1

DEPARTMENT OF HEALTH AND HUMAN SERVICES

FOOD AND DRUG ADMINISTRATION

CENTER FOR DRUG EVALUATION AND RESEARCH

PEDIATRIC ONCOLOGY SUBCOMMITTEE COMMITTEE

OF THE ONCOLOGIC DRUGS ADVISORY COMMITTEE

Wednesday, March, 17, 2004

8:00 a.m.

5630 Fishers Lane

First Floor Conference Room

Rockville, Maryland

PARTICIPANTS

ONCOLOGIC DRUGS ADVISORY COMMITTEE:

Donna Przepiorka, MD., Ph.D., Chair, ODAC

Johanna Clifford, M.S., RN, BSN

Pamela J. Haylock, RN,

Consumer Representative, ODAC

CONSULTANTS (VOTING):

Victor Santana, M.D., Chair

Peter Adamson, M.D.

Alice Ettinger, M.S., RN

Peter Houghton, Ph.D.

Eric Kodish, M.D.

C. Patrick Reynolds, M.D., Ph.D.

Susan Weiner, Ph.D.

Ruth Hoffman, Patient Representative

Barry Anderson, M.D., Ph.D.

Lee J. Helman, M.D.

Malcolm Smith, M.D., Ph.D.

Paul Meltzer, M.D.

Chand Khanna, DVM, Ph.D., DACVIM

Kenneth Hastings, Ph.D.

ACTING INDUSTRY REPRESENTATIVE (NON-VOTING):

Antonio Grillo-Lopez, M.D.,

FDA STAFF:

Susan Ellenberg, Ph.D.

Steven Hirschfeld, M.D., Ph.D.

Ramzi Dagher, M.D.

Richard Pazdur, M.D.

Patricia Keegan, M.D.

Pat Dinndorf, M.D.

Grant Williams, M.D.

C O N T E N T S

PAGE

Call to Order, Victor Santana, M.D., Chair 5

Introductions 5

Conflict of Interest Statement,

Johanna Clifford, M.S., RN, BSN 7

Safety Monitoring in Clinical Studies Enrolling

Children with Cancer:

Opening Remarks, Richard Pazdur, Director,

Division of Oncology Drug Products, FDA 10

Introduction of Issues and Agenda,

Steven Hirschfeld, M.D., Ph.D., Office

of Cellular and Gene Therapy, FDA 11

Protecting Children in Cancer Research: What Really

Matters, Eric Kodish, M.D., Director, Rainbow

Center for Pediatric Ethics 23

Legal Responsibilities for HHS Supported Studies,

Michael Carome, M.D., Office for Human Research

Protection, HHS 49

Legal Responsibilities for Studies with FDA

Regulated Products, Steven Hirschfeld, M.D.,

Ph.D., Office of Cellular and Gene Therapy,

CBER, FDA 61

Enrollment and Monitoring Procedures for NCI Funded

Studies, Barry Anderson, M.D., Ph.D., Cancer

Treatment Evaluation Program, NCI 70

Open Public Hearing:

Wayne Rakoff, Johnson & Johnson 87

Monitoring Procedures in a Private

Children's Hospital,

Victor Santana, M.D., St. Jude

Children's Hospital 88

Committee Discussion 116

Questions for Discussion 154

C O N T E N T S (Continued)

PAGE

Use of Nonclinical Data to Complement Clinical Data

for Pediatric Oncology:

What are Microarrays and How Can They Help Us

with Clinical Studies in Pediatric Oncology,

Paul Meltzer, National Human Genome Research

Institute, NIH 198

Advantages and Limitations of Cell Culture Models

in Pediatric Drug Developments,

Peter Adamson, M.D., Children's Hospital

of Philadelphia 210

Human Cell-Animal Xenografts: The Current Status,

Potential and Limits of Informing Us About

Clinical Studies, Peter Houghton, Ph.D., St.

Jude Children's Research Hospital 229

An Integrated and Comparative Approach to

Preclinical/Clinical Drug Development,

Chand Khanna, DVM, Ph.D.,

Tumor and Metastasis Biology Section, NIH 248

What Can be Learned About Safety,

Kenneth Hastings, Ph.D., CDER, FDA 263

Assessing Anti-Tumor Activity in Nonclinical Models

of Childhood Cancer, Malcolm Smith, M.D., Ph.D.,

Treatment Evaluation Program, National Cancer

Institutes, NIH 280

Committee Discussion 294

Questions for Discussion 306

1 P R O C E E D I N G S

2 Call to Order

3 DR. SANTANA: Good morning to everyone. I

4 know Dr. Kodish is on the line so good morning to

5 you too, Eric. I hope you can hear us well.

6 DR. KODISH: Good morning, Victor.

7 DR. SANTANA: This is a meeting of the

8 Pediatric Oncology Subcommittee of the Oncology

9 Drugs Advisory Committee and we are here today to

10 advise the agency on two issues. In the morning we

11 will deal with the issue of safety monitoring in

12 clinical studies enrolling pediatric oncology

13 patients. Then, in the afternoon we will address

14 issues related to the use of nonclinical data to

15 complement clinical data for proposed pediatric

16 oncology studies. So, we have quite a busy agenda

17 and I think we will go ahead and get started with

18 the introductions, and I am feeling so sorry for

19 Dr. Anderson who is sitting all by himself over

20 there, but we will go ahead and get started with

21 him and then move around.

22 Introductions

23 DR. ANDERSON: Barry Anderson, from NCI

24 CTEP.

25 DR. GRILLO-LOPEZ: Antonio Grillo-Lopez,

1 Neoplastic and Autoimmune Diseases Disorders

2 Research Institute.

3 DR. WEINER: I am Susan Weiner, from The

4 Children's Cause, a patient advocate.

5 MS. HOFFMAN: Ruth Hoffman, patient

6 advocate.

7 DR. PRZEPIORKA: Donna Przepiorka,

8 University of Tennessee, Memphis.

9 MS. CLIFFORD: Johanna Clifford, executive

10 secretary to this meeting.

11 DR. SANTANA: Victor Santana, pediatric

12 oncologist at St. Jude Children's Research

13 Hospital, Memphis, Tennessee.

14 DR. REYNOLDS: Dr. Reynolds, Children's

15 Hospital of Los Angeles.

16 MS. ETTINGER: Alice Ettinger, pediatric

17 nurse practitioner, St. Peter's University Hospital

18 in New Jersey.

19 DR. PAZDUR: This is Susan Ellenberg, who

20 has laryngitis. She is a statistician. I am

21 Richard Pazdur.

22 DR. HIRSCHFELD: Steven Hirschfeld, FDA.

23 DR. DINNDORF: Patricia Dinndorf, FDA.

24 DR. DAGHER: Ramzi Dagher, FDA.

25 DR. SANTANA: Eric, will you go ahead and

1 announce your name and affiliation for the record?

2 DR. KODISH: I am Eric Kodish, from

3 Cleveland, Ohio, Rainbow Babies & Children's

4 Hospital.

5 DR. SANTANA: Thank you, Eric. With that,

6 we will go ahead and have Ms. Clifford read us the

7 conflict of interest statement.

8 Conflict of Interest Statement

9 MS. CLIFFORD: Thank you. The following

10 announcement addresses conflict of interest issues

11 associated with this meeting and is made a part of

12 the record to preclude even the appearance of such

13 at this meeting.

14 Based on the agenda, it has been

15 determined that the topics of today's meeting are

16 issues of broad applicability and there are no

17 products being approved at this meeting. Unlike

18 issues before a committee in which a particular

19 product is discussed, issues of broader

20 applicability involve many industrial sponsors and

21 academic institutions.

22 All special government employees have been

23 screened for their financial interests as they may

24 apply to the general topics at hand. To determine

25 if any conflict of interest existed, the agency has

1 reviewed the agenda and all relevant financial

2 interests reported by the meeting participants.

3 The Food and Drug Administration has granted

4 general matters waivers to the special government

5 employees participating in this meeting who require

6 a waiver under Title 18, United States Code,

7 Section 208.

8 A copy of the waiver statements may be

9 obtained by submitting a written request to the

10 agency's Freedom of Information Office, Room 12A-30

11 of the Parklawn Building.

12 Because general topics impact so many

13 entities, it is not prudent to recite all potential

14 conflicts of interest as they apply to each member

15 and consultant and guest speaker. FDA acknowledges

16 that there may be potential conflicts of interest

17 but, because of the general nature of the

18 discussion before the committee, these potential

19 conflicts are mitigated.

20 With respect to FDA's invited industry

21 representative, we would like to disclose that Dr.

22 Antonio Grillo-Lopez is participating in this

23 meeting as an acting industry representative,

24 acting on behalf of regulated industry. Dr.

25 Grillo-Lopez is employed by Neoplastic and

1 Autoimmune Diseases Research.

2 In the event that the discussions involve

3 any other products or firms not already on the

4 agenda for which FDA participants have a financial

5 interest, the participants' involvement and their

6 exclusion will be noted for the record.

7 With respect to all other participants, we

8 ask in the interest of fairness that they address

9 any current or previous financial involvement with

10 any firm whose product they may with to comment

11 upon. Thank you.

12 DR. SANTANA: Thanks, Johanna. Anybody

13 else sitting at the table that wants to disclose

14 anything publicly? No? Dr. Adamson just joined

15 the group. Do you want to introduce yourself,

16 Peter, please?

17 DR. ADAMSON: Peter Adamson, from

18 Children's Hospital of Philadelphia.

19 DR. SANTANA: Thanks, Peter. Peter, do

20 you want to introduce yourself?

21 DR. HOUGHTON: Peter Houghton, St. Jude

22 Children's Research Hospital.

23 DR. SANTANA: With that, I will pass it

24 over to Dr. Pazdur for his opening remarks.

25 Opening Remarks

1 DR. PAZDUR: Well, I would like to

2 disclose something publicly, my disappointment with

3 Victor and Johanna for not mentioning this but the

4 disclosure is happy St. Patrick's Day.

5 [Laughter]

6 As you can see, we in the government have

7 provided you with green folders for the day and,

8 obviously, I am dressed in green but I would like

9 to remind you Pazdur is not an Irish name. The

10 other thing I would like to just emphasize is that

11 Donna and I, as compatriots from Chicago's Polish

12 community, would like to emphasize that St.

13 Patrick's Day is just a warm-up for St. Joseph's

14 Day. Okay?

15 [Laughter]

16 DR. SANTANA: Which is Friday, March 19th.

17 DR. PAZDUR: Thanks for pointing that out.

18 In all seriousness, I would like to go

19 back to why we are here today, and that is for the

20 subcommittee to discuss two important areas today,

21 one in the morning discussing safety monitoring in

22 clinical studies enrolling children with cancer and

23 then, in the afternoon, discussing nonclinical data

24 to complement clinical data for pediatric oncology.

25 We look at these as very important

1 thematic discussions to have. How these areas

2 impact on oncology drug development I think is very

3 important. One thing that I would ask the

4 committee to do specifically is to concentrate

5 really on the pediatric aspect of these. I know

6 that these areas have some tentacles to adult

7 oncology and to other areas of oncology but I would

8 like to remind you that the purpose of this

9 subcommittee is to focus on the pediatric

10 specificity of these issues and special

11 considerations of these broad issues in pediatric

12 oncology.

13 I would like to thank everyone for being

14 here. I asked Steve what number meeting this is

15 and we think it is the eighth. We may be wrong but

16 we are happy that the committee is meeting on a

17 regular basis. We intend to have the committee

18 meet on a regular basis here and to continue this

19 dialogue with the community. So, Steve, I will

20 turn it over to you.

21 Introduction of Issues and Agenda

22 DR. HIRSCHFELD: Thank you. It is

23 customary at the end of remarks to give the

24 acknowledgments but I wanted to give two

25 acknowledgments initially. The first one is to

1 someone who is in the room right now and I am

2 looking at her, and that is Johanna Clifford who

3 has done I think a marvelous job in helping to

4 organize this meeting, and we have had a number of

5 challenges to overcome along the way, so many

6 challenges that for a period of time we thought we

7 were working under a curse, but Johanna has been

8 steadfast, good humored, competent, rapid in her

9 responses and has been I think a driving force in

10 terms of having the meeting occur as it is and as

11 well organized as it is today. So, thank you,

12 Johanna.

13 I would also like to acknowledge someone

14 who is in this room, although not physically, but

15 someone who has had enormous influence on our

16 thinking and on our policies toward patients

17 enrollment in studies and in particular children

18 enrolling in studies, and that is Bonnie Lee who

19 has been with the FDA for many years and was

20 associated with the initial hearings of the

21 committee, which was mandated by Congress in the

22 1970s, to examine the role of children in clinical

23 research. Bonnie has been a particular guide and

24 inspiration for me and also a source of information

25 and direction, which I think has been an asset not

1 only to the agency but to the country and to all

2 patients. And, I wanted to dedicate the discussion

3 this morning in her honor. So, thank you, Bonnie.

4 As Dr. Pazdur pointed out, we are going to

5 be discussing the themes of safety and

6 extrapolation. Clinical research, which we have

7 discussed in some detail in this forum over several

8 of the meetings, has been recorded for at least

9 2,400 years. Children were often the first

10 patients for new procedures and interventions.

11 Part of this evolved from the concept that children

12 were the property of parents so it was rather easy

13 for parents to donate their children for whatever

14 questions might be asked. But along the way there

15 were some founding principles because,

16 unfortunately, children have also been the victims

17 of clinical research.

18 The founding principles of modern Food and

19 Drug Administration regulation were, in large part,

20 established for the purpose of protecting children

21 and, yet, pediatric therapeutic development has

22 never been as thorough and robust as adult

23 therapeutic development, and most of the people in

24 this room have been part of that process and

25 witness to these inequities. Many therapies are

1 administered to children without adequate studies

2 and, furthermore, many therapies are not made

3 available for pediatric study until after adult

4 marketing studies are completed and this is

5 particularly true in oncology. So, we have been

6 working to overcome some of these barriers and

7 challenges. And, the challenges are to assemble

8 sufficient data to establish efficacy and safety in

9 the relevant population. The relevant population

10 may be sufficiently rare that confirmatory studies

11 are not feasible, which is particularly the case

12 for many of the childhood malignancies.

13 There are concerns regarding the

14 implications of adverse events in children and this

15 has been a barrier to the further clinical

16 development of some products because of these

17 concerns. It is also important that there is the

18 establishment and maintenance of a framework that

19 would support systematic clinical investigations

20 for the relevant population. This has been the

21 case historically in pediatric oncology but that

22 framework has always been challenged and is always

23 competing with other priorities. So, it is

24 incumbent on us to make sure that that pediatric

25 research framework has the best resources, and the

1 best advice, and the best support, and the best

2 regulatory environment to do its job.

3 The particular issues regarding the safety

4 monitoring in pediatric oncology clinical

5 investigations are an acknowledgment that children

6 require special protections. Yet, on the other

7 hand, there is also an acknowledgment that risk

8 tolerance is higher in oncology therapeutics than

9 in other therapeutic areas. This sets up a

10 potential tension. Furthermore, there are no

11 detailed consensus standards on study monitoring

12 despite numerous international documents describing

13 what could be termed good clinical practice. We

14 will examine those in some detail during the course

15 of the morning. So, the charge to the committee is

16 to suggest ways to incorporate the fundamental

17 ethical and scientific principles in protecting

18 patients enrolled in clinical studies for pediatric

19 malignancies while providing clear guidance and

20 minimizing the resource burden.

21 We have a series of questions directed

22 toward the committee to help focus the discussion.

23 These are questions which are meant to stimulate

24 what we hope will be an informative exchange and do

25 not have a yes/no or a definitive answer.

1 The first questions revolves around the

2 principles, what are the principles that should be

3 addressed in safety monitoring of clinical studies

4 that enroll children with cancer? Dr. Kodish is

5 going to provide us with some background on that

6 particular topic. If the principles are adequately

7 stated in existing documents. statutes or

8 regulations, please identify the relevant documents

9 and sections.

10 The second set of questions deals with the

11 practice. Recognizing that particular populations,

12 disease settings and products may have specific

13 requirements, what general parameters should be

14 monitored for safety in all clinical studies? Or,

15 to rephrase that, what should the default position

16 be for safety monitoring?

17 Based on the response to the previous

18 question, how often should these parameters be

19 monitored? Again, just giving a framework or

20 guidelines.

21 Based on the responses to the previous

22 questions, who should do the monitoring? Is it

23 adequate to have the personnel involved in the

24 study be responsible for safety monitoring? When

25 we discuss this in detail we may parse this out

1 into the type of study, whether it is early

2 development or later development or the type of

3 disease or other risk factors.

4 What circumstances would benefit from a

5 data monitoring committee? And, are there

6 additional recommendations for safety monitoring?

7 The afternoon will be devoted to a

8 question which can be traced back to the principle

9 of extrapolation. Extrapolation has been a topic

10 of interest within the Food and Drug Administration

11 for many years. In recent years there has been an

12 FDA working group on pediatric extrapolation that

13 has identified four domains that may provide a

14 basis for extrapolation of adult data to the

15 pediatric population. These are nonclinical data,

16 pathophysiology, natural history of the disease or

17 condition, and response to therapy.

18 When our group, noted at the bottom of the

19 slide and some of the members are present here in

20 the audience, asked ourselves the question how can

21 we use nonclinical data to inform us about

22 pediatric clinical studies, and in particular

23 pediatric studies in clinical oncology, we realized

24 we needed further background and further discussion

25 before we could have an informed approach to it.

1 We recognize that the absence of

2 predictive or explanatory nonclinical models in

3 pediatric oncology is today's status quo. We know

4 that safety prediction based on animal studies is

5 estimated at approximately 65-70 percent for

6 cytotoxic compounds and it is unknown for other

7 classes of compounds, particularly the new biologic

8 therapies, gene therapies, immunotherapy, and

9 cellular-based therapies. Efficacy prediction is

10 unknown but low at best. The findings in clinical

11 studies, particularly negative studies, often

12 remain unexplained.

13 Therefore, further clinical studies that

14 entail resources and risks are undertaken to

15 further the field, and we are posing the paradigm

16 is there a mechanism by which we can use

17 nonclinical data to inform us and improve the

18 clinical research in pediatric oncology. There are

19 potential advantages of using the nonclinical data:

20 a lesser resource burden; the ability to answer

21 questions not amenable to available clinical

22 techniques. There might be ethical or, in fact,

23 legal considerations involved too; possibly a

24 faster time frame to generate data; a dynamic

25 interaction between clinical and nonclinical

1 findings that can enhance understanding and

2 confidence in results. When we only have a

3 sufficient population to do one definitive study,

4 and that study takes three to five years and it is

5 not feasible to do a confirmatory study, having

6 confidence in those results is critical. The

7 avoidance of non-informative and minimization of

8 negative outcome studies could be another outgrowth

9 and an opportunity for new study designs.

10 So, the charge to the committee for this

11 afternoon is to provide advice on what types of

12 nonclinical data are considered informative to

13 complement or supplement clinical results. What

14 should the characteristics or properties of

15 nonclinical models and data be to effectively add

16 to the clinical results?

17 If there are no satisfactory models that

18 exist currently, and we will hear some discussion

19 on approaches, what characteristics should a

20 nonclinical model have to confirm, extend or

21 substitute for clinical results?

22 Lastly, is there a set of postulates that

23 can be identified, or should a set be developed to

24 help us make the transition for data extrapolation?

25 So, the questions we are asking are what types of

1 questions that are of potential clinical relevance

2 but are not feasible or acceptable to answer in a

3 clinical study could be addressed by nonclinical

4 studies.

5 Examples may include the need for repeated

6 tissue sampling, always a contentious issue,

7 particularly in children; the assessment of

8 long-term effects of treatment; effects on

9 reproduction; access to critical anatomic

10 structures, and this is a consideration again

11 particularly for some of the pediatric brain

12 tumors; exposure to toxic reagents; evaluation of

13 non-monitorable or irreversible toxicities;

14 identification of biomarkers for clinical

15 monitoring; and many others which I am sure will

16 come up when we have our learned and motivated

17 panel discuss the issue.

18 What type of evidence and data would be

19 recommended in each of the following domains to

20 allow extrapolation from nonclinical data and be

21 informative for a clinical condition? There are

22 listed here a few but there may be others. These

23 include, but are not limited to pharmacology and

24 pharmacokinetics, safety, efficacy, behavior,

25 long-term effects, developmental aspects and others

1 which I am sure will come up.

2 Are there additional recommendations for

3 the effective use of nonclinical data? For

4 example, will open literature reports be generally

5 acceptable? Is documentation of compliance with

6 Good Laboratory Practice necessary to evaluate

7 animal data? Should nonclinical data be submitted

8 as an independent report with a presentation of

9 primary data sufficient for verification and

10 review? These are all practical questions and we

11 are looking for specific advice.

12 So, with this charge and these questions

13 before you, I would like to thank all the committee

14 members and our speakers and guests, and everyone

15 who has shown an interest here for participating in

16 this discussion, and I will turn now the further

17 presentation over to Dr. Eric Kodish, who will

18 discuss the fundamental principles involved in

19 clinical research and some of the issues of

20 enrolling children.

21 Dr. Santana, I think perhaps before we

22 have Dr. Kodish speak--we have some more members of

23 the panel that should be introduced.

24 DR. SANTANA: Yes. Anybody that joined us

25 a little bit late, could you please identify

1 yourself into the microphone by name and

2 affiliation, and any potential conflicts that may

3 have arisen since we started?

4 MS. HAYLOCK: I am Pam Haylock. I am an

5 oncology nurse and I am at the University of Texas

6 Medical Branch, in Galveston.

7 DR. SMITH: I am Malcolm Smith, pediatric

8 oncologist at the Cancer Therapy Evaluation

9 Program, NCI.

10 DR. SANTANA: Dr. Grillo, you had your

11 hand up?

12 DR. GRILLO-LOPEZ: Yes, a point of

13 clarification that I would like to propose to Dr.

14 Hirschfeld. On his first slide on the charge to

15 the committee, which addresses the morning session,

16 you used the phrase "providing clear guidance and

17 minimizing the resource burden" which clearly

18 applies to human resources and financial resources

19 but perhaps doesn't quite stress time. I would

20 suggest that part of your charge to the committee

21 should be that whatever recommendations we propose,

22 and however the FDA understands and decides to

23 apply those recommendations, should not affect the

24 time lines for cancer drug development which today

25 are already intolerably long, and we should be

1 concerned that the cancer patient in general should

2 not be subject to those too long time lines and

3 that anything we do should, in fact, try to reduce

4 the time lines for approval of new therapies.

5 DR. HIRSCHFELD: Thank you for your

6 comments, Dr. Grillo-Lopez. I think you touched on

7 one of the themes which is implied. I personally

8 have always incorporated in the concept resource of

9 time because time is, in fact, probably the most

10 precious resource and, if one looks at biology as a

11 broad spectrum, time is something which evolution

12 and biologic processes look to, to conserve in many

13 ways too. So, I thank you for calling attention to

14 the issue of time, and it is incorporated in that

15 specific charge.

16 DR. SANTANA: One of the philosophic

17 principles of stewardship is that it involves time,

18 people and money resources. So, I think those are

19 all encompassed in your comments.

20 With that, Eric, are you on line now? Can

21 we proceed with you?

22 DR. KODISH: I am on line, Victor.

23 DR. SANTANA: Good. Go ahead, Eric.

24 Protecting Children in Cancer Research:

25 What Really Matters

1 DR. KODISH: Good morning. It is good to

2 be with you virtually, if not physically. I

3 apologize for the inability to get to Washington.

4 We have, hopefully, completed our last big

5 snowstorm of the winter in Cleveland.

6 I am going to be speaking this morning

7 over the telephone and looking at a Webcast of the

8 slides and this is a work in progress so, please,

9 interrupt me if it is not going well and I will

10 switch to my Power Point presentation. I am looking

11 at the Webcast now and I don't see my Power Point

12 slides yet. What I plan to do is ask Johanna to

13 put on the next slide before I move through them.

14 So, let's give it a moment for me to see the first

15 slide.

16 I can introduce the talk by saying that I

17 have always thought I had a face for radio and this

18 is an example of that perhaps--

19 [Laughter]

20 I see my first slide. the title of this

21 presentation is "Protecting Children in Cancer

22 Research: What Really Matters."

23 Can I ask that we have the next slide,

24 please?

25 MS. CLIFFORD: You know what, Dr. Kodish,

1 if you just want to move on through your

2 presentation--

3 DR. KODISH: I have it now. Should I go

4 to the Power Point instead?

5 MS. CLIFFORD: Yes, that would be great.

6 DR. KODISH: All right, the Webcast didn't

7 work well and I will look forward to joining you on

8 the Webcast after I have done my talk.

9 MS. CLIFFORD: Okay, there just seems to

10 be a delay.

11 DR. KODISH: I figured that might happen.

12 The Belmont report I think articulates the key

13 principles of research involving human subjects.

14 My purpose today is to respond to the charge that

15 has been given to the committee and to paint in

16 broad strokes what the key principles are for

17 protection of children involved in cancer research.

18 I think it starts with the Belmont report and the

19 three key principles that are articulated there are

20 beneficence, respect for persons and justice.

21 The next slide, please. This slide shows

22 a concept of principles that move into practice. I

23 thought it was quite appropriate that the charge

24 for the first half of the meeting talked about both

25 principles and practice. I view the regulations

1 and their interpretation as a conduit, as a

2 mechanism by which we move from principles to

3 practice. I want to emphasize the word

4 "interpretation" here. I think that the current

5 set of regulations is subject to wide

6 interpretation, as has been pointed out over and

7 over again in the literature. I don't view this as

8 a negative. I think that it allows for thoughtful

9 IRBs, investigators, parents and others involved in

10 the research process to move from principles to

11 practice in an appropriate manner, and that

12 interpretation is really the key step.

13 The next slide, please. This slide should

14 show a triangle which points out that we are

15 talking today about pediatric research ethics and

16 that this is a more complicated system because of

17 the involvement of a child. The geometry of

18 pediatric research ethics involves parents, on your

19 lower left; the investigator, on your lower right;

20 and the child at the top of the triangle. If we

21 keep the best interests of the child in mind at all

22 points, I think we will be responding to perhaps

23 the most fundamental issue in research involving

24 children.

25 The next slide, please. This slide shows

1 a recapitulation of the Belmont principles with an

2 emphasis on beneficence in pediatric ethics.

3 Respect for persons and justice remain important in

4 pediatric ethics but it is my feeling that there is

5 a special place for beneficence when we are talking

6 about children, whether it is research involving

7 children or in clinical ethics regarding children.

8 In fact, more broadly in social policy regarding

9 children it is important to remember that children

10 are not able to vote; don't have economic

11 resources; and we owe an advocacy role I think on

12 behalf of children. It is very important and, to

13 me, prioritizes that beneficence as a concept for

14 pediatric ethics.

15 Can I have the next slide, please? The

16 principles of medical ethics then are different for

17 children compared with adults. I would say that

18 respect for persons, for good or for bad, has

19 become the dominant principle for adult ethics and

20 this is seen in research ethics where there is a

21 tremendous emphasis on informed consent, and this

22 is out of the derivative concept of autonomy which

23 comes from that principle of respect for persons.

24 By contrast, as I said, I think the best interest

25 of children has to dominate pediatric ethics and

1 justifies an population that takes beneficence as

2 the most important principles.

3 I don't want you to move slides back but,

4 if you recall a few slides ago, the slide that

5 shows moving principles into practice, I think

6 beneficence has to be the principle that drives our

7 interpretation of the regulations and our actual

8 practices.

9 The next slide, please. This slide

10 dissects out some text from the Belmont report.

11 The document itself talks about beneficence as an

12 obligation with two general rules. These are very

13 interesting. It had been sometime since I have

14 looked at them and in preparing for this

15 presentation I found the two general rules cited by

16 Belmont are do not harm and, secondly, maximize

17 possible benefits and minimize possible harms.

18 On the face of it, these two general rules

19 can be read as conflicting with one another. That

20 is, the charge do not harm is an absolute standard,

21 whereas in the second rule of minimizing possible

22 harms and maximizing possible benefits it is a

23 relative standard and it calls for a weighing of

24 benefit against harm. Again, to put interpretation

25 into play, I think it is the second rule that is

1 most appropriate for pediatric oncology studies.

2 That is to say, if one is talking about research

3 involving healthy children with no prospect of

4 benefit to that child, the first rule might be more

5 appropriate to apply, do not harm, period. But we

6 are talking about a balance in pediatric oncology

7 and I think the second general rule is more

8 appropriate.

9 Can I have the next slide, please? If we

10 are on the same page, this slide should continue to

11 cite the Belmont report which says that beneficence

12 is not always so unambiguous and goes on to say

13 that prohibiting research that presents more than

14 minimal risk without the immediate prospect of

15 direct benefit to the children involved limits

16 potential for great benefit to children in the

17 future.

18 This became, in some sense, the foundation

19 for the different categories of research in subpart

20 D that IRBs are able to approve and points out the

21 key ethical dilemma, as far as I am concerned,

22 which has to do with how we weigh benefits or which

23 benefits count when we are weighing risk and

24 benefit.

25 The next slide, please. The subtitle of

1 my talk today is "What Really Matters" and as I

2 thought about a way of presenting this I decided

3 that it could be divided in three phases, what

4 matters before a clinical trial begins; what

5 matters during the conduct of the trial; and what

6 matters after a trial has closed.

7 One of the members of the panel pointed

8 out the importance of time prior to the beginning

9 of my talk, and I guess this is another way of

10 looking at time as a divider for where the

11 different ethical obligations come in.

12 Speaking of time, I wanted to get some

13 validation from Johanna. Is the timing going

14 better now with the slides?

15 MS. CLIFFORD: It is fine, Dr. Kodish.

16 DR. KODISH: Going fine? Great! So, I

17 would like to now talk about what matters before a

18 trial begins and I could think of at least three

19 important issues. The first is that it be

20 significant science. Again, interpretation is a

21 key here. My view of significant science is that

22 it has the potential to help children with cancer.

23 I think it is important that I am very specific

24 about that. I think that if there are going to be

25 exposures of risk to children with cancer the

1 potential to help children with cystic fibrosis,

2 for example, may not be considered significant

3 science by this test. The potential to help adults

4 with Alzheimer's disease may not be significant

5 science by this test.

6 I think that we need to be cognizant of

7 the fact that research involving children with

8 cancer needs to resound back to help children with

9 cancer and that one should look for other avenues

10 to study other important diseases. It is difficult

11 to think of children with cancer as a resource, but

12 I think in some sense this really forces us to do

13 that and, by limiting the risk of exposure to

14 children to that which will come back to help

15 children--and I know that scientifically it is

16 often very difficult to predict in which direction

17 the work will go and how the results will, in fact,

18 play--ut but at the outset one can try to predict

19 and think about a definition of significant as

20 being that which has the potential to help children

21 with cancer.

22 The second thing that really matters

23 before a clinical trial begins is a risk/benefit

24 assessment. I think in the next several slides I

25 will talk more about what counts as risk and what

1 counts as benefit.

2 Finally, it is a study design that will

3 answer the question and that also does not

4 subjugate the interests of any single subject to

5 the overall needs of the research. Again, embedded

6 there are a couple of important ethical principles

7 that I think are perhaps specific--at least the

8 second one under study design--specific to research

9 with a vulnerable population and, as Dr. Hirschfeld

10 said, children certainly are considered and should

11 be considered.

12 The next slide, please. This slide shows

13 the criteria for the 405 category. As I think

14 everybody is aware, there are four categories of

15 research that can be approved by IRBs under subpart

16 D. Almost all cancer research I think is approved

17 under 405, that is, pediatric cancer research. It

18 is research that involves more than minimal risk

19 but presents the prospect of direct benefit to the

20 individual subject if the risk is justified by the

21 anticipated benefit to the subject; if the

22 risk/benefit ratio is less than or equal to the

23 alternatives; and if parental permission and assent

24 are obtained.

25 The next slide, please. As we weigh risk

1 and benefit in research ethics, it is important to

2 remember that risk means risk to the subject but

3 benefit may include benefits to the subject,

4 benefits to other patients, benefits to society or

5 benefits to an investigator or a sponsor. I think

6 what we are aiming for in research involving

7 children in some sense is limiting the benefits

8 that we think about in a risk/benefit analysis so

9 that the benefits that come to the subject are the

10 ones that we are thinking about as we weigh risk to

11 the subject, and that we avoid a situation where

12 children are used as a means to an end. To go back

13 to Emmanuel Kant and the idea that children are

14 valued and protected, I think it is inherent in

15 this sort of balancing.

16 The next slide, please. This is a slide

17 that looks at some of the issues in early drug

18 development involving children with cancer. There

19 has been a controversy over, what I have put in

20 quotes here, therapeutic intent. The point here is

21 that the prospect of direct benefit is the key

22 ethical and regulatory issue and, in my view, a

23 percentage view of what that potential for

24 therapeutic intent might be isn't that important.

25 That is, I think even a very low chance of

1 therapeutic benefit for the child should count as a

2 prospect of direct benefit to the child. Again, my

3 interpretation of the word prospect is a very broad

4 one, admittedly, but this is where the issue of

5 interpretation comes in. As the discussion goes

6 on, we can talk about how prospect ought to be

7 interpreted.

8 The second bullet point you see on this

9 slide has, in parentheses, the potential for 405

10 creep, that is, moving this issue of commensurate

11 experience that children with cancer have already

12 been through a lot so that it is okay to put them

13 through one more thing. This doesn't stand up in

14 my view as a valid justification for exposing

15 children with cancer to risk.

16 The alternatives is another key issue that

17 is discussed, if you recall, in the 405 criteria.

18 There needs to be favorable outcome for the child

19 compared to the alternatives.

20 The next slide, please. If we are on the

21 same page, this should be a slide that says options

22 on top. It has at least three different pathways

23 that families and children can seek out when a

24 child has refractory, untreatable cancer. On your

25 left is a Phase I study; in the middle is

1 alternative medicine and on the right is hospice

2 philosophy care.

3 The next slide shows further

4 considerations regarding Phase I oncology research

5 in children. The first is to point out that

6 subject selection is not a major controversy in

7 this realm, that is to say, Phase I studies are

8 done involving healthy children but it is not an

9 issue of wanting to do Phase I cancer research on

10 healthy children. That, to my knowledge, is not a

11 controversy but I put it here because it is

12 important to try to contextualize pediatric cancer

13 research in the broader picture of research

14 involving children. As I said before, I think that

15 Phase I research qualifies, in my mind, as research

16 with the prospect or direct benefit.

17 Most importantly on this slide, is that

18 potential for benefit mitigates but does not

19 eliminate the need for protection from research

20 risk. To be more clear about that, it is the

21 potential for benefit that is balanced against the

22 risk that mitigates it, but I think the charge to

23 the committee and the work we are going to do this

24 morning is still extremely important. The need for

25 protection from research risk is not eliminated by

1 the potential for benefit.

2 The next slide, please. This points out

3 some issues around alternative medicine. The

4 reason that I put this here is that I think there

5 is a yardstick of fairness that we need to keep in

6 mind. It is often the case that when research is

7 being done it is held to a higher standard or a

8 different standard than what is happening in the

9 non-research world, and it is very important I

10 think to the families and the children involved

11 that we try to put this in the lens that they are

12 viewing this off from, and to make it difficult to

13 access research or to have children participate in

14 well-designed, safely monitored research, in some

15 ways, runs the risk of shunting them to alternative

16 medicine where there are vulnerability concerns.

17 It is very prevalent phenomena for children with

18 refractory cancer. I think there are major ethical

19 differences when it comes to children getting

20 alternative therapy compared to adults who can make

21 their own decision. I think we have a very

22 important obligation to prevent harm when it comes

23 to children who are getting alternative medicine,

24 and I think it is extremely important that

25 alternative medicine possibilities be studied in a

1 rigorous and careful way. But the bottom line is

2 that we need to communicate with families and

3 children. The ones that the research community

4 encounters may also be taking alternative medicine

5 and if we don't know what medications are being

6 taken, then we won't have the ability to study drug

7 interaction with alternative medications and the

8 experimental agent, for example. I just think that

9 it is very important that we keep alternative

10 medicine in mind as something that is out there and

11 we shouldn't be blind to it.

