Summary




SLIDES & TRANSCRIPTS
Thursday, June 15, 2000

Advances in Imaging for Non-Small Cell Lung Cancer:  What's New?  How Could These Technologies Be Applied to Clinical Trials?
Edward F. Patz, Jr., MD
Slide 1:

DR. PATZ:   What I'm going to do this morning is just give a very general overview about tumor imaging.  We'll discuss where we currently stand and some of the directions where I think we're going in the future.  I'm going to present one perspective, but I think this is just a general direction.

I'm going to be fairly brief, because I would like to leave time for questions  It always seems like I get a lot of questions about how we are going to image angiogenesis, how we're going to image gene therapy, how we are going to image lots of these different new techniques.  I think we have to ask a number of questions when we are looking at that, and looking at these areas, so that we understand actually what the implications are and what we really can gather from imaging.

I think that noninvasive imaging can give us a tremendous amount of information, and tell us something basically about the tumor biology.  I certainly think we are changing, and we're trying to integrate our understanding of molecular biology, of tumor biology, and trying to integrate that with imaging, where imaging in the past has often traditionally looked at anatomy and morphology.  As we will see, that really doesn't give us the information that we are trying to get at this point.  You certainly want more information for your clinical trials.  We have new drugs coming along.  So what can we give you, and what can we tell you about doing tumor imaging?

We traditionally break tumor imaging down into several different categories.  The first, certainly from our perspective, is diagnosis, often presented with the radiographic findings.  Certainly the chest films are the most common radiographic study performed.  Often we see an abnormality, we don't know what it is, and we're faced with the fact, are we dealing with lung cancer?  Are we dealing with tumors or are we dealing with some other entity?  So certainly diagnosis is one of the first things.

The second, often once we make the diagnosis of lung cancer, is looking at staging, and I put staging and prognosis in the same category.  Then the third thing which we often want to do and which we're often asked to look at is the follow-up.  So here you have a patient.  You have treated them.  How are they doing, and what can we tell you about the tumor and the state of the tumor at that point?

It's very interesting, because we often divide and stratify these into different categories.  But in some ways what I would like to talk about, is even though this is how we have traditionally done this, what is interesting is these will come together and we will really look at something which will really be one area of doing just a general overall scheme of tumor imaging.

What do I mean by that?  Basically we are going to get the same information from diagnosis.  We are dividing staging and prognostic information.  And we will be able to use some of that information, as you will see also for follow-up and trying to understand what is happening with the tumor.


TOP
Slide 2:

So as we all know, we have chest films.  The mainstay of what radiology does for lung cancer is chest x-rays.  It provides a tremendous amount of information.  As we will see when we look at some of the new techniques and some of the new technologies, we have to ask ourselves “is the new technology any better than actually measuring the tumor here on a chest film?”  And it's not clear to me that that's necessarily going to be the case.

Sometimes we know the therapeutic targets.  We don't always understand the pathways, but we want to look at actual pathways with noninvasive imaging.  Yet, it may be no better than actually measuring a tumor.  We don't know that, and that's something that we have to prove once we understand what we are doing with tumor imaging.  But here we can see there is a right upper lobe mass.  There is bulky right hilar and mediastinal adenopathy.  We have a very good idea about what's going based on this film.  We can give you a very good idea once we see changes in therapeutic manipulation with the patient and interventions.  We can certainly see changes, and this does give us a very good idea about what is going on.

When you look at most of the trials with solid tumors, actually following radiographs has a very strong correlation with actually outcomes, and that's really the bottom line.  It may not tell you the pathway, but it will tell you something about the outcomes.

TOP
Slide 3:

After we do the chest film, we have CT.  CT again, provides us with very similar, more detailed information.  But what it provides us with is anatomic information.  Here we can see that same patient with the right upper lobe mass.  We can see the bulky right paratracheal adenopathy adjacent to the SVC here.  So again, it provides us with anatomic information.

There are some manipulations which we can actually now use with CT,

TOP
Slide 4:

and we can now go ahead, and we can do 3-D rendering.  We can color these.  We can do all sorts of fancy manipulations with the information.  But realistically when you look at it, this is anatomic imaging.  This still tells us the same thing.  We can see a right paratracheal lymph node here.  We can see that we can highlight it in green.  That still doesn't tell us what's going on.  Remember that when we look for CT, the CT is about 60% sensitive, and 60% specific when you look at mediastinal nodes.  So in patients with lung cancer who have these enlarged nodes, and they are actually reactive.  So we need to be very careful.  Just because we can highlight, does not necessarily mean that there is tumor.  We can show you these things better, but what we need to do is we gain more information about this.  What we want to do is understand something more about the biology of this disease and we need to pursue this further.

TOP
Slide 5:

One of the ways in which we have done this is to look at various different characteristics.  Not only does CT provide us with anatomic information, but also it provides us with morphologic information.  There have been numerous studies looking at the morphologic characteristics of lung cancer on CT.  So we look at spiculations, irregularity.  Again, this just does not provide us with the requisite information we need to make therapeutic decisions on these patients.

Here is one patient who has a spiculated mass.  We can look at the next patient here with CT.

TOP
Slide 6:

 It looks very similar.  Here we have two lesions.  They look very similar. You look at the anatomic, the morphologic picture, and when we look even histologically these both turn out to be

TOP
Slide 7:

squamous cell carcinomas, non-small cell carcinomas.  But one of these patients died in six weeks and the other is living and is now without disease at ten years.

