UNITED STATES OF AMERICA
+ + + + +
FOOD AND DRUG ADMINISTRATION
+ + + + +
ONCOLOGIC DRUGS ADVISORY COMMITTEE
+ + + + +
+ + + + +
NOVEMBER 28, 2001
+ + + + +
The Subcommittee met at 8:00 a.m., at the Gaithersburg Holiday Inn, Two Montgomery Village Avenue, Gaithersburg, Maryland, Dr. Victor M. Santana, Chairman, presiding.
VICTOR M. SANTANA, M.D., Chairman
PETER C. ADAMSON, M.D., Consultant (Voting)
FRANK BALIS, M.D., Guest (Non-Voting)
MARTINE BAYSASS, M.D., M.S.C., Industry Guest (Non-Voting)
MARK BERNSTEIN, M.D., Guest (Non-Voting)
JAMES M. BOYETT, Ph.D., Ad Hoc Member
SUSAN L. COHN, M.D., Ad Hoc Member
CHARLES A. COLTMAN, JR., M.D., Guest (Non-Voting)
ALICE ETTINGER, M.S.N., R.N., C.P.O.N., C.P.N.P, Ad Hoc Member
JERRY Z. FINKLESTEIN, M.D., Ad Hoc Member
STEPHEN L. GEORGE, Ph.D., ODAC Member
STEVEN GOODMAN, M.D., M.H.S, Ph.D., Guest Speaker
JOSEPH GOOTENBERG, M.D.
NANCY KEENE, Patient Advocate
ERIC KODISH, M.D., Ph.D., Guest (Non-Voting)
EDWARD L. KORN, Ph.D., Guest (Non-Voting)
J. STEVEN LEEDER, Pharm.D., Ph.D., Guest Speaker
JODY PELUSI, F.N.P., Ph.D., ODAC Consumer Representative
DAVID G. POPLACK, M.D., Guest (Non-Voting)
DONNA PRZEPIORKA, M.D., Ph.D., ODAC Member
WAYNE RACKOFF, M.D., Industry Guest (Non-Voting)
MARY V. RELLING, Guest Speaker
C. PATRICK REYNOLDS, M.D., Ph.D., Ad Hoc Member
ERIC KEITH ROWINSKY, M.D., Guest (Non-Voting)
MALCOM SMITH, M.D., Ph.D., Guest (Non-Voting)
CLINTON F. STEWART, Pharm.D., Guest Speaker
SUSAN L. WEINER, Ph.D., Patient Advocate
KIMBERLY L. TOPPER, M.S., Acting Executive Secretary
Richard Pazdur, Division Director, Division of Oncology Drug Products
Steven Hirschfeld, FDA Division of Oncology Drug Products, CDER
Call to Order 7
Chairman Victor Santana, M.D.
Conflict of Interest 8
Kimberly L. Topper
Richard Pazdur, M.D. 15
Charge to Committee
Steven Hirschfeld, M.D. 16
Open Public Hearing 25
Review of Ethical Principles as Applied to
Research in Children with Cancer
Eric Kodish, M.D., Ph.D. 25
Questions and Discussion 50
Review of Developmental Pharmacology
J. Steven Leeder, Pharm.D., Ph.D. 89
Questions and Discussion 109
Challenges of Pharmacokinetic/Pharmacodynamic
Assessments in Pediatric Oncology
Clinton F. Stewart, Pharm.D. 115
Questions and Discussion 143
Pharmacogenomic Considerations and
Opportunities in Trial Design
Mary V. Relling, Pharm.D. 148
Questions and Discussion 173
Open Public Hearing 206
What Can Bayesian Methods Do for Us?
Steven Goodman, M.D. 207
Questions and Discussion 225
Newer Kinds of Agents
Edward L. Korn, Ph.D. 242
Questions and Discussion 255
Summary of Experience with Phase II
Victor M. Santana, M.D. 266
Questions and Discussion 286
Summary Statements 311
Peter C. Adamson, M.D. 311
Question 1 - Phase I Design 312
Question 2 - Part 1: Extrapolation 339
Question 2 - Part 2: Product Labeling 335
Question 3 - Monotherapy versus
Combination Therapy 354
Question 4 - Phase II Window Design 360
Wayne Rackoff, M.D. 365
CHAIRMAN SANTANA: Good morning. Let's go ahead and get started.
For those of you that don't yet know where we are, it's November the 28th, and this is an Advisory Subcommittee Meeting of Pediatric Oncology to the FDA. The purpose of this meeting is a continued discussion that this group has had advising the FDA for the last year or so on issues relating to the implementation of the Pediatric Rule.
My recollection is that there has been at least two prior meetings in which -- three, four, how many? This is No. 4. So there's been three prior meetings in which we discussed some specific issues about some diseases in pediatrics and how they may relate to issues in adult oncology and the implementation of the rule. Today the specific purpose of the meeting is to look at study designs in pediatric studies and how those can be used to support some of the indications in the Pediatric Rule.
So, with that, we have an extensive agenda of various presentations, followed by some discussion, and then at the end of the day we will have a few questions that specifically the FDA wishes us to comment on.
So, with that very brief introduction, I want to go ahead and introduce the Committee to itself and to the public.
Oh, I'm sorry, go ahead, yes. Kimberly, go ahead.
MS. TOPPER: This is the conflict-of-interest statement. The following statement addresses the issue of conflict of interest with regard to this meeting and is made part of the record to preclude even the appearance of such at the meeting.
Based on the submitted agenda and information provided by the participants, the Agency has determined that all reported interests in firms regulated by the Center for Drug Evaluation and Research present no potential for a conflict of interest at this meeting with the following exceptions:
Since the issues to be discussed by the Committee will not have a unique impact on any particular forum or products but may affect the entire class of products with all similarly-situated manufacturers, in accordance with 18 USC Section 208(b), general matters waivers have been granted to each of the special government employees participating in today's meeting. A copy of these waiver statements may be obtained by submitting a written request to the Agency's Freedom of Information Office, Room 12A-30 of the Parklawn Building.
We would like to disclose that Dr. Frank Balis, an employee of the National Institutes of Health, has received a waiver from his institution allowing him to participate in today's meeting.
Further, Dr. Wayne Rackoff from Janssen Research Foundation and Dr. Martine Bayssas from Debio are participating in this meeting as industry representative, acting on behalf of regulated industry. As such, they have not been screened for any conflicts of interest.
With respect to FDA's invited guests and guest speakers, Dr. Eric Rowinsky, Dr. Charles Coltman, Dr. J. Steve Leeder reported interests which we believe should be made public to allow the participants to objectively evaluate their comments.
Dr. Rowinsky would like to disclose that he has grants from Pfizer, Abgenix, Diiachi, Enzon, AstraZeneca, OSI Pharm, Genentech, Schering, Janssen, Glaxo, Immunogen, Supergen, Aventis, Bristol Myers Squibb, Eli Lilly, Allergan, MGI Pharm, and Shire; is involved in research with the firms indicated and receives consulting fees from the firms indicated as well as from Pharmacia and BTG.
Dr. Coltman would like to disclose that he owns founder's stock in ILEX Oncology, Incorporated, and is on Bristol Myers Squibb's Advisory Board.
Dr. Leeder would like to disclose that between 1997 and 2001 he has served as a consultant to Abbott and Schering-Plough on issues concerning pharmacogenetics, Hoffman LaRoche, Bristol Myers Squibb, R. W. Johnson, PRI, and Abbott regarding issues related to idiosyncratic drug toxicity, and Eli Lilly on issues concerning pediatric clinical trial design.
In the event that the discussions involve any other products or firms not already on the agenda for which an FDA participant has a financial interest, the participants are aware of the need to exclude themselves from such involvement, and their exclusion will be noted for the record.
With respect to all other participants, we ask, in the interest of fairness, that they address any current or previous financial interests with any firm they may wish to comment upon.
The next thing is, because we have a person hooked in by telecon, it is critical that you all speak directly into the microphones. Just because they have this big holey pad does not mean it's picking up your voice. Please speak directly in. Thank you.
CHAIRMAN SANTANA: Thank you, Kimberly.
What I would like to do now is introduce the Committee for the public record. So please indicate your name and your affiliation, and if we could start with Wayne over here in the corner and go around, please.
DR. RACKOFF: Wayne Rackoff. I'm a full-time employee at Johnson & Johnson, Janssen Research Foundation, and currently working in oncology drug development.
DR. BAYSSAS: Martine Bayssas, an employee of Debiopharm in Switzerland, working in clinical oncology development.
DR. COLTMAN: Charles Coltman, President and CEO of the Cancer Therapy and Research Center and Chair of the Southwest Oncology Group.
DR. BALIS: Frank Balis, Pediatric Oncology Branch, National Cancer Institute.
DR. KODISH: Eric Kodish. I'm at Rainbow Babies and Children's Hospital in Cleveland and Director of the Rainbow Center for Pediatric Ethics, and PI for our COG activities.
DR. SMITH: Malcom Smith, Head of the Pediatrics Section of the Cancer Therapy Evaluation Program at the NCI.
DR. BERNSTEIN: Mark Bernstein, Pediatric Oncologist at the University of Montreal and a member of the Children's Oncology Group.
DR. STEWART: Clinton Stewart, Department of Pharmaceutical Sciences, St. Jude Children's Research Hospital, Memphis, Tennessee.
DR. LEEDER: Steve Leeder, Clinical Pharmacology at Children's Mercy Hospital in Kansas City.
MS. RELLING: Mary Relling, also Pharmaceutical Sciences, St. Jude's Children's Research Hospital in Memphis.
DR. ROWINSKY: I'm Eric Rowinsky. I'm the Director of Clinical Research at the Institute for Drug Development of the CTRC in San Antonio, Texas.
DR. GOODMAN: I'm Steve Goodman. I'm in the Division of Biostatistics in the Department of Oncology at Johns Hopkins School of Medicine.
DR. KORN: Ed Korn, Biometric Research Branch, National Cancer Institute.
DR. GEORGE: Stephen George, Duke University Medical Center, also Group Statistician for COGB, member of ODAC.
DR. BOYETT: James Boyett from St. Jude's Children's Research Hospital, Chair, Biostatistics.
DR. PRZEPIORKA: Donna Przepiorka, Center for Cell and Gene Therapy, Baylor College of Medicine, and member of ODAC.
CHAIRMAN SANTANA: Victor Santana, Pediatric Oncologist from St. Jude's Hospital. I need to inform people that I did not select this Committee. So the representation from St. Jude is purely by chance.
DR. FINKLESTEIN: Jerry Finklestein, Pediatric Oncologist, UCLA, and Long Beach Memorial Medical Center in California.
MS. ETTINGER: Alice Ettinger, pediatric nurse practitioner at St. Peter's University Hospital in New Jersey.
DR. WEINER: I'm Susan Weiner. I'm a patient advocate. I was a parent. I'm President of the Children's Cause.
DR. PELUSI: Jody Pelusi. I'm an oncology nurse practitioner at the Phoenix Indian Medical Center, and I sit today as the consumer representative.
DR. REYNOLDS: Pat Reynolds, Children's Hospital, Los Angeles, Hematology/Oncology.
DR. COHN: Sue Cohn, Children's Memorial Hospital in Chicago.
MS. KEENE: Nancy Keene, patient advocate.
DR. ADAMSON: Peter Adamson, Children's Hospital, Philadelphia; Head of the Children's Oncology Group, Developmental Therapeutics.
DR. HIRSCHFELD: Steven Hirschfeld, FDA Division of Oncology Drug Products, CDER.
DR. PAZDUR: Richard Pazdur, Division Director, Division of Oncology Drug Products.
DR. GOOTENBERG: I'm Joe Gootenberg. I'm the Center for Biologics, Oncology, Pediatric Oncology.
CHAIRMAN SANTANA: Okay, well, thank you to everyone and welcome.
I will now pass on the microphone to Dr. Pazdur for some very brief welcome comments, and then on to Dr. Hirschfeld.
DR. PAZDUR: As Victor stated, this is the fourth meeting of the Pediatric Oncology Subcommittee, and I just want to reiterate that this emphasizes a commitment of the Division to looking at development of pediatric drugs in oncology.
Not only have we had four Subcommittee meetings, but we have had an effort to promote pediatric oncology by hiring additional pediatric oncologists within the Division to, hopefully, review pediatric applications and form an improved dialog with the oncology community, both the pediatric oncology community, academic pediatric oncology community, as well as the industry representatives that deal with pediatric drugs.
I think that this will hopefully be a longstanding discussion and a longstanding Subcommittee, and we have plans for this to be a continuing effort to have this dialog with the pediatric community in general.
Really those are the only brief words that I have, and I will turn over the podium to Steve. Steve?
DR. HIRSCHFELD: Good morning. I'm Commander Steven Hirschfeld, U.S. Public Health Service, Pediatric Oncologist.
I also wanted to welcome everyone to this meeting. What I wanted to briefly touch on is rationale for why we are here at this particular meeting, which has to do with the application of the Pediatric Rule, but to understand the Pediatric Rule, we have to put it in context.
Clinical research, at least that's been recorded, has been occurring for the past approximately 2,400 years. The first clinical experiment that was recorded, at least in Western history, was Hippocrates, who was examining a man with a broken skull and poked around, in anticipation of Sheldon Penfield, and wanted to see which parts of the body were going to move. Hippocrates realized the implications of what he was doing and made a statement that: Research should only be carried out to the benefit of the patient.
That was then codified in the term "primum non nocere," "do no harm," but Hippocrates didn't actually write his work. It was written by his students, and that's probably why it came out in Latin.
So children have been involved in clinical research over the centuries. Children were the first participants in vaccinations, in intravenous therapy for cholera about 150 years ago, in x-rays, in the development of prophylactic antibiotics, in endotracheal intubation. In fact, almost every procedure -- general anesthesia, isolation wards -- has been first done in children.
Children were also involved in the very first randomized clinical trials in oncology at the National Cancer Institute in the 1950's. Also, the first studies which examined the use of adjuvant chemotherapy were done on children. There are many other firsts that were done using children.
Nevertheless, the therapies to address the health needs of children did not follow directly, and it was through the use of public events involving children that the founding principles of modern food and drug law were established. It was antitoxin contamination, which led to the passage of the Biological Protection Act in the United States, which was the first one in recorded history, at the turn of the 20th century. All these phenomena are 20th century phenomena and not prior.
It was a scandal that was published in Collier's Weekly of the poisoning of children through opiates in an elixir which was meant to calm colic which led to the establishment and the signing of the Drug Protection Act by Theodore Roosevelt.
It was the poisoning of children and adults by tainted sulfanilamide that led to the passage of the Food and Drug Act, the initially act, which established the principles of safety and labeling.
Then in 1962, again, a scandal involving children not in the United States, but globally, where children of mothers who took thalidomide had birth defects, led to the passage in 1962 of the Amended Food, Drug, and Cosmetic Act, and that established the principle of efficacy.
Although children played a critical role in the process of establishing protections for research, the children were not the direct beneficiaries. Many therapies were not made available for children.
May I have the next slide?
The paradigm for pediatric development was as an afterthought to adult drug development. There are typically pre-clinical studies which led to adult studies and then would follow with pediatric studies. What we would like to examine is a new paradigm to see if there could be development which could be concomitant pre-clinical studies which will lead to concurrent development of adult and pediatric studies or perhaps, if conditions warranted, pre-clinical development, which would lead immediately to pediatric development and then adult studies.
Next slide. It wasn't until the last decade that there were federal initiatives put into place to attempt to correct the inequities which were termed "children as therapeutic orphans" during the 1970's and 1980's. In 1994, there was a rule enacted which attempted to lower the threshold for achieving pediatric labeling for drugs, and it was a voluntary effort.
Unfortunately, it didn't have the anticipated success. So two other approaches were established. One was a section of the Food and Drug Modernization Act of 1997, which established an incentive program.
Next slide. And this incentive program, which we will not discuss further today, but just to place today's discussion in context, allows an extension of six months to the exclusivity, whether it is patent or some other licensing exclusivity, to a company. It has been a wonderful success in terms of stimulating pediatric research.
There is no linkage between the adult and the pediatric indications. It is a voluntary program. A sponsor may only proceed when receiving a written request from the FDA, and in oncology, as we have discussed publicly previously, it can be approached by a Phase I, followed by portfolio Phase II studies.
In the Oncology Division we have issued over 20 written requests of which approximately 17 have studies already underway. We consider this a highly successful effort and commensurate with the overall effort in the FDA, of which there are over 200 written requests, hundreds of studies underway, and there are over 40 products that have been granted exclusivity extensions.
In addition to this incentive program, there's a mandate program -- advance, please -- which is the 1998 Pediatric Rule that states that pediatric studies must be done if the indication for an application under review can be found in children and a therapeutic advance or widespread use are anticipated, and both these terms are explained in the preamble to the regulation.
It applies to drugs and biologicals. If the indication does not apply to children, a waiver can be granted. The default position is that the rule is applied and a waiver must be requested. However, recognizing the biological differences in some classes of disease, there are automatic waivers, and, in particular, now in oncology the diseases which are common in adults such as breast cancer, colon cancer, lung cancer, prostate cancer, receive automatic waivers.
What we examined in previous discussions is the circumstances of when the rule may apply. What we would like to examine today is, how should we apply the rule? So the general question for the Committee is: Given that circumstances that trigger the rule have already been invoked, how should the 1998 rule be applied? That is the particular charge which we would like you to address during the course of the day.
I would like to conclude with some acknowledgments. I would like to, first of all, acknowledge our Division Director, Dr. Richard Pazdur, who when he came to the FDA had an approach to pediatric oncology akin to W.C. Fields' approach, but since then has become exemplary and outstanding in his commitment to the development of therapies for children with cancer.
I would also like to acknowledge some visitors who have come from Europe particularly for this meeting as observers: Dr. Gilles Vassal from the Institut Gustave Roussy; Dr. Francesco Pignatti from the European Medicinal Evaluation Agency, the EMEA, and Dr. Anne Mathieu Boue from the Agence Francaise de Securite Sanitaire des Produits de Sante, which is the French equivalent of the FDA. We'll see how that comes out in the transcript.
And, lastly, but not least, I wanted to acknowledge Kimberly Topper, who did the administrative equivalent of leaping onto a galloping horse and has been refreshingly pleasant and professional, and because of her own personal qualities, has brought to the preparations for this Committee a depth of understanding and a particular sensitivity that we all appreciate. So we thank you, Kimberly.
CHAIRMAN SANTANA: Steven, I also do want to echo some of the comments that you made that I think I personally, from the other side of the table, have noted a greater sensitivity and interest on the part of the FDA in addressing issues in pediatric development. So I also want to publicly recognize the efforts that you and Richard and the other members of the Agency have to this revival of interest in this area that I think ultimately will benefit both our patients and clinical investigation.
With that, I want to go ahead and dedicate some time to an open public hearing. If there is anybody in the audience that wishes to address the Committee, I invite you now forward. There's a microphone here in the middle of the room. Please identify yourself by name or any affiliation or any conflict of interest. So I now offer that opportunity to the public.
CHAIRMAN SANTANA: If there's nobody in the public who wishes to address the Committee, we will move forward.
The first point of presentation and discussion, Eric Kodish from the Ethics Center for Children at Rainbow's, will address the Committee. Eric?
By the way, there are some handouts that you should all have in front of you.
DR. KODISH: Good morning. Thank you. Thank you, Dr. Santana. Nice to see you again, and thank you to the FDA and especially to Dr. Hirschfeld for inviting me here.
I am really pleased to be able to start a session with some discussion of ethics. I think it is where we should start. My remarks are going to be divided into four parts. I am going to talk at the beginning about some basic principles of pediatric ethics. I am going to share some data with you from our own research on informed consent in pediatric oncology. I am going to move then to some specific discussion of research ethics in pediatrics, and conclude with a few comments on the special ethical issues that relate to developmental therapeutics.
There we go, slide show. And the next slide.
The Academy of Pediatrics, in its vision statement, suggests that children are both vulnerable and symbolic of our legacy. This slide I think captures a lot of what we need to think about when we think about pediatric ethics. So I show it.
The next slide shows some of the principles of ethics, and beneficence is listed in red because, in my view, pediatric ethics really comes down to beneficence. That is not to undermine the importance of the other principles. Justice, for example, is often overlooked when we think about medical ethics and pediatric ethics. An example of the importance of justice would be the reason that we are here today, which is, in my view, to give voice to children who are a politically avocal audience. We need to speak for them.
So to advocate for justice for children, we need to get access for new medications for children. As Steve's original comments pointed out, children have long been left behind in this process.
When we think about beneficence, we need to think about both individual beneficence and collective beneficence. When I say "collective beneficence," I mean for the good of all children or of all people.
The next slide shows a little bit of what I think is the difference between medical ethics when it comes to general ethics and pediatric ethics. Over the past several decades, autonomy has become the dominant principle in adult medical ethics that's derived from this principle of respect for persons, and it is translated into an emphasis on informed consent as being really the key issue. So that competent adults can make their own decisions.
Pediatrics ethics is different because beneficence I think needs to be the dominant principle, and this is individual beneficence. The best interests of the child is really the first and most important issue in pediatric ethics, and I would argue in research ethics for children, too.
The next slide. So the geometry of pediatric ethics looks a little bit different than the geometry of other bioethics. What you see in this slide is my view of that with the child at the top of the triangle, quite intentionally, and often a triangular set of relations between parents and clinicians. For our purposes today, you might want to substitute the word "investigator" on your right for clinicians. I think many of those who are doing individual drug trials in children are clinician investigators, and parents are an important part of this decisionmaking process. So we will refer back to this slide as my talk goes on.
The next one. There are two common questions that come up in pediatric ethics committee meetings, I think in IRB deliberations, and in general ethics discussions in pediatrics. The first is, what should we do? Should a particular therapy be given? Should a particular child be enrolled in a study? That is a beneficence sort of decision. The other sorts of questions that come up are, who should make a consent decision? That is a sort of a procedural one on the bottom versus a substantive question on the top.
Again, having revealed my bias early, I think beneficence is really the most important question; that is, what should be done for a particular child? Sometimes we need to recognize that the answers are incompatible. That is, the question of what should be done, what is best for a child in the view of a physician, an ethics committee, a nurse, may be different from who gets to decide. I think too often we come down on the question in pediatrics of who makes the decision. So I would like us to focus more often on this question.
Next. This is an important point. Informed consent in some ways is a meaningless concept in pediatrics. It is really parental permission and assent, which I will mention in a bit, that come down. The autonomous authorization of adults on their own behalf is more robust -- that is, morally more relevant -- than permission for children by a proxy or surrogate, and the Academy, again, says this very nicely, I think. The pediatrician's responsibilities exist independent of parental desires or proxy.
Next slide. So parental permission is not the moral equivalent of informed consent. It is important, but it is not the same thing as consent. Surrogate decision is less authentic, and that is the primary reason. People, in making proxy decisions, whether it is a parent or a physician or an IRB, may want to think about two different standards of decisionmaking, best interest versus substituted judgment.
Next slide. Substituted judgment, again, is derivative from the respect of autonomy, and it is when a subjective decision is made on behalf of a child; the decisionmaker tries to put themselves in the shoes of the child and make a decision based on what that child would want, as opposed to a best interest, which is more of a mathematical calculation of risks and benefits for the general individual. It is more based on this idea of beneficence.
This is just to say that there are two different theoretical ways of making decisions on behalf of children. I think in practice we use an admixture of them all the time without even recognizing it. Those of you in the audience who have taken care of children with cancer can, I think, probably remember kids that you have cared for where you were trying to make a decision for them, and part of what you were thinking had to do with how that child seemed to be at the time, what they were feeling, what it would be like to be them. That is more of a substituted judgment sort of decision as opposed to maybe the more detached clinical decisionmaking, which is over here.
Next slide. So I alluded a minute ago to the assent of the child, and informed consent pediatrics is really a combination of permission of the parent and assent to the child in cases where the child is old enough to assent. It is important to recognize that for many children with stage four neuroblastoma, for example, they are too young to even be involved in this, and then you are left without this sometimes important part of the equation.
On the other hand, teenagers with Ewing's sarcoma or rhabdomyosarcoma, this is a critically important issue, and I will talk about that in the next couple of slides.
There are two different -- next one, Steve; thank you -- two different ways of thinking about assent also: a clinical definition and a research definition. The clinical definition I am going to show you here is from the Academy's Bioethics Committee, and the elements are very analogous to the moral elements of informed consent for adults. An awareness of the nature of a child's condition is required. The child needs to know what to expect with tests or treatments -- again, this is clinical -- under assessment that the child understands, including an assessment of whether there is undue pressure to accept or assent, and, finally, soliciting an expression of the child's willingness to proceed.
Before you go to the next slide, I just want to point out this asterisk, and the next slide shows that in the clinical context we need to note that no one should solicit a patient's views without intending to weigh them seriously. This is a morally very important issue.
When a child will have to receive medical care, despite his or her objection, the patient should be told that fact and should not be deceived. In fact, many have argued that the child should have an apology coming to them after this is done in the clinical context. An immunization for a young child might be an example of this sort of thing going on.
The next slide is the research definition of assent from the federal regulations, and it is a child's affirmative agreement to participate in research. Mere failure should not, absent affirmative agreement, be construed as assent. This is a critically important slide because it shows the difference between assent in the clinical context and in the research context. That is, in the clinical context it is reasonable to override a dissent of a child sometimes. In the research context I'm not quite sure it ever is.
Next slide. So research is supererogatory; that is, it is an optional thing. Assent/dissent should be determinative in the research context, but not the clinical context. There probably should be veto power for all three moral actors, as I showed on the triangle, in the research setting. For all studies, I think, the older the child, the more ethically justifiable the study is if assent is provided.
Now many of you are probably thinking, well, what about a situation where there's overlap between research and clinical care? Certainly those situations exist. Those are very difficult cases. In those situations I think the real question gets to be, what are the alternatives for that child at that point in time? We will get into that a little bit later in the talk.
Next slide. This is part two of the talk, where I am going to share with you some data that we have been collecting with funding from the National Cancer Institute to look at informed consent for childhood leukemia randomized clinical trials for newly-diagnosed patients. Obviously, this is our most common disease in pediatric oncology, and we have been really grateful to the NCI for the opportunity to look at informed consent in a very rigorous way. I am just going to share a little bit of what is a huge and fascinating dataset this morning.
We have been observing an audiotape in the consent process for children that are recruited to RCTs in the legacy group, the children's cancer group, and we have been interviewing parents and getting reports from clinicians. Here's what we have found in a nutshell:
The trial is well-explained to parents. We have good evidence that, with rigorous coding rules that I won't go into this morning, clinicians are explaining the trials to parents in a great amount of detail. Many parents do not understand their choice about the clinical trial. That means they don't understand that it is even a choice whether or not they go on study or not. About 32 percent of parents don't understand that, and this is an "n" of 108 parents that we have analyzed. It is a pretty big sample. Fifty percent of parents don't understand that their child will be randomly allocated to receive treatment.
Minorities and those in lower social position are at greatest risk for not understanding, and they ask fewer questions during the consent conference. I will show you some data about that.
It is important for this meeting to note that the data may not be generalizable in the relapse context, but I think this is a very serious potential concern for the Phase II window studies, where we have newly-diagnosed children. The reason it may not be generalizable to the relapse context is that parents and kids are experienced with the system by that point.
Next slide. One of the interesting measures we have used is a standardized validated measure for decisionmaking preference that we have adopted for pediatrics. This has been developed in the adult context, and this shows a decision choice that parents make in the interviews that we do with them after the consent process, ranging from up here to the doctor gets to make all the decisions, to down here I get to make all the decisions; in the middle is a completely shared responsibility, and then there are No. 2 and No. 4 variations on that, with shared but more emphasis to doctor up here, more emphasis to parent down here. The reason I show this is that I think decisionmaking is really the critical ethics issue that we're here to talk about today.
The next slide shows data from our 108 parents, and I think it is quite interesting, the No. 1 choice, of course, is a complete equal sharing of decisionmaking, but what I want to point out is to your left on this slide this group of parents who in general prefer doctors to have a little more power in the decisionmaking, and only one parent out of our 108 said, "I get to make the decision." So I think this is a context where parents really do want help from physicians and nurses in making the decision, their hard decisions.
Next slide. This is the number of question data, and what we found in the numbers of questions that parents asked during informed consent is that there is a very wide standard deviation; that is, we have some conferences where there's 150 questions and some conferences where there's 2.
This shows all cases by social class. This is the Hollingshead Index of Social Position for the social scientists in the audience. This is higher social class; this is lower social class, and you can see that there is a real dropoff in minority parents, but also an overall dropoff by social class in the numbers of questions that get asked.
Thank you. The next slide, my statistician suggested that we handle this by looking at the log of parent questions to help narrow that standard deviation, and we have run a ANOVA on that that shows a very significant decrease in the number of questions for all cases as social class goes down.
Parents asking questions is a symbol, but also an important way of getting information for the parents. So I think question-asking is a key variable, and we will be looking in the future at ways to increase the number of questions that parents ask.
The next slide. That is all the data I wanted to show this morning. I am going to get now to some comments on research ethics specifically.
The pediatric research ethics has a fundamental problem with the Nuremberg Code, if one takes it literally. The first precept is that the voluntary consent of the subject is absolutely essential, meaning that the person involved should have legal capacity to give consent.
The next slide shows rhetorically maybe that we have a problem. If the answer is, no, we can't adhere to Nuremberg, then children as a group will suffer. I think the answer needs to be, yes, and then the question is, how can children be adequately protected in studies?
The next slide shows three ways that we can do this. We can respect the principle of Nuremberg and still do pediatric research. We can use parents as surrogates, which I have said is parental permission. We can involve children in the decision, which is assent, and we can provide some societal protection. The most common way of doing this at least is IRB approval. I am not sure that the media provides much societal protection, but some might argue that those sorts of exposes would also provide societal protection. Certainly the government agencies like FDA can provide societal protection, and that is important.
As a research ethics investigator, I would have to say that this is a much more powerful, effective way of protecting children than these two things. That doesn't mean that these are unimportant. I think we need to make these better, but right now is the effective way that we protect children.
Next slide. So how do IRB's look at pediatric research protocols? Well, as most of you know, Subpart D of the federal regulations requires that children have additional protections; that children who are recruited to studies need to fall into one of four categories in order for IRB's to have even the ability to approve protocols.
The fact that a child falls into one of these four categories, or a protocol does, I should say, doesn't mean that an IRB has to approve the protocol, but those of you who have submitted these forms to IRB's know that you need generally to check off one of these boxes.
I think for today's purposes it is 46.405 that is the key category, and this is research involving greater than minimal risk, but has the prospect of direct benefit to the individual subject. IRB's can approve this kind of research if the risk is justified by anticipated benefit to the subject, if the risk/benefit ratio is less than or equal to the alternatives -- again, alternatives is a key -- and if parental permission and assent is obtained.
I just want to point out the benefit to the subject here, because the next slide shows that risk in research ethics always means risk to the subject, but benefit may include several different kinds of benefit. In adult studies all of these benefits can be added together to justify the risk.
In pediatrics we are limited to benefit to the subject, but other benefits may include those to other patients, to society. For example, the number of child life-years that are saved in pediatric cancer is astounding when you add up the benefits of our therapy economically. Benefits to investigators or sponsors are important benefits that may accrue in these studies. So in pediatrics we are limited to thinking about these benefits if we are going to follow the regs.
Next slide. This is probably the most important slide in my talk, and it is maybe the simplest slide, but I think what we are talking about is balancing the best interest of the child-subject against science which is there to benefit others. If we find ourselves ever getting to the point where this end of the teeter-totter is getting a little bit heavier than this, and the best interests of the child aren't represented as the most important feature, then I think we are getting into trouble. So if I could propose a line that we ought not cross, that is the line.
The next slide. This is part four of the talk, which gets to some of the specific issues in developmental therapeutics. Phase I oncology studies in general and in children, I think, for these studies there exists a controversy over therapeutic intent. I think that most of the developmental therapeutics community truly has therapeutic intent. I think that most practicing oncologists have therapeutic intent when they deliver therapy in a conventional Phase I study.
There are, however, IRB members, medical ethicists out there who hotly dispute the idea that there can be any therapeutic intent when the chance for benefit is 5 percent, 10 percent, those sorts of things. I think that we are best to stay away from language of intent because in ethics I think intent is always a difficult topic. It is hard to know how to objectively evaluate intent. So I avoid this controversy by moving away from it.
I think another one we want to move away from is commensurate experience. There is a category that some of you know, 406, where commensurate experience is allowed to justify research with a minor increase over minimal risk if there is no prospect of benefit to the subject. Now I am talking like a regulator, but you need to be conversant with this language.
I think that this category doesn't really fit Phase I research, and there is a tendency for this sort of thinking to creep into what should be 405 research; that is, research with the prospect of benefit to the child. It is not a valid justification that children have already been through chemo, so that it is okay to expose them to one more agent. If anything, I think, quite the opposite, we need to protect children from that sort of mentality.
So prospective direct benefit to the child is the ethical and the regulatory key to Phase I studies. I think there are problems defining benefit, and I think it needs to be more than a tumor measurement. I think strategically one of the reasons that IRB members and ethicists are concerned that 5 percent or 10 percent isn't enough to be benefit is that they don't always see the other sorts of benefits that children may get from participating in these studies.
I think the alternatives is a key feature, and the next slide shows some of the alternatives that relapsed patients with a poor prognosis and their parents may face: a Phase I option, an option for alternative or complementary medicine, and the option of hospice care. This is not to say that these are separate pathways. There is the potential for combination therapy, if you will.
The next slide, again, just briefly shows that I think for many individuals these are pathways to hope; these are ways of looking for hope, and I think hope is a clinically, ethically, fundamentally important issue here.
Next slide. So subject selection is not a controversy for Phase I in children, unlike studies that some of the PPRU's may be doing, where you can decide for a new antibiotic whether subject selection should be in a sick group or a well group, and you can talk about financial reimbursement. I think for Phase I cancer studies we are not dealing with the subject selection controversy. I think it qualifies as research with the prospect of direct benefit, and I think the potential for benefit mitigates, but it does not eliminate, the need for protection from research risk. I don't think we are going to be able to get around this. We need to view these children as needing some sort of protection from research risk.
The next slide has a few comments about alternative medicine. I have great concerns about the vulnerability of children to alternative medicine practitioners, and I think that it is a very prevalent phenomena, obviously under-recognized. It is harder to find exactly what alternative medicine is.
I sit on a Task Force for the Academy on Complementary Alternative Medicine, and we have spent lots of time trying to figure out exactly how to define CAM. It is a very difficult issue to define. There are differences for kids, though, in that, again, this is parents generally giving the substances to their children that may not be innocuous substances. We have an obligation, as Steven said, to primum non nocere, to prevent harm first. We need to study these, and I think we are doing a better job with that now. Most importantly, we need to communicate with families at the time of informed consent for these studies about whether and what sorts of alternative treatments these kids are taking.
Next slide has a few comments about hospice. It is not incompatible with a Phase I study. I think we need to be very proactive. Hospice care is underdeveloped in children, and we need to advocate for the benefits of hospice care for children with cancer. I think most ethicists would expect that Phase I investigators and hospice docs would be sort of separate sorts of individuals, but I don't think that necessarily has to be the case.
Hospice, when done right, I think, rejects the idea of a right way to die. It allows for each child and family to have the unique circumstances. I think it needs to be part of the consent process for Phase I studies. I think that is a responsibility that we have to dying children.
Phase II window designs, as I told Dr. Smith earlier, I look forward to the conversation on this. I think the subject selection here is more controversial. We need to define how poor is poor prognosis, and context here is everything. That is to say, a newly-diagnosed patient and their family is going to view the opportunity to be in a trial very differently than someone who has been through therapy and relapses six months or a year later.