12 The next slide, please. This slide has a

13 few words about hospice care for children who have

14 refractory disease. Now, some people I think have

15 the experience that those who come to Phase I

16 studies are self-referred, not interested in

17 hospice philosophy care, wanting to continue to

18 pursue anti-neoplastic therapy but, in my

19 experience, that is not the case. In fact, many

20 families who seek Phase I studies also are amenable

21 to having their child get hospice philosophy care.

22 So, the two are not incompatible. I think it is an

23 under-developed approach in children. It is not

24 the main focus of what we are here about today but

25 I felt that it would be incomplete to give this

1 talk without mentioning that hospice philosophy

2 care should be part of the consent process for

3 Phase I studies.

4 The next slide, please. This moves from

5 what really matters before the conduct of the trial

6 to during the conduct of the trial. The three

7 items that really matter during the conduct of the

8 trial are informed consent which, in my view, is a

9 communication process in addition to the

10 documentation that happens; ongoing monitoring via

11 a data safety monitoring board, if appropriate, and

12 I understand that much of the discussion later on

13 will have to do with when it is appropriate and

14 when it is not necessary; and ethical action to

15 suspend or stop a study at the right time. It is

16 easier said than done but in parentheses I thought

17 I would say not too soon but not too late either.

18 So, the question of when a study should be

19 suspended or stopped is a key ethical question that

20 happens during the conduct of a study and whether a

21 study needs to be stopped at all. I guess in most

22 cases there is no need to stop it but that question

23 needs to be always asked in the same way house

24 officers always need to ask themselves does this

25 child need a spinal tap. It is a question that is

1 part of the monitoring process as an embedded

2 function.

3 The next slide shows the Nuremberg code.

4 This is a quick bit about informed consent. The

5 Nuremberg code said that the voluntary consent of

6 the human subject is absolutely essential. These

7 are slides that I have shown at previous meetings

8 so I think we can go fairly quickly through them.

9 The next slide asks the rhetorical

10 question of whether we can do any pediatric

11 research at all, and just points out that if the

12 answer is no, that is, if we have to adhere to

13 strict interpretation of the Nuremberg or literal

14 rather than in the spirit of the law

15 interpretation, children as a group will suffer.

16 You saw in the Belmont quotation earlier that there

17 is a clear recognition that there needs to be some

18 research involving children so that we can both

19 protect children adequately but be sure that we

20 make progress in childhood disease.

21 The next slide talks about three ways of

22 respecting Nuremberg and still doing pediatric

23 research by using parents as surrogates and

24 obtaining parental permission; by involving

25 children when appropriate and obtaining their

1 assent; and by providing societal protection with

2 IRB approval as the most obvious but also meetings,

3 similar to what we are doing this morning,

4 investigator integrity and other things that

5 provide societal protection for children, we can, I

6 think, ethically do pediatric research.

7 The next slide shows the difference

8 between parental permission and informed consent

9 and, again, says that the autonomous authorization

10 of an adult--the difference between adult and

11 pediatric ethics is more robust than a proxy

12 decision and points out, from the Academy of

13 Pediatrics, that the responsibilities of a

14 pediatrician to his or her patient exist

15 independent of parental desires or proxy consent.

16 I think that there is a congruent statement that

17 one could make here that says that an

18 investigator's responsibility to his or her subject

19 exists independent of parental desires or proxy

20 consent.

21 The next slide shows that parental

22 permission is not the oral equivalent of informed

23 consent, and that surrogate decision-making is

24 necessarily less authentic. I am going to skip

25 past the next slide which shows proxy consent,

1 substituted judgment and best interests, because I

2 think this is familiar ground for most people and

3 we have already emphasized best interests.

4 I will go to a slide that says informed

5 consent in pediatrics equals parental permission

6 and the assent of the child. Here I want to say

7 that the combination of those two can potentially

8 be more powerful, if done right, than an

9 individual. This has to do with family centered

10 ethics that really seek to care for and do

11 effective communication with a family, which is a

12 dynamic and challenging process, admittedly. But I

13 think both of these issues are very important.

14 The next slide, please. This provides the

15 regulatory definition of assent, which is a child's

16 affirmative agreement to participate in research.

17 The key point here is that mere failure to object

18 should not be construed as assent. That is, the

19 silence of an older child for research

20 participation can't be interpreted as their assent.

21 Again, there is room for regulatory interpretation

22 here. There is a great deal of controversy around

23 assent and requirements for assent, and I think

24 there is likely to be a fair amount of variability

25 across IRBs with regard to this issue and I would

1 be happy to discuss this further during our

2 discussion.

3 The next slide, please. This slide shows

4 some differences between assent in the clinical and

5 research context, and points out the fact that

6 research is supererogatory, that is, as opposed to

7 a clinical context where there is a strong best

8 interests argument to be made. Generally speaking,

9 in research the decision is more voluntary and, for

10 that reason, assent is more powerful phenomenon, in

11 my view, ethically speaking in research than it

12 would be in the clinical context.

13 The bottom bullet point here is also

14 important I think as a principle perhaps for us to

15 consider, and that is the older the child, the more

16 assent contributes to the ethical justification for

17 the study. This is a problem for diseases that

18 happen in younger children certainly but, all

19 things being equal, an older child I think who can

20 participate in the decision gives us more ethical

21 justification for proceeding in research endeavors.

22 The next slide just points out a piece of

23 data. This is a scale that we did in our study of

24 informed consent about decision-making preference.

25 It shows everything from, number one, a parent who

1 wants to leave all decisions to the doctor and

2 perhaps to an investigator, and then a continuum to

3 number five, a parent who wants to make final

4 selection about which treatment their child will

5 receive.

6 The next slide shows a sample of 108

7 parents. The reason that I included it this

8 morning is to point out the variability among

9 parents and families when it comes to how they want

10 to make decisions. You see in this slide a large

11 number of parents in the middle, within the green,

12 red and grey columns, who fit into a shared

13 decision-making model. In my view, this is why

14 informed consent is important during the conduct of

15 research. Most people want a shared

16 decision-making approach whether it comes to

17 treatment or research participation and

18 communication. Effective communication is really

19 the key issue for informed consent.

20 The next slide. As I wind down the talk

21 and get to the conclusion, I want to make the point

22 that the over-interpretation of regulatory concerns

23 can prevent the ethically meaningful participation

24 of children in research.

25 Can you still hear me?

1 MS. CLIFFORD: We can still hear you.

2 DR. KODISH: Great! I heard a beep on the

3 phone. I am going to tell a quick story to

4 illustrate this point. Heather K was diagnosed

5 with a vaginal rhabdomyosarcoma at a children's

6 hospital in the Midwest within the past few months.

7 At diagnosis, Heather had a tumor that was causing

8 intestinal compression. Her pediatric oncologist

9 talked to the family about the diagnosis and then

10 subsequently discussed a Phase III non-randomized

11 study sponsored by the IRS/COG. The family

12 provided informed consent and signed a document at

13 6:05 p.m. The plan was to begin chemotherapy the

14 following day but the patient developed a bowel

15 obstruction at 11:00 p.m. and chemotherapy was

16 emergently started. At midnight nothing happened

17 that was ethically significant. Clinically, the

18 patient was continuing to get her chemotherapy.

19 But the next morning, when the CRA, the data

20 person, came to enroll Heather in this Phase III

21 study, the RDE, or the remote data entry system,

22 made enrollment impossible. The reason that

23 enrollment was impossible was that the date

24 chemotherapy was started was the previous date and

25 the form would not permit enrollment to happen if

1 chemotherapy had already been started.

2 So, what was a well-intentioned regulation

3 system designed to prevent people from being

4 entered on study if consent had not yet been

5 obtained--in fact, in this case everything went

6 perfectly from an ethical perspective but the

7 patient was not allowed to be entered on study. I

8 think that this is a cautionary tale and I wanted

9 to bring it to the attention of the panel today.

10 Next slide, please. We see many

11 well-intentioned regulatory protections and it is

12 important to realize that they can paradoxically

13 prevent the ethical participation of children in

14 cancer research and Heather's story is one example

15 of that. The physician then needed to go back to

16 the family and explain that, unfortunately, we

17 weren't able to include her as a subject in the

18 research. It wasn't going to change her treatment

19 at all but the future treatment of children with

20 rhabdomyosarcoma in some ways is harmed by the fact

21 that this regulatory mechanism prevented Heather

22 from being a subject in the study. The only

23 alternative would have been for the person doing

24 remote data entry to fabricate and to say that the

25 date chemotherapy was started was the day that she

1 was being entered on study, and that would have,

2 number one, been an unethical lie and, number two,

3 would have been picked up on an audit if the

4 subject had been audited subsequently though it may

5 have been, in fact, the ethical thing to do because

6 consent was obtained in an appropriate way, it is

7 an important study, and all of the things that we

8 have bee talking about, but the regulatory

9 apparatus prevented an ethical action from taking

10 place and I think it is a disturbing story.

11 The next slide shows a synergistic

12 approach. The protection of human subjects has

13 been done both through education and regulation and

14 we need to be concerned about developing too much

15 regulation at the expense of education and the

16 expense of thoughtful ethical action.

17 The next slide just has a few quick points

18 about what matters after a trial is closed.

19 Monitoring for late effects of therapy is an

20 important ethical issue after a trial has closed.

21 The publication of results and dissemination of

22 findings is ethically important. If the science

23 isn't disseminated, then it is like a tree falling

24 in a forest that nobody hears. Finally, the return

25 of results to the subjects who participated is an

1 ethically under-looked and I think very important

2 issue that symbolizes the partnership that we have

3 with subjects and their families, and I think we

4 need to do a better job than we are doing currently

5 after a trial has closed in getting results back to

6 the subjects.

7 The next slide shows conceptually the main

8 balance as a point of conclusion in pediatric

9 research ethics, that the best interests of the

10 child-subject are, in fact, balanced against

11 science to benefit others and we need to be

12 cognizant of that balance at all times and be sure

13 that the best interests of the child are not

14 subjugated.

15 The next slide shows a couple of

16 conclusions. The first is that beneficence, as

17 described in the Belmont report, is the key ethical

18 principle that I believe should guide monitoring of

19 patients in studies. Also, a risk/benefit

20 assessment by the investigator, by the IRB and by

21 others perhaps is more important than informed

22 consent, and that is because I don't think informed

23 consent has the ethical importance in pediatrics

24 that it does in adult medicine, and also because of

25 the relatively ineffective communication process

1 that is currently happening with informed consent.

2 I would be happy to talk more about that in the

3 discussion.

4 The next slide shows that the protection

5 of children from research risk and the imperative

6 to improve childhood cancer treatment are both

7 ethically important. The bottom point here is that

8 regulatory fervor intended to protect children

9 currently threatens the ethical conduct of

10 pediatric cancer research, as I tried to illustrate

11 in Heather's story, and we need to remember, I

12 think, that there is an ethical imperative to do

13 work in childhood cancer to improve the care of

14 children with cancer.

15 The final slide points out that children

16 are both vulnerable subjects who need protection

17 from research risk and a neglected class--and they

18 continue to be a neglected class despite our best

19 efforts--that need better access to the benefits of

20 research.

21 I thank you all for tolerating the virtual

22 reality nature of this talk and hope that I have

23 been able to make a contribution. Thank you.

24 DR. SANTANA: Thanks, Eric. Eric, are you

25 planning to stay on line for the rest of the

1 morning?

2 DR. KODISH: I am. The only question is

3 whether I should do it by phone or by Webcast.

4 DR. SANTANA: Okay, because if you are

5 going to stay, then we will just hold the questions

6 for the general discussion, if that is okay with

7 you.

8 DR. KODISH: That is fine.

9 DR. SANTANA: But I do want you to stay on

10 the phone line, if at all possible, for the

11 discussion because I think we can communicate

12 better that way.

13 DR. KODISH: Okay, what I will try to do

14 is watch but mute the sound.

15 DR. SANTANA: That is fine.

16 DR. KODISH: Thank you, Victor.

17 DR. SANTANA: Okay, good. I also want to

18 thank John for advancing your slides on your

19 behalf. Dr. Carome, you are next.

20 Legal Responsibilities for HHS Supported Studies

21 DR. CAROME: Good morning. I would like

22 to thank the subcommittee members for inviting me

23 to give a brief presentation on legal

24 responsibilities for studies conducted and

25 supported I think originally by the federal

1 government and since I speak on behalf of HHS, I

2 have limited it to HHS, the Department of Health

3 and Human Services.

4 What I am quickly going to do is go over,

5 first of all, the applicability of our regulations.

6 Then I am going to talk very quickly about the

7 major requirements of 45 CFR Part 46, Subpart A,

8 which are the general protections for human subject

9 research. Then I am going to finish up by talking

10 about the major requirements of 45 CFR, part 46,

11 Subpart D, which are the additional protections for

12 children involved as subjects in research.

13 Again, the regulations I am referencing,

14 45 CFR Part 46, are the HHS regulations for the

15 protection of human subjects. They have four

16 subparts. The regulations were last revised in

17 2001. One of the subparts, Subpart B, was revised

18 at that point but most of the regulations remain

19 the same as when they were promulgated more than

20 two decades ago.

21 So, what is the applicability of these

22 regulations? Our regulations apply in two

23 circumstances. The most common is research

24 conducted or supported by the Department that are

25 not otherwise exempt. That includes clinical

1 trials conducted intramurally by the NIH or funded

2 by the NIH, as well as many other agencies within

3 the Ddpartment. A second way in which research can

4 be covered by these regulations is research that is

5 conducted at an institution holding an applicable

6 assurance of compliance approved by our office.

7 So, any institution that receives funding from our

8 Department to conduct human subject research must

9 execute a written agreement in which the

10 institution pledges to comply with our regulations,

11 and in that document many institutions voluntarily

12 extend the same regulations to all research

13 regardless of sponsorship. In doing so, the

14 assurance comes to cover privately sponsored

15 research.

16 This slide demonstrates the relationship

17 and the overlap between the applicability of our

18 regulations and the FDA regulations. You can see

19 that there is in the middle an overlap. The

20 overlap may occur in two circumstances. One is

21 where NIH sponsors a clinical trial or other

22 clinical research, or any research, that involves

23 an FDA-regulated test article. Another

24 circumstance is where an institution, holding an

25 assurance with our office in which they voluntarily

1 agreed to extend that assurance to all research, is

2 engaged in an industry, privately sponsored

3 research, project involving an FDA-regulated test

4 article.

5 Very quickly, what are the major

6 provisions of Subpart A? As was previously noted,

7 the regulations, we believe, are clearly founded

8 upon an ethical framework that was articulated in

9 the Belmont report. Its three basic ethical

10 principles, and the fundamental provisions of the

11 regulations can be divided in three groups. One is

12 the provisions related to and assurance of

13 compliance. The second is those related to the IRB

14 requirements, institutional review boards, and the

15 third is those requirements related to legally

16 effective informed consent.

17 With respect to assurances, the

18 regulations stipulate that each institution engaged

19 in research covered by the regulations and which is

20 conducted or supported by the Department shall

21 provide assurance satisfactory to the HHS Secretary

22 that it will comply with the requirements set forth

23 in the regulations.

24 The regulations further stipulate specific

25 elements that must be part of an assurance. There

1 must be a statement of principles governing the

2 institution in the discharge of its

3 responsibilities for protecting the rights and

4 welfare of human subjects. And, the regulations

5 state that those principles must apply to all

6 research regardless of whether or not it is covered

7 by the assurance.

8 The assurance must designate at least one,

9 and many institutions designate more than one,

10 institutional review board and that must include a

11 list of the IRB members and their relative

12 capacities, and there must be a reference to

13 written IRB procedures. There are requirements

14 related to the IRB and they include specification

15 of what the IRB membership must include, such as at

16 least one person whose primary interests are in the

17 scientific area and at least one member whose

18 primary interests are in a non-scientific area, and

19 at least one member who is not otherwise affiliated

20 with the institution or a member of a family

21 affiliated with the institution.

22 The regulations have specific provisions

23 related to how the IRB should function and operate;

24 when it must conduct review in terms of initial and

25 continuing review. Then there are provisions

1 related to expedited review for certain categories

2 of minimal risk research and there are detailed

3 lists of specific criteria an IRB must find in

4 order to approve research. For example, the

5 regulations state that in order to approve research

6 an IRB must find that the risks to the subjects are

7 minimized and reasonable in relationship to the

8 anticipated benefits, if any, to the subjects and

9 the knowledge that is to be gained. Then, there

10 are other provisions for the records that an IRB

11 must maintain.

12 The last set or provisions in Subpart A

13 deal with legally effective informed consent. They

14 include an introductory paragraph that talks about

15 the general requirements. For instance, no

16 investigator may involve a human subject in

17 research unless the informed consent of the subject

18 or a legally authorized representative of the

19 subject has been obtained, except in certain

20 limited circumstances in which informed consent can

21 be waived.

22 The regulations go on to stipulate basic

23 elements that I think most people are familiar

24 with: the nature of the research; the reasonably

25 foreseeable risks; the reasonably foreseeable

1 benefits, if any, to the subject; and others, such

2 as alternatives that a subject may choose instead

3 of entering the research. The regulations

4 stipulate that consent must generally be

5 documented, except in some limited circumstances.

6 Then, there are waiver provisions both for

7 obtaining informed consent at all or for documented

8 informed consent, and I won't go into those in

9 detail.

10 Let's turn finally to the provisions for

11 research involving children under Subpart D, the

12 additional protections for children. Again, this

13 is a subpart that is unique to the Department of

14 Health and Human Services. Whereas all the Subpart

15 A provisions that I just went over have been

16 adopted by other departments and agencies, Subpart

17 D has only been adopted by the Department of

18 Education in addition to our department.

19 Subpart D applies to all research

20 involving children as subjects conducted or

21 supported by our department. It is important to

22 note that there is a specific definition of

23 children in the regulations, and they are persons

24 who have not attained the legal age for consent to

25 treatments or procedures involved in the research

1 under the applicable law of the jurisdiction in

2 which the research will be conducted. It is

3 important to note that in order to then understand

4 who a child is with respect to the research

5 regulations, you must understand state and local

6 law that defines who can consent to what and at

7 what age. Therefore, a child in one state might

8 not be a child in another state for the purposes of

9 these regulations.

10 The Subpart D requirements in

11 general--first of all, you have to satisfy all the

12 requirements of Subpart A. So, if a research

13 project involving children doesn't satisfy some

14 provision of Subpart A, then it is moot about the

15 additional provisions. The research would not be

16 approvable. But if the research is approvable

17 under Subpart A, there are additional requirements

18 of Subpart D which must be fulfilled and satisfied.

19 As Eric referenced, there are four

20 categories of research that are approvable under

21 Subpart D under our regulations. These are

22 primarily scaled to risk versus benefit as you walk

23 through each of these categories, and I am going to

24 do that very quickly.

25 The first category, 404, is research not

1 involving greater than minimal risk, and minimal

2 risk is defined in Subpart A. In order for this

3 research to be approved under this category, an IRB

4 must make one general finding. It must find that

5 there are adequate provisions for soliciting the

6 assent of the child and permission of the parents

7 or guardians, as set forth in Section 408.

8 The next category, Section 405, which Eric

9 went into more detail, is research involving

10 greater than minimal risk but presenting the

11 prospect of direct benefit to the individual

12 subjects. So, the benefit has to be tied to the

13 subjects as opposed to society in general and the

14 knowledge to be gained. Here, the IRB must make

15 three specific findings. The IRB must find that

16 the risk is justified by the anticipated benefits

17 to the subject; the relationship of the anticipated

18 benefit to the risk is at least as favorable to the

19 subjects as that presented by available

20 alternatives outside the research context; and,

21 again, the same provisions for assent and

22 permission apply throughout these four categories.

23 The next category, 406, involves greater

24 than minimal risk and no prospect of direct benefit

25 to the individual subjects, but likely to yield

1 generalizable knowledge about the subject's

2 disorder or condition. For this category there are

3 four criteria that an IRB must find. They must

4 find that, first, that the risk represents a minor

5 increase over minimal risk. Whereas minimal risk

6 is defined in the regulations, what a minor

7 increase means is not defined so that is left up to

8 the judgment of the IRBs.

9 Next, the IRB must find that the

10 intervention or procedure within the research

11 presents experiences to the subjects that are

12 reasonably commensurate with those inherent in the

13 actual or expected medical, dental, psychological,

14 social or educational situation of the child.

15 Commensurability is one of the factors that Eric

16 touched on but applies only in this category, 406.

17 The next two provisions--the IRB must find

18 under 406 that the intervention or procedure is

19 likely to yield generalizable knowledge about the

20 subject's disorder or condition which is of vital

21 importance for the understanding or amelioration of

22 the subject's disorder or condition. I think the

23 key words here are that you have to understand that

24 the child must have a disorder or condition, two

25 terms that are not otherwise defined in the

1 regulation and are of vital importance. So, it is

2 sort of a higher standard than the usual

3 generalizable knowledge standard that probably

4 applies to research under Subpart A only. Lastly

5 is the assent or permission provisions.

6 The fourth category and final category is

7 research that is not otherwise approvable under one

8 of these four categories which presents a

9 reasonable opportunity to understand, prevent or

10 alleviate a serious problem affecting the health or

11 welfare of children. For this, the IRB still must

12 review and assess the research with respect to

13 Subpart A and D, and must find that the research

14 presents a reasonable opportunity to further the

15 understanding, prevention or alleviation of a

16 serious health problem affecting the health or

17 welfare of children.

18 The project is then forwarded to the

19 Department. They come through our office and we

20 act on behalf of the Secretary to process these.

21 In order for the research then to be approved, the

22 Secretary, after consultation with a panel of

23 experts in pertinent disciplines and following an

24 opportunity for public review and comment, must

25 determine either that the research in fact

1 satisfies one of the other three categories, 404,

2 405 or 406 or, if not, three things must be met:

3 that research presents a reasonable opportunity

4 standard that I previously went over; that the

5 research will be conducted in accordance with sound

6 ethical principles, and hopefully that is something

7 that applies to all research conducted; and

8 adequate provisions for the assent of the child and

9 parental permission.

10 Finally, there are some additional

11 provisions of Subpart D that are provisions related

12 to soliciting assent, and assent is not always

13 required and an IRB may determine it is not

14 warranted, particularly under category 405. There

15 are provisions for soliciting permission of

16 parents, and the regulations speak to whether you

17 need both parents' permission. If the category is

18 405 one parent's permission is sufficient but for

19 406 or 407 two parents are required, except in very

20 limited circumstances.

21 It is important to note that there are

22 provisions for waiving parental permission or

23 guardian permission. Just like informed consent

24 can be waived under Subpart A for research

25 involving adults, parental permission can be waived

1 in certain circumstances and this is I think unique

2 to our regulations and not found in the parallel

3 regulations within the FDA.

4 Finally, there are specific protections

5 for subjects who are wards of the state or any

6 other agency, institution or entity for research

7 approved under 406 or 407. Among those

8 requirements, there must be a specific advocate

9 appointed for each child who is participating in

10 such research who is a ward.

11 In summary, I have quickly tried to go

12 over the applicability of our regulations and

13 contrasted that with the FDA regulations

14 applicability. I have gone over the major

15 requirements of Subpart A of our regulations and

16 finished up with a discussion of Subpart D, and I

17 thank you for your attention.

18 DR. SANTANA: Thanks, Dr. Carome. Dr.

19 Hirschfeld?

20 Legal Responsibilities for Studies with

21 FDA Regulated Products

22 DR. HIRSCHFELD: I would also like to

23 thank Dr. Carome and note that when he was wearing

24 a uniform which was a color more consistent with

25 the theme of the day, he was the head of the IRB at

1 Walter Reed Army Medical Center. I also want to

2 thank him for his efforts on clarification of the

3 regulations in ongoing discussions as they apply to

4 pediatric oncology, and he has taken a leadership

5 role in the Office for Human Research Protection in

6 that regard.

7 I am going to even more quickly, I hope,

8 go through the FDA regulations. One might ask what

9 is a pediatric oncologist doing talking about FDA

10 regulations, but that is one of the strengths of

11 the FDA, that there are wonderful opportunities to

12 be involved in many aspects or research in clinical

13 medicine, including the development of regulations.

14 I was on the working group that developed the

15 Subpart D and, in fact, wrote the first draft of

16 that document.

17 As Dr. Carome pointed out, there is some

18 overlap, and these slides have a lot of data which

19 is intended for reference and I will not go through

20 all the aspects of all the slides, but just to note

21 that there are laws synonymous with an act or

22 statute which are developed and passed by the

23 Legislative Branch and signed by the President and

24 these are published in the United States Code.

25 Then there are regulations synonymous with rule,

1 and these are developed and published by the

2 Executive Branch, the various departments and

3 agencies within the Executive Branch doing the

4 detailed work, and these are published in the Code

5 of Federal Regulations, which is referred to as the

6 CFR.

7 The FDA authority is derived from multiple

8 laws and regulations, and the focus is on product

9 and product use. There are a number of applicable

10 regulations for good clinical practice in the

11 research setting, and these include the human

12 subject protection, which is in 21 CFR, Part 50;

13 financial disclosures, which is in Part 54;

14 institutional review boards, which is in Part 56;

15 and investigational new drugs, which is in part

16 312.

17 Part 50 has actually three sections to it.

18 One is reserved for future use and Part D, you will

19 notice, is the additional safeguards for children

20 in clinical investigations, which is the focus of

21 the discussion now.

22 This is a catalog of all the various sections

23 within Subpart D of 21 CFR, 50. You will see that

24 there is mapping and harmonization between the

25 relevant sections of the HHS regulations.

1 Now, the relationship--and this is just a

2 textual representation of the schematic that Dr.

3 Carome presented--is that FDA regulations apply to

4 all research using FDA-regulated products. In

5 contrast, the HHS regulations apply to all research

6 that is supported by HHS. Research that is

7 supported by HHS using FDA-regulated products is

8 subject to both sets of regulations, and the

9 regulations are harmonized although there are some

10 differences which Dr. Carome elaborated on earlier.

11 The definitions, you will see, parallel those

12 definitions in the HHS regulations and put the onus

13 of interpretation on the local jurisdiction and on

14 the local IRBs, and that is the theme that persists

15 throughout these regulations. So, these

16 definitions are included here to show that there is

17 harmonization and in some cases, we believe, some

18 clarification because the scope of FDA-regulated

19 research is, in many ways, different and can apply

20 to domains where HHS research is not applicable.

21 So, it was important to have not only clarity on

22 the definitions but consistency and, therefore,

23 there are definitions that are included here so

24 that there is not, we hope, much ambiguity in terms

25 of how to apply and interpret these regulations at

1 the local IRB level.

2 Here, again, there is an emphasis on the

3 concept that Eric Kodish developed for us a little

4 earlier this morning, and that is children do not

5 actually engage in a consent process. Their

6 parents provide permission for them to participate

7 in the research. Then, there is the same emphasis

8 as in the HHS regulations that the child must at

9 least be approached for assent.

10 So, in addition to the other

11 responsibilities assigned to IRBs, the FDA

12 regulations ask that the IRB review clinical

13 investigations involving children as subjects

14 covered by Subpart D and approve only clinical

15 investigations that satisfy the criteria which are

16 described in Subpart 51, 52, 53 and the conditions

17 of all other applicable sections of Subpart D.

18 These are again mapped to the four risk

19 categories which were developed in the 1970s and

20 which, because of their serviceability and their

21 flexibility, have been maintained to this date.

22 These, again, discuss the concept of minimal risk

23 here with specific examples of how it applies to

24 pediatric research.

25 Since the IRBs are a conduit through which

1 research occurs, there are specific instructions on

2 when IRBs may approve clinical investigations, and

3 these are divided into the specific risk

4 categories. So, there is greater than minimal risk

5 under 50.51. In 50.52 there is greater than

6 minimal risk presenting the prospect of direct

7 benefit and the conditions, again, are analogous to

8 the HHS regulations; and 50.53 shows that the IRBs

9 can approve clinical investigations involving

10 greater than minimal risk and no prospect of direct

11 benefit but likely to yield generalizable knowledge

12 about the subject's disorder or condition, and the

13 same caveats about having a disorder or condition

14 and having the prospect of generalizable knowledge

15 apply, and these are addressed in some detail.

16 In addition, there are IRB approval

17 criteria which are explicitly stated and these

18 include not only minimization of risk and that the

19 risks are anticipated in relation to the benefit,

20 but that the informed consent process is adequate

21 and appropriately documented and looking for

22 safeguards. That is going to be theme which we are

23 going to look at in detail, what safeguards can be

24 and ought to be implemented.

25 Subpart D addresses this explicitly.

1 There is a paragraph devoted to monitoring which I

2 will quote briefly: While the level of risk in a

3 clinical investigation may change during the course

4 of a study, appropriate strategies may be included

5 in the study design that may mitigate risks. These

6 might include exit strategies in the case of

7 adverse events or a lack of efficacy, or

8 establishing a data monitoring committee to review

9 ongoing data collection and recommend study

10 changes, including stopping a trial on the basis of

11 safety information.

12 Part 56 addresses institutional review

13 boards, and the general provisions and organization

14 are discussed in the first part; IRB functions and

15 operations in the second part; records and

16 reporting in the fourth part; and the

17 administrative actions for non-compliance in the

18 fifth part.

19 Now we come to the IND regulations, 312

20 Subpart A, which are the general provisions which

21 are outlined here.

22 Subpart B, which are in essence the

23 mechanics of an investigational new drug

24 application and the obligations under those

25 sections.

1 Subpart C, which discusses the

2 administrative actions, and Subpart D which goes

3 into detail of the responsibilities of the sponsors

4 and investigators.

5 There is a Subpart E, which doesn't map

6 explicitly to other HHS regulations, which

7 addresses the drugs intended to treat

8 life-threatening and severely debilitating

9 illnesses which apply to pediatric oncology

10 studies. You will notice in the various paragraphs

11 here that in 312.87 there is a requirement for

12 active monitoring of conduct and evaluation of

13 clinical trials. It reads, for drugs covered under

14 this section, the Commissioner and other agency

15 officials will monitor the progress of the conduct

16 and evaluation of clinical trials and be involved

17 in facilitating their appropriate progress. So,

18 this places an FDA role in a dynamic way in the

19 research being conducted in the realm of

20 life-threatening illnesses.

21 In addition, 312.88 has specific

22 safeguards for patient safety which refer back to

23 the other sections that were discussed, Parts 50,

24 56, 312. We didn't discuss 314 which is the NDA

25 regulations and 600 which apply to the biologics

1 but there are analogous regulations in these areas.

2 I will just abstract from here that this

3 includes the requirements for informed consent and

4 institutional review boards, and that these

5 safeguards further include the review of animal

6 studies prior to initial human testing; the

7 monitoring of adverse drug experience through the

8 requirements of IND safety reports; safety update

9 reports for marketing and postmarketing.

10 So, our conclusions from this section are

11 that the FDA has authority to regulate clinical

12 studies using FDA-regulated products; that FDA

13 regulations incorporate both IRB and FDA oversight

14 of studies; that regulations exist for studies

15 using products intended to treat life-threatening

16 illnesses; and that regulations exist for providing

17 additional safeguards for children enrolled in

18 clinical investigations; and, as noted, HHS and FDA

19 regulations are intended to be harmonized. Thank

20 you.

21 DR. SANTANA: Thank you, Dr. Hirschfeld.

22 I think we will hold our questions until we

23 reconvene at the point for discussion. I think we

24 are just a few minutes behind time. We will take a

25 15-minute break--Dr. Hirschfeld wants a 10-minute

1 break. We will take a 10-minute break and try to

2 reconvene at almost 9:45. Thank you.

3 [Brief recess]

4 DR. SANTANA: We will go ahead and get

5 started with the second part of the morning

6 presentations. To initiate that, Dr. Anderson,

7 from CTEP, will be our next speaker. Barry? Eric,

8 are you back on board?

9 DR. KODISH: I am here.

10 DR. SANTANA: Thank you, Eric.

11 Enrollment and Monitoring Procedures for

12 NCI Funded Studies

13 DR. ANDERSON: I am Barry Anderson, from

14 NCI CTEP, and I want to thank the FDA and Steven

15 for inviting us to provide information about the

16 enrollment and monitoring procedures for

17 NCI-supported clinical trials.

18 For pediatric cancer clinical trials, the

19 appropriate enrollment of the individual patient,

20 the child who is going to come onto the trial, as

21 well as the monitoring of that individual patient's

22 experience during the trial and the cumulative

23 experience of all children who are involved in a

24 clinical trial I think are necessary components in

25 terms of trying to enhance the patient safety and

1 the scientific validity of the trial itself.

2 So, at the onset, from NCI's point of

3 view, it is important to work to assure that each

4 child accrued to a trial is receiving the

5 appropriate treatment within the clinical trial

6 itself, and that monitoring that is associated with

7 the trial monitors the toxicity and effectiveness

8 of the treatment intervention within each clinical

9 trial both for that individual child, as well as

10 for the trial overall.

11 The words "safe" and "effective" can be

12 applied to many of the standard treatments we use

13 in pediatric oncology to treat various childhood

14 cancers. These words have special meaning in

15 pediatric oncology. As Dr. Kodish mentioned, there

16 is a special sort of risk/benefit ratio that we

17 always consider because, while therapy for

18 childhood cancer is often successful and that is

19 something that differs from much of medical

20 oncology, the therapies that we use are always

21 toxic in pediatric oncology and they always carry a

22 risk of treatment-related morbidity and perhaps

23 even death in many cases.

24 So, selecting the proper treatment I think

25 is essential because compared with other serious

1 childhood diseases, such as asthma or cystic

2 fibrosis, childhood cancer includes many distinct

3 histologic diagnoses, and each tumor histology

4 requires a distinct treatment appropriate with its

5 own risks and benefits. The chances of cure also

6 diminish quickly if the proper therapy is not used

7 at the outset. That differs, I think, from some of

8 the other more chronic diseases that are serious

9 within childhood diseases but can have chances to

10 change the therapeutic approach over time.

11 In regards to enrollment, a question for

12 the clinical trials done in pediatric oncology is

13 who should be enrolled. Pediatric oncology has

14 evolved an approach of risk stratified treatment

15 regimens and within each tumor histology the

16 patient characteristics and the tumor

17 characteristics establish a risk of relapse. This

18 risk of relapse then is used to stratify the

19 treatment assignment for each child in terms of the

20 type of clinical trial or the specific clinical

21 trial they would be appropriate for. Using this

22 risk of relapse the intensity of the treatment that

23 the child receives--and for intensity you can also

24 say increased toxicity--is then set to best fit the

25 child's cancer. So, it is vital to treat the

1 child, as best we can ascertain at the time they

2 first present, according to the appropriate

3 treatment regimen.

4 By following this treatment stratification

5 approach, the goal in pediatric oncology is to

6 minimize the exposure to highly toxic therapies for

7 those children who don't need that much treatment,

8 in a relative sense, and also for the oncologists

9 to have some comfort in knowing that another child

10 who has a high-risk chance of relapse, that they

11 will in fact potentially benefit from using a more

12 intensive and more toxic treatment regimen.

13 To apply this treatment stratification

14 approach across an entire clinical trial, it is

15 important that the eligibility criteria within the

16 protocol by which all the patients are brought into

17 the trial--that those protocol eligibility criteria

18 are clear in regards to the clinical

19 characteristics of the patient and the pathologic

20 and biologic characteristics of the tumor--that all

21 these characteristics are clear and easy to

22 understand.

23 The pediatric oncologists that are

24 involved in the trial and who would be enrolling

25 patients must be properly informed on how to apply

1 the eligibility criteria that are presented in the

2 eligibility section of the protocol itself. If

3 anyone has ever had experience in trying to bring a

4 patient with rhabdomyosarcoma into a sarcoma trial,

5 it can be a be very complicated endeavor and many

6 mechanisms have been put in place to assist the

7 pediatric oncologist to make sure that the proper

8 decision is made in terms of treatment.