Why did we think that anatomic imaging would to give us the information?  Because it clearly doesn't stratify these patients into the appropriate categories.  You have two Stage I disease.  You have two patients who have the same histology, yet they are very different biologies, and we need to use something more than that.

It's very clear that if we are going to do this, and we're going to stage them, the staging implications are basically for prognostic and therapeutic implications, and we need something more than we have.  One of the things which we have done 

TOP
Slide 8:

has been PET imaging.  It has opened up a whole new branch, and this has told us something much more about the tumors than we could gain before.

Here is just one example from coronal image.  You can see marked FDG uptake here in the right apex of somebody who had a non-small cell carcinoma.  This is sort of the first time that we really got metabolic information.  We got biochemical information, rather than just anatomic imaging.  We got something about the tumor.

What we showed with this -- we certainly looked at it, and we had to make the diagnosis -- is an abnormality on the chest film.  We can help distinguish in the vast majority of the cases between benign and malignant nodules.  We can tell you whether it is lung cancer.  We can stage these patients.  We have actually shown in several different studies that before you treat them, the relative uptake in the amount in this tumor has prognostic information.  We have done a study, which has shown that even after you treat them, the PET does have prognostic information.  That those who are PET-positive clearly do worse than those who are PET-negative.  So it does give us some more information than just seeing a residual radiographic abnormality.

But it's like anything in medicine it’s not 100%. What this does is provides us with a model.  I think this is one of the model systems that we need to look at in the future.  This theoretically takes advantage of one property of tumor cells, which is increased glucose metabolism.  I think we start there.  That's one general property.  But what we need to understand is something more about the molecular biology and the tumor biology.  If we do that, then we can use this as a platform, as a starting model to then produce what we are going to do at the end, which I think is going to be the most important thing -- produce noninvasive tumor profiles to provide you with both diagnostic, therapeutic, and prognostic implications. 

TOP
Slide 9:

To do this I think what we have to do is go outside the so-called Ablack box.  We have this anatomic imaging,  exquisite anatomic imaging.  I will show you some images a little bit later from how we can do this.  We can look down at the 50 micron level, and we can look at a lot of things, but it's not clear that anatomic imaging alone is going to provide us with that information.  So we go outside of this black box.  We look at the fact that lung cancer -- and cancer is a disease of the genes – it’s a genetic abnormality.  We need to understand it.  So if we take it from the basics, we understand the basic abnormality here, if we can utilize that, then we can integrate that with some imaging techniques, and actually provide you with the tumor profile.

TOP
Slide 10:

To do this, what are we going to do?  But the first thing I want to show you is something which I think we need to think about for lung cancer, and all cancers in general.  This is a very interesting study, which some of this work was apparently presented at the ACR this year. 

This is a study from Markesh Jane's lab in Boston.  I show this because I think we need to be very careful about these tumors.  We have a very traditional model of how we think about lung cancer, particularly given a lot of the press recently on screening and finding them at small sizes. I think we need to be very careful about that.

I show this study because I think this is incredibly interesting.  This was a nude mouse xenograft.  They implanted several different tumors, several different cell lines subcutaneous in the nude mouse xenograft.  They then cannulated the vein and looked at the circulating tumor cells.  They then looked at these for several different assays.  They had a tumorigenic assay, an apoptotic fraction and a clonogenic assay.

The most interesting thing I think they found was the fact that when they looked at the tumor of a 1 centimeter lesion --  a 1 centimeter tumor which for lung cancer is fairly small, but still fairly late in the biology of the disease -- they found circulating tumor cells.  And in a 24-hour period they found between 3-6 million cells shed every 24 hours for every gram of tissue, for every 1-centimeter lesion.

I think if you realize that in some ways what we are going to do is look at this tumor, and we are going to look at these lesions as basically a systemic disease.  Now it is clear that the body keeps these in check.  What these assays showed was that clearly they had decreased tumorigenicity and clonogenicity when comparing the shed cells to the tumor cells, the primary tumor cells.  When they compared the tumor assays and clonogenic assays, they were significantly decreased comparing the shed cells to the normal cells to the ones in the primary mass, and they had an increased apoptotic fraction.

So this makes sense.  It actually makes sense, because we knew if this wasn't the case, all patients would basically die, and we'd have no real hope of curing it, because at a 1 centimeter lesion we know that they are shedding cells all the time.  They are a very different population than a primary lesion.

But this is really systemic to these and how can we understand in some ways some of the molecular characteristics of this to understand this property.  I think we need to use this, and to realize that this is going on, so we can get a better understanding, a better hold of what the real biology is of what we are dealing with.

TOP
Slide 11:

So there are several different ways I think we can approach this.  And again, this is just one approach.  This is one general scheme of things that we can understand, but it's a general scheme I think in the direction we are moving in.  The first thing that is particularly germane to the discussion, which we have had here, is imaging targets and molecular targets.  There are a number of different targets.  From an imaging perspective, actually we can choose a number of different things.  Once somebody discovers a target, then the onus is upon us to go ahead and try to figure out how we can image it.  But we can look at a number of different metabolic properties, as we have seen in the model with FDG and glucose.  We can look at membrane receptors.  Right now our lab is working on a very interesting model of mutant EGF receptor, which is really a tumor-specific receptor.  It's a genomic mutation, a very interesting model.

Then as we have also seen, when you section tumors, sometimes only 50 percent of the lesion turns out to be a tumor and there is all this support matrix and other infiltrating cells, whether it is the host response or whatever.  But there is a whole series of other elements and factors that go into it, so I just put this in the general support matrix.

So it doesn't really matter what we go after, but we need a target.  And a target is something that is either going to be over-expressed, something which is going to make it unique to the tumors or something which we can differentiate the tumors from the surrounding tissue.

