* * *


                ADVISORY COMMITTEE

                       * * *

              38th MEETING - TOPIC I

                       * * *


                   MARCH 3, 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


PRESENT (Continued):



      Voting Member


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


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

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


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

GAIL DAPOLITO, Executive Secretary










RAJ K. PURI, M.D., Ph.D.











                  C O N T E N T S


Introduction, Chairperson Rao .................. 5

Conflict of Interest Statement ................. 6

Welcome, Dr. Karen Midthun ..................... 8

FDA Introduction, Dr. Malcolm Moos, Jr. ....... 15

Evaluating Methods of Restoring Cartilaginous

      Articular Surfaces, Dr. Joseph Buckwalter 23

The Biology of Joint Surfaces, Dr. Frank

      Luyten................................... 58

FDA Perspective on Development of Cellular

      Therapies, Dr. Richard McFarland ....... 109

Preclinical Animal Models, Dr. Matthew Allen . 122

Open Public Statements:

      Dr. Gloria Matthews..................... 165

      Dr. David Levine ....................... 172

Committee Discussion of Preclinical Questions 181

               P R O C E E D I N G S

                                       (8:03 a.m.)

            CHAIRPERSON RAO:  Good morning and welcome.  My name is Mahendra Rao, and I'm the Acting Chair, I guess the Chair now, the Chair of the Biological Response and Modifiers Committee.

            Please note that the name has changed.  It's now called the Cellular Tissues and Gene Therapies Advisory Committee.  That name change happened just -- I think it's the first time it has been meeting under this new name.

            Before I turn the mic over to Gail, I just wanted to explain a few of the ground rules for having this meeting, and I wanted to remind people of a couple of things.

            So I want to remind people that this is an Advisory Committee meeting, and it's mainly involved with getting comments from the committee members and the invitees who are going to be ad hoc members of the committee.  Everybody else is welcome to participate, but you have to be recognized by the Chair before you can do this, and there will be a special time period where you can be recognized.

            If you need to make a special statement, then you should please contact Gail and see whether there's time to be able to make that statement.

            As you'll see, you need to use the microphone, and I'm going to ask all the members of the committee to remember that you should switch it off to avoid background noise when you put the microphone off after you finish speaking, and then wait for the Chair to recognize you before you start the conversation.



            MS. DAPOLITO:  Good morning.  The following announcement addresses conflict of interest issues associated with this meeting of the Cellular, Tissue and Gene Therapies Advisory Committee on March 3, 2005.  Pursuant to the Authority granted under the committee charter, the Director of FDA's Center for Biologics Evaluation and Research appointed the following individuals as temporary voting members for the committee discussion of cellular therapies for repair and regeneration of joint surfaces:  Drs. Matthew Allen, Richard Coutts, Wayne McIlwraith, Alan Nixon, Darwin Prockop, Sean Scully, and Rocky Tuan.

            Based on the agenda, FDA determined that there are no specific products being considered for approval at this meeting.  The committee participants were screened for their financial interests.  To determine if any financial conflicts of interest existed, the agency reviewed the agenda and all relevant financial interests reported by the meeting participants.

            The Food and  Drug Administration prepared general matters waivers for participants who required a waiver under 18 USC 208.  Waivers were granted to Drs. Richard Coutts, Mark Hochberg and Alan Nixon.

            Because general topics impact on many entities, it is not prudent to recite all potential conflicts of interest as they apply to each member.  FDA acknowledges that there may be potential conflicts of interest, but because of the general nature of the discussions before the committee, these potential conflicts are mitigated.

            We wish to note for the record that Ms. Alison Lawton, the non-voting industry representative for this committee, recused herself from the discussions of Topic I related to cellular therapies for repair and regeneration of joint surfaces.

            With regard to FDA's invited guest speakers, the agency has determined that their services are essential.  For the discussions related to cellular therapies for repair and regeneration of joint surfaces, the following disclosures will assist the public in objectively evaluating presentations and/or comments made by the participants.

            Dr. Joseph Buckwalter is employed at the University of Iowa, Iowa City, Department of Orthopedics.  He consults with Genzyme.

            Dr. Marc Hochberg is employed at the University of Maryland School of Medicine.  He receives research support and consults with firms that could be affected by the discussions.

            Dr. Frank Luyten is employed by the University of Leuven, Belgium, and Dr. Charles Peterfy is employed by Synarc.  He has associations with firms developing non-cellular osteoarthritis and cartilage therapies.

            We would like to note for the record that Dr. Matthew Allen is serving as a temporary voting member and a speaker making a presentation for Topic I.

            FDA participants are aware of the need to exclude themselves from the discussions involving specific products or firms for which they have not been screened for conflicts of interest.  Their exclusion will be noted for the public record.

            With respect to all other meeting participants, we ask in the interest of fairness that you state your name, affiliation, and address any current or previous financial involvement with any firm whose products you wish to comment upon.  Waivers are available by written request under the Freedom of Information Act.

            Thank you.

            CHAIRPERSON RAO:  So this 38 committee has two new members, and I'm going to ask them to introduce themselves first.  Dr. Tomford.

            DR. TOMFORD:  I'm Dr. William Tomford.  I'm a professor of orthopedic surgery, Harvard Medical School.  I am an active clinician/orthopedic surgeon at Massachusetts General Hospital in Boston, and I have an interest in bone and cartilage storage transplantation and preservation. 

            CHAIRPERSON RAO:  Thank you, Dr. Tomford.  If you would give the mic.

            The other new committee member is Sharon Terry.

            MS. TERRY:  Hi.  My name is Sharon Terry, and I'm the President and CEO of the Genetic Alliance, which is a coalition of 600 advocacy groups.

            I come to that as a parent of two children with a genetic disease, having founded PXE International to support research on that disease.

            CHAIRPERSON RAO:  Thank you, Sharon.

            I'm going to ask all of the committee members to introduce themselves as well, and we'll start from the left with Dr. Nixon.

            DR. NIXON:  Hi.  I'm Dr. Alan Nixon.  I'm working at Cornell University.  I'm Professor of Orthopedic Surgery there and the Director of the Comparative Orthopedics Laboratory.

            DR. COUTTS:  I'm Richard Coutts.  I'm an orthopedic surgeon from San Diego, California with the University of California, San Diego.  I've had a longstanding interest in cartilage healing and tissue engineering of cartilage.

            MR. McILWRAITH:  I'm Wayne McIlwraith, Professor of Surgery and Director of Orthopedic Research at Colorado State University.

            DR. SCULLY:  Sean Scully.  I'm a Professor of Orthopedics at the University of Miami with an interest in cellular biology.

            DR. TUAN:  Rocky Tuan, Chief of the Cartilage Biology and Orthopedics Branch at the National Institute of Arthritis and Musculoskeletal and Skin Diseases at the NIH.  My interest is cartilage biology and tissue engineering.

            DR. ALLEN:  Matthew Allen.  I'm a veterinarian.  I'm Associate Professor of Orthopedic surgery at SUNY-Upstate in Syracuse, and my work really focuses on the validation and development of animal models of orthopedic  diseases and conditions.

            DR. BLAZAR:  Bruce Blazar at University of Minnesota, Professor of Pediatrics, and my interest is in cellular and gene therapies related to inherited disorders and bone marrow transplantation.

            CHAIRPERSON RAO:  I'm Mahendra Rao.  I'm at the National Institute of Aging and I work on stem cells.

            MS. DAPOLITO:  Gail Dapolito, Executive Secretary for the committee.

            I'd also like to introduce the committee management specialist, Rosanna Harvey.

            Thank you.

            DR. TSIATIS:  Hi.  I'm Butch Tsiatis.  I'm a Professor of Statistics at North Carolina State University.

            DR. ALLAN:  I'm Jon Allan.  I'm a scientist at the Southwest Foundation for Biomedical Research, and I'm a virologist and I study pathogenesis of SIV in non-human primate models.

            DR. HIGH:  I'm Kathy High.  I'm a hematologist at the Children's Hospital of Philadelphia, and  my research interests are in cell and gene therapy.

            DR. HARLAN:  I'm David Harlan.  I'm the Chief of the Islet and Autoimmunity Branch of the NIDDK at the NIH.  My interests are immunotherapies for Type I diabetes.

            MR. STEVENS:  I'm Ted Stevens, another at Kaiser who's been delayed.  I'm the Chief of the Restorative Devices Branch in FDA's Office of Device Evaluation and the Center for Devices.

            DR. McFARLAND:  I'm Richard McFarland.  I'm a medical officer in the PharmTox Branch of the Office of cell Tissue and Gene Therapy of the Center for Biologics.

            DR. LEIBENHAUT:  My name is Susan Leibenhaut, and I'm a medical officer in the Office of Cellular Tissue and Gene Therapies, the Clinical Evaluation Branch.

            DR. MOOS:  I'm Malcolm Moos.  I'm a product reviewer in the Division of Cellular and Gene Therapies with interests in cellular therapies for repair and regeneration generally and research interests in the characteristics of cells that can be used for those purposes.

            CHAIRPERSON RAO:  If we could have people at the FDA introduce themselves very quickly, too.

            DR. PURI:  Do we have to use the mic?  Okay.

            I'm Raj Puri, the Division Director of Division of Cellular and Gene Therapy, the Center for Biologics Evaluation Research, FDA.

            DR. FREY:  Joyce Frey, Deputy Office Director for the Office of Cellular Tissues and Gene Therapy at CBER.

            DR. WITTEN:  Celia Witten, Division Director of the Division of General Restorative and Neurological Devices in the Center for Devices for two more days, and as of Monday, the Director of the Office of Cellular tissue and gene therapies.

            DR. MIDTHUN:  I'm Karen Midthun, Deputy Director for Medicine, Center for Biologics.

            CHAIRPERSON RAO:  So now that you've met all of the big chiefs at the FDA, I'll ask Karen to come and give a few words. 

            DR. MIDTHUN:  Good morning, and welcome.  I'd like to take this opportunity to thank Dr. Bruce Blazar, Dr. Katherine High and Ms. Alison Lawton, all of whom are completing their terms on our Advisory Committee. 

            We very much appreciate the expertise that you have brought to this committee and the time and dedication that you have given to us.  We recognize that is a very large contribution and very much appreciated.

            You've helped guide the center through many challenging issues in the gene and cell therapy arenas, and we hope that in the future we can also draw upon your expertise.

            And as a small token of our appreciation, I have some plaques that I would like to give to each of you.  So if you would please come up to the podium.

            (Whereupon, plaques were awarded.)


            DR. MIDTHUN:  And I just would also like to take the opportunity.  Fortunately Dr. Witten was able to introduce herself.  She is joining us as Director of the Office of Cellular Tissue and Gene Therapies.  We're very excited to have her joining us.

            She is coming to us from Center for Devices, where she has been the Director of General Restorative and Neurological Devices, and she brings with her a great background from Center for Devices and also a background in rehabilitation and physical medicine.  So we're very, very much looking forward to having her join us, and I'm sure you will have an opportunity to get acquainted with her.

            So thank you, and thank you for your contributions, and now we'll get on to some serious business.

            CHAIRPERSON RAO:  Thank you.

            I guess we can get on to the main part of the meeting here, and I'll start with asking Dr. Moos to give us the FDA's perspective of the questions and the issues that they would like us to discuss.

            I want to remind everyone that today we have a pretty long day, and so all of the speakers, please try to stay within your allotted time and give us room for a reasonable number of questions.

            DR. MOOS:  Thanks, everyone, for joining us for this discussion of cellular therapies for the repair and regeneration of joint surfaces.  This is an area where the agency recognizes an increasing amount of activity and significant opportunities, but also where we see that there are significant challenges, and in situations like this, we like to try not only to tell you what your problems are, but to take an active role in helping to guide participants in the field past those problems.

            One kind of activity that we've been involved with for some time but now has an official name and official sanction so that it's called the critical path initiative.  Originally we were going to hear a detailed report on this at this session, but to make room for what on your agendas is listed as Topic II tomorrow afternoon, that was dropped.

            However, I'd like to make first the point that information on this topic is available through our Web site, given at the bottom of this slide, and secondly, to call attention to the idea that advisory committees which conventionally have been used to assess specific products or proposals relatively late in the development scheme are also acknowledged as being useful much, much earlier, particularly in the areas of cell and gene therapies, as we've been finding over the past few years, to help us identify what scientific tools are needed to move products from bench to market.

            Now, the kinds of questions that arise sort of generically for cell products are several.  Initially there is often a focus on what is needed to get into exploratory clinical trials, and this depends on getting enough proof of concept data to justify that there's a rationale for pursuing these sorts of experiments in the first place and also enough supporting safety data from preclinical studies to conclude that the risks are acceptable in light of potential benefits.

            In addition, one needs to know enough about the product to have some idea of what it is, whether it is demonstrating reasonable integrity, viability, and so forth; whether it's pure and other specifications, such as enough stability to support the trials and shipping qualification, free stock qualifications and so forth.

            But those questions are not exactly the same questions as need to be answered to get the entire product development cycle through marketing approval completed.  So after initial trials have begun, it's important to continue to address issues relating to the product: of potency, viability, more studies on how to identify the product, the impurities that might be in it, how best to assess purity, and these other specifications.

            So it's very important in view of the difficulty with cellular therapies for several of these not only to keep going and trying to work hard while initial trials are underway, but to start thinking about these issues hard as early as possible in the development cycle. 

            As confirmatory trials are nearing their completion, it's important to have most of these issues ironed out, and I've listed some of them below so that licensure can be entertained, and perhaps next generation products can begin their development cycles.

            Now, these are questions that are generic to lots of types of products, and  during the discussion, I am sure  that there will be many issues that will make one or another of you think of these broad applications, and we recognize that one area of endeavor can inform another.

            I will show you a specific example or two of that tomorrow.  However, the Chairman and the organizing committee agree that we want to try as much as possible to stay focused today and tomorrow on those issues that are specific to cells that might be used for the repair of joint surfaces.

            And so Dr. Rao has been empowered to use all means fair and foul short of physical violence to steer you back on track as necessary.

            Now, the committee has been posed with a number of specific questions which have been set down in print in the packages that most of you have, and I'll just say a few general words about them.

            Conventionally the products of this type that have been written about and considered have begun with samples of predominantly autologous cartilage from a non-weight bearing surface of usually the knee, and the knee has been the focus for most efforts to date.  We hope that some of what we discussed today might have applications in other joints, but so far that remains for the most part in the future.

            Other collection sites have been entertained, the periosteum, the synovium, and there are other things that have been imagined and proposed as well, mesenchymal cells derived from bone marrow stroma or from bone marrow, autologous fat, even allogeneic sources of various types, and the committee will be asked to discuss some of the issues involved in identifying appropriate starting materials for the manufacture of these products.

            Once the tissue is collected, it is placed generally in culture for purposes of expansion with or without some other types of manipulation, and there are many questions about what happens outside of the normal in vivo environment that need to be addressed, and some of these are also going to be a focus of our discussion.

            Among the possible manipulations are the addition of other types of components, whether these might be substances produced by the cells themselves or artificial matrix components that might introduce specific considerations that will need to be factored into characterization of the products in development of release testing.

            A very important part in the development cycle is the use of the appropriate models for various different purposes required to support safety and proof of concept, and also to use them in the appropriate way.  Possibly more than one model might be needed to contribute to the overall picture of what might be expected in human studies.

            And finally, the products are readministered in some form back into the patient, and there are issues with the design of appropriate scientifically powerful clinical trials.

            During the course of the meeting, the organizing committee has considered that there are three main themes.  The first is scientifically sound clinical trials.  The second, capabilities and limitations of preclinical models, and the third is how to manufacture and test products so that we know that they are safe, effective, and consistent.

            As you heard previously, and I think it bears repeating one more time, this is a meeting to address scientific and medical issues and is not intended to address any specific product or proposal.

            More than usually, there is interplay between these areas, and we hope to develop how preclinical models, for example, might be used to qualify potential sets of release tests, and how the preclinical testing may also best serve the design of appropriate clinical trials.

            I'd also like to acknowledge and to emphasize that the conception of the meeting has been done jointly between the Center for Biologics and the Center for Devices and Radiological Health, and members on this list of the folks who have helped me put this meeting together appear from both of the centers, and we have tried very hard to develop a consistent scientific and medical framework for considering all of these issues, and in all of our written materials have endeavored to avoid language that is associated with one or another specific regulatory mechanism.

            And so if you hear us slip and say things like Phase 1, 2, 3 or pivotal feasibility instead of exploratory and confirmatory, I hope we can be forgiven.  The underlying concepts really are what are at issue here, and what our focus is intended to be.

            And a special thanks, of course, to our Executive Secretary, whom many of you have spoken with, and to Rosanna, who has helped in many of the logistical details.

            Now, to serve those goals, just a rough road map, today we're going to hear some invited presentations, and those will begin with an overview of clinical considerations by a Dr. Joseph Buckwalter from the University of Iowa, and an overview of a number of bench to bedside issues by Dr. Frank Luyten from the University of Leuven.

            That will be followed  by preclinical presentations both from the outside and by the agency and by discussion of preclinical questions, then by a presentation and discussion of clinical issues which we anticipate might go long enough it will have to wind that up tomorrow morning.

            Then we will discuss the product questions and wind up this portion of the meeting tomorrow afternoon.  The second topic will be addressed under the direction of my colleague Dr. Carolyn Wilson.

            And finally, I'd like to repeat a little bit, a few things that Dr. Rao said.  A record of this meeting which will contain many, perhaps if we are lucky, all of the slides and a transcript of the proceedings, and to that end, I would like to ask participants to please be sure and use the microphones, and as excited as I'm sure some of you will get from time to time, to take turns when talking so that our transcriptionist is able to keep up with the discussions.

            That concludes my introduction.  So I'd like to invite  Dr. Buckwalter to give us his overview with my thanks for making the trip.

            Dr. Buckwalter.

            DR. BUCKWALTER:  Well, thank you very much. 

            I really appreciate the invitation to be here.  The biologic resurfacing of joints is an absolutely fascinating subject and one that I've been personally interested in for 25 or 30 years, and I also appreciate the invitation to give a surgeon's perspective on how we evaluate the outcomes of biologic resurfacing of joints.

            But I stress "a surgeon" because surgeons bring different perspectives to procedures and how they evaluate outcomes, and certainly you see that in a diversity of procedures that are used for biologic resurfacing of joints, decisions about which patients you operate on, which procedure you use for which patient.

            And I think in putting that all in perspective, it's important to understand how surgeons evaluate procedures and how they make decisions about which procedure to use on which patient and the factors that influence their perceptions of the outcomes of results.

            And I say this because most of the information we have about surgical procedures does not come from randomized control trials or hard evidence.  It comes from a series of cases assembled by surgeons, and a number of things influence their perception of the outcome of results, and I don't think these can be underestimated.

            Certainly the first is in doing a procedure if a surgeon believes it works, if he or she has considerable skill and experience in doing that procedure, not only does their perception of the outcome improve.  I actually suspect the outcome improves.

            One of the elements of surgical judgment that is so critical is what patient do you operate on.  It's not so simple as looking at an X-ray and saying this patient needs a total knee or a total hip," or they need a cartilage resurfacing procedure.  Surgeons factor in patient expectations, life style, their psychological make-up, a whole variety of factors in making a decision about who they operate on, and with time, I believe most surgeons get better at selecting the patients they operate on.

            They also get better most of the time at least in their skills, but one of the most important is who do you operate on.

            In general, surgeons, particularly in orthopedics because it is such a physical and functional specialty, tend to look at outcomes in terms of the technical success of the operation.  When we fix a fracture, we look at how well we restore the anatomy.  How close did we come to the way that limb looked before it was injured?

            We have a child with scoliosis.  We look at how well we corrected the curve, and I would suggest that most of the time in cartilage resurfacing, we've looked at how well we restore the anatomy of the joint.

            Fourth, we tend to look at short-term outcomes, and that's obviously essential for many of these procedures, but when we look at many of the things we do at two, ten, and 20 years, we see different results at each decade.  It's not uncommon that procedures enter the orthopedic armamentarium, and I suspect in most surgical disciplines without evidence from a control or comparison group, for that procedure, and it's common that there are no validated measures preoperatively and  postoperatively.  So I think it's important to put that in context when we start looking at biologic resurfacing.

            At the present time, surgeons face a wide variety of choices in terms of biologic resurfacing of joints.  The classic approach is, including penetrating subchondral bone, soft tissue transplantation are relatively well understood.  There are a host of others that are entering clinical practice and some of which you will focus on that surgeons have to choose from in making a decision for an individual patient.

            In addition, we do autografts and allografts, and of course, there is the very exciting, emerging area that's sort of a fusion of these two, which is tissue engineering, growing an implant outside the body in the lab and putting it in the patient.  In some parts of the world that's already happening, and that's certainly a very exciting technology.

            Penetration of subchondral bone, it's a 50 or 60 year old procedure.  It's stimulation of the marrow to regenerate the joint surface.  Around the world and certainly in the U.S., the most common way of doing that, and certainly I believe the most common procedure by all of the data that I'm familiar with is so-called micro fracture, which is penetrating the articular surface and articular defect in the femoral condyle here with a sharp instrument or awl stimulating bleeding and clot formation and then stimulating repair of an articular surface.

            Now, this is a slide from the work of a surgeon named Lanny Johnston who performed a number of these types of procedures.  He actually did it by abrasion.

            Here's a patient that in 1980 had essentially no medial joint space.  He did an abrasion arthroplasty, 83 weight bearing radiograph.  They do have some sort of chondral repair.

            And Dr. Johnston harvested a number of these specimens over the years, and here's an example of one where you see the kind of regenerate tissue that formed following abrasion arthroplasty in the human knee.

            And I should have emphasized my task or my invitation was to talk about the human knee.  Soft tissue grafts have been done since the first part of this century, everything from chromic treated bladder to fascia, to periosteum, to perichondrium, and indeed, this is an example of a periosteal graft generating new tissue.

            Decreasing articular contact stress.  I'd like to point out that this is not just a biologic phenomena.  It's a mechanical phenomena, and recreating the right mechanical environment is extremely important, I think most surgeons would agree, and we know from studies of releasing, if you can imagine doing this, actually releasing the muscles that cross an injured joint and stimulating repair of the surface that way, performing osteotomies or, more recently, the concept of actually distracting a joint to stimulate repair.

            And this is just a finite element model that we're using to study regeneration of the articular surface in the ankle joint.  That's the talus, and this is a fractured tibial surface, and we calculate the forces and their concentration in different parts of the joint and then look at how the tissue regenerates in different parts of the joint depending upon the loading of the joint.

            Now, just to give you an example of this kind of phenomena, this is a lady 23 years old, had a distal tibial articular surface fracture; within two years developed severe and disabling osteoarthritis.  This is treatment of that by restoring the alignment of the talus under the tibia, distracting the joint, and here her pre-op film.  Here she is at six months and at two years that joint surface not only holds up,  it actually seems to improve slightly.

            And this phenomenon was first described by surgeons in Utrecht in the Netherlands where they've done a number of these procedures, but I show this just to illustrate the influence of the mechanics of the joint on any of these biologic procedures.

            There are, of course, a number of artificial matrices that are being popularized and used in different parts of the world for stimulating cell in-growth, cell multiplication, cell production of a new matrix, and regeneration of an articular surface.

            Chondrogenesis factors, a variety of them, are certainly already being used in a number of animal studies and in some parts of the world, even in Australia and New Zealand, being used in patients to stimulate repair of an articular cartilage defect.

            This is an illustration from some work that I was involved in in Switzerland in the '90s, looking at resurfacing of chondral defects using chondrogenesis factors.

            You're most interested in cell transplantation, and a dilemma is surgeons look at this, are a lot of choices, and more choices coming along.  Do you use chondrocytes or stem cells?  Do you use autografts or allografts?  Should it be embryonic, fetal, neonatal allografts?  Should it be mature cells?  A tremendous amount of information just here at this meeting in Washington last week actually at the ORS and the Academy about all of these strategies.

            This is, of course, the basic procedure of debriding the defect, covering it usually with a periosteal flap, although many surgeons and centers are using other materials to do this, and then injecting cells under the defect.

            Osteochondral autografts have been used since the 1950s, using patella, proximal fibula, and then more recently the concept of mosaicplasty where grafts are harvested from one part of the joint, transferred to other parts of the joint.  A number of papers coming out, actually transferring grafts from the knee joint to the ankle joint, and a variety of other such choices.

            Allografts also have a role in resurfacing of joints.  This is the shell allograft concept of resurfacing a defect with a small graft or resurfacing a major part of the joint with a large osteochondral graft, often performing an osteotomy.

            So how does a surgeon make sense of all of this?  And when you see a patient in the clinic that presents with a chondral defect, how do you decide which of this wide variety of biologic procedures makes the most sense for that specific patient?

            And obviously that hasn't sorted out, and what I put up here will no doubt be contested by surgeons in different parts of the world, but I've tried to give you a little sense of current practice around the world in terms of how do we handle this.

            And I picked four centimeters.  Some people pick two centimeters.  Some people pick three centimeters.  There are obviously a lot of choices, but I tried to put something out there that we could argue about or discuss.

            But most people would say if there's a small defect, debridement and micro-fracture or some sort of marrow stimulation technique would be reasonable, particularly in patients that have had no previous treatment.  And if they have a larger defect, if they've had previous treatment, if they have significant symptoms, many people would then go to autogenous cells.

            Some, however, would use autografts or allografts, and for very large defects most would use either allografts or a very complex combination of periosteal and cell grafts called a sandwich graft or other similar strategies.  If there's deformity, many would perform an osteotomy.  Not everyone, but certainly if there's malalignment of the knee joint, most would perform an osteotomy.

            So what are the results that surgeons are looking at in trying to decide which of these procedures should be performed on their patients?

            Well, part of what makes the choice difficult is all of the things I've shown you have been reported using the methodology I showed you in that second slide to restore biologic surface, but how do you decide for an individual patient which procedure is best?

            Well, if you look at most of our available literature, there are a whole variety of clinical disorders and patients that are being treated for this.  We have surgeons who are expert in every one of these techniques and tend to be strong advocates for that technique and are very skilled in that technique.

            There are a variety of outcome measures and how do you decide which is most appropriate?  And as you are well aware, there are very few prospective randomized studies.

            So what are the issues in evaluating results in the human knee for biologic resurfacing?  There are a variety of patient factors that will influence outcome.  Certainly defect size and type, severity of symptoms, age of the patient, and then there are a host of confounding variables.

            And I have put patient expectations up here because this particular strategy of resurfacing a joint probably is more dependent on what the patient expects in terms of the outcome than many things we do.  If somebody has a tibia fracture, we know what they expect.  They expect a realigned limb.  They expect a functional limb.  They expect to be pain free, et cetera.

            But what does somebody expect that has a chondral defect?  And I'll discuss that a little bit more in a moment, but then there are a variety of other issues, including activity level, body mass, joint alignment, previous surgery, and how well their muscles function.

            And then how do you look at outcome?  I'll come back to this point, but speaking as a surgeon, I'm most interested in is the patient happy.  Is their function what they want?  Is their joint function what they want?  They and I are perhaps less interested in the structure of their joint and what kind of tissue is replacing the defect.

            Now, I want to spend just a very short time on different types of defects because, as I say, I look at the literature, and I see a lot of mixed approaches to exactly what we're talking about, and I'd like to distinguish the traumatic chondral defects and the osteochondritis dissecans defect from the degenerative defect both in terms of pathogenesis, age of the patient, and the tissues that are involved.

            And this is a very crude, over-simplified analysis of a lot of data on human knee arthroscopy and cataloging what defects we're seeing mostly by a surgeon named Dandy.

            But what it shows you is in terms of the percent of total articular surface defects in the human knee -- and as I said, grossly simplified, if you will -- the so-called osteochondritis dissecans or osteochondral defects are most common in adolescent and young adults.  These chondral defects that I'll show you in a moment tend to be most common in mid-aged and young adults and then degenerative defects increase in frequency with increasing age.

            And this is an example of the typical osteochondral defect in the young adolescent or young adult.  This is an example of a pretty typical femoral condyle condylar defect in a young adult where there's a segment of the -- this is the articular surface.  This is the subchondral bone, and so you see the defect.

            This is, of course, a defect that involves both cartilage and bone, and then, of course, with age, cartilage changes.  It becomes less stiff, more friable.  There are alterations in the matrix, in the cell behavior, all of which change the biology of the joint, and it's for that reason perhaps more than any other that with increasing age we see increasing evidence of joint degeneration, osteoarthritis, and joint dysfunction.

            Now, why does that have any relationship to the issues of cartilage repair?  Well, let me just briefly summarize a lot that's in the literature comparing the two most common biologic approaches for small defects, as I showed you, micro fracture and chondrocyte transplantation.

            Here's a study for all of its imperfections, but it's similar to the data from the a lot of studies, and to be brief, I'm just going to show you one or two that are representative of a lot that's published.

            Looking at marrow stimulation, drilling the articular surface, trying to stimulate a new articular surface, a small series admittedly, but looking at osteochondral lesions in the ankle joint, the patients were under 30, 92 percent good results.  If they're over 50, 20 percent, similar studies on radiographic.  So a dramatic difference in terms of penetrating subchondral bone with age.

            An older study on perichondral arthroplasty showing that although it may work relatively well in people under 30 and perhaps up to 30s or 40s, in people over 40 it just doesn't work very well.

            Chondrocyte transplantation, this is just some data on cell function versus age, chondrocyte function versus age.  You can see there's a slow decay with age, and this is a study presented at the ICRS, and it's in press right now, looking at the baseline synthetic pattern of human chondrocytes with age, but then looking at how they respond to anabolic stimuli, and as you can see, there's a decay with age.

            So those are the issues.  Those are the patient issues.  Now, how do we as a surgeon want to look at the outcome of attempts to biologically resurface joints?

            And as I said, from our perspective, the most important -- my perspective, and I say most surgeons -- is patient function, joint function, joint structure, and chondral tissue, and I'll look briefly at those.

            I think it's worth thinking a little bit about what we mean by patient function.  Certainly if somebody is confined to a wheelchair and we restore to them the ability to walk, however limited, that is a spectacular result for that patient.

            But many of the patients I see coming into our sports clinic don't want to be restored to walking.  They want to go from doing 5K races to doing marathons, and they tell us that they have a little medial knee pain if they train too hard, and so forth.  So there's a range of perspectives in patients who have joint disease in terms of what they expect out of a procedure and what they are going to accept as a good result.

            Joint function.  Many of the studies do relatively little in terms of looking at joint function, although obviously they look at pain, but in terms of range of motion, how well the joint moves, its stability, these things are hard to measure objectively, but certainly once again from the patient's perspective and the surgeon's perspective, they're very important, and I think it's worth pointing out that if you want a spectacular result in a patient who has serious limitation of activity and serious joint pain and dysfunction, then a joint replacement is an excellent, excellent procedure.

            The patient is going to be very happy.  They're pain free.  They have excellent restoration of function.  So I think that helps put it in perspective.

            Joint structure.  Certainly an important underlying principle in terms of joint function and patient satisfaction.  There are a number of issues.  The most important, of course, in terms of biologic resurfacing is the articular surface, but I'd stress you also need to look at the bone, its shape, its density, its relationship to the joint.

            We don't look very often at the synovial membrane, but I look at a lot of these patients who had biologic procedures who are not doing very well, and many of them have synovial thickening and have problems with recurrent effusions, and of course, lastly evidence of osteophytes.

            I'm sure at 3:30 my good friend Charles Peterfy will dazzle you with the MRI imaging of joints, and certainly that's a very exciting technology to look at joint structure.

            Now, what about the chondral tissue?  These are the issues that we're concerned about.  How much of the defect was restored?  What is its integrity?   That is the repair tissue.  Is it fibrillated, soft?  Did it bond to the adjacent tissue?  Is it reorganized along the pattern something like normal articular cartilage?

            What about the matrix?  Well, most people are interested in how much Type II collagen there is.  What are the aggrecan or proteoglycan concentrations?  And then there are a number of other small proteins that have to do with the way the matrix is organized and the way it functions that we're also concerned about in terms of restoration or articular cartilage.

            Now, I show you this slide from a study that we did more than ten years ago on monkeys, on primates, because we thought at that time the best animal model for human osteochondral repair might be one with a knee joint that very closely resembled the human.

            So this is the tissue we find in the osseous and the chondral part of an experimentally created osteochondral defect in the knee of cynomolgus monkeys, and we applied a number of different treatments to try and stimulate repair, and this is based on looking at staining for proteoglycans,  Type II collagens, cell morphology, a number of other things.

            But just to simplify it, if we look at an osteochondral defect in those animals, most of them restore the bone defect very well.  Just like healing a fracture, it is very impressive that at six or eight weeks, and I anticipate the same is true in humans, although we obviously don't have the same kind of data, but they do a very good job of restoring the bony part of the defect.

            It's the chondral part of the same defect that is problematic.  In these fully mature animals, only about two thirds of the volume of the chondral defect was restored, and although a lot of it at six weeks was hyaline in terms of Type II collagen and proteoglycan concentration in chondrocyte morphology and others, still a fair amount of it was fibrocartilage and then even some granulation tissue.  And that's a fairly typical result.

            Well, last and perhaps more important than what I've talked about in terms of other ways of looking at chondral repair tissue is what are the mechanical properties of that tissue because, after all, we want it to perform as a mechanical bearing surface.

            And among those that I think are worthy of consideration are stiffness, permeability, tensile and compressive strength, and durability, and this is, again, just some data from our studies of primates ten or 15 years ago.

            And when we look at excellent osteochondral repair -- so the surface looks good.  Most of the volume of the defect is repaired and the joint looks very healthy -- and we measure the stiffness of normal articular cartilage in the same joint or the opposite joint and the stiffness of the osteochondral repair, it's about one third of the stiffness of normal cartilage to indentation testing.

            When we look at the permeability of repair tissue, excellent in all other respects; morphology, volume, integration, and so forth, but we look at the permeability of it, that is, how easily fluid flows through that repair tissue, it's nearly twice as permeable.  So it's softer.  It's less stiff.  It's less strong, the tensile testing, which I didn't put in, and it's more permeable.

            And that may have something to do with the issue of durability of chondral repair, which is another, if you will, mechanical factor.  These are some slides from Larry Rosenberg's work so many years ago looking at the osteochondral repair at two weeks, a lot of staining for proteoglycan, drill hole made here, articular surfaces restored, looks good.

            At about six months, the proteoglycan staining is lost, but notice that the bone has healed very nicely.  It's the chondral tissue that's problematic, and at six months to a year it becomes fibrillated and degenerated.

            Well, how do you measure durability of chondral repair?  And I put in this slide because it's the longest follow-up I know of any biologic resurfacing, imperfect though it may be, from Alan Gross and his colleagues in Toronto, and what you see using fresh osteochondral grafts, so these are cells and matrix transplanted into humans, and Dr. Gross has been doing this now for more than 30 years, but here's his 1997 follow-up.  A large number of these patients, human knee, tibial plateau.

            At five years, about 90 percent of those grafts have survived and are functioning well.  At ten years, it's about 70 percent.  At 20 years, it's just a little bit less than that.  So that's actually pretty remarkable durability in terms of the biologic resurfacing.

            Now, I notice you put the Knutsen study in these forms in the materials that were sent out, and I think that's very appropriate because there are very few, as I said, randomized controlled studies of any surgical procedure of any significance going back to the Vineburg procedure for revascularizing the heart and the Moseley study for debridement of osteoarthritic knees.

            So this is really a very important advance, and you have the material on that.  Eighty patients 18 to 45, isolated defects, a number of measures of joint and patient function, reasonable look at joint function, and then they also biopsied the tissue.

            Now, I picked just a couple of things out of that study to put in perspective, again, how surgeons want to look and need to look at what we choose for our patients, and I've just taken their data and put it in this form to make some points which you may agree or disagree with.

            This is the pre-op function of their patients on the physical components scale, the SF-36, and as you see, preoperatively these patients were actually doing pretty well, and when you look at other studies of chondral resurfacing or certainly joint replacement or osteotomies, they're doing much less well in terms of this however, again, imperfect it may be, the SF-36 physical component scale.  So these patients were doing pretty well pre-operatively.

            Then you notice that over the two-year follow-up there was some evidence of slight improvement and the improvement in the micro fracture as they reported reached statistical significance, but notice it's a pretty small difference.  And I think when you look at it actually drawn out this way, it becomes more apparent.

            These patients were doing pretty well to start with, and the improvement on the SF-36, perhaps because they started at a high level, was relatively small in terms of clinical improvement.

            On the other hand, the pain scores dropped pretty spectacularly, and they dropped over time, which is what we're seeing in a lot of our attempts to create the right mechanical environment to restore a joint.

            Some of these actually get better over time for reasons that I don't entirely understand, but it's an absolutely fascinating phenomenon.

            So their conclusion was that both groups had significant clinical improvement.  I mentioned the relatively small difference between micro-fracturing and cell transplantation.  Younger, more active patients did better. 

            There were relatively few complications, which, again, I was asked to speak from a surgeon's perspective.  One of the most devastating things that happens to us is to have a patient who is pretty healthy and has essentially moderate limitation of activity have a bad complication.  We don't like any complications.  Obviously that's what keeps us awake at night, but having something bad happen to somebody whose main problem, as you saw in the SF-36, is pretty small is very hard to live with.  So certainly the fact that the complication rate was low is reassuring.

            No significant differences in repair and no association between the histology and the functional outcome, and as I said, to be brief, I wouldn't throw at you the huge amount of data that supports that from osteotomy studies and other studies of chondral repair.  There's not a very clear relationship between the matrix composition and a number of other issues about the tissue and patient function.

            So, again, from the surgeon's perspective, looking at the Knutsen study, which is a huge advance in orthopedic research and particularly clinical thinking, these patients had relatively good preoperative function.  A lot of the patients we see aren't doing that well when they show up in our sports clinic or our knee clinic.

            Because of the nature of a good RCT, you can only look at a limited spectrum of patients.  What surgeons are facing is this whole spectrum of patients, some with huge defects and severe disabilities, some with small defects and minimal disability.  We still don't know what is best for many of those patients.

            And then one of the reasons there are so few studies like the Knutsen study -- and I'm doing two clinical trials right now that are NIH funded, and I can tell you these are problems.  Who's going to pay for surgery, anesthesia, implants, devices, hospitalization?  What if something goes wrong?  Who's going to pay for it?

            I deal with our hospital all the time on these issues trying to do clinical trials.  The skill and experience of the surgeons, surgery is a very technically demanding field of medicine and how well the surgeon does the procedure, and as I said at the outset, how experienced they are may have a lot to do with the outcome.

            And of course, I see what I'm sure many of you see.  Patients come who have already decided what operation they want, and they come for cell transplantation or they come for an autograft, and trying to run a randomized study at least in our clinic, I can tell you we are doing it, but it is very tough because they go on the Internet and they talk to their friends, and they look at the literature, and understandably and rightfully so, they bring a lot of expectations in terms of what they want to have done as an operation, and total joint surgeons can tell you patients come and want a ceramic hip or they want an in-growth or they want a cemented implant.  They don't want whatever you think.  So there's a real issue, I think, increasingly in how do you randomize.

            Last, there is another way that may be developing to look at this that deals with a lot of the issues I've talked about, and this is from the International Cartilage Repair Society, and I thought this would be interesting perhaps for discussion or to think about, but they have developed, and I have a little conflict.  I'm on the board of directors of this organization, nothing more than that, but they have developed a package for looking at knee joint injury, the surgical procedure, and the outcome, and I've just take a very small part of that, and that's the Web site.

            But one of their concepts is that this package would be Web based and surgeons all over the world who commit themselves to this would enter all of their patients into this database, and they would be able to look at a number of the factors that I have been talking about, the range of patients, the skill and experience of the surgeons, the differences in defects and patient expectation, and they have incorporated a number of the scales we're talking about.

            So last, again, I was asked to talk about this from the point of view of a surgeon.  What are we most concerned about?  That our patients are happy, that they're satisfied, that they didn't have a complication.

            And how would you measure that?  Primarily, objectively, by patient function measures, general health measures, certainly the SF-36, particularly the physical component most accepted.

            There are a variety of knee scales, and I saw the materials you sent out.  You did a great job of sending many of those out, and there are differences of opinion about which is best, which is most reliable, which works and which patient population, and I'd have to say that's the choice that investigators need to think about very critically when they think about what it is they're trying to measure.

            But just some of them are the Cincinnati, the IKDC, the Lysholm, the Tegner, the KOOS, all of these different ways of looking at knee function.

            Then I would say, important as it is, from the surgeon's point of view it's less important how their knee looks and what kind of chondral tissue because, as I said, the patient function and how satisfied they are with the result is what is most important to us and, I would suggest, to them, but we'll never make progress in understanding how this repair works or regeneration unless we collect this data.

            So even though from the surgeon's point of view this is what we're most interested in in terms of this field progressing, this is the critical information:  knee structure, restoration, the surface, what kind of chondral tissue, which procedure works best.

            So that's my quick overview, and thanks again very much for inviting me.  This is really an exciting opportunity.

            CHAIRPERSON RAO:  Thank you, Dr. Buckwalter.


            CHAIRPERSON RAO:  Are there any burning questions with regard to either Dr. Moos' presentation or Dr. Buckwalter's presentation from the committee?

            DR. MOOS:  One question that does jump out at me which has to do with the mechanical testing, which is quite striking, and the question I have is at what duration of follow-up and how might you expect those differences to increase or decrease at six months, two years, five years, 20 years, especially in view of that earlier slide that you showed that kind of suggests that after an initial short period of follow-up with I believe it was a micro fracture study or -- I'm sorry -- it was the 30-year follow-up with the osteochondral graft.  The patients appeared to stabilize. 

            So maybe there's a maturation process and some of those other characteristics might improve.

            DR. BUCKWALTER:  I think that's really an important question, and there is some interesting data from procedures, one which was called a cup arthroplasty, where the surface was abraded and an implant was put in.

            But some of these patients ten years, 20 years, 30 years actually seem to get better, and you know, I don't know whether it's, playing off your point, that in some patients, and perhaps it's the right genetics, the right mechanical environment, they actually seem to improve over time.

            And as I said, I'm fascinated as we do these distraction arthroplasties to see that some of these patients keep getting better; that whatever repair tissue is regenerated, it actually does seem over time to mature, and it's fascinating that it -- and this doesn't happen very often.  I'm not purporting that this is anything like 60 or 80 percent.  It's more like 20, 30, 40 percent, but why do some patients over time actually seem to get better and have a joint that works better?

            And, incidentally, you also see in some patients with severe osteoarthritis, a spontaneous repair and regeneration of the joint, and that has been written up a couple of times.

            So there are these repair or regeneration phenomena that occur in some people, where that regenerate tissue matures and improves and functions better over time.  If we understood that, we'd be along -- you know, what is that?  Mechanical environment, use?  I'm not sure.

