38th MEETING - TOPIC I, cont.



MARCH 4, 2005

“This transcript has not been edited or corrected, but appears as received from the commercial transcribing service.  Accordingly, the FDA makes no representation to its accuracy………”

      The meeting was held at 8:00 a.m. in the Potomac II and III Rooms of the Quality Suites, 3 Research Court, Rockville, Maryland, Dr. Mahendra Rao, Chair, presiding.



MAHENDRA S. RAO, M.D., Ph.D., Chair


MATTHEW J. ALLEN, Vet. M.B., Ph.D., Temporary Voting Member


RICHARD D. COUTTS, M.D., Temporary Voting Member



C. WAYNE McILWRAITH, BVSc., Ph.D., FRCVS, Temporary Voting Member


ALAN J. NIXON, BVSc., M.S., Temporary Voting Member


SEAN P. SCULLY, M.D., Ph.D., Temporary Voting Member

SHARON T. TERRY, M.A., Temporary Voting Member



ROCKY S. TUAN, Ph.D., Temporary Voting Member

GAIL DAPOLITO, Executive Secretary














Welcome and Administrative Remarks

      Dr. Mahendra Rao, Chair, Cellular, Tissue and

      Gene Therapies Advisory Committee......... 3


FDA Perspective on the Development of Cellular Therapies for Repair and Regeneration of Joint Surfaces - Manufacturing

      Dr. Malcolm Moos, Jr., Office of Cellular,

      Tissue and Gene Therapies, CBER, FDA...... 3


Committee Discussion of Manufacturing Questions 20







































                                         8:05 a.m.

            DR. RAO:  Good morning and welcome to another day at the FDA.  We have one new member on the committee and I am going to ask Dr. Prockop to introduce himself before we start.

            DR. PROCKOP:  I'm Darwin Prockop and I head up the Center for Gene Therapy at Tulane.

            DR. RAO:  Welcome, Dr. Prockop.  Just a brief reminder.  Remember to shut off your mike when you are done and wait for the Chair to recognize you so that we can have a reasonable flow.

            We are going to launch straight into the first part which is a continuation of what we were doing yesterday.  We will discuss the product issues related to articular cartilage.  Dr. Moos will give the update perspective first.

            DR. MOOS:  Good morning, everyone.  Thanks for sticking it out through to the second day.  By the end of yesterday I imagine we now know 75 percent of what there is no know about cartilage biology, not only 50 percent as Dr. Luyten suggested yesterday.

            Today some of what we may be discussing is the other 25 or 50 percent.  I think there are significant uncertainties that surround issues of manufacturing, product characterization, and testing.  I am going to hit on a few of those points in a fairly general way.

            By way of introduction just to review a bit of regulatory philosophy, while the oversight of things like devices and small molecule drugs is amenable to product testing, entities as complicated as biologicals have needed for a hundred years or so a sort of three-pronged approach where one considers carefully the starting materials, the manufacturing process, and how you take various kinds of precautions to make sure that it is consistent, and testing of the product. 

            And taking into account all three together allows one to move forward into new areas of endeavor despite significant uncertainties about the nature of the products and the key parameters that are required to determine their safety and effectiveness.

            Things that resemble devices or, for that matter, small molecule drugs, are more tractable in a number of ways.  They are well characterized.  The materials from which they are made are defined precisely.  Often we know their structures for small molecule drugs, where the protons are to Angstrom resolution and with devices.  All sorts of specifications.

            We know what to measure and we have analytical methods that allow us to do that.  We know which things that we're measuring actually matter, actually have a major influence on the performance of the product.  In contrast as we move to things that are biologically move into a gray area that's a little bit fuzzy and, on this slide anyway for us as well as for you, a little bit hard to read.  To try and throw some light on that, these products are very poorly characterized. 

            We used to be a decade ago concerned about how to characterize a single therapeutic protein.  Now we're talking about entire cells or collections of cells or products which have organized themselves with or without the help of an artificial substance or substances into a three-dimensional matrix.

            The nature of the product is such that it could be very heterogeneous.  The entities are complex in molecular terms.  In the case of a cell there may be tens of thousands of difference macro molecules to consider.  The distance between what we might want to measure and the capacity of existing analytical techniques to measure them is often quite wide by comparison with other types of products the FDA regulates.

            Finally, and I think most problematically there is insufficient scientific data that allows us to relate things that we can measure to what happens in vivo in biological systems. 

            Now, much of what I'm going to talk about today will not deal with the process.  The manufacturing process is generally considered to be something that someone who is coming to us has developed and is their proprietary matter and trade.

Nonetheless, it's worth nothing that issues of source and release testing can help support the process both to optimize it and to take measures to ensure that it's consistent.

            Although, as I mentioned yesterday, there is a lot of emphasis placed on getting into initial trials as quickly as possible for a variety of reasons.  Our experience teaches us that it may be useful to consider the entire strategic drug development process as early as possible from a global standpoint so that you, as the saying goes, measure twice, cut once.

            Sometimes there is a tendency not to spend enough time early in development and making sure that you have optimized to a reasonable extent the product that you are going to carry forward for more expensive kinds of analyses such as preclinical testing or very expensive clinical trials.

            It could be that up front investment in looking at your sources and playing with your manufacturing process may prevent a costly waste of time and resources and you may end up after some sort of a confirmatory trial just shy of an endpoint with a p-value that is tantalizing but maybe not quite enough. 

            Personally, one of my worse nightmares as a reviewer is for the idea to be fundamentally sound and there is something about the product, something that we are not controlling or we don't understand, and there may be a few patients where there is an outstanding success but we don't quite have the data that would support a license application.  The trial fails and suddenly the venture capitalists are nowhere to be seen.

            So it's my own personal feeling, and I think many of us in the agency would agree, that better characterization of the product early in the development program may make a number of things easier downstream.

            The first area is source control.  We heard a little bit yesterday about different anatomic sites which may include different sites within a knee as well as starting materials taken from other types of tissue or even other donors.

            The question of whether or not a donor site might be involved with the disease process could be construed by some and we did have this discussion at some length internally as a patient inclusion or exclusion criterion but, nonetheless, there are some patients who would meet inclusion criteria. 

            You go in and you look at the time of collecting tissue and you might see something that would tell you this is maybe tissue that isn't going to work to generate a useful product.

            Finally, there was some information yesterday speaking to the issue of cellular inhomogeneity perhaps in the starting material.  A graphical illustration of this are these scanning electron micrographs taken by Professor Stanley Erlandsen of the University of Minnesota. 

            This is a juvenile joint surface.  I think you can see that as you go through to different regions within a joint, there's substantial obvious heterogeneity from one cell to another just at a casual glance.

            At Dr. Rao's suggestion and with Dr. Luyten's permission I have put in a few of his slides from yesterday by way of review.  One very interesting piece of data that we heard yesterday is that even in freshly isolated chondrocytes there is a population of cells that in a model stressed with skeletal muscle injury by cardiotoxin is capable of developing markers for myocin.

            This one, and probably the simplest interpretation of this data is, indeed, the starting material is biologically heterogeneous and you can determine this with molecular markers.  Clearly even within a single tissue there might be characteristics that need to be considered in terms of subpopulations of cells that would have an influence on the performance of the final product.

            The next area is product testing.  Now, the general philosophy for developing a set of tests is to do a set of very thorough characterization studies to learn as much about the product as possible in the context of the manufacturing scheme that you have developed.

            From that set of characteristics further work is done to try and select a set of tests that could be useful and practical in a manufacturing environment so that you can do the test quickly enough, the assays are robust enough and so forth, that they could be used both to control the manufacturing process and to release the product.

            Ideally, you want them to be sensitive enough to changes in the manufacturing process that if there is a change your testing will help you detect it.  Similarly if there is something about the product that is either good or bad, your testing procedures will pick it up.  And, finally, in an ideal world your set of analytical procedures should predict what happens when the product is placed in an in vivo environment.

            Now, data in this field including what we heard yesterday suggest to me and to others on the committee that the focus that has been conventional in many areas of cell therapy on looking at terminal characteristics of the tissue that one desires to see formed after administration in vivo may not be the best ones to measure on the final product or during manufacture.

            For example, if we take chondrocytes out of their -- if we take cartilage out of its natural environment, dissociate the tissue enzymatically and put it on plastic, many of the characteristics that are normally associated with cartilage are lost.  So looking for those characteristics in vitro may not be helpful in contrast to other types of products where perhaps a terminally differentiated state could be expected.

            I think it's equally apparent from a variety of data that many of the terminal markers for several products including these may not predict biological function so there are plenty of examples of products which express type II collagen and aggrecan and don't express type X collagen which, nevertheless, fail in vivo.

            So let me suggest that perhaps the real task is not to identify something that has some characteristics of terminal chondrocytes in vitro so much as to identify functional chondroprogenitors.  By that I mean a set of cells which may or may not be a homogeneous population that can be qualified according to a meaningful biological response.  In other words, let the biology tell you what the tests should be rather than take a guess at what we think we should be measuring and see what works if we're lucky.

            Yesterday I promised you an example of this kind of idea from another area of cell therapy and that area is the area of pancreatic islet transplantation. 

            There is a certain kind of parallel in that in islet transplantation there was a sort of history of a few successes in a few patients who achieved a major benefit and a number of patients who achieved very little benefit and there was this sort of erratic one time to another variability that made everybody think there was something really real going on but that there wasn't a handle on some variable.  Work that has been done by pancreatic islet enthusiasts has taken a harder look at some parameters influencing product quality.

            The data I'm going to show next is curtesy of Klearchos Papas at the University of Minnesota and also Bernhard Hering at the same institution.  At an earlier meeting of this same group Dr. Hering actually showed us this data where a number of islet isolations were performed and by the conventional sort of measure that everybody does which is to look for exclusion of a vital dye, all of the products past.  They were all over 80 percent viable.

            What the groups then did is to apply a couple of different types of measures of cellular vitality, if you will, that looked at something other than the integrity of the plasma membrane.  The first of these that I am going to show is oxygen consumption rate. 

            Now, you would predict that oxygen consumption rate would be directly correlated with dye exclusion if dye exclusion was really a sensitive measure of how many intact fully living cells you have.  In fact, it's not correlated at all. 

            What is possible to do in this study was to then look at the performance of each of these isolations in vivo and separate groups where you had 100 percent failure or 100 percent cure with a few marginal ones in between cleanly based on this data, a 15-minute procedure that is telling you something useful.

            A similar type of test measured ATP levels on the same groups of cells but, once again, could cleanly separate cells which performed well in vivo and cells which didn't.  One thing that made this work very well was a clean in vivo assay, the diabetic mice made so artificially in a controllable way.  You could qualify these two tests against performance in vivo.

            To be clear, I'm not suggesting that either of these tests would be relevant to chondrocytes which might not depend on respiration anywhere near as well or directly as pancreatic islets.  Just to illustrate the parallel here between conventional methods not working and perhaps more refined methods and one could consider for chondrocytes looking at other things.  Perhaps caspase-3 levels, mitochondrial membrane potential sensitive dyes.  There's a whole list of things that could be considered.

            It's possible that in vivo models such as the one that Dr. Luyten showed us yesterday might be applicable and we can discuss the applicability of this type of model or perhaps other assays of this sort later on.  The idea here is the general concept that some kind of in vivo qualification assay that wouldn't be feasible in a manufacturing environment could be used to help you evaluate analytical tests that would be useful in a manufacturing environment.

            Now, the next question has to do with this idea of tests that would identify functional chondroprogenitors.  Many cell therapy enthusiasts, especially early in the history of these procedures, had the idea that a cell product could be administered and local instructions would cause the cell product to do exactly what was desired but they would be completely plastic, completely under the control of local environmental influences. 

            The analogy that I like to use, and people have heard me talk before are probably tired of, was given to us by American cartoonist Al Capp in 1948 who talked about pluripotential entities that he termed the Shmoo.  There's a transitive verb that has entered the popular vernacular that you now know the origin of.

            The instructor signals that they were subject to were simply the desire of the nearest human.  Whatever was needed by the nearest human the Shmoo would transform themselves into the finest butter, eggs, a certain variety of ham, and so forth.  This simplistic cartoonist's view, in fact, is one that developmental biologists have known since the 1920s probably didn't work. 

            In fact, an experiment that was cited for the Nobel prize demonstrated this quite clearly.  You could take a piece of an embryo from a pigmented strain and transplant it into a recipient embryo in an ectopic site and far from the expectation that the donor embryo would adopt the characteristics indicated by the surrounding tissue, in fact, the reverse happens. 

            What you see a couple of days later is an embryo that is almost entirely twin, Siamese twin structure, if you will.  If you take a section through the embryo in such a plane, what you'll find is, and I think we have a gamma problem and the colors aren't showing up on this computer, but what you see -- do we have a pointer?  Thanks.  What you would see if the colors were working is that in the primary axis the pigmentation would correspond to the recipient embryo.

            In the secondary axis a small amount of the tissue would correspond to what you would see in the donor but much of the tissue is, in fact, derived from the host.  In other words, the host tissues aren't capable of receiving instructions from the donor.

            Now, the message is clear, instructions can go both ways.  In fact, on another of the slides that Dr. Luyten showed us suggested in some of the discussion following this also raised the question as to whether in repair of tissue what you see is contributed entirely by the host, entirely by the recipient, or some combination of both.

            In fact, we know that for another product, and Dr. Tuan mentioned it yesterday, if you inject this product into a joint space, you actually recruit host tissue in such a way that it's clear that its fate has been changed by instructions from these therapeutic cell product.  This is a complexity that needs to be considered.

            Now, how cells may or may not either convey or respond to these types of instructions is an area of developmental biology that is concerned with what developmental biologists call competence factors.  These may be various kinds of monomeric or dimeric receptors or co-receptors. 

            They are soluble ligands, ligands which may be cell-associated, the signalling machinery intracellulary, and the gene transcription machinery that may represent a more long-term alteration and biological response that leads to something that could translate into an in vivo activity.

            Perhaps, may I suggest, that we might need to be looking for our sets of competence factors which could be positive factors or negative factors.  Again, from Dr. Luyten's data, we see examples of both of these things that are associated in a positive way with the expression of a stable chondrocyte phenotype and things which are associated in a negative way with such a cellular organization.  Perhaps that's the kind of question that we might consider.

            Indeed, this final slide from Dr. Luyten suggest that you may have sets of markers.  We can discuss his approach as well as alternatives that might be useful.  Certainly I don't think anyone suggests that this is the last page of the last chapter but it is, in fact, somewhat reminiscent of the oxygen consumption data that I showed you a few minutes ago.

