DR.BIRRER: Thank you, Bob. This is now the discussion part of the session. I would start off by throwing it open to the audience, if there are any comments or questions related to this presentation, and then we can redirect the discussion.
DR. OZOLS: I think the biomarkers, particularly CA 125, can certainly accelerate the evaluation of drugs, but you have got to get the FDA to buy into this. If they don't accept this, the drug companies and biotech companies are not going to do this.
DR. BAST: Let me reiterate, I think the problem is not the FDA. I think it really is with the drug companies and the investigators.
What I was suggesting was, for a successful phase II trial where you are going to carry a drug forward to phase III, you could invest the number of patients you need in RECIST criteria, so the FDA would be cool with that. I don't think there would be a problem.
The issue is, would you kill a drug based on a rising CA 125 in most of the patients, in the first 14 or 17 in the trial. That is really a decision for companies and for academic investigators, rather than for the FDA, I think.
DR. OZOLS: I meant those patients where the only measurable disease would be the CA 125. It would be nice for the agency to accept that. It would increase the number of patients eligible for those type of registration trials.
DR. BAST: Again, at least up front for the few patients, many patients who would be otherwise ineligible could be entered in those trials. I guess we don't have an FDA person present today.
DR. HOWELL: One of the things we could do as a group is to take on the challenge of validating CA 125 changes specifically for that FDA function.
In other words, we have to convince the FDA that this really is a valid biomarker for the purpose of approving drugs. We can accept that challenge.
I think Gordon Rustin has got enormous levels of data on this. I think there was a myth a few years ago that you couldn't measure taxanes with CA 125. I think that has largely been outweighed by increasing evidence.
I guess the other thing to point out is that whether you use CA 125 or some other biomarker for this, it would be important in preclinical studies to be sure that the drug you are evaluating didn't affect the shedding or the expression of the biomarker.
One of the things we have had trouble doing in tissue culture is changing CA 125 levels with most of the designer drugs that we are currently trying. So, for better or for worse, that doesn't help our laboratory studies, but it might be relevant to making it a more useful marker in the clinic.
DR. MARKMAN: I would like to directly address the point that Bob was making. The issue is not so much demonstrating how active a drug is, because ultimately that will require large trials, and comparative trials.
The issue is rapidly screening novel drugs to see if they have activity or if they don't have any activity that would lead forward.
This is an ideal marker. There is a very serious problem we currently have -- and I support it -- and that is, if you have a patient who is potentially platinum-sensitive recurrent disease, it is very difficult to justify taking a symptomatic patient with a large mass and not give them platinum, because you think they might have a 50, 60, 70 percent response.
In the recurrent setting, where an asymptomatic patient exists, but has a rising CA 125, that would be a very interesting window to take a novel drug and CA 125 criteria.
The point is, if there is activity, it says we ought to study it. If there is no activity in 15 or 20 patients, you will know that in four weeks with each of those patients, maybe six weeks, and you will abandon the drugs.
You can encourage companies to look at very novel drugs where they might otherwise not look at it, and we might get some very interesting ideas.
So, the idea is, in a limited number of patients, to rule out any activity in a setting where you would expect to see activity, and a marker like this would be very, very powerful.
DR. BOOKMAN: CA 125 as a marker between GOG and CTEP yesterday, there is interest in trying to move forward and allow that to be part of the decision rule in phase II trials.
The problem is, in a non-randomized, single arm phase II trial, where you are trying to see if a new agent crosses a bar for activity, we haven't really defined a bar with regard to CA 125.
If you enroll patients where CA 125 is really the only site of disease, or CA-125 with small volume disease that is not resist measurable, in general, looking at phase II studies that are published in the literature, they respond as much as twice as often to a given treatment in a patient with RECIST measurable disease.
So, we need to gather some data, build a database, do exactly what you are saying, validate the process, so we can know where we need to set a bar to have an effective screening tool and then be able to, with some authority and some experience, take that data to the FDA. We don't have that today.
DR. SEIDEN: Switching briefly to genomics, does the group think there is enough genomics data to start separating patients as far as clinical trials?
For instance, should clear cell be a totally separate disease entity, even if it requires international collaboration? Should BRCA carriers be a totally separate group?
DR. BIRRER: I think the BRCA-1 issue, I think that is actually important issue that it would be nice to have even partial consensus on, which is what Mike is voicing.