            DR. MOOS:  So that begs the second question that if you are comparing one procedure versus another, for example micro fracture versus autologous cells, the follow-up period at which you do it may have a profound effect or maybe a less profound effect, but a measurable one on your p-value.

            DR. BUCKWALTER:  I could very well.  As I said at the outset, when we follow out a lot of orthopedic procedures ten years, 20 years, we just completed a 50-year follow-up of scoliosis patients, and you really do learn a lot.

            But, of course, in terms of making any decision two years is a long time, and I guess I would say that in two years you have a pretty good idea at least in terms of biologic resurfacing what the potential is for the future.

            So, yeah, what you said is right, and we're a hot bed of 50-year follow-ups and 30-year follow-ups where I come from probably because people from Iowa never leave or something.  I'm not sure what, or I don't know why, but, yeah, I mean, I really believe in that.

            But I also believe two years is pretty long for most orthopedic procedures.

            DR. HIGH:  Can I ask?  You said in closing that for the field to move forward it was important to acquire the histologic and radiographic follow-up data, but in the Knutsen study, there seemed to be no correlation between the histologic data and the clinical outcome.

            DR. BUCKWALTER:  I didn't say histologic.  I'm sorry.

            DR. HIGH:  Oh, you said radiographic and ‑-

            DR. BUCKWALTER:   I meant -- maybe I misled you -- I meant restoration of chondral integrity and structure because partly to respond to this question.  If we're going to follow them just for two years, I'm reasonably confident even in those patients who might be doing pretty well on the SF-36 and an E score, if there's no repair tissue there and there's still a mechanical defect and problems, they will be the patients who deteriorate at five years and six years and eight years, and I should have said that.

            But histologically, I'm not sure.  I think there are some real issues, you know, at least personally about going back and opening up the knee again and biopsying somebody.  I don't know.  Again, I'm sort of looking at myself as the patient's doctor and surgeon, and is that something I would feel really comfortable with.

            DR. HIGH:  So then just to be clear, the data that you think it is important to collect then are the radiographic data and what else?

            DR. BUCKWALTER:  Well, patient function.

            DR. HIGH:  Right.

            DR. BUCKWALTER:  Knee joint function, knee structure, and chondral repair structure or volume or, I hope, and I guess as I said, I don't want -- Dr. Peterfy is probably going to talk about all of this -- noninvasive ways of actually looking at the -- well, I don't want to steal his fire, but amazing studies that lead me to think that high resolution imaging may answer that question without doing tissue biopsies.

            CHAIRPERSON RAO:  Thanks, Kathy.

            I had two questions as well.  One was:  is there consensus in the field that there's really no clear-cut predictive measure in terms of either selecting a patient or saying that this is what would be effective or beneficial therapy in this particular class of patients?

            DR. BUCKWALTER:  Well, I tried in that one slide to say there is, I would say, an emerging sense of which procedures work best for which patients, and that's often the way knowledge advances in medicine unfortunately is experienced bills, and you say for this group of patients, debridement, micro-fracture might be a reasonable first thing to do.  For this group of patients cell transplantation, et cetera, is appropriate.  Previous surgery, failed micro-fracture, and so on, and then for the patient with malalignment or a total tibial plateau or condylar defect, osteotomy and allograft might be reasonable choices.

            CHAIRPERSON RAO:  And is it a fair statement that the consensus in the field is it is very hard to do comparative studies of one procedure versus the other?

            DR. BUCKWALTER:  Well, I think what I was trying to say is in doing a randomized controlled trial, important as it is, and as I said, I'm doing it.  I think they are critically important.  I think it's also important to understand that they are expensive.  It's sometimes a problem in terms of getting patients to accept randomization, at least in my experience, and it's a challenge to do a randomized controlled trial, and certainly on a large scale it's a real challenge, and I guess I was trying to suggest maybe a little bit sub rosa that maybe there are other ways to collect data that may be helpful, not as good as an RCT, but given the time constraints and the cost constraints and the patient preference and surgeon preference, maybe there's other data that would be useful to have.

            CHAIRPERSON RAO:  And would the other surgeons on the committee tend to agree with that general statement?

            DR. COUTTS:  I always agree with Jody Buckwalter, but I think that if you're going to have a control group in a study, for a surgical procedure, it has to be another procedure.  In other words, sham operations are extremely difficult to enroll patients into when they have significant clinical deficits.

            And no treatment is another possibility that never seems to be entertained, but when you have progressive disease, it's very hard for patients to agree that they will go into an arm in which they will receive no treatment.  So I think your best chances are if you're going to have a control group, that it be an alternative treatment.

            CHAIRPERSON RAO:  Are there any questions for Dr. Moos, as well, from his earlier presentation?

            (No response.)

            CHAIRPERSON RAO:  So my sense from what Dr. Moos said was that it's clear that if you regulate cells as transplant, that each cell is a different type of cell, and that there are going to be issues of combining cells with devices, and so there will be, as the FDA seems to have already realized, that there's going to be issues with how they're delivered and what device and the different aspects that the agency might have to regulate this sort of process.

            And it seemed like this was a relatively new aim for the FDA itself in terms of this regulation, and the FDA was still trying to collect information.  Is that a fair assessment of what the FDA tried to put forth as its sets of questions?

            DR. MOOS:  More or less.  I think that there are, in particular, specific issues of the kinds of data in this field that can help us address issues that we've heard about already, and also there are some specific issues with the design and selection of appropriate preclinical data that we're going to hear a great deal more about in a little bit, and general issues that the whole field of cell therapy is dealing with that have to do with the best ways to characterize the products in the light of the kind of analytical techniques that are available now, and we'll discuss that in quite a bit more detail, but in broad essence, yes.

            CHAIRPERSON RAO:  If there are no more questions, then I'll ask the next speaker, Dr. Frank Luyten, to come up and tell us about the biology of joint surfaces.

            DR. LUYTEN:  Thank you.

            I would like to express my thanks to the FDA to give me the chance to come over here and present some of the work I have done and we have done the last ten, 15 years in order to try to move some of these approaches from bench to bedside.

            I'm a rheumatologist.  I've been involved in many clinical trials over the years, randomized trials with drug treatments, and although it is very difficult to run randomized trials, I profoundly believe it is possible to do this if both the physician and the patient are well informed of how this process is going on and are recruited properly, and there is obviously still a long way to go.

            As an introduction, I think it is absolutely essential that we are working towards well designed, prospective, clinical studies, trying to prove that there is a therapeutic benefit of our treatment that outweighs the risks and the potential side effects.  That has to be translated in outcomes, therapeutic benefit, and one can at least use some of the experience we have built up over the years in other joint diseases, such as rheumatoid arthritis and osteoarthritis.

            And we have learned from these diseases and from these clinical trials that you really need to aim for a combination of factors and outcomes.  There needs to be symptom relief.  There needs to be improvement in function, but also we need to be structure modifying.

            In rheumatoid arthritis, we have clearly learned this, that if you are not looking at structure modification, you will miss the boat, and you will get treatments to patients that may have them in terms of symptoms, but not at all in progression of disease and in ultimate outcome of the patient over the years.

            So there are, indeed, a number of important challenges for these clinical trials.  One of the big challenges in this field for this patient population, what are the kind of outcome scores that are sensitive to change.

            We've seen the SF-36 is obviously not a very good one because they start high and end up high.  So they are very poor in terms of sensitive to change.  So we have to, through our clinical experience and our clinical trials, try to design outcome scores that are sensitive to change in this patient population.

            We have then to define and to agree with each other.  What are the kind of improvements we would like to see that we feel and that the patients feel are clinically relevant?

            It is clear that in these applications at least in joint resurfacing protocols also, it is appropriate and important to compare this with the existing treatments, standard treatments, keeping in mind, in fact, and that's probably one thing which puts this field a little bit ahead of the disease fields.  We are talking about joint disorders in most cases, not joint diseases.  In principle, we are not entering yet the degenerative process.  We are trying to repair the joint surface defect in an otherwise normal joint, that potentially we may end up with what you call a large number of remissions, something that in rheumatoid arthritis we are much less successful to do where the remissions are in the order of ten percent or 15 percent.

            In this patient population, when we do a good job or when we have a successful procedure, we may end up with a large number of remissions, and detecting the outcomes that are identifying these  remissions is obviously one goal of developing the clinical trials.

            And obviously we would like to have long-term follow-up, and in order to be able to do this, identifying surrogate points, surrogate clinical points or structural points that may predict the long-term outcome is obviously something which is of great importance just like bone density has been used in the field of osteoporosis for the prevention of fractures with certain treatments.

            Although we've learned from that field that bone density by itself is not sufficient, required but not sufficient to reduce the fractures.

            So knowing that clinical trials are important in this field, regardless of what we think and how difficult it might be, we have to do it, and in order to do it, we are concerned about the sources of variability in our clinical trials. 

            What are the most important sources of variability?  Well, first, the patient population.  As you already highlighted, it's not so easy.  I will discuss this very briefly. 

            The surgical procedure, the postoperative rehabilitation protocols in my view are extremely important, protection of the graft and enhancement of the repair of the tissue and integration in the environment can be majorly influenced by rehabilitation protocols, and then the cellular products.

            What would you say if I tell you that I'm running a clinical trial in which I'm looking at the effect on biphosphonates on the prevention of fracture, but using five or six different biphosphonates in the study, but they would say you are crazy because you have five or six different products entering this.  They have a sort of common mechanism of action, but you are increasing tremendously the variability in your potential outcome.

            Indeed, with cellular products we have to make sure that we make our product as homogeneous or as well characterized as possible.

            Patient population variability, well, you would say you can do this by inclusion and exclusion criteria, as we do normally in clinical trials.  Well, there are still a number of variables in there which we are not going to control that well, and we may think about stratification of the patient populations.

            Symptomatic joint surface defects are in some cases deep cartilage defects, not underlying much of the subchondral bone.  In some other cases, there is clearly subchondral bone involvement.

            The duration of the disorder.  We know from RA that there is a big difference between early disease and established disease, but in this field when you are recruiting patients in a trial for symptomatic cartridge defects there are early patients in there which have had the recent accidents and up to a year ago no problems.  The accident, the cartilage damage, and they want to get -- but there's quite a group of more chronic patients that has established symptoms for five to seven years, have adjusted their life style to this new reality and are, therefore, measuring outcomes in that patient population may actually give you some different results as in the more early population.

            And then the etiology of these symptomatic defects will be different.  Some of them will be clearly traumatic, post traumatic by anamnesis, osteochondritis dissecans, and a number of them will be unknown.  You know, whatever you ask the patient how hard you try to insist there has never been a trauma in the past, in a number of young patients we cannot identify a precise moment where there was a trauma potentially leading to the cartilage defect.

            Surgical procedure, we need to try to standardize the surgeons and the procedure, training of the surgeons on animal models.  It is good in our experience to have a coordinating surgeon.  I call this a traveling surgeon who particularly in the first application of patients.  Trying to bridge the learning gap in these patients is very important, and they have to be -- the centers have to be dedicated to a large extent.  They have to be willing to go the way to do the effort to do this properly.

            Postoperative rehabilitation.  Well, there need to be at least standard operating procedures for the rehabilitation in our clinical trials.  The physiotherapists need to be trained.  Patients need to be educated.  Very important.  Tell them what is going to be required from them in order to get this procedure to be successful.

            A coordinating physiotherapist, good clinical practice, these are all important things.  There is one big limitation in the field of rehabilitation.  It is not because it sounds logical and that it's based on experience that we have not much signs actually behind the protocols and how they would involve and imply the outcome of the clinical trials.

            One thing is important.  What I'm going to discuss today is the characterization of the cellular products.  It is important we try to develop a strategy to insure product consistency and reproducibility.  I'm thinking and acting as a biologist from that perspective, having many years  in the sciences as a biologist.

            There are a number of challenges, and I will highlight four of these and give some examples which might identify and help you to orient yourself into this quite complex field.

            The joint surface is complex.  We are not just restoring one type of tissue.  We know already developmentally that joint formation is a complex process which very early on starts in development which distinguishes itself from the development of long bones and the endochondral bone formation processes and has its own signaling pathways guiding all of this.

            But post-natally, we are actually talking about the multilayer tissue, a superficial joint surface, skin which based on recent data or data from other groups probably contains a lot of our progenitor cells.

            The middle and deep layer, some of the aspects of these layers are reminiscent of the growth plate, and maybe some remnants of the growth plate, but is not replaced by bone.  This tissue is stable, stays cartilage throughout life.

            There is a tide mark in calcified cartilage.  We know the tide mark moves.  It's active.  It's not a dead place that stays there for the rest of our lives, and the subchondral bone plate is in its biology, but also at the molecular level directly involved in what the articulate surface above it is doing.

            So we have a quite complex tissue that interacts with each other at the different layers, and trying to restore this obviously probably is very, very difficult.  So we would like to put a repaired tissue in place that can at least take over most of the functions of the tissue that nature has created during development.

            I like to call the defects that are treated today in most cases the co-called superficial osteochondral defects.  The cartilage is gone, and here is a picture of it during surgery.  You see that there are very small places of blood coming through the calcified cartilage.  So there is always contacts in these patients with the subchondral bone.

            So the subchondral bone will directly influence the repair process.  Indeed, this is the challenge number two.  We need to integrate in a very complex environment.  This is a schedule, a sort of schematic representation of the environment.  We have the synovial joint fluid.  We have the neighboring articular cartilage, which unfortunately is not going to send many signals to the repaired tissue because it's far away from there and many of these things are entrapped in the matrix.

            We will probably get a lot of signals from the subchondral bone also, as there are remnants of the calcified cartilage with direct contact points with the subchondral bone.  There will be a covering plate in some cases, which will be, for instance, a periosteal flap or a membrane.  They will also signal to the tissue that we put in place there, and then the cell products which will be a cell suspension or a combination of several things.

            Well, we are implanting this tissue in an environment which is certainly not an autotopic one and could be considered as a heterotopic one.  That clearly gets a number of signals that they are not used to see probably in normal conditions during normal developmental processes.

            Thirdly, and this is already some data that Joe Buckwalter showed, and these are experiments I did with Joe at the NIH in 1989, together with Hari Reddi and Alan Lanzo, where we started to realize how important it is that the repaired tissue probably needs to be a stable one. 

            It is the so-called osteochondral defects in the rabbit.  This model is very well known to be a quite successful one for the spontaneous repair of the osteochondral defects, and indeed, three weeks after surgery, you see a tremendous influx from the subchondral bone layer with the fibrin clot starting to repair this tissue, with a quite impressive repaired tissue at three months, which in many characteristics has all the, I would say qualities, of the articular cartilage, the metachromatic matrix Type II collagen, even some form of columnar organization.

            So this is it.  I thought at that point we had the treatment.  You make an osteochondral defect.  In fact, at that time we filled it with a collagen matrix containing some BMPs, and we were very excited about it as a result.

            Even looking at the integration of this newly formed tissue into the neighboring tissue indicated by these arrows, everything was going great.

            Well, at nine months after surgery, we started to see the important problems coming up with sort of the release or cracks and fissures coming into the cartilage at the sides where there was integration of the new tissue and the old tissue.  And after 12 months after surgery, well, this was over.

            Well, we had this perfect repair tissue with many characteristics which were very exciting.  However, and in fact, independent of the treatment because we had control groups; we had collagen matrix groups and collagen matrix groups would be in p's; we had in many cases quite spontaneous successful repair which was however short lasting in these experiments.

            So dural repair leading to stable cartilage is something which is important.

            Challenge four, we need to have a sufficient cell number.  Indeed, we are talking biology here.  You don't have stray dose responses like with drugs.  You need to reach a certain critical threshold in order to have sufficient cells contributing to the repair process, whatever the mechanism of action is of  your tissue repair.  This is not only true for cartilage.  I feel it's true for bone.  I feel it's true for the repair of other tissues.

            We can only start from a minimal biopsy material.  We cannot peel off three grams of cartilage from the articular surface in order to generate some 20 million cells to fill one defect, which is only one-fourth of the joint surface.  So we need to be very concerned about the preservation of critical biological characteristics of the cells.

            But at the same time we need these cells to have some plasticity, some remodeling capacity which are required for solid engraftment and for integration into the environment.

            So many challenges coming up here, and I think it's truly the biologists that are, in the first place, the ones that can start to solve some of these.

            So thinking about all of these aspects, we defined that a cellular product that would probably be a reasonable one is a product that contains or that is capable of organizing a fairly young tissue.  It has to have a number of hyaline characteristics, hyaline-like characteristics in vivo preferably because we want it to be stable, and stable means that it is resisted or resistent to vascular invasion, to mineralization to replacement by bona fide tissue.  It has to be a tissue that once it has been made, it can maintain its homeostasis for a long, long time, and this is the reason why we believe that in vitro assays would probably not sufficiently address this issue of biology that we need to have successful repair in a cartilage defect.

            And, therefore, we define an in vivo assay, and this is a little bit of an in vivo assay which I would say comes out of the BMP field, working with BMP implantation ectopic bone formation.

            So ultimately I felt when we inject or implant our so-called cartilage implant in an ectopic site in muscle where there is massive vascularization and inflammation when you do this, if this tissue can establish itself, it has a cartilage tissue with a number of characteristics that are reminiscent of cartilage, including the hyaline-like nature, the resistance to vascular invasion, the stability during a long time.  Then probably we have an assay that would better reflect the biological characteristics we would like to have our implant to have in our patients.

            Stable means, indeed the hyaline-like cartilage matrix resistant to vascular invasion.  This is an enlargement of the tissue in which no single blood vessel can be seen into these implants.  We looked at molecular markers.  That's Flic1 for endothelial cells.  There is not one single endothelial cell entering this implant.

            You can stand at the right-hand side for Type II collagen.  It is very enriched in Type II collagen.  It is a very sensitive and specific assay.  Articular chondrocytes will do this for you, injected ectopically.  Epiphyseal chondrocytes, the ones that we call developmentally transient chondrocytes, ultimately going to be replaced by bone are, indeed, displaying this developmental cascade.  Actually the initial organization of an impressive amount of cartilage within ten days we see a massive invasion of blood vessels that you see on the left-hand side that are associated with precursor cells that turn into osteoblasts and the formation of mineralized tissue, the total implant turning into bone in about six to seven weeks.

            At the right-hand side, we use periosteal cells that we induce to be chondrogenic, implanted them in vivo, and essentially only found fibrocartilaginous fibrous tissue after three weeks.

            So this assay was quite specific for articular chondrocytes and was at least recognizing a number of biological functions that we cannot really measure in vitro.

            The only tissue source which for us has been successful so far to generate this type of tissue is, indeed, the biopsy of the articular cartilage, the cells derived from the biopsy.  Indeed, this tissue is phenotypically committed.  The cells seem to understand and have the molecular imprint to reproduce and make the tissue we want to have.

            Obviously, we don't like the fact that we need to take that biopsy, and we know for every cell, but certainly for chondrocytes that the differentiation loss of these characteristics is going to be the price we pay when we try to expand these cells in in vitro settings, which means an environment where the cells do not really recognize what is happening.

            And, indeed, here it is showed on this slide, and we also had it published already, that after two population doublings, when injected in vivo in the mouse assay, we still have the organization of a nice piece of cartilage.  However, after four population doublings, it was fibrocartilaginous tissue and blood vessel invasion occurred.

            So very quickly, to my surprise -- I would never have predicted it -- we already lost important in vivo biological characteristics.

            Obviously you cannot inject for every patient from every sample new mice to test for the cartilage formation activity, and therefore, we started to work on what I called the molecular profile of the stable chondrocyte.  We looked for a number of known markers and unknown markers.

            In this case you see very nicely after two population doublings, when we still have this cartilage forming activity in vivo FGFR3 called two-way and BMP2 are very strong, while Alk1, BMP Type 1 receptor is weak.

            As soon as we entered the loss of this in vivo biological activity, you'll get a steep decline of these markers and an uprising of what we call negative markers.  This is reflected in patient populations and becomes even more prominent in the oldest patient group, like in this 86 year old patient.  This is all patient material.  These are all human chondrocytes.

            Where they switch from the positive molecular markers and the good biological behavior into the negative one is a very quick and steep one, as shown in this second example.

            Once we've understood that, we've opened up what I call all of the modern molecular biology approaches, trying to distinguish the good from the bad cells, and, in fact, most of our data were generated with micro array analysis, and we discovered if we establish the molecular imprint of what we call the stable chondrocyte with positive indicators and negative indicators.

            The most surprising thing to me was that there were quite some genes and signaling pathways that I have never seen in the context of cartilage biology before, telling me that, in fact, we probably understand only 50 percent of cartilage biology.  Having been in the field for 20 years and Joe Buckwalter for 50 years, there were still so many things that we -- sorry to say --


            DR. LUYTEN:  There were still so -- well, you started there, ten you told me.

            There were still many things we didn't know of.  Interestingly, we also investigated the relevance of these markers that we identified, and studied them both in mice by using genetic models, in postnatal models of arthritic disease and doing comparative biology from mice to humans, which means when we have identified the signaling pathways, can we establish in all the models that, indeed, these pathways are relevant to the cartilage biology we are discovering here?

            One of the signaling pathways we discovered as being very practical was BMP signaling, especially BMP2 signaling to the Alk3 receptor, and in fact, today there are quite some solid data that are I would say not really validating, but are corroborating these findings that these are signaling pathways that are very critical to the stability of the tissue.

            David Kingsley, my friend in Stanford, just published shortly, a couple of months ago in PLoS Biology that the tissue-specific no cult (phonetic) of the BMP Type 1 receptor or 3 leads to postnatal loss of articular cartilage and the development of an OA-like joint disease.

            So he was using -- this is the Cre/lox system with the Gdf5 enhancer and promoter here regions, which is a market specifically linked and associated with the joint interzone.  When you take out your Alk3 receptor from these cells, you have a normal development that puts our bone with the normal articular surface, but as they further grow, the articular cartilage complete decomposes, loses its homeostasis, and ends up in an arthritic-like joint.

            We've shown, and we have these data now submitted for publication, that indeed, if BMP signaling is enhanced or you take part of it out of the BMP signaling out of the joint surface, that the response to injury by whatever means in arthritic models is clearly much more enhanced and much more dramatic when you have no BMP signaling.

            In fact, half of the reduction of BMP signaling are much stronger when you have more BMP signaling.  So in postnatal settings, regardless of the origin of the engineering in these arthritic models, BMP signaling is critical in the maintenance of the homeostasis of the joints.

            And as some of you might know, and some of these data have been published, both BMP2 and 5 have been associated with susceptibility to arthritis in humans, both in HART and knee of arthritis.

            Well, once we had and we continued to validate the other markers, and in fact, it is surprising how many of these are, indeed, becoming very critical in the maintenance of the homeostasis of the joints.  Now, once we wanted to further validate these markers -- let me go back to the former slide.  Sorry.  This is the previous on.  No.  Can I go back there?  This is it.  Yeah, I have it.  That's okay.

            We wanted to further validate a number of markers in a large group of patients and then, secondly, develop and optimize a culture method so that at least we obtained reproducible cell populations in the majority of our patients.  This goes spontaneously.

            So we basically then came down or boiled down to six markers.  We call that four positive markers and two negative markers.  When we have a strong signal for the positive markers and no, small and limited signal for the negative markers, we have a very good cartilage-forming activity in the assay.  When the positive markers decline, the negative markers still stay quiet.  We still have some cartilage activity.  When the reverse happens, your cartilage formation is gone.  You can do it semi-quantitatively.  We have now quantitated this by Taqman analysis.

            And here is the relation between the gene score.  We have a combined score or composite score of six genes and stable cartilage formation.  I think it is very clear.  Then when we have this grow spontaneously, this shifts spontaneously now.

            Yeah, I'm back.  So when we have a positive score, as you can see on the score, this is the green box.  Then we have a very good formation of cartilage in vivo.  When we have the negative scores, we completely lost the biologic activity.

            This implicated or implied that in the manufacturing process of the cellular product, one has to test all the variables versus this biological behavior, both the tissue source, the enzymatic release, the seeding density, the culture media, the harvesting of the procedure, the passaging, even the storage, and the shipment processes have to be validated every time because every step of the procedure will influence and can influence the stability of your chondrocytes.

            For instance, with culture media we have serum-free chemically defined medium which we can easily grow to cells.  When we compare this with the fetal bovine serum containing media, we clearly found a different gene score, resulting in very different biology.

            In this case, on the left-hand side the fetal bovine serum expanded cells and injected in the nude mouse model.  After two weeks at the left, you can see beautiful formation of cartilage which after seven weeks clearly progresses through maturation.  The serum-free expanded cells, being very healthy and viable made some form of early cartilaginous tissue, but collapsed on us after seven weeks in this ectopic model.

            So the gene score being differently, gene expression patterns being differently, different biological behavior.

            Well, obviously, this is an assay which tells you something about protecting and preservation of biological characteristics.  How does that behave now in the joint environment?  Because ectopically in the muscle may not be the same as in the joint environment.

            So we studied cell engraftment and the contribution of the cells to the repair process by labeling the cells, and we used a large number of goat  of cartilage defects. 

            Well, we had to work hard on the labeling because we found out that labeling cell procedures does affect biological activity of the chondrocytes.  You cannot label in any way.  In this case we used the MANDRA label and the  procedure.  We did not affect the biological characteristics.  The red dots are the cells we injected, and as you can see after three weeks in the cite of implantation the cells settle on the bottom, maybe grow partially and attach to the periosteal flap, but are negative to Type II collagen.  Panel C and D staining for Type II collagen, the neighboring cartilage very nicely positive, the implanted cells completely negative.

            So at this point in the game, the only thing we could say is our cells stays there.  The party engrafter, but somehow they may not yet contribute to the repair process.

            Well, we were glad to see that after ten weeks with the repair tissue that organized, which really looks like young hyaline-like cartilage and staining for Type II collagen, we already detected the cartilage-like characteristics.  Our implanted cells are the red dots.  The collagen Type II positive cells are the green dots when you superimpose the images.  A lot of our implanted cells did actually start to express Type II collagen, establishing that the cells seemed to contribute to the repair tissue and to the stable cartilage repair tissue in this animal.

            At the other site we found differences across species.  Indeed, the goat versus the human chondrocytes, you know, we became experts in goat's cartilage biology, and it's not with pleasure I did all this work.  The in vivo ectopic assay is preserved across species, yes.  We looked at bovine, porcine growth and more.

            But the behavior of the 1,000 expansion is different.  More passaging is allowed in goat before they lose the activity.  The nude mice we could even go up to Passage 5 in fetal bovine serum conditions.

            More strikingly, the molecular markers, although some were conserved, not all were conserved across species.  For instance, FGFR3, a critical marker in our hands, increases its expression level during passaging goat, while in humans we lose it very, very quickly.

            So going across species can give you some information on how the cells behave in the biological context, but we have to realize that the cartilage biology we see there may not be the same as we've seen in humans.

            Every time you shift the model, you have to answer the questions about basic cartilage biology in this model, and although some general principals may apply, many more detailed characteristics will clearly be different, certainly if we start to work with gene patterns.  This use saw that.

            Well, we have injected ectopically cartilage from ankle, hip and knee, and we did see formation of ectopic cartilage.  There were no differences between the three joint sizes.  However, we know and we saw also difference in metabolic activity.  This was already previously mentioned by Klauss Kuettner's group.  For instance, in ankle versus knee, it's not the same metabolic levels.

            More strikingly, that when we went into other sources, we don't like to take the biopsies from cartilage.  Let's take one from the synovium.  We saw important differences between in vitro and in vivo.

            Synovial derived MSCs can be easily established in vitro to become a chondrocyte by growing under certain conditions an advantage of beta, eventually in combination with BMP2.

            So they look in vitro like a chondrocyte, feel like a chondrocyte.  They make highly sulfated proteoglycans and full of Type II collagen 2.  So that says we don't need this cartilage biopsy anymore.  While injecting these cells in nude white mouse, very surprising what we found.

            Panel A, ACDCs, articular chondrocytes by injection ectopically in the mice, as you can see here, they make nice Type II collagen with the implant that we are used to seeing.  When we injected these cartilage cells, they are from synovial membrane.  Well, they were starting to make muscle.  Indeed, the myosin heavy chain was appearing, and we don't see Type II collagen.

            So as soon as these cells were put in another micro environment, they lost their phenotype.  We were not capable of locking them in in the stable phenotype.  So I thought it's because we are putting these as single cells we will implant the cartilage implant, the micromass.  It's a nice, tightly organized structure.  It has matrix.  As you see in Panel D, this is how it grows in vitro, nicely during many weeks.

            However, when we implant this in vivo, we see that we progressively lose this biological activity and that it is identifiable by additional histological stainings that this tissue is growing full of blood vessels and gets invaded.

            When we stain for apoptosis, the in vivo ectopic implant undergoes massive apoptosis, while in vitro everything goes well.

            Looking at product characterization, when we have now these cells that do well, FACS analysis and cell membrane markers, how do we all look based on our classical markers, the CD44, CD90, CD105 and so on?  While it is clear, indeed, whatever expansion protocol you use with important difference between the different protocols, we are clearly enriching in the precursor phenotype, and that has potential functional implications.

            Indeed, by using a model most of the regeneration injecting cardiotoxin in the muscle, where there is massive muscle damage and where the environment screams for "help me to repair the muscle tissue here," when we were injecting our two passage cells, so the chondrocytes that were minimally passage and abnormally make this beautiful cartilage in vivo, they still do it. 

            As you can see here, two populations, doublings I, this is in the tibialis anterior model of cardiotoxin.  They make cartilage, but at the same time, part of the cell population turns into muscle.  So there is a subgroup of cells clearly as we identify by FACS that starts to have precursor characteristics that when put in an environment where the environment screams for help, we see a change into the muscle phenotype.

            While at 12 population doublings, we completely lose obviously the cartilage forming activity, and it's mainly basically the formation of myosin heavy chain that appears.

            Is this only the result of the expansion?  No.  Freshly prepared populations of how the actual cartilage in the normal muscle makes the cartilage tissue with Type II collagen, in the damaged cartilage muscle environment, the cardiotoxin environment.  As you can see here, heavy myosin heavy chain is appearing in this model.  So at least some groups in the tissue is a precursor cell population which will listen to the environment and do what the environment tells them to do.

            So we believe the product characterization is a complex one where potency should probably be addressed very soundly and where we would like to use at least what we feel are relevant bioassays, potentially replaced by marker profiles.

            And obviously the product characterization is only part of a large process in which many aspects of generating the solid implant is going to be critical.

            And I thank you for your attention.


            CHAIRPERSON RAO:  Thank you, Dr. Luyten.

            We have a couple of moments we can take questions.  Go ahead.

            DR. COUTTS:  I have two questions for you.  How do you know what is the correct number of cells for any type of an implant?

            DR. LUYTEN:  Well, basically the way we determine how this might be useful for this application is, in the first place, the calculation of the cell density per volume tissue in the normal healthy tissue of that patient.  That's what we use in the first place as a sort of standard.  These are the kind of cells we would like to use.

            Then secondly, in our ectopic assay, we reduce the number of cells and determined at which level are we starting to lose.  What is the critical threshold at which we start to lose this cartilage forming activity, and that's where we discovered that adding the periosteal cell population increases quite impressively or reduces quite impressively the threshold.

            So you can reduce your number of expanded chondrocytes when you combine them with the periosteal material than when you inject them alone, and the combination of cell density that we would like to have in the ultimately repaired tissue, together with the critical biological threshold in our assay, established the order of magnitude of cells we would like to have.

            DR. COUTTS:  Do you not assume that cells will replicate once they've been implanted?

            DR. LUYTEN:  Well, we did some BrdU stainings, and there is limited cell replication going on.  We have not really quantified it sufficiently to see how much is really in these animal models.

            DR. COUTTS:  And then my second question is with regards to your assay, have you verified that a cell that performs well in your assay then performs well in a cartilage repair model?

            DR. LUYTEN:  Right, yes.

            DR. COUTTS:  And vice versa.

            DR. LUYTEN:  Well, that's doing well.  The opposite is much more difficult, and the reason why, as I just indicated to you, we wanted to replicate this experiment in a goat environment, and to our big surprise, first, the molecular profile didn't fit the human product, and then secondly, the behavior of the goat cells in the ectopic assay was quite different in terms of when you were losing that phenotype from the human phenotype, and that's a big issue.

            Secondly, when you try to still address this issue and push yourselves very hard in the goat to many cell divisions, like up to 20 cell divisions, the variability in the outcome of the animal model is large.  So we calculated the number of animals we would need in order to come up with a statistically significant outcome, and we came up with like 45 animals versus 45 animals and felt at that point that this was maybe going one step too far to ultimately define this issue and see how important this was.

            CHAIRPERSON RAO:  Go ahead.

            DR. HARLAN:  I first want to thank both you and Dr. Buckwalter for what I thought were both very balanced and open presentations. 

            With regard to your ectopic assay and, in particular, the cells that seem to have turned into chondrocytes in vitro but then turned to muscle when you injected them in muscle, have you tried injecting -- what if there are, and undoubtedly there are, local cell mediated or kind of paracrine cell signals, where if those cells were injected into a joint space, they would stay more chondrocyte-like.  Have you considered that possibility?

            DR. LUYTEN:  We injected some of that into the joint space, and they turned into bone.  You could have predicted that one.  So it just makes the point that these cells are still susceptible, more susceptible to environmental signaling than the well established chondrocytes which has that molecular imprint which seems to be imbedded in these cells.

            This is why I strongly believe you need the critical mass of these molecular imprint cells that don't need to be all of them, a critical mass to make sure that that message goes out to the repair tissue.

            CHAIRPERSON RAO:  Dr. Tuan.

            DR. TUAN:   Yes.  Frank, that was a very interesting and informative presentation.  I have a question concerning the fact that you saw these precursor cells going up or a percentage of precursor cells going up as a function of passage.

            So as you know, a number of workers have shown that there are multipotent cells even straight from the initial isolate.  So at what point or what criteria will you use to, I guess, prepare these articular cartilage derived cells to say the number of precursor cells are not going to be functionally important?  I mean, is there -- I mean, I know you look at this gene profile.

            DR. LUYTEN:  We look at the ectopic assay, and so what we did is we mixed -- you have your starting cell population, and then you add to this increasing amounts of what is considered as being precursor cells:  CD44, CD90, CD105, and these things.

            So when you increase that number of that, you very quickly start to overrun this in vivo cartilage formation assay.

            DR. TUAN:  But the relationship between the ectopic assay and the functionality in vivo, as Dick Coutts was talking about, that's not 100 percent there yet, right?

            DR. LUYTEN:  No, but in the joint surface repair model, we have not shown that the cells that have fully lost this activity are performing less in the goat model, and that is a little bit of a problem in the goat model since the cartilage biology in the goat model lists are more different when we go to what is I would say these more sophisticated labs.

            What we have done is we have used the synovial direct chondrocytes.  We have implanted these in the goat model and the results there are devastating.  I mean we have 20 animals which are absolutely a disaster with only fibrous tissue.

            What we have not done is the chondrocytes that ultimately lost the phenotype and the critical threshold for this changes.  Since this does not seem to be concerted or conspicuous, it's an experiment which is much more difficult to do.

            DR. McILWRAITH:  Yes, you showed a slide.  You had six markers there with Taqman assessment underneath.  Now, I presume you took the specimens from the repair tissue?

            DR. LUYTEN:  No, no, no.  This is from the sow product, huh?  This is not the repair tissue, right?  This is the characterization of the cell implant at the time of implantation.  You take the cells out of there, prepare the RNA, and determine your gene expression patterns.

            DR. McILWRAITH:  Yes, and so part of your implant.  So it's really an in vitro examination or before examination.

            DR. LUYTEN:  Yeah, because you cannot inject in that patient the cells in nude mice.  You would need too many of the cells you use for your patient to actually go in the nude mice assay.  So it's a sort of surrogate --

            DR. McILWRAITH:  So you're saying that those six markers are all useful indicators of chondrogenic potential?

            DR. LUYTEN:  Well, as I showed that on the slide, there is a very strong correlation between this combined gene score at one site and at the other site the ectopic cartilage formation assay, yes.  The correlation coefficient is 0.79.

            CHAIRPERSON RAO:  Dr. Blazar.

            DR. BLAZAR:  An implication of your study is that in order to assure that you're having cartilage and not myocyte or myoblast formation, you're going to need a strategy that will either purify away only the cells that are desired or to provide molecular or growth factor signals that prevent the outgrowth of nondesirable cells.

            So I'm wondering where you're going in terms of solving this problem because correlation coefficients are very helpful, and profiling will help you identify better conditions, but if you're looking at an all or none process, you're going to need something that will insure that you have your desired outcome.

            So I wonder what you're thinking in the future of what you would advise toward that end.

            DR. LUYTEN:  Well, again, since in the ectopic assay, in a normal muscle environment, since we don't have muscle formation there, and the cell population induces this cartilage formation, we don't have commitment into the muscle lineage.  The cells, even the precocious cells that are contaminating this final product are contributing to the repair process of the cartilage formation process.  It's only when the environment starts to be devastating that some of the cells that are around there have characteristics that will follow then the instructions of the environment and not do their own instructions.

            It is also not at all, in my view, a good idea to completely remove these cells with higher level of plasticity since there is evidence that these cells might abscond progenitives (phonetic) when they are in that environment, contribute also to the successful repair of the tissue and their engraftment.

            So the question is now what is the threshold of cells that you would accept to get in their hand, and that is where we see when you include the numbers of that above 30 percent, you are in major trouble, but you have a fairly broad range of, let us say, in that environment of acceptable behavior.

            DR. BLAZAR:  So one of the problems, though, is if you implant pure preformed cartilage, if there's going to be apoptosis in vivo and you need these other cells to contribute perhaps to their viability, you're still going to have to go back and figure out what the appropriate mix of cells is and under what conditions are you going to be concerned about differentiation of cells into the nondesirable type.

            DR. LUYTEN:  Right.

            DR. BLAZAR:  So how specifically are you recommending resolving that issue?

            DR. LUYTEN:  Well, this probable apoptosis only happens when you have an implant.  It does not happen when you work with suspensions.

            Also, in our animal model when we do the suspensions treatment in the cartilage repair, we have very little, if any, apoptosis happening there.

            When we inject or implant an implant in there, then apoptosis becomes a problem.  So, again, it has to be defined by the way you will locally deliver it, measuring the relative apoptosis there, and the contribution of the cell population. 

            When you work with the cell suspension, it looks like most of the cells are contributing to the ultimate repaired tissue.  By an increasing amount of precursor cells in there, that becomes increasingly difficult to maintain the cells, and then apoptosis will start to be more abundant.

            And from that perspective, the preparations we make are still very limited in apoptosis.  When we start to add the precursors to that, when you titrate an increasing amount of precursors of that, then that problem becomes apparent.

            CHAIRPERSON RAO:  Dr. Tomford.

            DR. TOMFORD:  Yes.  I enjoyed your talk.

            Could you comment and compare perhaps the use of autologous cells for repair as opposed to bone marrow derived cells?  That one article, as you know, compared the two, found not much difference, but you seem to feel or maybe I didn't understand the use of the autologous cells would be an advantage compared to bone marrow derived cells.

            Could you compare those methods in your --

            DR. LUYTEN:  Well, I haven't worked myself with bone marrow derived cells within these protocols.  So I cannot comment about that.

            However, we have worked a lot with synovial derived mesenchymal stem cells, which is based on the membrane markers and their behavior, have many characteristics in common with the bone marrow derived cells.

            The problem that I indicated with these cell populations is that somehow we seem to be so far at least into the conditions we are committing them into the lineage.  We don't seem to be able to commit them sufficiently well enough into this one single lineage to stay stable in that environment for a long time.

            And I guess that with the bone marrow derived cells we may encounter the same problem, but we haven't done that work with bone marrow derived cells.

            The reason why I stayed away from the bone marrow derived cells is precisely the rabbit model, because the repair of the cartilage in the rabbit model is coming from the subchondral bone marrow.  We making this beautiful cartilage to prove that we can also generate in vitro.  Yet this plug, this type of repair tissue, although sharing many characteristics of hyaline-like cartilage, doesn't seem to be surviving the environmental challenges over the longer time period.

            Therefore, at that time I decided it was probably not in the bone marrow that I'm going to find it.  That was my conclusion.

            There is another issue that I haven't discussed, that developmentally, developmentally the precursor cells that are giving rise to the articular surface are the GDF5 positive cells coming from the joint interzone, and these are present in the articular cartilage and the synovial membrane, and barely detectable in the bone marrow derived stromal cells.

            So when we work with precursor cells, we like to use cells that developmentally belong to the proper lineage, and bone marrow, we don't have so many of them.  They are more than rich in the synovial membrane and in the articular surface.

            CHAIRPERSON RAO:  Dr. Nixon and then Scully and then Dr. Moos.

            DR. NIXON:  Frank, very innovative and provocative talk.

            CHAIRPERSON RAO:  If I may interrupt.  Please shut off the mics because we get a back-feed.

            DR. NIXON:  Frank, thanks so much.  Innovative.

            I had some questions concerning the markers of cells.  Beautiful work there.  We looked at some markers and seeing cells deteriorate rather quickly and dramatically.  The implanted cells seem to go quite quickly.  I'll ask you a series of four questions, but we'll just do one at a time.

            Firstly, in long-term studies with marked cells, where do those cells actually go?  Are they apoptotic?  Are they diluted?  And if it is, in fact, a dilutional effect where you've got ingrowth of cells, how do you account for that to give you a realistic impression of the original cell pool that you implanted?  How many of those are still there?

            DR. LUYTEN:  The only data we have is up to, as I showed you, up to 12 weeks.  We don't have data beyond that, and the reason why we don't have data beyond that is if we want to label cells for a longer term period, membrane dyes are not the ones which stay around for longer than three to four months.  So you have to go over to other labeling procedures.

            We used retroviral GFP labeling, and that has a tremendous negative influence on the behavior of our chondrocytes in contrast to the mesenchymal stem cells which seemed to withstand that quite well, this procedure.  With chondrocytes, we completely lost a lot of our biology there, and therefore, we have animal follow-up for six months and one year.  In some of these animals, reasonably nice results, but I'm not sure that these cells are actually the cells we implanted.

            DR. NIXON:  When you do lose those cells, what impact do you think that the cells that are there that are implanted have on potential development or maturation of the cells that are ingrowing into the defect?  Do you think they have any impact on the autocrine, paracrine fashion?

            DR. LUYTEN:  Well, obviously, if we look into the mode of action of these cells, the only thing we know based on our leveling experiments is at least a number of these cells are contributing directly to the tissue repair.  In how far the paracrine effects are also important, we have not yet a what I call serious scientific idea and basis on this year.