            The last thing I want to discuss is the concept of new regeneration products that we heard some talk of yesterday where the cells are actually organized in three dimensional space.  If to the extent that is true, one may need to consider what is responsible for that organization and look into the factors which control that organization.

            This bright-field and fluorescent image of a gene that the laboratories of Lee and Kingsley and the mouse and Dr. Luyten and myself and a number of other species is a growth factor associated with forming joint space.  I think you can see that it is a remarkably specific marker that separates the joint space from other tissues, one example of the kind of thing to look for in products that may be organized three dimensionally.

            To take an even finer look, if one looks at that same growth factor and the molecule that we have found is necessary for its activity, you see in the forming joint space an overlapping region of perhaps one cell diameter corresponding to the forming joint surface.  It could be that this kind of a test or analysis might lead to optimizing and qualifying other more simple tests in a manufacturing environment.

            So I would like to leave you with a few general questions for thought.  What is going on early in the process whether we are talking about acceptance criteria for a homogeneous or mixed set of cells, whether there is selection for or against particular members of that population, or whether instruction is happening.

            As I said a bit ago, perhaps competence factors sufficient to define the cells that are useful might be thought about.  It's not necessary, may I point out that we actually understand the mechanisms as long as we can qualify them usefully in an animal model.

            As I just pointed out, distribution of characteristics in 3-D may be critical.  Finally, the theme that Dr. McFarland and I have tried to develop is that there is an interplay between how good your animal models are and how they can be used to tell you something and how easy it might be and how solid your basis will be for selecting particular sets of analytical tests. 

            The questions for discussion, in highly condensed form for projection, deal with the criteria for obtaining starting tissue:  What the characteristics might be for functional chondroprogenitor cells, or if we don't know what they are.  What would be useful approaches for finding that out.  What might be useful methods, analytical methods to explore for determining these characteristics.  What are useful approaches to qualify these tests using various kinds of preclinical models, which need not be disease models. 

            The special issue that is often neddlesome enough to cell therapists generally that it's worth breaking out separately of potency assays and considerations that might be specific for cells that are contained in naturally produced or artificial matrices.

            I will now yield the floor to our chairman so that we can discuss these issues more generally.  Thank you.

            DR. RAO:  Are there any specific questions for Dr. Moos right now?  In that case, we'll move on to the questions.  I want to make a couple of remarks right in the beginning when we consider this and just remind the committee of some things that already exist in terms of rules and regulations which have come from a long history of sort of cell and tissue sourcing from organ transplants or from bone marrow studies and from other studies. 

            So the FDA does have a guidance on tissue sourcing and that is available on their website and that provides some general guidelines in terms of consent issues, testing for certain human viral pathogens, etc.  That is not something that might be different. 

            What we really want to consider today in terms of sourcing is what are the unique aspects to sourcing of cartilage tissue or tissue which would form cartilage if you are doing autografts or if you are taking allogeneic tissue.

            Dr. Moos really raised some potential issues in terms of just sourcing.  Is it how you harvest?  Can you ship it?  Are there any sort of specific things that one needs to worry about in terms of taking articular tissue? 

            Is there a sourcing issue in terms of which region of cartilage one takes?   Whether if you take articular cartilage from the ankle or whether you take ear lobe cartilage whether there is any difference and are there any ways of measuring quality of the tissue in any sense?

            I am happy to have anybody lead off the discussion in terms of making points.  If nobody wants to, then I'll ask Dr. High to tell us a little bit about bone marrow just in general in terms of collection so that people can think about it a little bit.

            DR. HIGH:  Well, first of all, I would like to remind you that I think Dr. Blazar is probably a lot better qualified to talk about harvesting bone marrow than I am.  I actually wanted to ask a question about the issue in this first question.  Are we only talking about cartilage or are we talking about mesenchymal stem cells or any sort of starting product that might be used here?

            DR. RAO:  I'll let Dr. Moos answer that question.

            DR. MOOS:  Everything is on the table.  I did want to point out that even if we were talking about cartilage, there might be heterogeneities within even a single joint that could be important whether specific areas of pathologic involvement that might be apparent on arthoroscopy or even characteristics that might appear following collection where you might do some acceptance assays after starting the manufacturing process, arrays or PCRs or something like that.

            But, in addition, since there is discussion of using non-cartilaginous starting material if there are specifics there.  Now, we saw some data yesterday to suggest that material from synovium could be made to look like cartilage in vitro and totally fall apart in the mouse assay.  That is a clue that it may not be so simple to do it from other sources but that is not to say that it can't be done.

            DR. HIGH:  As I was listening to Dr. Moos talk, what struck me is that the most important thing here, and actually I was thinking this during most of the discussion yesterday, is to attempt to correlate the clinical outcomes with characteristics for obtaining product.  Those seem to be the central issues in the discussion here that actually overrode any preclinical considerations that we were spending time on yesterday.

            In other words, to attempt to correlate outcomes in the clinical trials with the characteristics of the product that were used.  It seemed to me that was something we hadn't heard a great deal about.

            DR. RAO:  I'm hoping we consider that in the next couple of sections but it's really absolutely very important and I think Dr. Moos has alluded to it when he said that we need to know some way of having a measure of potency.  Let's discuss that as potency.

            But just looking at tissue, I mean, is there any criteria?  When people take pancreatic islets, for example, when Dr. Moos pointed that out, you always worry about the time of isolation because it's a cadaveric source so there's a time window beyond which it is certainly not considered reasonable to expect to get good tissue and that is relatively unique to pancreatic islets in that that is the major source if you were to get cadaveric pancreatic islets as an issue.

            When you take biopsies in humans in terms of taking biopsy samples in the nervous tissue, you always worry about the source and what underlying pathological condition was present for which you could be allowed to take a biopsy so that becomes an important consideration in your tissue source material.  So is there anything like that as far as sourcing cartilage that one has to worry about specifically when one is considering a source material issue? 

            Go ahead, Dr. Tuan.

            DR. TUAN:  I think one issue, of course, is potential donor site morbidity for articular cartilage repair.  If you are taking it from articular cartilage we need to be concerned about that.  Particularly if we want to use low-passage-number cells we will need to get quite a bit of tissue. 

            If it's not from the articular cartilage, then where else?  And, also, how can we then really -- I mean, it's tied to the second question.  We can get whatever we want but if it doesn't work, then it's an issue.  I think one of the first concerns ought to be donor site morbidity -- potential donor site morbidity.

            DR. RAO:  So is it an important criteria the amount of tissue you get?  Is that like a really critical parameter that one needs to really know so that when you are sourcing and it's a single person source and it's an autograph that you have to get a minimum amount of tissue per se in terms of having any predictive value and quality or something that one needs to worry about?

            DR. TUAN:  I think perhaps the orthopedic surgeons can deal with this and how much is an allowable amount with some prediction of acceptable donor site morbidity.

            DR. RAO:  Go ahead, Dr. Tomford.

            DR. TOMFORD:  I think donor site morbidity is very important.  There are places in the knee, for example, that you can get cartilage that probably does not affect the joint over long-term so I think it's legitimate to consider that.  There is some work that shows that chondrocytes in the ankle are different from those in the knee so I'm not sure you should go to another different joint.

            DR. RAO:  Would there be consensus on that, that if you are taking from the knee people would only take tissue from the knee?  If you are working the ankle you wouldn't take cartilage tissue from some other source?  

            DR. TOMFORD: I don't know.  You would have to ask other members of the panel, I guess.  It may not be grossly different.  In fact, it may be that there is a difference that is advantageous.  I'm just saying there is a difference.

            For many years anterior cruciate ligaments were taken autologously from the middle-third of the patellar tendon and we now know that there are some problems with that.  Long-term there may be some problems with the donor site we don't know about but, as far as I know, not a lot of problems with that.

            I think as far as the numbers of cells are concerned we heard yesterday that the Genzyme Corporation actually increases the number of cells in culture so that probably either you have to take a lot of cartilage to get enough cells so that you don't have to expand them or you have to expand them.  Of course, when you expand them then you have to submit to other reviews.

            I do think we probably need some testing to determine what is the optimum number or the minimum number of cells that you are going to transplant.  I'm not sure we have a lot of data on that yet.  In particular, depending upon how large the lesion is.           In other words, lesions vary up to 4 to 10 sometimes square centimeters so can you get away with 10 million cells or if you get a larger lesion do you have to use 20 million cells?  I think there also has to be some evaluation of the number of cells corresponding with the size of the lesion because lesions come in all sizes.

            DR. RAO:  Dr. Coutts.

            DR. COUTTS:  I'm feeling very ignorant right now.  I don't think there is a lot of data on this which is what Dr. Tomford essentially just said.  I know that in the genzyme ACI methodology there's a minimum amount of tissue by weight that they request which I think correlates with the number of cells they anticipate harvesting from that tissue. 

            Obviously the more cells that you could provide them, the more they would be able to give back to you.  As a general principle, I think there's some correlation with the size of the lesion that you are attempting to treat with the number of cells that you ought to have.  But I'm not aware that this has been systematically studied and a lot of this has just kind of evolved like trial and error.

            DR. RAO:  Also if you harvest the tissue is there any measure of the quality of the tissue that one takes in terms of looking at this tissue?  Is there any parameter when you look at it and say this is good tissue or is there any worry what you get is what one takes?

            DR. COUTTS:  I think you are happy with what you can get.  You are obviously trying to avoid damaging critical parts of the joint so you are staying on the parameter.  The cartilage tends to thin as you go from the main weight-bearing surfaces out to the parameter of the joint.  It's quite possible that you are getting some fibrous tissue or fibrocartilaginous tissue as part of your biopsy.

            But it gets to this whole issue of how important is the cell that you're providing.  All cells start with the same genetic information.  We do tend to believe that the site in which it's placed has a tremendous influence on how it performs. 

            It could be that cells that are not chondrocytes that are provided and expanded in culture and then returned for implantation might not be -- may not have been chondrocytes to begin with but that when put into the joint environment that they will behave like chondrocytes.  There is some evidence to suggest this could happen but, again, I think there's a knowledge deficit here that hasn't really been carefully worked out.

            DR. RAO:  Dr. Scully and then Dr. Mc.

            DR. SCULLY:  I think that the over-simplification is even more than that.  It's not only that joint surfaces differ between ankles and knees but they differ within the joint, locations within the joint as Dr. Coutts has said. 

            Then the cells themselves differ within the depth of the cartilage.  I think all this concept of harvesting chondrocytes and putting them back in whether it comes from an ankle to a knee or from a knee to a knee there is a gross over-simplification.          I agree with Dr. Coutts when he's saying that they all have the same genetic information.  What we really need to do, I mean, if we had a magic wand and we could make this approach work is to recapitulate the developmental scheme so that we get normal articular cartilage back.  That magic wand hasn't been identified yet.

            DR. McILWRAITH:  We have done some work in the host with looking at donor site morbidity and use in sort of the equine equivalent of MACI taking 300 milligrams from just basically a Ferris-Smith rongeur from the lateral trochlear ridge and then doing follow-up pathroscopies up to 12 months.  There's no progression of the lesion.  There's minimal healing.             It varies but it's off the abaxial side.  Like Dr. Coutts was saying, getting off the nonarticular portion so we don't feel that we've got any clinical morbidity but it's always a feel thing.  You're basing that on going back to your question about quality or how do you assess it.  That's what we gave the company to process it because that's what they wanted, 300 milligrams, for instance. 

            You can look at the quality arthoscopically but it still doesn't take into account the points that Dr. Scully and Dr. Coutts are raising about difference in cell numbers.  But it does -- you know, then you've got the process of whether you culture or whether you put it right back in which is a newer technique. 

            Obviously if you put that cartilage right back into a new environment, a new location, then hopefully you are going to be able to modulate those cells effectively but we don't know for sure.  We just think about it.

            DR. RAO:  Dr. Moos, do you have something to add?

            DR. MOOS:  Yes.  Dr. McIlwraith raised one point that I would like to address explicitly when he said that you might see something when you scope a patient.  Is there something that might not be apparent before you scope that when you go in would tell you I'm not going to take this piece?  That's part one.  Part two, is there any data to support that your impression is correct or not.

            DR. McILWRAITH:  You mean to some predictor before you do the orthoscopy that you are going to have defective cartilage at the donor site?

            DR. MOOS:  Right.  Suppose you have a chronic knee with some degenerative changes.  The injury was seven years ago just for the sake of example and the patient is a candidate by inclusion criteria and you look and you see the surface of the knee and there is something there.  You say this is not going to work.  Let's just forget it and not charge the patient what it's going to cost for this second procedure and all the rehab and so forth and not even take the biopsy.

            DR. RAO:  Go ahead, Dr. Coutts.

            DR. COUTTS:  I could argue actually on the opposite site of that, that some pathologic tissue might actually be a better source of cells.  Studies of arthritic cartilage harvested at the time of total knee replacement shows that these cells are actually more response to TGF-beta, for instance so that their receptiveness to stimulatory growth factors is upregulated. 

            In fact, cells in fibrillated cartilage frequently are cloned which would suggest that they have been making an attempt to replicate and to try to heal the lesion.  These cells actually may be stimulated cells that potentially would be better repair cells than cells from a more quiescent area of cartilage.

            This is all hypothesis and conjecture.  I have absolutely no idea if what I've just said has any veracity to it but I'm just saying that you can go that line of thought if you want to.

            DR. RAO:  Dr. Harlan and Dr. High.

            DR. HARLAN:  I just want to make a small point.  I think everybody here is aware of it but the central paradigm of biology for so many years that all cells have the same genetic material and, therefore, that they all should have the same potential.  The Dolly experiment has shown us that's not true necessarily, that cells do have epigenetic changes that occur at various stages of their development that may be irreversible.

            The thought that not only an ankle chondrocyte might be different than a knee chondrocyte but even within a knee they may be different.  Those changes may not be irreversible.  They may be but they  may not be.

            DR. MOOS:  That was the point of my slide with the frog embryos.  Indeed, as you pointed out, it's a gross oversimplification to suppose that cells always respond to their environment rather than the other way around.

            DR. RAO:  Dr. High.

            DR. HIGH:  Just to start with a process that is already in place, the autologous chondrocyte implantation, what parameters are used now?  For example, does the material that comes in is it characterized as sort of wet weight of cartilage and then the final product is characterized as total number of cells?

            DR. RAO:  Can we get to that question after I get an answer to the sourcing question?  Maybe we can ask Genzyme to comment on that since they are doing this as a process.  To take this sort of sourcing issue is there any age at which you would say, well, I'm not going to harvest cartilage for this because it's a growing child and the epiphysis has infused and I don't want to touch the joint?  Is there an age at which you wouldn't do this at all?

