Wednesday, April 9, 2003


8:10 a.m.










Advisors and Consultants Staff Conference Room

5630 Fishers Lane

Rockville, Maryland




Michael P. Doyle, Ph.D.Chair

Susan F. Bond, M.S., Executive Secretary




     Robert M. Nerem, Ph.D.

     Harold Davis, D.V.M., Ph.D.

     Martin Rosenberg, Ph.D.

     Cecil B. Pickett, Ph.D.

     Josephine Grima, Ph.D. (Consumer Representative)

     Jim E. Riviere, D.V.M.

     Cato T. Laurencin, M.D., Ph.D.

     Katherine M.J. Swanson, Ph.D.

     Kenneth I. Shine, M.D.

     John A. Thomas, Ph.D.




     Norris E. Alderson, Ph.D.

     Robert Buchanan, Ph.D.

     Kathryn Carbone, M.D.

     Daniel A. Casciano, Ph.D.

     Lester W. Crawford, D.V.M., Ph.D.

     David Feigal, Jr., M.D., M.P.H.

     Jesse Goodman, M.D.

     John Marzilli

     Mark B. McClellan, M.D., Ph.D.

     Stephen Sundlof, D.V.M., Ph.D.

     Janet Woodcock, M.D.






Call to Order:

     Michael Doyle, Ph.D.     5

Waiver Statements:

     Susan Bond     8

Introductory Remarks:

     Norris E. Alderson, Ph.D.     10


Welcome and Overview of FDA's Initiative to Improve

the Development and Availability of Innovative

Medical Products:

     Mark B. McClellan, M.D., Ph.D.     13


Quality Systems Approach to Medical Product Review:

     Janet Woodcock, M.D.     41


Quality Teams to Improve Regulatory Processes:

     David W. Feigal, J.D, M.D., Ph.D.     50


Quality Systems for Clinical Pharmacology

and Biopharmacology Review:

     Larry Lesko, Ph.D.     59


Quality Systems for CMC Review:

     Yuan-yuan Chiu, Ph.D.     82


Questions and Discussion with Board/Presenters     94


Update on Pharmaceutical Manufacturing Initiative:

     Ajaz Hussain, Ph.D.     119


Update on Patient Safety Initiative:

     Kelly Cronin     139


Questions and Discussion with Board/Presenters     154


Open Public Comment     173


Fostering Technology Development--Pharmacogenomics

     Janet Woodcock, M.D.     174


Pharmacogenomics--Preclinical Studies:

     Frank Sistare, Ph.D.     193


Pharmacogenomics--Drug Metabolism/Dosage:

     Larry Lesko, Ph.D.     234


Industry Use of Pharmacogenomics and Regulatory Issues:

     Brian Speak, Ph.D.     245

C O N T E N T S (Continued)


Fostering Technology Development--Pharmacogenomics


     Janet Woodcock, M.D.     267


Ethical Issues with Regulatory Review of

Pharmacogenomic Data:

     Benjamin Wilfond, M.D.     277


Questions and Discussion with the Board/Presenters     283


Closing Remarks/Future Directions:

     Michael P. Doyle, Ph.D.     308


Call to Order

     DR. DOYLE:  Good morning.  I am Mike Doyle.  I am the incoming Chair of the Science Board and I want to welcome you all to this spring meeting of the FDA Science Board.

     We might begin by introducing the Board to you.  I am going to have each Board member introduce him or herself, briefly state what you do and where you are from.  Then we will move on with the meeting.

     I am Mike Doyle.  I am the Director of the Center for Food Safety at the University of Georgia.  I am a food microbiologist by training.

     DR. THOMAS:  I am John Thomas, Vice President Retired, Professor Emeritus, Pharmacology and Toxicology at the University of Texas Medical Center at San Antonio.

     DR. RIVIERE:  Hi.  I am Jim Riviere.  I am a pharmacologist/toxicologist at North Carolina State University.  I direct the Center of Chemical Toxicology.

     DR. GRIMA:  I am Josephine Grima.  I am from the National Morphine Foundation.  I am the Director of Research and Legislative Affairs and I am the consumer representative.

     DR. LAURENCIN:  I am Cato Laurencin.  I am Professor and Chair of Orthopedic Surgery and Professor of Chemical Engineering and Biomedical Engineering at the University of Virginia.

     DR. SHINE:  I am Ken Shine.  I am Senior Policy Fellow and Head of the Center for Domestic and International Health Security at the RAND Corporation and I am still going to keep an eye of what kind of food you will eat today because I am a cardiologist.

     DR. SWANSON:  I am Katie Swanson.  I am a food microbiologist.  I am currently Director of Quality and Regulatory Operations for Yoplait, Columbo and soon to be the Director of Global Product Safety for General Mills.

     DR. PICKETT:  I am Cecil Pickett.  I am President of Research and Development at the Schering-Plough Research Institute which is the R&D arm of the Schering-Plough Corporation.

     DR. ROSENBERG:  I am Marty Rosenberg.  I am a microbiologist.  I have recently retired from GlaxoSmithKline, Head of Infectious Disease.

     DR. DAVIS:  I am Harold Davis.  I am a toxicologist and pathologist.  I am Vice President of Preclinical Safety for Amgen, Inc.

     DR. NEREM:  I am Bob Nerem.  I am from Georgia Tech where I am Director of the Institute for Bioengineering and Bioscience and also direct the Tissue Engineering Center.  I am basically a biomedical engineer.

     DR. DOYLE:  Thank you.  As incoming Chair of the Science Board, it is difficult for me to start the meeting with what I consider to be very sad news, and that is, we have had with us for three years Ms. Susan Bond who has been what I call the glue that has held the Science Board together.

     Susan has been our Senior Science Policy Analyst for the Science Board and she is moving up.  She is now going to be the Special Assistant to Deputy Commissioner Crawford.  We are going to miss Susan, but we do want to, as a Board, give her some mementos of our appreciation for Susan.

     So, Susan, if you would join me up here, I will share with you some of our mementos.  Susan is a big fan of bulldogs.  We, at the University of Georgia, are also big fans of bulldogs.  So we wanted to give to her some bulldogs for all occasions.

     This one here--you can cuddle up at night with that one.  This one--you can keep warm in the wintertime with this one.  This one will keep you covered up in the summertime.

     Susan, we really appreciate all that you have done for the Board and do wish you all the best with Dr. Crawford.


     MS. BOND:  Thank you.  I appreciate it.

     DR. DOYLE:  Next, Susan is going to talk about the waiver disclosures.

Waiver Statements

     MS. BOND:  If you could just bear with me.  I have to read these for the public record.

     The following announcement addresses the issue of conflict of interest with respect to this meeting and is made part of the public record to preclude even the appearance of such at the meeting.

     The Food and Drug Administration has prepared general matter waivers for Drs. Nerem, Davis, Grima,  Riviere, Rosenberg, Doyle, Laurencin, Shine, Swanson, Pickett and Thomas.  A copy of the waiver statements may be obtained by submitting a written request to our Freedom of Information Office.  The waivers permit them to participate in the committee's discussion of the FDA's Initiative to improve the development and availability of innovative medical products and to discuss the agency's initiatives in pharmaceutical manufacturing on patient safety.

     The topics of today's meeting are of broad applicability and, unlike issues before a committee in which a particular product is discussed, issues of broader applicability involve many industrial sponsors and academic institutions.

     The participating committee members have been screened for their financial interests as they may apply to these general topics at hand.  Because general topics impact so many institutions, it is not prudent to recite all potential conflicts of interest as they apply to each participant.

     The FDA acknowledges there may be potential conflicts of interest but, because of the general nature of the discussion before the committee, these potential conflicts are mitigated.

     With all that said, I will just give you some housekeeping notes to that.  [Housekeeping notes.]  We have open public comment scheduled for 1 o'clock.  I would just remind everybody to turn your microphones on when you speak so that the transcriber can pick everything up.

     That's it.

     DR. DOYLE:  Next, we have some introductory comments from Norris Anderson.

Introductory Comments

     DR. ALDERSON:  Thanks, Mike.

     In addition to you members that introduced themselves, I do want to add some comments about each of those to give you an idea of the expertise and experience these four individuals to the Board.  Dr. Laurencin is the Lillian T. Pratt Distinguished Professor of Orthopaedic Surgery and Professor of Chemical and Biomedical Engineering at the University of Virginia.  He just moved there and he is telling me he really likes Charlottesville.

     He and I had an experience last night.  I wound up with his bags and he, by chance, just had my cell-phone number before I got away.  So we have a little something to share now.  But his experience is in chemical and biochemical engineering and orthopaedic surgery.

     Dr. Katherine Swanson is the Director of Quality and Regulatory Operations for the Yoplait, Colombo at General Mills and former Director of Microbiology and Food Safety at Pillsbury.  Her expertise is in food microbiology and food science and its impact on public health.

     John Thomas is Professor Emeritus of Pharmacology and Toxicology at the University of Texas Health Science Center.  His expertise is in toxicology and pharmacology.

     Last, Dr. Ken Shine is a Founding Director of the RAND Center for Domestic and International Health Security and former President of the Institute of Medicine and former President of the American Heart Association.  His clinical experience is in internal medicine and cardiology.  He has extensive experience working with global issues and emerging infectious diseases, bioethics and access to care.

     Certainly, we welcome these four new members to the Science Board.

     Mike advised everyone of Sue's departure.  I want to add to that.  I have been in this position now almost two years.  Sue is one of those people I have totally enjoyed during that period.  She and I have a lot of things we share.  She will tell you I am her chief advisor now and I keep telling her, now that she has moved to Dr. Crawford, that that service is no longer free, that I am starting to charge next week.

     But she still works with me until the end of the month.  Her other big responsibility, and a huge one, is this annual FDA Science Forum.  I do want to remind everybody of that.  It is April 24 and 25 at the new  Washington Convention Center.  We think we have an outstanding program this year with both Dr. Crawford and Dr. McClellan being major participants.

     We also have participants, members of CDC and NIH, and, hopefully, the Department.  We hope to know that next week.  But there is a large poster session of FDA research.  The three tracks of the presentation are Risk Management, Counterterrorism and Novel FDA Science.  We have included a copy of the program with your notebooks, so if you have questions about that and you are interested in attending, please see Sue or me.

     We are starting to recruit immediately to backfill Sue's position.  In the interim, until we get somebody named, if you have questions or issues you need to address for the Science Board, you can easily call me and we will take care of it for you.

     Thanks, Mike.

     DR. DOYLE:  Thank you, Norris.

     Next, it is a real pleasure for me to introduce our next presenter, Dr. Mark McClellan who is the Commissioner of the Food and Drug Administration.  We, at the Board, had the opportunity to have dinner with Mark last night.  For many of us, it was the first time to interact with Mark.  But I will tell you, I think the FDA is very fortunate to have someone with his capabilities in charge.

     So I, at least speaking for me, personally, Mark, appreciate you being the leader of FDA.

     Dr. McClellan is going to give us an overview of FDA's initiative to improve the development and availability of innovative medical products.  So, the stage is yours.

Welcome and Overview of FDA's Initiative to Improve

the Development and Availability

of Innovative Medical Products

     DR. McCLELLAN:  I want to thank all of you for coming this morning, especially Mike for his leadership of our critically important Science Board.  This is a very integral part of FDA's policy-development process, as many of you know, and that is why I am very pleased to see so many representatives from the public here today.  We are looking forward to hearing from you later.

     We need this kind of scientific input now more than ever.  I talked about this a little bit last night with our Science Board highlighting how the challenges that we face today are, in many ways, more complex than the challenges that we faced ever before.  But opportunities for addressing those challenges are also better.

     We have had some significant achievements on the legislative front in the past year with new authorities and new resources for both priorities like addressing terrorism threats and our challenges on approving safe and effective medical products more efficiently.  I am going to talk more about that this morning.

     But figuring out how to use these new authorities as well as the changes that are taking place more rapidly than ever in the fields of science is an important and ongoing challenge for the agency.  That is why this kind of interaction is a critical part of our efforts to do our job of protecting and promoting the public health as efficiently as possible.

     All of you on the Science Board have a major responsibility to play in this.  We very much appreciate your stepping up and being willing to take on that responsibility.  We look forward to the discussions here today and I am going to look forward to some more extensive interactions with you in the months ahead as we move forward on these critical initiatives for the public health.

     I, too, want to thank Norris and Susan Bond for their terrific work in putting together this effort.  I hope you all are going to find the day's presentation by FDA staff informative.  I know that they have put a tremendous amount of new thinking into the initiatives and ideas that they are going to be summarizing for you here today and they, like I, are very much looking forward to your critical feedback and appraisals of how we can fulfill these initiatives as effectively as possible.

     The FDA team is one that I have come to enjoy working with tremendously in my four months at the agency.  Les Crawford, in particular, has done a great job under difficult circumstances, working as Acting Commissioner for the better part of a year while I was getting on board and those types of things were happening. I will talk about this more in a minute.

     During that time, the FDA did manage to make some major strides forward on important issues.  Janet Woodcock has done a tremendous amount to organize the presentations for today.  Yesterday, you all got to hear about some of the activities going on in more detail at our Drug Center.  I hope you, like I, have found that work to be very impressive efforts to make the most of the limited resources that we have available to fulfill our mission.

     Our center directors and senior leadership represented here today will be participating in the meeting, too.  Again, I want to thank them for their ongoing efforts to help us fulfill our mission.

     I hope this meeting offers something for our Science Board in return.  We are expecting a lot from you in terms of input and ideas about how we can use the best available science to fulfill our mission effectively but I hope this is also an opportunity for you to hear back from us about some of the front-line challenges that we are facing in trying to think creatively about the practical implications of many of the new scientific developments and new public-health threats facing us today.

     As I mentioned, the challenges of promoting and protecting the public health today are greater than ever.  We are facing new threats of terrorism.  We are facing new opportunities to bring more complex but potentially much more effective medical products to the market and to the public in the months and years ahead.  For the sake of public health, we need to find the most effective ways possible to address these new challenges.

     We don't have unlimited resources at the FDA.  We don't have unlimited staff.  So we have to prioritize.  We have to identify the best opportunities for improving the public health and the best way to fulfill those opportunities.  In many ways, this is requiring a reexamination and updating of the way that FDA does its job.       This is true in each and every one of our centers, our Veterinary Center, our Food Center.  All of our centers are making some fundamental changes in the way that they approach fulfilling our mission given the greater challenges that we face today and some of the new resources and opportunities that I already outlined.

     What we are going to focus on today is illustrating some of the key ideas that are part of FDA's new Strategic Action Plan in the context of working to lower the cost of developing new pharmaceuticals, new devices, and other new medical products.  They must still continue to meet our standards of safety and effectiveness but we are trying to make sure we are using the best science, the best biomedical science, the best regulatory science, even the best economic science, in fulfilling this mission as efficiently as possible.

     This is a particularly important challenge today because of some of the changes that are occurring in the biomedical industry.  This is why I think it is a very good opportunity to illustrate to you some of the key ideas that  we are trying to implement as part of our goal of efficient risk management throughout our activities here at the agency.

     Why is this a special challenge today?  As many of you know, the approval of new drugs and biologics is now at its lowest level since the Prescription Drug User Fee Act was implemented over a decade ago.  The number of truly new drugs, the so-called new molecular entities, that have reached the agency is down around to half of what it was as recently as 1996.  The number of new biologics is down significantly as well.

     These challenges are not confined to drugs but are common to all medical technologies.  At our Devices Center, for example, the average review time for a major new device, as so-called PMA, is now almost fourteen months, probably not surprising given the staff level of the Devices Center has dropped by 8 percent since 1995 and we are facing more complex and difficult device review decisions than ever before.

     So, one possibility for the future is that this kind of trend continues, that the cost of developing new medical technologies keeps rising.  It is easy to see how this might happen.  The kinds of blockbuster drugs that were a hallmark of medical product development in the 1980s and 1990s are fewer and more far between today.  Many of the receptor sites that those drugs have effectively exploited, based on a good understanding of molecular biology in humans, have been exploited.

     We expect to see more of these drugs coming along in the upcoming years but not so many as before.  In contrast, while there has been a tremendous increase in biomedical research, a doubling of the NIH budget is being completed this year, over $27 billion.  Many in the public do not appreciate the fact that there has also been a substantial increase in private-sector research and development contributions.  For example, the research activities of the pharmaceutical companies has doubled since 1995.

     We are seeing a tremendous increase in research and development spending but, so far, it hasn't translated into new products being approved by the agency.  The decline in new-product approvals that I just talked about is a direct reflection of fewer new-product applications coming in to the agency.

     A lot of people think this has multiple causes.  One important cause that has been cited is that the kinds of products being developed now are going to be a bit  different.  Much of the new investment has gone into new, but more basic, types of biomedical research than has occurred in the past.  A new understanding of the genome and genomics, a new understanding of how proteins function at the individual protein level, proteomics.

     There are many other "omics" fields at well that are still at relatively young ages.  Remember, it was just a few years ago that the genome was actually sequenced.  So, translating all this new information into medical products is posing a challenge for product developers and it may also be a challenge for our agency, as I will talk more about in a while.

     So one possible course for the future is, as we get all these additional sources of information, we simply add to the cost and complexity of developing new breakthrough medical products.  Again, it is easy to see how this could happen.  If we don't get a solid and efficient understanding of what all this pharmacogenomic and all this proteomic information in data mean, it could mean additional preclinical tests or additional tests in conjunction with our usual review process.

     Because these enzyme-activity changes that are associated with potentially thousands of different testing sites don't have clearly understood implications for the human body and human systems, we may also end up adding on more testing at the clinical stage to just understand whether a particular abnormality or particular deviation in microarray testing has any clinical consequences for a patient.

     If we take that approach, we are going to be adding further to the cost and time for product development and that is already going up rapidly.  According to that Tufts University study group on this issue, the cost of developing a truly new drug is now over $800 million in present value terms.  While people may quarrel with what the exact number for this cost ought to be and how it should be calculated, nobody can quarrel with the fact that that number has gone up a lot.  According to the Tufts University estimates of a decade or so ago, it was less than half as much.

     A lot of the added costs are coming in at the clinical-testing phase but more costs are coming in in other parts of the development as well.  The time to get a product approved at the agency is going up, too.  So, all of these factors are contributing to increasing costs of product development.  If these trends continue, I don't think it would be a huge surprise that we wouldn't see a significant increase in the number of new products approved, especially if you keep in mind that these new products are not blockbuster drugs that are intended for a market of 5 or 10 or 50 million people but rather of more individualized treatments that are likely to be highly effective but that would be targeted to more particular segments of the patient population.

     It might involve not just a drug but a drug paired with a diagnostic test or a novel drug-delivery system in order to make the treatment particularly effective for a subgroup of patients.  If that costs $800 million or trillion plus, we are not going to see a whole lot of those breakthroughs.

     I don't think that the future needs to be that way but it does mean that we need creative solutions in all kinds of dimension.  First, policy makers need to find new approaches to the problems of health-policy reforms that we are facing today.  As healthcare costs keep going up, healthcare affordability is a front-and-center issue and is going to remain so.  So we need to find better ways in policy making to make healthcare more affordable for all Americans while still encouraging innovation.

     This is a very complex debate.  It involves a lot of money.  There is a lot at stake.  My only concern is that, if we don't find good, creative solutions here, we may, in fact, implement reforms that do not encourage this kind of innovation, all this potential for breakthrough and more effective treatments in the 21st Century.

     So we need to adopt, in the policy setting, ways to realize more value in what we do in healthcare in order to promote the availability of even safer and more effective treatments today and in the years ahead.

     Clinical researchers need to change the way that they are undertaking their activities, too.  I have had a number of very interesting discussions with Elias Zerhouni and other leaders from the National Institutes of Health in recent months.  They are talking, over there, about nothing short of taking steps to re-engineer the clinical-research enterprise to help get around some of the issues that I talked about before.

     We need clinical research that is integrated more than ever with some of these basic science biomedical breakthroughs in fields like genomics so that we don't end up having a longer and more costly and more time-consuming process for getting effective products from the stage of basic research ideas to the stage of proof of concept.

     But, beyond there, FDA has a critical role to play as well, so we need creative thinking in the area of regulatory initiatives as well.  Going from a proof of concept of a new medical product to a product that we can actually have confidence can be used safely and effectively by the public and can be produced using techniques that reliably get the treatment that is desired to the public is not a simple matter.  It requires solving countless more or less difficult and more or less challenging problems in the process of product development, effectiveness testing and manufacturing as well as more effective monitoring, I think, in the postmarket setting, too.

     So we have a major role to play in trying to facilitate this development.  FDA can't cause medical products to come to the public quickly and effectively by itself.  That is primarily the responsibility of the product developers.  But we do have a responsibility to make sure that our regulatory processes are keeping up with the changes that are occurring in medical technology and are responding to the very critical health-policy problems that we are facing today.

     Again, at the top of that list is healthcare affordability, making these treatments available, making sure they are safe and effective but making them available at no greater expense to the public than is necessary.

     So we are undertaking a lot of activities to try to fulfill our responsibilities in this area.  Many of these were announced as part of our medical technology initiative back earlier this year.  I know you all have some materials on that and have been hearing updates from us by e-mail and through other means as has been going along.

     I just want to review some of the main features of that initiative right now.  It had three prongs.  The first prong involved conducting a root-cause analysis of the so-called multiple-cycle product approvals that have occurred recently.  These are products that took more than one round of review before they could be approved.

     An additional round of review means at least an additional ten months or so of time before a product can be approved as well as millions of dollars of additional costs in product development.  In undertaking this effort, we found, in our preliminary looks, that, in many cases, these situations could not have been avoided but, in some cases, perhaps they could have been through earlier communication and feedback, for example, with the companies that are developing products.

     As part of our Prescription Drug User Fee Reauthorization, we are in the process of implementing some pilot programs to test whether frequent early consultation with product developers and whether reviewing applications a piece at a time, rather than waiting for the whole application to come in, can help us avoid some of these multiple-cycle delays by getting the information the companies need to understand their way through the regulatory pathway to, then, more quickly and more effectively, as they are designing their studies and the process, to get to a determination of safety and effectiveness.

     It is going to require some more resources up front through these earlier consultations.  We want to investigate whether that has an offsetting impact in terms of reducing our total amount of time and effort in the product-development process by requiring fewer repeat cycles and we want to see what the impact is on the time and the cost of developing products as well.

     Second, we are developing a quality-systems approach for our review procedures.  The idea here is to apply the best-management practices internally as effectively and widely as possible as we undertake reviews.  You are going to hear much more about this today.  David Feigal is going to talk about a quality-teams approach that is being implemented at our Devices Center.

     Larry Lesko and Yuan-yuan Chiu will be talking about initiatives in the two implement quality systems in clinical pharmacology and in our new drug-chemistry activities as well.  I want to emphasize that this is a fairly fundamental change that is ongoing in the agency and it is ongoing in all of our product-development centers.  So we will value your feedback on the initial steps that we are taking in this direction.

     A third part of this initiative is that we are working to publish new guidance documents in areas where we think regulatory pathways could be improved or better defined.  This includes some areas that we have identified as priorities for new product development where we don't think our regulatory standards have been defined or have been communicated as clearly to the outside world as they could be for diseases where the opportunities for reducing the burden on the public health is actually quite substantial.

     These include new guidances, for example, in such areas as obesity and diabetes treatment as well as many areas of cancer care.  We are also developing new guidances that will address emerging areas of product development.  So, just as a couple "for examples" here, these guidances may consider the use of bioimaging tools to help us more quickly and accurately map drug distribution in particular areas of the body where the drugs are intended to reach.

     To the extent that we can identify valid markers or valid biomarkers that are clearly related to important later health outcomes of interest to clinicians, that can help us in getting through the review process efficiently.  Validated biomarkers may be able to streamline clinical trials by allowing for shorter follow-up times and more confidence that a product is going to have the desired clinical effect and it may allow us to enroll patients who are more likely to respond based on a molecular signature from these kinds of biomarkers as well as to make sure that, when these treatments are actually approved, we can use them in conjunction with diagnostic tests and biomarkers to help make sure that patients are getting a maximal benefit from the treatments that they actually use.

     As part of this effort, we are going to be running a number of joint workshops with outside experts including clinical groups like the American Society for Clinical Oncologists and the National Cancer Institute to help in the development of these biomarkers.

     For other technologies such as cell and gene-based therapies and pharmacogenomics, we also intend to establish partnerships with outside experts including some with the National Institutes of Health to help guide research programs and activities that can lead to approval.

     I want to say a little bit more about pharmacogenomics in this context because this is a major focus of our presentations to you today.  This is one of the areas where we think guidance development can help and where we think there are significant opportunities for improving the regulatory process if we can use the emerging pharmacogenomics information effectively.

     Many people have argued that a stream of new medical breakthroughs could well be the result of integrating genomics information into medicine and making medicine even more of an information science than an art as it is today.  So you are going to be hearing a lot about these efforts in the agency to work with drug developers and other product developers to use these tools and this information to accommodate the transformation here.

     Right now, much of the kind of data that could be used from molecular genetics information technology and related areas is lost to the FDA.  We are not able to take advantage of it because either the product developers don't do certain studies that we think would be helpful or they don't submit the results to us for concern that this is an early science and it is unclear how the information fits into the regulatory process other than to potentially raise some red flags.

     It is true that most scientists, including our reviewers, still don't know what many of the pharmacogenomic patterns that we are starting to see in all these microarray studies really mean for their impact or their predictive value for the response or potential safety problems with the treatment.

     This is a problem.  We cannot improve our regulatory process unless we can use this new information efficiently.  Now, our hope is that this science will progress and that, one day, these kinds of genomic markers will be used to accurately predict a patient's propensity from suffering from a drug toxicology based on their genomic profile to being able to allow us to approve treatments and label treatments so that they can be targeted effectively to the patients who are most likely to benefit, again increasing the value of medical services by avoiding the use of treatments in cases where they are unlikely to have a benefit.

     So pharmacogenomics, if it actually works, is a tremendous tool for providing doctors and our reviewers with a science-based approach to risk management and risk assessment related to new products.  But, in order for us to be in a position to evaluate this science and to help support its productive development, we have to come along the learning curve on this new science along with product developers and researchers.

     We are starting to do this already with work that we are undertaking.  The National Center for Toxicological Research has a major library development program ongoing for genomic and proteomic information.  But we also need help from outside experts.  We need product developers and researchers to have efficient ways of sharing their results with us so that we can incorporate it into what we are doing in our review processes.

     We need to discuss this new science collaboratively, especially at this experimental-research phase of its development.  So, during this meeting, we are going to be seeking your advice about ways that we can make that happen.  Janet, Larry Lesko and industry and NIH representatives here today are going to all be making presentations on this issue.

     Another area where we think we can increase the value of medical products involves our application of risk-management principles to overhaul pharmaceutical good-manufacturing practices.  These are the regulations that govern how medical products can be produced in a way that is recognized by FDA as leading to safe and effective reliable treatments.

     Manufacturing processes often don't get as much attention in the areas of biomedical science as they deserve.  But we are working to change that thanks to Les Crawford's leadership on this issue when it was announced last August.  And Janet Woodcock is chairing--how many work groups do we have in this, like fourteen or something like that?  It is really not a small effort.

     But that is appropriate here.  Our GMP regulations have not been significantly updated or fundamentally updated in a quarter of a century.  Meanwhile, during this time, best-manufacturing practices in other industries that have virtually zero tolerance for impurities or errors have changed fundamentally.

     Think about the semiconductor industry, where it was in the mid-1970s versus where it is today.  Those companies were struggling along then and were wondering if there was even going to be a U.S. industry in this area.  Instead, they adopted some fundamental changes in their manufacturing practices such as 6-sigma manufacturing, total quality-control systems with a goal of zero defects in production.

     As a result, they have been able to march through significant improvements in the productivity of manufacturing, big improvements in output, reductions in cost, enormous improvements in productivity that have translated into better value for consumers without sacrificing quality.

     In fact, they have improved quality.  Their error rates, their precision problems, are lower than ever.  Those kinds of techniques have not been applied to nearly the same kind of degree in the area of pharmaceutical and biologics production as they have been in other industries.  We think there are lots of opportunities whereby, through regulatory reforms, we can facilitate these much needed changes to improve productivity and reduce costs of medical products.

     Ajaz Hussain will be talking to you about an update on the implementation of this program today and I am very proud to say that, even though this program has only been in existence for seven months or so, we are already implementing changes in our regulatory processes to take advantage of these new insights.

     Even if you develop drugs more efficiently, even if you produce them more efficiently, that is not the end of the story.  Too often, the drugs and devices and other products that we regulate are involved in preventable adverse events.  As many as 20 percent of Americans have experienced some kind of medical error and, according to the latest surveys I have seen, more than a third of Americans and an even higher percentage of doctors have family members that have been affected by significant medical errors.

     So we need to do more about that, about understanding how we can help our products to be used effectively not just under the idealized conditions of clinical trials but in the real-world conditions where they are actually going to be used.

     I am particularly concerned there about issues related to people who are using our products in conjunction with a number of comorbid conditions or other medications that are very difficult to test in a preclinical setting, minority populations, other special populations, where particular issues of safety or effectiveness may arise.  We need to do more in the postmarket setting to uncover these adverse events and understand better how to prevent them.       So we are working on developing a range of information-technology tools in addition to some regulatory changes that we have announced just in the past month to do a better job here.

     I would like to highlight, as well, that we have got new authorities in this area thanks to the Prescription Drug User Fee Act Reauthorization and the new Medical Device User Fee Program that we intend to take advantage of.  So, for example, we have announced partnerships with hospitals such as Columbia Presbyterian under our so-called Marconi Program.

     We are expanding the MedSun Program in our Center for Medical Devices to, hopefully, a couple of hundred hospitals and other healthcare institutions in the coming year to collect more data automatically, not just relying on the so-called spontaneous reports from product developers and health professionals but more automatic data collections based on modern IT systems to understand quickly when and why there is a problem with some of the products that we have approved and to give us a two-way street for providing more quick and effective feedback to the health professionals and to the general population that is involved in using these treatments.

     You will hear an update on this program from Kelly Cronin today.  This is, I think, a very important way to increase the value of the products that we regulate.

     Another agency priority involves providing better information to consumers and to patients.  We are not going to have much time to talk about this today given everything else that is going on, but, obviously, it is extremely important that people be able to get accurate information about risks and benefits of a product to help use it effectively and we need effective labeling for physicians as well so that they can get the information they need on how a product has been proven to be safe and effective in guiding their own treatment decisions.

     So we are undertaking a fundamental look at our product labeling requirements for physicians and we are also going to have some activities in the coming year on improving information available to patients.  We are working on some new guidance about risk/benefit information in the brief summary of direct-to-consumer advertising as well, lots of work in these areas.

     Generic drugs are another area where we can increase the value of products available to the American public.  We need to promote the availability of more low-cost safe and effective options for consumers in this area.  There are several hundred major pharmaceuticals that are coming off patent in the next few years that provide opportunities here.  Generic drug manufacturers provide medications that are just as safe and effective as brand-name counterparts.

     So here is an area where we will be expanding.  We propose some major budget expansions in this area and we are going to be implementing some fundamental reforms.  Many people don't know this but the actual time to approve a generic drug is significantly longer than the time to approve a new drug even though it is a less-complex process.  The reason for that is that multiple cycles of review are a way of life in the generic-drug approval process.