Phase II windows also qualify as research with the prospect of direct benefit to the subject, but it may not be as good as the alternatives, that is, multi-agent therapy, in allowing IRB's to approve this sort of research. IRB's should not be approving this research if they don't think that it is as good as the alternatives, that risk/benefit ratio I showed earlier.
I think that the new therapeutic paradigms may change the ethical acceptability of these studies. The potential for synergy between some of the newer approaches and the older approaches is something that, as an oncologist, I think has a terrific amount of potential. So I don't want to eliminate the possibility of Phase II window designs as the science changes, but I think that there are some serious problems.
Next slide has my conclusions. The first is that good ethics starts with good science. That is to say, if the science is bad, if it's not an important issue, if it doesn't have the potential to help children in this context, it is unethical; it's a waste of resources; it's putting children at risk. There are a lot of reasons that make it unethical.
That does not mean necessarily that good science is inherently ethical. I think that some research with tantalizing potential may need to be rejected on ethical grounds.
The accelerated drug development research I think needs to proceed, but I think we need to be very careful to build the system in such a way that we get long-term follow-up data, that we need to use stopping rules in a careful way, and DSMB's, and that we have an obligation from justice, the principle of justice, to proceed at a faster pace.
The next slide, and final one, makes the paradoxical conclusion that children are both vulnerable subjects in need of protection and a neglected class that needs better access to the benefits of research.
Thank you for your attention.
CHAIRMAN SANTANA: Thanks, Eric.
We now have plenty of time for questions for Eric and discussion.
DR. HIRSCHFELD: First, I want to thank Rick for that thorough and thought-provoking summary of the key issues.
I would like to note that the regulations that Dr. Kodish cited apply only to Health and Human Services-funded research, but I will also let you know that an adaptation of those regulations has been undertaken by the Food and Drug Administration; that they have been available for public comment over the last several months, are now in the process of being finalized. So that all research that comes under FDA regulation would have the benefit of having these regulations apply to them.
I have a question for Rick, though. Another triangle that is often stated in the research paradigm is that there are three components to participate in research. There's the risk, the benefit, and the consent. Ideally, those are all vested in the same person, but in pediatric research the consent is often taken out in a formal context, although there are assent procedures, and the benefit may be taken out, too. So the child participating in a study is left only with the risk.
I would like to ask you if you could comment at any time during the course of the day's discussion, the earlier in the development cycle that we are looking at a particular potential therapeutic product, the less we know about the risk and the less we know about the benefit. If you could give us some guidance as to how we might approach these difficult issues in examining trial designs?
DR. KODISH: I think risk and benefit both need to be in the equation. So I think with a novel approach one needs to -- and when I say "one," I mean both IRB members and parents and the older kids -- needs to take into account the numerator and the denominator of that equation, if you will. So the fact that the potential benefit to the child is also an unknown quantity needs to be factored into the decision.
I think an approach that looks only at the risks without looking at the potential to benefit that child is an impoverished way of looking at the situation. I think also that benefits need -- and this is where it gets tricky between the substantive and the procedural issues, if you will -- benefits in some sense can only be defined by that child and their family, because they are unique in the situation. So the meaning to that individual family unit of going on an experimental protocol is something that I am reluctant to say IRB's or investigators get to make those decisions.
That is why you need informed consent permission/assent as part of the process. It can't just be that purely objective thing. So I think it is a two-part process, and I think the sequence that we currently use is right, and sequence is important here. That is, IRB's have to approve this first before it gets to parents and kids to make a decision on.
Does that help?
MS. ETTINGER: I wanted to make a comment and thank you for considering the ethics of children. As a nurse and as a clinicians here, I think we know that we have always tried to include the children in age-appropriate language to explain whatever is happening to them, including their treatment, their side effects, what the disease is about. It's always been a struggle for all of us who have ever been in that position.
I have also noticed through the years that there are many more handouts and resources available for that. I think that is really important, many of them having been developed either on a local level or on a national level. I think that that has made it a little bit easier.
I have also noted that culturally, as well as socioeconomically, but culturally there are some cultures that really put obstacles in the way of the clinician in terms of discussing what is going on with their children. I have found that through the years that that has been an issue that we work with, particularly I have to say nurses because that is my discipline, but particularly to try to overcome that barrier.
I also just want to comment about my experience with hospices. I think that that is something -- I don't know how to overcome that barrier, but I have found that many hospice programs will not allow children to be in the hospice program if they are undergoing Phase I therapy. I think that that's to the detriment of the family and the child. If they are on any kind of medication or whatever, they can't, not even hyperalimentation in some cases.
So I think that there are many obstacles that we really do need to overcome.
DR. KODISH: Yes. I would like to follow up on that with just a very practical point. My talk suggested that hospice care needs to be part of the discussion for informed consent, but if you are in a context where hospice care is not possible, you need to do your homework first, obviously, before informed consent, and be sure that hospice care is a real possibility for that child.
Again, I think we need to advocate as a community of pediatric oncology care providers that hospices need to change these rules. I think many of those situations are adult hospices, and I think the more developed pediatric hospices are pretty comfortable with the idea of children getting anti-neoplastic therapy or supportive care and still get in-services from hospice.
DR. GOOTENBERG: Just in terms of definition, from my point of view, in terms of the term for hospice, in that case I don't think that if hospice is failing or if someone is in a situation where hospice is not yet really to where alternative care covers a bigger --
DR. KODISH: Right, or hospice philosophy, care, but, agreed; point well taken.
DR. GOODMAN: I thought it was a wonderful presentation.
DR. KODISH: Thank you.
DR. GOODMAN: I am going to ask a sort of provocative question, which won't necessarily reveal what I actually think, but I just wanted to get your thoughts. I will pose it sharply: How can you say that it is ever -- no, I don't want to say "ever" -- ever more ethical to give an untested and unproven therapy in a clinical setting as opposed to an untested and unproven agent in a research setting?
I mean, the alternative, one alternative you didn't actually explicitly write there -- you had the Phase I and hospice and CAM -- was giving a therapy that, in fact, had never been adequately tested in a controlled way. Its track record had been established in uncontrolled settings because it was impossible to mount administration in a research setting.
So I would like you to -- I mean, this has often been commented on in adult situations where doctors say, "It's permissible for me to give the agent to all of my patients, but I have to get special permission when I want to give it to half my patients." So I would like you to comment on that because you focus on the risks of research, but, of course, there are risks of giving agents in a clinical setting, in the absence of research as well.
DR. KODISH: Yes, I think the point that you're getting at, I have two responses. One is that, to sort of take your side and the way the question was framed, maybe that ought to be --
DR. GOODMAN: I didn't tell you my side.
DR. KODISH: I know. But maybe that ought to be considered alternative medicine. That is even more provocative perhaps, but maybe that is the category we ought to shift that into. That's the clinical trial lists answer to the question.
I think the other answer is probably better, and the concept actually was coined by my colleague, John Lantos, about neonatology research, and this is the inclusion benefit, the idea that children are benefitted from inclusion in clinical trials and the things that Dr. Murphy of Children's Memorial, things like that, have written that say that perhaps the fact that kids in studies do better needs to be part of the consent process.
I personally don't advocate for that in big Phase III randomized trials like the data I showed you for leukemia, but I think in the Phase I setting, as we get more and more experimental, in some ways it becomes safer and safer to be in the research context and less safe to be in the clinical context.
I think that the other key issue to talk about here is hope. I think what is being administered, whether it is in the research or clinical context, in many cases is high doses of hope.
DR. PRZEPIORKA: Thank you. A very nice talk. I just had one question I wanted to start with, and that has to do with, when would you trigger assent? I mean, according to current guidelines, it is when it is appropriate. How does one determine when it is appropriate?
DR. KODISH: Funny you should ask. In about two weeks I have a research meeting on pediatric research ethics where we are going to tackle that specific issue and look at some of the geriatric ethics literature, which has looked at people whose competency level is declining, with the thought that there is sort of some analogy there. That meeting is taking place in two weeks. So I'll probably have a better answer for you in a month.
At this point I would say that we don't want to overregulate assent. We certainly don't want to put a specific age on it. We need to rely on clinician investigators to use good judgment, because some children who have had cancer diagnosed at eight and then relapse when they're ten, for example -- I am talking about the hard age range -- will be on sort of accelerated developmental trajectory where they have the maturity of a 17-year-old and should be making that decision primarily themselves. But others, when their disease relapses, they will regress and become infantile almost in some cases. Certainly putting any specific age on it I think is going to be a problem.
So I think we need to continue to push the idea of assent as a concept without putting regulatory specifics on it. Many IRB's, I think, around the country now are getting more proactive in requiring assent forms, which is a very interesting move in pediatric research ethics. My personal opinion about that, as a PI, I hate the idea that we need to do more paperwork, but symbolically I think it is important because it requires a second signature; it requires that the investigator at least pay some attention to what is going on in the child.
We have data from the big study I showed you that was just accepted in Pediatrics -- it will be published sometime in the next year -- that looks at these cases and looks at what happens when the child is in the room for the discussion, the assent process. What I can tell you about that data is that children are not asking very many questions about the research. They're asking, "Is my hair going to fall out?", "When am I going to go back to school?" Those are the sorts of things that kids themselves want to know.
CHAIRMAN SANTANA: I also want to echo some of what Eric said; that I think one has to be careful in this whole process of applying a regulation regarding to assent too strictly because there's so much variability in the patient population, et cetera, their understanding and comprehension.
But I think one way to view it is that it is more of an information-sharing process, and then what level of information is shared is the critical factor. But there should always be some sharing of information between two or three parties. It is that level of information and that degree that I think defines when you trigger a written document or when you trigger a fast, hard rule. So that is one way that I have kind of tried to address this issue of assent in my own interpretation.
One thing that you did not mention that I would like for you to comment is this concept of viewing assent as a respect for the child in terms of his or her moral development by allowing them to participate in that information-sharing. Then when they become adults, you truly then have the respect for the person in terms of their autonomy. So it is an early process, I guess, of moral development, of weighing in the judgment of that individual, so that ultimately when he becomes an adult, he will have that experience or that capacity to make those judgments.
And you didn't comment on that. If you wish to comment, I would appreciate it.
DR. KODISH: I'm not sure I can say it any better than you just said it, Victor. The idea is that the 18th birthday is not a magical day when all of a sudden someone wakes up as if an epiphany has occurred and they're an adult. If we don't give children the ability to develop their decisionmaking capacity, we are not nurturing them. So I think you said it right.
CHAIRMAN SANTANA: Do you have a comment, Peter? Paul? Where's Paul?
DR. ADAMSON: If Mary's here, we'd have a group.
DR. KODISH: She is right across the table.
DR. ADAMSON: I have two unrelated comments. The first has to do with consent and assent, and I think an area that we as pediatric oncologists have to do a better job. Trials are becoming increasingly more complex with more biologic end-points, pharmacologic end-points. I think we need to do a better job in separating out for both parents as well as children what is of direct benefit to the child and what component of research is, in fact, of no direct benefit to the child.
In Phase I, I think it is very clear-cut that, when we administer an investigational drug, it is with the prospect of direct benefit to the child. However, when we obtain pharmacokinetic sampling, there is no direct benefit to the child.
Oftentimes, although I think we make it clear in Phase I, oftentimes we present studies as a package deal to a family. I think we have to be much clearer, saying, "These are the tests that are important for your child's safety. This is a drug and the risks that we think are going to be of direct benefit, and these are areas that are going to help us learn that in most circumstances are minimal risk or a minor increase over minimal risk, but that are truly a pure research component that's of no direct benefit."
I think that is where assent becomes highly critical for a child, because those components should be made very clear to a child, that this is optional; that you don't have to sit here for a day and have your blood drawn.
Having said that, in my experience, and I think the experience of others, most children want to help other children, and they are going to agree to do it. But they should absolutely have an affirmative assent that they are willing to do that. I think they derive benefit when they are given that component of assent.
My other comment was again related to assent, but one area that I would ask you to clarify. That is, assent is virtually always required for research, whereas it may not always be required for clinical care.
I think in pediatric oncology it may be that the two aren't always separable, and the upfront Phase III randomized trials for children with leukemia serve as a good point. Oftentimes what a child will experience in a randomized trial may essentially be no different. Independent of what arm they go on, they're going to experience leukemia therapy.
Yes, I think it is important to gain assent, but I don't think in that circumstance necessarily that the assent should be binding. In other words, if a child doesn't assent to go onto a trial, when it is an upfront trial with a prospect of cure, I think it is the parents' consent that will carry the day. We have to be careful saying, well, we're going to get your opinion on this; we're going to weigh it. But, ultimately, for a Phase III randomized trial, it's more, as Victor said, I think more of an informational meeting and to answer questions, and not to gather assent; whereas, in a Phase I study I think assent, in fact, carries a lot more weight and in many circumstances a child's assent may overrule a parent's consent.
DR. KODISH: Thank you. Very interesting comments.
I am going to get to the second part of your comments first and say that I think it is in the best interest of the child to get leukemia treatment, going back to the Chad Green case that many of you will remember from 20 or 30 years ago. I think that we do have a clinical obligation to act in the best interest of the child that overrides assent.
For standard therapy for ALL, I think we need to be very careful with the distinction between randomized clinical trial and standard-of-care ALL treatment for kids. I am sympathetic to the point of view that says research is the best treatment for children with cancer.
Personally, I believe that, but, as a matter of public policy, as a matter of the appearance of impropriety or conflict of interest, I think my vote would be to separate those out and say that, if a child dissents to randomization -- let's say it is a 17-year-old who understands randomization and says, "You know, standard therapy is embedded in May of 1961 and that's what I want." I think we need to give them that. I really do.
I don't think it is a huge loss to the research enterprise, to the best interest of other children, and I think it is a matter of respect for that child. I don't think it happens real often.
To reinforce a couple of comments you made in the first part of your remarks, this idea is referred to as bundling of treatment and research issues by IRB's sometimes. Some IRB's like it when these things are bundled, and others like them to be disentangled. I agree with you, it is best to try to disentangle them.
I think the issue of risk for that non-therapeutic component is key, and I think a PK study is clearly minimal risk. But if you are talking about an extra tumor biopsy, it is an order of magnitude higher, and the risk needs to be a key part of that.
Then to maybe put into other terms what you said about the kids themselves, I think what we want to do is foster altruism for those children. I think, as you said, many children are capable of it, and we want to provide a context where they are able to express that altruism in a way that also allows them to say, "No thanks, I'd rather be on the Internet."
DR. GEORGE: I had a question, I guess following up more on this assent issue. You presented data, but wasn't clear from that, did you ask any of the children what they understood about this process?
DR. KODISH: Unfortunately, no. Our study design was such that we interviewed parents, but we don't have -- we have observation of children during the time that they are in the room. We know the sorts of questions that they are asking their doctors and their parents, but we haven't interviewed kids.
DR. GEORGE: But it is a relevant point for what we are discussing, I think, with respect to assent. It would be very nice to know what kind of things the children are understanding and at what developmental stage.
DR. KODISH: Yes, and it is going to be in my competitive renewal application.
CHAIRMAN SANTANA: Dr. Patrick Reynolds.
DR. REYNOLDS: You present the dilemma that all of us face in Phase I trials, which is the prospect of benefit, which in the initial dose escalations is arguably extremely slim. At the same time you present the concept of hope.
Now given the tight linkage between the child and the parent, what I am not hearing, though, is that this prospect of hope for the parents on a Phase I trial is actually a component of benefit. It's not a benefit to the child, but it is a benefit to the parents. So can't that be part of the equation?
DR. KODISH: I mean, you get to a really interesting point in pediatric ethics, which is, do we have a narrow definition of best interest or do we have a more broad definition of best interest? Those of us who would focus on a narrow best interest definition would say that it is only the child, that we can somehow surgically remove the child from the family unit and view them as a separate entity, but, in fact, I think you're absolutely right, children feed off parents; parents feed off children. It would be, I think, incorrect to try to have a narrow best interest definition.
I think it is permissible to include the hopes of parents, especially for younger children. I think when it gets to older children, we need to have them a play a more significant role.
I don't want that to happen at the expense of suffering for that child, and that is why I say that Phase I investigators need to do a conscientious job in the consent process to talk about palliative or hospice care as one of the alternatives.
Recognizing that many parents will come into that conference not wanting to hear a word of it, and that can get ugly, but I think it's important to have that ritual of informed consent.
CHAIRMAN SANTANA: Nancy?
MS. KEENE: That was a great talk, Rick. Thank you.
DR. KODISH: Thanks, Nancy.
MS. KEENE: It was a pleasure to listen to, and I wanted to thank you for your last bulleted point under "Conclusions" that included the necessity for long-term follow-up data being collected and analyzed.
As you know, because we have known each other for several years, I don't usually use personal anecdotes in settings such as this, but I am going to use one to illustrate the point that I would like to make.
I'm the parent of a 10-year survivor of high-risk ALL who has multiple late effects, none of which have generated a single data point. At the institution that treated us, it only checked for recurrence of disease, did not check for late effects of treatment. So we went elsewhere at the end of treatment.
However, I called the data manager at that institution because, as you know, I am a big believer in clinical trials, and said, "Tell me what information you need from our subsequent health care providers for the trial." And he said, "Just call me back if she dies."
The reason I use -- social skills aside (Laughter), it does illustrate a good point. It illustrates that mortality has been the focal point for quite a long time, and that we need to make a cultural shift, at least for those diseases for which there is a high cure rate currently, to include the concept that late effects, indeed, are part of risk, and that acute risk does not define risk.
You could argue that one cannot make an informed consent if you're only notified about acute risks, and if we only collect information on acute risks, we are not collecting the information we need to give people the data they need on which to base an informed consent.
So I was the Chair of the first CCG Patient Advocacy Committee and then the Chair of the first COG Patient Advocacy Committee, and we have worked very hard to incorporate mandatory data collection on late effects, with not much success.
We scaled that back -- I've been sick for two weeks, and my voice, I'm losing it -- we scaled that back to a request for at least guidelines for known expected late effects of treatment to be incorporated in all new trials. There is a movement toward doing that. That would at least give families and subsequent health care providers information that they need to get necessary follow-up surveillance in the future.
So it is just sort of a plea on my part to all of you who are involved in development of future clinical trials that we make this a focus.
DR. KODISH: I want to take up one point from that, which is that informed consent in the context of Phase I or Phase II window studies, really I think we have an opportunity to do a much better job. I think parents, from my reading of our own data, are generally in a state of shock, and it is very difficult for parents to make an assessment of the short-term and the long-term issues.
I think in the setting that we are here to talk about today there is the opportunity to really do a terrific job with informed consent and trying to think of some of those short-term and potentially, if things go well, long-term issues.
MS. KEENE: Can I follow up on that? There are a couple of Phase I trials that I have reviewed the information for and helped to rewrite for CCG that are presented to parents of newly-diagnosed kids. There was one for a radioenhancer for kids with pontigliomas. Those families are at incredible risk for presentation of the information and understanding.
As you know, I told you when you were developing that study, I said, I can't wait to see your data, but I'm going to predict that it's parallel universes, that many of the physicians are going to be giving good explanations and many of the families are not hearing, and those that hear, some of them are not going to understand.
So I think that in the few situations in Phase I and the Phase II windows studies where families are newly diagnosed it is going to be very, very, very difficult to get a truly informed consent.
DR. KODISH: Yes, it's apples and oranges.
CHAIRMAN SANTANA: Dr. Cohn.
DR. COHN: I was just going to give a followup a couple of speakers ago to Peter's comment about how sometimes some of these studies are bundled in terms of what is truly research and what is of benefit to the child and what's not.
I just think that one of the things that we have developed, which I think is a much better informed consent form than what we have seen in the past, is something that we have recently developed in our Neuroblastoma Strategy Committee in the COG, which is that we have a biology consent form now that very much separates out what studies need to be done for clinical purposes, such as NMEC, and which studies are strictly research. Then we have checkboxes that parents and children can actually say, "I agree to this," "I don't agree to this," "I agree to the whole thing," "I only want the NMEC done," or whatever.
I think that that is something that probably should be done in more consent forms. I don't think bundling together is necessarily the appropriate way to go.
DR. KODISH: Thank you, Sue. I think consent forms have gone up a notch, actually, in my estimation, based on the data that I have collected. I came into the study thinking the consent forms were sort of a waste of time, but they can actually be very helpful tools when done the right way.
DR. WEINER: I wanted to thank you, Rick, for a wonderful talk and you, Steve, for leading this discussion today with the consideration of ethics.
I think that, though we are discussing matters of public policy, I think that it is very important to understand how that translates into an individual case. So the question of bundling touches on it, and also the issue of conflict of interest.
When you presented the triangle in your slide, you said that you listed it as clinician, but you said it could be the investigator as well. I just wanted to point out that there are many, many instances in which these roles are conflicted, and there are many things that can follow from that kind of conflict of role, which may or may not be in the best interest of the child.
The fact that we've led off this discussion, the day today with discussion of ethics I think is critical, but I think, to use a cliche, it matters where the rubber meets the road. There may be times when clinical considerations really have to be assigned apart from an investigator's role. I would like to know your comments about that?
DR. KODISH: I wish my colleague, Dr. Shurin was here because she's really developed a terrific expertise in these conflict-of-interest issues. My response is that, to the extent that procedural solutions would help, I am in favor of that. I think having the Phase I investigator be a separate individual from the treating physician makes a lot of sense. I think we want to be careful not to leave behind the value of conflict of interest.
I think it is especially critical for children with cancer. Look, it is not a big market. We're here at an FDA meeting, and at other FDA meetings I have seen, you know, a very interesting confluence of interests around big markets and potential benefits for patients. We are talking about still an orphan disease essentially, and we need to take advantage of whatever sort of academic/industry collaboration we can get, if it is going to benefit children.
I am sympathetic to procedural ways of trying to protect individual children, but I want us to understand the overall context.
DR. WEINER: I'm sorry, it is not that particular conflict necessarily that I meant to refer to, because, of course, it can have to do with the need to enroll more patients on the part of an academic investigator or the need for professional advancement or the need to make sure that a protocol is followed to the letter.
So those are instances in which there may, indeed, be a conflict between the clinical care of the child and what is important for the research.
DR. KODISH: Yes, and I think IRB is really the place where those sorts of things need to be decided because they're so center-specific. So I sort of trust the IRB risk/benefit assessment with risks to the child versus potential benefits to the investigator that you mentioned as being the place where those sorts of decisions are made.
DR. COLTMAN: As a parent with an acutely-ill child with cancer, I should be so lucky as to have my child have late effects of treatment.
CHAIRMAN SANTANA: Wayne?
DR. RACKOFF: Thanks, and I agree this was an excellent opportunity to have a discussion before the facts essentially.
I want to follow up on Susan's question, not because it was formulated as a followup, but it is a linked question and it is taking the conflict-of-interest question to the macro level.
What we are talking about here today is the application of a rule, the rule of law, and the question is, is it necessary, do you think it's necessary to in some way inform families that a study is being done as part of a mandated rule?
It comes back to what I thought was an excellent comment on your summary slide of, we start with good science. It's necessary but not sufficient to make it ethical. We always -- in our informed consents, if it is a corporate-sponsored study, there is a notice of sponsorship and indemnity, and the like.
I do not want to add anything, believe me, to informed consent documents, but as we think about application of the rule, do we need to think about information about the rule and how to disseminate it?
DR. KODISH: Yes, the reason I hesitate is that it is a very interesting question that I hadn't considered before. My first-blush answer is, no, that that ought not be a necessary requirement of the consent document because it is broad societal policy, I think. The understanding that our society is evolving toward doing better for our children sort of goes without saying. Parents I don't think are going to be, truthfully, all that interested in FDAMA and the six-month exclusivity and the Pediatric Rule and the sorts of trigger language that we have been talking about. I think they are going to be interested in what is best for their children and maybe helping other children, but I don't think there is any need to disclose something that's that ubiquitous.
DR. ROWINSKY: How would you distinguish that disclosure between industry studies or studies done because of other reasons -- for example, a large market where you might do a study first or just investigate or initiate a study when the risk/benefits are basically very similar to the child him or herself? So I don't really think that that needs to be mandated -- I mean, or disclosed.
CHAIRMAN SANTANA: Steve, do you want to shed some light?
DR. HIRSCHFELD: Yes. I think Dr. Rackoff raised a very important point, and it is never the intent on the part of the FDA to mandate a particular study or to mandate a particular family or child to be enrolled in a study, but, rather, to mandate if the conditions are met, that a particular drug be studied. I don't think it should play a role, because if it is a good scientific study and it is appropriate context, in that setting those are the critical factors. I think it would, in fact, confuse people and give a level of imperative that is not there and not intended.
While addressing that -- and I might ask Dr. Pazdur to make a comment, too -- it may sound semantic, but, from our point of view, there's no difference between what is called alternative medicine and any other type of medicine. From our perspective, there are either products that have data that supports a claim or they don't have data that supports a claim. We view all potential therapies as an equivalent universe or an equivalent cohort.
Dr. Pazdur, would you like to comment?
DR. PAZDUR: I pretty much agree with you, Steve. I would be worried that, if this was put in informed consent, that it could be interpreted as some false approval or urgency or need or some priority above other studies, and that really is not the intent of this. I would not want to get that confusion basically into the informed consent that, well, the FDA mandated this study, so, therefore, this is better than any other study. That is not the intent necessarily to create a priority here of trials for children to go on.
So I pretty much agree with the statements that have been made previously.
DR. FINKLESTEIN: Thank you. I have two comments regarding your excellent talk and then a question.
For the audience, I would like to comment a little further on Nancy Keene's statement. Late effects is part and parcel of what we do in pediatric oncology. We spend a lot of effort on late effects. So I would not like the audience to get the feeling here that we ignore it in pediatric oncology. But, in actual fact, it is highly emphasized.
I compliment you on not giving or assigning a percentage to risk/benefit. I have the advantage, as I look around the audience here, to state that I am probably the senior pediatric oncologist in this room. So I remember in the sixties when we were highly criticized for treating children with acute leukemia because the percentage was such that none of them were going to survive. As we well know, survival in acute lymphocytic leukemia over the past few decades has increased tremendously. So I compliment you for not using a percentage.
My question has to do with your informed consent data. You do assign numbers in terms of parents understanding their choice of a clinical trial and not understanding randomization. If we take away the term "parents" and become more global, what is the data for adults in general in terms of their own clinical trials? What is their understanding? Are parents any different when they're parents? Namely, are the adults different when they are parents versus when they are confronting the question themselves?
DR. KODISH: Thank you, Dr. Finklestein. The question of how parents do in terms of understanding the key issues compared to adults has not been studied. We are actually in the process of a very small pilot comparison of our data to a colleague of mine who is doing similar direct observation research in adults who are offered participation in colon and breast cancer trials. We haven't done the data analyses yet.
My suspicion is that there won't be a lot of difference. That is, there are significant barriers, I think, to understanding for adults who are thinking about participation in clinical trials.
The best dataset that I know of on this comes from the Advisory Committee on Human Radiation Experiments done in the nineties under the Clinton Administration, where they did a subject interview study, and there were in that dataset, which is a much larger dataset, significant barriers to sort of type 2 and type 1 errors, if you will; that is, people who are in studies who don't know that they're in studies and people who aren't in studies who think that they are. It is the former sorts of problems that I think are the most significant when they happen.
It wasn't all that common, but when it happens, when someone is in a study and they don't know they are in a study, I think that is a concern.
CHAIRMAN SANTANA: Dr. Boyett.
DR. BOYETT: I would like to follow up on a comment from Dr. Reynolds and ask you about the Phase I trial and the issue that the individual patient should have the prospect for direct benefit. I would interpret that as being the same direct benefit for each trial that is enrolled on a Phase I study.
I wonder if that is the correct interpretation in the setting where a Phase I trial has been completed in the adult population, establishing either the MTD, maximum tolerated dose, or establishing an optimal biologic dose for a particular disease. Then when we begin to start a Phase I trial in pediatrics, the tradition has been that we start at 80 percent of that dose from the adult trial. Does that mitigate the potential for direct benefit for the early cohorts of patients enrolled in a pediatric Phase I trial, and is that a problem?
DR. KODISH: I think that the question of dosing in the traditional Phase I paradigm is a problem in terms of direct benefit to the subject. I think that as we tinker with potentially new study designs that are going to start at 100 percent or 120 percent, or whatever the right level is to begin with, we need to be aware that we are increasing the potential for direct benefit at the same time that we're increasing the risk.
So it gets back to the comments of Dr. Hirschfeld at the beginning of this discussion that there is a numerator and a denominator that we need to try to consider. It is going to be, I think, especially with the wave of new approaches and new mechanisms of drugs that we're looking at, very, very important to do that. I think there is good reason to hope that some of the denominator issues, that the risks will be lower as we get away from conventional dose toxicity relationship issues.
Some of the studies that have suggested sort of choice to the subject, or in this case choice to the subject parent at picking their own dose level, I think are very intriguing in my mind, the idea that lower dosing may be lower in terms of potential benefit for the child. So it is a discussion I look forward to having as the rest of the day goes on.
CHAIRMAN SANTANA: I want to keep the meeting on time, so we will take one last question, and the person on my list was Dr. Bernstein. So you have the last question, Dr. Bernstein.
DR. BERNSTEIN: Well, mine was really more just a comment to say that, coming from a city where there are lots of immigrant populations, the questions become much more vexed at times in immigrant populations whose cultural traditions are very different and whose whole concept of who gives consent and whether the child should even be allowed to participate in the assent process are very different.
DR. KODISH: Right, and I conclude by saying we need to respect that, and that's why this issue of risk/benefit as an objective measure, in my mind, sort of trumps in pediatric ethics. We need to be respectful of those considerations and always do what's best for the individual child.
CHAIRMAN SANTANA: For the record, there's a point of clarification. Donna?
DR. PRZEPIORKA: If the FDA could comment as we globalize drug development, does the ICH have a statement on inclusion of pediatric participants in clinical research?
DR. HIRSCHFELD: There's an ICH document called ICH E-11 which addresses this specifically and has some, what we hope is an appropriate international advice on the ethics and on the consent/assent issues. It is available on the Internet and probably at your local convenience store, too.
But it is a widely circulated document that does address this.
CHAIRMAN SANTANA: Okay, moving right along, the next topic of discussion, I invite Dr. Steven Leeder for the review of developmental pharmacology and as it may relate to the ethics.
DR. LEEDER: Well, I can start off by saying that it is an honor for me to be here. Probably of all of the people in the room, I am the one that is least involved with cancer chemotherapy on a day-to-day basis, which is not to apologize for me to being here, but more to let you know that some of the issues that are arising with respect to dosing are ones that have to be dealt with in pediatric pharmacotherapy in general. The issues I am going to be raising are coming from the broader context of pediatric pharmacotherapy.
To start off, I have tried to make this presentation as concise as possible and yet deal with the major issues. The one that was presented to me is at the bottom of this slide. But what I am going to do is, first of all, just using some selected examples, review some general principles that are related to drug metabolism specifically in children throughout the developmental spectrum and try to raise some issues rather than provide concrete answers, because there may not be any just yet, related to the issue of choosing a dose for the right child at the right developmental stage based on available adult data.
I think it is useful just to refresh our memories as to how dynamic a process the first 15, 16, 18 years of life really is. In particular, we know that in the first year of life there is a lot going on. Weight doubles by five months of age and triples by one year of age. Body surface area doubles by 12 months of age. You need to understand, if there is that much growth going on, of course, there is a rather dramatic caloric expenditure to fuel that. It is estimated that caloric expenditures will increase three- to fourfold over that same time period.
Now the green arrow in the center of the slide and the question mark means that there is a lot that we don't know what's going on, because for healthy children the process of growth and development is largely marked by placing a dot on a growth chart. Then, of course, we head into that period of adolescence where those of us who are parents facing this really don't understand what's going on. The only thing that can be said with any certainty is that our children are far more intelligent than we are at this stage of life.
Whether it is cancer chemotherapy or pharmacotherapy in general, of course, the goal is to find the right dose/response relationship. In the era of therapeutic drug monitoring, the focus was on the right dose to get to the right target concentration. As we moved into the pharmacogenetic era, it was more the right dose for the right patient. As we enter the genomic era, it is really the right dose with the right medication for the right patient to give us the optimum response. Of caws, we often don't get it right the first time, and there's a feedback mechanism that allows us to alter the dose in response to lack of efficacy or excessive toxicity.
So then what are the key determinants that will help us establish what is the right dose effect/response? I am largely going to focus, almost exclusively going to focus, on the issue of drug clearance as it changes through a development because of the impact that this has on choosing an appropriate dose. I am not going to deal so much with the response end of things, although clearly the ontogeny of drug targets and the ontogeny of resistance mechanisms, for example, drug transporters, is something that certainly needs to be studied and is relevant to this discussion.
But I am going to focus on three main elements, and that is the acquisition of functional drug metabolizing enzyme activity. I am not going to discuss transporter activity today. I am going to address the dogma that drug metabolism activity is increased in childhood relative to adults, and going to raise the issue of metabolite shunting, and I will explain this a little bit as we move along.
Just to review the acquisition, the first element of the talk, the acquisition of functional drug biotransformation or drug metabolism activity, the fetus is largely devoid of activities that we would consider to be protecting the host from small molecular weight compounds, whether they be medications or environmental contaminants. There are some members of the cytochrome P450 family that are expressed almost exclusively in the fetus, one example being P450 3A7, where it likely plays a role in DHEA metabolism and maintaining pregnancy. Also, there are some sulfatransfereses agents that play a similar role.
After birth, most of our drug metabolism activities are acquired in isoform and probably tissue-specific patterns of expression. There are some cytochrome P450's, for example, where the onset of expression is measured in days; for example, P450 2C9 and P450 2D6. There are others where there is a little bit of delay, and the onset of expression is timed more in weeks. Then there are some, such as P450 1A2, where activity really doesn't level off until four to six months of age. I will show you some in vivo data that support these claims.
We generally consider activities to peak sometime in childhood and decline to adult levels at some later point. Much of these data have been gleaned from therapeutic drug monitoring studies. We will discuss this issue in a little bit more detail as well.
These are some in vitro data from France. The purpose of this slide is to show that at the fetal neonatal interface there is a transition of cytochrome P450 3A activity, so it's not unlike the switch in hemoglobin that occurs at this same time.
In the turquoise bars is the activity of P450 3A7, measured by a relatively selective substrate, the 16 alpha hydroxylation of DHEA. We can see that levels in the fetus are relatively high compared to after birth. The peak in activity, at least from these in vitro data, appears to be in the first week of life, with a decline thereafter.
On the other hand, the more mature form or the adult form, if you will, of the cytochrome P450 3A subfamily, 3A4, as measured by testosterone 6-beta hydroxylation, which has fallen off the slide, it seems, is relatively low in the fetus. In fact, some of this activity that is observed may actually be 3A7 activity, but after birth there is an increase.
The point I would like to draw your attention to on this slide is the fact that, even at three to twelve months of age, the activity observed in vitro is less than that observed at later ages.
Much of what we know or what we can infer concerning P450 2C9 activity can be drawn from the metabolism of phenytoin or Dilantin, which is a P450 2C9 substrate.
In this particular study, the investigators were looking at the appearance of saturable metabolism; that is, where the clearance of the drug is dependent upon the initial concentration, with lower clearance, slower clearance, being observed at higher concentrations.
In essence, these investigators found no relationship between initial drug concentration and a measure of half-life that incorporated the saturation metabolism that occurs with phenytoin in the first week of life. Between two and three weeks of life, or one and three weeks of life, they did see a linear relationship of saturation between initial phenytoin concentration and this apparent half-life measurement. In fact, at later ages the slope of the line is not as steep, implying that at a given concentration the half-life is lower in older infants. The point being here is that the appearance of satural metabolism, which is thought to be a cytochrome P450 2C9 activity, is acquired over the first two to three or four weeks of life.
Along with that is the fact that on the milligram-per-kilogram-per-day basis, children require higher doses of phenytoin than do adults to achieve the same target serum concentrations.
