Guideline medicine is relatively new, that is, the practice of following published guidelines based on evidence-based medicine. These guidelines are derived from consensus panels of clinicians, with whom insurers may or may not agree; however, these third party payors feel the pressure to cover diagnosis and treatment as outlined by physicians. Thus, the strength in collective expertise allows one to practice good medicine with greater ease…usually.
But there are problems, one of which is the fact that experts don’t always agree. Two sets of eyes can look at the same data and come away with different interpretations. And, when it comes to surgical and radiologic guidelines, things can really get murky because prospective, randomized, controlled trials cannot be purely blinded. Partial blinding can be done, but these trials are never as pristine as drug trials where the dummy-pill control looks just like the real thing, and neither patient nor doctor knows who is taking what. So, when it comes to surgical technique and radiologic interpretive skills, there is an unsettling tendency toward the “tall poppy syndrome,” that is, chopping the head off excellence in the name of uniformity.
Some clinical trials make no effort whatsoever to ensure quality care. The reasoning is this: your results must be “generalizable” to the community standard because you can’t expect everyone to become an expert. Centers of Excellence have thus lost considerable ground. When pre-operative MRI, for instance, was subjected to a so-called “high quality” prospective, randomized controlled trial (COMICE), the epidemiology might have been statistically sound, but the technology was substandard, the interpretations were substandard, cooperation and communication with surgeons was poor, and the outcomes widely misunderstood. To many, the p-values were the only important feature, and pre-op MRI was widely condemned. After all, it was a “prospective, randomized controlled trial” and the mere utterance of those words render magical truth.
Yet, for those who have worked hard to achieve excellence in the use of breast MRI, our outcome data is completely different than the COMICE trial results, and we can make a very strong case that MRI should be done routinely…IF IT IS DONE WELL. But critics say if you only study MRI at centers of excellence, your outcomes will have no external validity. Okay. So should we all aspire to mediocrity? “Bad MRI is worse than no MRI at all,” is an adage we have been using ever since the introduction of breast MRI into clinical practice. Yet, the pro-MRI data is routinely ignored, while policy-makers cling to substandard MRI results. A meta-analysis was even performed where mediocrity was studied in the collective sense, somehow rising above the “garbage in, garbage out” designation.
Most clinicians are familiar with the epidemiologic biases – selection, lead time, length time and overdiagnosis. But there are minor biases as well. One of them is called “file cabinet” bias or “shelf bias,” slang for the fact that enormous amounts of data go unpublished. What is the relationship between the 1% of data that gets published versus the 99% real life data that sits on shelves or in file cabinets? For instance, I have detailed data for perhaps the largest series of consecutive, routine pre-op MRI in existence. Most centers use pre-op MRI selectively or rarely, and their data is highly skewed – that is, there’s a reason why some women get MRI and some don’t, so any attempt to compare the two groups is highly biased, with the MRI group composed of younger patients, denser breasts, lobular histology, and extensive in situ components. When “no difference” in outcomes is reported with or without MRI, it may well be that the MRI group would have had much worse results without the MRI. Maybe not. However, the mere fact that a cavalier dismissal of MRI is the conclusion of these studies, in the face of glaring selection differential, is a stunning disregard of the power of bias.
But our data at Mercy Breast Center-OKC is unique. We can compare cancer yields in sub-groups to a degree greater than anyone (to my knowledge), by virtue of the complete absence of selection bias. The downside is that we have no real-time control group, so we cannot make claims about MRI vs. no MRI. With only historical controls for reference, we cannot provide high quality data as to what would have happened without MRI. But what we surrendered on that front actually strengthened our data on another front – sub-group comparisons.