Are we still in a discovery mode? I think Jeff summarized it very nicely. Are we still in the discovery mode, or are we moving into a validation mode?
I think it kind of depends upon the question being asked, my personal feeling is. I think we are beginning to get, from a predictive standpoint, for sensitivity or resistance to certain chemotherapeutic agents, and potentially prognostic.
We are accumulating a fairly large database that is now available, and we have gene lists. I think those are beginning to mature, such that we ought to start thinking about the next five years' validation studies.
For the biology, as Jeff pointed out, there may be certain questions which -- and certainly for activated pathways -- we are still in a discovery mode.
It is important to think about that, because that is going to dictate the type of trials we do, and the type of tissues we are going to collect.
Discovery can be done on tissues, I think, that one doesn't need quite as rigorous clinical databases, but validation requires a very carefully collected, preserved, stored set of specimens from patients that are treated, and the entire clinical database that follows that. So, I would be interested in other opinions.
DR. ROSE: Deb, I don't think, has made it here yet. She has a number of slides of us and patients. She had a 170B that maybe Bob wants to comment about.
DR. BIRRER: I can't hear you.
DR. ROSE: On GOG 170B, which was the Avastin trial, there were a number of patients who say CA 125 is elevated markedly with initial therapy and, therefore, you may not be able to rely on that in a non-cytostatic trial. In cytostatics it may work, but in biologics, it may not.
An alternative approach is to establish measurable disease, study a small number of patients, see that there is a response, then confirm the response, and open a larger trial, maybe a randomized phase II trial, using both CA 125 and measurable disease, because you have already proven that you have activity.
DR KOHN. I think that Peter's point has some merit. We are using -- I think there is some validity to questioning the validity of CA 125, especially against measurable disease.
In our serafinid/bev trial, if I had stopped when there was an elbow in my CA 125, my patient whose CT scan is dramatically improving, continually improving, would come off study.
I think we really need to validate, as Mike also said, the CA 125 in this setting, especially with the newer agents. We don't want to lose something good where we have documented benefit.
DR. SWANSON: I guess the thing that surprised me, being a physician, was how slow the clinical trials were in ovarian cancer. Like the average length of a trial is seven years, and we are only expected to live five years.
I look back on the HIV drugs, where we used to use time to disease progression and time to death, and it took forever to approve crixovan.
Now, as a researcher, we use biomarkers, and our drugs are shown to be good for some patients in 48 weeks. Well, as I stated, my CA 125 has never been above 19, but I do want to survive this disease, and I am sure there are medications on the way.
I would really like the doctors here to consider using the Gynecological Cancer Intergroup definitions of response to treatment, and the definition of disease progression in your clinical trials.
I think if you started doing that, the approval of drugs for ovarian cancer, especially in phase I and II, would really speed up.
You know, women like myself, we make 25 percent of the women up that don't have an elevated CA 125. You know, I would really like to see it done.
I don't really care if other women get help, because I think the fallout will help all, and I want to survive this disease.
I think it is really sad that it takes an average of seven years to approve an ovarian cancer drug. Anyway, I said my piece.
I did have one other thing to say, and that is, I wrote Felix Frueh, who is the director of genomics at the FDA, asking him to consider doing this.
So, if any of you want a copy of the letter, you are welcome to get it, and you might consider writing him, too. He said that, if he knew more about it, the FDA would consider using these markers in research.
DR. BOYD: I am going to exercise imminent domain and just speak. I want to follow up on two points that Dr. Seiden made in the context of the current discussion.
One is, should we treat BRCA patients different, and I think the answer is clearly yes. We should treat them all with platinum up front.
The reason that BRCA-linked hereditary ovary cancers have a longer progression free interval and a longer survival, and there is now no question that they do, is because they are exquisitely sensitive to the drug that is used as standard first line treatment, and that is because BRCA-deficient cells can't repair double strand breaks that are created by the cell in an attempt to get rid of the inter-strand cross link created by platinum. That is the reason they do so well.
BRCA-1 and 2 deficient tumors, this gene represents an Achilles heel of the tumor toward a particular therapeutic agent.
Then, back to the clear cell problem, I think there are at least three studies now that have consistently shown that clear cell tumors are different in terms of their gene expression profiles, in those studies that I am aware of, to the contrary.