My last example here is theophylline metabolism as a measure of P450 1A2 activity. Early on in birth, after birth, here on the axis we have post-conceptual age, and to make the math easy, you may want to subtract 40 to get post-natal age. But, early on, the newborn is highly dependent upon renal clearance to remove theophylline from the system. As post-natal age increases, there is a decrease in the amount of unchanged theophylline which finds its way into the urine.
Corresponding to that decrease is an increase in the amount of the 8-hydroxylation product, which is a function of P450 1A2 activity, implying that it is only after four months of age or so that 1A2 activity has been acquired.
Now we are very much interested at our institution in mapping the ontogeny and some of the other pathways, and the one of the ones, one of the pathways we were interested in was cytochrome P450 2D6. I have to say that the data that I am going to show have not been subjected to peer review. They have been presented in preliminary form, this preliminary form, at a number of meetings, but they have not been subjected to peer review.
But in this case what we are doing in a population of healthy newborns is to map the develop of the P450 2D6 pathway, and it turns out we think we are also seeing some developmental changes in a cytochrome P450 3A pathway as well. Later on, I can go through the ethical considerations of doing such a study in healthy infants, but for the next few slides we are going to focus on this yellow metabolite, which is an OD-methylated product of dextromethorphan, the DM component of cough and cold remedies, as a measure of P450 2D6 activity, and this turquoise metabolite, which is missing the methyl group at this position and the methyl group at this position, which is thought to be a product of both an initial 3A dependence step, followed by a P450 2D6 dependence step.
In a nutshell, when we look at the yellow product, which is the 2D6 product, and we look at the percentage of what we can recover in the infant's urine, at two week's of age we see that 80 percent of what we ultimately recover is the 2D6-dependent metabolite. And if one looks at the known 2D6 polymorphism, where 7 to 10 percent of the Caucasian population is actually deficient in this activity, we see that at two weeks of age, if a child's genotype says that they will be a 2D6 extensive metabolizer, that is, have functional activity, they do appear to have this activity at two weeks of age.
On the other hand, when we look at the proportion of metabolites that appear in the urine as this 3A, P450 3A-dependent metabolite, the mean is somewhere around 14 percent, but this increases to approximately 50 percent of what we recover by four months of age.
To put this into context, in adults roughly 30 percent of what we recover in the urine in a typical phenotyping study will be this turquoise metabolite. So around one month of age children have the relative contributions of 2D6 and 3A4 that adults have, but this clearly changes as they get older. At one year of age here, while the differences are not statistical significant, we are seeing a tendency for more of the dextromethorphan metabolite, more of the 3A-dependent metabolite showing up in the urine, and this exceeds what we see in adults.
Now this raises some interesting issues, particularly if one has a drug that requires bioactivation. So where I have "pro-drug" on this slide, for illustrative purposes you may want to convert that to codeine, and for "active metabolite," you may want to put in the word "morphine," and for "alternative metabolite," you may want to put in anti-methylated or non-pharmacologically active metabolites of codeine.
So, under normal circumstances, and this appears to be the case for codeine and also tramidol and other analgesics, the pro-drug itself is not pharmacologically active, but there is a metabolite that is generated, in this case by cytochrome P450 2D6, that is thought to have the bulk of the pharmacologic activity.
In a situation where competing pathways may actually be increased over the normal situation, so in adults the studies have actually been done with codeine with induction of the 3A pathway by rifampin, it can be seen that some of the active metabolite, or at least some of the pro-drug, the parent compound, is diverted away from the pharmacologic bioactivation pathway, so that you see less of the active compound being formed and you can also observe that the pharmacologic effects have been reduced as well.
In a pediatric context, I don't know that this phenomenon has been described, but I think as we learn more about how there may be pathway shifting, we need to bear in mind that a particularly unique pediatric consequence of these developmental changes in drug metabolism may be the fact that we divert drug away, that there may be a diversion of drug away from a potentially useful pathway.
Now for the last part of the talk I would like to address the issue of increased clearance during maturation and the requirement for larger doses in children on a milligram-per-kilogram basis relative to what adults require.
This particular issue has arisen from a number of therapeutic drug-monitoring studies in which people compared the doses of drugs such as theophylline or, the next slide I'm going to show you, cyclosporine, where doses have been titrated to achieve a particular target concentration. In this slide you see that the most common dose of theophylline to achieve a target concentration was between 10 and 14 milligrams per kilogram per day. In pediatric populations -- and these are children aged one to nine years of age -- the bulk of the individuals were between 18 to 26 milligrams per kilogram per day to achieve the same target concentrations.
Now one could see that if we just went directly from adult data and, say, selected a dose of 12 milligrams per kilogram per day, we would be down in this area of the dosage distribution and are at high risk of underdosing children based on adult dosing recommendations.
This is a more dramatic example of differences between children and adults. Cyclosporine is somewhat dependent upon P450 3A4 for its metabolism, and these are the doses at different weeks post-liver transplant required to achieve a target concentration.
The blue bars are the doses on a milligram-per-kilogram-per-day basis for children who had a mean age of 2.2 years compared to, in yellow, adults with a mean age of 42.3 years.
Now the easiest interpretation is that the clearance of cyclosporine is greater in children than it is in adults. It turns out that you can't just interpret these data directly because the transplanted liver in children does not have the gall bladder, and bile is required for cyclosporine absorption.
However, I also have a slide that shows FK506 dosing differences between children and adults. FK506 is not as dependent upon bile, and the same relationship holds, although it is not quite as dramatic.
But it is examples like this that have led us to believe that on a milligram-per-kilogram-per-day basis, children require higher doses of medications than do adults. Certainly there are data in other forms that suggest that perhaps something like P450 3A4 activity is, indeed, higher in children than it is in adults.
This is a study that was conducted in Denmark looking at the cytochrome P450 3A metabolite carbamazepine 10-11-apoxite and expressing the metabolite as a ratio of the parent compound, with the higher ratios implying higher 3A4 activity. We can see on the abscissas is post-natal age in weeks, and there is a tendency for the ratio implied here that the conversion of carbamazepine to the 10-11-apoxite, which is almost primarily a 3A4-dependent activity, shows higher values, certainly wider ranges, early on in life, but tends to decline as the child gets older, as the children get older.
On the other hand, there recently have been some data using warfarin, S-warfarin, that has been published. S-warfarin is dependent upon cytochrome P450 2C9 for its metabolism and the S-enantiomerase the one that's thought to have the anti-coagulant activity.
These investigators in Japan demonstrated that, yes, indeed, when the clearance of unbound warfarin was corrected for body weight, that the clearance was statistically significantly greater, and in this case around 40 percent greater, in prepubertal children compared to either pubertal children, and these were children age 12 to 18 years, or adults with a mean age of 60 years.
When you corrected the clearance values for body surface area, there was still a tendency for the -- and this is statistically significant -- for clearance to be higher in the prepubertal children mean age of 6 years compared to the adults. But when the data were corrected for liver weight, and this was estimated from pathological data, the statistical-significant relationship ceased to exist.
The implication here was that these developmental differences in drug clearance that necessitate the higher doses is simply a function of the change in the ratio of liver mass to total body mass. Indeed, this is a slide that I put together from using similar pathologic data. These are data that are found in pediatric pathology textbooks, and I have corrected them using the 50th percentile from the growth charts, but you can see that there is a spike or a peak in the ratio of liver mass to total body weight around the age of three to four years of age, and somewhere around puberty here things flatten out.
On the other hand -- and I have to be careful what I say here because these data came from St. Jude's, and right now I'm the meat in a St. Jude's sandwich over here on this side of the room (laughter) -- but these are data that were published by D. J. Murray and his colleagues at St. Jude's. Here, using antipyrine as a sort of measure of global P450 activity, yes, indeed, there was a statistically-significant difference in antipyrine clearance between children less than six years of age and post-pubertal kids. This relationship held even when the data were corrected for liver volume, and in this case by MRI.
So, to wrap up the talk, these raise a number of points that need to be considered not only by those of you who are involved in pediatric cancer chemotherapy, but those of us who are involved in pediatric pharmacotherapy, period. That is that the increased clearance or dose requirements of some compounds in children may be a function of this growth phenomenon. This particular issue may be most relevant if the liver is the predominant organ involved in the elimination of that compound and is likely to be enhanced if there is a single enzyme that is quantitative important in the elimination of the particular compound.
I say this because, in the case of P450 3A4/3A5, where it's often difficult to distinguish between the relative contribution of these two similar P450 isoforms, the fact that there is a high 3A content in intestine and also a high content in liver, or in kidney, it is not reasonable to expect that there would be a good correlation with liver mass, especially if a compound is orally-administered and must first get past P450 3A in the intestinal mucosa before it gets into the systemic circulation.
So, just to summarize then, drug metabolism pathways, and I've only dealt with the cytochromes P450 -- if we had an infinite amount of time, we could talk about glucotransformases and many other different drug-metabolizing enzyme families -- they appear to be acquired in isoform-specific patterns. It is no longer sufficient to say that cytochrome P450 is absent in the fetus and increases in expression over the first year of life. We really need to look at individual isoforms because now we have a better understanding of which specific isoforms are involved in the biotransformation of specific compounds.
I have not addressed the issue of tissue specificity, but it is likely that there are developmental changes in expression in the intestine. One would argue that we really don't need P450 3A4 in our gut until such time as we try to poison ourselves by eating our external environment. The same thing may be true for P-glycoprotein, that when we introduce solid oral foods, that's the time at which we need to protect ourselves from our environment.
Activities do appear to peak in young children. After all, we are starting from zero as a fetus and ramping up to 60 miles an hour probably in the first year of life. It is uncertain to what extent an increased liver-mass-to-total-body-mass ratio or increased functional expression per unit area of endoplasmic reticulum or volume of cytoplasm is involved in the variability that we see, but it is likely to be drug-specific.
Finally, this issue of shunting is a theoretical consequence of developmental changes in drug biotransformation, but probably is an issue that we ought to be mindful of as we look at the development of new compounds for new indications. Thank you.
CHAIRMAN SANTANA: Thank you, Steven.
We have two other presentations that relate to issues of pharmacokinetics, pharmacodynamics, and pharmacogenomics, and then we are going to have a period of discussion. But if anybody has any burning comments or questions, I think we will have a minute or two to take that now.
Anybody at the table? Eric?
DR. ROWINSKY: Maybe this is just, I guess, to possibly feed the fuel for further discussions later. That was an excellent presentation --
DR. LEEDER: Thank you.
DR. ROWINSKY: -- and it just illustrates how simple we have it in adult pharmacology and in adult medicine, just illustrating the dynamics of childhood metabolizing systems and the differences in really pharmacokinetics of drugs which portends dosing.
But I think that I would just like to -- and this is a premise for future studies, when we think about bridging between adults and children, potentially shooting for pharmacologic concentrations that might be effective or AUCs between adults and children to accelerate Phase I trials. But can we assume that pharmacodynamics between adults and children are less disparate, meaning the Cmaxs and the AUCs that we're going to potentially target in childhood studies, can we assume that at the maximum tolerated dose of a drug in children as compared to adults we have similar pertinent pharmacokinetic variables; that is, steady-state concentrations, Cmaxs or AUCs? Are the pharmacodynamics or the effect on relationships similar, which would make bridging studies a lot easier at least?
DR. LEEDER: My own personal bias is that, when trying to address the issues of effect, we are probably going to look at, need to look at, the developmental changes in the drug target, whatever that is. We may not need to achieve the same Cmax if there are differences in receptor density, for example, that are a function of development.
We do a lot of this dosing without knowing what's going on at the effect end of the spectrum. Part of it is because it is easier for us to measure a target drug concentration than it is to come up with some quantitative measure, validated quantitative measure of drug effect.
I am treading a little bit on thin ice now by trying to use a cancer illustration, but the issue that was raised earlier is that decrease in tumor size is probably one step removed from whatever it is that the drug is supposed to be targeting. I don't know just how things like receptors or intracellular signaling pathways differ with development.
DR. ROWINSKY: Well, I know that that's a very difficult question. I am not even asking you to think about the tumor, but just think about toxicity. Are the pharmacokinetic variables in children and adults similar at comparable toxicological severities, meaning at the MTD? Can we assume that?
DR. LEEDER: Well, that I don't know. For some medications we use similar -- again, I can only talk about things like phenytoin or carbamazepine. The targeted serum concentrations are the same, and we assume that those apply, but many of the ranges were actually derived from adult studies. So it is difficult to know.
Perhaps the other --
MS. RELLING: Both Clinton and I can jump in here. I mean, there's many examples in oncology where the indices might be the same. You might still want to look at AUC or steady-state concentration or time-above-some-minimum-threshold concentration, but kids tolerate higher drug exposures than adults do. They've got better protoplasm, and that means we may be able to push the concentrations higher and get a better effect. Taxol is a beautiful example.
DR. ROWINSKY: Well, that is a very important issue because at least we can have a goal. If we can assume that goal is stable, then it serves as a starting point.
CHAIRMAN SANTANA: I think this presentation reminded me that during a period of discussion one of the things, one of the goals that we have is to advise the Agency, when they are trying to implement this rule, what kind of studies and the level of rigor in the studies that they are going to require.
In your presentation I was reminded, having seen some recent protocols, particularly Phase I studies in which this concept of looking at PK in different age groups was introduced into the study. I would caution that we need to balance that against this issue that was discussed earlier this morning about minimizing risk, and maybe those kinds of studies should not be part of the Phase I design in the early dose groups, but should include patients maybe once the MTD has been defined or closer to the MTD, or maybe those different age groups should have PKs when the Phase II studies are designed.
I think that's a point that I would like for us to discuss later on. We will put it on the notepad here and come and revisit it later.
Donna, one last question and then we will take a break.
DR. PRZEPIORKA: Just one quick question, if you can give us two sentences on changes in renal function with age?
DR. LEEDER: Renal function appears to be mature by a year of age, and the kidney receives 20, 25 percent of cardiac output by that stage. Looking at it from the field perspective, immature newborns will acquire renal function at a similar rate as term newborns, but they start at a lower level. The two issues involved are nephrogenesis, which is not necessarily complete in the immature newborn, and recruitment of functional nephrons, which is acquired after birth. So probably by a year of age.
CHAIRMAN SANTANA: Let's go ahead and take a 15-minute break, and we will reconvene at half past the hour. Thank you.
(Whereupon, the foregoing matter went off the record at 10:12 a.m. and went back on the record at 10:35 a.m.)
CHAIRMAN SANTANA: If I can ask everyone to take their seats, please.
Kimberly has two brief administrative announcements. So, Kimberly?
MS. TOPPER: When we break for lunch, there is a table in the restaurant that is right behind us that has been reserved for the Committee, so that you can get in and get out, and we will get started back quicker.
Then somewhere over along here there is a transportation form. If you need us to arrange for the taxi to get you back to whichever airport, please indicate your airport, your flight time, and that information, and we will make sure the taxis are waiting when the meeting is over. Thank you.
CHAIRMAN SANTANA: Thank you, Kimberly.
We will resume, and I will invite Dr. Stewart to do his presentation; then after that, Dr. Relling, and then we will have a period of discussion.
DR. STEWART: Thank you very much, Dr. Santana.
I would like to thank Dr. Hirschfeld for his kind invitation to present today. My presentation is entitled, "Challenges of pharmacokinetic and pharmacodynamic assessments in pediatric oncology. Due to the time limitations, I don't really want to review all the pediatric oncology. I know pediatric oncology pharmacokinetic and pharmacodynamic studies; I know those of you in the audience who are very glad that I'm not going to do that.
What I would rather do, however, is to focus on the area of pediatric oncology, the PK/PD studies that we have spent the last eight to ten years studying. That is in the area of the topoisomerase I inhibitors, and in doing that what I would like to do is to just sort of generalize, where appropriate, on sort of the adult pediatric comparisons, generalize on issues dealing with model systems, and also talk about approaches perhaps that we might use for future pediatric studies.
Now in terms of the presentation, my outline is broken up into four parts that consist of -- the first part will be a summary of results of early clinical pharmacokinetic studies that we performed with topoisomerase I inhibitors at St. Jude. I guess what I should say before I move into that is that there have been a lot of studies performed by a number of investigators within this room as well as internationally with these particular compounds, but what I have chosen, again, with the time limitations, given the time limitations, I have chosen to do is to focus on the work that we have done.
Then what I would like to do is talk a little bit about some of the results from some of the non-clinical studies that we have done and how those were used to help in the design of some of our clinical trials of these agents, specifically some of the Phase I(b)(2)(a) studies, then summarize some of the results of these latter clinical trials, again the Phase I(b)(2)(a), and then spend one slide talking about some of my thoughts regarding design of the clinical PK studies of these targeted drug therapy approaches, and then have a final slide.
So, in terms of the application of non-clinical PK/PD studies to enhance anti-cancer drug development, I realize perhaps a lot of you in the room understand the different phases of drug development, but what I would like to do is to spend just a minute to perhaps get everyone up-to-speed on this; for those of you who may not think about this on an everyday basis, talk about the different phases of drug development, and in doing that, talk about how we use non-clinical PK/PD studies in a very general sense in terms of drug development.
So, obviously, drugs come from a variety of sources. At our institution they come from drug companies, from the NCI, from a variety of sources. But the studies that we conduct then lead to -- the results from the PK/PD studies lead to the design of the Phase I clinical trials. Then, not unlike a lot of other institutions, the data from those studies feed back into perhaps design of additional non-clinical PK/PD studies with the goal there to evaluate perhaps additional schedules, dosing, look at additional efficacy studies, so that that will feed into additional Phase I clinical trials with the goal to then move into Phase II clinical trials, looking more at the efficacy of the compound.
Now the results of these Phase II clinical trials often feed back into the Phase I clinical trials, where we are evaluating the clinical safety of new schedules, dosages, and combinations, based upon some of the results from the Phase II clinical trials, and oftentimes the results of the Phase II clinical trials will carry us back into the non-clinical models to evaluate additional aspects of the compound. Then, as you well know, the Phase III clinical trials and then the Phase IV post-marketing studies are conducted.
So, again, what I will want to talk about today are studies with the topoisomerase inhibitors, and for those of you who don't think about those on a daily basis, there are basically two of these compounds that are available for use in pediatric oncology, Topotecan and Irinotecan. What I have depicted on the slide is the camptothesin backbone molecules, the penylcyclic structure.
One of the things that we've had to face when we do these PK studies that is a little bit of a challenge, however, it is not something that we haven't been able to overcome, is the fact that this compound, the E-ring system, the lactone ring, which has been thought to be the active moiety, and to measure this we've had to stabilize this by doing a methanolic precipitation within a relatively short period of time. If this is not done, the compound, the camptothesin molecule undergoes a reversible PH-dependent hydrolysis to an open hydroxy acid form, which is thought, conventional wisdom is right now, that that's an inactive form, does not have anti-tumor activity. So that is one of the challenges that we faced with these particular molecules.
Now the two molecules, like I have said, Irinotecan is basically a pro-drug for the active moiety SN-38. It basically has a moiety that's cleaved off here, and then Topotecan is the dimethyl aminomethyl moiety up here at the R3 position. We will talk more about that later.
Most of the studies that I will talk about today will deal primarily with Topotecan, not because I have -- as Kimberly read, I don't have stock in Topotecan, but we've done most of our studies with Topotecan. I will mention a little bit of the work that we have done with Irinotecan. It is just a matter of the sequencing of the way that the studies have been done.
Most of the work that we have done with Topotecan we are planning now to sort of move into Irinotecan an do a lot of the same kinds of studies. So it is basically a paradigm that can be used in both, although I will tell you that Irinotecan is a little more complex to deal with. I will talk about that in a little bit.
So let me just move into some of the initial clinical trials that we deal with topoisomerase I inhibitors. The first study we did way back when, in collaboration with Dr. Charles Pratt, was a 72-hour continuous infusion. We saw Topotecan -- this was in children with recurrent solid tumors -- we saw anti-tumor activity, although, as I have asterisked here, this was not what we thought, based on our non-clinical studies, to be the optimal method of administering this particular compound.
The dose-limiting toxicity was myelosuppression. What was interesting for the PK part of this was that this provided the preliminary data for the derivation of a limited sampling model which we used for future studies. I will get into that a little bit later in my talk.
The other aspect of this particular study was that we did pharmacodynamics, and the pharmacodynamic relationship that we saw from this study was very similar to what had been published earlier by other investigators, Dr. Lewinsky and colleagues and others also.
The second study we performed was in collaboration with Dr. Wayne Furman. It was a 120-hour continuous infusion of Topotecan in children with recurrent leukemia. This is a little bit different from this Phase I study in that we used what was called, what Dr. Bill Evans has coined the term, "maximally-tolerated systemic exposure."
Perhaps the figure here will help me describe that, in that as opposed to escalating to dose to toxicity, what we did here was we escalated the exposure, the Topotecan plasma concentrations, to toxicity. So patients were enrolled in different concentration cohorts, and those cohorts were increased until they observed toxicity. So the dose was individualized for those patients based on what systemic exposure cohort they were enrolled in.
In this study we also observed an anti-leukemic effect; again, the asterisk meaning we didn't really think this was the right regimen to use. We could talk about that later on.
But this limiting toxicity here was mucositis, which was interesting because this 120-hour continuous infusion is really the only time the mucositis has been seen as the DLT.
The PK/PD observations were also interesting, and that is what I have presented in the slide here, because what I am plotting on the vertical axis is proportion of patients or proportion of courses versus the plasma systemic exposure, and the green line represents the oncolytic response, and the red or fuchsia or pink or peach-colored line represents the dose-limiting toxicity or mucositis.
What we observed from this study was that, once you got above a systemic exposure of approximately two, you really didn't get any more in the way of anti-leukemic effect or oncolytic response, but what you did do was you got more in the way of mucositis. So that was the results from that particular study.
Then we moved into a series of oral Topotecan studies where we evaluated 15- and 21-day dosing of oral Topotecan, showed that it was well-absorbed, saw wide interpatient variability, but also observed that it was less than the intrapatient variability.
In all of these studies we were fortunate that, for various reasons, patients had -- we were able to gets access, or we had access, to Topotecan or to CSF samples. We measured Topotecan and found that there was extensive penetration similar to what Dr. Frank Balis had published in the primate model, Frank Balis and Susan Blaney had published, in the primate model, as depicted on this overhead or this figure, where we are plotting Topotecan CSF penetration on the vertical axis. There was really no difference in the extent of penetration for the 30-minute infusion, 24-hour infusion, or the 72-hour infusion.
What you will note is that this is a really very high penetration for an anti-cancer drug, and we will take advantage of this particular characteristic of this drug in subsequent clinical trials.
Now the adults had moved forward with a short infusion given daily for five days, and this was the first study in which we, the pediatric community, did this. It was a Pediatric Oncology Group Study 92-75. Dr. David Tubergen was the principal investigator, and we did the pharmacokinetic studies at St. Jude. This was in children with recurrent solid tumors. It then moved to children with recurrent leukemia, and then there was a study in which we did it just in recurrent leukemia. We noted anti-tumor activity in this particular trial, DLT with myelosuppression, again, similar to what adults had seen.
Here was where we applied our limited sampling model, because this was a study that was conducted in a cooperative group and required, if we were going to do this, we had to simplify this. We had to make it where it was exportable, something that could be done and accomplished on a reasonable basis.
So we were able to export this, using a limited sampling model, and then in a subset of patients we were able to validate this particular limited sampling model. We observed a very wide interpatient variability in Topotecan systemic -- I'm sorry, Topotecan clearance, which led to overlap in systemic exposure because the differences in dose levels were so narrow.
So you have 20 percent differences in dose levels, and yet the difference in the interpatient variability and clearance was around 500 percent, let's say, at the 2.4 milligram-per-meter-squared level.
So it is interesting to me, when we talk about the fact that we want to have these differences in dose levels between patients, and yet no one, or it is very rare that you really hear people talk about the fact that there is pharmacokinetic variability which is going to lead to a difference in systemic exposure.
So you can change your dose from 1.4 to 1.7 to 2.0 to 2.4, but it is very likely that the systemic exposure that a patient will achieve is not going to be different between the different patients. So I think this is -- I have this opportunity to have a soapbox, so I got on my soapbox; now I'll get off of it.
So those are the studies that I wanted to talk about in terms of our early clinical trials for Topotecan. What I would like to do now is talk about a clinical trial that we did at St. Jude with Irinotecan.
This is Irinotecan. The Irinotecan molecule is in the middle of the slide. It is a compound that undergoes conversion by carboxglesterases to form SN-38, which is the active moiety that undergoes metabolism by the CYP 3A4 to form two inactive metabolites, the APC and NPC, and then SN-38 is converted by glucuronidation to SN-38G.
So this study was a 60-minute infusion in children with recurrent solid tumors. We used a schedule of daily times five times two, and I will talk about that in a few minutes, when I talk about some of our non-clinical studies where that particular schedule came from.
But what we did was we noted very significant anti-tumor activity for the compound on this schedule. As noted by adults that have reported data from this compound, the DLT was diarrhea. As you can imagine, the pharmacokinetics of this compound are very complex, and the metabolism, as I have depicted over here, is a very complex issue, although not something that can't be handled.
SN-38, also something else that has to be considered is that it is a very highly protein-bound compound.
Then, finally, this is a compound that is a pharmacogeneticist's dream with all the different metabolites and metabolic pathways.
So let me just spend a summary slide comparing the results of the adult and pediatric Phase I studies for the topoisomerase I inhibitors. So let's talk about the pharmacokinetics.
Topotecan lactone systemic clearance, and I think it is fair to say the Irinotecan lactone systemic clearance, have been similar between the adults and children. I qualify that by saying in early studies, and I asterisk that. I will clarify that in a subsequent slide. So let's not get too carried away with that. That is not completely true. So if you happen to doze off and don't hear the rest of my talk, it is not completely true that Topotecan lactone clearance is similar between the two groups. It was in the early studies, but it is not overall. Okay?
The problem with that is that in the early studies we were studying limited patient populations. We had small numbers of patients. I will tip my hand by saying the age ranges were fairly narrow in the early populations, and we didn't have the drug-drug interaction studies that we have had in subsequent studies.
What about PD? This is something maybe that Eric was alluding to. It looks like the relations, the PD relationships, between the two groups are similar for the most part. There is an interesting slide that I prepared that, if I had time, I would like to show you, but in the interest of time I wasn't able to.
The MTD, as Mary said, is higher typically for comparable schedules, but part of the problem we have is that our schedule, this daily times five times two, is different from what has been used in adults. So it makes for a problematic comparison. The dose-limiting toxicity between the two groups for the most part is very comparable.
Now let's talk about the application, the results, results from non-clinical studies of Topo I inhibitors to the design of clinical trials. Again, I will just refresh your memory about the first slide that I showed with this sort of paradigm of using non-clinical studies as sort of a bedrock for the design of the Phase I/Phase II studies.
So at St. Jude's, for those of you who have heard Dr. Peter Houghton talk, one of the models that we use quite extensively is the xenograft model. That is where one takes a tumor from a child and implants it into a immunocompromised mouse, currently a skid mouse, and evaluates both schedules and doses of different drugs.
The other aspect that we have done, or we have done quite a bit of, with the Topo I inhibitors is then to evaluate the pharmacokinetics in the murine model, the murine xenograft model, and compare that with humans.
Now there has been a lot of criticism about this particular approach. It is very justifiable if the pharmacokinetics for that particular compound of interest are different between humans and mice. We have been very fortunate for Topotecan that the PK between mice and man are very similar. The half-life, shape of the curve, the systemic exposure are all very similar.
Irinotecan is a little bit different in the sense that mice have quite a bit of esterase in their plasma. So they have a very different profile in terms of the production of the SN-38. However, we are studying now a different transgenic mouse, ES1-minus mouse, which is deficient in esterase, so it has no esterase in the plasma. It is a little bit different in terms of its production of SN-38. It may be a little more like the human in terms of the production of SN-38. So what we are trying to do is find the most appropriate model to be able to evaluate Irinotecan in this particular setting. The take-home message from this slide is this is a good model to use as long as you are cognizant of the differences in your species.
So what are the lessons that we have learned? Well, Pete's done a lot of studies with these drugs and these animals, and I am summarizing probably 12 years of his life in one slide. So bear with me.
What he has found is that these agents are very schedule-dependent. The duration of therapy is critical. The administration interval is very important, and that is what has led us to be a very large proponent of the protracted dosing schedule and saying that that is associated with very significant anti-tumor activity.
The fact that these compounds are very dose-dependent, such that at very high doses you don't get any more anti-tumor activity; you can't kill the tumor cell any more at higher doses. However, there is a critical threshold drug exposure that is necessary for anti-tumor activity, and that's depicted on this figure on the righthand side, where we are plotting AUC versus different neuroblastomas' intergraft lines, showing that once you exceed a certain AUC, you have a very good response rate.
So what this has led us to do is to develop in our clinical dosing schedule this low-dose protracted schedule, this Dx5x2 that I have alluded to a little bit earlier in a previous slide.
Now the other model that we use quite a bit, the other animal model we use quite a bit, is the non-human primate. Now Dr. Balis and Dr. Poplack both have used this quite extensively. We have used this to study Topotecan in CNS malignancies.
We have used it, in addition to studying the penetration, what we have looked at this model for is to evaluate the effect of Topotecan infusion rate on the CSF concentrations throughout the neuraxis, looking both at ventricular and lumbar concentrations, and using it as a prelude to the design of a clinical trial. Also, to try to help us generate a PK model that would describe the plasma and CSF disposition, so that we could then take the data from this particular model and use it to design a clinical trial to treat children with CNS tumors.
Now what I would like to do is to move into a summary of the results of some of our later clinical trials of Topo I inhibitors, talking about some of the 1B, 1B2A studies. I am really sort of pressed for time, so I really can't spend a lot of time talking about this whole concept of PK-guided dosing. A lot of you have heard me talk a lot about this.
Suffice it to say that dose intensity in clinical response for a lot of -- there's a lot of good rationale for it, for the appropriate kinds of tumors, but dose intensity doesn't equal systemic intensity, for a number of reasons. One of the most important reasons -- and Steve just alluded to this earlier before the break in his very good talk -- is that there is pharmacokinetic variability. A lot of it is maturation-related.
We have observed with Topotecan a lot of interpatient variability and systemic clearance. I have already mentioned the maturational changes, renal and hepatic impairment due to other concombinant drug therapy. Mary will talk a lot about the obvious problems with -- or not problems, but the considerations of pharmacogenetics, and then drug-drug interactions.
So there are a number of considerations we have for selecting drugs for pharmocokinetically-guided dosing, and let me just go ahead and get them all out there. There are general considerations, and they are listed here on the slide, logistical considerations.
Our dosage schedule really lends itself well, the Dx5x2 lends itself well to that. We have an assay method available.
The fact that we have done these earlier studies, we well-characterized our PK model. We have population priors available for a Bayesian analysis, and the fact that we have the limiting sampling model available makes it very easy for us to be able to do these particular studies.
So the selection of our initial dose, we used our non-clinical studies to be able to assist in doing that. The second thing that you have to deal with, and we would spend a lot of time talking about how one selects the pharmacokinetic metric to express your exposure. We have used for our studies the area under the concentration and time curve, but there are a lot of ways you could do it: Cmax, Cmin, time above a threshold. There's just a lot of ways to do it.
This particular slide just shows the setup of one of our pharmacokinetically-guided studies, where the drug is administered daily for five days. We do PK studies on day one, three, eight, ten, and twelve. This was our first study, which was kind of a feasibility study.
Then based upon whether the patient was in-target or out, we would adjust the dose to get the patient in-target, and this is basically sort of the general schema of how we have done most all of our studies.
So the first study we did basically was the feasibility study. We noted anti-tumor activity. We were able to achieve our target exposure, and we reduced interpatient variability. Then we have done subsequent studies, a couple of studies where we have done pharmacokinetically-guided Topotecan in combination with Vincristine. We have noted anti-tumor activity, myelosuppression. We had to use a little bit lower Topotecan target.
Let me just move on. Okay, so in a Phase II study -- this is the study that Dr. Santana is the principal investigator on -- we've done PK-guided dosing in this particular study in children with high-risk neuroblastoma. You will have to ask Dr. Santana about the clinical results, although I am sure that they are very good. The last data I had, the partial response rate was greater than 50 percent. We were able to achieve our target exposure and decrease the interpatient variability by doing this pharmacokinetically-guided dosing.
Now this is one of the things I really wanted to get into, and I brought this up a little bit earlier. We have studied on this protocol, and those of you who treat children with neuroblastoma, you are aware of this. This is a disease of children of a younger age.
We studied ten infants that were less than two years of age, and in this population of patients we noted that Topotecan lactone systemic clearance was significantly less than in other patients. The clearance was 12 versus 21 in other patients.
So what we have learned from this is a way to dose Topotecan in children that were less than 12 years of age. So it is, I think, a very important contribution.
We have done a study in children with high-risk medulloblastoma, where we used PK-guided dosing to attain drug exposure and a minor exposure compartment.
Manageable toxicities, what I would like to get to here is we have noted a couple of drug-drug interactions which I don't think have really been alluded to in adults. The enzyme-inducing anticonvulsant has for 9-AC and enzyme-inducing anticonvulsants and for Irinotecan, but the Topotecan enzyme-inducing anticonvulsants really hasn't been reported in adults. We also reported that dexamethasone increases Topotecan clearance. Both of these were observations that came out of this particular study.
So those were the results of some of our latter clinical drug development studies. The next-to-the-last slide are some issues that I would like to bring up as it relates to the design of molecular target-based anti-cancer drugs in children. I am afraid that what I have got here is I've got more questions than I have answers.
We are in the process of designing a lot of these studies ourselves at St. Jude, and so we have a lot of these sort of questions. Maybe I look forward to the discussion period so we can profit from the corporate wisdom in the room.
So, you know, probably the first question that would come to mind is, when you start talking about molecular target-based anti-cancer drugs, what is the target? I think that is an important question. So is the target just the expression of the protein in vivo or do you have to have an expression of the protein in vivo and data from an in vitro study that says that protein is actually sort of important? Or do you actually have to have some studies that say there's some prognostic significance to the protein that is your target?
So I think there's a lot of questions that remain to be determined about what a target is. Then I think if we could ever come on an agreement of what a target is, then I think there is a need for the development of a relevant model in which we could evaluate that target. The in vitro model is important, the xenograft model, the transgenic model, but I think, regardless of which model we pick, it is going to require that we have a complete understanding of the pathway or pathways that are involved.
Then, as we have these questions answered, which I don't necessarily think there are answers for them, but if we do come to some kind of consensus, it is going to be important to come up with some sort of pharmacokinetic metric, just as we face that same question with PK-guided dosing: Is it IC-50 in the plasma tissue, is it an AUC, is it some other measure of drug exposure? What do we use as a metric perhaps to convert between adults and pediatrics or between the lab and the clinic?
Then I think it is important to consider that pediatric tumors -- and I say "likely" -- what I should say here is "may have" different biological pathways from adults and that, therefore, they may have different targets. So that is just something, perhaps a provocative thought.
So I haven't really talked a lot about challenges per se because I am very fortunate that the resources and infrastructure at St. Jude are in place to be able to make these studies possible. However, because of a lot of the work of Mark Berstein and Peter Adamson, the infrastructure I think is in place, or is coming in place, in the developmental therapeutics community and the COG to be able to make these studies possible. I shouldn't just stop with Mark. I mean, there are a lot of people that have worked in COG and CCG to make these kinds of studies possible.
But I think the challenge for the future is to apply what we have learned in these studies of Topotecan to combination studies for the future. I just bring forth one example. This is a study that Victor and I and Julie Park have talked about a little bit, about a combination of Topotecan with cyclophosphamide in one aspect of therapy for neuroblastoma, and the question becomes how to dose Topotecan.
