Throughout my career, I’ve enjoyed the serendipity of Forrest Gump, strolling through history and meeting the right people at the right time. Working in a geographically isolated, non-academic community hospital, odds would have ordinarily kept me in lockdown as far as contributing anything to medical research. But to draw upon the overworked quote of Louis Pasteur, “Chance favors only the prepared mind.”
By studying two problems intensely, over the course of many years, the insight gained allowed me to enter two arenas of expertise – 1) breast imaging theory, with a focus on mammographically invisible cancers, and 2) quantification of breast cancer risk. As it turned out, the vast majority of experts were jousting elsewhere, and largely by default, I was able to claim expertise in two areas. Then, who would have predicted 25 years ago that these two agenda items would merge, in the form of “risk-based multi-modality imaging.”
Granted, some of my Forrest Gump experience was facilitated by my years in academia, but that only laid the groundwork. I did not walk onto the stage until I had been in the community setting for quite a while. I won’t name drop here (as I’ve done it excessively elsewhere), but my 3rd career evolved through the influence of many key people after I left my area of original training, i.e., general surgery. Career 1 was private practice in general surgery, Los Angeles, focusing on trauma. Career 2 was a dedicated breast surgeon beginning at my alma mater in 1989. Career 3 was a risk assessment and genetics expert (1st M.D. in Oklahoma to begin BRCA genetic testing, on Day One 1996), using multi-modality imaging based on risk levels.
Now, after 25 years of personal study, both mammographic density and risk assessment have been thrust to the forefront, and everyone has an opinion, it seems. As of 2015, all accredited cancer programs (by the Commission on Cancer) are required to provide risk assessment and genetic testing services. And, as of 2016 in Oklahoma, women are to be informed about the increased cancer risk and decreased sensitivity of their dense mammograms, according to new state legislation (see May 2016 blog). Unfortunately, there are no teeth in the legislation requiring insurers to pay for what needs to be done next.
As with most legislation, one problem is solved and many problems are created in the process. The good news is that women with extremely dense mammograms (10% of the female population where there is a near white-out) will be notified by letter of their “condition.” The bad news is that another 40% will be notified as well, when their risk and decreased sensitivity is not really that much different than the 40% one step down in density. A bell curve bisected is a false dichotomy, and those 40% of women just above the cut-off may be unduly alarmed, while the 40% below the cut-off will have a false sense of security.
If we use the 4 levels proscribed by the American College of Radiology, we have:
A = predominantly fatty replaced (10% of women)
B = scattered fibroglandular densities (40% of women)
C = heterogeneously dense tissue (40% of women)
D = extremely dense tissue (10% of women)
Again, a bell curve bisected. Nothing for A & B. “The Works” for C & D. While those at the extremes are pretty clear cut, the vast majority of women are bunched in the middle, and one radiologist might call you a B, while another would call you a C. Or, a single radiologist can call you a B one year and a C another year. (I won’t list the various bizarre scenarios that arise out of that degree of subjectivity.)
Density is far more complicated than the 4 traditional levels. In the past, quantification was tried: Level 1 = under 25% dense; Level 2 = 25-50%; Level 3 = 50-75%; Level 4 = over 75% dense. Yet, no matter how the definition has changed, radiologists still group patients into the same bell curve. The problem is that there is a strong qualitative aspect as well as a quantitative level of density, and this alone renders the pro-density activists on shaky ground.
Years ago, I began a practice that has never changed. Because our goal is to find cancer in the size range around 1.0cm, when assessing a density pattern in a new patient, I take a 1.0cm image in my head and move it around the X-ray to see how easy it would be for a cancer to hide. If there are large patches, it’s a concern, even if the overall density is less than 25%. And, the reverse is true. This is my attempt to practice so-called “precision medicine” (while the formal guidelines for screening that are promoted under this same moniker are often nothing of the sort).
The false dichotomy problem boils down to this – cancer can hide on low density mammograms. Yet now, those women who don’t get the letter are going to think they are home free. If legislators would review the screening MRI data, they wouldn’t be so quick to jump on the “dense breast legislation bandwagon,” which has now impacted 30 states (and counting). In those clinical trials where high-risk women had both mammograms and MRI performed, even the low density patients had a 50% miss rate on mammography.
High-density screening will usher in improvements through multi-modality imaging (mostly ultrasound) for those who garner a C or a D by the radiologist. But what I’m worried about is the 40% of women in the B group who are going to be tricked into thinking that mammograms are going to have a 90% detection rate. This is not true! The American College of Radiology used to admit this through their recommended reporting system that stated “sensitivity might be reduced” for women at that 2nd level. But that word of caution is no longer required. Now, only the C and D patients have this sensitivity disclaimer. In effect, this move only sharpens the distinction of the false dichotomy – A&B on one side, C&D on the other – no problem in the one group, trouble in the other. In fact, we’re dealing with a 0 to 100% continuum.
Complicating the matter are ethnic differences, where Asians have a higher mammographic density than whites, but a lower risk for breast cancer. Then, there is the obese patient who has a lower density level on mammography, but a higher risk for the development of breast cancer. This is not a straightforward issue, as usually portrayed. Even the pro-density activists are not promoting accurate information, falling into the false dichotomy trap.
In short, risk level is not the best way to determine who should do more or less screening. Density is not the best way either. A combination of “Risk and Density” (says Forrest Gump) is the best way to select patients for doing something more. As for doing less for women without risk and without density, well, it’s a tough question. Certainly, the 10% of women with “fatty replaced breasts” are going to do fine with annual mammograms and no adjunct screening modalities. But for everyone else, it’s a struggle to come up with a strategy that assures early detection.
Maybe you see now why I’ve spent more than 20 years helping scientists who are trying to develop a screening blood test to detect breast cancer, which would then prompt the need for multi-modality imaging if positive. And, why I’m working with computer scientists who are trying to develop image analysis systems that improve upon the human eye, again allowing better selection of patients for multi-modality imaging.
Things are looking up, though. The introduction of tomosynthesis (3-D mammography) is the first significant technologic advancement in the history of mammography, in that more cancers are clearly detected. Furthermore, these detections are taking place more often in dense breasts where “architectural distortions” are seen through the density by virtue of thin slices. Those patches of white, where my imaginary 1.0cm tumor can hide, are no longer so effective in camouflaging the cancer. The standard radiologist disclaimer, “it’s like looking for a snow man in a snow storm” is not as true as it once was. Tomosynthesis can sometimes see a vague outline of the snowman where 2-D mammography cannot; Ultrasound works like radar and can see the snowman by using sound waves; and MRI (or molecular imaging) lights up the snowman with the flare of contrast enhancement.
Now, if we can only figure out a way to make it all cost-effective. Odd that we really don’t need any technologic breakthroughs to find breast cancer early. The miracles of technology are already at our disposal. The only problem is trying to figure out who to put on what machine and when. And the final answer is not going to come through false dichotomies.