CHAPTER 23 — from Mammography and Early Breast Cancer Detection: How Screening Saves Lives by Alan B. Hollingsworth, MD (McFarland & Company, Jefferson, NC, 2016).
…a history of mammography leading to current controversies, a primer on the epidemiology of screening, and a polemic designed to endorse the need for multi-modality imaging in order to improve early detection.
LINK to Publisher’s Page: http://www.mcfarlandbooks.com/book-2.php?id=978-1-4766-6610-5
Shortly before the new year of 2004, an educator from Connecticut underwent her annual screening mammography, as she had been doing for years. Like before, the results were “All clear.” Six weeks later, she was diagnosed with Stage IIIC breast cancer. She was stunned to learn, for the first time, that she had very dense breast tissue on mammography, the feeble explanation as to why her cancer had been invisible (probably detectable several years earlier with other modalities).
Nancy M. Cappello, PhD has since joined the list of solo game-changers, that is, non-celebrities who have risen from anonymity to turn their misfortune into far-reaching upheavals that have permanently altered how we manage breast cancer. In the footsteps of Rose Kusher (ending the one-step biopsy/frozen section/mastectomy approach), Betty Rollin (First You Cry, the book and movie that introduced many to the emotional impact of mastectomy), Susan Komen and her sister Nancy Brinker (initially, the promotion of screening mammography, then breast cancer research in general), Nancy Cappello set fire to sweeping reforms as to how women are informed about their breast density.
As of this writing, nearly half of the 50 states have passed legislation requiring radiologists to inform patients when mammograms are dense (white on X-ray) and to describe additional imaging modalities that might help, most notably ultrasound. Federal legislation is being considered as well. Through the Are You Dense? Organization (RUdense.org) that Dr. Cappello founded, Connecticut became the first state to require that high density information be given to patients, then she has encouraged all states to do the same. Additionally, Connecticut became the first state to pass legislation requiring third party payors to cover ultrasound screening for women with dense breast tissue.
What added fuel to Dr. Cappello’s conflagration was the discovery that the medical profession knew about the problem of breast density all along, but did precious little to inform the public.
I sympathize without excuse. I do have mixed emotions about legislated physician practice, but this bundle of provocative data about density was dropped into a black hole, it seems. As an aside, it has always been a curious phenomenon that some information new to the scene is processed quickly and new standards adopted overnight, while other innovations or ideas are ignored. After decades of medical literature warning about the danger of breast density, it was like the boy who cried “Wolf,” with readers becoming numb to the data, if they paid attention at all.
And for those historians of breast density who have already groaned at the pun, it was Dr. John N. Wolfe who first drew attention to mammographic density patterns, publishing his classification system of N1, P1, P2 and DY that decorated the bottom of mammography reports starting in 1976(1). I read my first mammogram report with the Wolfe pattern noted in 1980, my first year in private practice. The reaction was straightforward at that time – so what? There were no good alternative methods for imaging, and the focus then was not on mammographically invisible cancers. In fact, Dr. Wolfe proposed his system entirely on the basis of imparted risk for breast cancer, yet there were no interventions available at the time short of preventive mastectomy. In fact, Dr. Wolfe was so convinced that women with the DY pattern were headed for breast cancer (45% lifetime risk by his calculations) that he felt they should seriously consider preventive mastectomies. And with the introduction of breast implants at about this same time, “subcutaneous mastectomies with implants” were already being performed for reasons far less impressive than Wolfe DY patterns.
But then, we entered a silent period, where at the clinical level, Wolfe patterns were gradually dismissed, sometimes as an old-fashioned “folly.” Certain investigators, though, were using different classifications schemes, but still coming up with the same conclusion – denser breasts translated to greater risk. Still, the entire focus was still on imparted risk. It was only later that clinicians began to appreciate that breast density was double jeopardy. Not only was density imparting elevated risk, but also it was responsible for hidden cancers missed by the very mammography that defined density in the first place.
What is most peculiar in re-tracing my steps during this era is that no one was arguing anything different. It wasn’t controversial, rather, it was esoteric. The risk data for density was consistent no matter how the density levels were described – the highest density breasts had roughly a 4-6 fold relative risk for breast cancer when compared to the lowest density breasts. And, lagging behind, only a handful of studies prior to 1990 showed the danger of density when it came to early detection. The vast majority of clinicians believed the “90% Sensitivity” for mammography, independent of density levels.