We are in the midst of doing a population pharmacokinetic study. Everything that I have told you so far is all single pharmacokinetics, single patients. We have done non-mini-mAnalysis of Topotecan, and what we have found from our non-mini-mAnalysis is that Topotecan clearance is related to BSA, concomitant phenytoin therapy, serum creatinine in age, and perhaps a model that includes these patient co-variants is something we could use to prospectively dose Topotecan, much like we dose carboplatin based on creatinine clearance.
The other aspect is PK studies will provide insight into differences in drug disposition, which can then be explained in many cases by genetic variations and drug metabolism or transport or the genotype approach.
So, with that, I will close, and I thank you very much for your kind attention.
CHAIRMAN SANTANA: Thank you, Clinton, for a rather extensive overview of this issue using the camptothesins as a model. I think we have a brief time for some questions while we change computers. Maybe I will ask the first question, which is kind of a little bit of followup of what Rowinsky was alluding to earlier.
Using the camptothesis as a model, because we're not really talking here about the drug specifically but as a model, and I want to make that clear for the audience and for the discussion. But using it as a model, do we have enough data in this class of compounds that we can address the question that Eric asked earlier, whether for this class of compounds systemic exposures that are seen in adults to some percentage should be what we use in children when we design our trials, or vice versa? How do we use the pediatric data in relation to what we know about similar exposures in adults?
DR. STEWART: Yes. See, the problem is, and I was actually giving that a lot of thought while I was putting the talk together, I think the problem becomes one of comparable schedules. So you can't look at exposures independent of schedule. So the schedule that most adult studies are on are daily times five. I think the study that Wayne did, Wayne Furman did, in the Pediatric Oncology Group, fairly convincingly showed that, and other studies that we have done fairly convincingly have shown that schedule is very important to the anti-tumor effect of the camptothesins.
So I don't think it is fair to compare a five-day schedule at some exposure to a ten-day schedule at some other exposure. So I don't think you can just -- it would be very easy to say, oh, okay, so the cumulative exposure is this and the cumulative exposure is this, so let's just start making comparisons.
You can't remove schedule from that comparison. Do you see what I'm saying?
CHAIRMAN SANTANA: Yes.
DR. STEWART: So I think that complicates it a little bit.
CHAIRMAN SANTANA: Donna, you had a comment or question?
DR. PRZEPIORKA: Just a quick question for either you or Dr. Leeder: Another class of drugs coming out now are the biologics and the monoclonals. Could either of you have any information on the pharmacology of monoclonal antibody in pediatric patients?
DR. STEWART: Steve?
DR. LEEDER: I don't.
CHAIRMAN SANTANA: Do you have any more detail?
DR. LEEDER: Do you want to comment? Malcom, do you want to comment?
DR. SMITH: I think it probably depends to some extent on the monoclonal. For example, for Atoxomab, there's an extensive body of experience in adults and some limited experience in children, but we are heavily building upon the adult experience to base our dosing and schedule and combinations that we use in children.
DR. ROWINSKY: There is probably less concern with antibodies, as antibodies generally behave in a very similar way, and the target may not be as important. So just knowing the differences in antibody clearance between children and adults, one can extrapolate. I think it becomes much simpler than the issue with drugs that behave in so many different ways and are cleared with so many different variables impacting. I think it is much simpler with antibodies.
DR. GOOTENBERG: I represent biologics, and with all due deference to Dr. Pazdur and Dr. Hirschfeld here, we think that biologics, monoclonal antibodies, and cytokines, and cellular therapies are a big wave of the future. So I think that your question is right, Donna, it is right on target, and it is not just monoclonal antibodies. That forms a very small part of the spectrum of biologics that will be coming down the pathway soon.
I think it is going to be a crucial question. Since the rule applies to these also, the differences in the pharmacokinetics and the pharmacodynamics, and wait until you try to work this out with cellular therapies and gene therapies and the different vectors that are being developed now. It's just somewhere where I think a lot of attention is going to need to be paid.
CHAIRMAN SANTANA: Malcom?
DR. SMITH: Just one comment. It was an excellent presentation.
You talked about targets and defining a target and whether a target needed to be a prognostic factor. I would point out something like BCR-able, for example, is an outstanding target, but with MPH-positive leukemias, it is not prognostic because every case has it. So it is not so key that it be a prognostic factor, but just that I would say it needs to be central to growth, survival of the cells; it needs to be intrinsic to some signaling pathways that are required for cell growth, like BCR-able, like mutated CKID and CKIT.
CHAIRMAN SANTANA: Other comments or questions?
CHAIRMAN SANTANA: Then, Dr. Relling, please.
MS. RELLING: Thank you, and I also appreciate the opportunity to be here. Obviously, I think this is an important topic.
So how do we relate pharmacogenetics to translating oncology studies to pediatrics? First of all, I guess the current interest in pharmacogenetics has been partly precipitated by the realization from the fruits of the human genome project, that every human gene is polymorphic. So, as we look around the room and we see how we all differ from each other, it is obvious that there are lots of genes that must contribute to all the different phenotypes that we are seeing around us.
It has now been proven that one single nucleotide polymorphism or genetic variation occurs about every 400 to every 1500 base pairs. That means that there's certainly at least one variation per gene. In fact, there are enough that there's almost certain a functional variation in every gene. Since the actions of drugs in children or anyone are going to be due to their interaction with the host genome and the tumor genome, pharmacogenetics is going to affect the action of drugs.
In oncology we have the added complication that cancer has acquired mutations. Of course, some of these tumor mutations are common to children and adults, although many, many, many are not. As Dr. Smith just alluded to, the 9;22 translocation that's proven to be such an interesting target for new agents in adults, CML is certainly present in the very rare disease of childhood CML, but also present in children and adults with acute lymphoblastic leukemia, and the function of that translocation seems to be affected quite differently in those diseases.
So even if we do identify common targets in adults and in children, we can't assume that they can be expected to respond to drugs in the same way. But the germline host polymorphisms are the germline host polymorphisms that are going to be present no matter what the age of the person is. Of course, they may express themselves a little bit differently in children than adults, but the principles are going to be the same.
So this just illustrates the fact that in cancer we have two genomes to worry about: the host genome with this at least one variation per gene characteristic and the tumor genome that by definition has acquired at least one, and probably many, variations that differ from the host tissue from which it arose.
Genetic variation in both the host -- and that's what we really mean by talking about genetic polymorphisms and pharmacogenetics -- as well as genetic variation in the tumor will contribute to the bioavailability of drugs; that is, their availability to the tumor that will affect the intrinsic sensitivity of the tumor to anti-cancer drugs and that will affect the host risk of toxicity.
By the relative size of the host versus the tumor genomes depicted here, as well as the degree of interaction, we get an idea, of course, that host polymorphisms and tumor polymorphisms are going to affect how much drug gets to the tumor. So a p-glycoprotein polymorphism is not going to only affect how much drug is absorbed and how much is excreted in the bile and how much is excreted in the kidney, but it is also going to have some baseline effect on the tumor. If the tumor has acquired mutations in p-glycoprotein, then that will also affect drug bioavailability to the tumor.
Of course, the risk of toxicity from anti-cancer drugs is largely determined by the host polymorphisms. So many of the clues that we get about host polymorphisms from adult oncology, we can certainly extrapolate or at least test in pediatric oncology.
So, given a similar schedule of drugs and similar regimens, pharmacogenetics should have similar implications for children and adults. In terms of host polymorphisms, the developmental changes that are expressed in semantic tissues certainly contribute to the child versus adult differences in pharmacokinetics that we just heard about from Dr. Leeder and from Dr. Stewart, but those germline polymorphisms should affect the hosts similarly in children and in adults.
For the tumor, we know that there are certainly many tumors that are quite different in children versus adults, but for purposes of today's discussion we are assuming that we need to test drugs in both patient groups, and the germline polymorphisms that are present could affect the tumor responsiveness or invasiveness similarly in children and in adults. So that if the polymorphism affects the degree of metastasis, the effect of the anti-angiogenesis, we assume that those things are going to be acting in both patient groups.
I am going to give a few examples of how pharmacogenetics has already been shown to have implications for anti-cancer therapy. To back up a second, I guess we all acknowledge that anti-cancer drugs are the one therapeutic area that is clearly going to benefit by optimizing the dosage of the drugs. That is certainly going to be true in children, where we want to give enough drug to have anti-tumor effect, but not so much drug to result in unacceptable host toxicity. So anything that we can use to more intelligently determine the way to give these drugs is worthwhile, and pharmacogenetics is going to play a part in that.
Polymorphisms in gene products involved in metabolism distribution and transport, receptors and targets of the host, which all affect toxicity and pharmacokinetics, as well as tumor receptors and targets, as well as polymorphisms in the pathogens, which still result in an incredible amount of morbidity and mortality in children with cancer, all have an effect on the risk of cancer development itself, on the risk of host toxicity, the probability of tumor response, and on the probability of severe infectious complications.
So I will just give you a few examples. The glutathione-S transferases, or GSTs, have been shown to affect the risk of toxicity from anti-cancer drugs as well as the chance for cure. Anti-cancer drugs often have metabolites that are free-radicals or electrophiles, and they can be conjugated with a tripeptide glutathione, and that conjugation is facilitated by glutathione transferases.
So that if patients have wild-type or normal glutathione transferases levels, they are likely to more efficiently inactive the drugs, and therefore, probably have less toxicity, but, of course, that may also mean that they have less anti-cancer effect. Conversely, patients with mutant glutathione transferases or low glutathione transferases activity will have less inactivation of the drugs, potentially more toxicity, but also potentially more anti-tumor effect.
There have actually been nice studies to demonstrate both of these principles published in the last couple of years. Stella Davies, as part of the CCG, published a nice study where they did a randomized trial in children with acute myeloid leukemia, where the question was very simple. They were testing a five-drug regimen of dexamethasone, Atoposite, AraC, thioguanine, and daunomycin, given on a standard timing schedule where the patient was allowed a bit of time to recover in between courses or an intensive timing schedule where one pushed on, despite the presence of toxicity, which, of course, we all recognize is often done in patients with AML. The question was: Which schedule is better? The overall results in these over 300 children was that there was a slight advantage for the intensive timing schedule, but whether you benefited from pushing the dosages of the drug depended very much on a single genetic polymorphism and glutathione transferase.
So they divided patients into those who received the standard timing that were wild-type or had the GSTT1 gene product present, standard timing that were null, so no GSTT1 enzyme present, and then the same genotypic groups in intensive timing.
You can see that in the intensive timing group there was a statistically-significantly inferior survivor, 43 percent versus 59 percent, in the patients who received intensive timing who were lacking in this single enzyme. This is a common polymorphism. So 15 to 30 percent of the American population is completely lacking; it's a total gene deletion in the germline of this GSTT1.
That translated into a threefold, almost a threefold, higher risk of death in remission from this intensive timing regimen. That risk was not present in the patients who were GSTT1 wild-type. So by looking at a single gene product, we may be able to start to get at individualizing therapy, and just the way that we give the exact same drug combination.
But this contrasts with studies published from Sweeney, et al., from the SWOG, who looked at patients with breast cancer. So these were adult women with breast cancer who received cyclophosphamide and anthrocycline-containing regimens, and they looked at a different form of the GST, the P1 enzyme.
Here the mutant form, this is looking at the proportion of women surviving of their breast cancer. They were more likely to survive if they did have a mutant form of the GST enzyme versus the patients who had at least one wild-type copy of the gene for that enzyme. So there the hypothesis was toxicity wasn't the main problem in overall survival. Having enough drug onboard to cure the breast cancer was.
So we can contrast these results, and I apologize that these colors are the same. It's not as effective as if we showed -- this is the result in the adult women with breast cancer where mutant GST was associated with a great anti-cancer drug effect in these women with breast cancer, never given the kind of intensive chemotherapy regimens we give to children with AML, but the opposite was true in children with AML, and that is, the mutant form of the enzyme was associated with the worst overall event for survival on the basis of unacceptable life-threatening toxicity in those who had the null enzyme.
So this illustrates that the effect of every polymorphism has to be evaluated in the context of the disease and the intensity of the therapy. So intensifying therapy in GST wild-type patients may be correct in children with AML, but not necessarily in adults with breast cancer.
Another example of polymorphisms affecting anti-cancer drugs is one that was discovered many years ago by D'Ozio, and others have established the molecular basis of this polymorphism: dihydrophyrimidine dehydrogenase is a gene product that metabolites 5-fluorouracil, and it inactives 5-fluorouracil, and therefore, the lower the DPD activity, the more parent drug is available to be activated.
This is on the basis of a single single-nucleotide polymorphism, a SNP, that's actually at an exon-enteron border, and the presence of that SNP affects whether the exon 14 is present in the gene product or not. So that single mutation results in the complete lack of exon 14 in the transcript, and therefore, a nonfunctional protein.
About 3 percent of patients are heterozygous for this mutation and are at very high risk for severe and life-threatening toxicity from 5-fluorouracil when they're given doses of this drug. Well, 5-fluorouracil has as its target, so it's metabolized by DPD, but its target in the tumor tissue is thymidylate synthase. Thymidylate synthase undergoes a common genetic polymorphism, which, of course, is also present -- it is present in the germline tissues, and so, therefore, affects host toxicity. It is also, of course, present in the tumor tissues, and therefore, can affect tumor responsiveness.
So, again, this is a case of two repeats of a 28-base pair section of the promoter versus three of these tandem repeats, and the individuals who have two tandem repeats have lower expression of the enzyme and lower TS activity. Therefore, there's less of the target that has to be inhibited by the 5-fluorouracil, and they have a better anti-tumor response to 5-FU. Those who have three repeats have increased target present, and therefore, they have a slightly worse anti-tumor response to 5-FU.
So this slide just puts it together, showing that the enzyme polymorphisms and the enzymes that metabolize the drug and polymorphisms in the target for the drug both will have an effect on an individual patient's probability of toxicity and their probability of efficacy from that drug. So that germline polymorphisms can affect tumor responsiveness as well as toxicity, and it illustrates that more than one gene product polymorphism is likely to affect drug efficacy.
Speaking of multiple gene products, I want to illustrate some polymorphisms affecting methotrexate, another commonly-used anti-cancer drug, widely used in many pediatric tumors. This is an extremely simplified diagram of the cellular targets and enzymes involved in metabolism of methotrexate, which is activated intracellarly and interacts with many different targets, all of which are probably involved in its anti-tumor effect as well as in its toxicity.
I am just going to focus on one polymorphism in the methylene tetrahydrofolate reductase gene product. This is, again, a very common polymorphism. Ten percent of us are homozygous mutant for this mutation that results in lower MTHFR activity. I'm one of that 10 percent. I'm at higher risk for cardiovascular disease and various neurological complications, so I'm taking my folic acid supplementation every day. Forty percent are heterozygen and 50 percent are homozygous wild-type.
You can tell folate metabolism, which is the target of methotrexate, is complex, so it is a little difficult to predict what the affected MTHFR might be on methotrexate effects, but the general idea was that MTHFR mutants tend to be lower folate patients, and therefore, they might be more susceptible to the adverse effects, and maybe also higher probability of response to methotrexate.
Published from Seattle this year in "Blood" is an analysis of the risk of oral mucositis, OMI, Oral Mucositis Index, in transplant patients who were given low-dose methotrexate as a preparative regimen. They showed that this one single nucleotide polymorphism that affects whether one is homozygous mutant, heterozygote, or a wild type has an effect on the risk of oral mucositis from this methotrexate. So that 10 percent of the population who are homozygous mutant were at a significantly higher risk from mucositis from low-dose methotrexate.
Of course, we give a lot of methotrexate at St. Jude's. So we have been curious as to what it would mean for us, but we have not found that MTHFR genotype status affects toxicity after high-dose methotrexate. That is higher doses that are given with a rescue agent called leucovorin, and this is looking at toxicity assessed as the delay in therapy after a dose of high-dose methotrexate. This is in several hundred patients, 50 percent wild type, 40 percent heterozygote, 10 percent homozygous mutant. You can see absolutely no difference in the number of days required to recover from that high dose, possibly because we're abrogating the effect of this polymorphism because we supersupplement with folate supplementation after high-dose methotrexate.
So another principle is that the effect of each polymorphism may be dependent upon the dose and the schedule of the anti-cancer agent. As Dr. Stewart just alluded to, there's lots of cases where we dose anti-cancer drugs differently in children than has been done in adults, not necessarily because we all couldn't benefit from learning from each other, but that is just the way it happens.
Another polymorphism in UGT1A1 that Dr. Stewart was alluding to has been shown to affect the risk of Irinotecan toxicity in adults with cancer. This UGT is a glucuronacil transferase. It is involved in inactivating the active metabolite of Irinotecan that's then excreted in the bile. So individuals who have low UGT1 activity, and that's on the basis, again, of a promoter polymorphism, so about 15 percent of the population has low activity, low expression, and low glucuronidation, and therefore, at higher risk for dose-limiting diarrhea and leukopenia from Irinotecan than the majority of the population who have higher or normal UGT1 activity.
We are starting to evaluate the importance of this in pediatric studies. Our colleague at St. Jude, Dr. Chris Cruz, has shown that in a pediatric schedule of Irinotecan, which is this very prolonged oral exposure, which again has been more tested in children than in adults, it doesn't seem that the UGT1A1 polymorphism will have the same important effect.
So when drugs are dosed to be below those KMs or those thresholds for saturation, polymorphisms and enzyme metabolism that may be present with higher-dose bolus doses may not be manifest themselves with low exposure to chronic doses.
Finally, I want to give you just a hint of our own experience with the thiopurine methyl transferase polymorphism. It illustrates, I think, several nice principles that have come up this morning.
6-mercaptopurine is one of the two backbones of ALL therapy. It was discovered and approved by the FDA at less than two years from its discovery in 1953. It has been used for treating childhood ALL for almost that entire 50-year time period, and I will submit to you that we are just starting to learn how to dose this agent. So while I am enthusiastic that we are talking about better ways to dose anti-cancer drugs, I am a little worried that some of these things may take a long time.
Mercaptopurine is a substrate for a polymorphic enzyme called thiopurine methyl transferase or TPMT, which inactivates the parent drug, shunting it away from its activation pathway by HPRT, where it is metabolized into TGNs or thioguanine nucleotides. These acts as false guanines, are incorporated into DNA and RNA, and that is part and parcel of the way that 6-MP kills leukemia cells. It is also part of the way that it causes toxicity.
One in 300 individuals is homozygous mutant, so both maternal and paternal alleles have at least one point mutation that inactivates the enzyme. Ten percent are heterozygote, and 90 percent are wild type or have normal high TPMP activity. That translates into an inverse relationship in terms of the systemic exposure to the active thioguanine nucleotide concentrations. The rare patients that are homozygous mutant have sky-high levels of TGNs. The majority who are wild type have relatively low level of TGNs. The heterozygote patients have intermediate exposure to these TGNS.
Those who are homozygous mutant are at increased risk of myelosuppression and I shall show you at increased risk of an unacceptable late effect of secondary cancers, whereas the wild-type patients are at lower risk for toxicity, but there is some evidence that they may be at increased risk of relapse, illustrating the tightrope we all know between efficacy and toxicity.
So we showed in a protocol at St. Jude called Total XII that accrued about 190 patients in the late eighties/early nineties, that the cumulative incidence or probability of requiring a dosage decrease was 100 percent in the rare mutant patients, and it was very rare in the majority of patients who are wild type. So when we give 75 milligrams-per-meter-squared 6-MP per day to kids with ALL, that comes from this majority of the population, but a significant proportion, 10 percent of patients, are heterozygote and will require dose decreases in their 6-mercaptopurine to be able to acutely tolerate 6-MP.
But, as was brought up earlier, we don't think about acute toxicities; we're also interested in long-term outcomes. We are really interested in event-free survival.
So when we divided patients into those who had at least one mutant allele for TPMT versus those that were wild type for TPMT, we did see a trend for improved event-free survival in the patients who had one mutant copy. That makes sense. They have higher exposure to TGNs; they have more active drug around, so they should be at lower risk of relapse.
But what we found was that several years out that we were seeing failures, and the failures were not due to relapse of the primary disease, but they were due to a development of a secondary brain tumor, a malignant brain tumor, in almost all cases a glioblastoma.
We looked very hard to find why we saw a high frequency of brain tumors on this protocol and what was it among the patients who did develop secondary brain tumors that was different than the patients who did not. All of the patients who have developed secondary brain tumors received cranial irradiation. So that was a necessary, but not a sufficient hit for the development of this devastating complication.
But the one factor that was statistically predictive of the risk of secondary brain tumor was this single mutation in this single gene. Having one defective allele for TPMT put patients at almost a 50 percent cumulative incidence risk of secondary glioblastoma compared to a still unacceptably high, but a lower risk of the occurrence of this devastating complication in patients who were wild type for TPMT.
So we have been giving irradiation to children with ALL for many, many years, and presumably 10 percent of the patient have always been TPMT heterozygote. So we had to look at why we were seeing this high frequency of this complication of patients with this secondary brain tumor on this protocol. Of course, the fact that TPMP was related made us look at the chemotherapy that was given along with the irradiation. Since thiopurine methyl transferase affects an anti-metabolite, we concentrated on methotrexate and 6-mercaptopurine intensity just during the two-and-a-half-week time period that patients received their cranial irradiation.
These are four successive protocols at St. Jude. On total, 11 where over 200 children received the same dose of irradiation, there's still not been a single secondary brain tumor, but every single dose of anti-metabolite therapy on that protocol was rescued with leucovorin, and there was no systemic anti-metabolite during the period of cranial irradiation, whereas on total 12 patients didn't receive a single dose of leucovorin with any of their intrathecal therapy during the irradiation and they received full-dose systemic methotrexate and 6-mercaptopurine.
Now four years ago you could have asked anyone and they would have thought that the safest drugs to give during the additional carcinogenic hit of cranial irradiation would be anti-metabolite therapy, but we and others have gone on to show that thioguanine nucleotides, especially in these patients, this 10 percent of patients, who have a defect in this single enzyme, are actually acting pretty much like alkalating agents and can be quite carcinogenic.
So an example of how a genetic polymorphism interacts with treatment, interacts with drug therapy, the polymorphisms can be in drug-metabolizing enzymes or obviously many other targets, and there may be non-drug influences, in this case cranial irradiation, but diet, many other things, that may all have to be present to result in an unfortunate intersection in this Venn diagram of risk factors that result in an unacceptable adverse effect.
Of course, what we really want to do is find those factors that will identify patients who will have an improved anti-cancer outcome. This just takes this a step further to show that this is true in many, many cases where drugs are interacting with germline polymorphisms and drug-metabolizing enzymes in targets and transporters and non-drug influences to result in patients at increased risk for thrombosis from asparaginase or increased risk of a fail arrhythmia from a simple drug like erythromycin, and that we really have to do a better job of identifying these germline polymorphisms and how they interact with drug and non-drug influences to more intelligently dose drugs in the future.
So another lesson that this teaches us is that elucidating the clinical implications of each of these polymorphisms can take a long time. We have known since 1980 that 6-MP was a substrate for these enzyme, and we're just starting to learn now how to utilize these drugs. I think, as Ms. Keene alluded to earlier, we don't know the unintended consequences and the long-term effects of many of the therapies that we are using, and that protocol-specific, long-term followup rather than specific protocols aimed at long-term followup are really required in order to understand the long-term effects of the therapies that we are giving to these patients.
So I guess I am a strong believer, and I'm risking what Boyett accuses me of, of being a true believer when I say this, I really think that pharmacogenetics should be incorporated into all clinical trials, not just cancer trials. Clinical trials are expensive. The hard part about doing a clinical trial is doing the clinical trial, enrolling the patients, administering the drugs, keeping track of the therapy, keeping track of the outcome, data managers, research nurses. That is what takes the money.
Getting a tube of blood from every patient is real cheap. They make plastic purple type tubes. You need one tube of blood, and we can genotype everything we need from that for the next few hundred years. I really think that we, as the public, should insist that NIH-funded trials incorporate pharmacogenetics.
Genotyping is expensive now, but it is going to get cheaper and cheaper and cheaper exponentially. It is important to get proper consent for future pharmacogenetic studies, so that we have the option to capitalize on the genetic revolution that's taking place, so we can do better and more detailed pharmacogenetic studies as we learn more in the future. Let's not pretend now that we have any idea what we should be looking at ten years from now or even two years from now.
Just to also say that these polymorphisms affect all elements of supportive care, which are still important for treating kids with cancer. We still lose a huge percentage of patients to infectious complications or nasty side effects of therapy. So this doesn't just affect oncology drugs; it affects everything we do.
I really like this quote from Gery Levy, who is the father of pharmacokinetics. It specifically addresses what we are talking about: that "emphasis should not be focused on population averages, but rather on providing prescribers with the tools to determine the most effective and the safest drug dosage for individual patients with a minimum of trial and error," and pharmacogenetics is part of what can do that. I hope that it will be incorporated into studies with children with cancer.
Thank you for your attention.
CHAIRMAN SANTANA: Thank you, Mary, for a very nice review of this issue of pharmacogenomics and how it impacts some of the issues in pediatric oncology.
What I would like to do is start the discussion. I know we are running a little bit behind time, but I think we do need to have a discussion on these three presentations.
Then I do want to bring the Committee to help answer one of the questions that the FDA has posed to us to answer as it relates to the topics we have discussed this morning: How do we take the information, how do we use this information in clinical trial design for pediatric studies that the Agency may be asked to evaluate in support of indications as the Pediatric Rule is implemented?
I want to finish the discussion on that subject, but I will allow some questions and comments on the presentations earlier. Donna?
DR. PRZEPIORKA: A question for Dr. Kodish, just to bring the early morning into the late morning: When Dr. Hirschfeld gave his presentation, he presented the current paradigm for drug development, which is to at least get some information in the adults on safety and efficacy before offering it to kids. That seemed a reasonable thing from an ethical point of view.
In the late morning we learned all about the tremendous differences between adults and kids, and how the information in adults may not be of that much value in kids, and we really do have to study it. If we delay things, we may delay the benefit to the children.
Dr. Hirschfeld also presented a second paradigm which is to go straight from pre-clinical studies into pediatric studies without the benefit of any information about safety and efficacy in adults. From an ethical point of view, how would you feel about that paradigm?
DR. KODISH: My overall feeling about it, it is, again, hard to generalize because I think there are going to be differences with different drugs and different diseases, but, as a general matter, I think that the prior approach is outdated. I think that we are in a time that it would be more appropriate to look at a paradigm shift that allows us to go directly to Phase I studies in children, especially older children who are able to be part of the assent process. The younger a child is, the more reluctant I would be to proceed along that line of thinking.
But, allowing for all the things that I mentioned this morning, and the special emphasis on the best interest issue for the child, I think it would be reasonable to abandon the old paradigm.
CHAIRMAN SANTANA: Eric, I would agree with that, with the caveat that the first paradigm, which there is some adult data, and we use some of that adult data interpretative in terms of the Phase I design in children. In the absence of that, and doing it in parallel, that you do Phase I trials in parallel or at the same time as you do in adults and children, is that you do have some substantiative data, pre-clinical data, that will give you some idea of where to start and where you are going. Because, if not, then I think we are ignoring the issue of safety, because the whole premise of the Phase I is to define a safe dose and some level of safety built into the clinical investigation that allows you to proceed in a manner that is ethically and scientifically valid.
So I would agree with you that I think we do need to shift and we need to think that maybe it is time to start doing these studies earlier in children or in parallel as they are happening in adults, but with the caveat that I think there has to be a scientific rationale and there has to be some pre-clinical data that would support where we start. Because traditionally what we have done is started based on the adult data, but now, if we don't have that, we are going to have to have some data, probably from pre-clinical models, to support that.
Other than that, I do agree with you. I think we need to start shifting in our thought.
DR. LEEDER: On the other side, I would argue that the older the children, the closer they are -- in terms of strictly from a drug metabolism or a drug clearance and dose requirement issue, the closer that the children are to being adults, the smaller the difference one would expect, and where the largest difference is and the biggest problems are going to be, in the youngest children.
CHAIRMAN SANTANA: Eric?
DR. ROWINSKY: Just being rhetorical, is there any reason -- and maybe we should think about encompassing some of the older children, the teenagers, in adult studies and shifting our infrastructure to incorporate them, perhaps working together, adult and pediatric oncologists, in the same studies.
We are often faced with referrals without any scientific or pharmacological reasons to preclude those teenagers from entering the studies. It seems like that would be one starting point.
CHAIRMAN SANTANA: Malcom or somebody in that corner want to comment on that? Or maybe Peter over here? Mark?
DR. BERNSTEIN: I think that it is an interesting idea that in practice I think it turns out to be very difficult. I think the other thing that has occurred practically, and an issue that I faced and that Peter faces, is that it is going to take us years to catch up with the drugs we haven't had access to, for which there already are adult Phase I data. So that it would be an interesting idea to consider doing Phase I studies in children much earlier than we have done them in the past, but our major problem has been not having access to drugs, for a variety of reasons, which hopefully we can address here, where there are adult data and we just haven't been able to get a hold of the agent for study.
CHAIRMAN SANTANA: Peter?
DR. ADAMSON: I think that we can do a better job of coordinating with adult trials and utilizing adult data to bring Phase I studies into pediatric patients at an earlier time.
I know Frank has proposed doing a combined adult/pediatric Phase I study in certain circumstances where adults will lead the way, but one doesn't need to wait until completion of a trial before one even begins considering a pediatric study.
I think once you have some exposure information in adults, and you have some biologic effect observed in adults that are telling you that you are entering an arena where you may observe biologic effects in children, one can safely begin pediatric studies. I think we may have to move away from some of the traditional dose escalation schemes in children where we can better utilize pharmacologic data as well as adult data to say, "All right, here's our exposure in children. Do we need to take 30 percent increments or should we, in fact, catch up to where the adults are?"
Right now, as Mark has pointed out, the greatest challenge is the tremendous lag in our initiating studies relative to adults, where we seem to be waiting an endless period of time before we even have access.
I think we are going to have to move more rapidly in starting pediatric trials, and therefore, getting some adult data, but, in fact, not necessarily waiting until drugs are on market before we approve performing a pediatric Phase I trial.
CHAIRMAN SANTANA: Malcom?
DR. SMITH: There really is a balance here, though. We want to get good drugs to children as quickly as we can. The balance comes in, when do we have enough information to know that this is really something that is going to be a good anti-cancer drug?
The risk of starting early is that we pick the wrong horse; we pick drugs that are, in fact, going to turn out to be too toxic. There are examples of that where both drugs that have entered adult trials have been determined too toxic, never studied in children, or cases where pediatric trials were started early and where they had to be stopped because, in fact, the drug turned out to be too toxic in adults. So it was dropped for further development.
So I think the key is having enough data from the adult experience and more pre-clinical data relevant to pediatric tumors to pick the good drugs that we really want to study in children. Our only problem in terms of studying drugs in children isn't that we have had delayed access.
Another challenge to us, and one that we don't bemoan but are thankful for, is that there are a much smaller number of children available to participate in Phase I trials and Phase II trials than there are for adults. The article that Steve distributed by Dr. Bruce from Germany made this point very well.
So we can't study every drug that adults are choosing. We have to use information from their studies to pick the ones that, in fact, are going to be best and to use that information to make our pediatric Phase I trials and Phase II trials as efficient and as quick as possible.
DR. GOODMAN: I certainly agree that we need to speed up the -- well, in any case, speed up the process, but it seems to me that what's critical is that we also modify the design or include end-points in the adult Phase I studies to gather information that's relevant to the extrapolation.
I think what happens now is that a lot of pharmacokinetic and pharmacodynamic data is gathered in a pre-clinical setting. Then the Phase I trial does not include continued measurement of these parameters to see the relationship of, say, Cmax or AUCs, or whatever, to toxicities. This can be absolutely critical to the extrapolation and the rapid evolution of the adult knowledge into the pediatric population. We can't just use the MTDs. We need to know what is behind the MTDs.
So it seems that, if we are going to move quicker, we need to be designing both the pre-clinical and the Phase I and the Phase II studies with an eye towards the critical information that's going to be necessary to rationally design the pediatric trials.
CHAIRMAN SANTANA: Eric?
DR. ROWINSKY: It seems to me there has to be a synthesis of two major issues. One, which Malcom brought up, is the selectivity of drugs. I am not so certain that we are at the point that pre-clinical models will ever sort of supply us enough data, at least within the next several years, to really justify a rationale for studies in children, but if there is a drug that we definitely have a good gestalt that it might benefit children, one potential compromise with regard to where we start would be to at least obtain a point in adult studies in which we are seeing some biological activity. That could be the toxic dose low, the point at which we stop our accelerated accrual in accelerated accrual schemes, when we start to see consistent Grade 2 drug-related toxicity; define some element, some target element, be it one of the PK maxims, an AUC, a Cmax, whatever might be valuable for that particular drug, and then hand it off to the pediatric studies with regard to a target.
DR. BALIS: I think the point of what we are here for today is to talk about diseases that are comparable in adults and children, meaning that we would be developing the drug for both populations. Hopefully, we will know better pre-clinically where we are going to target those agents before we start trial, so we have a strong scientific rationale to take it to that disease before there may be a lot of clinical data in adults.
If that is the case, then probably the most important thing that we glean from doing separate trials in children and adults is: What are the differences between them? I think, as Peter brought up, we had proposed doing simultaneous trials to try to overcome some of the barriers to doing that with the current setup. That is, separate trials are done in kids and adults at separate institutions at a separate period of time -- oftentimes, previously, with maybe different definitions of dose-limiting toxicity or MTD, using different labs to assay drugs, different scheduling times to do that, maybe even on different dosing schedules. Certainly the dose levels are always different because pediatric trials are started at 80 percent of the adult MTD, which is usually not a dose level that was studied in adults, and then escalated on a different schedule.
So we ended up defining a dose, not looking at the same dose levels, maybe using different definitions as to how we define an MTD, and not in the end being able to compare either the pharmacokinetics or the clinical data that we derive from those trials.
So I think if we're thinking about developing drugs for a disease that occurs in both populations, we need to have the forethought at least to make the trials, if they're separate, designed in the same ways in terms of all those definitions, and maybe even coordinate them so that the pharmacokinetics can be done at the same places with the same sampling times.
DR. PRZEPIORKA: Just to get back, if we do choose to go directly to the pediatric population, obviously, there is going to be a problem with the first dose of the drug for the first patient in the first study. In the adults there are guidelines for how to extrapolate from the animal models up to the first dose for adults.
Has anyone looked at whether or not that guideline is appropriate for pediatric patients, if you've gone backward from later Phase I studies in kids to what the adult MTD or what the adult first dose was? Is there enough information that we have now that we could actually make similar guidelines for pediatrics?
DR. STEWART: I'm not sure if that data has been published, but I am sure the data -- I am not sure if the data for that analysis has been published, but the data is available to do that kind of analysis. So, typically, for adults, Eric, what is it, 10, 20 percent MELD?
DR. ROWINSKY: Jerry Collins published sort of the landmark paper --
DR. STEWART: Right.
DR. ROWINSKY: -- for adults.
DR. STEWART: At NCI, 1990?
DR. ROWINSKY: Well, I think it's even cancer treatment reports back years ago, where they --
DR. STEWART: I'm not that old. I don't remember that far back.
DR. ROWINSKY: -- where he looked at the ratio of starting doses and doses in which we finished and determined that one-tenth of the LDT was grossly safely for most agents.
I think that definitely can be done with children's studies very easily.
CHAIRMAN SANTANA: But the answer is we don't know. Malcom, do you have a comment?
DR. SMITH: Just a comment on Eric's proposal. It is an interesting idea and it does protect against some things. You are now starting at a dose that is probably sure to be inactive.
When we have done this, the concern or the problem has been that, again, because there's so many more adults with cancer and so many more adults entering Phase I trials, that a center like Eric's is going to have patients lined up and ready to go. So that Phase I study is completed relatively quickly.
