The American Society of Breast Surgeons announced a new policy statement at its annual meeting in April 2017, regarding the use of breast MRI pre-operatively for women newly diagnosed with breast cancer. Unlike most guidelines that describe what physicians ought to do, this one is quite the opposite – what NOT to do. That is, don’t routinely perform breast MRI pre-operatively to map tumor extent.
The operative word here is “routinely,” with the implication being that surgeons are free to order MRI in selected cases, thus avoiding the more oppressive guideline that has been suggested by some – don’t use MRI at all. This new policy statement sounds perfectly reasonable at first glance. But at second glance, we’re facing an unusual problem – no one has clearly demonstrated specific subsets of patients where MRI yields are higher, that is, more easily justified. We all believe we know where MRI is most likely to help, but so far, attempts to demonstrate these subsets – based on patient age, histology, breast density, etc. – have not shown sharp distinctions in MRI yields or outcomes. It is very difficult to predict in advance when MRI is going to be of benefit.
Once upon a time, in a faraway medical world, there was an ethereal entity called a “Center of Excellence.” Physicians aspiring to this status would band together and outline a strategy of superior performance, publish results, and then allow the medical community to analyze these outcomes and adopt the protocols if true excellence was demonstrated. As a result, quality of care improved accordingly. Today, this approach has been replaced with Guideline Medicine (allegedly evidence-based medicine – however, with enormous gaps in our facts, guidelines are often a matter of opinion). There is a subtle distinction between these two approaches. Guideline medicine tends to bring the Average Joe up to a basic standard. As for rising above that standard…not so much. Instead, guidelines have the potential to fall into the trap of the “tall poppy syndrome,” that is, cutting the heads off more vigorous poppies to make sure that all are the same height.
One other oddity about Guideline Medicine is unique to certain disciplines, especially radiology and surgery – one can have vastly different outcomes in studies due to variations in interpretive skills and surgical technique. Prospective, randomized trials work wonderfully when evaluating a new drug because the manufacturing process of that drug is standardized, and the participants (patient, physician and sponsor) are blinded. Not so for radiology and surgery. Imagine what clinical trials would be like for new chemotherapeutic agents if each participating hospital were responsible for manufacturing their own drug, and if all participants were unblinded.
My point is that the words “prospective, randomized trial” in surgery and radiology do not have the depth of validity as when studying a drug in double or triple-blinded prospective RCTs. This is especially true when it comes to the technology behind MRI and the very difficult interpretive demands. After the introduction of breast MRI to the clinical sphere, it took no time for breast MRI experts to warn us all: “Bad MRI is worse than no MRI at all when it comes to the breast.” This warning went largely unheeded, and the introduction of breast MRI was every bit as chaotic as the introduction of mammography (which prompted government intervention and FDA monitoring that has been in place now for nearly 25 years).
In the early days of breast MRI, experts routinely refused to accept outside studies, and any patient referred for biopsy was re-evaluated with another MRI where biopsy was often not indicated. At our facility with breast-dedicated MRI (Aurora Imaging), these repeat MRIs were often normal. Yet, during this era (which is not completely behind us), the surgical literature was flooded with outcomes – both good and bad – based on substandard MRI. Some of these oft-cited studies include patients who underwent research MRI 25 years ago, a full decade before its clinical introduction.
Back to the recent meeting of the ASBrS. Through coincidence, I happened to be speaking on the topic of pre-op MRI at this same meeting where the new MRI policy was announced. My talk was delivered in one of the pre-meeting courses devoted to Advanced Imaging. In my presentation, I showed 5 examples of FAKE NEWS and ALTERNATIVE FACTS about pre-op breast MRI. Time constraints, however, limited me to a discussion of only 2 of the 5. As occurs whenever I give this talk, attendees tell me they are surprised by what these “landmark” studies actually reveal, or fail to reveal. Other than the obvious problems of terrible interpretations or “research” MRI in its early years (or unilateral MRI performed in one study often quoted for its mysterious lack of contralateral benefit), listeners to my analyses usually agree that these studies are nothing more than a string of caveats held together by gaping holes.
