SLIDES & TRANSCRIPTS
Monday, June 17

Pathology of Soft Tissue Sarcomas

Christopher Fletcher, MD

So, all of my talk is focused on trying to address where we are now, the things we have done in recent years in pathology, what are the shortcomings of what we do, and what do we need to do as we go forward. I am going to try to take care not to overlap with Marc Ladanyi's talk.

TOP

Compare if we had held this meeting 10 years ago. There are big changes now. We are much better at diagnostic sub-classification. We tend to not lump things together in the same way.

We have learned to appreciate -- as clinicians, this is true as well -- we have learned to appreciate that there is a biologic continuum in some tumor types.

GIST -- since it is the thing that drove this meeting -- is one perfect example; but it is really very, very difficult to prognosticate. Clearly, because we have a continuum, you might think of SFT, solitary fibrous tumor, the same way.

The various grading schemes from the NCI and the French system have been validated and tend to be useful. We have begun to incorporate the new and molecular techniques, and there have been big, big studies that have actually validated the correlations between cytogenetics and morphology. We have also demonstrated that molecular testing can be useful, but we haven't learned to use it widely, and I will come back to that later.

The other big improvement, I think, is that now when people set up big clinical trials, they realize the importance of centralized pathology review. There were many trials conducted in the past without central pathology review, I think almost certainly with numerous misdiagnoses or at least heterogeneous diagnoses included.

TOP

If we just look at the current status of where we are now in pathology, I think one thing we have to address -- I don't know how we get around it -- is that there is enormous geographic variation in the available resources and how they are used and, therefore, in the diagnoses rendered.

It relates very much to case volume. There are people -- everybody in this room is interested in sarcoma and could reasonably classify themselves as an expert --there are people in the room who may see 1,000 or 2,000 new cases a year, either pathologically or clinically, and others who still feel they are an expert who only see 100 or 200. There is a huge difference in perception of disease according to the volume that you see, and that is very hard to address in terms of understanding biology.

The fact is that, as we go forward, we continue to refine our morphologic classification. Whether you like it or not, we keep inventing, and learning to recognize new morphologic subdivisions which are often clinically very relevant. I mean, if you think about desmoplastic small cell tumor 10 years ago, low-grade fibromyxoid sarcoma -- the thing that takes 25 years to metastasize -- those things we begin to recognize and make your jobs easier.

We do classify things increasingly meaningfully, I hope, and more and more cases are now graded. Yes, there are still reports that you get that don't have a histologic grade in them, but not huge numbers. This last, diagnostic variability, occurred in the past because major textbooks have helped to propagate decent diagnostic criteria.

Immunohistochemistry has become a very routine sort of part of diagnosing sarcomas, and people have begun to use cytogenetics and molecular genetics, but perhaps not as much as they should, and we will come back to that.
Implicit in all this -- and it is something that people have struggled with, and this is not me and my own bandwagon, it is a fact -- MFH as a diagnosis is rapidly diminishing.

We now recognize that it is a term that can only get used for the ragbag of what is left of unclassified pleomorphic sarcomas, and that is reflected in the new WHO classification that will come out in the fall.

There really isn't anybody left who can define MFH as a reproducible entity. Those of you who have huge databases with hundreds of MFHs in them are not happy about that. They will bawl and they will quiver and their chests will pound. Especially if they are a surgeon, there is lots of chest pounding. The fact remains: with those data, you have no idea what those tumors are. So, you are going to have to learn how to address that. You will have to pay poor slobs like me to go and reclassify them I suppose;

TOP

but it is going to be critical to moving the field forward. I know that is something that people like Marian Sand feel quite strongly about and that is why they are reappraising a lot of their cases.

It is hard to collect large case numbers of a single histologic type outside of the main sarcoma centers. In order to do that, you often depend on consult material, which results in big issues in variability and treatment and the way the tissue has been handled and so on.