9 As technology has advanced, eligibility

10 criteria have moved beyond what they have been in

11 the past, just being tumor histology and perhaps

12 the staging of the patient. As histologic and

13 biologic characteristics of tumors are better

14 defined and refined, we also are incorporating in

15 many cases in pediatric oncology central input on

16 the pathology and biology, such that central review

17 of the patient's tumor pathology and diagnostic

18 biology assays are used to improve the likelihood

19 that a child receives the best available therapy

20 for their specific tumor pathology and for their

21 risk of relapse.

22 This has been used in a variety of tumors

23 in pediatric oncology in the recent past. With

24 rhabdomyosarcoma there is central review of

25 alveolar versus embryonal rhabdomyosarcoma

1 pathology that is used basically in real time so as

2 to assure that the patient goes on the proper

3 risk-stratified treatment regimen. For

4 neuroblastoma there are a variety of biologic

5 characteristics that make amplification and other

6 genetic changes that are characteristic to each

7 tumor, and that is also looked at in real time.

8 For Wilms tumor there has been a central review of

9 that tumor histology for favorable histology versus

10 focal or diffuse anaplasia that all distinguish

11 patients for their appropriate trial, and there are

12 a variety of genetic studies that are done, both

13 centrally and locally, to establish the appropriate

14 treatment for children with acute lymphoblastic

15 leukemia, the most common diagnosis in childhood

16 cancer.

17 Phase I and pilot studies also have

18 specific eligibility criteria. In these cases, it

19 may not necessarily be the case that you need to be

20 concerned about the tumor histology so much,

21 especially in Phase I where a child has already

22 received treatment, but it is important to ensure

23 that those patients who are enrolled in a trial

24 have no other treatments that provide a reasonable

25 potential for cure or substantial clinical benefit.

1 For patients who have newly diagnosed tumors but

2 have a type of tumor that historically has a poor

3 response to therapeutic interventions, we want to

4 make sure that any sort of pilot treatment

5 interventions that have been tried balance

6 appropriately the benefits and likely risks in the

7 child's prognosis. So, before considering trial

8 monitoring we consider that getting the right

9 patient on the right trial is vital given the

10 stratified approach we have to treatment in

11 pediatric oncology.

12 NCI supports a variety of investigator

13 groups to do clinical trials in children with

14 cancer. The largest is the Children's Oncology

15 Group, which pretty much every pediatric oncologist

16 in North America is a member of. That is the group

17 that does the Phase III studies primarily as well

18 as Phase II studies and pilot studies. There is

19 the COG Phase I Pilot Consortium that is a smaller

20 group, about 20 institutions, that is assigned to

21 do Phase I studies. The Pediatric Brain Tumor

22 Consortium I think is around 10 institutions as

23 well. Their focus is on newer therapies for brain

24 tumors in children. The new approaches to

25 neuroblastoma therapy is a program project grant

1 that NCI supports that is now 12 or 14 institutions

2 I think, focused on early phase studies for

3 children with neuroblastoma, high risk

4 neuroblastoma. There are also individual grants to

5 investigators that may include clinical trial

6 research.

7 All these, because of the nature of

8 pediatric oncology and the relative lack of number

9 of patients, are usually multi-institutional.

10 Given that they are multi-institutional, that

11 brings on special responsibilities in terms of

12 trying to conduct a trial at multiple sites

13 simultaneously and trying to have all the

14 investigators that are enrolling new patients and

15 treating ongoing patients aware of what is going on

16 with the trial. So, the NCI has worked with these

17 various groups that we support to facilitate this

18 sort of intake of information and distribution of

19 information.

20 The investigators that are part of these

21 various groups are committed to report toxicities,

22 the regimen delivery and the ability to deliver the

23 regimen as defined in the protocol and the response

24 data in a timely fashion. Some things such as

25 remote data entry have been put in place now to

1 help facilitate that. There is a data center

2 assigned with each of these groups that we support

3 that is capable of readily receiving the data,

4 analyzing the data and then reporting important

5 data trends to the investigators, be it the study

6 committee and perhaps beyond if necessary. There

7 is an operations office component. They are able

8 to communicate with investigators continuously

9 throughout the clinical trial by email, by web

10 site, by the phone, etc. There is sort of this

11 continuous back and forth going on between the

12 investigators at the local institutions and a more

13 centralized body that is helping to run the trial.

14 In terms of monitoring, again it starts, I

15 think just like enrollment, at the individual child

16 level where there, is within the protocol, guidance

17 provided to the local institutional clinicians as

18 to what sort of laboratory results for

19 tumor-related or treatment-related abnormalities

20 need to be done and at what interval. There are

21 radiologic characterizations of the tumor and the

22 consequent organ dysfunction that are also asked

23 for in terms of the initial diagnosis of the child

24 and then subsequently during their course of

25 treatment. Then there are interval evaluations to

1 establish the tumor response to the treatment

2 interventions that are being conducted during the

3 study.

4 The protocol--and we look for this at NCI

5 when we review the protocols that come to us--must

6 provide sort of a consistent and uniform approach

7 to all these aspects of monitoring of the

8 individual patient. The frequency by which these

9 studies are performed would be consistent with or

10 greater than good clinical practice. Because the

11 children are on a clinical study, oftentimes they

12 get more frequent monitoring of some of these

13 aspects than they would if they received standard

14 of care treatment off the protocol. But, again, it

15 depends on the intervention that is being

16 undertaken and the specific tumor diagnosis under

17 consideration.

18 When you accumulate all this information,

19 the monitoring and the clinical trial itself, that

20 is where some of the infrastructure that NCI

21 supports comes into play because, as I mentioned

22 before, it is very important that patient data is

23 submitted at protocol-defined intervals; that the

24 data is accumulated, analyzed and then reported;

25 and then that the significance of this data, be it

1 the toxicity data or the effectiveness data, is

2 interpreted so that appropriate patients are being

3 accrued to the study; that treatment toxicity is

4 acceptable and that there is some efficacy of the

5 treatment interventions as defined in the protocol

6 beforehand.

7 There is some debate and discussion and

8 variability in terms of who and how often this data

9 that is accumulated and reported on is reviewed.

10 Within NCI, we work with the guidelines established

11 by NIH for data and safety monitoring and these

12 requirements call for the oversight and monitoring

13 of all human intervention studies to ensure the

14 patient safety and the validity and integrity of

15 the data itself for the study. The monitoring in

16 the study is to be done at sort of a level that is

17 commensurate with the risks and size and complexity

18 of the clinical trial.

19 The oversight monitoring under Phase III

20 clinical trials, which many of the pediatric

21 oncology trials are, calls for the establishment of

22 a DSMB. The DSMB, according to NIH, is also

23 appropriate for Phase I and Phase II clinical

24 trials if the studies have such things as multiple

25 clinical sites, are blinded or masked or employ

1 particularly high-risk vulnerable patient

2 populations. In pediatric oncology we sort of hit

3 throughout this so we call for sort of the default

4 to be towards some sort of formalized monitoring

5 committee for most of the studies that we do.

6 The NCI, in response to NIH sort of

7 formalizing its approach to data and safety

8 monitoring, in the not too distant past has

9 finished reviewing all the data and safety

10 monitoring plans for the cancer centers that NCI

11 supports across the country. That was I think an

12 education for both NCI as well as for the cancer

13 centers, for them to really kind of fess up and

14 look at what they actually do in terms of the

15 monitoring; what goes on in their human subject

16 clinical trials within their cancer centers. But

17 they all submitted them and they were all reviewed.

18 Some of the key, essential elements for

19 these monitoring plans that we had to consider, and

20 that then subsequently have also been extended to

21 some pediatric groups, are the monitoring and

22 progress of the trials and safety of the

23 participants; the plans for assuring compliance

24 with adverse event reporting; and plans for

25 assuring that data accuracy and protocol compliance

1 are performed.

2 As I mentioned, while in pediatric

3 oncology basically we don't work from a cancer

4 center model, we work more in a multi-institutional

5 approach so it is a more distributed coverage in

6 terms of who is performing the trials.

7 Nevertheless, these particular essential elements

8 were taken on by pretty much all the groups that we

9 have that I mentioned earlier that NCI supports in

10 one form or another, again, moving to the default

11 of having some sort of more formalized data

12 monitoring committees for all the trials.

13 The composition of the DSMB and the

14 various data monitoring committees may differ

15 between the different groups that I mentioned that

16 NCI supports for pediatric oncology but the goal is

17 the same, and it is to have capable and informed

18 observers be responsible for the oversight of the

19 trial. The reviewers are people that are outside

20 of, and in addition to the study committee, and

21 they evaluate the trial data at regular intervals

22 to monitor the treatment toxicity and the

23 effectiveness of the treatments that are being

24 used. Then, the review determines whether the

25 continued accrual to the trial is safe and

1 appropriate. COG itself has two DSMBs, one for

2 solid tumors and one for the leukemia and lymphoma

3 studies, and they meet twice a year, each one of

4 those DSMBs, to go over the studies. Actually we

5 go over pilot, Phase II and Phase III studies in

6 those sessions. The Phase I Consortia also has a

7 DSMB that meets twice a year to go over all those

8 Phase I studies. In addition to the Phase I

9 Consortia, the PBTC and the NANT, all of which have

10 a DSMB type of component, have more frequent

11 discussions with the groups that are beyond just

12 the study investigator and any sort of data

13 personnel or statistician directly involved. They

14 have a discussion of their studies sometimes on a

15 weekly basis, sometimes on a monthly basis, and

16 sometimes it also includes people from outside the

17 group itself to overlook what is going on with

18 their particular studies.

19 In terms of compliance with adverse event

20 reporting, another one of the essential elements

21 that NCI has, NCI-funded studies use the adverse

22 event expedited reporting system, or the AdEERS

23 system to report toxicities. This is a

24 computerized system that is available now to all

25 the funded groups with which they can fairly easily

1 report adverse events that occur during their

2 clinical trials. That data can then be accumulated

3 easily within their group, but also important

4 things can be sent off to the FDA or to drug

5 sponsors or the NCI as appropriate, especially for

6 studies that involve IND agents.

7 Then, it is the institutional principal

8 investigator that is ultimately responsible to

9 assure that the AEs are reported in a timely

10 manner. Whenever we review the cancer center

11 approaches, they list out that sort of the CRA

12 should submit this and then there is a nurse

13 practitioner or someone that is behind the CR to

14 make sure it gets submitted, and at some interval

15 the principal investigator locally is responsible

16 to make sure that all the AEs that may have

17 occurred had been properly reported.

18 Finally, for assuring data accuracy and

19 protocol compliance, the cooperative groups and

20 these consortia practice ongoing quality control

21 and interval quality assessments such as by using

22 institutional audits. This has been something that

23 has been ongoing throughout the creation of each of

24 these groups.

25 In summary, NCI has worked to establish a

1 framework to allow appropriate monitoring and

2 oversight of pediatric oncology clinical trials.

3 To address some of the issues that Steven had

4 brought up before in terms of the general

5 parameters that we look at, we first want to make

6 sure that the enrollment of patients is appropriate

7 to the diagnosis and risk of relapse for the

8 patient or the availability of standard treatments

9 for recurrent and relapsed disease, and that

10 laboratory and radiologic monitoring for toxicity

11 and response to treatments is established within

12 the protocol before any patients are accrued.

13 The frequency of monitoring would be equal

14 to or greater than standard of care for the

15 individual patient that is enrolled on a clinical

16 study, and there would be continuous protocol

17 monitoring by the study committee because they

18 receive this data on a daily basis. There would be

19 interval protocol monitoring on a monthly to

20 biannual basis, depending on the risk and specifics

21 of the trial, by a group outside of the study

22 committee itself.

23 Who does the monitoring? The daily

24 monitoring is by the study committee itself. The

25 interval monitoring usually involves concentrations

1 and statisticians that are not directly involved in

2 the trial.

3 When is a data monitoring committee

4 needed? For Phase III studies you need a DSMB.

5 For multi-institutional trials you need to have a

6 monitoring committee for high-risk populations.

7 You need to have a monitoring committee for complex

8 treatment. For studies with early stopping rules,

9 which many pediatric studies have, you have to have

10 a monitoring committee. With conflicts of

11 interest, which may not be as much of a case in

12 pediatrics as it might be in medical oncology, you

13 need to have a monitoring committee.

14 I think that with pediatric oncology

15 trials we hit many of the points that are brought

16 up by various agencies of situations where a

17 monitoring committee is required so that virtually

18 always in pediatric oncology some sort of

19 monitoring committee is involved in the oversight

20 of the practices of the group, as well as the

21 conduct of individual clinical trials. Thank you.

22 DR. SANTANA: Thanks, Barry. Before I

23 stand up to give the last presentation of the

24 morning, we have an opportunity for an open public

25 hearing. So, if there is anybody in the audience

1 that wishes to address the committee, this is the

2 opportunity to do so. I would ask that if you are

3 going to do that you come to the front of the room

4 to the podium and identify yourself by name and

5 affiliation.

6 Open Public Hearing

7 MR. RAKOFF: Wayne Rakoff, Johnson &

8 Johnson. Just a quick question, that came up this

9 morning that I would like to hear discussed during

10 the discussion, is with regard to the FDA guidance

11 on data reduction in oncology trials. It would be

12 important to us to know if there are any variances

13 in that with regard to pediatric studies.

14 DR. SANTANA: Steve or Rick, do you want

15 to address that now or do you want to address it

16 during the discussion period?

17 DR. HIRSCHFELD: We can address it in a

18 little more detail but, in brief, that is a global

19 commentary and there isn't a specific pediatric

20 component to it. I think that is a good suggestion

21 that maybe we should consider in the future, a

22 pediatric specific component.

23 DR. SANTANA: Any other comments from the

24 audience?

25 [No response]

1 Monitoring Procedures at a Private

2 Children's Hospital

3 DR. SANTANA: First of all, I want to

4 thank Steve, Richard and the rest of the FDA for

5 always bringing the pediatric oncologists to set

6 examples in these initiatives. I am personally

7 very appreciative of all the efforts that we have

8 had on behalf of the issues that we deal with in

9 pediatric oncology.

10 My task this morning, as I was charged to

11 do, is to bring a perspective from a private

12 institution with the caveat that St. Jude really is

13 an NCI cancer designated center so a lot of what we

14 do in terms of our own monitoring is reflective of

15 what we have to do to comply with the NCI

16 regulations.

17 What I would like to do over the next 20

18 minutes or 25 minutes or so is talk to you about

19 two issues. One is how we set forth monitoring of

20 our St. Jude studies--not the cooperative group

21 studies for which we still have to comply with COG,

22 but our own intra-institutional studies that follow

23 a parallel system to the NCI monitoring plan, and

24 what that monitoring plan involves and what

25 parameters we have designated for monitoring.

1 Then, a bigger part of my talk will be on a project

2 that Don Workman and I worked on in terms of trying

3 to handle adverse event reporting within the

4 institution and tried to develop an interactive

5 web-based model to try to get a handle on that.

6 With that, I will go ahead and get

7 started. As Barry has already said, monitoring of

8 trials is really an ongoing, continuous review of

9 the conduct of the trial. For the purpose of

10 distinction, I will make the note that to me

11 monitoring occurs while the study is ongoing.

12 Whereas a lot of people use the word auditing, to

13 me auditing is a post facto thing that happens

14 after the study has been completed. Then you go

15 back and see if the study was conducted the way it

16 was supposed to be; if the data is good enough; if

17 there is quality in the data; and if there have

18 been any other issues that occurred during that

19 post facto process. So, to me, monitoring occurs

20 real time whereas auditing occurs after the study

21 has been completed.

22 Monitoring is really a shared

23 responsibility of many individuals. We always talk

24 about monitoring being the responsibility of maybe

25 one particular group but at St. Jude we have the

1 notion that this is really the responsibility of

2 the research team. We always talk about the

3 principal investigator but it is really the

4 research team. The research team has many

5 components to it of which, hopefully, the principal

6 investigator is the lead person but there are

7 research nurses, there are CRAs, there are other

8 members of the study team who also have

9 responsibility for this process.

10 Institutional officials have a major role

11 in this, not only in terms or providing

12 infrastructure resources to conduct some of this

13 monitoring, but also to set a culture and example

14 that is transparent to make sure that things occur

15 very openly and that everybody is knowledgeable

16 about what is happening. Then, the oversight

17 committee--you heard a little bit about DSMBs which

18 I won't talk about and IRBs and other committees

19 that may be involved in this process.

20 Eric had a little figure this morning of a

21 triangle. I didn't know he had a triangle so I

22 brought a triangle too, but my triangle is a little

23 bit different. It makes a different point. The

24 point of this triangle is that in the center of the

25 process are the participant in the research but

1 there are many other people involved in this whole

2 process in which, as I mentioned to you earlier,

3 the partnership includes the investigator, the

4 research team, the IRB, other oversight committees

5 and then institutional officials. So, I view this

6 more as a partnership, not just the responsibility

7 of one individual.

8 One of the things I want to cover is point

9 number one and point number three on this slide,

10 which is how can we systematically approach some of

11 these problems in terms of monitoring and adverse

12 event reporting.

13 So, I think the first step whenever you

14 deal with a promise to define a problem in this

15 case is what needs to be monitored and what needs

16 to be reported. I think that is a good point to

17 start and I will talk about that in a minute; then,

18 dividing the role, the different committees that

19 provide some of this oversight and I really won't

20 go into detail on that although I could during the

21 discussion if anybody has any questions; and,

22 lastly, developing an infrastructure to allow this

23 to happen so that the reporting occurs, that there

24 is a process of evaluating the reports, and then a

25 process of acting in a timely manner when there are

1 concerns. So, that will be the latter part of my

2 talk.

3 As I mentioned to you, we are an NCI

4 cancer designated center so we also had to comply

5 and submit an institutional data safety monitoring

6 plan to the NCI a few years back that was reviewed,

7 approved, etc., etc., and now we provide our

8 monitoring under the umbrella of what that plan

9 says.

10 So, the first thing was to define what

11 elements we were going to monitor. So, we have

12 kind of followed the parallel system that the NCI

13 designated in the clinical data update system of

14 what data should be collected. We look at patient

15 specific data, the demographics, date of birth,

16 gender, those things that we have to collect; the

17 date of entry into the study; the treatment status,

18 if the patient has been previously treated, on what

19 protocols and what therapy the patient was on; and,

20 if they were off therapy, for what reasons. All

21 that gets captured as part of the monitoring of the

22 patient on the study.

23 Then, there are subgroup data elements

24 that are also captured. Barry mentioned, very

25 appropriately in his talk, the issue of eligibility

1 and determining that the right patients go on the

2 right studies. One of the things we have done at

3 St. Jude in the last ten years is we have

4 established a separate office, which is called the

5 protocol office which is actually an office that

6 provides the infrastructure to help investigators

7 deal with many of these issues. The protocol

8 office, obviously, is manned by a group of people

9 and one of the responsibilities, for example, is

10 that when an investigator enrolls a patient on a

11 study we have to fill out electronically an

12 eligibility check list. The eligibility check list

13 gets faxed to that office and a patient-specific

14 consent is generated for that patient on that

15 study. So, right at the beginning there are some

16 checks and balances in terms of the eligibility of

17 the patient so that the right patient is put on the

18 right study and the correct consent is used for

19 that patient. So, that is an ongoing process that

20 occurs early on during the trial and the patient

21 enrollment of the trial.

22 Once the patient receives the therapy,

23 they monitor the cycle or the course of therapy.

24 If is a Phase I study, what dose level the patient

25 is currently being treated with; the start date;

1 some other parameters like BSA and weight. They

2 monitor, particularly in Phase I studies, the

3 agent; the dose of the agent; if there have been

4 any modifications, why there have been

5 modifications. We will talk a little bit about

6 adverse event reporting later on. Then, as part of

7 the monitoring during certain periods of the trial,

8 the patients will be monitored in terms of response

9 because the trials will have stopping rules based

10 on response, not only in terms of toxicity but also

11 in terms of response so a Phase II trial that has

12 some response built-in stopping rules will be

13 stopped at the right point once the monitoring is

14 occurring in terms of the response that has been

15 achieved.

16 I tried to summarize this in two or three

17 slides. This is kind of how we do it at St. Jude

18 in terms of our own institutional Phase I/Phase II.

19 We don't do many Phase III but we do have an

20 auditing plan for Phase III studies and for some

21 studies in which we hold the IND.

22 So, for Phase I studies the central

23 elements in terms of demographics, eligibility and

24 informed consent, that is monitored continuously.

25 It is monitored continuously because I told you

1 that there is a check at the beginning in terms of

2 eligibility and in terms of informed consent that

3 occurs in real time when the patient gets

4 registered. So, that is done continuously as the

5 patients go on a study in a Phase I study.

6 The protocol office also is monitoring the

7 study in terms of the data elements for the study

8 so there are templates very similar to the RDE

9 system that is developed by COG, templates of data

10 capture forms. Those data capture forms are

11 electronic and the monitor on a monthly basis that

12 he or she is assigned will go through those and

13 will see if there is data that is missing. If

14 there is data that is missing, a report is

15 generated to the principal investigator that data

16 is missing on a monthly basis. So, it is a good

17 system in terms that it keeps the research team

18 kind of continuously on top of making sure the data

19 is being collected.

20 On a quarterly basis for a Phase I study

21 there is a report that is generated. I will show

22 you in a minute where the reports go but, in a

23 nutshell, it goes, obviously, to the principal

24 investigator and to the research team, and then it

25 goes to the subcommittee of the scientific review

1 committee that also oversees monitoring to make

2 sure that they are separate from the protocol

3 office and from the investigator looking at this

4 data.

5 Then, for every Phase I study that we are

6 the primary sponsor of at St. Jude, the first three

7 patients enrolled in the study are monitored.

8 Then, once the first three patients are monitored,

9 one additional patient per dose level is monitored

10 in real time. The idea of doing the first three

11 patients is that in many studies usually within the

12 first three patients you know if your systems are

13 in the right checks and balances so that you want

14 to monitor those first three patients very acutely

15 so if there is a problem with the system, with the

16 templates, with potentially things not going right,

17 you can pick it up very quickly and make the right

18 adjustment so that for the subsequent dose levels,

19 if you monitor one patient in real time, you should

20 have resolved all of that.

21 We do a lot of Phase II studies at St.

22 Jude and we also do the eligibility, essential

23 elements and consents as outlined here. We also do

24 missing data reports on a quarterly basis.

25 Obviously, in Phase II, just like in Phase I, you

1 are interested in adverse events and those are

2 reported quarterly. Then, on a semiannual basis

3 the monitors will verify the coding of response so

4 that the studies can be stopped if the response

5 criteria for stopping rules have been met. There

6 are reports semiannually or more frequently or less

7 frequently, as defined by the protocol, in terms of

8 the individual monitoring plan that the protocol

9 may have.

10 In Phase II we always monitor the first

11 two patients plus at least--and the clever word

12 here is "at least" ten percent of the total

13 patients that are being accrued. It could be

14 greater than ten percent. It depends obviously on

15 the resources that you have available and the

16 workload that the specific monitor may have but at

17 a minimum ten percent of the patients on any Phase

18 II study at any given time should be under active

19 monitoring.

20 We don't do many Phase III studies at St.

21 Jude but we do have a marching plan in the event

22 that there is a Phase III study and it parallels

23 the Phase II monitoring plan, with the exception

24 that there may be other primary objectives in the

25 Phase III trials that also require some monitoring.

1 St. Jude holds INDs or IDEs for a few

2 products so under those circumstances, they could

3 be Phase I or Phase II trials or whatever, but

4 separately from those, if there is a particular IND

5 or IDE for which St. Jude is the "sponsor" then

6 there is a specific monitoring plan that is

7 assigned to that study, and it will depend on the

8 risk, what is known about the IND drug, what is

9 known about the device, etc., etc., and may be more

10 strict but at least it will be just like Phase I or

11 Phase II studies I described to you before.

12 Usually, under some circumstances like some novel

13 therapy, it may be a little bit stricter in that

14 the studies are being monitored a little bit more

15 aggressively.

16 So, this is kind of in a nutshell how we

17 kind of agree with the NCI in our data safety

18 monitoring plan and how we would monitor our

19 studies. Having said that, there is also auditing

20 that occurs. So, there is a different auditing

21 plan that I am going to give a lot of detail about,

22 but for most auditing plans the monitors, once the

23 study is done, will make sure that at least 20

24 percent of the patients have had a full audit of

25 their records. But that is after the study is done

1 and that occurs over a long period of time. It is

2 not as active as the actual monitoring which is

3 occurring in real time.

4 I want to switch now and talk a little bit

5 about the issue of adverse event reporting which

6 has to do with monitoring and safety. We, at St.

7 Jude, also have struggled with this issue and we

8 struggle because there are a lot of problems in

9 reporting. There tends to be a lot of

10 over-reporting. That is, anticipated adverse

11 events that are known in the investigator's

12 brochure or known from other clinical trials are

13 being reported on a continuous basis and that

14 creates a big backlog of data that is important but

15 not important in real time in terms of monitoring.

16 As you all know, there is increased

17 research in new drugs and biologics. There is more

18 oversight and scrutiny by federal agencies. Just

19 like in many other places, we tend to get

20 saturation effects. There comes a point where you

21 see so many reports that it doesn't ring a bell; it

22 doesn't ring any whistles or anything like that.

23 So, we have to be careful that we don't over-report

24 because then it gets us into the saturation effect

25 and we don't react appropriately when there are red

100

1 flags that we should be paying attention to.

2 But one of the problems we have at St.

3 Jude, which is very common for pediatric

4 institutions, is that there are no denominators for

5 how to make any sense of this; what constitutes a

6 red flag? Where do you cut the line to say this is

7 important or this is not important? There is no

8 normative data for each of the populations that we

9 have to deal with for Phase I studies, for Phase II

10 studies and for the studies I mentioned to you.

11 So, trying to approach this problem, we have tried

12 to deal with this I think in a prospective way.

13 In terms of review, there are a lot of

14 external events that we get from study sponsors.

15 If there happens to be a drug that we are doing a

16 study with but the drug is being used in adult

17 studies or in other institutions, you know, the

18 sponsors package a lot AEs and send them to you and

19 we have to deal with those too. The problem with

20 those is that sometimes the information is very

21 sketchy and there is no opportunity for

22 clarifications or for questions so that then you

23 can put that in the context of your own experience

24 with your own patients at your own institution.

25 The other thing is that the IRB is not a

101

1 DSMB. A DSMB has a very specific role; the IRB has

2 to deal with a lot of other issues. They have to

3 deal with adverse events and they should be looking

4 at them and they should be judging them, but it is

5 clearly in the context of the whole package of the

6 research, whereas the DSMB has very specific roles

7 and responsibilities.

8 The IRB is not the FDA who holds the IND

9 file for the drug and knows everything. So, the

10 IRB over here is getting little pieces of

11 information and trying to make sense out of it in a

12 more global sense. Then, the IRB also needs to

13 rely on the local investigators to interpret the

14 meaning of the adverse events that they are

15 receiving from the outside, from the sponsors,

16 because clearly the IRB doesn't have the expertise

17 or the knowledge to put that in contextual features

18 in terms of the study as it is being conducted at

19 other institutions.

20 So, at St. Jude we decided to approach

21 this problem first by doing quality improvement

22 projects, trying to figure out where the problems

23 were and where we could attach the problems. One

24 of the first issues that we addressed is that at

25 the beginning the PI or the research team needs to

102

1 report and categorize the events, but there was no

2 systematic way of doing that. I mean, it was being

3 done in paper form; there were different versions

4 of that paper form.

5 One of the things that Don Workman and I

6 recognized is that at least if at the beginning we

7 could make this a standardized way and force

8 everybody to do it the same way, then five, ten

9 years later we actually would have a system in

10 place that would provide a lot of the normative

11 data that we would need in order then to do some

12 process improvement.

13 So, the first thing that we did is to

14 create this electronic submission that I will

15 describe to you in a few minutes. This electronic

16 submission is pretty neat I think, to use words of

17 my nephew--it is pretty neat because it allows you

18 then to disseminate that information very quickly

19 to all the key players in the field and then they

20 can do their own assessment the same time that the

21 IRB is doing their assessment. So, the IRB will

22 get a copy of this electronic adverse event and the

23 IRB will do their own assessment of the adverse

24 event and certainly give feedback and follow-up to

25 the investigator. At the same time that it goes to

103

1 the IRB, it goes to our office of regulatory

2 affairs which is also charged with making sure that

3 agencies that have to be notified about these

4 adverse events are also notified. So, it kind of

5 takes the IRB and the investigator away from that

6 responsibility of having do to that paperwork but

7 it goes to a central office that then now deals

8 with all the external agencies that have to look at

9 this data.

10 Internally, it goes in a different

11 direction. It goes to the vice president of

12 clinical trials for internal reporting and internal

13 processing so that the St. Jude DSMB or what we

14 call our scientific review council which is called

15 the CPSRMC, the clinical protocol scientific review

16 monitoring committee, is really the scientific

17 council which also has a function in terms of the

18 cancer center doing monitoring. They also get a

19 copy of the report and then they deal with it

20 internally and then they can give also feedback to

21 the principal investigator.

22 Don and I were very concerned with the

23 first step in this process to try and make it

24 uniform and to try to make it normative so that we

25 could then create a system that, hopefully, would

104

1 help us in retrospect. So, we started this about

2 18 months ago. The first thing we did is we said

3 let's create a form that is standardized. We can

4 then make sure that people understand what is

5 important in that form before we convert it into an

6 electronic format. Then we were able, as we

7 designed the form, to start thinking prospectively

8 of how that same data could be captured

9 electronically.

10 Then we developed a flow diagram as a

11 quality improvement project of where this web-based

12 report could go, which is a little bit of what I

13 just showed you. We had to deal with some issues

14 of security access and then we also had to deal

15 with some issues of electronic signature that we

16 eventually resolved.

17 One of the key features of this, which is

18 a recurrent problem in adverse event reporting, is

19 that there are databases and the databases don't

20 talk to each other. So, one of the key features

21 that we wanted to cover in this was to make sure

22 that this adverse event electronic reporter was

23 talking to the other databases in the hospital and

24 was capturing information from the protocol office

25 in terms of the protocol that the patient was

105

1 registered on and the additional protocols was that

2 the patient was registered on because there could

3 be some cross-talk between adverse events on

4 different protocols or different PIs. I will show

5 you an example at the end.

6 We also wanted to make this user friendly

7 and make sure that anybody who is part of the

8 research team could do this at any place in the

9 hospital. Through a security pass they could

10 access this web site and could potentially feed in

11 the information in a very quick manner, without

12 having to go to a dark office somewhere and grab

13 papers and try to do it. So, there were some

14 security access issues that got resolved but it was

15 made available to anybody on the research team

16 electronically.

17 We then tried to address the issue of

18 internal reporting, that is studies in which

19 adverse events are occurring in our patients at our

20 institution versus the information of adverse

21 events that are occurring at other sites that are

22 being fed into our protocols in terms of the

23 cooperative group studies, and so on and so forth.

24 So, one of the things that we had to address is how

25 we could link protocols so that the information

106

1 could be identified very easily. If a patient was

2 registered on one protocol and the adverse event

3 occurred on that protocol, we wanted to know what

4 additional studies that patient was enrolled on so

5 that when the IRB or the subcommittees reviewed

6 this they could begin to get trends if there were

7 complementary adverse events that were occurring

8 from complementary studies and there could have

9 been a red flag there that we needed to address.

10 In addition, we could share the

11 information with the PIs of the other studies

12 because they also have to be kept in the loop in

13 terms of what is happening to patients that

14 potentially may also be enrolled in their own

15 studies concomitantly, for example therapeutic

16 versus non-therapeutic studies.

17 Then, for external reports we wanted the

18 investigators to help us sort that out because we

19 couldn't sort it out. So, the investigators had to

20 invest some time at the beginning sorting out

21 external reports before they submitted them to us

22 so that they would be more meaningful to us.

23 Then, the functional outcomes would be

24 that there would be real-time reporting and that

25 the IRB would acknowledge that through some

107

1 electronic time stamping mechanism. There are

2 forced choices so that everybody has to do it the

3 same way; no incomplete data submissions so we

4 wouldn't have to address the issue of going back

5 and asking for more clarification and more

6 questions; easy access so it would be friendly;

7 ability to generate single incident reports;

8 ability to generate reports in a given time period.

9 If you were noting a trend that something was

10 occurring in a particular study over some period of

11 time, you could capture that and, as you will see

12 in the end, provide cumulative data that you could

13 sort out to look at trends that potentially could

14 be occurring. Quicker reporting times; ability for

15 the IRB office to generate reports based on

16 protocols; specific events across subjects, across

17 protocols to give us some functionality at the IRB

18 level to look at the data in different ways;

19 generate internal denominators of trends that we

20 wanted to look at; use standardized NCI toxicity

21 tables for the oncology trials; and be able to

22 record the IRB actions and updates from

23 investigators onto previous reports. So, it wasn't

24 a dead system. It was a system that the

25 investigator could go back and add more information

108

1 or, when the IRB reviewed it, could add more

2 information so it became a living document as the

3 report was being done.

4 Let me give you an example of how this

5 works. I couldn't get it electronically. It was

6 going to cost me money to be able to do this

7 electronically so I did some snapshots of what it

8 looks like.

9 So, this page is accessible to anybody who

10 is identified at St. Jude as a principal

11 investigator or a member of a research team. So,

12 if you are listed on the protocol as the nurse for

13 that study, as the statistician for that study, as

14 a pharmacist for that study, automatically you get

15 access to this through a user ID and your own

16 password. So, it is available to anybody who is

17 part of the research team.

18 This is how you log in. Here I logged in

19 and it says, "welcome, Victor Santana." Then it

20 gives a listing of all the events that have

21 accumulated during a particular period of time. It

22 gives the event ID which is an internal working

23 number. It gives the event date. It gives an

24 identifier that I have erased here for a particular

25 patient. It is usually a numerical number. If it

109

1 is an external event, then there is a way to code

2 that to an external number. Sometimes you get an

3 event from a sponsor and it is coded ABXY235, well,

4 there is a way that you can code that the same way

5 here so you can track it and use the same codifier

6 if you ever have to go back to the data.

7 The status tells me, as an investigator,

8 whether I have reviewed this or not. So, when I

9 copied this the other day I only had one adverse

10 event that I had yet to review that somebody sent

11 to me for comment. Then, it tells me the date that

12 the event was reviewed by me or that I modified it

13 or I did anything to it.

14 Very quickly, it goes through a couple of

15 screens that provide some general information. It

16 tells you whether it is a St. Jude patient or not

17 because if it is not, it throws you in a different

18 direction in terms of the data that you need to

19 capture because, clearly, the data is being

20 captured for external adverse events a little bit

21 differently than it is for internal. There is some

22 information here in terms of the patient.

23 Then, it begins to do its own internal

24 processing once it identifies the patient. It

25 tells us, as you see at the top of the screen, all

110

1 the protocols that this patient is registered on.

2 So, it goes back and talks to the data warehouse.

3 If this patient is enrolled on ten studies, it will

4 pull and identify all those ten studies. Then it

5 will ask me, as the person putting in the

6 information, under what study am I following this

7 report. So, it identifies primarily the study and

8 the adverse event, but it also tells me all the

9 other studies the patient is on, and this is

10 critical because this report will go to the PIs of

11 all those other studies too. You will see it at

12 the end for their comments. So, it provides a

13 little bit of a cross-talk among studies.

14 Then, it clearly identifies the type of

15 adverse event that is being reported. You have all

16 seen this in different variations. For adverse

17 events that require a CTC code it takes you to the

18 CTC code so there is a link too so you don't have

19 to scramble through 50 books looking for those

20 codes but automatically it links you to those

21 codes. Then, it allows you to put the descriptor,

22 etc., etc. So, it is all being captured in a

23 uniform language.

24 Then it goes to a page that allows the

25 person who is submitting the information to do some

111

1 attribution on the adverse event. It is a click

2 system but it reminds people, because we all tend

3 to forget, what each one of those words means. So,

4 it reminds me that I need to read when something is

5 serious; when something is unexpected. It defines

6 it very clearly because there are always a lot of

7 questions from members of the research team what

8 constitutes something that is unexpected versus

9 expected. Well, there it is. It is, hopefully,

10 black and white and then you select, based on your

11 interpretation. It allows you to do one selection

12 across lines horizontally for each one of those.