That doesn’t mean I have unequivocal proof of outcome benefit with pre-op MRI. It only means that the criticism of pre-op MRI from commonly quoted available studies is severely limited, and you can’t overcome this with the revered “meta-analysis,’ which is merely an example of “garbage in – garbage out.”
Nevertheless, two days after my presentation, I heard the justification for the society’s new policy on MRI. The 2 studies quoted were the very 2 I had analyzed in my FAKE NEWS talk days earlier, but with the usual lack of in-depth analysis. We were told that pre-op MRI “causes more mastectomies, doesn’t alter re-excision rates, delays patient care, etc., etc.” Sadly, this is true if one takes certain anti-MRI literature at face value, ignoring those of us who have published in-depth critiques of these studies, while at the same time offering our different outcomes at “centers of excellence,” that today are being converted to “level poppy fields.” At my facility in OKC, there is no delay at all with pre-op MRI, there is no increase in mastectomies and, improbably, the conservation rate is even higher after false-positive findings on MRI!
Our surprising (and counter-intuitive) results come from the fact that all newly diagnosed patients are told by the radiologist and the nurse navigator that the purpose of the MRI is to help confirm candidacy for conservation. This simple introduction is a game-changer. Because the breast-dedicated MRI is located in the breast center (and not used for any other purpose), we are able to perform the MRI within days of diagnosis when the patient’s head is still spinning, and therapeutic options have not even been discussed yet. Post MRI work-up is completed quickly and before presentation at interdisciplinary conference. This approach is contrasted to the patient who is presented at a conference without MRI, and then the surgeon decides later if she needs an MRI, whereupon the study’s pivotal role is greatly magnified. Here, false-positives cause anxiety which, in turn, result in mastectomy.
But using our approach of “up front” MRI, the false-positives are handled so quickly that the opposite is true – the patient who was leaning toward mastectomy sees how much effort went in to proving conservation safety, and she actually switches her choice from mastectomy to conservation. As we published nearly 10 years ago (Am J Surg 2008; 196:389-397) in what was the largest series of pre-op MRI at the time (n=603), our overall conservation rate was 60%, but when we had a false-positive MRI, the conservation rate was 70%! And for those who criticize our work at Mercy-OKC because we have no control group with routine consecutive use – yes, there are certain claims we cannot make with regard to the impact of MRI. However, there are inter-group comparisons that are extremely helpful with regard to cancer yields and outcomes, as exemplified by this remarkably paradoxical 70% that goes against what the vast majority believe about breast MRI. And that result comes with ZERO selection bias, by virtue of our policy of routine pre-op MRI. (Note: the definition of a false-positive was a mere callback, with or without biopsy. For those patients who went on to a biopsy that proved benign, the conservation rate was a surprising 86%, although now we have selection bias in favor of patients strongly committed to conservation.)
No published study suffers more from excessive selection bias than the oft-quoted COMICE Trial (one of the 2 fake news studies I mentioned). While I’ve heard speakers at the podium describe this as a trial of “all comers,” nothing could be further from the truth. Ignore the fact that MRI was poor quality, interpreters were not expert, biopsies of enhancements were often not performed (sending patients straight to mastectomy), and surgeons were barely engaged – yes, ignore all that – the mere study design in Materials and Methods is bizarre. First of all, patients were selected as possible candidates for conservation and referred to study coordinators, but then, a staggering 70% were excluded from the study. Even though the vast majority of women are candidates for conservation, somehow 70% still didn’t make the cut. The breakdown goes like this: 1,360 did not meet inclusion criteria, 1,173 refused to participate, and a remarkable 1,338 were excluded for “other reasons.” What other reasons? Incredible. When the great majority of women qualify for conservation, why were these women excluded? In all, 3,871 patients were excluded from the trial, while enrolling 1,623, and by the time we get to the number that underwent MRI, 816 were analyzed (not many more than our local experience in OKC at the time). This 70% exclusion rate (after referral) is given startling perspective when compared to the landmark NSABP B-06 that proved the equivalency of lumpectomy, wherein a mere 12% were excluded. Wow.