TOP
Slide 12:

When we look at that, and again this is just one immunohistochemistry stain from our lab, looking at the mutant EGF receptor, you can clearly see that it was very positive.  So we chose this, because we know that this is a very tumor-specific receptor.  It's very different from the wild-type EGF receptor.  Again, it just doesn't matter, but you need to choose something and focus on that receptor, at least as a model system.  Then once we get enough of these targets, hopefully we'll be able to build this tumor profile.

TOP
Slide 13:

The second thing we need is techniques.  I'm going to talk about this very briefly relative to what we have right now.  I mention this because I had some discussions with several of my radiologic colleagues yesterday, talking about new techniques coming on.  I honestly cannot give you a lot of information about things such as optical imaging, because I don't know that much about it, and we are just learning.  But what I can tell you is something of what we have right now.  What we have right now typically is the imaging techniques that we are using, CT and MR and nuclear medicine.  One of the reasons we have gone to nuclear medicine is because of what I show here.  If we want to direct it against tumors, there are some very interesting properties which we must recognize if we are actually going to target tumors directly, use a specific target, use a technique, and then try to use tumor-specific ligands.

The interesting feature is that if you look at CT/MR and you are trying to use a tumor-directed ligand, for CT, approximately 1 milligram per cc just to raise the so-called 20 HU when you enhance it, it turns out that you need about 109 molecules per cell.  This is an awfully hard number to get if you are looking at a tumor-specific agent.  We are not talking about intravenous now.  You can get intravenous contrast to get at this concentration.  We're talking about actually tagging tumors, to tag the tumor cells or some other target.  So this becomes a very difficult proposition, and it's slightly less for MR, but still we're dealing with millimolar quantities.  So it makes it very, very difficult to direct a tumor-specific imaging agent for either CT or MR.

When several people have approached me about doing this, we have tried to work the numbers, and as far as we can tell, it's going to be an extremely difficult proposition to actually get a specific tumor imaging agent.  This doesn't mean that we cannot get other surrogate properties such as diffusion or perfusion, and that we want to look at the vascularity of the tumor.  But we are looking at a marker of that.  We are not directly looking at some of the endothelial cells.  We are not getting a tumor-specific agent at that point.  So we have to understand the differences.  But when you look at nuclear medicine studies, we have calculated you need about a microcurie per cc, or it turns out about one molecule per cell.  So we are orders of magnitude more sensitive when we are looking at nuclear medicine techniques.

I am actually a thoracic radiologist.  I mean I was trained basically doing mostly CT and MR and anatomic imaging.  But we have gone much more to nuclear medicine and certainly PET, because of the issue of sensitivity if we are going to direct this for tumor-specific.  Now when we look at CT –

TOP
Slide 14:

this slide was lent to me by Steve Swentz at the Mayo Clinic.  He has done some very nice studies looking at differentiating nodules with contrast enhancement.  There have been a number of studies now and with this slide Steve sent me you can see before contrast enhancement, and after contrast enhancement.  What they showed, and has now been shown in multi-institutional trials, is that, yes, contrast enhancement can most of the time make the distinction between benign and malignant pulmonary nodules.  It is not 100 percent specific, because some benign nodules will have increased uptake.  But this, remember, is a very general property.  This has to do with intravenous contrast.  This does not have to do with specific tumor tagging.  So while we can use this, and it gives us some information, I think we are lacking in some of this information.

Can we also look at other ways? 

TOP
Slide 15:

Another property, which I show you in this slide from Al Johnson at our center for in vivo microscopy, these are really pretty amazing anatomic images of angiogenesis.  We can look down -- and these are 5-millimeter images -- at the vessels and get a very good idea about angiogenesis here.  In an animal model, this is exquisite.  Again, I'm not sure that by finding this in an anatomic site, it will give us all the information that we need, nor is necessarily any better than simply measuring the tumor at this point.  I don't know that, but this is something that we have to pursue.  These areas are very interesting but I think there is more that we can do.

TOP
Slide 16:

The next area which we look at often are ligands.  We have a target, we have techniques, and I think if we are working towards that molecular characterization noninvasively, often we are going to need to develop new tumor imaging agents.  And there are several different properties which we often look at, very different from drug discovery for therapeutics, even though that's the same sort of similar game that we're in, but we have a lot of different properties which we have to look at.

First, obviously low dose and low toxicity.  We need it to be easily metabolized and excreted, and again a lot of times we have learned the lesson with monoclonal antibodies that we would like something which is not immunogenic, so we can repeatedly give this to patients.  There are ways of manipulating this, and looking at small molecules which actually can accomplish all of this.  And  if we choose the targets appropriately, we will be able to do this sort of imaging.

There are several different models that I would like to show you. 

TOP
Slide 17:

One of the studies here, which I think is a very elegant model, comes from Memorial, and they have done some exquisite work looking with gene therapy.  Now again, I'm a little bit confused sometimes about looking and knowing always what I'm looking for when we want to so-called Aimage@ gene therapy, because I know that you can transfect cells, and you can have very high transfection rates, but it doesn't always correspond with the clinical response.  Just because I can image it, doesn't necessarily mean that that is going to tell you what the clinical response is going to be.  We need to do the studies.  But there are models out there that tell us that we can do it, and this is a very interesting animal model.  What they do is put a so-called marker gene, the TK gene, next to the gene that you are looking to transfect the cell with.  So whatever gene you are interested in introducing, you put a TK gene next to it.  And wherever that's expressed, once this gene is expressed, then you can inject intravenously a marker substrate.