     Only 7 percent of generic-drug applications are approved the first time around.  We are going to change that through some fundamental reforms in our generic review process.  We are also in the process of implementing some regulatory reforms in the Hatch-Waxman law that governs generic competition to make that work more efficiently as well.

     So these are just some examples of what we are trying to do here at the agency, to apply the principles of efficient risk management throughout our activities in the context of making safe and effective, better, medical products available at a lower cost.  That involves our premarket review processes.  It involves our manufacturing regulatory processes.  And it involves our postmarket activities to address medical errors and adverse events as well as getting better information to consumers.

     This kind of comprehensive approach is something that the expert staff at FDA have been doing a tremendous amount to implement.  That is why I think the schedule here today is so packed and we really are just focussing on one major, but only one, area where we are trying to apply these principles of efficient risk management throughout the agency.

     I am going to stop my remarks there.  I wanted to give you an overview of what you are going to be hearing about today and, again, give you a major plea for critical and useful input as we work to meet these new challenges facing the agency and these more critical challenges for promoting and protecting the public health that we face today, challenges that are more difficult and complex than ever before.

     Thank you very much for listening to me and thank you again for your contributions to the tremendously important work of this agency.  We very much appreciate it.

     DR. DOYLE:  Thank you for that excellent overview.

     Next we are going to hear from Dr. Janet Woodcock who is the Director of the Center for Drug Evaluation and Research.  Dr. Woodcock is going to tell us about the quality-systems approach to medical product review.

Quality Systems Approach to Medical Product Review

     DR. WOODCOCK:  Thank you and good morning.


     As Dr. McClellan said, one of the parts of our new initiative on innovation--one of the major efforts in our Improving Innovation Initiative that was announced earlier this year is the idea of, "instituting a continuous quality improvement or quality-systems approach throughout the premarket review."  I am quoting from the announcement of the initiative.

     So, first, I would like to run through what we actually mean by that because I think that probably has confused some folks.  David Feigal--I was interested to see, David, in your talk you just handed out that the Science Board review recommended this to you all in Devices as part of the scientific review; is that right?  So, apparently, this is not a new idea and it was already reinforced by the Science Board in their previous review.


     What do we mean?  Well, a review is a scientific assessment of submitted documents and data where we draw conclusions about conformance to scientific, technical and regulatory standards.  So the review that we are doing within the agency on the medical-product side is really a scientific activity of checking conformance to standards.

     In the review document that we produce, we have to produce documentation of this.  These activities are often controversial and we have to write our assessment down.  We have review documents that are written documentation.


     However, the issue with the review process is it is not just this gentlemanly process of doing reviews.  It is a production-scale activity.  For new drugs, this is just a ballpark estimate that I pulled off our tracking systems.  We do an annual number of scientific reviews just in the new drug review side.  About 21,000 scientific reviews are produced.  We issue about 2,500 letters out of a new drug side which have multiple, usually, scientific issues within them.  The generics program estimates they produce about 5,000 scientific reviews a year.


     Although this is mass production, each of these reviews must be scientifically correct, apply the regulatory and scientific standards that have been established appropriately.  They are subject, many of them, to intense stakeholder and scientific legal scrutiny for various reasons because they are implementing standards.

     We need to incorporate any new scientific findings or new regulatory guidance as we go along in these reviews, so we need to make sure each of these 21,000 reviews, for example, has brought in any new principles or any new guidance or anything that we have developed over the recent past.


     How do we assure quality of these scientific activities?  That is really the question I am addressing here.  Obviously, execution is very important in this area, in the regulatory area.  If we have scientific advances, we have to make sure they are incorporated in a uniform and high-quality manner into our regulatory process.

     What has been done traditionally, and I don't know whether this is because this is a medical setting or for other reasons, but, traditionally, there has been sort of a craft or guild approach where there is a hierarchical system of control.  So we have successive or serial checking by different levels of expertise or management over each one of these reviews to assure its quality.

     As a model of quality, that is 19th Century or earlier, I would say, as far as how to insure quality.  It is most successful as this craft model for one-of-a-kind types of products.


     As was already alluded to, there has been a revolution in how to insure consistency in mass production.  That started out with standardization issues back with Henry Ford and others and has moved through quality control, then the concepts of quality assurance and, finally, to quality systems and quality-management approaches which are systemic, systemswide, approaches to assuring quality.


     I know this is the Science Board, but I would just like to take you through.  Some of you are familiar with this, but some of you may not be and I am not going include any jargon here.  I am just going to say what the principles are sort of in a general English-language way.

     I like this set of statements about it.  The quality-systems approach is; say what you do, do what you say, prove it and then improve it.  That applies to the system as a whole.  What does that mean?


     This sounds simple, but it is actually much more powerful and I think we may illustrate with some of our presentations how this can be applied.  To say what you do, you need to identify what your vision and the purpose of your organization is and who the customers of your organization are, and even subparts of the organization.

     You need to define quality.  We are struggling with that right now in the GMP world, how do you define the quality of a manufactured product, a pharmaceutical, another medical product.

     Then you define the attributes that you are going to measure the quality by and you define the processes by which you are going to produce this quality.  That is saying what you do.


     Doing what you say, you measure and produce things that have quality attributes and you also have another important step which I think the FDA has pretty much accomplished in many areas, which is process management; that is, you define the process you are going to follow.  You standardize the process.  You track things within the process and you control the process.


     To prove it, you make sure that whatever your customers are expecting that you are producing that.  You do trend analysis and have other metrics of whether you are achieving the quality attributes that you set out to achieve.  And you do audits and evaluation.


     Then improve it has been the most challenging and difficult, I think, for anyone involved in the quality area because that is where you need to do corrective and preventive actions.  You need to do feedback and training.  If it is about science, you need to incorporate that science in and prove--the organization needs to learn and change as you are moving forward.

     So that is sort of a short description of quality systems free of any particular jargon.


     But you might say to me, and that is what the reviewers say to us, "Review is an intellectual activity.  The review product is a scientific document."


     How can it be reconciled?  How can we apply on to this a quality-system framework that really did originate in mass manufacturing?  How can we bring these things together?  That is what we are here to ask you because we don't know totally, precisely, either.


     But, in fact, the systems approach works very well for many processes.  Sure, it originated in mass manufacturing but it is now being applied many places.  I read this morning in the newspaper that a healthcare system won the Malcolm Baldrige award this year for applying quality-management principles to healthcare demonstrating, demonstrably improving the quality of the care within their system.

     There are a variety of tools and methods that can be selected for any specific application so we don't have to pretend we are making widgets and then apply quality principles of widgets to our process.  We are exploring examples of how these approaches can be used in scientific activities; in other words, getting people in to talk to us about how they are applying these principles in science.


     I think David is going to talk a little about CDRH and the situation there.  In the Center for Drugs, we have completed considerable work on process-management aspects such as procedures, tracking and training.  We have also done considerable work on some work-product standards.  I already told you our work product is a review.  So we have gone through and we have established templates for the review documents, what they should look like, and some directions on good review practices; in other words, what the review activity should be, what items should be covered and so forth.

     We have done some work on quality assurance mechanisms and feedback.  We are instituting audits and so forth to see how well these standards are followed.  We have done very little on systems aspects such as peer discussion and assessment, CAPA, organizational learning and also the metrics, the overall metrics.  So some of the presentations we are going give you this morning are to show where we are in a couple of different areas, not every discipline, and then get some of your feedback.


     The next presenters are going to give examples of some progress and also some proposals for moving forward in this area.  Our question to you is really what steps are most important do you think, as we move forward in this, for integrating evolving science into the review process.

     There are many things we need to do, obviously.  Which ones do you think will be most important for us to implement first with a focus on integrating the new science, making sure that our reviews contain the up-to-date scientific information and apply that to the review process.

     So our next speaker will be David Feigal.  He is going to talk about what the Device Center is doing in this area.

Quality Teams to Improve

Regulatory Processes

     DR. FEIGAL:  Thanks.


     I would like to actually tie some themes together and this is a bit of a progress report.  Fifteen months ago we had a presentation to this Board of Science Review that was led by Robert Nerem.  It was a process that the Center spent about a year preparing for, and we took the recommendations of that group very seriously and actually promised to perhaps even come back with a more comprehensive follow-up report and actually pull the committee back together.

     But there were two of the recommendations of the science review to the Center and one of them was about quality systems.  CDRH should develop and implement a quality evaluation and improvement program, and this should include metrics in addition to timeliness of reports.  We have always had metrics for timeliness.  We produced about a 40-page annual report that slices timeliness about a hundred different ways.  But it has been more difficult for us to grapple with quality, and this is one of the things that the Science Board recommended to us.


     It is also a theme that is really a theme government-wide and the whole issue of how do we measure performance, how do we measure value added--this is just a clip from one of the federal government trade presses about the government-wide effort to be able to actually quantify the value that we add.  Historically that has been done in terms of review times but as the Center began the process of having user fees to speed review times, the Center became even more interested in showing that we can improve the quality at the same time we improve the speed.


     This fits in with a number of themes.  It fits in With Dr. McClellan's initiatives that he spoke of this morning.  It fits in also with the strategic plan that the Center developed, and as part of that strategic plan we have begun a process of developing a scorecard where we will actually be accountable, and describe in real time the areas that we think are key to our performance and key to some of the changes we need to accomplish.


     I just wanted to show you the areas that we think are areas for key results and how we will indicate these.  Some of these relate to our central mission of public health protection.  We want to look at the effectiveness with which we identify hazards and resolve them.  Some of them are on promotion in terms of timely availability of novel products.  But in terms of operational accountability and quality, which is where we have had most of our metric, there we want to actually look at our application activities, all of the conformity assessment activities including our field programs and compliance activities and even business accountability and look at the quality of those systems.


     One of the things that we did as a strategy to do this, we actually began working with an author in this area, Dr. Richard Cheng, and one of the things that we contracted with him to do is to actually train our staff in a method of continuous process improvement to actually take to some of our staff.  We have identified 18 people and they have worked half time in training for the last year to actually learn how to be leaders and coaches in one of the methods.  Many of the quality systems have similar elements and this is one that Dr. Cheng uses for teaching materials, but to actually have an intellectual framework for this and to have a team that actually can move out into the Center.  So, working with one contractor we actually leverage that off 18 of our staff to begin this process.


     What I just wanted to show you is one of the processes.  I am just clipping three slides from the middle of the 20-slide presentation of one of the teams.  What the teams do, they pick a continuous process improvement area and they take the review process and break it down into the components.  They take many of the elements that Janet talks about, not just looking at the process but also looking at the quality.

     For example, they took a look at PMA filing decisions.  This is just simply the facts, we didn't file 11 percent of them in the study period, which was an 18-month study period.  But what was done is a little bit different.  Instead of just looking at the ones that weren't filed, and that is an activity all the centers do because those get appealed and those are under scrutiny, but they also did peer review analysis of the applications that were filed because there is often the sense that there is tremendous pressure to file and can we identify whether or not the filing clearly met our guidelines and standards or whether there were some issues with that.


     The primary issue, and this will actually help us work with industry to improve the quality of the submissions--the primary issue, even in the applications that we filed, were concerns about inadequate clinical studies in the applications and disorganized applications; then, less frequently, lack of information about the product.  I only bring this up as an example of a process.  This is one of 18 projects that are implemented in the Center by this team.


     One of the ways that they take this issue apart, and this is again just focusing in on the filing issue, is to look at what are the opportunities to really make improvements; what are the root causes of the problems.  Some of them deal with our staff; some of them deal with relationships between different units in our group; and some of them deal with industry.


     These are just examples of the way that this team approached this problem, looking at various options that could be taken both to staff and to the outside stakeholders.  Specifying minimum filing criteria, we have this.  The question is, is it up to date.  Providing feedback, we actually don't do that very much about the quality of applications, except in an informal way.  Education.


     For turbo filing we actually have successful applications that are computerized and that have been nicknamed turbo, so that is what turbo refers to.  Then, the issue of actually having operational accountability for this, a lot of times there is a feeling, particularly when time frames are tight, that standards have fallen or things have changed.  If we are actually measuring this and monitoring this in an ongoing fashion we can actually talk about this.  Of course, we are very interested in the stakeholder collaboration.


     So, I show that just as an example to highlight the strategy that we took, which is to develop some internal competence in this area.  As busy as we are, and all of these people had day jobs and useful things that they did, one of the things that struck us was that we needed to develop some internal resources.  One of the questions to you is are there opportunities to actually learn from business or other opportunities to learn from university systems that do similar things.

     The other thing I wanted to give you some quick feedback on is a second science review recommendation which I think also relates to quality and peer review in a perhaps more indirect way.  That was a recommendation about the use of expertise, particularly external expertise.  Dr. McClellan alluded to this a little bit as well.


     Since the science review we have developed a device fellowship program.  The types of people that have participated in this range from the very senior, including a sabbatical that is planned this fall with a chairman in surgery from a medical school, to more mid-career faculty, to fellow, residents, medical students, engineering disciplines and other health science disciplines.  This is a program that we have already had formal and informal relationships and fellows join us from all of these different universities and, actually, more contacts to come with this.


     We have had cardiology fellows from Brigham and Women's come and spend six months with us.  We have begun co-op programs with engineering schools and working with faculty.  So, this is something that we are actually quite excited about and take quite seriously.


     One of the things that was very helpful for this board--this is just a list of the kinds of expertise that we are looking for.  One of the ways that this board was very helpful and the science review was very helpful is that when it came time to actually negotiate the user fee bill, the science board really very clearly laid out the need for resources for external expertise.  In the language of the letters that went back and forth about what was expected from the user fees it was actually specified that we would actually try and develop enough external expertise that would consume about 20 FTEs of staff.  To put that into perspective, the Center has 300 special government employees on its advisory panels.  There are 17 panels.  They each meet about twice per year and that is one of our major sources of external expertise.  All of that time from those 17 panels each year only adds up to 3 FTEs.  With bringing in the fellows program, we have already added in part-time people the number of hours equal to all of the consultation from those panel meetings.  But that still only is about one-sixth or one-seventh of where we intend to go with this and what we intend to develop.

     So we appreciate your support.  I think that the strong recommendation to do this and the strong interest in external expertise from industry combine to actually make this part of the user fee bill so that it wasn't just focused on review time with us.


     Implementation--we always feel in the hot seat, and we appreciate your support.  Thanks very much.

     DR. DOYLE:  Thank you, Dr. Feigal.  We are going to have a break and then two more presentations and then we will get into a discussion for about 30 minutes.  So, at this point we can take a 15-minute break and reconvene at 9:30.

     [Brief recess]

     DR. DOYLE:  Next we are going to hear from Dr. Larry Lesko, who is with the Office of Clinical Pharmacology and Biopharmaceutics with CDER.  He is going to tell us about quality systems for clinical pharmacology and biopharmaceutics review.

Quality Systems for Clinical Pharmacology and

Biopharmacology Review

     DR. LESKO:  Thank you, Dr. Doyle and good morning, everyone.


     Thank you for having me here to talk about quality systems for clinical pharmacology and biopharmaceutics.  I look forward to hearing what your comments are after the combined presentations that you are going to hear this morning.


     What I am going to do in the time I have is cover five topics.  I will talk about the organization and responsibilities of our Office.  I will talk about what we have instituted as quality systems approaches to reviews; the benefits to customers, both internal and external; the metrics of improvement, although these won't necessarily be hard numbers; and some of the future goals that we have in the area of quality systems.


     Let me begin by introducing the Office to you.  You can see in the organizational chart, down to our divisions and teams, we are structured very much like any other office with an upper management level responsible for the strategic management and implementation of the program.  We have three divisions, each in three separate locations throughout the Rockville area.  Then, we have 15 therapeutic teams, each led by a team leader with anywhere from three to nine reviewers per team depending on the business of that therapeutic area.


     This will give you a sense of the demographics of our scientific staff, at least with respect to their educational background.  All of our reviewers have advanced degrees.  Many of them have post doctoral experience of anywhere from one to three years.  Several have dual degrees.  Most of those dual degrees are M.D., M.P.H.D. degrees.  We have a fairly active recruitment program because it is necessary.  There is a relative shortage of clinical pharmacologists within the profession and we are in deep competition with the pharmaceutical industry to get talented people.  We recruit actively at colleges of pharmacy, from clinical pharmacology training programs, schools of medicine, and while we don't recruit from the pharmaceutical industry, we average about three scientists per year coming to the Office from the pharmaceutical industry with experience.

     Our current staffing is about 95, and 70 of those represent the scientific review staff.  On the right-hand side you can see the breakdown of our review staff in terms of their degree.  Most of them are Ph.D. but we have a sprinkling throughout our three divisions of both Pharm.D. and M.D. as terminal degrees.


     Let me define for you briefly what I mean by clinical pharmacology.  It is a science dealing with human properties of the drug substance--and I have underlined "drug substance" to differentiate it from the drug product--which occur after the release of the drug from the dosage form.  So, these are pretty much systemic properties of a drug substance, as well as nonclinical characteristics of the drug substance that relate to these human properties.

     To give you an example, clinical pharmacology would encompass the pharmacokinetics of the drug substance, the absorption, distribution, metabolism and excretion properties either in healthy volunteers or, for example, patients with renal disease.  Another example of clinical pharmacology data that would be nonclinical would be the solubility and permeability characteristics of the drug substance that are important prerequisites to getting the drug absorbed.


     Compare that and contrast that to biopharmaceutics, this is a body of science that deals with the in vivo performance of the drug product once the drug is formulated into a drug delivery system and the in vitro properties that relate to the drug product.

     I don't mean to separate the two sciences.  They are integrated and they are overlapping but they do have distinguishing features.  Examples would be bioavailability of a tablet.  It is the percent absorbed or perhaps bioequivalence of two capsules.  Another example might be the rate of dissolution of the drug product.


     This is an important aspect of our quality system because it represents the flow of information throughout the review process and it is categorized in terms of the scientific and information content of what we receive in the Office.  It beings at the top with data that comes in as part of the NDA.  I consider data simply facts.  It is raw measurements and some of the examples of that are illustrated on the slide.  The review process, however, takes that data and begins to make connections between it, associations, cause and effect.  I call this information and this is what is important in terms of understanding what is going on with the drug product.

     As a review proceeds, we move down into the third hierarchy of knowledge, and this is very important because it really represents the context for applying the information that comes out of the review process.  This is what we know to be relevant to the therapeutic situation, and this is also a prerequisite to an effective review in terms of the risk management aspects that we have heard about today.

     All of this informational content, all of these hierarchies are in the past because they rely upon the review of the NDA.  We try to go a step further in our review process and look at the knowledge component or the fourth category which is more futuristic.  We attempt to understand things a little bit better, to synthesize new information that may be relevant to future reviews.  In this category of knowledge we will frequently apply modeling and simulation, predictive tools to try to look at "what if" situations that may be relevant to the current review or may be relevant to some future review.  So, as we proceed through the review these are the stages that we hope the review process takes on.


     Where the information comes from is, of course, the drug development process.  I have illustrated it here in a very simplistic linear fashion.  But, as you can see, the clinical pharmacology and biopharmaceutic information comes from all of the clinical phases of drug development as well as the nonclinical portion of drug development, and under each of those boxes of clinical pharmacology and biopharmaceutics I have highlighted just a sample of the types of information that comes out of the drug development process that is subject to review.


     The regulatory review process then is intended to move us through the data to the knowledge hierarchy and convert that information to customer-related knowledge.  I am referring here to the internal customer.  I will talk about external a little bit later, but this is the internal customer who I see as the other disciplines that we integrate our review with.

     We expect that the reviewer becomes an interpreter of data.  While they may assess all data in an application, we ask them to prioritize and selectively analyze only that data that has the most impact on the risk/benefit assessment.  We ask them to look at the factors that most likely impact efficacy and safety and, just as important, we ask them to look for data that is missing, the gaps of information that we will have to deal with in the product label or in a Phase IV commitment.

     We emphasize mechanistic understanding at all levels from the cell to the patient, and we view our goals as translating science into therapeutics.  We view clinical pharmacology as a means to an end, and the end is an effective risk/benefit assessment when coupled with the other disciplines.

     Our primary scientific focus in the Office currently, as it has been, is on adverse events and managing the risk of adverse events and understanding the likelihood of those adverse events.  For that reason, we focus in particular in our review on exposure-response relationships, drug-drug interactions which, we understand, is a major cause of preventable adverse events; and on the integration of new technology and evolving science into the review process.


     Let's get to the quality systems approach to review and speak in terms of the customer.  This is the internal customer.  Our primary customers are the medical officers who rely on our assessments, along with their own, to make some judgments about risk/benefit.  Our systems approach involves placing as much emphasis on reviewing and describing the connections between data and studies as on reviewing the data in studies themselves.  In other words, we emphasize an integrative type of review.

     The figure on the left-hand side shows our Office as a matrix.  it covers the five ODEs in the Office of New Drugs.  On the right-hand side, under good review practices, are the ideal goals of our review.  It really reflects our vision and mission within the Office.  We want to generate knowledge as part of our good review practice.  We want our reviewers to be decision makers.  We want our reviewers to understand patient context for what they are doing and to recognize the medical need for knowledge that medical officers have and, most importantly, to communicate the science in a very clear and useful way, and not in the jargon of the individual scientists.


     Our GRP is based on something we call the question-based review.  It occurred to us some years ago that when NDAs come into the agency they don't come in with studies in any particular order.  They may come in the order that they were conducted in for drug development, or they may come in the order that the sponsor would want us to look at.  Instead of reviewing studies one by one in sequence as they appear in an NDA, we felt it was more important to put those aside and focus on the pertinent questions, not the studies or the sequence of studies.

     So, we developed a question-based review that is intended to integrate knowledge across different nonclinical and clinical studies in a way that addresses what the reviewer perceives as the key safety and efficacy issues.  In addition, the reviewer looks at the information in a way that links it to the claims being made by the sponsor in the label of the drug product.

     As I said, there are often gaps in information so it is important for the review process to know the risks associated with uncertainty or gaps of information that are in an NDA and use appropriate label language, or perhaps Phase IV commitments, to manage those risks.


     Our good review practice timeline is not that old; it is still evolving.  But, as you can see, we began using a question-based review in July of 1999 and went through a series of educational steps.  We had an Office retreat to talk about it within our Office.  We had a voluntary implementation of it and a formal requirement of it in October of 2001.  We are currently at the point where we are going to stand back and assess our GRP and question-based review for compliance and for effectiveness and determine where else it might be applied within the review process.


     These are the five quality subsystems of our systems approach, and I have numbered them in a clockwise direction to give you a sense of the quality subsystems.  It begins at the top with a template.  We call it the clin. pharm./biopharm. review template.  It is a very important template because it brings uniformity and consistency to the review process across three divisions and 70 reviewers.

     What we like about it, and I hope the users of it like about it is that it standardizes the order and placement of the subject matter.  It doesn't dictate the content.  That is still an individual scientific endeavor.  But when you pick up one of our briefing packages, over time and in different therapeutic areas, it should look the same in terms of the format and consistency.

     We build our quality system not only around the regulations but the domestic and international guidances that relate to our area.  Coupled with the regulations, these form our review standards.  We have guidances for reviewers we call MAPPs or SOPs.  These are both scientific and a process that creates the construct for the review process.  In those MAPPs are frequent decision trees that aid the reviewer in analyzing their data and making decisions.  We stay on top of new issues and new science through our team leader meetings, from which oftentimes new scientific policies emanate.

     Then, number four, we get into a very important part of our quality systems, and that is our NDA clin. pharm./biopharm. briefing package.  This package, which represents the product of the review, is posted on the Internet.  It is a very open, transparent process.  It contains something for everyone.  It contains an executive summary, a listing of key issues and important unresolved questions and recommendations related to the approvability of this section of the NDA.

     Finally, the quality systems approach ends with the NDA clinical pharmacology/biopharm. briefing.  This is the highlight of our quality systems.  This is a formal presentation by the primary reviewer of the NDA.  They make a presentation to an inter-disciplinary audience.  The meeting itself is characterized by a very good interactive dialogue and it represents in many ways the vehicle that we use to assess the quality of the review and a reality check in terms of its relevance to the needs of other disciplines. Following that briefing, the review package is reconsidered as necessary and then finalized for the records.


     As I said, we are talking about internal customers and what have been sort of the expectations and the benefits to the reviewers with this approach.  The quality systems approach has been excellent.  I think what it has done for us is that it has standardized our method of delivering a comprehensive review.  This has led, I believe, to a higher quality of review.  Our reviews usually meet and usually exceed and frequently anticipate our internal customer expectations, and I believe this has built over the years a significant amount of trust with our other disciplines.

     I think our review process collectively has a greater ability to derive clinical inferences from the data.  This has helped communication and I believe it has led to a more effective review.  I think the process has enhanced the critical thinking of our reviewers through the question-based review and this has led to some efficiencies in our review process.

     Finally, I think there is an ability to recognize efficiencies more in our clin. pharm./biopharm. data set than we had in the past.  I think this has instilled in our reviewers a large amount of confidence.  I think it brings a large amount of job satisfaction and high morale to our staff as they go about their business of intellectual reviews.


     As I mentioned, the briefing is really a critical component of our quality system.  Because it is a formal presentation, I believe it compels reviewers to think very deeply about their review.  There is individual accountability as you stand up in front of a group of 20 or 30 individuals.  There is a pride of ownership.  The feedback and the seeding of new ideas from our briefing is also critical.  The attendees, as I said, are medical officers, toxicologists, chemists and this is critical to our Office in terms of continual learning and knowledge sharing.  There aren't many opportunities to do this within a review process.

     For the attendees, I think in this setting there is a greater consideration of the comments before speaking because it is a showcase of people's comments and I think, therefore, the quality of what is being discussed is high.  We also invite other reviewers.  They may be new; they may be inexperienced.  We invite them to the briefing.  It is a good teaching tool and it is a form of professional growth and the presenter, the reviewer can lead by example in this presentation.


     I have talked pretty much about internal customers.  Let me turn to external customers and what we think about in a quality systems approach here.  We think about their needs and expectations, and I am only going to use one example, drug-drug interactions, because it is such a predominant part of the adverse drug reactions scene.  With health professionals, we know and they know that drug interactions represent a high percentage of preventable ADRs and are an important contributor to ER visits and hospital admissions.  We know that from the literature; we know that from hearing from them.  Patients, what do they worry about?  Their top two concerns are getting the wrong drug and getting two drugs or more that interact.  So, the problem is very real.

     We feel the work product of our review is the delivery of prescribing information in the label that addresses the needs of health professionals and patients.  In the past the drug interaction section was very much a fact-based presentation of pharmacokinetics, area under the curve, Cmax.  What does that do for the physician and the patient?

     There are many potentially harmful drug combinations.  This information was often poorly utilized by consumers.  Number one, they had a hard time finding it in the label.  Number two, when they found it and they looked at two labels the presentation may have been inconsistent and unclear.  Finally, what is the significance of increases in area under the curve, and what is the significance of this increase of 20 percent versus that increase of 40 percent?  So, it was a real difficult interpretive problem.


     What we have tried to do over the last year in particular is to bring more added value to the work product and improve the labels, using drug interactions as an example.  We still, by law, by regulation, have to provide fact-based changes in pharmacokinetics in the label.  However, along with the agency's label initiative, that information is now being moved to a more prominent part of the label so it isn't lost in the shuffle of clinical pharmacology information so it is moved to the top of the label.

     We have begun to establish a common language to describe drug-drug interactions.  We now classify drug interactions as being strong, moderate or mild.  That gives the physician and the patient a sense of risk when they use two drugs together.

     Along with that, we have also a parallel process of standardizing the method of interpreting drug interactions so that we can say something about what to do in the label, and this usually involves an analysis of pharmacokinetic changes using what is known about the exposure-response relationship for the drug.

     We are also moving towards expressing results in terms that are meaningful to clinicians such as probabilities or odds ratio so that somebody can make an informed decision.  We don't know the success of this.  It is in progress.  We hope to survey customers at some point to assess the value.


     As far as metrics go of quality systems, our metrics of process improvements since we have introduced our question-based review, good review practices are maybe soft metrics but valuable nevertheless.  We feel the consistent format of reviews that we have brought to the process has made our review packages very readable, which is a very important aspect in a busy scientific environment and, as I said, various levels of detail.  It is a predictable package.

     High quality reviews that I feel we produce address issues of relevance to therapeutics, and I think this has resulted in more informative questions of sponsor data that comes in an NDA.  I have noticed over the last five years a significant increase in attendance at our briefings.  People only attend meetings if they are valuable and I think this is indicative of a metric.  We have had greater staff participation at advisory committee meetings.

     I think also through different means--CDER awards and other things like that--we have achieved greater recognition of the expertise and leadership in our Office.  What I like best to say is we have frequent requests from the Office of New Drugs for more reviewers so I hope we are providing a need--additional invitations to speak at advisory committees and an obviously increasing role at industry meetings.


     Some other metrics--efficiency and informativeness of drug development.  I think what we have done over the last five years is actually identified studies that may be more germane to regulatory decision-making and public health.  As an example, through our efforts and our interactions with industry we see an increasing number of high quality drug-drug interactions which is a major issue in public health, but along the way we have also been able to identify and convey to industry the idea that some studies aren't necessary for regulatory decision-making.  I don't know if the companies did them for reasons that they had of their own or they thought we wanted them but, for example, the number of bioequivalence studies in applications has decreased significantly since we indicated clearly, through guidances and one-on-one meetings with companies, where they are necessary and where they are not.  I think this allows us to shift emphasis.

     Our most recent initiative is to emphasize exposure response studies.  While they are good, they could be much better and much more informative in regulatory decision-making and we are currently communicating that to industry through guidances and public meetings.


     Let me just say a few words about this.  Dose response and PK/PD represent exposure response studies and we feel they are really the core or the hub question in our good review practice as it applies to the submissions from sponsors.  It is a quantitative approach to assessment of efficacy and safety.  The reason that it is important is that it allows in the review to assess safety and risk of dose selection in quantitative ways.  It allows us to evaluate the risk of exposure changes in special populations because of drug interactions or disease and come up with a rational basis for dose adjustment.  As formulations evolve during drug development, it allows us to assess the clinical significance of differences in formulation.

     In order to implement this in the review process we have had to make use of extensive pharmacometric tools such as modeling and simulation and other statistical approaches.  It has been an intellectual challenge for us.  In the near future we will have a guidance for industry that describes a lot of the principles that I am referring to.


     This has had a significant impact on quality of our review process and I think on drug development because, if I look at three categories of impact of our involvement with exposure response, I find that we have case studies that have demonstrated improved study design that assess benefit and risk during the end of Phase II review.  In the long term, this can increase efficiency in drug development through greater interactions at this point in drug development.  We know that some of our case studies support approval of doses that are deemed effective and safe during the NDA review, particularly when the doses are different than those that the sponsor proposed, or sometimes without additional clinical efficacy and safety studies which is a powerful impact on efficiency of drug development.

     Thirdly, our examples of impact have identified absent exposure response information that, had it been available, could have supported efficacy and safety and avoided delays in approval that resulted from conducting additional clinical trials.


     Let me conclude with saying that we have future goals.  We are currently in the process of a QC/QA check of our NDA reviews.  We want to check two things.  One is the compliance with our good review practice MAPP.  We want to get reviewer feedback on the experience of the last 18 months and make revisions or tweaking of our process as necessary, and also focus on more added value areas, both internally and externally.