Does 70% exclusion make sense? No. But understanding why the COMICE trial was organized in the first place might shed some light. This was a trial in the U.K. where the National Health Service was growing weary of the high re-excision rates after lumpectomy. They wanted to see if they could get this rate below 10%. This was the ultimate goal of the COMICE Trial. Is it possible that the extraordinary exclusion rate was to identify patients with a very low probability of re-excision, that is, trying to get as close to 10% as possible without MRI, and then to see if MRI could push the number below 10%? I don’t know the answer to that question, but if so, it stacks the deck heavily against any MRI benefit through the selection of patients with very low re-excision potential. My point is that COMICE was not a trial of “all comers.” Both groups had a 19% re-operation rate, whereupon one of the principal investigators announced to the media, “This ends the debate. MRI offers no benefit,” as if there were only a single endpoint in the universe.
Then, critics of MRI jumped on board by announcing to their shock and dismay: “MRI caused more mastectomies in the COMICE trial.” Again, wow. 100% of patients entered the trial as conservation candidates. Possible outcomes of MRI were unidirectional. That is, there was no other outcome measured for MRI – it was impossible to look at the impact of MRI in converting a patient headed toward mastectomy to breast conservation instead. COMICE re-created the exact scenario I described above wherein false-positives cause anxiety which cause mastectomy, by putting the MRI at the very end of the process and magnifying its role. This is in sharp contrast to how MRI can be used in the newly diagnosed patient, up front and before options are discussed.
(Let me give a nod here to my radiology partner, Dr. Rebecca Stough, with whom I’ve co-directed the MRI program – surgeon + radiologist – since its inception. Note the word “program” where we have multiple quality measures in place that are still not generally considered routine. It was Dr. Stough who first hypothesized the “up front, radiologist-directed MRI” to explain why our data was different than other reports.)
Is Mercy-OKC the only poppy with its head chopped off? Hardly. Recognizing that COMICE bore no resemblance to real world experience (and the prospective RCT called the MONET trial had its own serious limitations), EIBIR-EuroAIM/European Society of Breast Imaging designed a trial to look at the overall impact of MRI in BOTH directions – conservation-to-mastectomy and mastectomy-to-conservation. Furthermore, they decided to minimize selection bias by including “all comers” (this time for real) in a study of consecutive patients, some getting MRI, some not. Prospective, but not randomized. “Lower level evidence,” the pundits will say, but the real world functions at this lower level whether we like it or not. Furthermore, to put the critical mind at ease, one of the harshest critics of pre-op MRI anywhere on the planet – public health expert Nehmat Houssami, PhD – was on the Steering Committee. So let’s give a little credence to this study where results were announced in the month prior to the new policy statement by the American Society of Breast Surgeons.
Although not yet in print, well-known expert, Dr. Franceso Sardanelli of the University of Milan presented the data in Vienna (European Congress of Radiology) on March 1, 2017. Known as MIPA – Multicenter International Prospective Meta-analysis of Individual Woman Data — this study of 2,425 patients is old school – rather than using the Average Joe for interpretations, they followed the “Center of Excellence” pathway, selecting 34 centers in 14 countries where MRI expertise was in place. Of note, the U.S. site was in Northwest Arkansas under the direction of noted MRI expert, Dr. Steven Harms (who had been our mentor at Mercy-OKC). Too, this was a study of modern MRI (2012-2016), as opposed to primordial MRI done in many studies that are often quoted. Of special note, the MRI was ordered early, up front, by the radiologist in 68% of cases. The goal was to study the overall impact of MRI, multiple endpoints, and in both directions. This is not easy to do. One must know what the patient was planning prior to the MRI, which is a challenge when the MRI is ordered early in the process. Nevertheless, the study focused heavily on this Bi-directional Analysis, something that the single-focused (“get those re-excisions under 10%”) COMICE organizers had no interest in whatsoever.