As a group, pathologists aren't that good at laying out their diagnostic criteria. I now find, in reviewing any manuscript, I want to see whether they have defined how they collected their cases?

They say, "We retrieved 200 synovial sarcomas from our file." I want to know what criteria they used to know they were synovial sarcoma. In the past, some of these things were pretty sloppy. Now, they are much, much stricter.

We don't always use objective tests to validate diagnoses. There is still a tendency to say, "It is what it is because I say it is", and that is no longer good enough.

Now, one wants to see objective support -- either from immunohistochemistry or molecular genetics, if available and if relevant -- to support the diagnoses.
I have mentioned already the big problem in variability of primary treatment. So, you are going to have big problems in prognostic studies, whatever happens.

TOP

What about just practical issues, day to day? These are all supposed to sort of provoke you to think of things to discuss later in the morning.

How much tissue do we, the pathologists, need to make a diagnosis? That also depends on how good the pathologist is, which is not sort of entirely fair. In real life, there is big variability between an open biopsy, a needle biopsy, or even a fine needle aspiration, which brings us to this, what is the most appropriate biopsy technique?

You would like to be able to ram a needle into somebody in the outpatient clinic. It is quick, it is easy, it is cheap. You don't have to book time in the OR, but it makes it much, much harder for the pathologist to render a sensible diagnosis, because of the amazing heterogeneity of sarcomas. If you put a needle in a breast cancer or a colon cancer or a lung cancer, in large part it looks the same, wherever you stuff the needle. In a sarcoma, you can routinely find five or six different appearances, depending on where the needle went. If the people doing the biopsies don't understand that, then that obviously limits the ability of the pathologist to make a diagnosis.

The other big problem, of course, is that if you do use needle biopsies and then you insist on giving preoperative neoadjuvant therapy of any kind -- doesn't matter if it is radiation or chemo -- you should not imagine that you are going to get a sensible diagnosis on the resected specimen. Once they have been radiated or have had chemo, then forget it. You can no longer classify most of them and any grading is worthless unless, of course, they haven't responded at all, but then you wouldn't believe that ever happened.

So, this is a big issue. We have to think about this because we are going to go back to having endless sarcoma NOS. That is just needle equals sarcoma, let's roast it, and then all of a sudden you have got hundreds of patients you have no idea what you have treated.

Now, if you do give us some tissue, we have to decide how best to triage it. Should we be giving all those tissues -- for cytogenetic analysis? Should we always freeze some for molecular analyses? Should we be sending stuff for EM and so on? This needs to get discussed. We need to agree on what essential data should be there in the final report. That has largely been addressed in reports and recommendations from various organizations; but we need to get people to stick to them.

I have already addressed the problem of variability in individual experiences, that diagnoses rendered depend on how many cases you see, just like the surgery depends on how many cases you see, and we have to find a way around that. You can't -- clinicians haven't succeeded in mandating that all sarcomas get treated by sarcoma experts, even though we all know that patients would do a hell of a lot better.

The same problem applies with pathology. You would love it if you said they all have to be sent to expert diagnosis centers around the country. It is never going to happen. Yes, pathologists are just as bad as your surgical colleagues, who screw everything up three times before they refer the patient to you with a recurrent lesion.

What are you going to do about MFH? We have to address that. Those people who have got huge databanks full of MFH are going to have to think, how are we going to readdress that problem? Then, I have mentioned the impact of preoperative therapy already.

TOP

Now, prognostication, which I thought fell into pathology, but Murray here talked about it for a minute. I will just mention the things that pathologists can do.

Histotype is still probably the single most important predictor of prognosis, and that is reflected in Murray's nomogram as well.

Histologic grade: we know the French system and the NCI system work pretty well. The French one is best.
Proliferative indices and cell cycle indices, whether you are talking about things like T67 or p53 and Rb, have been things that have been proffered. In reality, they rarely improve upon decent grading or mitosis counting.