     We want to look at the application of good review practice to INDs and supplements.  We have only focused on NDAs, for the most part.  We want to establish a standardized method of assessing experience response, particularly for dose adjustment when intrinsic and extrinsic factors change exposure.

     Finally, we think it would be valuable, and we have said this publicly in other settings, to interact with the industry at a much earlier stage of the drug development process to talk not only about specific studies but about drug development plans and to jointly maybe think about what is necessary and what is not necessary for the decisions we have to make when that NDA eventually comes through the door.

     With that, I will conclude and say thank you for your attention and turn it back over to Dr. Doyle.

     DR. DOYLE:  Thank you very much, Dr. Lesko.  Next we are going to hear from Dr. Yuan-Yuan Chiu, who is the Director of the Office of New Drug Chemistry at CDER.  She is going to address quality systems for CMC review.

Quality Systems for CMC Review

     DR. CHIU:  Good morning.


     It is my pleasure to be here to present to you our CMC product quality review.  CMC stands for chemistry manufacturing controls.  Usually this is what people call our reviews.  I am going to give you a brief summary of the current status of the CMC review and our responsibility, and then I am going to explain to you what we plan to do and what we are hoping we can achieve in the future.


     As Dr. Lesko said, his office is structured based on the therapeutic class of drugs and, in the same way, we are have 19 chemistry teams and each team is responsible for drugs of a specific therapeutic class.


     We also have three divisions.  Our review staff are called chemists because they all have the basic training in chemistry.  However, many of them actually are specializing in very different fields and they have academic training, some in chemical engineering, biochemistry, molecular biology.  This is because we review a diverse variety of product types so we need people with different expertise.  We hire people not just right out of university.  They all have working experience.  Some of them actually were professors and some have research experience, and a small number of them come from industry.  We are not very competitive in salary and when they come, at minimum, they have to have their pay cut by 30 percent.  That is why we are never successful in recruiting people from industry.


     The purposes of the CMC review are multiple.  The first one is to assure the quality of the investigational new drugs.  We want to make sure that clinical trial material is safe for all phases of the clinical studies.  We also want to make sure that clinical material is properly characterized and with adequate consistency and the data generated from expanded studies are reliable.   We don't want the products to be not properly monitored with inadequate quality so that you cannot be sure whether the batches, the material taken by the patients actually is what it is supposed to be.  We also review the NDAs and we also monitor some of the possible changes after a product is marketed.  Therefore, we really monitor the quality of the products throughout their life cycle.


     We also see ourselves as the facilitators of product development and approval.  So, we have written guidances provided them to industry for various phases of drug development and for NDA submissions.  We participate in international activities to standardize the NDA submissions.  We encourage dialogue with sponsors to discuss their drug development programs.


     I want to give you a flavor of the diversity of the product types we evaluate.  In terms of drug substances, we have very simple compounds sometimes.  However, we also have very complex compounds.  In fact, our Center is the one that approved the very first biotech products.  So far we have approved more than four NDAs in the biotech area.  We also have products which are complex mixtures of many, many substances derived from natural sources.


     The diversity of dosage forms is also very important.  We have simple oral tablets for immediate release but we also have very complex dosage forms which are targeted for an action site or which are designed to reduce toxicity, such as liposomal products, inhalation products.  Surprisingly probably to you, we also evaluate some devices.  We evaluate IUDs.  We have a number of NDAs for IUDs.


     So our scientific evaluation of the drug substance includes a lot of different studies, data and information.  I have several slides to explain that.  I am not going to go into detail in the interest of time, however, you can see that even within a product the knowledge required to evaluate it will be very important because the product types are so different.  Even within a product, the data and information will cross multiple disciplines.

     I also want to point out that even though in a submission we have the manufacturing process and control description, however, the majority of that data and information actually is the responsibility of our field offices.  They assess the process validation, demonstration of product batch analysis, and those data are evaluated outside.


     With all the data submitted and the studies submitted to the NDA, and with analysis such as the impurity profile and analysis methodology validation, study and clinical batches results, the NDA establishes specification for a drug substance at release time and also through shelf life.  I would also like to point out that most of the time the NDA specification eventually becomes the published standard at a later time.


     We evaluate the stability, the design of the study and the results so that we can establish the storage conditions and expiry or retest data for a drug substance.


     The second major product of our work is to evaluate the drug product information.  There are also multiple elements which are listed in the slide but I will not go over them piece by piece, however, it is important to point out that recently, because of the implementation of ICH, we are able to get some knowledge from the data for pharmaceutical development.  Therefore, now we can understand the formulation design, the dosage form characteristics and the critical components and attributes.  So, this will help us to have a critical evaluation of the process, of the specification we are going to set.  Just like a drug substance, the manufacturing and the process control information are mostly a field evaluation function.


     In the same way, the NDA specifications for a drug product eventually also become public standards.


     Some of the container closure systems actually are used as a delivery device and in those cases we will evaluate that the container closure system is appropriate for the use.


     We establish the shelf life for the dosage form and, because of the study, we will know whether you need special handling instructions.  Whether it needs to be protected from light, things like that.  We also review all the container labels and certain sections of the package insert.  So, we want to make sure we have clear and accurate information for patients and the healthcare providers.


     I think this gives you a flavor for the diversity of the data and the information we have to evaluate, and the challenges we have been facing over the years.  The first one is, as I explained, that some of the information is actually reviewed by the field office so, therefore, the reviews actually lack of full manufacturing information to make a final decision.  The CMC review is separate from inspection, with sometimes limited or no communication between the reviewers and the investigators.  I will explain later how we rectify those things.  The reviewers have no access to inspection results.  The inspection observation of deficiencies we don't get to see.  Turbo 483, we are not on line and we don't see established inspection reports, the final report.

     In the past few years, two or three years, we also experienced change.  Because of the compressed drug development time, because the companies would like to launch the product as quickly as possible, there is not enough time to really do a thorough manufacturing process development and optimization becomes a post-approval activity.  Therefore, in our submissions when we do the review, often we do not know the manufacturing capability.


     As I mentioned, the data and information cover multiple scientific areas and the emerging new science and technology in formulating design, manufacturing process and novel dosage forms require the reviewers to be on the cutting edge of science and technologies.  Reviewers are expected to have broad knowledge to address all scientific and technical issues for all types of products.


     We also observe an increased work load of generic products.  There are many, many interests of industry to launch generic products as soon as possible so there will be an increasing need for our Office to coordinate with the generic chemistry review.  There are changes in patient involvement in medicine and patient expectations which we have to fulfill.  We have to be able to meet their needs.  The last remark is that this is consolidation of CDER and CBER products.  We need to have the same review standards.


     Facing those challenges, what have we done?  In our Office we have participated in ICH activities, and because of the change in product process development and the shortened time or the compressed development time, we don't have enough data so what we have done is to implement interim specification at approval time.

     We use the limited clinical data and we use statistical approaches and set an interim specification.  The final specification would be set post-approval when we have additional data.  We also implement skip lot testing and sunset testing because, once we know more about the process or we are confident about the data, a certain test may not be necessary to be repeated every time and certain tests may be able to be dropped.

     We also implement limited involvement in the pre-approval inspections so our reviewers will be able to resolve some of the issues on site with the firms and we will be able to evaluate some of the data on site.

     We also have expert review teams for special products.  Because of the uniqueness of biotech products or metered-dose inhalation products and botanical products, we have special people for consultation review or for primary review.

     In terms of sharing information, the lessons learned, we have symposia.  We also have peer review.  We select certain INDs and NDAs and present them to the entire staff at the office level or at the division level or at the team levels so we have multiple levels of peer review.


     Like any other office, we have the good review practice which we implemented more than a year ago.  I am proud to say the Center has evaluated the implementation of GRP across the offices and our Office received the highest score.  In addition, we also established some unique programs so we can continue to train our reviewers in terms of manufacturing processes, in terms of learning the cutting edge of science and technology.

     We have coordinating committees to address common issues across generic drugs and new drugs.  We participated in the 21st century GMP initiative, and we are very hopeful about integration of quality review and inspection to bring us additional knowledge so we can make more intelligent decisions.


     The GMP initiatives mapped out the desired states of product quality, the product review and the inspection.  Dr. Hussain will give you a detailed explanation of the desired state so I am not going to go over that.  However, in order to reach the desired state we believe we need a better quality system for our review.


     The benefits of a quality system are multiple.  I am not going to go over them one by one but I think a couple of them are very important.  First and foremost is that we will have improved scientific underpinning in reviews.  We will have question- and risk-based review and the risk assessment will be based on science and the decisions will be based on critical analysis and thinking.


     In addition, I think a quality system will make us dynamic; make our causes dynamic; make our review dynamic.  Therefore, it will be easy to adopt changes and have agility for the future.


     So, what are our initiatives?  To have a quality system; to have multiple internal conceptual discussion.  We have formed a steering committee.  The committee members consist of the management of Office of Pharmaceutical Science and our Office.  We also formed a working group consisting of a strong team, the best and the most dynamic people in our Office.  We have high hopes that this team--they are actually in the audience--will be able to bring us to the future.


     We also actually already hired a quality systems specialist as a consultant and he has started to interview the members.  We would like to do certain things.  A few of them are listed here.  We would like to identify our customers.  We know that patients have care providers, our customers, and we would like to identify all of them.  Then, we would like to survey customers to determine their needs and to establish metrics for accomplishment and to establish milestones, short-term and long-term goals.  With that, we would like to determine an implementation process and, hopefully, at the end we will have a continuous improvement program.


     This concludes my presentation and I thank you for your attention.  I am looking forward to having feedback from all of you.

Questions and Discussion with Board/Presenters

     DR. DOYLE:  Thank you very much, Dr. Chiu.  This now brings us to the time where we can ask of the presenters any questions or present any comments that we may have.  We have had a couple of questions presented to us that we should try to address as well.  Specifically, what steps are most important for integrating evolving science into the review process?  And, are there opportunities to learn from business or universities in improving the review process?  With that, do we have any thoughts or comments?  Yes, Jim?

     DR. RIVIERE:  I have one question for Dr. Lesko.  I am just curious, on the template that you have provided for the structure of the review, have you made that available to the sponsors applying and what those criteria are for the information in that document?

     DR. LESKO:  That is actually a very good question.  I think we have made it available informally through presentations at professional meetings.  We have certainly made it available through our meetings with PhRMA.  Whether it is publicly available on an Internet site, I don't believe it is.

     DR. RIVIERE:  Very similar to an NIH current review, you have very specific areas and you know a study section is looking for that information.  So, you know, the whole point of this is to improve the approval process and improve the quality of the data.  If that is now formalized to the point where your reviewers are looking for this information in that format, I would think it would be very useful for PhRMA to look at that so they know exactly that the information they provide is consistent with your goal of early communication.

     DR. LESKO:  Right.  I think it is a great idea.  It has been somewhat of an experimental phase for us using this approach, and it looks like we are going to move to make it more firmly established in our process.  That might be the time to do exactly what you are saying, to make it available on an Internet site as opposed to an internal site and continue to advance it within our public contact with the industry.

     DR. DOYLE:  Dr. Shine?

     DR. SHINE:  First let me congratulate you on the efforts to bring continuous quality improvement to the process.  I think this is a very important development and it is one that should be strongly supported.

     I would just make a couple of observations.  First of all, Janet, when we started to worry about quality in the healthcare system we were told it couldn't be measured.  Scientists will tell you that there are all kinds of ways in which people resist the notion of addressing these issues.  I think the key in many regards is data, and I want to come back to that in a moment.

     I guess my concern is that for a long time the efforts in improvement in quality in the healthcare system were very much focused on process.  JCAHO kept looking at the processes of hospitals and things of this sort.  One of my concerns in these presentations is that there has been so much emphasis on process and not as much on outcomes, as I always like to see.  Outcomes can be measured in a whole variety of ways and I am not talking about simply the time frame in each area, but outcomes can be measured when you want to talk, for example, about the confidence of your customers or how can one set up a way to measure that.

     We heard about different ways in which new drug applications come in.  Are the outcomes different depending on how, in fact, the new drug application comes in, whether it comes in, in an orderly way or whether it comes in, in bits and pieces, whatever?  Depending on what the data shows, what approaches can one take in order to improve the quality of the submissions and what are the outcomes you want to measure in that regard?

     But I want to emphasize that there are many, many ways to define the outcomes in each segment of the activity and I would urge you to look much more closely at that.  Again, we heard metrics used but in some cases the metrics were about the quality of the meeting.  Again, that is soft.  Can you come up with any better outcome measure in terms of helping people to tell whether things are getting better?

     The second observation I would make is that we have certainly learned in the healthcare environment that it is systems that count.  As you know, ever since we brought out the patient safety report at the IOM, our principal message has been "it's the system, stupid" in terms of drug safety and not blaming an individual.

     Again, we saw some approaches to subsystems.  Dr. Lesko presented a subsystem.  But I would submit to you that each of those items, one through whatever it is, five or six, could be described internally as a system, that is, the very system which allows one to move within each of those areas.  And, I would like to see some real attention given to the microsystems, if you will, that underlie each of those because those microsystems are, in fact, where you have the greatest leverage in terms of improving quality, and if you have a relationship between some outcomes and how that microsystems works, then you expose, if you will, the way in which the interstices work to a much better extent than I think the broader conceptual activities.

     Finally, I think one of the things we have learned, at least from some industries, is the notion that one can encourage the worker, in this case the staff, to innovate and to do things in different ways provided you have an outcome measure.  In other words, in the context of some kinds of outcome measures you can give people the opportunity--and it has to be open and transparent and explicit, but you can give them a chance to do things somewhat differently and then demonstrate that they can produce a better, or if it doesn't work a worse, outcome but at least you have a context in which you create the opportunity for innovation.

     One of the things we learned, that you may or may not know, is that on the cardiac advisory committee that has been reporting on the outcomes of cardiac surgery and angioplasty, pediatric care and so forth, in New York State, what was remarkable was the pride that people took, once they got the data, in improving the system.  It wasn't about the doctors choosing better hospitals or the patients choosing better hospitals.  It was when one group saw that they weren't doing as well as another group that they tried to figure out ways to improve.  So, I am suggesting that thinking about a way to do that could be constructive.

     DR. DOYLE:  Excellent points.  Thank you, Dr. Shine.  Dr. Pickett?

     DR. PICKETT:  I also want to congratulate the group for thinking about quality systems.  The industry also has given, as you know, a lot of thought to quality systems. One of the things that I did not hear today, because in my own experience it has been difficult implementing quality systems without increasing head count--I heard absolutely nothing about what the plans were to increase head count, I assume, to meet the demands of implementing quality systems across your organization.

     DR. MCCLELLAN:  I just want to take a moment to respond to Dr. Pickett on head count.  We are increasing head count.  The agency is larger than it ever has been and this is a direct result of legislation passed last year.  On our food side, we have added over 700 new people, primarily in the field to help improve our inspection activities, and the like, but that creates some opportunities for taking a fresh look at how we are undertaking our food security and food safety programs, which we are doing.

     Here, on the medical products side, we had legislation last year that implemented some new programs and provided some new resources as part of the Prescription Drug User Fee Act and the Medical Devices User Fee Act.  The PDUFA new authorizations with increased fees is going to provide some new personnel, including personnel in key areas like IT and in post-market monitoring, that will help us implement these activities.  The Medical Device User Fee Act, which is off to a little bit of a rocky start because of some issues about the right level of appropriated funds to go along with it, is also providing some new resources and we are fully committed to implementing that program effectively too.

     We are actively pushing right now for new legislation on animal drug user fees and there is considerable bipartisan support for that.  I think it is mainly a matter of just getting the focus in Congress to get it on through, given all the other very urgent priorities that Congress has to deal with.  These programs together provide some new resources as well as some new authorities to undertake these activities.

     But I don't want to kid you, the financial situation facing the government is tight.  It looks like it is going to be tight for a while to come.  So, I think we are going to have to try to make the best progress we can without huge new infusions of new resources and that is difficult given the increasing complexity and the increasing number of products that we are facing as a result of all the increases in research funding that I was talking about earlier, and all the information that is still at a stage of being transformed in understanding, as you heard in the presentations this morning.  So, that is an area of challenge.  It is one that we are trying to respond to as best we can given the constraints that we face.

     DR. DOYLE:  Yes, Dr. Shine?

     DR. SHINE:  Yes, I just would emphasize the world is filled with quality experts.  I know that Cecil is not implying this, but you have to have some support people in terms of facilitation but this kind of quality improvement in this kind of an organization has to be generated internally by the people who are doing the work, who know how it has to be done, who can interact with each other and do it well.  So, I think the notion of additional help is important in terms of facilitating that but they are not going to do the planning; they are not going to do the quality improvement themselves in a professional organization like this.  I think people have to buy into it and they have to be committed to it in terms of their own part of the organization.

     DR. PICKETT:  Could I respond to this for a second?  I don't disagree with Ken.  The quality obviously starts at the bench and individuals at the bench really are critical for quality.  But there are certain other parts of the organizational structure--quality control; quality assurance--that do need expertise and will require additional head count if you want to fulfill that aspect of the overall quality system.  I think that has certainly been my own personal experience in industry and I think those types of individuals are necessary.

     DR. SHINE:  I would agree with that, not disagree.

     DR. DOYLE:  Dr. Thomas?

     DR. THOMAS:  Mike, I would like to comment on one of your opening remarks with regard to how to integrate the new sciences, and perhaps a question directed to any of the previous speakers.  How much effort is actually devoted to what I will call test validation, replication, particularly in the area of biopharmaceutics or agro. biotech. products?  Without this validation and replication of these assays that you are going to be looking at in the future, and you probably are already looking at some, I think you need to have some internal expertise within given units to assess these things.  I don't know who to direct the question to specifically, but I would like to know what sort of effort or energies are devoted to this type of activity.

     DR. LESKO:  I think that is a good point.  In particular in our clinical pharmacology area when we think about things like pharmacogenomics or things like modeling and simulation, it would be unrealistic to think that everyone in the Office is going to have those skills.

     What we have tried to do over time is identify those skills that we think are relevant in proportion to sort of the integration of that science of drug development and tried to stay ahead of that a little bit and, through the recruitment process or through some internal training, try to get that expertise built up and sprinkled throughout the Office and the respective divisions in a way that gives us the coverage appropriate for the therapeutic areas.

     The reality is that all therapeutic areas don't need all of the new science and it always happens that certain therapeutic areas sort of advance faster than others in terms of integrating some new science such as pharmacogenomics or such as modeling and simulation.  So, we try to do it that way and begin to utilize on site--by on site, meaning in a division--experts to begin to mentor other people in that division on the skills that we are talking about and really come to the reality that what we have is sort of a puzzle where each piece doesn't necessarily equal each other but together they provide us the coverage that we need in terms of assessing the science coming in.

     DR. THOMAS:  Thank you.

     DR. DOYLE:  Thank you, Dr. Lesko.  Kathy?

     DR. CARBONE:  Hi.  I am Kathy Carbone, sitting in for Jesse Goodman who is unavailable for this morning.  At CBER we have a fairly extensive program where we actually do some in-house testing.  This is assisted by Sherry Lard and Deborah Jensen, who are focused on quality issues and specifically laboratory quality issues.  We are working with all the laboratories which do testing to make sure that the systems are up to quality assurance and quality control laboratory testing standards.

     We also have an extensive interaction with manufacturers where experts on board actually assist in development of tests that have good utility, from our point of view, and consistent products.  That is driven, of course, by the complicated nature of the products and the way in which biologic products are produced that require specific interaction.  So, we have a fairly extensive program in-house.

     DR. DOYLE:  Thank you.  Dr. Crawford?

     DR. CRAWFORD:  I believe, Dr. Thomas, you also asked about ag. biotech.  That is done, as you know, in our Center for Food Safety and Applied Nutrition, and was put in place in terms of reviews of these products as they were being developed in 1992 when we formed the Food Advisory Committee which I was on.  Since that time, they have refined those procedures I think very well under Jim Maryanski from the Center for Food Safety and Applied Nutrition.  I would characterize it as a very reproducible system that is working well.  For those that they have cleared, they haven't had any recalls or any product cancellations as a result.

     But they are continuing to look at it because the next generation of those kinds of products will be those that are specifically doing something to change the food.  In the past it has just been processes that cause product to grow faster, and so forth.

     DR. THOMAS:  Not substantial equivalence?

     DR. CRAWFORD:  Yes, substantially equivalent to the products already on the market.  As the science gets more refined there are going to be products that will have more vitamin C or whatever and, as you know, those products will have to be labeled.  So, the enterprise will have to be ratcheted up a number of steps and quality control is going to be a critical part of that.

     DR. DOYLE:  Thank you, Dr. Crawford.  We have time for one more.

     DR. CHIU:  I would also like to answer the question about new methods, new assays.  Whenever there is a new technology or new methodology developed by the pharmaceutical company for new products, our laboratory will repeat the tests and verify the rigidity of the tests.

     DR. DOYLE:  Thank you, Dr. Chiu.  Dr. Nerem asked for the next question so we are going to respect that.

     DR. NEREM:  Thanks.  I didn't want David Feigal to get off without any questions, but actually my questions could be submitted to any of the centers.

     David, through the CDRH science review it was very clear that one of the challenges is to bring the right science to bear, and there were a couple of things that were talked about and I would appreciate your providing some kind of an update.  One was that as you look ahead over the next five years and you have a number of people, for whatever reason, leaving that you actually developed some kind of a plan as to what you want your science to look like five years from now as opposed to simply falling into the trap of when someone leaves replacing that person by someone that looks sort of like that person.  So, I am wondering where you are in terms of having such a plan.

     The second thing was in terms of using outside people.  You indicated the 20 FTEs.  Part of the suggestion from the group was that you have in critical areas where you don't have the science expertise, at least not to the extent you would like, have people that are sort of consultants on call because when you need to talk to someone you don't want to wait weeks for paperwork to go through the bureaucracy.  One of your staff needs to be able to pick up the phone and simply talk to the person then.  So, where are we on these kinds of issues?

     DR. FEIGAL:  A couple of great questions.  On the issue of how do we plan for the future, let me describe what we did with planning for the new hires with the user fee positions.  We didn't simply look at the organizational structure and see which groups were busy, where there was need and proportionally distribute the resources according to the organizational chart.  What we did instead was that we recognized that we had divided the product areas into six different working groups.  For example, one of our busy groups is the cardiovascular group.

     What we did, we asked each of the division directors responsible for a product area to actually pull together a team from across the whole Center, not just from new product review but also from our laboratory programs, from our post-marketing programs or compliance programs, and for them to look at the kinds of things that need to get accomplished by the user fee goals, as well as the critical needs of the Center from our initiatives, from the initiatives from Dr. McClellan and from the Department and some of the President's objectives.  What we asked each of those groups to do, Bob, was to actually give us a prioritized list of the hiring they needed without respect to organizational structure.  So, the cardiovascular group could, in fact, say, well, they need to have some expertise in biomaterials; they need to have some expertise in electrophysiology; they need to have an engineer who does computer software without respect to where organizationally those people sit, and prioritize them.  We essentially asked them to take responsibility for our overall responsibilities for the whole product area.

     Those requests are ranked and voted on, just like a study section, by the office directors.  That is the way we do our prioritized hiring.  So, it isn't on the basis that, well, if you have lost someone you can back-fill that position.  In fact, you have to actually request it through this process, that there is a critical need to back-fill that position and time to stop to think do you want to change and reshape the organization, and also not to get groups into the bind of, well, I lost a person but the only kind of people in my unit are statisticians, or the only kind of people in my unit are engineers and clinicians.  We invited people to actually think about the whole product area.  I think that is actually going to allow the Center to identify needs for the future.

     DR. NEREM:  This will be a continuing process?

     DR. FEIGAL:  A continuing process.  As the resources become available through the collection of fees, which is just starting now, we will release a certain number of slots for hiring in batches because you never know exactly who you are going to find and in what order, and have authorization to hire sort of from a pool of identified needs.  But it is a way of actually getting involvement of the whole Center in thinking about where does the Center need to go rather than simply asking each organizational unit to do its best.  Historically, that hasn't worked so badly but this actually allows you to not tie the needs to the organizational structure.

     Your second question, our biggest pool of outside experts that are ready-made are actually advisory committee and advisory panel members like yourselves who are already qualified as special government employees who are available, if used singly, for consultation, which we nicknames homework in the Center.  We will often actually send out parts of a review or parts of a protocol, sometimes the whole thing, to an advisory committee member and ask them to provide expertise.  It is not unusual to invite a single advisory committee member to participate in a telephone conference call with a company over an issue.  So, that gives us a pool of about 300 people who have been selected for their expertise on the advisory panel system to start with.

     Then, above and beyond that, we have actually looked at one of the things that we have in the works which I think will be out in the next month or two, and that is a process for actually contracting with universities, with individuals who will be available to us on a part-time basis as essentially part-time staff, except located in universities.  We think this might work particularly well with more junior faculty that don't have as complex relationships with industry as more senior faculty often do.  We have started in our very busy cardiovascular area to see if we can develop a contract proposal for universities to bid on where, in fact, a young cardiology faculty member could review applications, could review protocols as a special government employee.  We would probably hire them through an IPA, or there are many, many different mechanisms.

     But those are some of the things we are pursuing and moving ahead on.  The thing we had to do to make progress was to hire someone whose job it was full-time to make this happen.  You know, as long as we talked about an interest in this nobody had time to get through all the contracting details and punch these things through.  Dr. Susan Homire has been very effective in putting these programs in place over the last year.

     DR. DOYLE:  Thank you, Dr. Feigal.  Dr. Buchanan?

     DR. BUCHANAN:  Thank you, Mike.  I would like to also follow-up a little bit on Bob's questions and reflect on our experiences in the last few years in terms of the evolution of our scientific staff and our needs.  Strategic planning for your scientific hires is a critical component but the lesson that it has also taught us in the last few years, as we have had tremendous changes take place in our program orientation, is that we also have to try to hire some of the brightest people we can and have a learning organization.  So, it needs an equivalent commitment, for example, to an FDA university where we can quickly retrain people to fill in the needs that we have immediately to get started on critical attributes.  We found that you can't do one or the other.  You can't rely on future hires.  You also have to be able to respond quickly with your own internal learning organization.

     DR. DOYLE:  Dr. Rosenberg?

     DR. ROSENBERG:  I just have a comment and maybe a question.  Dr. Lesko mentioned the initiative by the agency to get involved earlier in the process of interaction with some of the people that are starting to do that science and to participate much earlier with industry in the process of going through clinical development.

     I think I applaud that initiative but it strikes also of moving kind of to a dynamic role of review rather than kind of just a static role of review.  I was wondering if the agency has considered some of the problems that they may run into.  It sounds good on paper and I think it would be applauded.  The problem, of course, is that these processes tend to be fairly long processes.  They can go on for years and, therefore, the agency is going to have to maintain very careful consistency in how it works with industry through that process.

     I think as Janet mentioned, the more you think of this as a guild or an artist in science, of course, then scientific opinion and as people turn over in these groups, as things go on in time, particularly with people turnover, you end up with different opinions along that process.  Therefore, it will be very important that when you start a conversation or begin that process with industry you have a way to maintain the consistency you are going to need so that what they are hearing at the end of the process is consistent with what they heard at the beginning of the process.  It is something that I think people in industry have been concerned about in the past in terms of maintaining consistency in conversations with the agency.

     DR. LESKO:  I agree with what you have said.  I think we have the context with industry now and interface in meetings where decisions and discussions are made about the Phase III program where that is equally important.  But I think you are right, there are things that have to be thought through for these earlier interactions.  I believe we should be dynamic in the sense of interacting with companies earlier in drug development as opposed to living with what we get at the end of the day with the NDA, where we have unanswered questions or perhaps even some wasted dollars in terms of generating information that perhaps the sponsor thought we needed and we didn't need.

     In particular in the area of dosing strategies for drug development programs, I believe dosing strategies are set very early in the drug development program, even preclinical, and that is a critical area in terms of regulatory assessment later on in terms of efficacy, safety and dosing adjustments.

     So, I think we should do both.  I mean, I think we should be both dynamic and facilitate efficient and informative drug development, but also be in our traditional role of review.  This has implications and I would expect not all companies would welcome this sort of early interaction and, like any meeting, I think it would be up to the sponsor to determine whether this has value to them or not.

     It also requires resources on our part.  This is not a small issue.  But I think it is worth exploring.  I think it is worth exploring in selected cases or with very clear roles in mind to pilot something like this to see what the value of that would be, both for us internally as well as for companies that engage in this activity.

     DR. DOYLE:  Thank you, Dr. Lesko.  Dr. Sundlof?

     DR. SUNDLOF:  Thank you, Mike.  I think Dr. Rosenberg raises a very critical question.  The Center for Veterinary Medicine has had what we call a phased review process for about ten years now in which sponsors meet with us in the very beginning, even before they have decided whether or not they are going to go forward with the development.  All of the problems that you have just mentioned we have run into but we have been able to work our way through all of those problems.  Some of them we are still working on, just to be honest.

     Some of the things that we have done have been that we have chopped it up into various technical sections, the review process.  When we get through with each one of those sections we send the company a letter that says this is a technical section complete and you can move on.  You don't have to worry about that.  One of the disciplines that we had to maintain is that with our reviews, especially as we have turnover and new things come up is that once we make a commitment to a company that we have reviewed their protocol and have come to agreement we will stand by those.  We are not going to change unless it is something that would put the public health at risk.  So, we do have some caveats to that but primarily, once we have made that commitment to the company, then we honor that commitment and that is very important to us.

     It does have some additional problems just because the submission is broken up into so many different pieces and it is going in so many different directions throughout the Center, it is hard to maintain centralized control over where all the pieces are.  So, that has been our biggest challenge in this phased review process.

     DR. DOYLE:  All right, thank you, Dr. Sundlof.  I think we need to move on if we are going to have lunch.  Next we are going to hear from Dr. Ajaz Hussain, who is the Deputy Director of the Office of Pharmaceutical Science in CDER.  He is going to provide to us an update on pharmaceutical manufacturing initiative.

Update on Pharmaceutical Manufacturing Initiative

     DR. HUSSAIN:  Good morning.


     I am very pleased to be here to share with you a progress report on our manufacturing initiative.  There are actually two initiatives but I think my goal is to show you just one initiative as we move forward from here.


     What I would like to do today is to share with you a progress report on the process analytical technology initiative that we first discussed with you in November of 2001 and subsequently in April of 2002.  Then came the cGMP initiative for the 21st century.  And, how these two sort of come together now, and to move forward I would like to sort of spend my time discussing a desired state.

     We have a number of activities, a number of programs, a number of initiatives ongoing and unless we define very clearly what the desired state is, I think alignment and the challenges associated would be greater.  So, I would like to get some input from you on how we have articulated our desired state for the future and, hopefully, you will provide us some ways to improve that desired state.


     The PAT initiative, the process analytical technology initiative, was, as I said earlier, first proposed to you in November of 2001.  We presented this as an emerging science issue in pharmaceutical manufacturing.  We presented several different perspectives.  We had an academic and an economic perspective by G.K. Raju from MIT and Doug Dean from Price Waterhouse Cooper, who essentially outlined the opportunities for improvement, looking at the current efficiency levels in manufacturing; some of the challenges; some of the reasons, root causes for why things are as they are.