A word about 4-6 fold relative risk. Recall from the Number Games chapter that relative risks (RRs) are fractions with a numerator and a denominator. When the word hit the street about the 4-6 fold risk, many patients were terrorized by these numbers. The problem with this particular RR is that the number applies to the highest density compared to the lowest density, the latter being present in only 10-15% of the population (hardly the average woman). In epidemiology-speak, the “referent” was this low density group, not the average woman. Indeed, the “average” density patient is at 2-fold risk for breast cancer when compared to predominantly fatty breasts.
How can someone be “average risk” and “2-fold risk” at the same time? By switching out the denominators. That’s why they are called relative risks. The RR will change if you alter either numerator or denominator. Forget the math if it’s not your thing – if we compare women with extremely dense tissue to the average patient (not the low density group) we get a more acceptable RR of 2.0. This degree of risk is more in line with having a first-degree relative diagnosed with breast cancer at age 45, whereas an RR of 4.0 would be like having two first-degree relatives diagnosed with breast cancer at 45.
While nearly all the focus was on the imparted risk of mammographic density, pioneering radiologists interested in ultrasound saw immediate application for screening women with dense breasts. Radiologist Dr. Thomas Kolb, then at Columbia University in New York City, was certainly at the forefront and became an activist for screening ultrasound. Within a decade, at least 10 large studies, totaling 60,000 patients, had been published(2), all showing similar results – the number of additional cancers discovered after negative mammograms and negative clinical exam ranged from 2.71 per 1,000 to 4.61 per 1,000. Roughly speaking, this is a relative 50-100% improvement over mammograms alone. The Society of Breast Imaging made its recommendation for ultrasound screening accordingly.
The evidence for MRI screening began trickling in about this same time as well, so at my facility, we created a “Breast Density” brochure in the early 2000s, describing the double jeopardy of density as well as the recommendation to consider multi-modality imaging with either ultrasound or MRI. At the time we initiated this program, there were no screening guidelines for multi-modality imaging at all, so we developed a scoring system that combined risk and density for patient selection, giving equal weight to both. After all, if one is going to recommend a second tier of imaging, then it should be based on the probability that the first tier is going to fail(3) – and the most powerful predictor for first tier failure is breast density.
While the primary interest in breast density seemed fixated on the associated cancer risk rather than the hidden cancer rate, even the risk agenda failed to generate momentum. To this day, one has to seek out maverick modifications of our mathematical models (e.g., Tice modification of the Gail model) in order to incorporate density levels into formal risk assessment. This, after hundreds of articles have confirmed the relationship of density and risk.
In my presentations about multi-modality screening, primarily MRI, I would offer the overwhelming evidence that mammographic density is a risk factor, then call it, “The Rodney Dangerfield of Risk Factors.” It doesn’t get any respect. Yes, it’s talked about all the time, but try to work the density level into your formal risk assessment program, and you’ll need to have those maverick models at your disposal. (Addendum: After publication of this book, in 2017, the Tyrer-Cuzick model released version 8.0 that included breast density in calculating breast cancer risk.)
In 2009, a study published in the Journal of the National Cancer Institute(4) described a meta-analysis of 47 studies of breast density related to breast cancer risk, involving 28,521 cancer patients and 3 different ways to categorize density levels – all 3 methods showed the same thing – a 4-fold risk for breast cancer when the top category is compared to the bottom category, and approximately a 2-fold risk when compared to the average patient. Once we add the possibility of invisible (or “missed”) cancers on top of the risk problem, the Rodney Dangerfield reference was really an understatement. Everything has changed now, with heightened awareness of density, not via 47 studies combined into one meta-analysis, but through one woman trying to reach 50 states.
The double jeopardy concept seems to trip up even the experts. When the American Cancer Society issued their 2007 guidelines for breast MRI screening, they treated mammographic density as a risk factor, placing it in the “insufficient evidence” category along with “15-20% lifetime risk” and other modest risk factors. Density was not considered as a determinant of “first tier” sensitivity for women, that is, an independent predictor of mammographic failure. Fortunately, the recommendation for adding ultrasound to screening high-density women was adopted by the Society of Breast Imaging, but then again, unfortunately, experts are no longer the voice of authority. The Society of Breast Imaging guidelines have gone largely unheeded, while we wait for “neutral epidemiologists” to weigh in.