That is something that you may find strange. The fact is, just gazing down the scope, counting mitoses, turns out to be just as prognostic and a lot cheaper than staining something for T67. Of course, it is harder to publish because it is not novel.

In Scandinavia, there have been big studies that have shown that vascular invasion is prognostically important; but we don't all routinely look at that, and we have to consider whether that is worth doing.

Fusion genes increasingly appear to have prognostic relevance, which Marc may talk about; but we have to decide, are we going to use those in clinical practice?

We could begin to predict treatment response in different ways. For example, if you are trying to distinguish between synovial sarcoma and PNST and fibrosarcoma, if you know for sure it is synovial sarcoma, you know there is a fair chance it is going to respond to ifosfamide, whereas the other two might not.

You could say, looking at mutational analysis and GIST, there are clearly different responses to Gleevec according to what kit mutation you have. That could become part of a routine pathologic assessment, and has become so in Oregon.

One thing we don't do in soft tissue is to assess treatment response. In bone sarcomas, everybody talks about looking at the resected specimen, how much necrosis is present after it has been preoperatively treated.

My big problem with that is, to this day, you cannot reliably distinguish between spontaneous necrosis and treatment-induced necrosis. We have no good way of assessing treatment response in sarcoma. We have to decide whether that would be useful.

Tumor size: we can all measure them. Actually, you have to remember that they get measured more accurately on a bench than they do by an MRI.

TOP

So, there is the grading that we know splits things very nicely. The good thing about the French system is that it dumps fewer cases into the intermediate grade category, which you don't like because you don't know what to do with them. So, the French system does actually work better than the NCI system.

TOP

Murray's nomogram, I think, is great, particularly for two reasons. You will see the two factors which are most discriminate. In fact, they spread right across from the zero to sort of the top end, a histotype and histologic grade.

Those are the two single most important discriminates, as was acknowledged in Murray's paper.

TOP

What are the problems with prognostication? Well, we have to admit it, we are not reproducible in histotyping.

There are a lot of MFHs and fibrosarcomas in the older Memorial diagnostic sort of categories, whereas those diagnostics virtually never get made at Dana-Farber, haven't been made for the last 10 years. So, how are you going to compare the data?

The grading systems are not perfect. Some tumors are not susceptible to grading, whether we like it or not. There are various histologic types that you cannot apply the objective data -- things like clear cell sarcoma, epithelioid sarcoma, and so on.

We don't want to measure treatment response. One of the biggest problems is, right now, there is very little routine molecular testing. We all love to talk about it. Half this meeting will be devoted to it. How many of you are using it on a routine basis? I would say it is probably no more than 10 percent of even the people in this room who are sarcoma fetishists.

TOP

So, we have to think about that. So, the question is, do we need to develop histotype-specific grading schemes? If so, are we going to have enough case numbers to do so?

Does fusion gene or breakpoint testing have a wider role? If we decided that the different EWS fusions in Ewing's sarcoma are prognostically important to us, as several major studies have shown, then why aren't we doing it? Can we find ways to measure treatment response? I suspect not, but we have to think about it.

TOP

Let's just think about the feasibility of molecular testing in routine practice. What we are really talking about now is the identification of all those reproducible translocations with fusion genes that have been very well characterized over the last 15 years, and which we know we can use to validate diagnoses, and also increasingly for prognosis -- things like Ewing's, alveolar rhabdo have fusion genes which correlate with outcomes.

The same with synovial sarcoma but not true in myxoid liposarcoma.
If we are going to do it, where are the resources going to come from? In which laboratories are you going to put it? You can't put it in a routine surgical pathology laboratory, where it is all dirty and mucky and stuff. You need to have nice clean benches when you are going to play with DNA. Somebody is going to have to pay for it. If you bill for doing this testing right now, I can tell you, you lose money hand over fist. So, most pathologists don't want to do it.