            On the other end, is there an older age at which you wouldn't take this because we know that cartilage ages and we contrarily get a large number of cells even if we take a biopsy of a size that we can take in terms of being any viable use?  Is there a range that there is consensus in the field which people do or don't do?

            DR. COUTTS:  I think I would willingly take cartilage from a child because a child has the capacity to heal it.  Because of this fact that they are still in a growth phase, that imparts to them a repair capacity that we lose when we reach adulthood.

It's rare that children present for this.  I think probably because they do have this healing capacity, although the issue of osteochondritis dissecans borders on the child/adult age range. 

            The issue is more on the age side.  It's clear that the ACI technique and the microfracturing technique work better the younger the individual.  We know that there are significant changes that occur in cartilage with aging.  The cells die off.  The cartilage becomes thinner.  The cells are, for lack of a better description, less robust in terms of their intercellular machinery and manufacturing capacity and their responsiveness to various cytokines all declines with age.  I hate to say this because it's happening to all of us. 

            I think that the current methods of cartilage repair are pretty much reserved for the young middle-aged people and that for the elderly it hasn't been an option and total joint replacement really is the option for them.  There's an arbitrary cutoff of 50, although the definition of middle age keeps changing. 

            Either you personally change your own definition or society is changing it.  It's clear that people age 50 today typically are much more active than they were a few decades ago.  That may impart some degree of preservation and may slow the aging process but clearly there's an age factor.

            DR. RAO:  Bruce.

            DR. BLAZAR:  It would seem listening to this that if there's an analogy to the bone marrow transplant field, we take biological materials.  They have some heterogeneous characteristics but they are from generally single site.  What has happened in the field is really to look at the outcome data first and then try to look at correlations with cell dose or CD34 content, etc. 

            But it's only been going from the biological outcome to correlative parameters to try and find surrogate potency assays that we are then able to determine the answers to some of the questions that are being posed. 

            Unless you have a correlate that you know will predict outcome whether it's cell number related to metabolism or phenotype, these at best, I think, are going to be best attempt speculations but are really not going to be able to answer the questions in the absence of those clinical outcome correlates. 

            We'll get into that discussion later but it would seem that it is trying to almost answer the question before you have the outcome data to allow you to back correlate and determine what makes the most sense.

            DR. RAO:  We'll try and get a little bit into that but we should keep that thought.  Before we lose that, maybe I can ask Genzyme to comment on when you get soft tissue, what do you do?  Do you look at wet weight and do you have a clinical prediction on number of cells in terms of getting --

            DR. WILL:  Hi.  Jackie Will, Senior Scientist in Manufacturing at Genzyme.  We asked the surgeons to give us a full-thickness biopsy so that we capture all of the layers of cartilage and we are looking for a beginning weight of at least 200 milligrams.  We say aim for 200 to 300 milligrams.  We will process whatever the surgeons send us so we have started with as little as 20 or 30 milligrams of tissue to grow the product. 

            Then during the biopsy processing steps our technicians are trained to remove any extraneous tissue from sources other than cartilage so if there's bone or if there's synovium in the biopsy we won't process that.  We are aiming to just process what looks like hyaline cartilage.  then our process is monitored throughout the growth phase of the cultures.             We have certain time frames that we expect cells to achieve certain density markers and the cells are monitored for morphology during the entire growth process.  Then the final product is actually tested for trypan blue dye exclusion and our product is based on the number of cells that we are shipping out.

            DR. RAO:  Live cells or are they shipped as a frozen vial?

            DR. WILL:  They are live cells.  They have a shelf life of three days.

            DR. RAO:  So a flask of live cells.

            DR. WILL:  A small vial of a very concentrated cell suspension.

            DR. HIGH:  And what range of numbers of cells are typically released?

            DR. WILL:  Our average is somewhere in the range of 2,000 to 3,000 cells per milligram of tissues that we process.

            DR. SCULLY:  If you get a small sample coming in, not the 200 milligrams you asked for, do you go through more passages to end up with a larger product or do you do the same passages and they just get less product out at the end?

            DR. WILL:  We only go up to three passages so what happens typically in that case is that the cells are in culture a little bit longer so they take a few extra days to get to the confluencing markers that we are looking for.

            DR. RAO:  So that's an important point.  They may go through more population doublings even if they don't go through passaging.

            DR. WILL:  Right.  They definitely do go through more population doublings.

            DR. BLAZAR:  What surrogate studies do you do at Genzyme or what studies are done on the site to look at differences in the type of products?  You have a cell number but what are the attempts being made in the field to correlate the product itself with outcome?  We didn't get a good feeling for that other than --

            DR. RAO:  Let's hold that for a little bit later.

            DR. BLAZAR:  Okay.

            DR. PROCKOP:  Point of information.  How many population doublings do you go through?

            DR. WILL:  We typically see four to five population doublings per passage so 15 all together.

            DR. PROCKOP:  And how long does it take?

            DR. WILL:  Three to four weeks.

            DR. McILWRAITH:  Maybe my question is premature, too, but you mentioned markers so it's probably a little bit the same as what Dr. Blazar said but do you have some markers you use?

            DR. WILL:  We have done process validation where we have taken the cells in our final product and put them into suspension cultures in agarose and alginate.  Then we look for collagen II and aggrecan expression.

            DR. RAO:  Let's hold that, though, and we'll ask this again maybe if you can come back for that.  I still want to try to focus on tissue sourcing so if you have a question specifically on sort of the size, amount, patient sort of issues, then let's get that through first before we get to all these other really important issues that Kathy and Bruce and everyone has raised about markers and how you qualify the number of cells that you get.


            DR. TUAN:  Yes.  So my understanding was that we are not only just talking about chondrocytes.  Is that correct?  So there are these other sources that we need to also think about such as bone marrow, such as adipose and other tissue sources, placenta and so on, so forth. 

            I guess the placenta is not as critical in this discussion but in terms of the adipose and the bone marrow I think those need to be considered because there are reports, for example, when you take the bone marrow, which is commonly from the iliac crest, that multiple aspirations may not work actually as well as single aspirations so that's a tissue site issue. 

            In terms of the adipose -- one more thing is, of course, concentrating the putative progenitor cells and how much one can do that and that relates to how much you want to aspirate.  Then, of course, the adipose tissue how do we, I guess, quality control and then determine how much to take out, what is a good piece of adipose and so on.  I guess we should also discuss those.  I just wanted to raise that.

            DR. RAO:  Dr. Luyten, do you have a comment to make?

            DR. LUYTEN:  One piece of information that I think might be very useful also for clinical outcome is that it is published the number of cells per milligram wet weight in patients is published and decreases with age.  That is in the order of starting at 45 and 1,000 per 100 milligram wet weight and declining with age.

            In osteoarthritic samples the number of cells -- again, this is published work -- per milligram wet weight is clearly less, about two to four-fold lower so in the preparation of your cell population from your initial biopsy you are getting a cell number that is, indeed, low on wet weight. 

            This might be an indication that we are treating here a patient with osteoarthritis, another patient with a focal cartilage defect in an other normal joint.  The outcome in that patient may actually be different from the ones you obtain.  That kind of information may actually, in my view, be very useful also for clinical outcomes in the future.

            DR. BLAZAR:  Just a comment for bone marrow harvest.  We have very different constituencies of materials depending on age of the patients, whether or not they are smokers, whether or not we take samples from the beginning of a harvest toward the middle to the end.  The counts can vary two, three, four-fold just in terms of total nucleate cell counts.              The amount of fat tissue that comes along with the bone marrow harvest and the number of T cells and hematopoietic progenitor cells varies throughout the harvest whether you do a 1 or 2-cc pull or you are doing 10 or 15 cc-pulls or if you are putting it through a grated filter set which we do. 

            We put it through three sequential filters to filter out bone particles, fat, and probably MSCs are getting filtered out.  There are so many differences in the procedure itself and then the patient characteristics going into the procedure that are major determinants of the final product that need to be taken into account.

            DR. RAO:  To get back to what Dr. Tuan said, it seems to me just from listening to everyone here that the big issues if you're taking cartilage and using it as a source to get cartilage cells seems to be that you need to worry about donor site issues and you worry about the age of the donor because that's going to have some predictive value. 

            And you have a limit on the size of tissue sample you can take dependent on the patient and that seems to give you some prediction just from the size of wet weight of the number of cells you possibly would be able to get. 

            Nobody seems to think it is very important to do a biopsy of that material because, as Bruce pointed out, there is really no clear-cut predictive measure saying that if you take this sample and it has this quality, this is what results you'll get downstream.  And it doesn't seem to be part of what everybody does in the field.  You get tissue.  You get whatever you can from the same patient if you are doing it and that seems to be true for cartilage. 

            That is relatively unique for cartilage because that wouldn't apply to a lot of the other tissue sources which may be potential donors for this sort of transplant.  Does that seem like the state of art?  Would Genzyme agree that is the state of art with what they do in terms of current clinical practice?

            MR. O'CALLAGHAN:  There is only one thing I would add and that is about this donor site morbidity issue.  I don't think -- we've treated more than 10,000 patients.  There's really no evidence of significant donor site morbidity. 

            If you compare this, for example, to a nose procedure where people are taking cores and punching holes all the way through, you saw some pictures yesterday, this is trivial.  I mean, the amount of cartilage in 200 megs is about the size of a half of a Tic Tac so when you think about donor site morbidity, I think you have to put it in context.

            DR. RAO:  Dr. Tomford.

            DR. TOMFORD:  I just wanted to add that Dr. Minas in Boston who has done a lot of these has an upper-age limit of 40 that he uses to procure -- that he does the procedure on.

            DR. RAO:  So there's a range but in general it's middle-aged people and there seems to be some upper limit that there may or may not be consensus on.  Is that fair?  Dr. Harlan, did you have a comment?  It seems that is pretty much what is done with cartilage.  Are there any sort of specific issues that one worries about when one is harvesting any other tissue sample? 

            It's really too broad for us and that's why I didn't want to have a discussion on that first but just sort of look at one particular tissue.  Is there any sort of specific burning issue in terms of taking tissue when you are using it for cartilage repair?  For example, Bruce brought up a point that when you harvest bone marrow you do it many different ways and there are many sort of gradient protocols that you use. 

            Maybe some of those cause you to lose your mesenchymal stem cell population which you think might be the way to go.  It's important in those cases to say when you harvest tissue it has to be harvested by a particular procedure.  Is that true for adipocytes as well?  Is it true for any of the other sources that people use for synovial cells?

            DR. COUTTS:  The attractiveness of the other tissues is its relative amount is much greater and you don't have to be concerned about taking away vital tissue.  I don't know anybody who objects to having their fat removed and the bone marrow is plentiful.

            However, the work that Arnie Caplan has done who has been one of the principal promoters of mesenchymal stem cells is that as you age, the number of mesenchymal stem cells in the marrow diminish so you have the same age issues from that source.  I don't know if similar studies have been done on sources of MSCs in fat. 

            You can also get it from muscle but muscle is a little bit more vital tissue and people object somewhat to having it taken away but you have a lot of excess there.  You can tolerate some removal of muscle, too. 

            But I think the attractiveness of the mesenchymal stem cells is presumably you are getting a cell that is designed to be activated and to go through a proliferative and differentiation pathway that would lead to recapitulation of the formation of articular cartilage given the right stimuli and that these other sites have much more plentiful sources of these cells.  But there is a lot of work to be done yet in terms of how this would work in the clinical paradigm.

            DR. RAO:  Dr. Prockop.

            DR. PROCKOP:  Well, on the matter of getting mesenchymal stem cells from bone marrow aspirates, we have gone through about 300 donors now.  We take a sample of about 2ml from the left iliac crest and the right iliac crest at the same time or within a half hour of each other. 

            We really don't find that much difference between one donor and the next in donors ranging in age from about 20 up to one donor is 70.  The difference is more the left iliac crest than the right.  It's a sampling problem as far as we can tell and we have trouble seeing any relationship with age and that.

            We find, yes, that not all samples grow well but the average we get two out of four, one out of two, one out of three growing extremely well regardless of age or sex.  Beasty turns out to be a problem, though.

            DR. RAO:  I want to ask the FDA before we go because as many members of the committee have already pointed out, it's the next set of questions which is really an important issue and it's a hard issue in terms of thinking about it.  Do you feel that in terms of sourcing material that you have an answer that you were looking for?

            DR. MOOS:  Let me just summarize what I think that we have in the way of an answer.  At present there really is limited data that suggest that there is a sourcing issue or that there isn't a sourcing issue. 

            I would just like to comment that this is, therefore, part of a manufacturing that is as likely to be as important as in any other type of biologic and to the extent that processes and results may be perceived as mixed, an important and possibly fruitful area for further work, and this leads into the next set of questions in that if there are types of data that could be collected on incoming material for evaluation by correlation to outcome later, especially analytical procedures that wouldn't require too much of the incoming tissue.  It might be very valuable for participants to consider this.

            DR. RAO:  Kathy, you had something to say?

            DR. HIGH:  No, I'm agreeing with Dr. Moos that it seems to me that is really the critical issue is to try to characterize better about the source material and then what comes out on the other end and correlate that with outcomes as Dr. Blazar said earlier.

            DR. RAO:  Let's try and tackle that issue then.  Let's imagine for a moment that you've taken acceptable tissue, you have a graph, and it's pretty clear that for most purposes, at least what is being considered here, these cells will have to be grown in culture for some time period.  Once you've grown them for a certain time period they are going to be shipped in a while of a certain number of cells which is what you are going to implant. 

            Clearly one can imagine this is the product.  You have a vial of cells and that is your product.  In any product you really need to know what are the characteristics of the product.  You need some basic characteristics of what you're sending to people, right?  What is the number, how many cells are alive, what do you need. 

            As everybody pointed out, you also need to have some idea in terms of its potency in some fashion.  Is it the same as what you send for other people if one lot is the same as another lot?  Is there any predictive value to this therapy in terms of any same cells?  That is all very critical in terms of having any kind of product. 

            We need to consider, at least in terms of articular cartilage or any cell product, what would be the issues.  I know I held off people making comments on this until now but maybe we can go back to the same people and ask them to enunciate. 

            Before they do that, I just want to remind people that this is something that has to be done for every cell product and different fields have evaluated different ways to do this.  You heard a little bit from Dr. Moos on what's done and what may work and what may not work so dye exclusion is not a great idea in pancreatic islets. 

