Epidemiology and Prognostic Assessment of Soft Tissue Sarcomas

SLIDES & TRANSCRIPTS
Monday, June 17

Epidemiology and Prognostic Assessment of Soft Tissue Sarcomas

Murray F. Brennan, MD

Slide 1:

DR. BRENNAN: Thank you, Dr. Borden, Dr. Saxman. Let me add my thanks to all of you for coming. I think this is the 27th year that I have been actively involved in the management of sarcoma, beginning with the trials that we began at the NIH more than 25 years ago, looking at amputation versus limb-sparing surgery, a subject of the past and, b) adjuvant protocols in both radiation and chemotherapy under the leadership of Dr. Steve Rosenberg.

I am going to spend 15 minutes just describing where I think we are in terms of prognostic factors for outcomes.
What I would like to talk to you about briefly is how we came to some of the ideas that we have about particularly prognosis and staging, and that won't take us a great deal of time

Slide 2:

As you know, the original staging system was described in 1977. The important thing about that staging system, of course, was that it was totally grade dependent.

So, if you had a high-grade lesion, then you were in Stage III and were expected to have about a 30 percent survival.

As we and many others showed, of course, that this ignored size other than within the context of gray. So, a less-than-five-centimeter high-grade lesion sat in the stage III category.

Slide 3:

Many people showed, including ourselves, that for both primary soft tissue sarcoma less than five centimeters,

Slide 4:

both local disease-free survival and, of course, more important, distant disease-free survival in primary treated patients was certainly not 30 percent in five years, but was somewhere on the order of 85 to 90 percent.

So, that made the issue of the small lesion being treated quite different than it had been in the early 1970s, when we first got started.

Slide 5:

We began a prospective in-house sarcoma database in 1982, when I went to Memorial.

These are, as of the end of last year, now over 5,000 cases that were treated with in-house treatment. This does not include consultations.
So, all of these people underwent some form of treatment, usually surgical followed by chemotherapy and radiation therapy.

As you know about as well as anyone, about 50 percent of sarcomas occur in the extremity, 30 percent in the lower extremity, and about 15 percent in the upper extremity. So, about 50 percent of these lesions are extremity-based.

Slide 6:

The important thing about that is -- and this exists in no staging system of course -- the fact that site itself is a prognostic factor for outcome.

It is an important prognostic factor for outcome because virtually all staging systems or descriptions are all based on extremity lesions, and only 50 percent of patients have extremity lesions.

We know that if we look at local disease-free survival and we look at the retroperitoneal sarcomas, there are about, as you can see, somewhere between 1,000 and 2,000 cases on each of these curves, and at 10 years there is a very significant local recurrence rate in the retroperitoneum. Everybody who manages these patients knows that, and it is a constant decay continuing on out as long as we have followed these patients.

However, despite that local recurrence rate, and the local recurrence rate of visceral lesions, about which we will talk later -- which is considerably less and much more like what happens in the extremity --

Slide 7:

death from sarcoma, disease-specific survival for both visceral and retroperitoneal lesions is remarkably the same.

That makes the key point of course that, in fact, in regard to local recurrence of retroperitoneal sarcoma (usually liposarcoma) death can be from local progression whereas death in the other lesions is almost all due to systemic recurrence.

So, we are immediately confronted with an issue that this is essentially a local problem and maybe it should be addressed with questions of a local therapy, whereas all of these other lesions in terms of death from sarcoma are systemic problems. None of that appears in staging systems.

Slide 8:

You are aware that the crude distinction between high grade and low grade has held up over 20 years. We, of course, have been surgically minded and consider tumors either high or a low grade.

Others, of course, have had all grades one through four.
Surgeons don't particularly like Stages I through III, because we feel everyone puts them in intermediate grade and no one can answer any questions, but this has held up reasonably well -- high risk of metastasis, low risk of metastasis.

Slide 9:

What do I mean by that? Well, if you take only primary extremity lesions and you look at disease-specific survival, then the likelihood of dying from an extremity low-grade sarcoma at 10 years is certainly less than 15 percent as opposed to, in the patients who have high-grade lesions -- and this continues with late recurrence -- at 10 years, somewhere between 55 and 60 percent will actually die from soft tissue sarcoma, immediately distinguishing these two types.

Slide 10:

The important issue of lymph nodes, of course, I was going to make the point that the presence of lymph node involvement only in this significant database is of the order of about three percent.