When it comes to sub-group comparisons among those women who have undergone pre-op MRI, our data is unique and invaluable. This is one reason we have persisted with routine pre-op MRI as long as we have, given the apparent absence of a comparable data bank anywhere. I’ve personally kept the extensive database for 13 years, logging in over 2,000 studies, with close comparison of final path to what the MRI predicted. Yet, the data beyond our first 603 patients remains unpublished (at this point), so it doesn’t even count. And when we published our results after 603 patients in the American Journal of Surgery (Vol 196: 389-397), we had no idea that MRI was going to have its feet held to the flames in a way that was never done for pre-op mammography (for which there is not a shred of evidence, btw, that pre-op mammograms for palpable tumors alter the outcomes demanded of MRI). As a result, we didn’t spend a lot of time in that publication addressing the sub-groups, namely age, breast density, and histology.
So, when I see guidelines that say, “Pre-op breast MRI is an option for women with dense breasts and/or lobular histology,” I have to respond: “Based on what evidence?” Even though I agree with using MRI in these instances (at a minimum), the belief that MRI cancer yields are higher in dense breasts and lobular histology is not as clear-cut as most believe. If you look at our (unpublished) data, you are drawn to this conclusion – either do MRI routinely or not at all. The sub-groups targeted by the vast majority have nearly the same cancer yields as those where MRI is thought to be unnecessary. Yes, there are differences, especially with lobular histology, but not a clear cut-off point. What irony! “Selective use of pre-op MRI” is based almost entirely on a “gut feel” without empiric back-up, and our decision to keep performing pre-op MRI routinely is based on actual data, the gold standard of evidence-based medicine. Yet, we are the ones under fire.
In the community setting, we have no resources or residents to help with the publication process, so what we’ve managed so far is done at considerable burden in the private sector. But even when we do publish, our results have been largely ignored, perhaps because they run counter to the party line. Witness what should have been a highly provocative article that we published in the use of pre-op MRI in patients newly diagnosed with DCIS. At the time, it was the largest series of DCIS and pre-op MRI ever reported (The Breast Journal 2012; 18:420-427). Furthermore, our implications involved survival differences in DCIS patients, unlike invasive disease where survival should not even be an endpoint with regard to the use of pre-op MRI.
Many believe that pre-op MRI is of limited value in invasive disease when it comes to better local management of the index lesion, and then absolutely worthless in DCIS. And, because these critics are focused entirely on the index lesion, as are nearly all publications, I cannot disagree when it comes to the known area of DCIS. DCIS is so hard to define during surgical excision that a good road map doesn’t help much, if at all. But we asked a different question entirely, based on data from MD Anderson (Dawood et al. Ann Surg Oncol 2008: 15:244-249) where they reported a large observational series of 799 DCIS patients, treated unilaterally between 1976 and 2005 (no MRI is assumed for the vast majority, if not all), with the finding that the most likely event after treatment was invasive cancer (at a rate of 3.9% after only 2.9 years), usually in the opposite breast. Even more concerning was the associated increase in disease-specific mortality following the second event. Commentators (Drs. Lagios and Silverstein) pointed out that with such short follow-up, the contralateral invasion was probably present at the time of DCIS diagnosis, but remained undiagnosed.
Our series was intended to address that possibility. We ignored the index DCIS lesion, we ignored other areas of DCIS found on MRI, and we focused entirely on “elsewhere” sites of invasion that were not connected to the known DCIS, and in fact, were either in separate quadrants or were contralateral. Our results were nearly identical to the MD Anderson numbers, with “elsewhere” invasion present in 3.5% of our 285 patients.
Occult ipsilateral invasion might be treated incidentally with whole breast radiation (untreated if partial breast radiation is used), though 4/5 patients with ipsilateral “elsewhere” invasion turned out to have Stage IIA disease. But most unsettling were the findings in the contralateral breast where there will be no treatment other than possible endocrine therapy, generally considered inadequate as sole treatment for invasive disease. And again, there was a surprising proportion of Stage IIA disease, this time in half of the patients. It is very difficult to call untreated Stage IIA breast cancer “subclinical” as was the adjective of choice of a vocal MRI critic in describing all MRI discoveries.