            For bone marrow there are some surrogate markers which work really well in terms of having prediction.  You can look at CD34 expression of the number of CD34 positive cells in your sample and that seems to be a reasonable correlate.  One needs to know whether you have those two types of measures as far as a cartilaginous product is concerned.  Maybe, Kathy, you can --

            DR. HIGH:  Well, no.  Just yesterday we heard Dr. Luyten present a panel of six markers RT-PCR that had at least some correlation with biological activity.  We heard the Genzyme people say that they check expression of collagen II and aggrecan, I think, in terms of the product.  So it sounded to me listening yesterday as if this was all still in a state of evolution in terms of what markers are useful and how they may correlate with clinical outcomes.

            DR. RAO:  Bruce, do you --

            DR. McILWRAITH:  Dr. Rao, I've just got a question first.  You sort of alluded in your introduction there saying that you have to have some markers for product like product needs to be consistently tested, or at least have some baseline properties?

            DR. RAO:  I think that is pretty much a requirement most times you have a product.  The FDA can correct me if I'm wrong but there are two issues generally in terms of having some kind of product definition.  If it's a manufactured lot which is going to be used in many patients, then one would have to have some kind of definition that says that every aliquot of that lot is the same. 

            If you have indication for which you are going to use cells and a lot is a single batch, then you have to show that each lot will have more or less the same predictive properties for the indication that you're using it.  Otherwise, it's very hard to define what you're giving so that's going to be critical. 

            Go ahead, Bruce.

            DR. BLAZAR:  It seems that the field right now might be in the state that there is a minimal requirement for some sort of marker or other assay to say that the product contains chondrocytes.  Then there will be an evolution of the field that will allow one to say this is the critical constituency of the product that will be associated with the most favored outcome. 

            It just seems that even with the very impressive molecular correlations right now that we saw yesterday, we are still at a stage of correlation coefficient appointee which means a variability of 36 percent in the product. 

            Then trying to relate that to outcome in patients is still a huge missing link.  Maybe the question should be divided into what are the minimal requirements now to say the product has the potential to be therapeutic, versus what is the optimal product.

            DR. RAO:  Thank you.  That's really very useful, I think, in terms of that division.  We should consider that.

            Dr. Tuan.

            DR. TUAN:  I would like to make a comment also related to, say, the chondroprogenitor cells that one can get from, say bone marrow aspirates.  We have also done a lot of human bone marrow aspirates a little bit differently from what Dr. Prockop said.  We actually took the bone marrow from OA patients, patients who have elected total joint replacement, most often for degenerative joint diseases, sometimes for osteoporosis related things. 

            We've tried to see -- the comment was made by Dr. Coutts earlier that perhaps the cells from that patient who has a disease state may actually be more responsive because of more acute sort of alertness to the disease itself. 

            We were trying to see if these OA derived bone marrow cells were able to do all the things that mesenchymal stem cells are supposed to do and they certainly do.  I don't know if they divide slower or not.  I don't think so because we have some non-OA patients as comparison.  So the bottom line is that they can do it.

            Then the question here is what are the characteristics of functional regenerative cells.  It seems that there are cells that we get from certain patients that don't do anything.  Very rarely but sometimes that happens.  We have been paying attention to why they don't do certain things.

            First of all, morphologically they look very different.  It's hard to say.  It's sort of an eyeball thing but they do look different.  Their proliferative rate is a little bit different.  They are usually a little bit slower than the general population of OA patients -- cells of OA patients. 

            Then the other thing is that they also seem to differ a little bit in terms of the cell surface markers.  I think there are some hints out there.  We're not quite there yet.  I think as we do more and more of these cells and find the ones that don't work, then we begin to say, well, those are maybe the ones that we shouldn't use.  That would be my comment.

            DR. RAO:  Bruce also mentioned the same thing, that one should divide it into two parts, right?  One part, one can imagine, would be collecting information which might be really helpful to do, the second part of what is being said.  That information could be a whole lot might be needed to define an optimum product.

            There is also a minimum set that one needs to worry about in terms of being able to say since these cells are already being used in some fashion.  Let me give you what I have often thought about this in terms of putting cells into the brain. 

            When we think of putting cells into the brain, and I have a sample of cells, my big worry apart from saying, well, I've got cells which can make neurons, I also want to know what else do these cells make because I can't have cells making muscle in the brain. 

            It's a huge problem if that happens.  I can't have connective tissue cells which are making too much ECM.  When we test the cell sample, we have to test it very vigorously not just for the ability to do something.  Not just say, for example, the ability to make neurons but also the lack of ability or the absence of certain things and that becomes a very critical piece that we need to worry about. 

            We also have to worry about relative proportion in our case when we are putting cells back in the brain.  We want either to make mostly neurons or sometimes we want to make the support cells of we want to make glia so we need to have some way of being able to say what we are putting in as those characteristics. 

            The way we harvest cells in the nervous system it's impossible to get a defined population which is pure so that I can say it is 100 percent of these cells.  But that doesn't stop us from putting cells back, right?  So what people do is take one step back and say, "Well, we can't tell you exactly that we have a pure population but we can tell you in what different cell types we have based on an accepted set of markers." 

            That is the strategy, for example, in bone marrow, right?  When you do bone marrow transplants you say, "I have this percentage of CD34 positive cells and I have X number of other cells which also went in." 

            Or we don't want to put RBCs so we think these cells are not important so we have depleted that particular cell population and here is the criteria by which we have depleted them.  It seems to me that even in a general basis one can maybe make a general statement that one can't just look at a cartilage product to say it makes cartilage. 

            One maybe needs to worry a little bit about what it should not make or what should be absent and one should have some way of saying when I give you 10 million cells, out of these 10 million cells 80 percent will be homogeneous for the expression of X number of markers and there will be 20 percent of cells which are contaminating population with the Y characteristic.

            Maybe that's not important right now because we don't know whether the Y cells are better than the X cells but maybe we do need to know that.  Does that to the people in the cartilage field seem like a reasonable thing?

            Go ahead, Dr. Coutts.

            DR. COUTTS:  What you are describing sounds very rational.  I hope that at some time in the future that there will be some marker that would be identifiable that would have high predictive value for success.  I would just want to point out that making that correlation is going to be extremely difficult because there are so many other confounding variables.

            You have the defect itself which is being repaired.  It varies in size, location, severity.  Then you have the patient factors, their activity, their conformance with recommended rehabilitation methodologies and perhaps maybe some restrictions on activity, a wide variability in their activity levels.  And you have surgeon variables. 

            Some surgeons are better technicians at doing this than others so there's going to be a range of outcomes based on these other variables and your particular marker may not be able to get through all that noise.

            DR. HIGH:  Yesterday when we were talking about this I was wondering what was the utility of doing MRIs since it was not completely clear that the result on the MRI would correlate with patient improvement but at least that would provide you some data to help understand how well the cells had filled the defect or had grown. 

            It would be some objective measure that might help you filter out some of the noise possibly.  If you're just trying to look at numbers of cells and how well they covered over the defect. 

            DR. RAO:  Bruce.

            DR. BLAZAR:  I think we've dealt with some of these same variabilities in the transplant field.  There are a multitude of diseases, private practitioners, academicians.  I think over time one still is able to look within disease groups, within different settings related to the number of transplants done in the center per years of predictive outcome for some diseases. 

            The support of care varies from center to center.  Yet, I think as the field evolves it will continue to collect data.  At least as an aggregate you'll still be able to do some correlations with what is a reasonable type of minimal criteria for cells going in. 

            It may not be perfect but, you know, I would take a little more optimistic view in the long run that with enough aggregate data you will be able to look at prognostic factors of practice and diseases that will allow you to get as close as possible to a predictive outcome since we've gone through the same sort of experience in the last 20, 30 years now in the transplant field.

            DR. RAO:  Let's extend that also one more step which is a worry that came through to me from yesterday's presentation as well is that are markers even going to be enough given what we heard from Dr. Tuan and Dr. Luyten that the cells look more or less the same except for maybe small differences in cell division.  But if you pass them one more extra passage or you use this sort of nude mouse assay in terms of its ability to form cartilage, that was a much better predictor. 

            On the other hand, that is a lot of additional work in terms of being able to do it and is that an absolute requirement because otherwise there's no way to say what the cells are going to be able to do and is that the kind of information that is going to be critical?  Before you respond, maybe I'll have Dr. Harlan.

            DR. HARLAN:  I have kind of two thoughts.  One is, I think, Bruce would agree that with bone marrow transplantation we've gained knowledge simply by doing things we didn't know but we had the advantage of the diseases that were being treated having such a severe prognosis that the risk was worth the experiment.  The risk of those patients was worth the experiment.  So that is one thing I want to bring into this, just how risky are these various therapies for patients with defects in their joints.

            The other thing that I think I keep hearing is that we would like for minimal criteria but if I'm hearing the practitioners correctly, we don't know enough to even begin to establish what those criteria would be.  Is that true?

            DR. RAO:  This might be the right time to ask Genzyme to see what is the minimal criteria they use in terms of using the cells for therapy.

            MS. LAWTON:  Unfortunately, I don't think we actually have anymore advice or experience to give to you other than what was previously said about the number of cells, the viability and morphology.  The only thing that I would comment on is during the presentation yesterday with the biomarkers, I think it was very interesting work but we heard yesterday that even between mouse and goat and certainly humans we don't know whether those biomarkers are helpful in a clinical setting still.  Unfortunately, we don't have anything else to add for you.

            DR. McILWRAITH:  You got a bit cut off at the end there before when the other person was saying about type II collagen and aggrecan expression.  That is very logical if you are worrying about getting hyaline cartilage so how do you do it or when do you do it?  After they are finished doublings and you just do RT-PCR or what do you do?

            DR. WILL:  We did the agarose and alginate assays on final product during our process validations because it takes several weeks for the cells in these suspension cultures to redifferentiate and start expressing COL2 and aggrecan.  We can't do it as a release test on each lot of product.

            DR. RAO:  While you're there one more question.  When you release a lot do you keep an aliquot on the shelf for future testing?  Do you have a mechanism?

            DR. WILL:  We keep a retain of each lot that goes out.

            DR. RAO:  And roughly what is the size of the retain?

            DR. WILL:  I believe it's a million cells.

            DR. RAO:  A million cells.  Go ahead, Dr. Harlan.

            DR. HARLAN:  I have a question, too, before you leave.  You described earlier a 10-fold magnitude difference in starting material.  Have you ever tried to correlate how many cells you started with with clinical outcome?

            DR. WILL:  No, we haven't done that.

            DR. McILWRAITH:  Is that retain what you use and you then put that into agarose?

            DR. WILL:  No, we would actually take a sample of the final product.  Then the retain is actually frozen down so it's got DMSO added to it.

            DR. TOMFORD:  Well, we heard from Dr. Luyten yesterday and it's pretty clear in his article the expression of alpha1 type II collagen and FGF receptor correlated positively with our in vivo assay.         It seems to me the minimum we would like to know is are these cells capable of making type II collagen and do they have some of these receptors on them because if we're not sending the clinician back, or let's look at the patient who is the final recipient of this, something that's going to make articular cartilage, why are we doing this in the first place? 

            I think we have to have a standard that says that the product that's being developed and that the company, whatever company it is, is producing should at least meet the standard of making type II collagen and perhaps some of these other markers that were expressed in this paper.

            DR. RAO:  I guess you are enunciating some of the frustration that Kathy and Bruce and others have tried to point out, is that there is not enough information to say any one of these markers is predictive.  Even in Dr. Luyten's data it was predictive only for that sample in a nude mouse model.            There wasn't a follow-up correlation which said that same sample would be also predictive of making articular cartilage with high efficiency in humans.  The thought is correct.  I think we need markers.  The question to me is it's hard for me to say which markers because it's also possible that mesenchymal cells which don't make any of these cartilage markers may be great once they go into the appropriate issue in making samples. 

            That is sort of the conundrum, I guess, that Bruce has mentioned several times.  The thing is that's the worry.  Maybe we can make a general principle but I don't think we can make a specific list.

            DR. TOMFORD:  Well, it seems to me that you could do tests perhaps in other animals to see if these cells that are mesenchymal stem cells under experimental circumstances would form cartilage.

            DR. RAO:  Hold that thought because I would like to get to that immediately after we get the comments in the same section.

            Go ahead, Dr. Scully.

            DR. SCULLY:  I think it's really important that we use the right terminology.  There's been no technique that has shown us that it can make articular cartilage.  Right?  We can talk about maybe hyaline cartilage versus fibro cartilage but nobody has shown us they can make articular cartilage and that was one of the questions I asked frank when he made his presentation yesterday.  Because a cell makes type II does not mean it's going to make articular cartilage when you implant it.

            DR. RAO:  Dr. Harlan, do you have any additional comment on that?  Dr. Moos?

            DR. MOOS:  A quick point which is that sometimes negative and positive results can be interpreted differently.  In the context of a specific manufacturing process, and for the sake of example, one which begins with cartilage tissue, a release test that looks for type II collagen, as we have just been discussing, and shows a negative result can be interpreted that product will fail. 

            I think everyone would be comfortable with that conclusion.  For other types of cells like MSCs, as you point out, perhaps, we don't know this, but perhaps all bets are off.  On the other hand, the expression of type II collagen, I think we would all agree, is not sufficient.  There is some difference between negative and positive criteria at this point.

            The other thing is a little bit of discomfort I would like to express.  From what I'm hearing the sentiment seems to be implicit in some of what I've heard that the only useful qualification system is going to be the human being.  I think we may come to this in a minute.  Perhaps there are some intermediate types of analyses that we could consider in a minute.

            DR. SCULLY:  It's not clear to me that because a cell makes type I collagen that it's going to fail in an implant.  No one has given us that data.

            DR. MOOS:  If it doesn't make type II.

            DR. SCULLY:  Do we know that?  I haven't seen that data.

            DR. MOOS:  That's the sort of thing that will be -- well, that's the type of a negative that will be difficult to prove definitively for logical reasons.  However, if a cell that starts as a chondrocytes loses that capacity, it's suggestive of a more dramatic perturbation in its fate specification mechanisms than is consistent with the preponderant manufacturing experience with these products.  In other words, most such products perform a certain way in culture.  If they couldn't make a type II, that would be a multiple sigma deviation from existing experience.

            DR. McILWRAITH:  I think another important aspect there, though, is time.  We did the work with early expression changes after microfracture and at two weeks those cells were not producing any message for type II collagen but at four and six weeks they started producing it.  There's going to be a window whereby you could miss out on knowing what that is.  By eight weeks they are producing lots of it but it goes up like on a curve.

            DR. RAO:  Dr. Tomford, then Dr. Prockop, then Dr. Allan.