Again, everybody in the room, you see a lymph node metastasis from a liposarcoma, you go back to the pathologist and say, are you sure of that diagnosis, because in our experience that is less than one percent and, overall, in adults, one to three percent.Isolated subtypes, epithelial types, we will see a little higher.

However, once you have a lymph node metastasis, whether it is a lymph node with other areas of metastasis or a lymph node alone, essentially the outcome is the same.

So, this is metastatic disease. You can see a small area on this lymph node alone curve where isolated patients are salvaged by lymph node dissection.

Slide 11:

We reviewed, back in 1994, 1,000 sequential patients with extremity sarcoma, looking for prognostic factors for outcome, and that was at the time that Dr. Peter Pisters was one of the senior fellows, and he put together a manuscript from this data.

Slide 12:

The important summary of that is, with 1,000 cases, we can now summarize significant adverse prognostic factors.

Local occurrence, these four, not surprising: if you have a positive margin and already come with a local occurrence you are more likely to get another.

Distant recurrence is grade dependent, size dependent, both small and large, deep location and the presence at the local recurrence at the time of the initial presentation, which we will argue about on several occasions, but which I believe is a biological event.

Slide 13:

So, the staging system was changed in 1997, and now it was made a little more difficult because it is hard to organize. We had low-grade lesions in Stage I, low-grade and high-grade lesions in Stage II. You can see that these lesions here -- high-grade, large, and superficial -- that is a relatively rare event.

Slide 14:

That staging system held up. It was no help in local recurrence -- not surprising -- but it did hold up reasonably well.

Slide 15:

However, patients with 2L, as I said, any superficial large tumor is a rare tumor. So, the numbers that contribute to that aren't very helpful. What I did simply here was to lump those together and then break IIIA and IIIB out. This becomes a very simple clinical way to look at it.

These are low-grade tumors. This is risk adjustment. These are low-grade tumors. These are high-grade small tumors, these are high-grade five to 10 tumors, and these are high grade greater than 10. You see clinically it makes it much easier, and clinicians like that kind of thing.

Slide 16:

The reason that I did that, of course, was, as I emphasized, the patients who have large superficial lesions, whether low grade or high grade, those numbers are small and they clearly distort the reality of systemic disease. So, that is a simplified system.

Slide 17:

However, now, of course, we have a nomogram. The nomogram has now taken on about 2,500 cases. We have data now long enough for it to be a 12-year, sarcoma-specific death rate. It is a point system and now it begins to include some of the factors you know -- size, site, histology, age, low grade or high grade.

Slide 18:

It can be placed in a nomogram. It shows you very tight probability, predicted survival, actual survival, significant numbers for all of these tumors.

Of course, it looks terribly busy on paper.

Slide 19:

The correlation is good. This is the actual sarcoma-specific survival, this is the predicted. You can see, for the group, it is very tight. Obviously, for the individual patient, there is some range about the mean; but you can certainly now put that on a nomogram.

Slide 20:

I have this on my palm pilot, if you want it. It just says, here is an example. You take the screen, you pick a high-grade, deep tumor, greater than 10 centimeters, it is an MFH, it is lower extremity. The age is 35.

You compute 12-year sarcoma-specific death probability 48 percent, or 52 percent survival.

That can be done repetitively, and obviously requires validation, we hope, by other databases, but that is easily placed on the website.

Slide 21:

Now, we act as if molecular markers are new. We began our studies, at least, in looking at various gene products, surface oncogenes, in 1990.

This was the first paper we published in the New England Journal in 1990. Bill Cance was a surgical fellow working with Carlos Cordon-Cardo.

What we showed -- and of course, the New England Journal loved it - was that if you took a group of patients with soft tissue sarcoma and you took some tissue, and you looked at the expression of the Rb gene product, some expression of a tumor suppressor gene--well first of all, we had about 40 patients in this study. We showed that if you looked at the two-year survival for the homogeneous expression of this gene product, people did very well. If they were absent gene expression, they did very poorly. This was the two-year survival comparison between the two.

This, of course, was all comers. The tissue was taken from various age groups, various metastases. Then we said, now we have published in the New England Journal, why don't we do it correctly. So, we took the next 100 sequential, high-grade lesions and looked at that marker. Of course, it looked exactly like this and, of course, we had enormous trouble getting that published.

Slide 22:

We have continued a number of studies. Of course, it is a group that I work with, a significant group, and there are lots of people involved in these studies. By the strength of the database and the availability of various molecular markers, we have looked at a lot of other things.