            One would, however, predict these chondrocytes are, for instances, are secreting BMP2.  We know BMP2 contributes to the stability of the articular cartilage both shown in genetic human and mouse models.  So we guess that the BMP2 secretion at reasonably high levels is also going to be a paracrine effect in the tissues.  So my guess is we have a direct impact and we have a paracrine impact.

            But how the real distribution is, I cannot tell you right now.  We don't have the quantification for that.

            DR. NIXON:  Could that work both ways?  I noticed in your slide that lots of the labeled cells were right next to the articulate age where it was impacted.  Is there something in the original cartilage that's impacting on survivability of transplanted cells?

            DR. LUYTEN:  Well, it may be, but I'm not completely sure about that because in goat, the big problem we have in goat is the protection of the graft.  So you have to release these goats after three weeks to start basically walking around, and in the center of the defect there is a problem that the full loading in that area, if you go and load it, can quickly and in general quickly damages that graft there.

            So I would be careful in interpreting this that there are more cells at the borders than in the center.  It could be the result or the disadvantage of the model and the control of the protection of the graft in any animal model we have.  We can just not tell this animal model to behave properly while we can do this in our patients.

            DR. NIXON:  One last question if I could.  The difference between allograft and autograft survivability of cells is really on my mind particularly.  Do you have any labeled cell data that suggests that the allograft cell is hugely inferior to an autograft as you'd expect, or is there not that much difference?

            DR. LUYTEN:  Well, I cannot tell you yes for cartilage, but for bone it clearly is the case.  We have the data for bone that we and other groups have data.  For bone, I would agree with you.  For cartilage I don't have the data, not sufficient data to feel comfortable with that statement.

            CHAIRPERSON RAO:  Dr. Scully.

            DR. SCULLY:  Frank, that was a really interesting talk.  Can you make a comment though on the difference potentially between different stem sources and their ability to recapitulate the developmental process of the tissue architecture?

            DR. LUYTEN:  Well, we use what I call the developmental logics in this case, that is, that GDF5 expressing cells are the joint interzone.  It's an exit marker for the joint interzone. 

            David Kingsley has demonstrated by using the Cre/lox approaches and following these cells postnatally that the GDF5 positive cells are ending up and constituting most of the cell surface, part of the meniscus, and they're also left in the synovial membrane, PLoS Biology, 2004.  I think it's in October.

            So developmentally, the GDF5 positive cells which break down in deep postnatally fine back in the human adult synovial tissue, are cells that at least developmentally know when to get the proper signals, how to enter these differentiation pathways, and that's our logic to basically address or try to enrich for this population of cells because at least they would have the intrinsic machinery to eventually go through the proper developmental process.

            DR. SCULLY:  I appreciate that.  I guess my question was more along the lines:  do you have any insight in terms of whether you can actually stimulate that recapitulation of the developmental process or are these just cells that are sitting there putting out matrix components, but the matrix components are not really being organized into the most effective tissue architecture.

            DR. LUYTEN:  Well, they show the developmental cascade that we see in cartilage formation, but don't forget that for the articular cartilage we are still very weak.  Recognizing the articular cartilage from the transient cartilage based on molecular markers, we are not very strong from that perspective developmentally.

            So although in our epiphyseal cartilage and the growth process towards maturation we are very strong, my personal feeling is that the articular cartilage is constituted of three different layers for which the bottom is a remnant of the growth plate cartilage.  The intermittent Type 1 is that very special hyaline-like cartilage resistant to vascular invasion, and the superficial one is our stem cell layer and makes out the skin.

            So we don't have the developmental data that actually identifies the maturation of the articular cartilage and make it clearly distinguishable, based on markers.  There are a few of them that are suggested, but we don't have many of this unless hooking to one would reveal today that he has a number of other markers that are specifically indicating the formation of the articular surface.

            CHAIRPERSON RAO:  Dr. Moos for the last question.

            DR. MOOS:  First, I'd like to thank you for the wonderful news that we now understand 50 percent of cartilage biology.


            DR. MOOS:  There are a few points on which you seem to echo Dr. Buckwalter almost precisely.  Now, we get very excited when we begin to sound anything -- when we hear of anything that begins to sound a little bit like consensus.  So let me just go down those points and check and make sure I heard correctly.

            The first point, I think you made a quite emphatic statement that though prospective randomized clinical trials are difficult, they're essential.

            DR. LUYTEN:  Oh, yeah, I would because a lot of these consensus statements, Malcolm, are the result of experience, but not really the result of randomized trial information, and we've learned from other fields in medicine that this may be temporarily helpful, but it may not really give us the information that is valuable.

            DR. MOOS:  Well, good.  Next is it probably does matter significantly what sorts of patients you include, and that there may be subtleties to patient inclusion criteria and what procedure is best for which patient, which I think Dr. Buckwalter also articulated.

            And the final thing that struck me was that both of you seem to agree that although perhaps structural assessments may not be the primary decision making piece of information, nonetheless, to evaluate the long-term potential for any kind of treatment under evaluation, that structural measures of one kind or another may need to be evaluated at a number of time points and perhaps long term.  Is that fair?

            DR. LUYTEN:  Well, yeah.  I'm very strongly in favor of that.  I think we need to work out the sort of combined score in ICRS or in ORS 30 or 40 that establishes improvement in symptoms, in function, and certain characteristics of tissue repair that to get it judged as being clinically relevant for the patient population we are treating here.

            And there is work to be done, but obviously if you don't do the files you are never going to know it.  So you just have to do it.

            DR. MOOS:  And one more point also concerning the potency assay.  You went directly to the immunocompromised mouse and the ability to retain a stable piece of organized cartilage.  There are other assays which have been used by workers in the field.  Can you comment on how those compare, what their advantages and disadvantages might be as tests to use to qualify the types of molecular markers that you showed us?

            DR. LUYTEN:  Well, as you know, I've worked for 25 years on the in vitro cartilage tests, and I would say all of the in vitro cartilage tests are addressing specific aspects of cartilage biology.  The gross assay is the anklage independent growth, is a useful assay.  It reflects a specific characteristic of a chondrocyte which most other cells don't have, within my view is not sufficient.

            Agro surviving cells are doing fine when we grow them in fetal bovine serum after a number of passages, which are completely flatized (phonetic) in the in vivo assay.  So to some extent every single assay we have in vitro, proteoglycan matrix synthesis, the characterization of the protoglycans, the secretion of Type II collagen and so on.

            The accumulation of all these markers is obviously what we want, and that's why we believe that probably this in vivo assay, besides that, is also measuring other characteristics which in vitro you cannot measure.  It's like vascular invasion is something that in vitro is a very challenging test, and at least as inin vivo assay it is also measuring these aspects.

            So we believe we are probably addressing or looking at evaluating 30, 40, 50 percent of cartilage biology while every single in vitro test is maybe only addressing ten percent of cartilage biology.  So this is what I would call a more comprehensive assay than all these single, individual in vitro assays.

            CHAIRPERSON RAO:  Let me sort of try and get a summary statement.  One of the points that I think was addressed by a lot of the questions that was asked by several of the surgeons here was it seemed that when you looked at transplants of cells, that even though you felt that a significant number of those cells contributed to the region rating cartilage, that it was not clear that that was the principal focus or function that they performed.  It was quite likely that they may have modulated the surrounding cells or the surrounding tissue to enhance that repair function.

            And did you have a sense or could one make any prediction to say given what markers exist that this was ten percent of its function or it was 90 percent of the function that they performed?  Was mobilizing the endogenous cells as opposed to directly contributing?

            DR. LUYTEN:  In this case, mobilization of the endogenous cells in my view is relatively minor to the contribution of the implant to the tissue repair.  It's rather on the order of ten to 20 percent, while in bone forming assays with other types of cells, it's just the opposite.

            CHAIRPERSON RAO:  Thank you.

            We'll take a short break now, ten minutes, and we'll try and be back here in ten minutes.


(Whereupon, the foregoing matter went off the record at 10:07 a.m. and went back on the record at 10:19 a.m.)

            CHAIRPERSON RAO:  So welcome back.

            You've already heard a little bit about the problems with characterizing cells, and you've already heard a little bit about selecting patients and what would be the difficulties in having a clinical trial and its necessities.  You're going to hear a little bit more about animal models, but before you do that, you'll hear from Dr. McFarland from the FDA's perspectives about the issues that they face.

            DR. McFARLAND:  Thank you.

            So my task today is to give you the FDA perspective on preclinical issues in these products.  What I'm going to discuss first, I'm going to discuss somewhat about the CBER-CDRH collaboration that we've continued to do with joint surface repair products, a little bit about the framework of what we do with preclinical review of these products, and then work into the introduction to the preclinical discussion questions, which we'll be doing later this morning and early this afternoon.

            So a few years ago, the two centers internally decided that we really should be talking together a little more about these products so that we were consistent and knew what each other centers were doing, and that movement got a big push last fall at a critical path initiative stakeholder meeting in the fall, and we heard unmistakably in some of the breakout sessions that our stakeholders thought it would be advisable if CBER and CDRH were collaborating on these products so there was a uniformity of information and approach.

            Subsequent to that, there was a CBER-CDRH joint team established late last fall, and in some sense this Advisory Committee can be seen as also an outgrowth of that critical path initiative in our attempts to collaborate so that there's some uniformity in our review of these products.

            I'll spend a little bit and tell you a few more details about the CBER-CDRH joint team.  It's made up of members of both centers who are actively involved in product review, preclinical review, and clinical review of these products.  We meet roughly monthly in collaborative meetings where we share information across centers of what's going on at each individual center.

            We also have some shared educational activities, initially just internal educational activities where we were educating each other about our respective centers and what our processes are and what the institutional histories are.  We have plans to bring speakers in not just to the joint team, but to the two centers to inform us of what's going on in the field on an ongoing basis and also to present outreach to the community on FDA perspectives on the field.

            The last thing on this slide that we do is this creation of the CBER-CDRH joint team has increased the degree to which CBER members are aware at all levels of review, pre-IND stage, IND stage, and licensing stage of what's going on in CDRH by serving as a consultive nature on review teams, and likewise CDRH is serving on CBER review teams on these types of products with the understanding that many of these products, although the focus of this meeting is cellular, many of these products have device components and CDRH has similar products for similar indications that are purely devices, and it behooves us to discuss across centers.

            So with that cross-center communication, I'm presenting a general overview of the product focused review that preclinical reviewers in both centers do.  The level of review is really dependent on the characteristics of the specific product.  Is it purely a cellular product?  Is it purely a device?  Is it a combination of device and cells?

            And the specific questions that are required in each product is specific to that product, although there are some general requirements.

            And we do also look at preclinical studies designed to support designed to support the use of a specific product.  Because it is the FDA, they are framed by regulations, and a subset of regulations are listed down below, but I want to make the point that although we have different regulations in the two centers, the underlying precepts of the regulations are to respect subject safety and to develop data that allows us to move products to market.

            This is a general side of goals of preclinical evaluation.  This list could actually be three or four or five slides and is applicable not just to products that are being discussed at this meeting, but many products, but I'd like to highlight a few points.

            One of the goals of preclinical evaluation is to provide a rationale for the proposed therapy, develop a preliminary risk-benefit assessment, and recommend safe starting doses and escalation schemes for humans. 

            Other goals of other preclinical tests can be to characterize products and compare to normal tissues and to qualify analytical tests used during manufacturing.  So actually preclinical tests, although our regulatory focus and review is off into identifying safety concerns, there are multiple uses in pertinent part development for a preclinical evaluation.

            But as I said, preclinical review, we asked two questions that always come up, and the first, this may seem obvious, but it is a question that needs to be addressed because our experience has shown that sufficient information to assess the risk is not always included in an IND application, IDE application.

            So our first question is:  is that adequate?  If it's not adequate, were the adequate preclinical studies preformed and the data just wasn't submitted in sufficient detail?

            We need to conduct an independent review of the data, and this is a concept that is sometimes lost on fields that are doing cutting edge research, is that our review is somewhat different than an NIH type of review because we have to actually conduct independent review of the data.

            Now, if those data are present and sufficient, then we get to ask the interesting question:  are the risk of human subjects from products administration reasonable?

            Where can we get those data?  Well, we can get them from safety assessment in an animal model and a GLP compliant toxicology study is classically the gold standard, although well controlled animal studies conducted in house are perfectly acceptable.  We can get those data from cross-reference to data on identical or similar products previously submitted to FDA.

            Some of those data can come from in vitro studies, genomic studies, previous human experience or published data in peer reviewed journals, all of which can be potential sources of data.

            I do have one slide on the perils of using published animal or human studies as sole support for the initiation of clinical trials.  One of the problems we see with this is that often the posed reports are not designed to answer a toxicologic question and, therefore, adequate toxicology endpoints may not have been incorporated into the design of the study.

            Additionally, published reports often don't contain sufficient detail on study design, toxicities, study dropouts, et cetera, to really allow us an independent review.  And in many products that limits the use of published reports for primary support.

            What are the potential animal study designs one could conduct?  Well, classically they've been divided into pharmacology or proof of concept studies in an animal model of disease.  We've heard some of these already this morning, and we'll hear some of these later in the day, and toxicology studies in healthy animals.

            CBER and CDRH find that actually their very efficient model is a hybrid pharmacology-toxicology study, that is, a study that is based in an established animal model of disease, in this case a cartilage injury model, onto which toxicology endpoints have been added to gain toxicology data necessary to make a safety risk assessment in a relevant model.

            You can also add in these functional and biomechanical considerations and assays to collect those data as well in the animal model of disease.

            What are we looking for in toxicology study design?  This is a very, very brief overview.  We're looking at appropriate controls for a pivotal toxicology study, that is to say, a toxicology study that is designed primarily to support a clinical trial. 

            We are looking for appropriate controls, that is, controls so that we can tell that the toxicity that occurs is really due to the product administration and not just to, for instance having made the animal model of disease, having made the injury in this case.

            We'd like the trial to mimic clinical treatment as closely as possible, including product, route of administration, formulation, device, dose, regimen, reasonable group size so that a single animal that does extraordinarily well or extraordinarily badly doesn't unreasonably skew data and give you an unreasonable sense of safety or concern for a particular product.

            And the classical toxicology endpoints, some of which are listed here, mortality, clinical observations, hematology, serum chemistry, gross pathology, histopathology, body weights, et cetera, really need to be included in a study that will be used solely for support of toxicology study design.

            Now, if we had an ideal world, what would the potential characteristics of an ideal animal model for cellular therapies in joint surface repair?  Well, they would have similar anatomy, biomechanics, pathophysiology, cell biology to humans, and would also be immunotolerant of human cells so that we could actually try the human cellular product.

            However, an ideal model doesn't exist.  Therefore, we, both FDA and the sponsor communities, must understand both the capabilities and the limitations of the available models.

            Now, there have been numerous animal models established for joint service repair products.  We've heard about some of them now, and as they vary in species, they introduce inherent interspecies variations, including body and joint size, anatomic features, cell biology, and we've heard about some of that already.

            And luckily for you and for me, Dr. Matthew Allen of SUNY Upstate had graciously agreed to review animal models right after me.  So I won't spend any more time, but just to underscore what is the obvious.  Because of these interspecies variations the applicability of any one animal model for regulatory use will vary and likely one model will not be good for all regulatory uses.

            One complication that all established, most established animal models for joint service repair products, certainly in situ repair, share is that the fact that because human cells for the most part cannot be used for the joint service, you must use analogous cellular products from animal source to get the data for your assessment of safety. 

            Certainly in the cultured products, potential processing formulation and storage differences between these animal products and human products exist, and limited characterization of the animal cellular product introduces uncertainty in our ability to extrapolate any findings we find in animals to human studies.

            So it behooves us to spend time to characterize both the animal product and the human clinical product.

            So I now have arrived at the point where I'm going to discuss the issues for discussion this afternoon.  The first question deals with specific animal models for prediction of safety and clinical activity, and this is a bullet point extracted from the questions which you have in your package and was in the briefing documents, and we'd like to ask you to discuss many of these specific issues. 

            And I've already heard this morning that, you know, we're posing difficult questions, and we appreciate that.  And your discussion this afternoon will aid us in our day-to-day job of trying to evaluate applications as they come in.

            We'd like you to talk about dose and allometric scaling.  How do you scale from an animal to a human?  What's the significance of the interspecies differences for use of analogous animal cells to model human chondrocytes?  We've already talked about that a little bit.  We'd like to touch on it in the discussion.

            Methods for evaluation of interarticular toxicity and/or cartilage formation, what's the role of noninvasive imaging modalities of biomechanical tests, arthroscopic biopsy in the animal models?  What's the role in assessing toxicities both in the animal models and as models for clinical trials?

            And we'd like you to also touch on the potential need for tumorigenicity studies of culture chondrocyte cellular products.

            The second preclinical question specifically deals with pivotal toxicology study design.  We'd like you to discuss animal model or models and study duration that we should be requesting of sponsors to support exploratory clinical trials.  Likewise for licensing application.

            And we would like some discussion on the appropriateness of measures of systemic toxicity, such as we mentioned before, for these types of products, for the more traditional cultured cellular products and for the modified cellular products that may secrete molecules capable of producing systemic toxicities.

            Ex vivo transduced cells with growth factors and the potential for those growth factors to get into the circulation.

            Lastly, it was touched on a little bit before the break.  What potential additional safety concerns beyond those posed by autologous products should be addressed with allogeneic cellular products, and can those be addressed in a similar animal study add-on or are there specific additional studies that we should be asking for when we get allogeneic cellular products?

            And I'll thank you in advance for your discussion, and I will guarantee you that the results of the discussion will be taken into consideration by the agency and if past history is any indicator, will be used to guide what we tell sponsors in the future.

            CHAIRPERSON RAO:  Thank you, Dr. McFarland.

            Before we go to the next speaker, let me take this opportunity to ask two members of the committee who have joined to introduce themselves as well.  Dr. Kaiser and then Dr. Murray.

            MR. KAISER:  I'm Mr. Aric Kaiser.  I'm the expert biomedical engineer reviewer in the Center for Devices and Radiological Health and specifically in the Restorative Devices Branch.

            CHAIRPERSON RAO:  Dr. Murray.

            DR. MURRAY:  I'm Tom Murray.  I'm president of the Hastings Center, a research institute on ethics and medicine and the life sciences. 

            I apologize for being late.  I was involved in a briefing on Capitol Hill this morning.

            Thank you.

            CHAIRPERSON RAO:  Thank you for taking the time.

            Maybe I'll ask Dr. Allen to give us a summary.

            DR. ALLEN:  All right.  Good morning.  I'd like to thank the FDA and, in particular, Rich McFarland for the invitation to come speak to you today.  And I'd also like to thank Gail for getting me organized so I could be here.

            It's a pleasure to come talk to you about this.  This is a large part of what I do, is develop, validate preclinical animal models.  And at the start of this I'd like to make it clear I am by training a veterinarian.  So I recognize the value of animal models, but I'm also very aware of the ethics associated with it.

            So the only reason to use animal models is if, much in the case with clinical work, the outcome justifies the costs.  So really one of the problems that we'll see with animal models, I feel, is that we really haven't identified what the design criteria need to be.  We don't really know what the valid outcome measures are, and I think we'll see this is a theme through the animal model's discussion.

            So the other view of my talk then is, first of all, to discuss why we use animal models at all to study cartilage repair and then to look at the questions that are being asked.  I know there are a variety of things that one might want to get from an animal study.  There may be a mechanistic study in terms of a basic biological study.  Why does this happen?  What is the process of development, for example?

            Or it may be a proof of concept, just a preliminary concept that a certain material may support cellular growth.  It could be a more complicated preliminary dose finding study for either a gene therapy or a drug, or just a cell product.

            And then, of course, it could be just the pivotal safety and efficacy study that we hope is going to give us the confidence to put these into patients without detriment to them.

            So a large part of this then, the really critical step for me as somebody who is involved as a researcher who gets approached by industry to develop and run animal models is to identify the most appropriate animal model for the question that's being asked.

            And unfortunately, as is often the case, each of the commonly used models has advantages and disadvantages, and if nothing else out of my talk today, hopefully I can give you some of the insight that we got through experience of what works, what doesn't work and sort of the considerations that you have to bring to bear when you're discussing animal models.  Things that may seem perfectly logical in a human clinical setting just aren't feasible with a 100 kilogram pig.

            So those are some things, just the practicalities.  I think it would be good to highlight those.

            Also, very importantly, you've got to describe the rational selection of outcome measures and time points, and here we have a problem because, again, it's clear from the clinical setting that what the patient and the surgeon may need.  A good repair by functional description may not look like what we think a good repair should look like.  It may not look pristine.  There may be gaps around the implant, but does that really matter?

            If we can produce a surface that survives 20 to 25 years and take a patient who's 45 at entry who becomes 70, and if we can improve joint replacements to the point that they are now where we're getting very good success rates, does it really matter?  Should we really be trying to have lifelong results from a single procedure?  What's appropriate?

            Let me give you some practical recommendations.  Again, they're based on my experience and my opinion only.  So they are obviously contentious potentially, but I'm going to suggest some potential ideas for initial screening tests, and then the pivotal testing for safety and efficacy, and then I'm going to describe the obvious caveats and the limitations of current animal models.

            So why use them at all?  Why use animals at all?  Why not do everything in cell culture, in the classic test tube?

            Well, you can do a lot with cell culture.  You can look at a lot of interesting things to relate it to the biology of chondrocyte differentiation of function, and you can characterize interactions between cells and potential matrices that you're looking to explore, whether it's just how you get the cells to stick to the matrix, whether you modify the matrix to do that.  All of that can be done in cell culture.

            But at some point before this goes anywhere near a patient, we need some sort of confirmatory information from animal testing.  We need to determine what the effects of the articular environment are on cell growth and differentiation.  Look at the local and systemic safety, the stability of cell function, the expression of these candidate markers, and particularly importantly, the mechanical properties of any near cartilage that's formed.

            And then I think also we need to look at the long-term effects of tissue remodeling both within the site that's being repaired and also adjacent to the repair site because there can be knock-on effects.

            The purpose of the study then that's being designed, it may be a mechanistic study of a basic biological process, and it could, as I said earlier, be a preliminary dose finding study.  It could be to work out how big a drill hole should be, what spacing there should be, et cetera.  It could be a cellular therapy.  Which cell is better, a chondrocyte versus a stem cell, et cetera?  Or it could be to look at gene or growth factor therapies.  What number of cells expressing an inserted gene?  What dose of recombinant protein, et cetera?

            A lot of what I do in my day job is what I call A versus B studies, where we take a device and compare it with a predicate device, and we're hoping to prove that it's at least as good as the existing therapy, and I think that's sort of a reasonable framework to go from with this sort of work we're talking about now with cartilage, is let's look at what's being done right now, and if we say that that's our baseline for successful or reasonable outcome, then we should at least be able to compare it favorably with that.

            And then pivotal safety and efficacy studies for new treatment.  Well, in this context, the cartilage researching, what constitutes efficacy?  In our animal models we're unlikely to be able to say that the patient is happy.  The up side of animal studies is we have no problems with the recruitment.  Animals are recruited fairly easily, and we don't have too many problems usually with compliance of the surgery, but we have rehab issues.  We have no really good way of knowing whether the animal is functioning well.

            We can use some outcome measures, and I'll describe those that are somewhat subjective, and there are some better objective measures now available as well, but there are limitations.

            We've also got to look at what the difference is between statistical and biological significance.  The fact that we can demonstrate that something is significantly different doesn't necessarily equate with clinical outcome, and that's another problem.

            And then we've got to look in terms of the safety studies at the accuracy, sensitivity, and relevance of the particular models.

            So what I'm going to do is briefly overview each of the available models through rodent right through to, as Richard said, everything from a mouse through to a horse.  So no small task.

            I'm going to skip a few species in between, but just as an overview, in general terms in musculoskeletal repair models, there are several critical issues you need to think about at the start.

            Age and it's effect on skeletal turnover.  Generally young animals heal well, whether it's a mouse or a human.

            We have issues of bilateral versus unilateral surgery.  In many cases for a surgical procedure on a limb we'd like to use the other limb as a control.  We may want to perform the index surgery.  The comparison surgery may be, for example, a micro fracture type study on one side and then a micro fracture plus a matrix on the other leg. 

            But there are issues with doing bilateral surgeries, both from an IACUC, animal care and use committee, ethical perspective, but also there's growing information if you do surgery on one leg, the other leg is not really a pure control.  So that needs to be dealt with.

            Whether you do one stage or two-stage surgeries, particularly with autologous cells, is an issue for IACUCs.  Most IACUCs do not like you to do two survival surgeries on an animal understandably.

            The nature of the implant, whether it's autogenous or allogeneic cells.  Obviously, if it starts getting into allogeneic cells, then you're worried about immune responses.  Then there are options like the mouse that you can use or the immunocompromised rat, for example.

            But many of the allogeneic cells seem to be quite effective even in mismatched species. 

            So we can look then and vary the size in terms of area and depth, as well as the location of the lesion.  We can look at partial or full thickness injuries, critical or non-critical size injuries, and they can be placed in weight bearing versus non-weight bearing locations.  So we can vary all of those.

            But it is fair to say that all of the animal models that we currently have are limited in size, even the horse.  The largest lesion we do in the horse is still not really in the general domain of the area of a human lesion unfortunately at this point.

            And then as I say appropriate surgical controls.  You'd like to have the same site perhaps from the contralateral limb to compare apples with apples. 

            So which is the best model?  Well, the best in this context should really refer to relevance to human, but it is also impacted by practical considerations, and unfortunately cost comes fairly high up those for many sponsors.  Short studies in small animals are least expensive.  However, for the larger animals, which many of us feel might be the way to go, there are limitations with the facilities, equipment, and the skill of the personnel who do these studies.

            There are differences between the anatomy and joint function in different animals.  We're aware of some of these differences.  We're absolutely blissfully ignorant of many of them, I'm sure, and that's an area for active work that's going on now.

            Surgical access to the articular surface.  Larger models generally have better access, and that's useful because you can make bigger lesions.  You've got more cartilage.  You can do multiple sites.

            But it's also really important because if you want to apply some of the methodologies we developed to humans, then we'd like to be able to do those in larger animals.  For example, arthroscopic procedures, which are really not feasible in the very smallest of our animals.

            Rehabilitation protocols.  Again, a lot of the rehabilitation protocols for these cell based therapies are very, very intensive, and in many cases not all that practical for research animals, depending on the species.

            We have to look at the most appropriate outcome measures both in life and post mortem and recognize that one model is likely not ideal for every stage of testing.

            So he said to do master up.  So these are just a color montage of this is a human, goat, sheep, pig, dog, rabbit, rat, mouse.  Now, notice obviously we don't have really, really large mice in New York.  These are clearly scaled appropriately.


            DR. ALLEN:  If I put the mouse on here so you could see it, the horse would be off the screen.  So I just scaled them down to make it easier.

            But what I want to just demonstrate is a couple of thing.  Firstly, generally speaking, they look pretty similar.  You know, there's a fairly straightforward articulation.  There's a femoral tibial and a femoral patellar and patellar-femoral joint.

            But in point, notice there are some differences in the angles between these bones.  The tibial surface here, the articulation here on a human is sloped very gently backwards, and it's much more aggressive in the dog, for example.  The slope back here is 25 degrees.

            What tends to happen is as the dog weight bears, there's a tendency for this bone, the tibia, to shoot forward, and that's restricted by the presence of cruciate ligaments, and so these animal species are more cruciate dependent than the human, which works in our favor when we're doing transections of the crucia and inducing osteoarthritis, but it's something to bear in mind.

            We really have a poor understanding at this time of the loading conditions within any of these joints.  We can probably say that the goat is going to be more aggressive than the mouse, but it's really just supposition.  We know the loads are higher, but we just don't know the number, and we certainly don't know how they compare to humans.

            But generally speaking, the skeletal anatomy is fairly consistent. 

            Surgical access.  I have an interesting life and I operate on everything from this little guy right through.  So this is a 20 gram nude mouse, and this is a 70 kilogram Dorsett sheep, and you can see the issues we face.  You've got very nice surgical access here.  This is a rating, the middle, and this is a mouse.

            Well, you can just about implant something in there, but you really don't get very much cartilage.  So that has to be borne in mind.

            If you want large lesions, and unfortunately for human therapies we are -- even the smallest human lesion is really quit large.  You need to be at the higher end of the animal, at least the higher end of the size range.

            So there are some advanced uses for using mice for certain things, and Dr. Luyten has identified that in his very elegant talk, and there's great value in using mice as a sort of an in vivo tissue culture system for looking at the potential for cells to develop and to look at the biology.

            There are controlled sources.  They're inexpensive to purchase and house.  You have got the opportunity to have immunocompromised, transgenic and knockout animals to look at biological processes.

            But from the perspective of doing repairs on joints, they're absolutely tiny joints with almost no cartilage, and even if you can do that, the imaging on these animals if you want to do anything noninvasive, it's tremendously expensive, involving very, very, very expensive magnets that most of us don't have.

            The rabbit gets to be a little bit more reasonable.  Again, there are nice sources.  They're controlled sources of inbred rabbits of specific disease free conditions available.  They're relatively inexpensive at about $300 to purchase and relatively inexpensive to house.  They're very easily managed within the laboratory setting, and they have a good track record in surgery as surgical models.

            They still have relatively small joints and relatively thin cartilage, and one of the big concerns of the rabbit is certainly in the context of burrowing assay, also in the cartilage structure, is that there are some concerns that they may heal too quickly, especially when they're young.

            The size of them limits the usefulness of full sized human implants or fixation devices, and it's hard to scale those down.  Companies who provide you with implants often get upset if you ask for a scaled down version because it means retooling at huge cost.

            And the loading conditions may not be very aggressive.  They also spend a great deal of their time with their knees in extreme flexion, which may not be ideal either.

            The dog.  Actually, the dog again, slightly larger animal.  Its controlled sources are, again, available.  It is feasible to scope these dogs, particularly their knees.  They have a good track record of cartilage research in OA through the Pond-Nuki transection model.

            They're generally docile.  They tolerate casts and external fixators as well.  It is possible to do controlled rehabilitation on these animals.

            The big disadvantage with the dog is the ethics of using dogs.  They are companion animals, and they raise the ire of really anybody who has any animal rights issues at all, and even in our facility, the animal care people when we get dogs there get very upset.  It is very easy to bond with these animals.  It is not something I as a PI enjoy doing, dog studies, and we do do them, but it is not easy to do, and I would try and avoid them if I could.

            They are very expensive to buy, at approximately $1,000 each, and they're expensive certainly also to house, and they're housed indoors.   They're not in runs outside.  They don't get a huge amount of exercise.  They do get exercise each day, but it's not a very natural existence for these animals.

            And most of the animals, the dogs, the research dogs that are being sold, are going to medical device, non-orthopedic medical device studies where skeletal maturity is not an issue, and for that reason it's often hard to get skeletally mature animals because they just aren't available.

            The sheep.  Relatively inexpensive to buy and house.  You can get them from either commercial or non-commercial sources.  A range of sizes is possible, and they have pretty good anatomy similar to the humans.  It's feasible to do arthroscopy, but tricky because they've got a lot of fat in their knees.

            But they do have some advantages in terms of the post-op care.  They're relatively inexpensive to manage because you can put them out at pasture and they get lots of exercise.  They can be with their friends, and it's a nice, natural lifestyle.

            There are some disadvantages, there can be variability in quality.  They take longer to get to skeletal maturity than the smaller animals, and there are some communicable disease issues with sheep that you need to be aware of, particularly Q fever and toxoplasmosis.

            Their cartilage is of variable quality and also thickness.  It's generally thinner than the goat which we'll come onto, and unless you get very docile sheep, it's hard to rehab them safely and carefully.

            The goat is rather like the sheep, relatively inexpensive to buy and house.  From an ethics perspective, rather like the sheep, it's considered a food animal, and whether rightly or wrongly, it's considered less unfair to use these animals than the dog by many.

            I don't really see the great difference, although it is harder to get attached to these sheep, but it's still a living animal.

            They are available in a range of sizes.  You can get dairy or non-dairy goats.  They have similar anatomy to humans.  They are slightly finer boned, in general, and slightly less heavy than sheep, and that can be an advantage.

            And in comparison with the sheep, they do have slightly thicker cartilage, particularly on the femoral condyles, and it is still feasible, although technically a little demonic, to scope them.

            There is a disease in goats in certain areas of the country that is problematic, caprine arthritis encephalitis virus, which can be problematic.  You need to screen for that.

            There are all sorts of rumors about how difficult they are to anesthetize.  People will tell you when we started using goats they told us we should expect to lose 40 percent to anesthetic deaths and sepsis.

            If you're careful and you are fortunate enough to have people who know what they're doing around you, that should not be an issue.  Like sheep, they have relatively late skeletal maturity, and the biggest problem with goats is after they have surgery, they like to get at everything, including any pen you put anywhere near them for recording their vital statistics, any pieces of paper in the next pen.  They'll get up on walls.  They'll do anything they want.

            So they are naturally inquisitive animals, and that can not be helpful.

            So the pig.  We have not a lot of experience with the pig.  There has been a lot of work on the pig in Europe.  The minipig now is becoming more appropriate because their growth is a little bit more controlled, having a typical porker in your animal facility does not engender you to the animal care people:  the smell, the noise, the size.  It's all good.

            So there is some advantage though to having pigs.  They have much better student and facility acceptance than domesticated companion animals.  They have very similar anatomy, and it's widely touted for cardiac, GI tract, and even bone work now that they are very relevant.

            You can get them in a range of sizes, and it is possible to arthroscope them potentially, as I say, but we have not had a lot of experience, and that has not really taken off as yet.

            The horse is the largest of them all as it's commonly used.  It's possible to get large defects or even multiple defects.  The size of the horse means that you can imagine that as an aggressive loading condition inside that joint.  The cartilage thickness is the closest to the human that we have.  It is absolutely feasible to do arthroscopy and biopsy after surgery.

            There is a lot of baseline information on their healing characteristics after surgical repair, and importantly, there is a clinical need.  There are osteochondrosis lesions in horses.  We do see these and veterinarians do see these as a clinical problem.

            So this animal model makes sense from a clinical perspective. 

            Again, ethically, there are issues.  They are companion animals, but it is also address an equine disease.  So that maybe tempers that a little bit.

            There are variable sources.  There's not really a commercial source of highly controlled horses available.  So generally they may be donations or bred on research farms, and the biggest limitation to their widespread use, I feel, is that you need special imaging facilities, special surgery facilities, special necropsy facilities, and it's not generally feasible or acceptable to try to bring your horse into the hospital MRI.

            Personnel safety is a real issue with these animals unless you know what you're doing, and really at this point it's limited to vet schools, and as a veterinarian, I think that's perfectly fine, but I don't want to bang on that drum, but it is a very, very valid model.

            What outcome measures can we use?  Well, clinical function, we can put animals over force plates, before surgery and after surgery and assess their return to function.  We can now do fairly sophisticated visual analogue pain scores.  We can use goniometers and things to determine range of motion, and we can  look at visual assessment either by a second look at arthroscopy and biopsy in the large animals, such as the horse, or unfortunately at necropsy with most of our smaller species.

            Imaging both by radiography and MRI, typically requires general anesthesia.  That's something that isn't an issue for humans, but is an issue for our animals, and again, the costs of doing that are quite high sometimes.

            Histology.  There needs to be standardization of sampling and scoring schemes.  Mechanical testing and marker analysis.  Marker analysis is something that's really starting now.  There are some assays available for looking at what's going on inside synovial fluid from around diseased articular surfaces.  It's becoming increasingly used in the management of osteoarthritis.  It may have a role in looking at the repair of these lesions, but also you can look at molecular phenotyping if you'd like, or genotyping of tissue samples as well.

            I'm not going to talk to you too much about MRI and cartilage because Dr. Peterfy is going to talk about that hopefully, but there are some fairly dramatic advances now in using MRI to noninvasively assess not just the anatomy, but also the functionality, if you like, of cartilage and those are already very appealing, certainly in the human context, but for animals it may still be problematic because in many cases they require extremely high field strength magnets and also because of the nature of the scanning and the delay imposed, essentially you scan them off the delay, and that's problematic because it means our animals are hanging around, waiting for MRI for a long period of time, potentially anesthetized.

            So the feasibility is still to be determined.  For histology, you've seen in the packaged materials there are at least two widely used scoring schemes, the O'Driscoll and Pineda.  It's my opinion that the histology should focus both on the repair site and the so-called healthy tissue around the repair, particularly in long-term studies, to see if there's any knock on their fat.

            It's critical that the center of the lesion, which you'd imagine is the hardest to have dirt, paracrine, the effect from local tissue is probably the hardest to heal.  So that should definitely be included in the area that's being studied.

            And you should have a contemporary, not historical controls whenever possible, and as we've seen, I think, in previous talks, both tinctorial, i.e., regular, histology stains, and immunohistochemistry, if you bring them together can really provide value.

            Mechanical testing, the problem in mechanical testing is we can measure things, but do we really know what we want it to be?  We ultimately don't at this point.  We can see differences.  I think Dr. Buckwalter elegantly showed those earlier, but you can look at both compression testing type of tests or an indentation test which may be more useful because you may be able to do that through a scope both in a large animal or in a human.

            But the question is:  what are we looking for?  Are we looking for some number?  Are we looking for some comparison with the normal cartilage in that patient?  What are we looking for?  And that, I think, the jury is still out on exactly that.

            We pretty much know what's bad, but we don't really know what's good.

            Allometric scaling.  This has been well recognized since really this work of Kleiber here in 1947.  This relationship between the size of the animal and basic biological processes, and we know that many of those, for example, metabolic rate, are related in a linear fashion.

            But what about cartilage?  Well, if you look at cartilage -- this slide actually took forever to get together because it relied on getting a lot of information.  Actually, if you read around, there's quite a lot of information, but there is tremendous variability in what's reported, but this is my best guess estimate at this point.

            For the sorts of thicknesses you can see going from a human, a horse, a goat, a sheep, a pig, a dog, a rabbit, notice that the rabbit looks like it's smaller than a rat.  I haven't put arrows on here because you can't easily get standard deviations out of the literature, but suffice to say the rabbit and the rat, there's quite a lot of overlap.

            The bottom line is beyond the sheep there's not much cartilage.  If you look also then at chondrocyte morphometry, there's very eloquent work from Hunziker's lab looking at the rabbit versus the Gottingen minipig.  There are tremendous differences in the volume of the chondrocyte, here, looking at around 1,000, over here looking at about 4,000.  You're looking at cell density.  This is 50,000; this is 5,000.  So there are great differences between species in terms of how big the cells, the volume of the cells, not just the chondrocyte itself, but its perichondrocyte domain:  how big those are in different species.

            That has a huge impact because ultimately in a human, I would think that what we want to do is reconstruct a tissue that has similar cell density, similar micro architecture to the human.  We're not worried about what it is in other species, and if we do our animal model, then what do we do?  Do we try to make it look like the normal horse, or do you try to make it look like the human?  That's a key issue.

            Scaling volumes and areas.  Cartilage lesions represent a volume, not an area of tissue loss.  So you can't just do a 2D measurement.  For a standardized lesion in a given animal species, it is fair to look at the area as an index, but I think because none of our animal models get anywhere near the human sites, we really can only use the animal models to look at relative rather than absolute information on healing rates for new therapies.

            And, again, this is from Hunziker's lab.  This shows basically in a human, rabbit, and a goat essentially morphometry of the cartilage, showing the differences in thickness.  This is all cartilage here, and what we have as an issue here is this.  This is the same size defect in these three species, showing that this lesion here is not even full thickness in a human, whereas over here in a rabbit it has 95 percent contact with bone marrow elements, and in the goat slightly less.

            Nevertheless, they are completely different lesions.  This one has no access to the subchondral bone and to bone marrow, and this one is bathed in the stuff.

            Where you put the lesions, again, the trochlear groove versus femoral condyles here.  That has a huge impact.  There's talk about some of these sites being non-weight bearing.  These sites are all loaded to varying degrees.  We don't really understand yet what the loading is across any of these sites.

            So I'm just going to talk very  quickly about a couple of procedures.  Autologous chondrocytes and also autograft and allografts.

            In terms of animal models, modeling ACI generally requires two procedures as it does in humans, an initial tissue harvest and an implantation of cells alone or with a scaffold.  One possibility in very small animals that are inbred is to use syngeneic cells.  Just take another animal from the same strain and use its cells.  That would get around that.

            But in the larger animals that's not really feasible.  There has been some progress in that area, though there is now the recent ORS meeting last week actually.  There was a report of using diced up cartilage basically as a one-step surgical procedure which still has to, I think, be evaluated more closely.

            ACI has been performed in a whole bunch of different species with variable success, rather like it has been in humans with variable success.  We don't have a good handle yet on what the failure is associated with.  We certainly know that in some cases there's displacement of the periosteal patch, but it's also likely that there are intrinsic differences between our animals in terms of how well they do, and there may also be surgical differences in who's doing the surgery and how it is being done, et cetera, and the age of the animal, et cetera, isn't well controlled.

            I'm going to skip this slide because you saw this earlier this morning.  This is the basic procedure.  We know that there are differences in biological response based on the graph location, as it has been documented, say, in humans.  Right now if you read the literature, there's no clear guidance for lesion size or cell density in our animal models.  So studies really are needed to evaluate a range of cell doses since that's essentially what happens in clinical ACI.  You get a vial of cells and you put a vial of cells in.  The guidelines on what size lesion to treat are not all that clear.  So you're really scaling without knowing it, and it's absolutely uncontrollable.

            So future work should really focus on enhancing the proliferation survival and/or perhaps function of the transplanted cells, and in this context ex vivo therapy appears to offer tremendous potential.

            We've heard earlier about bone marrow stromal cells.  Again, there has been.  There has been very limited work on bone marrow stromal cells in animal models of actual lesions.  It's an appealing concept because you can expand them, manipulate them, and then reimplant them, and they are self.

            However, there are concerns, as have been expressed about the use of these marrow stromal cells, and unfortunately in our animals the collection of bone marrow stromal cells certainly by marrow puncture often involves another essentially surgical procedure anyway.

            Allogeneic cells have been widely used, especially in the horse, using a number of exogenous growth factors that have been introduced genetically, including BMPs, IGF-1, TGF-beta.

            The application of allogeneic or xenogeneic cells is potentially appealing because it eliminates the need for two surgeries, and really also allows you to use the cells from young donors, and it's pretty clear that young donors do seem to have a higher potential for healing, but there are significant potential drawbacks which we understand only a little at this point, including immunogenicity and also the ever existing concern about disease transmission, and any animal model that uses those, really needs to address those at some point in its development.