            DR. TOMFORD:  Well, I think the point about whether it makes type II collagen or not, this was a discussion we had a little bit yesterday about will fibro cartilage suffice.  My question would be if you're not going to use type II collagen as a standard, then type I collagen may suffice do you need an implant.  I mean, the microfracture makes type I collagen so do you really need an implant to make type I collagen.  Studies suggest that you don't really need an implant to make type I collagen.

            DR. RAO:  Dr. Prockop.

            DR. PROCKOP:  Well, I would like to come back to what Bruce says.  I think we have to just relate -- we have to get some markers I think is clear of the standards being released, themselves being released.  You've got to look at the outcome.  I think everybody is aware we're on the brink of huge technology for characterizing cells so it's difficult and expensive. 

            I learned on Monday that for $21,000 you can get a company to measure every message in the cell and give you the copy number quantitatively.  It's a technique that's been around for about 10 years but now a company will do it for you so you don't have to do it in your own lab. 

            There are now ways of getting at the message levels inside the protein in various kinds, intracellular and extracellular.  The idea that one marker like CD34 defines a cell to me doesn't go very far.  I think we have to think in a more sophisticated way.  In fact, we have two problems.  One is we are now on the brink of getting technology that fully defines what a cell is doing at time A. 

            That is a bit like looking at horses at the start of a race and trying to predict who is going to win.  We have to watch these cells as they change so we've got a difficult problem there.  We need a game where we look at the cell.  We start and we see how it responds to environments.  Then we know what the cell is going to be doing. 

            I think we are only at the stage now of putting that kind of data in our laboratory books and in the end coming back and putting in the clinical correlations in spite of the problems with those clinical correlations.  I think that's really where we are.

            DR. RAO:  Thank you, Dr. Prockop.

            Dr. Allen.

            DR. ALLEN:  I think just expand on that.  Unless I'm missing the point, essentially these cells are taken out of the collagen and put on plastic and allowed to de-differentiate.  Then we re-differentiate them back into chondrocytes within the defect.  We are already talking about the fact that these cells are susceptible to paracrine/autocrine effects from the tissue. 

            It seems kind of naive to expect that necessarily we can do much with the cells that shift where essentially you need to know within a very short period of time what they're doing.  What you really are interested in is what the cell is doing when it's been in that defect for a week or two weeks.  Predictive tests, I think, are going to be very muddy.               I don't think necessarily we can ask companies like Genzyme to do a whole lot more because what they are sending out is a cell that potentially has the ability to express those genes but whether it will choose to switch and do that depends on a whole bunch of facts which we don't even understand so that comes back to animal moles and what we can do in those.  I think markers out of a vial of de-differentiated cells is probably now the way to go.

            DR. RAO:  Rocky.

            DR. TUAN:  Yes.  So this is a question actually also for Genzyme.  You said you measure type II collagen and you also measure aggrecan expression, probably both gene mRNA and protein level.  Is there a cutoff?  I mean, what point?  Do you do real-time PCR? 

            I mean, at what level of type II collagen do you say, "Well, this is good.  We'll ship it out."  Or if it's below that, you would say, "We don't ship it."  Because that's kind of important and whether that actually correlates with any of the clinical outcome that you have tried to maintain.

            DR. WILL:  The assay that measures aggrecan and COL2 is a ribonuclease protection assay for message RNA.  We run an 18S standard with that to normalize for the amount of RNA in our sample so we do have a minimum signal that we are looking for.  These tests aren't used as release tests because of the length of time that it actually takes to grow the cells and re-differentiate them.  It's not something that we do on every lot.

            DR. TUAN:  So I guess my question is there must be lots that you deem unshippable or not worthy of being shipped.  There must be some cutoff.

            DR. WILL:  We look at the growth of the cells in culture.  We have certain windows that we expect each passage to grow in and if the cells aren't growing fast enough to actually meet that spec, then they are not released for final product.

            DR. TUAN:  So it's more based on the proliferative characteristics as opposed to the type II and aggrecan.

            DR. WILL:  Right.  We look at the morphology as well as the growth in culture.

            DR. RAO:  So do you look at things like LDH or cytochrome oxidase or number of dead cells in the culture or just a growth rate by some other measure?  MTT maybe?

            DR. WILL:  No, we do counts of the live and dead cells.

            DR. RAO:  With trypan blue exclusion.

            DR. WILL:  Not during the culturing phase.  It's phase contrast microscope.

            DR. TUAN:  While we are trying to figure out the characteristics, it's important to know what has been done and try to correlate back to what we could do in the future.  Since we actually don't have the type II collagen and the aggrecan data in evaluating the outcome it's hard to draw any conclusions at this point.  Like Dr. Prockop said we are in the starting point of the race.

            DR. RAO:  Go ahead.  Dr. Coutts first and then Bruce.

            DR. COUTTS:  I was just going to say that -- carrying on what Dr. Tomford said.  You could be shipping out fibroblast.  It may be send out fibroblast as the product because we don't know whether or not a fibroblastic repair is good enough.  It may be perfectly adequate.  We could do an MRI and find that the defect had been filled and that the fibrous tissue had attached to the surrounding articular cartilage. 

            If you could look at the tissue, you would find that it is pure fibrous tissue and it could function quite satisfactorily.  In fact, there's evidence to suggest that it would.  Having the cells expressing type II collagen and aggrecan may not be essential for satisfactory clinical outcome.

            DR. RAO:  Bruce.

            DR. BLAZAR:  Since there is outcome data and since there are cell growth characterization data along with whatever else is collected, what is being done to try to merge this to at least say here is a group of patients that does well versus another group that doesn't do well. 

            Here is the data we've collected for whatever stage it is because I think the animals are still developmental and they will integrate into this paradigm as it is tested.  Right now there are outcome data and there are at least passage data, cell characterization data.  There may be other correlates that can.  What I haven't heard in the last two days is what is being done to integrate the available data with the available outcomes.

            DR. COUTTS:  Maybe Genzyme would like to comment on this because they do have a registry that they have put a lot of effort into maintaining so they have some well-characterized patients that they could use to correlate against some of their culture methodologies.  I don't know.  Do you want to comment on that?

            DR. RAO:  Before you respond, Dr. Allan, do you have a specific comment?

            DR. ALLAN:  My comment is sort of related to that.  I just sort of get frustrated when I sit here and listen to the conversation because we are discussing something we have no data really.  Even with what Genzyme has already approved, there doesn't seem to be a lot a data to have any correlations. 

            I'm reluctant to even discuss like other products that could potentially go to market when we have no data to suggest what would work or what wouldn't.  I mean, in an ideal situation what I would say is like we need to develop a nonhuman primate model. 

            We need to develop all of the parameters that you would use to test those products on nonhuman primates and have some sort of correlates that you have some assurance that they would correlate if you did the same thing in humans.  I would be reluctant to even discuss what kind of parameters cells would have that you might consider in humans.

            DR. RAO:  Does Genzyme want to respond to Dr. Coutts?  Mention anything before we take up this important issue which has come through?

            MR. LEVINE:  I think the answer is in terms of the practicality of now retrospectively taking some of the scientific data from the manufacturing process and trying to correlate it to clinical outcomes, I think you just have so many variables I'm not sure you could really get a meaningful scientific result at this point. 

            I mean, just the clinical variables alone would be so confounding that I don't think you would really be able to get that kind -- I don't think you would be able to answer the type of question you're looking for.

            DR. RAO:  So maybe I want to emphasize this and maybe then Dr. Harlan can add to this.  I think it's really important to put down that there is a sense of discomfort among the entire committee in the sense that the amount of data collection even to be able to make some kind of prospective correlation seems to be not there and there doesn't seem to be consensus on what kind of collection should be undertaken as well. 

            It's hard to discuss this in terms of even making specific suggestions because there is no consensus or clear-cut consensus in the field.  All one could possibly say is that it's really, really important to collect additional information but we can't tell you what information.  Does that seem like the sense of the committee?

            DR. BLAZAR:  Maybe there needs to be a working group of clinicians to decide what are the important clinical parameters the way we've done in bone marrow transplant to try to do the same thing.  Then that could, at least, lead to a database.  It doesn't have to be an enormous database.  There could be some very important questionnaire type that goes into a database to allow this analysis.  If it doesn't move in that direction, I think we are always going to be having the same discussion.

            DR. RAO:  Dr. Scully.

            DR. SCULLY:  Dr. Levine, I hate to put you on the spot here but you have 10,000 patients in your database.  Some of these are approaching 10 years of follow up because you told me you started in 1995.  It seems to me there must be some way to farm the database statistically to come up with some type of data that correlates with outcome. 

            You can dismiss it and say statistically if we look at enough variables something is going to be statistically significant like every 20 variables one is going to be statistically significant.  It seems to me that somebody should be looking at that database and farming that database so that we have this type of data that this group is searching for and we have nothing to hold onto.

            MR. LEVINE:  Right.  Let me clarify one thing I said yesterday.  There's 10,000 implants but in order to get outcome data on a patient you need to have their voluntary participation in clinical research and you need to have their consent.  When we initially started the Carticel project in 1995 there was an attempt to enroll 100 percent of the patients in a registry type follow-up. 

            While the compliance was high early, it was not 100 percent and as time has gone on it's never been 100 percent.  Then as I mentioned yesterday, there was the additional impact of the HIPAA privacy law and we were actually very dedicated to try to continue to follow these patients. 

            It would have been impossible to contact a patient post-HIPAA without getting a new HIPAA consent consistent with the privacy laws.  We invested an enormous amount of time and money to reconsent patients consistent with the HIPAA privacy law.  I don't know the exact number off the top of my head but it's on the order of over 1,000 patients agreed to reconsent to additional follow-up post-HIPAA in a registry type follow-up. 

            Right off the bat there are not 10,000 patients available with the ability to contact in a post-HIPAA environment with clinical outcome so it's a smaller number of patients with clinical outcome.  Then I think within those 1,000 patients also going back to the discussion yesterday, if you look at the treatment algorithms and all the variables that just affect clinical outcome, I mean, I think there is a law of diminishing returns as you start looking retrospectively trying to find correlations between clinical outcomes and other variables. 

            Again, you know, I don't have this off the top of my head but my guess would be the variability within the type of scientific data that you're looking at with the variability of all the clinical factors just on the face of it it just seems to me that with this database with where we are right now you're unlikely to get a scientifically meaningful result that's going to further inform you and help you with the challenges you're asking about.

            DR. RAO:  While I understand that the clinical availabilities can confound the data, I mean, it does seem to me still a very important principle that one should have some way to be able to predict, at least say when I'm sending some cells, or a vial of cells, what they contain.  Right?  At a minimum set, apart from saying that they are live and dead by exclusion that I have a certain number by cell count.  Right? 

            That should be consistent when I'm sending it because if fibroblast work as well as cartilage precursors, then, you know, I'm going to save a lot of money harvesting fibroblast and it would be a different product in my mind then a product which is cartilage biopsy from a particular place.  Right?  So one would want at least that as part of a minimum set even if one didn't have a clinical correlate. 

            The issue I think with what we're asking even is was there any sort of data that's available which would allow one to make even that kind of really gross correlation.  Well, one percent of the samples turned out to be not what we thought they were because when you grow cells it's a mixed population and those are the patients which didn't show a response in respect to what was the clinical condition that we did it in.

            That's our hope that these cells are performing a certain function, cells of a certain kind.  If any cell can do it, then we need to know that at minimum.  Is there a possibility of getting even that minimal sort of information from the data set?

            MR. LEVINE:  I guess standing here in front of you I guess I'm not smart enough or fast enough to design this clinical experiment that you're asking about.  I guess what I'm suggesting just sort of on the level of face validity is that there's an enormous amount of process validation that went into the original Carticel process. 

            I think a lot of the test Jackie was talking about were the process validation.  We know the cells for every single lot.  Every single patient that's been treated have gone through a validated process and that the patients have met the existing release criteria. 

            Otherwise, the product would have never been shipped.  I don't know what the testing -- if you took these -- I think what you're suggesting is could you take these retained specimens somehow subject them to some molecular testing, and then try to correlate that with the subject of patients who are HIPAA compliant and try to correlate that with the subset of patients who are HIPAA compliant who have outcome data and see what the outcome is. 

            I mean, I kind of understand at a high level what you're asking and it has a scientific appeal.  That said, I haven't heard today, and I'm not smart enough to tell you today what is the molecular test you subject these retained samples to and would that meaningfully correlate with the clinical outcome.  Until we had a chance to think this through much more carefully what are the available molecular tests and is there a possibility to correlate some clinical outcome I don't think I can answer your question.

            DR. RAO:  Would you agree it's a point of concern?

            MR. LEVINE:  I guess I would agree it's a point of concern, yes.

            DR. RAO:  Thank you.

            DR. COUTTS:  David, I might suggest something simpler.  I know that you have the area of defect that was treated in each patient.  That's in the database.  At least it was an estimate by the surgeon.  You could do a regression analysis to see if there is a correlation, say, between the number of cells shipped and the size of the defect and determine whether or not there is a certain minimum number of cells that you would have to have. 

            Let's say that on average 10 million cells are shipped out and if you had 10 square millimeters of defect and you had the outcome of that patient versus someone that had a larger lesion, same number of cells, and see whether or not that patient also did well or did not do well, it would give some assurance with regards to the question of number of cells required to come up with a successful outcome as opposed to trying to find molecular markers that would be predictive.  But some of the more simpler questions could be studied with the data that you've got.

            MR. LEVINE:  I just don't know standing in front of you today without knowing the exact numbers of patients that would meet the criteria to answer any specific question whether it's even feasible.  I understand the type of question you're asking and I understand the desire to have that answer. 

            I think part of what you are asking me to answer for you today is it feasible to do this with the existing database and I can't answer that for you because without knowing the exact number of patients in the database with a HIPAA compliant consent that are available for providing outcomes information for any question you might ask, be it relationship, the question you asked or any other one, I can't possibly even tell you if it's feasible.

            DR. RAO:  I completely agree.  My apologies.  The idea was not to try and put you on the spot here in terms of saying there must be markers or something like that.  I just wanted to see what one would need to do in terms of product when one considers something which is going to be transplanted and how does it compare with what is being done, say, with bone marrow or being done with some of the other tissue cell types which are being transplanted.


            MR. TSIATIS:  I just had a comment.  Again, I don't know this data set but I do know there's been a lot of advances in statistical techniques to analysis data where there is relatively small number of patients and lots of variables per patient.  These are machine learning kind of techniques. 

            It seems to me that a thousand patients is a very rich data set even if there's lots and lots of variable that you have to shift through.  Again, I don't know all the data so I can't comment exactly but I don't think one should just throw up their hands and give up without looking at this more carefully.

            DR. RAO:  David.