This is just an analog of ganciclovir.  So what happens is you inject it intravenously.  Once the enzyme is expressed, and theoretically this should only be in the cells where the TK gene is expressed, it phosphorylates this, and it traps this agent.  So here they have placed subcutaneous tumors into the nude mouse xenograft.  Here you can actually see that it uptakes into those cells which have the TK gene in it, and it does not into the wild type.

Basically this is a very elegant model of actually looking at and knowing where the gene is expressed, or at least knowing that the gene that you are interested in is next to the TK gene, and we can actually follow this.  This is a very interesting model and a very powerful model that we can do right now.  Again, I'm not sure, but I think what we need to do is correlate that in the future.  Just because the gene is there, does that mean it has any real biologic effect?  That is something that we need to do and need to understand.

TOP
Slide 18:

The next model here is a very interesting situation.  The paper was published by a group in San Diego in PNAS recently. We have seen this, and this is basically where they transfect the cells with the GFP or the green fluorescence protein.  If you need to know whether you have a response or not and you want to know if it's tumor cells, here is a way of actually looking at them.

Again, it is a very interesting nude mouse model, where you can section them and look at them in the axial images, or do this with actual optical imaging, which is what they did here.  Again, a very nice model for looking at tumor cells.  This gives us some information about whether a tumor exists or doesn't exist.  The previous study tells us basically what we are seeing there is the gene being expressed, but I still think we need to take it a little step further. 

TOP
Slide 19:

One of the things that we are working on, and which we found very interesting, is looking at concommitory libraries and phage display.  I find these to be incredibly powerful and robust techniques.  Here this is just a cartoon illustrating the fact that you can have random peptides.  You put it into a phage vector next a gene 3, which expresses this gene, expresses a protein, which basically at the end once you have put these in the bacteria, you get the virus out of the bacteriophage.

They have this CAP protein, which is a presenting protein, and then you have your random phage out here.  This is a basic technique that has been established.  Although it's not very easy, it seems a lot simpler when we look at the cartoon.  But what we have done is use this to look at tumor antigens.  As I have mentioned before, we have chosen the mutant EGF.  It's just a model system.  What we have done now is we’ve screened, and we have panned a number of different libraries, trying to develop new tumor-specific imaging agents directed against this tumor target, against a very specific tumor target.

I think this is a very powerful model, because once we develop the targets, what we are going to be able to do then is come up with a series of agents.  And again, the same way that you had the gene chip arrays, we are going to hopefully be able to do this noninvasively, and provide you with the noninvasive array, a molecular array, and provide you 

TOP
Slide 20:

with what I think is the most important thing which we are going to look, which is tumor profiles.  We are going to be able to do this noninvasively.

What I'm hoping to do is tell you it is not just in non-small cell carcinoma.  It's a tumor which has certain properties, and whether we are looking at p53, ras, we are looking at EGF, whatever we decide are the appropriate targets and can give you enough information to provide you with that information which you need for both therapeutic and prognostic information, that's what we will come up.

So we’ll see, we'll come up with the same way that they colored the microarrays, we'll be able to do the same thing.  We'll tell you what the pattern expression is of that tumor.  And if we could tell you noninvasively the molecular pattern, it almost won't matter what the stage is, because you will know that it has this pattern.  Once we correlate it with the biology, and correlate with epidemiologic data, we will know that these patients will have a predilection for brain metastasis or for bone metastasis.  We will know what their therapeutic outcome will be.  We'll know that they respond or don't respond to this.  And this is the most important thing that I think we are trying to work on.

Again, it almost doesn't matter what the surrogate markers are that you use, or whether it is a marker directly, directed actually against genes, directed against proteins, directed at looking at angiogenesis, as long as we can then correlate that with the outcomes that then give you a very good idea of how to stratify these patients in the appropriate categories.  As we saw, it is not appropriate, or at least right now the most appropriate way in which we look at them is by stage.  But we know that that actually doesn't work.  That doesn't work well enough, and we need something better.  And I think that this is what we are going to be doing in the future in trying to create these molecular tumor profiles.  Again, going outside that black box, going outside anatomic imaging, and provide you with this sort of information.


TOP

Slide 21:

So I would just like to leave you with  some information and a general overview that I think the chest x-ray CT and MR are here to stay.  They provide a tremendous amount of anatomic and morphologic information.  It's unbelievable, and they give us a lot of information, probably a lot more than we realize.  But I think we are going to look for more tumor-specific targets to give us biochemical, metabolic, and genetic information.  We are then going to develop tumor-specific imaging agents.

TOP
Slide 22:

We are then going to create these molecular tumor profiles to again provide diagnostic, therapeutic, and prognostic information. 

I always put this at the end, to decrease mortality, or improve outcomes, and that is the bottom line. As much exquisite anatomic imaging as we have right now, I hate to say that our outcomes for lung cancer still have not significantly changed.  We have to be very honest about that.  We can do a tremendous amount with imaging, but we have to be realistic about the bottom line.  Once we have both the diagnostics and the therapeutics, I think we will hopefully in the near future, will come up with something to decrease lung cancer mortality.

I would like to thank you very much, and stop here and entertain any questions.

[Applause.]

TOP
Slide 23:

DR. GANDARA:  Ed, tell us about currently available other imaging agents besides FDG.  For instance, yesterday we talked about a new class of agents, hypoxic cytotoxins.  So being able to look at hypoxia and changes with administration of an agent.  And of course you mentioned earlier the anti-angiogenesis agents.  What things could we conceivably put into clinical trials now, not necessarily on a national basis, but at selected centers that might have expertise in functional imaging?