            Osteochondral grafting has been done in goats, sheep, dogs, horses, with variable success, but generally speaking the sort of mosaicplasty approach appears to do reasonably well, but again, as in humans, there are issues with autograph, donor site morbidity.  What do we do?  We just take an area that we're going to repair, and we sacrifice an area that was perfectly healthy before?

            It doesn't seem entirely logical no matter whether it's, quote, unquote, weight bearing or non.

            There are concerns over the long-term survival of implanted chondrocytes.  Perhaps that's not an issue because as is being shown with these cell therapies, maybe the long-term survival of the cells is not necessarily critical to the long-term function.

            There are still relatively poor guidelines on how long you can store these things for before you use them, and we can use some animal studies to look at that as well, although, again, recognizing that animals don't necessarily behave the same as humans.

            This is osteochondral grafting, and this is mosaicplasty essentially in the goat here.  A plug taken from here, placed here, lovely healing of the bone bed here, still a gap here.  Is it important?  We don't know.  Not in this duration study.  It needs to go out longer.

            So as I was reading a review and was getting ready for this, I read this article, and I just thought this was really kind of a really -- it was saying better than I could do what I was thinking, and essentially, although the repair of articular cartilage defects has been studied in many species, including rabbits, goats, and sheep, there's no consensus on the most appropriate animal model.  Absolutely true.

            None of these species replicate the anatomical cellular and biomechanical properties of the human knee, and then this was the bit that killed me.  "Therefore, we selected the most closely related species, a non-human primate, that may exhibit a healing response most similar to that of humans."

            Well, that's the problem.  Why is a primate any closer to a human in the context of cartilage with anything else?  It's not necessarily.  We don't know is the problem.

            This animal, the animals they use are about nine kilos.  They are animals that are essentially quadrupedal 60 percent of the time.  So 40 percent they're bipedal.  Is that better?  We don't know.

            So I think this is exactly what I want to try and get over.  We have a huge amount that we don't understand.  Yes, it gives us comfort that we can put this into a living situation and look up tissue, and we can see collagen Type II, but I think it's really naive for any of us to think that we're going to get anything more than that from these animal models.

            The only way to really determine how these things are going to perform is to put them into patients.

            So you can do preliminary short-term proof of concept studies in small animals, and I would recommend the rabbit is the model for that.  It can be a short term; it can be a three-month study. 

            You can perform definitive tests of efficacy for small lesions in an animal such as, I think, the goat.  It's feasible at most institutions.  It is already established as a standard model.

            But to do confirmatory studies of efficacy for grafting large lesions, you're really down to a horse, and therefore, the limitations associated with a horse and where you can do that and the facilities that are needed, et cetera.

            There are several issues in closing that I just want to just highlight.  We don't know the appropriate scaling factor for cell-based therapies.  Right now you put about ten -- I think you put 12 million cells into an ACI procedure of which the viability is about 80 percent.  So you figure you're putting in ten million cells into a defect that, let's say, is four centimeters square.

            If it's four centimeters square, it's 400 square millimeters. Multiply that by the three millimeter depth and you've got a 1,200 cubic millimeter defect.

            You scale that across to a horse.  The average horse defect of 15 millimeters is about 150 cubic millimeters in volume.  It's eight times.  Do we scale the cells down eightfold?  What do we do?  We have to have some guidelines on what we should do.

            Are there inherent differences between chondrocytes from different species?  We know young versus age there are.  We also know that if you look at a cow versus a rabbit versus a horse that chondrocytes respond to growth factors differently.

            What is the idea rehab protocol?  And is it practical in the species?  Are clinical outcomes in animals predictive of those in humans?  Even if we can measure it, does it mean anything?

            And how long do we need to follow animals?  And I would argue that for the study of most things, for proof of concept in a rabbit, it could be three months.  To determine essentially the initial response on tissue remodeling, most animals you could probably go out to six to 12 months and probably call it quits to that.  Going out long-term in an animal model may not be necessary because ultimately we just have to be comfortable that it's going to be safe in the first instance, and then maybe under very controlled settings that's when we do clinical trials.

            Immune responses are an important issue any time allograft materials are put in.  It has been determined in dogs that you do see both cell based and  humoral responses, and certainly in terms of doing allogeneic cells we need to pay attention to those in our study designs.

            So in summary no single animal model is ideal for all stages of testing and this is something we face all of the time.  That's typically true for anything we do.  So a rational approach should be to screen strategies in small animals and then confirm them in large animals with the emphasis being on safety and confirmation that the strategy is at least as effective as the current benchmark, whatever your current benchmark is.  So first do no harm.

            And ultimately controlled clinical trials are always going to be needed to document long-term efficacy.

            Thank you very much.


            CHAIRPERSON RAO:  Thank you, Dr. Allen.

            We have time for a few quick questions.  Go ahead.

            DR. ALLAN:  Since I work with non-human primates I thought I'd come back to the non-human primate issue.  I understand that using non-human primates has enhanced ethics problems associated with them and also costs.  I mean, not being in this field, what I see is that typically we had cardiac surgeons here last year or whatever, and they used certain particular models that they're familiar with, and it seems as though that with the orthopedic surgeons they tend to use specific models.  They're more familiar with those models.

            When you start moving to cellular therapies and even gene therapy, you sort of push the need closer towards non-human primates, I would think, especially if you get to gene therapy when you'd like to use the same construct in humans, and it may work relatively well.  The same construct could be used, let's say, to express certain growth factors, that those would work in primates and they wouldn't work in goats or mice.

            So it sounds to me as though there may need to be some sort of development in the non-human primate model for at least cellular therapies.

            DR. ALLEN:  I mean, I think that's an excellent point.  I mean, I think you're exactly right.  There are all sorts of reasons why we typically don't use non-human primates.  One of them is, you know, availability costs.  Many of us don't have them in our facilities because of facilities that are there, and there's also, you know, in terms of some of the Old World animals there are serious concerns about doing surgery in communicable diseases and Herpes B and all sorts of other things rolled in.

            But you're exactly right.  The fact that we have been reluctant to do it doesn't mean that we shouldn't do it.  My concern is that simply saying that a non-human primate is the most relevant because it's the closest to us.  I just don't think we have the data to support that, but you're exactly right in terms of gene therapy.  It's not a suggestion we shouldn't use them.  It's just, I think, every animal model potentially has a place at the right time, but we do need to do more work on it.

            Maybe they are the closest.  Maybe we should prove that.  We should just have some data I think now.

            CHAIRPERSON RAO:  Go ahead, Dr. McIlwraith.

            DR. McILWRAITH:  That's a really nice review. 

            I had one question.  Earlier on in your talk you mentioned that you can't emulate where the clinical location is in humans in any of these models.  What about central weight bearing area of the medial condyle.  That's where we're addressing a lot of the clinical lesions.

            DR. ALLEN:  Right.  If I said that I apologize.  That wasn't my intention.  I think actually exactly the opposite.  I think we are able to get medial femoral condyle beautifully in most of these species.  So I think that's absolutely right.

            One thing we don't do in these models generally, and you've obviously got a huge amount of experience with the horse in natural disease as well, but one of the things we do recognize in these animal models, they are normal joints that we then make bad and fix acutely.  So you know, one way of dealing with some of this, maybe we should be doing some lesions and then leaving them a period of time to get chronic and then going in looking at the reparative capacity of that first lesion and then using that, if you like, as a covariate for how well the lesion subsequently is healed.

            DR. McILWRAITH:  I think that was going to be my other question.  So really you anticipated it because some people, there's quite a few people sort of addressed that.  You know, we obviously use acute lesions because we create the lesion and do the therapy at the same time, and so people are saying, "Well, why don't you go in and create the lesion and then wait a couple of months, then go back and do your technique?"

            And you alluded to the main problem is that at least our ACUC committee would not tolerate two major surgeries, and they'd consider both of those procedures, creating the defect and doing the treatment as major surgeries, whereas we're allowed to do follow-up arthroscopies on a relatively unlimited basis because of the relatively noninvasive nature.

            But that's the main.  But do you really think that it makes a difference?

            DR. ALLEN:  I don't know.  I mean, I think that's the problem.  I don't know.  I don't know.

            CHAIRPERSON RAO:  Bruce.

            DR. BLAZAR:  That was a very nice presentation.

            Can you tell me which animal model would you recommend then for the immune studies.  The rabbit is a little more difficult to understand the immune mechanisms, and what would you propose that if we move to allogeneic cells?  What's known about what interventions might be necessary and the risks of those interventions?

            DR. ALLEN:  That's an excellent question.  I think it depends if you're using allogeneic cells.  If you're using allogeneic cells in the context of a full thickness lesion down to a subchondral bond with minimal exposure of marrow elements, then you could use a large animal such as the goat.  I mean, I think there's a big distinction between right now we seem to have lesions that are classified as full thickness lesions in a rabbit that really are osteochondral lesions, and I think you can't really do partial thickness or even full thickness lesions easily in a rabbit model.

            So for looking at structures, for example, if you want to just load cells onto a polymeric scaffold with abilities then to plumb that as a plug, the rabbit is fine.  If you want to look at cell based therapies with an intact subchondral layer, then you're into the goat or the horse.

            So in general terms, I mean, in the context of immune responses, you know, again, I'm not sure I've come here with a whole load of answers.  I mean, I think that we just need to raise the issues.

            I am not an immunologist by training, but I recognize the need to look at these things.  So I'm not the right person to ask, but I think it's an excellent point.

            CHAIRPERSON RAO:  Dr. Tuan.

            DR. TUAN:  And that was an excellent presentation.

            I'd like to make another point, and that is  in considering which animal model is a good model, perhaps the information concerning the disease frequency or the extent of diseases, such as osteoarthritis or rheumatoid arthritis in these animals, be they companion animals or farm animals, which of the animals did you discuss or even others has the kind of disease frequency in an age related way or what have you, bears some resemblance to what we see in the human population?

            And perhaps that's another criterion that we can use to select an animal model.

            CHAIRPERSON RAO:  So, Dr. Tuan, do you mean something like osteoarthritis which occurs, you know, rheumatoid arthritis which occurs in an animal which --

            DR. TUAN:  Right, exactly, right.

            CHAIRPERSON RAO:  So you wouldn't be doing a lesion; is that --

            DR. TUAN:  Right.

            DR. ALLEN:  Again, an excellent question.  The incidence of these various diseases in a species we don't know, I think, to the same extent as we do obviously in humans.  Certainly we do see osteoarthrosis, osteoarthritis in dogs, in horses.  Certainly if you apply trauma, we don't really get older sheep.  I mean, it's just we don't have those animals going over that period of time.

            I can tell you Drs. McIlwraith and Nixon will tell you about OA in the horse.  We see it in the dog.  We see rheumatoid in the dog.  You know, we do see these things, but most of the cases, you know, and they'll correct me if I'm wrong, but certainly in the companion, the smaller companion animals, most of the OA cases that are seen are secondary to trauma, and classically in the dog stifle, in the dog knee joint, that would be secondary to some ligament disruption, usually an anterior cruciate, and then some meniscal, much like it is in the human, but it's traumatic rather than necessarily idiopathic.  They'd be much better to talk to you about the horse, again, sheep, pig, most of these animals.

            We do see osteochondrosis in the pig as a genetic.  I mean, you do see these things, but I'm not sure that the incidence is sufficiently high that we can really gauge which is more relevant and which is less relevant, but it is something that you need to watch out for.  If you have an older sheep entering a study, you certainly need to radiograph its joints because they can have fissuring of their cartilage.  They can have secondary changes, and that can preclude entry into the study because they're not normal.

            DR. TUAN:  Are there specific breeds of any of these animals that could give us some more insight in the validity?

            DR. ALLEN:   Gee, that's a good question to which I don't have a good reply.  So I'll stop talking.

            I think, quite honestly, I don't know.  It would be dangerous to speculate.

            CHAIRPERSON RAO:  So if there are no further -- go ahead.  Take the last question.

            DR. NIXON:  Matthew, a beautiful overview.  It's difficult to really gauge which of the animals go to something around that size is satisfactory.  When you're working either to leave the calcified cartilage there or take it away, is there any sort of upper launch point where you can reliably take away calcified cartilage and reliably leave calcified cartilage to form an experimental lesion.

            Which animal, I guess, is that launch point?  Is it the dog or the goat?

            DR. ALLEN:  Yeah, I think it's the sheep or the goat, and even then I think it's more predictably the goat than it is the sheep because generally the goats do have thicker cartilage.

            So you know, as is the case in many other species, if you do start doing these procedures where the full thickness is removed, there may be pinpoint bleeding, but I think your best bet for having an intact subchondral bone base is probably something like the goat.  I would be leery about going smaller.

            The dog has variable thickness and is pretty thin, and everything below that is ludicrous really for thickness of cartilage.  There have been, you know, some reports of using very, very tiny, little micro blades to create partial thicknesses, but I think most people would have problems with those.

            CHAIRPERSON RAO:  Before we get to questions, we will have an open session where members of the audience can also ask questions.  You need to be recognized by the Chair, and you need to identify yourself to Gail so that you can get time.

            Before we can do that, however, I need to make an important announcement, and that has to be made before all open hearing sessions.

            And that is both the Food and Drug Administration and the public believe in a transparent process for information gathering and decision making.  To insure such transparency at open public hearing sessions of the Advisory Committee meeting, the FDA believes that it is important to understand the context and individual's presentation.

            For this reason, the FDA encourages you, the open public hearing speaker, at the beginning of your written or oral statement to advise the committee of any financial relationship that you may have with any company or any group that is likely to be impacted by the topic of this meeting.

            For example, the financial information may include the company or a group's payment of your travel, lodging, other expenses in connection with your attendance at the meeting.

            Likewise, FDA encourages you at the beginning of your statement to advise the committee if you do not have any such financial relationship.  If you choose not to address this issue of financial relationship at the beginning of your statement, it will not preclude you from speaking.

            I hope that all made sense.

            One group has already asked to address the committee, and that's Genzyme, and they have a statement prepared, and I'll ask someone from Genzyme to go ahead and make that presentation.

            DR. MATTHEWS:  Good morning.  My name is Gloria Matthews, and I am the senior scientist in charge of preclinical cartilage repair research at Genzyme.

            And I'd like to take this opportunity to thank the FDA for allowing us to speak in this forum.

            We want to share with you today ten years of experience that we have with cell based cartilage repair in order to help potentially guide some of the difficult decisions that need to be made in this field.

            We have engaged in extensive preclinical research that spans four different species of animal models.  We have experience with the design and implementation of cartilage repair clinical trials and currently have the first and only FDA approved BLA for cartilage repair, which is Carticel, approved in 1997.

            To support this, we have an FDA approved cell processing facility from which we have generated over 10,000 implants applied to patients to date, and I'm going to briefly talk about some points to consider in the preclinical arena, and then I'm going to turn it over to my colleague, Dr. David Levine, to talk about some of the clinical considerations.

            On the preclinical side, I intend to review the role briefly of animal studies in the development of cartilage repair products, and I think that this talk will dovetail very nicely with the excellent presentation by Dr. Allen.

            I'd also like to suggest some guidelines based on our experience for animal model use, and on the clinical side, Dr. Levine will review the complexities of clinical cartilage repair and briefly discuss some of the clinical trial design considerations and suggest guidelines for a regulatory approach, again, based upon some of the experiences we've had.

            Some of the types of studies that we look to animal models to address are those in the far left-hand corner of the table, and we've left the mouse off of this for the time being because it is not well characterized at this point.

            As Dr. Luyten pointed out, there is little correlation between the mouse data and either the in vitro data or the goat or human data, and also, in fact, some of Dr. Luyten's data actually is in conflict with some of our in vitro data, and certainly with some of our clinical human experience, as well.

            So for the time being I'm just going to talk about the four species with which we have the most experience, and as you can see from the safety proof of concept and mechanism of action parts of the table, based on the green circles, there are available animal models that are suitable to address these particular study types.  It's only when you start to get to efficacy, and Dr. Allen has touched on this nicely, that you reach considerable limitations with animal models and particularly large animal models, which is mostly where I'll focus my talk.

            I know the FDA is well aware of some of these limitations, but when you expand those studies to more durability of effect studies, to the eight month to 12 month and even beyond range, that's when you really start to capitalize on these limitations, and in fact, it has been our experience that humans are the best test for efficacy and long-term durability of effect once, of course, an appropriate safety profile and proof of concept and early efficacy has been established in the animal model.

            So at this point in our experience, we would question the utility of long-term animal efficacy studies.  There is no validated cartilage repair animal model.  I think that became very clear in the previous talk.

            And some of the limiting factors that we have found for applying animal to human experience include what others have brought up, the difference in the mechanical forces, the recovery environment.  Certainly the rehabilitation and postoperative mobility is very difficult to control in these animals, and the variations in cartilage, thickness, and cellularity.

            And I love that demonstration of the human versus some of the other species because the human really has the thickest cartilage of most of the species that we deal with, and also one with some of the least cellularity.  So it's very different.

            And one thing that I know has been touched on, but I think tends to slip through the cracks when we look at animal studies for cartilage repair is the very different outcome measures, the typical animal study for cartilage repair takes the lesion and looks at histology and says, "How close have we gotten this repair to perfect hyaline cartilage?"  And that is the basis upon which success is judged.

            Whereas in humans, although certainly there are a few biopsy series and there is some MRI data, primarily the important human outcome measure is clinical outcome, and Dr. Buckwalter made a big point of saying, "I care what my patient feels and what they perceive.  That's what's important to me as a doctor."

            What we would propose based on our experiences based on 43 separate studies in four different animal species ranging in time points from six weeks to 12 months, as well as a comprehensive and perpetual literature review is that studies for efficacy in large animals should be a minimum of four months.

            We and other investigators feel that if there's going to be a treatment effect, you should be able to detect it at four months.  However, if you're looking for an appropriate time frame to demonstrate a little bit longer term efficacy, we would suggest four to eight months would be the ideal time point.

            There really is no definitive data at this point to support implementation of longer preclinical studies for predicting efficacy in humans.  It's not that there is no correlation.  It's that we just don't know.

            I'd like to focus in a little closer on a large animal model comparison from our perspective, and we use the goat and the horse -- we really don't have that much experience with the sheep -- to point out some of the advantages and disadvantages, some of which have already been touched upon.

            Certainly both species are feasible for athroscopic models, and the four pictures on the left on the bottom are from a goat model from my lab, Genzyme, showing a cartilage repair model that involved an in situ polymerization technique, and the two slides on the right are from an equine similar procedure.

            The big advantage of the horse over the goat is that there is thick cartilage and subchondral bone and as Dr. Allen alluded to, there's a lot of variation in the reported thickness of the goat.  In our experience, two millimeters in goat cartilage would be extraordinarily thick from what we've seen.  We get more like one to one and a half millimeters on the femoral condyle, and that can be challenging for some of the models that we try to implement.

            The horse is much more appropriate to the human situation, but has some limitations in that large sample sizes are not as feasible.  Most of the institutions that can do equine work are not necessarily amenable to being able to do these studies in a GLP environment, and the breed and source consistency also is an issue, as Dr. Allen discussed.

            The disadvantages, of course, are common to both, and that is that the post-op mobility and rehab are very difficult to control, and long term we just don't know whether these are predictive of human outcome.

            In summary, the utility of long-term animal efficacy studies for cartilage repair is limited.  The medium term study, sort of the six-month range in the four to eight-month time point are thought to be valuable for assessing efficacy based on our experiences and the experiences of our colleagues.

            Correlation with long-term clinical outcomes still remains to be established, and we would currently recommend, as have others, animal studies for proof of concept, safety, and early efficacy mechanism of action.

            The human clinical experience still remains the true test of efficacy and durability, and I'm going to hand it over now to Dr. Levine to discuss some of the clinical considerations, and I will remain near the podium for questions.

            Thank you.

            DR. LEVINE:  I'm Dr. David Levine.  I'm a Vice President for clinical research a Genzyme, and again, I'd like to echo Dr. Matthews' thanks to the committee for the opportunity to present our experience in this field and some recommendations based on that experience. 

            Cartilage injury is clearly a significant public health concern.  If you go look at one of the landmark studies in this field by Karl , et al., the prevalence of arthroscopically diagnoses chondral injuries is very high, and subsequent studies have confirmed that.

            In aggregate, there's a very large number of patients that are found to have chondral injuries at the time of knee arthroscopy or other procedures.  The clinical indications, as I think some of these speakers earlier this morning pointed out, are heterogeneous, and what has happened over the past ten years is that there has been an evolution of lesion classification and treatment algorithms which now attempt to use the best available clinical evidence and experience we have to classify these different lesions in multiple factors and recommend specific treatments for specific patients.

            And in terms of the implications for clinical research, I think one of the key points is though in aggregate there are a large number of patients that may present in any orthopedic practice or center with chondral injuries.  When you start to look at specific indications, specific types of lesions and specific patients, the populations for a particular clinical study may actually be quite limited.

            And a consequence of that is a single trial design does not apply for all cartilage repair indications.

            There are multiple factors that go into the decision making in these algorithms for the treatments for different patients, their patient factors, lesion factors, co-morbidity factors, and procedure factors, and perhaps there should be another factor that I think Dr. Buckwalter alluded to, which is the surgeon factor.  What is the surgeon's experience and belief as well?

            My aphorism for summarizing this slide is you need to consider the whole patient and not just the hole in the cartilage.

            This is a very simplistic diagram trying to summarize the complexity of this field and where different treatments might fall along the treatment spectrum.  There actually are some very nice published treatment algorithms to attempt to put all of this information into sort of a decision tree type analysis and diagram, and for purposes of this presentation, I would sort of simplify that to really fold all of these different patient factors into a spectrum of severity going from mild to severe, where all of these different factors, such as lesion size, acuity, symptoms, prior treatments, are all rolled into sort of the severity factor.

            And when you map the different available treatments or perhaps next generation treatments against the severity index, one of the things that's striking is there is not that much overlap across the spectrum, which I think has important implications for how you design clinical trials and how you pick comparators for clinical trials.

            This complexity makes designing these trials quite challenging.  We need to choose clinical trial designs that are scientifically sound, but also have equipoise for both the patient and the surgeon and are clinically relevant.

            With our own clinical trial work, we have developed a systematic framework for assessing these different factors and balancing them to apply it to do study designs that meet this criteria.  We currently have a very large ongoing registry database which has produced several studies in this field, and we have an ongoing control clinical trial called the STAR study, which stands for study of treatment of articular repair, and it has a separate smaller study that's looking at the role of MRI and evaluating these patients.

            This is a very complex field, and for purposes of today's discussion, I just want to focus on three critical elements of trial design, the endpoints, the duration, and what might be appropriate comparators.

            In terms of endpoints, I think we echo what previous speakers have said today, that the primary endpoint really should be based on a patient focused clinical outcome, the fundamental logic of this being that that's really the whole purpose of the intervention, and that's what we should be trying to measure.

            And I think it's also important to emphasize that the methodology exists in their validated instruments to reliably measure these outcomes.

            We think MRI may be useful for secondary outcomes or for addressing specific questions, such as lesion fill or graft detachment that are supportive of the clinical outcomes. 

            We think basing studies on histology outcomes is problematic for several reasons, one being the invasiveness of the procedure and the ethical questions that that raises, but also there appears to be a lack of standardized interpretation and how to relate some of these histology outcomes to the clinical endpoints.

            In terms of duration, I think this is another area of clinical trial design where there's a need for flexibility, and the appropriate length of a trial to the primary endpoint should really be based on the product or procedure characteristics, the existing data, and the clinical indication.

            It has been our experience for patients that are at the severe end of that treatment spectrum that I presented earlier, these very large, chronic, and symptomatic lesions, that you should be able to demonstrate a clinical benefit within a year.  These are patients that typically present with pain and difficulty with activities of daily living, and once they've completed their postoperative rehabilitation and recovery, you can usually see the treatment effect within a year.

            We believe you can consider post approval follow-up to demonstrate the long-term durability of that initial clinical benefit.

            Again, in terms of comparator, this is an area where we think it should be indication specific, and just to give some high level examples of some of the challenges of picking the comparator and why there is not necessarily a single comparator that will work for all trials or all indications, if you're comparing a treatment that's fundamentally for a second line treatment for patients who previously failed, let's say, a more noninvasive initial treatment, it's very difficult to do a head-to-head comparison and very difficult to have patient and surgeon equipoise to compare to something that patients already failed.

            If you just look at one aspect of the complexity of the clinical decision making, small versus large lesions, again, patients have very different preferences and surgeons have different opinions about where the indicated treatments at the small lesion end of the spectrum versus the large lesion side of the spectrum.  So it's very challenging to pick the appropriate comparator.  And also associated with size of lesion is a surgical approach.  So it's very difficult to compare an arthroscopic procedure to a procedure that requires a full arthrotomy or a mini-arthrotomy.

            Those are not only obviously difficult trials to blind the patient and/or evaluator as to what therapy the patient received, but it's also difficult to enroll patients to trials that have such fundamentally different initial interventions and rehabilitation procedures.

            In summary, we believe that a single trial design does not apply for all cartilage repair indications.  How the patient presents has implications for the treatment options for that particular patient, but also very importantly for the trial design for those types of indications, and we would urge you to consider product specific trial design to insure that there's equipoise, the product fit within the treatment algorithms, and that the trial is appropriate for the clinical indication for that particular approach, and again, that the primary endpoint should be the patient clinical benefit.

            Thank you.

            CHAIRPERSON RAO:  I'm going to propose to the committee unless they have a very specific question because lots of issues that are raised are going to be what will be discussed in the questions which the committee discusses anyway.  So unless there's a very specific question, I'd like to thank Genzyme for the comments and hope that they're around so that we can ask you specific questions later.

            Thank you


            CHAIRPERSON RAO:  So I guess it's time to launch into looking at the questions. 

            Before I do that, is there anybody else who would like to make a comment from the public?

            (No response.)

            CHAIRPERSON RAO:  Well, if there's no one, then we can go move ahead.

            I'm going to ask someone from the FDA if they can put up the questions on the projector if possible.

            So while they're doing that, I just want to try and emphasize or remind the committee of a couple of things, especially the new members and the invited ad hoc members, is that this is an Advisory Committee, and the advisory committee is not trying to make very specific recommendations.  It's trying to express an opinion, and there's no urgency or requirement that there be a consensus of opinion, but rather that all opinions be fairly aired so that the FDA gets a full representative feel for what people who are experts in the field think.

            And there are some general issues which the FDA understands, and we don't want to try and focus on them, but rather focus on the specific issues as related to cartilage therapy and repair.  So obviously there are going to be concerns, such as, well, you know, you have to have good tissue sourcing and you have to know how you isolate the tissue and you have to know how to grow cells and standard sort of manufacturing controls that you have to perform for cells, but those are not the issues we want to worry about as much in the committee today, but rather look at model systems and the specific questions and see what are the specific issues related to cartilage repair and see if there's any kind of burning issue or kind of absence of consensus or a debate on what can and cannot be done and what's feasible in terms of designing a study.

            So I'll read out the first question and we'll try and see if we can do it in this particular order, and we'll try and do it exactly like we tried to do it earlier, is please wait to be recognized and make comments and then we'll follow from there, and we'll try and stick to the issues at hand.

            I may ask you to hold your thoughts while we try and get all the issues on the table at any one time.

            The first question that the FDA wanted us to look at was:  discuss the limitations and capabilities of available animal models for predicting safety and clinical activity, focusing on the following, and there's a list that's down there.

            And before we start with that list, I thought maybe I can make an attempt to go to the extreme to say whether there seems to be reasonable consensus.  It seems to me from listening to what we heard before, and I'm certainly not an expert in the cartilage field, was that there are multiple animal models, but there is no consensus that there is any one perfect model that everybody could say unanimously that that's the model to use.

            It also seemed to me that there was clear consensus that you need animal models.  It's not that we have perfect tissue culture models or readouts that you can just say, "Well, we can do something in culture."

            So is there any sense from the committee that that seems to be a basic starting point that is reasonable?  I'd ask the surgeons if they would agree with that statement.  Dr. Nixon maybe, Dr. Coutts, would you say that that's reasonable?

            DR. NIXON:  My apologies.  I'm a veterinarian by trade.  I'm not an orthopedic surgeon in the human field.  Dick is much better qualified to answer that than me, but I think we've heard quite a lot on modeling and from the standpoint of animal research, let me just refocus Dr. Allen's comments and say that the models that he has listed there as a rabbit for the early efficacy, some of the guides for medium term efficacy and the horse for the final kind of longer term efficacy seems to follow very well with what we do in the laboratory environment in the animal research environment for looking at various cellular products and gene therapy combination products with cell therapy.

            So we haven't compared notes, but that's the way we run our experiments.  We use those three species.  We're mostly in the high end in the horse for long-term efficacy studies.

            CHAIRPERSON RAO:  So let's then --

            DR. COUTTS:  Even though I make my living as an orthopedic surgeon, I have also operated over the years on hundreds of rabbits and lesser numbers of sheep and goats in an effort to solve some of these questions with regards to tissue engineered cartilage, and I used to believe that in order to go from animals to humans, that what was needed was 100 percent efficacy in the animal; that the model you were using needed to give a repair 100 percent of the time.  And this in an ideal world I think is a nice goal to have, but in reality you're not going to get it.

            I've never seen it.  I've never seen it from anybody else, and we now have the experience of a clinically applied methodology that went to humans after what I would consider to be fairly minimalist animal experience.  In fact, they went from rabbits to humans.

            And that opened my eyes to the fact that there are differences between animals and humans, and they're fairly significant differences.  So I would support, I think, what I'm hearing from the other speakers this morning, which is that the animals are useful for very limited goals; that you can certainly get information with regards to the safety of the product, and you can get some idea of efficacy.

            But I think to put a high standard of 100 percent efficacy is unreasonable and will be stuck in animals for the rest of our lives and never do anything for humans.

            CHAIRPERSON RAO:  That's an important point and we'll get back to it as well.

            Did you have anything specific to add, Dr. Scully, on just the general issue and the basic sort of consensus?

            DR. SCULLY:  Well, I think that the issues are in vitro studies versus in vivo studies versus human studies, maybe in three groups, and I think that in vitro studies are very important, but it's very hard to model all of the factors that are involved in tissue regeneration in vitro, especially when you're talking about biologic factors and mechanical factors that Jody brought up and things like that.

            So I think it's very important to go to in vivo studies.

            I agree with Dr. Coutts that in humans we don't have 100 percent outcome on anything.  So that I think that we need to be realistic in terms of what we measure as outcome variables, what time frame we measure it over, what goals we're looking for in terms of human outcome.  Are we just looking to delay conversion of total joint for five years?  Are we looking for a lifelong reconstructive procedure?  And be very clear about each of those things.

            CHAIRPERSON RAO:  So with those thoughts in mind, let's just go to the first subname or the subquestion in 1(a), which is:  how should questions of dose and allometric scaling be explored in animal models?

            And I should say that let's use this as an extreme and say that is it necessary to explore given the points that you've already made or is it just simply saying that one uses a large animal model and that that's sufficient because there is no absolutely perfect model?

            I will leave that open and ask anyone who would like to volunteer to go first.  Go ahead, Dr. Allen.

            DR. ALLEN:  I think the problem with large animal models, it's an availability issue.  I mean, you're talking about companies who are now going to have to go straight in with large animal models and spend a lot of money without a great deal of proof of -- I think there's still -- no matter what we do, whatever the device, the product or whatever else, there's a huge amount of comfort comes from doing it in a couple of species. 

            First, he tends to suggest there's a biological effect that's real, and secondly just from an economic standpoint, I think you've got to -- if you're going to invest the sorts of money that you will invest in even a goat study, I mean, you're looking at a lot of money to start a goat study.  It's a couple hundred thousand dollars to do anything sensible.

            That's a lot of money to put forward if you really don't know that the base of your construct is going to stand up.  So I would be very hesitant to suggest that large animals are the only thing one should do.  I would be even more hesitant to suggest that large animals should not be done.

            CHAIRPERSON RAO:  So in one extreme then one is imagining that just for dosing, right, just thinking of dose, are you suggesting that it's going to be absolutely important in your mind that the dosing be done in a large animal as well as a small animal or --

            DR. ALLEN:  I think you need to be able to repair large lesions in large animals fairly effectively before you go into even the smallest situation in humans because the smallest human lesion is a large animal lesion, straight and simple.  I mean, it's the sorts of techniques that if you're looking at very, very tiny lesions you can already manage those reasonably, it seems, with some of the current marrow access type techniques.

            So for the sort of medium to larger lesions, which seem to be the bigger problem in the long term, those have to be a large animal model, I think.

            CHAIRPERSON RAO:  Okay.

            DR. ALLAN:  Just to sort of clear things up for me because I don't know where the field is headed, maybe it's hard to tell, but it seems to me as though, I mean, sophistication is going to happen in terms of the cellular therapies you're going to be attempting to use in humans, and those will be genically modified.  There may be different types of growth factors that may be used.

            All of those things are going to end up precluding the use of probably several of these animal model systems where you can't really address either -- you may not be able to address efficacy, and you certainly wouldn't be able to address toxicity very well if you're looking at a human product.

            So it's hard to really say at this point, you know, what's going to be useful if you don't really know what the therapy really is, what the cellular therapy is because there may be significant differences when it comes to the types of therapy.

            CHAIRPERSON RAO:  But it does seem to me, and maybe Dr. McIlwraith will want to comment on that as well, is that it seems that if it's possible to do it in a large animal model because these are large defects, it would be important to do the dosing in a large animal.  Does that seem like --

            DR. ALLAN:  As long as you're talking about something that's cellular therapies that are pretty mundane in some sense.  In other words, they're just sort of derived from the --

            CHAIRPERSON RAO:  Where feasible.

            DR. ALLAN:  Yeah, where feasible.

            CHAIRPERSON RAO:  Go ahead.

            DR. McILWRAITH:  Well, with a number of products, you know, dosing is not so much critical because they've already been prepared beforehand, or their manipulation of endogenous techniques, but with regard to the question of scaling and size and shape, I think that, you know, I'm biased because we use the horse and we use the horse for a number of reasons, but one is we can justify it because horses get a lot of OA.

            Back to Dr. Tuan's question, 60 percent of horses, athletic horses, get retired due to joint disease, osteoarthritis.  So it helps a lot, but back to the point.  We can make large defects, and a lot of these have already been discussed, but we can make large defects that emulate the human situation.

            The cartilage thickness, as has been talked, is the same, and there is an issue of expense, and we can do them GLP, by the way.  We have done GLP long-term studies in the horse.

            But I think whether it's 100,000 or 500,000, it's a waste of money if you're not going to answer the question, and we feel pretty confident that I think the correlation between equine studies compared to, say, rabbit and small animal studies give you a more appropriate prediction of what's going to happen in people, and therefore, the issue of size is important.

            CHAIRPERSON RAO:  On the same dosing issue, and maybe I can add the sentence and then we can do it; on the same dosing issue for me, there's also an issue which happens across species, is you make a defect and you have a certain volume, and human cell size doesn't scale with the size of a joint, for example, right?  The size of cells scales at a much lower level than does cells in a cross-species.

            So when you're testing human cells in an animal model in terms of a dose, how does that correlate with what you might do in humans?  And is there any way to find that out, or this is not a question that can be answered in an animal model and should one be ignoring it, not highlighting it?

            And then maybe before anybody responds to that, we can get Dr. Tomford to.

            DR. TOMFORD:  Well, I agree with the previous speakers that one of the issues here in humans is aging of the cartilage, and I do a lot of arthroscopy on people with arthritis, and the problem is not so much very small, confined lesions.  It's degeneration of cartilage over wide areas.

            So the one advantage I see in a large, mature animal is that you would tend to have aged cartilage in that particular type of animal which would more closely mimic the human situation.  You know, I have a feeling that when we simply put some isolated cells into a rabbit lesion and see what happens over six weeks, all we're really doing is culturing it in an animal model.  We're not really looking at what we're trying to do to repair arthritic lesions in the human.

            I think there's a difference there.  So I think the large animal model would be advantageous.

            CHAIRPERSON RAO:  Dr. McIlwraith, you had made a point earlier where you said that, well, it's the femoral condyle, for example, which is the target in most cases of therapy, and is there an appropriate sort of pressure bearing model in any of the large animals that people would be comfortable with.

            You made the point about whether that's true or not true.  Would you want to comment on that in terms of scaling or dosing or a site in a particular place in an animal model that you prefer?

            DR. McILWRAITH:  Yes.  Well, what I was asking before, because I had misunderstood what had been said apparently, but we actually use two different models, and Dr. Nixon uses one that's slightly different, but we've used a defect on the central weight bearing area of the medial condyle of the femur that we can create arthroscopically, and we started doing that because we felt that it emulated the clinical condition in people that micro fracture was being used on in the first instance and also ACI was being used on.

            But we've done a number of studies with tissue engineering techniques where we've gone to two 15 millimeter defects on the medial trochlear ridge of the femur, and that's because we could have two defects of large size and in the same horse, and bilateral lesions as well.  There doesn't seem to be any significant morbidity.  We've got these horses exercising at two months on a treadmill, and so which goes back to the biomechanics, all of it supposition.

            Horses are big.  They weigh 1,000 pounds.  We extrapolate and say, well, that should be certainly an effective test, which was pointed out before, but we have no data on the forces.

            We've done some modeling work on other joints where we get a lot of clinical instances of cartilage loss that we do surgery on, and we have some idea of the forces, but, for instance, in the carpus which is halfway up the front leg, just with flexion we get 4,000 pounds across the articular surface.

            So we know that the horse's medial femoral condyle weight bearing is a significant area of force, but that's work that, back to your question, would be great if we could get those forces worked out.

            We have done contact studies.  So we know the articulation limits.  For instance, we have defined different forces over our proximal defect compared to our distal defect on the medial trochlear ridge.  So we need to randomize for that, but we know contact, but we don't know the forces exactly.  We just -- again, this fits into the don't know category.

            CHAIRPERSON RAO:  Go ahead, David.

            DR. HARLAN:  I just have a question.  One of the rules I've learned in my life is that one of the troubles with making rules or guidelines is that then you have to follow them, and I wonder based upon this discussion I've heard if by even strongly suggesting that any therapy be tried in a large animal model, if there's any potential therapeutic agent, cellular agent that that might limit, or if there are agents that might be worthy of testing in a clinical model for which a large animal model just wouldn't be predictive.

            I ask the question because I don't know.

            CHAIRPERSON RAO:  Maybe we can just get a really brief statement from people on that because I think it's a really important point.  It seemed that the consensus from what I heard from everyone today was that a large animal model seemed to be required for a lot of issues, at least in terms of cartilage repair, and so is there any context where one would say, "Well, it's just simply impossible or it would be really worthwhile not having to have a large animal model or not being required to go through a large animal model.

            Go ahead.

            DR. TUAN:  I think the point I'd like to make relates to what Dr. Allen was mentioning before, and that is the life style of these animals differs significantly, which therefore impacts on the extent, articulation, the weight bearing regime on all the joints.  So you can't make animals do what they don't do.

            So as a result you can only go to another animal which expands the horizon, expands the spectrum.  So that's justification with different animals, number one.

            And, number two, the other thing is, as pointed out again by Dr. Allen, the age of skeletal maturity ranges significantly, and it is definitely the case that an aged articular cartilage contains a totally different population of cells, many of which are actually apoptotic and so forth and so on, and in terms of healing and so, the response is very different.

            So unless we do both, namely, look at growing animals, as well as aged cartilage, we're not going to be able to get even the initial step we want, which is some type of early efficacy, and I think that those are some of the basic reasons for using more animal models, large or small.

            CHAIRPERSON RAO:  So that brings us to one really important -- sort of the next question is:  well, is that if one has to do an animal study and one has to do it in more than one animal, are there issues of having to test across species?  Should one be testing goat cells in goat or horse cells in horse as a comparison for what human cells will do in humans, or is it -- and are there issues with doing that because are they a good mimic for human cells or should we be testing only human cells in a horse or a goat, where we have to worry about if they're not because now they're xeno.  Whether we have to worry about all of the immune issues which we have pretty much ignored in that sense.

            And it will be nice to get some viewpoints from the experts.  So go ahead, Dr. Coutts.

            DR. COUTTS:  We've actually looked a this issue of allogeneic versus autogenic cells in animal models, and if you put naked cells into the joint, you'll get immune responses to them.  So it's only if you allow the cells to produce a cartilaginous matrix around it, which then gives them some degree of protection will the immune issue be mitigated.

            But getting back to the bigger question really of this animal model and its applicability to the humans, I think that the predictability of repair strategies in an animal model has a very low predictability for how it's going to behave in humans.

            We have evidence of that already, and so to me the principal role of the animal studies is safety.  You develop your strategy, prove that you have a pretty safe product, that you're not going to do any harm, and I think you can show that in the intermediate sized animals, and then go to the ultimate animal model, which is the human.

            CHAIRPERSON RAO:  Go ahead, Dr. Blazar.

            DR. BLAZAR:  I just wanted to get a clarification of that.  Are there studies in animals that suggested something using the best appropriate animal model for the human context in which the cells were delivered, that that was not predicted?

            Because we've heard all of the limitations of the different animal models, and is your statement based upon choosing the absolute best state of the art, given what we've heard today and showing that it doesn't predict for that exact indication?

            DR. COUTTS:  Well, we don't have a lot of examples of this, but we have to thank the Genzyme people and the investigators who did the work before them for demonstrating this. 

            They actually did not show what I would consider to be particularly good efficacy in the rabbit studies.  They showed that they could grow cartilage, but they grew it inconsistently, and it had a high degree of variability, and to a certain extent that has carried over into the human experience, but that methodology has worked much better in the humans than it did in the animals.

            Does that answer your question?

            CHAIRPERSON RAO:  Dr. Allen.

            DR. ALLEN:  I just have a comment related to that.  I think probably the problem with that is that if you do ACI inter-rabbit model, then you take the bone out at the bottom.  That's not really what you do clinically.  So you know, the whole issue is what is the most appropriate animal model maybe for that procedure, and that's not the most appropriate.

            I don't believe that the rabbit is the right model for ACI.  I think ACI has to be done in an animal model the way you do what you do in humans.  Otherwise, I mean, Hunziker's work has shown a lot of things, but one of the key things it has shown is that there is a huge issue if you expose this neocartilage that's forming to the subchondral bone and the elements below it.  So I think that you're dealing with a completely different model. 

            So I'm not surprised that the ACI and the rabbit isn't predictive of what it is in the human.  What would be more of a concern is if the larger animal models were not predictive, and I think we've got less data on that because, you know, the procedures aren't being done now.  So we've got the ultimate animal model.

            But there is potential by then going backwards and looking at things and trying to work out what is the best animal model, but it's certainly not -- I would never want to suggest that all procedures have to be done in both animal models.  You know, you have to use discretion just like everything else and pick appropriately.