            DR. HARLAN:  So I've heard -- it seems like we're all agreeing that we need a large database to analyze the data but that there's three significant hurdles that I've heard.  One is that we don't know what measures to measure so I would propose that we cast a very wide net there.

            Another one is the regulatory concerns.  We can at least state from this panel, it seems to me, that there has to be some way to overcome the HIPAA concern so that we can gather this data.

            The third one is that somebody is going to have to pay for all of this.  I haven't heard any mention of how this would all be paid for.  I think that we can cite from Framingham and Ansel Keys and any number of database studies that very important biological variables are uncovered this way and that we as a committee can recommend that some solution be found for those three hurdles.

            DR. RAO:  Dr. Prockop, did you have a comment? 

            Let me make a suggestion to the committee and then maybe I can have you comment.  Lot of people in tissue therapy or cell therapy face this sort of problem when they have a product which is relatively ill-defined and you really don't know which component of that product may be really the absolutely important component but you do need to know in some fashion what you are putting into patients. 

            In bone marrow when it wasn't quite clear which was the self-renewing stem cell population people said, "Well, we'll do analytical facts.  It's kind of wide net of markers but we will get a sort of profile of what we put in and it will be consistent with the broad profile." 

            As information was collected it became clear which subset of markers was absolutely critical.  That could be done with a small number of cells and could be done relatively quickly.  That became, I think, in many ways de facto standard. 

            I'll ask my colleagues who do this a lot more to comment on whether that would be a possibility in terms of taking a sample at release and saying this is what it comprises of so that you at least have some component data.

            DR. BLAZAR:  So we went through this in the bone marrow transplant field and started with colony assays and we did CFU-GM and CFU-mix, etc., and we found that some of those correlated with some of the early events but not the later events.

            We then went into a more extensive flow cytometry profile based in part upon a learning curve from the rodents and larger animals to say these subsets of cells may be important in certain positive and negative biological outcomes.  Through the years now there are very well-established cutoffs for the number of T-cells that are infused that can cause certain disease complications. 

            The number of CD34 cells that, in fact, are infused per kilo in the context of cord blood that have a very strict relationship to outcome and measured in survival despite differences in different centers, etc.  There have been more unifying characteristics as we have begun to collect the data.            One of the advantages of the current state of the art with Genzyme supplying materials is that the difficulty has been for CD34 counts, for example, variability from lab to lab.  But a centralized distribution site providing cells outward gives a wonderful opportunity of collecting a limited set of data based and directed upon in vitro biology and animal outcome data to then try to integrate into an algorithm or an outcome data set that will prospectively allow us to understand how best to treat the patients and who will have good outcome. 

            I think a very powerful part of this is the centralized distribution laboratory that would have a single quality control set of standards within their own distribution center, something that we don't have as much luxury in the bone marrow transplant community.

            DR. RAO:  Alison, you wanted to make a comment.  When you make that, would you also tell me whether it could be possible to do something like that in terms of a practical process of just looking by facts.

            MS. LAWTON:  Yes.  I'm Alison Lawton.  I'm Senior Vice President for Genzyme Corporation.  I just wanted to make a general statement because we're talking about our product, Carticel here.  But at Genzyme we also have a number of other products, recombinant enzyme replacement therapies for orphan diseases. 

            We have many patient registry programs where we believe that's a very important aspect to collect information, ongoing information for our patients and understanding the safety and the long-term efficacy of our products.  These are very expensive endeavors we're talking about. 

            I think one of the challenges that I know you all understand here is it's one thing to know what you are looking for when you are collecting this information but it's a completely different endeavor to start collecting information in the hope that when you do know what you're going to want to look at that you can go back and take a look at that.

            I think the feasibility is certainly a huge issue here.  It's easier in these smaller patient populations where less is known about the disease as well and people want together to understand that more.  I think at the moment obviously as a company we're the only company with this kind of product to be able to collect that information. 

            I think it is important as we move forward in the future that we try to pool this information and understand from a broader base whether the data actually tells us something as well. 

            I'm not sure, again, how feasibly you do that.  We've had these discussions many times about long-term follow-up of gene therapy and other such issues about how do you have some kind of pooled or national database.  It's a very difficult endeavor as to how you set that up for these types of products.          I don't have any answers.  I just want to caution the expense and the feasibility.  Obviously we all want to collect more information to understand and progress the science.  At the same time we don't want to get to a point where it is so expensive and undoable that it stops people from actually even getting those therapies available.

            DR. RAO:  Go ahead, Bruce.

            DR. BLAZAR:  Maybe one in between strategy is to pick a certain number of patients that fall into good and poor outcomes in a particular very defined disease setting so that it doesn't have to be every patient getting every product.  If you know that half patients will do well and half don't, you can do a statistical analysis to come up with a very limited data set. 

            Once you validate it in that data set, maybe you'll find prognostic markers and then you can interrogate other small data sets in other disease groups without the expense of maintaining and analyzing every entry into a database.  That may be a very realistic compromise situation.

            DR. RAO:  Bruce, is it fair to repeat what you said earlier that it's important perhaps to think of a working group of physicians in the field to come together with some minimal set of markers or collection of data that should be considered in terms of these cell types? 

            Does the FDA feel that is really an important conclusion, then that's realistically an important conclusion which can come from the limited amount of data that we have in terms of which markers work and what predictive value there is.

            DR. MOOS:  First I would simply like to comment that it isn't a huge surprise to FDA that the consensus of the committee is that, as I think Dr. Allan articulated, there's not a lot of data and we wish to push very hard to see what data people can agree on and it isn't much.  Then to ask, that being the case, how can we get something that is useful and comes to the level of the types of characterization and release test and process controls that we might like to see. 

            Dr. Prockop mentioned the fact that we are at a stage in technology development where methods that could be used to characterize cell products are becoming much, much more powerful and, indeed, in our question 7 we do talk about methods that could be used in addition to the data that might already exist.  Dr. Harlan used the words "cast a wide net."  There are discovery based techniques for analyzing mRNA expression levels. 

            We heard a bit about that yesterday.  There are also proteomic technologies and phosphoproteomic screening technologies that promise a lot of power in addition to now conventional things like analysis of cell surface markers.  I would definitely also echo Dr. Prockop's comment that we aren't looking for a marker. 

            Indeed, I use the term "sets of confidence factors" anticipating that what's going on in the dish, as Dr. Allan indicated, is not the same as what we want to have happening when we administer the product. 

            Still, the product may have characteristics sufficient to determine what Bruce has suggested which is to do a comparison between a set of known home runs, if you will, and known flat tires and apply some discovery-based techniques to make a comparison, do data mining, as a couple of people have suggested, and look for candidates. 

            Then in a sort of -- well, I don't know if it would be a really a Monte Carlo, but an out-of-sample testing procedure to see if that works.  It seems like a very reasonable approach.

            DR. RAO:  I also would agree that it's a reasonable approach given the history we have from other stem cell fields or other fields where the number of good candidate predictive markers which can be used for such things turns out to be relatively small.  It's not that you have to always be looking at 200 genes.  It's something that turns out to be a relatively small number subsequently once you've gone through the discovery process.

            DR. MOOS:  Right.  I think it's important to emphasize that.  It isn't that you need to do 200 or 2,000 on a routine basis but you may need to do a comparison once and do it well and that can let you narrow things down and then go back to a clinical database that may not need to interrogate all 10,000 that's in the registry but a feasible number because we acknowledge feasibility is an important issue and what we really need is -- well, what is needed is just as much as will give us an answer.  It doesn't mean we don't need to assay the entire database if we can get the right statistical power short of that.

            DR. RAO:  Before we move on to the next question, there was one piece which we haven't discussed in any detail and I just wanted to get the committee's sense on this and maybe the clinicians can tell us.  I raised this issue of knowing which other set of types are there.  In other systems that can be of great significance. 

            Is this true in the cartilage field?  Is it important to know what is not there or what is there as a cell type that you don't want that is a contaminant?  Is there a way to be able to test that in an easy way or is there consideration that should be -- maybe first we should ask whether that should be done. 

            Then, (b) one can ask whether that's possible in terms of whether it can be done.  Anyone would like to tell me whether they think it should be done?  Do you want to make a comment?  Dr. Tuan.

            DR. TUAN:  I'll make a comment.  Many cells that give rise to connective tissues tend to have the tendency to be able to switch or de-differential and then go do something else. 

            Some of the chondrocytes that we are currently using in the Carticel procedure perhaps can also do something else.  The point you made is an important one and that is someway of making sure that we are only going to be making something that can function like an articular cartilage and not something else is an important issue.

            DR. RAO:  Dr. Harlan.

            DR. HARLAN:  Well, my only thought is that if we have the reagents to look at these different cell types, that would be fine.  It seems to me that these gene chip approaches are extremely democratic in what they look at.  They look at everything and that would be a way to begin to look for things that might be targets to follow.

            DR. RAO:  And, again, these gene chips are relatively inexpensive.  Is that a reasonable --

            DR. HARLAN:  No, they are expensive.  That's why I said money is a big part of this.  But we've got to start somewhere it seems to me.  We don't know what the important -- at least, I haven't heard anybody say, "We know what the important parameters are," so we have to be fairly democratic in what we look at it seems to me.

            DR. RAO:  Dr. Scully.

            DR. SCULLY:  I agree.  That goes back to what Dr. Prockop said.  If we had an RNA phenotype on every cell, then we could tell exactly what that cell is and maybe do a gene chip analysis on these residuals would be a way of doing that but it is -- I mean, forget about the cost.  The amount of information involved is mind boggling. 

            It seems to me that you have a database.  It may not be -- it may have the RNA phenotype in it but that has to be examined.  It has to be mined and look for factors that could be predictive.  Then that could be supplemented with molecular markers if we think that's important.

            DR. RAO:  I agree with what Dr. Blazar said.  That is, that one needs to be focused on a limited cell.  It's not that this has to be done as a release criteria or any sort of routine analysis.

            DR. McILWRAITH:  One other comment on the gene chip.  We've got some experience with OA in the horse and came up with, as you say, you go mining and you start off and it is quite expensive because you've got to start with a lot of sequences and then decide what ones are useful. 

            The ones that were significantly upregulated with OA, at least in the horse model, were 20.  When you went back and identified what those genes actually were, there's a lot of surprises. 

            You can relate it to the disease but they do work as surrogate markers.  If you are going to get consistent change with disease, then you can later focus on designing a chip with just those genes on it.  Then it becomes a lot less expensive and becomes less of a bio statistical nightmare as well.  I agree it's a way to go.

            DR. RAO:  Sharon.

            MS. TERRY:  So a simple lay person's point of view here.  I think what I'm hearing is there's a lot that is unknown.  From what I've learned here and in other places correlation doesn't always equal causation.  We are looking at surrogates for outcomes that are clinical and human and that is always a very difficult pathway. 

            I see resolution decreasing as we move through those pathways sometimes.  I think that the questions we're asking are very global and apply to a lot of systems when we try to move basic science to clinical outcomes. 

            There is sort of this messy in between that says knowing globally whether or not this technique or therapy works is important and then refining that.  But during the refining there's got to be give and take between what's cost effective and also what then alters the outcome, whether it's a surrogate outcome or a clinical outcome in the end. 

            Those are very complicated questions that I think require not only large databases that probably should be federally supported in some way, not on just this issue but many issues, but also combining the various aspects of statistics and gene expression and cell growth, etc., in a way that we probably don't often do because those disciplines stay somewhat siloed.

            DR. RAO:  Go ahead.

            DR. HARLAN:  Just a quick comment.  I fear that I'm stating the obvious all the time but we're talking about so many variables here and I've heard several people say, Dr. Coutts and Dr. Murray yesterday, that it's going to be hard to tease out all the variables. 

            The thing that really peaked my interest is when they start talking about 10,000 patients undergoing this procedure because those are the types of numbers where you actually can start seeing signal through all the noise.

            DR. RAO:  Go ahead, Dr. Tuan.

            DR. TUAN:  Yes.  So I didn't realize we were talking about analytical methods already.  I guess one of the things that in addition to the molecular and proteomic and genomic data sets, one of the hallmarks of the hyaline cartilage or articular cartilage ideally is, of course, the structural properties which is very unique. 

            There have been a lot of studies out there that have attempted to I guess score the chondrocytes according to their ability to form in a three-dimensional way a structure that has mechanical properties that somewhat approximate, or even just approach that of articular cartilage.  Many nano or micro methodologies are now in place, indentational type of analysis to look at stiffness and Young's modulus and other types of mechanical properties. 

            If there are cells, candidate cells, and they need to be evaluated for the possibility of articular cartilage repair, I think one method prior to a use would be to somehow assess the ability to form a three-dimensional structure that has some type of mechanical property that appears to be favorable.  Those methodologies are available.

            DR. RAO:  I have a quick question for the committee.  Should we proceed with this discussion and the next two questions because that's a start, or should we take a quick break?  Everybody looks like they are ready to work so we don't need a break.  Is that right?  Let's take a five-minute break and then we'll be back.

            (Whereupon, at 10:18 a.m. off the record until 10:28 a.m.)

            DR. RAO:  If we can have everybody back in their seats, we'll start.  It should be short.  If we can all take our seats.  This is sort of a continuation of sort of thinking about different aspects.  I think a lot of the issues we've covered are highlighted as discomforts are probably going to be true in this next set of questions as well.

            This is really in terms of tests at release.  You had a licensed product and you were going to be sending it out were there any special additional tests that one would really absolutely need or one would recommend or one would look at.  Rather than just summarizing I'm going to read out what that question says.

            "For licensed biological products each lot of final product must be tested for identity, purity, and potency prior to clinical use.  Please discuss what analytical tests and acceptance criteria should be applied to each of these parameters to provide reasonable assurance of adequate product performance in vivo."

            We really did have a long discussion and it's pretty clear that there is as such no clear-cut predictive marker right now.  Right?  We have gone through this many times and there has been a sense of discomfort that there isn't one.

            The question, though, is should there be one or is there something that one can identify as a reasonable marker or a set or a test.  Dr. Tuan started by saying that it's cartilage so you should be really looking with some kind of matrix and culture and looking at the mechanical properties of the cartilage that are there.

            While that's not a test that can be performed during the time you are growing the cells and shipping the cells and using them is that a test that should be performed post hoc in some fashion with aliquot.  I feel, from what I've heard, that it's not clearly predictable in terms of what happens with those assays and what the function of the cells is in vivo. 

            If it's not absolutely predictable, then maybe it shouldn't be a release criteria test.  It should be something which is much more basic science in terms of understanding whether that could become a release criteria test but, as such, it doesn't provide that function. 

            Is there any test -- I mean, I would like your comment on that as well -- that would function that way which would be a reasonable prognostic -- I mean, which should be done in animals or which should be done in culture and in nude mice or any other model system?