Just for the U.S. members of the room, which is 90 percent: for example, in our lab, we bill $300 for doing a FISH analysis to validate an X18 fusion gene. We get reimbursed $30. We genuinely lose tens of thousands of dollars a year in running a FISH service. Where are we going to find techs that can run all this stuff? How are we going to decide what technique to use? You have all seen published data based on RT-PCR; but we are completely lacking, based on the fact that we have sort of, let's say, questionable RT-PCR with contaminated data. Should we be using RT-PCR or should we be using FISH? These are very different techniques. They require different sources.

Should they be done in specialized centers only or does everybody want to do it? If so, will it be meaningful if everybody does it? The short answer to the latter is no, it won't, because people who don't do it often enough won't do it right and there will be lots and lots of hideous errors.

TOP

Then, there is the new world we now have in pathology that is a whole different thing. This isn't so much to do with diagnosis. This is now target identification. It is the thing that gives every med onc in this room a little quiver, because they can chuck drugs at a target.

Yes, instead of saying, let's kill cells, we don't know quite how we do it and let's hope we don't kill normal ones, you can now go for targets, which is great. It is wonderful. It is a really exciting new phase in sarcoma molecular oncology.

How are we going to try to do this? You have already seen, those of you in this room who have already addressed this problem -- that is probably many of you -- the poor reproducibility of things like kit staining. Is it a GIST or not, or has something other than GIST got kit positivity so that I can give it Gleevec?

The short answer to that is almost invariably no, and I wish you didn't ask, because it is not -- how to put it? For tumors in which there is no good evidence that kit is mutated or activated, there is no logical reason to stain it.

I know we all have to deal with patients who are seven-eighths dead with untreatable disease and you are willing to try to give them anything; but it destroys the credibility of targeted testing if you say, "Please, will you test this patient with ovarian cancer or dedifferentiated liposarcoma, for kit?" Because that suggests that one doesn't understand the biology of what we are dealing with.

I think it will apply in due course to all these other ones. It is going to be people in this room that are going to provide leadership in how that is going to be achieved.

What are going to be the best techniques for demonstrating these targets? If immunohistochemistry isn't reproducible? Right now there are no decent antibodies to PDGF receptor.

Any of you who think you are getting immuno results from commercial labs that are believable for PDGF receptor, you are fooling yourselves.

The pathologists in this room have tried repeatedly -- and all reasonable pathologists. The antibodies are garbage. The whole slide goes brown. You shouldn't be basing your testing on that.

If so, should it be Western blot? Should it be people like Jon Fletcher running Western blots instead?

Should we be having mutational analysis for kit mutations because it is prognostic and more reproducible? If so, who is going to pay for it? Where is it going to get done other than Oregon? Because they can't do every GIST in the world.

Should we be thinking about downstream targets? As these tumors become resistant to the initial, for example, Gleevec therapy, should we know more about targets further down the line that can then be switched off with different drugs?

TOP

So, there is potential study that we have to think about in pathology -- rigorous analysis and comparison of different prognostic parameters. Right now, we don't know which is best -- histologic grade, Murray's nomogram, gut feeling from experienced pathologists, or fusion genes. I mean, it is that sort of wide a spectrum, and we really don't know which is best.

We don't know what is the best way to go about attaining wide genetic testing -- FISH, RT-PCR, specialized labs, everybody's labs.

We have got to learn whether it is worthwhile to measure treatment response, but we have got to find ways of defining treatment response first, because that may prove to be useful. It appears to have been useful in osteosarcoma and Ewing's; but those are two relatively less common diseases in bone, and bone pathology is different from soft tissue pathology.

We really do have to address what sort of biopsies are ideal, particularly if you are going to start using more and more preoperative therapy. You have to know you are losing a huge amount of information if you go for needle biopsy after neoadjuvant juices.
TOP

Now, what about just some practical research issues? We know -- and it is a problem that plagues all of us in the room, and it is one of the reasons there isn't a sarcoma SPORE and so on -- is because it is a relatively rare disease, because there are more limited case numbers, that is one of the big things that has made it harder to get funding for a lot of research in sarcoma.