No one has attempted an ultrasound screening trial where the endpoint is mortality reduction, given all the problems we’ve seen already with mammography, not the least of which is obsolete technology by the time the requisite follow-up is complete. Instead, we have relied on prospective trials (non-randomized) where participants undergo multi-modality imaging, and then cancer yields are recorded for each method individually – mammography, ultrasound and MRI – and in combination. Improved cancer yields are reflective of a mortality reduction, though purists will be quick to point out that we can’t rule out the impact of the Big Four biases.
We have very good offerings for imaging beyond mammography available now, barely utilized, though with much improved outcomes:
Mammographic density and no other risk factors = ultrasound
Mammographic density plus additional risks = ultrasound or MRI (perhaps alternating)
Women at very high risk, regardless of density = MRI
Nancy Cappello saw the problem shortly after her own diagnosis. Breast density is double jeopardy. It is a risk factor, and it is a predictor of mammographic failure.
Several questions often arise with regard to breast density. First, “How did I get it?” And, “what can I do about it?”
Baseline breast density is a product of your genes, though environmental influences do cause alterations. Interestingly, some of these alterations are associated with a similar change in breast cancer risk, raising the question as to whether or not density can be used as a surrogate measure for risk-reducing strategies. Each pregnancy lowers breast density a bit, and breast-feeding does as well, both known to be protective factors when occurring in the younger age groups.
Other factors don’t fit so well. Body fat (BMI = body mass index) is a confounding variable that doesn’t always match up with breast density risk. In fact, for many years, it slowed down acceptance of breast density as a risk factor. Ethnicity doesn’t always match either. Asian women, in general, have a much higher density level than African-American women, yet the breast cancer risks are higher for African-Americans. Postmenopausal hormone replacement therapy, especially estrogen and progesterone, can result in an increase in mammographic density. And, the use of SERMs (Selective Estrogen Receptor Modulators – tamoxifen or raloxifene) can lower breast density, perhaps reflecting when they are also preventing breast cancer.
Taking “statins” to prevent cardiovascular events and death is widely accepted, even though, as with any preventive health measure, it is the minority who benefit. Yet, the surrogates – blood lipid levels – have become endpoints unto themselves. “We’ve successfully treated your hyperlipidemia” ignores the fact that the goal is something else entirely – that is, reducing the probability of cardiovascular events and death. The lack of such a surrogate may be one of the reasons why so few women accept a recommendation to take the SERMS, which are FDA-approved to reduce the risk of breast cancer. Fairly good evidence suggests that the same women who have a lowering of their breast density pattern while on SERM therapy are the same ones benefitting most from the drug. Perhaps someday, breast density will become an official surrogate, and pharmacologic risk reduction will enjoy greater popularity.
A common misconception is that “young women have dense tissue, while older women do not,” thus explaining the superior efficacy of mammography in older women. While these differences may be true in general, exceptions are quite common. Some women who start screening in their 30s will have low density mammograms, and we see 80 year-olds with mammograms where we can’t see a thing. Some women gradually become less dense after menopause, others do not, especially if they take estrogen-plus-progesterone hormone replacement therapy. And, there is no sharp loss of density at menopause. In fact, if you look at density in the 40s as a whole and compare it to density in the 50s, there is very little difference. The point is that breast density is a highly individual situation. I’ve included an assessment of density in my risk assessment program for 20 years. My reasoning for this is based on the double jeopardy issue, with a determination of the individual’s estimated sensitivity for breast cancer detection should it occur, a separate issue from density as a risk factor.
Now, one quibble with the Are You Dense? educational efforts. As often happens in medicine, we are confronted by continuums for which we must create artificial (and subjective) classification schemes. Are You Dense? has opted to use the dichotomy approach – two groups, dense and non-dense, with 50% density as the dividing line, an approach used in some clinical trials as well. The problem here is that the 50% point is where subjectivity is at its worst.
Picture a bell curve, which is what we nearly have in breast density, from 0 to 100%, with most women bunched in the middle. Then, you draw a line straight down from the peak of that curve, and you have the majority of women clustered right at the dividing line. Not only do radiologists routinely differ at this point in the dividing line, studies have shown that the same radiologist will call the density level different from year-to-year in the same patient, even when the density level is unchanged. This is no one’s fault – it’s simply the nature of subjectivity applied in a quantitative fashion to a phenomenon that has a strong qualitative feature as well. In other words, it’s not merely how much density is present, but what is the nature of that density – small patches of white? Large patches of white? Diffuse haziness? Net-like strings of white? Net-like strings of white interconnecting small white patches with a diffusely hazy background?