DR. PATZ:  I can tell you that we do a tremendous amount of PET imaging, probably more than anybody has done in the world, and it's exclusively done with FDG.  I can tell you that the only other real agent out there is octreotide, which looks at the somatostatin receptors.  And again, we don't have a lot of experience with that, because we had the advantage of actually having PET.  And that can be used as a SPECT agent, a single photon agent.  I really don't even have any experience with that.  I can tell you from my experience right now the only thing we really have is FDG, which people have available.  There are a number of other agents, I understand in talking to John, and seeing what was presented to the Society of Nuclear Medicine recently last week, that there a number of people working on them.  But there is nothing ready for prime time right now.

I'm not sure that anything is better than FDG at this point to give you any more information.  And again, I think what needs to be done when we do these studies, is if you have a target, when we develop the agents, there need to be very hypothesis-driven studies.  And we need to show this in small pilot trials before we go out and put something in a large trial.  I can tell you from my perspective, there is nothing out there that I know of which is ready for prime time besides FDG.

DR. BUNN:  To take that further, in the angiogenesis session yesterday, we had a discussion about the fact that in lung cancer it is unlikely that anybody is going to have tumor tissue before and after the administration of an angiogenic agent to assess what happens.  Therefore, the only way to predict and assess is going to be with some imaging tests.  So one of the recommendations was going to be that considerable resources be put into the development of agents.  The problem was nobody in the room really knew what the various agents were for assessing angiogenesis.  So maybe you can just give a few comments about what the various options are with ultrasound and contrast, with MR, with PET.

DR. PATZ:  I will make a few comments, and I see some of the other radiologists also want to make a comment.  I'll make a comment from my perspective, from a nuclear medicine perspective, and tell you things that I know about CT and MR.  First of all, imaging angiogenesis itself, looking at the vessels, the endothelial cells is an extremely difficult proposition.  We have gone through this.  I have gone through this with several different companies saying we want to look at the endothelial cells.

Unlike drug discovery for therapeutics, diagnostics are a very different challenge.  First of all, if you are looking at receptors, and that was one of the things that we went through, we looked at the tie-2, the FLK, the VEGF receptors.  You need a certain number of receptors to actually do this imaging.

I can tell you for several years that we found and we studied opioid receptors in lung cancer.  We found a compound for imaging which had picomolar affinities, very strong, high affinities, yet there were only 10,000-20,000 receptors, and there was no way in the world  we could actually see it.  We just don't have enough to actually get a signal.  You have to look at certain things to understand whether you are going to get a signal from doing it by nuclear medicine.

It turns out when you look at some of the angiogenesis, again the VEGF receptors, you don't have enough.  And it's so variable, there is so much background, that when we did the studies -- and we have done the preliminary studies in the cell cultures, and we have looked at this -- how you are actually going to image it?  I would be very interested.  I have not been able to figure out how you are ever going to get a signal out of this, because there is so much background, because you just don't have enough signal, until we find a more specific target, which we don't have right now, specifically for endothelial cells and for angiogenesis, saying that, on the other side of it there are some techniques with MR which can look at perfusion.  Certainly you can look at vascularity, a general property of how much is being perfused.  You can look at contrast enhancement with CT.  Whether that will correlate or not, I think the right thing to do is do it -- we have animal models.

We can do it in animal models first.  We can try it, and look at it -- you will have some tissue, and I don't know how much from the lung cancer beforehand --  and do some correlations.  And whether it will hold up or not, I just don't know, to tell you the truth.  I think it's a very difficult proposition, actually, imaging angiogenesis per se, and I think you need to be very careful about it.  I think there are other surrogate markers and different techniques.  Matt or Ed may want to comment on that also.

DR. FREEDMAN:   The first question was asked, can we determine whether or not cells as a group structure, in other words, an area within tissue, are ischemic.  The answer is that there are MR techniques that suggest that one can do this, from other organ systems.  In fact, both of them are related to MR spectroscopy.

MR spectroscopy can show you changes in the amount of lactate within tissue.  Lactate is obviously anaerobic metabolism, an  effect of anything that's anaerobic, but ischemia will do that.  The second thing is that, in the brain, in some work being done at Georgetown, it has been shown that we can detect changes in pH by ratios among different metabolites in MR spectroscopy.  This is proton spectroscopy.  It has only been done for the pH as far as I know in the animals, rather than in humans, but that is also something that changes with ischemia.

Looking at angiogenesis is a problem we are working on at the moment, and we do not have answers, but we have leads.  We cannot tell you whether or not vessels are growing or regressing.  What we can tell you is the amount of perfusion of tissue.  According to the people that I'm working with, anti-angiogenesis agents result in an diminution of newly formed vessels.  So can't tell you that a vessel has not formed, but we can show you a change in the amount of vessels in tissue, perhaps.  And that perhaps is important, because we don't yet have the data to know that this is what we are really seeing.  We can look at vascular flow.  We can look at a combination of perfusion and diffusion of our agents into tissue.  But we don't know which is the agent or agents that will show us exactly the thing that correlates with histologic change.

We can use magnetized blood, which is something that we are developing, which involves no injection in the mice.  We have not yet replicated, but there are people who are injecting gadolinium, looking at perfusion of tumors in nude mice.  If there were a change, and you could accurately quantify the amount of injection and the rate of injection, you should be able to see a change in blood flow.  You can label RBC's I believe with carbon monoxide.  That has PET-labeled carbon monoxide attached to hemoglobin, and you can monitor the blood flow into this.