     We heard from Norm Winskill and Steve Hammond from Pfizer about their perspective on new technology and innovation in manufacturing.  Essentially, they outlined for us that industry has adopted a "don't use" or a "don't tell" approach to new technology and innovation.  Under the "don't use" scenario, essentially because of the regulatory uncertainty, industry says let's not use any new technology.  Or, if they desperately need new technology they will do it, but then they will provide the routine type of information to the FDA and be in the "don't tell" scenario.  We were extremely concerned by this because this is not in the interest of public health and this is not in the interest of the country.  So, we really needed to address this.

     In April we had a presentation by Ray Scherzer, who is senior vice president of manufacturing at GlaxoSmithKline.  Essentially, he came and sort of challenged the industry itself, and his presentation was titled "challenge to PhRMA industry--quality by design."  The perspective he shared was that manufacturing, although being treated as a stepchild in some cases, is a significant part of the industry; a critical part of the industry, as well as that the technology that exists for manufacturing and improving the efficiency of manufacturing exists already outside and not within the pharmaceutical manufacturing arena.

     Therefore, I think we actually posed a question to you, is it appropriate for FDA to take the lead on this initiative and actually facilitate introduction of innovation and new technology in manufacturing?  Essentially you sort of endorsed that proposal and we moved forward from there.


     Following our proposal to you, we set up an advisory committee under the Advisory Committee for Pharmaceutical Science, a PAT subcommittee, and we held three meetings.  These meetings and deliberations were focused on definitions of what process analytical technology really is in the pharmaceutical context, benefits and scope.  Essentially, the discussion led to the scientific underpinning of process understanding.  PAT is essentially the science of manufacturing and understanding the process itself, not just having a sensor on-line, and so forth.

     We identified perceived and real regulatory hurdles.  I think the point I want to make here is that we have perceived hurdles that have a way of becoming real hurdles.  I think that we saw that in the internal hurdles a company has.  I was actually a bit pleased but I sort of held myself back at the end of the third meeting, when the industry representatives on the PAT subcommittee said that FDA is not the hurdle; industry itself is the hurdle for us.  I would like to believe that but I think we are all part of the same system; we are all part of the problem; so we all have to be a part of the solution so I think we all have to work together to remove these hurdles.

     We also identified the need for cross-discipline communications.  Pharmaceutical manufacturing in particular essentially works from the art of pharmacy compounding and to some degree I think our thinking is somewhat similar in that context.  To give an example, I think we make tablets the same we made tablets a hundred years ago.  The scientific underpinning of that in terms of predictability and generalization has not really occurred.  So, we really need to bring pharmacy, chemistry and engineering together as a discipline of pharmaceutical engineering.

     We also identified approaches for removing these hurdles.  I will share some of those with you.  We also had excellent case studies presented by company representatives such as Bristol-Myers Squibb, Pfizer, Glaxo and others.  We also discussed a general approach to process validation.  But I think most importantly, we actually developed a training and certification curriculum for our staff.  We are proceeding with that.


     But before I sort of outline some of this to you, let me just share with you the people involved in this initiative.  This is the Office of Regulatory Affairs, Center for Drugs and Center for Veterinary Medicine initiative right now.  We have a steering committee, a policy development team, training coordination, as well as a team of reviewers, inspectors and compliance officers that are being trained and being certified in some of these new technologies.  The certification program is based on the curriculum we developed through our discussion and deliberations of the PAT subcommittee.

     Three universities, University of Washington, Seattle, University of Tennessee, Chemical Engineering and University of Purdue, Pharmacy have been brought together to provide this training.  This training has both didactic and hands-on experience at these sites.


     In summary, what the consensus of these deliberations was is that process analytical technology provides an opportunity to move from the current "testing to document quality" paradigm to a "continuous quality assurance" paradigm that can improve our ability to ensure quality was built in or was by design.  We felt that this was actually the ultimate realization of the true spirit of cGMP.

     This provides greater insight and understanding of the process itself--at, on, or in-line measurement of performance attributes.  I want to sort of distinguish that from measurement of, say, pH, temperature and so forth.  We are not talking about those type of measurements.  We are talking about measurements that can predict product performance.  Real-time or rapid feedback controls focus on prevention; potential for significant reduction in production as well as development cycle time; and minimized risks of poor process quality, thereby reducing risk itself and also reducing regulatory concerns.


     What we were able to do was to create a conceptual framework for PAT, not piecemeal.  This is quite a complex slide and I am not going to walk you through this slide.  The slide is just to illustrate that we have addressed every part of the manufacturing process, including development optimization, continuous improvement and how do you bring a multivariate systems approach to risk classification and mitigation strategy to be part of the PAT system.  This is the framework that we are using to develop our draft guidance.  I will not, as I said, walk you through all of the elements of this.

     But I would like to sort of emphasize that, as we move forward, product and process quality has to abe based on knowledge.  I have distinguished knowledge from data.  I would sort of submit to you that I think many of our decisions are data-driven decisions where we cannot generalize, we can not predict and, therefore, the level of sophistication is not what it can be.

     So, if the question to the FDA review staff as well as inspection staff is to assess whether quality was by design, what is the information that they use to make that assessment?  I think that is the key.  The information that is submitted to us in the review or when it is held at the site tends to be more data driven.  What I mean by that is that you will have bits and pieces of data arranged to form information and we say, all right, this drug is stable, therefore, it is all right.  But when there is a change in any manufacturing process you either have to repeat and retest every package to say the change did not impact and in a knowledge-based system you will have understood the critical formulation variables, and so forth, and be able to say this is not a critical change and, therefore, you can make a decision in a different way.

     Today I would submit to you we are at the bottom of this knowledge pyramid where our decisions are driven by data that are generally derived by experiments and as we move up this knowledge pyramid, going towards mechanistic understanding and first principles of these systems, I think that is the direction we want to move in.  The PAT process actually brings us in the middle of this pyramid where actually we are moving towards establishment of causal links that are a predictor of performance.


     So, the regulatory framework under which this initiative will sort of conclude in terms of a draft guidance is that the PAT tools are not a requirement.  The current manufacturing processes do provide quality product but they are not as efficient, but the technology is really new so it cannot be a requirement.  Industry and companies will need to decide whether it makes sense for them, whether they have the know-how to move forward.  So, that is an important question that we have posed to you and you sort of have endorsed that.

     Research exemption I think is an important part of this because one of the biggest hurdles we have is if you put new technology on an existing manufacturing line you will see problems, or you will see trends which might be considered as problems, whereas, if you had not done so, there would not have been an issue at all.  So, continuous improvement without the fear of being considered non-compliant is a major hurdle and a research exemption framework.  Again, Dr. Woodcock presented a case study on how we intend to use sound statistical principles for addressing some of these issues that will be part of this draft guidance.

     Regulatory support and flexibility during development and implementation--this is the PAT review and inspection team.  This is to eliminate the fear of delayed approval and actually dispute avoidance as well as resolution for some of these issues.

     Science- and risk-based approach--low risk categorization based on a higher level of process understanding.  So, if you link regulatory scrutiny to the level of process understanding you are able to provide better incentives.


     The strategy for moving forward has been to conduct workshops, and we have done several workshops now, both in the U.S. and Europe.  We have participated in some workshops in Japan also.  I forgot to mention that.  The scientific discussion and debate that occurred following our PAT subcommittee was very important and it brought in debate across disciplines--pharmacy, chemistry, chemical engineering.  And, some of this debate was emotional because I think you are dealing with different disciplines and some of the common issues and how different disciplines address that.

     We had debate across organizational units between development and manufacturing.  For example, manufacturing folks say you develop a product and throw it over the wall and we have to deal with it, that type of debate and, then, the regulatory departments and what sort of regulatory policy and process there should be.

     So, the general guidance that we will issue in a few months as a draft will sort of encompass all these aspects.  It will not be a technical guidance; it will be more of an information guidance on process and so forth.  We plan to bring other guidances out on the scientific and technical aspects.  But when we release this guidance we will bring together the engineers, the chemists, the pharmacists together as part of the workshop for training.


     I think the most important point that I am very proud to share with you is that although FDA initiated this initiative, we cannot champion it.  We need champions to take over from us, and the champions have been created.  Champions to drive this initiative towards a shared desired state are coming from industry.  Some examples of companies which have come forward are listed there but, more importantly, academia.  We started with the MIT presentation by Purdue, Washington, Tennessee, Michigan--you can look at the list of schools in the U.S. where we have established contact and they have already introduced PAT in their curriculum as well as research programs.  But also London, Bradford, Basel have already started working towards their PAT-based curriculum and research program.  We plan to work with several universities in Japan this summer.  But more importantly, I think you have pharmaceutical engineering programs as part of clinical engineering departments now, and PAT has been introduced in these programs at Purdue, Michigan, Rutgers and so forth.

     One challenge that we are still working on is that the instrument manufacturers, the sensors and so forth, do not have a common voice.  We want to bring them together as an association to deal with the common issues.


     Moving forward, we have hired several experts.  We have an intramural research program focused on our needs.  We are learning from other industries, especially linking to ASTM and their organizations where other industries have already done this.  We are in the process of finalizing a collaborate development agreement with Pfizer on issues with respect to chemical imaging and other on-line technologies.  We are in the process of finalizing an inter-agency agreement with the National Science Foundation to be part of the Center for Pharmaceutical Processing Research.


     I think the main point I want to make here is that I think this started as a small initiative but it could not have proceeded to its ultimate goal without the cGMP initiative.  So, this is going to be part and parcel of the entire initiative that was introduced, and this will be an example of how a science- and risk-based system approach could be brought to bear.


     The cGMP initiative was introduced on 21st of August of 2002.  I would like to call that initiative a drug quality system for the 21st century initiative because it is broader than just GMPs.  It also includes review.  It also includes all aspects.

     The objective here was to sort of take time to step back and evaluate the currency of our current programs on product quality, such that the most up to date concept of risk management and quality system approaches are incorporated while continuing to assure product quality.  The latest scientific advances in pharmaceutical manufacturing and technology are encouraged.  The management of the program encourages innovation in the pharmaceutical manufacturing sector.


     The submission review program and inspection program operate in a coordinated and synergistic manner.  You heard from Dr. Yuan-Yuan Chiu about some gaps that we need to fill.  Regulations and manufacturing standards are applied consistently and FDA resources are used most effectively and efficiently to address the most significant health risks.


     The scope and timeline of this initiative--this is veterinary drugs and human drugs, including human biological drug products.  Organizations involved are ORA, Office of the Commissioner, CBER, CDER, CVM with involvement of CFSAN and CDRH depending on the issues to be discussed, for example, electronic records.  We have identified 14 task groups.  We have 13 active.  Dr. Woodcock is chairing that.  We have a two years timeline for this.  The immediate goals were to complete certain tasks, which we announced in February of 2003.  We have intermediate and long-term projects.


     In your handout packet you have a summary progress report so I will not go through, line by line, all the activities but I want to share with you the major highlights.  We have issued a draft guidance on 21 CFR Part 11.  We have issued a draft guidance on comparability protocols for CMC information and manufacturing changes in a more proactive way than what we do today.


     We are working on different aspects:  Center involvement in the cGMP warning letters, a process for the Center to review that.  A technical dispute resolution process for cGMP disputes is being considered; emphasizing risk base appropriate to the work planning process; including product specialists on the inspection teams.  I think there are some best practices that CBER has that I think we can learn from and bring best practices to work so we want to sort of evolve a program which integrates the review and inspection in a synergistic manner; improving the operations of team biologics; enhancing expertise in pharmaceutical technologies; pharmaceutical inspect rate; quality management system; international collaboration; and holding a scientific workshop with stakeholders.  As I said, you have a progress report in your handout.  I will not go through each one of those but if you have any questions I will be glad to try to answer those questions.


     I would like to sort of have your feedback on the desired state.  It is I think very important that we all define the desired state for pharmacology manufacturing for the 21st century, and this desired state has to become a shared vision for public health not only from an FDA perspective, but from an industry perspective as well as from academia.  I think it will not only help us move in a synergistic manner towards this desired state, but also identify scientific and engineering gaps that need to be filled and how to fill those.


     In our announcement on February 20th we articulated the desired state, and the principles that this articulation was based was as follows:  We recognize that pharmaceutical manufacturing is evolving from an art form to one that is now science and engineering based.

     Effectively using this knowledge-and I underscore knowledge--in regulatory decisions in establishing specifications and for evaluating manufacturing processes can substantially improve the efficiency of manufacturing as well as our regulatory processes.

     The initiative that we have started is designed to do just that through an integrated systems approach to product quality regulation founded on sound science and engineering principles for assessing and mitigating risks of poor product and process quality in the context of the intended use of pharmaceutical products.


     So, that is our framework to sort of defining the desired state which is as follows:  Product quality and performance is achieved and assured by design of effective and efficient manufacturing processes.  Product specifications are based on mechanistic understanding of how formulation and process factors impact product performance.

     I want to sort of underscore that.  If the information available in submissions is limited, for example, if you have done seven or eight pilot batches and that is what you have used for clinical development and these are your specifications for those batches, if you don't have a mechanistic understanding you say, well, this was your lowest dissolving tablet so this is your specification.  So, you are setting specifications based on the capability of your pilot batches and then you set up a system later on because you cannot achieve that capability during routine manufacturing.  So, if you move away from that modality of specification setting to more mechanistic understanding, it really will improve the process.


     Continuous real time assurance of quality, that would be sort of the manufacturing perspective.  From an FDA perspective, how can we facilitate that?  I think we can do that as follows, if our regulatory policies are tailored to recognize the level of scientific knowledge supporting product applications, process validation and process capability.  Today I think we have done an excellent job of harmonizing and coming up with minimum standards, but in doing so we do not have good means for saying this company has done better science than this.  So, we treat the same company the same without recognizing the level of scientific understanding underpinning their application in this area.  So, we could sort of open that up and actually understand and reward or provide incentives for good science.

     So, risk-based regulatory scrutiny would lead to a level of scientific understanding of how formulation and manufacturing process factors affect quality and performance, and the capability of process control strategies to prevent or mitigate risk of producing a poor quality product.  So, you can see the regulatory incentives removing the hurdles of good science sort of comes through some of these statements.


     I will stop here and would really welcome any suggestions that you may have on how we may improve our articulation of the desired state; what considerations we should have when we sort of discuss this with our stakeholders; and how we should align our processes better.  Thank you.

     DR. DOYLE:  Thank you, Dr. Hussain.  It is heartening to see the progress that you are making in this area, and thank you for that update.  Next we are going to hear from Kelly Cronin, who is Senior Advisor for the Office of Policy and Planning at FDA.  She will provide an update on the patient safety initiative.

Update on Patient Safety Initiative

     MS. CRONIN:  Thank you and good morning.


     I am going to spend the next 15 minutes trying to give you an overview of all that we have been working on in recent years regarding patient safety, but with a particular focus on our recent efforts with strategic planning.


     I am going to start by just giving you an overview of this strategic planning effort, and while I go through the various areas of our focus I will give you an update on ongoing initiatives, many of which you might have read about recently in our announcement last month.

     I also want to try to articulate our vision for the future and get your feedback on that, as well as articulating some of the challenges that we are going to have in trying to make that a reality.


     I think as mentioned briefly this morning, there are five key areas to the strategic plan:strong FDA; risk management that has a focus on really pre-market activities and programs; better informed consumers and there are many different initiatives trying to get more information out to consumers; and patient safety, and most recently we have expanded this to consumer safety with the intent that we would like to be broader than medical products and also including foods, dietary supplements and animal health products.  Then, the last major area of this strategic plan is counter-terrorism.  But this will give you a sort of overall idea of how this fits into the larger effort.


     Our primary goal with patient safety is obviously to improve both patient and consumer safety by reducing risks of regulated products, and by regulated products as I just mentioned, we mean medical products, devices, drugs, biologics, vaccines, blood, as well as products for animal health.  We have three primary objectives within this effort.  We want to enhance our ability to more rapidly identify risks associated with these products.  We want to increase our capacity to analyze these risks.  We also want to be taking appropriate actions to communicate risks and actually correct problems that are identified and known.


     In order to think through all these issues and come up with a thorough strategic plan and action plan that will be implemented in the next two years, we have put together four different agency-wide working groups.  There are 55 people involved across five centers at this point.

     The first one is focusing on how we can improve our current reporting systems.  So, the adverse event reporting systems that are across CFSAN and CDER or CBER, CVM are all our focus.

     We are also trying to take a more careful evaluation of all the external data sources that exist to better identify risks.  We currently have access to many in-house healthcare databases from payers, from the CDC, from AHRQ that we do actively use, but we are trying to do a more comprehensive inventory and address how else we could be accessing data that is outside the agency.

     We would like to improve our risk communication efforts both to consumers and healthcare professionals.  Historically, we have been very reliant on certain types of dissemination vehicles and we are trying to think more broadly now about how we can partner with people and implement some new ways of communicating directly to stakeholders.

     We are also trying to develop other approaches to control risks and reduce risks.  A good example most recently is the bar coding rule.


     Relevant to the adverse event reporting systems, we recently announced the proposed suspected adverse drug reaction rule which has been many years in the making.  It was largely developed in conjunction with industry and many different foreign regulatory bodies.  In essence, it proposes new standards to improve both the quality and the usefulness of the data that we are collecting, but it is also going to reduce burden in that it does allow for one common set of definitions and procedures for reporting across all of these countries.

     There are many different features to this rule.  We will be getting expedited critical safety information.  We will be getting more volume of information on medication errors.  But it is a 500-page rule and we don't have time to get into it in too much detail today.


     Again, with adverse event reporting systems we have been continuing to implement the MedSun program.  We are currently across 80 facilities in the U.S. and we hope to expand to additional 100 in the next year.  For those of you who are not familiar with this, it is an Internet-based reporting system that is allowing more direct interaction with healthcare professionals to get device-related adverse event reports in.

     We are also trying to identify potential partners that are already active in the patient safety arena, that are collecting various types of reports related to our products.  One company in particular we have recently met with has 50,000 reports already collected across 100 different hospitals in the country, and we would like to be able to develop the capacity to be able to get this information in on a regular basis since it is really above and beyond anything we would normally have access to.


     As I mentioned before, we are also trying to look carefully at the databases we currently do have in-house to identify and analyze risks, but we are thinking beyond that now and trying to look across 24 different databases, both public and private, to ascertain how we can better measure exposure and risks associated with all of the products of interest.  Many of them we have identified and tried to target to address the concerns that we will likely have under products specifically approved under PDUFA III since there will be some funding available to do some surveillance of those products.  We would also like to plan forward, not just with 03 but plan to tie in to the risk management programs that will be proposed under PDUFA III but to look at future years and see how we can meet these needs.


     Another important project we have going on related to external data sources i the Marconi Project.  This has been recently mentioned in announcements.  It is coming out of a private-public collaboration with the Marconi Foundation and an initiative called connecting for health that is involved with setting clinical standards to enable exchange of healthcare information.  Essentially, what we are doing is partnering with healthcare facilities that will be sending in periodic reports, but it will be automatic.  So, if a woman on thalidomide is tested positive in a pregnancy test, it will automatically facilitate a report that will go into the agency.

     We actually plan to get data starting in June and this really will be our proof of concept study.  We hope to expand this effort with the healthcare providers that are participating now  and look more towards signal detection and trying to identify unexpected events, as well as try to perhaps tie this into some active surveillance of the products that will be marketed or approved under PDUFA III.


     We are also going to be developing a strategy specifically to facilitate this ability and we have the interest of many of the providers that are already engaged in this effort. There have also been other activities ongoing in the Department of Defense in this area.  So we hope to try to learn from everyone who has already been involved to try to use the tools and come up with a better way for real-time assessment of product risks.

     Of course, there are some concerns with HIPAA given that we will likely have access to electronic medical records.  We have attorneys with expertise in this area who are very much involved with the working groups and are thinking through these issues carefully.


     In risk communication, we have recently completed an inventory of all the types of risk communication efforts that we have going on across the agency to try to get a better handle on our baseline.  We think there are some known weaknesses but we do really need a more careful evaluation of some of our key ways of communicating to better understand how we can improve them and also build on them.  For example, we have known historically that the package insert has a lot of very valuable information but it is perhaps as user-friendly as it should be to most physicians.

     We have also started exploring new ideas as to how we could perhaps get more information out on a regular basis to providers, in particular, who are prescribing products with some known risks.  While we do have many different mechanisms for doing that now, we would like to try to develop better ways of doing that, perhaps emailing certain groups of physicians who are prescribing products, and doing that based upon the idea that we would be facilitating the collection of additional information so we could better ascertain the true risks.


     We also have an ongoing initiative, called the DailyMed.  This is in conjunction with the National Library of Medicine.  This is also related to two different rules, both the physician labeling rule as well as the electronic labeling rule.  What we hope to accomplish with this initiative over the next couple of years is providing healthcare information systems with product labeling that will be updated on a real-time basis so that that can be fed into various types of tools like decision support technologies that could be used at the point of care.

     We also have a variety of proposed initiatives.  Right now we would like to get better information out over the Internet and we don't have a very well coordinated place for people to go that would give you all the information that would be relevant and important for all of our products.  We would also like to expand on our electronic communication of tools we have already developed, like the patient safety news, which should be relatively low cost and easy to do.


     I think I mentioned before that we have recently proposed the bar coding rule, several weeks ago.  This will require bar codes on all prescription drugs, over-the-counter drugs packaged for hospital use, vaccines, blood and blood components.  This will allow for unique identification of these products.  They will be able to scan them in and identify them through NDC codes.  This is expected to facilitate the uptake of scanners across the country at hospitals and, in essence, prevent a lot of medication administration errors which are very prevalent.  So, things like the wrong dose, the wrong drug or the wrong time to administer a product could be avoided.

     I mentioned before that our last group has focused on risk prevention and control.  We are identifying many different actions that we could take under our current statutes to try to correct or prevent problems.  Most recently we have been talking about how to take a more risk-based approach to recalls.  While these ideas are early in our thinking, we are thinking that this could perhaps improve in a more systematic way how we handle a lot of our actions.

     We also have recently released concept papers on risk management and pharmacovigiliance and there is actually a series of public meetings ongoing this week downtown to discuss these concepts.  We hope that in conjunction with these efforts we will be able to better prevent risk given that, at the time of approval, it is more likely that the industry will be engaging with us to figure out how we can work together to prevent risks.


     We have various different partnerships that are under way or at least being explored.  With CDC we have had several discussions about how we could improve our data collection by adding on specific modules for drugs and devices and potentially foods through their existing surveillance systems.  There seems to be a lot of interest and support of this idea, and it really could drastically improve our ability to collect information related to device-related infections and adverse events that occur in the emergency room.

     With AHRQ we have been exploring different ways that we can facilitate data collection specifically with adverse events.  There has been a portal that has been under discussion for quite some time now that could perhaps facilitate the collection of data from hospitals.

     We also like to share the expertise that we have in this area.  Since legislation is pending on patient safety right now, that could create a national patient safety database.  We would like to work with them and be involved in that effort in thinking through how we might design and access data that would be coming into that database.

     With the VA we have been under discussions as to how to sort of continue our collaboration with them.  Historically, we have used their pharmacy benefit management database which has been very useful for specific projects.  But we would like to expand our efforts with them and use their data sources on a more regular basis.  They are actually quite ahead of the curve in terms of electronic medical records.  They have a variety of health services researchers and pharmacoepidemiologists who are very interested in collaborating with us.


     We have also been trying to think about how we can work with different outside organizations to improve communication.  Most recently we have been discussing this with the Joint Commission.  We have several projects that we would like to pilot with them.  In fact, we have a meeting this afternoon with them.  They, in particular, have a lot of interest in root cause analysis associated with medical errors and would like to be a part of getting risk messages out to various organizations.


     In the future, we would like to obviously build up our spontaneous reporting systems but, given the voluntary nature and the fact that it is passive surveillance, we would really like to move to a system that would be much more automated and less reliant on healthcare professionals to take time out of their busy days to report to manufacturers and to us.  So, we really would like to be using more automated tools to collect data, to analyze data and the Marconi Project is really the most relevant example of an effort there.

     We also need to make sure that we are going to have a flow of information across information systems.  Right now there is a lot of effort under way to develop standards to facilitate this but we feel we really need to be part of this effort to make this type of thing work.  Again, we want to be continuing to use healthcare databases and other outside sources so we can improve our ability to monitor risk on a more real time basis.


     We feel that the better we become at rapid detection of adverse events and medical errors, the better able we will be to take steps to actually prevent them.  We really do want to improve our risk communication with the intent that informed healthcare professionals and consumer are going to be able to make better decisions about their role in the healthcare system, and also be able to prevent obviously the many adverse events, morbidity and mortality and associated healthcare costs.


     As I mentioned before, we have many different challenges.  Of course, we have budget constraints.  We have our two-year action plan that is focused on being resource neutral, meaning that we are not anticipating any additional funds.  So, we are trying to establish what is feasible given the FTEs we have across the centers and the expertise that we have across the centers, but we also realize that we are going to have to rely on partnerships with healthcare providers, with other government agencies, with accreditation bodies to really carry a lot of this through.

     Our new initiatives are also very IT intensive.  We realize that we have to be thinking about a lot of new and innovative ideas in the context of consolidation which is being directed by the Department.  We really do at this point have less control over how we are spending our IT resources when we really need to be thinking creatively about how we are going to be making electronic labeling a reality and getting that directly to the provider.  Also with some limited resources, we have some issues with trying to retain and recruit talent.

     We also have legacy IT systems that we are going to need to transition out of but, yet, we can't interrupt the functioning of our programs.  So, these are all considerations that we need to keep in mind as we are moving forward.  I think that we are trying to deal with our barriers and our obstacles as best we can, and we are being very creative in our use of partnering and what-not, but these are challenges that are not going to go away.  Thank you.

Questions and Discussion

     DR. DOYLE:  All right, we are getting a little bit close to lunch time so, with Dr. Woodcock's permission, we will save your presentation until after lunch and initiate some discussion of our last two presentations.  So, if anyone has any thoughts or comments, questions?  Dr. Hussain did give us questions that he would like for us to comment on relative to how FDA might improve their articulation of desired state considerations for communicating the desired state to stakeholders, and any additional considerations for aligning FDA's activities to ensure efficient progress.  So, with that, any thoughts?  Dr. Shine?

     DR. SHINE:  A question for Dr. Hussain and also for Kelly.  Dr. Hussain, there is a little paragraph about team biologics in your report but the overall flavor of the report is about drugs and drug manufacturing, and the expertise that you described was pharmacy, chemical, chemical engineering.  Clearly, one of the areas that has been most troubling has been manufacturing processes for vaccines and the biological soups that have to be created to do that, and also the concerns about what has happened with vaccine manufacturers not being in a position to continue production.  Could you give us some sense as to the extent to which that is a focus for this initiative and what we are doing with regard to addressing that particular element?  Then I have a question for Kelly.

     DR. HUSSAIN:  My presentation started with the PAT initiative which is focused more on traditional pharmaceutical manufacturing, and within that context I said, you know, chemistry, pharmacy, engineering.  So, that was the focus of the PAT initiative and that continues to be the focus.

     I think in many ways the opportunities for improving efficiency are tremendous but with respect to the level of complexity we have in terms of scientific understanding and so forth because vaccines, and so forth, are far more complex systems so the PAT initiative started with that focus in mind.  The GMP initiative actually is trying to address all pharmaceutical human drug manufacturing including vaccines.  So, within that context that is being addressed, but with respect to the PAT, we have kept that limited at this time to the traditional pharmaceutics.

     DR. CARBONE:  We do have representation in that group, and I think the encouraging component is what was mentioned by Dr. Hussain, CBER's plans include product specialists in inspections, and this is a plan that is under consideration for adaptation and adoption.  So, I think that the biologic specific issues, although not prominent, certainly will be part of the consideration and can be increasingly so as the need arises because you are absolutely right about the complexity and specific issues in manufacture of biologics, and vaccines are an example.

     DR. SHINE:  Given the challenge in this area, Mr. Chairman, I would hope that sometime in a future meeting we might address the question of the status of vaccines and vaccine manufacture.  I think with the crises that have occurred with regard to childhood vaccines, vaccines for terrorism and so forth, it is a really pressing issue that I think the agency has a great deal to contribute to.

     I just wanted to ask Kelly a couple of things.  First of all, congratulations on getting bar coding.  As you well know better than any of us, the VA have been doing that for a long time.  There are two issues.  One, Marconi is a hospital-based experiment?

     MS. CRONIN:  Right now it is, although there are some providers that are involved that have vertically integrated healthcare systems where we could potentially be getting data across settings of care.

     DR. SHINE:  That would be terrific, and the VA is another example of being able to do that because we really don't know very much about what is going on in the ambulatory arena, and beginning to get some meaningful data there would be extremely useful.

     The other question I have is, as you are probably aware, when you try to get healthcare providers to pay attention to the future the hospital administrators are interested in a microsecond but the docs are not.  The docs are resistant.  Recent polls have demonstrated that only about a quarter to a third of docs really acknowledge the importance of errors, and so forth.  Part of the reason is they believe it is something that takes place somewhere else.  The most successful programs have been ones in which a hospital has looked within its own organization and identified the error rate in their institution and being able to say to the medical staff we have just as much of a problem here as they do at the hospital across the street where you think everything is happening.  Is there any mechanism, as you develop this information--I am not talking about public reporting or whatever; I am talking about the question of any kind of feedback mechanisms so that we can get information to people that will get them to take seriously the notion that there is a problem in their environment as opposed to somewhere else?

     MS. CRONIN:  We are thinking about various ways we can facilitate communication and feedback once risks are identified.  We haven't really worked through all those ideas yet.  We hope to discuss that at an upcoming brain-storming session with many of the providers that are involved in Marconi.  But we realize that quality improvement and prevention of risk is really an integral part of this effort and in order to really get people to buy into this effort, since it is really above and beyond any regulatory effort and is really just relying on their interest and their participation, we will have to be communicating what we learn back to them in an effective way so that they can then turn to prevent risks and improve quality in their systems.

     DR. DOYLE:  Dr. Woodcock?

     DR. WOODCOCK:  Yes, I have a comment about this.  One of the more promising discussions I feel we have had has been with some of the payers because for the data you are referring to, you have to talk to the practitioner about their practice pattern.  Some of the payers are actually willing to collaborate with us in feedback mechanisms around prescribing patterns and consequences, and so forth.  That type of feedback is very personal, of course, but it also is very effective in changing behavior compared to exhortations or general communications, and so forth.  So, there could be some interest in actually communicating to patients who are on specific medicines about the risk that would pertain to that medicine.

     DR. DOYLE:  Dr. Thomas?

     DR. THOMAS:  I have a question for Dr. Hussain.  Where do you think you are going to get the next generation of PAT experts or professionals?  For example, in schools of pharmacy it is very difficult to find a so-called physical pharmacist.  Schools of engineering, while some may have very strong chem. engineering may not be interested in this sort of manufacturing process.  Also departments of chemistry, I mean, the medicinal chemist obviously isn't going to fulfill the criteria that you need for a PAT.  Has anyone done any manpower surveys?  Has anyone looked at curricula within the U.S. academic institutions with regard to this type of training?  It used to be more prevalent but, for whatever reason, it is not something that young people are going into.