Once breast density became an accepted risk factor and predictor of mammographic failure, an entire industry arose in order to quantify it in a meaningful fashion, using software that spits out exact percentages of density, a number that can be translated to both “risk levels” and “sensitivity levels.” The resilient Dr. Dan Kopans, surfacing again on this issue, has spent considerable effort trying to educate the world about the subjectivity of breast density, pointing out the extreme complexity that underscores all attempts to simplify the problem, that is, the qualitative issues are every bit as important as the quantitative.
For the past 20 years, the American College of Radiology has required that interpreters of mammography describe the degree of breast density, dividing into 4 categories. The definition for each category has changed slightly over time, but what used to be Levels 1-4 are now called Levels A-D, generally based on quartiles of density: 0-25%, 25-50%, 50-75%, 75%-100%, though with some modifiers. Unfortunately, these percentages have not until recently been listed in a straightforward fashion on the radiology report. Instead, radiologists were directed to dictate “in code” using a formal lexicon. For instance, “scattered fibroglandular densities” was code for a Level 2 (now called Level B) mammogram, or 25-50%. Interestingly, only Level A does not include a disclaimer about reduced sensitivity for the detection of cancer. Levels B, C and D each have different wording that describes increasing concern about lowered sensitivity with increasing density. Primary care physicians grew immune to the redundant terminology that was actually “code,” but you can imagine Nancy Cappello’s shock to discover that this density information rarely made it to the patient.
In my practice, as I am reviewing the mammograms for the double jeopardy, I attempt to incorporate the quality of density as well as the quantity, the latter having already been addressed by the radiologist with A, B, C and D levels. In my mind’s eye, I picture a small invasive cancer, about 1.0 to 1.5cm, and I move that imaginary cancer around both the MLO views and the CC views while asking myself a simple question: Is there anywhere this (imaginary) cancer could hide? Invasive lobular carcinoma can prove problematic to this approach, as the diffuse growth pattern in some lobulars allows them to hide anywhere they want. This approach is still subjective, but it’s not a false dichotomy. Generally speaking, the overall percentage of white reflects the imparted risk level, while the qualitative pattern of the white allows me to predict the probability that a small cancer will be detected (Sensitivity).
Importantly, this “roving” 1.0 to 1.5cm (imaginary) tumor approach to mammographic review is based on the premise from the previous chapter that overall density pattern is only an indirect predictor of invisible cancers – the real problem is the density immediately adjacent to the tumor that could completely encase the cancer. This is why even Level B density (scattered fibroglandular densities, or 25-50% dense) was originally accompanied by a disclaimer on radiology reports that sensitivity could be compromised, while the dichotomy approach would call these patients “non-dense.” When you add the possibility that the origin of cancer might be primarily within these patches, it can render some uncomfortable conclusions about mammography used alone.
I would be remiss if I didn’t briefly mention the controversial entity of DCIS (ductal carcinoma in site) that is a product of screening mammography, usually presenting in the form of a calcium cluster, that is, tiny white dots clustered into a group on X-ray. Calcium appears on mammography better than ultrasound or even MRI, so mammography has held the top spot in the hierarchy of screening. But now that DCIS is dropping in popularity as a worthy goal, ultrasound has gained momentum in that a higher percentage of ultrasound-discovered cancers are small invasive cancers, rather than DCIS. To be specific, mammography-discovered cancers will be DCIS about 25-40% of the time, while ultrasound-discoveries are DCIS only 10-20% of the time. So, some observers claim superiority of ultrasound over mammography in the very high density groups, given one or the other. This introduces the speculative, but attractive, notion of screening with ultrasound alone in selected patients.
Certainly, mammograms alone in this highest density group (greater than 75%) have failed to deliver. I once encountered a breast center that had a disclaimer at the bottom of its reports: “4% to 8% of breast cancers are not visible on mammography.” I did not make any friends when I openly stated that the only way to make that statement true in a patient with extreme breast density was to add a zero to both numbers, that is “40% to 80% of breast cancers are not visible.” People thought I was kidding, or at least exaggerating. I was not. Years later, when DMIST demonstrated the sub-group of film screen technique in young women with dense breasts, and only 27% Sensitivity (a 73% miss rate), I was vindicated, although still not popular.
With the great advantage of lower cost and greater patient comfort when compared to breast MRI, whole breast screening ultrasound has some distinct advantages. The drawback is specificity, with more benign biopsies (note: I do not use the pejorative term, “unnecessary biopsies”) than occur with mammography or MRI. Most studies show that an ultrasound-generated biopsy will be malignant only 5-10% of the time (compared to 20% for mammography and 30-40% for MRI).