So when we start the Phase I study in pediatrics, we start and the adult study has gone three or four dose levels ahead, and you are constantly saying, okay, we need to amend this study to jump up, to skip two or three dose levels, because the adults got ahead. So when you get to the end of the game, in fact, you've basically waited for the adults to determine the MTD; you've adjusted your dose schedule to the adult MTD, and you haven't gained a lot of time, because the adult Phase I studies are inherently conducted quicker in almost all cases than a pediatric study could be. So that is the downside.
The other thing it doesn't protect against is the possibility that the toxicities that you hadn't anticipated in the adult study crop up and the drug that looked promising, in fact, had some anticipated toxicity, and whatever time and effort and pediatric patients that had been entered is all for loss.
DR. ADAMSON: I think there is, unfortunately, a window of opportunity that we can perform Phase I trials in an efficient manner. One of the byproducts of an increasing number of agents on market now is that there are an increasing number of children being treated off-label. Although that is probably a topic of discussion for a different time, what that results is every time a child is treated off-label is potentially one less patient who could have been treated on a Phase I or Phase II study, where we could have learned something.
If we only embark on Phase I trials after a drug is on market, and I am not saying that people are advocating that, but the longer we delay in starting our trials, the greater the risk is that there is going to be an increasing population of children who are exposed to drugs without any information, where we learn nothing from.
So I agree with Malcom it is a fine -- we have to strike a balance between when is it safe to start versus waiting until all the data is in, and then ultimately what we have is a shrinking population of patients who haven't already been exposed in an uncontrolled setting to the drug.
DR. BOYETT: To go back to the question about the existing data that is out there that might relate the pre-clinical model that was used to choose the starting dose in adults that you might be able to relate to pediatric, I think I am less enthusiastic that you may be able to do that. The reason is I think what you have is biased data.
The adult studies that started and got some horrendous toxicity, et cetera, those drugs are out. The only ones that you have data on are the ones that went on and had some success in adults, and then you did them in pediatrics. So I am not as enthusiastic that you are going to have unbiased data to assess that model.
DR. ROWINSKY: What you can do is you can basically look at some of the drugs that have been valid drugs of impact in both diseases and look at the toxic dose low in adults and relate that to the dose where we ended up in children -- here may not be too many of those agents -- to give us an idea of how many, I hate to use the term "wasted resources."
I mean, I think that these trials, Phase I trials, in general, we may be talking a lot about very small numbers of patients who really get ineffective dosages if we utilize that proposal. We are not talking about scores and scores of patients. I mean, entire adult trials reaching the MTD, even in modified Fibronacci conservative dose escalation schemes, generally, about 20 to 30 adult patients, and that is probably a high guesstimate.
I would imagine that we proposed a way in which children could get onto trials, at least an adult toxic dose low with subsequent escalation, it is really not going to subject too many children to ineffective doses that might be unethical or construed as being unethical.
DR. BOYETT: We also could consider some other models like the CRM for studying those.
CHAIRMAN SANTANA: David? Dr. Poplack should be on teleconference. Are you there, David?
DR. POPLACK: Yes.
CHAIRMAN SANTANA: Okay, that's fine. I just want to make sure for the public record that he is listening in.
Donna, did you have another comment?
DR. PRZEPIORKA: Yes. Our good friends at the FDA have recognized that we sometimes cannot interpret the CFR and provide us with guidance documents instead. If one had the opportunity to contribute to a guidance document when it comes to dosing in pediatric studies and PK studies and designing these, I have heard a lot this morning about the difference in surface-to-volume ratio as kids grow up, and the difference in pharmacokinetics.
My question would be: Would you prefer to have all drugs dosed per kilo versus per meter squared in these studies? How many patients would you have to study PKN in order to say these are valid PK?
CHAIRMAN SANTANA: Dr. Coltman, do you want to address that?
DR. COLTMAN: I just wanted to make another point, that with targeted therapies, we have targets in adult diseases that are present in such diverse clinical situations as chronic myelogis leukemia, gastrointestinal stromal tumors, myeloid dysphasia, ovarian cancer, prostate cancer, all targeted by a single molecule that has potential extraordinary effect. So our concept about the differences between adult and pediatric tumors, while they morphologically look different and may behave differently, we should be addressing the target question.
With this targeted therapy, while there is some degree of toxicity, they are in no way comparable to the level of toxicity you see in the standard cytotoxic therapy. I think that is the direction we are going to be going in, although I certainly wouldn't want to muck around with the successful management of pediatric lymphatic leukemia, but there are other issues that need to be addressed, looking toward the future with more targeted therapy.
CHAIRMAN SANTANA: Let me see if I can summarize what I have heard to satisfy the requirement that the FDA has of us.
DR. BAYSSAS: I just wanted to say that for targeted therapies, in adults at the moment, even at Phase II, the dose is not established. So I don't know, the definition of the OBD itself has not currently even in adults been well-established. So I don't see how you can extrapolate from adults to children. You have to wait until the end of Phase II in adults to know which is the OBD in adults that you can extrapolate.
CHAIRMAN SANTANA: Susan?
DR. COHN: Yes, I just wanted to say again, to follow up in terms of all these ideas of trying to speed up getting these drugs into pediatric trials, to go back to what Malcom had mentioned, which is, you know, you don't want to -- fortunately, we have relatively few patients. So you want to make sure that whatever drug you use, you make sure that it is being used in a patient with a disease that will potentially respond, the disease will potentially respond to the therapy.
So the other plea that I think that we ought to consider is in the pre-clinical trials that you don't just use, as pre-clinical trial models, breast cancer, colon cancer, and lung cancer, but that we include some pediatric cancers in these pre-clinical trials. I think that is where we really can get together in terms of making sure the models are representative of pediatric diseases, when you are testing initially these new agents. Then when you do get a little bit of information from the adult studies, we will know that that agent is something that potentially will be effective in some of these pediatric cancers.
DR. ROWINSKY: I hope that no one will be offended by these remarks, and I am glad that Dr. Houghton is not here, but the question is: When is a breast cancer pre-clinically a breast cancer clinically, when is a neuroblastoma -- and I don't think we are at that point yet where we could really -- I think when we see activity pre-clinically that might portend for something of impact in the clinic, but where that tumor, what tumor is going to impact, I --
DR. COHN: Right, no, and I absolutely agree with you, but the reality is that a lot of these drugs are initially tested in these pre-clinical ones. The ones that potentially look -- it may or may not be at all effective in the patient or you may have a totally separate -- but that's, indeed, where we start. I am just saying that if, indeed, we really do want to try to move these drugs faster into clinical trials for children, I think that, whatever basis is being used to define a drug that potentially should be moved into an adult Phase I trial, whatever that laboratory data is, that we ought to try to incorporate those same experiments with pediatric cancer cells.
CHAIRMAN SANTANA: Pat, Dr. Reynolds?
DR. REYNOLDS: I think one of the things that you are missing when you think that there is no value to pre-clinical data is that --
DR. ROWINSKY: I didn't say that.
DR. REYNOLDS: -- whereas you may not be predictive of a response in patients, if you under ideal conditions in a pre-clinical model have a drug that may work in breast cancer, but doesn't do anything at all in neuroblastoma, you can probably think that maybe you don't want to do any trials in neuroblastoma. If it doesn't do anything in a panel of pediatric tumors, then you could think, that's really an adult drug, not a pediatric drug. So it is going to provide us some guidance that we really should seek.
DR. ROWINSKY: I am not certain that we have the knowledge to even imply that those models will even -- that they will be even slightly selective for pediatric tumors. I would be afraid not to try a drug that was inactive in certain pediatric xenografts, or vice versa. I just don't think we are there yet whatsoever. I think pediatric tumors may be oversensitive or may be undersensitive with regard to selectivity.
DR. COHN: Yes, but I was going to say the reality is, though, as we were saying, because we have so few patients and there are so many drugs, we have to prioritize.
DR. REYNOLDS: That's the point exactly. Pediatrics is not the adult community where you have the ability to say, well, let's make sure this is not an active drug and do a study. We have to be able to select before we get to the patient. The only way to do that is intelligent pre-clinical data.
DR. ROWINSKY: Well, you're preaching to the converted, but I'm not so certain that the models, the xenograft models, are really the key.
DR. REYNOLDS: I agree with you, but I think that that's why we need to study pre-clinical models and find out what is going to work. Unless we do that in the context of or together with clinical trials, we will never learn anything.
CHAIRMAN SANTANA: Let me see if I can summarize because the hour is running late, and certainly I hope I express the view of the Committee when I make the summary.
First of all, I think we recognize that the world is imperfect. All these models and all these trials are not perfect. It limits our ability, when we are making decisions of how well we ultimately end up at the end of the day.
But recognizing the limitations, I think I get a strong sense from the community here that, as the FDA decides, and other groups decide, what studies should be done in pediatrics, that pre-clinical data that is relevant to the diseases under consideration in children are paramount. I think that is one very important point.
That doesn't mean that we have to go out there and do every single drug on neuroblastoma xenografts, but there should be some scientific validity in the pre-clinical models as it relates to the clinical condition in children. I think that is one concept that I think was fairly well expressed by the Committee.
The second, I think we have limitation of resources, and by "resources," we not only mean economic, we also mean in terms of patients. It is a very limited population. We cannot do everything that we think we need to do. Given that, I think we have to allow the clinical investigators who are experts in this field, to allow us to make those decisions in terms of what drugs based on the relevant information are the drugs that they want to prioritize and they want to test.
That obviously means that there will be some drugs that will not be tested, but I think we do have to have some confidence in the clinical investigators and scientific community in pediatric oncology of what drugs they want to prioritize, given the limited resources.
Having said that, then I think there is potentially no fast rule. There may be different models that the Agency could use in terms of applying the rules in terms of some of the studies that they will request. I heard Frank say that, if the disease is similar in adults and children, then I think under that model scientifically it may be appropriate to allow Phase I studies to occur concurrently, because there may be some differences in toxicity. So just waiting until the adult trial is done is probably, given the same disease, it is not something that probably should do.
But in those scenarios, probably parallel studies or concurrent studies for the same disease, the same biology, those studies should occur concurrently. In all the others, I think history has served us well. I mean, I think some adult data and some pre-clinical data has allowed us to define some dose levels which are reasonable for us to start. That will be completely different with biologics. I don't think those rules automatically apply to biologics. I think in biologics we may have to think of a completely different paradigm, maybe doing studies concurrently or some other way. I am not an expert in that area of biologics, but I think in biologics, which is a topic that I think we do need to discuss maybe further this afternoon, the paradigm may have to be a little different.
That is what I think I heard the Committee say this morning.
DR. HIRSCHFELD: Thank you, Dr. Santana.
I would want to, just for the purposes of focusing the discussion, clarify again the conditions where the rule is triggered. That is, the diseases would be considered the same or essentially the same, sufficiently similar, plus the prospective therapy should be considered a therapeutic advance. That is, the rule is not meant to be triggered for -- and I don't want to malign any particular class of drugs, so I will try to avoid that, but it should not be triggered for any "me-too" drugs.
That is, if they are already -- and the way the Agency as a whole has interpreted -- if there is already a drug of the same class that is labeled for children, then the bar becomes much higher for subsequent drugs of that class in order to have this triggered.
So, within those constraints then, I think we can then focus our discussion, and the broader discussion of Phase I studies and how one relates diseases is in the background, and this is a special case.
CHAIRMAN SANTANA: If there are no further comments, we will adjourn for lunch, and we will try to reconvene at one o'clock, so we can keep ourselves on time. Thank you.
(Whereupon, the foregoing matter went off the record for lunch at 12:17 p.m. and went back on the record at 1:14 p.m.)
CHAIRMAN SANTANA: Okay, let's go ahead and get started.
Just in case there is anybody new in the audience or at the table, let's go ahead and reintroduce ourselves. Dr. Rackoff, can you start from the corner over there, please?
DR. RACKOFF: Wayne Rackoff. I'm a pediatric oncologist in oncology drug development at Janssen Research Foundation.
DR. BAYSSAS: Martine Bayssas. I work with Debiopharm in Switzerland. I'm a medical oncologist.
DR. COLTMAN: Chuck Coltman. I'm a medical oncologist from San Antonio, Texas, and Chair of the Southwest Oncology Group.
DR. BALIS: Frank Balis, Pediatric Oncology Branch, National Cancer Institute.
DR. KODISH: Eric Kodish, Rainbow Center for Pediatric Ethics in Cleveland, Ohio.
DR. SMITH: Malcom Smith, Cancer Therapy Evaluation Program, NCI.
DR. BERNSTEIN: Mark Bernstein, Pediatric Oncology at the University of Montreal and the Children's Oncology Group.
DR. STEWART: Clinton Stewart, Department of Pharmaceutical Sciences, St. Jude Children's Research Hospital, Memphis, Tennessee.
DR. LEEDER: Steve Leeder, Clinical Pharmacology at Children's Mercy Hospital in Kansas City, Missouri.
DR. ROWINSKY: Eric Rowinsky, Medical Oncology at the Clinical Research Institute for Drug Development in San Antonio.
DR. GOODMAN: Steve Goodman, lapsed pediatrician, now biostatistician -- (laughter) -- at Hopkins Oncology Biostatistics.
DR. KORN: Ed Korn, Biometric Research Branch, NCI.
DR. GOODMAN: Stephen George, Duke University Medical Center and ODAC member.
DR. BOYETT: James Boyett, St. Jude Children's Research Hospital, Chair of Biostatistics.
DR. PRZEPIORKA: Donna Przepiorka, Center for Cell and Gene Therapy, Baylor College of Medicine, Houston, and ODAC member.
CHAIRMAN SANTANA: Victor Santana, Pediatric Oncologist from St. Jude's.
MS. ETTINGER: Alice Ettinger, pediatric nurse practitioner from New Brunswick, New Jersey.
DR. WEINER: Susan Weiner. I'm a lapsed developmental psychologist, was a parent, and am now a patient advocate.
DR. PELUSI: Jody Pelusi, oncology nurse practitioner, Phoenix Indian Medical Center, and I sit as the consumer representative and also an ODAC member.
DR. REYNOLDS: Pat Reynolds, Hematology and Oncology, Children's Hospital, Los Angeles.
DR. COHN: Susan Cohn, Children's Hospital in Chicago.
MS. KEENE: Nancy Keene, patient advocate.
DR. ADAMSON: Peter Adamson, Children's Hospital, Philadelphia, and Children's Oncology Group.
DR. HIRSCHFELD: I want to yield my time to my distinguished colleague from Los Angeles to identify himself, and then we'll return.
DR. FINKLESTEIN: Thank you, sir. Jerry Finklestein, Pediatric Oncologist, Long Beach, California.
DR. HIRSCHFELD: Steven Hirschfeld, FDA.
DR. PAZDUR: Richard Pazdur, FDA.
CHAIRMAN SANTANA: Thank you.
This afternoon we are going to cover two topics. One is issues of clinical trial design as it relates to statistical design and validation of end-points, and then I will briefly talk a little bit about Phase II window studies, and then we will have a discussion.
So, with that, I will introduce Dr. Goodman. Who's going to go first? You are? Dr. Goodman, please.
DR. HIRSCHFELD: While the screen is set up, I believe Dr. David Poplack is on the telephone, too, and should be identified.
CHAIRMAN SANTANA: Okay, David, can you hear us, David? Well, he's on the telephone. He can't hear us.
I also forgot to mention, we do have some time for a public open hearing. If there is anybody in the audience that wishes to address the Committee, please come to the microphone and identify yourself.
CHAIRMAN SANTANA: If there is no one, then we will proceed with Dr. Goodman's presentation. Thank you.
DR. GOODMAN: I want to thank Dr. Hirschfeld very much for inviting me. It would seem that you heard from my intro that I had spent my whole life preparing just for this meeting, even though I am not a member of the panel, since I started off as a pediatrician and then decided that I would serve the health of the world's children by not touching them anymore.
Went into biostatistics and clinical trials and oncology and also ethics of clinical trial design.
CHAIRMAN SANTANA: Dr. Goodman, let me remind you that anything you say is public record, so be careful what you say.
DR. GOODMAN: It's all in my C.V.
But one area where I never worked was at the intersection of all these things, which this Committee represents. So I have found this discussion very, very interesting.
Now when Dr. Hirschfeld asked us to talk, I think originally he did propose the topic there, which just like the intersection was one area where I didn't feel particularly expert. So we actually had a back and forth about what we would talk about, and this is actually literally from the email. So I just wanted to show what the charge was for us to talk about.
It will not be specifically on the title that was given. This is what he wanted me to comment on: Can one apply Bayesian analysis where the a priori data comes from an adult population and the new data come from pediatric population? So that's actually what I will be spending a few minutes talking about.
He sent this email on Halloween. Maybe the other title was the trick or treat version. We'll switch.
So the title is "What can Bayesian methods do for us?" I will just be talking very, very generally. I will warn you ahead of time, I'm just going to show one equation; it's not to scare you. It's not the Halloween component.
So, first of all, what are Bayesian methods? Well, the simplest definition is they are methods based on Bayes' theorem. So what is Bayes' theorem. Here is the scary part.
You can just forget about this. I will just translate this into words. It is some prior knowledge plus data from the study that you're doing which gives you your final summary knowledge. So that is the simplest way to summarize it.
Now in English, we can talk about it in a variety of ways. Bayesian methods are approaches that combine information of different types in a statistically-justifiable way. Another way to look at it is it provides a formal way to make statistical inferences from a clinical trial by incorporating prior knowledge.
Different people have a different perspective on what the calculations involve. You can look at it as a calculus of uncertainty; that is, that the most important thing it does is it properly represents our uncertainty at the end of the day, given how uncertain we were at the beginning.
You can look at it as a calculus of belief, that it tells you what you should believe at the end of the day, given what you believed before you started looking at the data.
And, finally, you can also look at it as a calculus of evidence; that is, a proper way to measure the strength of the data and how strongly it points to one hypothesis or not.
Different people fall into different schools, but it has components of all of these things, which is the important thing to recognize.
So in what settings have Bayesian designs or analyses been used? Well, I will tell you, if you will look at the statistical literature, you wouldn't imagine that it is this somewhat poor cousin of standard methods that it is in the clinical literature, because these days Bayesian applications and Bayesian methods occupy probably pretty close to 50 percent of what you find in a table of contents in any modern statistical journal.
The actual applications in medicine have been pretty much across the board. The heaviest representation, you see if you do a search on Medline, is in the area of pharmacokinetics. It has also been applied in the Phase I arena; mainly, in the form of the continual reassessment method, which I will just talk about very briefly later in Phase II. These are just representative authors who have written about this. In Phase II studies it has been used combining what is known historically about response or cure rates to a particular trial. It has been used in Phase III studies by a whole host of statisticians, and it is also used extensively in med analysis.
This is just a representation of the number of articles that use the "Bayes" that's appeared in the medical journals over the past really 40 years, and you see that it is pretty much exploding, and this doesn't necessarily capture all the articles that do.
But, on the other hand, you also see that the scale here, which represents 250 at the top, is pretty tiny if you compare it to the number of articles that are published in the medical literature. So depending on what discipline you are in, it is actually, except in the area of pharmacokinetics, fairly unlikely that you will run across actually a published application of Bayesian methods. So I wouldn't be surprised if many of you might be unfamiliar with them except through the CRM, the Phase I design.
Now let's start off, before I start telling you what they can do, let's start off by saying what Bayesian methods cannot do with respect to the charge. Thank you very much. What they cannot do is tell us, in the absence of information, how alike children and adults are and how relevant adult information is for children. It cannot make this extrapolation for us. It is sort of, sometimes Bayesian methods are looked at as a way to produce knowledge where other sources of knowledge are not available. What it is useful for is encoding or representing knowledge that we actually have. So it is not going to help us make this extrapolation if we can't do it biologically or we can't do it clinically or we can't do it empirically.
So the charge again: Can we apply Bayesian analysis where a priori data comes from the adult population? Yes, but only if you make an a priori judgment about how relevant the information is for children from the adult population. We have heard a lot this morning about the foundations for those sorts of judgments, but we have to make those judgments first before we can apply the calculus.
Now what sort of information do we look for coming from adults? I was sort of alluding to this in my comment before about the kinds of things that we need to include in adult studies if we are going to have an eye to extrapolate them to children. Obviously, we have information on the pharmacokinetics, although for every one of these we could probably, and we have already had, a talk on why these things are different.
But the issue is not that these parameters are different, but how they are related. They can be different, but we can know consistently for a certain class of drugs that, because they are metabolized in a certain way, the average dose in children should be something greater than that in adults. That guess plus its uncertainty can be reflected in a Bayesian calculation.
We heard about pharmacogenomics. We can learn about dose toxicity relationship from adults, about the types of toxicity and frequency of toxicity. Again, we don't know things perfectly, but even the imperfection can be represented to some extend. We learn something about efficacy. We learned about the effects of patient characteristics, whether they are genetic or clinical on all of the above. Finally, we learn about the uncertainty in all of the above. We rarely know any of these things, even in adults with certainty.
So what allows extrapolation to children? Well, I don't need to tell you this. Obviously, empirical comparisons, and we heard some suggestion about the kinds of studies, and we got articles on comparison of MTDs in adult or pediatric populations; basically, just outcomes comparisons, knowledge of mechanisms in adults versus children, known adult/child biologic and clinical properties of analogous drugs, and known sensitivity of children to specific toxicities, and we could probably make an almost endless list if we went around the room.
Now how is prior information represented in Bayesian analysis? They're represented as probability distributions on key parameters that express both our best guess and our degree of uncertainty. These days there is a lot of emphasis not just on a single representation, but on doing sensitivity analysis; that is, representing uncertainty by showing a whole range of possible representations of uncertainty. Because for every curve, while we might say it represents a certain amount of uncertainty about a main effect, the flip side of the coin is it will represent certainty as well; that is, our certainty that an effect of a given size is "X" probable.
You could look at that as certainty or uncertainty. If you say that something has a 50 percent chance of happening, that actually can be certainty. If I say that this coin has a 50 percent chance of landing heads, that's actually a pretty precise estimate. So there are two sides to that.
So what are these key parameters that we might have guesses about? We might have guesses about the MTD, about the response or survival rate, about a toxicity rate, about the shape or slope of a dose toxicity curve, which is particularly the case which was used in the Continual Reassessment Method, or we can have guesses about pharmacokinetic parameters.
Here's an example of the kinds of curves that we might draw. Here we have on the "X" axis the pediatric over the adult MTD. So one represents an MTD that is exactly the same in the two populations measured per kilo or per BMI, or whatever, or per-meter-squared. This would represent a pediatric dose that was twice as high; down here would represent a pediatric dose that was half the adult dose, et cetera.
So one possible representation might be this: This is just a hypothetical curve which would say that our prior guess is that the MTD is the same as the adult MTD, but we think it could be -- and this would actually represent what we would call a fairly informative "prior" because it would restrict the range of plausible values from about 2, or somewhat less than 2, about 1.8, to about 1.5. So it would say that a priori our guess is that the pediatric dose does not vary by more than half the adult MTD and doesn't go below one-half or above twice the adult MTD.
This curve would be formally incorporated into the calculations and combined with the accumulating data from a study that you were doing. Now if you didn't have that much confidence, you might draw your prior probability curve like this, where you say, well, my best guess is that they are equal, but I will allow some probability that they are anywhere up to three times, the MTD is three times as high in the kids or three times lower, or you could draw almost anything else that you wanted.
Similarly, you could reflect some of the information that we saw this morning about drugs that are metabolized in certain ways by saying our best guess is that the pediatric MTD is twice as high as the adults, going down as low as half as big and going up to six times as high. Obviously, this would be deemed to be somewhat improbable, but it allows that if the information from your study accumulates strongly enough, you will allow that possibility, and a more informative guess would look like that.
So this is how prior information gets fed into the Bayesian machine, by starting off with curves like this. Now a lot of times what I hear, and the main place where I interact with clinicians in doing Bayesian analysis is with the Continual Reassessment Method, where we talk about what the desired toxicity is and what appropriate balance is, and what the toxicity might be. A lot of times the investigator will say, "Well, I really have no idea. I don't come with that knowledge," and that, of course, is the Achilles' heel or is thought to be the Achilles' heel, of Bayesian analysis: that you have to have some sort of prior idea of what you are going to see.
The fact is that most people do have some pretty good guess about at least what are the extremes; that is, what is implausible. I will give you an example. Actually, well, this is just to say that these prior probably distributions are pretty much equivalent to information from prior individuals. Some people describe it as made-up data.
But I want to make the point that actually weak knowledge corresponds to a lot of individuals, particularly when we are talking about pediatric trials. For example, if we are talking about inference about a response rate or a survival rate, if you have pretty high confidence that the cure rate lies within a 40 percent range -- that is, you say it's unlikely to be less than 20 percent and it's unlikely to be greater than 60 percent, and often clinicians can, I would think people around this table, make statements like that with a fair bit of confidence.
This is roughly equivalent to 25 patients' worth of experimental information. Now that is a lot of information. That is not nothing.
This would be a typical situation where the clinical investigator might come and say, "I really don't know anything." But, in fact, they know a tremendous amount compared to truly nothing.
The confidence of the cure rate lies within a 20 percent range; that is, somewhere between 20 percent and 40 percent, corresponds somewhere in the vicinity of about 100 patients' worth of experimental information.
So if you come to a study with this knowledge, based on either knowledge of other treatments of this disease, the disease, the drug, or whatever, it is actually equivalent to and potentially saves a fair number of subjects from subsequent experimentation.
So what do these methods do for us? I am just going to sort of summarize this in a very broad-brush way.
First of all, they properly account for uncertainty and knowledge in both previous and current experimental data.
They minimize the amount of information necessary from the current experiment, but this, of course, is only of value if your "priors" are reasonably accurate. If your "priors" are total guesses, then you will find that the amount of diffuseness that you have to introduce to accurately represent your prior uncertainty ends up producing sample sizes of roughly the same order that you would get now, because you're operating from nothing. So you have to have a certain amount of humility sometimes, all the time.
Another thing it does is it promotes -- I shouldn't say research treatments -- research designs and the choice of treatment for any particular subject and choice of dose that reflects, hopefully, as closely as possible, our best guess about what would be the best for the child based on all prior information. It allows a certain flexibility in design because all Bayesian designs can be adaptive; that is, responsive to data as it comes in.
Now I have found the most useful thing that both Bayesian design and analysis does is that it encourages extremely valuable discussions about prior knowledge on uncertainty and about the goals of the study. It is actually this discussion that is more valuable than anything else that goes on, and it is not necessarily discussion that is stimulated by asking questions about, well, how much power would you like for this effect side, which I find to be a fairly empty exercise.
When I send people back to talk with their colleagues about what sort of toxicity frequency would be acceptable, they come back, or if I attend, I see a really fascinating discussion around the table talking about things that they have actually never formally talked about with their colleagues before. They find it very, very highly informative. Sometimes they will tell me to start, well, we'll accept a toxicity of up to 30 percent of the patients. Sometimes they will come back, after discussion with their colleagues, and say, well, no, it's really 5 percent, or vice versa.
So it is these sort of discussions about what the real content of prior knowledge is and what different people bring to the table. The most valuable "priors" are, of course, those based on collective expertise, not just one person's hunch or interpretation. This is an extraordinarily useful exercise. It brings out a lot of things that sometimes are not brought out when sort of the cookie-cutter, fill-in-the-blanks methods are used for designing experiments.
Now I want to say, in a nod to all the superb statisticians, many of whom I have learned from, who are at my right here and my left, that most standard approaches should flexibly and with common sense, which is how anybody who has worked in clinical trials more than five months is forced to operate, can become operationally indistinguishable from Bayesian ones. So it is very, very possible to get the same sort of operational results by not hewing to any extreme philosophies about how studies should be run.
But sometimes this requires a bit more "ad hoc-ery." That is, the standard methods don't necessarily have formal ways of representing prior information. We sort of make up ways or we put implicitly into the design our beliefs about what we think are plausible effects, and we do it in the form of also how many control patients we might choose, et cetera, et cetera. Those methods don't always have a coherent theoretic foundation.
So, to bring it full circle, can we apply Bayesian analysis where the a prior data comes from the adult patients? I would say, as I said before, yes, but only if the adult data is deemed relevant or informative, and more empirical studies of this relevance need to be conducted, and they need to be ongoing; that is, this needs to be a continuing area for study.
As every new agent and every new mechanism comes out, we are going to have a whole new set of principles upon which we base our judgment about this extrapolation, and it is those principles that will guide the way we represent our knowledge in Bayesian analysis that borrow strength from the adult studies and are used in the pediatric studies.
So, with that, I will stop, and I guess I will ask for any burning questions right now because Ed will not be talking directly on this subject.
CHAIRMAN SANTANA: So any questions for Dr. Goodman or comments? Jerry?
DR. FINKLESTEIN: From a clinical point of view, if you use Bayesian analysis and if some of the unpublished data gets verbalized, that perhaps the Phase I data on adults is very close to the Phase I data in children, as we now -- we have been using the 80 percent rule, but if it is pretty close to 100 percent, how would the Bayesian analysis help us to fortify this impression?
DR. GOODMAN: Well, I'm going to think of the Bayesian approach as, I think what that would do -- I mean, much of what I'll say is just common sense; you don't require Bayesian perspective to implement it.
The first thing you do, if you actually had pretty high confidence that the MTDs in the two groups were the same, you would start at the same MTD in children that you had in adults. It might allow you to -- you would also represent that confidence in the form, and I showed the "priors" before, in the form of "priors" that were fairly tight around the hypothesized MTD that was equal to the adults.
So you would have more confidence about starting, which, again, you don't need Bayesian analysis to tell you, but it might lead you to stop the trial perhaps a bit earlier because you would essentially have, in having a very tight high confidence, that is, in a sense, equivalent to adding subjects to Phase I study.
Now if you don't think that is legitimate or you think that this particular agent doesn't operate in the same way as the agents upon which that original guess was based, that very high confidence that you might have to start might not be justified. So you have to look very, very carefully at the basis for that confidence (a), and (b) at whether this new agent that you are trying actually falls within the class of agents or class of mechanisms that the studies upon which that confidence is based. That is the best way I would say it.
So if you think that this is the same kind of agent that has shown equal MTDs in the past, you will get a Phase I study that is smaller in general than you would if you did it traditional method, which is sort of stand alone and we'll use the same sample size. But, of course, that is only of value if your prior guess is right and reasonable.
CHAIRMAN SANTANA: Wayne?
DR. RACKOFF: Operationally, the real question for this set of meetings has been how to get drugs to kids sooner and then get them in and out of trials faster. So, as I read your two papers and hear you speak, the question that remains is: What would you do operationally with Bayesian analysis that would, access question aside, expedite the clinical trials process? Because, as you said, if the assumptions are made reasonably, standard methods and Bayesian methods come together. So what advantage would there be in terms of speeding trials along?
DR. GOODMAN: Well, the kinds of things that you would have -- I mean, I will leave it to some of my colleagues here to maybe suggest other things, but the most direct way of adapting a standard method to do in some sense operationally what a Bayesian method would do would be to have a set of hypothetical or real data which you incorporate into your current analysis, just by averaging it in or pulling it in. It is in a sense making believe that you have a larger experiment than you actually have.
And the way this is typically done is just taking prior clinical trial data or Phase I data, or whatever, and giving it a certain weight relative to the weight of your own trial. So, in that sense, you can simulate it.
Bayesian, as I said here, in my mind, Bayesian methods offer a better and more flexible way to represent, first of all, multiple sources of uncertainty and represent certainly in a variety of different ways. But they will come close, but you have to be able to hypothesize that the information you have or the experiments you have are already relevant to this question.
DR. RACKOFF: In this setting, if I can follow up, Steve, in this setting that is almost a given, because to invoke the Peds Rule, the assumption is already in place that there is some linkage between the adult disease and the disease in children.
DR. GOODMAN: When you say "linkage," you have to be very, very precise. When you say "linkage," I mean, the efficacy of the drug, the mechanism of the drug, the toxicities of the drug, survival, I mean --
DR. RACKOFF: To take an example, if AraC were being developed today, in adult AML you would have data and now you are moving it into pediatrics. I mean, my sense is that if you use the prior information, given the effect sizes, you would probably have trials that would require fewer subjects and, therefore, be finished sooner. Is that correct?
DR. GOODMAN: Yes. Yes, if you believe.
DR. RACKOFF: Right.
DR. GOODMAN: Absolutely. I mean, that is essentially, if you add in statistically information in the form of subject that you haven't experimented on, you have effectively a larger sample size, even though you haven't experimented on more subjects. But, of course, it only buys you -- it is only advantageous if that prior information is relevant, and that is a judgment that has to be made by people who aren't statisticians.
DR. HIRSCHFELD: This actually segues, I think, with Dr. Rackoff's question, and I wanted to ask Dr. Goodman to make a point that he stated earlier and will rephrase it. In terms of practical implications, if there were coordination among the people designing the adult Phase I studies and the pediatric Phase I studies, and the appropriate data were being collected in the adult Phase I studies that could be utilized in analysis for pediatric studies, would that, then, facilitate this type of approach?
DR. GOODMAN: Yes.
DR. HIRSCHFELD: Thank you.
DR. ROWINSKY: I mean, the methodology for the proposed pediatric trial, as far as dose escalation, would be entirely unconventional in that you would be selecting doses based upon a Bayesian method. So I am just trying to understand operationally.
DR. GOODMAN: Yes.
DR. ROWINSKY: And you are going to be drawing, actually defining a precise dose toxicity curve which will allow you to discern an MTD --
DR. GOODMAN: Right.
DR. ROWINSKY: -- with a certain order of confidence with smaller numbers of patients?
DR. GOODMAN: Right. One other thing, I mean to address the question before, this is maybe getting beyond where we want to use them, but Bayesian methods can be very, very effective in certain settings at combining different sources and different kinds of information that all have a bearing on the inference you want to make.
For example, if the drug is similar, if the basis for our extrapolation is similarities in metabolisms, similarities in drug targets, or variations thereof, you can construct Bayesian hierarchical belief networks that sort of amalgamate all this information in ways that are very, very difficult to do using standard methods. I mean, when I said make up patients with a certain outcome, that's assuming that your uncertainty is focused on just one particular outcome, like the MTD or the AUC, or something like that.
But it sometimes the case that we learn that a number of things are related, and that kind of information, again, because the Bayesian methodology is sort of a belief propagation or an uncertainty propagation model, is easier to represent, much more easy to represent, under a Bayesian model.
However, those models then become more and more dependent on assumptions, and you may or may not want to design a trial or save many patients based on those assumptions. This is what I meant when I said, depending on the kind of uncertainty and the levels of uncertainty, it is much more awkward to start constructing Bayesian equivalents using standard methods because standard methods don't fundamentally allow you to express uncertainty about these unknowns in the same way that Bayesian methods do. It is an issue of calculus.
CHAIRMAN SANTANA: Dr. Boyett?
DR. BOYETT: Actually, not being a Bayesian and not wanting to endorse the methodology broadly, I will say that in a Phase I setting I think it is entirely appropriate. The traditional Phase I design that we have used, three and six, et cetera, and saying that we know what the maximum tolerated dose is, that's not well-defined. It doesn't define anything.
On the other hand, the CRM method, in terms of actually realizing that what you are doing is estimating a dose toxicity response curve, if you will, and the method allows you to model that dose response, and it is not picking an MTD -- it is saying that I want to estimate the dose at which I have some confidence that maybe 20 percent of the patients might experience, unacceptable toxicity, or 30 percent of the patients might experience unacceptable toxicity. I think that is a far improvement over what we have done in the past with our Phase I design.
Secondly, I can see how that, if you had the actual data from a Phase I trial in adults, that you use that to get a prior estimate of what this dose response curve might be in children, though you might weight it only 50 percent of what it would be if you actually had children there, I can see how that can definitely be an advantage.
