The Alchemy of Lumpectomy Margin Guidelines

First, it was lumpectomy margins for invasive disease, and then in late August 2016, margins for DCIS. In the former case, it was “no ink on tumor,” and in the latter, “2mm is adequate as long as radiation is planned.” After a quarter-century of breast conservation, with absolute chaos reigning the entire time as to what constitutes an adequate margin, the majority are shouting, “Amen!”

Now, go back and read my July 2016 blog “Guidelines Morphing Into Canon.” Guidelines are great, as long as they are used as guidelines. Instead, guidelines have a way of becoming “standard of care,” and from there, it’s a small leap to canon, with any departure being a grave violation of patient care, and certainly not worthy of reimbursement by the insurer who only has the patient’s best interest at heart.

Lumpectomy margins are probably the most random, irreproducible, inaccurate measure we have in breast cancer management today. I sometimes joke that “pathologic margins are only a surrogate for, well, true margins.” But somehow, if you collect enough experts together in the same room, representing three major cancer organizations (SSO, ASTRO and ASCO), then wave a statistical wand over a plethora of clinical studies, you can turn lead into gold!

Having spent a year in the pathology lab, and being intimately familiar with tissue processing, I have a different view. From the time the lumpectomy specimen is being removed from the patient, aberrations begin to occur that give phony confidence when the margin is assessed. From the specimen radiograph that can knock off the outer rim of fat, to the handling of the gross specimen, to the ink that slides down into the crevices, to the ink that can be dragged into the specimen with the scalpel, and on and on, I can tell you why false-positive margins emerge. And then, on the false-negative side, there are even more reasons. The point is that margins are highly unscientific. So, how do p-values correct this? They don’t, but they give the illusion of certainty and the delusion of evidence-based medicine.

There are so many variables that should enter into the decision for re-excision that I can’t begin to list them here. But the first problem with this new standard is that “DCIS” is addressed as though it were a single entity. But a 5.0cm DCIS where the extent is not seen well on imaging, and with margins of only 2mm at multiple sites around the specimen…well, it’s not even in the same category as the 0.5cm DCIS that fits on one slide were you see the entire lesion.

As DCIS grows, the margins become less and less reliable, and imaging of all types can be unreliable as a roadmap as well. How many prospective, randomized clinical trials have been performed to test the value of radiation in women with Grade 1 or 2 DCIS that measures under 1.0cm? Answer: None. Yet, these women are told to undergo radiation because prospective RCTs have shown benefit to radiation, even in “good DCIS,” failing to mention that so-called “good DCIS” in those trials, when Grade 1 or 2, can be up to 2.5cm in the study protocol, with 3mm margins. That’s a different animal than the 0.5cm DCIS that was also 0.5cm on imaging that I described above, especially if you can accomplish 1.0cm margins. In this latter situation, recurrence rates with wide excision alone are close to zero, making it very difficult for radiation to add anything other than trouble.

Take note, by the way, that if one thinks this through (called rational thought, often a passé approach in today’s climate), and you get only 2mm margins on your first excision of a 0.5cm Grade 1 DCIS, the Triumvirate would spare you the re-excision (and this is where they are being praised), but insist upon radiation. In contrast, you could perform re-excision and spare the patient radiation therapy. Which can cause more trouble – a re-excision or unnecessary radiation? I hope the answer here is self-evident.

The congratulatory comments have been pouring in for the New Triumvirate of Margins – SSO, ASTRO and ASCO – with other societies chiming in their support. The take-away message is that this 2mm directive is a wonderful advance in that it will cut down on the need for re-excisions. In fact, the leader of this entire “margin guideline” movement is a breast surgeon who, only a few years before this crusade, published one of the highest re-excision rates ever reported – 60%. The reason for this reinvention of one’s self is unknown, but it’s remarkable that one can go from assuring the world that a 60% re-excision rate is okay, to the extraordinary steps taken these past few years to minimize re-excisions.

Regardless, this movement to cut down on re-excisions is no doubt admirable. We are trying to minimize overtreatment these days. Critics point out that DCIS should not be called cancer at all, but there’s a more elemental question than that – Many of the small, low grade DCIS lesions I see in consultation are not DCIS but ADH (atypical ductal hyperplasia), best termed a “borderline lesion.”