This 2% risk of untreated contralateral invasion introduces an intriguing scenario. What do you do with this bit of knowledge? Do you perform 98 routine MRIs to capture the 2? But if you don’t, you are in essence, performing “wrong side surgery” in 1 out of every 50 DCIS patients. Take that wrong side surgery and apply it anywhere else in the body, and you’ve got gross malpractice. But in DCIS management, you have 2% of 60,000 women every year in the U.S, or 1,200 patients, who have their DCIS treated on one side, while life-threatening cancer is left in place on the opposite side (ignoring the trend toward bilateral mastectomies). I don’t have the answer to the best approach here, but I would run from dogma that says “Don’t routinely perform pre-operative beast MRI.” 1,200 lives are at stake every year with untreated (unknown) contralateral invasive cancer, and the current standard of care says, “Don’t worry about it. Stick your head in the sand. MRI is bad enough for invasive disease and plays no role whatsoever in DCIS.”
Enter the new organization called “Choosing Wisely,” http://www.choosingwisely.org, which asks specialty societies to list 5 things that should NOT be done in their specialty. An organization to which I belong, the American Society of Breast Surgeons, fell for the trap and opted to include “routine pre-op MRI” as one of the five “Don’ts.” This is a new level of dogmatism we haven’t seen since Halsted, something that has no place in science. Rather than guidelines that suggest proper treatment, we are now slipping into different territory, that is, condemnation. The very definition of wisdom decries dogma and eschews condemnation. So, as one who has a strong handle on our own extensive data plus the published literature, it is disconcerting to hear this new school of thought, which says, “WE can’t make it work, so YOU have to stop doing it.”
Our DCIS article should have started a buzz, but it got no attention at all, and to my knowledge has never been referenced. I don’t have the answer, but 2% wrong side surgery is deeply troubling. Some of the mysterious breast cancer-specific deaths of women who have only been diagnosed with DCIS, but then show up with metastatic breast cancer (landmark article in the October 2015 issue of JAMA Oncol 1:888-896 by Narod et al) could easily be due to undiagnosed invasion in the opposite breast. There are other explanations for the bizarre results in that study, of course, but I’m just tossing another possibility into the ring.
Remember, it is not our data that raised the issue of reduced mortality after DCIS and second events. It was the MD Anderson observational data that documented diminished survival in those women who were diagnosed with invasion disease within a mere 2.9 years of their DCIS, most commonly in the opposite breast. Would MRI up front at the time of DCIS diagnosis improved survival? I don’t know, but it is certainly possible. A known delay in diagnosis of invasive breast cancer of even 6 months buys you a guilty verdict in a malpractice trial, but we really don’t know what the consequences are of ignoring the 2% who undergo wrong side surgery for DCIS, when it ought to be invasive cancer treatment for the other breast.
“Choosing Wisely” needs to reconsider their use of the word “wisdom.” Perhaps, they should call themselves “Choosing Efficiently.” And perhaps, the American Society of Breast Surgeons should think again about their willingness to say, “Don’t.” In a sense, knowledge shrinks as wisdom grows (not mine – it’s a quote from philosopher Alred North Whitehead). We need more policy makers who have read Nobel laureate Dr. Richard Feynman’s essays on scientific wisdom. Of the many insightful quips of this transcendent individual, one of my favorites is paraphrased like this: “When your results verify your hypothesis and excitement abounds, stop and think through all possible alternative explanations.”
This is no small issue. Guidelines are just that. Even Johnny Depp in Pirates of the Caribbean, when challenged about not following the “pirate code,” told his adversaries that pirate laws were “really more like guidelines.” Unfortunately, when some policy-makers use the word “guidelines,” what they intend are “rules,” and nothing is more demonstrative that guidelines are morphing into canon than the word “DON’T.” Halsted would be delighted to see this resurrection of dogma. Others have said it before me – dogma is the enemy of science.