13 Then, there is a page that allows you to

14 provide more information. One of the problems

15 always with electronic information is that

16 sometimes you can't capture everything in a unique

17 format. So, there is a page that allows you to do

18 a little more narrative form of how this all

19 happened, and so on and so forth, so it can give

20 you some additional data that you can comment on.

21 Then it asks you do you think, based on

22 your interpretation of what has happened with the

23 adverse event, that there is a follow-up that is

24 needed. If you say there is a follow-up needed,

25 then it links back to a reminder within 30 days

112

1 that you owe us a follow-up. The IRB reviews it

2 and they also communicate directly. But if you

3 think you have enough information and you want to

4 submit a follow-up, within 30 days you will get a

5 reminder that you owe us a follow-up.

6 Then it tells you something about what

7 happened to the patient based on that adverse

8 event. Then it asks the investigator or the

9 research team to make some judgments based on the

10 information that they have on that particular

11 adverse event, and in terms of what they know is

12 going on in the study does this alter the

13 risk/benefit ratio for the other participants.

14 Does this require modifications to the protocol or

15 to the consent? And, does this provide additional

16 information that we should be sharing with other

17 people that are participating in the study? So, we

18 ask the investigator to specifically address these

19 issues with each adverse event.

20 This is an example of a summary page. All

21 that data is generated in the end into a summary

22 page. Obviously, I have whited out a lot of stuff.

23 There is a doctor that is called "Dr. Teddy Bear."

24 That is a famous doctor at St. Jude that we always

25 use whenever we do electronic examples of things.

113

1 But it gives you a nice summary of who is doing

2 this; who reported it; the protocol which was

3 reported; the PI of that study; the date it was

4 reported; when the adverse event started. It will

5 list all the studies, based on that warehouse

6 capture of data, that the patient was on. It will

7 quickly generate all that data into specifically

8 designated toxicities that were reported as part of

9 the adverse event. The attribution and nature that

10 you selected gets summarized; additional medical

11 history; treatment prognosis; patient outcome.

12 Then it tells us at the end--this all goes

13 to the IRB--it tells us at the end how the

14 principal investigator judged this in terms of his

15 own interpretation, that it doesn't alter the

16 risk/benefit ratio; does not require modification,

17 and so on and so forth.

18 So, it goes electronically--only focusing

19 on the IRB part of this, it goes electronically to

20 the IRB and there is a designated person in the IRB

21 office who will certify that he or she has received

22 this report, and will certify it electronically

23 down here with the date. Then it allows, at the

24 end, to add additional information when the IRB

25 actually reviewed it. So, the IRB will come back

114

1 at the end of the meeting and put in there the date

2 that it was reviewed by the IRB so it provides a

3 tracking record of when the IRB looked at it.

4 Another very neat thing I think, and I

5 like to use that word, with this project was that

6 it allowed cross-communication among investigators.

7 In that example I gave you, the message that there

8 was an adverse event reported in August, '99 will

9 also go to all these other studies that that

10 patient was enrolled on. So, the PI of the SD/01

11 protocol will also get the message and will get the

12 summary report, and the PI of that study has to pay

13 attention to that report and then make a decision

14 whether he or she thinks it may or may not be

15 related to his study too because there could be

16 complementary toxicities and they are the only ones

17 who are going to know that, not the IRB unless it

18 gets reported through a different mechanism.

19 So, it forces all the PIs of all the

20 studies that the patient is enrolled on to also

21 critically review the adverse event and make some

22 judgment about whether it is related or not related

23 to their own research. If it is, then it takes

24 them back to make some comments to the original

25 report that I submitted on my study. So, there is

115

1 a page that allows the other PIs to come back in

2 and give additional information.

3 This doesn't project very well and I

4 apologize, but all this data then can be captured

5 in different ways. In this particular page there

6 is data on one study and all the adverse events

7 that have been reported on that study within, I

8 think, a six-month period. Each one of those cells

9 can be manipulated to provide you different ways of

10 looking at the data. So, you could ask the data to

11 be cut only at grade 3 or grade 4 or only deaths on

12 that particular study. You can ask the system to

13 report all deaths on all patients across three

14 studies to see if there are complementary problems,

15 and things like that.

16 So, this is where we are right now. We

17 established this about 18 months ago. The next

18 phase of this project is actually now beginning to

19 mine the data so that we can create some normative

20 rules of when we should be setting lines that raise

21 red flags that we should pay more attention to.

22 So, I think that is the strength of this, that now

23 it unifies it in a certain way so that now we can

24 go back and make some sense of all the data, and I

25 think with that I will stop. Thank you.

116

1 Oh, obviously I didn't thank everybody

2 that was part of the team. Don Workman, our IRB

3 administrator, was very involved with this. Donna

4 Hogan, from the IRB office, is in charge of the AE

5 reporter. Then, two individuals from clinical

6 informatics were the ones who put all these ideas

7 to work. Thanks.

8 Now I think we have some time for

9 questions before we go into the discussion.

10 Committee Discussion

11 DR. HIRSCHFELD: I have a question for Dr.

12 Anderson. Dr. Santana discussed the goal of the

13 project at St. Jude to get some normative data on

14 what types of events one can expect and, perhaps by

15 implication, what needs to be monitored and what

16 doesn't need to be monitored, and when things do

17 occur how serious they are. Does the NCI have such

18 a program? If it does, are there any analyses that

19 you are able to share? Or Dr. Smith could answer

20 the question.

21 DR. ANDERSON: I am not aware of a

22 specific program for pediatric oncology, you know,

23 with the AdEERS system for bringing in information.

24 There is data trial by trial, especially for IND

25 agents. That sort of information is accumulated.

117

1 But to provide sort of a baseline level of here is

2 what to look for over time, I don't know that that

3 is a specific project that is under way right now.

4 DR. SANTANA: Yes, in fairness to the

5 question, we are not doing that right now. We have

6 the capability based on this project after we have

7 been into it 18 months because we thought about

8 that when we tried to build the electronic format.

9 We now have the ability to do that but, in fairness

10 to the question, we have not done that. We are

11 just establishing the data and, hopefully, at some

12 point we will begin to analyze it once we have

13 enough data to make some sense out of it. We are

14 really only, particularly right now, focusing on

15 the St. Jude studies, studies where we are the

16 primary sponsor.

17 DR. SMITH: Steve, as Barry said, we do

18 have the AdEERS system that is an electronic

19 reporting system. So, you know, there is the

20 capability if there is a question about cardiac

21 toxicity or other organ toxicity to pull up all of

22 those reports for a particular toxicity. But in

23 terms of what to look for, you know, if the

24 question is what toxicities are occurring in what

25 types of trials, then the Phase I and Phase II

118

1 databases of the Phase I Consortium and the

2 Pediatric Brain Tumor Consortium are more relevant

3 because if the AE reporting is being done

4 correctly, then it is the unexpected events. You

5 know, what you would really be interested in is the

6 whole universe of events and, as well, the

7 denominator of how many patients were in those

8 trials. So, I think I would approach one of the

9 consortia for early phase trials or COG for later

10 phase trials if the question was what type of

11 events are occurring, how frequently they are

12 occurring, etc.

13 DR. SANTANA: Dr. Przepiorka?

14 DR. PRZEPIORKA: Back to Steve, if I could

15 turn the question right back to you, does the FDA

16 have enough information or a database on SAEs in

17 pediatric trials to actually do that same study?

18 DR. HIRSCHFELD: Short answer? No. We

19 would like to but we don't have a database that

20 captures premarketing adverse events. We only have

21 a database for postmarketing adverse events. That

22 is mined in a fairly rigorous and maybe even

23 imaginative way to look at frequencies of what one

24 can expect but, again, it hasn't been examined

25 sufficiently on the basis of pediatrics and, even

119

1 more specifically, on pediatric oncology. So, we

2 don't have the data and that is one of the issues

3 and one of the reasons for having this discussion

4 this morning.

5 DR. DAGHER: Dr. Santana, another question

6 about your presentation which also may impact on

7 the NCI perspective, one of the challenges you

8 identified was the situation where a patient is

9 enrolled on several studies at the same time. I am

10 curious to know how often that happens and whether

11 that is somewhat unique to St. Jude, or is that

12 something that you also see across pediatric

13 studies that NCI supports?

14 DR. SANTANA: The way the system is

15 designed is that it will pick any protocol that the

16 patient is still currently enrolled on. It doesn't

17 mean the patient is on active therapy on those

18 other studies; it may be that they are in follow-up

19 for those other studies for example but the patient

20 has not been taken off those additional studies.

21 We did that on purpose in terms of thinking outside

22 the box, that if there were long-term issues with

23 patients that had been enrolled on other studies

24 and then you began to see trends that were

25 complementary to a group of studies that together

120

1 created something in the future, we could go back

2 and capture that.

3 So, your point is well taken. The primary

4 study that is generating the adverse events is many

5 times the active study that the patient is being

6 treated on. But we also wanted to make sure that

7 we were able to capture data on studies where the

8 patient was not actively receiving therapy but was

9 still technically enrolled on that study.

10 Having said that, we also wanted to

11 capture non-therapeutic trials so it will list any

12 trial. It won't make any distinction whether it is

13 therapeutic or non-therapeutic up front.

14 DR. DAGHER: Supportive care--

15 DR. SANTANA: Yes. On the last page there

16 was one trial which was a behavioral medicine trial

17 on which the patient had been enrolled that had

18 nothing to do with the primary therapeutic trial,

19 and that showed up too.

20 DR. HIRSCHFELD: May I ask for just one

21 more clarification on your presentation, Dr.

22 Santana? You said that the system in use at St.

23 Jude will bring up the relevant definitions for a

24 serious adverse event, unexpected, etc. What is

25 the source of those definitions? There are several

121

1 places that are source documents, including ICH

2 documents.

3 DR. SANTANA: I think what we did was an

4 amalgam of the different definitions and tried to

5 make it into a definition that people could

6 understand without having to pick up a dictionary

7 or call the IRB administrator. So, it was really

8 looking at all those documents and coming up with

9 some definitions that were kind of a semi-practical

10 way that people could relate to and then choose the

11 right box. Dr. Grillo-Lopez?

12 DR. GRILLO-LOPEZ: I have a suggestion for

13 future meetings on this subject, and that is to

14 invite a representative from the pharmaceutical

15 industry to make a presentation because although

16 you might argue that the FDA knows very well how

17 the pharmaceutical industry functions in terms of

18 monitoring adverse event reporting, on the other

19 hand, others around this table and others

20 participating in the Webcast or viewing the tapes

21 later on might not. The fact is that there is

22 extensive experience with clinical trial monitoring

23 in the pharmaceutical industry and, likewise, with

24 adverse event reporting. Although I see a great

25 parallel and even consistency in terms of the

122

1 procedures and methods that are used in your

2 institution representing an academic experience,

3 and at the NCI particularly with the cooperative

4 groups, there are some points that are different

5 and that would merit discussing and presenting

6 because they might present opportunities for

7 improvement.

8 DR. SANTANA: Are you in a position to

9 highlight some of those points?

10 DR. GRILLO-LOPEZ: Well, one thing that

11 just came up in the discussion was the subject of

12 denominators. Certainly, when you are conducting

13 research with a new therapeutic agent the database

14 at that pharmaceutical company contains the most

15 information regarding the safety experience with

16 that agent at any given point in time. Of course,

17 all of that database is transferred to the FDA as

18 required. But investigators participating in

19 multicenter trials could certainly call the project

20 clinician who would have access, through his

21 biometrics group, to that database and would be

22 able to provide information about what the

23 experience has been with other events of that

24 nature.

25 DR. HIRSCHFELD: Dr. Santana, if I may

123

1 just respond to the initial suggestion, and I want

2 to thank Dr. Grillo-Lopez, one of the reasons you

3 are at the table is to provide that. Previously at

4 the meetings of this committee we had multiple

5 representatives from the pharmaceutical industry

6 and had routinely asked for presentations but there

7 was a policy decision made outside the group that

8 you see here today to restrict that. So, since you

9 are new to the process--had you been involved

10 earlier you would have seen what you are

11 suggesting--maybe you can help us restore that

12 previous mode of interaction because we found it

13 helpful also.

14 DR. GRILLO-LOPEZ: Yes, that was an

15 unfortunate decision and I, of course, didn't know

16 about that. I think it is a three-legged stool or

17 a triangle, as you were saying, with the

18 participation, on the one hand, of the NCI and

19 cooperative groups particularly, individual

20 academic institutions and the pharmaceutical

21 industry as sponsors in conducting research. We

22 should not forget that third leg of the stool

23 because a lot of the research that is conducted

24 with new agents particularly is sponsored by the

25 pharmaceutical industry and the pharmaceutical

124

1 industry holds the databases for the results of

2 that research, and one particular institution may

3 have had a lot of experience with a new agent but

4 not necessarily all of the experience because many

5 other institutions might be participating and they

6 may not be communicating between themselves but

7 certainly the database of the pharmaceutical

8 company holds all of that information.

9 DR. SANTANA: Dr. Adamson?

10 DR. ADAMSON: A couple of comments, first,

11 I want people to be aware that what Dr. Santana

12 presented, which I think is something academic

13 institutions should strive for, is not the norm.

14 Most institutions are many steps behind what St.

15 Jude has done and is capable of doing, and in most

16 institutions what you are looking at are piles and

17 piles of paper. So, your colleagues are to be

18 commended on beginning to address what is a problem

19 for all academic institutions.

20 I wanted to comment that I think the

21 current SAE and AE mechanism--and this will echo

22 and build upon what Victor said--has some

23 significant flaws. I mean, we can be inundated

24 with reports that we cannot interpret, and what I

25 would say is that the large majority of external

125

1 reports, when it comes to the cover letter,

2 "because of regulation blah, blah, blah, you are

3 required to submit this to your IRB"--the large

4 majority of those reports, as a member of an IRB as

5 well as an investigator, one cannot interpret. It

6 gets down to knowing the denominator and you said

7 there are large databases but the problem is you

8 need real-time access to that database in order to

9 interpret it. There is too large a line of these

10 reports coming in for the investigator to call and

11 track down every report--is this relevant? Has the

12 risk/benefit ratio really changed for my patient?

13 Rick said earlier we should try to focus

14 on pediatrics so I will. As one moves

15 forward--because this is the problem and it is not

16 limited to pediatrics but is a problem across the

17 board that one can't interpret the large majority

18 of these reports and we are fooling ourselves if we

19 think simply by submitting the document to the IRB

20 you have fulfilled your obligation. That is not

21 improving patient safety. You may have fulfilled

22 the regulatory obligation but you have done nothing

23 to improve patient safety. We need access to the

24 type of data you referred to that industry has to

25 interpret this.

126

1 To focus on pediatrics, I will give you an

2 example. We did an industry-sponsored study and

3 there were many studies of this investigational

4 agent. The large majority of reports were about

5 myocardial infarction in a 76 year-old. That is

6 important but it is not particularly relevant to

7 the pediatric population. So, when we move forward

8 and ask for data, I think we need to have some

9 depth to that data, that is, not only the frequency

10 of the event but somehow to categorize what

11 population that event is occurring in. Because if

12 an event is occurring in a 30 year-old--and my mark

13 of what I think is young is continually shifting

14 upwards--

15 [Laughter]

16 --but if an event were to occur in a 30

17 year-old you might spend a little more time looking

18 at that event as far as, you know, was it a

19 cardiovascular event relative to someone who is

20 more elderly. So, I would hope that one looks at

21 the regulations and makes it that you don't just

22 send the report, but the report has to be in

23 context and the context is what is happening

24 globally with the safety of this drug, focusing on

25 its particular toxicity, but within that have the

127

1 depth to say this is the breakdown of the

2 population that we are looking at. I mean, you

3 don't need to get it down to all 12 year-olds, all

4 13 year-olds or all 20 year-olds but give us some

5 sense of what is happening. Otherwise, I don't

6 think we are doing anything for patient safety in a

7 meaningful way for the large majority of these

8 reports.

9 DR. KODISH: This is Eric, in Cleveland.

10 I hope my timing is okay and you can hear me.

11 DR. SANTANA: Yes, Eric, go ahead.

12 DR. KODISH: Thanks. I want to add an

13 idea to Peter's idea which I think is very

14 important and relates to the point I was trying to

15 make about regulation actually harming patient

16 safety on some level.

17 A 76 year-old who has a myocardial

18 infarction on a drug that we are testing in a

19 pediatric population compared to a 30 year-old

20 compared to a 20 year-old I think gives us the

21 ability to maybe, rather than contextualize which

22 would be great--ut maybe a more simple idea is to

23 provide some sort of sorting function so that we

24 are not just, for regulatory or prevention of

25 litigation, trying to download all of these reports

128

1 to our IRBs to say that we have fulfilled

2 regulatory requirements, but that at some level

3 there could be a sorting function so that the

4 events that are going to be relevant to children

5 are presorted, if you will, and not disseminated

6 across the country automatically. I think we do

7 need to be concerned about the paradoxical effect

8 of everyone feeling that because we have filed all

9 these adverse event reports that everything is

10 going to be okay.

11 DR. SANTANA: Eric, just to play devil's

12 advocate with your comment and Peter's comment, who

13 defines what is relevant to our population? It is

14 us. And, I think that is probably why we are here

15 today. We have to define in our studies, either

16 prospectively when the study is being created based

17 on what we know about the agent or whatever is

18 going to happen in the study or during the conduct

19 of the study as we review things--we are the ones

20 that have to define what triggers that it is a

21 pediatric issue that we need to address. If not,

22 then we just rely on these big data warehouses that

23 have data that are not relevant, but we have to

24 define the relevance up front or during the conduct

25 of the study.

129

1 DR. KODISH: I agree with that and I think

2 that maybe the discussion could focus on how we

3 sort those that are and those that aren't, maybe

4 starting with something as simple as an age

5 cut-off.

6 DR. SANTANA: Well, one of the things that

7 Barry mentioned in his presentation, and in

8 retrospect I wish he had given more discussion to

9 his point, was this issue of how some of the

10 consortia--and it is the PBTC Phase I group or

11 maybe it is the COG--that those committees

12 electronically and telephonically and through

13 computers meet on a regular basis and they review

14 real-time data of those patients that are on Phase

15 I studies. I presume, and I think correctly so,

16 that there also is, as part of that review, the

17 toxicity and the adverse events occurring in those

18 patients. So, that whole arm of this process,

19 which we didn't discuss in great detail, I think is

20 very strong because it relies in part on the

21 research team to very actively monitor this in

22 their own hands.

23 We have to have checks and balances

24 through other groups too but the beauty of that is

25 that it allows the research team who is actually

130

1 conducting the research in real time to be able to

2 communicate and evaluate these and then

3 prospectively, even as the study is being

4 conducted, define what are the parameters that

5 trigger the normative data that we are looking for

6 because in reality it is an experiment. Until we

7 do it we are really not going to know the whole

8 scope of things that may happen. We kind of can

9 predict based on what are the things that may

10 happen or what are the things that would really

11 worry us. Right? If somebody dies we all worry,

12 or if something unexpected occurs we all worry.

13 But for the majority of things, things are

14 happening and it creates a lot of noise. I agree

15 with you, Peter, it creates a lot of noise. So, I

16 think we have to go back to the research team and

17 address what their role and responsibility is to

18 help us at the other end figure out how this data

19 may be interpreted or may be incorporated.

20 DR. ADAMSON: I think you are right.

21 Phase I, in many respects, is somewhat easier

22 because the data is being monitored in real time

23 and the numbers are small. I think it is a

24 multi-level process. It begins with the treating

25 institution and the team at that institution

131

1 recognizing and identifying the event and reporting

2 it to the study principal investigator.

3 What we do in our consortium is once it is

4 to the study investigator it immediately comes to

5 us and then on a weekly basis all the events on

6 every study are reviewed. What we have the ability

7 to do and what we focus on is that we don't just

8 look at the serious ones because the serious ones

9 are usually pretty straightforward. We look at the

10 non-serious toxicities to look for trends because

11 we are doing a dose escalation and so we want to

12 know. Okay, we are starting to see some grade 2s

13 in an area that wasn't described that are not

14 triggering any alarms but, in fact, maybe we need

15 to do more careful monitoring of hepatic function

16 because we are seeing a lower level of toxicity.

17 So, it is a multi-level review but we have the

18 ability to look at all the toxicities on a study as

19 a function of dose, as a function of severity and

20 that gives us the context to interpret it.

21 DR. SANTANA: Dr. Grillo and then Dr.

22 Carome.

23 DR. GRILLO-LOPEZ: I find that we are

24 talking about adverse events in general but also

25 about serious adverse events and perhaps not always

132

1 making a distinction about the different reporting

2 requirements for those. Certainly, in the

3 pharmaceutical industry we collect each and every

4 adverse event but the reporting requirements are

5 different if it is a serious adverse event. It

6 might help if someone from the FDA would just

7 summarize what the requirements are for reporting

8 to an IRB and reporting to the FDA.

9 DR. HIRSCHFELD: The definitions are

10 essentially ICH definitions, International

11 Conference on Harmonization, that the FDA adopts.

12 The requirement essentially is if it is serious an

13 unexpected according to both triggers, then there

14 has to be what is called a rapid report filed.

15 That can be filed by a number of mechanisms. The

16 time frame typically is within 15 days and

17 sometimes, depending on the circumstance, can be 7

18 days. But that is still not what could be called

19 real time. It is essentially informing. All other

20 adverse events do not have to be reported to the

21 FDA, other than in the annual reports which are

22 required. The annual reports are due within 90

23 days of the initial filing of the IND.

24 DR. GRILLO-LOPEZ: How about to IRBs?

25 DR. HIRSCHFELD: The IRB requirements work

133

1 on multiple levels. So, the IRB can set their own

2 policy but in the FDA regulations, in 21 CFR 50,

3 the reporting requirements for IRBs parallel those

4 of reporting to the FDA.

5 DR. GRILLO-LOPEZ: If I may, in that vein

6 I would make the point that we have to be very

7 precise, very specific and very timely in reporting

8 serious an unexpected adverse events. At the other

9 extreme, there is a multitude of minor events that

10 are still adverse events and need to be in the

11 database at some point without creating this

12 backlog, this bureaucratic mass of paper and

13 electronic data coming at you without denominators,

14 which doesn't make much sense at any one given

15 point in time for one patient.

16 However, many times we find at the end of

17 the development of a new therapeutic that when we

18 have to put together the documentation to submit to

19 the FDA, one of the things that we, in industry,

20 have to do is to do an analysis across all of the

21 experience with that agent, Phase I, II, III, all

22 of the studies ever done, which is called the

23 integrated summary of safety. Many times it is

24 only then that certain trends become significant

25 that were not significant earlier on when you only

134

1 had the Phase I or the Phase II experience. That

2 is why it is important to report each and every

3 adverse event but not necessarily make it a

4 bureaucratic jungle where you just get so entangled

5 in paper and data that it doesn't make sense at any

6 one given point in time.

7 DR. SANTANA: Dr. Grillo, you represent

8 the pharmaceutical aspects of this. As somebody

9 from that group specifically focusing on pediatric

10 oncology issues, how would you advise your group of

11 things that we need to have access to, and in what

12 time lines would you advise your group that we need

13 to have access to those data so that we can

14 complement that with what we want to do?

15 DR. GRILLO-LOPEZ: I may not be the right

16 person to respond to that question because I am an

17 adult oncologist, not pediatric oncologist. In

18 fact, in over 20 years in industry, I never did a

19 pediatric study, ever. So, I have zero experience

20 and I have to be the first one to admit to that,

21 other than my rotation through pediatric oncology

22 when I was a fellow.

23 But I think there is a variety of ways and

24 systems and procedures that the pharmaceutical

25 industry utilizes to follow-up, collect and be able

135

1 to analyze adverse events. It begins with the case

2 report forms coming in from the different sites

3 participating in a multicenter study. I can tell

4 you that for all of the studies that I was ever

5 related with, I would personally look at each and

6 every piece of paper coming in from the different

7 sites, or the safety officer responsible within my

8 group would do that even before it went into the

9 database. So, if there was a major red flag that

10 was apparent even just from the experience in one

11 patient, we would see that. Of course, immediately

12 that was entered into the database and periodically

13 we would print out tabulations that would indicate

14 if there was any trend that was becoming obvious.

15 So, there is a variety of checks and

16 balances that are in place within the

17 pharmaceutical industry to follow-up on these

18 issues. Again, I would suggest that investigators

19 who are participating in multicenter pharmaceutical

20 industry sponsored studies, that their point of

21 access or one point of access might be the project

22 clinician within the pharmaceutical company who is

23 the person responsible and/or the safety officer

24 within that company when issues arise or questions

25 arise regarding a specific adverse event.

136

1 DR. SANTANA: Dr. Carome?

2 DR. CAROME: I will just note a couple of

3 things. I think it is important for this

4 subcommittee to be aware that this discussion is

5 occurring elsewhere. The Secretary's Advisory

6 Committee on Human Research Protections had a panel

7 on adverse event reporting and they are going to

8 continue that discussion at their next meeting.

9 They had a panel in December and they are going to

10 continue the discussion in their March meeting,

11 coming up in a couple of weeks. And the discussion

12 is exactly the same. I mean, the types of comments

13 being articulated are verbatim what you hear

14 repeatedly.

15 I think the Department recognizes that

16 there is a need to make adverse event reporting

17 more meaningful and less burdensome in order to

18 better protect human subjects, and there are

19 ongoing discussions between our office, the FDA,

20 NIH and other federal departments and agencies on

21 how best to do that. So, it is recognized to be a

22 problem and developing strategies is complex but we

23 believe important.

24 If you look at our regulations, just the

25 HHS regulations CFR 46, there is no adverse event

137

1 reporting requirement. There is a requirement for

2 reporting what are called unanticipated problems

3 involving risk to others. It is our view that most

4 adverse events that occur in clinical trials do not

5 fall into that category and, therefore, under our

6 regulations the vast majority of adverse event

7 reports do not need to be reported under our

8 regulations. Those that we particularly care

9 about, and we have articulated this at the

10 Secretary's advisory committee in December, are

11 those that represent unexpected, serious harms to

12 subjects, which are words that come from another

13 part of our regulation. Those are the types of

14 events we think should get to IRBs and that we care

15 most about.

16 DR. SANTANA: So, Mike, where do you think

17 the confusion comes that all these reports are

18 being generated and submitted to IRBs? Where do

19 you think the communication breakdown is in terms

20 of what the regulatory agencies want versus what

21 the sponsors or we, as investigators, see that you

22 guys want and need to comply with?

23 DR. CAROME: There are probably multiple

24 reasons. It is clear to us, and I think to others,

25 that the greatest burden comes from these external

138

1 adverse events that don't occur at your site but,

2 because we do research at multiple sites, the

3 sponsors deliver those reports or ask that they be

4 delivered to the investigators at all sites. So,

5 now we have 100 IRBs maybe receiving the same event

6 so it is those external events that are being

7 multiplied to multiple IRBs where the burden has

8 been articulated to us as being most severe, and if

9 the letter reads that under the regulations you

10 must deliver these to your IRB, that is certainly

11 one source. It is not our regulations that are

12 demanding that and I would posit that a close look

13 at the FDA regulations probably doesn't justify all

14 those events going to the IRB as well. But FDA

15 would have to comment on that. So, that is one.

16 I think it is driven by fears of

17 litigation liability. You know, who makes this

18 initial assessment about unexpected and serious?

19 We think that at one level the sponsor and the

20 investigator can be doing that. There are some who

21 think those are conflicted parties and maybe we

22 need an independent body making those decisions so

23 people are driven to having an independent body be

24 the IRB looking at them. So, I think there are

25 multiple reasons. Those are a couple that I would

139

1 highlight as perhaps driving it.

2 DR. SANTANA: Rick, do you want to comment

3 on the FDA?

4 DR. PAZDUR: Well, I just want to comment

5 in general. Could there be attempts to try to give

6 investigators more guidance on specifically what

7 needs to be reported? I think there is a tendency

8 to report a lot to cover oneself because we don't

9 have good guidance on exactly what those words

10 mean. Maybe we need to look into that. You know,

11 if you go over your phrase that you gave, there is

12 a lot of interpretation here and somebody could say

13 that it might be the index case; they are not even

14 sure of the attribution issue, and I wonder if we

15 really need to give more guidance to perhaps cut

16 down on some of this. I don't know, do you want to

17 comment on that?

18 DR. CAROME: I think for us, we believe

19 guidance is essential and it is the most important

20 step. We have had discussions with FDA. We are

21 prepared to draft guidance that articulates in more

22 detail what I just articulated to you and I

23 previously articulated at the Secretary's Advisory

24 Committee on Human Research Protections in

25 December. But we think, yes, guidance is the

140

1 important step. We think because adverse events

2 are primarily referenced in FDA regulations the

3 guidance needs to come out of both of our offices

4 or entities.

5 DR. SANTANA: Dr. Smith, I think you had

6 your hand up?

7 DR. SMITH: One point, and, Victor, I

8 think you made it, there is over-reporting of

9 adverse events, expedited adverse events, despite

10 FDA's stated requirements, despite their statements

11 in protocols of what does require expedited

12 reporting. So, I think one of the initiatives that

13 we want to undertake in the next few months is an

14 educational initiative to try to limit the

15 over-reporting of things that, in fact, just do not

16 require regulatory reporting. So, this will

17 decrease some of the burden at the institutional

18 level.

19 It doesn't address the issue, however,

20 that Peter raised about when you get a letter from

21 a company saying that this event occurred. I

22 wonder if there is a role for the Phase I

23 Consortium itself or the COG itself to play the

24 filter role that Eric Kodish was talking about in

25 terms of saying we have reviewed this, and our

141

1 recommendation to IRBs, when they look at it, is to

2 say this isn't applicable for pediatrics.

3 DR. ADAMSON: We are actually now doing

4 that, Malcolm, when it comes, you know, a COG

5 trial. When we disseminate it we usually give a

6 recommendation that, in our view, this does not

7 change risk/benefit or, in our view, it does change

8 the risk/benefit ratio and it should be reported.

9 So, we try to put it into context but, of course,

10 every investigator has the ability to interpret the

11 data and make their own decisions.

12 DR. SANTANA: Dr. Przepiorka, you had your

13 hand up?

14 DR. PRZEPIORKA: Yes, I clearly remember

15 sitting through multiple discussions at the

16 initiation site visits with sponsors regarding the

17 definition of an SAE, and I recall a few years ago,

18 after the incident at Penn and the FDA sent that

19 Webcast to all the academic institutions with a

20 long, drawn-out discussion on what is an SAE, and I

21 sat here and I think we listed them, although I

22 didn't see them on any of the slides this morning.

23 I won't go through them but I don't see any that is

24 very specific to pediatrics and I am wondering if

25 there is any SAE that should be added to the list

142

1 specifically for pediatric groups. I am thinking

2 about long-term cognitive dysfunction or something

3 like that.

4 DR. SANTANA: Ruth?

5 MS. HOFFMAN: I was just going to mention

6 that I sit on the IRB at Children's National

7 Medical Center, as well as the DSMB board there,

8 and from a lay perspective it is very difficult to

9 get lay people to continue with the responsibility

10 because of the burden of time commitment. There is

11 no monetary compensation. I don't get paid and I

12 am not an employee of Children's National Medical

13 Center. I spend three days a month totally related

14 to IRB-related work between the protocols and the

15 SAEs and AEs. I mean, it is just a stack of paper

16 and usually the check-off is that the AE has

17 nothing to do with the protocol at all and, you

18 know, maybe you can eliminate that whole column

19 and, again, reduce the workload. But, I mean, they

20 have a very hard time to even recruit members from

21 the community and that is a requirement of the HHS,

22 to have a lay person on the committee. So, the

23 guidance document would be great. It would

24 certainly help from our perspective as well.

25 DR. SANTANA: Dr. Grillo-Lopez?

143

1 DR. GRILLO-LOPEZ: I would suggest that

2 further guidance is not necessary, that what we

3 need is education. The guidance that is already

4 provided by the FDA is very specific and very clear

5 as to what is a serious and unexpected adverse

6 event, and what we need is for those involved in

7 research, and particularly at the IRB level, to

8 have an understanding of what that means. I think

9 it is education. Generating one more document to

10 file away does not help anyone.

11 DR. SANTANA: Peter?

12 DR. ADAMSON: I agree that the FDA

13 guidance on the definition of an SAE is clear.

14 What I think the point is, is that it is not

15 particularly functional in that it is generating an

16 incredible amount of paperwork for institutions.

17 What we get are, indeed, SAEs by the definition.

18 That information I don't think is improving patient

19 safety and that is why I would actually agree that

20 we need to re-look at what we are requiring to be

21 reported to IRBs across the country because it is

22 not only a multi-institutional trial, it is when

23 you have multiple trials of an investigational drug

24 that affects all those trials.

25 DR. SANTANA: Dr. Keegan?

144

1 DR. KEEGAN: Yes, I was wondering if we

2 could go back to the concept that was discussed

3 before about having a central body that looks at

4 all the adverse events because, as you say, every

5 individual institution is going to be unable to

6 look at a single adverse event out of context with

7 the rest of the data. So, to what extent are there

8 really plans in place to have a central point that

9 has all the data that could make reasonable

10 interpretations that have people with the

11 background information who could interpret the

12 adverse event information in the context of animal

13 and nonclinical studies and other things to make

14 relevant decisions? Because you mentioned that in

15 the instance of the consortium but it doesn't seem

16 that that is a general theme. To some extent, I

17 don't think any one individual is ever going to be

18 able to make a conclusion on the index case but we

19 certainly can't ignore the index cases because that

20 also puts patients at risk.

21 DR. SANTANA: Patricia, in follow-up to

22 that comment, what kind of body were you thinking

23 of? What ideal body, if you had to come up with

24 that, would you propose?

25 DR. KEEGAN: Well, it sounds like that is

145

1 sort of the model that the consortia are working on

2 and I thought maybe there could be more discussion

3 of whether it could be that sort of model, where

4 the consortium looks at adverse events and then

5 sends out their interpretation as a central

6 repository analysis, much like a medical monitor

7 would do at a drug company to perform that same

8 function.

9 DR. ADAMSON: Well, I think it is easier

10 for smaller studies, and the key thing is you have

11 access to all the data. So, when we do an

12 NCI-sponsored study where the NCI is cross-filed

13 there is a drug monitor at the NCI that has access

14 to all the data and, in fact--correct me if I am

15 wrong, Barry--usually when an AE comes out there

16 also is a recommendation of an interpretation when

17 it happens. I can't say that is the case uniformly

18 for industry-sponsored trials. But the multiplying

19 effect I think is a difficult effect for when

20 events are occurring really that are distantly

21 related to the study that you are doing.

22 DR. SANTANA: I think the advantage that

23 we have in pediatric oncology is that it is a

24 smaller universe and most pediatric, if not all

25 pediatric oncology studies are really conducted in

146

1 the list that Barry showed, plus a few others.

2 Right? So, we are a much smaller universe so that

3 if we adopted a model similar to what is happening

4 in the Phase I consortium and expanded that to all

5 the participants in those groups because it is a

6 small universe, we could at least set that model

7 and see if it works for us. Because that is what

8 we are really here for, right? For pediatric

9 oncology, not to ignore or belittle the other

10 important issues that are occurring with adults but

11 we have that advantage and maybe we should think of

12 that model as a test case for reviewing adverse

13 event reports to make it more functional and

14 timely.

15 To me, the issue is time. So, what if

16 something happened six months ago? It doesn't help

17 my patient who is on the study now. Right? So,

18 maybe we have that advantage. We are a small

19 group. Malcolm?

20 DR. SMITH: The other possibility is, Pat,

21 we are at the earliest stages of setting up a

22 pediatric central IRB and so, you know, could that

23 be a body that is somehow constituted so that it

24 could play that role nationwide and then other IRBs

25 could use that information if they chose to?

147

1 DR. SANTANA: Since you raised the issue I

2 am going to try to explore it a little bit further.

3 There has been some recent discussion I think in

4 some of the things I have been reading about

5 whether DSMB should play some of this role. Do you

6 want to comment on that?