            CHAIRPERSON RAO:  So the important point seems to be that there is no clear-cut data that an animal model or transplants done in animal models are predictive of what will happen in humans.  Is that a reasonable statement, that there's no data either way right now and it would be useful to have that data, and one would hope that one would have data because the limited data that's available in animal models and rabbit is not a direct mimic of what of what's done in the human studies.

            DR. McILWRAITH:  Just to add to that I would say that there has been a lot of studies done in rabbits and smaller animals that have then gone on into human studies and haven't correlated, whereas if you want to use the horse model as an example, it's early days yet for that confirmation, but we're getting close to there.

            DR. NIXON:  That was going to be one of my points, that I think that there aren't as many goat and horse studies that have gone through in cartilage compared to a clinical trial, and I'd just like to thank Dick for pointing out again and Dr. Coutts for pointing out again that we have no better data set than the results of clinical human trials, especially if they go into significant numbers.

            That ultimately is the data that we're looking for.  Experimental animals all aside, I think the results of the final application are really key in whether you go forward with that particular procedure or, as some of them are done, they're abandoned after multiple years in human clinical trials.

            If I could, Mr. Chairman, just one other issue.  You're on the issue of heterogeneous or allograft cells or autograft cells, and I think from the standpoint of the research environment, we rarely ever entertain a cross-species transfer of cells.  I'm not aware of human cartilage cells being implanted in any of the animals that we routinely use.

            Generally we avoid that if we possibly can.  The issue of allograft reaction versus an autograft is where we're focusing much of our attention at the moment because one would like to use an allograft cell if it were available.  It could be tissue banked.  It could be used very quickly, potentially within hours or days of the need arising in the OR for cells.

            And a lot of our work in cartilage repair has been using allograft chondrocytes for that reason.  The cells are usually derived from neonatal or in some instances even embryonal tissues, and we have for many years maintained the issue that those cells were in a rather protected environment in a cartilage repaired defect. We had plotted them in vehicles which to some extent isolated them, and by and large we saw none of the normal immune responses in the longer term, the four, the six, and the eight month studies.

            However, in the last couple of years, we have turned our attention to some of the very early defect changes, four weeks, eight weeks, and I think it would be fair to say at this point that we're beginning to develop some uneasiness about using allograft chondrocytes in that situation because you don't see a lot of immune response, but you will occasionally see pockets of lymphoid tissue in the subchondral bone and very rarely in the synovial tissue in that defect.

            And I think because of that many of the issues about culturing autogenous cells need to be essentially reborn again because there is a great need, I think, for an autograft cell not necessarily a chondrocyte, but a cell from some other sources that could also be quite simply harvested and directed down a chondrocyte pathway.

            DR. SCULLY:  So let's hold that thought about allografts being different from autografts, and there might be an additional concern as far as that goes because that's going to be an important part of the question.

            But in terms of the specific question that I raised, if one has goat cells and puts them back in a goat, how predictive is that for what human cells would do, given what Dr. Luyten, for example, pointed out that, you know, passaging cells causes changes, but they're not species identical in some sense, using the same cell or taking it as a punch biopsy from the same region.

            So there are those differences.  Is it a concern in terms of how one interprets data?  Should it be as one said that one just looks at safety type studies, and one looks at sort of basic things in animal models and does not worry about efficacy at all, or is it a concern in any way that there are these sort of species differences?

            DR. NIXON:  And you're asking a question about autograft goat cells, which I think is a highly relevant model, and I think the information you're going to get is only really limited by the size of the defect you can make in either a goat condyle or a goat trochlear groove, which is somewhere around the seven, potentially eight millimeter defect size, and you're beginning to get off the edge of the condyle, off into the edge of the trochlear groove.

            So there are physical limitations as to the size.  On a percentage basis, it's quite a relevant size defect, the percentage of the defect versus a cross-sectional area of the condyle.

            I feel that the goat is highly useful model to do the early efficacy data because if it's working in that situation, the final analysis can be moved forward into the largest animal species, in the horse.

            But if it fails in a goat, it's not worth moving forward, in my opinion.

            Similarly, if you're looking at strict dosing, the goat is not a bad option in that situation either.  The rabbit is very good, but the goat is also a useful screen for dosing.  We don't do normally dosing studies in the equine.  Because of the expense and the size of them, you usually go in with a specific dose either of cells or of growth factors or some other protein additive to that mixture and do a strictly head-to-head comparison of no peptide or no cells versus the set doses you've already achieved through either in vitro work, which is highly important in setting dosing, or some of the smaller ruminant defect size where you can also do dosing titration studies.

            CHAIRPERSON RAO:  Dr. Scully.

            DR. SCULLY:  With respect to the necessity of large animal models, I agree with Dr. Coutts that there's very little hard data out there, and a lot of what we work with is speculation.  But I was wondering if it would be worth listening to the Genzyme people because they're in a little bit of a unique position where they went from rabbit to humans, and then in the past seven years they've also looked at large animal models.

            And I wonder if by doing that if they've changed their clinical operation on the basis of what they've seen in large animal models compared to rabbits.

            CHAIRPERSON RAO:  Would you care to comment?

            DR. McPHERSON:  My name is John McPherson.  I'm Senior Vice President, Research and Development for Genzyme Corporation.

            We actually began working on autologous chondrocyte transplantation in collaboration with Anders Lindahl and Lars Peterson in 1995 or 1994 actually, and they had actually initiated human clinical testing prior to collaborating with Genzyme, and as was pointed out, their initial data was from a rabbit study or a group of rabbit studies that were published in 1989 in the Journal of Orthopedic Research which showed that autologous chondrocyte transplantation could generate hyaline-like cartilage in a rabbit articular defect.

            We used a combination of the rabbit data.  We confirmed  those experiments.  We used the human clinical data and moved forward as all of you know with regards to seeking licensure and approval of this product.

            During that time we also looked at larger animal models, including the goat, and our experience with the goat in the context of autologous chondrocyte transplantation has been that, first of all, the results are highly variable in the goat.  We've seen excellent repair in some animals, and in some animals the repair is less impressive.

            And the basis for that is really not clear.  It may have to do with difficulties in rehab, in managing the animals following surgery.  We simply don't know.

            Additionally, at the end of the day, we've seen positive trends in the goat after having done many, many studies that Gloria alluded to, but in terms of, you know, consistent statistically significant effects, we have not been able to reproducibly achieve --

            CHAIRPERSON RAO:  How about in the horse?

            DR. McPHERSON:  In the horse we have done some experiments with Dr. Nixon.  Some of that data has been submitted to FDA.  In that situation we have seen a positive and in some experiments statistically significant effects as well in terms of autologous chondrocyte transplantation, improving cartilage repairs judged histologically.

            One of the challenges here is obviously a point that Gloria brought up, and that is you need to ‑- I think it's important to understand what we're judging repair from a histological or histomorphometric point of view in animal models.  In people we're judging it from the point of view of patient outcome and using Cincinnati rating scores and pain scores and things like that.

            So it is a very challenging proposition.  I think our sense is more or less a sense of many of the people on the committee here on the panel in that animal studies can provide you information about safety.  They can demonstrate the plausibility of the approach from the point of view of developing proof of concept data to say that, indeed, you can generate tissue that one would expect to be appropriate and helpful in the context of articular cartilage repair, but in terms of demonstrating definitive results, as you can do, for example, with ACE inhibitors in rat models of renal hypertension, that doesn't exist.

            CHAIRPERSON RAO:  So while you have the mic, can I maybe also get an answer to one more question which I thought was pretty straightforward, given what we know, and that's F on your list here.  Tumorigenicity studies needed for cultured chondrocyte cellular products.

            DR. McPHERSON:  Yeah.

            CHAIRPERSON RAO:  And given that you have extensive clinical data as well, is there any reason to believe that there's a concern?

            DR. McPHERSON:  Well, I think Dr. Mankin actually during the discussions when Carticel was being reviewed made the best point, and that is while there are obviously situations where you can get chondrosarcomas, I do not think there are any data from any published literature, peer reviewed data, that would suggest that articular chondrocytes have the capacity to become malignantly transformed themselves.

            And so in our experience in animals, we've never seen anything like that and nor have we seen evidence of malignant transformation in any of the patients we've treated.

            CHAIRPERSON RAO:  Would the clinicians on the panel have any opinion against that statement?  Would they believe that that's a consistent statement with what would assume to be scientifically sound?

            DR. COUTTS:  The evidence to date would suggest that tumorigenicity is not one of our greatest concerns.

            I would also maybe, to carry this conversation just a little bit further with regards to the animal models and to ask my good friend Dr. Nixon here just exactly what he considers success or failure in an animal model because I think there's a broad spectrum here that one could accept as being success or vice versa, as failure.

            Any growth of tissue in a defect might be considered success, whereas if you're looking at it very carefully from a histological perspective, you might say that this tissue is far from being a cartilage, hyaline cartilage.

            So it could be that the use of these animal models will be dependent to a great extent upon just exactly how rigidly you want to define success or failure in them.

            CHAIRPERSON RAO:  That's all.  I think, Dr. Harlan, did you have something relative to the earlier topic --

            DR. HARLAN:  Yes.

            CHAIRPERSON RAO:  -- before we move to that?

            DR. HARLAN:  While the Genzyme representative was at the microphone, a general statement that I've heard over and over again is safety, with which I agree, and potential efficacy with which I agree.

            But another thing our animal models are extremely useful, and I'll bet Genzyme is using them for this reason, is that we can then more carefully follow the mechanism of the beneficial effect in an animal model and thereby possibly improve it.  So another reason for animal models, I think.

            DR. McPHERSON:  Absolutely.

            DR. ALLAN:  I hadn't really seen any data.  Since you asked about F, I hadn't seen any data on tumorigenicity studies or any discussion of that so far, and so I was a little bit sort of wondering about tumorigenicity.

            And you know, you said that you didn't see that in the chondrocytes that you've implanted.  There's also a time frame and also when using humans it's a good study.  But when you start to manipulate those cells or do anything beyond the norm, that's when you may begin to start to worry about tumorigenicity.

            DR. McPHERSON:  Yes.  I mean, if you were doing ex vivo gene therapy and doing transfection of cells, to me that's a  very different proposition from using cells that are not genetically manipulated per se.

            CHAIRPERSON RAO:  Yes, and I didn't really mean to imply that one shouldn't be considering too many studies at all.  All of those were specifically for the chondrocytes.  Was there any available data which suggested that this should be of high concern or not.

            And before I take you there --

            DR. SCULLY:  Can I just follow up on the tumorigenicity question?  I'm an orthopedic oncologist.  So it's something I deal with on a regular basis, and I think the things that need to be brought to light are I was a Fellow with Henry Mankin, and I remember when I was a Fellow he was asked to write a chapter in a book about neoplasms of the joint, and in a typical Henry fashion, he wrote on one page in the middle "none" and mailed it back into the editor.

            So tumors don't happen in joints very often, and the risk of chondrosarcoma in the general population is on the order of one in 150,000, one in 200,000.  So if you've done 10,000 implantations, although that's reassuring, it's probably --

            DR. McPHERSON:  But to show statistically significant effects that the procedure is inducing chondrosarcomas would be a challenging proposition, indeed.

            DR. SCULLY:  Right.  Exactly.  But the fact that it's negative so far isn't necessarily --

            DR. McPHERSON:  No.  Maybe I misstated it.  We have no evidence.  The lack of evidence isn't a demonstration that it couldn't happen.  I will concede that, but a concern about using autologous chondrocyte implantations from the point of view of inducing malignant transformation to me is at this stage at least not a major issue.

            CHAIRPERSON RAO:  You can extend that even to allogeneic or to other cell types, and that's an important distinction that we should keep as well.

            DR. HARLAN:  Well, the only thing I heard this morning was when Dr. Luyten presented.  I thought you said that even after five passages the cells were very different.  So that suggests to me that something is happening to these things in vitro, and it raises at least the concern of malignant transformation.

            DR. McPHERSON:  Well, I mean, just as a point of clarification, when chondrocytes are cultured on plastic, they do go through a de-differentiation process.  This is a reversible process.  It was first published by Joyce Schaeffer and Paul Benyon in 1982 in a Cell paper.  The rabbit chondrocytes when enzymatically released from cartilage and cultured on plastic de-differentiated, but when they were put in suspension culture, they had the capacity to re-differentiate and develop the same biosynthetic profile as a freshly isolated chondrocyte as judged by Type II collagen expression and proteoglycans like aggrecan expression.

            And we've verified that with human cells.  We have published that in the Journal of Orthopedic Research.  The human cells behave very similar to your rabbit cells in the context of expressing a different phenotype, extracellular matrix phenotype when cultured on plastic, more similar to mesenchymal fibroblasts, but when they're put in suspension culture they re-differentiate, and to us that's one of the key sort of issues or points to consider as it relates to the potential utility and how chondrocytes, propagated chondrocytes can in an appropriate environment produce extracellular matrix consistent with more hyaline-like cartilage.

            CHAIRPERSON RAO:  Thank you.

            Is it related?

            DR. LUYTEN:  Yeah, it's related to this point.  For sure, the reexpression of the phenotype as judged by Type II collagen and sulfated matrix synthesis looks okay.  However, if you do analysis of gene expression patterns and compare the reexpressed chondrocytes associated with the primary chondrocytes, you do see another weld coming up there.

            CHAIRPERSON RAO:  So that point is important, and I think what Dr. Harlan, I think, was alluding to is that while we can make the statement for first passage or early passage cells, you cannot generalize and say, well, it's chondrocytes and we're now looking at tenth-passage cells and we don't have to worry about issues.

            Cells change, and since they're a different population, we'll have to consider them.

            You know, I want all of you to look at C, D, and E, and I'm going to try and see if we can combine them in one way because it seems to me that there has been an issue which has been based throughout in all of these conversations, is that the animal models don't seem to be very predictive in terms of efficacy.  Right?

            You can show that the cells have a particular phenotype and you can look at safety, and you can even imagine things happen.  So the question here seems to be what should we studying, and all of these sort of tests seem to be related to efficacy, right?

            Should one be looking at arthroscopic biopsy in terms of cartilage repair, in terms of assessing the cells if you transplant them?  What should be biomechanical tests of the quality of cartilage that's repaired, or should one just be doing noninvasive MRI screening in terms of assessing the quality of cells, or is it useful in terms of doing any of these?

            And maybe I can first ask the clinicians to make comments directly, and then maybe I can ask you too.

            DR. MOOS:  Well, can I make a really quick comment before you get started?

            I think we need to make the distinction when we're talking about efficacy in an animal model.  I think we're really talking about biological activity in that context, and we have a distinct meaning for clinical efficacy, and I don't want the two to get confused in our discussion.

            CHAIRPERSON RAO:  Yes.  I stand corrected.

            So maybe I can have maybe -- Doctor?

            DR. COUTTS:  I think we all are anxiously awaiting the continuing development of noninvasive modalities for the evaluation of articular cartilage repairs, and I think that we have enough evidence to date to suggest that the CT, but more importantly, I think the MRI will be a very useful tool for the evaluation.

            In terms of their use in animal models, I would see them being used as a way of validating their ultimate application in the human.  So that would be a very important thing.

            CHAIRPERSON RAO:  So a separate function from the question that's asked here?

            DR. COUTTS:  Yes.

            CHAIRPERSON RAO:  I see.

            DR. COUTTS:  Of course, one reason that one does animal studies is that you can harvest the tissue and study it in a much more detailed fashion and thereby avoid the uncertainties that are produced by these less complete evaluations.

            MRI has suffered from a lack of resolution up to now, but it keeps getting better and better, and it looks like they will be able to give us useful information about the health of these repairs in the future.

            Biomechanical testing has typically shown very poor performance of articular cartilage repairs in the early time periods, and I don't see that changing even if we end up having a super repair that actually grows normal hyaline articular cartilage and completely restores the articular surface.

            Early on it is going to show a fairly high indentation and permeability.  So I don't see it as being terribly useful in short-term studies.  In a long-term study it could be very useful and particularly if you had a very good repair methodology it would be a way of validating that it is, in fact, a good repair.

            Arthroscopic biopsy is I find of little value.  It's trying to figure out the size of the elephant by feeling its leg, and so in an animal model you had better get the whole thing and look at it in a more complete fashion and really tell what's going on.  There's little useful information to be obtained from the biopsy.

            CHAIRPERSON RAO:  So that would impact, say, if you were looking at a model, an animal model, in which arthroscopy was difficult.  You would not consider that a major hindrance in terms of choosing their animal model.

            DR. COUTTS:  I would not.

            CHAIRPERSON RAO:  Dr. Scully, would you have?

            DR. SCULLY:  I don't have a lot more to add to what Dr. Coutts said, with the exception that I think that just as different animal models provide different information, different endpoint evaluations can provide different information, and I think what you have to do before you say this is useful or not is to identify what clinical endpoint you want to improve, and then you can talk about whether this can be useful in helping you understand it.

            CHAIRPERSON RAO:  Go ahead, Doctor.

            DR. TOMFORD:  I just wanted to add that it depends upon the answer you're looking for.  It seems to me if you want to prove that it's hyaline cartilage, you have to do certain tests.  Biomechanical tests are not great, as Dr. Coutts said, but they're pretty good.  You need metabolic tests to see what's in the cartilage and arthroscopic tests, I think, depends upon the size of the animal.

            But if your endpoint is or the standard you're going to hold the process to is reestablishment of hyaline cartilage, really the only way to prove that is with these types of studies.

            DR. RAO:  From all that I heard, and again, you know, as I said, this is sort of an Advisory Committee, it doesn't seem like this is a good endpoint to hold anybody to, right?  And if that's the case, then this is not a good or a required test to do.

            We learn a lot of the biology of the cells, and that's always important in terms of basic science, but the FDA also sort of has to weigh what's an absolute requirement and necessary in terms of safety and application.  I want to get a sense from the clinicians whether that seemed to be something that they would recommend as an absolute requirement.

            And it doesn't seem to me that that was what they were saying.  Would you agree with that?

            DR. TOMFORD:  Well, not quite.  In other words, I think the discussion of the endpoint that you're going to require is another discussion that we really haven't had.  But I'm saying that if the endpoint is hyaline cartilage or that is the human cartilage, articular cartilage is hyaline cartilage, and is that what we're trying to replace, we have to use certain tests to prove that that's the case.

            If we're not interested in proving that that's the case, then why do the test?

            DR. RAO:  Go ahead, Dr. Tuan, and then Dr. Mule.

            DR. TUAN:  I think a point needs to be made also that it depends on the cell based therapy that you're doing.  If you are using a completely cellular non-matrix associated  implantation, then I think these tests as they are stated is probably, you know, more or less a look-see type of thing.

            On the other hand, if you were going to do an ex vivo methodology where you develop something that is almost cartilage, then it's important to follow the performance of that implant, which is cell based to make sure that it is indeed not deteriorating, for example.

            CHAIRPERSON RAO:  Hold that thought.  Dr. Mule.

            DR. MULE:  This is an admittedly naive question.  What we're interested in in the end is whether we come up with products that improve human function and diminish the symptoms of severe injuries, degenerative injuries also.

            Number one, is it possible to look at, to get reliable measures, meaningful measures of function, including, say, lack of pain or reduction in pain from any of these animal models?

            And, number two, would such measures provide us any reasonable basis for then generalizing to humans?

            CHAIRPERSON RAO:  Go ahead.

            DR. McILWRAITH:  Just back to C, your question after the clinicians answered.  You know, I'd add in there diagnostic arthroscopy, not necessarily biopsy, but we've certainly found it useful to get an idea of functional tissue versus nonfunctional tissue, whether the tissue is in there, well integrated, or whether it's coming out merely with arthroscopic probing.

            We have done work with the arthroscopic indenters, and this was mentioned before by Dr. Allen.  They give you an objective number that correlates pretty well with your arthroscopic probing.

            But I think that certainly does cite horses, it means you don't need the number that you need before.  It still begs the question that I'm sure we're going to get to at some stage as to what's the gold standard for repair tissue, but for functional tissue, tissue that is comparable to articular cartilage or noncomparable, depending on the case, I think the diagnostic arthroscopic exams -- we do them at four, eight, and 12 months -- are really useful.

            CHAIRPERSON RAO:  That was really useful, and maybe I should add one more question to this piece in terms of doing this and then get back to what you asked as well.

            It is how long, right?  I noticed that when we were listening to the talks earlier, people said and Dr. Luyten, for example, show data that you can look at six weeks and eight weeks and things look great, at least in the rabbit, and then you look longer and it doesn't look like it's that great.

            And then when Genzyme presented, they said that, you know, eight months is a reasonably long time, but you know, going to two years may be really hard in terms of the animals and the expense.

            And we did talk about all of these tests, but there's a corollary to these tests, is when do you do these, especially when you're thinking about MRI and how you're following them.

            And, again, it would be nice to get comments on what one would think would be a reasonable duration, which would be reasonable in the sense of giving one useful answers about the cells that you're using or the therapy that you're proposing.

            DR. McILWRAITH:  Well, we've done -- based on some studies we've done, there's a variation depending on what technique you're looking at, but when we compared four month postmortems at 12 month postmortems in looking at micro fracture, for instance on the femoral condyle, the four month assessments were very predictive of what we saw at 12 months.

            On the other hand, when we've been looking at some tissue engineering techniques, we saw trends at six months, but the differences were clear at 12 months.  We feel you have to go out 12 months with those just based on the work we've done.

            And we've gone out to 18 months, and we did show differences between 12 and 18 months when we looked at scaffold alone compared to scaffold plus cells, just to give an example.

            In other words, the technique that failed failed worse at 18.  Like it looked terrible at 18 months, but we could see that at 12.  So I was discussing this with the other guys in the lab yesterday, but we feel 12 months is certainly enough, but I think you need to go, based at least on our stuff, with the tissue engineering technique, you need to go out to 12 months.

            CHAIRPERSON RAO:  Dr. Nixon.

            DR. NIXON:  Let me just add one or two things because I think it's vitally interesting to look arthroscopically at many of the defects, and when you're not using a scaffold-based system and you've got more soft cells, you've got more cell direct implants rather than on tissue engineered products, it's also very useful to look early as well. It answers many questions about the survival of a cell there, and it also allows you to look and develop a scaling system for an arthroscopic score and use other devices.

            Dr. McIlwraith has already talked about the hand-held mechanical devices which, by and large, contain a lot of variability, but there are improvements to those that are happening, and there's also other things coming down like the optical coherence tomography, which is a vitally interesting tool to be able to noninvasively develop essentially a histologic cross-section.

            So I think arthroscopy will undoubtedly have a huge role.  We have a very bad feeling about arthroscopic biopsy.  I'd certainly echo what Dr. Coutts has said.  It's not an easy tissue to analyze when you get it out, but it also has an impact on the surrounding tissue, and we thought it was a great idea for a little while, but it's very clear that if you biopsy at four or eight weeks and you go back at 12 months or eight months and look at the defect, you can still see where you took the biopsy.

            And it's hard for me to see that it's not having an impact much beyond the one-by-two or one-by-three millimeter defect you made to biopsy.  So we've stopped doing that.  We're very much in favor of arthroscopic repeated analyses of whatever mechanical devices you want to put on it, but certainly the ability to look and carefully score those lesions gives you a pretty good predictability of what you're eventually going to see.

            And the longer you look, I think eventually you're going to want to analyze the tissues, and if you've got a large defect, you've got at your disposal then tissues that you can do both histology and molecular analyses and biochemistry as well, and perhaps not so much mechanics anymore.  The answer I see to questions about mechanical tests are that we don't do them very much anymore.  We have many other mechanisms to analyze tissue, and by and large, the mechanical capabilities of them even under the best of repairs.

            We slowly keep edging up as far as repair quality, but mechanically they're still far deficient compared to normal cartilage or even the cartilage surrounding it.

            So perhaps mechanics has been put on hold for a decade, or whatever you'd like to look at, but I think at the moment we're mainly focusing on the integrity of the cartilage, its structural integration at the base and at the perimeter, and the hyaline qualities of it and to some extent the biomechanical composition.

            I mean, we're always assaying proteoglycan and collagen, and we see we're slowly edging out collagen Type II, and to some extent aggrecan, although it seems to be the big challenge right now is getting aggrecan levels anywhere near normal cartilage, which I think gives you the answer why mechanics are still rather poor in these defects.

            But if I were at this point looking at serial arthroscopy with optical coherence tomography or MRIs, serial MRI certainly gives you a very effective way, particularly if GEMRIC studies in MRI and some of the intense MRI software program give you a really pretty good answer as to the quality of the repair.

            CHAIRPERSON RAO:  Did you have something to add to that, Dr. Allen?

            DR. ALLEN:  Yes.  I just had two things, one about early time points.

            I see the real value of doing early time points perhaps in a small animal as being a very good predictor of negative outcome.  I'm not sure you know that very much is -- you don't get a lot from doing okay, but you do get a lot by doing badly because it means that if you're doing something like a goat or a horse where it's a longer term study, and if you're not doing serial scoping or some other MRI or something, you've got a whole bunch of goats going out there with problems.  That's an expense, et cetera.

            So I see the value of the short term as much as anything as being let's weed out the bad stuff quickly, and a lot of what I do as a job is in terms of preclinical testing.  You know, a lot of what I do is joint replacement.  I'm unlikely to find something that's going to make a huge difference to more than probably one percent of the patients by finding something new that may be better, but if I find something that's bad, I have the potential to impact a whole bunch of people.

            So, you know, I see the ability to predict negative outcomes is very valuable.

            In terms of MRI, I think one of the real values of MRI in the animal models as well as we have the potential to do MRI, and especially if you do it around the time of necropsy and make it in a sense sort of maybe a nonsurvival event.  You can apply some of these newer technologies and you can get the pictures on MRI, and then you can directly correlate them with the histology of that particular lesion, and that gives us comfort as we move forward then to using these noninvasively in humans, and it should absolutely obviate the need to take any slices of anything through an arthroscope.

            If you know what that tissue -- if you can show the transition of that tissue towards a hyaline architecture and show that on the GEMRIC or whatever other technology you're using.  As the resolution gets better and the slice thickness goes down, we really do have the ability to do in vivo essentially sort of histology, and we can get small pieces of what is a large lesion.

            So I would not rule it out.

            CHAIRPERSON RAO:  Before I get to you, Dr. McFarland, just a quick question.

            Dr. Tuan, did the discussion on different tissues and types address your issue which you had raised a little bit earlier in terms of scaffolding and so on would cause different issues?  Was that Dr. Tuan?  Did it answer your question?

            Yes, go ahead, Doctor.

            DR. COUTTS:  Well, I'll just echo what was just said because when I first answered the question about MRI, CT, and ultrasound, it was within the context of should this be a modality used for evaluation in animal studies.  It's a different question of using them as a way of validating those methodologies for human application.  I think that's a very, very valid point.

            With regard to arthroscopy, I think that is useful in animal studies, obviously only in larger animal studies because the small joints don't really lend themselves to it, but I'm sure we'll get into the subject of should it be used in the human studies, which is an entirely different question, and even though I think we could justify it in the animals, we're going to have a knockdown/drag-out, I think, on the issue of this.

            CHAIRPERSON RAO:  Yes, and I'll ask you to comment on that when we get to looking at those sorts of questions.

            So on that note, let me ask the FDA:  do you feel that we've addressed the issues or you've gotten what you need in terms of input on Question 1?

            MR. McFARLAND:  Generally, yes.  Eric?

            MR. KAISER:  I agree.

            MR. McFARLAND:  I do have one specific question I want to pose, going back to Part A, and what I heard was that doing dosing studies in equine is probably not a feasible idea.  Goats might be a good model; rabbits might be a reasonable model.  What I didn't get is an answer whether we should be concentrating on cell number, scaling by volume of lesion, or the like or is that product dependent and that when a sponsor comes to the agency and said, "We did X volume and we want to scale it," give me some advice on how I should say that they're scaling has some rationale about it?

            CHAIRPERSON RAO:  I'm going to ask Dr. Allen to comment first just because he made the point about the size of a defect and the volume of that defect in terms of -- and we can calculate back from the number of cells you can reasonably pack in that area, which gives you a certain absolute limit of what you can reasonably expect to get.  So maybe --

            DR. ALLEN:  I thought you would ask me because I was just finishing a mint.


            CHAIRPERSON RAO:  I noticed that.

            DR. ALLEN:  Yeah.  I was hoping you were going to slow down.  I'm all done.

            I mean, I think it's a key issue.  I think the critical thing is the animal models, all of the animals have different cell densities, and they're typically higher than the adult human.  I guess my perspective would be that we have a volume of a certain size, and we have in the native cartilage a certain cell density, and in our animal model we should recreate the normal tissue for that species with the understanding that hopefully the humans will do the same.

            I think that's as much as I would want to do.  I think there should be some emphasis placed on scaling, but it should be based on what is normal for that species.  We should not try and create low density cartilage in a goat or high density in a human.  I'm not sure.  I don't know that that makes sense.

            So absolutely, I think you've got to pay attention to the size of the defect and how many cells you're going to put in it to achieve then -- you know, you're fracturing in the chondrocyte volume, the domain volumes and stuff; you can make some predictions.

            So I think it's a very important point, and I think it would be very useful in animal models to look at, you know, create a range of sizes rather than everybody seems to do a different size defect, pretty much as big as you can get.  One of the problems with doing different sizes of defect in a controlled setting is there is a critical threshold.  I often say for the rabbit it's three millimeters.  You can't go that much more and still achieve -- see, we have the whole spectrum squeezed down.  It would be really nice to do, you know, a critical sized defect, 20 percent more, 40 percent more, 60 percent, and scale our cell, if it's a cell based therapy, scale it up and show linearity, and then we could maybe extrapolate.  But it's a very hard study to do.

            So the best I think we can do is shoot for restoring normal anatomy for that species and then do a controlled clinical study and hope we got it right.

            CHAIRPERSON RAO:  Maybe here might be a good time to really ask Genzyme if they have a comment on how they decide on their clinical patients when the supply cells or what number or size is there.  Is there something they use?

            DR. McPHERSON:  I mean the reality is in terms of the dosing, it was established empirically, again, based on experiments that were done by Anders Lindhal and Lars Peterson in patients many, many years ago, and so there has not been -- I mean, the truth is there has not been, never been to my knowledge a systematic sort of dose ranging kind of study looking at efficacy using autologous chondrocyte implantation.

            It's based on the empirical data that -- I'm sorry -- I mean the data that were generated from the empirical approach that was used by the Swedes, you  know, 15 years ago.

            CHAIRPERSON RAO:  Is there a range at all?

            DR. McPHERSON:  In most patients the range is from about ten to 30 million cells at the time of implantation.  In the defects range, the average defect is approximately four square centimeters, but there is a broad range from probably two or three square centimeters up to 12 to 15.

            CHAIRPERSON RAO:  So is it a fair statement then to say that a kind of dosing range is kind of an important thing that needs to be done in some fashion?

            DR. McPHERSON:  I think it's a fair comment.

            CHAIRPERSON RAO:  Do we have a range based on what's already been done in the clinic in terms of getting that?  Dr. Moos?

            DR. MOOS:  Yes.  Is there any evidence from your experience that too many cells could ever be bad?

            DR. McPHERSON:  No.

            DR. MOOS:  There's a practical limit, of course, and maybe you've not reached it, and when you talk about ranging studies, you know, you generally like to --

            DR. McPHERSON:  In animal models where we've probably done -- looked at the widest range, we've not seen a toxic effect of cells, for example.

            CHAIRPERSON RAO:  Do you want to address that?  Any specific comment and then maybe we can break for lunch.  I see lots of people looking at their watches.

            Go ahead, Dr. Allen.

            DR. ALLEN:  I've just got one comment coming back to the question about pain in animal models.  I think two things.

            Firstly, we can assess that with visual analog scales, limb usage, measures of joint function, weight bearing, et cetera, but I think the important thing for us to bear in mind is that our patients or our research animals come to us with perfectly healthy joint function and no pain in the morning that gets better as they exercise.

            Our patients that the clinicians see have in many cases a chronic history.  Their expectations of a satisfactory outcome may be very different than our animals.  Our animals may respond.  If it's not perfect, they may be painful, but I'm not sufficiently familiar with the data.  It would be interesting to see what the people who do this in the horse, how uncomfortable those animals are early and if that gets better and how they score that.

            But certainly, I think pain is a critical issue ethically for the animal models, but also as a very good indicator of how these things are performing actually.

            CHAIRPERSON RAO:  Thank you for reminding me that we had forgotten to answer that question, as well.

            So if there are no further questions, then we will break for lunch, and we'll reconvene here.  Maybe we can try and do it at 1:40.

            MS. DAPOLITO:  There is lunch arranged in the atrium upstairs for the committee and the participants.


            (Whereupon, at 12:49 p.m., the meeting was recessed for lunch, to reconvene at 1:44 p.m., the same day.)

            DR. RAO:  As you may have gathered looking at the agenda, that they're going to try and continue with some of the pre-clinical questions, and you see them up on here on the board.  And we have about an hour to try and get through them, so I went through the questions, and I'm going to make a suggestion to the committee, just because we want to give full time to the first two issues, which I think are quite important.  And perhaps, that's why the FDA listed them first.  But I don't want to omit any discussion on the last two issues, but since I feel they will be shorter, I'm going to suggest that we start with those first.

            And I'll read them.  And Question 3 here is from an allogeneic cellular product for articular repair; what, if any, additional safety concerns beyond those posed by an autologous product that should be addressed in an in vivo study prior to initiation of clinical trials.  And to me, this is somewhat of a complicated question.  On one end it's sort of simple, is that people do do transplants which show some kind of mismatch, and there are issues that the FDA has learned when they do bone marrow transplants and looked at all of that, but can we study those sorts of issues in animal models?  And is there a reasonable way to do it?  And if so, what would be the additional things that you'd want to look at in an animal model above and beyond what you would look at if you were just doing an autograft? 

            And I'm going to ask Dr. Scully to lead off on that discussion, and see where we go with it.

            DR. SCULLY:  Well, I think that one of the issues about safety of allogeneic material has always been transmission of disease, and I think that we have a fair amount of knowledge and procedures built into the tissue banking industry that probably will protect against that.

            The second is the immunologic aspects of transplantation, and can we develop autoimmune disease from it?  And the truth is, I don't know what the answer to that is.  The data that comes to mind is some dog studies where they looked at bone transplantation - Sharon Stevenson - and with progressively genetically mismatched dogs, there was progressively worse incorporation of the bone graft.  But I'm not sure that they saw any autoimmune disease that came out of that.

            DR. RAO:  Does anybody else have any specific comments on this?  Dr. Harlan.

            DR. HARLAN:  Nothing that can be addressed in animal models; although, the statement about transmission of disease, I think to be perfectly clear, if the source for the donor tissue is a cadaver, our screening methods are imperfect in that situation.  So for a life-threatening disease, certain the transmission or the transplantation of allogeneic tissue makes sense.  But for the reasons that Dr. Buckwalter said this morning, if we're using cadaveric tissues for something that's not life-threatening, it would be a cause for concern.

            DR. RAO:  Should there be, perhaps, a record kept of the MHC typing of the cells, if you were going to use them, because this was not going to be autographed?  Is there any sense from the community on that?  Dr. Tomford.

            DR. TOMFORD:  Do you mean in animals, or do you mean in humans?

            DR. RAO:  It would be humans.  So would that be somewhat of a necessity in terms of just keeping track, since it's hard it seems from what I hear in terms of looking an immunogenic responses, because most of the ?? let me back up a step and say that it seems to me from listening, is that in the animal studies most people look at animal cells back in the  atom.  Right?  And it's not a good idea to take human cells and put them back in an animal to study them because there aren't standard immune suppressive models in a large animal that you can study, so you will have some data.  And those will generally be an inbred or syngeneic transplants that you can look at.  I'm not sure that we can match any kind of rejection phenomenon that you might see for small MHC class mismatch, for example.  So if that's the case, should there be a specific issue if somebody were to come in and say I want to do a clinical trial and look at putting cells into humans.  And since we can't maybe do any of this animal study, is there at least some basis for data keeping that one should perform; is that, well, we have to keep track so that we have a retrospective analysis of what happens if you do that.  Should we be keeping track?

            DR. TOMFORD:  First, we know that human chondrocytes have transplantation antigens on them, so they certainly will stimulate an immune response.  Whether or not it's worth undergoing MHC testing I'm not sure.  I know to match human bones in a bone bank takes about four or five thousand bones or tissue to match one human.  You have to have about four thousand product array or range, so I'm not sure that that would be very feasible.

            DR. RAO:  Dr. High, do you have anything to add to that?

            DR. HARLAN:  Well, I understood your question to be should we just keep the data?  I agree with Dr. Tomford that it would be difficult to find a match, but I think there's a minimal cost of keeping the data and potential benefit from doing so, so I would support collecting that data.

            DR. RAO:  Dr. Tuan.

            DR. TUAN:  I was just going to make a comment.  Again, I see that we're primarily referring to the chondrocytes, but I believe if you were to refer to other allogeneic cell types, such as the nesencymal stem cells from various origins; from placenta, bone marrow, et cetera, et cetera, there have been many studies in animal models; sheep, goat, I believe, and I think Osiris may have actually reported some of those things last week at their orthopedic meeting showing that there was no detectible immunological reaction based on both cellular as well as antibody ?? cell-based, as well as antibody-based analyses.  But I think that's important to keep.  I think having a database of MHC pheontype I think is important.

            DR. RAO:  And would you say that the MHC data even that exists may be specific to that specific class of cells, but can't be generalized to other stem cell populations or any other population?  

            DR. TUAN:  Very good question.  The thing is that once they introduce ?? if I remember correctly, I don't know if there's anybody from Osiris, but the point is, I think in that study they were trying to correct meniscal defects, so the cells were simply injected into the joint.  And other studies have shown that when you do that, the cells actually will differentiate into various type of cell types, various cells.  And if so, somewhere along the way, they are  no longer MSCs; but yet, there was no detectible immunological reaction.  So again, one can make the argument that say a chondrocyte that came from an MSC is not the same as a articular chondrocyte.  But the fact is that remains, at the time you analyze it, they've become something else already, so that information suggests that it is important to keep that in mind, because it's an important issue.

            DR. RAO:  So here's another question.  At least, when I think of cartilage, and I think of autoantibodies and sort of arthritis that can occur, and that's because you're exposed to antigens which you would not normally be exposed to, and that often happens when cells die or they are damaged in some fashion.  So is there any logic or reason to be thinking of looking for autoantibodies, or even when you do animal studies to look for antibodies that react to transplanted cells in any fashion?  Is that something of concern?  Is that something one should be considering as an additional or a different issue that one needs to worry about when one does allogeneic as opposed to autografts?

            DR. TUAN:  Just another comment on that is that some of the degenerative joint diseases, of course, are caused by autoimmunity to say some extracellular matrix molecules that are associated with chondrocytes, so in that regard, that could be a matter of some concern if you are going to place a construct that is going to make exuberant amount of cartilage which was missing not too long ago.  Would it be possible that this will elicit another round of reactivity by the host?  That's a possibility.

            DR. RAO:  Dr. Coutts.

            DR. COUTTS:  There were some studies in the past done on autografting of cartilage bone constructs into dogs that Jay Rodrigo did.  Now he was transplanting combined tissue, it was bone and cartilage, but he looked for evidence of antibody production and reaction to this tissue.  And he found that there was a response from the dogs that indicated that there was an immune reaction to the cartilage and bone.  Now we know that the bone will elicit that, but it would lead to a breakdown of these transplants.

            This has not been the case in humans.  And I don't know whether Alan Gross has looked for antibodies in his patients.  He's a Canadian surgeon who has done a lot of osteochondral allografts, but somebody did show his data in terms of survival, and they do break down slowly over time.  And it's thought that some of that may be an immune mediated response to the cartilage that has been implanted.  Now cartilage has typically been described as being immuno-privileged, that it doesn't generate that type of response, but I think there may still be some openness to that question.

            DR. RAO:  I mean, we all thought many places in the body were immuno-privileged, and it's turned out not to be the case. 

            Did you also say that when you did some of your goat transplants, you saw pockets of lymphoid infiltration in some of it?

            DR. COUTTS:  I think that was ?? Alan Nixon said that. 

            DR. McILWRAITH:  He'd done that in the horse.  They saw that in the horse with allogeneic chondrocytes.

            DR. COUTTS:  Right.  But those were cells that have been transplanted.  That was not in tact cartilage. 

            DR. TOMFORD:  Let me just say that collagen will occasionally stimulate an antigenic response.  There are few people around who have responses to bovine collagen products that are on the market right now.  So if these cells produce a collagen that's somewhat different from that of the host, we might see some immune responses to those, so I think it would certainly be worth analyzing the fusions from the joint, things like that.

            DR. RAO:  Go ahead, Dr. Harlan.

            DR. HARLAN:  Well, just as you well know, there is an animal model for inflammatory arthritis, to immunize Freund's adjuvant with collagen, and so we know that this occurs in animal models.  And I think the trouble with being in the clinical setting, we don't know who might be predisposed or what the antibodies would be, so maybe just an FDA mandate that  recipients of cells have sera stored.  I don't even know what we'd measure, but if a clinical situation came up, you'd like to have pre and post sera to look for antibodies.

            DR. RAO:  So given the relative paucity of information, is it fair to say that the limited available data does suggest that these cells are not immune privileged in any major fashion, and that there is likelihood that they could induce some kind of immune response?  And that it should be measured, and there should be some way to be able to assess it?  Go ahead.

            DR. McPHERSON:  Sir, could I make a comment; from Genzyme Corporation.  There's a tremendous amount of human clinical experience with bovine and porcine collage-based materials, both collagen hemostats, collagen membranes that are used for fascia replacement, injectable collagen in the face, so millions and millions of people have been treated with collagen-based biomaterials.  And a small portion of these patients, and immune response does evolve, and it's been well-characterized in the literature.

            At one point in time, in the late 1980s, there was concern that these antibodies might cross-react with human collagen and induce RA or in the specific context of injectable collagen in the face, there was concern about induction of dermatomyositis or polymyositis.  And there was a very, very long investigation by FDA to see whether the frequency of these diseases in patients that were treated with collagen-based biomaterials was any different than the population as a whole.  And the conclusion of that by the FDA is there was no significant difference, so this ground has been well-tread.  It's been investigated in great detail, and I think while you will see antibodies in some patients that are given bovine collagen or porcine collagen, the clinical implications of that are, at best, not particularly provocative.

            DR. RAO:  Yes, I agree.  I mean, but here the issue was just can one, if you have antigens expressed on a cell type, it's not necessarily collagen.  Right?  Is there a cause for concern, or is there a reason to be able to monitor them?  My understanding of this is that collagen degrades when you put it in, and so it's, therefore, a relatively short period of time.  While here we are imagining cells to be present for a really long period of time, and being able to present in a different way in terms of an antigen than what you would have with inert extracellular matrix, as well.