There are also other experimental agents, both nuclear agents and MR agents that are not yet proven, but are designed to go specifically to characteristics of tumors.  These are proprietary things that various companies are working on, and they may become available if they pan out in studies.

You have another method, which is related to ultrasound.  You can look at vessels down to the arteriolar level with Doppler and power Doppler ultrasound in mice, and in people.  You can also use what are currently for this purpose experimental ultrasound contrast media that basically have several different properties.  They come in different sizes, so presumably if you knew exactly what you were doing, someday you might be able to characterize the size of the vessels by knowing whether or not these agents are trapped within the vessels.

You can look at the rate of flow into tissue in cardiac models in mice.  Therefore, you probably can do it in tumor.  If they can tell you where ischemic areas are, and the relative ischemic areas, we should be able to do the same thing in tumors in mice, and obviously in people.

There are agents there that are trapped by capillaries.  There are agents there that in various ways diffuse into the tissue.  So there are a lot of opportunities.  There is no answer yet for the particular model of angiogenesis, and how we can detect the change or ischemia in tumors, but there is a lot of evidence from other organ systems, and a lot of promising leads that I think should be pursued.

DR. MABRY:   I found your concept of using phage display generated peptides to look for specific molecules a really interesting idea.  Are there any differences if one is looking at proteins in the cytoplasm versus nuclear proteins versus extracellular receptors in terms of efficiency?

DR. PATZ:  We don't really know that yet to tell you the truth.  We have taken a very simplistic approach, and that is getting it to the cell surface right now.  Getting it into the cell is another whole problem.  We do it with FDG.  Theoretically, that's how FDG works.  But I can tell you that we really haven't done it with the small molecules.  We have focused on the receptors, because we find that to be the easiest actually to get it into the tumors and just sit on the cell surface.  But I don't know of any definite difficulties as long as you can get it into the cell, and you are targeting specific property, no.

DR. MABRY:  So you mentioned, for example, p53 as a potential target.  So there is no direct data showing that you can get that in in a reproducible way yet?

DR. PATZ:  No, there is not.

DR. HOFFMAN:  I just wanted to make a comment.  I'm chief of the Molecular Imaging Branch of NCI, and we have put forth several programs, funded many centers -- many of you are actually associated with those centers -- molecular and cellular imaging centers to deal with many of these issues.  This is a new and evolving area, obviously very complex.  And it needs a lot of work -- again, much as Ned said, the targeted approach.

We can do a lot of the more generic imaging of perfusion/diffusion, looking at downstream effects, but it doesn't really get at the real subtleties of the situation. Even with angiogenesis one of the ultimate abnormalities is there may be reduced FDG uptake if the therapy is effective.  And one can do that.

So there are these different techniques.  We had a workshop in Lake Tahoe in February where we got the worlds experts in angiogenesis together, and we are coming up with a document that will be available on the current state of angiogenesis imaging.   Imaging is obviously very powerful, and it is going to be very important, because you do it serially and noninvasively, and its power is obviously there.

But again, there needs to be a lot of interaction in these molecular and cellular imaging centers, trying to get these people to begin interacting with each other, and discovering these specific powers.

I wanted to make one other comment.  There are  various hypoxic agents.  These have primarily been done with PET, but as Ned mentioned, they are really not ready for prime time.  Washington University, as well as the University of Washington have developed several of these compounds.  But again, their widespread clinical applicability just isn't there yet.

DR. CURRAN:  For the purposes of the group assembled here today, do you see any value in exploring MR spectroscopy further in thoracic imaging?

DR. PATZ:   I have had numerous discussions with our MR spectroscopist.  I'm not an expert in this, but trying to cover all the bases. I can tell you that spectroscopy is on a continuum.  In some ways like FDG, whether it's hot or not hot, whether it's there or not there, it's even much more difficult.  Spectroscopy looks at various different parameters, and various limited parameters in some ways that we are looking at.  I don't think that it's going to give us the entire spectrum.  I don't think it's going to tell us what exactly we need to know to give us the information which I have showed you, some of the molecular characteristics of the tumor.

I don't think it's going to be able to do that.  I really think that it's very limited in its capabilities.  It can provide some exquisite information, but I do not see that being the way that we are going to be going.  In particular, remember that sometimes we are dealing with a 1-centimeter lesion.  Now trying to do spectroscopy in a lung on a 1-centimeter lesion is a very difficult proposition.  I know we are getting better resolution.  I understand this.  But I think it's going to be a very difficult proposition to actually molecularly characterize that lesion by MR, by doing spectroscopy in the situation, at least right now.  I think that maybe in the future, but I think again, we are looking at a continuous spectrum, and it's not clear to me, particularly given a lesion that is 50 percent tumor and the rest are lymphocytes and everything else, you are going to be swamped with other signals when you are looking at spectrum of things.  So I would say be very careful about doing MR spectroscopy.  I haven't seen any data to show that it will work, and would give us the information that I think we require.

DR. ROTH:  In our breakout group one of the conclusions was that we need intermediate endpoints for lung cancer clinical trials, surrogate markers for survival, and PET scan was mentioned as one of the possibilities.  I wondered if you could comment on the limitations of PET scan, and the scenario of multimodality therapy.  Most of our patients are going to be getting chemotherapy and radiation, so there is going to be a lot of metabolic activity in the tissues.

What are the limitations of PET scan here in post-treatment imaging in a patient who's been very heavily pre-treated?  Is there any kind of optimal window for imaging these patients?  And finally, are there plans for large validation studies to begin to look at PET, either alone as a possible intermediate endpoint, or in combination with say pathologic response, or other radiologic markers for endpoints?