            DR. TUAN:  So to go back to the -- I mean, there are criterion.  I think some of the ones that are currently used such as collagen type II, aggrecan, and other cartilage related markers, perhaps the FGF receptors and so on, I think there should be due diligence in at least presenting the expression level of these markers as part of the package in describing the cells.  Those are readily doable and expressing these type II and aggrecan and so forth, maybe type IX collagen, these are hallmarks of chondrocytes. 

            DR. RAO:  These would be different for different set of types that you're using. 

            DR. TUAN:  Correct.

            DR. PROCKOP:  Well, I think it would be nice to get that information but with mesenchymal stem cells or, as we call them marrow stromal cells, we are finding the differentiation capacity of those cells is not the important thing in many situations.  It's what they are making in the way of chemokines.

            In the spinal cord models, I think there is now agreement among five laboratories that you do get improvement in axon regeneration if you inject cells into an injured spinal cord but it's not the differentiation of the cells.  It's clearly the neurotopins they put out. 

            These cells we're dealing with under some circumstances pull out huge numbers of 50 or 60 different kind of chemokines.  I don't know if that applies to cartilage or not but, in a sense, we think for some applications we don't want the cells differentiated.  We want to know what they can make in the way of supporting other cells.

            DR. RAO:  So, Darwin, is that another way of saying that one should have markers which assess what you think is the predicted mechanism of action.  If you think mesenchymal cells are working by releasing chemokines, then your release criteria should say this is the chemokine release because this is how I think they work. 

            If your thinking is they are going to make cartilage, then this is the amount of type II collagen that they make at that stage.  If they are a precursor cell, then these are the number of precursor cells.  Whatever mechanism you postulate for those cells, that should be one of your criteria in a release.

            DR. PROCKOP:  No, I don't think so.  I don't think that's what it's saying.  I come back to what Bruce was saying earlier.  We should have some criteria now that is reasonable and practical.  We ought to be collecting together data and later come back to correlate that with outcomes.  I think that was what Bruce was saying earlier.

            DR. RAO:  That's true that we have to collect information but there is a separate set that you have to say for each lot.  Right?  Every time you manufacture it, you are manufacturing the cells but by the same protocol you have already verified that protocol and now you want to release those cells.  Every time you make a lot it might be slightly different.  There might be some other criteria that causes subtle changes.

            DR. BLAZAR:  Darwin, I think when you process cells you have a very good -- you collect information on passage number, time for a passage, what the cells look like, and the differentiation potential.  I think if you knew that a certain marker was associated with a predictive value to be able to do something in vivo, you would be collecting that.              There are probably easy things to collect now which are passage number, growth characteristics, etc., that are being collected but aren't necessarily related to the outcome data yet.  If you knew that you had a multi-potent differentiation potential of an MSC and you had a marker for that, you could certainly collect that information at least to say here is the quality of the product going into this inductive environment that we realize will change the characteristics of the cells in that environment. 

            But at least we know what we have infused has a multi-potent inductive potential.  It's not as if you're infusing differentiated or predifferentiated cells into that environment.  You still are able to collect some basic characteristics of cell growth.  And you still potentially if you had a marker able to collect the multi-potentiality. 

            If it was Alk4 in ES cells, you would be collecting that information to say that market is associated with the potential ability in vivo to have a certain biological function.  I don't think it has to be so complicated as to try to predict what biological markers in vivo have to then translate into the infused product so much as the capacity of that cell to have a biological activity once it's in the right environment. 

            I think we can take this down a level of sophistication as a quality of the product going in if we collect, you know, more rudimentary data than the final predictive value.

            DR. PROCKOP:  Well, I'm not competent with our cells we have those markers yet, those indicators.  Growth curves, morphology of cells and such, yes, we have.  We have lots of data now but I'm not sure we have the right markers we're looking at.  I'm not sure.

            DR. BLAZAR:  We should at least be working toward trying to understand the product that you're infusing rather than having to deal with the complexities of the inductive environment that is going to change the cell to have its desired characteristic is still a way to quality control what's going in to that environment and to that patient so that you can say here is what I put into there. 

            I think if we get too far in this process, it's going to get mind boggling to try to come up with a predictive gene or a transcript because it's too complicated but we still need to deal with the quality control of the product going into the patient.

            DR. RAO:  Let's separate the markers into two kinds.  One would be what identifies the cells that you're putting in in some fashion.  Right?  At whatever stage that you do, you define that stage because you have to make a commitment to what you are going to put in and you want to make sure that's what you do each time. 

            There would be a separate set of markers which would tell you or may be predictors of potency and we hope that those exist.  There seems to be consensus that right now we don't have any predictors of potency that we can use in an easy way. 

            That actually leads to the next part of the question which says if you don't have other animal models where you can take out a core of the cells and test which would give you a predictor of potency but we won't get there right now.  In terms of identity, it seems that should be reasonably feasible and there should be markers that exist and that should be part of a large release criteria in terms of being able to do it. 

            If you say it's chondrocyte precursors that you are providing, then you better know that it's chondrocyte precursors you're providing.  If you say it's MSCs, then by some criteria that you define them as MSCs you should have that as part of your definition of a lot.  To me that seems pretty obvious.

            DR. PROCKOP:  Not to me.  So each cell expresses about 18,000 genes on the average.  Right?  Cells differentiate to the level of expression of any one or whole bunch of those genes go up and down.  How much data do you want to define your cell?

            DR. RAO:  At the time you give the cells to someone to implant they should have a certain characteristic which is similar when you ship it to patient B as well or you should know.  Right?

            DR. PROCKOP:  How complete does the information have to be?  18,000 genes and the precise level expression at the protein level?

            DR. BLAZAR:  I don't think you have to get that complicated.  I think you could say that there are certain cell surface phenotypic characteristics whatever they are.  A limited panel that isn't fully predictive but at least as the frequency cells expressing those cell surface determinants or if they are producing a protein. 

            At least they have these characteristics so I can relate lot A to lot B not at an 18,000 gene level but at least at a production level so that you could have some assurances or be able to correlate in the future with outcome so there's got to be some panel of markers the way we do in other circumstances that at least say here is a biological characterization of a product.

            DR. RAO:  Dr. Tomford.

            DR. PROCKOP:  So we are agreeing and then disagreeing.  We are doing both.  Okay.  I think the analogy from bone marrow transplants is a little too easy for you guys.  You had a really easy time of that.  All you had to do is do it when the buck system cam back again.  Okay?  We've got a much tougher job here.  We've got to repair cartilage and that's not an easy game.

            DR. BLAZAR:  I'm only talking about the product.

            DR. PROCKOP:  I agree we should probably have a handful of markers to make sure each lot looks the same as the next one but to say we should have the markers to tell which cell is going to become a better cartilage cell or better cytokine producer, that's too complicated.

            DR. BLAZAR:  I'm not saying that.  I just wanted the first.  The first part we are agreeing on.

            DR. PROCKOP:  Right.  Yes.

            DR. RAO:  Dr. Tomford and then Dr. --

            DR. TOMFORD:  I think your point is well taken.  I would agree we are kind of lucky in cartilage in that we are not going to have a lot of other cells.  We are going to have chondrocytes now.  Admittedly there are certain levels of cartilage, as Dr. Scully said, so the transitional way of cells may be different from the deeper cells but, nonetheless, are all considered chondrocytes.  We don't have to worry about a whole lot of different cells.

            Secondly, I'll go back to what Dr. Luyten says in his paper.  "Under our conditions this set of molecular markers appear to be predictive of the in vivo capacity to generate cartilage tissue reproducably."  It seems to me that's a standard that we ought to consider, that it reproduces cartilage.

            Once you transplant this tissue, then all bets are kind of off, I think.  In part, from what Dr. Levine said, all the variables of the surgeon, the size of the lesion and what not but we're just talking about the product before release and I think we do have some criteria that we can use.

            DR. RAO:  So, Dr. Tuan, you had a comment on that?

            DR. TUAN:  Just a short comment.  I think a key issue is at what point are we shipping -- what is this product for?  Is this a completely precursor of cartilage or is it already almost cartilage?  The label or the information will be very different for these two types.  I think for the latter I think we actually have reasonable even early chondrocyte markers and so forth so we can do that so there is reasonable confidence that when this product is used it will lead to cartilage.

            The other one is a lot more complicated so maybe an effort that we need to make is to make sure that the label for the precursor product really justifies or predicts what is going to happen later.  Those are two different things.

            DR. RAO:  Go ahead, Dr. Coutts.

            DR. COUTTS:  I would like to charge Dr. Luyten to go back to the laboratory and to take those cells that were positive for cartilage in his subcutaneous injection model and those that failed and demonstrate that difference has an impact on the ability of those cells to effect a repair in a joint environment.

            If he could produce that bit of information for us, then we would have a bioassay that we could use for the cell products and it would be no different than injecting BMP into a subcutaneous pouch and showing that it forms bone.  I think we are kind of drawing on that analogy to suggest that this is a good bioassay but there is that one little piece of evidence that is still lacking.

            DR. RAO:  You made a really important point.  Before you comment on it, I would like to close out the first part which was just this idea that as a lot release I think when you define a cell as a particular population of cells for a particular function, then you need to be able to say not just that I have so many cells which are alive and have a certain set of dye exclusion criteria, that the oxygen consumption is this much, but you also need some set of markers which define what is the product that is being released so you need to have some identity or character of those cells which is consistent with whatever is in that criteria. 

            The selection of those markers should be consistent with whatever function you expect those cells to perform.  If you say it's a precursor cell that's performing that function, then you better have some assay or cellular markers for that.

            DR. COUTTS:  Right.  And, in fact, if he is able to identify markers that predict the behavior in the subcutaneous pouch, then we could pass by the bioassay and just do the bench testing of those markers and that would be even more terrific.

            DR. RAO:  So does the committee agree with that minimum consensus?

            DR. ALLAN:  Part of the question also says potency and you're not addressing potency.  You're just saying what the markers are for the cells.

            DR. RAO:  Immediately after we -- even if you have consensus on one small piece, then that is the next question is potency because of what we said.  That really leads to that and you enunciated it and Dr. Luyten is going to respond to that.  Okay, now you've got cells which have a certain capacity or identity but how do you know that they will do what you want them to do. 

            Do you have any surrogate measure of potency other than saying we have to put them into a human patient and follow them in some way for over a period of two years before we can say that set is okay or not.  That is a little bit hard to do because even if you find out it didn't work, it's too late to do anything about it. 

            We really need to have some measure and I guess everybody has been really anxious to get to that potency measure.  I thought it would be easy because it seems clear to me from all the discussion that there is no direct measure that we have for potency that anybody agrees on.  We can start with that and I'll let Dr. Luyten make a comment first and then Dr. Harlan.    

            DR. LUYTEN:  Well, as I said, the reason why we are a little slow in the large animal experiments is, first you want to make sure that there is a sufficiently long time to underscore the statement.  Secondly, we discovered whole new aspects of cartilage biology in the goat that don't seem to correlate very well with humans so there is lots of work going on. 

            But we will provide you these data and, at the same time, we have an ongoing and just randomized 120 patients in a prospective trial so we will be able to work from the nude mice model into the human model and see if that correlation is correct, too.  By the second quarter of 2006 we probably will be able to provide you this data.

            DR. HARLAN:  I want to endorse what I think I hear Dr. Prockop saying only because I haven't heard -- I heard some dissention among the experts about whether we have any idea what characteristics will predict outcome.  The minute you start to say the product has to have these characteristics, what I heard yesterday was that the only relevant outcome is what the patient tells you. 

            That is the only thing we know right now.  If we start saying they have to have these characteristics, I'm afraid we could throw the baby out with the bath water.  Have a very well-defined product that doesn't give the desired outcome.

            DR. RAO:  Go ahead, Bruce.

            DR. BLAZAR:  I think I'm still pushing for having a well-defined product and then determining whether it correlates with outcome.  You still need a well-defined product going into the patient.  I think we can't avoid that.  That is a really different question than what is the optimal product.

            DR. PROCKOP:  But, Bruce, just to challenge you on that point, CD34 people, I thought, are using now CD133 cells that are negative for CD34.  Is that true?

            DR. BLAZAR:  That's true, and CD34 is not, at least in rodents that are CD34 negative, they have stem cells but the issue is that has been correlated statistically and clinically with outcome data so I think it's collecting what we thought were potential correlates and then doing the outcome analysis to determine whether or not they have predictive value.               If you don't collect the data, you can't know if that has predictive value.  That doesn't say that the best cell type to infuse a patient is a purified CD34 positive cell.  It probably is not but we need to collect the quality of the graphs going into people to be able to determine what correlates with outcome, even if we don't understand all the biology of the graphs that we're putting in.  We still need a graph-to-graph ability to compare one graph to another graph.

            DR. RAO:  Dr. Harlan.

            DR. HARLAN:  So, Bruce, would you be -- I'm agreeing that we should characterize each cell prep as extensively as we can.  Unless we have data that indicate that this should be a product release criteria, we shouldn't specify those.  Why should we have -- I mean, other than things that would affect safety it seems to me we shouldn't artificially create criteria.

            DR. BLAZAR:  Well, it goes back to viability or mycoplasma or anything else that is being collected.  Does that have a full correlation with outcome?  Well, not necessarily but I think it tells us what the quality of the graph is that's going in.        We really need to, I think, collect that data.  If there is a limited panel of markers or proteins that we can say that the graph should have reasonable viability and be infection free, there have got to be some other markers that we could collect to say that this is the intended product. 

            As we have made 10 batches these 10 batches all share those characteristics so when we have batch 11 that doesn't, is that product reasonable to put in or not?  I think if we don't collect this data, even though we don't know, we are never going to get anywhere.

            DR. PROCKOP:  So let me kind of rephrase that.  I think we're agreeing, okay?  Our plans are for trial in spinal cord injury and the way we're going to go about it is we are going to prepare large numbers of cells with patients.  Then we have about eight hours, maybe 15 hours, to get those cells into the patient. 

            There are lots of things we can't do.  We can within those eight hours do about four or five cell markers for proteins, different proteins antibodies.  We are going to keep our retainers and we're going to do microarrays and microfluidics and that's kind of the way we are going about it.  We can't have microfluidics or microarrays data available before the cells are put in the patient.  It can't be done.  It takes three days.

            DR. BLAZAR:  I think that's very reasonable, Darwin.  The only issue is you're collecting the markers.  Do you have a criteria as a cutoff for what is a reasonable product?  If you are collecting them without a cutoff, then you'll do correlative data.  That's fine.  But if  you have 10 batches and batch 11 doesn't look like the other 10 batches and you've collected the markers, will you be infusing the product?