            DR. McPHERSON:  But most of the transplant immunology issues are around differences in match in MHC Class 1 and Class 2.  While there are polymorphisms that are different between different individuals, I think that again there's not much evidence to suggest that's the major basis for histological mismatch in the context of transplantation.  So while it's fair to ask the question, are there likely to be in an allogeneic setting mismatches?  Almost certainly that's the case.  It's seen all the time in bone marrow transplants, and in kidney transplants, and cardiac transplants, and so forth.  But again, that has been pretty well studied in transplant immunology, and the implications of that, I think, are reasonably well understood.

            DR. ALLAN:  So the question that we're addressing is whether is not humans, but is allogeneic model, animal model for allogeneic.  And so I guess what you're getting at is, we don't have enough information on allogeneic transplantation of chondrocytes or some related product even in humans, and so I guess the question I ask is, what animal model system?  We talked about Osiris, and they used animal models.  Is that right?  And what animal models did they use?

            DR. TUAN:  It was a sheep or goat model.  I don't remember exactly.

            DR. ALLAN:  So there are some studies that are similar to that.

            DR. TUAN:  Yes.

            DR. ALLAN:  I mean, I would suggest that based on the question, that indeed you would want to use an animal model system to be looking at this issue.

            DR. RAO:  Would the surgeons have any specific thing to add to that?  Do you feel they have a sense of where the committee is at in terms of allogeneic versus ?? I mean, it seems to me pretty clear that people think that there will be differences between autografts and allogeneic cells.  And there will be some concern that you have to keep in mind.

            DR. MOOS:  I think we recognize that there is this perception of relative immune privilege, and have some questions about how absolute that is.  And I guess that our perception is exactly correct, there remain these questions.  And I think that's about as far as we'd be able to take it.

            DR. RAO:  I think we should add two things; that MCSs may be different from other cells, so we can't generalize, even if there was spectacular data that MCSs are immune privileged, it may not be true for most of the cells, including chondrocytes in terms of being able to do that.  And that there would be some issue which is, I think, maybe of potential concern, but there's no data on that; would be, if you have cells which die, which have a small mismatch, would there be antibodies against those particular cell types.

            DR. MOOS:  So it would probably be reasonable for whoever does this first, at least, to do some sort of small qualification study.

            DR. LUYTEN:  Just one small comment.  There have been recently some data released that MSCs are, indeed, immuno suppressive, and one would consider that as a positive element.  On the other side, by using animal models, they have demonstrated that as a result of that, tumor growth was going faster.  So there is a concern that the defense against, for instance, tumor growth might be affected, and that the MSCs might have a very negative role by their suppression.

            DR. RAO:  MSCs are special. 

            DR. BLAZAR:  I was just going to say on the tumor growth, that was when they put MSCs into an animal at the same site as a tumor that was growing, so I'm not sure that that, for a local cartilage implantation, is necessarily relevant to a systemic immune response.

            DR. RAO:  So if that covers that ?? go ahead.

            MR. NIXON:  There's one other thing I'd just like to add; and that is, there may be a difference in the immune response, albeit minor response, to allograft chondrocytes in a full thickness defect that penetrates the subchondral plate in the open marrow spaces versus a partial thickness where you have calcified cartilage in tact.  And in that circumstance, I think you're in a much more isolated environment, and you stand a much reduced chance of seeing an immune response.

            DR. RAO:  That's a really important point.

            DR. LUYTEN:  Yes and no, because in rheumatoid arthritis, the outgrowth from the pilus which is growing over the cartilage is only happening when the cartilage is still there.  When you remove the cartilage, you have much less proliferation of the pilus going out from synovium, so although there is no contact with the subchondral matter, we have to be careful for the synovial, there have been reactions with the cartilage as a target tissue.

            DR. RAO:  That's a fair statement, but the confusion one can draw is that there are differences depending on whether it's full thickness or partial thickness, and the exact condition in which you're going to use the cells. 

            Let's go to Question C - that's 2-C.  And the reason I want to do that is I think one part of it, some aspects of it we've covered already.  And that's traditionally in vivo toxicology studies include measures of systemic toxicity, such as clinical pathology tests and histopathology of major organs.  Is this approach warranted for toxicology studies with the following categories of products?  It lists cellular products and then modified cellular products.  And I think the FDA had a reason for separating them.  And to me, this is a very practical issue if you're doing the experiments. 

            We just talked a little bit earlier about saying large animal models are really important, and it would be reasonable to do it.  And one can imagine trying to do 100 micron thick sections on every tenth section in a large animal, and saying well, we have to look at all organs.  And I think that would be a little hard to imagine anybody would be willing to do it.  And I'm going to ask somebody who really does this to make a comment on this first.  Dr. Allen, and then maybe we can take it from there.

            DR. ALLEN:  I think you're exactly right.  The practicalities of having a large animal model that is both specific for your condition, also sensitive to the potential complications, however rare, is an unlikely mixture.  So our approach is always whenever we've done this, I mean there's perfect rationale when you're doing a toxicology study on a drug, which is systemically distributed to loc all around the body, makes perfect sense.  But when you're dealing with an potentially injectable or an implantable material that's going into a joint, I think certainly in the large animals you're not going to have high sensitivity for detecting systemic effects.  So our approach has always been for a local implantable device that's not releasing anything, to do the local tissue response.  And in the context of a joint, we would take local draining lymph nodes based on regional distribution, and any tissue, any major organ that shows any signs of pathology would be sampled.  And depending on who the sponsor is and what their interest is, we may take and archive parenchymal organs as a matter of routine, and keep some serum bank, but that's about it.  That would be if it was cellular product.

            If it was a modified cellular product, I think you're in a whole different game.

            DR. RAO:  Hold that on the modified cellular.  So if you're just thinking of cellular products, you just heard an outline of what might be done.  Would anybody think that that's too little, or that's too much?  That's good.  There seems to be some consensus here.

            DR. MOOS:  We can live with that.

            DR. RAO:  So maybe we can think of this as a somewhat more general statement, and see whether that seems to make sense to the committee.  If you're implanting cells or a defined product which might include scaffold, and you're looking at that in a local tissue, and you have a defined number which is relatively small compared to what people do in bone marrow or anything else, that the chances of having reasonable measures which would detect systemic pathology don't really exist.  And what one needs to be concerned about is looking at sentinels or what would be local, so you look at local tissue or local lymph nodes which drain from that.  And that will be possibly the most reasonable thing to do.

            If one was concerned over specific affect on these cells, then one may want to consider archiving tissue samples, much like we said maybe keeping a record of MHC antigen expression.  Does that seem to the FDA to answer the question?  There certainly seems to be a consensus that it's actually feasible.

            DR. McFARLAND:  That seems reasonable.

            DR. RAO:  Then let's get back to the second part, which was a whole different ball game.  Maybe I can ask Dr. Allen to continue with his thoughts on that.

            DR. ALLEN:  This is the one that goes south.  Cellular products, I think one of the key issues practically speaking from animal models is, when you're dealing with say a BMP.  Say a BMP has been inserted, it would be nice to know what recombinant purified protein over expression does systemically, so if you've got that as a guidance for you, it can help you.  But I think when you're dealing with anything that's producing a diffusible agent that's bioactive, then that's when I would want to do much more aggressive sampling.  But it would really, I think, have to be guided by what the toxicology profile is for that particular agent, if we know what it is.  And that's sort of an approach we've been taking with bone bioactive agents, is to look at the safety profile for the agent, and then sample accordingly.  So this is a bit more, I think, of a ?? but it's certainly one where I would not just want to sit on local tissues.

            DR. RAO:  Dr. Allan.

            DR. ALLAN:  Yes.  I'll come back to what I was saying earlier.  If we're at 2(c)2, which is I think we are, modified cellular products, now we've elevated.  And I would even suggest that you might even consider going to non-human primates for this, because of the nature of what you're probably going to be expressing in human cells.  And although monkeys are not humans, many of the growth factors work in non-human primates.  They're very closely related in that sense, and so you might get a sense of if things are going to go haywire, that they may go haywire in monkeys.  And it could be at a distant site, you could have a lymphoid tumor or something else related to the modified cellular product, so that's what I would be leaning for.

            DR. RAO:  Maybe I'll try ?? that may be a specific comment.

            DR. MOOS:  Just a quick point there, which is that if may depend ?? on the point of going to non-human primates, it may depend somewhat on the biology of the particular transgene.  Certain kinds of cytokines of exquisite specie selectivity and you might need a non-human primate.  Others, I think, I just heard mention of the BMPs go across most phyla, and it might be less of an issue.  And so there is some room for discretion, I might suggest.

            DR. RAO:  Dr. Nixon.

            MR. NIXON:  If I might muddy the water even more and suggest that there are recombinant peptides essentially not in the same category at all as any of the recombinant ?? any of the transgene induced changes.  And that even from that standpoint, the use of various vectors, to include retroviral and lentiviral are much more of a concern than the use of other vectors; particularly, ADNO or some of the AAV group - well, let me say particularly AAV group.  So I think in that circumstance, you may have tailor the sorts of tissues you have as some potentially uses of sentinel structures like the synovial membrane, which for us is the major indicator of any particular immune response, or any particular toxic response within the joint itself.  It's really very much marked at the synovial membrane, or draining lymph nodes if you want to go a little farther downstream.  But I think from that standpoint, if we were using a recombinant protein in a cartilage repair system, I don't think you need to do any specifically enhanced toxicological studies other than the sort of regional lymph nodes and the synovial structures.

            DR. RAO:  So I'll tell you our concern, given we don't work in cartilage, but just looking at what's happened in the nervous system, for example.  Everybody thought well, you give a growth factor, and it's going to be behind the blood-brain barrier.  You're not going to really see any systemic effects.  Maybe you'll see something in the CSF, and you may be able to look at that, and that's what it will be.  And you're putting really only a small number of cells.  And people have done that, and the net result was that they got pretty obvious clear-cut systemic effects, which were significant enough to call for cessation of therapy.  So to me, it seems that there's evidence that in systems where you're giving a recombinant product, as long as the levels are high, and as people pointed out, depending on what we know about its biology, we will have to look systemically, unless there is clear-cut data which proves otherwise.  And that would be my sense, and it would worthwhile to ask if other people would agree that that's their sense from what's been done in systems.  Dr. Harlan.

            DR. HARLAN:  I just agree.

            DR. TUAN:  I agree, too.  I think if you're going to be producing a recombinant product, I think you're going to have to follow the same guidelines used in any of the recombinant factor administration and gene therapy guidelines, which is systemic toxicity, vis a vis ??

            DR. RAO:  Including generation of autoantibodies, cross-reactivity.

            DR. TUAN:  Exactly.  Yes, even transgenerational-type thing, because I remember in the case of BMP, you worry about transplacental transport, et cetera, et cetera.  So I mean, it depends on the amount, obviously.  It's one of the first things, figure out the local concentration and systemic local concentration, and then use what is known in the literature to make the best judgment.

            DR. RAO:  Dr. Coutts.

            DR. COUTTS:  Well, just sort of a devil's advocate with regards to this.  If we're talking about a cartilage tissue replacement, and that the cells that have been transfected and are expressing this gene that has been put into it, it doesn't exactly have ready access to the vascular system.  It's going to have to get out of the cell, into the extracellular matrix, and then it's going to have to get squeezed in and out through joint action, and into the synovial fluid, and then it's going to have to be absorbed by the synovial membrane, and ultimately get into the blood supply in that fashion, or maybe into the lymph drainage.  So it has a somewhat circuitous route to take to get systemic.  And I'm not saying that it wouldn't, but just that it's tougher for it.  And I would tend to support Allen Nixon's perspective, that I would still be more concerned locally than I would be systemically.

            DR. RAO:  I absolutely agree, and maybe I should make that emphasis clear, is that one can't assume there won't be systemic effects, and one has to be willing to monitor those, if one uses recombinants.  Does that ??

            DR. McFARLAND:  Yes, I think we can work with this.  I think it's a very rational suggestion.

            DR. RAO:  Now we come to somewhat harder and more nebulous things, I think.  And that was let's provide specific comments on the following with respect to pivotal animal toxicology studies, so let's keep that in mind, that we're considering a toxicology study that is designed to support a clinical trial of a cellular cartilage repair product, so it's a cartilage repair product.  And what animal models and study duration are needed to support exploratory clinical trials?

            And we did get a little to that from our earlier discussion where people said that well, at least in looking at the cells in animal models, we need to look at between four months to twelve months, at least.  The question is, is that a reasonable time limit, as well, for toxicology studies?  Should it be longer, should it be shorter, or it doesn't seem that there's any obvious way to make that kind of decision?  Go ahead.

            DR. ALLEN:  In the context of secretory proteins and things, it seems like the first question is ?? it seems there's some pretty reasonable data that these cells don't survive in great numbers for a very long period of time, so it would be interesting to know how long we really do have these cells metabolically, proliferatively active inside the lesion before predicting when one would want to look for effects.  It seems that a lot of the effects is probably an induction of local tissues rather than necessarily the cells that you put in initially actually being proliferative and repairing the lesion, so I guess I'd want some clarification on what the cells are doing before trying to determine how long to look for bad effects of those cells.

            DR. RAO:  So if the cells are surviving for 18 months, let's say just a number off the top of my head, would you want to look at least for 18 months, or would you want to look at nine months, or do you say I want to look for one-year after that?  I mean, just in general.

            DR. ALLEN:  Well, I'd like to know what the number is.  I mean, that's the problem, there's a balance between what you might feel safe doing, and what would be optimal to do.  I mean, I think these pivotal pre-clinical studies have to be of a reasonable duration, as in both reasonable enough to see an effect, and reasonable enough that a sponsor is going to do it.  So I would say 18 months seems a long, long time to me, when ultimately our real experimental animal is the human under controlled settings.

            DR. RAO:  And in the light of there being some data in humans, as well, would that affect

your ??

            DR. ALLEN:  Yes, absolutely.  

            DR. RAO:  What would the clinicians feel?  Dr.  Coutts.

            DR. COUTTS:  I think it depends on what product we're talking about. 

            DR. RAO:  Consider just cartilage cell product.

            DR. COUTTS:  Not modified, no cellular manipulation.

            DR. RAO:  Yes, short passage, not modified, no transgenic gene.

            DR. COUTTS:  I'm not aware that there's anything toxic coming from these products.  We do have some experience, again thanks to Genzyme.  They've been implanting cells of this nature, and there has been no evidence that there's a toxic effect from this.  I'm not even sure what I would measure from these cells.  Do you have any idea what to measure?

            MR. O'CALLAGHAN:  Yes.  Michael O'Callaghan, Vice President of Pre-clinical Biology at Genzyme.  We've debated this question quite a lot, not just on cartilage repair, doing other systems using cells.  And we've sort of come up with a sort of a general rule of a three-day, three-week, three-months sort of scenario where you for any acute effect in the synovium in particular, in that first immediate period that might be related to something in the product, or might not be the cells themselves, but maybe something else that's noxious.  And in three-weeks to sort of look at a general change in that cellular response population, because there will be some response.  And in three-months, to sort of understand whether there's any immune responses that we can get a handle on.

            We don't feel that there's that much difference between three months and any of the longer periods in terms of the toxicity, so that's our sort of general approach.

            DR. RAO:  So a toxicity of like three to four months is what ??

            MR. O'CALLAGHAN:  We believe that's a reasonable approach.  Now there may be other opinions, but we're still trying to work to do this in the most efficient manner, and look for changes that are relevant, and to try and understand just exactly what is going on from a pathophysiological point of view, because some reaction is expected.  And we're trying to understand whether that reaction is really relevant or not.

            DR. RAO:  And you did say something about just looking mostly locally initially in a short time period?

            MR. O'CALLAGHAN:  Particularly locally at the synovium and draining lymph nodes, as was mentioned.  We would generally in those cases reserve tissue, bank the tissue from most of the organs anyway, and if we had any questions, we might do a full tox histopathology on a small animal.  We would prefer to do this in a small animal, in the first instance.  If there's anything that arises, we might then look at a larger animal.  We would prefer to do that in small animals to keep it to a reasonable scale, and also to get some statistical significance.

            DR. RAO:  Dr. Tomford, would you have anything to add to that? 

            DR. COUTTS:  Just a follow-up question.  I mean, basically so what you would be looking for would be a reaction to the cells themselves, and something of an immune response, or evidence of lymphoid tissue indicating that there was something noxious there, that it was trying to get rid of?

            MR. O'CALLAGHAN:  I think we take a more general sense here.  WE're trying to really understand what response you might have to cells and anything else that might be in the product that is delivered.  And within a few days, if there's something really serious, you will see it.  At about three weeks, this is only the reason for those types, at three weeks you'll be able to see the waning acute response, maybe anything that's chronic that's starting to develop.  And at three months, you should be able to see whether there's any delayed types of effects, like autoimmunity or something like that.

            DR. COUTTS:  But you're looking for an inflammatory response.

            MR. O'CALLAGHAN:  Well, a variety of things.  We're looking for any modification to morphology, histologic morphology that might indicate something else, as well.  Like we were looking for indicators of problems.

            DR. COUTTS:  It's easier to find things if you know what you're looking for.

            MR. O'CALLAGHAN:  Of course.  But one of the problems in preclinical biology is that you often are looking for those serious things that might later be a problem, and so we always have to keep that skeptical approach.

            DR. RAO:  Dr. Murray.

            DR. MURRAY:  Another question for you.  With small animals with a relatively short life span, so for a mouse with a life span of two-years, three months is an eighth of its life.

            MR. O'CALLAGHAN:  Correct.  Small animals is rabbits, essentially.

            DR. MURRAY:  Rabbits.  Okay.  Is there any reason to think one might want to expand those time horizons in larger and more long-lived animals?

            MR. O'CALLAGHAN:  Perhaps, but I think if you're looking at truly acute and early chronic changes, we don't feel that the scale of the animal really matters.  Now there may be different opinions on that, but we feel that an animal like the rabbit is probably suitable for that.

            Now, obviously, when you get to looking at efficacy and all of these other proof of concept, it's a little different.  But for these toxicity issues, we don't feel necessarily, unless there's something that turns up, that a large animal is necessary.

            DR. RAO:  Thanks.  Let me ask the committee the same question in a slightly different way.  So the sense I see is that the tox studies, as far as one can tell, there doesn't seem to be any reason for them to be any longer than what one is looking at in terms of the cells and their function in an animal model, which is between four to twelve months is what we've talked about.  Is this going to be true for most cell types, or should we really be wondering and saying well, I don't know what's this mixture of my cells, so maybe it'll be different in some other cell type, or can we generalize in some fashion and say, if we are putting cells in, and we're putting them into a cartilage and it's defective for certain type and a certain class of patients, and we are using this animal model which is similar to that by whatever justification one has, that this would be probably true?  So I could compare mesenchymal cells with articular cartilage and so on, or whether this would be well, I don't know what my mixture of mesenchymal cells is, because I may have this rare population which may do something different, and so my tox studies will be different.

            Is there any reason a priori to suspect that there will be some major difference that one needs to consider, or have a red flag or something that the FDA needs to keep in mind; that well, this is true for most cells.  It's not going to be true if you have a mixture which is completely ill-defined.  I know it's a hard question, but go ahead, Dr. Harlan.

            DR. HARLAN:  Well, you set up this whole discussion by saying low passage number of chondrocytes, and everybody has answered to that.  The further you get from that, the higher concern I have.

            DR. RAO:  Exactly.  That's why I wanted to separate them and just see whether that would be ?? so there would be concerns.  So, Dr. Nixon, did you have something to add?

            MR. NIXON:  Well, we were laughing because you're basically talking about moving the stem cells, and I guess there's one thing that's true about stem cells, there really isn't any predicting what stem cells are going to do.  So if now we're transplanting stem cells, mes cells or some sort, then there may be a need for greater exploration of the cells, where they target, how they leave the joint, where they go when they do leave the joint.  Because as compared to chondrocytes, which apparently from the work we've been looking at synovium membrane, don't seem to depart on this peripatetic trip around the body; whereas, the stem cell has the capability of doing just that, or inducing some other immune recognition cells, which then will turn up in opposite limbs or structures in the opposite portion of the body.  So I think there may be a cell issue here as to which cell you're using, and whether you need to explore more.  I think most of the studies are going to finish up having a toxicologic study done at some point on a small animal, a rabbit or whatever we've just heard about, and then as you move up in the species, and the size, and the defect, and the challenges of increased size of defect healing, less and less focus on toxicology, and more and more focus on efficacy.  So I think my earlier comments were really not addressed to whether we need to do toxicology at all just because of a cartilage cell implant, so much as you get up into the higher efficacy studies, whether you need to repeat these extensive toxicological studies.  And in that situation, the sentinels of the synovium membrane and the draining nodes, if you wish, are perfectly enough in my mind.

            DR. RAO:  Thank you.  I think you summarized it really well in terms of what I think is really important to annunciate, is that when one offers advice to the FDA, one wants to try and be as specific as possible.  But by the same token, we can't just say you can only do this for this, and some sense this is our opinion, so we want to see what the boundaries of extrapolation are.  And you clearly pointed out that this is true for certain classes of cells, but you have to have caution before you can extrapolate to any other cell type, particularly for cell types which have known properties which might merit different sorts of attention.

            Does that in some sense answer most of the preclinical questions that the FDA raised?

            DR. McFARLAND:  Yes.  I think the committee has done a wonderful job answering the huge majority of the questions directly with something that we can work with.  And we appreciate that considerably.  A couple of comments.

            One is that one thing we didn't discuss in much detail, but I don't know that we necessarily have to, but I want to point it out; is that, some of these products are cells with matrix and scaffolds, and the like.  And so the same concerns, I think that the Center for Devices would have if it were a scaffold alone, we would have as a combination product. 

            The other thing is, in this last discussion about toxicology, I think our discussion doesn't preclude the use of collecting some toxicologic endpoints, and the proof-of-concept biological activity studies in large animals so that you can potentially answer two questions at once with those long-term data.

            DR. RAO:  That's great.  And if the committee has no other additional questions or comments on this specific topic, we can move on to the next set of questions, which we should try and address, after we hear from a couple of speakers who should summarize the field for us.  Dr. Nixon, do you have a comment?

            MR. NIXON:  No.  I apologize.

            DR. RAO:  So we will hear from Dr. Susan Leibenhaut on the FDA perspective on the development of cellular therapies for repair and regeneration of the joint surfaces and the clinical aspects.

            DR. LEIBENHAUT:  Hi.  As Dr. Rao said, I'm Susan Leibenhaut, and I'm a medical officer in the Officer of Cellular Tissue and Gene Therapies, and I'm here to provide a brief overview of the major regulatory concerns in the clinical development of cellular products for repair and regeneration of joint surfaces.

            First, I will note that the major purpose of our clinical discussion today and provide information concerning some basic regulatory concepts.  Secondly, I will highlight the major regulatory considerations and clinical study design issues in the clinical development of the products under discussion with emphasis on the use of these products in the knee.

            The purpose of our clinical discussion today is highlighted here in the first bullet on this slide.  Today we hope to obtain insight that may be useful in the design, conduct, and analysis of clinical studies for cellular products proposed to be used in repair or regeneration of articular cartilage.

            The next bullet on this slide notes an important caveat for our discussions.  We are not asking any questions that pertain to a specific cellular product.  The presentations here serve as an example of types of data that have been used in clinical studies for similar indications.  Our discussions may be enlightened by this information, but the information presented should not form the substantive basis for decision-making on any specific product.

            This slide highlights the observation that the clinical development of a product may be broadly divided into two categories; exploratory and confirmatory.  The first bullet is exploratory studies, the first stage of clinical development.  Exploratory studies, also commonly referred to as Phase 1 and 2 studies, or pilot studies, help generate a specific hypothesis.  This hypothesis is then tested definitively in confirmatory clinical studies, which is the second bullet on the slide.  These studies are often referred to as Phase 3 or pivotal studies. 

            Exploratory studies are supported by data from product and animal testing.  They are designed to detect major safety concerns, determine feasibility, and work out issues such as optimal surgical technique, and provide information such as potential range of concentrations in doses of the clinical product for use in humans. Exploratory studies may use a variety of study design features, including uncontrolled design and dose ranging studies, and may use an early version of the investigational product.

            Confirmatory clinical studies must provide scientifically convincing data demonstrating clinical benefit and provide an accurate risk-benefit assessment of the product.  They should provide adequate data to support a description for directions for use of the product in market labeling. 

            Two important elements in design of a confirmatory clinical study are the choice of endpoints or clinical outcomes, and the choice of control group.  Clinical outcomes or endpoints must provide a measure of clinical benefit.  IN order to determine the duration of the clinical benefit, the times at which the endpoints are measured are important considerations. Certain endpoints may not measure clinical benefit directly.  Instead, they measure outcomes that may correlate with clinical benefit or predict the benefit.

            Some of the most commonly cited clinical outcomes regarded as important for cartilage repair and regeneration products include measures of changes in patient signs and symptoms, such as pain and function.  Similarly, physical examination changes, such as changing in walking times and distance may be indicative of clinical benefit.  We believe that measures of clinical outcomes are important because they are direct measure of clinical benefit. 

            The last bullet on the slide is to emphasize that the clinical benefit should be demonstrated to be maintained over a clinically important duration of time. 

            On this slide is a list of some outcome measures that are potentially important, but are not generally regarded as a direct measure of clinical benefit.  Instead, they measure outcomes that may correlate with clinical benefit or predict clinical benefit, but are not in themselves a measure of clinical benefit.

            The first bullet refers to the appearance of the joint using non-invasive imaging techniques, such as MRI.  The second bullet refers to joint surface appearance or other measure, or description of cartilage using an invasive technique, such as arthroscopy.  And the third bullet refers to the histological appearance of a biopsy specimen, a biopsy being more invasive than the first two procedures.  The last bullet on the slide is to emphasize  that the timing of the evaluation is an important aspect of the ability to obtain meaningful data.

            Another important component of a clinical study is the choice of control group.  An appropriate control group is required in order to discriminate outcomes caused by the study agent from outcomes due to other factors, such as the natural progression of the disease, observer and patient expectations, and other concomitant conditions and treatments.

            The choice of a control group is one of the major challenges in the design of confirmatory clinical studies.  This slide highlights some of the type of control groups.  As shown here, control groups may be broadly grouped into two major types; concurrent controls, and non-concurrent controls. 

            Some examples of concurrent control groups are cited by the three green bullets.  The first is active treatment control group.  For example, a study that compares a new joint repair product to an established procedure, such as microfracture, or to a currently marketed product. 

            The next is a placebo control, which in the case of products that we are discussing today may be a sham surgery control.  This type of control is sometimes considered problematic in the setting of orthopedic surgery.  An additional type of concurrent control design is different dose groups, where the study intent is usually to show a dose response effect. 

            The non-concurrent control is limited to what is commonly referred to as historical control.  The next slide highlights some of the important aspects of control groups in the clinical study of cellular products for cartilage repair and regeneration.

            For the study of cartilage products, control groups should have comparable demographics.  The nature of the lesions and the concomitant conditions should also be comparable, and treatment such as perioperative care and rehabilitation should be standardized between groups.

            The last bullet on this slide notes that the use of historical or external controls limits the ability to ensure comparability between active and control groups.  The goal of attaining comparability between groups is most readily achieved through the use of randomized studies that use concurrent control as shown on the next slide.

            Randomized control studies have the greatest ability to demonstrate valid differences between groups because of the minimization of bias associated with the use of randomization.  The use of blind evaluators, as highlighted in the second green bullet, exemplifies another design feature that a randomized control study may use to minimize bias.  As previously noted, historical controls or external controls are not sufficient to control bias due to the inherent selective nature of the controls.

            In summary, the purpose of the clinical discussion is to obtain insight from the committee regarding the elements of confirmatory clinical design that I have discussed above.  Before I introduce our speakers, I will provide a summary of the clinical questions.

            The committee will be asked to discuss the extent to which each of the outcome measures listed on this slide would be considered a clear, meaningful measure of clinical benefit.  These are some of the possible outcome measures that I mentioned previously; joint function, pain, clinical examination findings, joint surface appearance, histological appearance of a biopsy specimen, appearance of the joint on imaging such as MRI, or other non-invasive techniques.

            Next, the committee will be asked to discuss the importance and limitations of certain aspects of clinical study design, and highlight those situations where flexibility may be acceptable, and identify any ancillary considerations that might optimize study design.  These study design elements are the nature of the control group, the importance of blinding procedures, and the duration of the clinical studies.

            The next two speakers will discuss some of the endpoint measures that have been used in studies of musculoskeletal conditions that affect pain and function.  These measures have been studied extensively in osteoarthritis, and may have applicability to current and future studies in cellular products for cartilage repair and regeneration.

            The first speaker will be Dr. Marc Hochberg from the University of Maryland, to discuss clinical outcome measures.  And the second speaker will be Dr. Charles Peterfy from Synarc, who will discuss perspectives on imaging outcomes for cartilage repair.

            DR. RAO:  Before Dr. Hochberg starts, I'd like to take this opportunity to ask two members of the panel who have just joined to introduce themselves. 

            DR. RIEVES:  Hi, there.  This is Dwaine Rieves.  I'm Clinical Branch Chief in CBER's Office of Cell Tissue and Gene Therapy.

            DR. DAWISHA:  I'm Dr. Dawisha.  I'm a Medical Officer for the Center for Devices and Radiological Health, Office of Device Evaluation.

            DR. RAO:  Welcome. 

            DR. HOCHBERG:  Thank you very much.  Can everybody hear me all right?  Okay.  So I was asked to talk on perspectives on clinical outcomes of studies of products for use in cartilage repair.  And I'm going to approach this from the point of view of starting with osteoarthritis, which I think is what we probably want to try and prevent in treating patients who have focal cartilage defects with the devices and techniques which have been discussed somewhat this morning, and will be discussed later on today, and tomorrow morning.

            So I thought I would first review the current draft guidance document for the development of products for the treatment of osteoarthritis.  And in particular, look at the potential indication for prevention of the development of osteoarthritis, which is a potential indication for a product, or drug, or device, and then talk about some newer methods for measuring symptomatic outcomes.  And these include state measures which allow you to categorize individuals as either being better or not better, or being good or not good. 

            And then talk a little about the KOOS, or the Knee Injury and Osteoarthritis Outcome Score, which has been used not only in osteoarthritis, but also has been used in the area of cartilage repair.  And then Dr. Peterfy, I'm sure, will be talking about newer methods for measuring structural outcomes, so I won't touch that.

            So let's start with the draft guidance, which was published in July of 1999, and is available to you on the web.  And it outlines three potential claims for osteoarthritis.  First being symptomatic treatment of pain and function, the second being delay in structural progression, which I am not going to talk about.  But if you want to hear a talk on structural progression, I would suggest that you stay in town and tomorrow afternoon those of you who are not involved in the discussion might want to come to the American College of Rheumatology course on innovative therapies which is going to be held downtown.  There's a session on cartilage and bone tomorrow afternoon.  And then the third is the prevention of the occurrence of osteoarthritis.

            Now the issues with regard to symptomatic treatment of pain and function as an indication of osteoarthritis are somewhat relevant to the human experience here with cartilage defects, because the efficacy endpoints look at pain and function, which were just mentioned as two of the clinically relevant endpoints that we should think about measuring.

            So there have been several groups which have proposed recommendations for outcomes measures.  And the sort of alphabetical letters up here, OARSI, stands for the Osteoarthritis Research Society International, and OMERACT stands for Outcome Measures in Rheumatology Clinical Trials, which is an international group that meets under the umbrella of the International League of Associations of Rheumatology and the World Health Organization.  So these groups have met and as many as about eight years ago decided on some recommendations for outcome measures in osteoarthritis trials.

            They suggested that the measurement of pain and the measurement of function should be disaggregated, so that there should be separate scales to measure pain, and separate scales to measure function.  And we may see that when we look at some of the outcome measures that have been used in let's say athletic individuals, that these, in fact, don't disaggregate these domains, some do.

            Now patient global assessment should be considered as an outcome measure.  And that in clinical trials of osteoarthritis, there should be measurements of structure if the trial lasts 12 months or longer, and this was for purposes of risk-benefit assessment; although, it sort of blends into the issue of looking at structural progression.

            And at the time, the recommendation was to get x-rays.  Now the recommendation might be to get magnetic resonance imaging, instead of x-rays.  And then there were some comments about effects on non-signal joints, which I won't go into.  So I'm going to skip delay in structural progression, and go to the prevention of the occurrence of osteoarthritis, which seems like a reasonable long-term goal in the treatment of patients with focal cartilage defects.  And this is defined as preventing the occurrence of incident symptomatic osteoarthritis using both clinical and imaging criteria. 

            For the purposes of the discussion today, would represent new joints in persons at risk for osteoarthritis, since we recognize that people with knee injury are at risk for developing knee osteoarthritis. 

            So how do we measure clinical outcomes in patients with osteoarthritis for purposes of studies?  We most often use the Western Ontario MacMaster, abbreviated WOMAC, Osteoarthritis Index.  And used to but don't now use the Lequesne Algofunctional Index.  Michelle Lequesne is a retired French rheumatologist who developed an index which combines pain and function for the knee, and a separate one for the hip.  These were actually demonstrated to be reliable, valid, and responsive after they were published and utilized, so they sort of were studied methodologically after they were already being used.  But because they combine together pain and function, they're really not used for purposes of studies now.  And these are just three citations that support the use of these measures.

            Now to sort of move ahead as to what's new, what's new is the use of state measures, so to categorize people as either responding or non-responding, so OARSI came up with responder criteria which were published in 2000 in the Journal of Osteoarthritis and Cartilage, and these were derived from a data set which included 14 randomized placebo controlled clinical trials involving almost 2,000 patients with either hip or knee osteoarthritis.  And there are a variety of intervention studies in these trials, so some were oral interventions, some were NSAIDs, some were so-called osteoarthritis-specific drugs, and then there were some intra-articular agents.

            And there were two sets of propositions developed, Prop A and Prop B.  And the cutpoints to determine a responder differed by the proposition.  They differed by the joint that was being studied, whether it was a hip or knee.  They differed by the type of intervention, whether it was oral or intra-articular, and they differed as to whether you had high improvement on one criterion, or moderate improvement on several criterion.  So this was felt to be somewhat cumbersome.  It lacked simplicity when looking at it.

            There was a question about really the increase in precision when this was applied to the data sets in terms of being able to improve statistical power and reduce the sample size.  And almost two-thirds of the studies that were provided by the pharmaceutical companies were not included in the analysis because they didn't have all the data that were needed in order to be able to develop the algorithms. 

            So an attempt was made to try and improve this, and this is the so-called OMERACT-OARSI Responder Index, which was developed at I think the sixth OMERACT meeting, which was about two and a half years ago.  So this was published in the Journal of Rheumatology, and here the goal was to develop a simplified set of criterion.  Here the six different scenarios were compared using the original database, and then validated in a so-called revisit database, which include 15 studies with about 8,000 patients.  And then the data were presented to an expert group like those of you sitting around the table, and people voted with their hands to pick which was the best criteria or algorithm.

            And this is the best algorithm, and it's in one of your handouts.  So to be a responder, you can either have high improvement in pain or function, which is 50 percent change, and an absolute change of 20 units on a naught to one hundred scale.  So this means that in order to get into the study, you really need to have a minimum level of 40 in terms of a pain score or a function score in order to be able to be a high responder improved 20 units and 50 percent.

            If you don't meet that qualification, you can still be a responder by basically improving in two of the other three so-called primary outcomes; pain, function, and patient global assessment, so there are the two ways of becoming a responder.

            Now this has not been validated in another data set, so we actually attempted to do this, so I'll share with you these data.  They've been presented at the Annual Congress of the Osteoarthritis Research Society, but have not as yet been submitted for publication.  And this comes from a data set.  So this is the objective of this analysis, and what we did was to take a data set which was a randomized three-arm controlled trial or traditional Chinese acupuncture in patients with osteoarthritis of the knee.  The results of the trial were actually published in the Annals of Internal Medicine in December, 2004.  So there were 570 patients with symptomatic osteoarthritis of the knee enrolled in the study, and 236 or 41 percent of these patients fulfilled the OMERACT-OARSI responder index at the end of the study.  This was a six-month study.  And that's using the intent to treat population.  If you use the completer population, it's about 60 percent of the completers.

            So I'm going to show you a series of histograms here, where the people who fulfilled the responder criteria are shown in light blue, and the non-responders are shown sort of in purple.  And you can see here that the scores in normalized units, and this ranges from naught to a hundred, for those who were responders are lower than those who are non-responders.  But this is obvious because the WOMAC is used to actually define the response. 

            But if you look now at the Health Assessment Questionnaire, which is an item which measures pain on a 15 centimeter visual analog scale, and disability based on 20 questions, and a score from naught to three, where naught is having no difficulty performing any of the activities, and three as having severe difficulty performing activities in each of the eight domains.  You can see that the pain score is measured by the visual analog scale is lower in the responders than the non-responders at the end of the study, as is the measurement of disability.  And this is, in fact, a low level of difficulty in performing activities in both groups.

            If you look at Quality of Life as measured by the EruoQol, which was included in our study because it's useful for doing cost-effectiveness studies, as well, because it's a utility measure, you can see that either measured with five questions, or measured with a Visual Analog Scale reporting quality of life, those who are responders have better quality of life than those who are non-responders.  And if you use the SF-36, which was used in the one randomized comparator trial that I'll talk about at the end, you can see that the responders have higher scores on all of the eight sub-scales of the SF-36 than the non-responders.

            So this is a valid state measure of response in terms of looking at pain and function in a trial.  And actually, in separate analyses this does improve the precision of this trial so that you're able to use fewer people in terms of smaller sample size.

            Now what about newer state measures?  These were actually just published in the January issue of the Annals of The Rheumatic Diseases by Tubach and colleagues.  The Minimal Clinically Important Improvement or the MCCI, which is the smallest change  in the measurement that signifies an important improvement in a patient's symptom.  So what they did was to take patients with symptomatic knee osteoarthritis, measure them at entry, give them a non-steroidal anti-inflammatory drug for one month, and then measure them at four weeks time.  And patients had an opportunity to say at the end of the month what was their improvement, it was either none, poor, fair, good, or excellent.  So of those who were good or excellent improvement, the 75th percentile of the amount of change was picked as the minimal clinically important improvement.  So 75 percent of the people who had good or excellent improvement had this amount of improvement or more.

            The other state measure that was looked at was the Patient Acceptable Symptom State, so this is the value of the measurement of a patient symptom beyond which the patient considers herself well.  So here they asked at the end of the study at four weeks, are you satisfied with your current state of your knee osteoarthritis after your therapy?  And among those who were satisfied, this is then the 75th percentile for the score on each of these measures for those that are satisfied.  So 75 percent of people had scores at this level or below, among those that were satisfied.

            So then we applied these cutpoints, the amount of change and the satisfaction point, to our 570 patient data set, and we found that between a quarter and 40 percent of patients had a minimal clinically important improvement.  This equates to about a score of 32 on a zero to one hundred range for the WOMAC scales.  And about a third to 45 percent reached a patient acceptable symptom state, and this equates to a score below 30 at the end of the study.

            And in fact, there were highly statistically significant associations between achieving an OMERACT-OARSI response, and having either a minimal clinically important improvement, or being in a patient acceptable symptom state for each of these three domains. 

            So that sort of summarizes where the field is with regard to measurement of pain and function in studies of osteoarthritis.  There's one other state measure that I didn't present to you, which is called the BLISS, which is Bellamy's Low-Intensity Symptom State, obviously developed by Bellamy.  And you can use a BLISS cutoff of 5, 10, 15, 20, and 25, depending upon where you want to set the level of satisfaction.

            So let's move now towards the issue of focal cartilage defects.  In 2003 in Arthroscopy, Marx published a paper looking at the reliability, validity and responsiveness, sort of summarizing the data on knee rating scales that had been used in studies of athletic patients.  And these are the eight scales which were included, and in that paper there are primary citations to the papers which present the data for each of these eight scales.  This slide is actually not in your handout material.

            So the one that I'm going to talk about is the Knee Injury and Osteoarthritis Outcome Score, it's third up from the bottom.  But the others, I'm sure, are familiar to the orthopedic surgeons sitting around the table.  They are not, however, generally familiar to rheumatologists who do research on osteoarthritis.  But let me talk about the Knee Injury and Osteoarthritis Outcome Scale, which is referred to as the KOOS.

            This evaluates both short-term, as well as long-term consequences of knee injury.  This is a 42-item questionnaire which has five separately scored domains.  So it fulfills the requirement of disaggregating pain and function, so it has a pain scale.  It has function in daily living as a scale,  it has function in sport and recreation as a scale, it has other symptoms that occur in people with knee problems like stiffness, buckling, locking.  And it also has questions regarding knee-related quality of life; how cognizant are you of your knee during your daily activities?

            Now the KOOS has been validated in several populations.  It's been validated in patients who undergo surgical reconstruction of anterior cruciate ligament tears.  It's been validated in patients undergoing knee arthroscopy, in people who have had meniscetectomies 16 years previously and are now screened for osteoarthritis, in patients undergoing total knee arthroplasty, and in patients undergoing autologous cartilage transplantation.  The latter data set are available in a thesis, and they're not in the peer review literature, but the reference to the thesis is obtainable through the documentation for the KOOS. 

            Now the KOOS is reliable, it's responsive.  The effect sizes are greater than one, so this is a large effect for all five of the sub-scales in patients who undergo total knee arthroplasty and tibial osteotomy for knee osteoarthritis.  And the effect sizes are greater than 0.5 - these are moderate size effects for all five sub-scales in patients undergoing ACL reconstruction, and arthroscopic meniscetectomy .

            The KOOS actually contains all 24 questions of the WOMAC, so it expands the WOMAC.  It deals with the problems that we have with the floor effect of the WOMAC, and it will deal with the problems that orthopedic surgeons would have with the ceiling effect of the WOMAC, because we score the WOMAC in opposite directions. 

            For rheumatologists, a score of zero is the best you can have with the WOMAC.  For orthopedic surgeons, a score of 100 is the best you can have with the WOMAC.  It's just the convention.  So all the WOMAC questions are embedded within the KOOS, and that there are additional questions which deal with the issues of sport and recreational function that are not covered in the WOMAC, the knee-related quality of life, and the other symptoms.  And, indeed, the KOOS is more responsive than the WOMAC in younger subjects who have knee injuries who don't have osteoarthritis, and even more responsive in older subjects who undergo total knee arthroplasty because they can return to some recreational functions that are not captured in the WOMAC, and report these knee-related quality of life scores.