So, one thing that we could do, for example, as a field is to mine these data, attempt to find a potential BRCA-1 analogue in clear cell cancers that may represent an Achilles heel to a particular therapeutic agent.
We all know that clear cell tumors don't respond well to the typical chemotherapeutic regimens. They probably almost certainly would respond well to something.
We are now starting to generate databases of information that could be mined, in an attempt to discover what they might respond to.
DR. BAST: There is also a literature suggestion that mucinous are clearly different. I think GOG has come to the same conclusion clinically. Again, cirrus and endometrioid may be the group that is most typical of our usual philosophy.
DR. HAMILTON: I support what Jeff said. I have to say that I used to study DNA repair, and was very interested in trying to get pharma to develop agents against different aspects of DNA repair as opposed to gap filling.
The problem with that, it seemed to me, is that what you are asking pharma to do is to develop agents that make platinum better. That doesn't seem to be a very popular thing for other pharma to do.
DR. CIBULL: I think both speakers alluded to a study from the breast community looking at a platform of limited genes in terms of prognosis, in a subset of breast cancer. That study, I think, was done by the NSABP in conjunction with industry.
I just wanted to emphasize the fact that I believe that used paraffin embedded material…paraffin embedded material, I think, is, while fresh material is a great resource and important in terms of discovery, in terms of translation to clinical care, anything we can do in paraffin is far superior to things that require fresh tissue.
I think that we need to keep that in mind when we are developing strategies in terms of both screening and developing prognostic tests.
DR. BAST: I think that is really an important point. Genomic health, I think, has really taken a step forward in doing that.
I guess, just to clarify any confusion, there are actually three different studies we are talking about. There are a couple of expression array studies where, when you are looking at the pattern on the array, they haven't necessarily been confirmed.
The 21-gene individual panel from paraffin has got at least two or three NSABP studies behind it now. I think, in the best of all worlds, there needs to be one more study to be absolutely sure this is right, before it is widely applied in the community.
So far, those have been completely consistent, and it is probably the most encouraging panel, at least in my view, of anything that is out there for breast cancer at the moment.
DR. BOWTELL: I would just like to pick up on the point about the adequacy that Mike raised, the adequacy of the studies that have been done so far.
We are getting lists that are coming out, but I think we still are in the discovery phase. Particularly if you compare it to breast, that study that you just mentioned, that was originally published in the New England Journal, I think there were something like 560 cases in that particular case.
When I last looked at the genomic studies that have been published for ovarian cancer, I think that, by my reckoning, at least for expression studies, the largest so far still is about 113 samples, which was published in 2002 and included an array of different histologic subtypes.
So, when you look at the prognostic studies for ovarian cancer, they are on the order of about 50 to 60 samples that are being analyzed.
On top of that, most of those studies have some issues -- not issues, but are complicated by the nature of the disease, that it is histologically diverse, that the patients are diverse in terms of debulked status and so on, and these, we know, are confounding issues.
There was a paper that was published in Nature Methods a couple of months back by the Toxicogenomics Group. They took a single set of RNAs in a standardized kind of experiment to measure the difference between expression platforms and differences between laboratories.
They have been using just a single set of RNAs. They could see that there was a lot of variability between the different platforms, as I think we all know now, and that this could be improved substantially by using SOPs.
I think what we have got to recognize is that, for those things that are predicting response to treatment or survival, there are a few things that are stacked against it.
It is pretty clear that the signal that is in these groups is small, and that there isn't a really strong biological signal that is going to come through.
The small numbers, the lack of SOPs, the nature of the disease, the confounding variables, the lack of consistency between groups in the way that we record patient details, the inconsistency between platforms, the small signals, all these things are just going to make it very difficult for us.
I think that what we need to do, there will be real signals there, but they are going to be subtle, and the way we are going to find that is by using large numbers of samples in some sort of consortium effort on a consistent platform with a consistent SOP in terms of the molecular aspects, and some consistency in terms of the criteria that we evaluate patients, that we measure response, and that we record clinical details. I think that, until we do that, we have to accept we are probably still in the discovery mode, is my feeling.
DR. BAST: I guess the question is, if we are going to move it forward, what would be the requirements. Can we get hundreds of patients with ovarian cancer with frozen tissue up front for both the research and development, or are we going to have to wait until we can boil this down to where we can look at PCRs out of paraffin in order to be able to validate the studies, or is it somewhere in between.