7 DR. PAZDUR: I would just say that I had a

8 side conversation with Susan, here, and one of the

9 issues is that usually DSMBs are single trials.

10 Now, would one consider, for example, kind of a

11 super-DSMB not for the trial but for the drug that

12 is being investigated by a commercial sponsor?

13 Could a commercial sponsor, for example, if they

14 are investigating drug X in, you know, 50 diseases

15 in pediatrics, in geriatrics, and whatever, to have

16 a coordinating center to look at this and then

17 issue some type of report on these individual

18 toxicities?

19 Again, I understand exactly where Peter is

20 coming from and the comments, having been there.

21 You know, you get all this morass of information

22 which is almost useless because nobody knows what

23 to do about it and we are just generating paperwork

24 with a pretense basically that we are doing

25 something to further not only children but also

148

1 adult clinical trials.

2 Here again, you know, although we are

3 talking about pediatrics, this does have obviously

4 ramifications for adult medicine and adult clinical

5 trials. Although we may want to kind of say, well,

6 some of the adult toxicities may not protect for

7 what may go on into childhood toxicity, again, that

8 is another level of clinical judgment and

9 subjectivity that comes into play here. Many of

10 these drugs, especially with the IRBs, are not

11 solely being looked at in children. In many of

12 your hospitals, since you practice exclusively in

13 children's hospitals, that may be the case and your

14 interest may be in that group, but for a garden

15 variety IRB at a university hospital they may have

16 ongoing studies in adults with breast cancer, colon

17 cancer and pediatrics. So, they need to look at

18 this so it isn't that helpful sometimes to the

19 larger IRBs, the university IRBs.

20 DR. SANTANA: Dr. Reynolds, you had a

21 comment?

22 DR. REYNOLDS: I wanted to make it clear

23 that the DSMB process is a little different in the

24 pediatric setting. You have one for your

25 consortia, don't you, that look at all the studies

149

1 that are going on and that are not specific to a

2 drug. But I think taking that in the context of

3 what we are hearing from Ruth, and I hear this

4 continually from a lot of people, the burden that

5 is placed on the IRBs at the institutional level is

6 substantial.

7 Just taking a round number of 20

8 institutions in your consortia, Peter, you have 20

9 different IRBs looking at each one of these adverse

10 events. How many people are on each of those IRBs?

11 Certainly, the total number far exceeds the number

12 of patients on a study by an order of magnitude or

13 two. So, it is that process, yet we have a

14 centralized DSMB process. So the real central

15 issue though comes down to the responsibility that

16 the IRBs have under the regulations to be the

17 ultimate and final arbitrator of whether or not

18 this is going to be safe and appropriate for the

19 patients in their institution.

20 Somehow we need to use the word that I

21 first learned in the context of Steve Hirschfeld,

22 "harmonization." I see it over and over again with

23 the regulations you are harmonizing. I think we

24 need to somehow harmonize this process so that we

25 can then decrease the workload for these poor

150

1 people in the IRBs that, as you have heard, are

2 volunteering their time and they are a precious

3 resource that we could exhaust and then we wouldn't

4 have anymore volunteers.

5 DR. SANTANA: I want to follow-up on a

6 comment related to the previous issue of whether

7 there should be another body that could help us

8 review these things and probably give better

9 knowledge to practicing oncologists. You know, one

10 of the concerns I always have about creating

11 another body is that you don't destroy mass; it

12 doesn't go away; you are just shifting it to

13 another group. If we do that, I think we run the

14 same risk without clear guidance of what that group

15 needs to be doing. They are going to be getting

16 the same paperwork we are getting now. So, unless

17 there is guidance at the first step, which is let's

18 clearly define what we should be looking at and

19 streamline that, it doesn't really matter where it

20 goes to, whether it goes to an IRB, to a DSMB or to

21 another group or another consortium.

22 I think that may solve part of the problem

23 but it is really shifting a little bit of the

24 responsibility and what I want to get at is that we

25 should probably encourage ourselves more to define

151

1 the responsibility and the process rather than

2 creating another group. That is just a general

3 comment. It is not meant to be a criticism. It is

4 just something we need to think about.

5 DR. KEEGAN: Actually, going towards

6 that--what you say, it doesn't help to create

7 another group that is duplicating effort so it

8 would only be effective if, in fact, the other

9 groups then would agree to accept the information

10 provided by the central group. So, I think that

11 has been the issue with central IRBs all along.

12 While IRBs are crying out that they are

13 overwhelmed, yet, they also refuse to defer that

14 part of their responsibility to another group or to

15 a central IRB. Do you think that for a central

16 pediatric IRB there is more willingness to do that,

17 Malcolm? I mean, are they willing to say, okay, we

18 will allow somebody who is going to make an

19 integrated analysis to do that and we will accept

20 their judgment?

21 DR. SMITH: It will vary by institution.

22 You know, the adult IRB has a facilitated review

23 process and when a local IRB accepts the central

24 IRB as the IRB of record, then the central IRB is

25 responsible for the review of the adverse events

152

1 relevant to that study. In pediatrics, based on a

2 survey that the Children's Oncology Group did,

3 there is a high level of interest in a central

4 pediatric IRB, both among PIs as well as among IRB

5 chairs. But when it comes to implementation, some

6 institutions will accept it wholeheartedly and some

7 won't. But those who do will certainly be saving

8 in terms of the effort expended on this.

9 DR. SANTANA: Peter, one last question.

10 DR. ADAMSON: I just wanted to follow-up

11 on that. So, it is not only the IRBs who are

12 sometimes unwilling to give up the ability, it is

13 the institution. The institution more often than

14 not will actually tell the IRB, you know what, we

15 need an independent IRB; we are not going to accept

16 it. So, they may even take it out of the hands of

17 the IRB as far as whether they are willing to or

18 not. So, IRBs are looking for ways to cut down

19 their own work but it is not always coming to them.

20 DR. SANTANA: With that final

21 comment--Ramzi, I will defer to you.

22 DR. DAGHER: Just very briefly, you seem

23 to have identified a sense of challenges in terms

24 of the filtering. One is how to decide how

25 relevant an adverse event is, and that is not

153

1 really just specific to pediatric oncology or

2 oncology, for that matter. The second one, which

3 Peter Adamson was trying to focus on, is how do you

4 filter out the adult oncology experience or other

5 experience that is submitted to you in terms of how

6 relevant that is or isn't to the pediatric oncology

7 setting.

8 Now, you mentioned age and the nature of

9 the adverse event. Those are two potential

10 criteria. I am curious to know, and probably we

11 will get into this more in answering the questions

12 from Peter Adamson, Victor or others who have dealt

13 with this, what criteria do you use in making

14 decisions about filtering the adult oncology

15 reported events and deciding how relevant they are

16 to your specific studies?

17 DR. SANTANA: I think with that question

18 we will go ahead and try to address the questions

19 for the committee because I think we will cover

20 that.

21 DR. PRZEPIORKA: Can I just ask one more

22 question?

23 DR. SANTANA: Yes, Donna?

24 DR. PRZEPIORKA: You had indicated that,

25 if I recall, your institution does not take

154

1 patients off protocol so that you get long-term

2 follow-up. I was wondering if you thought that was

3 appropriate for everybody to be doing in the

4 pediatric population or if there is some time

5 limit, like by age 35 we are not going to look

6 anymore, or something like that?

7 DR. SANTANA: Well, if we are conducting

8 active research on those patients, those patients

9 would come off their primary therapeutic protocol

10 and get enrolled on a non-therapeutic protocol,

11 which is an umbrella protocol we have for long-term

12 follow-up. So, they would still be research

13 participants and we are collecting data on

14 long-term effects, survival and things like that.

15 So, the patient would come off the primary

16 therapeutic protocol once they are transitioned

17 into the long-term follow-up protocol on which

18 research is being conducted. So, those active

19 protocols will not show up in the reporter but the

20 long-term follow-up will show up in the reporter

21 for that patient.

22 Questions for Discussion

23 Let's go ahead and try to address the

24 questions that we have before us. Just for the

25 purpose of the minutes and the documents, I will go

155

1 ahead and read the questions to the committee, the

2 introduction, and then we will take one question at

3 a time.

4 The tolerance for risk in cancer

5 therapeutics is different than for most other

6 medical therapies. It is also recognized that

7 children are a particularly vulnerable population

8 and regulations and procedures have been

9 implemented to provide protection to children

10 participating in clinical research. The following

11 questions relate to the setting of children with

12 cancer participating in clinical trials.

13 Under the heading of "principles" the

14 question is, what are the principles that should be

15 addressed in safety monitoring of clinical studies

16 that enroll children with cancer? If the

17 principles are adequately stated in existing

18 documents, statutes or regulations, please identify

19 the relevant documents and sections.

20 Barry or Malcolm, from the NCI

21 perspective, do you have any comments on existing

22 regulations or documents that we could reference

23 to?

24 DR. ANDERSON: In terms of the DSMBs, the

25 composition of DSMBs, that sort of information is

156

1 provided in OHRP. In terms of the frequency of

2 monitoring and the exact nature of the monitoring,

3 what is monitored which is part of the discussion

4 we had, I don't know that that is laid out as

5 clearly. We have guidelines that we work with at

6 CTEP and NCI but I don't know that that is in

7 regulatory form at all.

8 DR. SMITH: Yes, there is the overall

9 policy on data monitoring. That is really not very

10 prescriptive in terms of here is what you have to

11 review; here is how often you have to look at it;

12 and here is, you know, who should be looking at it.

13 It says you need to have a plan but it is not very

14 prescriptive in terms of what the plan is. Each of

15 the institutions has their own data and safety

16 monitoring plans, particularly for Phase III

17 trials, and those tend to be more prescriptive and

18 detailed in terms of what is happening. But in

19 terms of early phase trials, you know, I am not

20 aware of kind of NIH-generated documents that

21 provide detail about what, how, when and where this

22 needs to be done.

23 DR. SANTANA: Go ahead, Barry.

24 DR. ANDERSON: And having been on the

25 panel of people who looked at the cancer center

157

1 data and safety monitoring plans that they had to

2 submit, previously I think a lot of people would

3 recognize that for early phase studies it was the

4 investigator and their research nurse that looked

5 over the data with the most frequency. A lot of

6 times I think there was not a lot of oversight from

7 outside of that small group. It was clear from

8 looking at the different cancer centers that there

9 is a huge spectrum of what in reality they were

10 doing and when you told them, you know, you need to

11 formalize this what they presented us with what

12 they thought were acceptable approaches. From our

13 point of view, we had these essential elements to

14 work from but they are very general and it took us

15 a while to kind of gear up to say here is exactly

16 what we think--well, not exactly but here is a

17 range of possibilities that are acceptable as an

18 approach, and I think it does vary by the type of

19 study that is actually being considered. That was

20 one of the criteria, for Phase I studies we would

21 do this; for pilots, this. For Phase II and Phase

22 III there were different levels of monitoring that

23 seemed to be appropriate for each of those, both in

24 terms of the type of monitoring and the frequency

25 of kind of review of the data and that type of

158

1 thing.

2 DR. SANTANA: As a follow-up to that, in

3 the non-NCI cancer center umbrella, all the other

4 groups that NCI supports like the consortia, are

5 there also specific requirements for DSMB plans for

6 those consortia?

7 DR. SMITH: The overall NIH requirements

8 apply to all NIH-sponsored research. Again, those

9 require a data monitoring plan, not a particular

10 form that that plan has to take for implementation.

11 I guess one question here is does FDA want kind of

12 the form and the details, or is it a question of

13 principles, you know, whatever the plan is, it

14 should adhere to these principles?

15 DR. SANTANA: I think with that comment, I

16 will ask Eric--are you still on the line?

17 DR. KODISH: I am here.

18 DR. SANTANA: Eric, can you comment on

19 that in trying to address the issue of global

20 principles, other than specific detail?

21 DR. KODISH: I would opt for flexibility--

22 DR. SANTANA: Eric, can you speak just a

23 little bit louder, please?

24 DR. KODISH: Yes. I would argue for

25 flexibility. I think that the different contexts

159

1 of the particular clinical trials involving

2 children with cancer that we are talking about

3 would dictate that it makes more sense to allow a

4 plan based on principles, such as beneficence or

5 such as filtering serious adverse events compared

6 to those that are not as impactful, and I wouldn't

7 try to prescribe the format so much. That would

8 lead to bureaucratization that could actually

9 paradoxically harm the ethical importance of

10 research.

11 DR. HIRSCHFELD: I would like a

12 clarification from Dr. Kodish. So, would you then

13 say that the principles of, let's say, beneficence

14 and respect contained in the Belmont report and the

15 principles that are annunciated in the ICH

16 documents, for instance particularly the one that

17 applies to pediatric research, E11, are a

18 sufficient statement of the principles?

19 DR. KODISH: I would.

20 DR. HIRSCHFELD: I think we can move on.

21 DR. SANTANA: Before we get to that

22 question though, because I want to make sure that

23 we cover the whole loop of this point, do

24 pharmaceutical sponsors in their DSMB plans have

25 any specific requirements for pediatrics, or are

160

1 pediatrics dealt with in monitoring plans as the

2 greater universe of adults? Or has that ever been

3 discussed, that they should develop specific plans

4 for pediatrics?

5 DR. GRILLO-LOPEZ: Not to my knowledge

6 but, again, I may not be the best person to address

7 that. On the other hand, I would like to comment

8 on the subject of DSMBs because I would not like

9 the FDA to come away from this meeting thinking

10 that there is an endorsement for DSMBs to be

11 required and/or regulated in any way, shape or

12 form. I think that there may be a need for some

13 consensus agreement at the level of professional

14 societies, the NIH and so on, on how different

15 DSMBs might be constructed and when they may or may

16 not be required, but allowing for the flexibility

17 that several around the table have mentioned.

18 DR. SANTANA: That was my interpretation

19 of the discussion too. I don't think there was any

20 endorsement from this group that we should be

21 moving towards a model DSMB to solve some of the

22 problems.

23 DR. GRILLO-LOPEZ: I see Dr. Pazdur

24 agreeing with that and I am glad to see that.

25 DR. SANTANA: I want to clarify that that

161

1 was my interpretation too. That is not what I

2 think the comment was all about. Eric, did you

3 want to add anything else? I am sorry, I think I

4 interrupted you. No?

5 DR. KODISH: No, that is fine.

6 DR. SANTANA: So, we will move on then

7 from question one--oh, Malcolm, I am sorry.

8 DR. SMITH: I think those are good

9 principles but I think one can get a bit more

10 detailed without being prescriptive in terms of

11 what the principles of study monitoring should be.

12 For example, the principle that study monitoring

13 should be performed by experienced experts and that

14 that review should be timely, and that whatever the

15 system is, it should have those characteristics.

16 And, study monitoring should be done in a way so

17 that conflict of interest issues are addressed, and

18 that study monitoring in whatever setting,

19 especially in Phase III settings but even in Phase

20 II settings and others is done in such a way that

21 the integrity of the study and the confidentiality

22 of data, when that is important, are addressed.

23 So, I think there are principles of ethics that we

24 need to adhere to and there are principles of

25 monitoring that I think need to be clearly stated

162

1 so that you can benchmark how you are addressing

2 those basic principles of monitoring.

3 DR. GRILLO-LOPEZ: If I may, most of that

4 is already covered in GCP and in other regulations.

5 DR. SANTANA: So noted. I would only add

6 to that that I think an essential element to that

7 is this concept that I advocate, that there has to

8 be an open communication with the research team,

9 that monitoring doesn't occur in isolation from the

10 actual research team that is conducing the study.

11 I am not implying that the research team should be

12 doing their own monitoring. It shouldn't be

13 interpreted that way but the research team should

14 be integral to that process. Dr. Reynolds?

15 DR. REYNOLDS: Malcolm, could I just ask

16 you to elaborate on what the role of that DSMB is

17 in the conflict of interest monitoring that you

18 were talking about?

19 DR. SMITH: What the role of the DSMB is

20 in conflict of interest?

21 DR. REYNOLDS: Did I hear you correctly?

22 Were you saying that they are really involved in

23 that role?

24 DR. SMITH: No, that the monitoring is

25 done in such a way that conflict of interest issues

163

1 are addressed.

2 DR. REYNOLDS: In other words, that the

3 DSMB is a separate body and is not subject to

4 conflict of interest. That is what you are saying?

5 DR. SMITH: Well, that is one way of

6 addressing it but not the only way of addressing

7 conflict of interest issues, but that those issues

8 are considered, both the financial and intellectual

9 conflict of interest that may lead people to ask

10 questions about decisions that are made.

11 DR. KODISH: This is Eric, in Cleveland.

12 Another way of saying that I think is that

13 transparency is an important principle, perhaps the

14 idea that whatever the monitoring plan is that the

15 appearance of the fox watching the henhouse won't

16 be something that people can interpret as having

17 gone on.

18 DR. SMITH: The Pediatric Phase I

19 Consortium and the Pediatric Brain Tumor Consortium

20 both have independent data monitoring committees,

21 and these are early phase clinical trials. They

22 are not so much looking over the day to day

23 activities of the consortium and every independent

24 decision, but at intervals they are looking at the

25 overall conduct of how these studies are being done

164

1 and are an independent body that tries to address

2 some of the conflict of interest issues, in this

3 case particularly intellectual conflict or kind of

4 ownership conflict issues, and to make sure that

5 the research team is appropriately making decisions

6 as they are conducting the studies. They are there

7 to provide guidance if difficult decisions arise

8 about what their advice would be about how to

9 address these difficult decisions.

10 DR. SANTANA: If there is no further

11 comment on that we will move on to number two. The

12 next series of questions are more related to

13 reality and practice. Recognizing that particular

14 populations, disease settings, and products may

15 have specific requirements, what general parameters

16 should be monitored for safety in all clinical

17 studies?

18 DR. HIRSCHFELD: I should say all

19 pediatric oncology clinical studies, just to be

20 clear about that.

21 DR. SANTANA: So noted. Peter?

22 DR. ADAMSON: I will take a stab at that.

23 I think it very much depends on the phase of the

24 study. In pediatrics I think we have some

25 advantages in that for Phase III studies there is

165

1 probably a general standard of care that we follow

2 whether a child is or is not on study as far as

3 frequency of monitoring. I would say that that

4 would probably be the minimum threshold for Phase

5 III studies.

6 As one marches down from Phase III to

7 Phase II and Phase I, I think this is where Phase I

8 cancer is different than Phase I "the rest of the

9 world" because we conduct the Phase I studies in

10 patients with the disease. So, I don't think you

11 can layer the same level of monitoring as you do in

12 other studies where volunteers are locked away for

13 two weeks and are plugged into every known device

14 to see what happens. We can't do that.

15 I think we need to look at preclinical

16 data as far as what potential toxicities are, and

17 in children we have the advantage of looking at the

18 initial adult Phase I experience to see what the

19 relevant additional monitoring might be required.

20 We shouldn't be getting PFTs, echoes, EKGs, stress

21 tests, all the way down the line if, in fact, that

22 is not relevant to a particular drug. So, I think

23 we have the advantage of looking at the Phase I

24 adult experience. Then, we always have to balance

25 the level of monitoring, recognizing that these are

166

1 patient volunteers and not normal volunteers as far

2 as trying to strike a balance.

3 DR. SANTANA: Pamela?

4 MS. HAYLOCK: I am not sure how relevant

5 this is but you keep talking about monitoring and I

6 think a lot of this has to do with expanding the

7 definition of safety and monitoring in regards to

8 concepts that involve long-term and late survivors.

9 Your institution is maybe somewhat unique in having

10 long-term survivorship programs, but not all places

11 which do pediatric research have such things, and

12 now we are ending up with adult survivors of

13 childhood cancers who are 10, 20, maybe 3 or 4

14 decades out who are experiencing surprise long-term

15 related effects and I think somehow the parameter

16 of safety and monitoring needs to be expanded. I

17 don't know how to do that but I think the late

18 effects need to be a consideration.

19 DR. SANTANA: Actually, cooperative groups

20 and other pediatric consortia are addressing that.

21 I mean, I think there is a big effort at the

22 cooperative group level to look at long-term

23 survivor issues in pediatric oncology patients.

24 Obviously, it is in different stages but I think we

25 all recognize as pediatric oncologists that that is

167

1 an issue, and I think it is being addressed at

2 different levels. Malcolm and then Donna?

3 DR. SMITH: It is a critical issue. The

4 challenge with it is that you are looking 10, 20

5 and 30 years up the road so the infrastructures,

6 like the children's hospitals around the table,

7 need to reach out to a lot of other institutions

8 and to the survivors in order for that work to be

9 done. So, there are different ways that the

10 Children's Oncology Group, the childhood cancer

11 survivor study are trying to address that, and it

12 is recognized as an important issue that we have to

13 address.

14 DR. SANTANA: Donna?

15 DR. PRZEPIORKA: I just wanted to ask, the

16 organized groups and the major institutions clearly

17 have a plan but what about industry? I mean,

18 industry does do pediatric trials. What sort of

19 guidance do you give to them, and what is the basis

20 for that guidance? I mean, what has come out of

21 the St. Jude experience monitoring long-term

22 survival in their patients, and is it really worth

23 mandating that the pharmaceutical

24 industry-sponsored trials do long-term follow-up?

25 DR. SANTANA: I think the issue of

168

1 long-term survivorship follow-up and data needs to

2 be considered by the pharmaceutical industry when

3 they are developing a drug in terms of the

4 long-term issues that may be particular to that

5 drug. The problem comes there that the sponsors

6 themselves are limited to a period of time in terms

7 of when they are doing the project with you. Once

8 the project is over, then the responsibility of

9 monitoring patients long term becomes the

10 responsibility of the treating institution. So up

11 front, at least in my experience in all the studies

12 that I have participated in with pharmaceutical

13 industry, I have never really seen, within the

14 context of the protocol, any plan for long-term

15 issues that may arise as a result of follow-up of

16 these patients. Once a study is done, it is done

17 and then it becomes the responsibility of the

18 treating institution to decide what they are going

19 to look for, how it is collected and how it is

20 analyzed. So, there is a little bit of a dis-link

21 there in that we have never really required or

22 asked pharmaceutical industry to address that in

23 the context of the front-line trial that is being

24 developed. Peter?

25 DR. ADAMSON: Again, pediatrics in this

169

1 respect differs from adults because where you

2 really get the long-term ability to look at late

3 effects is in or following Phase III. I am not

4 aware of any industry-sponsored Phase III studies

5 in pediatric oncology. They are almost universally

6 done within the cooperative groups. There are

7 industry-sponsored Phase I and Phase II studies,

8 without question. I think our ability to really

9 ask late effects questions in that population is

10 severely limited so it really becomes the burden of

11 the NCI and the cooperative groups when conducting

12 Phase III trials and, as Malcolm said, there is a

13 whole separate late effects effort. So, I don't

14 think it is something that realistically we can

15 burden industry with because of the likelihood of

16 getting that data in a Phase I or Phase II study.

17 If the environment were to change and we would

18 dream that industry would support a Phase III

19 randomized study in children, then I think we would

20 have to look at the willingness to look for

21 long-term effects.

22 DR. SANTANA: I will correct myself. I am

23 aware of one study that I have seen, which is an

24 antibiotic study that is actually being sponsored

25 by industry, looking at some issues of long-term

170

1 effects of the use of that antibiotic in a

2 pediatric population. It is a very long-term

3 study. It is a very costly study too. So, I am

4 aware of that example that came to mind as I was

5 hearing the discussion but that is kind of unique.

6 DR. ADAMSON: And it is not

7 anti-neoplastic therapy.

8 DR. SANTANA: No, it is not. It is an

9 antibiotic study. Any other guidance we can give

10 you on this question, Dr. Hirschfeld or Dr. Pazdur?

11 Yes?

12 MS. HOFFMAN: I think integral to

13 monitoring safety also in terms of when a child is

14 on treatment is also monitoring participation and

15 entering into the study, and I think we need to

16 monitor informed consents and parents'

17 comprehension of randomization, especially in Phase

18 I studies. Are they really understanding what they

19 are getting into? Also, monitoring waiver of

20 consents because I think there is potential

21 conflict of interest there. The waivers that are

22 coming to the IRB are coming from the PI who is

23 often the clinician as well of the child and,

24 again, there could be conflict there. So, again, I

25 think it is a safety monitoring issue.

171

1 DR. SANTANA: I will try and summarize

2 what I heard as committee discussion of this

3 question. I think the committee was pointing out

4 that in a certain way we have a little bit of an

5 advantage in that there may be some adult data

6 before pediatric studies are initiated, and a lot

7 of the safety issues and monitoring that we would

8 want to do in pediatrics have to be put in the

9 context of what data already exist in the adult

10 population that has received those drugs, but also

11 considering that there may be specific niches that

12 pediatrics would provide that we have to look for

13 that may not have been identified in the adults. I

14 heard that comment.

15 I heard the other comment, that it has to

16 be developmentally phase dependent in terms of what

17 type of study you are talking about, that the issue

18 of safety monitoring is very different in a Phase

19 III trial than it would be in a Phase I, and that

20 there are different mechanisms of reaching those.

21 In a Phase I it may be more the research team, the

22 consortium group continuously looking at that data

23 and making safety judgments, whereas in a Phase III

24 it may be a DSMB or may be other regulatory bodies

25 that can define what safety issues need to be

172

1 looked at and how they are evaluated. I heard that

2 comment.

3 I think the third comment I heard was

4 about this issue of paying some attention to the

5 initial enrollment of patients on studies,

6 pediatric oncology studies, and how we can more

7 effectively not only monitor their involvement but

8 get some degree of understanding of what people

9 really are hearing and their assessment of risk and

10 what they think they are participating in.

11 Those are the three comments I kind of

12 heard around the table. Susan?

13 DR. WEINER: I have one more, which is

14 that I really haven't heard any discussion this

15 morning of the notion of safety in trials of

16 biologics where toxicity may not be what you are

17 looking for in a Phase I trial, and it is not clear

18 to me how we might approach that in this context.

19 DR. SANTANA: That is a good point.

20 DR. HIRSCHFELD: Noted.

21 DR. KEEGAN: I think you also should

22 consider that it may be toxicity, it may be other

23 examples but one shouldn't exclude the fact that

24 toxicity could also be a component even in biologic

25 trials.

173

1 DR. GRILLO-LOPEZ: I was just going to

2 reinforce what Dr. Keegan said. You know, in the

3 past two years having developed two biologics, they

4 were both associated with some toxicities that were

5 important. So, one has to be careful, going into

6 the development of a biologic, not to think that

7 there might be fewer, lesser toxicities. So, one

8 really has to do the same monitoring that one would

9 do for a chemotherapeutic agent until one is sure

10 of what the toxicity profile is for that particular

11 biologic.

12 DR. WEINER: Or expand those definitions.

13 DR. HIRSCHFELD: I think we all agree that

14 the spectrum and the severity may vary but there is

15 no intervention that is risk free.

16 DR. KEEGAN: Yes, I think the principles

17 Dr. Adamson mentioned were, you know, looking at

18 the nonclinical and adult data to guide what would

19 be used for biologics and even for a lot of

20 traditional drugs, you know, small chemical drugs

21 that are targeted in some way.

22 DR. SANTANA: Yes, I want to add that

23 there was another point that was made as a general

24 consensus point as advice to the agency that had to

25 do with the issue of neurocognitive development,

174

1 and that that may be a particular issue in terms of

2 safety that should be addressed in safety

3 parameters in pediatric oncology trials. In

4 contrast to some of the things that we could

5 capture from adult trials, that is particularly

6 unique to pediatric trials and we should pay some

7 attention to it. Donna?

8 DR. PRZEPIORKA: Actually, just to

9 follow-up on that, the one other piece of

10 information that I think is very easy to obtain and

11 to analyze is growth.

12 DR. SANTANA: Any further comments on that

13 question? If not, we will move on to the next

14 question. Based on the response to the previous

15 question, how often should the parameters be

16 monitored?

17 Here I would say I think we need to be

18 careful. We don't want to get into a prescription

19 plan that everybody has to do kind of in the same

20 way in terms of what things get monitored, at what

21 particular time intervals and how often. I think

22 the idea that I proposed when we looked at our plan

23 at our institution is that it is phase dependent.

24 Once again we go back to the phase issue of the

25 type of study that you are conducting. So,

175

1 particular Phase I studies may be monitored more

2 frequently than other Phase I studies. Maybe some

3 biologic studies, gene transfer studies that are

4 Phase I need to be monitored more frequently than

5 an oncology Phase I study.

6 The point I want to make is that although

7 it is phase dependent, I think also in the formula

8 has to be included the specific agent that you are

9 testing in that phase in order to decide how often

10 you are going to monitor it. Peter?

11 DR. ADAMSON: Yes, I would echo that.

12 Again, going back to the adult experience, it gives

13 us an advantage as far as what to expect and when

14 to expect it. But the other thing that we

15 sometimes err on is that we have to look at our own

16 definitions of toxicity and what we consider either

17 serious or dose-limiting because when you look at

18 those definitions, you then look at how frequently

19 you are monitoring and you realize you will never

20 be able to meet those definitions. So, as I said,

21 perhaps a simple starting place if you want to get

22 some idea of what the spectrum is, there a number

23 of cooperative groups or a number of single

24 institutions that conduct this and my guess is you

25 will find a common thread in the backbone of those

176

1 that apply across the board for Phase I and a

2 different set for Phase II and then it becomes very

3 agent dependent beyond that.

4 DR. PAZDUR: I have a question as far as

5 the toxicity criteria for children, are there any

6 differences between that and what we use for

7 adults, other than perhaps physiological

8 differences that may exist with growth parameters?

9 What I am after is some of our adult toxicity

10 criteria have some subjective elements as far as

11 elements of daily activity, fatigue, etc., and how

12 do you figure that into toxicity assessments with

13 children? Or, do they have difficulty in assessing

14 some of these toxicities in children? You know,

15 for some of our activities for adults neurotoxicity

16 might be difficulty in buttoning your shirt or in

17 adult activities of daily living in a sense.

18 DR. SANTANA: Alice, it looks like you

19 wanted to respond to that.

20 MS. ETTINGER: Well, I think we all

21 understand that for kids we have to look at them at

22 an age appropriate level and many times that would

23 be school attendance, how they are functioning in

24 school, certainly measurements of that sort. I

25 think in terms of fatigue, we are way behind in

177

1 measuring the actual fatigue level that we may be

2 seeing in children, not only little ones but

3 certainly as they grow up. Often in filling out

4 the forms for doing the criteria, I feel that there

5 may actually need to be other criteria that we look

6 at and that we measure for children.

7 DR. SANTANA: It is a good point. The

8 issue with those criteria is that as yet they

9 haven't been validated so it is very hard to apply

10 them across studies but there is actually a lot of

11 research going on in that field that, hopefully, in

12 the next few years will give us some guidance. But

13 that is the problem, those criteria are soft and

14 they haven't been validated so it is very hard to

15 apply them. So, in oncology we kind of rely on the

16 standard toxicity criteria that was developed by

17 NCI, etc., in terms of what we look for and how we

18 code it.

19 DR. HIRSCHFELD: I will just add to that.

20 There have been questions raised about having some

21 pediatric specific scales, but it was the absence,

22 as Dr. Santana pointed out, of having validated

23 assessments that has precluded from formally

24 incorporating those. So, that is an area that

25 still remains under discussion and has had some

178

1 interest for some years.

2 DR. ANDERSON: And in the current version

3 of the CTC, the updated version that just came out,

4 where possible, all distinctions between pediatric

5 and adult criteria were eliminated because

6 basically we generalize the grading. I can't

7 remember exactly what word you used, Dr. Pazdur, in

8 terms of the degree of toxicity. You know, just

9 having treated patients with different pediatric

10 cancers and actually having heard from people who

11 are trying to set up studies with certain

12 dose-limiting toxicities, in pediatrics a lot of

13 times I think the dose-limiting toxicities that we

14 accept are greater than are accepted in adults.

15 They will stop an adult trial or they will change

16 an individual adult's treatment much sooner than we

17 do in pediatric oncology and I don't know that we

18 have different measurements of toxicity but we

19 would move a grade further perhaps, or half a grade

20 further in terms of maybe the duration of the

21 tolerance of a toxicity than happens in medical

22 oncology.

23 DR. SANTANA: And those are usually

24 specifically defined within the context of a

25 protocol. So, for some studies we would accept up

179

1 to grade X and in others we wouldn't. So, I think

2 there is a lot of variability and it is really

3 driven by the protocol and the question you are

4 trying to answer and what you know about that drug

5 beforehand.

6 The next question is based on the response

7 to the previous question, who should do the

8 monitoring? Is it adequate to have the personnel

9 involved in the study be responsible for safety

10 monitoring? Susan?

11 DR. WEINER: The issue of the conflict of

12 role between the investigator and the treating

13 physician is something that has been discussed over

14 the past few years in a variety of contexts. I

15 think that applying that notion to this, it becomes

16 obvious that such a team is insufficient.

17 DR. SANTANA: Peter?

18 DR. ADAMSON: I guess I would disagree

19 with that to an extent. It very much depends I

20 think on the phase of the study, and the number of

21 children who are at risk, and what the goals of the

22 study are. From a practical standpoint, for a

23 Phase I where the study is a real-time study that

24 is the role of the study team. They are making a

25 decision on a patient to patient basis. Having an

180

1 additional layer of oversight to make sure the

2 study team is meeting its obligations I think is

3 helpful and is important but, from a practical

4 standpoint, you can't convene a data safety

5 monitoring board with every dose escalation step;

6 you never would end up conducting the study.

7 Having said that, it is important to keep

8 in mind that the goal of a Phase I study is to

9 recommend a dose and so the study is going to be

10 successful really no matter where you stop as far

11 as an investigator conflict of interest. I mean,

12 they will meet their study endpoint. Having said

13 that, when you come to a Phase III, you really do

14 need additional layers of monitoring because then

15 you really want to prove is this drug effective and

16 there is a lot riding on the outcome of that study.

17 So, the level of monitoring I think very

18 much depends on what the phase of the study is.

19 But I think, without question, you need to know

20 what the data safety monitoring plan is. I mean,

21 investigators need to be very clear and very

22 specific up front about how this study is going to

23 be monitored. I will come back to what Eric Kodish

24 had said earlier, you need to have some flexibility

25 as far as what the level of monitoring is and who

181

1 does it. If it is a cytotoxic and there is a lot

2 of experience in developing the cytotoxic, that may

3 lead to one level. If it is an entirely new

4 modality of treatment being put forward, you may

5 want to consider another layer of monitoring. So,

6 there has to be some flexibility within the system.

7 DR. SANTANA: I would echo what Peter

8 said. I think it is a graded system and it depends

9 on the type of study you are doing and what

10 elements are being monitored. For example, if you

11 want to get into the nitty-gritty details of

12 monitoring enrollment and informed consent, I think

13 that has to be independent of the research team.

14 There is no other way you could do that; it has to

15 be a separate monitoring group that does that,

16 whether it is the protocol office or another group

17 of people. But in a Phase I study if the central

18 question is the toxicity, that should be monitored

19 by the study team because that is what is going to

20 define how the study progresses. Then you may have

21 intervals in which that data is shared with a

22 central Phase I group, etc., etc.

23 Whereas, in a Phase III study you are in a

24 completely opposite direction. For a Phase III

25 study most of the elements for safety that you want

182

1 to monitor have to be done independent of the

2 investigator. They are large group studies with

3 data collection. There may be some safety issues

4 that have to be reported to the safety data

5 monitoring boards so you have to use those

6 resources.

7 So, I don't see it as black and white. I

8 see it as a graded system in which the elements

9 that are going to be monitored, the safety and how

10 that is done may incorporate different groups and

11 you just have to find the right fit for the study

12 that you are considering. I hate to put it in

13 black and white; it won't work if it is black and

14 white. I think the beauty of some of the stuff

15 that Peter mentioned in terms of what the Phase I

16 and the COG Consortium is doing is that they are

17 doing it in real time. I mean, they are looking at

18 that week by week, maybe two weeks or however

19 often, so they have the advantage of doing that in

20 real time so that they can intervene if they have

21 to. Whereas, I think that would be impossible to

22 do in a Phase III study. You just couldn't get

23 people to do that. Dr. Reynolds?