            So there are numerous options for assessing clinically relevant outcomes in trials of products used for cartilage repair, and it appears, at least by my superficial look at the literature, that the KOOS is probably the recommended self-report measure for pain, function, and quality of life, because it includes all these in disaggregated scales, and it's been shown to be very responsive in multiple populations.  So if you want to get to the KOOS, which is available in the public domain with no charge, you go to www.KOOS.nu.

            Now I mentioned the goal of preventing the occurrence of incident osteoarthritis, and the definition of incident osteoarthritis usually requires some structural change.  Dr. Peterfy will talk about magnetic resonance imaging, and maybe a little bit on ?? nothing on radiography.  But arthroscopy is another way of defining the structural changes in osteoarthritis.  And there also are reliable, validated, and responsive scoring systems for arthroscopy, including one developed by the French Society of Arthroscopy, which is referred to as the SFA score, and then the International Cartilage Repair Society, or ICRS score.

            And then, of course, to have incident osteoarthritis, you have to have symptoms, not just structural changes, because we're not sure what the clinical implications are of just having structural changes of osteoarthritis. 

            Now finally, let me comment on one randomized controlled trial which was provided to me, but also we came up with separately by my asking one of my orthopedic colleagues who doesn't operate any more, but does a lot of research in orthopedics.  And I said well, I'm giving this talk here at the FDA.  And he said oh, there's one paper which is compare two different types of cartilage repair.  So he sent me the abstract of the paper, and when I finally got the paper it was the same paper, so we got the same paper from two different directions.

            So this is a randomized controlled trial which involved 80 patients.  The patients were followed for two years.  And these were patients with focal cartilage defects without evidence of osteoarthritis either radiographically or at the time of their initial arthroscopy.  They were randomized to autologous cartilage implantation or a microfracture technique.

            The randomization was done at the time of arthroscopy, so the knee was arthroscoped.  When the scope was in the knee and it was determined that the patient did not have osteoarthritis and was eligible, an envelope was opened in the operating room, and the randomization was done. 

            The microfracture technique was then carried out, or the biopsy was done to remove the piece of cartilage from the non-weight bearing segment, which was then sent to Cambridge, chondrocytes were isolated, expanded, and then a second procedure was done approximately four weeks later to implant the autologous chondrocytes.

            So the outcomes were measured at two years.  There was some loss to follow-up, which didn't differ between the two groups.  The vast majority of patients were available for follow-up in terms of questionnaires, and the primary outcome measures, which were the Lysholm score and the VAS pain score did not differ between the two groups.

            And there were power calculations included in the method section, and the study was actually powered to show what would have been a clinically important difference of about 10 normalized units in terms of VAS pain scores or Lysholm scores.  There was, however, a significant difference found in the SF-36 physical component score, which favored the microfracture group over the ACI group at two years.

Now there were differences in baseline.  This was adjusted for in an analysis of co-variants, and the differences remained statistically significant.

            When they looked at structural outcomes which were also discussed, this was done at the time of arthroscopy, so arthroscopy was done, the cartilage was looked at, the ICRS scale was used, and then biopsies were taken which were examined histologically.  There are beautiful pictures in the paper, but there were no differences between the groups either in the arthroscopic score or in the histologies, and the citation is shown here.  So this is the only study that I could find which actually compared two different techniques in a randomized design.

            So I'll conclude now.  I thank you for your time and attention, and I don't know if you want me to answer any questions now, or wait until after Dr. Peterfy's presentation.

            DR. RAO:  Does anybody have any specific questions now?  Can they wait until we've heard from the second speaker, as well?  Thank you once again.

            So we're now going to hear on the Perspective on Imaging Outcomes for Studies of Cartilage Repair by Charles Peterfy.

            DR. PETERFY:  Thank you very much for inviting me here to talk to you about what's probably the most exciting area of development in musculoskeletal radiology today.  And there have been numerous advances made in imaging of articular cartilage over the last several years, but most of that up to now has been in osteoarthritis.  So for those of us who have been working in this area for some time, and quite frustrated with how slowly these technological innovations get assimilated into mainstream clinical practice, it's really exciting to see these first steps being taken, and actually taking off in cartilage repair.  So I'm going to talk about the status of some of these developments as they would be applied in clinical trials of cartilage repair.  And particularly in cartilage repair, imaging brings a few advances or advantages to the table over the other approaches that one might use to follow post-operative development.  And the first, of course, is that it does this non-invasively, and also throughout the entire joint, not just at a specific site of a biopsy or at the limit of the articular cortex.  And then finally, in addition, morphological information imaging can provide compositional information about the integrity of various other tissues, and again non-invasively and throughout the entire joint.

            Now in any clinical trial, imaging can play three fundamental roles.  First of these is patient selection.  It can be used to detect and characterize the cartilage defects in terms of their depth, their area, and other measurements, the quality of the tissue there, in terms of matrix integrity, perhaps, not compressibility, unfortunately, although those are probably related, and by location.

            In addition to characterizing the lesions themselves, it can also provide information about all of the other structures in the joint.  And since those are all important to maintaining the functional integrity of the knee, changes in those, their current status, until we understand their inter-dependency should be accounted for.  And these include meniscal abnormalities, ligamentous changes, et cetera, numerous bone abnormalities, and even periarticular features like cysts and bursae, et cetera, that can affect the results.

            The goal here is really to enrich or homogenize the study population with patients that are more likely to progress rapidly, or respond to a particular therapy, or less likely to develop complications.  And another important consideration is aligning with the target market.  In addition to that, monitoring treatment response at the repair site, most important features are felt to be the degree of defect filling, surface contour, the integration of the repair tissue with the surrounding cartilage and bone, the quality of that repair tissue, and its durability over time.  In addition to the repair site, again the changes that are taking place in the other structures in the joint are important to take note of.

            And then imaging is also useful for assessing safety and detecting adverse events at the repair site.  And there are numerous of those.  And then also, at the donor site, in which cases that's applicable.  And then once again, throughout the remainder of the joint.  So imaging can contribute in each of these different roles in a clinical trial.  And then if one wanted to compare the performance of different imaging technologies and imaging markers in terms of these different roles, and probably one could do that in the following manner with respect to their relative validity, discriminative power, and feasibility.

            With the validity always being the most important, and it's important to distinguish, however, between pathophysiological validity that links the imaging measurement to the clinical outcome of interest, how the patient will feel or function, and technical validity linking that imaging measurement to the exact morphological or compositional feature of interest.  And the degree certainly that one needs with these really depends on the role that the study is going to play in the program, whether this is a confirmatory study.  The endpoints there must be fully established, and that's the most stringent requirement.  But in exploratory studies that are aimed primarily at internal decision-making, of course, one can take advantage of more novel markers that may, perhaps, have performance characteristics that offer certain advantages and responsiveness to change, et cetera.

            Then the discriminative power in terms of the rate of change in the disease of the therapy and how precisely that change can be measured essentially together determining the sensitivity to change.  There are three basic sources of variation that one must consider; the biological variation and the only control you really have over that in this case is in terms of how we select the patients and homogeneity of the study population.  And then variations associated with the image acquisition itself, and these are particularly problematic when dealing with multi-centered studies.  And they relate, of course, to site capabilities in terms of the technologies that they have, but also in terms of the technologist's competence and interest, et cetera, and how forgiving the imaging protocol can be for dealing with some of these difficulties.  It should be kept in mind that right now there is a worldwide shortage of radiologists and radiology technicians, so everybody is quite overwhelmed.  And if one has a protocol that interferes very much with the daily routine, it'll show up as a drop in performance, and can be a very serious matter, so this source of variability has to be taken care of, and the trend is making it harder.

            Then there are variations that occur at the time of the actual image measurement.  One of the advantages of imaging is that it can be centralized, and so one can contain that source of variability by centralized assessment and leveraging whatever computer-assisted automated methods one has at one's disposal.  One can use more sophisticated methods when centrally analyzing the data.

            And then finally, feasibility.  This relates to the safety issues to the patient with imaging.  This means ionizing radiation, and the effects of any contrast agents.  These are usually not too severe.  Convenience from the perspective of the patient, the patient tolerance, how much travel there is involved, scheduling conflicts, et cetera.  At worst, this results in patient drop-outs, so it's an important consideration.  And then as I mentioned earlier, from the technologist's perspective, the ease of the procedure and how disruptive it is to the clinical routine actually makes a difference in what can be done, and how well the results will come out.

            Finally, multi-center and multi-national logistical issues, availability of the technology, IRB issues, language, et cetera, are also considerations.  And finally cost, and in this case with respect to the entire development program.

            And so if one were to compare the various imaging technologies that are available right now through this filter, one would find that magnetic resonance imaging would come out so far ahead that in the interest of time I won't cover all of the other ones, but I'd be glad to discuss them if anybody wants to.  Just to recap some of the strengths that MRI has in this context, it has very high technical validity because of its direct, very high-resolution delineation of the articular cartilage tissue.  And this is just an example that my friend, Doug Goodwin, gave me of a specimen of patella cartilage.  It's imaged at 1.5 TESLA but with a high-resolution quota, but just to illustrate the precision and accuracy of that anatomical delineation of the cartilage specimen.  You can even make out some of the fine striation infrastructure on that specimen.  And I think it illustrates the potential diagnostic accuracy of the technique.

            It offers comprehensive anatomical coverage throughout the joint in contrast to, for example, ultrasound, where acoustic shadowing from bones is going to obscure certain articular surfaces, and certainly anything beneath the articular cortex.  And allows direct visualization of other important articular structures, allowing for the first time a "whole organ" evaluation of the joint.  And as I mentioned earlier, it offers not only morphological but compositional information about tissues, and I'll illustrate some of these in a moment.

            It has good discriminative power because it's easier to control the image acquisition in a multi-center context.  This may not be true for some specialized techniques, and I'll mention those in a moment, but certainly for the bulk of the techniques that are being used these days in clinical trials.  And this allows us to deal with the reproducibility primarily at the level of the image measurement which allows us the opportunity for centralization to help with that.

            And although the unit cost is generally quite high, the improved sensitivity to change can actually reduce the number of patient sites, and the study duration and, therefore, the cost of conducting studies.  There's no radiation exposure associated with MRI.  There's relatively good patient tolerance, if you stick to a one-hour time limit.  That does limit what you can get done, and it's actually a pretty harsh constraint, because we always want to do much, much more than we can squeeze out of one hour, but this is a worthwhile limit to stay within.

            The routine techniques are very straightforward and don't interfere with clinical flow for the most part.  And at least today, now the systems are quite widely available.

            The two favorite techniques for conventional imaging hardware and software in the knee are these two that I'm showing you here on the left.  This is a  T-1 weighted 3D gradient echo image with fat suppression.  You can see that it depicts the articular cartilage as this isolated high-signal intensity band, and basically everything else is low in signal intensity.  This is the femur, the tibia, and the patella, and there's the cartilage along it.

            On the right, the alternative is this two-dimensional Fast Spin Echo sequence with or without fat suppression.  In this case, without fat suppression, and you can see in this case the articular cartilage is dark.  There's a broader range of contrast on that image.  And both of these techniques have been in use for well over a decade in almost unchanged form from the way they are today.

            The gradient echo technique offers higher spatial resolution because it's three-dimensional, but it has lower contrast properties in terms of matrix abnormalities in the articular cartilage, and detecting abnormalities in other tissues.  It also takes a relatively long time, from six to twelve minutes depending on certain nuances in the pulse sequences, and it's vulnerable to metallic artifacts which can be a problem post-operatively.

            The Fast Spin Echo scan is faster.  It's four minutes a scan per knee, and can detect matrix abnormalities in the articular cartilage.  It's less affected by metal in the knee, and offers a very good assessment of other joint structures. 

            Here's just an illustration of the difference in susceptibility to metallic artifacts.  Here is a small micrometallic particle from a prior arthroscopy that does obscure the articular surface mildly right at this point, and it's not even visible on the Fast Spin Echo image, just to illustrate that difference.

            Three-dimensional DESS, or dual echo steady-state is another 3D gradient echo technique that offers the same high resolution that the SPGR technique did, but with a more favorable contrast.  You could see the cartilage is darker, you can make out some of the striations in the cartilage.  The fluid is bright, and it also offers contrast that allows you to evaluate the menisci, ligaments, and other structures.

            This is the technique that we're actually using in the osteoarthritis initiative, which is a public-private partnership exploring the natural history of osteoarthritis in 5,000 patients followed over five years.

            Other techniques on the horizon are the so-called SSFP, or steady-state free procession techniques, one of which is illustrated here, developed by Garry Gold at Stanford, a femur fluctuating equilibrium MRI, shows that you can get very high resolution images with the favorable contrast properties that the Fast Spin Echo has, but in a fraction of the time of either of those.  And these sorts of changes, although not widely available in a bug-free form in the clinic right yet, promise to really speed up the acquisition time, and make some of the patient tolerance issues much better in the near future.

            Well, if we go back to the first two sequences I showed you, there's been several studies that have shown both of these techniques to have very high sensitivity and specificity for detecting focal cartilage defects that at least are Grade 2 or larger based on arthroscopic or direct cadaveric confirmation.  And in terms of that Grade 2 score, it's based on this ICRS score, which you can get a copy of very easily over the web at the ICRS website.

            In this score, which focuses really on depth, a macroscopically normal cartilage would be a Grade 0, and cartilage with mild fibrillation of the surface or small fissures in the surface, so-called nearly normal, 1-A and 1-B. 

            Now the techniques that I showed you earlier in their most widely available form are relatively insensitive to those two mild superficial changes, although in many of those cases, one might see a focus of signal abnormality in the otherwise low-signal intensity cartilage as shown here.  This signal change is related to a loss of collagen matrix, which allows the fluid, the water in the cartilage tissue to become more mobile and behave more like the rapidly mobile synovial fluid water and, therefore, have a higher signal.

            Here's an example in the deep cartilage with potentially early delamination of that cartilage, that otherwise on the surface may appear completely normal.  But when a defect is deeper than that, but less than half the thickness of the articular cartilage, it's regarded a Grade 2.  And here's an example of a small less than 50 percent partial thickness defect in the lateral tibial plateau.  And even though this is a pixelly image, you can pick that out, and you can feel quite confident that this is less than 50 percent; so just to illustrate the level of accuracy that one could get with these techniques.

            Defects that go deeper than 50 percent for Grade 3, and here's an example where you can easily say that this is deeper than 50 percent.  That grade is actually broken into four categories depending on whether the defect goes up to or through the calcified layer, whether it has blistering.  Unfortunately, MRI does not confidently discriminate the calcified layer of cartilage and, therefore, can't discriminate among these three variations of Grade 3, although it can detect blistering cartilage, usually with some matrix abnormality showing high signal as in this example.

            And then, of course, there is the defects that enter into the bone with or without cyst formation.  And, of course, MRI can detect that very well.  Here's an example of a cyst, and note also the associated bone marrow edema in this case, not in the scoring method, but something that's detectible and always watched for with MRI in these settings.

            Now in addition to these subjective or semi-quantitative analyses, one can use various segmentation techniques to quantify the volume, the thickness of cartilage, or of individual lesions.  And there are numerous techniques that have been developed.  None of them are fully automated, and there's never been a well-validated automated technique, so they all require some human intervention.  And most of them are quite tedious to do, but they can generate thickness maps like this, and one can even register serially acquired images and subtract them, and form a number of sophisticated measurements like this quantitatively.

            This is probably overkill, particularly in the setting of a imaging prior to debridement, which obviously will under-estimate the final defect size, but it can be used to survey the remainder of the cartilage, as I mentioned earlier, while assessing the associated findings.

            And scoring methods have been developed to do this.  This is one called Whole-Organ MRI Score that looks at 13 articular features.  Five of those include the cartilage, the marrow edema, cysts, bone attrition, osteophytes in 15 different locations in the knee, and eight other features including the menisci, cruciate ligaments, collateral ligaments, synovial effusion and some periarticular bursae and cysts.  And in some preliminary studies it's been shown to have relatively high inter and intra-reader reproducibility.  And it's currently being used in several large epidemiological studies of osteoarthritis, so information will be coming out about the relative importance of these different articular structures.  And the preliminary results do suggest that there are some correlations with pain and functional outcomes, although the interdependencies of these features, as I mentioned, are not well-characterized yet, and the longitudinal performance is also not yet well-characterized.  So that's the state of that type of a scoring method.

            In terms of monitoring change, as I mentioned earlier, these five features are the ones that have been focused on the most in the literature, and these have just some face credibility in terms of their merit, but rigorous clinical correlations obviously haven't been carried out at this level as in osteoarthritis.  But here are some examples of what these types of scores might look like.  This is from a proposal from Marlovits on a four-point scoring grading system, like the one I mentioned earlier, where it would be less than half, or more than 50 percent of thickness loss, or hypertrophy as in this example.  One can get a sense of the level of reliability that one would have with something like that, and although the reproducibility hasn't been firmly established under different circumstances, the ICRS repair assessment scoring scheme actually is relatively similar, just breaking it down by quarters instead of halves in this case.  And this hasn't been tested in terms of its performance, however.

            There haven't been too many longitudinal studies, as Marc mentioned earlier.  One that has been published recently by Wendy Brown at the Hospital for Special Surgery, involved 122 patients, 34 of them with autologous chondrocyte implantation, 80 microfracture patients followed up for 15 months and 13 months respectively, and in this study found a superior performance in terms of defect filling with the chondrocyte implantation relative to microfracture at this average duration. 

            Interestingly, she published a 63 percent incidence of ACI hypertrophy, which is much larger than most other series have shown.  Usually, it's in the 10 to 20 percent range, and these usually showed up quite early.  The study is limited, however, in terms of the follow-up intervals not being standardized, so it's difficult to quantify the longitudinal changes.  And another point that's relevant to some of the discussions that we've been having here is the difficulty blinding the readers.  As you look at it, you can tell right away which type of repair one is dealing with, so this poses a greater challenge to scientific integrity.   And there are ways around it potentially, but it's a particular challenge to cartilage repair imaging.

            Another study by Henderson involved 57 patients with autologous chondrocyte implantation who had MRI at three months and twelve months post-op, starting to give us a sense of the time course of some of these changes.  And as you can see, over the course of this study, the proportion of repairs that had 100 percent filling increased quite dramatically over that three to twelve month time period, and the number of poorly filled lesions also decreased.

            And although the time intervals were slightly different, and the proportions were different, these are by 50 percent, and Wendy Brown's was by thirds, the numbers still come out roughly in the same ball park for proportions.  There weren't too many good images in those articles, so here are some just to illustrate this phenomenon.  These are sagittal gradient echo images in a patient who's had microfracture.  You could see the site of the procedure here two months post-op.  There's incomplete filling of that with depression of the surface that by seven months has filled out nicely, and shows a good congruency of the surface.  And if you look on the coronal fast spin echo image in this patient, you could see there's some heterogeneity to the signal of that articular cartilage, but otherwise, it's a good filling of the defect.

            This example of autologous osteochondral transplantation, you can see that there was quite a bit of subsidence of the osteochondral plugs on this time point with the fast spin echo here, and the gradient echo here.  That shows a fairly remarkable recovery by three years with restoration of that articular surface over that period.  So in terms of trying to decide what the right time interval for evaluating efficacy is, and there is some data starting to accumulate to give some guidance on that, but nothing too firm yet.

            Other parameters that ICRS focuses on include the integration of the border zone, and this is simply fluid signal tracking along the perimeter of the graft, and the extent of that scored in a semi-quantitative way, and the surface smoothness in terms of fissures and fibrillation.

            Here's an example of delaminating graft that isn't integrating well, and you can see the fluid signal tracking up beneath it like this.  In this earlier example, there is high signal intensity linear streak at the margin between the graft and the adjacent native cartilage.  And this may be a lack of integration, although there are some cases in which it appears that this is actually quite in tact, and it recovers over time, and it may represent just simply granulation tissue.

            In that study by Brown that I noted earlier, only 15 percent of the 34 patients that had had ACI actually delaminated within the first nine months of the follow-up.  And in the study by Marlovits, 16 patients had matrix-induced autologous chondrocyte implantation without a periosteal cover, and again, a relatively low proportion that detached over approximately five weeks of follow-up, with 88 percent completely attached and filled.  This is just some of the follow-up data that's in the literature today.

            Another feature that has drawn a lot of attention is bone marrow edema in cartilage repair, as it has in osteoarthritis.  It is the finding of high signal intensity on fast spin echo or stir images, fast spin echo with fat suppression, in particular, and you see examples of it here.  It's a striking feature.  It's quite easy to pick up.  And it's very common in the early post-operative period, but regresses with healing.  Time lines aren't well mapped out, but this one study by Henderson found it in about 61 percent of the subjects that had ACI within three months, and more than half of those seemed to have resolved by twelve months, but it's still a fairly good proportion, still have bone marrow edema after one year.  Most of that being mild, but 9 percent with moderate bone marrow edema.  And there's some concern that persistence of bone marrow edema or intensification may indicate failure of incorporation.

            Brittberg and Winalski have proposed that a scoring method for intensity and the extent of bone marrow edema with intensity being a relatively simple scale referenced against muscle on the same image, and the extent being graded in terms of the proportion of the distance between the articular surface and the physeal scar. 

            And once again, not to forget to check the donor site.  In this example six months post-op, you can see where the plugs have been harvested and the bone is starting to fill-in, although the gaps from the articular cartilage are still there.

            In terms of the quality of the repair tissue on MRI, this shows up on T2-weighted images or the fast spin echo images like this as a area of high signal intensity in the graft tissue compared to the normally low signal intensity adjacent articular cartilage.

            Again, although the water concentration in articular cartilage, approximately 75 percent that of the adjacent synovial fluid, the presence of collagen  immobilizes that tissue water and promotes dipol-dipol interactions that causes a decay of signal intensity.  So as that collagen breaks down, the water becomes more mobile, and begins to show MRI characteristics similar to that of the free water in the synovial fluid.  And that's why these areas of high signal intensity that we see in articular cartilage, and there's been quite a bit of histological and biochemical correlation of this, represent collagen damage.  And in this case, you can see the blistering surface over that area.

            In that study by Brown, 100 percent of the 34 patients that had ACI showed hypertense graft in the first six months, 30 percent of those or so normalized within a year and a half.  She didn't give the numbers for the microfracture, but a high percentage were also hypertense at baseline.  And we see here an example of one case that's six weeks post-op shows high signal intensity that normalizes as that graft matures.  In osteoarthritis, we usually see it going the other way.

            And in this study by Henderson, once again at baseline you can see 96 percent had high signal intensity in the graft site, and the majority of those normalized over one year.  Now one can look at this semi-quantitatively, as I just showed you, or actually quantify T2 relaxation in articular cartilage.  And this can be depicted in image mode, as here, or graft mode.  And there's been quite a bit of work done on this by Bernie Darnzinski in Cincinnati in humans, in children with juvenile rheumatoid arthritis, but very little work has been done so far with cartilage repair.  There's been some work with animals, but to date we don't have that much experience with this so far.  It's a technique that can be done in a multi-center context, but it uses up some of the imaging time.  And if one wanted to evaluate the tibial femoral cartilage, typically one has to go to a higher field strength like 3 TESLA, and that is less available.

            And just one more matrix marker that I want to mention because it's a very elegant technique called the GEMRIC, delayed Gadolinium-Enhanced MRI of Articular Cartilage.  This is a technique really spearheaded by Debbie Burstein and Martha Grade, MIT.  And it's based on the fixed negative charge of the glycocyamine or glycans in the normal articular cartilage.  And when the negatively charged Gadolinium DTPA is introduced into the blood, it gets in the synovial fluid, and subchondrally it's repelled by these fixed negative charges and, therefore, distributes in the articular cartilage in inverse proportion to the GAGs.  And by measuring the effect on T1 relaxation, one can actually quantify the glycocyamine or glycan content in the articular cartilage.  And this relationship is actually linear over quite a large glycocyamine and glycan range.  So it's an interesting technique.  Most of the work has been done at single sites, and in various settings, but very little in cartilage repair.

            Here is one study that was reported a little while back that did find, however, using this GEMRIC that the glycan content in the grafts was lower in the first six months than it was in the subsequent six months, where these grafted areas gained by the glycocyamine and glycan content of the adjacent cartilage in the same knee.

            The technique is somewhat cumbersome and requires an injection, and then 90-minute wait, and then a 20-minute scan, so it's something that is more difficult to do in a multi-center setting, or with a long series of other desires for your protocol, but it still shows a lot of promise.  It's a very intriguing marker.

            Anyway, in summary, conventional MRI as it currently stands has a lot to offer as a tool for non-invasively evaluating cartilage repair, both in terms of selecting the right subjects, monitoring the treatment response, and identifying the safety issues and adverse effects.  And they can accurately identify and grade the articular cartilage defect.  Scoring methods have been developed for most of the key features that are believed to be important in successful repair and adverse outcomes.  Longitudinal data is starting to accumulate, but we still need more standardization and performance characterization of these metrics. 

            Currently, it's also difficult to blind readers, and some thought has to be given to that.  The novel compositional markers, I just mentioned two of them but there are several others, are also very promising but need more development before they'll be easy to use in a multi-center context.  And, of course, to remember to include assessments of the other structures in the joints, as Marc mentioned, since these aren't things that we want to lose sight of as we go forward.  Thank you.

            DR. RAO:  Thank you.  Any questions for the doctor?

            DR. MOOS:  Just one.  A little bit off the track, but I see we're well ahead of time, that may be famous last words, but we were reminded a bit ago about the proclivity for certain types of cells, like MSCs, perhaps to migrate outside of the joint space.  I'm wondering if you might be able to comment on the use of relaxation techniques to keep track of cells that might be tagged paramagnetically or something like that, to kill another bird with this very powerful stone of your's.

            DR. PETERFY:  That sort of thinking has so many applications it would be interesting to do.  It's harder to do than it sounds, particularly with MRI, usually because of signal-to-noise ratio issues in PET or scintigraphic techniques are usually easier to do.  But I can't think of anything that's actually been put together up to now for that particular purpose.

            DR. TOMFORD:  In my hospital, the reports I get from the radiologists seldom mention the cartilage.  They say there are severe degenerative changes.  What is happening in the clinical field right now in terms of routine changes, or routine differences that the radiologists may be taking in the future?

            DR. PETERFY:  Yes.  What the radiologists look for is really driven by treatment, and we've seen the same thing in rheumatology.  There's virtually no radiologist who knows how to do sharp scoring, for example.  It's not used clinically, and by the same token, I think now that cartilage repair is starting to find clinical application, radiologists are paying more attention to cartilage.  But before that, there was no real therapy for it and, therefore, the details about the anatomical variation wasn't relevant to clinical practice.  And these pulse sequences have been around for 10 years.  In fact, they're even often used in routine practice, but actually mapping out where all the defects are takes a long time, and how people are actually going to use that information is not very clear, so I don't think it's routinely commented on in great detail.

            DR. RAO:  Dr. Coutts.

            DR. COUTTS:  How well do these techniques work on thinner cartilage, such as might be in an animal model?

            DR. PETERFY:  Well, you can do rabbits at 1.5 TESLA, and the trade-off is really time.  And typically, with animals one goes to a higher field strength to save time, get the spatial resolution.  The smaller the cartilage, obviously the higher spatial resolution, and the longer the time it takes.  Paradoxically, it's easier to image humans than animals with this technique.

            DR. COUTTS:  Right.  But in a rabbit where the cartilage is half a millimeter thick at the most, probably closer to a third, can you get down to that degree of resolution and see something that thin?

            DR. PETERFY:  You can get a few pixels through it.  Usually people report very high field strength imaging, but we've done 2 TESLA scans on rabbit knees and quantified cartilage volume.  There was a guy named Tom Link that has studied looking at focal defects in rabbit knees at 1.5 TESLA, and found pretty good results, actually.  It's a matter of how much time you have to image.  The animal is anesthetized so you can usually do things that you wouldn't do in humans, even with that lower field strength.

            DR. RAO:  Dr. Nixon.

            MR. NIXON:  Clearly, GEMRIC is a very useful way of semi-quantitating the proteoglycan content.  I just wondered how wide-spread it had become among radiology facilities, firstly; and what are the limits to adopting the program for routine examination of cartilage?

            DR. PETERFY:  With GEMRIC?

            MR. NIXON:  Yes.

            DR. PETERFY:  Well, the Gadolinium DPTA has been around for more than a decade and used routinely daily in most hospitals.  But the GEMRIC itself is not routine in most sites.  It's not that easy to do.  It requires, as I mentioned, an injection, and then exercising the knee briefly, and then waiting 90 minutes, and then re-scanning.  And then the sequence that one has to do is called inversion recovery sequence, so it's not a typical sequence, and you have to actually map out the T1 parameter from it, so it's not routine from that standpoint.  And, in fact, the multi-center reproducibility hasn't ever been really worked out, so all the reports I've seen have been single sites, usually small numbers of patients.  But it shows great promise.

            DR. HOCHBERG:  If I could just comment; there is a multi-center study now going on which is using the GEMRIC to look at effects of pharmacological intervention in osteoarthritis, and the effects of the intervention on proteoglycans.  I think it's going on in six centers.

            DR. RAO:  I should add that if you have questions for Dr. Hochberg too, you should ask them now.

            DR. McILWRAITH:  Just going back to the question Dr. Moos posed, that there are a couple of publications on ferritin-labeled stem cells being put into joints and showing in MRI menus to show where they localize.  So back to that question, are you saying that you don't think you could have the resolution in humans to do that?

            DR. PETERFY:  I've done it in rabbits.  I took ferritin and put it into inflammatory models of rabbits, and it localized in the synovium, and even later you could find it in the macrophages.  That's an intriguing point, but I don't know if the chondrocytes would pick something like that up.  And this had to be phagocytosed really to make the contrast.

            DR. McILWRAITH:  In one of the papers it was small animal, but it showed unfibrillated cartilage, or showed label unfibrillated cartilage.  I thought it  did, anyway.

            DR. PETERFY:  I didn't read that.

            DR. RAO:   I have limited experience with ferridox label, but it seems to me that it depends on the cell types.  There's quite a lot of variation on which cell type picks it up to give you adequate signal.  And if you have dividing cell, then there's a dilutional effect that you have to worry about, as well.  And so there have been quite variable reports on labeling in these types of issues.  Go ahead.

            MR. O'CALLAGHAN:  Michael O'Callaghan from Genzyme.  Charles, you didn't mention anything about spectroscopy.  Do you have any sort of insights into that?  Like I know that Garry Gold was exploring that at some stage, and it looked fairly promising.  I just wonder whether you could comment.

            DR. PETERFY:  He was using it to measure T2 and water content in articular cartilage a few years ago, but it didn't go much farther than that.  That's the most that I've seen.  He also did diffusion weighted imaging, which is an interesting thing.  I had that up there, so there's a few techniques that still need further development.  And the two that are probably closest to application are the T2 relaxation and the GEMRIC.

            DR. RAO:  Go ahead, Dr. Tuan.

            DR. TUAN:  Charles, I have a question for you.  I remember Martha Gray did a series of studies where she took explants of cartilage and looked at I think it was IOM beta induced degeneration and monitored with the GEMRIC.  I wonder if anybody has provided evidence correlating the GEMRIC signal and the mechanical property of the cartilage; which, of course, is related to proteoglycan content, but I'm just wondering whether there is ?? using the signal that you have, can you actually say something about the mechanical integrity of the cartilage?

            DR. PETERFY:  I haven't seen any study like that, but that obviously would be a terrific correlation to have.  Same with the T2 changes.

            DR. RAO:  So is it fair to say that it's easier to do MRI in humans, and it's more freely available right now, but given the size of the animals and the necessity of anesthetizing some of the larger animal models, that MRI is much more difficult in animal models?

            DR. PETERFY:  Well, it depends what you're trying to do.

            DR. RAO:  Just in general.

            DR. PETERFY:  In general, you have a larger joint, so you always have an easier time with it.  But if you have a high field magnet, then you can do some very interesting things with animals.  But the clinical magnets are ?? you can do a lot more with humans because of the size.

            DR. McILWRAITH:  In the human machines, well the conventional ones like the GE one and a half TESLA, there's limitations with how far you can get the horse's leg in.  But we have actually ?? we've just recently installed an extremity scanner that works well.  You have to anesthetize the horse, but that's obviously got limited availability at the moment.

            DR. RAO:  If there are no more questions, thank you, doctor.

            DR. PETERFY:  Thanks.

            DR. RAO:  I guess it's time to go back and look at the questions and see what we can say.  Maybe I can poll the committee.  Do you want to have a quick break before we start?  Let's have a quick break, 10 minutes.

            (Whereupon, the proceedings in the above-entitled matter went off the record at 3:58:05 p.m. and went back on the record at 4:12:27 p.m.)

            DR. RAO:  So we're going to try and consider two major questions, and this is related to clinical studies, whether they are confirmatory studies or initial studies.  And before we actually look at those questions, I want to try to ?? we had some wonderful talks today, which tried to give us some overall summary of what needs to be done.  And I'm going to try and make a statement before we go and look at these questions, and see whether the committee agrees in terms of focusing on these.  But my sense from what I heard, especially with Dr. Buckwalter starting in the morning, was that it's really important to look at outcome measures which include patient response, and pain, and improvement is a very critical outcome.  And we need some measure of that sort of outcome which will be included in any sort of study that's done.

            It also seemed that you needed some kind of measure of anatomical improvement.  It wasn't clear to me from listening to the presentations whether that had to include arthroscopy, or whether it had to include biopsy.  In fact, it seemed to me that there was a feeling that that was not what should be included, but there had to be some kind of imaging or some kind of measure of the cells of the biology that would help in evaluating the cells.  And that alone would not be enough, and it also seemed like there was a sense that in any of these studies, one had to look at it over a certain time period.  And does the committee feel that that's a fair assessment, at least in the initial pass, in terms of when we consider these clinical questions?  If I don't give anybody a chance to comment, I guess we can move on.  Dr. Coutts.

            DR. COUTTS:  I'm not so sure that I agree with your statement that we need to characterize the tissue.  It would be nice to be able to characterize the tissue.  And if we could do it non-invasively, that would be very good.

            When you start talking about characterizing the tissue, it essentially means in my mind that you have to invade the body and get a specimen, or even just look at it.  And when you start talking about that, I put myself into the position of an individual being asked to participate in the clinical study.  I don't put myself in the position of the physician asking, but of being a patient being asked.  And if you tell me that you want me to come back so that you can do another operation on my knee, and perhaps maybe put a needle into it and take it out, take out some tissue, I don't get very excited about that.

            DR. RAO:  Point well taken.  I wasn't implying that that's what I felt the sense was.  In fact, as a point here, I think the sense is that that may not be what the committee feels.

            DR. COUTTS:  Right.  But to the extent that you could do it, say with MRI and characterize the tissue that way, I think that would be very, very beneficial.

            DR. RAO:  Did you have something to add?

            DR. LEIBENHAUT:  Just to point out that the clinical question number 1 is related to the confirmatory studies, not exploratory.

            DR. RAO:  So let's ?? go ahead, Dr. Tomford.

            DR. TOMFORD:  I just wanted to make the point that I think there probably is a correlation between the type of tissue and the longevity of the successful transplant.  In other words, if the transplant ends up mostly fibrocartilage, I don't think that will last as long or provide as good pain relief as one that is hyaline cartilage.  So I couldn't disagree with Dr.  Coutts, I don't want a third or fourth arthroscopy, and particularly not a biopsy, but if we had ways of determining the true measure of the tissue, I think that would be helpful.

            DR. RAO:  Let's keep that thought in mind.

            DR. McILWRAITH:  Can I just ?? you just brought up the really big issue of the gold standard of repair.  It's often quoted that if you only have fibrocartilage, it's not going to last, but where's the proof of that?  Like if we look at long ?? even in our long-term equine studies, we certainly test durability.  And I know we've got the caveat of extrapolating from horse to human, but these horses get galloped on the treadmill five days a week from two months to eighteen months in this one particular study I'm thinking of.  And the tissue certainly - I think it's a fair test of durability.  Now admittedly, it's not 10 years out, but eighteen months in a horse is quite a while, particularly if you're exercising it athletically.  And it's fibrocartilage.  We've got hyaline elements in the base, and we don't really improve the quality if we look at histomorphometry and compare our amount of fibrocartilage, versus hyaline cartilage, versus fibrous tissue.  It is fibrocartilage, but it does stand up to wear.  And I'm sort of prompted by a talk that I heard on Saturday at specialty day at the American Orthopedics Sports Medicine Association, where the person got up and denigrated a certain cartilage-healing technique, and then denigrated another one on the basis that they didn't form hyaline cartilage.  And I don't know of any technique that forms hyaline cartilage yet, or I haven't seen the proof.  And so I'm sort of interested how this has sort of come to be biblical in terms of fibrocartilage not being durable.

            DR. RAO:  Before we discuss that ??

            DR. TOMFORD:  May I say something?  I think your point is well taken.  I don't perhaps know the literature exactly in terms of whether fibrocartilage will hold up; however, I think that ?? well, first of all, I'm not saying it has to be hyaline cartilage.  I'm saying that the ideal would be hyaline cartilage.  I think that's true.  I would rather have hyaline cartilage in my knee than fibrocartilage. 

            But secondly, when I look in a knee and it has lesions on the tibia, on the femur, on the patella, all of which are probably fibrocartilage already, and this patient if I follow them for two or three years ends up having to need a knee replacement, then I ask myself is that fibrocartilage beginning to deteriorate more, and they're beginning to get bony lesions?  So I think there is a progression from, at least in terms of a pathologic spectrum, articular cartilage to fibrocartilage to degenerative joint disease.

            DR. RAO:  Before we discuss that, it's going to be part of the question which is third down on our list, let's get through the first two, which are really on some rating scores.  And there was a point made earlier about the rating scores, one wants to disassociate function from pain.  And I'd really like some of the clinicians who have experience with this to maybe perhaps comment on the rating scores, and see if there's any clear-cut obvious choice, as Dr. Hochberg pointed out of the many that he presented.  It's okay as long as one uses a particular rating score, and one should disassociate pain and function in some measure.  Anyone would like to go first?

            DR. COUTTS:  I'll bit the bullet on this.  In our packet that we received beforehand, there was an excellent article about the different rating scores, and I thought that kind of put it into perspective.  And it mentioned that some scores were better for measuring pain and function in patients with osteoarthritis versus athletic injuries and the like.  And once again, Genzyme seems to have led the way here using the modified Cincinnati score, which seems to lend itself to the patients with focal articular cartilage defects.

            Now if you're not treating focal lesions, then I think you'd want to pick one of the other scores, which would be more applicable say to patients with degenerative arthritis, if that's the type of lesion that you're treating.  And so I think it would have to be individualized to the type of patient selected for your study.

            DR. RAO:  Some kind of score, and that would be dependent on ultimately that was in terms of ??

            DR. COUTTS:  Yes, I think so.

            DR. RAO:  Dr. Hochberg, would you want to comment?

            DR. HOCHBERG:  If I was providing, let's say, a text for recommendations that you would be voting on, I would say that you'd want to use a reliable, valid, and responsive instrument, which would have a separate score for pain, and a separate score for function.  And you could choose any instrument that has those metrics.

            DR. RAO:  You did point out in your presentation that there was this society, OMERACT.

            DR. HOCHBERG:  Well, there's the society called OMERACT.  OMERACT hasn't looked at the issue of cartilage repair, and if there was a group that was interested in having OMERACT look at it in combination with orthopedists and rheumatologists, et cetera, that certainly could be put on the agenda for the next meeting.

            DR. RAO:  But there is currently no internationally accepted ??

            DR. HOCHBERG:  Other than what's, I think, proposed by the International Cartilage Repair Society.  And I think there was one other one, there's an International Knee Documentation Scale.

            DR. RAO:  So if I were to put forth to the committee that there seems to be consensus in the field that some kind of score which disassociates between function and pain needed to be used, but there isn't a clear-cut consensus that there's any one scale; but rather, that's dependent on that scale being appropriate in terms of the measures it has for the kind of repair or defect that you're going to look at.  But that's what would be an important consideration for the field.  Does that seem ??

            DR. COUTTS:  Yes.  Because, for instance, people with focal articular defects, they tend to be younger individuals, and their functional demands are higher as a rule than patients with osteoarthritis, who simply want to be able to walk without pain; whereas, those with focal lesions and being younger want to be able to get back to playing sports.  So you have to take that into consideration and base your selection on those criteria, because the modified Cincinnati form asks questions with regards to function in a more athletic sense than say the WOMAC does, which looks pretty much just at activities of daily living.

            DR. RAO:  And does it seem as if just a questionnaire type of scoring which was the self-scoring system is useful or should there be consideration given to a physical measure, say quantitative measure of testing range of motion, or stepping tests, or walking style, I mean, placing feet.  Is that ??

            DR. COUTTS:  Obviously, quantitative data is much easier to evaluate than subjective data.  But the quantitative measures that you can use are surrogate measures for function, for being able to do certain activities.  And when you come right down to it, that's really why these individuals are seeking care and attention, is so that they can obtain restoration of the function that they have lost.  So asking them about what functions they are capable of doing beforehand, and then seeing if they can do them after your intervention, I think is pretty valid.  It's hard to put a number on it, but that's what these measurement tools attempt to do.  And I think they're reasonably good evaluators of the effect of your treatment.

            DR. RAO:  Have there been any issues of sort of placebo effect in these sorts of questionnaires which is common with any sort of questionnaire-based scoring system that's not been an issue in these systems, people feel it's reasonably reliable?  Dr. Hochberg.

            DR. HOCHBERG:  Well, the questionnaires are designed to be self-administered.  They are completed in a very short period of time.  They're highly reliable in terms of internal consistency, test/retest reliability, so you can put a number on it.  And as was commented, it's the sort of patient centered outcome as to how I function, how much difficulty I'm having with regard to functioning, how much pain am I having? 

            There are some, certainly, correlations between findings on clinical examination and the questionnaires that ?? the scales of the questionnaires that deal with functional limitation or disability, so one question you all have to deal with  is whether you want to have measurements on clinical examination, in addition to questionnaires. 

            We sometimes do physical performance tests, as well as questionnaires to actually measure performance, and we use timed up-and-go tests, we do walking time.  Rheumatologists don't do a lot of range of motion any more because we didn't feel, at least in osteoarthritis it doesn't seem to add anything to the measurements with questionnaires. 

            There's certainly a placebo response.  I guess the article that people tend to point to now was the randomized three-arm sham control trial that came out of Texas, and was published in the New England Journal a couple of years ago, where there was a response in the sham group which was similar in magnitude, and not distinguishable from the response in the active groups, which may represent the natural history of the condition.

            Oftentimes, when people enter into a trial, they're at their worst, and just following them over time, there are some so-called regression to the mean, that there is improvement.  But the questionnaires that are responsive are able to discriminate effective interventions from what is seen just in a comparison group which is followed over the same period of time, which receives either a sham intervention or a placebo intervention, or a wait-list intervention.

