DR. BOYD: Thanks, Mike. Well, I certainly have had a long career, and I appreciate the opportunity to be here.

I have been asked a couple days ago to do something that I found -- I thought would be relatively straightforward, but it turned out to be extraordinarily challenging, which is to summarize in five to 10 minutes what turned out to be about 70 publications that I could find.

What I am going to try to do is to simply provide a couple of slides of bullet points that, to the best of my ability, summarize what we have accomplished over the last five or six years in this field.

This work has been primarily in the area -- when we say genomic, of gene expression profiling, there has been work, of course, at the DNA level, but not to the breadth and extent to which we have addressed expression profiling as a means of delving deeper into the biology of ovarian cancer, and attempting to develop new strategies for diagnosis, prognosis, and treatment.

I will further preface my comments by saying that much of the work that I am going to summarize in the next couple of slides is a direct result of initiatives begun and funded by the National Cancer Institute, through both the Director's Challenge, that has already been mentioned this morning, and the SPORE programs in ovarian cancer.

I think it is safe to say that the majority of the work I am going to summarize is a direct result of those two programs. So, we are obviously very grateful to have that funding opportunity from the NCI.

TOP

Just to make sure we are all on the same page, when we are talking about expression profiling, this is micro array based expression profiling using one or another technology.

I think the field has more or less coalesced around the platform of oligonucleotide-based expression profiling using one particular platform manufactured by the Affimetrix company.

I have no personal interest in the company, but I think it is safe to say that one challenge in moving forward is certainly going to be to assimilate the data that has been generated, and move forward and, to do that, I think we are going to have to adopt some standard technologies and some standard approaches.

For better or for worse, that is going to involve using the same, or a very similar, platform for the work that we do, to continue the work that has already been initiated.

TOP

So, there have been several areas, the low-hanging fruit, if you will, that have been addressed over the last few years, and I have just tried to summarize these areas here.

In terms of just simply profiling ovary cancer compared to some normal control, there have actually been many studies. In a paper I wrote recently, I cited 13, I think, addressing this particular topic.

It is clear that some genes and pathways are consistently implicated in these dozen or so studies. The problem, of course, that we face with this approach is that you have always got this issue of the whole ovary cancer specimen versus a micro-dissected tissue, not only in terms of looking at cancer cells, but in terms of perhaps stromal epithelial interactions and the role that plays.

What is a normal control, what is a normal ovary, what is the comparison to, which differs, essentially, in every paper that has been published in this area.

I think a key question is, what have we learned from this exercise over the past few years, and what next? What do we do with these genes and pathways that are consistently implicated as up- or down-regulated?

I think that gets to a fundamental distinction between the two types of studies that I am trying to summarize here, those that provide more insight into the basic biology of ovarian cancer, and those that may provide a more direct translation into the clinical realm, and I think certainly both are important and necessary.

There have been several studies looking at the presumed biologic differences between the different histologic subtypes of ovarian cancer.

I think one conclusion that it is fairly safe to make now is that clear cell cancers are different, which is perhaps not terribly surprising.

What is somewhat surprising to me is that the endrometrioid tumors didn't really fall out as a separate category in separate studies.

What is more interesting is that the clear cell ovarian cancers cluster very closely with clear cell endometrial cancers, but the other histologic subtypes don't cluster with each other, which I thought was an interesting finding from several groups.

In terms of prognosis, again, several studies, and I think we need to consider overlaps in terms of gene lists between these studies, and the utility.

What is the utility of being able to distinguish two prognostic groups in stage three ovarian cancer, for example? Is there any?

I think clearly at this point we very much need similar data pertaining to early-stage ovarian cancers, where I think there would be a better opportunity to directly translate differences that we might find into the clinic.

TOP

Drug sensitivity, response to front line therapy, is certainly an obvious problem that can be addressed using this technology.

There have now been several studies positive in the sense that individual laboratories have found differences, when they have looked to primary resistant to primary sensitive ovarian cancers, and generated gene lists that are statistically significant in terms of response to therapy, both front line and salvage therapy.

What do we need to do next to make these data clinically useful, and I will come back to these questions on my third and last slide.

There have been at least two or three studies looking at expression profiles associated with cytoreducibility of primary ovary cancer.

Unfortunately, the data are conflicting. One group has found that they can show significantly different gene profiles that associate with optimal versus suboptimal cytoreducible tumors.

We can't find any difference in our own laboratory, which is perhaps not surprising. The hypothesis itself, in my mind, is quite possibly at odds with clinical reality.

I know that the fraction of ovary cancers that are optimally cytoreduced differs substantially from one institution to another.

Even within a single institution, such as ours at Memorial Hospital, over the last several years, the number of advanced ovary cancers that have been rendered optimally cytoreduced has increased substantially as the surgical approach has become more aggressive.

What does that say about the biology of the disease, if anything? Is it a surgical problem? Is it a biological problem? Is it a combination of the two?

I think it is an unusually complex problem, and will require more than the amount of thought and effort that has gone into this issue at this point.

Finally, I think I would just like to conclude by mentioning that an interesting finding from the Director's Challenge has been that sporadic ovary cancers all tend to cluster as BRCA-1 or BRCA-2-like, in terms of their gene expression profiles, which I thought was an interesting observation from a biological perspective.

TOP

I will just conclude, where do we go from here in order to do the job that I was asked to do and initiate thought and conversation.

Let's take Principles and Practice of Gynecologic Oncology, for example, the standard textbook. If you look at any of the paradigms in terms of predicting outcome or prognostic factors for a particular cancer, and you look at the data in table three, and there are typically 10 or 12 studies that have been published in prominent journals that have defined, for example, the prognostic importance of depth of myometrial invasion and endometrial cancer.

It is very unusual, in laboratory science, to see, for example, in publishable studies on gene expression differences between optimal and suboptimal ovarian cancers, do we need eight more studies to really answer the question of whether there is a difference or not, or, do we need to mine the existing data in order to refine and validate specific markers and move forward.

Do we need to concomitantly or perhaps wait for a while to design prospective correlative studies through some of the Cooperative Groups based on the existing putative markers or marker panels that have been generated?

As you are probably all aware, there is a lot of attention give to a putative breast cancer model that predicted T1 and 0 tumors that would occur, two very prominent papers published in Nature and the New England Journal of Medicine. Then, some prospective studies that attempted to validate these findings found that it just simply didn't hold up.

So, at some point we need this type of approach. Whether now is the appropriate time or not remains to be seen.

I would like to just conclude by editorializing a bit and say that high tech doesn't necessarily equal high speed in terms of moving ahead and translation of genomic medicine into the clinic. I think we still have some of the challenges we have always had in clinical and translational research.

It is going to take time to come to valid conclusions, and I think we have got to be willing to accept that and take our time and do this correctly. Thanks for the opportunity to speak.

TOP