     DR. HUSSAIN:  Just to sort of reflect on that fact, before I came to the agency I was on the faculty of a school of pharmacy for nine years and I saw the erosion of the physical sciences from those curricula.  I think the sensitive issue is in the sense that schools of pharmacy have a very important function from a patient care perspective and that is where the focus has been and will continue to be.  So, the gap that remains is industrial pharmacy, physical pharmacy and where this information will be utilized to train individuals.  The trend has been that schools of engineering, especially chemical engineering, have picked that up.  For example, the Rutgers School of Pharmacy essentially has moved away from industrial pharmacy program but the School of Chemical Engineering now has a very mature pharmaceutical engineering program.  So does the University of Michigan.  Purdue has still maintained its strong focus on industrial pharmacy with the PAT-based applications.

     So, there are now at least three focused programs that at least we have direct contact with which have already moved in this direction.  But what I would like to say is that from a chemistry perspective, process analytical chemistry actually has evolved over the last 30 years as a mature science.  So, you do have opportunity to tap into that pool of talent and then sort of combine that with the pharmaceutical know-how.  I think it has to be a team approach.

     So, in the short term the solution is to bring a team concept to this with engineers from other sectors which have the know-how for some of these.  Eventually what I see is that some of these programs will mature and will provide the know-how or the talent base for the U.S.  But that is a major concern when we have spoken to a number of companies about where the new plants are going to be, PAT-based, many have said it is going to be Germany, not U.S., and that is precisely the reason why.  The German education system has maintained a strong base and is providing that know-how.  But I think the concern I have always had is that we need to have that base in the U.S. also.  That was the reason for sort of partnering with the National Science Foundation and creating some opportunities to have this talent pool.

     In short, I think we will have the talent pool as a team concept but eventually this will take off.  The University of Michigan actually will have a distance learning program based on PAT systems and so will the University of Purdue.  So, those curricula will be established very soon.

     DR. THOMAS:  Thank you.

     DR. DAVIS:  You mentioned, Dr. Hussain, the use of case reports to help work up the desired state.  I would suggest, sort of thinking out of the box, because you don't have a lot of programs, training programs that you might try case reports as a teaching tool to provide information to those schools that aren't so strong.  You know, I am not business or an MBA person but I often look at the Harvard Business Review from a learning perspective.  It serves as a great way to get information out using case reports to study.  So, I would suggest scientists need to consider doing the same thing where we have training programs that are deficient in some specific region so instructors might take this information and at least use it to talk about.

     The question I have for you though, I notice you had ORA as a part of the team.  How are you anticipating these new technologies in CFR Part 11 from the start?  I was very pleased to see you had ORA.  I am just wondering what is the input there in using CFR Part 11 with these new technologies getting started at the ground level?

     DR. HUSSAIN:  Let me go to the first question.  I think the point is well taken in the sense that I think case studies, case reports would really be examples.  What has been quite gratifying is that companies have stepped forward with some examples and have shared those publicly.  For example, I think Bristol-Myers Squibb came up with a complete case study presentation at our PAT subcommittee, and this is how it can be done; this is how they are doing it.  So have Pfizer and others.  So, those really have helped us sort of conceptualize the concept itself.  At the same time, I think the curriculum that we have developed for PAT staff training and certification actually has become a model for academia to adopt.  So, in this instance we are actually providing a curriculum to academia.  So I think that was a positive step of our deliberations on that advisory committee.

     On the issue of Part 11, I think we did realize that the interpretation of Part 11 could have created some of the hurdles for new technology and innovation.  Therefore, as part of that I think we should draft a guidance as part of the cGMP initiative.  As we move forward, I think the integrated team concept is the only way forward and I think the centers are working hand-in-hand to get there not only with respect to computer validation, software validation and Part 11 issues but every aspect of science because we will be moving to a more sophisticated statistical base analysis and I think that is a learning curve that needs to be sort of understood as well as applied appropriately.

     MR. MARZILLI:  This is John Marzilli, from ORA.  I wanted to echo that sentiment and just to add that Dr. Hussain, with Dr. Woodcock's leadership, has been involved with our senior management staff at our past two senior staff meetings where we have presented this to our regional directors and district directors.  On the GMP initiative since its inception, Dr. Woodcock and company have worked closely with our Division of Field Investigations and our field staff from across the country to our drug field committee.  Our Chair, the District Director, Doug Ellsworth from the New Jersey District, as well as our Regional Director, Susan Suderberg from the Central Region and members of that committee have worked closely with these work groups and we have members on each of these work groups.

     Because we are dispersed across the country, we try to have our folks come in as much as possible to participate and generally we have been active participants from day one in these initiatives.  I want to thank Dr. Woodcock for bringing us on board, because it is an important aspect of the field organization, at the table from day one.  So, we have been there.  Thank you.

     DR. DOYLE:  Did that cover your points, Dr. Davis?  Dr. Shine?

     DR. SHINE:  What kind of metrics are you going to use to evaluate this program in the future?

     DR. HUSSAIN:  Well, I think we have been giving some thought to that.  One clear metric is actual real-life applications and moving towards adoption of some of these technologies, in that regard, bringing technologies that are already existing in the "don't tell" mode and actually having some regulatory utility and benefit from that.

     In one sense, I think it would be prudent on industry's part to adopt a move in this direction, and that essentially would indicate that the hurdles that are perceived or real are sort of being removed.  The metric for that would be simply the applications that will come through.  That will happen over a period of time.  But most of all, I think sort of a metric in terms of what has already occurred is the almost emotional debate of the current state of pharmaceutical manufacturing in terms of what it was and what it should be.  The workshop that we have held I think brought that debate out, moving towards a consensus to the desired state.  I think we have established that.

     But in terms of collaboration between the centers and the ORA, I think with respect to technical issues and disputes that might come about, whether we resolve those quickly or we actually prevent those issues from happening, those are all a potential source of future metrics that we could come up with.  So.

     DR. SHINE:  For example, will there be any attention paid to cost issues, to interruptions in production, some of the things which are sort of the end measures of what the outcome is--quality of product?  I just want to get some sense at the end of the day of what does this do for the industry.

     DR. WOODCOCK:  Yes, or for the public.  We are going to look at that.  We have an evaluation group that is trying to devise how we would measure we are achieving the desired outcome of this initiative.  So, your suggestion is a good one and we agree.  I mean, we would hope, like Ajaz said, for some work examples from industry because the preliminary indications we have, which were presented previously at the Science Board, is that there could be a major reduction in cost after the initial investment.  So, we look forward to those.  Those are real-life applications that would demonstrate some of the benefits.  But we also need to further develop how we measure whether it is successful or not.

     DR. DOYLE:  Thank you, Dr. Woodcock.  Dr. Laurencin?

     DR. LAURENCIN:  This is for Kelly Cronin.  How do the adverse event reporting strategies that you have compare to other countries, say, like Canada?

     MS. CRONIN:  We actually haven't considered comparing our systems to other countries.  I think with the SADAR rule there are probably going to be more similarities than differences, once that gets finalized.  But we do hope to improve our systems in terms of trying to have better outreach and increased awareness of the importance of reporting.  We also hope to come up with better ways of signal detection through the adverse event reporting databases that we have in-house.  So, we have a variety of different initiatives that we are thinking through now.  Whether they will differentiate us more from other countries, you know, I am not certain.

     DR. WOODCOCK:  We have actually been talking to Canada.  They want to get onto our database and add their reports and be able to analyze because they don't have the power with a much smaller population to be detecting rare adverse events.  Their system is somewhat similar to ours.  European countries are very difficult for pharmaceuticals at least because they have a national healthcare system so much of their reporting is because they are kind of administering the payment of drugs through one system.  In Japan for pharmaceuticals, I don't know for devices, it is a totally different healthcare system where the practitioner actually dispenses the drugs.  They have put in place some adverse event reporting systems that are focused on the practitioner when a drug is newly launched and their early experience with dispensing that drug to patients.  So, we are talking about a lot of different kinds of systems.  It is very interesting but it is hard to make direct comparisons because of the differences in the underlying healthcare systems.

     MS. CRONIN:  Yes, I think it is also fair to say that they are all voluntary as well so it is passive reporting.  Several years ago I did look at data across various different countries that reported to the WHO collaborating center in Sweden.  The data, at least for this particular class of products, was sort of spotty across many different countries.  So, I don't think that any other system is going to have a better way of identifying the true safety profile given that they are all passive systems.

     DR. DOYLE:  Dr. Feigal?

     DR. FEIGAL:  It is interesting to look at the system in the U.K. for medical devices.  In the U.S. 90 percent of our reports come from manufacturers.  In the U.K. 90 percent of their reports come from health professionals.  So, they have a very different system and it is in part related to the fact that the same public health system that runs the hospital systems and the primary care systems also is responsible for the MDA, which is the regulatory authority for devices.

     They have the ability to make some linkages that we don't.  For example, they look at recalling devices as something where they have responsibility for the whole system.  So they have recall coordinators within hospitals that are accountable to the same body, the device authority.  So there are some interesting different systems.

     There is another layer that exists probably for all products, which is communication between regulators of the problems they are analyzing.  It is one thing to have robust systems for signal identification and triaging and data mining and refining systems, but one of the benefits of the harmonization efforts of pharmaceuticals and devices has been more opportunities to share between regulators.

     For example, a few years back problems began developing with a new heart valve and we had reports that we heard from the U.K. and from Canada.  It was a product that had an earlier launch in Europe than it had in the U.S. that alerted us to the problem while there were still clinical trials ongoing and early marketing in the United States where we could look at this system together.  So, I think it is going to be necessary to take a whole systems approach to the situation and recognize that it is really a global marketplace, that these same products are appearing throughout the world in somewhat different orders, and if we want early signals we are not going to be able to get them just by turning up the detection in our own systems.

     DR. DOYLE:  Any other questions or comments?  I think we are getting hungry.  So, I think we are going to take a break now and reconvene at one o'clock.

     [Whereupon, at 11:55 a.m., the proceedings were recessed to reconvene at 1:00 p.m.]

- - -

A F T E R N O O N  P R O C E E D I N G S

Open Public Hearing

     DR. DOYLE:  I guess we are ready to reconvene.  Next on the agenda is public comment.  We don't have anyone officially registered for public comment but is there anyone in the public audience here who would have a comment?  I see no hands so I guess no public comments.

     DR. BOND:  I do have one written comment that I am just going to submit to the transcriber for the record.

     [Written comment submitted for the record:  "It is clear that FDA is not protecting the public from cancer or other illness risks.  It is clear that FDA is allowing industry to sell any old product to the public so that the seller makes money, and the public gets sick.  The FDA is there primarily to protect the public and all standards should do so.  If it takes 20 years of testing to prove a product safe, that should be the standard, not anything less.  These are my comments for inclusion in the public record since I cannot attend Washington conference on scientific standards.  B. Sachau, 15 Elm St., Florham Park, NJ 07932."]

     DR. DOYLE:  Very good.  We will submit that for the record and move on.  Next we are going to hear from Dr. Woodcock again.  We are going to make her earn her keep today.  She is going to talk about fostering technology development in relation to pharmacogenomics.

Fostering Technology Development--Pharmacogenomics

     DR. WOODCOCK:  Thank you.


     Good afternoon, folks.  I hope we will have a lively discussion this afternoon and keep everybody awake after lunch.  This continues the theme of facilitating the introduction of new technologies through regulation.  This afternoon we are going to talk about pharmacogenomics, pharmacogenetics.  I want to say that many of the centers are involved in different aspects of this.

     As I said earlier, NCTR obviously, as was said earlier, is extensively involved in certain scientific assays and so forth.  The Center for Biologics has a very extensive genomics/proteomics program.  But today what I wanted to talk about is the relationship of genomics and, by extrapolation, really the new proteomic techniques and similar techniques.  I am going to talk about the genomic techniques but really these issues apply to all these techniques in the development and regulation of drugs.  So, that is the "pharmaco" in the genomics.


     As I said, this is really about translating innovative science to bedside medicine.  We know this is often a very rocky road and there are a lot of bumps along the road, and part of the initiative that Dr. McClellan spoke about this morning is trying to smooth that pathway and help these new innovations translate into benefit to patients.  As Dr. Shine would say, that it the outcome measure we want to see here.  We want to see the actual benefit to patients.


     What is the issue we are talking about here?  This new science of pharmacogenomics, and I will define this later, and proteomics and other similar technologies are actually being applied extensively by the pharmaceutical industry in drug development.

     They have the potential to revolutionize a number of the processes.  Most of the data that is being generated is not being seen by regulatory agencies, partly out of concern for how the data will be used.  All right?  We need an approach to this new data that is going to be generated, and is being generated right now, that will enable the free exchange of information between regulators, companies and, hopefully, to the extent possible the scientific community at large.  It will help advance the science and technology and, very importantly, aid in the timely development of appropriate regulatory policies because we cannot have this revolution upon us before we start formulating our policy development.  There are many scientific and technical issues that are going to have to be solved and some of the later speakers are going to address some of these.


     Basically what am I talking about here?  I am going to give you a little bit of background, if you will, on why this is important.  There is tremendous variability in how people respond to medicines, and that is such a commonly understood fact that most people don't even think about it but that is really a major barrier to having effective therapeutics.  It is also a major barrier to developing drugs and biologics because we can't predict who is going to respond.


     This is true for effectiveness.  For many drugs, if you leave aside antibiotics, antivirals and so forth that are directed at some other organism, other than a person, for many drugs the size of treatment effect when we do randomized trials is less than 10 percent of some outcome measure.  It is a very small effect.  Many people conclude from this that the effectiveness of drugs is very small.  Even some of our own staff will tell me this drug doesn't work; it is such a small effect that we observe in the population.


     You expose a population; you randomize them and you see something like this.  You see that the mean response of the drug is enough over the placebo that, in fact, you can demonstrate a statistically significant difference in a population but it is not anything to really write home about clinically.  This is very common and one of the problems in drug development because you have to have very large trials to show effectiveness of drugs.


     If you did trials in a different way and you set up the responders, what often you may find is that most of the subjects on the drug arm look just like the placebo subjects.  In other words, they don't respond or, you know, some of them get better, some of them get worse, just like the placebo.  All right?  But there is a little group of people that respond very much to the drug.  This is a common observation, except we don't usually conduct trials using this type of hypothesis and we don't usually report them in medical literature according to the magnitude of the response; we look at the mean difference between populations.  One of the goals and one of the issues here is to predict who these people are, the small group of people who actually do respond to a specific drug in a very positive way.


     What about drug toxicity?  Again, this observation is so common we don't even think about it.  Often not everybody gets the drug toxicity.  If you study drug versus placebo you see each drug has a consistent pattern of side effects, and that is repeated over and over again in different trials when you compare it to placebo.  That is observed with the common as well as the rare events.  Some of these effects are attributable to known pharmacologic effects of the drug and others, in medicine we have always called them idiosyncratic.  Idiosyncratic is not defined this way in Webster's but what we mean is we don't know what this means at all; you know, it just happened.  Actually, from a science perspective you have to say there was a reason; there was a reason for this.  There is a reason that some people suffer more from the pharmacologic effects of the drug than others.  It can't be just random but that is how we have thought about it.  Random is another word for we don't understand the causes; we don't understand the underlying cause.

     So, the current physician medical approach is kind of at the level of organ function.  We say, well, people have compromised liver function and you have to be careful here, or whatever, or we just observe the patients and we wait for something to happen that we try to prevent from getting worse.  We can't really often predict who are the people who are going to experience these adverse events and why; what is the underlying cause that they get them and some other people don't get them?


     It is commonly agreed, and we will probably convince you in our presentations today, that there is an inherited or a genetic component to variability in drug response.  Here are your definitions.  Pharmacogenomics we are defining, just for the purposes of this meeting since these are controversial definitions but this is what we are going to talk about today, pharmacogenomics: application of genome-wide RNA or DNA analysis to study differences in drug actions or drug effects.

     The earlier term, pharmacogenetics, is by many people considered to be study of genetic basis for individual differences in pharmacokinetics.  The reason for that distinction is pharmacogenetics was discovered fairly early.  We have known for a long time that different people metabolize drugs differently, the same drug differently.  There was even early study of this and some things have been known for a long time, using probes and so forth to find out people's phenotypes.  So we were able to learn about pharmacogenetics a long time ago for some drugs but we are learning a lot more now because we have better methods of studying it.  Today I am going to talk about the broad issue of pharmacogenomics and some subsequent speakers will focus more or less on the drug metabolism issues.


     What genetic background issues could affect efficacy and what do we know about this?  Well, we know there can be genetic diversity of disease pathogenesis.  So people have studied, for example, hyperlipidemia and there are many different genetic differences in people and how they handle lipids--lipid transport, lipid receptors, all sorts of things.  In some cases there is a suggestion that perhaps that may, in fact, alter their responsiveness to certain drug interventions, and that makes sense depending on where the variability is and how they handle the lipids.

     Variable drug metabolism is very well understood as a cause of lack of efficacy.  There are some people called hypermetabolizers.  Larry Lesko probably talked about this.  You give them the usual dose of drug and their body chews it right up so they don't really have any drug around and so they can't respond; they need a much higher dose.  So that is a cause of lack of efficacy and, depending on how big the population of hypermetabolizers is for that particular group, you have a certain group of people who are genetically fated not to respond to that drug at that dose, at the usual dose.

     As I showed you earlier, we often determine mean population effects so we determine a mean dose for the population and that is what everybody gets.  Well, that doesn't work when the drug is one that is variably metabolized in a human population.

     Then, there are genetically based pharmacodynamic effects.  What does that mean?  Well, that means that the beta-adrenergic receptor may be an example of this.  People have genetic differences in their beta-adrenergic receptor.  That is not believed to cause them to experience asthma at a higher rate but certain people whose adrenergic receptor is different may not respond the same way to beta-adrenergic agents.  In other words, their drug responsiveness is modified by the fact of their different genetic background.

     Most of these things I have been mentioning here are somewhat controversial.  There are studies to support to this; studies that don't support this.  But basically the story is that there are probably three different ways where efficacy can be impacted by genetic differences.


     What about toxicity?  Again, there are underlying genetic contributions to the variability in the toxic response.  For example, if you have certain inherited syndromes that affect cardiac repolarization and then you are also given a drug that happens to also affect cardiac repolarization you are going to be much worse off than a person who didn't have that genetic background.  This, along with a long QT syndrome is something you can identify phenotypically.  In other words, you can measure someone's electrocardiogram and find out about this but there are probably many underlying genetic problems where we can't easily measure them with our current measures and so we give them a drug and there is a bad interaction.

     Again, differences in drug metabolism also go the other way.  We have people who are outliers in the population who metabolize drugs very slowly.  A good example is thiopurine methyltransferase.  Drugs like azathioprine or 6MP are metabolized through the pathway and the population has normal metabolizers--it has various metabolizers but there are people who are very slow who lack the ordinary metabolic path.  Their AUC will be about ten times higher on administration of these drugs than the normal person--well, these are all normal variants--than the most common person.  This is very important because 6MP for example is a cancer chemotherapy agent, and so forth.  It is used in the treatment of inflammatory bowel disease, chronic use.  So, it is a very important thing that may result in various drug toxicities.  There is a lot of information about this in the literature and similar cases.

     Another example of a different kind of genetic contribution might be what I call a toxicodynamic interaction.  That is not where you have a genetically based abnormality or you don't have a metabolism difference; you have some normal existing state that somehow interacts with the drug that can cause a severe reaction.  People have postulated that a hypersensitivity reaction to abacavir might be caused by association with a certain MHC type, for example.  Again, these are still being studied.  We are not really sure of the association but this is the kind of prediction that could be made.


     How important are these differences?  You could say, well, they exist but they are not that important.  How much of the variability could we predict if we knew the genetic substrate?  That is an important question.


     At the level of an individual--and that is what people always ask us, they say what drug is right for me?  What dose is right for me?  They want individualized therapy.  Obviously, at the level of an individual a genetic difference may actually determine the drug response.  I am using a very obvious one here, enzyme deficiency.  Obviously, if you pick the wrong replacement enzyme for somebody then they are not going to respond and that may not be completely clear phenotypically.

     Also, it may highly influence drug response.  That is what I have already talked about, the polymorphic drug metabolizing enzymes.  In some cases these are not that important because there are multiple pathways; because the therapeutic index of the drug is very broad; for a variety of reasons.  In other cases these are extremely important and will predict to some extent who gets toxicity; who fails to get efficacy if you don't control the dose correctly.


     Unfortunately, we have to recognize that many responses of people to drugs, just like everything else in life, are going to be an emergent property of multiple gene products interacting with each other and with other environmental factors so you get a system where you have multiple gene products.  They all can have genetic differences.  They interact with each other and it can be very difficult to predict how that will affect their drug response.

     So, many individual genetic differences or even the patterns of differences, that Frank Sistare is going to talk about, even transcriptional differences may not be that highly predictive of drug response.  So, we are going to have to sort this out over time.


     This is kind of a picture that was just in Science about the body and actually how it is organized.  The point is that there is a level--we are talking at the genetic level, the RNA level and the protein level, expression level and so forth--the body though has larger systems imposed on top of that to become a very large-scale complex interacting system.  So, we are probably naive, if we look at a few things at the bottom, here, that we are going to tell how the system is going to read this out.  All right?  So, genetics is not determinative, except in some cases.


     Let's go back to drug development, the background of how the genetics could affect drug development.  Currently, in drug development we have a satisfactory way to determine efficacy.  As I said, it has to be done on a population basis.  We randomize a whole bunch of people; we look at what happens and then we decide whether a drug works or not.  It is the same with a determination of drug toxicity.

     It is observational.  First we expose a bunch of animals--and I apologize to the toxicologists here, I am sure it is much more scientific than that, but basically you expose a bunch of animals and then we expose a lot of people and, again, it is observational.  We expose a lot of people to the drug and we write down what happens.  If we are lucky we compare it to a placebo or something so we can tell which ones are real.  In fact, even for approved drugs the carcinogenic potential and repro. tox. potential is based on in vitro and animal studies because obviously we are not going to find that out in people.


     So, that is drug development.  It does well; gets stuff on the market.  But how can we--we, meaning the whole enterprise--use pharmacogenetics in drug development?  Number one, and that is being done now and this doesn't really involve the FDA, we can improve candidate drug selection.  Obviously that is in drug discovery.

     We could develop new sets of biomarkers, and that is being worked on too and I think that is what NCTR is working on as well as others for toxic response in animals and humans which will eventually perhaps decrease or minimize animal studies, or at least make them much more predictive and be able to understand the underlying mechanisms.

     Perhaps we could predict who will respond to a drug at some level; predict who will have serious side effects and rationalize drug dosing, which would be a very helpful thing to do.


     In fact, Ajaz and I were talking about this, this has a lot of parallels to the PAT GMP initiative we are doing.  The pharmacogenomics has the potential to move us from a current empirical process--and that is really what drug development is right now; it is empirical, not mechanistic--to a mechanism-based process that is hypothesis driven.  Really that is pretty far in the future, I admit, but we ought to be trying to go there because that is the science we are seeking.

     This could in the future result in a lower cost, a faster drug development process that would, nevertheless, result in more effective, less toxic drugs but for smaller a population, the people to whom it was targeted.


     How is pharmacogenomics being used now by the pharmaceutical industry?  We are going to have a later presentation by a member of the pharmaceutical industry so I will skip over this pretty quickly, but I want to just give you a little bit more of the background before we get to the next speakers.

     As I said, it is being used in discovery and identification of lead candidates, identifying targets, evaluating cellular or animal responses to different candidates.  But it is hard to make a lot of sense right now, I think, out of a lot of this data and this is not part of regulatory submissions, not required.  This is in the discovery and nonclinical part.


     Within the nonclinical it is in the path toward a regulatory submission.  There are a lot of exploratory studies going on.  There are a lot of directed studies of genes of interest.  Sometimes we see what I call an explanatory study.  Well, what is that?  How is it different from exploratory?  Explanatory studies--well, you find a toxic effect in animals.  You don't think it is related to people or other animals and you can look at the responses elicited during that toxic response and you can perhaps show they are different than in other animals or maybe even in human cells, or whatever, and that explains that this can't be generalized.  Then, people are working on predictive toxicity patterns.


     What about human studies?  Well, the real hope is that we can sort disease syndromes into subgroups based on genetic differences and pathogenesis and these subgroups may have different responses to treatment.  This is undoubtedly true because many of our diagnoses of disease are, again, observational based on syndromes and these syndromes may contain many different disease pathogenesis groups.  Almost by definition, it is like these are going to respond differently to different drugs also in humans, looking at genetic or phenotypic tests for metabolizer status to predict dosing, rational dosing, and see how metabolism interacts with dosing.


     Then you can go the other way and you can look for genetic differences in responders versus non-responders.  Some of that is being done.  People respond to a drug.  Are they somehow different and you find out in the genetics versus non-responders?  Can you look at RNA expression and look for pharmacodynamic differences, different pharmacodynamic responses?

     There is a hope that soon--and I know nothing about this and whether this is a reasonable hope or not that some day soon we could find some genetic explanation for the severe or catastrophic side effects that people experience nowadays from drugs.  This is really a bad problem because we have drugs out there that really help people, help large numbers of people and, yet, every once in a while, and completely unpredictably from what we know now, they cause catastrophic side effects that are fatal or terrible.  We don't have a way to predict who is at risk right now, and that is a huge problem I think for medicine.

     This again is a typo and I am sorry.  This is supposed to be SMP and gene expression screening.  Other folks are going to talk about this more but these are the kind of things that are being done in people.

     What I am going to do now is pause and I am going to have the other speakers come up.  We are going to hear from two FDA speakers next about what we are seeing and what we are doing from our side.  Then we are going to hear from industry about what they are doing and what their concerns are about the sharing of these data.  Then I am going to come back and present our proposal to deal with this and then we are going to hear from a medical ethicist from NIH, who has kindly consented to talk about this problem, and then we can have a general discussion of the proposal.  Thank you.

     DR. DOYLE:  Thank you, Dr. Woodcock.  Next we are going to hear from Dr. Frank Sistare, who is the Acting Director in the Office of Testing and Research at CDER.  Dr. Sistare is going to share with us a presentation on pharmacogenetics in preclinical studies.

Pharmacogenomics--Preclinical Studies

     DR. SISTARE:  This is a good time slot because we saved an hour for the open public discussion so that means I have an hour to talk--


     I think I have 40 slides here.  I am going to breeze through a bunch of them but it is going to be a challenge to get through in 30 minutes but I am willing to take on that challenge.  Let the record reflect it is 1:27!


     In fact, I am going to alter my original charge.  My original charge was going to focus just on preclinical but we realize that there would be a big gap missing if I focused only on the preclinical and we decided that a fair divvying up of the action was for me to focus on RNA and Larry is going to focus on DNA.


     What I am going to do is give you an overview first of the technology and get everybody up on the same page, and then talk about some of the medically related applications of pharmacogenomics relevant to CDER's responsibilities.

     I am going to highlight some of the concerns and issues that we have already heard or that we have already experienced ourselves, and I categorize those in three major areas, technical, biological and procedural, that have been raised at that triad interface between patient care, drug development and regulatory oversight.  Finally, I am going to talk about what CDER is doing to address those issues presently.


     Again, to get everybody on the same page, this is a very simplistic cartoon that shows the double helix DNA molecule.  From the DNA, RNA is made and then from RNA protein is made.  They are not all made in the same place obviously.  There is a little trafficking that needs to go on there.  But what I am going to talk to you about is the RNA component of this process.


     RNA is measured in a number of different ways.  RNA can be converted to cDNA and cDNA or DNA can be amplified through a process of PCR.  This has been a well tried and well tested and highly depended upon method of measuring RNA in a quantitative way.


     But in 1996 what really stimulated the focus on measuring RNA molecules as a function of disease and drug intervention was the application of the DNA microarray.  Again, this cartoon is very simplistic but what it shows is a two-dimensional surface to which DNA sequences are attached.  These serve as probes.  A sample can be prepared.  RNA is isolated from those samples of interest and they are labeled in some way, copies of those things are labeled in some way.  Then, the labeled copies of the samples, together with the two-dimensional capture matrix, are hybridized and you can measure a signal ultimately.  There are a lot of technical aspects to being able to get the data out of this process but this is essentially the process.


     There are a number of different platforms, a number of different specific technologies that have been used to measure RNA or gene expression.  There is the dual color approach in which a control and a treated sample or a normal and a diseased sample can be hybridized to the same exact same platform.  That platform exists in several different ways.  You can have a long 500 to 2000 base pair length probe sequence attached to that surface, called the cDNA microarray.  You can have small oligonucleotides attached to that capture surface, on the order of 30-60 length of nucleotides.  Or, it can have in situ synthesized oligo microarrays.  These all exist out there in the marketplace.


     Then there is another whole species of microarrays.  There is the GeneChip from Affymetrix which is a high density probe array and 500,000 specific 25mer oligonucleotides are synthesized, and this is a single color approach.  So, controlled, treated, diseased or normal are applied to two different probes.

     I said that just as background to get into some of the other issues I will be talking about later about technical aspects of the technology that need to be established.


     What I am doing here is just giving you a feel for the fact that these things are probably not a flash in the pan.  We are not talking about cold fusion here I think.  This is a technology that is on the rise and it is probably here to stay, and it is going to continue to evolve.  There are probably a lot of different ways of making this point.  We can look at publications; we can look at commercial sales.  This is just showing GeneChip sales and the sales of the machinery to analyze these have quadrupled in the last four years.


     Why are people really so interested in measuring RNA?  Essentially, the RNA is serving as a protein function signal.  Proteins are what is making alterations in activity of the cell.  So, something is perturbing the cell, working through proteins, to alter changes in RNA concentration and we have the capability of measuring tens of thousands of these things simultaneously--an incredible capability.

     Well, what are you going to do with tens of thousands of measurements at one time, in one experiment, at one dose, at one time in one animal?  You have to measure collections of these things.  You really need computers obviously.  You need a database.  You need to be able to really efficiently analyze the experiments, individual sets of data across patients, across animals, across times.  You need a relational database to help you with that, to identify changes and the function of what the proteins are doing.


     I am going to spend some time, as I mentioned, and not just talk about the nonclinical but I am going to talk about some clinical.  I think this really kind of excites the imagination and I think it tells you that the future is now.  Okay?  We are being challenged presently with some very exciting data that is out there in the peer-reviewed public literature.

     So, very quickly I am going to go over the next six or eight slides just to impress upon you the reproducibility of this information.  This is a publication, a collaboration of two groups, The Netherlands Cancer Institute and researchers at Rosetta Inpharmatics, who have recently been bought out by Merck.  The focus, as you can see from the title of the publication in The New England Journal is on gene expression signature as a predictor of survival in breast cancer.  There was also a paper in Nature and I am going to actually present data from both of those.


     So, the question that these investigators are asking is can gene expression profiling be used to improve prediction of clinical outcome as it relates to breast cancer?  The specific aims were to identify patients at risk to develop distant metastases and to accurately select for adjuvant therapy--who to treat; can we determine better who to treat, who not to treat or who to avoid over-treating?


     Proof of concept in a sense here is this first unsupervised cluster analysis of breast cancer expression profiles.  What we mean by "unsupervised" is we are looking at correlation between and among gene transcript changes as a function of the samples they derived from.  You can see across the top, there, we are looking at 5000 genes.  Which genes behaved similarly; which ones were not together; which ones went down together; which ones correlated, in a sense, as a function of the tumor samples.  Of course, on the left side you can see which tumors are behaving similarly as a function of the gene expression profile that they exhibit using a microarray.