24 DR. REYNOLDS: Peter, acknowledging the

25 challenges you put forth that a data safety

183

1 monitoring board in a Phase I study--that it is not

2 practical for them to convene and review, I think

3 we should acknowledge that there are some

4 significant advantages to having such a board for

5 the day to day people that are monitoring to go to

6 with questions about study design amendments that

7 might make it more acceptable from a safety

8 standpoint, and having that group that is external

9 to the people who are actually conducting the

10 study. It is a small world in pediatrics, so

11 having that separated out, at least from the NANT

12 perspective, is a great advantage in being able to

13 bounce things off these people externally.

14 DR. SANTANA: Dr. Smith?

15 DR. SMITH: We talked about NANT trials,

16 COG trials and we are very restricted to that.

17 Would there be a separate answer for

18 industry-sponsored Phase I/Phase II trials? Is

19 that a different situation?

20 DR. SANTANA: Usually in Phase I

21 industry-sponsored trials, at least the ones I am

22 familiar with, there is a research team that is

23 identified. It is usually the PI at various

24 institutions; it is a medical officer or monitor

25 from the pharmaceutical company or contact person.

184

1 I think the same functional principle can be

2 applied, that that research team should communicate

3 frequently and often enough as the study is being

4 conducted to make ongoing decisions about the

5 safety of the study. So, I think that may already

6 be happening. We just don't know about it. If it

7 is not happening, we should probably extend those

8 things that we are doing in some of these consortia

9 to those. I think they are practical and they

10 don't require a lot more work.

11 DR. GRILLO-LOPEZ: If I may expand on what

12 you said, which is absolutely correct, there is a

13 research team in a pharmaceutical

14 industry-sponsored study. Beyond that team, within

15 the company itself, there is also the equivalent of

16 a data monitoring board which usually consists of

17 the project clinician, the safety officer and the

18 statistician as a minimum. The data is looked at

19 very frequently. In addition to that, there are

20 periodic presentations of the safety data to larger

21 committees within the company and then there is an

22 opportunity to also present that data, if there are

23 any red flags, to the scientific advisory board of

24 external advisors which usually meets three to four

25 times a year depending on the situation.

185

1 DR. SANTANA: Donna?

2 DR. PRZEPIORKA: Actually, it sounds like

3 industry has a separate oversight; the organized

4 groups have a separate oversight; NCI-sponsored

5 studies will have a separate oversight. What we

6 haven't discussed is individual

7 investigator-initiated studies at single

8 institutions. I think under those circumstances it

9 might not be too disruptive to say, you know, at

10 some point see if there is somebody who can give

11 you an outside reality check before you go on to

12 the next level. It may not require convening an

13 entire board but just sending a member to the IRB

14 or to whatever institutional data safety monitoring

15 committee might be available.

16 But, you know, having conducted Phase I

17 studies, one can get lulled into, okay, I have five

18 more patients lined up; let's go to the next level

19 before I really have all the data collected on

20 safety. It may be just enough to actually improve

21 patient safety at that one institution.

22 DR. SANTANA: Yes, I am glad you mentioned

23 that. We tried to address that at St. Jude. As an

24 academic institution, we tried to address that too

25 with some of our own Phase I studies. So, we

186

1 operated very similarly to what the Phase I

2 Consortium is doing, and that is that if it is an

3 institutional Phase I study the research team meets

4 frequently to review, as the study is being

5 conducted, what the safety concerns are; what is

6 going on with the next escalation, etc., etc.

7 Then there are two separate groups that

8 also look at that. There is a separate Phase I/II

9 planning group that we have that includes

10 disciplines from solid tumors, leukemia,

11 transplantation and biostatistics, all the basic

12 science people and they are also supposed to meet

13 every month but in reality they probably meet every

14 six weeks and all the studies are also actually

15 presented very briefly. So, the whole group knows

16 where each study is going and what is happening

17 with toxicity; what is happening with issues of

18 accrual. That is not truly separate because it is

19 constituted by individuals from the same

20 institution.

21 The third layer is that even for Phase I

22 studies-- if you saw in my flow diagram where data

23 went, all the adverse events, independent of any

24 type of study, also get reviewed by the clinical

25 protocol scientific review group subcommittee which

187

1 does not include any of the Phase I PIs. They also

2 make a judgment in terms of how that study is

3 going; in terms of dose escalations; in terms of

4 safety. So, it is very similar and kind of a

5 little bit of recapitulation of what the

6 cooperative group is doing in terms of having other

7 people look at it. It is not totally independent

8 in the sense that there is an outside group that

9 looks at it.

10 Having said that, also in some Phase I

11 studies, like the gene transfer studies--when we

12 get to the question of DSMB committees I was going

13 to mention that, we have a definition of what gets

14 referred to DSMB and one of the definitions is if

15 there is a Phase I study that includes gene

16 transfer or a biologic that is potentially

17 problematic, that will go to the DSMB although it

18 is a Phase I study. Barry and then Susan?

19 DR. ANDERSON: Also, being part of this

20 review board at the cancer center data safety

21 monitoring plan, anybody who is receiving a grant

22 that might involve a clinical study as an

23 individual also has to provide a data monitoring

24 plan in order to receive the money for the grant.

25 DR. SANTANA: Susan?

188

1 DR. WEINER: Just a point of

2 clarification, just to make sure that the following

3 case is covered for Phase I and perhaps for Phase

4 II in pediatrics, let's say a network of

5 institutions that are doing combination therapy

6 trials, pharmaceutical trials, and they are not

7 being supported by NIH--presumably the institutions

8 have assurances, etc., but the monitoring of that

9 particular kind of trial.

10 DR. SANTANA: Do you want to address that

11 because it is primarily coming from the issue of

12 industry-sponsored small trials within two or three

13 institutions? Am I correct, Susan?

14 DR. WEINER: Or more.

15 DR. SANTANA: Or more. Do you want to

16 address that?

17 DR. GRILLO-LOPEZ: From the safety point

18 of view, they are monitored in exactly the same way

19 that I mentioned earlier.

20 DR. WEINER: Well, just in terms of the

21 external terms. So, the company sets up some

22 external monitoring to review safety concerns--I

23 mean, if it is two drugs--

24 DR. GRILLO-LOPEZ: Well, if it is a Phase

25 I or Phase II trial usually there is no external

189

1 review, external to the company review, other than

2 that the company has to report to the FDA. So,

3 that is an external third party. Also, the company

4 has the possibility of presenting the safety

5 information to the scientific advisory board which

6 is also an external review board.

7 If there is a Phase III randomized study,

8 particularly a blinded study, most companies are

9 opting to have an external independent data safety

10 monitoring board following that study, or if it is

11 a Phase II trial that is already randomized and

12 blinded.

13 DR. SANTANA: Any other comments or

14 questions? Then we will move on to the next

15 question which is asking us for advice on what

16 circumstances would benefit from a data monitoring

17 committee/data safety review board oversight?

18 To try to address that, I think Barry had

19 in one of his slides what some of the

20 recommendations are from NCI regarding--or was it

21 COG? I don't remember that.

22 DR. ANDERSON: Recommendations from NIH.

23 DR. SANTANA: Do you want to expand on

24 those, Barry?

25 DR. ANDERSON: In pediatrics the default

190

1 seems to have some sort of monitoring committee, a

2 more formalized monitoring committee because the

3 recommendations were if they were complex--and if

4 you have ever looked at an ALL study or anything

5 else, they are pretty complex, and every study

6 basically, if it is multi-institutional, which

7 pediatrics for the most part usually are--if it is

8 a vulnerable patient population, and we have our

9 own separate part of the regulations just because

10 pediatrics is a vulnerable patient population, and

11 high-risk treatments--you know, a lot of the

12 treatments that we use with stem cell transplants,

13 etc., etc., are high risk. So, because of all

14 those issues coming up in a lot of cases, a data

15 monitoring committee is involved. It may be

16 different than a DSMB that you were talking about

17 for a Phase III randomized study because some of

18 these monitoring committees also work for

19 single-arm studies that may have early stopping

20 rules that need to be interpreted, and that sort of

21 thing as well.

22 DR. SANTANA: I would add two additional

23 items to the list that Barry proposed. As an

24 institution, there are two other types of studies

25 that we would refer for an independent data safety

191

1 monitoring board. One is any study that involves

2 any type of gene transfer or biologic that

3 potentially could present a hazard to children in

4 the future. Then, the second is a very unique type

5 of study which is what we call the window study

6 where an experimental therapy is given prior to

7 conventional therapy and there is a limitation of

8 time in which you can really do that to provide

9 safety for the patients. So, those kind of studies

10 we would also refer to DSMB to provide oversight.

11 Any other comments or questions on that?

12 Yes?

13 DR. GRILLO-LOPEZ: A clarification, when

14 you made your presentation you mentioned the makeup

15 of your data monitoring board and you said it was

16 the staff involved in the study itself, the

17 principal investigator and perhaps some others

18 around the principal investigator, and then some

19 additional members outside of that group. But

20 should I interpret "outside" as within the

21 cooperative group or completely external to the

22 cooperative group?

23 DR. ANDERSON: It depends on whether you

24 are talking about a DSMB or a DMC. I mean, there

25 has been some distinction there. The DSMBs would

192

1 be probably reflective of what industry uses when

2 they have an outside independent one. For the COG

3 DSMBs there is a member or maybe two members of COG

4 that are part of that but there are statisticians

5 from other adult cooperative groups. There are

6 outside lay people that are part of it, and there

7 is a government representative there. It is set up

8 such that the vote could never be carried by COG

9 members. And someone who would be perhaps a study

10 investigator for a particular Phase III study, they

11 would not be involved in discussions of their study

12 if they happened to be also a COG representative to

13 the group.

14 For other data monitoring committees--I

15 can't speak for Peter's group but for the NANT that

16 data monitoring committee has one representative

17 from that group or institutions that are conducting

18 these early phase studies. A number of people are

19 COG members but they don't participate in these

20 studies. There are other people who are retired

21 pediatric oncologists. We have statisticians from

22 outside the group. We have lay people from outside

23 the group. So, again, we are not looking at Phase

24 IIIs, we are looking at early studies. Again, the

25 predominant role is that you are outside of the

193

1 people who are doing the investigations. That,

2 again, is for the interval of about every six

3 months of formal review but also being there as a

4 resource ongoing.

5 The reviews that go on in the NANT group

6 sort of on a more frequent basis could involve

7 study investigators but it is usually the bigger

8 group of other investigators that are part of the

9 group but not responsible for that particular

10 study. So, there is some oversight in the sense

11 that it is within the group but it is not the

12 person who has the most vested interest that that

13 single study succeed in one way or another.

14 DR. GRILLO-LOPEZ: It is probably

15 worthwhile to mention that in industry today most

16 Phase I and II studies have an enrollment period

17 that ranges from 6 months to 12 months and perhaps

18 not more than that. So, the value of an external

19 data safety monitoring board is limited because of

20 your ability to actually give them trend

21 information and so on when you have actually

22 completed enrollment on the study.

23 DR. PAZDUR: I would just like to mention

24 that we have a draft guidance on data safety

25 monitoring.

194

1 DR. SANTANA: I think with that we will go

2 to the last question, which is an open-ended

3 question, are there additional recommendations for

4 safety monitoring? Peter?

5 DR. ADAMSON: I think the only one that

6 came up earlier is that institutions don't have

7 adequate resources to do this job well. That is

8 not unique to pediatrics but every layer of

9 monitoring that gets put on an institution and

10 investigator--you have to look if the resources are

11 there to truly meet it. I think in most

12 institutions the resources are inadequate right

13 now.

14 DR. SANTANA: I would echo that. I think

15 we started this morning's session with a comment

16 about stewardship and I think stewardship includes

17 financial resources so I think the regulatory

18 agencies need to be very cognizant that if we are

19 going to do this, there has to be a mechanism to

20 provide monies to do this well. There can't be

21 mandates without monies to actually carry this out

22 well. Susan?

23 DR. WEINER: I have one additional

24 comment, and that is that I think that the term

25 "lay member" is fine but it seems to me that when

195

1 one is reviewing pediatric trials there really

2 ought to be a family member who is that lay person

3 to help assess the safety of the situation.

4 DR. SANTANA: Good point. Any other

5 comments? Any other guidance that the FDA wishes

6 from us on this session? If not, we are adjourned

7 for the morning. Thank you. We will try to

8 reconvene at about 1:15.

9 [Whereupon, at 12:25 p.m., the proceedings

10 were recessed for lunch, to reconvene at 1:20 p.m.]

196

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

2 DR. SANTANA: Let's go ahead and get

3 started with the afternoon session in which we are

4 going to talk about preclinical models and other

5 data that we could extrapolate in terms of helping

6 us design clinical studies. Before we get started

7 with the actual presentations, we need to go around

8 the table again and re-introduce ourselves because

9 there are new individuals who have joined the group

10 and, hopefully, not many others have left. So, can

11 we start with Dr. Anderson, please?

12 DR. ANDERSON: Barry Anderson, from NCI

13 CTEP.

14 DR. HOUGHTON: Peter Houghton, from St.

15 Jude Research Hospital.

16 DR. ADAMSON: Peter Adamson, The

17 Children's Hospital of Philadelphia.

18 DR. HELMAN: Lee Helman, Pediatric

19 Oncology Branch, National Cancer Institute.

20 DR. SMITH: Malcolm Smith, Cancer Therapy

21 Evaluation Program, NCI.

22 DR. GRILLO-LOPEZ: Antonio Grillo-Lopez,

23 Neoplastic and Autoimmune Diseases Research

24 Institute.

25 MS. HAYLOCK: Pam Haylock, oncology nurse

197

1 and ODAC consumer representative.

2 DR. PRZEPIORKA: Donna Przepiorka,

3 University of Tennessee, Memphis.

4 MS. CLIFFORD: Johanna Clifford, executive

5 secretary for this meeting. I am just curious, is

6 Eric Kodish still on the line? No.

7 DR. SANTANA: Victor Santana, pediatric

8 oncologist at St. Jude's Children's Research

9 Hospital, in Memphis, Tennessee.

10 DR. REYNOLDS: Patrick Reynolds,

11 Children's Hospital of Los Angeles.

12 MS. ETTINGER: Alice Ettinger, nurse

13 practitioner at St. Peter's University Hospital in

14 New Jersey.

15 DR. WILLIAMS: Grant Williams, Oncology

16 Drugs.

17 DR. KEEGAN: Pat Keegan, Oncology

18 Biologics.

19 DR. HIRSCHFELD: Steven Hirschfeld, FDA.

20 DR. DINNDORF: Pat Dinndorf, Oncology

21 Biologics.

22 DR. DAGHER: Ramzi Dagher, Division of

23 Oncology Drug Products, FDA.

24 DR. SANTANA: Thank you. With that, we

25 will go ahead and get started with the first

198

1 presentation, Dr. Paul Meltzer.

2 What are Microarrays and How Can They Help Us

3 with Clinical Studies in Pediatric Oncology

4 DR. MELTZER: What I am going to do is to

5 very quickly give the members of the committee a

6 tour of some of the clinically relevant

7 applications of genomic technologies involving

8 microarrays which may have a bearing on some of the

9 issues that you are considering today. I will do

10 that in the fashion of a very brief overview of

11 technology in a few specific examples, and give you

12 my impression of some of the issues that would have

13 to be overcome for this information to be evaluable

14 in clinical trials.

15 Array technologies have now been around

16 for several years, and the ones that I am going to

17 talk about mainly today are actually becoming

18 rather mature, and it is now possible to generate

19 data with these technologies which can be

20 considered sort of archival quality that will serve

21 as a long-term source of information about the

22 diseases that are being looked at.

23 There is some the excitement around these

24 technologies, as indicated by this slide which just

25 shows the number of citations in PubMed on

199

1 microarrays from the inception of the modern

2 technology for microarray expression profiling

3 through last years. There has been an exponential

4 growth in the number of publications that cut

5 across all areas of biomedical research. There has

6 been a tremendous amount of interest and activity

7 in data generation, importantly, for you to

8 consider.

9 The reason that this momentum has built up

10 has been based on the availability of the human

11 genome sequence which now allows a whole genome

12 approach to identifying the genes expressed in

13 tumor tissue samples or in the context of other

14 types of biological samples. Of course, this will

15 include drug targets and, indeed, it should include

16 every conceivable protein drug target, as well as

17 gene expression signatures which represent a

18 cellular readout that is associated with important

19 clinical or biological properties of cancers. I

20 will try to explain this concept with just a few

21 examples in a moment.

22 There are a number of different microarray

23 technologies and I am just going to be touching on

24 the two that are underlined because these are the

25 ones that are in most widespread use today really

200

1 throughout the world. At the top of the list, and

2 mainly what I will be talking about, is expression

3 profiling, measuring the expression of large

4 numbers of genes in parallel in a given biological

5 sample.

6 It is important to note that there are

7 other array technologies coming along which are

8 likely to have a role of some type in clinical

9 application, and that includes microarrays to

10 determine DNA copy number in tumors, or CGH arrays,

11 microarrays which can determine DNA polymorphisms,

12 commonly referred to as SNP chips. I am going to

13 touch briefly on tissue microarrays because they

14 have emerged as a very important confirmatory

15 mechanism for the RNA-based expression arrays which

16 are also potentially of clinical importance. Of

17 course, protein arrays, various forms of

18 proteomics, are important and I am not going to

19 talk about that.

20 It is important for you to realize that

21 there is a tremendous amount of gene expression

22 data, mainly from adults, which has already been

23 generated with these technologies, and a great deal

24 of this is already publicly available in databases

25 that are universally accessible.

201

1 So, this is just what one form of

2 microarray looks like, basically a glass microscope

3 slide on which DNA probes have been deposited. I

4 won't dwell on the technology, other than to point

5 out the important feature, and there are several

6 different embodiments of the technology but the

7 important feature is that we now can look at the

8 entire human genome, or animal genome if you are

9 talking about an animal model, cramming in the

10 entire genome on a single microarray chip and it is

11 possible to interrogate this chip, to use it to

12 interrogate a biological sample to look at

13 expression of all the different genes in the human

14 genome in a biological specimen. For those of you

15 who are into gene expression, you know that there

16 are subtleties involving, let's say, splice

17 isoforms and, indeed, that is being looked at with

18 this type of technology as well. So, you can

19 really get a very detailed picture of expression

20 across the genome at the RNA level with this

21 technology, and one that is actually remarkably

22 accurate and carries with it quite a nice snapshot

23 of an individual biological sample.

24 So, what are some of the potential

25 connections between this information and cancer

202

1 therapeutics? The first I would mention is to

2 increase the precision in tumor diagnosis to

3 complement additional pathologic techniques, and

4 perhaps to identify and define subsets that haven't

5 been previously recognized in previously thought to

6 be homogeneous tumor groups; to measure the

7 expression of drug targets; to recognize

8 signatures, and I will expand on this in a minute,

9 which might be associated with the activity of a

10 particular drug target; to identify features in the

11 gene expression profile which might be related to

12 drug sensitivity or resistance; and potentially to

13 monitor or predict toxicity.

14 Now, there are subtleties and, in fact, it

15 is actually an extremely complex topic, the

16 analysis of microarray data that I am not really

17 going to touch on, but it is important for me to

18 point out to you that aside from simply scoring in

19 a simple kind of plus/minus way for the presence of

20 a given gene or target, all of these types of

21 analyses require a training set of tumors to

22 identify the relevant genes and to develop a

23 scoring algorithm which can be used to look at

24 these various types of readouts.

25 Another very important feature of this

203

1 data is that if you have full genome data it is

2 comprehensive. It is intrinsically comprehensive.

3 There are only so many genes in the human genome;

4 there won't be more in five years than there are

5 now, or in 10 years or in 20 years. That is why

6 the data has a very nice archival quality to it.

7 So, it can be reanalyzed in the future with respect

8 to novel targets or signatures that might be

9 identified so you basically have data that really

10 won't go stale so long as it is collected in a

11 state-of-the-art fashion and is appropriately

12 archived.

13 This slide just outlines the strategy that

14 is used in microarray studies. You start with the

15 whole genome and look at a very large number of

16 genes, so tens of thousands of genes across many

17 samples to develop profiles that occur in a

18 particular clinical situation. Then you go through

19 some process of gene selection to identify those

20 genes which separate tumors or patients into groups

21 according to the particular question that is being

22 asked, whether it be drug response, toxicity or a

23 diagnostic question, genes that are associated with

24 a particular target activity, and so on.

25 You then have to go through a process of

204

1 validation, frequently involving a new sample set

2 and reiterating this process to validate it, and

3 also probably validating it with other technologies

4 such as RT PCR, quantitative PCR or

5 immunohistochemistry or RNA in situ hybridization

6 or something like that to validate the results.

7 You might want to proceed to a clinical assay

8 development, and it is very important to point out

9 that much of the momentum in the development of

10 clinical assays based on this type of information

11 involves not microarrays but other forms of

12 multiplex gene analysis which might involve, for

13 example, a PCR-based method.

14 So, this is the overall approach and here

15 are going to be a couple of very quick examples.

16 This is from a study we published several years ago

17 identifying groups of genes that separate for

18 common pediatric cancers, Ewing's sarcoma and

19 neuroblastoma, rhabdomyosarcoma and Burkitt's

20 sarcoma. Color-coded here and at the top of this

21 clustergram, each of these little groups of red

22 squares represents groups of genes that separate

23 these groups of tumors and can be used to diagnose

24 them with a high degree of accuracy.

25 The important point about this slide is

205

1 that out of a large number of thousands of genes,

2 the genes that were necessary to give a perfectly

3 accurate call involved a very small number of

4 genes, about a 100 genes, 96 to be precise, which

5 were identified by a process of gene minimization.

6 So, that is the bottom line of everything that you

7 will see in the literature or hear about, that one

8 doesn't need tens of thousands of genes to answer a

9 question. If it is possible to answer it, usually

10 a very small number of genes, less than 100, will

11 be sufficient to accomplish what you want and

12 sometimes as few as two.

13 I am going to give two quick examples that

14 illustrate these features in detecting therapeutic

15 targets by microarrays, one in gastrointestinal

16 stromal tumor, or GIST, and the other is breast

17 cancer which involved a couple of studies that were

18 from our lab.

19 In the case of GIST, here were are seeing

20 the separation of gastrointestinal stromal tumor

21 from non-GIST sarcomas with, again, the minimal

22 number of genes necessary to establish the

23 separation. The important point for today's

24 discussion is that when we looked at the top genes

25 we found that the KiT oncogene was actually the

206

1 number one gene. So, we both could score the

2 presence and assess its relative importance in

3 characterizing this particular tumor in one

4 process, and one can do this in respect to any

5 property of a tumor that you choose. So, this is

6 an example of scoring a single gene out of

7 microarray data.

8 If you will forgive me for introducing an

9 adult example, I will now give you an example that

10 indicates how you might work--

11 Oh, this is just to show how KiT looks on

12 a heat map of GIST versus non-GIST. You see this

13 very uniform pattern of KiT expression.

14 I will give an example now of how you

15 would look at a signature for gene expression using

16 the estrogen receptor in breast cancer which, of

17 course, is a very nice molecular target widely used

18 in breast cancer therapy. The point here is that

19 there is a distinct pattern of gene expression in

20 breast cancer that separates the positive from

21 negative tumors very sharply, and everybody who

22 looks at these tumors has found exactly the same

23 result. It is the strongest feature in gene

24 expression profile of breast cancer.

25 Importantly, it is possible to actually

207

1 predict the value of the protein measurement for ER

2 in a tumor specimen from the gene expression

3 profile using a number of genes to make that

4 prediction excluding the estrogen receptor itself.

5 So, you can actually plot on this figure the actual

6 ER level in the little magenta circles, and the

7 predicted value based on the gene expression

8 profile based on a group of several hundred genes

9 in these tumors. So, there is a signature that

10 goes with the presence and function of this

11 particular drug target that can be read out using

12 multiple genes. Similar observations have been

13 made for other targets. So, this is an example of

14 a multiple gene predictor.

15 The bottom line here is that microarrays

16 can measure therapeutically relevant genes either

17 as individual genes or as complex signatures, and

18 expression profiling then can reveal both the

19 presence of a target and measure relative abundance

20 within the cell at the RNA level. Finally, a

21 signature related to target function can reveal its

22 level of biological activity, as in the ER example.

23 I just want to take a couple of moments to

24 talk about tissue microarrays because I think these

25 are very important and very accessible from a

208

1 technological point of view. A tissue mircoarray

2 is simply an array taken from paraffin blocks from

3 patient samples, assembled into an array which can

4 then be sliced to produce many slides that can be

5 assayed for various markers. The power of this

6 technology is that, in contrast to the DNA

7 microarray in which we measure thousands of genes,

8 for each tissue specimen in the tissue microarray

9 we can measure one gene in thousands of specimens

10 very rapidly. So, these are very powerful tools

11 for the validation of findings for genomic surveys

12 and potentially for translating them into clinical

13 studies.

14 Just to emphasize the tremendous advantage

15 that we gain from using these arrays, it arises

16 from taking a large number of paraffin blocks and

17 condensing them down into one very affordable,

18 economical package where we can survey single

19 tumors with a slice from any individual tissue

20 microarray. So, it is a very powerful technology

21 that I think can be quite useful.

22 So, how might these technologies be

23 implemented in clinical trial designs? I just want

24 to take a moment to give you some perspective.

25 First of all, to reiterate, detection of individual

209

1 targets is really simple. That is not difficult

2 and is very straightforward and should pose no real

3 challenge. However, in terms of using this for

4 pediatric cancers, we have a problem in that so far

5 only limited data is available on pediatric cancers

6 in the public repositories and that would be one of

7 the major obstacles. Indeed, very minimal data

8 exists relative to any question of toxicity, and

9 these are issues that are just beginning to be

10 seriously looked at in adults and, to my knowledge,

11 haven't been examined in children at all. As far

12 as I can see, implementing tissue collection

13 protocols and microarray analysis as part of

14 ongoing trials would be a necessity to overcome

15 this limitation.

16 Tumor tissue sampling is essential to get

17 a picture of the tumor but I am not sure that it is

18 necessary to have serial sampling. It would be in

19 principle nice to know what happens in the residual

20 tumors of patients who don't respond to therapy but

21 in principle this should be predictable from the

22 initial signature.

23 It is also interesting to speculate that

24 useful information regarding toxicity may

25 potentially be obtained from blood samples for

210

1 example, but the data to support this concept is

2 extremely limited at the present time.

3 Finally, again to reiterate, complex

4 questions such as the prediction of response or

5 toxicity require a training set and can't be

6 answered a priori or predicted from a bunch of

7 array data. So, if we want to talk about taking

8 array data from an archive and predicting what

9 might happen in those patients in response to a

10 particular agent, we really don't have a way to do

11 that at the present time. The only way we can

12 really examine that is to have samples annotated

13 with respect to that clinical question. So, that

14 is basically what I had to say. Thank you.

15 DR. SANTANA: Thank you. We will have

16 some opportunity during the discussion period to

17 address some questions. I think Dr. Peter Adamson

18 is next. Peter?

19 Advantages and Limitations of Cell Culture Models

20 in Pediatric Drug Developments

21 DR. ADAMSON: For those of you who

22 remember Monty Python and now for something

23 completely different, whereas microarrays are

24 approaching their tenth birthday, cell culture

25 models are probably approaching retirement age.

211

1 So, what I thought I would do is speak briefly

2 about some of the advantages and limitations of

3 these models. Historically, they have been

4 controversial as well as helpful. I think many of

5 the issues that occurred historically are still

6 issues today.

7 To really understand that, I want to take

8 you through a very brief history of cell culture

9 models in the context of drug development. In

10 looking back, probably the clonogenic assay is a

11 good starting point as far as how these models have

12 been used. This was work done by Hamburger and

13 Salmon, published back in 1977 in Science. What

14 they were able to show was that they could take a

15 number of primary human tumors and grow them up in

16 a cell culture matrix.

17 This is a photo micrograph from their

18 publication. Definition tumors have different

19 colony formations but the concept was that these

20 represented tumor stem cells, and stem cells were

21 the renewal source and they served as a seat of

22 metastatic spread, and cytotoxicity in this assay

23 was going to be proportional to cytotoxicity in

24 vivo. If you didn't get at the stem cell, you

25 weren't going to have an effective anti-cancer

212

1 treatment.

2 The way the clonogenic assay worked when

3 it came to cytotoxicity is you would expose your

4 culture media to various concentrations of drugs

5 and then look at the effect on colony formation,

6 look at the clonogenic assay.

7 Predating the clonogenic assay there were

8 other mechanisms to try to look at cell growth and

9 behavior in vitro. The tritiated thymidine assay

10 was probably the most common one. This was a

11 pretty straightforward approach where you would

12 tritiate thymidine and measure the incorporation

13 into dividing cells. It basically was a

14 measurement of S-phase cells and it quantified that

15 simply by counts per minute with a radioactive

16 label.

17 There were clearly limitations really to

18 both of these approaches. The clonogenic assay was

19 very labor intensive and there were a number of

20 investigators who, despite that hurdle, ran an

21 incredible number of assays looking for activity of

22 cytotoxic agents. But the reality was that it was

23 really not readily amenable to high throughput.

24 Conversely, the tritiated thymidine,

25 although there were the limitations of just using

213

1 the radioactive label, was also a non-clonogenic

2 method. You are looking really at a different

3 endpoint.

4 Then the field began to change and began

5 to change based on a paper by Mossman, an

6 immunologist, in The Journal of Immunologic

7 Methodology, in 1983 when he described what is an

8 assay familiar to almost everyone, the MTT assay

9 which was a colorimetric assay for cellular growth

10 and survival. In this assay a salt, MTT, when

11 incubated with viable cells in the mitochondria

12 undergoes a ring opening and produces a purple

13 salt, formazan. Then you solubilize this; you get

14 a purple color and you put this in a plate reader

15 and the intensity of the optical density is

16 proportional to the cell number. This assay really

17 began to change a lot of what was happening in the

18 world of cell culture and cytotoxicity.

19 Perhaps in part where it had a great

20 impact was at the NCI which, at this time, was

21 looking at moving from their historic way to screen

22 compounds for anti-cancer activity to what became

23 known as the NCI 60 cell line screen. This is a

24 typical output on a plot of logarithmic

25 concentrations of a drug as well as survival. As

214

1 many people have noted in the past, the 60 cell

2 line incorporated a number of

3 malignancies--leukemia, non-small cell, small cell

4 and so forth, but there was ne'er a pediatric

5 malignancy on this list. There were many efforts

6 made to try and change that and probably, in

7 hindsight, it was probably just as well that we

8 didn't.

9 Nonetheless, in the late '80s, early '90s

10 and even today there are a large number of

11 clonogenic assays that were based on the MTT, XTT.

12 The SRB assay, sulforhodamine blue, was the one

13 that the NCI eventually adopted; historically

14 trypan blue uptake in viable versus non-viable

15 cells; and the list goes on and on. Each of these

16 has various advantages and various disadvantages

17 but ultimately they are all measuring a very

18 similar endpoint and these are non-clonogenic

19 assays.

20 At this point it is helpful to step back

21 and say, well, what are non-clonogenic assays, when

22 it comes to drug development, really telling us?

23 What principles do they rest on? Taking some

24 liberties, I think these are the assumptions that

25 are made. As you can see, many of these

215

1 assumptions are supported by data, others less so

2 as we work down the list. But the non-clonogenic

3 assay is really a measurement of viable cell number

4 and almost all the non-clonogenic assays do that to

5 a reasonably good degree.

6 Many of these have been correlated which

7 is considered in vitro the gold standard, the

8 clonogenic assay. Again, not all of them, and it

9 is very cell line dependent how well that

10 correlates. But then one starts making larger

11 leaps. That is, that the clonogenic assay somehow

12 is correlated to in vivo cell growth and in vivo

13 cell growth that is somehow correlated to the tumor

14 growth in the patient. So, when you start up here

15 you have a long list to go down as far as what we

16 are asking an assay to do as far as being able to

17 predict or not predict what is going to happen in a

18 patient.

19 Let me talk about some of the potential

20 uses. I mentioned drug discovery and this is an

21 output from a more recent NCI screen. This has

22 advanced as far as the type of information that

23 comes back. There is a compare algorithm that can

24 talk about mechanism of action, and so forth, but

25 if you put it in the broader context of drug

216

1 discovery, this is not how drugs are discovered

2 today. I mean, in industry today you have a

3 target; you develop an assay for a target and you,

4 hopefully, have an assay that is amenable to high

5 throughput. For the most part, outside of the 60

6 cell line screen, this is not how drugs are being

7 discovered.

8 But cell culture models are still useful

9 in a number of areas. You can study cellular

10 pharmacology. You can certainly study mechanism of

11 action of drugs in these models, as well as

12 evaluate drug resistance.

13 Now, as pediatric tumor models, they have

14 historically and continue to serve at some level as

15 a screening for drug activity, but you can also ask

16 dose or, more appropriately, concentration schedule

17 dependent questions in cell culture models and one

18 can evaluate drug combinations in these models.

19 There are, not surprisingly, limitations.

20 Some of these limitations are unique to in vitro

21 models; some can be transferred over to in vivo

22 models. We know that cell lines undergo

23 transformation to allow for in vitro growth. For

24 in vitro drugs that require metabolic activation or

25 have active metabolites, you are likely to miss

217

1 that. You are not likely to be able to pick that

2 up given the nature of the in vitro model.

3 There are clearly potential differences in

4 drug exposures in these in vitro models. They can

5 range from differences in protein binding. Drug

6 disposition is incredibly difficult to try to model

7 in vitro. You basically dump the drug in and you

8 let it sit there for a period of time. That is not

9 what happens in a patient as far as how drug is

10 cleared. There are certainly differences in tumor

11 micro-environment or lack of vascularization and

12 hypoxia. There are methods, and Pat has looked at

13 some methods, to try to compensate for that in in

14 vitro models to try to better reflect what is

15 happening in vivo, and there are many other

16 limitations.

17 With that background, there are still some

18 advantages to these models. Relatively speaking,

19 these are not labor intensive models. They are

20 relatively low cost and they are amenable to

21 moderate throughput. In addition, because of

22 these, you have the ability to study multiple cell

23 lines and I think, perhaps as we move forward in

24 product oncology, the ability to study multiple

25 combinations of drugs.

218

1 One advantage of the in vitro model I

2 think over other models is that it is probably the

3 only model system that is mathematically amenable

4 to defining synergy, additivity or antagonism. It

5 becomes very complex in other systems to really

6 know if something is synergistic or not. There are

7 a number of accepted methods to do that in an in

8 vitro system.

9 So, let me start there and I am just going

10 to share three very basic examples of in vitro

11 models and what they can do, and I think I will be

12 commended then for picking up the pace as far as

13 getting us back on whatever time line we should

14 have been on.

15 The first one is determination of synergy.

16 I know folks in the room know this, there is a

17 problem with a simple addition method. If your

18 drug A kills 15 percent and drug B kills 25

19 percent, well then, if the combination kills more

20 than 50 percent it is synergistic. Well, it

21 doesn't take much to realize that you run into a

22 problem pretty quickly if drug A kills 70 percent

23 and drug B kills 70 percent. You can't simply add

24 them up. We can't just say, aha, it is

25 synergistic; it is more than the sum. That is what

219

1 we are sometimes left with, with in vivo models but

2 it is very difficult to know that. There are a

3 number of mathematical approaches and these get

4 debated constantly in journals that I don't like to

5 read--

6 [Laughter]

7 --but they do get debated. One of the

8 more accepted models is the median effect model.

9 There is now software that really can make this

10 very user friendly and straightforward. But if you

11 have different drugs you first look for a rational

12 effect as a concentration of dose and you do that

13 with one drug; you lay on the other and you lay on

14 the third, and then you realize you can't see what

15 is going on. So, you transform the data and you

16 get what is called a median effect plot. From the

17 median effect you can calculate what is called a

18 combination index. Please don't try to figure this

19 out from the graph, but let me tell you that the

20 software will basically tell you, yes, it is

21 synergistic or it is additive, or no, in fact, it

22 is antagonistic. There are other methods and

23 probably all of them are reasonable methods to look

24 for whether a combination is going to be

25 synergistic.