DR. PATZ: To answer your first question, I think there are some studies -- and we published a study --  and I know David Gandara talked yesterday about it --  in AJR last year.  We looked at patients post-treatment.  I didn't want to pigeonhole patients into who had this treatment or that treatment, and it's two weeks after, six weeks after the PET.  We just did a very generic study.  What we did was look at everybody who had non-small cell lung cancer and had been treated, regardless of the time, regardless of the therapy, and we had a very clear, statistically significant separation of those who were PET-positive, versus those who were PET-negative.

The question often comes up, when do we know that the PET is going to turn negative, and when can we use that?  Nobody knows the answer to that.  I think to do that appropriately there needs to be a study, and there needs to be the study where you serially look at PET imaging to know when it actually has prognostic implications.

We know that once you have been treated, normally we wait 6-8 weeks after treatment to try to do the PET.  Whether that is right or not, I honestly don't know.  We have probably done more than anybody else has.  I don't think we know the right answer right now.

I can tell you from our study that it's very clear that once you have been treated and you're PET-positive, you have a significantly worse prognosis than those who are PET negative.  But again, the time course of when to look at it, I don't know.  We have often been asked this question and that was one of the reasons to do this study --  can we use PET in a similar way that we use gallium?  After two cycles, if we know that you are getting better, or if you are getting better to continue it after two cycles, that you are getting worse, do we change therapy?  Again, these are all things that need to be studied.  There has been no study.

On your second question, basically I think there need to be some validation in a larger trial, although I think the data in even the smaller trials has been that PET, in most cases, has been very consistent.  In particular, in some of the trials that we have done, we've looked at survival data.  We have looked at outcome, so not always having pathologic correlation, but then looking really at the bottom line, which was outcome there, and looking at some of that data.  I think the data right now for some of these scenarios, and some of the cases where you want to use PET, is fairly clear but certainly larger validation studies could be done.  I don't know that any are constructed right now.

DR. BUNN:  There is a study, which is the BLOT neoadjuvant study.  Dr. Shields from Wayne State is the PI on an ACRIN supported study.  And patients have a PET scan before neoadjuvant treatment.  Then they have an optional PET scan after one cycle.  Then they have a PET scan just prior to surgery.  Then both pathologic and clinical response are obtained and will be correlated, not only with the changes in PET, but SUV values are also being obtained.

At the ASCO presentation, when the question of timing was asked, they said, we did them at two months and it worked.  And somebody said, why did you do them at two months?  And they said, because that's when we did it, and we don't know whether that's the best time.  But two months worked in that study.

DR. PATZ:  We normally suggest, and what we found is, 6-8 weeks, but there is no hard data on that.  That was an arbitrary selection.

DR. OKUNIEFF:  We have lots of attempts to look at living cells after we give treatment.  But maybe even more important would be an imaging approach that looked at dead tumor cells.  So that very early after you give your first course of chemotherapy, you don't have to wait several cycles.  You could see if it was working, or the same with radiation or combination treatment.  It would also potentially provide a very rapid surrogate marker.  Is there anybody looking at imaging of the cells that are killed, or the fraction that are killed?

DR. PATZ:  Not that I know of.  I can tell you that there are so many different agents.  I started looking through abstracts of people and what they are looking for, I imagine they are looking at a huge number of properties.

DR. FREEDMAN:  There was one paper using MR spectroscopy to document apoptosis in experimental animals, monitoring this every two hours I think for 48 hours.  That was a specific study.  It has not been replicated to my knowledge, but it looks like a very careful study, and we are going to try to replicate it.

DR. PATZ:  So it was looking at cell death or hypoxia?

DR. FREEDMAN: This was looking at cell death.  This was correlated with apoptosis, which is cell death.  So it was correlated with the TUNEL assay.

DR. HOFFMAN:  An indirect measurement is actually FDG.  You can get a volume.  You can get what the activity is.  I mean FDG represents in many ways the bulk of viable metabolizing tumor.  So indirectly Rich Wall has done some of this.  If you have treatment, and there is a significant reduction in FDG, you can do some calculations, and probably get some idea of the volume of very ill or dead tissue, but not directly.

DR. ADJEI:   There is compound and it's in vivo, which is a mark of apoptosis.  And as far as I know, it hasn't been tested clinically in patients yet, but in animal models with tumors that have been treated.  And in vivo imaging it correlated nicely with apoptotic cells.  So imagine the animals getting the cells out and doing in vitro assays as a nice correlation.  I believe there are some studies going on with that compound.

DR. PATZ:   I'm going to put on my clinical hat right now and say that we have a lot of these that are very interesting compounds.  They are very interesting techniques of doing this, maybe worth it certainly when we are validating some of these compounds in animal models or even early clinical studies.  But I think we have to remember one thing.  When we look at it, is this actually going to be any better than looking at a chest film when we are looking at the outcomes?  I think we need to be very careful about doing that, because we have all these expensive techniques right now.  But we need to then go ahead at look at the bottom line and say, “is this better than what we have actually got right now?”  And I think we need to be very careful about that.  I just want to make that point clear.

DR. STAAB:  While we are trying to develop all these new markers and ways to go, and try to learn what is good and what isn't, I don't quite understand what the reluctance is to repeated sampling of tissue.  I think that we can do this with fine needle aspirations with very low morbidity -- in fact, probably none in today's service.