CHAIRMAN SANTANA: Donna?
DR. PRZEPIORKA: I guess if we get down to the nitty-gritty, if we usually 9 to 18 patients in a standard 3-plus-3 design, what is the smallest number of patients that you would expect in a CRM design, if, as Dr. Finklestein says, the MTD in the kids is going to be very close if not exactly the same as in the adults?
DR. GOODMAN: Well, in the typical CRM designs that I use, actually, I use very broad "priors," very little prior belief. It can be shown in lots of simulations and in reality that the CRM designs tend to end at roughly the same time as standard designs. They are most reliable when they go up to about 24, between 20 and 24 patients. That is with no prior belief.
I have actually never run one or simulated one, sort of assuming that we had the equivalent of 10 patients of information already, but some of that time is used in building up from lowest dose, and some of that time is spent in making the estimate more accurate when you start with no estimate, no information about what the estimate should be.
So if you are starting very, very close to the dose, that is, either at or very close to the dose, within one dose level, and you are starting with information that is equivalent to something on the order of five to ten patients, then I would expect that the study would end in more like 12 patients. You know, this is really an off-the-cuff guess.
CHAIRMAN SANTANA: Dr. George I think had a comment, too.
DR. GEORGE: I would just like to comment, and maybe Steve would like to comment on this, too. I think the Bayesian approach does have a lot of attractions in pediatric oncology as well as oncology in general, but the main advantages in terms of the basic problem we are facing here with very small numbers is that its main advantage would appear, as you just mentioned in your example, when you have very precise or more precise prior information, and that's fine, although therein lies the risk.
Something else you said earlier was mentioned briefly the sensitivity kind of analysis; that is, that is where your real worry is. If you think you have real precise information and, in fact, you don't because you are making some faulty assumptions, you can run into problems. That is presumably why you use very diffuse priors in your CRM approach.
I just wondered if you have any comments on whether the Bayesian approach would, in fact, be able to save time and patient resources, particularly in the regulatory setting. In the scientific setting I can see how it might, but if you are trying to, if we are talking here from the FDA's perspective, I think we may be talking philosophically, but not practically.
DR. GOODMAN: Well, I would actually argue the opposite. I would say that to conduct analyses without formal incorporation of everything that we know is to create, first of all, a lot of ethical tension because there is this clinical sense, and sometimes it is hard to articulate, and statisticians can help make it more formal, that starting at a particular dose or treating with a particular agent does not represent what they know biologically or clinically, is not what they would choose for their own child if the child had the same illness.
If, to the extent we can, we incorporate all that we can rationally incorporate into the design, reflecting our prior knowledge, it is much more likely that the child will be treated either at doses or with agents that are most likely to be effective. So I think it reduces the ethical tension. It, in the end, also reflects our -- I mean, the reason we are here today is because we believe that there is relevance of the adult information to the children. Otherwise, we wouldn't have to -- we would always be starting anew. We wouldn't even be necessarily starting with the same agents. We would be testing, you know, have two completely different, independent panels of testing. So we believe that they are highly related.
I think that, because we are so sensitive -- I mean, this beneficence argument really translates into trying to do the best we can to, even sometimes at the expense of, I would argue, being -- this is going to sound wrong -- of being in the long term correct, that is, this allows you to choose what you think is best, given everything that you know now, for the child with less emphasis on the value to society. If that is the emphasis of the pediatric clinical trials, then the emphasis switches from having the maximum sample sizes to incorporating as much information as possible in the treatment of each individual patient, and then making as much use of the collective of information as we can at the end, but the priority is on beneficence for the individual.
I think that the Bayesian approach is the most coherent way -- I'm not saying the only way, but the most coherent way to represent all that we know when we come to the next subject. So it really does have to do, I think, with this balance between societal interests and the interests of the individual.
As I said, if you use smaller numbers of patients, in the end you are going to be based with empirical data that is based on smaller numbers than we have for adults. Part of that is necessary because there may be a fewer number of children subjects.
One might argue that that is less reliable knowledge, but the choice for each child will be more likely, given everything that we know now, given what any rational parent or clinician would do, to reflect what we think is best for that child. So I do think it represents a different balance of interests than we would have in adult populations, but the kind of balance that, as was suggested this morning, we might want to strike for children.
DR. HIRSCHFELD: I would like to add to that that it is our goal, and I hope our practice, that there is no distinction between the best science and any regulatory policies or actions, but that they are synonymous.
CHAIRMAN SANTANA: Malcom, do you want to comment?
DR. SMITH: Yes. I would just say we support Phase I trials, pediatric trials, both with the CRM method that Jim has taken the lead in as well as the standard 3-and-6 method.
I feel less of the ethical tension that you describe. I think if we were starting at doses that were very low, then I think it would be a much stronger argument that we needed something like the Continual Reassessment Method. Starting at 80 percent of the adult MTD though, we are at a dose that, in fact, is very close to the adult Phase II dose, and there is enough experience on both sides, the Phase I doses in children being both lower and higher than that in adults, that I think, either way, either the standard 3-and-6 method or the CRM method is an acceptable, ethical way of conducting a Phase I trial that really does provide as reasonable a chance of benefit for each participating child as we can hope, given the context of the Phase I trial.
DR. GOODMAN: This is getting away from Bayesian issues and getting more into Phase I issues. I would just say I disagree only to the extent that the 3-and-6 method does not allow you to smoothly or coherently calibrate the target toxicity that you want to reach. That is, it doesn't allow you very easily to say the optimal balance here is represented by a toxicity rate of 5 percent or 50 percent. It is very, very difficult to do that. In fact, it is essentially impossible.
DR. ROWINSKY: But you can't do that in a Phase I trial anyway, given --
DR. GOODMAN: You can with a CRM.
DR. ROWINSKY: Not in terms of efficacy and activity. You can't --
DR. GOODMAN: No, no, no.
DR. ROWINSKY: Only with respect to toxicity.
DR. GOODMAN: With respect to toxicity, yes. So, in that sense, that is one of my stronger arguments for the CRM, not necessarily the technology, but the fact that it can be tuned properly to reflect what the investigator thinks is the appropriate balance of risk and benefits.
CHAIRMAN SANTANA: Okay, I would like to go ahead and invite Dr. Korn to the podium for his presentation.
DR. GOODMAN: Dr. Korn may not agree with that.
I just have this sneaking suspicion that I'm going to keep the microphone.
DR. KORN: Well, first, I would like to thank Dr. Hirschfeld for inviting me to come talk, and we are going to continue with the bait-and-switch --
CHAIRMAN SANTANA: We can't hear you very well. Can you check your microphone or move a little bit closer?
DR. KORN: Did you turn it off on me?
So we are going to continue with the bait-and-switch approach, and I'm also not going to talk about what was on the agenda.
But that's only because Dr. Hirschfeld asked me to talk about Phase I trials, and I don't completely disagree with Steve.
Let me start with cytotoxic agents, where things are a little bit simpler. So we usually treat cohorts of patients with escalating doses until unacceptable toxicity is seen and then back off. The rationale, of course, is that increased dosage of an agent will offer more anti-tumor benefit, provided the dose has acceptable toxicity.
Then we have heard talk about the standard design where you use cohorts of three or six patients. There are accelerated designs. I consider the CRM perhaps an accelerated designs, where you treat less than three patients in each dose level to begin with, or possibly you have bigger jumps between the dose levels, including some not dose-limiting toxicity, but some Grade 2 toxicity, and then you treat more patients at the dose levels or narrow the distance between the dose levels.
As Malcom just said, and which I agree with, I think these designs are most useful when you have no good idea of a starting dose level, and you don't escalate through ten dose levels before you start seeing some biologic activity.
In the present situation, as I understood it before the meeting started anyway, we are in a setting where we are looking at a case where we think we have efficacy data for adults. So I am sure we have already at that point gone through the Phase I trial for the adults, and so we probably have a fairly good idea of the starting level. I don't think that the accelerated designs have that much to offer.
Of course, you are going to still have to do some Phase I design to make sure you have an acceptable dose. In fact, I think you might even want to do a larger Phase I trial than you might normally do if you are planning perhaps on skipping directly to a large, randomized Phase III trial.
So even though I am talking about Phase I trial designs, let me just say that I think, to me, that is perhaps where the big gains are here. If you already have efficacy results in adults, since most agents don't work and here you have one that does work in adults, you might feel comfortable in jumping from Phase I to Phase III in children and not doing Phase II trials in children. It seems to me that would save a lot of time and patience on trial. But if you were going to do that, you would probably want to make your Phase I experience a little bit larger to make sure that you have the right dose.
So that's actually all I wanted to say about cytotoxic agents because Dr. Hirschfeld asked me to talk somewhat about the newer kinds of agents, where you are not interested in getting to a maximum tolerated dose. For these agents, you think that a lower dose may be just as effective as going up to the maximum tolerated dose and have less toxicity.
Well, if you are not going to use toxicity in your Phase I trial, you have to use something. Different things you might have -- if you had some blood concentration of the agent or perhaps some PK levels, you might use that. I am going to talk about that. I am also going to talk about, when you don't have that, then you might want to use some sort of molecular targeted biologic response to try to decide what the appropriate dose is.
So let's say you are in a situation where you have a minimum effect of blood concentration of the agent or its metabolite, and you know that, and here's a situation where you might have some feeling from the adult data what that is. Normally, when you are doing the adult trials, it's based on pre-clinical data which you may be a little less comfortable with.
So you could treat a cohort of patients at a dose level and measure their concentrations. Depending upon these concentrations, you would either treat additional cohorts at higher or lower or the same dose.
So, for example, if you treated five patients and you saw concentrations like that, well, if the minimum effective level is known from the adult data, say, to be 80, then it looks like you might treat the next cohort of patients with a lower dose. But if it was 130, based on adult data, that you are under it; you haven't reached it yet. So you would want to go up.
Of course, if the minimum effective level was 100, based on your adult data, well, then it is a little unclear here. Your observed mean is 110. Your lower 90 percent confidence level for the mean is 102, which sounds good. Eighty percent of the observations are bigger than 100, but the 90 percent confidence interval for the true proportions that are above 100 is only 49 percent. So if you wanted to really be sure, say, that 80 percent of the patients were going to have levels above 100, you would have to treat more patients at this dose level.
Sort of a bottom line here is you may into somewhat larger trials, if you really want to be confident that a good percentage of your patients are achieving a certain concentration, a larger number of patients than in the usual Phase I trial.
Now that's a situation where you have some blood concentration. Now suppose you don't have that, but you have some biologic response. I am not going to even attempt to define that, but I am thinking of some measured level of molecular target or change in level of molecular target that you think is associated, potentially associated with some clinical benefit. Obviously, that is a difficult question what that response should be.
But if you had it, you could design a trial around it. So, for instance, if you treated 11 patients at a dose level and all 11 had this response, biologic response, then the observed response rate is 100 percent, and you could be 90 percent confident that the true response rate would be bigger than 81 percent, which you might feel is comfortable and that is a good level to be at.
Of course, if you only observed 10 out of the 11 responses, even though your observed response is 91 percent, you could only be 90 percent confident that your true response was bigger than 69 percent. So, again, if you wanted to be really confident that it was above 80 percent, then you would have to treat more patients, and we are already up to 11. So, again, if you are trying to achieve this kind of goal, you are going to be into larger sample sizes than what we usually see with Phase I trials.
Now partially because of that, we have, my colleagues and I at NCI have been working on trying to do some other things that would require less patients. So one possible way is finding what we have been calling a biologic efficacious dose. In the context of dose escalation, rather than trying to ensure there is a minimum biologic response rate, we are going to only ensure that, if the true response rate is low, then there is a high probability of escalating, and if the true response rate is high, there is a low probability of escalating it.
So I give you one possible trial design here which is similar to a standard 3-6 phase escalation. You could also do this in a Bayesian way. You initially treat three patients at a dose level with zero or one of these biologic responses. You escalate the dose for the next cohort with two or three responses. You expand the cohort to six patients; with five or six responses, declare this dose to be a biologically efficacious dose.
So the statistical characteristics of this are that, if the true response rate was less or equal to 40 percent, there would be a 96 percent probability of escalating it to the next dose. If the true response rate was bigger than 90 percent, there would be only an 11 percent probability of escalating. If you were willing to accept those characteristics, then this would be a scheme that would only require three or six patients per dose level. This is just one way to do this. There's a lot of ways you could do this.
Now a question you can ask is: Is there any dose response relationship between the agent and the biologic response? This question is not really trying to determine adults for further studies, but occasionally people are interested in this, obviously, for scientific reasons, proof-of-principle reasons.
So a typical trial design might be treat 20 patients at a low dose, treat 20 patients at a high dose, and compare the response rates between the two dose levels. With these sample sizes, you could reliably detect a difference in true response rates of 50 percent versus 90 percent between those two doses.
So, again, I should probably cross out "power" here, so Steve doesn't get offended.
DR. GOODMAN: I like the "alpha equals one."
DR. KORN: You like the alpha equals one? Okay. We'll just cross out that one.
However you look at it, this is the kind of sample size you would be required to detect this kind of difference reliably, which is actually a fairly large difference, 50 percent versus 90 percent. So sometimes one sees studies with these very small sample sizes that are looking for dose response curves; well, they are kind of fooling themselves.
If they are fooling themselves, the people who say they are going to find the optimal biologic dose are even fooling themselves more. You occasionally see studies like this, too. Well, first of all, it is not clear what the optimal biologic dose is defined as, but let's just say we want to really assess the shape of the dose response curve. Well, you are into even larger sample sizes here.
What I did was I simulated some data, and I'm not even a Bayesian and I make up data -- (laughter) -- where I know what the true response rates are, and the true response rates go from 50, 60, 70, 80, and 90 percent, corresponding to the five dose levels here. I generated simulated data; 10 patients reached dose level. So this is a 50-patient study, which is, we'll agree, is not small.
So the true response rate is a straight line. So this is the first time I did it, and that looked pretty good. But then I did it again and said, well, now, if you just saw this, you might say, gee, things are kind of leveling off at the third dose level. Maybe that is my optimal biologic dose. And I did it one more time. Well, here it looks like things are actually going down, so maybe we should stop it. Biologists have a word for this which I have forgotten.
When you look at this kind of data, you can easily get fooled, even with 50 patients. So if you really want to find the shape of the curve, you are talking about hundreds of patients.
DR. HIRSCHFELD: Is that close enough for government work?
DR. KORN: So one way to not fool yourself, of course, is to put confidence intervals on the proportions, and that gives you a better feel that, gee, I don't really know what is going on here with the shape of this curve.
So let me summarize here in two slides. For cytotoxic agents, I think standard designs should work well, since you typically know about where to start from the adult data. I have no objection to Bayesian designs. I mean, I think you are going to get, if you start about the right place, you are going to get to about, you are going to use about the same sample size and end up about the same place.
For non-cytotoxic agents, things can be hard or easy, depending. If you actually have an effective blood concentration to target or something similar to that, that's good and you can go after that dose. If you have a targeted biologic response available, you can try to use that to determine the dose in the different ways I described.
Now I should say that usually using these targeted biologic responses is problematic at least in the adult -- in the first studies of agents. The reason tends to be that the assays and the techniques for measuring the response are being worked out simultaneously with the clinical testing of the agent. So no one feels comfortable in using those assays and techniques to decide what dose to use.
So we have these discussions all the time: "Gee, wouldn't it be nice, since this is a targeted agent, if we used a target to determine the dose?"
Then somebody says, "Well, how are you going to measure the target? Do you really believe it's that? Do you think it's this or the other thing?"
But it seems like in this present pediatrical setting we have a big advantage. If a bunch of the adult studies have already been done, you actually may have worked out some of these techniques, and you may really be able to put your hands on the target. Surely that would be a more optimal approach than basing it on -- excessive toxicity would be to use the targeted response so that you could actually measure it.
Thank you. I think I will stop there.
CHAIRMAN SANTANA: Thank you, Dr. Korn.
Any comments or questions for Dr. Korn? Malcom?
DR. SMITH: Ed, in your 50-patient trial to determine the optimal biologic dose, the variation that you were describing was just statistical variation?
DR. KORN: That's correct. I didn't add any biases or anything. That's random variation of what it looks like with ten patients per dosage.
DR. SMITH: Okay. Now the other kind of variation that you can get is in the assay itself. Any assay looking for a biologic target is going to have a certain variability and a certain imprecision. When you factor that in, how does that affect your enthusiasm for the optimal --
DR. KORN: Okay. Well, actually, that is a good point. I mean, I was pretending that the response was a yes/no binary decision. If you do that, then the assay variability is sort of already factored in, in that I was saying that the true response is going from 50 percent to 90 percent.
Another way to do this, of course, is to actually measure something on a continuous scale and try to use that to determine the optimal biologic dose. That, from a statistical point of view, is attractive because you are kind of using more information rather than dichotomizing things into yes/no, but then you do get into these whole issues about, well, how reliable is the assay? In that situation, the more unreliable the assay, the more spread you see about these points, and the larger the sample size which results.
CHAIRMAN SANTANA: Eric?
DR. ROWINSKY: Just a comment on that: I think we should only be so lucky to have assays that are validated at the end of Phase I to give you hand-off -- even binary assays.
DR. KORN: But my understanding is you are not only done with Phase I, you are going into the FDA asking for an indication; you are probably done with Phase II. You still don't have the assays?
DR. ROWINSKY: We still don't have the assays, and there are rare drugs that we have developed -- I am trying to think of them here -- in which we would be able to hand off information to you.
CHAIRMAN SANTANA: Martine?
DR. BAYSSAS: One of your comments, you said you would like to go from Phase I to Phase III. Looking at what you have done, I don't know where is the Phase II here, because, okay, it's a Phase I. To some extent, you have a certain end-point which you hope will relate to efficacy, but you still don't have an efficacy end-point. So I don't see how you go from Phase I to Phase III. You need to have a Phase II built into Phase III or some element of Phase II into Phase I, but I don't see how you --
DR. KORN: Yes. Well, first, my easy answer is Dr. Hirschfeld only asked me to talk about Phase I.
But the more serious answer is that, I mean, Phase II is a lot of times just used to screen agents to find the most promising ones to go to Phase III. If you already have -- I mean, here we can think of using the adult trials to be the screening trials. The adult trials can do all these tests to these agents, and most of them are negative, and they find a good one. Well, I would have no trouble taking that good one directly to Phase III for children, provided that evidence was there from adults.
The only reason to go to Phase II, I think, is if you had so many adult agents that were promising from Phase III trials that you didn't have enough patients to test them all, then you would have to say, okay, well, maybe I need to do Phase II trials in children to pick which ones to go to Phase III. But I don't know that there's that many promising agents that have shown efficacy in Phase III that are ready to go for children that we couldn't perhaps skip Phase II.
DR. SMITH: Ed, could I just clarify? You are assuming that the Pediatric Rule has been invoked and, hence, there is an adult tumor that is similar to the pediatric tumor?
DR. KORN: Yes.
DR. SMITH: Okay.
CHAIRMAN SANTANA: Yes, I think we are talking in that context, not beyond that.
DR. BALIS: For I think a lot of practical reasons, in looking at other trial designs, the most likely one that we would be pursuing in pediatrics, for a couple of reasons, would be the one where we were trying to achieve a minimally-effective of effective concentration, only because, first of all, it is much more difficult to do biologic end-points in pediatric patients unless you are willing to accept a surrogate tissue instead of a tumor as the site that you look.
Secondly, because it is very possible that by the time we do the trials that relationship would be defined at least in different types of cancers in adults. If that is the case, then I think that the way that you would approach that would be different than doing a dose escalation with the intent of defining what dose it took to achieve that concentration.
Firstly, if you knew the concentration was effective, that would be the way you would dose all patients. You would individualize them to achieve that, as Clinton was saying.
Secondly, unless the kinetics are nonlinear, which would be an unusual situation, you would be able to get information from every patient you treated. So, for example, if you gave a patient a dose and they were half the target level, then you could assume, knowing what the kinetics were in adults, that if you doubled the dose in that patient, it would give you an effective concentration.
I don't know that you would need to go in and take another patient and prove that necessarily.
DR. KORN: Right, but don't you have to start, wouldn't you want to even start the first patient at a targeted lower concentration than the adult? I mean rather than put 10 patients on at once?
DR. BALIS: Yes, what you would really be doing I think in a trial like that is defining, is determining whether the concentration you defined as your active concentration was tolerable.
DR. KORN: Right.
DR. BALIS: You wouldn't want to be defining what dose it took to achieve that in some percentage of the patients.
CHAIRMAN SANTANA: Steve?
DR. HIRSCHFELD: Again, I think there is a seque between Frank Balis' comment and a question I wanted to ask, which is a followup on a comment that Dr. Coltman made this morning.
If you have a drug that has as its target in the modeling that you were discussing, let's say, BCR-able in one disease, CKIT in another disease, PDGF, binding site in another disease, and perhaps even a fourth target, how do you go about thinking about your dose-finding studies? Would you do four different dose-finding studies for each target or would you look at a panel of different assays for each one, or should you think of a totally different paradigm than trying to look at the targeted dose?
I don't expect you to have an answer for that --
DR. KORN: Thank you.
DR. HIRSCHFELD: -- but I wanted to raise the question and think about it in terms of both modeling and the types of advice that is given. Anyone is welcome to respond.
DR. PAZDUR: Could you perhaps have a different dose for different tumors, you know, aimed at the molecular target that you have? Generally, we have not done that, but if you are evolving to that situation where you have multiple targets here, perhaps one has a target-specific dose that they are going to be looking at. I don't know. Here again, that is quite hypothetical, but I think this whole discussion has some hypothetical connotations to it at this time.
DR. HIRSCHFELD: Well, I think it is quite real, and that would be the sequential model then. For each tumor, you would have a dose-finding study, depending on each target. I just wanted to raise the issue, if there were other ways to approach it. Maybe Dr. Rowinsky has some thoughts.
DR. ROWINSKY: I can give you a hypothetical answer, which would be measuring, assessing the tumor and titrating your drug to some level of inhibition that you can measure, but in the real world you are most likely going to escalate to a dose that you are pretty confident that is going to suppress all those targets, and then figure it out later. I mean, I'm sorry to have to say that, but that's usually what happens.
CHAIRMAN SANTANA: Dr. Boyett?
DR. BOYETT: Yes. I was going to comment that a similar issue is that traditionally in Phase I studies we put heterogeneous types of patients on these trials and look to toxicity. When you start talking about optimal biologic response modifying dose, it may be worse than the 50 that you saw up here because then they have to be all alike. They may have to have some particular polymorphism and you are looking for the optimal dose in that particular set.
So, similar to what you are saying, I think in some settings you are going to have to study different types of diseases separately to find the optimal dose that you want to use in those diseases.
CHAIRMAN SANTANA: Dr. Bernstein?
DR. BERNSTEIN: Well, just to comment on the last point, that assumes that our drugs have gotten so good that we are not limited by toxicity, but rather we have the luxury of actually targeting the dose to the tumor target as opposed to being limited by the toxicity incurred. That would be nice.
CHAIRMAN SANTANA: Dr. Reynolds?
DR. REYNOLDS: I think one of the issues in this that we haven't discussed is that, if you are talking about this theoretical magic bullet that Mark just alluded to, and the discussion earlier of going for a maximal practical dose or biologically-effective dose rather than an MTD is that, when you go from one disease to another, you are going from very different tissue distributions that you need to effect the tumor.
So, for example, the penetration of drugs into bone marrow to effect leukemia and the lymph nodes and blood is going to be very different than the penetration in the solid tumors. So I think you have to be very cautious about thinking about those issues, and that what works in one setting may not be what you need to get into the next setting and get the job done.
CHAIRMAN SANTANA: Dr. Boyett?
DR. BOYETT: Yes, one comment about physician belief, and this is a recent experience I had with soliciting belief to use CRM or a traditional model. I had this one particular study where we chose to continue to use the 3-and-6 rule because the belief was we were almost there -- six dose levels later.
We have another drug that was assured to be nontoxic. We got about six or seven dose levels we're going to go up to. We had DOT at the first dose level.
So, Steve, maybe the people you work with have better belief patterns than the ones --
CHAIRMAN SANTANA: Any further comments or questions?
CHAIRMAN SANTANA: Okay, thank you, Dr. Korn.
Dr. Hirschfeld asked me to address this variation in Phase II design that's called window studies. I am not going to summarize the experience. That would be three hours' worth of talk. But what I am going to do is to try to summarize some of the rationale behind this variation in the Phase II design and the pros and cons, and then briefly touch on some of the ethical issues of this kind of design.
Okay, so I am going to be talking for the next 15 minutes or so about this variation in study design for Phase II studies that are called Phase II upfront windows.
In my recollection, and there are certainly others at the table who know this better or just as well as I do, probably the first indication of this design was about 10 years ago. There have been now a number of studies, primarily done in the pediatric arena, using this type of design.
For the uninitiated, in essence, what this design calls for is an end-point of response, like any other Phase II study; also, trying to assess toxicity as is relevant to any Phase II study, but the difference is that the population of patients that are being studied is different. These are populations that are receiving these -- these are patients that are receiving this agent or this combination of agents early on in their therapy, prior to them receiving standard treatment. So, in essence, it is a window because it is an opportunity, prior to receiving standard therapy, and it is upfront because it is occurring temporarily prior to these patients getting their extended therapy.
Then, for the purpose of discussion, I wanted to contrast a little bit some of the issues of the classical Phase II design versus the window Phase II design. The patients may be different. In the classical Phase II design, usually these are patients that have a failed, a prior therapy, so they're very highly selected, and they are also highly selected in the sense that, because they have had prior therapy, their end organ, their toxicity issues are unique, and some of those patients obviously would not be eligible for the classic Phase II design based on some eligibility criteria. So they are usually a very small number of patients.
In the contrast, for the window trials, these are patients that have not received prior therapy and are going to get this experimental drug or these experimental drugs prior to the standard therapy. It can be a very well-defined population. For example, the high grade brainstem gliomas, and that entire population can then be the subject of this type of design. So it could capture a much larger population of patients with a particular disease, because they don't have some of the other limitations that this group of patients would have.
In terms of response, like any Phase II study, we are interested in assessing the response of the particular agent or the combination of agents, and the argument has been made that in the Phase II window upfront the response is most representative because these are patients who have not received prior chemotherapy, who do not have issues of tumor resistance and/or toxicity.
So these patients probably will give us a better representation of the true activity of this agent or this combination of agents, whereas patients in the classical Phase II design, these patients may have lower responses because their tumors may have acquired resistance or they may have had issues of toxicity because of prior therapy, whereas in the window setting these patients are truly virginal to prior therapy. So the toxicity represents a unique observation of the true toxicity of that agent or that group of agents.
So this is just to contrast a little bit of the differences between these two designs. Having said that, I think we need to remember that the intent of the Phase II window design is to provide some effective therapy; that is, to produce some tumor response.
But I think the ethical tension in these kind of designs is that then we very carefully have to assess the risk and benefits of these designs in the context of two major issues. One is that window therapy is occurring in the context of a larger trial. So the patients are not only getting the window upfront therapy, but that's followed by some additional standard therapy.
That's relevant, as you will see later on in my presentation, because that raises the issue of the potential negative impact of the window therapy on the ability of the patients to get the standard therapy, but I think it has to be considered as a whole, not separate, in terms of the therapeutic intent.
I think the other thing that was briefly mentioned this morning is that in pediatric oncology this concept of standard of therapy usually is in the context that patients are participating in an investigational trial. So Phase II window trials are not occurring alone; they are occurring in the context of a larger trial that has another component that follows the window. Okay?
Now let's talk a little bit about the scientific validity of these kind of designs in contrasting them to the classical Phase II design. While in the classical Phase II design one of the rationales for doing upfront windows is that, by doing an upfront window, you really get a better estimate of the true activity of a new agent, for the reasons that I have exposed before.
In contrast, in Phase II classical designs one may overestimate the toxicity because one has a population of patients that have been exposed to many other agents, and therefore, the agent may be declared as highly toxic just because of the population we have chosen in the classical design.
So an argument has been made that the window design allows us then to more accurately get a better estimate of the true activity of the agent in a patient population.
Now one of the criticisms for the upfront Phase II window trials has been that so far there has been no conclusive evidence that these studies improve or impact on survival. I think that is a valid criticism.
The counterpoint to that is that a similar situation occurs in the classical design, that there have been many agents used in Phase II classic studies that ultimately were incorporated into therapy, and there was no demonstration of the impact on survival of incorporating those agents. So I think that criticism is valid for both types of designs.
The relevant issue in the upfront window is, how do we balance that against issues of tumor progression and other risks that the patient may be exposed to? We will come back and visit those later.
The other issue is that, although some of these trials have now, the initial trials were initiated about 10 years ago, and there have been quite a number of them, many of these trials require a lot of time. I have given you some examples here of some studies that we did at St. Jude with the window of ifasfamide in osteosarcoma, something similar that was done in POG, some of the issues of ifosolin osteosarcoma in a randomized trial in a cooperative group and how long it took to complete that trial, and then some of the trials that are currently ongoing with Irinotecan in pediatric randomized sarcoma.
So it is really very hard for me to stand here today and truly give this group a final conclusion about the impact of using this trial design in terms of the impact that it has had on the survival of pediatric oncology patients because of the long time that it takes to complete these studies.
Let me focus a little bit on this issue of risk and benefit, because I think a lot of the discussion that has occurred around this design has been relevant to that ethical tension of whether these trials pose some unique ethical questions.
So let's talk about benefits because, whenever we offer a new therapeutic drug to a patient, I think our intent is to, hopefully, offer some benefit. Well, potentially if the agent is proven to be active, you could get a very prompt tumor response. Most of these agents that are being tested potentially have non-cross resistance to some of the other classic agents the patients are potentially going to receive.
I think the example of that is some of the issues of camptothesins that Clinton talked about earlier this morning, which these compounds clearly have some activity in a wide variety of pediatric tumors. I think there is some literature in pediatric oncology to support that the promptness of a tumor response ultimately gives us some idea of the potential for that particular therapy to improve the survival of patients. So if the agents are proven to be effective, prompt tumor response is a potential benefit.
There's no clear indication in many studies that there is a benefit in terms of improved outcome, but there are some data that at least in some combinations there may be an improved outcome.
Then the other issue is the issue of decreased toxicity, that clearly because these patients are now virginal to prior therapy, they have not been heavily pre-treated, potentially their toxicity spectrum is much different, and there could be decreased toxicity in these patients also.
The other issue is that, as far as I know, of all the Phase II studies that have been conducted so far, there clearly has not been a major issue with tumor progression in these patients that have participated in the Phase II window trials. However, I say that with a grain of salt because, clearly, that data is evolving.
One of the problems that we need to address in this design, though, is which agents we bring up to this type of design, because clearly this type of design has to be very selective in terms of the agents that we are going to test and the patient population that we are going to test. I think as we talked earlier this morning, I think we do need some pre-clinical data to help us sort out which of these agents we should be using in these kind of designs.
As was mentioned briefly this morning, I think to the present, we are very dependent on some of the pre-clinical xenograft data that represents particular pediatric tumors, testing those agents in that setting, and then from that, selecting those agents that potentially could be used in these Phase II window trial designs.
The other issue is, in terms of selecting what agents we are going to incorporate into this design, like we mentioned earlier, there is a limited number of patients. There may be more drugs than patients or ideas that we have. So we have to be very selective in terms of how we prioritize which agents we adapt to this design.
I have outlined here at least two concepts that at least we use at St. Jude when we consider whether an agent should be incorporated into an upfront window. One is that it either has a novel mechanism of action or it is an analog of an agent that we know that is effective, but potentially has an improved toxicity profile. So by incorporating that earlier in the therapy, we may negate some of the issues of toxicity while still, hopefully, providing an effective therapy.
The other point is which patients we should select for these kind of unique trials. I think this is one of the questions that Dr. Hirschfeld wanted us to address at the end of this afternoon. This is one that I don't have a quick answer or a threshold answer; it is truly open for a lot of discussion.
One of the concepts that we have used at St. Jude, when we accept these designs in in our trials, is that we try to define which patients ultimately would be at the greatest risk of treatment failure with the conventional standard therapy that we have. Those are the patients that I think ethically and scientifically then one could justify incorporating into these kind of designs.
I give you a brief example: Patients with malignant brainstem gliomas have a survival of less than 10 percent. So I think in these group of patients there's an imperative to assess new therapies that potentially could impact that disease.
And then the counterpoint to that in terms of patient selection is that we have to carefully assess the risk and benefit for the individual patient. We have to have some prior experience with the agent, whether it comes from adult data or whether it comes from pre-clinical data, and that we also have some idea about the potential toxicity profile of that agent.
Having said that, I think we need to recognize that many of our pre-clinical models do not really help us in defining toxicity for patients. So a classic example of a Phase II window trial that, unfortunately, was not very successful was the use of melflan in rhabdomyosarcoma in which the animal model did not predict that myelosuppression was going to be so excessive. Unfortunately, these patients that got melfalan prior to conventional therapy had excessive toxicity.
So I think we need to look at toxicity, if it exists in humans in other settings, and not in animal models, because it will not help us in selecting the best approach for these patients.
Let me briefly finish by talking a little bit about some of the issues that Eric kind of challenged me this morning to discuss. He and I did not talk before this. So I wasn't aware he was going to bring this up.
It is the whole issue of, how do we approach the consent for these patients? I want to outline three basic principles, and then I am going to come back to a little bit more discussion about this, based on a consensus meeting.
First of all, like any study, I think we need to recognize that enrollment is voluntary and that this has to be carefully explained to the parents and to those patients that are ultimately going to participate, and that I think we need to carefully educate our parents and our local IRBs in the concept and the scientific and ethical rationale between theses kind of trials, so that they can understand what the ultimate reason for performing these trials is.
Having said that, about four, four-and-a-half years ago, CTAP, Malcom Smith and other investigators at the NCI formed a consensus meeting that actually was held here in Washington, if I remember correctly, in July of 1997, and this Committee produced a paper on this issue of trying to assess the ethical validity of these Phase II upfront windows. I think this is available on a website, and it certainly is available through CTAP.
I want to focus on two issues that they specifically discussed that I think help us in terms of minimizing the risk that these patients potentially could be exposed to when we apply this kind of design.
One of the risks is obviously the risk of tumor progression. So we need to be very careful when we do these upfront window designs that we have very strict rules about stopping the study based on tumor progression. That has occurred in some of these trials, but in other trials we have been fortunate that tumor progression has not been an issue. So tumor progression does need to be addressed in the statistical design of this kind of study in terms of the stopping rules.
We need to recognize that there may be unique toxicities. Since these are patients that have not been previously treated with any chemotherapy, their toxicity profile may be very different than patients who otherwise have been treated, and the impact of the window therapy prior to the standard conventional therapy in terms of the future therapy that the patients may receive, I gave you kind of a negative example in terms of melfalin producing myelosuppression and then not allowing those patients to get effective standard therapy. So we do need to pay some attention to the impact of the window in the context of the standard therapy or the other investigational therapy that the patients may get subsequently.
I mentioned to you briefly in terms of the benefits, the early response, and also that if patients respond to this agent during the window phase or the upfront window, that I think serious consideration should be given to incorporating this agent into the subsequent therapy that these patients receive. I would be happy to hear other opinions about this issue, but I truly do believe that, if an agent is proven to benefit the patient early on in a window trial, that that agent should be seriously considered in the subsequent therapy that the patient will receive.
Then the duration of the participation in the upfront window should be kept as short as possible. If the other mechanisms in which we can identify responses by maybe potentially looking at other surrogates, we should also attempt to do that, so that we can limit the exposure of patients to the minimum while giving them the maximum benefit of participating in that trial.
