provider had not screened for the illness that subsequently killed the patient.

Statistical evidence against the wisdom of the routine public health screening of low risk patients does not make up for the sadness (and possible malpractice litigation) of having lost someone who could have been saved.

And that is where the USPSTF’s recommendations were inadequately sold — and why they (apparently) did not work to change clinical practice.

© 2011 Peter Free, The US Preventive Services Task Force (USPSTF) 2009 Mammography Screening Recommendations Were Understandably Ignored by the Medical Establishment and Its Patients, PeteFree.com (01 May 2013)

The Canadian study gives the USPSTF recommendations against annual mammograms very clear evidentiary support, but the medical provider’s dilemma, nevertheless, remains the same.

Look at this way — for a dose of percent-based patient reality

If health care costs are not an issue for medical providers and their patients, the question becomes:

Should we detect smaller tumors with mammograms — at the apparently sole cost of taking on the risk of subjecting 1 in 424 (female patients (aged 40 to 59) being mistakenly diagnosed with a tumor that either:

(a) was not one (false positive)

(b) was (true positive), but should not have been operated on — given either its low-grade nature or the patient’s overall health?

This, I think, is an easy question to answer for most women, who are legitimately fearful of such a devastating disease.

They are going to continue to want annual mammography. Despite the fact that 22 percent of the supposedly invasive cancers in the annual mammography group were “overdiagnosed” — meaning that the subsequently taken medical procedures were unnecessary and probably harmful.

Why would patients ignore this 22 percent bad result?

Because the (only apparently) more relevant number is the 0.24 percent of the whole mammography group that was mistakenly diagnosed — a ratio of 1 women per 424 of the women screened. This small percentage occurs — in contrast with the large 22 percent — because we are looking at ALL the women in the mammography group, not just the ones with apparently invasive detected tumors.

Looking at the whole group — a 0.24 percent chance of being overdiagnosed with apparently invasive cancer is apparently trivial. And patients and their physicians are going to ignore that tiny chance of misfortune with seemingly just cause.

BUT — what the Canadian study does NOT explain, and what most physicians and patients will not recognize, is a statistical twist that flips the above “I am not convinced” reasoning on its head

Quite by chance, I came across the following outstanding video about statistics and breast cancer detection.

The following Sci Show statistics video does a better job of showing the volume of potential harms attributable to annual mammography (for women in their 40s) than both the 2009 USPSTF recommendations and the Canadian study.

Sci Show, How to Predict the Odds of Anything, YouTube (17 February 2014)

Be warned, you will have to put your numerically-minded brain in gear before watching the video’s racehorse speed explanation.

What the Sci Show’s Baysian analysis implicitly shows about the potential risks of annual mammography

If you are a thoughtful woman in her 40s, pay attention to this. It may change your opinion regarding the intuitively presumed merits of undergoing annual mammography.

Basic breast cancer facts

1.4 percent of women in their 40s have breast cancer

If a woman in this age group has breast cancer, there is a 75 percent chance that mammography will detect it

If a woman in this age group does not have breast cancer, there is a 10 percent chance that mammography will mistakenly say that she does have cancer (false positive)

Assume we have a randomly selected sample of 1,000 women in their 40s

14 will have cancer

1.4 percent chance of breast cancer x 1000 women = 14 women with cancer

986 will not have cancer

1000 minus the 14 women with cancer = 986 women without cancer

But what will the mammograms say about this group of 1,000 women?

Only 10.5 of the 14 women with cancer will actually be detected

14 women with cancer x 75 percent chance of detecting cancer = 10.5 women with true positive mammography screening

98.6 women will be mistakenly thought to have breast cancer

986 women without cancer x 10 percent false positive mammography screening rate = 98.6 women with false positives

What are the chances of a woman, who is undergoing a mammography screen, of being accurately diagnosed with breast cancer?

9.7 percent

10.5 women with true positive + 98.6 women with false positive = 109.1 total positive results

10.6 women with true positive results (divided by) 109.1 total positive results = 9.7 percent chance of receiving an accurate mammographic detection of existing breast cancer

What is important (for women in their 40s, who do not know their breast cancer status) about the above calculations?

Two things should catch your attention:

Your chance of being accurately diagnosed (via mammogram) of having breast cancer are essentially equal to your chance of being inaccurately diagnosed with the disease.

10 percent false positive mammography result — versus — 9.6 percent chance of being accurately diagnosed with breast cancer

Worse, because overwhelmingly most women in their 40s do not have breast cancer, this means that numerically way more women are receiving false positive diagnoses than those who are receiving accurate detections of breast cancer.

Recall that 98.6 women in the 1,000 women group are being told they may have cancer, when they do not

And only 14 women in the same screening group actually have cancer — and, of these, only 10.5 are getting the correct mammographic diagnosis

So, mammograms are delivering 10.5 correct diagnoses for every 102.1 (out of each 1,000 women screened) that they get wrong.

In absolute terms, 98.6 women per each 1,000 are going to undergo potentially harmful follow-up tests and interventions that they do not need.

And 3.5 women out of each 1,000 are going to be mistakenly told that they do not have cancer, when they do.

If now we do not see analytical data to back up the USPSTF and the Canadian study’s cautionary recommendations against annual mammography, we are not thinking realistically about the risks that accompany mammography screenings and false positive follow-up interventions.

That said, just as the Sci Show video indicates, the overwhelming majority of people are not accustomed to thinking in numerically analytical ways. Consequently, I am reasonably confident that the Canadian study will not alter the medical establishment’s behavior any more than the 2009 USPSTF recommendations did.

The moral? — Nothing will change, until insurance companies get serious about controlling costs

Only when insurance providers get on board with the USPSTF’s and Canadian study’s data-based perspective will a clinical paradigm shift take place. When women have to pay for screens that insurance will not cover, they might become more amenable to the USPSTF and Candadian study’s possibly valid logic.

But until that day, no reasonable medical provider is going to voluntarily subject herself to assisting in potentially killing off a small, but noticeable number of her patients.

In thinking about this health care cost problem, remember that there are essentially no penalties for practicing overly cautious medicine. The American paradigm seems to be:

(a) the more screening the better — even when the evidence clearly argues against that proposition

and

(b) the more procedures and interventions the better.

So far, there has been effectively no price to pay for the cost of what the Canadian study calls overdiagnosis. Nor, too often, has there been a cost for practicing essentially purposeless medical interventions.

The bottom line — in addition to the unavoidable scientific ambiguity that much of medicine necessarily involves — is that the medical establishment makes a whole lot of money from feeding and supplying the infrastructure that supports medical interventions — whether necessary or not.