     Red means up and green means down.  The transcripts are up or down and you can see that there is a pattern there.  It isn't just random noise; there is a pattern there.  In fact, there are 98 tumor samples there.  The 20 at the bottom are derived from women with germ line mutation in BRCA1; the 70 above are not.  You can clearly see that pattern in a totally unsupervised approach.


     So, you dig down deeper and you look at those 78 tumors to which the women did not have the BRCA1 mutations.  You can whittle those 5000 genes down to 70 significant prognosis genes using a standard T-statistic based approach.  This is a supervised approach, so you look at phenotypic outcome.  These are tumors that were derived from women over a period of 10 years, starting 20 years ago.  So, you had outcome on these women in terms of whether they developed distant metastases and also whether some of them died.  I will show you some of the plots of that data.  But you can clearly see patients with what we call good signature, with that red section of genes up in the right-hand corner, there, as opposed to women with a bad signature with a poor outcome.


     These are patients that are lymph node negative, the same 70 prognosis genes lymph node negative.  I am not a clinician.  I don't know a lot about the practice of medicine but I have a sister-in-law and she was diagnosed with breast cancer and she was lymph node negative and I breathed a sigh of relief.  I think a lot of people in the family breathed a sigh of relief.  What this is saying, however, is that even amongst those women that are lymph node negative, 14 of 40 of those women went on to develop metastases and they had a poor signature.  Of the ones with the good signature, only one of 33 went on to develop metastases.  So, it is looking like we have something even better than current clinical practice.


     Patients lymph node positive, sort of the flip side of the coin--women with poor signature, yes, the majority of the ones that had metastases had poor signature but there are some that had a good signature there even though they were lymph node positive, and 42 out of those 44 had good outcome.


     Again, just putting it all together and looking at the outcome, good signature; bad signature--I don't want to spend a lot of time on the study; I don't have a lot of time to spend on this study but you can see that same red/green segregation of what is going on.


     Here are the Kaplan-Meier plots, again showing that for prognosis the good signature versus poor signature outperforms current medicine.  That is just an example.  That is an example I spent a lot of time on.


     There is prostate cancer; there is brain cancer.


     Here we have GI cancer of some sort.


     We have kidney cancer.  There are a lot of different cancers these things are being applied to.  I believe Brian is going to show an example not of cancer but psoriasis where, again, you can use this approach.


     So, what is the approach?  Thematically, this is the approach.  Based on phenotype, you can visually or by some objective accepted method determine two different ends of the spectrum, green and blue, metastases, no metastases within a defined framework, and then look for differences.  Can we define differences using expression profile?


     What you find is that when you do those expression profiles, yes, you can define differences.  Not only do you see these global differences in just those two broad phenotypes, but you find subclasses within those observable phenotypes as well.  This is a process that goes on in an iterative way getting better, and better, and better at distinguishing what look like the same disease or the same people using the technology that we have now.  So, this is saying we have a very powerful technology and we can do things better.


     This is just a visual graphic that says everybody might look the same but we can clearly classify them in different categories using some of the new technologies, genomics, proteomics.


     Let's think about what we just did.  We took 5000 genes we applied in the clinical scenario and we reduced that down to 70 genes, which really did all the work for us.  That was looking at outcome over a ten-year period.

     What if we could do that same thing in toxicology?  We don't have to wait ten years.  We can treat animals with or without drug and look at differences that occur in a much faster time with a lot faster throughput, and we can go through that same process of turning 5000 genes down to what are the critical genes that actually serve as the business end to help us make these kinds of decisions or classifiers.  I hate to use the term prediction at this point in time.  Drug developers have to predict; clinicians have to predict but toxicology reviewers I think classify and integrate data.


     This is shown on a slide where we are looking at signature projection scores.  It allows accurate clustering of compounds and allows identification of subparts of a compound's effect.  One more point, 5000 genes, whittled down to 70 left 4930 genes that didn't contribute a whole lot there.  I think that is where a lot of angst comes when we start talking about toxicology.  There are a lot of genes that are changing and a lot of signals there, and we have to do a lot of good science to figure out where the signals are that we need to pay attention to and what we shouldn't get into an uproar about.

     But we have the capability of doing that to a certain extent already, and I think this slide from our collaborators and Iconix makes this point very clearly.  This class of compounds, statins--you can see all the statins along the bottom line.  Each one of those squares does not represent one gene but represents a signature of 6-20 genes and they have an overlapping redundancy, some overlapping genes which go into that signature and some which that are not totally overlapping.  So it is like that 70 gene score, only now we have different sets and we are talking about 6-20.  But you can see the consistency with the statins.

     Here we are looking at pharmacology.  The statins are doing something desirable.  If you put another compound in there and it performs like a statin you should see that signature come up and that tells you something about the pharmacology.  So, we have proof of concept established here.  With the NSAIDs you can see performing signatures there.  Clearly, the dose defines the toxin.  You can push these things.  That is the purpose of a toxicology experiment, you have to push the toxicology.  We won't be happy in the FDA unless you tell us what the toxicity is.  With the statins, if you push them high enough, you also see a hepatotoxicity signature.

     You can see an estrogen signal over there, and then you can see with DNA cross-linkers carboplatin, cisplatin and you can see the cross-linkers significant there.  So, the signatures are relating expression changes to pharmacology, toxicology, pathology, clinical chemistry, the chemical structure of the compound and the biology.

     How can we do this?  This is somewhat simplistic but it makes the point how can we come up with a signature of a statin?  Here we are showing the biochemical pathway for how a statin works.  Statin works pharmacologically because it inhibits H and G CoA reductase.  We learned in biochemistry that there are feedback loops, substrates, downstream factors that will regulate a biochemical pathway.  If you inhibit H and G CoA reductase all those enzymes downstream don't see the same substrate concentration.  They are trying to drive and up-regulate expression of genes to compensate for that.  So, we are seeing an attempt on the part of the animal, a compensatory response, to compensate for the inhibition of H and G CoA reductase.

     So, you see a pattern.  all those red spots indicate that with simvastatin all those genes went up, and it makes sense.  It makes biochemical sense that is telling you there is some pharmacology going on.  And, it wasn't just one transcript.  There was a combination of transcripts which gave you the confidence in what you are seeing here.


     Here, making the point, are a lot of different platforms and there are a lot of different steps you have to go through to get knowledge out of that whole process, that data generation process.  You have to worry about the RNA.  Sampling integrity is critical.  There are a whole lot of enzymatic processing steps--amplification, labeling, array hybridization, washing, array scanning, a lot of steps in the process there.

     You turn that data into information through statistical processes, data transformation.  There is normalization, scaling and, again, statistical processes you apply to select the key transcripts.

     Finally, there is the new and exciting, burgeoning and very much in need of field of bioinformatics.  If someone can put bioinformatics behind their name I think they are on a road to success.  People are hungry for these people to explain all this information, to tell you about all these cellular processes, biochemical pathway analyses referencing previous experiences.  Database comparisons enter into that and there is no substitute for just downright good homework, getting into the literature and trying to figure out what is going on.  Knowledge of biological truth is what we all want here, and this is where the decision-making level is in terms of the regulator.  This is where it has to occur.


     So, what concerns and initiatives have been raised about that triad, drug development, regulatory oversight, patient care?  At the technical level, I have alluded to this already, with so many variables and options in data capture--I didn't get into the details about the probes; probes can be at different ends of the genes on these different platforms but it is an issue.  RNA sample processing, hybridization processing, data analysis--what is the true signal?  Are we measuring true biology reproducibly and accurately?  Or, is there too much systematic experimental error to reasonably contend with?  Are the data misleading us?  Can we get the same true answer from data sets on different platforms and different laboratories?  Can universal reference standards help us?

     These are all concerns that were raised in the May workshop that we had.  It was an open public workshop.  What is a reasonably detailed and practically useful relational data set to constitute a regulatory submission that defeats healthy skepticism concerning the integrity of the data?  We need to define that.  Will these data be appropriate, ever be appropriate for an FDA database to draw from and develop institutional knowledge against?


     What about the category of biological concerns and issues?  How would the agency react if an oncogene was activated?  I am finally going to answer that question for you.  I am going to have fun with this presentation; I am really enjoying this.

     Would the sponsor have to notify the FDA, clinical investigators and IRBs?  This is a question that was posed to us in May.  To be fair, that question got some answers and all those answers weren't all that pleasant for drug sponsors to hear at this workshop in May, and it raised a lot of concerns and we have to hit these concerns head-on and that is what we are going to try to do here, or at least get started with that today.

     Will they be--and I put this in quotes--"more sensitive" because I am not really convinced yet that they are more sensitive; they are earlier clearly but are they more sensitive in terms of dose response?  Will "more sensitive" gene expression changes drive drastically lower clinical trial starting doses and prolong Phase I clinical trials?

     Which expression alterations--and this is a difficult one and, again, this is a critical biological question--are reliable and biologically relevant classifiers or biomarkers of--and I have to separate these out--which are desirable effects, which are undesirable but clearly tolerable, which are healthy and fully compensatory responses to the exposure that we can adapt to and not be concerned about, and which are the special category of intolerable drug actions that are going to lead to irreversible outcomes?  How are they all biologically relevant?


     I think Janet has a very good plan and she is going to get this out to you today to get some feedback from you in terms of some concerns and issues raised that we can maybe deal with procedurally.  Would all these tasks have to be done under GLP conditions?  Would they all be interpreted by FDA as relevant to human safety?  Is a research information package approach feasible?  What data are appropriate for that mechanism, what data are inappropriate?  These are extremely important questions that we need to resolve.

     How will the FDA prepare itself to work with these huge data sets in a timely manner?  That is an internal process that we have to deal with and we are dealing with that, talking about that.  How are we going to ensure that individual reviewers do not "prematurely" interpret, generate their own hypotheses, or over-interpret those expression alterations whose biological significance have not been scientifically established?

     How will the FDA communicate which expression alterations have reached a scientifically mature level of understanding and can rationally be considered relevant safety biomarkers?


     Here is the sort of traditional drug development process.  You think you understand something about disease, the biology of disease and you feel you have a validated target.  You put a compound there and you see if you can alter that target, if your drug is doing what you wanted to target.  Then, you develop a lead compound to inhibit that process and you come up with a better and better lead compound.

     Well, there is a line sort of between that discovery process which Janet alluded to and then the process which FDA really focuses in on.  But in my own discussions with individuals in drug development and industry, I think it is not always clear now that some of the stuff they do in those early phases of development do not need to be shared with the FDA.  That needs to be clarified.  There are some grey areas in there.  Is it a no line; is it a dotted line; or is it a solid line here?  I think that is something that we need to work out and be very transparent and clear to the industry about because it doesn't quite seem to be a totally level playing field.


     This kind of captures in a sense if that dotted line appears between that solid green box--if there isn't a solid line here, then why not allow a lot of imprecision?  Why not allow a lot of hypothesis generating?  Maybe the platform doesn't need to be very validated.  And, let the scientific researcher generate the hypothesis, look at the consistency and keep things wide open.  As opposed to, as you progress the concept here being this is what we envision, that there is going to be more and more fine tuning and more of a demand eventually on the precision and maturity of the signals and interpretability of the biology as you move closer and closer to the clinic.


     What is CDER doing to address the issues?  First, building the internal capacity infrastructure.  One thing we did early on is establish a core expertise within the CDER laboratory.  I like to think of that as an effective process.  Having that core expertise within the CDER laboratory, we have contaminated the rest of the Center and we are bringing our knowledge, our experience and our skills and getting more people excited about establishing committees.  You know, we have a neat toy.  We had some money and with that money we are able to buy a cDNA array scanner.  We got some cDNA arrays and just went to town, and through a beautiful collaboration with International Life Sciences Institute we were able to get up to speed very quickly and to be able to compare and learn a lot through that process.

     What did this do?  This enabled us in a sense to sort of establish credibility.  We were able to leverage the Affymetrix GeneChip system and we have a wonderful research agreement with them.  We are working very closely with them on a more profound biological issue, and also working with Rosetta on a research agreement so we can look across platforms with the same samples and try to understand the biology.

     We were also very fortunate, and I want to thank the Office of the Commissioner, we have an Office of Science grant from the Office of the Commissioner and we are using that to develop consensus toward RNA standards development initiative.  This involves all the medical centers within the FDA in a collaborative mode and involves the NCTR as well, to look at can you establish a benchmark data output across platforms using, for example, a mixed tissue standard.  We are focusing on the rat as the fundamental toxicology model.  We had a tremendous workshop, working through NIST, to bring this out and to get more collaboration.  NIH is going to join us as well.

     Expanding core expertise for CDER review enhancement--we have established a nonclinical pharmacogenomics subcommittee.  The focus there is on regulatory decision-making practices, procedures and policies as nonclinical expression data come to us.  the database I referred to through Iconix, the DrugMatrix database, leveraged that.  Again, a concept the previous commissioner brought to us and pointed to this as a way we are going to be operating in the future.  But if you think of it, you know, how are we going to get up to speed very quickly?

     This database all of a sudden gave us access to 550 pharmaceuticals and how they perform.  These are things in the PDR so we can turn the clock of time in a sense and say what if we had this capability 10, 15, 20 years ago and 550 pharmaceuticals which are now in the PDR?  What would we have done if we had seen that microarray data?  So, we have something to compare it to.  This is stuff we have experience with that is being used, and we feel comfortable with these things.  We can ask about oncogenes.


     Very briefly, the point I want to make here is that the same sample was run at two different sites on the same platform, Affymetrix, FDA--blue is Affy.  They are showing that you don't always get the exact same genes identified as statistically significant.  But what this is saying is, okay, if you don't get the exact same ones sort of surfacing above a p value, what happened on the other platform when it was significant on one platform?  Well, this shows you there is tremendous correlation.  It behaved pretty much the same way, p less than 0.001; 8000 transcripts; 0.001 times 8000 may be 8-10-fold.  Over here, you see in quadrant 2 and 4 there are about 8 or 10 spots--amazing statistics that kind of play out there.


     If you lower the p value, 0.01, you have a lot more genes to work with if you are in the hypothesis-generating mode.  Look at all the genes in quadrant one and three.  You have more in two and four "false positives" maybe to contend with but if you are in the hypothesis-generating mode why not go down to this level?


     You don't have this in your slides, sorry.  This is saying another story, and that is if you use a slightly different statistical approach to identify the genes of interest--Affy has a unique thing, increased/decreased calls where you look at exact probes across results from controls to treated animals and I don't want to go into details of that but look at this correlation.  It is amazing; a much better correlation.  So, the statistics that you use to identify those transcripts are equally important in terms of how you define, and this is something we need to work together on.  Each platform sort of has their own approach to identifying transcripts of interest and this is something we need to work toward together.


     Here is the answer to this question.  What if a reviewer sees an increased expression of oncogenes in a product submission package?  We used the Iconix database to ask this very simple question.  You have 72 genes on the left-hand side.  They all have the term "oncogene" in them.  We are really talking about proto-oncogenes here; these are not mutated genes.  They have oncogenic potential if they are mutated.  The list of oncogenes is growing every day.  Nature came out with an issue a month ago with, you know, this list of what could be oncogenic if it is mutated.  It is going on and on.

     But here you can see diphenhydramine clustering.  We have 9 known human and rodent genotoxic carcinogens; 5 human and rodent non-genotoxic carcinogens.  There are 14 carcinogens proven to be carcinogenic in man and we have 14 non-carcinogens, at least tested in rodent 2-year bioassay.  Do we see oncogenes going up with these rodent non-carcinogens, and do they cluster themselves all the way from all the carcinogens?  There is no pattern here.  There is no clustering of the non-carcinogens away from the carcinogens when you just look at oncogenes.

     You can see mannitol and up-regulation of oncogenes there.  You have aspirin.  You have diphenhydramine.  Oncogenes are going up.  This is a message more to our reviewers than anything else.  You see an oncogene go up, no single transcript is going to derail and identify a safety issue.


     I mentioned that efforts have been initiated by all the medically related product centers, along with NCTR, along with NIEHS and several microarray stakeholders, Affymetrix, Rosetta using Agilent-based platform, Iconix using the Amersham platform to develop standards useful for evaluating platform features and, as well, experimental performance if you want proficiency of the end user.  Ultimately, bottom line, biologic conclusions that are independent of platform and represent biological truth is what we want to help these standards do for us.  That project is ongoing.


     I mentioned we have a PTCC nonclinical pharmacogenomic subcommittee and you can see here what the various goals are of that group.  We want to effect an appropriate infrastructure for pharmacogenomic data review and integration.  That is our initial and primary goal right now.

     A question came up during one of the discussions, what are we going to do to explain to sponsors how we want the data submitted to us?  That was a question that was posed to Larry.  This is our first goal, to get that out and open to them if they choose to submit microarray data right now, at this point in time, if they choose to generate it.  If they choose to generate it and want to submit it to us, how should they submit that data?  And, this is what we want to be transparent with them about.


     What else are we doing?  Establishing a network to assimilate reasonable consensus; develop mechanisms to communicate and deliver our needs.  I mentioned working with NIST to develop RNA standards.  A workshop has been held; an initiative has begun.

     We have a number of collaborations with NCTR, a tremendous resource to the agency down there.  FDA inter-center communications group, white paper.  There is a National Research Council committee, a government liaison group, again, to explain to them what our needs are.  This is really something that NIEHS has thrust and tried to develop to get input from a lot of different regulatory centers in terms of how they can help us as well, and we are involved with that.  MIAME-tox, minimal information about a microarray experiment as related to tox, specifically working with MAHS on that and also with ILSI.  There have been FDA-PhRMA workshops.

     In sort of a non-collaborative mode, engaging external expertise.  On June 10 we are going to convene that group for the first time that we are going to ask to help us with what should the data submission be that we receive from sponsors.  What is workable?  What is reasonable?  What is compatible with current practice within the drug development industry that doesn't add a lot of burden and can work with the current systems that are currently being used?  Then, what is reasonable for experts in the field in terms of how broad does the data have to be?  Do we have to see images?  Can we have scaled normalized data?  So, these are all things we need to work out.

     We are going to practice a mock data submission.  We need to develop our prowess at this.  We can ask and maybe be careful what you ask for, but we are going to see how this whole process works and try to work out some of the kinks.


     I am doing pretty well.  I am only three or four minutes over and I am at the end.  CDER shares the vision that applications of pharmacogenomic technologies will improve recognition, understanding and assessment of pharmaceuticals by recognition; easier, quicker, more accurate identification of efficacy, efficacy potential toxicities and toxicity potentials.  By understanding, we mean modes or mechanisms of drug actions; by assessment, the significance and relevance of the findings to humans.


     CDER has a responsibility to enable, not force nor impede evolution, vision, desired state, preferred future, whatever the best term is there.  So, if we think about the drug development process and where it interfaces with FDA, lead compounds going into analytical trials and perhaps animal trials can be reduced and shortened if we work together and we get to the crux of the biological value of this approach.  Perhaps our clinical trials will have a lot more options to pursue and we will have potentially shorter, more focused kind of trials with clearer indication of efficacy going into them, and potentially more product approvals.

     Just one last quote struck me and I am going to change it a little bit.  This is a quote from a very brief piece that appeared in Nature Genetics.  The individual is Chris van Ingen:  As the impact of new technology on the quality of human life increases, so too do the responsibilities of appropriate and reasonable regulation.  I think we have a good plan to work with people on that and I hope you give Janet a lot of good feedback.  Thanks.

     DR. DOYLE:  Thank you very much, Dr. Sistare.  Next we are going to bring Dr. Larry Lesko back and he is going to address Pharmacogenomics from the perspective of drug metabolism and dosage.

Pharmacogenomics--Drug Metabolism/Dosage

     DR. LESKO:  Thank you and hello again.


     As you see in the binder that you have there, I have more slides than Frank had but I promise not to show them all.  I also promise not to show any microarray data.

     What I will talk about, however, is pharmacogenetics, which many people think is a subset of pharmacogenomics.  I will be speaking about it from a fairly narrow perspective, that of drug metabolism and dosage.

     As you have heard already today, there has been a paradox in the modern drug development process.  Clinical trials provide evidence of efficacy and safety at usual doses in populations.  On the other hand, physicians treat individual patients who can vary widely in their response to drug therapy at the usual dose.

     Well, in clinical pharmacology we try to take care of this paradox in part by looking at subgroups of patients defined by demographics, disease or other types of factors.  While that is not quite dealing with the individual patient, it does deal with individual subgroups of patients with respect to drug dosing and drug therapy.


     You will see in any discussion of pharmacogenomics many definitions.  This is the practical definition that I will provide for our discussion: systemic genomic analysis in populations of treated subjects to identify variants that predict drug response including the occurrence of adverse events.


     Leading from that definition are three broad areas of application.  You will hear more about this, but drug discovery is obviously an area where genomics can be used to better understand the cause of disease and, therefore, identify new drug targets.  On the far right is drug selection.  Drug selection results from using genomics to differentially diagnose something that we call hypertension or diabetes and then align a specific drug in a class with that subtype of the disease.

     I am not speaking about either one of those.  I am speaking about the middle channel, talking about drug therapy and how drug therapy and the knowledge of genomics can be used to tailor the dosing to an individual or to a population.


     This is a cartoon that illustrates the pharmacogenomic strategy as applied to practice of medicine.  It is a bit of a simplistic presentation because it takes all patients with the same diagnosis.  Based on a single nucleopolymorphism, or SNP, the idea is to remove non-responders or toxic responders in advance and then, based on other SNPs identify responders or patients not predisposed to toxicity whom we can treat.  It is idealistic in that it treats non-responders and responders as two different groups when, in practice, it has actually turned out that it is more probabilistic than that and there is probably an overlap in terms of response between the responders and the non-responders.


     In any case, in terms of pharmacogenetics we know already that there is a pharmacokinetic basis for differences in drug response that define those two groups I just showed you.  A way of categorizing those differences is based on extrinsic factors such as the environment, smoking, diet, alcohol, drug interactions Rx, OTC and herbal products taken together.  Also intrinsic factors and intrinsic factors can be demographics; could be disease states; or the subject of our discussion today, pharmacogenetics.

     An analogy would be just like we define subgroups of patients based on renal disease or hepatic disease which is phenotype, the thought is that we can also define subgroups of patients based on their phenotype by looking at polymorphisms in the genes that encode metabolic enzymes.


     So, I am going to change that cartoon a little bit and talk about the pharmacogenomic strategy applied to drug metabolism and dosing.  Notice, on this slide I don't have responders and non-responders.  Rather, I have a group, group A, that has a genetic profile for toxicity with the usual dose of the drug.  It isn't indicated that this group not receive the drug, but it is indicated that this group receive the drug but perhaps at a lower dose.  That is the implication of what I am referring to in pharmacogenetics and drug metabolism.

     As you can see, groups 2 and 3 continue with the theme of responders but link the responders more to the drug selection process than to identifying a drug target.  So, I am really operating under conditions of number 1 in the presentation that I am making today.


     This illustrates what I am referring to when we talk about pharmacogenetics and drug metabolism.  This is the case where I have the usual dose given to a population of patients, but those patients between them have different plasma concentrations.  They may have those different plasma concentrations for many reasons, as I indicated, environmental or intrinsic factors.  But I show on this slide patient A with a certain wild type sequence in their DNA that encodes for the activity of the enzyme.  Patient B has a polymorphism that creates a mutation in the variant.

     The net result, if you go to the far right, is a plasma concentration time profile in the wild type patient that is right within the boundaries of that therapeutic range.  When you go below that, because of the slower metabolism imparted by this polymorphism, the exposure goes above the therapeutic range and that patient, or patient subgroup, is predisposed to toxicity.


     We know from history, probably twenty years of history, that many of the cytochrome enzymes are, in fact, polymorphic.  I could pick several of them but let me use as an example one of the most popular, and that is cytochrome 2D6.  On the left I show sort of the portfolio of cytochrome enzymes.  I could pick for this example 2C9, 2C19 or many others that are polymorphic, but I picked on 2D6 to illustrate the example because this is an important enzyme in terms of prescription drugs.  It is responsible for metabolism of about 40 percent of prescription drugs.  That means over 300 million prescriptions for drugs with polymorphism in the 2D6 per year.

     The example on the bottom represents the terminology in the cytochrome enzyme literature.  I have a family called CYP2.  I have a subfamily called 2D6, and then I have a gene defined by, in this case, a deletion at 5 2D6*3.  If you happen to have that 2D6*3 you are going to be characterized as a phenotype called poor metabolizer.


     What are the clinical implications of that?  Well, this shows a theoretical but somewhat real dose-response curve for nortriptyline, and on the left-hand side I have indicated the therapeutic window which is somewhat well defined.  You can see that there is an overlap between the exposure response curve for therapeutic effect.  That would be the therapeutic response curve.  There is also an overlapping curve for toxicity and that window isn't very wide.  It is about three-fold in practice and, in fact, nortriptyline levels are monitored in therapeutic drug monitoring labs quite often.

     If you think about nortriptyline and the implications of having a wide range of exposure from the usual dose, you may be moving into that efficacy world of reduction in anxiety and symptoms of depression, or you may be moving into that safety world of tachycardia, arrhythmias and drowsiness.  That is all possible from the usual dose without taking into account metabolic patterns of a patient.


     This shows a little bit more.  On the left are illustrated nortriptyline plasma levels as a function of frequency.  It shows the bell-shaped curve.  The majority of patients are called intermediate metabolizers.  Their plasma levels are somewhere around 60.  The poor metabolizers may be on of those 2D6*3 on the right hand side and you can see their exposure is significantly higher.  Then you have people with extra activity that are on extensive metabolizers side who have lower blood levels and, in fact, may require a lower dose to achieve a therapeutic response.  Translated into functional dosing, on the right, the dose in the PDR for nortriptyline is 25-300 mg.  You can see the wide range and that wide range is a result of the wide range of requirements that patients have based on their metabolism.

     But you could ask what is the equivalent dose if I were to achieve the same exposure in all patients or the patients who are a mix of poor metabolizers, intermediate and extensive?  That is what the graph on the right shows.  In order to achieve equivalent exposure I would have to give, for example, to the extensive metabolizer 120 mg.  To get the same systemic exposure the dose would have to be reduced in the poor metabolizer to 25 mg.  That is the implication that we have for therapeutics, the wide dose requirements defined by the genotype of the patient.


     Well, if this is a wonderful part of science, one might ask the question what has been accomplished over the last twenty years with the integration of this at the bedside.  I did a search of the electronic 2003 version of the PDR which had 2000 entries and identified 51 labels that contain pharmacogenomic information.  That doesn't seem like a lot of advancement over the course of time in looking at the field.  In most cases, in fact, of the 51 labels the information about genomics in the label was not easily translated into clinical practice.


     Here is an example of one label that came up in the search.  It was thioridazine, Mellaril.  In that label I thought the information was fairly informative in terms of communicating the risk.  In the contraindication section of the label thioridazine was contraindicated in a subgroup of patients which are seven percent of the Caucasian population who are known to have a genetic defect leading to reduced levels of CYP 2D6.  In the warning section, to further provide information to the prescriber, there are certain circumstances that may increase the risk of torsade because of long QT in patients with reduced activity of 2D6.

     So, here is a factual piece of information, no different than other factual pieces of information in the label.  It does not go on to say what to do or how to identify those patients in terms of dose reduction or in terms of laboratory tests.


     A recent example that illustrates the inclusion of genomic information in a label is that of atomoxetine or Straterra, which was approved in January of 2003.  This goes a step further in terms of informative information in the label.  For example, it includes in the human PK a statement that a fraction of the population are poor metabolizers resulting in...and the label goes on to describe that.

     There is good news here.  The inhibitors of 2D6 in extensive metabolizers increase exposure, however, if you are a poor metabolizer you are not prone to that interaction.  In the adverse reaction section the following ADRs were either twice as frequent or statistically significantly more frequent in PMs compared to EMs.  Finally, in the label the laboratory tests are available to identify CYP 2D6 poor metabolizers and, in fact, this is a relatively new development in the field in terms of the widespread availability of laboratory tests to identify poor metabolizers.  So, this is probably the most informative label that we have at the moment.


     Brian is going to talk a little bit more about pharmacogenomics during drug development.  Talking to Brian and other people in the industry, we have been told that 80 percent of new clinical trials include collection of samples for DNA analyses.  We haven't been told what those analyses are or how they are going to be applied.

     This graph kind of illustrates the number of clinical trials under way.  So, if you multiply those clinical trials by the number of samples collected you can see why we have so many gene banks springing up.


     But what is going on from our experience at FDA, we undertook a survey, a very informal survey that doesn't claim to be 100 percent accurate, of the appearance of pharmacogenomics in INDs and NDAs.  We started this very recently, and you can see the time course of this survey.  Over that short period of time the number of INDs and NDAs that contain genomic information has gone up very dramatically.  It has gone from 5 to 70.  I expect this is an underestimation of the actual utilization of genomics in the drug development process as perceived from the IND/NDA standpoint, but it gives one a sense of what is going on.


     What do these 70 applications represent?  They represent predominantly genotyping of the cytochrome enzymes that are responsible for drug metabolism.  You can see the breakdown in the pie chart.  By and large, most of the genotyping is related to the enzyme I picked as the example, CYP 2D6.

     On the bottom of that slide are some bullets that represent other things that are being genotyped that are much more exploratory than the confirmatory data on the cytochrome enzymes.  For example, we are beginning to see genomics as applied to receptors in an attempt to understand differences in drug response.  This is very preliminary.  It is very exciting to see it but it is not in the confirmatory world like the metabolite enzymes are in.  We just don't understand that as much.  So, there is a lot of activity going on, as evidenced by the trend in the data that we are seeing.


     On a lot of people's minds is the issue of drug safety and the fact that adverse events are a major public health issue, as well as a major pharmacoeconomic issue.  So, the question is logical, how does this relate to pharmacogenetics?

     We have asked the same question and we were piqued by the article that appeared recently in the literature that really linked several databases but provided us with some interesting information that the top 27 drugs frequently cited in adverse drug reaction reports, including nortriptyline, 59 percent of those drugs are metabolized by at least one enzyme having poor metabolizer genotype; 38 percent of those drugs in the top 27 drugs are metabolized by 2D6, mainly cardiovascular and so on.  So, this is inconclusive but it is circumstantial evidence that polymorphism in drug metabolism matters when it comes to drug safety.


     It raised an issue for us about how pharmacogenetics can improve existing therapy.  The emphasis, by and large, when people talk about genomics is the future.  How is this going to help us develop better drugs, more targeted drugs and reduce adverse effects?

     According to Dr. Collins, at NIH, mainstream genomics is still eight years, ten years away.  But I think we are on the verge of possibly doing something sooner that can benefit public health.  Thus, I ask this question, how can pharmacogenetics improve existing therapies, which I will define as all medicines that have been approved by the FDA for prevention or treatment of any disease in humans under patent or not?


     As an example of one case study we can look at thiopurines and more specifically 6MP.  This is a drug that has been approved for acute lymphocytic leukemia mainstay therapy for over 50 years, and clinical studies over the last 20 years have focused on various phenotypical as well as genotypical methods of refining optimal dose.

     To look over the history of 6MP to the present time, there have been incremental advances in using pharmacogenetics.  Efficacy is fairly high for an oncolytic application.  Adverse effects are also associated with that.  In the case of 6MP, the risk of myelosuppression is the primary limitation to drug dosing.  Blood counts have been traditionally used to monitor therapy, however, there is evidence to suggest that pharmacogenetic testing can significantly reduce, although certainly not eliminate, this risk.