DR. BOWTELL: For the Australian study, which is just one, we are now up to 1,600 cases. From that, when you boil down the numbers, currently there are about 350 cirrus invasive cases where we have got fresh frozen material, and probably about that again where we will have paraffin separate to those ones where we have got fresh frozen.
So, that is a cohort currently that stands at about 700 cases of cirrus invasive and, by mid next year, we hope to be up to around 900.
Now, I would be very happy to see that cohort added to other cohorts that are similar, to that we can actually look at hundreds of hundreds of cases in some sort of larger consortium.
DR. COUKOS: I wanted to comment a little bit on this. I wouldn't be as pessimistic as our Australian colleague is, in terms of the reach and the opportunities that genomics offer.
I think that what we are realizing is that, clearly, we are facing a tremendous heterogeneity in tumors, and that is one of the major obstacles.
Perhaps if we take into consideration the recent advances that show that tumors are, indeed, particularly addicted to one major pathway, which could represent their Achilles heel, perhaps identifying that pathway and identifying the signatures for that pathway might be a reasonable and rational way to do genomics.
For example, we could identify, based on CGHRA data, major pathways, or major oncogenes that are altered, and then look at the tumors that only have that oncogene amplified, and see what the profile and the signature of that might be, and then start classifying patients based on these pathways.
Then the challenge will be to, once signatures are identified, to see if, in fact, these are validated prospectively in the context of biological therapy.
For example, in my laboratory we are very interested in immune response in ovarian cancer, and how can we use signatures to classify patients for immune therapy.
We have been able to identify the signature of T-cell infiltration in ovarian cancer in solid tumors. So, taking now this signature, we can go forward and try to predict whether patients with that signature might be eligible for immune therapy.
I am sure similar signatures might be identifiable for patients with increased VEGF expression of vascularization that might be eligible for, for example, antiangiogenic therapy, or tumors that have, for example, PI-3 kinase dependence and so forth.
So, I think one of the major challenges might be to try to identify those with functional genomics, and then the reverse, try to validate them in trials in selected patients.
DR. MCSHANE: I would like to go back to the comment that was made referring to the genomic health 21-gene assay.
I think it was a very important point that was made, that they did develop their assay so that it could be done using RT-PCR-based methods on paraffin embedded specimens.
I would urge everyone to look very carefully at the whole development process that was gone through to develop that assay, because I think it is a very good model that was used, and it is instructive to look at the steps they took.
In particular, how did they use microarray data? They actually did start out by looking at several different microarray studies.
They looked for the commonalities across those studies, came up with a set of about 250 genes that looked reasonably promising, and then took some additional data sets to try and narrow that list further.
Another important thing that they did during the development process was to try to narrow the set of patients for whom they thought this assay was going to be applicable and useful.
Namely, it became clear, as they were going through their analyses, that this assay was probably going to be most useful for estrogen receptor positive breast cancer patients who were treated with tamoxifen.
So, I think we have to realize that the notion that you can just do microassay analyses on any mixed bag of patients and come up with useful things may not be a good model to follow.
As I said, I think it is very instructive to go through all of the steps that GHI and NSABP followed, to see where they started narrowing things down, and how they made the conversion from the microarray data to data that could be actually obtained on paraffin-embedded specimens that we have in great supply.
DR. BIRRER: One issue that came up early on, I think Jeff mentioned it, and I know my own lab is interested in this, and I think is probably in the short term the place where we could make the biggest impact in terms of managing patients with ovarian cancer, from a chemotherapeutic standpoint, is trying to identify those patients with early-stage disease who are actually at risk for recurrence versus those who do not. We know that we grossly over-treat that patient population.
The problem that we have had -- and I think this is a general problem -- is that basically these samples don't exist.
We utilized the entire GOG tissue bank, we went to collaborators throughout the United States and have only been able to get about 40 or 50 specimens, and I wonder if anybody had any comments about that.
DR. MANNEL: I think one of the big problems with early-stage ovarian trials goes back to, there is plenty of tissue out there and plenty of survival. It is just not going to be fresh frozen.
The reality is that a lot of those patients have their primary surgery done by people who are not part of the investigative surgery, and they may have their ultimate staging surgery done at a research institution, or they may go in, thinking that it is mostly likely not a cancer, and then have a GYN oncologist call in to complete the staging surgery.