Then in terms of the informed consent, that consensus meeting touched on some very important points. I think the issue of the upfront window therapy should be clearly identified in the consent separate from the other elements of the consent that relate to the standard therapy or to the other investigational therapy, not necessarily that there be a separate consent for the window, but within the informed consent document and within the informed consent discussion, that this be identified as a separate, unique component of the trial; that patients and parents be given the opportunity not to participate in the window component, if they so choose and so desire.
It has been very interesting because I think at St. Jude, as we have done a number of these trials, we have come to appreciate that now a greater number of parents and patients opt not to participate in some of the window trials that we have designed. So I think people are becoming -- I think we are doing a better job in terms of the informed consent and providing alternatives to patients.
I think, very importantly, the consent document in the process should clearly identify how this window differs from the subsequent therapy. That should be clearly identified. Then I think a potential risk is that of delaying the standard therapy or making the patients ineligible for other future therapies. I think all this needs to be included, in addition to all the other requirements and the elements of the informed consent.
Another issue that needs to be addressed in the informed consent for these window trials is the issue of the impact on quality of life. If there is any additional procedures that the patients may require, if we are using some surrogate end-points during the exposure to that agent, and then the potential impact on quality of life of extending the duration of therapy, because if we are now giving a window trial that lasts six weeks, that may be six more weeks of total therapy that the patient would receive. So I think that needs to be considered, too, in the discussions of the consent.
We need to clearly indicate in the consent document any pre-clinical or clinical data that would support the use of that particular agent or agents, and, as we all know, treatment alternatives should clearly be identified and, as necessary, provisions for assent or refusals should also be part of the elements of informed consent in these window trials.
So I am going to finish that by saying, up to the present, I think there is some scientific evidence to suggest that these trials are scientifically-justified. I think given the constraints of some of the discussion and points that I made, they are acceptable under certain conditions. I think, more importantly, scientifically, they also provide us an effective mechanism to identify active agents within this issue of the developmental program of drugs for children, but, unfortunately, we are still too very early in this process, and the number of trials is still very limited, that we can have a final conclusion on the ultimate impact of this design on treatment outcome of patients that participate in these trials.
So, with that, I will finish, and I will entertain some comments or questions. So I guess, as the Chair, I get to choose, right? Eric?
For the purpose of discussion, I am going to sit down, okay?
DR. ROWINSKY: I think you have presented very cogent argument, pros and cons of these trials. You alluded to something. Phase II trials are really screening trials, and I don't think that they're just screening trials to screen for drugs that are active. You suggested that perhaps active drugs in a Phase II window should be considered for incorporation into frontline therapy, but I think what we are trying to do now, especially now with so many agents, is not only to find active agents, but agents that agents that are really going to incrementally impact.
I think that one disadvantage of the windows screening trials is that when you find a drug that is active at the back end in patients who are refractory, I think you are more inclined -- that drug is much more likely to demonstrate an impact, an incremental impact on disease.
I think that you are selecting the best situation, of course, patients that are most apt toxicity-wise and host-wise, but I am not so certain that we are really -- in the upfront window situation we may be screening for "me-too" drugs, many drugs with very similar mechanisms of actions as opposed to the back end. At least from the adult experience, drugs that have been shown to be active in patients who are totally refractory are drugs that have made an incremental impact. So it's one of the arguments that is a con argument for that design.
CHAIRMAN SANTANA: It's a valid argument. I am not going to take a position.
DR. ROWINSKY: No.
CHAIRMAN SANTANA: Mark?
DR. BERNSTEIN: I think that I agree as well. I think, however, there are some times when there are biological differences in tumors at diagnosis and then recurrence where it is still possible that, for instance, methotrexate versus trimetrexate in osteosarcoma is something we are hoping to look into. There may be reasons that tumors are resistant at the time of recurrence where they may not be at the time of initial diagnosis, and still adding an agent at the time of initial diagnosis may have an incremental impact. It needs to be tested, but I think that it may be possible.
CHAIRMAN SANTANA: Peter?
DR. ADAMSON: To throw perhaps some controversy into this, Victor, you concluded with four points: that Phase II windows are scientifically-justified, ethically acceptable; they can identify active agents, and it is too early to judge whether that works. So I will go out on a limb and say, to varying degrees, I disagree with all of those conclusions. Let me tell you why.
CHAIRMAN SANTANA: It's the first time we've disagreed today, right?
DR. ADAMSON: It had to happen sooner or later.
I think it is important to look historically as to how Phase II windows first evolved and where they ended up, and why we are sort at the crossroads we are today. You spoke about Mark Horowitz' trial with melfalin. I believe the main thrust then was we would do a Phase II window so we wouldn't dismiss a potentially active agent. For others, melfalin in the relapse setting had uninteresting activity, and in a Phase II window setting had interesting activity, and it set the paradigm that we will do Phase II windows so we don't wrongly dismiss active agents.
What then evolved is that we did a host of Phase II window studies of drugs that we knew had activity. So I would challenge you that you say it gives us a better response estimate, and I would say, what do we do with that information? So if we have set a threshold for relapse patient of 20 percent and for newly-diagnosed patients a threshold of 40 percent, we are still left with we have an active agent. Does it improve curability? I would argue that there has not been a Phase II window trial that has ultimately impacted on the need to do a Phase III study, nor has it impacted on identifying an agent that's active.
I don't think anyone can come up past melfalin with an example where we have identified an active agent that we didn't already know was active. So the whole problem that we find ourselves in with ethics, and that is probably where I agree with you most, is that in certain situations these are ethical, but I think scientifically and how we move forward, my concept of when is a Phase II window going to be scientifically-justified and productive has gotten, that window has gotten a lot smaller. Knowing that VP ifos is active, ifos carbo is active, still hasn't answered the question, are these agents going to improve curability?
CHAIRMAN SANTANA: I have to agree with you that I think one of the problems with this approach is that, to date, I don't think there has been any conclusive data that it impacts survival. But I would say that that is true of the Phase II design in general, that I think we've got to be careful that we define the end-point. The end-point of a Phase II trial is not to impact survival of the patient. The end-point of a Phase II trial is to give us some idea about the efficacy of this drug or this combination of drugs in a particular tumor system, in a particular patient group. It is only until you take it to the next level that you clearly can demonstrate whether there has been an impact of that drug in the disease or in the process.
So I think we've got to be careful because I think we have been maybe a little bit too critical of the Phase II window design in terms of, well, we shouldn't be doing it because to date it hasn't impacted on survival of patients. I would argue that the same argument could be made for the classical Phase II design, that the number of drugs in pediatrics that we have potentially brought to Phase III that have had impact on survival is still very limited. So I think the argument is for both.
I agree with your points. I am taking both positions here because I also wrestle with some of these Phase II window design studies, but I think in certain patient populations, in special conditions, I think this is a unique way that maybe would allow us to benefit the patients to a certain degree that otherwise they would not, because of prior history of treatment, and so on and so forth.
DR. ADAMSON: But I don't think that the Phase II window study is a prerequisite for Phase III, for doing a Phase III study, and that, in essence, is what it has not evolved into, in that we're saying, well, before we move to Phase III, we need to a Phase II window study. My argument is we are not going to learn enough from the Phase II window study to help us decide whether to do the Phase III or not.
If we have an active agent in a relapse setting in a classic Phase II, we can sit down and decide, is this a high enough priority that we should do in Phase III? We don't need to repeat the experiment and learn that it is more active in newly-diagnosed than in relapsed patients.
DR. COLTMAN: The sole example you gave was brainstem glioma with a 10 percent survival rate at best. Now what is the boundary of responsiveness to therapy that you use? Is that the bottom or the top of the criteria?
Furthermore, if there is a treatment that is available, even if it has a 10 percent survival, and since you have been at this for 10 years, one should be in a position to look at what the impact on standard therapy is that is bound to follow this, and has this Phase II window upfront in previously-unresistant patients generated resistance which might negatively impact on the standard treatment, even though it is only a 10 percent five-year survival?
CHAIRMAN SANTANA: The data that I know that potentially could indirectly answer your question is derived from a POG study in neuroblastoma in which there were sequences of windows, and maybe Dr. Korn or Dr. Reynolds can comment, or maybe Dr. Cohn, since she, I think, participated in this trial.
My interpretation of the end result of that trial was that none of the windows negatively impacted on the ultimate survival of the patients after they received the standard therapy. That is kind of an indirect way of looking and trying to answer your question.
DR. COLTMAN: Well, I think that is very important, because if this upfront window has a negative impact, unless they are dated to show that it doesn't, then I think it is not ethical to do it.
CHAIRMAN SANTANA: No, and I think you are correct, but that is the only, I think, big body of data that at least compared different windows. So there was some variability in the windows, and then the impact of those windows on the subsequent treatment that all the patients received, whereas all the other trials -- and Malcom and Mark or Peter can correct me -- all the other Phase II window trials that I have been familiar with have been single windows followed by some standard therapy. That trial I think was very informative because there were different windows for different patients, and then they all got the same therapy subsequently. So I think that trial gave us some idea that at least there was no negative impact on the ultimate survival of the patients.
The issue of thresholds is very important. I think that needs to be clearly defined when one undertakes a Phase II window trial in terms of the population of patients, and what you would find acceptable in terms of the ultimate survival of those patients in the absence of the window, to define that population very carefully.
So this is a very small group of patients. This is not promulgated for the larger number of patients, but for a very unique patient population.
Sue, do you want to comment on that?
DR. COHN: Sure. I don't know the data intimately, but I do know that there were four Phase II upfront windows. There was no negative impact in terms of ultimate survival, depending on which window somebody was initially randomized to.
Of the four upfront windows, there were definitely differences in terms of response. So because of that, some agents have been subsequently further studied; others have been dropped in neuroblastoma.
It also, however, depending on which of the four arms, not only was there no negative impact, but there was also no difference in positive outcome. So Peter is right in that regard as well. So there was no advantage to have gotten even one of the more activations, like Topotecan. Had you gotten the two cycles of Topotecan prior to your standard therapy, there was a very beautiful response with Topotecan upfront, but the ultimate survival of the group who got the Topotecan versus the group who got nothing versus the group that got a different Phase II was not different.
DR. COLTMAN: Of course, if you intercede with a Phase II window, then you had substantial lead time bias, so that you would have to be in a position to know that at the time of the initiation of the standard treatment going forward, was there an impact on that? And these studies didn't address that. They looked at overall survival.
One can imagine that the Phase II window may have had an impact, but it inhibited the way they responded to the standard treatment, and therefore, the overall survival was not different, but the impact following the initiation of standard therapy may have been worse. That is the question that needs to be addressed.
I can tell you in adults we have lots of tumors that don't respond to therapy. There are no standard therapies for many of those tumors. We have been doing Phase II windows in that population upfront because we don't have good therapies.
So, early on, therapy in a Phase II agent in patients with incredibly refractive disease, refractory disease, is something that has been done for years. As a matter of fact, when I heard Archie Blair first present this at the Vail Course, I pointed out to him that my first Phase II window I participated in in the Southwest Oncology Group in 1964, and that was the use of hydroxyurea upfront in the treatment of malignant melanoma for which there was no treatment.
CHAIRMAN SANTANA: Dr. Finklestein?
DR. FINKLESTEIN: I would like to submit the data is not in yet for pediatrics. As Dr. Cohn knows, many of us believe we have not really impacted on the overall survival of -- Malcom is going to smile because I have been saying it for decades -- the overall survival of neuroblastoma since day one. So I am not sure, no matter what you do to neuroblastoma, you are going to change the survival.
So I think, however, many of us would like to see the four windows, if that is the analogy, with some other tumor before we would be able to say that it is in for, at least the data is available for pediatrics. I think neuroblastoma is the wrong tumor in that regard.
CHAIRMAN SANTANA: Malcom, do you want to make any comment? You don't have to if you don't want to.
DR. SMITH: I will let Eric go first.
CHAIRMAN SANTANA: Okay, Eric?
DR. KODISH: Thank you, Victor, for your comments on Phase II windows. It is an extraordinarily complex, ethical issue. I think to do a good, moral analysis of the area, one needs to doa thought experiment essentially and unbundle, if you can, the scientific appeal of this study design from the clinical care of the child.
The thing that struck me in your comments as most exciting was the possibility of incorporating the Phase II window agent later on in the design for that particular child. So morally that resonated as something that might tip the scale to balance and justify it. But, short of that, I am not sure that we are going to find a lot of examples where a Phase II window study can be morally justified.
I also have to make the Tolstoy point that all happy families are alike, but unhappy families are unalike in different ways. The difference between a child with a brainstem glioma, though it might high-grade, versus a rhadbo or a neuroblastoma or Ewing's, these seem to be very important points here. I think it would be dangerous to try to put all Phase II windows into one group because of that.
CHAIRMAN SANTANA: Wayne?
DR. RACKOFF: One comment, Victor, and one question. It was a very nice summary of a very complicated area.
I think that one comment is that, with some of the newer agents, because of one of your specifications, that the window be of short duration, I think we have to be careful. Some of the agents are, by their very design, expected to work only with prolonged duration therapy. Actually, a couple of them have been subjected to windows already, and I am skeptical of that. So that is just a comment.
The question is: Putting the ethical issues aside and assuming some scientific validity, because I think there is some to the approach -- it is at least intellectually attractive, if not empirically validated yet -- how does it accelerate drug development in pediatric patients?
CHAIRMAN SANTANA: If I could share my thoughts on it, I think potentially an agent in which there is some pre-clinical data of efficacy and some limited adult data could quickly be moved to previously untreated patients before it gets tested in previously treated patients, to identify its response characteristics and its toxicity profile earlier on.
I think, to me, that would be one potential advantage of doing these kind of designs, to more quickly identify the true spectrum of activity and toxicity of an agent.
Let me go down the list. Donna? Go ahead.
DR. RACKOFF: And your sense is, because these studies actually take longer in some cases, that that's why I ask. I am trying to see how it really accelerates the overall timeline of a drug coming to an indication in pediatrics, as opposed to a classical Phase II, which may be done along the same timeline, I guess.
CHAIRMAN SANTANA: That has been the issue I think that Peter was addressing earlier. I think both things are happening, and they're happening at least in selected tumors concomitantly. The issue is, is there an advantage of one versus the other, and are we really gaining any more time by doing the window in terms of ultimately identifying a drug that we may use in a Phase III setting?
I don't know, because I think the ones that I have the most experience with are the Topo I inhibitors. Clearly, Irinotecan was introduced probably a little bit earlier into the rhabdo trials than it would have, I think, if an historical approach to a Phase II setting would have been done. But I didn't participate. I mean, I am not a member of that Committee; I don't know how that came about. Maybe Mark or Malcom can comment.
Certainly in Topotecan there was already data in the relapse setting concomitantly to the study that Sue referred to. So those studies kind of were occurring concurrently.
So you're right, I think the proof is still not there that it in any way has accelerated our ability to move an agent or a group of agents any quicker.
DR. BAYSSAS: Is it possible to reserve the support, for example, to some eliminate the cytostatics of this approach? Would that be a possibility, and to select for which agents you would apply?
Also, you know, if you take patients with potential cure versus patients that first relapse, and another question I have is that, if you had in the CTAP thing, I have heard about Phase II, that it was sort of abducing maybe: first, taken only one cycle, very short, to have some characteristics about the drug, and it could maybe kind of screen for further combination therapy? So I wonder if this could also be an approach in this type of design?
CHAIRMAN SANTANA: Yes, I do think that the upfront window doesn't have to be restricted to a single agent. It can be a combination of agent. So it doesn't have to be a single drug. I think it could be two drugs that potentially one wants to investigate in combination.
The second point, or your first point was, how this design potentially could be applied to some of the newer biologics. I think that is a challenge. This design is going to be a challenge because, clearly, many of these biologics are cytostatic or they require a long period of observation before one sees a clinical response. I think the potential way to get around that is that one would then have to look at some other marker of potential efficacy and activity, so that one would not have to wait for six months of therapy in order to define the role that this agent may play.
So that is going to be a challenge, too, if one were to use biologics, use this design to test some biologics, and one would have to think of what end-points then one would have to ask.
DR. ADAMSON: To in part answer Wayne's question, I think the only way Phase II windows will impact upon the speed of drug development, if we are willing to put into Phase II windows drugs where we have no activity data or the classic Phase II activity data is inactive. Now the past decade we have been unwilling to do that.
But just doing Phase II windows, where we have activity data, all it is doing, in my opinion, is delaying the decision whether to move into Phase III. It is not impacting our decision to move into Phase III.
Which circumstances can you put an inactive agent in classic Phase II in upfront, and where you have no data into upfront, I think carries still the same ethical considerations, but as far as speeding drug development process, until we are willing to do that, I don't think the Phase II windows are going to speed the drug development process.
CHAIRMAN SANTANA: Peter, I will take note to your comment that it has delayed. How has it delayed the process?
DR. ADAMSON: Perhaps I will temper that.
VPI fos we knew was active. Or let me take perhaps a more recent example. Topo cyclo we know is active, and we know we have to do a Phase III study of it to determine if it is going to impact. Waiting for Phase II window data -- and I think the reality is we probably didn't wait; we only have so many opportunities to do Phase III. So I perhaps should restate it and say, I don't think it has accelerated our ability to do Phase III trials. In many cases the reality is we only can do so many Phase III trials in pediatric oncology and what to do in the interim.
CHAIRMAN SANTANA: Yes, I agree with you; I don't think, based on the data that we have so far, that we could definitely say that it has accelerated the process, but I think we have to recognize that it also has not negatively impacted on it.
DR. SMITH: I have several points. Chuck Coltman's question was, has it affected outcome? That is an easy question to ask, but a very hard question to answer because every Phase II window is different. If you amalgamate four or five or ten different trials, there may be no difference, but one trial there could have been a difference in outcome. So it is a very hard question to answer, and they are small numbers, small differences that may or may not be significant, just because we don't have the numbers. So it is quite hard to answer that question with any confidence. We can't say they do or they don't.
The other point that you brought up is the question of parameters such as disease progression, and that is one thing that we can say with some confidence, is that disease progression is more likely in the single-agent Phase II window setting than it is when therapy has begun with conventional multi-agent therapy that we know is able to induce remissions or responses in neuroblastoma or Ewing sarcoma or osteosarcoma. So that's come from several different trials.
A third point, and one that I thank Steve for circulating our commentary on the Phase II windows, but a point we made in that was that at first relapse many of these issues are the ethical concerns are reduced. The effectiveness of salvage therapy is diminished compared to the effectiveness of upfront therapy. Parents know about cancer treatment and are better able to assess whether they want to participate.
And Eric's point, these are tumors that have come back in spite of our known effective therapies, and presumably they are enriched for the clones that we most want to kill and we most want to identify active agents.
So I think in terms of looking to the future, and perhaps accelerating pediatric drug development, making better use of that, the first relapse, and looking at some of the new agents in that setting, potentially in a Phase II window setting, before proceeding to a more conventional salvage therapy has a number of advantages.
CHAIRMAN SANTANA: I will take one more question because I want to make sure we stay on time. So we will let Dr. Cohn comment.
DR. COHN: Thank you. I just wanted to make one comment about the biologicals, and that is many of them I don't think would, even if we are ending up now deciding the Phase II upfront windows would not be a good thing to do, I think Pat's study, where he used retinoic acid, demonstrated very nicely that many of these biological agents work best in a setting of minimal residual disease. So to use them upfront, when you've got disease from head to toe, probably isn't the best way to approach some of these biologic agents anyway.
CHAIRMAN SANTANA: Susan, one last comment or question?
DR. WEINER: Yes, I had the dubious privilege of being involved in that consensus panel, as you describe it, a number of years ago. It comes as no surprise that the parents at St. Jude, with the kind of consent form that we had recommended, are not agreeing to participate in window trials for new patients, for newly-diagnosed patients. The parents' perspective is really give it the best shot you've got, and a Phase II window trial is not exactly equivalent to that.
My question also extends to Drs. Smith and Hirschfeld, and that is that it was my understanding from this panel that this represented a set of guidelines by which to judge and decide on Phase II window trials in general. I wonder, what is the relationship was between these document and our discussion today and what guidance the FDA may take from that?
DR. HIRSCHFELD: We, I think, will defer that until after the break, when we have a very specific question which I think Dr. Santana will pose to the panel to help frame a response.
CHAIRMAN SANTANA: Well, with that, we will take a 10 -- Malcom, do you want to comment?
DR. SMITH: Well, just to address Susan's question in terms of, that guidance has really guided CTAP's review of Phase II window studies, and specifically in terms of the informed consents now do contain all of those things, and really the number of Phase II studies that have been initiated, single-agent Phase II window studies has been small, quite limited, since the meeting. There have been some. They have conformed to the guidelines from the Phase II window meeting.
CHAIRMAN SANTANA: Yes, and at St. Jude, where we initiated some of these Phase II window trial concepts, I think we have been very, very attuned to following the guideline as much as possible. I think it reflects what you mention in terms of where we are with the informed consent process for these studies.
With that, I want to take a 10-minute break and resume half after the hour, so people can relieve themselves, and then we will get started and do the questions.
(Whereupon, the foregoing matter went off the record at 3:18 p.m. and went back on the record at 3:32 p.m.)
CHAIRMAN SANTANA: Okay, let's go ahead and get started, so we can finally get to some advice to the FDA on the questions that they have posed. They have posed four questions for us.
For the public record, what I will do is I will read the brief introduction to the question, pose the question, and then we have asked Dr. Adamson and Dr. Rackoff to comment. The way we will do it is Dr. Adamson is going to take each one of the questions and give his perspective. We will have a public discussion about the question. We will move on to two, to three, to four, and then Dr. Rackoff will come at the end and give us an overview and his comments. Okay?
So the first item is the paragraph that's introductory to the questions that relates that, "The common approach for selecting starting dose for Phase I studies in children with cancer in cytotoxics is to begin at 80 percent of the adult maximally-tolerated dose, the MTD.
"Children who currently enter Phase I studies tend to be more heavily pre-treated with other therapies than historically had occurred. In addition, many newer therapies are not cytotoxic in the manner that previously-developed therapies were or may have different modes of action, including modulation of cellular-signaling pathways."
So the first question is: "If a potential therapy has an established dose in adults based on the optimal biologically-effective dose, the OBD, what principles should be applied to designing studies in children? For example, should the starting dose be a percentage of the adult dose, as historically has been done? Should the same exposure or AUC be targeted? And what role should pre-clinical data play in these kinds of study designs?"
DR. ADAMSON: I will preface this by saying I tried to take the questions that Steve had originally posed and foresee what some of the comments might be. As you will see from these slides, I have been more successful on certain occasions and less on others. But it, I think, will serve as a bridging point for discussion.
So here are my way of restating the first questions. The first one that we are going to talk about is the Phase I design issues, optimal biologic dose, and then we will get to the following three. I will sum up and then turn it over to Wayne, talking about the importance that we face in prioritizing agents, and then ultimately when should the rule be invoked.
So the first question, Phase I design, when one has the optimum biologic dose, how is this going to impact pediatric Phase I design? I think the challenge that we are going to face, if we try to implement this, is one of tissue acquisition. Frank Balis had commented on this earlier.
Essentially, the procedure must be minimal risk to be acceptable. So for leukemia studies, getting the leukemic blasts I think was going to be pretty straightforward; tumors invading the bone marrow, again, pretty straightforward. When we start moving beyond that for soft tissue tumors, immediately we are going to get into a more difficult area in order to obtain tissue. Ultimately, we will still be left with a question, as Eric posed, of: Are we really measuring the correct end-point? Not only is it important to know, are we measuring the correct end-point, but what is the specificity of the drug?
We have learned more often than not that a drug we think is working in one way turns out to have worked in additional ways. So the mere fact that we are measuring one end-point, it may not be the most meaningful end-point to explain the biologic effect of the drug. In many respects, the fluorouracil inhibitors are a good example of that, where they have activity beyond mutated RAS.
I foresee that we are going to be using pharmacologic data as a surrogate more often than not, and in this case the AUC. The advantage of this is that by targeting exposure, it is really independent of what the specificity of the drug is for the target. In some respects it may be independent of what the mechanism of action is. If we have an exposure that is correlated with an effect, the challenges become and the dangers become, we don't always or we rarely know what the correlation between the plasma concentrations that we measure is with the target tissue exposure. Ultimately, we assume that there is some correlation, and it remains undefined. The hope, in fact, is that AUC turns out to be a good surrogate for tissue exposure.
How might this impact starting dose? I think even if we were fortunate enough to have data on what the optimum biologic dose truly was in adults, it may differ between adult and pediatric tumors. So the optimum biologic dose may not be the same in a pediatric tumor as in an adult tumor. Ultimately, what this means is we are going to need pre-clinical data at least to give us a relative comparison of what an optimum biologic concentration may be. If not an absolute, we may be able, if we use similar models, to say, for this tumor, we are going to need twice the exposure that we do for an adult tumor. Absolute exposures probably carry more dangers with them, but relative exposures may be a reasonable way to go.
Again, to emphasize that chances are the optimum biologic dose may not be known following adult Phase I or Phase II trials. Right now we are faced with, well, how long do we wait to have this data? If we truly wait until the adults are confident they have defied an optimum biologic dose, we are simply going to increase the lag time to initiate pediatric clinical trials.
So I think, although we need to look to the future, to a time when we can rapidly define an optimum biologic dose, over the upcoming years it may be premature to think that we are going to be able to do this. To delay pediatric trials excessively in an effort to define this in adults, I think is going to be doing children a disservice and is going to be setting pediatric drug development back further.
Finally, to come to pharmacology, and here I just want to build on what Mary and Clint and Steve have presented, but I want to take this emphasis: Phenotype matters. There is no question that pharmacogenomics, pharmacogenetics are critical pieces of information, but in fact, as Mary pointed out, genotyping is simply going to get easier.
What is not easy is defining phenotypes. In this case, phenotype, Clinton very nicely displayed some of the strategies where we can obtain phenotyping data with limited sampling methods, and I think we are going to have to look to do that more frequently.
Certainly, in Phase I trials we routinely define phenotype, and I think Mary gave compelling reasons why we must start looking at obtaining genotype, because this is going to be the one clear situation where we are going to have both phenotype and genotype data available.
More importantly, I think, the Phase I pharmacokinetic/pharmacodynamic data that we generate will set the stage for Phase II and III clinical investigations, and I think there is going to be an ongoing need to obtain genotyping data and then phenotyping likely through limited sampling methods, not only in Phase I, but in Phase II and III.
So that's my few cents on question one, and maybe I will turn it back to Victor then.
CHAIRMAN SANTANA: Okay, further comments and discussion on question one? Dr. Bernstein?
DR. BERNSTEIN: I would just like to support Peter's point that really we don't want to wait until an optimum biologic dose has been defined in adults in order to initiate pediatric trials. We really want to get in sooner, and, in fact, as Eric was explaining before or suggesting before, that perhaps the place for us to get in is as early as the dose level at which Grade 2 toxicities are being seen in adults, because at that point we know that there is at least a dose that has biological activity, and maybe we can think about initiating pediatric trials at that point, without even waiting for completion of the adult Phase I trial, and build into our study designs the fact that, if the adults are able to escalate very rapidly, that we can then skip a few dose levels in order to catch up, so that we don't have to duplicate things that have already been done.
So I would just like to suggest that what we want to do is work on study designs that enable us to get trials going sooner, to get trials completed in a reasonably short time, rater than waiting for as much information as we would need in order to define what is truly a biologically-active dose.
CHAIRMAN SANTANA: Clinton?
DR. STEWART: Peter, I heard you describe the optimal biological dose, and earlier I guess it was Ed talked about the biologically-effective dose. Frank talked about a minimally-effective concentration. Peter, you talked about measuring AUC. We've got all these parameters floating around.
It seems like, I don't know, maybe that reflects sort of the confusion in the system. It seems to me if we were to pick -- this whole dependence on dose is a little bit concerning to me, and it goes back to the slide that I showed during my talk about the study that we did, the POG 92-75 study, where you escalate dose, but you are not escalating exposure. So you can change the dose until the cows come home, but you are not changing the exposure that the child is getting.
So I think we've got to be careful about this dependence upon changing dose and thinking that you are changing what kind of exposure that the child is getting. So I like the idea of the AUC or the minimum effective concentration that Frank talked about, but if we are going to do that, then we are going to have make the commitment of resources and the infrastructure to be able to follow through with that. It also means that I don't think we can do that with every agent that comes down the pike. I think we are going to have to be selective about that, because it does require a great deal of work and commitment of resources to be able to follow through with that. So that's a comment.
The question I would ask you, Peter, this is a slightly different one, but you were talking about genotyping in Phase II and Phase III studies. Assuming that a patient is a genotyped in a Phase I study, you wouldn't genotype them again, would you?
DR. ADAMSON: No.
DR. STEWART: You are not talking about rephenotyping or regenotyping?
DR. ADAMSON: No.
DR. STEWART: Okay, I just wanted to make sure about that.
CHAIRMAN SANTANA: But let me follow up on Clinton's comment. I think the advice that I would propose, and certainly this is my opinion and certainly not that of the Committee, is that you have different scenarios. You have a scenario where there is a classical way of doing Phase I studies, where you start at some level that has either been predefined based on adult data or concomitantly as the adult data and the pediatric study are being done.
I think that will apply and has worked very well for the majority of the things that we have done with cytotoxics. Continuing with the cytotoxic story, but there may be certain drugs, which Clinton was referring to, in which we may have pre-clinical data that supports that a different concept be applied to defining the escalations or how the Phase I study is done. I can't give you a broad recommendation on that. It is going to depend on what pre-clinical data exists, either from relevant pediatric models or other models that may exist that would then indicate that maybe a different end-point, like the AUC or the systemic exposure, may be more relevant in applying your design to the Phase I concept.
So I think for cytotoxics we want to move to No. 2, but we have to recognize that there is not a lot of infrastructure to support that at present, and that we have to be very selective in which drugs we could apply that principle to. Although ideally it should be with every drug, I don't think realistically we are at that point yet.
For biologics, I think it is a completely different story, in my view. I think the principles of the traditional Phase I escalation I think may not, as you heard earlier today, may not apply. I think in those settings I would argue more like Frank was saying earlier, that in those settings with biologics, not cytotoxics, that we do the parallel scenario studies, because, clearly, the differences in activity or potential toxicity may be very different in the populations.
DR. HIRSCHFELD: May I just ask for a little clarification? I thought I was hearing that it might be appropriate in the Phase I study to look at, as an end-point, the first doses which you see Grade 2 toxicity, independent of whether there is a target or some other biological assay, but that your Phase I study would guide you toward a toxicity, and that, in turn, would be informative.
CHAIRMAN SANTANA: Do you want to address that? I think you were the one --
DR. ADAMSON: I didn't raise it, but, Ed, if you want, I think the intent is, when you see a Grade 2 in adults, is that a reasonable time to say, okay, this is a reasonable dose to start in children, and then we can catch up to wherever the adults are when they have gotten to their MTD or closer to the MTD? At least that was my interpretation.
DR. KORN: Yes, I just want to add, I mean, I didn't say it, but it should be understood, of course, that if you are using a molecular target or some targeted response, that if you see toxicity, you have to stop escalating and back off.
DR. HIRSCHFELD: Thank you for that clarification. Just so that we understand, a potential direction for Phase I, an adult study would have as one of its end-points some toxicity, independent of whether a targeted dose or some bioassay was going to be invoked, and that that toxicity, in turn, could be a guide for initiating pediatric studies.
CHAIRMAN SANTANA: Malcom?
DR. SMITH: Two comments: One is that, in terms of the AUC and that as an end-point, you know, we have done that with 06-benzyl bromine, and so there's precedence for doing that, and I agree completely with Peter that I think we will be doing that more and more in the future as we have a systemic exposure that has the appropriate effect in adult patients, and then we want to make sure that, when we use a similar dose in children, that we are attaining that systemic exposure. So we have done that, and I think, as Peter says, we will be doing that more in the future.
The second comment is I would caution, we can do what Steve just asked and what Ed said; we can start at the first Grade 2 toxicity. I would caution our goal isn't to start pediatric studies as quickly as we can. Our goal is to complete them as quickly as we can. If we add five or ten patients below the MTD, using that design, whereas using another design we add all of the patients closer to the MTD, and that takes your patients, then, in fact, we are studying more drugs; we're completing perhaps more studies more quickly, and the point I made earlier: that because adult Phase I trials can be completed so quickly with dose levels occurring very quickly, the rate-limiting step with the design that Ed described is most commonly getting to the adult MTD, so that you can jump over three or four levels in the pediatric study to get to that.
So I think we need to keep our eyes on the prize, and in terms of Phase I that is the number of drugs that we can study, hopefully, picking the very best ones that we can.
DR. ADAMSON: Yes, I agree, Malcom, to an extent. Part of the problem, at least over the past few years, has been that we haven't had enough drugs to study for children, and that we have had children who simply have not gone on to study because there were no open studies.
So it is always a balance as far as -- it is not only completing studies on time, but it is making studies available in a timely fashion. I don't think there is a hard-and-fast rule as far as when to start. Clearly, for some of these, we have started, in my opinion, much too late, and that is when the drugs have been on the market.
But there will have to be a balance to when we think it is appropriate, and certainly, if all patients are going on to open studies where the adult MTD is known, then there will be less pressure to open studies earlier.
DR. PRZEPIORKA: During lunchtime I was encouraged to re-ask two questions that were deferred from this morning, one of which had to do with the dose when moving from an adult to a pediatric study, and whether or not you should continue on a milligram-per-kilogram dosing versus milligram-per-meter-squared, especially in light of the fact that we recently have a drug that went out with a milligram dosing in adults without any basis on weight whatsoever? So no one is sure how to do this in children.
Secondly, just to readdress the question of what is clinical and what is research, and we have now the option for participants not to participate in the pharmacokinetic studies. How many patients do you really want to get pharmacokinetic studies in this heterogeneous pediatric population before you're certain that you really have the dose to get the right AUC?
CHAIRMAN SANTANA: Clinton?
DR. STEWART: Well, I would actually like to address the second question first, and that would leave Steve with the first question, which is the more difficult one.
The issue of clinical versus research, as far as it goes with pharmacokinetic studies, I think is a very important one and actually one we talked about at lunch today. A Phase I study, let's talk about that specifically first. Whether it is part of the primary objective or part of a secondary objective is one way that we look at it at St. Jude, whether or not is considered in the informed consent aspect, is whether or not a child is to be -- to participate in the study, to receive the study drug, that they have to participate with the pharmacokinetic studies. If it is a secondary objective, it is part of a checkbox and they are allowed not to participate. I guess that is just one comment.
CHAIRMAN SANTANA: Could I follow up on that, Clinton?
DR. STEWART: Yes.
CHAIRMAN SANTANA: I think the point is that if the end-point is that you are going to define the toxicity based on a pharmacokinetic parameter, you can only realize that end-point if you get PK; whereas, if PK is a secondary issue to the design, that the main design is a completely different question, then I think in those circumstances then the option of yes or no PK is something that you can allow.
DR. STEWART: Right.
CHAIRMAN SANTANA: But in the first scenario, if you are really trying to define the toxicity based on AUC, how can you not get PK? In that scenario it is absolutely crucial to that study. So you either participate in the study or you don't.
DR. STEWART: Right. So then the second part of your question is, how many patients do you need? And that is something that I have been listening to as we talk about the number of patients in a Phase I study. I mean, obviously, one of the things you want to do is to minimize the number of patients that receive what is considered -- and I hesitate to use this term, but a subtherapeutic dose and maximize the opportunity that a child would have for a therapeutic response.
So you want to try to minimize the numbers of patients in a Phase I study, I would assume, and yet I counterbalance that with the need to learn about the disposition of a drug in the pediatric population. One of the references that Steve sent out was the paper from Elizabeth Eisenhower, who was an author on it out of JCO. It was a consensus conference about Phase I studies. My colleague, Mark Ritane, wrote in that about the need for studying additional patients, and one of the ways he suggested doing it was to study more patients at the MTD. So you had 20 to 30 patients -- this is in the adult population -- 20 to 30 patients at the MTD.