Furthermore, there is a bigger difference in technique and interpretations with ultrasound than mammography, very much related to experience and skill. To that end, a new development (albeit in research status for 40 years) is “automated whole breast ultrasound,” which gives a standardized picture of the entire breast that should be very desirable for the screening setting. While targeted, handheld ultrasound will still be better for diagnostic problems, the automated whole breast ultrasound approach should be a great addition for the screening tools available for women at average risk, but with dense mammograms. That said, preliminary data suggests that a physician using hand-held ultrasound will detect more cancers than the automated approach.
Although single-center studies and multi-center studies have revealed that screening ultrasound increases cancer yields by 3-4 per 1,000 above the 5-7 cancers found by mammography, I’m only going to review one of the most important studies – ACRIN 6666 (5).
The American College of Radiology Imaging Network designed a trial to study ultrasound as a complement to mammography in high-risk, high density patients. Notice that participants had to have both – traditional risk factors in addition to the inherent risk of breast density. Although there is not a control group in a study like this for cancer yields (a no imaging at all group), care was taken to account for as many variables as possible. For instance, patients were randomized as to which study was performed first (mammography or ultrasound), and the radiologists were blinded as to the results of one study when interpreting the other.
Kudos to the trial designers who addressed every possible criticism that I can dream up concerning current multi-modality guidelines. First, the study designers realized the age-discrimination inherent when lifetime risks are used as a sole criterion. The 60-year-old patient with risk factors, and at peak short-term incidence of breast cancer, often won’t qualify for MRI because of fewer remaining years in her lifetime. So, the ACRIN 6666 team came up with short-term risk calculations, as well as standard long-term calculations. Then, designers also realized the “qualitative” problem of density, allowing certain patients to qualify for the trial if there were large patches of white even though overall density was less than 50%. And then, one of the most subtle, but perceptive, entry requirements allowed an adjustment in traditional risk requirements based on density level, that is, the extremely dense patient did not need as many traditional risk factors to qualify, and conversely, the lower density patients needed higher calculated risk. I can’t say it made any difference in the outcome, but I will offer personal testimony to this – these entry requirements were so sophisticated in design that they could be used for all forms of multi-modality imaging. Enough praise to Dr. Wendie Berg and her team, moving on to results.
From 2004 to 2006, 2,662 women (with both density and traditional risks) at 21 sites underwent 3 rounds of double screening with mammograms and bilateral, whole breast ultrasound, at 0, 12 and 24 months. From the boatloads of data generated, let’s go with this:
33 cancers were detected by mammography alone, 32 detected by ultrasound alone, and 26 were seen on both mammography and US. Clearly, the modalities are detecting different aspects of cancer, as less than a third of the cancers were seen on both modalities. This suggests a powerful complementary role to ultrasound. In fact, ultrasound appears equal to mammography, if picking one or the other.
A breakdown of tumor size shows that mammography detected tumors with a mean size of 1.15cm whereas US detected tumors with a mean of 1.0cm. This close comparison in size indicates that US is not finding cancers much earlier than mammography, it is capturing the cancers that ought to be large enough to be seen on mammography, but were missed, presumably due to density. Yet, there is an important difference: 75% of the cancers were invasive by ultrasound, while only 52% were invasive by mammography. Because these ultrasound-discovered invasive cancers were small and usually node-negative (4% with positive nodes, compared to 33% node positivity for mammograms alone), ultrasound could again be declared the winner in head-to-head competition. That said, ACRIN 6666 was not looking for a “winner” when it came to these two modalities.
Beyond digital mammography, ultrasound generated 5.3 cancers/1,000 the first year and 3.7 per 1000 in each of the second and third screens. What about sensitivity, documenting the benefit from a different angle? Discounting the first screen (as a prevalence screen), the next two incidence screens revealed a sensitivity of 76%…combined! That’s right. Both mammography and ultrasound together missed 24% of cancers. Sensitivity of mammography alone was 52%. Granted, this is not your average patient population, but “risk levels” have no influence on sensitivity levels. Density, on the other hand, is the primary determinant of sensitivity.