     In addition to its approved use, there are many off-label uses of 6MP, inflammatory bowel disease, various autoimmune disorders.  If one were to look at the prescription use of this drug, it wouldn't be surprising if the use off-label exceeds the use on-label.  Nevertheless, it is a mainstay of therapeutics in today's practice of medicine.


     What we know about the genetic basis of 6MP is fairly well established.  This is not exploratory but, rather, confirmatory that a certain enzyme TPMT, thiopurine methyl transferase, catalyzes the inactivation of the compound.  If you happen to be a deficient patient you accumulate excess of thioguanine nucleotides in your hematopoietic tissues and this accumulation in RBCs leads to an increased risk in all patients of severe and possibly fatal myelosuppression.


     Let's look at integration of genomics via this flowchart.  What this flowchart shows is the prevalence of genotypes for TPMT in the population.  I start at the top with all newly diagnosed ALL patients per year.  That is 30,000 and 3000 of those are pediatric; the rest are adult.  By and large, that population can be subdivided on their genotype into three broad categories.  On the far right is the wild type.  Most patients are of the high activity type.  In the middle is the heterozygote deficient, intermediate activity, and on the left, homozygote deficient, the one that we probably would be most concerned about.  That individual would have two mutant alleles.

     It is further well-known that three major SNPs, or single gene polymorphisms, define these mutant alleles.  It is quite common to find in the population *3A, 3C and *2.  I have shown in parentheses the prevalence of these alleles.  There is a rare *3B, 120,000.  There is also one more that I don't have on there, *3D but it usually travels in association with *3A so one doesn't have to measure it specifically.  If you are in that at-risk category on the homozygotes, low activity, the prevalence is 1/300; for the heterozygotes it is about 11 percent of the population.


     In order to think about a genetic test to guide therapy and guide dosing, one has to look at clinical utility.  We have tried to do this by looking in the literature at the incidence of toxicity.  What this graph shows on the left-hand side is the cumulative incidence of dose reduction due to toxicity in a patient population receiving the usual doses of 6MP.

     Notice in the homozygote deficient, the v/v patients, 100 percent of the patients need a dose reduction because of toxicity because they can't metabolize the drug.  The middle category, heterozygotes need a dose reduction in 35 percent of the cases because of toxicity.  As one would expect, the inherent toxicity of the drug in the high metabolizers is around 7 percent.

     What is very interesting though, if you move to the right-hand side, is you can see the different dose requirements that these patients require.  These are the average final weekly 6MP doses.  If you are a wild type allele, on the far right, the dose is pretty standard.  However, if you are on the far left, homozygote deficient, you can see the dramatic reduction in dose necessary to avoid toxicity.

     The problem in dealing with this drug dosing of 6MP is that you have to interrupt therapy based on the toxicity, and interruption of therapy lessens the intensity of treatment which is very important in the ALL patients.  Furthermore, if one has multiple drugs in the regimen, as is typical for an ALL patient, reduction of the 6MP dose allows for full dosages of the other chemotherapeutic agents.  In fact, many of the drugs in a regimen for an ALL patient have overlapping toxicities so for a physician to figure out which of the drugs is causing toxicity a test can assist that.


     In order to construct the framework of how to think about the integration of a genomic test into existing therapies, I have borrowed some of the guidelines from the Secretary's advisory committee on genetic testing from their report.  I think these are very reasonable questions to ask.  Is the phenotype relatively common?  Is the impact of the phenotype serious?  Does early detection of a genotype alter therapy, in this case drug dose?  Are there accurate, reliable tests available and, once those tests are performed, is there counseling available for the physician or the patient as necessary?

     In going through this questions one by one as they apply to 6MP or as they apply to a drug metabolized by 2D6, I would hope we could come up with a rational answer as to how to move forward in improving existing therapies using pharmacogenomics.


     You heard about many working groups within FDA, well, here is yet another one.  We are going to have a summit one of these days to talk to each other about working groups, but this is the one I have been chairing since June of 2001.  This group had, as you can see, a cross-section of individuals from all of our centers.  It also has a cross-section of disciplines represented on the group.


     Our goals are listed on this slide.  We came together originally in 2001 to organize a public workshop to at least get on the table a discussion with the industry and others about the issues of pharmacogenomics.  That occurred in May, 2002 and the workshop report was released this month in The Journal of Clinical Pharmacology, in the April, 2003 edition.  It makes for some interesting reading, if you haven't seen it.

     We also used the working group to make presentations at public meetings, primarily professional associations, to begin to provide some regulatory perspective, which audiences generally want, and to begin to engage audiences from academia and industry in discussions of issues.

     We are developing a draft guidance for industry on pharmacogenomics that will emphasize primarily the clinical side of pharmacogenomics, the efficacy and safety trials.  It will discuss statistical issues.  It will discuss labeling recommendations and it will discuss in part some of the issues revolving around diagnostic tests and nonclinical pharmacogenomics.

     What it will eventually discuss at the end of the day depends in part upon coordinating the activities of this working group with the others, and we are still some time away from that guidance.

     We have also used this working group as a forum to discuss the scientific details of a submission requirement for PG data that Dr. Woodcock will talk about in the next set of remarks that she is going to make.  We have tried to develop a number of case studies that would represent data that would be typically in a standard review stream, versus that data that would perhaps be outside that review stream in terms of submission to the FDA.

     With that, I will conclude and turn it back over to the Chairman, Dr. Doyle.  Thank you.

     DR. DOYLE:  Thank you very much, Dr. Lesko.  We will move on.  Next we are going to hear from Dr. Brian Spear who is Director of Pharmacogenomics, Global Pharmaceutical Research and Development, at Abbott Laboratories.  Dr. Spear is going to share with us a perspective of the industry's use of pharmacogenomics and regulatory issues.

Industry Use of Pharmacogenomics and

Regulatory Issues

     DR. SPEAR:  Thank you very much.


     I appreciate the opportunity to participate today and to put forward an industry perspective on the use of pharmacogenomics in drug development.  Now, what is this industry perspective that I have?  Good question.  Part of it comes from my experience at Abbott Laboratories where I am responsible for pharmacogenomics, including pharmacogenetics and cellular molecular toxicology.  It also comes from active interaction with a number of people in the industry, primarily through either formal or informal groups such as the industry pharmacogenetics working group which has been working closely with FDA, for instance on last May's workshop, through PhRMA, through FPIA which is the European equivalent, and other groups that involve industry, regulators and academics who are focusing on scientific and, to some degree, regulatory issues related to pharmacogenetics.

     But the views that I am going to put forward are my own.  I think that they correspond with those, and I have had feedback saying they correspond with those of other people in the industry but they are still my own and I will take full responsibility for them.

     I should say that within the industry the participation among the major pharmaceutical companies is 100 percent.  All of the companies are involved in pharmacogenomics.  This is not something that a company here and a company there have taken a chance on.  This is now a standard part of the drug discovery and development process in every one of the drug discovery and research companies.


     What I would like to do is two things.  One is to just go over, in very brief detail and by example, some of what the major activities are within the drug industry in pharmacogenomics to give you a sense of where we are putting our resources and what we think is important.  Then I would like to raise some issues that are important to us in terms of how our work is done, how the results are going to be used, and what the regulatory consequences are to try to lay a bit of groundwork of where we would like to see the interaction between the industry and the FDA.


     The primary uses of pharmacogenomics within industry relate to clinical trial results, to data quality, to study design, to biomarkers.  Some of the results here will not be reported; some of the results will.  Some of them will eventually go on to drug labeling.  Some of them may lead to specific labeling to direct a genetically defined group for therapy.  However, targeting drugs at genetically defined groups, that is a drug strategy where you are going to make a special drug for a special genetic group, is not a primary focus in industry.  You hear this from time to time.  You seldom hear it from industry.  Rather, we are still engaged in conventional drug development but now we have a new set of tools that we can use to direct the drug development or interpret the results in new ways.

     What I would like to do is to describe three areas which seem to involve the greatest degree of effort within the pharmaceutical industry: clinical genotyping, pharmacogenetics, preclinical gene expression and clinical gene expression.


     First of all, clinical pharmacogenetics--I am not going to go into a lot of detail, but there are areas where understanding patients' phenotypes can help considerably in interpreting or designing Phase I studies.  First of all, there are outliers that appear in Phase I studies and when you are looking at 6, 8, 12 patients one outlier can have a considerable effect.  In some instances, understanding patient genetics can help you understand the pharmacokinetics much better, and I will show an example in just a moment.

     Secondly, you can use this to exclude certain patients or you can use it to include certain patients, depending on what you are trying to tease out in a particular Phase I study on healthy volunteers.

     You can use it to normalize genotypes.  What do I mean by that?  Well, as Dr. Lesko pointed out, poor metabolizers of 2D6 appear somewhere between six and eight percent of the American population.  In a small trial you would hope that you will have genotype frequencies which represent that so that you are not, because of small sample size, missing the boat completely or overloading on the wrong patients.

     Finally, bridging the other populations.  We have had no discussion so far about ethnic or racial differences in allele frequencies.  But some racial groups or ethnic groups have high frequency of poor metabolizers for enzymes, low frequency of poor metabolizers, high frequency of hypermetabolizers.  Because of that, if you have not allowed for these frequencies in your trials, you may find it difficult to translate data, let's say, from the United States to Japan or China, or vice versa.  By normalizing the allele frequencies to account for these differences in different national populations, you may make it possible to bridge much more easily.


     Let me just show you an example of a Phase I trial.  This is a study with investigational drug.  We are looking at the effect of the investigational drug on the pharmacokinetics of desipramine.  This is a fairly straightforward trial that one would run.  But in this particular case, because desipramine is metabolized almost entirely by 2D6, as is nortriptyline, patients who are poor metabolizers do not metabolize at a rate which will appear normal in your Phase I trial.

     What we did in this case is we genotyped people ahead of time.  We removed those who were poor metabolizers.  I think there were 3 in a population of 36.  Then we carried out the interaction study only on those who had active genes to metabolize desipramine.

     There are two major reasons to do that.  First of all, patients who are poor metabolizers do not typically tolerate desipramine.  So, they would be at risk for injury, especially at the dose levels in the trial protocol, and they would probably drop out.  If we did get PK data from those patients, since they don't metabolize it, it is going to be confounding whether you are looking at the genotype effect or the drug-drug interaction effect.  Therefore, we excluded those patients.  It is a safer trial and it is a trial that is going to have a cleaner outcome.

     But when you look at the results you can see that there is still an outlier.  If you look at the rate of elimination you have one with a long half-time.  It is being metabolized slowly.  Why is that?  Here genotyping helped us again.  This individual turned out to have the unusual phenotype *6, which is one of those nonfunctional null alleles, and then *9, which is a partially functional mutation.  This is an extremely rare genotype.  Nevertheless, we just happened to get one.

     So, you can see that most of the patients cluster with a consistent, readily interpretable PK, and then we have one outlier.  We have interpretation of that outlier and we can go on to say that, due to studies we have done in our ethnic diversity panel, and published figures, to say that this *6/*9 occurs about 1/250 in the population.  We can account for it.  We have the mechanism.  Therefore, we don't have to concern ourselves significantly with that anymore.


     What about pharmacokinetics in Phase II and Phase III studies?  First of all, I am going to use pharmacogenetics here not quite in Dr. Woodcock's way but we are using DNA sequence variants as they relate to drug response.  There are genetically identifiable groups whose disease can be differentiated by genotype.  These may be more rapid progressors.  They may have more pronounced disease.  These are very good populations that can be used to carry out targeted small clinical trials, as has been done with celecoxib with patients with colon cancer.

     Secondly, we can include or we can exclude patients from these trials, include if we would like to carry out a proof of concept study where we would like to be able to demonstrate that it really does what it is supposed to do in a likely population; exclude if we are concerned about at-risk populations and we want to prove the concept before we go and expose those patients.  We can stratify studies according to genotype, either by clinical response or risk of adverse events.  In some cases we may wish to proceed with drug development for a genetically defined group.


     Also, in Phase II and III studies it is possible, when you have a large enough number of patients, to use those patient samples to discover new genetic markers related to disease outcomes.  This is an example I have taken from Alan Roses at GlaxoSmithKline.  These are patients, in this case, with Alzheimer's disease who have been scanned using SNP markers in the genome to determine where in the genome there is association of a genetic variant with the condition.  This is able to demonstrate the effect of APOe alleles on Alzheimer's disease.  The same approach, using large number of patients and large number of genotypes, can be used to discover genes associated with drug effect or with drug side effects.


     Going on to preclinical gene expression, there are two ways primarily that people are interested right now in preclinical gene expression.  One is toxicogenomics, which I will come back to.  Briefly, toxicogenomics is attempting to predict toxicity of candidate compounds or identify mechanisms of toxicity.  The second is to identify potential biomarkers that reflect drug toxicity or drug efficacy, and these are biomarkers that one can use later in clinical studies.


     I would like to go just a little bit more into toxicogenomics.  This is the "everything on one slide" result.  This is a very broad toxicogenomic study that was carried out between Abbott and Rosetta.  This is the same sort of heat map or clustering analysis that Dr. Sistare showed, where across the top you see 3500 different genes.  The lines there in the dendrogram indicate how similar those genes are in terms of response of rat liver to those genes.   The longer the line the more disparate they are; the shorter the line the more similarly they respond.  Down the right side are 52 different hepatotoxic chemicals or pharmaceuticals in multiple doses.  On the left you see another dendrogram which indicates how similarly they respond to each other.  At every intersection of a vertical and horizontal line there is a result, and that shows whether gene expression was increased, or decreased, or remained the same when a rat was exposed to that compound and then was analyzed for that gene.  You can see a wealth of information on this slide.  You can especially see it if you are not red/green color blind, as I am.


     But even with that, it is apparent that there are some genes that respond very similarly to each other.  There are some drugs that respond similarly to each other.  From this, you can determine patterns that are indicative of hepatotoxicity or particular types of hepatotoxicity.


     This is now taking results from one drug on that graph on one page of 28 pages of results.  This is just put up here to give you an idea of the wealth of information that comes out of one of these studies.  You can see in that magnified section that there are some genes--and I don't even know what these genes are; with 3500 of them I am not even going to try to--but you can see that there are some whose variation is at a significantly great level and some where it is not.  This is a treasure-trove of data mining.  There are any number of post hoc experiments that one could do with this data.  When we talk about the concern we have about interpretation of this data, it is this exact post hoc data mining which first comes to mind.


     Let me briefly describe some work going on in clinical gene expression now taking samples from patients in clinical trials and looking at gene expression responses.  The primary effort here is to find biomarkers that we can use to indicate drug response or to indicate early signs of drug toxicity.  It is also very useful, if you have been doing the sort of rat or mouse studies that I showed in toxicogenomics, to be able to compare human responses to the animal model responses to see if we are getting similar effects or different effects.

     We can use this to identify genes that seem to be key for particular responses, which may then become genetic markers if we find that there are polymorphisms in there.  Since each one of these genes that is expressed indicates a protein which is being expressed, this now may be an indicator that there is an important protein which has an effect in the drug response or as part of the pathogenesis of the disease.


     Here are some results from Andy Dorner, at Wyeth, looking at gene expression in patients with psoriasis who were treated with cyclosporin or IL11.  They looked at 7000 different genes--much the same sort of analysis that we looked at in rat liver.  They have access to a human sample, which is skin.  They found that there are 159 genes whose expression is related to the incidence of psoriasis.  Of those, they found 142 which respond to treatment with psoriasis.  You can see in the two graphs here, starting at 1, which is the incident level, that with these particular genes the expression level decreased when they were exposed to the drugs, and they responded well.  Whereas, patients who are not responders, whose psoriasis did not get better, as you can see on the right panel, there is no improvement in the gene expression pattern.  So, we have a very clear indicator from these genes that there is a correlation of these genes with drug response.


     Using these high density approaches, human clinical gene expression raises some challenges.  The greatest, as I already mentioned, is we generate huge amounts of data, more than can be readily interpreted.  As I said, it is open to many different types of interpretation.  The statistical methods that we are using now themselves are experimental methods.  We are still developing methods that can generate true p values, that can show true association of patterns with outcomes.

     Once we reach a conclusion, we have yet to work out good methods to confirm those conclusions.  We can do the same thing again and use the same interpretation but that doesn't necessarily confirm it.  So, these are still in process.  As I said, there are multiple interpretations.  Two people can take the same data, use different interpretation algorithms and reach different conclusions as to which genes are important.  But at the bottom of all of this is, yes, we have the human genomes sequenced.  We have a sense of how many genes there are.  We have names for many of them.  But the correlation between our knowledge of the genome and our knowledge of clinical outcomes is still rudimentary and we are not at a point yet where we can clearly say if this gene does this, the clinical outcome will be that, and that is at the bottom of much of the work that is going on and much of the uneasiness we have with people who would want to draw facile conclusions from the data.


     There are proposals being considered for ways in which this data could be submitted to the FDA so that the FDA can learn from the data, can develop methods and test methods on the data which would, at the same time, not jeopardize us to this post hoc analysis concern.  So, what I would like to do is just reflect on some of the comments that have come from people in industry to the suggested proposals and what we are looking forward to in proposals for how this could happen.

     Because of the concern about post hoc analysis, and you have heard already from Dr. Sistare, there is a reluctance by pharmaceutical companies to submit data on investigational drugs to the FDA.  In many cases that is because we are reluctant even to carry out the experiments on investigational compounds because we recognize that in IND updates it would be appropriate that you would ask for that information.  This is hampering our research in the same way it is hampering the transfer of information.

     Why is that?  Well, the same issues--we are using analytical methods that haven't been validated.  There are multiple interpretations.  The reviewers may not be trained in any way related to this data but may, nonetheless, recognize the names of some key genes.  It is possible the results could be over-interpreted; they could be misinterpreted.  Even without those problems, the sheer analysis of it could lead to lengthening of timeliness while people work through what it all means.

     The perception on our side then is that submitting this data could jeopardize a drug development program and, as a consequence, we don't want to do that.  The risk may be low but the consequences may be high.


     In considering a process by which these types of results might be exempted from normal review, and we will hear from Dr. Woodcock shortly what that proposal is, there are certain things we would be looking for.  One of them is to lower our risk.  We would like to be able to do the work without the risk that we are going to jeopardize our programs.  That is the bottom line.

     We would want data to be evaluated by experts.  We would like an evaluation that is consistent from drug to drug, and we would like to find out what those results are.  We would hope that the review would be independent of the drug review timeline.  That is, the drug review is not going to be held up while people are trying to figure out what to do with the genomic data.  We would hope to work closely with the FDA in determining what is the best process to represent the science and get the best regulatory approach, continuing the type of work that we have been doing with the FDA so far.


     There are some things that seem to be up in the air, and I think they can be worked out.  Just what would fall under this research exemption?  How would it be defined?  There will be a definition but as yet we are not sure what it is, and we hope that there is some input to that.

     There would certainly be situations, and it was mentioned at some point this morning, where there may be a public health issue that would overrule this exemption.  We would like to know what that public health issue is.  Certainly, if it is a real public health issue we would be certainly in agreement but this needs to be made clear.

     It is also not clear what sort of feedback would be, the process by which we would get information back from the FDA once they have looked at data from one company and perhaps compared it to data from others.

     There are also things that we would find unfavorable in a process like this.  One that comes up repeatedly is, well, what if we have this program for three years and then we decide to rescind it?  And, everyone is concerned about that.  But I think perhaps a more real concern is will this process then lead naturally to a request for more and more data?  If there is a process by which pharmacogenomic data may be submitted, does that lead to the assumption that pharmacogenetic data should be submitted?  That is a decision that we believe should be made based on the specifics of the drug and the drug responses, not simply because there is a process by which it can be carried out.


     This matter of submitting high density data is important.  It is one of many issues that are facing industry right now with respect to pharmacogenomics.  I would just like to bring up some of the other issues that are important to us in this field.  Of course, the biggest thing is not industry-FDA, it is nature which is not easy.  Biological systems are very complex and genetics is only part of it, and even when it is, that is not a clear yes/no answer.  This is something we are always dealing with.

     From a drug development standpoint, in carrying out a pharmacogenomic study generally you have no idea what the result is going to be.  You don't know the result and you don't know its value until you have finished.  If these get to be expensive and you are trying to convince people that you are going to carry out a study but you don't know what the result will mean, this is a natural impediment to conducting these sorts of studies.

     There is also no clear regulatory pathway for what to do with the data.  If you look at drug labels where genetics is found in drug labels, it is in all different sections from one drug to another.  So, it is not clear to people who are designing drug development programs that there is a role for genetics in all of this.  It seems to be an ad hoc thing.

     In industry, where we have significant financial constraints, a program which requires up front investment with uncertain outcome, uncertain even as to whether the results will help the drug or hurt the drug, those programs are not going to be looked on favorably when competing for resources.  So, these are built-in impediments.


     Let me just bring up some of the issues which have been surfaced by people within industry that we think are questions that we would like the industry and FDA be able to discuss more fully.

     First of all, what is the reasonable expectation of the role of genetics in drug responses?  Dr. Lesko was saying, well, it may be large; it may be small.  If the expectation is that most drugs have a genetic component, then that tells us one thing that we need to do in terms of conducting our studies.  If the assumption is that few drugs have a genetic component, that tells us something else.  It is not that I know the answer to this, although we have been trying to address it, but we need some clarification of what those expectations are, and those expectations have to be consistent with the best scientific information.

     In a practical sense of conducting clinical trials, in the hypothetical case where a drug in its pivotal studies shows a benefit and has a reasonable safety profile, such that that drug might reasonably be expected to be approved, if we had carried out a genetic study and found that maybe 30 percent of the people don't respond and, yet, the remaining 70 percent do, are we now going to have to have a restrictive label that says don't treat the 30 percent non-responders?  That is, are we going to be penalized in our label because we did the extra work?  This is a question that comes up all the time from drug development people.

     This next question I have here sounds almost trivial but it is also very important to a lot of people.  If we collect DNA, is the FDA going to expect us to do something with the DNA?  Because if the answer is, since you have the DNA, then why don't you come in and genotype something, the response is very simple, fine, we don't collect DNA.


     If we look at different genotypes in the population--let's say we have three different genetically identifiable groups, homozygous wild type, homozygous mutant, heterozygous, and we want to evaluate the drug, does that mean we now have to redesign our trial where we power it to the smallest group?  That can make for a very, very large clinical trial and that also is an impediment to carrying out that sort of study.

     We would like to know what the regulatory requirements are for a test that might accompany a drug.  Let's say we have a drug for which we want to exclude a particular genetic group, what are the regulatory requirements for that test?  Does it need to be an FDA approved in vitro diagnostic?  Can it be a home brew?  What about the tests that are used just for conducting clinical trials?  What are the regulatory requirements for those?  In order to do this we need some clarity.

     Finally, if we only test a drug in a certain genetically defined group of people, what is that going to do to the label?  Again, there are many answers.  We just need clarity on that.


     Just to end this, three points.  First of all, pharmacogenomics is now an integral part of drug development.  It is an emerging part but it is already taking place within all the drug companies.  Secondly, we have had excellent cooperation with the FDA in trying to develop the scientific basis for using these procedures and for regulating drugs.  Thirdly, the greater the clarity in FDA's expectations of pharmacogenomics, the greater will the progress be in use of these technologies.  Thank you very much.

     DR. DOYLE:  Let's take a ten-minute break.

     [Brief recess]

     DR. DOYLE:  Next we are going to hear from Dr. Woodcock, and she is going to share with us the FDA proposal on pharmacogenomics.  So, Dr. Woodcock?

Fostering Technology Development--Pharmacogenomics


     DR. WOODCOCK:  Thank you.  I think you have gotten a very good overview of some of the issues that we are facing and why we are here, before you, today to discuss how to move forward on introducing the technology into the regulatory arena and actually making this come to pass.


     Drug regulators, according to the framework that has been set up, have basically two kinds of obligations.  We have to determine if drugs are safe and effective for marketing, and if they remain safe and effectiveness.  We also have to protect human subjects who are enrolled in trials.  That is the regulatory framework that has been set up for us and we have to fit this new information into this framework.


     I am not going to talk about the PHS Act.  It is slightly different for the biologicals but the Food, Drug and Cosmetic Act defines safety for a new drug application as the FDA will evaluate reports of all methods reasonably applicable to show whether or not such drug is safe under the conditions in the proposed labeling.  That is the requirement for safety.

     Effectiveness is even more vague in a way.  We need adequate and well-controlled trials to show that the drug will have the effect it purports to have under the conditions of use.


     These obligations and framework have been translated into regulations.  For an IND the sponsors have to submit pharmacology and toxicology information, and here we are talking about the animal data, on the basis of which the sponsor has concluded that it is reasonably safe to conduct the proposed clinical investigations.


     For the new drug application for nonclinical studies, it says submit the studies that are pertinent to possible adverse events.  Those that are pertinent.  For the clinical information for a new drug application, submit data or information relevant to an evaluation of the safety and effectiveness of the drug product.  That is the legal and regulatory framework.


     The question that is before us and that we want to have a discussion on is when and how to use this developing pharmacogenomic information in regulatory decisions.  That is kind of the crux of when is the information reasonably applicable to safety.  It is a statutory standard.  Under what circumstances is this information required, either by regulation or statute, to be submitted to the FDA?  These are the questions we would like you to talk about.


     But we have a proposal.  We are not just asking you to come up with something.  This proposal is that FDA will establish policies on categories of pharmacogenomic studies, broad categories.  This will be hard.  Let me give you some examples.

     We will have a submission requirement, an explicit submission requirement as part of the policy and the submission requirement will be of several types.  A submission might not be required for some types of data that are being generated.  I think the various presenters presented some kind of data that wouldn't be required to be submitted to the FDA--early drug development data and so forth.

     A submission may be required under the current statute and regulations right now for some other types.  We would sort that out in this policy.  But some of the submissions would not have regulatory decision impact.  This is the proposal.


     What do I mean by that?  We would say in our policy related to the categories what the regulatory impact of the study type would be.  Results from some study categories would not have regulatory impact.  In other words, they wouldn't influence the decision making.  We wouldn't say to Abbott you have to do ten more clinical trials or five more animal studies, or whatever.

     Other study results though will have to be utilized as part of the safety and effectiveness evaluation.  That is clear right now and I think that is clear going into the future.  The question is about the difference between these two types of categories.


     A possible threshold determination or a proposed one that we could use: does the genomic information represent a valid biomarker with known predictive characteristics?  That is an important threshold that I want you to think about.  I know Dr. Spear said, well, maybe if this information were to be submitted and it was analyzed by the FDA, it doesn't mean necessarily we would require that a drug would have to be used in a subset, or something, to respond to your hypothetical, where you did a trial in everyone and some subset seemed to respond better but you would have to submit that information.  That is the proposal.  But that wouldn't mean that you could only market that drug for that subset.  So, those things aren't linked necessarily.

     The FDA is proposing we would develop this threshold and this set of policies using a public and transparent process with advisory committee insight, and we are starting this process now.


     Possible procedures--procedures are important in the transparency.  We would establish an interdisciplinary pharmacogenomic review group probably across the centers.  There is expertise in different places.  We would also categorize these studies and develop internal procedures and publish them in a guidance so they would be publicly available.  Our guidance procedures call for us to publish a draft, get public comment and then go to a final guidance.  So, we would have a lot of opportunity for public discussion.

     We could propose that the results would be submitted to the IND or the NDA as a research information package for review by this group.  These are the ones who are not going to have regulatory impact.  This would allow FDA to get this information where we have determined it is not a valid biomarker, but it is useful information that we need to learn about, and that Frank and his group needs to learn about, and Larry's people need to learn about.  We could get this; we could review it centrally and develop our knowledge base without negatively impacting on the clinical development programs or the review process.

     We would have to have periodic public re-evaluation of this decision tree that we would construct to allow all this to happen, to make sure you are still on the same track, which gets to the issue of would we renege on this at some point.  We would have to have a public discussion.


     Examples I think that are fairly clear that do have some impact in the  sense that you would need to submit doesn't mean, as I said, that there would be impact on the label necessarily in some way but trial enrollment by genotype, if you are going to stratify people in your trial by genotype, you want to enrich for responders, you want to avoid bad outcomes in a trial so you are going to eliminate certain people from a trial.  That is something that needs to be submitted as a part of your clinical development plan in this proposal.

     Selection of dosing based on metabolizer genotype, I think we would all encourage these type of activities to be done during drug development where there is a hypothesis, say, that a metabolizer genotype may make a big difference.


     Where there is a safety rationale based on animal genomic data, say, explaining why a toxic finding in animals is unique to that species and is not expected in humans, other species, and so on, in general, we think that results intended to influence the course of clinical development process should be included in the safety and efficacy evaluation.  Does everybody follow that?  So, if you are randomizing patients based on these data, if you are modifying their dose, if you are doing those sorts of things, this should be included.  This has gone into mainstream drug development as part of the safety and efficacy evaluation.


     Potential results that could be done and not have regulatory impact, and I am not saying whether or not these would have to be submitted; that is going to depend on the interpretation of the regulations that I showed you, but what if you found a new transporter gene and your drug is metabolized by it and you want to look at the diversity of this gene versus the response, the clinical response in clinical subjects?   We really wouldn't know anything about what this gene means or is, and so forth.  So, that type of study would be one that wouldn't have regulatory impact, decision-making impact.  Something that is being done frequently now is what I call sort of fishing expeditions where you are looking at markers in clinical trial subjects during the trial.  That is hypothesis generating.  That would have no regulatory impact.  The same with microarray screening in trial subjects where you are simply trying to find out correlations.  The same with gene expression microarray screening in animal toxicology studies.

     Again, I am not saying whether or not these would have to be submitted.  I am saying until these have been established as far as the validity of information from these studies, they would not be figured into regulatory decision making.  We already have good tools, animal toxicology, traditional nonclinical development.  We have good tools to ensure the safety of clinical subjects, and so on.


     We are going to have one more presentation but then these are the questions that I would like maybe the discussion, if possible, to focus on.  Is this approach reasonable?  We have to move forward.  You have heard the horns of the dilemma that we are on.  These studies are being widely done and they may have tremendous promise.  There is extreme concern about how they are going to be used, and that is legitimate because they are not really ready for prime time, many of them, for regulatory decision making.  We need to find a way to get the information in, develop our policies, develop a regulatory framework for this information, and help to move this field along.  So, is what we have proposed reasonable and feasible, or should we go back to the drawing board?

     Will it achieve its objectives?  Obviously, we need to achieve the objectives via policy.  We need a free exchange of data, and this is how I started the afternoon although that was a long time ago and you probably don't remember.  We need to have free exchange of data.  We need to have the ability of the FDA scientists to begin developing the framework for the new findings so they can be integrated into regulation.  We need to advance the use of the new science, not impede the use of a new science in drug development.

     With that, I will turn it over to the last speaker and then will come back.

     DR. DOYLE:  Thank you, Dr. Woodcock.  For our last presentation we have Dr. Benjamin Wilfond, who is with the Medical Genetics Branch of the National Human Genome Research Institute at the National Institutes of Health.  Dr. Wilfond is going to address ethical issues associated with regulatory review of pharmacogenomic data.

Ethical Issues with Regulatory Review

of Pharmacogenomic Data

     DR. WILFOND:  Thank you.