            DR. PROCKOP:  We'll have a cutoff but I'm saying those markers are not really predictive in my mind of what is going to happen in the end.

            DR. RAO:  Let me reemphasize we didn't say they are predictive.  They are just for consistency.       Go ahead, Dr. Scully.

            DR. SCULLY:  Well, I think the issue then as we get into this discussion is are we going to ask for certain data on these retained tissue samples and what is reasonable to ask for.  Do we do a gene chip analysis in every retained sampling, get all this data, and then not know how to use it and the expense and all those things we talked about?  I mean, I agree with what you said, Bruce, 100 percent.  I think there is a lot to be learned from the bone marrow literature but we also have to be realistic about how to focus what we're asking for.

            DR. BLAZAR:  I just want to say I'm not making -- I've never said that anyone should collect all 18,000 genes or even do gene chip analysis.  It's really we're not at that level until we could be guided by the animal studies or by interrogating a database with very big positives versus very big negatives and then asking what surrogates can we start to look at to have a focused interrogation.  I'm not anywhere near close to those statements.

            DR. SCULLY:  I agree.

            DR. RAO:  Let's try in the short time that we have think about potency in some ways.  I want to throw out one statement here given what we've heard before.  It seems that there's no consistent predictive potency assay.  We might have one by mid 2006 perhaps.  Right?  But there isn't a really clear-cut potency assay which one can use which will be predictive of the function of the cells once you transplant them in.  Is that a fair statement right now given the state of the field?

            Dr. Luyten, is that true?  Dr. Coutts?  Is there a likelihood that there would be a reasonable assay?  Would it be right now is the data which would suggest that it would be markers or does it really look likely that it would be an assay which would require implantation of cells in some kind of small animal model in terms of being able to look at the property of the cells, whether it's a nude mice implant or whether it's a rabbit model, something, but that would like be a potency assay?  Or even that's unlikely in the time period from the available data as a predictor?

            DR. SCULLY:  I think potency is a funny term.

            DR. RAO:  Efficacy and potency.

            DR. SCULLY:  I think the issues that Rocky brought up just before the break, probably the critical ones, which are mechanical properties of the tissue.  Not of the cells but of the tissue.  Then the other thing that we haven't really talked about is durability.  If you could put those things together, that is what I would equate to potency.

            DR. RAO:  So we know what a potency assay should look like in terms of what information you want back from it but it's not clear that there is any one simple straightforward potency assay that we can point to and say that is the way to go.  Does that seem to be a consensus?

            DR. ALLAN:  Well, I guess it's because it doesn't seem to me that there's a product that you can use that has known efficacy that you could design a potency assay for.  I guess the question I'm asking is I'm a little uncomfortable with talking about, well, we're going to have markers but we're not going to have any potency assays so we're just going to go ahead because since we don't have potency assays, we'll just go ahead and put the cells in.  Is that what you're saying? 

            DR. RAO:  No.  I just want to try and make sure that the problem is illustrated in a clear-cut way and whether there is a clear-cut solution or there isn't.  If there isn't, then that's a problem that has to be sorted.  Luckily it's not the committee's job to solve that problem.

            DR. BLAZAR:  I agree with what you said but let's remember that for the bone marrow transplant field we have no potency assays.  There just are none that have ever been developed that are predictive of long-term outcome.  We do the best that we can.  We collect information and we do correlates related to outcome to validate them.  It would be great to have potency assays for everything but we are still in a field that's more than 30 something years old without a potency assay for that product.

            DR. RAO:  Can I get Alison to make a comment and then Kathy and then Dr. Moos.

            MS. LAWTON:  Thank you.  I just wanted to make one comment.  Obviously it relates specifically to autologous cell therapies and it may do to other cell therapies as well.  I just wanted to remind you that if you are talking about an animal test for potency release, you are doing every single lot is every patient and then you are testing that each time in an animal.  Again, the practicalities and the feasibility of doing these types if they are going to be release assays.

            DR. RAO:  You are absolutely right.

            DR. HIGH:  Well, just getting back to Bruce's point, I know that cell number isn't potency but, I mean, for example, that we do know about bone marrow transplant.  If the numbers are too low, then the graph will fail. 

            Similarly could that be something like a surrogate for a potency assay eventually in this sort of thing?  I heard the comment yesterday, "There's no such thing as too many cells."  Could there eventually be criteria based on cell number per square centimeter of defect or something like that?

            DR. RAO:  Clearly that's absolutely right.  It's just what do the surgeons feel would be a good measure for them in terms of being able to say, "Look, when I put these cells in a patient and I can't take them out, I need to know whether they are going to work or not work and I need to be able to make some prediction of whether I want to do it or not.  Is there any measure, an ideal system, or anything that could come in the near future or that could be expected from every lot?  It seems to me the answer is no.

            DR. COUTTS:  This kind of gets back to the question that I was asking David Levine in terms of what they could tease out of the database that they have and knowing the number of cells that were shipped to any one patient and try to do something of a dose response curve. 

            As the receiving surgeon, you get the this vial with the set number of cells but you have the variable of the size of the lesion that you're treating, or lesions.  If you have lesions, in other words, more than one, you have to divide up those cells to treat each of those lesions separately with a smaller dose. 

            This kind of gets back to the potency thing because if the cells are capable in a lower concentration to still effect a repair, I think that is a good sign of potency.  But if you divide them up and you end up with failures, then it suggest that there's a critical number of cells that you have to have for any one repair. 

            It would be nice to know.  Since the surgeon has mapped the lesions that he or she is going to treat at the time of the biopsy, there ought to be communication with the culturing people and say, "I'm going to need x number of cells," if you knew what that number is based on the size of the area that you're going to be treating.  This would be good information to have. 

            I think that would probably be how I would address this issue of potency.  If you could come up with a number of cells per square area that would be a minimum necessary in order to effect a repair.  It's difficult data to get, mind you, because the other variables are also going to apply and it's not going to be a direct line relationship.  It would be worth the effort, I think.

            DR. RAO:  Let me make a statement and then let you comment on it.  It seems to me that nobody seems to think that it's really critical in some way with each lot to do cartilage prep, you know.  Put it in an animal model and show that you made this much of hyaline cartilage or do some kind of assay of that kind for each lot. 

            It would really be important at the very least to have some idea of what numbers do so that one would be able to make some prediction on that basis.  That would be what one would consider for each lot.  Not that one wouldn't want information in basic science about doing this.

            DR. HARLAN:  My comment is just to follow-up on what you just said, Dr. Rao, and what Dr. Blazar said earlier.  The only reason why we can do these things in the bone marrow transplant setting and why it makes sense to continue with these studies and cartilage repair is because the risk benefit ratio appears to be satisfactory. 

            Bruce's patients do well and I assume that the orthopedic patients do well.  But we have to admit what we don't know.  We can't enforce a potency assay when we haven't a clue what we're looking at.  I view potency and release criteria at this stage as in the same infant state.

            DR. RAO:  Dr. Scully.

            DR. SCULLY:  I don't disagree with what Dr. Coutts said.  My problem is that there is an assumption that the number of cells -- if you have an increased number of cells you can have increased matrix and better biomechanical properties.  I don't think we know that.  We don't really want the cells.  We want the tissue.  Right?  I mean, that's where our mechanical properties come from.  You may be right with that assumption but you may not be.

            DR. RAO:  On that note, and maybe you can make a comment on that given the time constraints as well.

            DR. TUAN:  A very short comment and that is most of the assays, even for developmental biology questions with regard to the ability of a cell to give rise to a cartilage tissue, requires three dimensionality and also artificial matrix of some sort, hydrogel, what have you.  So there is a way to get some kind of generic potency. 

            Not necessarily specific potency for each lot.  For example, we are going to go for whatever CD number you want to choose and you want to know whether those cells are potent meaning they can form a piece of cartilage.  There are assays that are not that long, maybe two or three weeks long at most so you can do them while you are preparing the cells.  You can have some idea of what the chondrogenic activity of those cells.

            DR. RAO:  Dr. Allen.

            DR. ALLEN:  I just have one comment.  There are obviously a lot of things we don't know but one of the things that the animals maybe ought to give you is some predictability.  Even if you don't use them as efficacy in terms of how good the final product is, you can still do a model way.  You look to issues such as the relationship between numbers of cells and the size of the defect as I mentioned yesterday. 

            You could do those experiments.  I mean, the logical thing to do even if there are differences in biology, I would start all of the stuff that we're talking about in humans is prospective.  Starting at some point in the future we may start collecting markers. 

            Well, the logical experiment to do is to take autologous cells from an animal, look at some of those candidate markers.  We have the ability to do full strings on those things.  Then look at some predictive endpoints. 

            If we are saying on MRI in humans that if we get fill at 12 months and that's appropriate, then let's do a goat or some other model, maybe a nonhuman primate or some other model, horse, whatever, use those surrogate markers that would be predictive of fill, and let's just see if we can fill defects of different sizes more efficiently by putting in more cells. 

            That's a start.  Then go backwards and look at what the markers may or may not be.  At least do the basic science at the same time as thinking about doing the human science and not just sit and do nothing waiting to get something from the humans.  We'll never know what the animals will do unless we start doing them.

            DR. RAO:  I agree with that and I think all the committee members at some stage or another have enunciated this that we need to connect the dots in some fashion.  We had a long discussion yesterday exactly on that point in some sense and saying, "You know, the way we collect markers in humans, the work we look at in humans is different from what we do in animals and it's different from what we are talking about in doing in cell culture and biomechanics. 

            None of the markers we talked about as biomarkers have a data correlation because the data simply doesn't exist.  There's a lot of basic science that needs to be done but, at least for this particular purpose for a release criteria, it doesn't seem like there is a potency assay and that, I think, there is consensus on. 

            If it's a brief comment, then we can go to that.  Otherwise, we need to wrap up on the other two questions.  Go ahead, Dr. Coutts.

            DR. COUTTS:  I just wanted to make the observation.  I'm hearing what sounds to me like statements that we have a predictive animal model.  We do not have a predictive animal model in which we can put cells and test them.

            DR. RAO:  Dr. Tomford.

            DR. TOMFORD:  I just wanted to add that I think from lot to lot given the fact there are chondrocytes pretty much and they can be determined microscopically, the product will be fairly uniform so I think the potency depends upon a number of cells.  That's my suspicion.  We don't have that but I think in terms of individual cell potency they are probably pretty much similar. 

            Perhaps vary a little bit with age, although going back to Dr. Luyten's study he said it did not vary with age.  I think the potency would depend in some degree on it.  I've heard people from Genzyme say sometimes they get these biopsies that aren't even articular cartilage.  Short of the data that Frank has promised us in another year or so, there may be some sort of intermediate things to look at.

            The other point, just to reassure Alison, I don't think -- certainly I intended to suggest that one needed a lengthy animal assay for routine release.  I suggested the possibility, and we may be short of having that be feasible at this point, of using certain types of animal assays to qualify more tractable tests. 

            We did talk about first linking the nude mouse or something else to the goat and/or, even better, the human.  Then we can go back with one of those assays and check the molecular characteristics which could be used as surrogates.  I think the feasibility aspect of things is something that we all appreciate.

            DR. RAO:  That really brings us to almost the end of the time we have for this and also to the absolute last question that we have on this list of questions for the committee.  I'm going to make an observation and see if everybody agrees.  If that's the case, then we have done all the discussion we have on this in the time we have.  If people disagree, then we will leave the FDA with that impression.

            The last question is, "Many products in this category consist of cells within biological artificial matrix."  Are there any additional concerns or criteria that one might need in that case?  The quick answer, and we've already heard from Dr. Tuan, for example, several times that's going to be different. 

            The way you implant the cells is going to be different.  It's going to be a tissue rather than cells.  The way you have to follow the predictions of function of potency will be different.  Yesterday we heard that the matrix itself has specific immune responses and other changes and, as a result, those sorts of tests will be different so one can't simply say I've done A and B and we can put them together and say that we have a result or a prediction because we've tested the cells. 

            I think apart from being able to do anything more than that, we really can't say what else one could do as a specific test much like what we can't say with cells.  It would be hard to say that with matrix given the lack of information we have.  I'm going to use that statement and ask if the committee agrees with it or disagrees and if they have any specific comments to add to that and then I'll ask the FDA whether they feel that's a reasonable summary of that question.

            DR. TUAN:  Maybe just a comment.  So you are 100 percent right.  If it's simply taking an FDA approved matrix material or non-FDA approved -- yet-to-be-approved FDA material and then you just add the cells, then it's a little bit different from having done some ex vivo longer-term cultures.  Now you end up with cells plus the artificial matrix and whatever the matrix themselves have made.  I think it's more complex.  I agree.

            DR. RAO:  Dr. Moos, do you feel we addressed most of your issues and concerns?

            DR. MOOS:  I think to the extent that it's reasonable to expect.  It's actually been very helpful.  As I connect the dots looking at pages of notes I've taken, I think it's a fairly complete picture of what we don't know. 

            That said, I would like to recall the very slide I showed this morning where when one thinks about product issues in biologics, you think about not just testing the product which we have been focused on in the last several minutes but on the source material which still has questions, and on the process itself and the things all fit together. 

            One of the reasons that we look at products like this in that way is because each of them contributes, each element of that triangle contributes importantly to trying to ensure a safe and effective product.  In fact, Genzyme indicated that they have undergone a great deal of effort to validate their process. 

            Whatever it is they're doing regardless of what we can argue about various kinds of tests or datamining and so forth, they are doing it pretty much the same way every time.  I think this is what Bruce is trying to bring up, that you have to have some minimal definition that allows us to do something other than sit on our hands. 

            It's not perfect but considering where we are it's workable.  Part of the reason that we asked you all to come the last day and a half is so that we can help everyone figure out how to make it better from where it is.  What that puts, I think, square on everybody's radar screen is that this distinction that I heard some people make between translational questions and quotation marks basic science is very artificial because some of these so-called basic science questions are square in the center of where we need to go to translate these products to the bedside.            With that final statement, I would like to extend my tremendous appreciation for all the effort and I think very informative discussions that we've had from all the panel and the presenters.  Especially to Genzyme for being will to contribute their experience and respond with such good grace through such enthusiastic questioning from time to time.  Thank you, Mr. Chairman.

            DR. RAO:  With that, thank you everyone and we'll declare the meeting closed.  We are going to reconvene for a second session for all the people who are involved with that and that will be in half an hour at quarter to 12:00.

            MS. DAPOLITO:  There is lunch for the committee upstairs again today.

            (Whereupon, at 11:13 a.m. the meeting was adjourned.)