Forty years ago, mammograms detected 90% of breast cancers. Or so we were told. Oddly enough, the more that the technology improved, the worse that mammograms performed. That is, the data on sensitivity started falling apart under scrutiny. The 90% became 80%, which more recently added this modifier: “…but with dense breasts, it’s more like 50%.”
Those of us practicing medicine in the dense breast space, or more accurately, the dense breast wilderness, have long known that the dense breast problem is much worse than both clinicians and patients have been willing to admit. In fact, for Level D density, 50% is a gift. Level D gets to hide behind the skirts of Level C, and when both groups are managed as one (the usual situation), then yes, 50% sensitivity works pretty well. But isolate Level D, and…ugh. The 10% of women with Level D have not been singled out for more specific information about the reliability of their mammograms when used alone.
At least, not until now. We’ve been eagerly awaiting the initial results of the prospective, randomized DENSE Trial from The Netherlands for several years. This is a unique study using breast MRI in women with Level D density. Very straightforward patient accrual strategy – Level D and you’re in. The study will be generating data for the next 10 or more years (including the stringent endpoint of a mortality reduction, the only endpoint recognized by the U.S. Preventive Services Task Force in breast cancer screening).
The interesting thing about these trials is that you can predict results fairly well with some basic information: Sensitivity of mammography in Level D patients, Sensitivity of MRI in Level D patients, and disease prevalence in the study population. For longer studies (DENSE is only 2 years of screening), you also need to know disease incidence. In the case of disease prevalence and incidence, one must remember that Level D patients are at a modestly increased risk for breast cancer, such that we face a double whammy – higher probability of cancer and higher probability that cancer will be missed if screened only with mammography.
It would have been nice to include Ultrasound as an alternative to MRI, but a 3-limb study requires many more patients and much more money. Nevertheless, it has been my view for a long time that for Level D patients (what we used to call Level 4, and what was originally called “extremely dense”) mammography alone is inadequate for screening and that supplemental imaging is mandatory for acceptable practice.
The current grassroots movement on breast density awareness would have done well to avoid the dichotomy of “dense vs. non-dense.” They should have come out swinging, stating: “Supplemental imaging is mandatory for Level D patients, and is strongly advised for Level C.” The Level D data would have poured forth freely, and as a result, it would have become obvious that mammograms are not enough. Gradually, this would have spilled over to Level C patients, or at least, selective Level Cs, and we would be where we need to be. As it is, we have a staggering amount of legislation that is largely edentulous because insurance won’t cover the supplemental imaging in most states.
It doesn’t help that the Task Force has ruled on breast density in one of the most ludicrous policies they’ve ever generated, that is – No evidence to support imaging the dense breast with 3-D mammography, ultrasound, MRI “or any other modality” beyond 2D digital mammography. It doesn’t matter to them that if one ranks all known breast imaging modalities according to sensitivity levels in dense breasts, mammography would come in dead last. For the Task Force, no mortality reduction equals no endorsement when it comes to breast cancer screening. Though the DENSE trial won’t matter to the U.S.P.S.T.F. until the mortality data is available, at least we have the preliminary results for other endpoints.
Traditional risk factors didn’t matter one way or the other as the sole criterion for entry was Level D status. Kudos to the Netherlands for their design. All prior MRI screening trials have been based on risk levels (without a care in the world as to breast density). It could be claimed that the well-designed ACRIN 6666 took density into account, but not without traditional risk factors, and with its focus on ultrasound, while MRI was secondary.
To me, it’s always been a two-factor problem – the level of risk is only half the story. The other half of the story when discussing a second modality for imaging should be: “What’s the chance that the first modality will fail?” It is inconceivable how this second issue has been cast aside. It all started with the original 2007 guidelines for MRI screening from the American Cancer Society. There, breast density was treated like an isolated risk factor that “needed more research,” when, in fact, density should figure into the equation for recommending supplemental imaging in every single patient, high risk or not. Density dictates efficacy. And density is a feature that should be integral in our recommendations for supplemental imaging.
So what do we know from the DENSE trial preliminary data, published this past week in the New England Journal of Medicine https://www.nejm.org/doi/full/10.1056/NEJMoa1903986
Interval cancers were cut in half (from 5 per 1,000 to 2.5 per 1,000). This is a major point (in fact, it’s the primary endpoint in the study) because screening alone can find low-grade cancers which evoke accusations of overdiagnosis. If you are detecting clinically important cancers, then the number of cancers popping up in between screens must go down. Later, you also need to show fewer advanced stage cancers in the MRI-screened group. If you detect more cancers with MRI, but the interval rate stays the same and the advanced cancer rate stays the same, then you are likely detecting such low grade tumors that the patient would have done just fine waiting until her tumor was found on mammography or palpation. In the end, however, both interval cancer rates and advanced disease rates are surrogates for the top priority — mortality reduction.
The cancer detection rate (CDR) in the mammography plus MRI group was 16.5 per 1,000 screened, approximately 4-fold the number ordinarily found with mammographic screening alone. So, if mammograms are detecting 90% of cancers, then MRI is detecting 360% of cancers? Of course not, but this is the incongruity we’ve been living with for years.
About 15 years ago, TIME magazine ran a spread on breast MRI screening, and within the article, it was clear that mammography was only detecting 40% of detectable cancers when compared to MRI. That’s 40% sensitivity. Then, in a sidebar about mammograms in general, the comment was made: “Mammograms detect 90% of breast cancers.” Where was the cognitive dissonance? I can’t figure it out. The dissonance didn’t seem to bother anyone! There was apparently no cognition.
Pro-screeners have shot themselves in the foot on discussing sensitivity across all levels of density – the only way they can promote much-needed multi-modality imaging is to admit that, for many years, they have overstated the effective sensitivity of mammograms alone. And if we combine Level C and D, then we’re talking about misinformation for nearly one-half of the female population who opt for mammographic screening!
For the record, 3D mammography has been a much-needed leap in mammographic sensitivity, but still, ultrasound will detect the 3D discoveries in most all cases, plus more cancers beyond that, with MRI detecting even more. (Contrast-enhanced mammography is another matter, and beyond my scope here.) Nevertheless, the time has come for supplemental imaging to be routine for the high-density patient and “standard” for Level D patients, reinforced strongly with the prospective, randomized DENSE trial results – not to mention the powerful data we knew already about ultrasound screening and MRI screening from earlier non-randomized studies.
PS — I’ve been singing this same song for so many years it’s really sort of pitiful. But I did get all my thoughts collected in a magnum opus and published in the American Journal of Surgery recently. Here’s the pub med link to the abstract https://www.ncbi.nlm.nih.gov/pubmed/30739738 or you can write me for a copy.