     It is a pleasure to be here.  I certainly enjoyed hearing the prior presentations.  There are a lot of questions which I wish I had time to answer but I won't address.  What I will do is try to focus on Dr. Woodcock's specific proposal and make some comments about some of the ethical issues related with that.

     For those of you who aren't familiar with ethicists, basically all we do is ask more questions--


     --but on a more serious note, we try to refine the questions to give the questions greater clarity.  So, that is what I hope to do today so it will be a way of beginning the discussion in a more earnest way.


     We have heard from all the speakers that the ultimate progression from pharmacogenomic research is to improve clearance applications.  But it is a challenge to get to that place.  The question that is being asked from a regulatory perspective is when is it appropriate to use the data, either for the drug approval process or for clinical decision making.


     The FDA's proposal is to divide this into two current approaches.  The first is the regulatory approach which, as Dr. Woodcock described, would be based upon when a sponsor intends to use the data either to demonstrate safety or efficacy from clinical trials, or if they would recommend specific clinical applications of the data.  However, it would then put aside other categories as research uses which might include unique genes or gene expression profiles that would not be part of the regulatory process.


     One of her slides had this quote which I wanted to discuss because I wasn't quite sure exactly what it meant, which is develop thresholds and policies using public and transparent process with advisory committee oversight.  I understand all the words.  I think the challenge is to figure out what exactly does that mean, or what exactly would that do.

     I don't mean to say this facetiously.  I think it is actually a very difficult question.  What was mentioned is that we need to have some threshold for these issues of biologic validity and clinical relevance, but the challenge is what process will we use to decide in a particular instance when this threshold has been reached.  I think that is going to be the hard question.


     One of the ways we can think about this is to think about two different types of decisions.  One I will call policy decisions, which would be thinking about categories of studies, types of data, and whether we would use of not use those types of data.  We can also think about thresholds for inclusion from a conceptual point of view in terms of biological relevance or clinical significance.  But again, as I pointed out, the challenge will be the application specific decisions.  Those are the ones that I think will be the most challenging, and those will be the ones that either will be used for making further requests for more data or to be used for evaluations for decisions about regulatory approval.

     Both of these categories would benefit from review of ongoing research data.  So, the general idea of trying to ask industry to provide data to the FDA could be useful in either of those two processes.


     From my quick understanding of what is being proposed for the pharmacogenomic review group, it is going to be primarily focused on policy.  So, the question is will this group, or should this group also consider specific drug application decisions itself or might it, as part of its activities, say that a decision needs to be made and a second group ought to do it?  I think that is one question that I think is important to get some feedback on about what the role of this pharmacogenomic group should be.  Dr. Woodcock may have a particular answer for that, I don't know, but that is a question I would like to know about.

     The second issue, and this is the second main point that I want to make, as we think about this group or another group, is the procedural options we have for how this group should practice.  I wanted to identify three specific categories of options that we would need to consider.  One is going to be the issue of the confidentiality of the group.  Would this be a public group or a private group?  The second would be is this independence?  Is this a group that is internal to the FDA or external to the FDA?  The third is authority.  Is this an advisory group or is this a determinative group?


     I think these are very important questions and there is not one clear answer to them but it might depend upon what you wanted this group to do.   To illustrate this point, I have this slide that shows IRBs, data monitoring committees and FDA advisory committees along these three points.  You can see that each of these has different features among these characteristics.  All of these are appropriate for what they do, but the question will be, as we think about this group, where might it fit in to these categories?

     So, with regards to confidentiality, clearly a public group will have greater accountability and transparency.  With a private group there will be a greater willingness of industry to share what otherwise would be thought of as confidential data.  An internal group will have greater familiarity and expertise with all of the issues related to the drug approval process, but an external group might have greater objectivity and not feel so constrained.  With regards to authority, groups that provide recommendations are often much more acceptable, except that it still would require the designation of who ultimately would be making the decision.


     I think I would like to end by saying that a choice between these procedural options includes ethical considerations of these three categories and the tradeoffs between them, and that this information could be used to decide whether or not this review group would decide about specific applications or refer to a specific group and, secondly, to decide about these issues of confidentiality, independence and authority.

     What is missing from the slide is perhaps the hardest question, which is what criteria ought this group, whether it is the review group or second group, use for making decisions in specific cases because, as much as we would like to say that this will only be for research, we have to sort of think down the line and realize that at some point we will cross that threshold.  Thank you.

Questions and Discussion with the Board/Presenters

     DR. DOYLE:  Thank you very much.  Now it is in the panel's bailiwick.  It is now time for us to respond.  The proposal is in front of us.  Anyone have any comments or suggestions?  Yes, Dr. Laurencin?

     DR. LAURENCIN:  First I want to thank the speakers today.  This was really a great session and great information.

     My comments are that before we consider the world of pharmacogenomics we should consider a group that we can identify right now that are suffering healthcare disparities possibly due to drug effects, and a group that could possibly benefit from the monitoring or higher level monitoring in terms of drug effects, and that is the African American population in the United States.

     If we think about pharmacogenomics, it is interesting that pharmacogenomics as a discipline probably has, as part of its roots, the observation that certain antimalarials when given to African Americans cause hemolysis, which led people to uncover problems with G6PD deficiency and, again, led to the realization that different drugs have different effects on people.  We know now in the United States that there are a number of drugs that are given to the African American community for problems that have high prevalence in that community, yet, the numbers of patients that are tested in terms of clinical trials, those numbers are low, leading to questions about the effects.

     Now, with more and more acceptance of data from European trials, one would expect that the numbers of African Americans that suffer from some of these diseases, in terms of clinical trials, will be decreased.

     I know the Commissioner is sensitive to this.  I have spoken to him about it and he is very sensitive to this issue, but I think that before we explore the world of pharmacogenomics and get committees and councils set up to start to explore this or subgroups, I think that we need to really take a step back and also look at what is going on in populations in America right now whose healthcare disparities may be affected by the manner in which drugs are regulated right now in the United States.

     DR. DOYLE:  Dr. Shine?

     DR. SHINE:  Can I ask Janet some questions?

     DR. DOYLE:  Sure, absolutely.

     DR. SHINE:  First of all, just for clarification, Janet, when you made the statement "free exchange of data" I presume that means between the FDA and industry, and that it doesn't mean that it goes on the web.

     DR. WOODCOCK:  In principle, it is most desirable that this information all be published but I believe that a fear of regulatory action is inhibiting publication, although also inhibiting publication, obviously, is the proprietary nature of some of the data and its use in drug development.  To answer your question, no, I meant free exchange between the regulators and the regulated.

     DR. SHINE:  Thank you.  Again, I understand that you are proposing a plan but it just would be helpful to be a little bit more explicit.  Under the proposal, there would be a threshold determination of genomic information which represented valid biomarkers with known predictive characteristics.  First, does that imply then that those biomarkers would be then obligatory in terms of all the testing that is done, or are we talking about biomarkers that sponsors happen to collect as a consequence of it?  I am trying to understand how prescriptive this would be.

     Secondly, some sense of what you mean by valid biomarkers.  I can easily see a requirement, for example, for looking at genetic evidence with regard to metabolism.  I could see that even ultimately being a requirement in terms of how one does it.  On the other hand--and this goes to Cato's comments--will there be an obligatory requirement that studies be done in such a way as to determine biomarkers for particular populations?  So, some help with regard to that?

     DR. WOODCOCK:  What we are talking about right now is studies that companies are doing for their own purposes, and whether or not that data needs to be submitted, number one, and if it is submitted whether it would be used in regulatory decision making.  We are not talking at all about requiring companies to do any kind of genetic testing during drug development.

     As Larry Lesko showed, we are seeing a lot of the drug metabolism polymorphism data right now, and that is something that clearly is a valid marker, a predictive marker for many systems, many enzyme systems of what is going to happen to metabolism of that drug.  So, that is kind of one category.  But we are not requiring that people be genotyped.  We have other ways of getting that information through pharmacokinetic data, phenotypic data, all sorts of ways of determining how people metabolize drugs.

     So, no, we are not talking about establishing a system whereby we would develop new requirements.  We are talking about all this information being generated out there and what should be submitted to the FDA, and what of that universe that is submitted to the FDA, which part of that universe should be used for regulatory decision making.  The threshold that is being proposed is the universe where we understand what the information means.

     DR. SHINE:  For example, if they have information about a genotype related to metabolism, that might be required whereas something else might not be required.

     DR. WOODCOCK:  Required?  You mean in a submission?

     DR. SHINE:  Yes, if they had the information.

     DR. WOODCOCK:  Yes, that is correct.

     DR. SHINE:  Finally, I am curious, you talk about the research information package for review by IPGRG.  Now, information package--I would imagine, from what I understand about the discovery process and so forth, there is a huge range of research that may go on with relation in a company, a laboratory or whatever, some of which bears directly on that study; some of whiich is interpreted or inferred for the purposes of that study.  It seems to me there is a huge potential spectrum there.  What kinds of criteria would you use for defining what this package would look like?

     DR. WOODCOCK:  Well, I think Frank Sistare was trying to get to some of that in his talk.  We have to work with industry in that particular instance to define the parameters of what would be submitted because, obviously, these results are platform dependent.  So, we would work with industry on the specifics of a submission.

     DR. SHINE:  But what would be the principles that underlie that discussion?  I mean, what boundaries would one want to place around that in terms of what we would be talking about?  I am trying to understand what the underlying thesis would be in terms of the scope of the activity.  I understand you would negotiate the contents but there needs to be at least some definition of the area.

     DR. WOODCOCK:  Well, the area would be at some point during drug development, when you have a drug that is going into people and you are submitting that information in the IND, say, to the FDA or you are going through clinical development with a drug and you are going to do genetic testing on animals or people, that type of testing might be submitted to the FDA.

     DR. SHINE:  But it would be specific for that drug?

     DR. WOODCOCK:  Absolutely.  Yes, everything we do is more or less product specific, yes.  So, it would be specific tests in animal or human or maybe cellular systems in some cases on that drug that is actually in an IND, is actually in drug development.  Other submissions might be voluntary.  People were telling me at the break that their folks are doing natural history studies--you know, there is a lot of other information.  Hopefully, that will be published and we would also like to see it as well.

     DR. SHINE:  The last question goes to the discussion by our ethicist.  As I understood your presentation, you in fact do separate the role of the IPGRG from the regulatory process.  As I understand it, they would have nothing to do with the regulatory process per se?

     DR. WOODCOCK:  They probably would talk to the folks making the regulatory decisions but they would be separate.  It would be a separate review.  We do that in other cases, such as animal carcinogenicity studies.  These are centrally reviewed at CDER by a committee of expert toxicologists and statisticians.  They give a recommendation back to the division about what that study means.  In this case probably very little would go back to the reviewing division, although they might talk to them.

     DR. SHINE:  But you used two words, "talk to them" or "recommendations."  Which would it be?  Is it a discussion or does this group make any kind of recommendations?

     DR. WOODCOCK:  The current proposal is that we would establish the categories up front.  That is very important.  Once the categories are established up front, that submission comes in and if it is not a regulatory impact submission it goes to the multi-disciplinary review group to look at.  They don't make recommendations about regulatory action.  They do not.

     DR. SHINE:  Thank you.  Thank you, Mr. Chairman.

     DR. DOYLE:  Dr. Thomas?

     DR. THOMAS:  I would just like to ask the experts if they feel comfortable with platform uniformity and standardization, harmonization, as well as the assimilation in terms of a repository or library for this technology based on the state of the science before we get down to that next step.

     DR. DOYLE:  Any response?  Frank?

     DR. SISTARE:  I am not an expert so that is why I don't feel comfortable answering that.  I don't know who is an expert on that topic, but I can speak from some level of experience.

     One of the points I tried to make in the slides is that platform comparisons, even the same platform, the same sample, in the hands of different investigators, if you are looking at a venn diagram, is not going to perfectly overlap.  But the question is what does overlap or even what doesn't overlap, does that represent biological truth?  Is it accurate information?  That is the key.  So, what we are actually trying to assemble is what actually is out there in the literature to try to address that question.  When people for the most part have tried to confirm the results of transcript alterations using alternative methods, say RT-PCR, there is very good agreement.  So, that suggests that when you do see changes, they are real.  Very few publications have looked and asked when you don't see changes, are those non-changes real.  In the few that have been published, again, there is very good agreement.

     Now, let me answer this question from another perspective.  If I look at one platform and look at another platform, again, do I get perfect overlap in those transcripts?  The answer is no.  A venn diagram won't be perfect.  If you do the correlation coefficient, like I did, quadrant 1, 2, 3 and 4, you are going to see some things in quadrants 3 and 4 where one platform says they went up and another platform says they went down.  Is one wrong?  Maybe not because there may be spike variants and you may see real changes there.  They may both be giving you accurate information.

     We are actually involved in a very nice collaboration right now with the same samples being looked at on two different platforms.  The bottom line right now that we are seeing is you can come to the same biological conclusion--you can come to the same biological conclusion, and that is the bottom line, can you get to the same knowledge level when you look across platforms.  A lot has been said, a lot of nay-sayers have said confusion, havoc, a lot of craziness by looking across platforms, but the bottom line is, by careful investigators following established protocols, making sure they are doing everything right--new tools in the hands of fools, you are going to get garbage out--but if you have careful investigators looking very carefully, really testing what they are doing and validating it, you are going to get similar biological conclusions at the end of the day.

     But it still is a burning question and we need to address that.  We are addressing that exactly.  We have this Iconix database that we have access to, 550 compounds.  We are generating data on different platforms and asking the question if I take this different platform data and I relate it to this database or for the Amersham Motorola platform can I get the same biological answer?  So, those experiments are being done and they do need to be done.  There is a lot that needs to be done to calibrate platforms across different laboratories.  We are hoping a universal RNA reference standard will help that process.  We are not at the point where we can rely on this for regulatory decision making, but if we envision the day when we can use this information and obviate the need for some of the very lengthy experiments we do right now using very traditional toxicity testing, we have to go through this process of making sure we get accurate answers.

     DR. THOMAS:  Thank you.

     DR. DOYLE:  Go ahead, Dan.

     DR. CASCIANO:  Another response to the question, I don't think the mystery is any different than with any other assays that we have developed over the last 25 years.  Even so-called standardized assays like the Ames assay provides us data that is not relevant to so-called gold standard until you go deeper into the analysis and understand well what may be true for a mouse may not be true for a rat.  So, it depends upon what you mean by standardizing and what is the gold standard.  So, I don't think we are that much different.

     DR. LAURENCIN:  I am not sure I would use the Ames assay as an example.

     DR. CASCIANO:  I only used it because it is a standardized assay.

     DR. LAURENCIN:  Yes, okay.

     DR. DOYLE:  We have a few other questions here.  Dr. Rosenberg?

     DR. ROSENBERG:  First of all, Frank, I agree with you about gene expression data.  That is why I want to concentrate maybe on something that is a little more relevant in real-time today, which might be the genotyping data, and talk a little bit about threshold determination and criteria that the agency may want to consider.

     It would seem to me that in trying to relate genotype what is going to be very important for you to consider is a term that geneticists use called penetrance.  Where penetrance is very high, and the threshold of penetrance is probably going to have to be thought about differently for toxic phenotypes versus efficacious phenotypes, and you will probably have to break them down differently, but normally what one would consider, of course, is that if a genotype is highly penetrant with its phenotype, then it meets the kind of criteria that the agency would want to consider as being meaningful because if genotype and phenotype correlate, then you can get decent data.

     The problem, of course, is that for very few loci is that the case, and when you are dealing with the human population and you have things that have lower penetrance where many people, or a number of people are carrying genotypes but express different phenotypes from that, you have to be very careful about making regulatory decisions about who should or who shouldn't be taking drugs because their phenotypes have to be correlated with the background of perhaps many other genes that are being expressed against a background of the genotype you are measuring.

     Therefore, genetic penetrance I think is a very good criterion and threshold concept that you should be looking at in order to kind of start to make or think about decision making in terms of which of these genotypic changes are relevant to the agency starting to make real decisions about.

     DR. DOYLE:  Dr. Woodcock?

     DR. WOODCOCK:  Thanks.  I think that is a very good point.  I think we have one other step that we have to go.  It can't be just phenotypic expression.  It actually needs to influence drug response or we have to think it is predictive in some way.  So, there are kind of two thresholds that genetic information has to make.  I mean, many of the drug metabolism polymorphs, sure, affect drug metabolism because metabolism has multiple routes and they aren't that predictive of one level, for example, so you know, we don't make a big deal out of that.  So, I think your point is extremely well taken.

     DR. DOYLE:  Yes, Dr. Pickett?

     DR. PICKETT:  Yes, I also would like to just compliment the group on the presentations today and to also hear about some of the thinking here, at the FDA, on pharmacogenomic approaches.

     I think the procedures Janet has outlined, to me, make a lot of sense on the path forward.  What concerns me are examples that implied regulatory impact versus those that do not, and how one can really clearly define that.  To me, it is obvious in terms of metabolizer data to regard to 2D6 polymorphism, etc. and that that should be an important component.  For example, if you are developing a drug against HIV and against the C*5 receptor it would be wise to genotype patients that actually express that receptor before you initiate therapy.

     What is unclear to me is transporter gene diversity versus response in clinical subjects.  You said that that would be pharmacogenomic data without regulatory impact.  To me, it would depend upon the data that obviously comes out from that type of study before one can make that determination.  So, going into the process I think it will be a challenge to clearly define pharmacogenomic data that will be subject to regulatory impact versus that which is not subject to regulatory impact.

     DR. WOODCOCK:  Could I make a comment about that maybe to stimulate some more discussion?  I think we agree with you.  There is a difference though.  Say, a firm is studying a transporter gene and wishes to use it, that is a completely different situation than being compelled to use that information that you collect.  In any case, if a firm wishes to use genetic data as a biomarker to shape the design of a clinical program, then that is completely fine and we welcome that.  That is a challenge obviously, but that is not the subject of this.  It is where this kind of data is being collected for research purposes to see perhaps, if you wish to do that, but you are not doing it.  Does that make sense?

     DR. PICKETT:  Just one other comment.  Because pharmacogenomics DNA arrays are only part of the overall biological picture, it is not clear to me how other technologies, such are proteomics--for example, I mean, if I was focusing on oncology and trying to understand signaling pathways, more than likely I would look at phosphorylation of certain proteins in a signaling pathway to give me more information versus a DNA array experiment.  So, it is not clear to me how the agency is thinking about utilizing that data to generate really a total picture of biological response.

     DR. WOODCOCK:  Is it fair, I will ask the assembled people here from the FDA, all the different people in the back there?  I think it might be fair to say we are looking at all of this in a similar way.  This is a spectrum of markers and folks will be collecting these data, and so forth, so it is very similar.  I mean, proteomic data is, in my mind, more empirical than the genetic data because it is more of an observational empirical pattern recognition type of thing.  But conceptually, as far as this scheme we are proposing, it is very similar.

     DR. SISTARE:  Yes, I would add that I agree 100 percent.  Again, one of the points I was trying to make is that the RNA is a response to what the proteins are really doing there.  It is an indirect measure because we can measure it so tremendously right now.  It is much more difficult--we don't have the protein arrays to the point where we can get fidelity of those kinds of measurements right now.  It is happening; it is developing; and definitely we need to be on board with that.  There is some extremely provocative data coming out of some of the CBER labs right now.  For example, the HER2 NU story and that is only part of the story.  Part of the downstream of that is from HER2 NU receptor signaling.  Looking at phosphorylation of proteins and signaling pathways is adding another level of information in terms of who to treat; how to measure responding.  So, I agree with you 100 percent.  We need to be prepared for these kinds of input data as well.

     DR. WOODCOCK: Can I just say one more thing?

     DR. DOYLE:  Sure.

     DR. WOODCOCK:  But once the developer makes the decision to segregate patients based on one of these markers, then we are in a different game.  That is part of the development program and we all have to look at that and consider it.

     DR. DOYLE:  Thank you, Dr. Sistare and Dr. Woodcock.  Dr. Riviere?

     DR. RIVIERE:  As time is running out here, I just want to make one comment.  I think you are going along well.  You just really need to be very careful as to what you actually require.  I think you have brought up all the cases.  The point is that that data is existing and it is going to be generated, and really you have to start developing a framework upon which you can even accept the data into the system.  Right now you can't handle any of this data.  As you work on this just keep it very, very flexible because, again, I would think that the proteomic data might actually show more information in some cases than the genomic data.

     I also agree with you that if the genomics data is actually being used to design a clinical trial, then obviously that data somehow has to be presented.  If it is not, and if it is being interpreted to, say, an animal toxicology study or to rule out a response in another species, you need to at least have some mechanism of bringing that data into the picture because right now you don't.  I think it is a process.  In a year or two a lot of the outstanding questions right now will not be questions or we are going to have a bunch of new questions.  I think you are doing fine; just keep going.

     DR. DOYLE:  Dr. Davis?

     DR. DAVIS:  As someone who has watched this and been to most of the meetings that Frank and PhRMA have had together, I would like to applaud the effort.  You know, we started out with a great deal of fear that the agency was going to try to force something down industry's throat that we weren't ready to present and how it was going to be interpreted.

     Clearly, Janet has presented, to me, a well thought out proposal.  I think there is still a long way to go but, clearly, there is a sense of recognition that there are these two camps of data, one being the research grade data that we were very much concerned about that, all of a sudden, lead optimization data was going to need to be dumped into a regulatory package.  I think most of us are aware, I believe, that if this is data that safety decisions are being made around, then you have to share that data and we will have to debate what it means.  But if we are making decisions with it, we are drawing some conclusions from it so it would be crazy not to think that we would have to share it with our FDA colleagues.

     But the question will be what are the grey areas, and what to submit with no impact versus submit with regulatory impact and I think we just have to work that out.  But I think the groups are in place to work on that, the ILSI, PhRMA groups, some of the workshops that are going on.  So, I am quite pleased with the proposal.  Thanks.

     DR. DOYLE:  Any other comments regarding this proposal?

     DR. SHINE:  I thought the representative from Abbott was going to say something.

     DR. DOYLE:  Brian?

     DR. SPEAR:  I will take the opportunity.  I was intending to respond to the question about having a standardized platform.  It is very important that we have sufficient standardization that we can have data from one experiment to another and compare them.  The concern I would have would be if standardization involved something that would inhibit technology development.  That is, saying that one device procedure or method was the acceptable device procedure or method could significantly stop development of new devices procedures, methods and so on and that would be very antithetical to what we would like to see in research.

     DR. DOYLE:  Dr. Sistare?

     DR. SISTARE:  We totally agree with that.  Maybe the best way to characterize what we would like to do is to provide universal calibrators as opposed to standardizing a procedure, but offer universal calibrators so people could see that we are on the same page.

     I wanted to also, if I can, come back to something that Dr. Rosenberg started with when he started talking about penetrance, focusing on DNA.  I want to bring that discussion to RNA.  One thing I think I failed to make clear--you know I don't think I heard one work of Dr. Lesko's presentation because I was going through my presentation and what I didn't say and what I did say but, anyway, I have heard it before and I have the notes--but one of the things I didn't make clear and I kick myself for not making it clear, is the data that you saw at The Netherlands Cancer Institute to day, and from this day forward, that information from those arrays is being used to make clinical treatment decisions.  Okay?  They have made the decision the data are powerful enough now to decide who does and who doesn't get treatment based on expression array.

     I would ask our ethicist, in a sense, to ask the question, posed with data like that is it ethical to not use that information to treat a lymph node negative patient with chemotherapy in a situation like that when you have data that look that compelling.  So, when I said the future is now, it is now.  This is happening; they are using expression patterns to decide who does and doesn't get treatment.

     DR. ROSENBERG:  Frank, help me then because in the data you showed clearly, again, there wasn't 100 percent separation in those groups.  Therefore, if there are people who are in the not treated group but they still are at risk, and you showed they are at risk, to deny them treatment because they fall in some RNA expression group I would say is just as damning.  You can't deny such a person treatment because you have shown, in fact, that they have a statistically significant probability of having the same problems as people in the high risk group and, therefore, I think it is very hard to use that data to make good clinical decisions.

     DR. SISTARE:  There is no doubt about it, it is hard to use the data.  There are two ends of the spectrum where I think some decisions are easier to make.  There are grey areas where it is very difficult to make decisions.  And, even at those ends where you will see that people don't conform, it means that there are other variables.  If you want to talk about penetrance, there is not complete penetrance, if you will, for RNA expression patterns which means we have to do more work to define what is different about those individuals.  We are not perfect yet, but it is better than what we are doing now in terms of--again, I am not a clinician.  The question you are asking, I agree with you 100 percent and I don't know what they are doing in The Netherlands in terms of deciding not to treat patients.  That is a much more difficult question I think than to treat someone that otherwise looks clean.

     DR. DOYLE:  Thank you Dr. Rosenberg and Dr. Sistare.  Dr. Shine?

     DR. SHINE:  I would just comment that ultimately it is going to be a decision made by a patient with a doctor, given the relative risk, the options and whether they want to be treated and what the complications are.  So, I agree entirely with Marty that we have to be very careful from this point of view.

     I do think that the pharmacokinetic data becomes critical in terms of what you are talking about.  So, I am comfortable that if you do good pharmacokinetic data, that will provide a lot of the information with regard to the genotype.

     Mr. Chairman, do you want a sense of the committee on this matter?

     DR. DOYLE:  Yes, I think that would be a good thing to do.

     DR. SHINE:  Well, I would just comment that the devil is in the details and seeing how, in fact, this is articulated and elaborated is the critical thing.  But I think the sense of the committee, which they can express in any ways they want, is that the approach that is being proposed is sensible.  It is well thought out.  It offers some promise both to improve the regulatory process and the educational process for the agencies, and I would support moving forward with this proposal.

Closing Remarks/Future Directions

     DR. DOYLE:  I sense that is exactly what the committee is thinking right now.  So, there is general agreement so thank you for articulating that, Dr. Shine.  Does the committee have any further comments?  If not, let me move ahead and try to summarize these conversations.

     First of all, relative to the quality systems presentations, I think the Board applauds the FDA's efforts in the quality systems area.  The Board encourages FDA to emphasize metric measurable outcomes when approaching what one can do to improve the systems.

     With respect to FDA personnel, the Science Board encourages the worker--I can't read my writing--that the worker provide these outcome measures and needs to consider not only total buy-in from the internal work force but also consider increasing head count and/or outside expertise.  We encourage the FDA to continue strategic succession planning not only for attrition of personnel over time, but planning for what type of science you will have in the future and what type of expertise you will have on hand.

     The Board applauds FDA's efforts in getting involved with sponsors early in clinical development but asks the agency to emphasize careful consistency.

     Relative to pharmaceutical manufacturing, the Board is interested in hearing about similar efforts with regards to vaccine manufacture.  The Board encourages the use of case reports as a learning/training tool.  Thirdly, the Board encourages measuring outcomes and benefits of PAT.

     Relative to the patient safety initiative, the Board encourages FDA to facilitate feedback and communication.  Are there any other comments or follow-up that I have missed in that regard?

     DR. SHINE:  Pursuing the ambulatory data.

     DR. DOYLE:  Yes, pursuing the ambulatory data.  Thank you.

     Relative to pharmacogenomics, the Board applauds the FDA's efforts in this area.  It encourages the FDA to step back and look at specific populations, that is minority populations, as we look at pharmacogenetic populations.

     FDA needs to explore and consider what needs to be done about the standardization of different microarray platforms.  They need to consider penetrance and phenotyping and genotyping.  Marty, would you like to further clarify that?

     DR. ROSENBERG:  No, I think Janet made exactly the correct point, that it is good criteria to use and that you have to pick the right phenotype because it has to be the medically relevant phenotype to be measuring.

     DR. DOYLE:  Thank you, Dr. Rosenberg.  We need to clearly define which examples determine a regulatory impact and which do not.  We need approaches that are sensible and offer promise to improve the regulatory process and encourage FDA to move forward on this issue.

     Does anyone have any disagreements with that or any additions?

     DR. NEREM:  There was the statement that Ken made, and I think we should emphasize that the devil is in the details.

     DR. DOYLE:  All right, good point.  Dr. Shine?

     DR. SHINE:  Before we adjourn, to come back to the Commissioner's initial talk this morning when he talked about the substantial decline in the number of new drugs coming through, the new drug applications and, at the same time, we have been talking about pharmacogenetics and pharmacogenomics.  I am just curious given that there are people here from very diverse backgrounds, including industry, as to how much of this reflects the fact that although we may know lots about the genome and lots of genes, we don't know what the targets are and we don't know how to attack those targets in a predictable way given the multiple genes that regulate a variety of processes.  We have the problem of polygenic involvement in many of the common illnesses, and so forth, and I am wondering--you know, it reminds me a little bit of the computer revolution.  For 20, 25 years in the '60s and '70s we were going to see this huge revolution in computing and it wasn't until the mid to late '80s and the early '90s that it just exploded and you began to see real differences in the way companies operated, middle management being laid off, and so forth.  I am wondering whether there isn't going to have to be a very prolonged period of development in this area, and then an inflection point when some of the fundamental questions about these targets are addressed.  I am worrying about whether the movement from chemistry to genetics may, in fact, account for some of this lag.  I am just curious, since we have some very knowledgeable people here, whether they would have any comments about the Commissioner's observations which I found very interesting.

     DR. DOYLE:  Any thoughts.

     DR. DAVIS:  Well, it is too bad Cecil left given the position that Cecil holds with Schering.  I think you are absolutely right.  I think everything you elucidated is part of the issue or the perspective.  Genomics has given us a lot of targets that we don't know what they do, and also even after you have a target, being able to put a drug on that target and have it have an effect is not always an easy thing.  Some of those targets are not as easy to get to.

     So, I think we are in a place where there is a plethora of information but we aren't sure yet what to do with some of that.  On top of that, as the Commissioner mentioned, there is the increasing review time.  So, I think it is a combination of a whole lot of stuff that has really got us in the position that we are in.  I am optimistic, being in industry, that in a couple of years there is going to be this deluge.  We sure hope there is, anyway and I hope we don't have to lay off a lot of people, like the computer people did.

     DR. DOYLE:  Just a few last comments.  First of all, I want to thank Commissioner McClellan.  He just did a super job in sharing with us his vision for the FDA and I think, by all standards, at least I feel very comfortable that the FDA is in good hands.

     Secondly, I want to thank all the presenters for the excellent presentations that they have given and the forward thinking that was provided.  I want to also in particular thank Dr. Woodcock and her team for the very informative overview that we received of their research program yesterday.  That was very well done.  Finally, I want to thank Susan Bond for assisting this group for the last three years, and wish her all the best.

     With that, happy trails and we look forward to seeing you in the fall.  Excuse me, one last thing, I want to let either Dr. McClellan or Dr. Crawford have the last word here.  So, do you have anything you would like to add?

     DR. MCCLELLAN:  I don't have anything to add.  This has been a terrific discussion of a whole set of complex issues.  Thank you for your comments about where we seem to be headed.  To the extent that we have coordinated vision and are actually making progress in getting there, it has nothing to do with me and everything to do with the very talented and dedicated staff in this agency, and I am pleased to see that the Science Board is able to have some impact on the direction in which we are heading.  We are going to need that help more than ever so thank you very much for what you are doing.

     [Whereupon, at 4:17 p.m., the proceedings were adjourned.]

- - -