220

1 Other examples, and this is probably where

2 this has been most widely used, that is, is this

3 drug that is being developed in adult malignancies

4 relevant to pediatric malignancies? Does it have

5 activity in pediatric tumors?

6 So, I chose a relatively recent example

7 that Beth Fox is working on at the NCI, epothilone

8 B, a Bristol-Myers drug. This is an analog that

9 binds tubulin. It stabilizes microtubules by

10 inhibiting tubulin depolymerization, blocks mitosis

11 and causes apoptosis. Interestingly, this drug is

12 cytotoxic in Taxane resistant tumors, as well as in

13 cell lines that over-express MDR. So there was an

14 interest certainly in the pediatric community as

15 far as is this a drug that we should be looking at.

16 So, what one can do is one can look in

17 vitro. In general, it is always helpful to have

18 some sort of reference base to compare your drug

19 with. In this case, we compared it to other

20 microtubule toxins, paclitaxel, vincristine and

21 vinorelbine and looked at the concentrations that

22 were required to produce cytotoxicity in an in

23 vitro model. You can look at these and you can

24 say, well, for these drugs, in fact, these are

25 concentrations that fall within the range achieved

221

1 in patients, and then you look at the drug in

2 question and say, well, these are the

3 concentrations that, if this model is predictive,

4 one might anticipate needing as far as a relative

5 effect and one can ask if there is adult Phase I

6 data or are these relevant concentrations.

7 In addition, one can do some

8 pharmacodynamic work. In this case, one can look

9 at the concentrations that were effective. Were

10 you hitting your target in a very endpoint type of

11 way before cytotoxicity? What was the effect on

12 the polymerization versus non-polymerization? That

13 is what Beth did in this study. So, it is helpful

14 as far as an inexpensive way to look across a panel

15 of cell lines to get some idea that this drug may

16 have some relevance.

17 I think an area that we probably need to

18 do more work on is integration with new agents. I

19 am going to choose leukemia as an example here.

20 For those of you who don't do this on a regular

21 basis, this, in one slide, is what childhood acute

22 lymphoblastic leukemia therapy looks like with

23 different phases of therapy from induction through

24 consolidation, interim maintenance, all the way

25 through maintenance to just over three years.

222

1 As you can see, in each of these phases we

2 treat children with anywhere from six to eight

3 different cytotoxics. Then, on this backbone of

4 very successful therapy that is toxic and is not

5 curing all children, along comes a new drug that

6 has made its way through Phase I and Phase II and

7 clearly has efficacy. The question is, is it going

8 to improve outcome? The question is, aha, here is

9 our new drug, and this drug in this case is the

10 prodrug 506U, and now what? And "the now what" is

11 not an easy question to answer. Where do you put

12 it? What are the risks and benefits of putting it

13 in any one place? We actually were confronting

14 this problem, and still are with this drug as far

15 as how do we integrate this into successful

16 front-line therapy to ask a Phase III question?

17 Well, we have the advantage that 506U is

18 actually a drug that is a very old drug that has

19 only clinically come to our attention in the last

20 decade. Work done by Trudy Allen many, many years

21 ago, beginning in the '50s and extending through

22 the '60s taught us a whole--and a number of other

23 investigators. And, one thing that came to light

24 with anti-metabolites was that there was a

25 potential drug interaction, a negative interaction

223

1 with asparaginase. It turns out that for other

2 drugs there is a very sequence-dependent drug

3 interaction. So, we asked ourselves, okay, we are

4 using asparaginase at a number of points in this

5 therapy, is that a potential problem?

6 You can look in vitro and begin to get an

7 answer to that. So, in this set of experiments we

8 did sequential exposure. Nelarabine is 506U, so

9 first exposing to nelarabine and then following

10 with asparaginase, in this case, because this is in

11 vitro and asparaginase is an enzyme, simply

12 changing over to asparagine-deficient media and

13 then asking the reverse sequence question at least

14 in one cell line--and this is early work that is

15 going to be presented at AACR in a couple of weeks,

16 but in this case there is, indeed, a red flag. If

17 you expose cells to asparaginase before you expose

18 them to 506U you are going to have as much as a one

19 log decrease in effectiveness. So, this is an

20 important piece of information when it comes for us

21 to try to determine what we should attempt to do

22 and what we should avoid doing. This is far from

23 comprehensive and, again, there are only two cell

24 lines and one cell line really didn't have a

25 significant effect. We have to do more work. But,

224

1 again, these models might help us in trying to

2 understand how to integrate new agents on the

3 backbone of effective therapy that we currently use

4 in children.

5 I want to just share a few perspectives in

6 closing. I will preface it by saying this was not

7 a comprehensive talk on cell culture models and

8 these are as much opinions as they are accepted

9 fact.

10 In vitro models are a cost efficient

11 method to search for activity, but

12 mechanistic-based approaches likely will have a

13 higher yield. In other words, drug discovery has

14 moved on from screening I think in cell culture

15 systems.

16 In vitro models can, however, further our

17 understanding of drug action in pediatric tumors,

18 and the moderate throughput is advantageous,

19 especially when studying drug combinations. I

20 showed you that for leukemia we treat with eight or

21 nine drugs. It will become a nightmare trying to

22 figure out all the combinations but with in vitro

23 models you at least have a chance of grappling with

24 some of the major issues there.

25 For most cytotoxic agents, if it does not

225

1 work in vitro it will not work in vivo. So, the

2 negative predictive value for most cytotoxics is

3 pretty good. If you can't kill the cell in the

4 dish you probably shouldn't invest a lot of energy

5 if this is a cytotoxic agent.

6 Correlated to that, if it takes a

7 super-pharmacologic concentration in vitro to have

8 an effect, it will likely not fare well in vivo.

9 For the most part, you can kill cell cultures with

10 anything if you put enough in so you do have to put

11 it in the context of are these concentrations

12 relevant concentrations.

13 Lastly, and this is where we probably fall

14 down most often, if it works well in vitro there is

15 a reasonable likelihood that it will do absolutely

16 nothing in vivo. That is true of a lot of models

17 and it is certainly true of cell culture models.

18 So, I will stop there and let the program continue.

19 DR. SANTANA: Thank you, Peter. We have a

20 few minutes for questions because we have to do two

21 things, we have an open public hearing if anybody

22 wants to speak and we also have to switch laptops.

23 So, there is opportunity to address any questions

24 to Dr. Adamson and Dr. Meltzer now. I have a

25 question for Dr. Meltzer, you kind of hinted at the

226

1 end of your talk about an issue of peripheral blood

2 and, I read in between the lines surrogate use of

3 peripheral blood. Can you expand on what you

4 meant? Did you mean that you would take the tumor

5 diagnosis, establish a profile, and do it also with

6 peripheral blood and diagnosis but then only

7 monitor peripheral blood as your surrogate? Please

8 go the microphone.

9 DR. MELTZER: What I really meant was

10 monitoring toxicity, and the example that I know of

11 that has the most effort is in actually monitoring

12 for radiation toxicity. There are patients who are

13 extremely sensitive to radiotherapy and have severe

14 toxicity and there are some tantalizing preliminary

15 data from Stanford that suggest that you can tell

16 the hypersensitive patients by gene expression

17 profiling of their peripheral blood. That is an

18 approach that, to my knowledge, has not been really

19 applied to chemotherapy and there may be an

20 opportunity to do that. So, I was really

21 speculating.

22 DR. SANTANA: Dr. Reynolds?

23 DR. REYNOLDS: Peter, I think there are a

24 couple of comments I want to make about what you

25 said about the predictive value of these. One is

227

1 that I think there was a very interesting panel

2 discussion at the AACR ERTC meeting in Boston this

3 year about the predictive value of models in

4 general. It wasn't just in vitro, it was talking

5 about animal models. The conclusion was that they

6 were basically non-predictive and, you know, no one

7 had any magic models.

8 At the same time, when you look at the

9 publication that is coming out of the NCI 60 cell

10 line screen, what they are saying is that the one

11 thing that was somewhat predictive is if they have

12 activity in multiple different cell lines, then

13 that tended to give you some predictive value. So,

14 more is better in that setting.

15 The third is that there are some

16 well-established principles that have been

17 discussed in the literature and often ignored that

18 say that if you really can get two logs worth of

19 activity, whether it is in an animal model or in an

20 in vitro model, that may be somewhat predictive.

21 In other words, there is a two-log threshold, which

22 you didn't address. And, I think when you talk

23 about IC50s we clearly are not talking about

24 multi-log assays or in the MPT system.

25 So, I guess what I am suggesting is that I

228

1 think that one reason why the predictive value of

2 some of these has been less than we would all like

3 is that, first of all, I think the systems still

4 aren't optimized and I think they need to be done

5 in multi-log systems and, secondly, as you pointed

6 out, a number of us are studying things like

7 physiological hypoxia and the impact on this.

8 Certainly, as you pointed out very astutely, there

9 must be consideration of what the pharmacological

10 parameters you are going to see in a patient are

11 when you approach these.

12 Third, I think what we really need is to

13 be doing them in more cell lines, not just one, two

14 or three but we need a lot of them. Once we get

15 the right panels of biologic reagents in these

16 systems and the right systems we might see the

17 predictive value go up, and I don't think that we

18 should exclude that possibility when we consider

19 these.

20 DR. SANTANA: Dr. Grillo?

21 DR. GRILLO-LOPEZ: Another comment that I

22 would like to make is that many of these models

23 have been developed for chemotherapeutic agents and

24 when you are dealing with a biological they may

25 have no applicability whatsoever.

229

1 DR. SANTANA: Any other comments?

2 [No response]

3 We have a few minutes for an open public

4 hearing so if there is anybody in the audience who

5 wishes to address the committee, could you please

6 come forward to the podium and identify yourself

7 and any potential conflicts of interest, and make

8 your statement?

9 Well, if nobody is going to take the

10 opportunity, then we will invite Dr. Houghton to

11 proceed with the next presentation.

12 Human Cell-Animal Xenografts: The Current Status,

13 Potential and Limits of Informing us About

14 Clinical Studies

15 DR. HOUGHTON: I would like to thank Steve

16 for inviting me. When we were given the mandate or

17 the subject of this afternoon's session, it was

18 actually a clarifying moment to think about what

19 sort of preclinical data is required or is of any

20 use. I think there are two ways of looking at what

21 sort of preclinical data can be of use. That is,

22 use for us in sort of designing clinical trials as

23 opposed to perhaps the information that would be

24 required for the FDA to make some sort of decisions

25 regarding the potential use of an agent.

230

1 So, I think we can look at early drug

2 discovery largely within defined standardized

3 environments, either in drug discovery groups

4 within companies where you have set protocols and

5 set criteria for establishing whether an entity has

6 adequate activity to progress to the next stage, or

7 in the NCI screening program where, again, there is

8 a set of protocols that drive the criteria for

9 advancement of the compound. The problem is that

10 pediatric cancers are represented in neither

11 entity. They are obviously not going to be a focus

12 of the pharmaceutical industry and, as Peter

13 alluded to, despite multiple attempts they were not

14 included in the NCI screening program.

15 The consequences of preclinical data using

16 pediatric models is generated essentially in an

17 uncontrolled or non-regulated environment where

18 everyone uses their own pet models, their pet

19 design of experiments and, in fact, their own

20 criteria for assessing whether or not they regard

21 something as being active. So, such data derived

22 from experimental systems that are not validated,

23 using experimental designs that are, again, not

24 validated and interpretation of those results lacks

25 consistency and rigor.

231

1 So, taking advantage of the approaches

2 that have been taken in industry and the idea of

3 developing a consistent, criteria-driven approach a

4 group, many of whom are represented here, under the

5 leadership of Malcolm Smith, Barry Anderson and

6 Peter Adamson, met during 2002 to consider what

7 sort of screening program would be useful to

8 implement that would allow us to identify drugs

9 that are in the early clinical or just at the late

10 preclinical stages from industry that might be

11 useful in identifying drugs that would have

12 specific application and perhaps should be

13 prioritized for pediatric clinical testing.

14 The schema is shown here and I am not

15 going to go into detail on this because Malcolm is

16 going to deal with this in somewhat more detail in

17 his talk. But the idea is to set up a panel of

18 models so tumor A may be medulloblastomas and tumor

19 B may be neuroblastomas, a panel of six to ten of

20 these comprising either xenografts or heterografts

21 of human cancers in immune incompetent mice, or

22 where there are transgenic models to implement

23 those within the screening program. But the idea

24 would be that we would have a framework where we

25 can set the criteria for experimental design, set

232

1 the criteria for assessing responses that may be

2 more consistent and may generate data that would be

3 of use not only to us as a group that are

4 interested in developing clinical trials, but

5 perhaps more appropriate use to a federal agency

6 such as the FDA if they wanted to use such

7 preclinical data.

8 So, we were asked to look at the following

9 categories of nonclinical data, and I am going to

10 concentrate on pharmacology and pharmacokinetics

11 efficacy and the aspect of using the models to

12 identify pharmacodynamic endpoints that may be more

13 amenable to analysis within the model systems than

14 they are, certainly, in patients with solid tumors.

15 The other aspect is to ask where such data fits in

16 terms of development of drugs in the pediatric

17 cancer realm.

18 I think the models can be useful in

19 identifying active agents and perhaps better

20 analogs to optimize the administration schedules,

21 or to look at drug combinations in vivo, to

22 prioritize agents for Phase I trials, to make

23 rational decisions within the pediatric consortia

24 as to whether to continue to develop drugs or

25 whether, at some point, we should drop those drugs

233

1 in further development, preferably at the Phase I

2 to Phase II transition to allow us to potentially

3 focus a drug in treatment of certain tumors for the

4 specific activity against certain models in the

5 pediatric clinical screening program, and the

6 potential to relate target inhibition to biological

7 response which is going to become progressively

8 more important as we deal with more agents that are

9 inhibitors of specific signaling pathways.

10 So, the data that suggests that some of

11 these preclinical models may be useful is shown

12 here. This is from rhabdomyosarcoma models, and

13 this is data that was gathered over about a 10-12

14 year period in my own lab which identified

15 vincristine, cytoxan, dactinomycin D and Adriamycin

16 as having good activity against panels of

17 rhabdomyosarcoma xenografts. On the right column

18 are sort of the response rates that have been

19 gleaned from the literature that was available.

20 On the other hand, it shows on the bottom

21 that norfolan is a very active agent in the

22 preclinical models and, indeed, is very active in

23 rhabdomyosarcomas. However, there is a cautionary

24 note here. Although we can identify drugs that are

25 active in model systems and potentially active in

234

1 the clinical setting, it doesn't necessarily mean

2 that this is going to be a good drug. The

3 limitation of norfolan is that it causes cumulative

4 toxicity to bone marrow and subsequently limits the

5 ability to deliver standard therapy to those

6 children.

7 So, we have to look at these results as

8 being promising in terms of being able to

9 retrospectively identify drugs that we know are

10 active in the clinical setting and to prospectively

11 identify drugs that may have activity. That

12 doesn't necessarily mean to say that that drug is

13 going to be potentially a very useful drug in the

14 clinical setting. So, there is a limitation to the

15 models even though they are very promising.

16 Ultimately, the value of the entity itself has to

17 be determined in clinical trials. We can merely

18 point in that direction.

19 On the other hand, you can take the same

20 drugs and run those against colorectal

21 adenocarcinoma xenografts, again, in

22 immunodeficient mice and you see that the drugs

23 that are very active against pediatric

24 rhabdomyosarcoma essentially have no activity

25 against the colon xenografts. So, that gives you a

235

1 little bit more confidence that it is not the fact

2 that you have heterografted a tumor into a mouse

3 that dictates its response.

4 Coming to some more recent data, we have

5 established a series of Wilms tumor xenografts, WT1

6 through WT10, favorable histology Wilms tumors, and

7 SKNEP is a cell line that was derived from a

8 diffused xenoplastic Wilms tumor and the more

9 pluses there are, the more sensitive the tumor is.

10 So, anything that is greater than a 4-plus is an

11 objective response in this model, so 50 percent

12 regression in tumor size.

13 So, you can see with vincristine, WT1

14 through WT10, is 6-pluses which means that these

15 tumors completely regress and do not regrow within

16 a 12-week period of time. Similarly, cytoxan has

17 very good activity in most of the tumors.

18 Prospectively, the model identifies the

19 camphotecan, topotecan and irinotecan as being very

20 active. The important thing here is that topotecan

21 and irinotecan are administered at doses that give

22 relevant systemic exposures to humans.

23 The relative exposure is perhaps the most

24 important change in the way we are thinking about

25 how to look at efficacy. Efficacy in animal models

236

1 is defined as the anti-tumor effect, let's say the

2 ability of a drug to inhibit growth by 50 percent,

3 divided by the dose that causes 10 percent

4 lethality. The problem is that the mouse is not a

5 very good model for human toxicity. The mouse may

6 be either less tolerant to a drug, in which case

7 you may under-predict the activity against a human

8 tumor, or may be much more tolerant than a human,

9 in which case the drug looks fantastic against the

10 heterograft but ultimately fails in the clinic

11 because you can't achieve systemic exposures in the

12 patient that are consistent with the tumor

13 regression in the mouse.

14 The data shown here is the responses of

15 different neuroblastoma xenografts to the drug

16 topotecan against systemic exposure or area under

17 the curve in nanograms per milliliter, showing that

18 if we target, where the arrow is, 100 ng/ml we

19 would expect to get in this case four out of the

20 five tumor lines to give some response. In fact,

21 the total data set was the sixth line which is also

22 completely resistant to topotecan. So, we would

23 predict if we targeted 100 ng/ml that we would have

24 a response rate of four out of six or around 60

25 percent.

237

1 These sort of data are interesting but

2 ultimately you have to prove or validate that this

3 approach does have some merit. That has been done

4 in a clinical trial that was headed by Victor

5 Santana, and the pharmacokinetics was done by

6 Clinton Stuart, at St. Jude. The idea here was to

7 target the same systemic exposure that we set in

8 the mouse, the 100 ng/ml of topotecan-lactone, and

9 the design of the clinical trial is shown at the

10 bottom.

11 The drug is given for five days on two

12 consecutive weeks, which is the schedule that is

13 most effective in the xenograft models, and on day

14 one there are pharmacokinetics taken and then the

15 dose is adjusted to hit this target dose. We are

16 getting quite good at doing this. In this

17 particular trial there were 113 courses of drug

18 administered and 92 percent were in the 100-plus or

19 minus-20 ng/ml range.

20 The results are shown here, where for 28

21 evaluable patients we had approximately a 60

22 percent response rate which is very close to that

23 which would be predicted from a limited number of

24 xenograft models, again suggesting that the idea of

25 using pharmacokinetics as the metric against which

238

1 anti-tumor activity is measured is perhaps more

2 appropriate than using mouse toxicity per se.

3 We looked at the retrospective analysis,

4 again, about ten years work, and we see that for

5 drugs that really didn't progress from Phase I any

6 further, the area under the curve at the mouse

7 maximum tolerated dose versus that in the human is

8 somewhat higher in the mouse than the human. So,

9 in this example, the mouse is about 80 times more

10 tolerant than are humans. Yet, when one looks at

11 the effective dose range, the effective dose range

12 being reductions from the maximum tolerated dose in

13 the mouse at which point you lose objective

14 regressions in your tumor models, we see that the

15 effective dose range for these drugs is relatively

16 small, between two and three.

17 So, where you have such discrepancy in the

18 tolerance between the species and, yet, a very

19 narrow window of true activity against the model

20 systems which are human, one would anticipate these

21 drugs would not necessarily achieve adequate

22 concentrations to give tumor responses in a

23 clinical situation.

24 Drugs that work are shown here. Norfolan,

25 despite its limitations, is very active and, again,

239

1 the pharmacokinetics in the mouse and the human in

2 terms of tolerance are very similar. There is a

3 reasonable dose range of three to four before you

4 lose activity. Similarly, for topotecan the

5 effective dose range spans the differential between

6 mouse and human, as does irinotecan which is very

7 well tolerated in terms of the active metabolite in

8 the mouse and has an extremely wide therapeutic

9 range within this model system.

10 Looking at a more recent drug, irofulvin,

11 NGAI114, again, anything that is more than a 4-plus

12 is causing objective regressions. We have looked

13 at some 18 models. It looks very active. If you

14 dose reduce you see that the MMT is somewhere

15 between 4 and 7. So, let's say you have 14 out of

16 18 tumors, independent tumors show activity in

17 terms of objective regressions. As we reduce the

18 dose further, it is 8 out of 18; reduce the dose

19 further, it is 3 out of 15; and the lowest dose

20 evaluated, only 1 out of 14 different tumors showed

21 objective regressions. These include tumors

22 derived from brain tumors, neuroblastoma and

23 rhabdomyosarcoma.

24 The problem here is that even at this dose

25 the systemic exposure to this drug in the mouse

240

1 exceeds ten-fold that which can be achieved in

2 human trials, again, suggesting that here is a drug

3 that looks dramatically active in a model system

4 but when you relate that activity to the ability to

5 achieve systemic exposures of the drug in human it

6 would suggest that this is a drug that would not be

7 of high priority to undertake clinical trials, or

8 at least progress from Phase I to Phase II clinical

9 trials.

10 Similarly, I think we can address issues

11 of schedule-dependent anti-tumor activity. This is

12 old data with topotecan, but topotecan is given for

13 5 days for every 21 days over 3 cycles, or given

14 for 5 days times 2, so it is Monday through Friday;

15 Monday through Friday at half the dose. So the

16 cumulative dose in both of these trials is exactly

17 the same but the outcome in terms of tumor response

18 is very different.

19 I think this sort of data, where it is

20 derived in a substantial number of tumor models to

21 show that this is a consistent finding, may also be

22 quite valuable in leading us in the design of

23 clinical trials especially where, as was mentioned

24 earlier today, one tends to get one shot at doing a

25 large clinical trial and we might as well give it

241

1 the best chance we can.

2 The other aspect is the discrimination

3 analogs of a particular class of chemical. I think

4 the models again can be quite useful if you apply

5 this in the context of the achievable systemic

6 exposures in humans versus the mouse. This shows

7 some data from an osteosarcoma xenograft which is

8 particularly sensitive to carboplatin and

9 cisplatinum but oxaliplatin, which is a drug that

10 is of current interest in the pediatric oncology

11 world, shows essentially no activity. I think if

12 one extends this data to, say, 6-10 osteosarcoma

13 models and sees that, in fact, oxaliplatin has very

14 little or no activity against these models, this

15 can be factored into how we develop this drug in

16 the clinical setting.

17 These are classical cytotoxic drugs and,

18 obviously, over the next few years there is going

19 to be a progressive shift to drugs that we fondly

20 call molecularly targeted drugs, even though

21 perhaps they aren't quite as specific as we think

22 they are. But under those conditions we have to

23 generate models that very accurately recapitulate

24 the activity of, for example, signaling

25 transduction pathways. This raises the question of

242

1 whether the conventional subcutaneous model is, in

2 fact, going to be useful or whether we will have to

3 go to models where the tumor is implanted into the

4 more physiologically relevant sites, such as brain

5 tumors into the brain, Wilms tumors into the

6 kidney, etc.

7 One way of addressing whether this is the

8 case or not is through expression profiling and

9 proteomics profiling, as alluded to by Paul

10 Meltzer. We have been looking at nearly

11 established models and I will show you a couple of

12 examples here where we have looked at Wilms tumors

13 when we transplanted them into mice as xenografts

14 and have done profiling from the primary tumor from

15 which this xenograft was derived and the xenograft.

16 So, what we are looking at here is the

17 expression profiles for about 6,000 genes that are

18 expressed at reasonable levels in the xenograft

19 versus the primary tumor. As you can see, there is

20 a very high level of concordance, with about 20-30

21 genes that are expressed greater than one standard

22 deviation from the mean. The data suggests very

23 strongly that the expression profiles that are

24 observed in the primary tumor are very largely

25 recapitulated in the early xenograft studies.

243

1 This means two things. It gives us the

2 first real metric to say this model is

3 representative of the parental tumor because

4 previously we have looked at histology and maybe

5 measured a few antigens to see whether they are

6 retained or not. Now we can do this by profiling

7 25,000 to 30,000 genes.

8 Then having this data set, we can do two

9 things. One is, as these tumors are serially

10 passaged in mice, from one mouse to another, we can

11 ask a very pertinent question, at what point do

12 these models start to deviate from the original

13 tumor and, thus, may have much less relevance,

14 particularly for screening or evaluating activity

15 of molecularly targeted drugs.

16 The second use is that if you have really

17 consistent profiles like this, and these are

18 maintained for multiple generations in the mouse,

19 then we have the ability to look at the effects of

20 drugs to perturb these profiles and start to get

21 molecular signatures that may relate to biological

22 outcome, that is, tumor response.

23 One of the uses that we have made of this

24 data is in collaboration with GlaxoSmithKline in

25 cytokinetics, who had data, shown here, that the

244

1 gene for the mitotic KSP was expressed at

2 relatively high levels in tumors and particularly

3 in Wilms tumor. So, the arrow shows the levels of

4 expression in normal kidney, which is extremely

5 low, and also in clear cell carcinoma of the kidney

6 and transitional cell carcinoma the expression of

7 KSP is very low, but in Wilms tumor, which is

8 circled here, it is extremely high.

9 It allowed us to ask the question whether

10 high levels of expression of KSP did, in fact, make

11 this a drug target. We have looked at one of the

12 analogs of an anti-kinase inhibitor that is an

13 analog of the compound that is currently in the

14 clinic, and we have looked at this against a panel

15 of Wilms tumors. The bottom line is that this is a

16 very active agent against favorable histology Wilms

17 tumors that over-express KSP. It is,

18 unfortunately, not particularly useful against the

19 diffuse anaplastic variety, here. That is based

20 upon a single xenograft and we are trying to

21 establish further models and will see if that is,

22 in fact, the case.

23 In terms of the anti-tumor activity, if we

24 can just focus here, this is tumor volume versus

25 time after starting treatment. Control is here.

245

1 This is the KSP inhibitor inducing complete

2 regressions, with only 2 out of the 5 tumors

3 regrowing during the 12-week period of observation.

4 The limitation of this particular analog

5 is that whilst it works very well at the highest

6 dose, there is a very steep dose-response curve and

7 there are much less active fractions of the MTD.

8 So, again, this is going to be a drug where the

9 relative pharmacokinetics between the mouse and the

10 human are going to be really quite critical in

11 determining whether this is very likely to have

12 therapeutic benefit in these tumors.

13 The final part of this is really this

14 aspect of pharmacodynamics. As all of us know, to

15 look at target inhibition and, more specifically,

16 target inhibition and target recovery in patients

17 with solid tumors has been, and will remain to be,

18 a very difficult proposition. Multiple biopsies of

19 tumor at various times before and after treatment

20 is in most cases not really possible.

21 I think the models can be quite useful in

22 this respect and I will illustrate that in terms of

23 the signaling pathway that we will be looking at in

24 the context of a therapeutic trial of a rapamycin

25 analog, CCI779. This particular analog targets

246

1 serene kinase, and it is very easy to monitor the

2 effect of this drug by looking at downstream

3 effectors and whether they are phosphorylated

4 downstream.

5 The problem is that target inhibition is

6 only the first part of the question that you really

7 want to ask. That is, you are really asking at the

8 drug doses that I am giving am I inhibiting the

9 target? That is the first part. But what you

10 really want to know is does the inhibition of

11 target correlate with biologic readout.

12 I think the model systems are going to be

13 very useful to link the pharmacokinetics to target

14 inhibition to biological readout in terms of

15 anti-tumor activity, but even more so in terms of

16 developing concepts of molecular signatures that

17 may be much more important in predicting the

18 outcome for treatment than merely looking at the

19 target inhibition per se.

20 Malcolm Smith will discuss this but the

21 developing initiatives at the NCI include to

22 systematically characterize tumors at the molecular

23 level using both genomic and proteomic arrays. The

24 second is the Pediatric Preclinical Testing Program

25 where we hope to establish models to identify new

247

1 active drugs.

2 I think in terms of using preclinical or

3 nonclinical data we have to standardize our

4 experimental procedures. This is going to be

5 difficult, but in the context of the proposed

6 consortium that I have described it is difficult

7 but it is a realistic goal, and I think once we

8 have a group that is doing this on a large scale

9 under consistent conditions, then I think others

10 outside of that consortium who are doing similar

11 work may adopt the same criteria for looking at

12 tumor response and the design of experiments so

13 that their data and our data can be compared and

14 normalized. I think we have to be careful that we

15 use standardized criteria for assessing drug

16 activity and, again, I think this is something that

17 will come out of the consortium or the PPTP

18 initiative, whoever carries that out.

19 One of the other questions that was being

20 raised is should we be using animal data that is

21 derived under Good Laboratory Practice compliance.

22 The problem here is that if we do this for the

23 cancer screening program, then my understanding is

24 that the entire vivarium within an institute or

25 university also has to function under GLP

248

1 conditions and this aspect of the work may be a

2 very small percentage of the total work that is

3 being done in a vivarium per se. It would

4 certainly increase the costs quite dramatically so

5 I think we have to think about the prospect of GLP

6 in the context of who is going to be doing this

7 work and whether this would increase the cost of

8 animal experimentation not only for the work that

9 is being focused on cancer, but also for non-cancer

10 related work that is ongoing in the same

11 institution. Thank you.

12 DR. SANTANA: Thank you, Peter for a very

13 thorough overview of this issue. I am going to ask

14 Chand Khanna to go ahead and do his presentation.

15 After that we will take a break and then we will

16 come back and reconvene and finish the last two

17 presentations and have our discussion and

18 questions.

19 An Integrated and Comparative Approach to

20 Preclinical/Clinical Drug Development

21 DR. KHANNA: I want to thank everyone for

22 the opportunity, specifically Steven, to come and

23 speak to you today.

24 As Peter suggested, the convention to drug

25 development, as you all know, is to include

249

1 preclinical models to evaluate promising agents and

2 then move those promising agents through clinical

3 development. To continue Peter's theme, what I

4 would like to present is a vision towards an

5 integrated approach wherein preclinical models can

6 be helpful and informative, both at the preclinical

7 level and during various phases of clinical

8 development, and the spin that I would like to

9 provide is one that includes a number of novel

10 models, models that have not been used very much in

11 drug development, and those include naturally

12 occurring cancers that are seen in both genetically

13 engineered mice and, more specifically, in pet

14 animals in our communities that can, again, be

15 included in translational and biological cancer

16 research.

17 What I am going to do is to bring this to

18 you from my efforts within the Comparative Oncology

19 Program of the CNI, which is a new initiative

20 within the Center for Cancer Research, and my work

21 with the Pediatric Oncology Branch where my focus

22 is on sarcoma biology and metastasis.

23 As Peter has alluded to, there are a

24 number of modeling options, and the ones he has

25 focused on and shown us really are how we can best

250

1 use the xenograft models, but there are also

2 opportunities for us to include syngeneic or mouse

3 cancers that are transplantable into mice, and

4 genetically engineered mice that can be used for a

5 number of important steps in the translational

6 process. Lastly, what I want to focus on is the

7 use of pet animals in the drug development process.

8 This is a schema that you are familiar

9 with, wherein small animals are used early in the

10 development. Primarily for toxicology we use large

11 animals, whether they be non-human primates or

12 dogs, and then we move into clinical development.

13 The question is how can we use first a small set of

14 examples genetically engineered mice to inform this

15 process? Largely, I think, because they are more

16 complicated and challenging, we can use them in the

17 evaluation of interesting findings from traditional

18 transplantation models.

19 So, if we look at the historical

20 perspective, genetically engineered mice have been

21 problematic for basically three primary reasons:

22 One is that they are conventionally associated with

23 very rapid tumor progression. They are

24 historically associated with hemologic malignancies

25 and the cancers that emerge usually emerge in a

251

1 number of sites synchronously.

2 Recently there have been novel modeling

3 approaches which have provided us an opportunity to

4 study genetically engineered mice across a range of

5 cancer histologies, almost all histologies.

6 Through efforts including conditional expression of

7 genes, somatic expression of genes within a

8 selected pool of target cells, there are now very

9 good mouse models for most human cancers. The

10 advantages that these genetically engineered mice

11 provide through the translational process are that

12 after you induce the genetic change in the mouse

13 the cancers that emerge, emerge spontaneously.

14 That is one.

15 The second is that the tumor that emerges

16 is syngeneic from the tumor to the tumor

17 micro-environment to the host, and that is

18 something that I think provides opportunities

19 specifically for targeted biology-based therapies.

20 The genetics of the cancer are modifiable and are

21 relevant and, although it is more easily said than

22 done, the biology of these cancers can be

23 controlled now so we can have opportunities for

24 therapeutic evaluation during the course of

25 progression that these genetically engineered mice

252

1 have.

2 There are limitations, and the limitations

3 that we see with traditional animals still exist

4 with these genetically engineered mice. There is

5 heterogeneity within a specific population of mice.

6 There is heterogeneity in the genetics of the

7 cancer and I think that is a value. It adds to

8 what we get out of more or less homogeneous

9 populations seen in the transplantation settings.

10 Experimentally, these are very difficult and

11 complicated designs to pursue from the standpoint

12 of translation but they can be done. They are

13 expensive, time consuming, and we don't really know

14 yet about their predictivity.

15 The most important issue about their use

16 is a series of patents that have been provided to

17 Dupont exclusively that really extend to all

18 genetically engineered mice. Any activated

19 oncogene in a mouse is covered by the OncoMouse

20 patents. The result of these patents is really the

21 limitation of their use in the pharmaceutical

22 industry. So, unless this issue can be dealt with,

23 I think the use of these genetically engineered

24 mice in the pharmaceutical industry will be

25 limited.

253

1 What I want to move on to is ways for us

2 to include naturally occurring cancers in the

3 translational process in the drug development

4 process. Again, within the Comparative Oncology

5 Program what we plan to provide are opportunities

6 to include these models in drug development. So,

7 pet animals have a number of interesting cancers

8 that are relevant from the standpoint of pediatric

9 cancers, including lymphoma and then dogs with

10 osteogenic sarcoma.

11 Dogs in the community are developing these

12 cancers. There are 65 million pet dogs in the

13 United States, 6 million will develop cancer in a

14 year and the pet owners of these dogs are seeking

15 out advanced care and, in many cases, are very

16 interested in including their dogs in trials that

17 evaluate new therapies. So, what this provides is

18 an opportunity to include these large animal models

19 in drug development and this has been done largely

20 within the pharmaceutical industry.

21 The advantage that these large animals

22 provide is, in fact, that they are large outbred

23 animals, unlike the small animals that we

24 traditionally use at the preclinical level. The

25 genetics of the host, the dogs, have been shown by

254

1 the recent completion of the canine genome to be

2 quite similar, very similar in fact, to humans.

3 They are naturally occurring cancers. Then, within

4 given histologies the genetics of the cancers are

5 very similar to the genetics of the same human

6 cancers. Very importantly, one thing that these

7 models provide is that within a histology there is

8 considerable genetic and individual variability

9 that is, in fact, captured within populations of

10 humans and often is the problem as we move through

11 clinical development. This heterogeneity is not

12 captured in other models.

13 If you look at histology responses, for

14 example lymphoma, the drugs that are effective in

15 dogs with lymphoma are effective in people with

16 lymphoma. The drugs that are not effective in dogs

17 with lymphoma are not effective in people with

18 lymphoma. To a large extent, that parallel is true

19 for a number of histologies with classical,

20 conventional cytotoxic drugs. The biology of

21 metastases within these models is faithfully

22 reproduced for specific histologies. Lastly, I

23 think an important point is that these cancers are

24 characterized by resistance or recurrence and this

25 is really the problem that we face with pediatric

255

1 patients and adult patients. The biology of

2 recurrence or resistance is difficult to model in

3 most small animal settings.

4 So, if we look at this table that I have

5 taken from Shadner's recent review in JCO, he has

6 listed out preclinical through clinical development

7 of the number of agents at various phases at one

8 point in time. What I have done in red is just put

9 the number of agents that are active per year. By

10 looking at this, you can see where opportunities

11 exist to improve the process of drug development.