Now Dr. Boyett is over there squirming in his seat because, you know, there is no way to say that is going to provide a, quote/unquote, "statistically-valid" estimate of clearance in that particular populations, but what it will do is it will give us a better handle on the estimates of clearance than studying two or three children at a particular dose.
So how many children do you need? I can't give you that number, but I can tell you we do need to study an adequate number of children, more than, say, six or eight or ten. I would be an advocate of being sure that in our Phase I studies we don't sacrifice the knowledge that we need to gain from those pharmacokinetic studies that we can use as we move into the Phase II, that we just don't sacrifice that knowledge.
CHAIRMAN SANTANA: Dr. Bernstein?
DR. BERNSTEIN: I won't disagree with Clinton about the pharmacokinetics, but I would only disagree with the setting and say that what we can start to do more of is, as we take drugs into Phase II, especially if there is a limited sampling strategy that has been designed, is that we can look at the pharmacokinetics in the initial Phase II population as well to complete the number that is required to actually get an idea of the things like Clinton is talking about.
CHAIRMAN SANTANA: Have we satisfied this question, Dr. Hirschfeld?
DR. HIRSCHFELD: I think Dr. Leeder was going to make a comment --
CHAIRMAN SANTANA: Oh, I'm sorry.
DR. HIRSCHFELD: -- about dosing, yes.
DR. LEEDER: Well, these discussions have been recorded for all posterity, and I really don't want to go down on record as the individual who sets policy for converting adult doses to pediatric doses. But there are a couple of points worth making.
One point is that, until November the 19th, which is when I got the email message from Steve that kind of laid out what might be expected of me, and this is specifically this issue of converting adult doses to pediatric doses, I had not really seriously considered this whole aspect of the ramifications of correcting clearance for body weight versus body surface area versus liver mass, or whatever it is.
It is my impression only, without having a detailed slog through the literature, that the apparent differences in clearance between adults and pre-pubertal children tend to be less when corrected for body surface area than for kilogram body weight. Now I don't want to completely cop out on an answer, and what I am going to suggest is that perhaps some of the information is available to us, and some of the notes that I made have been to go back and look at some of the existing pediatric data for compounds such as medazalam, carbomezapine, which are pretty well accepted to be measures of P450A4 activity, for example, where there are a lot of pharmacokinetic data in children.
I think one thing that needs to be borne in mind is that pharmacokinetics classically have described the disappearance of the parent compound. In cases like Irinotecan, where it is an active metabolite, what is important is probably the exposure to the parent compound. And the point I am trying to get at is the fact that perhaps the pediatric clinical oncology arena maybe ought to take advantage of some of the information that is available for some compounds in the pre-clinical testing phases with respect to in vitro drug metabolism.
Certainly for a lot of compounds there is a process that is called reaction phenotyping, where a battery of human liver microsomes that have been phenotyped for particular activities, recombinant heterologously expressed human enzymes, are used to map the drug metabolism pathways that are involved in the disappearance of the parent compound, and perhaps the formation of the pharmacologically-active anti-neoplastic compound, with the intent of at least identifying polymorphic pathways or perhaps some would say avoiding them. Maybe you don't want a compound that is a substrate for 2D6 which is deficient in 5 to 1 percent of the Caucasian population.
The other reason this process is completed is to characterize and perhaps minimize drug-drug interactions. The point I am trying to make is that, if we go back through the literature and look at this issue of clearance with respect to the metabolic pathways, the specific metabolic pathways that are involved, how those metabolic pathways change during development, whether or not they are better correlated to thing such as a correction for body weight, body surface area, or liver mass, that it may be possible for specific compounds to put all this information together and come up with some sort of a rational conversion factor.
But, as a starting point, again, I would like to repeat that my impression is that the differences are probably least or less with body surface area than with kilogram or total body mass, and I think it is just a little bit premature to pursue the liver mass at this point in time, until we do a few more studies.
CHAIRMAN SANTANA: But how do we address the issue that Donna was trying to, I think, hint at, which is, we have a product or an agent that's been approved on a milligram basis? How do we really relate that in terms of the type of studies and where we start in pediatrics? Because to me that's an issue. Can you shed some light on that? Because I think we are going to be seeing that more and more with some of these biologics. It is not going to be milligram per kilo, milligram-per-square-meter. It is going to be some total dose.
DR. LEEDER: I find, and maybe Clinton can comment on this as well, I find it very difficult, just from the concept of dumping a specific mass of drug into a volume of varying ranges, to be able to predict what concentration is going to come out. So I would probably pick any correction over a straight milligram dose.
CHAIRMAN SANTANA: Clinton?
DR. STEWART: So, as you probably know, we are faced with that with one of the upcoming clinical trials that we are going to have. One of the things that I have proposed, and Peter and I have talked about this, is then the first three to five patients, what we will do is we've got the dose; the dose was decided in adults in milligrams. We have converted it using a 1.73-meter-squared typical adult to a meter-squared, milligram-per-meter-squared dose in children. So what we will do is we've got the assay up online, and the first three to five patients we'll measure the concentrations and we will see what the levels are in our children relative to what they were in the adults, and we will be prepared to go from there.
So, I mean, that doesn't help the first child that is enrolled in the study very much, but that is how we are going to be prepared to react to that situation. So, I mean, that is one thing I can offer up.
One other thing I would like to say in terms of our experience with Topotecan, and I tried to allude to it a little bit earlier with the infants, the children less than two, it does appear that, if one were to normalize the clearance for body weight per kilogram, it does seem to normalize the differences between the infants and the older children.
So what I am saying there is I think that we have to be very careful when we talk about children. All children are not the same. You've got the little children, and then you've got the bigger children. I think that, as Steve pointed out, there are maturational differences in the way that they eliminate drugs. So we've got to be careful about just making the straight -- I mean, it is a difficult conversion from adults to children, but then even within the children, quote/unquote, "children" population, we've got infants and then those kinds of considerations. So I hate to make things even more complex, but we have to consider that.
CHAIRMAN SANTANA: Steve?
DR. LEEDER: If I could make one comment that might help simplify things, again, it goes back to this point that the closer the children -- for example, adolescence, post-puberty, chances are when you look at -- let me back up a bit and say maybe we can get some insights, for example, from the shapes of growth curves -- we all have access to the growth charts -- and look at when the growth charts start to flatten out as being the points where there are fewer changes with age.
So one of the areas where the growth charts flatten out is after puberty. So perhaps it might be possible to go directly from an adult dose, using an adult dose on -- correcting for whatever index you wish -- in a post-pubertal individual.
Before that point, it gets a little bit more difficult. Certainly the most difficult region would probably be the first year of life, I think where it is such a dynamic process that I would hate to predict what the appropriate scale ought to be in converting an adult dose to an effective but not toxic, or an optimum, dose in that age range.
CHAIRMAN SANTANA: Pat?
DR. REYNOLDS: I think one of the things we haven't considered in this is that you have very young children. You can't often take the formulations that are developed for adults. We certainly see this with the oral retinoids as being problem.
So the bioavailability of these agents, if you are dumping it out of the capsule and trying to mix it in with applesauce, or whatever it is, is going to change. So that is another parameter that needs to be considered in this. You can't necessarily directly translate these formulations, and some formulation development may be necessary.
CHAIRMAN SANTANA: Okay, I think we have pretty much covered this, Dr. Hirschfeld, unless you wish for us to make any comments?
DR. HIRSCHFELD: I thank you for all the input.
Did you have something you wanted to add, Dr. Rackoff?
DR. RACKOFF: Pat made the point I was going to bring up.
DR. HIRSCHFELD: Okay, excellent. Okay, let's move on then.
CHAIRMAN SANTANA: Yes. So the second question has to do with, and I will read the question, but it has to do with issues of extrapolation and then what efficacy data would be necessary for product labeling for pediatric indications. So the question reads, for the record:
"If tumors that are considered to represent the same disease or condition are found in two different populations and/or share a common biological mechanism that is supported by a body of scientific evidence that is generally accepted, and a therapy targets that mechanism, what type of studies would validate extrapolation of efficacy findings from one population to the other?" So this is the question of extrapolation of data.
Then the second part to the question is: "Product labeling to support a marketing claim usually is dependent upon demonstration of a patient benefit and an assessment that it is safe and effective for the intended use. If activity is noted in adults, and the same tumor type, based on generally-accepted criteria, such as histology, cytogenetics, common biological mechanisms, et cetera, exist in children, what evidence would be needed to establish efficacy for product labeling; that is, to make a market claim for children with cancer?"
DR. ADAMSON: I have very little to put forth, but I will throw out a suggestion. I think efficacy extrapolations are going to only apply in limited situations, for all the reasons you have simply stated in that question, where biology, and so forth, is the same between adults and children.
However, we all can name situations where that occurs. In my opinion, we are still going to need Phase I, and likely some Phase II data, in pediatric patients. However, if the pediatric Phase II data is consistent with the adult Phase II data, I would argue that that ought to be sufficient to extend labeling, if the adult randomized trial has shown efficacy, and that we need not attempt to repeat, nor would we likely have the ability to repeat, a Phase III randomized trial in children.
My only other comment along these lines comes back to, Steve, one of your opening comments about when the rule would be invoked. If I understood correctly, you said for certain diseases, an automatic waiver would be granted: prostate cancer, breast, lung. I would argue that granting an automatic waiver for that would be detrimental to pediatric patients.
The reason I say that is that one can envision that there is a new agent, a signal transduction inhibitor that's highly effective in prostate, and that signal transduction pathway we learn is a relevant pathway for pediatric tumor. If a waiver is granted, then we will be doomed in that situation.
So maybe I will leave it there and let the discussion follow.
DR. HIRSCHFELD: Well, Peter, first, I want to thank you for the advice, and I think the advice you propose there is at least consistent with our thinking on the question.
In terms of, we'll take the prostate cancer paradigm. If one is defining the indication, and this would represent a paradigm shift, if one defines the indication as, we'll say, a certain Gleason stage prostate cancer or hormone refractory prostate cancer, that's one circumstance. But if one is defining the indication as tumors that are dependent on a certain signaling pathway and have other characteristics, then the Pediatric Rule could be invoked.
When the rule was first developed, we didn't have, or didn't anticipate, the tools to make these, what could be called, cross-histologic diagnoses or mechanistic-based. But our thinking has evolved as the science has advanced, and what we would need would be a very firm scientific basis to make that extension, but if there was a firm scientific basis to make that extension, then in that case I think the rule could be triggered.
I would like to ask Dr. Pazdur if he would add some comments?
DR. PAZDUR: What you want and what we may want has to be supported by the law, because basically this is a mandate. Okay? So it is not an optional thing or it's not like, "Would you please study this perhaps?" It is a requirement that the companies study this, study the indication.
So, as Steve mentioned, it would have to be kind of a paradigm, king of sea change, where the academic community, treating oncologists in general, and a general scientific acceptance would be that these are the same diseases or entities that would represent identical populations with the same disease.
It couldn't be just that we're studying a drug that has overexpression or targets topoisomerase 1 and, therefore, we are going to then make the company study all tumors that have overexpression of topoisomerase 1, do studies in pediatric tumors, so express that.
So it is really a level, since this is a requirement and a mandate, that has a legal implication that we would have to be very careful in applying and would have to have a scientific foundation that is well-accepted by the community.
CHAIRMAN SANTANA: Just as a followup of that in terms of my own perspective is I think in terms of these issues of two populations and what data can be extrapolated from one or another, and although the question is posed as an efficacy question, I would caution that safety also has to be a part of the equation. So when one tries to extrapolate data from one population to the other, one has to look at certain parameters that may indicate a different scenario of safety may be similar to some of the PK discussions we've had earlier today.
So it is just not a matter of saying the populations are very similar in terms of the diseases, in terms of the biology, and therefore, the efficacy should be the same, but there may be some minor differences in safety that also should be a part of the equation.
DR. HIRSCHFELD: I appreciate that comment, Dr. Santana, and that is a major concern not just in oncology, but across all of pediatric therapeutic development in terms of, how do we know that we are getting adequate safety data, that something could be labeled?
Just, Peter, one other point: We have another program, an incentive program, for those circumstances where we would like to encourage development in pediatrics but we can't invoke a mandate. That program has been spectacularly successful in general. In oncology we have found that companies are now learning about it and have not been shy about making proposals.
CHAIRMAN SANTANA: I want to also make a comment regarding the second issue, which is that, once you have established this commonality in terms of the disease and the biology, then what additional information would you want to legally support a product claim for children with cancer? My position would be that, if the agent that you are discussing, that you are targeting, is a biologic with the common pathway, and that indication exists also in children, that you would have the same requirement, that you would require a Phase III study that would indicate the impact of that, wouldn't you?
DR. HIRSCHFELD: Not necessarily. We have been exploring broadly in pediatrics the issue of extrapolating efficacy findings. I think the paradigm that Dr. Adamson was discussing is something which might be acceptable. I don't think there would be a regulatory requirement to repeat an efficacy study with the definition -- and I'm looking at our colleagues over in that corner -- that an efficacy study is a study which is designed to establish the efficacy rather than just demonstrate an exposure/response relationship. I think the exposure/response relationship could be sufficient, but I will ask Dr. Pazdur.
DR. PAZDUR: Yes, I think I would agree with Steve. Let's take a specific example. If we were dealing with Hodgkin's Disease and somebody did a randomized study and proved the efficacy and combination of a drug and proved what we would be looking for, probably a survival advantage or a curability clinical benefit, probably since the disease would be quite similar in a child, the childhood population -- and correct me if I am wrong since I am not a pediatric pediatrician -- but if that was a similar disease, we probably would accept Phase II data, since it probably would be very difficult to do and replicate a Phase III trial in that identical population.
Here, again, it really depends on how comfortable we are that these are exactly the same diseases, but we don't want to be overregulatory or overburdensome in this regard. I think there could be a kind of bridging that could occur with Phase II data here, again, emphasizing the safety that you previously mentioned for a pediatric population and look at a surrogate end-point such as response rate toward a particular disease, and making sure the duration of that response was meaningful.
CHAIRMAN SANTANA: Thanks for a clarification.
Other comments or questions? Jerry?
DR. FINKLESTEIN: This question is either to Rich or to Steve. At the last meeting we spent a lot of time in comparing adult and pediatric tumors. In fact, we made the comment maybe histology was on the way up because molecular biology was on the way in.
My question is whether the incentive program would apply to those pharmaceutical companies that would like to use molecular biology as a way of studying pediatric tumors. Using your example of prostate cancer, for example, with some signal, could that be invoked and would you accept that? So I am asking it, actually, from an industrial point of view.
DR. HIRSCHFELD: The incentive program doesn't require any relationship between the adult tumor and a pediatric disease. In fact, within the FDA in general we have adult indications which were approved in one Division, in one disease area, and the pediatric indication that is being studied is in a completely different area. So there's no need for a biological or mechanistic linkage in any way.
However, if the rule is invoked, that does not preclude qualifying for the exclusivity extension. That would then require further discussion. It is not an automatic trigger or something that one can assume. It would require specific discussion, but that's plausible.
DR. PAZDUR: I think it is also important to put in context the rule as it can be applied. What we are dealing here, from our previous meetings, is a relatively limited number of diseases that really go back and forth. You could talk about acute leukemias, for example, some lymphomas, some brain tumors.
But, remember, the sponsor has to be studying the drug in that indication in the adult population, and really when you take a look at what indications are being studied in adult populations, for example, they are the big diseases. They are breast cancer. They are colon cancer, prostate cancer, pancreatic cancer. We have very few applications where the rule really can be nailed down, where we could have this linkage between and exert our regulatory authority in that regard. So I think we have to put this in some context of the real world, and that is why, as Steve already mentioned, the exclusivity arrangements probably have a greater degree of flexibility really to encourage drug development in pediatrics.
CHAIRMAN SANTANA: Dr. Boyett?
DR. BOYETT: Just a point of clarification: Suppose you take your Hodgkin's Disease example and you've got efficacy data in adult and you do a Phase II trial and you show some response in pediatric Hodgkin's Disease. So now you grant a labeling indication in pediatrics.
You know, there are very few pediatric cancers that are cured by single agents. How to incorporate one of those single agents into a regimen, that's a different question. If you grant this labeling, what is going to be the implication of COG, for instance, trying to learn how to use this particular drug in a regimen to treat newly-diagnosed Hodgkin's Disease patients?
DR. HIRSCHFELD: Our history of approvals is actually that very often in the label the product that is being granted the claim is being granted the claim as part of a regimen. A recent example is the approval of Irinotecan, where it isn't Irinotecan that is approved; it is Irinotecan in the use with combination with other drugs in a particular setting.
So in the case where a new Hodgkin's therapy would come along, boy, if we had something that is as a single agent could treat the disease, that would be fairly spectacular, and it would warrant that. But, in general, the data that are being submitted are data where it is not for monotherapy but as part of what, hopefully, is a therapeutic advance over previous regimens.
Rick, do you want to comment?
DR. PAZDUR: Yes, because probably what we would do in the adult indication, it probably would be used in combination with other drugs, and that same combination would be studied in the pediatric population, looking for similar response rates between the adult population and the pediatric population. Here, again, it is a bridging aspect. We wouldn't just take a single agent and say, well, you have 20 percent response rate and, therefore, have activity in Hodgkin's Disease and say, well, that's sufficient for approval in --
DR. BOYETT: I guess I was misinterpreting the use of Phase II here because your intent in --
DR. PAZDUR: Let me clarify that.
DR. BOYETT: Yes.
DR. PAZDUR: Probably a more accurate description would be a single-arm trial looking for response rate.
DR. HIRSCHFELD: An exposure/response study is what we usually use in broader pediatric term, not Phase II in the narrow sense.
CHAIRMAN SANTANA: Dr. Korn, one last comment on this issue, please.
DR. KORN: Well, that's going to be kind of tough now, isn't it? So you're going to have a combination therapy, a small Phase II trial, and you are going to try to show that response rate, whatever, is consistent with the small Phase II adult study. I mean, I'm not sure what that means. You almost can say, well, why bother, because you're not going to find anything out from doing that anyway?
DR. PAZDUR: Well, here, again, I think there is some elements of safety that we would consider by looking at it in a Phase II study, which I think is important, as Victor implicated. Remember, these diseases -- and here, again, the link of how comfortable we feel between the disease and the child and in the adult is one that is really going to mitigate what degree of information that we would want from the sponsor and the clinical trial. If we really believed that these were identical diseases, to make a sponsor do a separate Phase III study with a combination against the standard therapy of "X" hundreds of patients may not be warranted here to show either superiority or non-inferiority.
DR. KORN: Right. Well, the alternative is not to do the Phase II study, but do a small safety study then.
DR. PAZDUR: Here, again, the Hodgkin's disease might not be the best example. It is an example that I threw out, but, here again, a lot of these factors are mitigated by how comfortable the feeling is and the scientific underpinning between extrapolations.
DR. HIRSCHFELD: I would just add to that that I think we're ethically bound to do the studies only in patients with the diagnosis. If we did even some modest study and saw no responses, I think everyone would begin to get worried. So although the purpose of the study wouldn't be to establish efficacy, you would certainly want to make note or want to be reassured that you were getting the same type of response.
DR. KORN: Right, but I can understand for mono-therapy but for combination therapy, you are going to get responses. So that's not going to really be an issue.
CHAIRMAN SANTANA: That would be the added value.
Okay, let's move on because I think we have covered this one pretty much.
So the third question is relevant to mono-therapy versus combination therapy. That is, "If a therapy is intended to be used as part of a combination, are monotherapy studies in children advisable? If so, what types of studies should be implemented prior to initiating the combination studies?"
DR. ADAMSON: I'll just comment. We are more limited in pediatrics than we are in adults in this situation. If a single agent has no prospect for direct benefit, and is greater than minimal risk, we cannot perform a single-agent study, I think. At least that would be my interpretation of the regulations.
However, I think -- and I kind of shudder to have used the word "window" here, but I think one can consider a single-agent window, and let me expand what I mean by that. Where one starts with a new agent, determines acute toxicity, defines some pharmacokinetics, and continues with a combination, and I have a couple of examples that I will share with you, which is where we have taken this approach.
The first is a study that Frank Balis and Kathy Warren led with RMP-7 and carboplatin, and the subsequent one will be a study that COG will be starting soon with the antisense compound Genasense. So RMP-7 is a drug that modulates the blood/brain barrier. There was a fair amount of pre-clinical data and some adult data that it could potentially increase the efficacy of standard cytotoxic agents, but by itself RMP-7 was not known to have any anti-cancer potential.
So the pediatric trial design was the following: The first day of the first cycle of therapy, the patient received Cereport as a single agent to determine its acute toxicity, its acute tolerability, as well as the pharmacokinetics. Then it was immediately followed in day two and day three of the first cycle by the combination of carboplatin and Cereport, Cereport being administered at the time of peak carboplatin exposure.
All subsequent cycles were then two-day exposures of just the combination of the two. So in this case we know we were not getting data as far as any chronic toxicity of single-agent Cereport. This is a drug, however, that essentially only has acute toxicities, but we were getting a fair amount of information at the same as allowing patients a full benefit in a given cycle of being exposed to an active combination.
Similarly with Genasense, which is a BCL2 antisense compound, the trial design that we are proposing, and that CTAP will soon review, is that there is a lot of data, first of all, that to get maximal decrease in BCL2 expression, you need a sustained exposure to the antisense compound. So there will be a seven-day continuous infusion study of the antisense compound, and during the initial four days the patient is only going to be exposed to the antisense compound as a single agent.
We will get some acute toxicity information here. We will get some pharmacokinetic data here, and then during that first cycle they will then get exposed to the combination of some standard cytotoxic agents, in this case doxyrubicin and cyclophosphamide, in conjunction with the agent that by itself would have minimal activity as a single agent.
So we have addressed this in pediatrics, and we have done it in situations where it has been, I think, relatively straightforward to do, and that is when we are either anticipating only acute toxicities or when the biology of the drug action warrants administration of the drug prior to combination.
However, during a trial where there is just a cycle of drug with no potential for therapeutic benefit, I do not think it is going to be feasible in pediatric patients.
CHAIRMAN SANTANA: Yes, I want to expand on that. I think that the intent defines the study that you want. So if we know pre-clinically or from some other data that the agent by itself doesn't play a major role, I think it would be both scientifically and ethically invalid to do a single-agent trial if ultimately that is going to be needed to be done in combination.
So I think one has to be very specific about the agent that one is talking about, and what one knows about that agent a priori before one mandates that that agent be studied as a single agent, based on the intent.
DR. PAZDUR: How does one know actually that, that the agent has no activity? You know, here again, there is activity and activity; one could be activity measured as response rate for a cytotoxic drug versus time-to-progression for a more cytostatic therapy, and just to say, well, because the drug does not produce response rate, one does not have to demonstrate single-agent activity would be somewhat hard necessarily to swallow for us.
CHAIRMAN SANTANA: No, I agree with you. I think with cytostatics, it is very controversial and it is very difficult. With cytotoxics, I think it is a little bit easier in terms of answering the question. With cytostatics, you are faced with the issue that you really may not know if by a single agent given over a prolonged period of time you do get some activity.
So I think in that scenario you are really going to have to rely very heavily on some pre-clinical data and what data you may discern from adult studies before you mandate a single-agent pediatric study.
DR. BALIS: The area this most applies to is modulating agents, agents that modulate resistance, for example, to anti-cancer agents, where giving them alone obviously makes no sense.
But the other, the converse is also true. There are a number of studies with MBR inhibitors in adults where the agent, the chemotherapy agent, was given alone, and then on a second cycle it was given with a modulating agent.
I think I have the same difficulty doing that type of trial in pediatrics, in that many of the diseases that we treat tend to progress very rapidly if not given effective therapy. So even doing a chemotherapy alone followed by a cycle later with modulating agents I think is a difficult trial design to undertake in our population, if there has been efficacy shown for that modulating agent in adults.
CHAIRMAN SANTANA: Any other comments or advice on this issue?
CHAIRMAN SANTANA: Okay, so let's move to the fourth question which is the Phase II window design. For the sake of time, I won't read the paragraph, but I think the question is: "What circumstances (for example, types of diseases, expected results with available therapy, prognosis, types of patients) would warrant a Phase II window design?" And I'll let Peter comment on that, and then I will give my insight, too.
DR. ADAMSON: I think I made my opinion known. So, in fact, I don't even have a slide on this one, but I would echo what Malcom had said earlier. I think in order to accelerate the drug development process, we need to start believing our own data. In the relapse setting for that vast majority of childhood cancer, not all but the vast majority, including high-risk ALL, we, in general, don't have meaningful salvage therapy coming off of current frontline protocol.
I think a Phase II window study that either a modulating agent or, in fact, a novel agent in certain circumstances may be more appropriate in the relapse setting than it is going to be in the upfront setting for diseases where there is no known effective standard therapy, and I would probably limit that to brainstem glioma at this point. It may be acceptable to do it there. However, beyond that, I think we run into many of the ethical issues that have been discussed, as well as the value of doing them that I had mentioned earlier.
CHAIRMAN SANTANA: Yes, my comment was going to be, as you heard earlier, that I think the consensus document from four or five years ago has been a good tool and a good guideline for answering this question. Until that document is revisited, I think that document should serve the basis to answer this question.
So for those of you who haven't read it, I would encourage you to read it because I think it does provide some insight into the potential type of patients, the expected results, and some of the relevant issues regarding the ethics of these trials. So until we revisit that guideline and that consensus, I think that serves as a good parameter for us to answer this question to give advice to the FDA.
DR. BALIS: Yes, I think we are moving into targeted therapy. The one potential area where it could be useful is if we want to define a biological effect of a drug that has a specific target that is completely separate from any effect that standard therapy has on a tumor, because so many of our tumors now are treated neoajuvantly, and we have the opportunity potentially to get tumor tissue, because we are doing a clinically-indicated procedure in those patients at some point later during their therapy that we could administer drug maybe even in combination with standard therapy during the neoajuvant phase, where we could measure a biologic effect when we go in and actually remove their tumor.
DR. BERNSTEIN: I think that Peter is a little overly restrictive about disease categories. I think there are a variety of sarcomas, metastatic at diagnosis, for instance, where the prognosis is sufficiently poor so that one could consider the addition of new agents early on rather than waiting for recurrence.
I won't push the point too hard because I actually think that, for the immediate question, which is, how are we going to give guidance to the FDA, I think that there will be rare circumstances in which window trials are actually going to be very informative in terms of the things that you're looking for.
I agree with what Victor said, that I think the document that was produced by our consensus meeting, which with Susan I had the privilege of attending several years ago, I think is reasonable as a guidance.
CHAIRMAN SANTANA: Do you want to make some final comments and then, Wayne, because I think we are running short on time?
DR. ADAMSON: Yes, I am going to turn this over to Wayne. The only one I want to talk about, and I am going to jump to that, is on prioritization.
The challenge for us is that we simply cannot study all the drugs that are in the developmental pipeline in children. So we have a choice. We can do it randomly or we can look at pre-clinical models that we know in most cases have not yet been validated, but at least may give us some basis for helping to prioritize.
What I find is a remarkable situation now -- and, again, this may be a little bit overstated, but with FDAMA in the final rule, we are faced with a situation whereby it may be easier to administer a new drug to a child with cancer than it will be to administer the new drug to a mouse. I say that because intellectual property issues now -- and this is a two-way street; this is not just industry; this is academia working with industry, and basically lawyers going at it with lawyers -- where we have no pre-clinical data in pediatric tumors at a time when we are ready to embark on pediatric studies.
This is a situation that we are going to have to address, and we are going to have to improve upon because the problem is just going to expand as the number of agents in the pipeline increase. Until we solve the issue of overcoming the intellectual property debate, where we can get new agents into models that we openly will say have not yet been validated, but which we intend to validate over time, we are going to be operating in the blind as far as being able to prioritize.
That is simply stated here, that I think pre-clinical models for pediatrics, we are likely to have to rely on more heavily than one will have to do it in the adult situation.
So let me turn it over to Wayne, who I think is --
DR. RACKOFF: In the interest of time, I will just do it from here because I've only got three or four slides, and I'll provide them, if you want, but they are fairly straightforward and really address the last two points.
I just want to warn people that the last Advisory Committee I took part in was in 1977, the National Health Insurance Advisory Committee. I was a staff member, and you know how successful that was. So I hope this project will be a little more successful in coming up with specific guidelines.
I am speaking now, it is not the opinion of Johnson & Johnson; it is not the opinion of any one group. Raj Malik and I, who chair the COG Committee, have conferred a little bit during the breaks. What I would like to do is provide in five minutes not an industry opinion, but a perspective on what's gone on actually in the last four meetings.
I think that the overall goal here is a little different from what's stated in the law. It's been there in the air in all the meetings. The goals are early access to new agents, to accelerating the process of drug development for new agents, and, finally -- and this is the industry perspective -- in trying to get some consistency in that process.
Now I think there are three things that have to be taken into account that have come out at these meetings. One is that pediatric rule and exclusivity, although we have tried to divide them for these meetings, I'm a "lumper" and that's why I like these meetings, because the rule lumps adult and pediatric cancers together. I think the Pediatric Rule and exclusivity have to be considered in toto as moving toward that goal, much in the same way as the Bill of Rights, the Civil Rights Act, and Voting Rights act are not standing alone doing what they are supposed to do.
So I think that as the Agency moves forward, our advice, my advice is that we consider these things working in concert, and that's where Peter established a gap, but I think, again, that pediatric exclusivity and the way these guidances are written around pediatric rule can help to fill that gap.
Second, and I agree with Peter and I think it's the major issue that we face in pediatric drug development in cancer, is that the patient-to-agent ratio is going down, not going up, because I think Pediatric Rule and pediatric exclusivity have been effective. There is anticipatory action being taken by a number of companies. Here I will speak about my company where there's a Vice President for Pediatric Drug Development and a whole Pediatric Drug Development Group.
But we are dealing with small populations, and Rick Pazdur alluded to this. Where the Pediatric Rule is applied, it's always going to be applied in small adult populations and small pediatric populations because those cancers, with very few exceptions that are linked in the Pediatric Rule specifications, are small tumor burdens, public health burdens.
So we need to link, but we need to link for the sake of the adults who have AML and for the sake of the children who have AML. I think there are two ways to link. They have come out at these meetings. One is sequential development, and I think there are advantages to that. What needs to be stated, I think, in the guidances are what allowances will be made for prior probabilities, not necessarily using a Bayesian approach or a classical approach, but what allowances will be made in the development process for prior knowledge. I think to the extent that those can be specified clearly, they will encourage further anticipatory development without invocation of the rule.
Second is there's parallel development, and a lot of people, I think, on the academic side and in COG would like to see us do parallel development. In parallel development you'll get faster development, too, but then you need allowances for combining data in studies run in parallel or linking the studies themselves. Here I think the COG and the NCI come into play more than the Agency.
That brings to one of the other points, which is that, in setting priorities, we have to have COG, NCI, CTAP, the Agency, and industry at the same table. Now the industry people can't be at the same table at the same time because there are intellectual property issues. But I think that there has to be some sense in the guidances that provides a fair and consistent way to combine all four of those forces to be able to set priorities, because that's the only way we're going to deal with this patient/agent ratio being very low.
So, with regard to specific invocation of the rule, and this is the final set of points, I think that the guidances should set out, or the letters, or at the meetings, that go out after the meetings, what bridging studies are required, what specific bridging studies will be acceptable, to the extent that the Agency can do that within the law, and, finally, what will be a significant treatment advance is going to be a very important point in setting priorities.
To the extent possible, if the guidance that comes out of these meetings can set out what is significant treatment advance, I think that that will help, again, determine how much anticipatory activity there will be on the part of the industry drug developers.
So, in summary, again, I think the rule and exclusivity are working. So I think to the extent that these guidances build on that, they are going to be very important. I think that the major issue from both the company's standpoint and I think from COG and every other standpoint is that patient numbers are limited, and we have to have specific guidances that will combine the forces of NCI, COG, industry, interested members, and the Agency, to set priorities in a way that makes sense and that is consistent and fair to both industry and to patients.
CHAIRMAN SANTANA: Thank you for those comments, Wayne. As you note, there is that effort underway. There is a group that meets at COG that has representatives from FDA and industry that tries to address some of these issues. I think there's still a lot more work to be done.
DR. RACKOFF: One more point that came up, if you don't mind, real quickly: There is work going on, too, on this issue of material transfer for pre-clinical studies. It is going to take some time and it is going to take some battles among lawyers sitting down in a room, locking them up, to come up with what might be a master agreement that's acceptable, but under Malcom's guidance that effort is underway.
CHAIRMAN SANTANA: I want to thank you for those comments, Wayne.
DR. HIRSCHFELD: I also want to thank Dr. Rackoff for his participation previously as an observer and now at the table with us. We hope that we can continue to have industry representatives on the table for this Committee forever essentially.
We will also, I think, have to acknowledge that, of the products that are available to treat patients with cancer, whether they are children or adults, it's been the pharmaceutical industry that has done the brunt of development work and has taken the risks and the distributions and maintained the quality control, and I think should be acknowledge for the contributions that are made in that regard to the public health.
I wanted to pick up on the theme of the matrix because we all feel we are part of a matrix. We should in no sense be perceived as adversarial or that one has a barrier to overcome, or if we can only get around the regulatory hurdles, but rather that we view the regulatory mechanism as a way to ensure high-quality products for patients, ethical and consistent scientific development.
We have been working ourselves with international colleagues, and I think both the pharmaceutical industry and the regulatory community have started on a path which I hope there's no return from. That is to get greater international cooperation, and it is through international cooperation I think that we can address some of these issues of limitations of numbers and prioritization. I am very heartened that we have at this meeting some international representation and that we look for further development in this arena as well.
CHAIRMAN SANTANA: I want to echo that. I think it has to be a conversation that includes many different parties. I was encouraged to see that there were colleagues from across the ocean who came today and provided some of their effort and time at this meeting. So I want to personally and publicly thank you for that effort.
DR. PAZDUR: But I think it's important that we realize that drug development, in essence, is an international, global development process that occurs. So we don't approve drugs in the United States in isolation. In fact, our regulatory actions have great implications not only in Europe, but throughout South America and Asia.
CHAIRMAN SANTANA: That is correct.
DR. PAZDUR: Just to echo Steve's words, it is basically we have to be cognizant of our more widespread regulatory activities.
Nevertheless, getting back to this idea of how do we prioritize drugs, I think this is one of the things that we will be using this group as in future areas, to give us the scientific information both using the Pediatric ODAC Subcommittee as well as individual members, in consultation with us on individual applications.
There obviously are many drugs. Which ones to study in pediatrics needs to be really addressed by the people that are studying them and treating the patients. Not all drugs are appropriate, obviously, to be studied in a pediatric population, and we need to have that conversation with you on a long-term basis.
CHAIRMAN SANTANA: And we look forward to providing whatever help and guidance we can to the Agency in that regard.
If there are no further comments, I want to thank -- I'm sorry. I'm sorry, go ahead, Susan.
DR. WEINER: I thought that today's meeting was really a very good one, Steve, and I thought that the discussion was authentic. I very much appreciate the notion of the matrix, that is, of all participants, since we have an important charge.
But I guess what I would like to add is that I think that at every step of the way we have to be cognizant, every component of this matrix has to be cognizant of how to cut corners, how to make the process more efficient, where things can be more consistent, and how to promote appropriate uniformity. There are certain variables in this process which we can control -- sample sizes, ethics, et cetera. Where we can control it, I think it's really an obligation to the kids and families. Thank you.
CHAIRMAN SANTANA: Thank you, Susan.
I think we have fulfilled our goals as best we could, Dr. Hirschfeld and Dr. Puzdur. So, with no further comment, I want to thank everybody and declare this meeting closed. Thank you.
DR. HIRSCHFELD: Thank you.
(Whereupon, the proceedings concluded at 4:50 p.m.)