Yet, even though they are low, you all know they actually carry disproportionate biologic significance, not only because of their relatively greater clinical prevalence in children, adolescents, young adults -- which often scores lots of brownie points with grant-giving bodies -- but also because of these mechanisms.

It has been this major source of these mechanistic paradigms -- the specific translocations with relatively simple genotypes that are susceptible to analysis, to work out these pathways of signaling that actually cause these diseases and may determine their behavior.

The fact that we have these well-characterized fusion genes, the fact that we are now beginning to recognize all these receptor tyrosine kinase pathways -- like in GIST, and undoubtedly in others -- EGF and synovial and whatever -- doesn't matter.

The fact is, we are learning to identify a whole new set of pathways that can get treated as well; but we need to find a way to get funding for this work.

The other problem -- particularly in pathology anyway -- is certainly pathology study sections do not support translational research, period, and that is not an exaggeration. It is a statement of fact.

If the conclusion of this meeting was, all right, we need to collect huge numbers of sarcomas and we need to get a group of pathologists together who are going to decide whether RT-PCR or FISH is better or ELISA -- because these techniques are prognostic across large numbers -- one could not get the money to do that work, because it is too clinically focused.

It is somewhat easier in clinical study sections, where there is a different emphasis, but in pathology in particular there is no funded translational research through NIH.

TOP

Some needs or questions we can finish up with. We need to work on how best to achieve even greater diagnostic reproducibility. You want to know that, if I see it or Sharon sees it or Brian sees or what Pancras sees, you want to know that we are looking at the same thing, that there isn't that much variability. Otherwise, data is questionable.

I think pathology has come a long way in that regard; but we are not perfect yet, because the fact is that not everybody is Sharon or Brian or Pancras. Yes, there are still millions of pathologists out there who don't do quite a perfect job.

How are we going to get better at prognosticating in those tumor types with a biologic continuum? And that is a real issue, and GIST is the perfect paradigm for that. How are we going to better prognosticate in those tumors that only give rise to metastases infrequently? The ones in which you know there is some small metastatic risk, but you don't know which patient it is going to happen to.

We need to decide whether molecular genetic testing is feasible more widely; but it has to be reproducible. How are we going to do it, and will it be worthwhile doing?
We need to look at how we can optimize getting diagnostic and prognostic information for what, at our end of the food chain, appears to be increasingly limited biopsy material.

It is quite interesting, actually, if you stop and think. In the centers that are truly multidisciplinary, where sarcoma groups have got equal representation from surgeons, from med oncs and pathologists -- most of those centers still use open biopsy, and it is all to do with specialties talking to each other.

The centers where maybe you don't have difficult pathologists nagging you saying, "That is not enough tissue", then you get away more often with a little biopsy.

TOP

The fact is, as Murray's talk and in this I hope to show, is that right now you all know pathology is central to your diagnosis and prognosis of each of your patients.

So, we need to optimize how we are going to go about this, and we want to try to minimize variability. We have to try to work toward this.

We need to decide about how we are going to introduce molecular genetics or functional analysis in a routine setting. Since we all talk about it, we are paying lip service right now to a fantasy, and we are not doing it. We need urgent, larger-scale validation. Otherwise, we are like the king with no clothes. Yes, we talk about it, but we just don't do it.

The development of these new, targeted therapies is creating a whole new role for pathology, and we have got to find a way of standardizing for that. Also, we have to decide how to respond to what is now consumer-driven.

It is not even you guys demanding tests, but we have patients phoning up the whole time, "Please stain my tumor for kit."

You have to sit there on the phone and say, "Actually, that is not a good idea." And you have to explain to them why not. That should in reality be coming from the clinic, rather than the pathologists.

The fact is that, to do most of these things, we are going to need larger collaborative case numbers. So, there is going to have to be more and more emphasis on bringing large groups together. Thank you.

TOP