DR. PATZ:  We could.  It's unusual for us.  Obviously, it's the unusual circumstance where a patient will agree to have biopsy afterwards.  We can do that, but remember, when you do that, we are looking at cytology.  So you need something.  We are not getting any properties.  We're not getting any stroma.  We can do a core biopsy.  But the other issue that often comes up with lung cancer, I can tell you, is that when they have been treated, I don't know what the viable cells are, and where is tumor in fibrosis.  I'm not always clear when I'm biopsying some of these patients to prove what they have.

So we can do that, and absolutely in studies of patients who are willing to, and if we need to prove some of these things, it may be a very viable option.  But you have to remember, sometimes we can get a core, and sometimes we can't.  Sometimes we don't know what exactly is tumor, and what is fibrosis, so you may be biopsying somebody, and you just have to be careful about it.  But clearly, we always say in our department if we have a needle and you have a lesion, we can stick it.  We can biopsy anything, and that is certainly possible.

DR. WONG:  This is more a question for Ed Staab or John Hoffman from the NCI's Biological Imaging Program.  Ed was talking with me yesterday about BIP had an initiative last year in getting small animal models for imaging the new molecular targets.  Ed, can you give the audience a little information about that program?

DR. HOFFMAN: We funded five centers to do small animal imaging resource programs, as they are called.  We will be reissuing that at some time in the near future.  And we are working closely with the mouse model of the Human Cancer Consortium to try to get imaging integrated into all of their cancer models.  And again, many of you are at institutions where they have these particular funds.

I think that that will be an extremely powerful technology, because animal imaging capabilities now are quite exquisite, the new generations of the small animal imaging PET system are actually about 1 cubic millimeter in resolution.  This is extremely high.  There will be a lot of development in this area, small animal MR techniques.  And then there will be the translation into the human counterparts of cancer.  It is very elegant, because you can scan the animals serially, get all of your pathology correlations, and  that is one of the real issues of correlating what do we see on the imaging with what the pathology is.  And this is one of the missing components right now.

DR. BUNN:  Since Ned brought it up at the end --  and he's taught me everything I know about radiology, which is why we have PET in Colorado, but is also why I get chest x-rays on all my patients and read them myself -- two of the most annoying things that happen are basically the FDA wants for a response, CT scans.  And many companies want to do a CT scan every cycle, which is the biggest joke in all mankind.  But now we have a new RECIST system.   It shows that a unidimensional measurement is just as good.  And the question is, what RECIST says is also you still need a CT.  The question is, are we ever going to get back to where a chest x-ray will work?

DR. PATZ:  Unfortunately -- I'm just going to take one minute to answer this question, because Dan Sullivan had originally asked me to look at that as a consultant.  My name actually is on the back if you look at all those people.  I had written that this was a very arbitrary, capricious selection of numbers, and that actually when you go back and you look at tumor measurements, tumor measurements come from no scientific data whatsoever.  It was an arbitrary selection by a physicist who decided 50 percent meant something.  It means nothing.  It is based on no scientific data.  In fact, I know of no studies in solid tumors that show that unless you're a CR, if you are 5 percent better or 10 percent better or 90 percent better, it doesn't matter as far as your survival.

I said why don't we get back to where we look at basically what some of these things mean, and this means nothing.  Whether it's unidimensional or  bidimensional, it means nothing.  You are looking at an arbitrary selection of volume, 50% are tumor cells, 50% are inflammatory cells, 80% -- I don't know what this means.  I think it's all nonsense.

So basically, the bottom line is I can look at a chest x-ray and I can measure it.  I can tell whether they are doing better or not.  I think the rational for the RECIST was that they wanted to try to systematically standardize everything.  That was the response I got.  It wasn't to ask questions about how we do it.  It was just so that we could come up with a standardized criteria so that we could eventually learn to compare them.

That was the impetus behind RECIST, and I think that's why they wanted to do CT’s.  They said you need contrast.  That's a bunch of garbage.  Actually, we did the study.  You don't need contrast in these people.  It makes no difference whatsoever.  So they based this on absolutely no real foundation, but simply just to standardize things so we hopefully will get the data, analyze it, and then we can come back.  The response criteria again is we use it.  We use it for drug companies who want to look at response.  The FDA requires it.  I've been to the FDA.  So this gives a so-called objective measurement.

DR. BUNN: Why do they actually say in there that you need a contrast CT?  How could they possibly say that?

DR. PATZ:  They are wrong -- I did the study.  I wrote the study showing that it makes no difference, contrast versus no contrast.  I told them that.  And there are no other scientific studies whatsoever out there.  If anybody has got any other data, I would be glad to look at it and see it, but I can tell you that the only study, particularly in lung cancer, looking at staging management. It made no difference, contrast versus non-contrast.  I don't know why they did it.

DR. STAAB:  Ned, do you have any firm feelings about this?

DR. PATZ:  Do you want to know about screening now?

DR. SAXMAN:  Thank you.  I don't know how to follow that exactly.  The rest of the morning we are going to spend with the summaries of the breakout sessions.  For those of you who weren't here at the last meeting, this is not meant to be a passive exercise.  I'm not sure I need to tell this group that.  But to stimulate further discussion, questions, arguments, disagreements about how we should begin or how we should integrate these new therapeutics that we discussed yesterday afternoon into the treatment of patients with this disease.

The first breakout session will be breakout session A, which was the anti-angiogenesis session.  The moderators were Dr. Paul Bunn and Dr. Curran.  Unfortunately, Dr. Curran had an emergency yesterday, and couldn't make it.  He is, however, with us this morning, so he'll be able to participate in the discussion.  So Dr. Bunn from the University of Colorado is going to present the results of that session.

TOP