Slide 23:

This just says the exact thing. It took us five years to get the right study published as a negative study, in which, whether you had uniform expression or absent expression in primary high-grade tumors -- in other words, taking the molecular prognostic indicator which, in generic terms, appeared important, correcting it for a simple known variable, taking only primary tumors from an extremity, and the value is not so great.

Slide 24:

We have done numerous studies looking at various other things

Slide 25:

in which it always holds up, but low grade and high grade distinguish even more in the same study when we begin to look at other markers

Slide 26:

-- for instance, p53 presence or absence.

We can show, usually on univariate analysis, some variation, but the moment that we include other than the other than sort of straightforward primary lesion, most of that individual significance disappears.

That will be one of the challenges for this meeting, to say "how do we evaluate some of the prognostic markers for outcome that are molecular-based, in the context of what we already know." We will revisit what was done in breast in terms.

That, of course, becomes quite difficult because of the content of the type of patient that leads into the study. That is the second part that we all hope will come out, that the availability to the tissue banks that we have around the country, where there exists clinical validation of material will become important.

Slide 27:

We have looked at other combinations. Carlos, for instance, published a paper looking at the relationship between MDM2 and p53 and the concept of one modulating the other, and looked at what the outcome would be, and we were able to put together groups.

Slide 28:

As you can see, this is at best working very hard. MDM2-negative, p53-negative; both positive -- yes, there is a statistical difference, but you can imagine that if you put that in terms of overall survival as opposed to disease-specific survival, the likelihood that just two simple combinations of molecular markers will make major prognostic difference is low.

Slide 29:

One thing that has held up as best we can tell, and usually as a surrogate or directly proportional to grade, is the expression of Ki67, and that is a recurring theme.

Again, the differences between the two curves, as you might imagine, are small and unlikely to make a huge difference in outcome.

Slide 30:

If you look only at primary, high-grade, extremity tumors, where distant metastasis by size, as I showed very early on, consistently and almost invariably predicts outcome.

Slide 31:

You try to then ask that question in terms of the relationship to p53 -- in other words, defining the clinical presentation and then looking at the molecular marker. Then individual markers, as you would expect, disappear, and I don't understand why we are surprised by that.

Slide 32:

This demonstrates the same thing for MDM2, just saying that exact same thing.

Slide 33:

This just says the same thing occurs, as I told you before, which is an interesting concept.

The distant metastasis rate, by margin status, by the presence of a positive margin, at the time of the original operation, is a predictor of outcome. We could talk about that for a long time.

It is amazing, biologically, if you take patients who are out at five years disease-free, and you look at them, nine percent that will then die of sarcoma in the next five years, the most powerful predictor of that event is a positive microscopic margin at the original time. It must be biology. It cannot be anything else.

Slide 34:

This just shows, Ki67.

Slide 35:

This, Dr. Ladanyi will tell you about, because it looks like, initially, exactly the same concept, that if you look at the SYT-SSX -- and he will talk about that -- in patients only with localized disease at presentation, it looks as if there is, indeed, the value in a molecular marker in a particular tumor at a specific site.

Slide 36:

However, it does not reflect outcome when you look at all patients in terms of overall survival, again making a suggestion that these molecular markers for outcome are related to some of the non-prognostic factors.

Slide 37:

So, our challenges, I believe, are with known current prognostic factors for outcome, that are well defined for clinical pathological variables -- and I have just listed them -- we have model nomograms now that improve staging, still limited by size despite the 2,500 database, and the duration of available prospective data.

As we look at soft tissue sarcoma with this late delay and late recurrence rate and begin to ask questions of prognosis based on initial markers, then we need long-term follow-up. This is a challenge, I think. It is the integration of molecular markers into the prognosis indices, and that requires -- because we haven't even commented on it -- improvement in assay quality, with everything from immunohistochemistry and PCR among other of these.

You are going to hear about multiple assays and multiple array analyses -- A) How do you analyze these? and B) How do you put them into factors for outcome?
It needs, again, validation in large clinical cohorts, not just cohorts from a single institution such as ours, and they do need to be done in the context of prognostic factors. That is one of the strengths of the nomogram. It is a mathematical model.

So, you can introduce new factors constantly and challenge the nomogram. That is much easier than introducing it to a staging system that only has two or three variables.

Of course, the true challenge later in the day will be to define whether we can define treatment based on those outcomes. I apologize for the delay in getting started, but the operation is finished. Thank you.