            DR. RAO:  Go ahead, Dr. Coutts.

            DR. COUTTS:  With regards to the placebo effect, there's no question that any time you intervene, if you operate on a patient, there's a huge potential for placebo, and it can even fool the patient.  In clinical practice, I think that's great.  I don't care how it works, as long as the patient is better, so maybe placebo is a good treatment.  But I don't think that we're ever going to be able to get away from that.  That's an issue, but that will become uncovered as you study patients over time, because if it is a placebo effect, it won't last.

            DR. RAO:  Dr. Hochberg also pointed out, he said that clinical examination doesn't add much when he's looking at osteoarthritis patients.  Would that be true, you think, in patients in which you have ?? younger patients where you have a small localized defect to filling in, or whether it would really be important to do a clinical exam or some measure in addition to looking at the questionnaire-based study?

            DR. COUTTS:  Well, I didn't interpret his comments as saying that you wouldn't do a clinical exam.  He said he wouldn't do range of motion, but there are other things that you can see on clinical examination, such as a swollen joint, fluid in the joint, particularly if we're talking about the knee still, evidence of instability, watching the patient walk and see if they limp.  And if range of motion is severely restricted, it will have a functional consequence to it, so particularly if they have flexion contracture - this is something that you would want to observe and note.

            DR. RAO:  Clearly, a clinical exam or some measure of range of motion is going to be important.  Some clinical parameters are going to be important, and those would depend on, again, the effect that you're going to look at.

            DR. COUTTS:  We do it anyway, so why not collect the data?

            DR. SCULLY:  One more thing on range of motion, though; that at least in terms of orthopedic intervention, the best indicator of post-operative range of motion is actually pre-operative range of motion.  So if somebody comes to you with a restricted range of motion, you don't want to guarantee them they  would have a better range of motion when you're done, because it's probably not going to happen.

            DR. RAO:  There's another issue here which sort of came up earlier just in looking at these measures, was everybody said that it's going to be hard to consider double-blind placebo control trial, given how one has to treat patients who come in with this.  And also, I think Dr. Buckwalter also pointed out earlier in his talk that it's hard to take a person who has failed one arm to then use him as a control for the second arm.  So is there a difficulty in designing a trial where one would have a reasonable control given one knows that there is a placebo effect, and while one may be happy with the placebo effect, it could be pretty expensive to be treating patients to get that placebo effect.  So in the clinicians minds, then if one had to design a study where one is studying one method, what would you use?  Would it be a comparative study where you're comparing it with an alternate method, or have to be done somewhat differently?

            DR. COUTTS:  This is a really tough question, and so much hinges on it, because you're going to approve or disapprove a product based on how it performs, the pivotal clinical study.  Good science dictates that there should be a control, and there are a variety of ways at arriving at that.  And it depends, to a certain extent, on what it is that you're studying.

            We now have the ACI methodology, which could be used as a control for any new technique that comes along now that purports to heal articular cartilage, and patients could be prospectively randomized to that.  Blinding is extremely difficult if there's a big difference in the nature of the operative procedure.  It's pretty hard to blind somebody if you have to biopsy them, and then bring them back for an open procedure, and what you're studying is some arthroscopic technique which is done say in one stage.  So you just have to accept that as an inherent limitation of your ability to apply the ultimate scientific methodology.

            Cohort studies might also work.  Again, it's not as rigorous scientifically, but it does give you an opportunity to create a control group.  It's essentially looking at an alternative treatment, but doing it say with a different surgeon, or at a different institution where the patients don't have to be randomized. 

            When it comes to a surgical procedure, randomization is difficult.  Again, put yourself in the patient's position, and regardless of the merits of a particular technique, that patient may have a bias in terms of what they want to have.  And they'll refuse randomization if it looks like they'll get something that they consider to be inferior.

            You can sometimes influence that by the way in which you present the question to them for participation in the study, but you're talking about invading their joint, putting them through a significant period of disability and rehabilitation, and they'd like to know that they're going to be better after this.  So I think the message here is that it's important to be flexible in regard to the issue of controls, and be somewhat tailored - let the investigators tailor this the way that they think it best meets the needs of the scientific need for data, but at the same time recognizing the limitations.

            DR. RAO:  You wanted to make a comment?

            DR. LACHENBRUCH:  Yes.  I'm Peter Lachenbruch. I'm head of the Division of Biostatistics at CBER.  One alternative when you have a failed therapy is to have two control arms, and the patient could be randomized to a control arm that they had not failed on, or the treatment of interest. Now there would be some statistic issues in developing the appropriate analysis, but I don't think they would be insurmountable. 

            DR. TSIATIS:  I just want to speak against not doing a randomized clinical trial, if that was an implication that was made.  We heard earlier this morning that a lot of physicians actually choose their patients, and they've gotten good at choosing patients.  And that would definitely tend to bias any kind of non-randomized control, even as best as you try.  And so I would very much caution against even thinking that way.

            DR. RAO: It also seemed to me that a couple of studies were presented with ACI, at least, would seem that they have done randomization, but that compared to two arms in which they looked at ACI versus debridement in terms of looking at improvement.  It can't be truly blinded, but there is still a randomization in a trial, so that certainly seems possible.  But truly blinding, I think, seems to be difficult to do in this situation.

            DR. TSIATIS:  As far as the blinding goes, obviously, if we could blind, that would be incredibly important, especially since the primary endpoint seems to be clinical endpoint such as pain and function.  I think the issue about time was important, as well, which is if there is an effect of knowing what treatment you're on, I would suspect that that effect would go away over a period of time, and so I think the time aspect is also very important in doing these trials.

            DR. RAO:  So let us consider what we've said so far.  In some sense, we said that if you were to have a study, you would have to have at least two arms to the study, and you'd have to design it well in terms of the readout you'd have in a kind of questionnaire format, or the questionnaire you'd use.  And you would certainly have to have some kind of clinical measure, which would maybe include range of motion, or which would be dependent on what you would need, depending on the symptoms that the patient came to you with.

            There were two other issues in terms of readouts that people had mentioned, and there didn't seem to be much enthusiasm among the committee, and people who have actually performed these studies; and that was to look by arthroscopy, or whether there was any need for taking a biopsy.  And I know that we've already gone over this, but it may be useful to have a reiteration or a summary at this stage for those aspects of that question before we move to the next part, as well.  And so, would one of the clinicians want to ?? does Dr. Scully want to say something on that?

            DR. SCULLY:  I don't have anything to add that hasn't been said before, I don't think. 

            DR. RAO:  So maybe it would be useful to restate that, as well.

            DR. SCULLY:  That I think intervention is ??

            DR. RAO:  Arthroscopy, perhaps.

            DR. SCULLY:  Any intervention, any invasive intervention, either arthroscopy or biopsy, it would be advantageous not to have to do that in terms of the outcome.

            The other thing I guess I've been kind of sitting here thinking, and I was thinking about surgeon bias in terms of entry into trials and things like that.  One of the problems, if you start thinking about doing a trial like this, the inclusion criteria will be very, very difficult to absolutely state, because you're going to have to talk about etiology of the lesion, you're going to have to talk about size of the lesion, you're going to have to talk about depth of the lesion, does it involve subchondral bone or not.  I mean, there's a lot of inherent problems here.

            DR. RAO:  Dr. Coutts, and then Dr. Tomford.

            DR. COUTTS:  I know of ?? I have some orthopedic friends who consider arthroscopy an extension of the physical examination.  And it doesn't bother them to scope into people.  They do so much of it that they even lose sight of the fact that it is a potentially dangerous thing to do, but anesthetizing a patient, or even doing it under local and sticking instruments into a joint, every once in a while things go bad.  And whether it's an infection, or whether it's an arthrofibrosis, or whether it's damage to intra-articular structure which causes the patients to be worse afterwards, that is a possibility.

            I think this Knutsen study that was reported in the Journal of Bone and Joint Surgery is quite remarkable, in the fact that they did randomize patients to do different treatment arms, and then they got them back for arthroscopies and did biopsies.  I'm not so sure where the biopsies help them in evaluating the two procedures, and that the clinical measurement tools that they used were probably the major determinants of their conclusions with regards to the efficacy of the two treatments.  So I'm sure that if I were a study sponsor, if I could avoid the expense of an operative procedure, I would like to do that.  And it certainly would help my enrollment if I didn't have to do it.

            Now another thing that we haven't talked about is using the patients as their own control.  You can measure where they are before your intervention, and then using the same measurement tools, you measure again afterwards.  And this is kind of the classical way that orthopedic surgery has done things, where we don't get high marks in the scientific community for that type of thing, and our literature is considered to be rather weak because we don't have many of these Type I studies.  But still, there is some validity to that.  You might get the placebo effect, but if you follow the patients long enough, I think you probably would get meaningful information from that route.  It does mean that you don't have a control group in the classical sense, but you're using the patient as their own control.

            DR. RAO:  Dr. Tomford.

            DR. TOMFORD:  I just wanted to ?? someone mentioned diagnostic arthroscopy I think as a means of assessment of the joint rather than biopsy.  And I know two or three surgeons who have had studies where patients have agreed to a second-look arthroscopy, so I don't think it's impossible.  It is difficult, but  taking long enough, you can probably find people who are willing to undergo second-look arthroscopy.  It really doesn't provide much information other than the appearance, the feel of the cartilage, things like that, but it isn't impossible, I don't think.

            DR. RAO:  So that leads to sort of the obvious question then.  So while it's not impossible that one can recruit patients to do it, and you can convince them that that may be necessary, is the information you gain from it critical in terms of helping you predict, or plan, or be able to read, or interpret your results downstream?  Does anybody feel that that will be the case?  Do you feel that that would be critical?

            DR. TOMFORD:  I don't think it would be critical.  I think it would be helpful, but I don't think it would be critical.

            DR. COUTTS:  I agree. I think again your clinical measurement tools, a function of pain will be your principal determinants.  You'd like to know what the tissue is that you formed, and it looks like we may be able to do that with MRI. 

            DR. RAO:  Dr. Murray.

            DR. MURRAY:  The Phase 1 and 2 trials are efforts to find out if something may work, and to find out particularly how it may be working, and so that kind of information seems more valuable in those sorts of trials than in a Phase 3 trial where you're trying to find out really does it, in fact, work compared to some reasonable control.

            DR. RAO:  Go ahead, Dr. Harlan.

            DR. HARLAN:  As an outsider listening to all this, I would just say that by far the most important endpoint that I've heard are the functional ones, but in a randomized trial, for the same reason that Dr. Murray just said.  That's the only way you can really know.

            DR. RAO:  So that also brings up one more point, which both of you alluded to just a little bit earlier, is that given we think MRI may work, is that then a strong endorsement that some kind of imaging which is non-invasive would be an ideal way to go to get a measure of what's happening within a joint?  Is that true?  And if that's the case, are there any specific measures that one would imagine?  Is it always useful to get a measure of cartilage thickness, is it always useful to look at osteochondral edema, which you can measure, or is it important to do GAG GEMRIC studies in terms of looking at what's happened, or it's hard to say.

            DR. COUTTS:  There's several critical parameters that you would like to achieve with your repair.  You want to fill the defect, and I think that you should be able to measure with MRI.  You want your repair tissue to attach to the bone and to the surrounding cartilage, and I believe that MRI will be able to tell you that.  And then you want to recreate a surface that is in line with the surrounding cartilage, and that is reasonably smooth.  And I think MRI is going to be able to tell you that, too.  So those would be the key parameters.  And whether or not that tissue is hyaline cartilage, or fibrocartilage, or a mixture is probably less critical, as long as everything else is working the patient, i.e., they don't have a lot of pain, or any pain, and they're able to function.

            DR. RAO:  Before we get to you, Dr. Murray - so does it have to be MRI, or can it be any other imaging modality?  Could it be CT Scan, which is perhaps cheaper, but maybe give you the same level of resolution?

            DR. COUTTS:  Anything which would give you those measurements.  Ultrasound can give you some of that, as well.  I don't know how good it would be in terms of evaluating the surface of your repair, but it could certainly probably give you whether or not you filled the defect.

            DR. RAO:  Dr. Murray.

            DR. MURRAY:  I appreciated what was just said, because it is, after all, the patient's pain and restoration of function that we're after.  So for me, the key question about which of any imaging technologies would be desirable would be a function of how well do they correlate with, and give us insight into relief of pain and restoration of function.  I'm not sure I've heard much data about that, but that would seem to be critical.

            DR. TSIATIS:  Go ahead, Dr. Tsiatis.

            DR. TSIATIS:  I just want to be clear that what we're talking about is just gathering ancillary additional information with this MRI that might help us, and not that this is being proposed as a substitute endpoint.

            DR. RAO:  Yes.  That was not the way I hope I framed that question.  Go ahead, doctor.

            DR. COUTTS:  Just in answer to that, I think the problem is that the resolution of the MRI, and the ability to come up with this information has only recently developed.  The GEMRIC is a relatively new thing, and so are these newer techniques for clearly identifying the cartilage, so I think that this correlation is coming.  There may even be some reports in the literature now, but it's relatively new, and that's why we don't have that correlation.

            DR. RAO:  So one more question just on a practical aspect as you pointed out, Dr. Murray - so from what I heard from what Dr. Peterfy said, was that the MRI scans that were proposed have been around for a reasonably long period of time, and they're not an undue burden in terms of the time or the kind of MRI you need, or in terms of the T2 relaxation protocols that you need to set up; that none of that would be so difficult to set up that it would not be widely available, or it would be available only to a few unique centers.  Is that a reasonable statement, as well?

            DR. PETERFY:  Yes.

            DR. RAO:  Dr. Tuan.

            DR. PETERFY:  The GEMRIC is still more complicated, but it's just a matter of attention to detail.

            DR. TUAN:  Just a question relating somewhat to science, I guess.  So we can collect all the images and perhaps as a function of time and so forth, and of course, all the patient's performance evaluation and so on.  I guess the point of this is that I would like to see this as a database using which we can then evaluate whether some of the earliest images that we get post-operation can predict the long-term outcome, because invariably there will be situations where the images appear different, but the immediate scores appear to be the same.  But then the long-term scores are different, so the idea is that with this database, it would be crucial to use this to have some predictive parameters for performance long-term.

            DR. RAO:  So let me make sure I get this clear, and correct me if I get this wrong.  But it's really been useful to have a joint, for example, like a joint clinical trial database where one can use meta-analysis procedures to follow-up on trials which have occurred in the past because there's a common data set that can be used, or there's a common format for collecting them.  Is that right?  And so, if one collected this information, it would be very useful even on a longitudinal long-term follow-up, even if it was done independently of any particular study.

            DR. TUAN:  Correct.  So I'd like some science to come out of this, that's what I'm trying to say.

            DR. HARLAN:  I heard Dr. Tuan's point slightly differently.  I think it's a good one, but for many studies, diseases with which I'm familiar, the endpoint was life or death, but we had surrogate endpoints which we later found are much quicker to give us answers.  And the CT scan may not be the input we're after right now, or an MRI, but if we get the answer in that in two weeks rather than five years, that would be good data to collect.

            DR. TUAN:  That would be a collateral benefit on that.

            DR. MULE:  It would also be highly desirable to collect that data, if it turned out to be very well correlated with function and lack of pain, which would make it a less susceptible to bias measure under designs that would, for example, not require sham surgery.  It would be less of a temptation to go to those kinds of designs if we had endpoints that were not to susceptible to subject or observer bias.

            DR. RAO:  I think that's what Dr. Tuan emphasized, and Dr. Coutts, in fact, pointed out, as well, was that this would be in addition to, rather than ?? go ahead.

            DR. TSIATIS:  I was just going to say, yes; I agree that collecting this information, or any information that we think might correlate long-term would be useful.  I do want to caution about thinking in terms of maybe developing a surrogate marker for future studies, which is, I think, some of the thinking here, because you have to be a little careful with that, because a correlate, even a strong correlate does not necessarily imply a good surrogate marker.  And in order to be a good surrogate marker, treatment has to mitigate itself through that marker to its ultimate effect.  And that's a high standard for any surrogate marker to hold, and so I just want to caution about what one thinks they're going to get out of these kinds of analyses.

            DR. RAO:  Does anybody have any ?? go ahead, Dr. McIlwraith.

            DR. McILWRAITH:  Well, it's pretty hard to sort of argue about surrogate markers when we're still not really unified on what tissue we want.  Not to bring that up, but when I was talking about fibrocartilage before, we were talking about focal defects, but I think imaging seems to be the best that we have at the moment.  But based on some work we've done in the horse, we're not there yet; like we don't get excellent correlation between our MR images and the histology and filling of the tissue.  But filling is critical, if we can't exactly quantitate the amount of filling, like Dr. Coutts talked about, I think it's really critical.  We've got to be able to do it with really good sensitivity and specificity, because the thing that makes the difference, at least in our studies, between treatment techniques is the degree of filling, and the degree of attachment, which Dr. Coutts also said, rather than comparative histomorphometry, so we've got to be able to image those two parameters.

            DR. RAO:  So let me see if I can summarize this first question, so that it seems to be pretty much a clear-cut consensus that you need more than one single measure.  There's no one absolute measure that's reasonably predictive, that that's all one needs to use.  A critical measure is patient response or gain of function, or change in question by some kind of assessment score.  And that score needs to disassociate between function and pain, and it has to be specific to the kind of defect or lesion that you're attempting to treat, so that you can't have one kind of score which fits all either.  You can just have a score, but you need to have some kind of clinical measure of filling a quantitative way in terms of function, range of movement, or some other measure of that form.  And in addition to that, one should consider some kind of imaging, because that may be independent assessment of what happened with the cells that you put in, in terms of measuring them, in terms of did they fill the defect, was there a uniform surface, was there any infiltration of media, something below that.  MRI seems to be a good measure, but it's not a substitute for the other measures, and other imaging measures which are non-invasive may work as well, as long as they provide this kind of information that one needs.  And these measures are relatively commonly available, or freely available, and they're not an undue burden in terms of either the time or the stress put on the patient.  And as such,  it would not be difficult to collect that sort of information.  And there would be many advantages to that kind of information if that was collected.

            However, there didn't seem to be any consensus or requirement that people felt, or a necessity for, while it might be good, to do routine arthroscopy, or require arthroscopy, or require biopsy, or require some assessment or measure of the material in terms of the biochemical parameters of the material in terms of assessing it.  Does that seem like a summary of some of the statements that people have made on different aspects?

            DR. COUTTS:  You're right.

            DR. RAO:  Does the FDA feel that that has answered their questions to some extent?

            DR. RIEVES:  Dr. Rao, that's exactly right.  The conversation, in general, I think we all agree is remarkably consistent.  It's very useful.

            DR. RAO:  So let's see if we can get to the second part.  And the advantage is again, there seems to be a lot of consensus from some of the discussion that we've had already.  And we did look at it a little bit in terms of A and B, in terms of the control group, and the importance of blinding.  And let's skip that for a minute, and we'll come back to that in trying to put this all together, was the duration of the clinical studies as it relates to assessing short-term, as well as long-term benefit in time-weighted or landmark sort of analyses.  And specifically, at what time point should important endpoints be evaluated in order to assess the success and durability of a treatment effect? 

            This is really hard, because we don't know which cell you're going to put in.  We don't know exactly what the defect is, and we don't know how the patients have been selected, so I understand that we can't really make any very, very specific statements.  But maybe we can try and think of some of the points that people have already made in the presentations here.

            It does seem clear that you need to have some kind of short-term evaluation as to what happened to the cells when you put them in, so there's some kind of short-term evaluation.  It also seemed to be clear that you also needed some kind of evaluation to see what happened over a reasonable time period, because we know that there's a loss or change of function over a reasonable time period.  And there was some argument that what happened after four months, and what happened after twelve months seemed to be pretty much the same, and so those seemed like rough two time measures that were there. 

            And then also in conversation, it seemed what came up was that eighteen months starts getting to be really long in terms of being able to look at outcome measures, or being able to recruit, or being able to follow-up.  And it's not clear what additional benefit one would get. 

            So using that as a start basis, maybe we can ask the clinicians if that was true, how would one look at a trial or time points in terms of some of these measures that we've already talked about should be assessed?  Does anybody want to go first?  Dr.  Scully, Dr. Coutts.

            DR. SCULLY:  And we're talking about clinical studies.

            DR. RAO:  Clinical studies.

            DR. SCULLY:  I think that they're difficult questions.  Control groups is a difficult issue.  At this point, I don't think you have an option for a lack of ?? oh, the duration?

            DR. RAO:  Yes.

            DR. SCULLY:  I don't know.  I think that the long-term outcome of any type of damage to joint surface is going to be degeneration that may ultimately be treated by total joint arthroplasty, and I think that continued observation of these patients that are treated long-term is important, to know whether you've been able to delay and/or avoid arthroplasty.

            DR. RAO:  But continued observation should be prevention of progression, for example, or should it be measurement of repair?  When one says continued observation, what would you consider as a general sort of readout to look at?

            DR. SCULLY:  I think early time points are fine in terms of the measurement of the repair process.  I think long-term outcomes are going to be the delay or prevention of arthroplasty.

            DR. RAO:  Sort of a measurement score, just like a clinical patient assessment score, in your mind?

            DR. SCULLY:  Certainly, you can do that.  You can either work through a patient-based assessment, or you can work through a functional assessment.

            DR. RAO:  Sort of clinical assessment.

            DR. SCULLY:  Right.  And then the other is just you can do a survival analysis in terms of lifetime of the joint. 

            DR. COUTTS:  Well, if I were running the FDA and we're dealing with this particular product, I would feel comfortable taking one-year data, because I think that the repair process will have taken place.  And to the extent that it's going to take place, it will take place within a year.  And the patients will have probably achieved a reasonable level of functional improvement, if they're going to improve, by one year.  And I would feel comfortable making a judgment of safety and efficacy on the basis of one-year data, but I wouldn't be satisfied with one-year data, because there are significant long-term issues for this.  And that's what you really want to know.  But I think it's unreasonable to hold a company hostage for five years in order to get that data.  So I would think that a continued follow-up of the patients at intervals where you would have them fill out these evaluation forms, and assess their clinical function, as well as getting the MRI at different time intervals would then give you the final piece of information that you really want to have for this particular product.

            It's not too dissimilar from what we're dealing with now with total joint replacement.  Almost all total joint replacements are safe and effective in the short-term.  And most of the design changes are for the purposes of extending the longevity of these prostheses.  And we're trying to get them to go 20 and 30 years predictably, so that really requires post market surveillance.

            DR. RAO:  Before I get your comment, can I ask one more question here?  Would it change your opinion in any way if you knew that the cells were surviving for a long period, versus they were being lost in a short time period?

            DR. COUTTS:  If I knew that I had filled the defect, and that the tissue that had formed was attached, and that the surface had been restored, it wouldn't matter to me whether or not the cells that had been implanted were still there or not.  Something is there, and that's what we're going to be dependent upon.

            DR. RAO:  Okay.  And you would still consider then a long-term follow-up would be useful?

            DR. COUTTS:  Without question, long-term follow-up is mandatory.

            DR. RAO:  Dr. Murray.

            DR. MURRAY:  I'm trying to think about it from the point of view of the patient.  I'd want to know relatively soon after how well patients tolerate the procedure, so there would want to be some relatively short-term relevant measures of discomfort, whatever the patients felt were the most relevant things.  And then I would want the, whether it be a four month or a six month measurement.  And I also would encourage - I would be comfortable with the FDA making a decision based on say one-year data, but I would want to see data carried forward for many more years to follow the long-term progression.

            However, I would not use time to arthroplasty as a measure, because I think that whether or not one has knee replacement will reflect many things other than the therapy will reflect, among other things, tolerance for pain, the ability to take off a month from work if you're working, and what kind of health insurance you have, if any.  So since those are all factors that would affect going to knee surgery or not, I would suggest that that can be measured, but it not be taken as a primary measure.

            DR. RAO:  Dr. Tomford.

            DR. TOMFORD:  I think that's good point about the arthroplasty.  Dr. Minas, who does probably more ACI transplants than anybody in the country, does osteotomies on 40 percent of the patients that he does ACI transplants, suggesting that alignment is a key feature of the treatment of the patient.  And malalignment is also a key feature of the deterioration of the joint, so I think that's an important point.

            I've looked at the Knutsen article here.  They really don't show too much difference between one and two-year follow-ups.  I think one year seems reasonable. 

            The third thing I'd like to say is in my experience, I can wash a joint out and give someone three to six months relief of pain, even if I don't practically do anything at all, so I think we have to at least go to a year.  And long-term follow-up would be helpful, although if the procedure is working, you'd like to get it into the general population, so I don't think we ought to hold them to long long-term follow-up.

            DR. RAO:  Dr. Harlan.  Actually, go ahead.

            DR. DAWISHA:  I just had a question about the post ?? it has been suggested that post market follow-up might be useful, and I was just curious what minimal ?? if there is any opinion about what minimal amount of post market follow-up would be reasonable for these types of products.

            DR. RAO:  Hold that thought, and we'll just get back to it after we get Dr. Harlan's comment.

            DR. HARLAN:  Well, I just was going to try to resurrect the knee replacement as an endpoint, simply because if it's a truly randomized trial, the factors that Dr. Murray raises would come out in the wash.  It's true there are factors, but in a randomized trial, those factors should affect both groups the same.

            DR. RAO:  So let's get back to what you said about a longer term follow-up.  For example, with gene therapy, it's always been assumed that you will have a lifetime follow-up, but part of the follow-up has often been well, after five years we can have a postcard with a questionnaire, and then that can be returned.  And that's been considered a reasonable follow-up, as long as you document that that's been done, and that's appropriate, because that seems to be a reasonable level of follow-up that's needed.

            Now in this case, as you pointed out, that it's important to consider what happens even post market in terms of follow-up and looking at patients, what kind of measures would one want to look at if you had a questionnaire?  You pointed out that MRI on occasion might be useful.  Is it useful to have a questionnaire saying did you have to go in for additional surgery?  Should it be some kind of specialized questionnaire that's put together, should be a pain measure, should it be a change in function that one might want to consider?  It is all of the above?

            DR. COUTTS:  I would use the same measurement parameters that were used in the first year.  Certainly, the measurement tool that measures pain and function, I would have that filled out again, and I would have an MRI done at that time, as well.  And whether or not the patient required subsequent surgery I think is an important consideration.  The patients are not going to have further surgery unless they have a problem and need it, so I think it is a good measure.

            And in terms of your question, lifetime would be ideal, but I'm thinking five years would probably be a pretty good measure of how effective the treatment is.  And I can't give you a good scientific reason as to why I think five years would be the appropriate time period, but that certainly gives the repair plenty of time to break down, if it's going to.

            I personally am of the opinion that anything that we put into the human body that is expected to last the lifetime of the individual should be tracked, that there should be registries.  I think every pacemaker, every total joint, every heart valve, every lens implant, we should know where it is.

            Ford Motor Company can tell you if you have a defective part in your car, and yet we can't do that for patients.  And that would be such a huge, powerful tool if we had mandatory registries of all parts implanted in humans that are expected to work and function for the lifetime of that individual.  And that would be kind of a great way for following all of these.

            DR. RAO:  Dr. Scully.

            DR. SCULLY:  But also probably on an annual basis, it would be plenty.  Once a year is probably plenty for evaluation.

            DR. COUTTS:  In fact, you could probably string it out.  You'd have them come in at two years, and maybe three years, and five years, something in that fashion.  And as time went by, and we became more comfortable and familiar with a particular treatment, and found that it worked, then you could probably stretch out evaluation periods.  And the profession will drop it if it doesn't work.  They'll find out.

            DR. RAO:  Does Genzyme want to comment on any of this?

            DR. LEVINE:  Hi.  This is Dave Levine.  I think we're generally in agreement with the sort of statements that have bene made.  I just make a couple of comments on the registry experience that we have had.  In 1995, a registry was established, and it was sort of a pioneering effort because nothing had been done at that time.  That was before the time I was involved, so I give credit to the people who established that.  I think they had the foresight and picked good outcome measures to follow patients.

            It actually takes tremendous effort to follow these patients the farther out you get from the procedure.  We make extensive efforts to find them, and have reasonably good success. I think one of the questions was whether patient-based outcomes alone over the long-term was sufficient, whether you needed some additional outcome measures.  For the early cohort of Carticel patients in the registry, there was both a physician examination and evaluation, and a patient evaluation.  And that initial cohort was followed that way for, I believe, four to five years.  And in each of those years, there was a very close correlation between how the physicians rated the patients, and how the patients rated themselves.  So I think based on that experience, I think you could rely on patient-rated outcomes for the long-term follow-up, and would actually get much higher compliance if it could be based on a patient mailing, or something of that degree.

            DR. RAO:  Dr. Harlan.

            DR. HARLAN:  I really like Dr. Coutts' Ford Motor Company analogy.  Ford Motor Company, though, has the advantage of all vehicles having a VIN, and having to go get a new license plate every year.  Could you comment on your success?  You said it's increasingly difficult to keep up with patients.  How do you keep up with them?

            DR. LEVINE:  There's actually two issues there.  One is, another branch of the federal government enacted HIPAA, and that's actually had an enormous impact on the registry effort. 

            DR. RAO:  A negative impact, or a positive impact?

            DR. LEVINE:  Negative impact.  We were actually very committed to following these patients, and got legal advice, in order to do that in the post HIPAA environment, even though the registry was established pre-HIPAA, that we had to reconsent every single patient.  So we actually went through that effort, and obviously there was some fallout in the registry because patients didn't want to fill out a HIPAA consent.  But to the extent possible, in compliance with the privacy regulations, we've attempted to do that, so I think that's one very practical barrier.

            The other issue, either pre-HIPAA or post HIPAA, at the end of the day, patient participation in clinical research is voluntary, and the patients have to consent to it.  So Ford can stamp that VIN at the factory, and sell you the car with the VIN, whether you want it or not.  We can't force the surgeon and the patient to get a registry identification number and follow them.   I think we make those efforts.

            My joke has been, if we can't find you, you're in the federal witness protection program, because we actually make extraordinarily extensive efforts, both through direct mailings to the patients, through the surgeon's offices, and have actually used search firms to try to find patients.  So as a practical matter, I would emphasize to you, even with the best intentions and the hardest efforts, I don't think you're ever going to achieve that ideal world where you know where every implant for every indication has ever been.  But I think certainly for the initial cohort of patients who are in a controlled clinical trial for product approval, trying to follow them for a post approval time period to see what the durability and further outcomes are, is something that's practical.  But then I would also emphasize that I think the patient-oriented outcome is probably the way to go for that.

            DR. RAO:  Go ahead, Dr. Coutts.

            DR. COUTTS:  I just would want to make a distinction between the registry suggestion and outcomes.  They are actually two different things.  Following patients and getting outcomes data is a much more difficult project, where you really have to interface with the patient.  But in a registry-type of circumstance, when the patient shows up at the hospital to have surgery for that failed previous procedure, that's when you capture them.  And that gives you your long-term outcome.  So that, I think, is quite doable, and the only thing required would be for Congress to mandate that it happen.  Another one of those unfunded mandates that they're so famous about doing, but it would happen.

            DR. McPHERSON:  Mr. Chairman, may I make one additional comment?

            DR. RAO:  Go ahead.

            DR. McPHERSON:  John McPherson from Genzyme.  One of the very confounding variables that we have to address in the context of durability response is that after we've treated a patient, we really have no control of what they do.  And many of the patients we treat are athletes, either profession or amateur athletes, and so if you treat a patient and they go out and start playing soccer again, or doing skiing on the Black Diamonds, it has an effect on the ultimate outcome and the durability of response.  So we've encountered a similar situation when we were developing products to treat chronic non-healing wounds in diabetic patients.  You can get the wounds to heal, but if the patients don't adhere to off-loading of those wounds and proper care of their feet, you're going to have recidivism rate that's very high, but it has nothing to do with how well the product works.  So I think in the context of durability response, it's important to remember that there is this very confounding variable that's very difficult to control.

            DR. RAO:  Circumstances beyond your control.  So it seemed to me from listening to everyone here, was that it's important to follow patients, and you have to do it at short time intervals to look at acute problems, as well as at intermediate, and then at long-term.  And long-term was defined as at least a year.  Less than that wouldn't be very good, and more than that would probably not be necessary, at least from the available data that was there.

            Once you pass that one year, you still wanted to have some kind of long-term follow-up, and you wanted to have some kind of registry, if that was possible.  And in the long-term follow-up, a questionnaire at the very least would be a good way to assess it.  And perhaps one could capture what had happened in terms of progression to additional disease if there was some kind of registry, some kind of follow-up that could do that, either in a questionnaire or some other form.  Does that seem like a reasonable assessment of what one would need? 

            This actually brings up for me, at least, an important sort of conundrum, and that's the following; the kind of things we emphasize in a clinical trial when you're looking at patient evaluation and clinical assessment, and MRI scan; and we just went through and said that for the animal studies we can't really have a questionnaire, and MRI scan is probably not the best way to go with some of the animals, so that we have no clear-cut equivalent correlate in terms of what you'd study in animals, and what you'd study or follow-up in patients.  And so there is a little bit of lack of overlap.  Is there any concern among the clinicians if they were to be making a decision, and then following up with a completely different set, or partially overlapping set of parameters.

            DR. TOMFORD:  Yes, there's a concern.  I think your point is well taken.  I'm not a veterinarian, obviously, but it sounds like there are certain ways to evaluate animal outcomes, perhaps that are not as reliable maybe as the outcomes of the humans, but it sounds like there are some.  But I think that brings up this question of the histology of the tissue.  The animal is the ideal area where you can evaluate the histology of the tissue, but it's not such a good evaluator of the functional.  But I think, nonetheless, I would hope that the procedure is first tried in animals, perhaps to make sure that nothing goes completely wrong.  In other words, it would be sort of an evaluator of things that it's not a disaster, or it's not completely worthless, and then use that data to move into the human realm.

            DR. RAO:  You make a very good point, and we should ?? I want to re-approach that, and get other comments first.

            DR. COUTTS:  I think what you just brought up is what we alluded to earlier when we were talking about the animal models, and how the correlation of the animal model to the clinical performance in humans is not a straight line.  And how the animal models serve a useful function, as Bill just said, for looking at safety and some measure of efficacy, but that the ultimate proof of efficacy will be in the human.

            DR. RAO:  Well, I completely agree with that statement.  I mean, I'm sort of trying to think about this from the FDA's perspective.  Go ahead, if you need to make a comment.

            DR. McILWRAITH:  Well, not disagreeing with either of the previous speakers, but in certain species we can do pretty well with clinical exam.  Like Dr. Coutts mentioned synovial effusion and pain, and we assess all those in the horse, both subjectively and objectively, and you can do it in the dog.  It's impossible, I think in the goat and the sheep, so we look at the clinical parameters.  And I think that the statement I made before about MRI, it's just is the work in progress.  I think we're ultimately going to be able to provide the same imaging that we can in, and that we will be able to in humans.  And so it's not arguing against the fact that we're looking at different things a lot of the time, but there is some ?? the clinical correlations are good.

            DR. RAO:  So is that then an important component that the FDA should emphasize in terms of developing, that there should be some kind of equivalent measure as well, or is that something that's too difficult to do, or something not worthwhile given all the constraints we already heard before?

            DR. McILWRAITH:  In my opinion, I think the clinical - and, of course, we have the imaging, the radiographic aspects, as well.  I think they're very important.  And it's an argument towards assessing doing preclinical studies in an animal that you can objectively assess clinically.  I don't think it's a deal breaker, but I think it's important clinically, because that's what we assess our clinical patients on too.  If we do cartilage restorative procedures, or arthroscopic surgery on horses, like the clinical results are the critical ones, clinical results and performance.

            DR. RAO:  But I mean, by implication does that mean you do the animal studies in horses, or would as one say well, we have to develop a measure of range of motion in goats, for example, or we have to have a substitute measure, we have to make sure we have MRI data in goats?  I mean, is there any sense that any of that's actually necessary, or it's a worry, or it's a concern, or that's probably not?

            DR. HOCHBERG:  While the preclinical models don't do a good job filling out questionnaires, you actually can measure gait and lameness in a lot of these models, so I'm sure as you measure performance in the horses after doing the cartilage repair procedures, that this has been done, I know, in other preclinical models in ACL transection for development of osteoarthritis, tracking abnormalities in gait in those animals, and then looking to see if the abnormalities in gait respond to interventions.

            DR. RAO:  So I just absolutely agree, and I just want to try and get a sense.  I mean, Dr.  Coutts also pointed out earlier, I think, that improvement ?? animal models should not be considered for measures of efficacy or improvement.  That's not a good way to use the animal models, but they're really useful as a quick assessment of what kind of behavior the cells in terms of their safety, et cetera.  So maybe the answer is that it does not really make sense to have such a tight correlation, or have to have an overlap of measures, and that's okay.  I just want to make sure that it's been considered, and that the FDA has consensus opinion of some kind, even if a non-consensus, but some kind of opinion on that front.  And before Dr. Nixon, maybe we can get ??

            DR. MATTHEWS:  May I make one quick comment?  We, too, do lameness evaluations in goats, as well as range of motion measurements and joint width measurements, that's been reasonably successful.  And others have done force plate measurements in dogs with a fair degree of success, so I think you don't have to be limited to horses.  I think those outcome measures are probably going to be something you need to consider in the future.

            DR. RAO:  That's important to know.  Dr. Nixon.

            MR. NIXON:  Yes.  I think there's a tendency to get the parameters for evaluation between the clinical and the pre-clinical research actually rather close together.  I don't see there's such a disparity really, particularly as the MRI becomes more extensively used for research on the animals, for serial MRI exams.  We'll know a lot more, and it will prevent some of the serial or extended animal programs where you're doing harvest of cartilage or tissues at various time points.  It's fairly expensive, and it's obviously a lot of lives.  And the use of MRI certainly limits that.  It allows you to go out long-term with some intervals, where you're actually get a good assessment of it, and that can be done in goats.  It's a little easier to get when you get thick cartilage, so a goat is one of the choices.  A horse is obviously a good choice, but it's limited by the size of the scanner you have.  And that, too, is being resolved, as Dr. McIlwraith mentioned.  It's getting to the point where you can essentially get almost all the entire stifle of a horse still hooked to the horse  with a scanner, which is ultimately where this is going. 

            That would allow you then to have intervals of assessment with MRI, and obviously the clinical findings in multiple animal species.  And then have a harvest and do all the multiple analyses of the tissues.  So it's a big difference really then as you get to see the ultimate result, and perhaps turn the correlate back to what you saw with your MRI or other tests, as well.  But it doesn't seem that different really to the sorts of assessments, other than the sort of patient-delivered styles of performance and their own levels of pain, which you're not going to get out of your animals, or at least not most of them.  So that's the only real difference, so I don't really see a big assessment disparity between clinical and preclinical work.  And certainly, that is actually going to narrow as the time goes on.

            DR. RAO:  That's really important to know, and thank you for that clear summary.  Do ahead, Dr. Harlan.

            DR. HARLAN:  I just wanted to ask any of the three vets that we heard from, Drs. Allen, McIlwraith, or Nixon, do you ever see an apparently good functional outcome in a large animal model that then doesn't correlate with what you find when you look?  If you don't, then your concern is really moot.  If the animals have a good clinical outcome, then it's a good model.

            DR. McILWRAITH:  In my opinion, talking about research cases, the answer is no, we don't see miscorrelation.  And in clinical cases, the clinical signs are still our best parameter, I feel, of the amount of damage you've got in that joint.

            MR. NIXON:  I'd also add the changes, the clinical signs, the sorts of pain, the sorts of swelling you see are usually only useful for the first several months.  Beyond four, perhaps six months, you rarely see persisting pain levels, persisting effusion, so at that point you need to go to ancillary tests.  Simple patient data sets aren't going to do it for you.

            DR. RAO:  Go ahead, Dr. Luyten.

            DR. LUYTEN:  I would like to have one comment.  I think in the animal models, we want to certainly get information on safety. I think that issue, do not harm, is a very important one.  Secondly, we would like to understand more about the mechanics of action, how does this kind of treatment work in the animals?  As I indicated, and as we all know, the cartilage biology is different across species, so it should be two species in which you study and you try to find some commonalities in the mechanics of action of your repair methods.  If you have safety done reasonably well, and you have indications of how it works, then I feel it's unreasonable to ask for an additional third model, a large animal like a horse, in which you are never going to be able, actually, to prove efficacy. 

            Let's not forget that in the field of osteoarthritis, we are running today clinical trials with medications in Phase 2 without any preclinical animal data that are successful in animal models.  I think we should remember that also, so we do already a tremendous effort now in this field of cell-based therapies, and going in at least two models.  If a third one is required which is extremely expensive, not available in most parts of the world, like the horse model - well, how far do we need to go at the end of the day, knowing that cartilage biology is different across species when we start to look at that.

            DR. RAO:  Good point.  Does anybody have any additional comments?  Sharon, as a consumer representative, do you want to say anything for what you've heard?  Do you feel we've covered patient concerns in an adequate way?

            MS. TERRY:  Yes, I always speak up if I need to, and I have not.  You covered them in a very good way.

            DR. RAO:  Does the FDA feel we've covered this issue in a reasonable way?  Do they feel they've gotten enough input?  Well, that's great.  Go ahead.

            DR. RIEVES:  I was just going to say, Dr. Rao, it was very impressive.  Our second question especially was somewhat engineered, but it was very focused, because one of the main questions we get from sponsors is it's so very difficult to demonstrate treatment effects, so the focus oftentimes in clinical development is a bond of the questions of demonstrating the primary import of the treatment effect, if you will.  But the panel picked up very quickly that that is not the sole consideration in clinical development.  The safety should also be considered there, so that was very impressive, and we appreciate that information.

            DR. RAO:  Well, I guess it's the clinicians you chose to represent on the panel.

            DR. MOOS:  If I may just follow-up on that, maybe it might sound like a minor point, but as I indicated at the beginning of the day, what we really need to begin exploratory human trials is, as Dwaine just said, and I think it's been unanimously articulated, very strong data on safety.  And then what amounts to a reasonable rationale for placing subjects at risk in exploratory trials.  And there's more than one way, if I may use the expression, to skin that cat.  And it is not the same to do that, necessarily, as to require that in an animal model you meet the same endpoints as you might in a confirmatory clinical trial.  And I hope that's a reasonable summary of what you've been telling us, and things like comprehensive histology requiring sacrifice will tell us a great deal about what's happening, especially if we can find commonalities across a couple of species that would give us substantial level of comfort to proceed into the clinic absent some of the more difficult endpoints that I think especially Dr. Coutts was articulating.  Fair enough?

            DR. RAO:  Well, on that note we can say we can declare this session today at an end, and see where we convene tomorrow.  I think tomorrow it's at 8. 

            (Whereupon, the proceedings in the above-entitled matter went off the record at 5:30 p.m.)