But that’s not where the story ends. ACRIN 6666 designed a sub-study, where women could undergo a single breast MRI at the conclusion of the 3 screens with mammography and ultrasound. Only 612 women of the 2,662 moved ahead with this aspect of the trial, so a separate set of statistics was created for this group. An additional 9 cancers were discovered on the single MRI (8/9 invasive with an average size of 0.85 cm, all node negative). Converted to our 1,000 standard, this was an additional cancer yield of 14.7 per 1,000. Had all 2,662 women participated, an extrapolation would indicate that 39 additional cancers would have been detected, in a group that had been cleared as “good to go and cancer-free” with 3 sets of negative mammograms and 3 sets of negative ultrasounds.
The smaller tumor size and negative nodes with MRI is a little more impressive than the difference between ultrasound and X-ray, now comparing 0.85cm with MRI to the 1.15cm of mammography. So, MRI has lowered the threshold of detection, which automatically hurts the other two forms of imaging when it comes to sensitivity. To the point, when sensitivity for mammography and ultrasound are now recalculated while including the MRI-detected cancers, then the sensitivity for mammography and ultrasound combined was only 44%. A 56% miss rate using clinical exam, mammography and ultrasound together three times over the course of 24 months!
These study results are both sobering and confusing, if those two descriptors can fit in the same sentence. As a result, the 2012 headlines surrounding this landmark study were mixed, as though the investigators interviewed were baffled by their own data, leaving the journalists bewildered, yet eager as ever to report.
Consequently, some headlines pronounced ACRIN 6666 a clear victory for screening ultrasound, while others proclaimed the superiority of breast MRI. But one headline was never used and never considered: “Mammograms alone are good enough.”
Wendie Berg, MD, PhD, breast radiologist, was the Study Chair and Principal Investigator for ACRIN 6666. In January 2014, two years after the trial results had been released, she underwent a digital mammogram with 3-D tomosynthesis, which showed Level C density (50-75%), but no cancer. Because of her family history, she decided to proceed with adjunct breast imaging, and in April 2015, she publicly announced that a 0.9cm invasive carcinoma had been discovered using breast MRI. A new activist group was formed – DENSE (Density Education National Survivors’ Effort) (6).
- Wolfe JN. Breast patterns as an index of risk for developing breast cancer. AJR 1976; 126:1130-1139. Dr. John Wolfe (1923-1993) was a professor of radiology at Wayne State University School of Medicine in Detroit when he published his landmark paper. His N1 pattern today would be called “predominantly fatty” (estimated lifetime risk for breast cancer – 2%), with the P-1 pattern being less than 25% prominent ducts, P-2 being greater than 25% prominent ducts, and DY (dysplastic) for dense fibro-glandular tissue. Acknowledgement for the bad pun goes to Karla Kerlikowske, MD who wrote a New England Journal of Medicine editorial in 2007, titled, “The Mammogram That Cried Wolfe.” I should have known a pun so obvious would not be original on my part, but in truth, I found Dr. Kerlikowske’s editorial in my files after I had written this chapter. So, it probably was not an original thought on my part, but instead, a subliminal repository from which I drew.
- While all pioneering authors on breast ultrasound screening deserve mention, space limits us to the first authors on papers that led to the Society of Breast Imaging recommendations: Paula Gordon, Thomas M. Kolb, W. Buchberger, Stuart S. Kaplan, Isabelle Leconte, Pavel Crystal, V. Corsetti, W. Berg (see #5), and Kevin Kelly. Dr. Thomas Stavros played a key role in the development of breast ultrasound, as well as many others.
- Hollingsworth AB, Stough RG. Breast MRI screening for high-risk patients. Semin Breast Dis 2008; 11:67-75. In this article, we gave our initial experience with MRI screening, using a point system that selected patients for auxiliary imaging equally weighted for risk and density. The scoring system also delineated the MRI interval, that is, MRI performed annually, every 2 years, or every 3 years. While this was intended for MRI, the principles are the same for ultrasound. Our preliminary findings were unsettling in that our MRI-discovered cancers were almost entirely in patients who would later prove not to qualify for MRI based on American Cancer Society guidelines introduced in 2007.
- Cummings SR, Tice JA, Bauer S, et al. Prevention of breast cancer in postmenopausal women: approaches to estimating and reducing risk. J Natl Cancer Inst 2009; 101: 384-398.
- Berg WA, Zhang Z, Lehrer D, et al. for the ACRIN 6666 Investigators. Detection of breast cancer with addition of annual screening ultrasound or a single screening MRI to mammography in women with elevated cancer risk. JAMA 2012; 307:1394-1404.
- Dr. Berg joined with JoAnn Pushkin and Cindy Henke-Sarmento in forming DENSE, their educational website called DenseBreast-info.org.