I really think that, where it is clear that fresh frozen tissue offers a lot of benefits, this is a great example of that is a block in the road that you are probably not going to overcome.
If you really want to move forward on early-stage disease, or that movement between benign, borderline, early-stage disease -- and there is a lot of debate on that whole issue -- you are going to have to look at paraffin.
DR. SATHYAMOORTHY: I thought I would take a minute to talk about work from a DCB grantee. The name is Manoora Rathmone and he is based in Toledo , Ohio .
They are working on a biomarker that they feel has great promise for early detection of ovarian cancer. This is the soluble form of the folate receptor alpha.
They have found that they have a very sensitive method to detect the folate receptor in ovarian tumors, and they have seen success in about 90 percent of ovarian cancer tumors. I thought I would just talk about it so that -- his work has recently been published in Cancer Research as well.
DR. BIRRER: Thank you. Perhaps that gets us a quick venue into the dreaded field of proteomics. Elise, would you like to comment on the relevance of that technology in this discussion?
DR. KOHN: The relevance of proteomics is something that needs to be determined. I think that one of the approaches that our group and many of the groups represented around the table and the GOG have taken is that we are going to develop well annotated, carefully collected sample sets initially, so that we can perhaps get around some of the concerns that were just raised around the table about some of the array data, where the methods and the quality of the samples are not crossing over.
So, in terms of proteomics there is probably perceived promise, because we are looking at not a single event in the blood, but representing multiple events.
The true value has to be determined with a series of trials that are beginning. The pelvic mass study through the GOG, the recurrence trial through the intramural programs, SPORE collaboration, are good places to start.
DR. MANNEL: Another focus of this was looking at endpoints, future trial design and so forth. The bottom line is, we all realize that the endpoint is survival. I think one of the problems is correlating any of this with ultimate survival.
There are two main problems with that, the first being, as there are a number of agents that can be utilized, if I, without any biomarkers or any help at all, choose an agent and it doesn't work, I switch to another agent that does, have I substantially impacted negatively that person's survival? Or, will these arrays help me to improve survival?
What happens, the positive impact of any type of biomarker gets diluted out over time, because the average American now is, I think, up to 5.2 cycles of different chemotherapy before they succumb to their disease, and that is only increasing with time.
So, at some point, we have to ask a fundamental question and that is, how do you correlate the knowledge of the biomarkers with survival, not their predictive value?
If I know about this biomarker, does it ultimately change the survival in this patient? Those are two very different questions.
So, that then tells you that anything you do in this arena ultimately has to include survival. These patients are surviving even stage III disease.
After debulk now, a median survival is over six years. So, now you are talking about an infrastructure that has to not just acquire tissue and do the biomarker study, you have to have an infrastructure that is willing to follow patients for six to 10 years. That is where the funding issue becomes a bit of a problem.
DR. BOOKMAN: I just want to underscore that point, but from a different perspective. You all recall that we have had a number of methods over the years developed to evaluate tumor drug sensitivity or resistance, based on ex- vivo testing.
Several of those now are actually approved for insurance reimbursement. Yet we do not have any prospective randomized data to show that that really impacts outcomes and taking care of patients with disease.
We are finding ourselves potentially recapitulating that experience if we don't act very thoughtfully and incisively at this juncture with the knowledge that is being assembled.
DR. BAST: Needless to say, as I think everyone here is aware, given the heterogeneity of patients, given the fact that some people respond to nothing, other people respond sequentially to six or seven drugs, designing and executing those trials for survival is really a challenge.
I guess the hope -- and it still is a hope -- is that as we develop better and better targeted therapies, perhaps intervening selectively in signaling pathways that are actually abnormal in particular patients, that that is going to have a greater impact on survival.
It is the paradox of “good biomarkers depend critically upon effective therapy.” I think we have seen that with choriocarcinoma.
If you have got a disease that is exquisitely sensitive to chemotherapy, HCG is an extremely useful marker, but we are clearly not there yet for ovarian cancer.
I think, looking down the road, most of us believe that that day will come. In the meantime, I hope we don't get bogged down with managed care issues, whether we can monitor patients or treat them beyond a few cycles. I think we will move on to the next session, or to our break, sorry.
TOP