Results? There was a statistically significant lower re-excision rate with MRI – 8.3% with MRI vs. 13.4% without MRI (p<0.001). Interestingly, these numbers are nearly identical to what we have been reporting for many years. Our re-excision rate has been 9% for 15 straight years, never varying more than 1% for a given year, in spite of a 100% turnover in surgical personnel. Furthermore, in the MIPA Trial, MRI did NOT cause a statistically significant increase in the number of mastectomies. Yes, 13.7% underwent more surgery (mastectomy), but as we have been squawking about relentlessly for the past 15 years, that’s only half the story – 12.7% underwent less surgery than had been planned prior to the MRI (the 1% difference not statistically significant). Ideally, those changes in management will be for the better, that is, those converted to mastectomy should have needed to do so, and those converted to conservation should have better results as well (time will tell).
So, if you thought this debate was over, I’m putting my poppy head back on, and will start crowing again about 9 endpoints in the evaluation of pre-op MRI benefit, 2 of which have never been studied – 1) the multiple re-excision rate (3 or more operations), and 2) the rebound mastectomy rate after positive margins.
But as to our original question as to whether or not we can properly select patients as to who needs a pre-op MRI, my reference is a high quality database that I’ve kept meticulously for 15 years, wherein I compare final pathology to pre-op MRI in real time (n=1,800), and I’ll confess – I can’t tell. (Some hints, yes, but this will require formal analysis).
I’m fine with those who practice breast surgery without MRI. I’ve never claimed that it should be the standard of care. In fact, Ben Anderson, MD, has made the point that, in underdeveloped countries, you can practice good conservation surgery without pre-op mammography. This should be no surprise. After all, newfangled mammography was not a requirement for entry to the B-06 trial, and the independent impact of mammography on the management of palpable cancers has never been studied to the degree we are demanding of MRI. Fisher Theory was a pre-mammography construct. The attacks against pre-op MRI based on the notion of biology trumping anatomy apply to tumor mapping with mammography as well. How can an unsuspected 1.0cm invasive cancer found in the opposite breast on pre-op mammography prompt lumpectomy, radiation and node sampling, but if that same 1.0cm tumor is found on pre-op MRI, it is deemed “subclinical?”
While I am completely comfortable with surgeons who do not use pre-op MRI, I do bristle with borderline restraint when those who are opposed to anyone having pre-op MRI say, in effect: “We can’t make it work, so you have to stop doing it.”
So how about the middle of the road – selective use for pre-op MRI, the original topic of this wordy blogatorial? I’m fine with those who opt for selective use for MRI. It’s an alluring proposition. My point is to remind everyone that, at this point in time, we are not basing selective use on published results. At present, the difference between young and old, ductal and lobular, dense and non-dense, simply doesn’t jump out of my database, nor has it been clarified in the published literature where others have failed to define the best candidates for MRI. Pre-op MRI is a decision best left to the individual physicians and/or institution.
As a post-script: Our surgeon-radiologist combo directing the MRI “program” at Mercy-OKC once wrote about our methodology as being more important than the exact technology (brand) utilized for breast MRI (Hollingsworth AB, Stough RG. Conflicting Outcomes with Preoperative Breast MRI: Differences in Technology or Methodology? Breast Diseases: a Year Book Quarterly 2010; 21: 109-112) I believe any reader of this article will be surprised at the exhaustive steps we took to avoid the inherent danger of false-positives, specifically to avoid unnecessary mastectomies. Our approach has been anything but the Average Joe.