Many clinicians would be shocked to understand the subjectivity that comes with DCIS, esp. small, low grade lesions. The articles have been published (most recently by Elmore et al, JAMA 2015), but the implications are so overpowering that surgeons and radiation oncologists just tuck the information away in their individual safe places. But the fact remains, we have known since the days of the BCDDP study in the 1970s that there are lesions for which there is no agreement on the diagnosis. My solution to this problem is staggeringly simple – admit these lesions are “borderline” and treat them all with wide excision, NO RADIATION. This alone would spare more overtreatment than this entire Margin Triumvirate trend, which has been focused on a lesser endpoint. My solution is not original. A small minority of pathologists have been trying to draw attention to the problem for over 50 years, but clinicians will have none of it. They want black-and-white from a medical discipline (pathology) that happens to have shades of gray as part of its core.

So, here’s how I rank the most urgent problems today with regard to DCIS:

  1. Unnecessary radiation therapy for a non-life-threatening condition and for which there is no survival difference after treatment (the current Margin Triumvirate was not able to reach a conclusion or recommendation on this most important issue).
  2. The correct diagnosis, which ought to be “borderline lesion” in some cases, the working definition being whenever two experts disagree (this phenomenon is so well known in the hidden world of pathology that practicing pathologists will even tell you, correctly, that “Dr. X would call this lesion atypical ductal hyperplasia, while Dr. Y would call this DCIS. So which diagnosis do you want?”
  3. How do we handle the “STOP Pre-op MRI” movement when, for the DCIS patient, 2% (our publication) will have a life-threatening invasive cancer on the opposite side of their known DCIS. In other words, without MRI, one in every 50 patients undergoing some minimal treatment of their DCIS is going to have an untreated invasive cancer on the opposite side. Imagine anywhere else in the body where “wrong side surgery” is performed as a matter of routine once every 50 times! Yet, pre-op breast MRI is maligned from every angle possible, and especially so by the leadership of the Margin Triumvirate. Remarkably, because pre-op MRI in DCIS does not help much with the index lesion, it is maligned even more than when used for invasive disease. Yet, for our tiny 2% (nothingness to a statistician), it’s a matter of life and death, bringing up the possibility that MRI is even MORE important for DCIS than invasive disease. After all, we don’t alter survival with pre-op MRI in invasive cancer, but what about our 2% in of women with DCIS? We’re talking about 1,200 women each year diagnosed in the U.S. with DCIS who actually have mammographically invisible invasion on the opposite side.
  4. How do we accept the notion that nothing is important beyond 2mm when a vast body of data exists that show a direct relationship between margin size and local recurrence rates? Do we ignore the entire body of work by Silverstein and Lagios (USC/Van Nuys Prognostic Index), replicated by others? Do we reduce the seemingly infinite clinical presentations of DCIS and call it a single entity and offer simplistic solutions in the name of “generalizability.” DCIS is a complex array of entities, and there are many breast surgeons who put an extraordinary amount of thought into re-excising as needed, not when guidelines dictate a one-size-fits-all approach.


Re-excisions are not the most pressing problem with regard to lumpectomy, but of all the outstanding issues, why did this group try to tackle the least scientific? Why did they focus on turning lead into gold?

I have said it many times before – when it comes to prospective, randomized trials that involve surgery or radiology, there is no such thing as pure science, and that means evidence-based medicine cannot thrive here like elsewhere. Why? Inadequate blinding and huge differences in quality.   Excellence is put on the back burner, and the focus is on “generalizability” – or, stated alternatively, the acceptance of substandard care into the trials.

In contrast to this poorly controlled situation, a miracle drug can be tested with triple-blinding (even the study sponsors don’t know who is taking what), and the pill being studied is standardized with a high degree of quality control. Imagine the absurdity of a clinical trial for a new drug, with each participating hospital responsible for manufacturing its own drug, free of quality controls, with these directions: “Whatever works best for you. We want the results to be generalizable.” That’s exactly what happens in surgery and radiology trials. If there is a more unscientific parameter than MARGINS that has been subjected to evidence-based medicine, I’d like to know what it is.

Oh, that’s right. The American Cancer Society subjected the clinical breast examination to the rigors of evidence-based medicine, and decided it should be stopped! I guess there really is no limit to the absurdity to which statisticians and their death-to-rationalism clinicians will follow.