12 Certainly as Peter suggested, there is room for us

13 to improve this initial step but as we move along,

14 I think there are great opportunities for us to

15 take Phase I agents that are not burdened by the

16 hurdle of maximally tolerated dose and inform

17 decisions towards Phase II. I think there is an

18 opportunity for these large animal models, for

19 genetically engineered mice to take that role of

20 informing towards Phase II and potentially

21 informing towards Phase III.

22 So, this is the integrated approach that I

23 would like to suggest wherein pet dogs--we have

24 largely done this work within the pharmaceutical

25 industry to assess activity, toxicity,

256

1 pharmacokinetics and pharmacodynamics and used that

2 information to lead towards Phase I. Well, perhaps

3 as important, use these tumor-bearing dog studies

4 to define dose regimen schedules towards Phase II

5 to validate, potentially to identify but really

6 more appropriately validate biomarkers, define

7 responding histologies, and then provide a rational

8 system in which we can demonstrate that

9 combinations should be considered towards Phase II

10 and potentially Phase III.

11 So, I would like to give you a couple of

12 short examples. Thrombospondin-1 is a very large

13 protein with a number of receptors and a number of

14 effector domains. The second type-1 repeat has

15 been associated with significant antiangiogenic

16 activity. From the second type-1 repeat a series

17 of small peptides, non-amino acid peptides, are

18 being pursued as anti-cancer drugs, antiangiogenic

19 drugs. The problem with the development of this

20 class of drugs and specifically thrombospondin-1 is

21 that although we can show within mice that these

22 agents are antiangiogenic and although we can show

23 that they do have anti-cancer activity, the leap

24 towards the clinic has been difficult.

25 So, the question was whether or not we

257

1 could use dogs with naturally occurring cancers to

2 help us make that step. What I would like to show

3 you is a simple example of how we have done that.

4 The experimental clinical trial for pet dogs

5 included dogs with any measurable malignant cancer,

6 no concurrent therapy, and the endpoints really

7 were to assess toxicity, a limited attempt to

8 evaluate PK, and then to look at response, keeping

9 in mind that response was going to be assessed

10 against bulky disease using a single-agent

11 antiangiogenic drug.

12 The first point that I want to bring up is

13 that accrual is achievable. In a short period of

14 time we can enter large numbers of dogs in these

15 clinical trials with the support and interest of

16 their pet owners. Toxicity has always been

17 evaluated, in fact, in dogs. An interesting and

18 important point is that pet dogs that bear cancer

19 have different toxicity profiles than beagle dogs

20 that are evaluated in the research setting. In

21 fact, in many situations the toxicities that are

22 seen in pet dogs are much more similar to those

23 toxicities seen in patient populations.

24 I will show you some of the responses.

25 This is a dog with a maxillary squamous cell

258

1 carcinoma. This is the lesion after 30 days on

2 therapy. It is perhaps a little clearer here.

3 After 60 days the lesion is much more active. It

4 is hemorrhagic. Through a 60-day period of time in

5 a human clinical trial, Phase I trial, it is

6 unlikely that you would continue this patient on

7 therapy with progression. But we did continue this

8 dog and after 90 days, the lesion is now no longer

9 present. We can biopsy this site and there is

10 squamous cell carcinoma that is persistent there

11 but the lesion is not actually assessable there.

12 So, this dog continues to do well, free of disease

13 that is measurable within the mouth, but not a

14 histological regression.

15 I have several other images that I could

16 show you to suggest, in fact, that the agents are

17 active and they can result in regressions. The

18 responses include stabilization which we feel are

19 real but, in fact, objective regressions of lesions

20 that cross a number of histologies.

21 The other thing that this points to is, in

22 fact, histologies that we wouldn't have predicted

23 activity in. So, lymphoma was found to be quite an

24 active site and now, in Phase II, these drugs are

25 moving ahead. Of interest to the group, sarcomas

259

1 were particularly responsive histology.

2 So, what did we learn from these dog

3 studies? Antiangiogenic peptides can be active

4 against bulky disease. They need time. Because of

5 the results that we were able to generate in dogs,

6 the Phase I trials in Europe extended their

7 observation times and they did see objective

8 responses in patients treated for 60 days.

9 Agents are active against histologies we

10 wouldn't have predicted, like non-Hodgkin's

11 lymphoma. A very important point is that all dogs

12 that continue through therapy develop resistance on

13 therapy so combinations are going to be necessary

14 and, as we look towards the use of these agents, we

15 are going to have to keep in mind that resistance

16 will be an obvious problem. Most dogs don't

17 respond to therapy and, therefore, there is an

18 opportunity for us to define markers that predict

19 responsiveness within a heterogeneous population of

20 dogs and, in fact, predict when responses will be

21 seen. That work is being done and thus far

22 circulating endothelial cells seem to an interest

23 and will move on into the clinical setting as well.

24 This is, again, the perspective that we

25 have and I think there are some examples from the

260

1 thrombospondin-1 studies that show how we can

2 inform towards Phase II. I want to end with

3 another brief example and it speaks to this

4 pharmacokinetic/pharmacodynamic response question

5 that Peter brought up.

6 So, Cheryl London, who is at UC Davis, is

7 evaluating small molecule inhibitors of the split

8 tyrosine kinase receptor family. What she was able

9 to do in a very similar trial design, treating dogs

10 with bulky disease, is actually do tumor

11 pharmacodynamics using phospyl KiT as the target;

12 do serial biopsies in dogs evaluating the diversity

13 of KiT mutations in dogs with nasal tumors and

14 define the dose that is required to modulate the

15 target in vivo to validate surrogates that could be

16 more evaluated in human clinical populations

17 against this tumor target, and then move those

18 things into the clinic.

19 She was able to show that the dosing

20 schedule, an every other day dosing schedule, was

21 valuable and able to achieve threshold receptor

22 inhibition of KiT. This information was translated

23 directly into the development of products in

24 clinical trials. The every other day dose was

25 suggested for human development but the human

261

1 development required input from marketing and

2 marketing didn't want to pursue every other day

3 dosing. The drug trials predicted daily dosing

4 would be toxic and, in fact, was toxic in people.

5 I am just going to jump ahead. So, what

6 we are interested in being able to do within the

7 Comparative Oncology Program is provide a reagent

8 kit that can allow biology-based questions to be

9 answered in these trials. This has been a

10 difficulty for dog trials thus far in that we just

11 don't have reagents to study dogs in a rigorous

12 way. We now have a validated canine oligoarray, a

13 17K element array. We are validating proteomics

14 approaches in dogs with cell signaling. We have

15 screened specific antibodies for cross-reactivity

16 to dogs and we have made good progress there.

17 Multicenter collaborations are going to be

18 required for us to be able to do trials in a short

19 period of time, and allow that short period of time

20 to inform towards clinical development of the same

21 drugs, and to be able to help with decisions of

22 when these models can be used and when they should

23 not be used in development. There are times where

24 really the questions are not appropriate to ask

25 within these dog studies.

262

1 I will just end with a list of histologies

2 that I think are relevant. Osteosarcoma is

3 obviously an area of personal interest and we have

4 actually published randomized, prospective,

5 placebo-blinded trials in dogs with osteosarcoma

6 looking for opportunities in the clinic.

7 We are interested in lymphoma. There are

8 other histologies. But important to note is that

9 within each of these cancer histologies are genetic

10 changes that can be modeled and can be targeted.

11 So, it doesn't have to be histology based.

12 The weaknesses of these models are the

13 cost. Drug costs are a primary concern; the cost

14 of managing the trials and time. They are longer

15 models than what we would see with typical small

16 animal studies although the time is much shorter

17 than what you would have in the same clinical study

18 in a human population.

19 With that, I will conclude. I will

20 acknowledge our initial group in Comparative

21 Oncology, and the slide also includes Lee Helman

22 and the people in the Pediatric Oncology Branch.

23 UC Davis and Cheryl London has been doing a lot of

24 these translational studies. Now, with the

25 interest of CTEP and the CCR, we are pursuing some

263

1 trials with 17DMAG to answer some of these

2 questions that will inform towards Phase II.

3 DR. SANTANA: Thank you, Chand. I will

4 seek the advice of the FDA. Should we take a

5 ten-minute break and try to get back on schedule

6 because I know we are going to have people dropping

7 off as the day progresses. So, why don't we just

8 take a ten-minute break and reconvene at 3:00,

9 finish with the two presentations and then take

10 questions and discussion and try to get out of here

11 on time?

12 [Brief recess]

13 DR. SANTANA: I will invite our next

14 speaker to come to the podium. Dr. Kenneth

15 Hastings will address the issues of what can be

16 learned about safety using different models.

17 What can Learned About Safety?

18 DR. HASTINGS: Well, my task, after these

19 really nice scientific presentations, is to give

20 you the regulatory spin on things so your task is

21 to stay awake.

22 What I want to talk about today is the use

23 of neonatal and juvenile animal studies for

24 determining the safety of drugs for use in

25 pediatric patients and, obviously, this is going to

264

1 apply to pediatric oncology.

2 The specific guidance that really led to

3 the development of guidance on juvenile animal

4 studies was the Pediatric Exclusivity Act under

5 Section 505A of the FDC Act. The specific language

6 that is included that refers to nonclinical studies

7 is that the FDA may request nonclinical trials

8 before completing pediatric studies in humans.

9 Certain toxicology studies in immature animals may

10 be necessary to evaluate the safety of use in

11 pediatric conditions.

12 Also another regulatory background

13 document has been referred to previously, and that

14 is ICH E11, clinical investigation of medicinal

15 products in the pediatric population, and once

16 again the decision to proceed with a pediatric

17 development program involves consideration of many

18 factors, including any nonclinical safety issues.

19 The need for juvenile animal studies should be

20 considered on a case-by-case basis. Then it refers

21 to ICH M3, which is the document that outlines the

22 timing of nonclinical studies vis-a-vis clinical

23 studies.

24 Finally, there is a draft document that

25 was published in February, 2003, nonclinical safety

265

1 evaluation of pediatric drug products. We now have

2 the final version, after comments were made to the

3 docket, and we hope to publish it sometime this

4 spring or summer, and we took into consideration

5 the comments that were made. This document

6 provides guidance on the role and timing of animal

7 studies in the safety evaluation of therapeutics

8 intended for the treatment of pediatric patients,

9 and it also provides specific recommendations based

10 on the available science and pragmatic

11 considerations.

12 Why did we get into the issue of juvenile

13 animal studies? Well, in assessing the use of

14 drugs for pediatric use the basic assumption that

15 we have proceeded with over the years has been that

16 under most circumstances the safety and efficacy of

17 drugs approved for use in adults predicts pediatric

18 use if you make the appropriate dose adjustment.

19 Now, in the past we have used things like

20 relative body surface area. We consider that to be

21 a good default measure for dose adjustment. But

22 generally this is less informative than data you

23 would get from a clinical pharmacology study. That

24 is really what we are after, being able to make

25 dose recommendations based on actual ADME

266

1 pharmacokinetic studies.

2 Neonatal and juvenile animal studies to

3 enable clinical studies are needed basically to

4 support the safety of studies in pediatric

5 patients. The origin of the guidance really was to

6 provide information and what we call triggers on

7 the need for nonclinical studies. Basically, what

8 we are saying here is that you don't need to do a

9 juvenile animal study every time you want to do a

10 clinical trial in a pediatric patient population.

11 What we were trying to do is to find out what are

12 the sorts of things that we could observe or

13 already know about the toxicology or the safety of

14 a drug that would tell us that maybe you need to do

15 a pediatric juvenile animal study to support the

16 safety of a pediatric study.

17 Also, this guidance contains advice on the

18 conduct of the studies and provides information on

19 how the results of these studies would be used in

20 designing pediatric drug trials and, in fact, in

21 deciding whether or not they would be safe.

22 Now, we recognize that there are

23 differences in the drug safety profiles between

24 mature and immature systems, and these include

25 differences in susceptibility to insult and

267

1 differences in toxicity-related ADME parameters.

2 We recognize that some physiological systems are

3 more vulnerable than others, especially those that

4 undergo extensive postnatal development.

5 When you think about it, you know, that

6 doesn't exclude much. There are a lot of things

7 that undergo significant postnatal development.

8 So, really more than anything else what we would

9 think about are those that might be particularly

10 susceptible to insult, such as the developing

11 nervous system, maybe the developing immune system,

12 the kidneys, perhaps even the gut. So, those would

13 be potential triggers for asking for a juvenile

14 animal study if we knew from adults, from clinical

15 practice or from mature animal studies, that these

16 are target organs of toxicity.

17 Now, I want you to keep in mind two basic

18 concepts that toxicologists use all the time. They

19 have to do with how you look at the usefulness of

20 studies, what it is that you intend to get out of

21 the study. Actually, I have them in reverse order.

22 The first are studies that are designed for hazard

23 identification. Basically, the idea behind hazard

24 identification is that you demonstrate that a drug

25 or a candidate drug has the potential to cause an

268

1 adverse effect. An example of hazard

2 identification would be something like an Ames

3 assay or a discovery toxicology study where you

4 administer a drug by intraperitoneal injection.

5 You are just trying to find out if a drug can cause

6 a toxicity.

7 Pertinent to our discussion today, under

8 certain circumstances adverse effects in mature

9 animals might not be predictive of adverse effects

10 in developing systems. So, some studies that you

11 might conduct, some juvenile animal studies you

12 might conduct actually might be for the purposes of

13 hazard identification, and I am going to talk about

14 how that plays into the design of studies a little

15 bit later.

16 Risk assessment, of course, is that you

17 are trying to look at all of the parameters of

18 toxicity--systemic exposure, route of

19 administration, length of exposure, all of the

20 parameters that determine whether or not what is a

21 potential toxicity is actually going to be manifest

22 as a toxicity in the use of the drug. Basically,

23 this is one of the assumptions that we make when we

24 say that for studies conducted in mature animals

25 the effects will predict what happens in neonates.

269

1 What you need to do is determine what parameters,

2 particularly ADME parameters might alter that risk.

3 I want to just mention very briefly the

4 differences in pediatric versus adult patients or

5 subjects with respect to ADME because that really

6 was the driving factor in looking at juvenile

7 animal studies to start out with. In humans, if

8 you look at ADME, there are differences with age as

9 far as distribution of drug dose. The receptors

10 come and go; they develop and certain

11 age-restricted ranges and, therefore, what you

12 observe in younger systems may not be applicable to

13 older animals and, obviously, extrapolating this

14 clinically.

15 As far as absorption of an orally

16 administered drug, you have to consider that in

17 infants they have a larger volume of distribution,

18 larger surface area to body weight ratio, and the

19 body composition is different. Infants and

20 children have higher gastric pH which will affect

21 the absorption of basic and acidic drugs, larger

22 absorption of the basic drugs; less absorption of

23 acidic drugs. GI motility is different. In

24 infants and neonates GI motility tends to be fairly

25 low compared to adults. In children the motility

270

1 tends to be high compared to adults. So, the

2 actual achievable AUC for a particular orally

3 administered drug may be different if you just do

4 your extrapolation based on body surface area.

5 And, there are certain other things to consider,

6 such as unique routes of exposure such as through

7 mother's milk.

8 A very difficult issue is metabolism. We

9 know that as a general rule there are certain

10 metabolic systems that appear to be more functional

11 in pediatric patients versus adults. I am not

12 going to get into a long discussion about

13 differences in metabolism except to say this, with

14 respect to P450 enzymes, if you look at juvenile

15 animals and if you look particularly at rats which

16 is a model that we use quite often, we actually

17 don't know a lot about the relative development of

18 the P450 enzymes. There is probably one exception

19 to that. We thought that there would probably be a

20 lot of information on this. It turns out that

21 actually there is not in the published literature.

22 Finally, another thing to consider is

23 excretion in juvenile animals--actually, I am

24 talking clinically but in children you have lower

25 glomerular filtration rate, lower tubular

271

1 secretion, resulting in slower clearance and longer

2 half-life. Once you get up into the child range

3 you have rapid clearance and shorter half-lives.

4 So, once again, pharmacokinetics may not be

5 predictable based on body surface area.

6 One thing to consider is how valuable are

7 animal models for ADME comparisons. Well, an

8 obvious advantage is that in animals you can do

9 experimental manipulations that might help you

10 define ADME parameters. But a not so obvious

11 advantage, as I have mentioned, is the lack of

12 comparative information in animals, particularly

13 with respect to metabolizing enzymes. One thing to

14 consider though is that if you can associate PK

15 parameters with adverse effects in animals, this

16 might be useful in clinical trials. So, that is

17 one real advantage to a juvenile animal model.

18 Really ADME was what originally drove the

19 consideration of doing juvenile animal studies.

20 Obviously, the other thing that we are interested

21 in is toxicity. Are these studies going to be

22 safe? The things that we need to consider are the

23 relative maturations of physiologic systems. These

24 are probably better understood in animals but we

25 could have a debate about that. If adverse effects

272

1 are observed in mature animals, then the juvenile

2 animals could be used to demonstrate increased or

3 decreased susceptibility, and you may be able to

4 understand how ADME might affect that. Once again,

5 however, extrapolation to clinical trials may be

6 less certain because of the variations in, for

7 instance, metabolism that we don't really

8 understand as well as we should in animals.

9 Let me lay out a couple of scenarios where

10 juvenile animal studies might be useful for the

11 purposes of toxicology studies. One thing, you may

12 need a juvenile animal study if you already have a

13 pretty good handle on the adverse effects and you

14 have a pretty good idea about the ratio of toxic

15 dose to efficacious dose, and particularly this may

16 be true for short-term use drugs like antibiotics.

17 However, and this was mentioned earlier--I

18 believe Dr. Santana mentioned this, sometimes even

19 with acute exposure you might need long-term

20 follow-up studies. The classic example for this is

21 the fluoroquinolones. What happened here, as you

22 probably are aware, fluoroquinolones are associated

23 with a very troubling effect called crippling

24 arthropathy. It was originally discovered or

25 described in puppies, in young dogs. The question

273

1 was the clinical relevance of these studies. There

2 is a lot of talk about this and I don't want to get

3 into that debate but I think that most people

4 nowadays consider that fluoroquinolone use in

5 children is something you approach very carefully

6 because this may very well be a serious adverse

7 effect that would persist into adulthood.

8 One of the ways that we have looked at

9 answering this question was simply to do this, to

10 dose juvenile dogs, beagles, with fluoroquinolones

11 over a course of, like, two weeks at, say, doses

12 equivalent or maybe higher than what you would use

13 clinically, producing AUCs equivalent to higher

14 than clinical doses, and then just let the dogs go,

15 let them mature and then, at about six months of

16 age, you would look at the dogs again and do

17 clinical evaluations, to histopath on the affected

18 bones and see if there are any changes in those

19 animals; see if the effect gets worse; see if it

20 improves; see if there are any associated lesions

21 that appear to be caused by this juvenile exposure.

22 In fact, what we now know about

23 fluoroquinolones--to cut to the chase--is that

24 actually these effects tend to persist. They

25 probably don't get worse but they do persist. That

274

1 is an important thing to learn in deciding whether

2 or not to conduct a clinical trial, let's say, for

3 something like otitis media, and also to look at

4 the follow-up. In fact, that was used as an

5 argument for the long-term follow-up of children in

6 clinical trials with fluoroquinolones, and I think

7 this is something you should take into

8 consideration when you think about oncolytics used

9 in pediatric patients. I think it is a pretty good

10 comparison that you might want to think about.

11 When you talk about long-term use,

12 particularly, let's say, a drug that has never been

13 developed for use in adults, then you might think

14 about what we would call a shift to a hazard

15 identification type of study. What you would do

16 here is you would start with juvenile animals. You

17 would dose them all the way through adulthood, look

18 for adverse effects and then, if you do see adverse

19 effects, you can go back and do window of

20 vulnerability studies where you try to find out

21 where, in the development of that animal, this

22 occurred and this could help you in understanding

23 where the vulnerable windows would be in a clinical

24 trial. In other words, you can build risk

25 assessment into what is in fact, when you think

275

1 about it conceptually, a hazard identification kind

2 of study. You can also build in pharmacokinetics,

3 obviously, and safety pharmacology studies such as

4 effects on blood pressure, cardiac function, renal

5 function and things like that.

6 I just want to make one mention about

7 efficacy models. We have had a lot of talk about

8 efficacy models; very good talks. I just want to

9 say that you can build safety determinations into

10 efficacy models, particularly large animal models

11 where you can do serial blood levels of biomarkers

12 or AUC for the drug, things like that. So,

13 although we haven't in the past typically looked at

14 efficacy models for safety information--we do our

15 toxicology studies in otherwise health animals,

16 efficacy models probably can be used for this, and

17 I think there is at least some experience with that

18 in looking at biologics.

19 Now I am going to mention the animal rule.

20 The animal rule was passed, I believe, in 2001. I

21 think that is when it was finally codified. This

22 allows for use of animal studies to demonstrate

23 efficacy for where clinical trials would be

24 unethical and/or not feasible. It applies to new

25 drug and biologic products. It is used to reduce

276

1 or prevent toxicity of chemical, biological,

2 radiological or nuclear substances. Obviously, I

3 think we can sort of see what the animal rule is

4 really designed for, and that was for development

5 of drugs to treat things like anthrax. Basically,

6 we are talking about counter-terrorism measures.

7 You know, antidotes for nerve toxins and things

8 like that. That is what it is really designed for.

9 Drugs considered should have demonstrated

10 safety in humans. That is one thing that is built

11 into the animal rule. Now, whether or not that

12 would apply to oncolytics, that is a different

13 question and I think that is something for the

14 panel to discuss. If possible, clinical activity

15 in a relevant disease, although lack of clinical

16 efficacy data shouldn't prejudice against

17 consideration under the animal rule. We have had

18 sponsors come in and propose to pursue a drug under

19 the animal rule where there was no activity data in

20 clinical trials in adults, let's say, as applied to

21 what we are considering today. The important thing

22 to consider is in what way can this principle be

23 applied to pediatric oncology drugs. I think this

24 is something that maybe would be worth discussing.

25 Juvenile animal studies can be useful for

277

1 safety determinations. They are not prohibitively

2 challenging to conduct. You can dose rat pups from

3 day seven on. In fact, people have even looked at

4 beginning with birth, transferring drug in mother's

5 milk and then starting to dose after weaning.

6 There are all kinds of ways you can manipulate

7 neonatal animal studies.

8 The available data doesn't indicate that

9 juvenile animal studies need to be routinely

10 conducted, but they might be needed under certain

11 circumstances, as I have mentioned previously. But

12 the database is limited and this conclusion could

13 change. I don't think it will but, as with

14 anything, as we start seeing more juvenile animal

15 studies we will start looking back at these and

16 deciding whether or not we made the right decision

17 in our recommendation.

18 So, thanks and I appreciate your

19 attention.

20 DR. SANTANA: Thank you. I am going to

21 take the chair's prerogative and ask you two

22 questions because I don't want you to leave the

23 podium without addressing these. One is, can you

24 give us an idea of the universe of where this is

25 applied? I mean, how many times when there is a

278

1 new drug, either in development or a drug that is

2 already out there, are we going back and doing

3 either retrospectively, when is the drug is already

4 out there or as part of the development plan, some

5 of these studies addressing specific issues of

6 toxicity? Is this a common thing that happens?

7 DR. HASTINGS: In juvenile animals?

8 DR. SANTANA: Yes, is this common or

9 uncommon? That is the first question. Then a

10 corollary to that is, are there specific animal

11 models that address specific systems? So, is there

12 an animal model that already looks at neurologic

13 toxicity? Is there an animal model that already

14 looks at cardiac? Or, is it really just this model

15 and then we look the nervous system or we look for

16 the heart system, and so on and so forth?

17 DR. HASTINGS: Well, the first question,

18 yes, we have seen a number of juvenile animal

19 studies. Dr. Karen Davis Bruno, who is the chair

20 of that committee, has been keeping a running

21 tabulation. Karen, do you know what the number is

22 right now?

23 DR. BRUNO: [Not at microphone; inaudible]

24 DR. HASTINGS: Also, as I understand it

25 from sponsors, when it was understood that we were

279

1 working on this guidance, if they were going to

2 pursue pediatric development before they understood

3 that we were looking at developing a for-cause

4 guidance, in fact, a number of sponsors just did

5 them. I mean, they basically just said we are

6 going to anticipate that FDA is going to ask for

7 them. So, yes, there are a number of them and some

8 of them have been quite informative. I didn't

9 really get into that because, frankly, I am not

10 aware of a case in pediatric oncology.

11 As far as a preferred animal, well, no. I

12 wish we could say that there is. You know, we have

13 standard models in toxicology in drug

14 development--rats, beagle dogs, cynomolgus monkeys

15 and it is almost like those are the better models

16 simply because we have just developed so much data

17 with them that we understand what is going on

18 there. If it is neurological though, you are

19 probably wanting to think more in the line of a

20 non-human primate like cynomolgus. But for, like,

21 immune parameters probably rats would be a better

22 model simply because we have the reagents to do

23 that kind of study.

24 DR. SANTANA: Thank you for answering

25 those two questions. I think they were relevant to

280

1 what you were trying to address in your

2 presentation. I will invite Malcolm--he has the

3 daunting task of being the last speaker.

4 Assessing Anti-Tumor Activity in Nonclinical

5 Models of Childhood Cancer

6 DR. SMITH: I would like to thank Steve

7 and colleagues at the FDA for sponsoring this

8 meeting and for the invitation to speak here this

9 afternoon.

10 I will be talking about NCI's initiatives

11 to develop nonclinical models for pediatric

12 oncology. Throughout the talk I will slip between

13 nonclinical and preclinical. The slides are

14 variably labeled that way but you will know what I

15 mean. The three major things I will be focusing on

16 are, one, why we need to be working in this area;

17 two, why we are doing what we are doing; and,

18 three, why we think it has at least some chance of

19 providing useful information.

20 I have shown this slide at I think

21 previous pediatric ODAC meetings, but it is the

22 drug development pyramid and it makes the point

23 that there are more agents entering Phase I studies

24 in adults than we can move into children; then,

25 more agents during Phase I in children than we can

281

1 conduct Phase II studies for; then only a very

2 limited number of Phase III studies that we can

3 conduct. We are not limited now at the Phase I

4 setting. We actually could study more drugs in the

5 Phase I setting. Where we really are limited is in

6 moving to Phase II and doing all the Phase II and

7 pilot studies that we need with these new agents,

8 and then especially moving into Phase III studies

9 and the one neuroblastoma Phase III study or

10 rhabdomyosarcoma Phase III study that we may be

11 able to do in the next three, four to five years.

12 To make a concrete example of this

13 neuroblastoma and looking at the agents under

14 evaluation now, these are all in pediatric Phase I

15 or Phase II trials--a demethylating agent,

16 decitabine, fenretinide, interleukin-12, the Trk

17 tyrosine kinase inhibitor, oxaliplatin, HDAC

18 inhibitors and then BSO. Those are the single

19 agents or we could combine those with standard

20 chemotherapy agents in different regimens. We can

21 combine them with each other and try to inhibit

22 some of the different pathways jointly that these

23 agents inhibit. So, how are we going to pick which

24 of these agents, which combinations to bring

25 forward for the one neuroblastoma Phase III study

282

1 that we will be starting in two or three years? It

2 is a daunting challenge to try to get data that

3 informs that decision.

4 Hence, this is a primary need for some

5 help with that from the preclinical or nonclinical

6 area. If we had predictive nonclinical methods, it

7 could contribute to prioritizing agents for

8 evaluation against specific types of childhood

9 cancer. To do this, we need a systematic approach

10 opposed to what really has been a haphazard

11 approach over the past twenty years. The

12 systematic approach is required to assess the

13 predictive value of pediatric nonclinical models.

14 In recognition of the need for such a

15 systematic approach, the NCI board of scientific

16 advisors approved committing ten million dollars to

17 this effort over the next five years through the

18 Pediatric Preclinical Testing Program. I will

19 describe this in a bit more detail later but for

20 now suffice it to say that this will be a

21 systematic approach, primarily based on in vivo

22 testing with xenograft models, but also having an

23 in vitro component and making use of genetically

24 engineered models when those are available and

25 applicable.

283

1 So, the questions I am asked about this

2 when I have talked about this are, well, why are

3 you doing this? Don't you know that adults have

4 used xenografts and xenografts don't

5 work?--analogous to Pat Reynolds' question earlier.

6 I would respond to this by pointing out three

7 papers, and I will start with the last one, a

8 review article that I would refer you to for

9 marshaling of the arguments that xenografts can

10 contribute to drug development and the take-home

11 message there is better than commonly perceived but

12 can be improved.

13 The first reference was a paper from the

14 developmental therapeutics program at NCI, and the

15 conclusion there was that although maybe a breast

16 cancer xenograft didn't predict for activity in

17 breast cancer, activity across a range of

18 xenografts predicted that that was an agent that

19 had a good chance of being successful when

20 transferred to the clinic, not necessarily for the

21 tumors that weren't in the xenograft models but for

22 at least some cancers having activity.

23 The second paper, a more recent paper

24 published last year in Clinical Cancer Research,

25 made the point that using panels of xenografts for

284

1 a given tumor type increases the likelihood for

2 correct prediction, and we will be focusing on

3 panels of xenografts in our preclinical testing

4 program.

5 This shows two figures from that paper.

6 If you look at the one on your left, each of the

7 squares represents a drug that was studied in the

8 clinic. There is the Phase II activity, the

9 response rate. And, it was studied in a panel of

10 xenografts, and the readout there is the mean

11 treatment to control. So, a low treatment to

12 control indicates a high level of activity in the

13 preclinical setting and high response rate in the

14 Phase II, of course, indicates high activity there.

15 So, you see the predictive value for at least two

16 of these xenograft panels where activity in the

17 preclinical setting in these ovarian xenograft

18 panels and the non-small lung cancer xenograft

19 panels predicted for Phase II activity for these

20 agents.

21 The other point that I make when

22 justifying why we think this has some reasonable

23 chance of being successful is that we have the

24 advantage of being able to make use of pharmacology

25 to enhance a predictive ability of preclinical

285

1 models. We will be able to make comparisons

2 between mouse pharmacology to human pharmacology

3 and this can rule out the trivial explanation for

4 activity in xenograft models. That trivial

5 explanation for an agent being active in a

6 xenograft model being that the mice tolerate much

7 more of the agent than humans do. So, a human

8 cancer implanted in the mice is going to be exposed

9 to much higher levels than we will ever seen in the

10 clinical setting and there is a good chance that

11 activity will be seen but it won't be replicated in

12 humans.

13 In the pediatric preclinical setting we

14 can use both the activity of the agent in our

15 pediatric preclinical models that test results, and

16 also the comparison of the mouse PK of the agent

17 with the PK of the agent in the initial adult

18 trials. We will be studying these agents or we

19 will be making our decision at a time after we have

20 some initial adult experience.

21 So, the most promising agents then will be

22 those that have activity in the pediatric models at

23 serum levels that are actually achievable or

24 systemic exposures that are achievable in humans.

25 Peter Houghton gave examples of this and I will

286

1 just reiterate two of those. The topo-1

2 inhibitors, irinotecan where incorporating PK led

3 to positive prediction for the activity of

4 irinotecan against neuroblastoma. Then,

5 incorporating PK correctly predicted inactivity for

6 another agent that he described, sulofenur.

7 Peter mentioned the data that we have that

8 support the potential for prediction, and I just

9 list those, the data that he described for activity

10 of agents in rhabdomyosarcoma xenografts mirroring

11 the clinical activity of these agents, the correct

12 prediction of activity for topo-1 agents against

13 both rhabdomyosarcoma and neuroblastoma. Another

14 point is that models now are not just limited to

15 rhabdomyosarcoma and neuroblastoma. Peter

16 described the Wilms tumor and some of the

17 predictive supportive data there.

18 Importantly, we also have xenografts for

19 acute lymphoblastic leukemia. Since this is the

20 most common cancer in children and a major cause of

21 mortality among children with cancer, it will be

22 important to also look at this in an in vivo

23 preclinical setting.

24 This is work from Richard Lock, published

25 in Blood a couple of years ago, just showing the

287

1 blast cells in the patient and then growing in the

2 NOD/SCID mice.

3 This is a table from that work showing

4 that when these lines are transplanted into mice

5 with no treatment there is a reasonably consistent

6 growth pattern. With treatment with an agent known

7 to be active against some childhood ALL cases there

8 is substantial growth delay for some cases;

9 moderate growth delay for other cases; and no

10 growth delay for some. Importantly, this in vivo

11 sensitivity to vincristine correlated with what we

12 know is an important measure of sensitivity in ALL,

13 the duration of the first complete remission. So,

14 we have the capability now to look at these ALL

15 xenografts to address this important disease.

16 An important contribution of the

17 preclinical models now is in the area of

18 molecularly targeted agents, and the ability to

19 make preclinical pharmacokinetic and

20 pharmacodynamic comparisons. Peter mentioned this

21 and I will reiterate it. Especially important in

22 this era of molecular targets, we can use these

23 models to identify the degree of target modulation

24 that is associated with anti-tumor activity, 50

25 percent inhibition, 75 percent, 90 percent, what is

288

1 needed in order to achieve anti-tumor activity; how

2 long does target modulation need to occur to

3 achieve the desired effect; and then particularly

4 important for children, what are the serum levels

5 or systemic exposures of the agent that are

6 associated with the requisite levels of target

7 modulation because it is going to be very difficult

8 for most childhood solid tumors especially to be

9 able to biopsy repeatedly tumor specimens to

10 measure this in children so we can understand the

11 pharmacology in children and target the systemic

12 levels that we have shown in the preclinical models

13 to achieve the desired level of target modulation.

14 This is also an opportunity to correlate anti-tumor

15 activity with gene expression profiles and protein

16 expression profiles.

17 One area that we are working in to try to

18 facilitate the evaluation of molecular targeted

19 agents is a project called POPP-TAP, or the

20 Pediatric Oncology Preclinical Protein and Tissue

21 Array Project. This is a collaboration between

22 NCI, both intramural and extramural, and Children's

23 Oncology Group researchers. The objective of this

24 collaboration is to develop tissue and cell arrays

25 and protein lysate arrays of pediatric preclinical

289

1 cancer models, primarily focusing initially on

2 xenografts and we are going to have close to 100

3 xenografts, different xenografts for which we will

4 have these tissue arrays available for study by

5 researchers. Also, Kahn's laboratory is

6 determining the gene expression profiles for these

7 pediatric preclinical cancer models, again focusing

8 initially on almost 100 xenografts for this. Then,

9 these data will be available for researchers as

10 well. We hope that this project will facilitate

11 the conduct and interpretation of preclinical

12 testing of targeted agents in childhood cancer

13 models.

14 The kind of complicating factors in

15 testing molecularly targeted agents--the comment is

16 sometimes made, well, you know the target is there,

17 just go after the tumors that express the target.

18 It is not that easy. One of the complicating

19 factors is the promiscuity of agents. A targeted

20 agent may hit multiple targets, some recognized;

21 some not. The Bay compound is one of many

22 examples. It was initially a raf kinase inhibitor.

23 So, there is promiscuity of agents in terms of

24 their targets.

25 There are multiple biological effects of

290

1 modulating a particular target of these so-called

2 molecularly targeted agents so farnesyl transferase

3 inhibitor in all the pathways that affects; the

4 proteasome inhibitors in all the different pathways

5 that that affects; Hsp90 inhibitors, all the

6 pathways affected there. And, it is very hard,

7 kind of on first principles of tumor biology, to

8 predict a priori what the potential applicability

9 of a particular agent such as this is to a

10 particular childhood cancer based on just its

11 biology. The preclinical testing then can allow

12 identification of previously unrecognized or

13 unsuspected activities that may have clinical

14 relevance.

15 I am often asked, in terms of addressing

16 preclinical activities, well, what about mouse

17 genetic models? Why aren't you focusing solely on

18 mouse genetic models? They have certainly made

19 critical contributions to our understanding of

20 cancer pathogenesis. In order to use genetic

21 models for testing, not all models will be

22 appropriate for testing. Really specific

23 properties are needed, particularly short latency

24 and high penetration for feasible testing are two

25 characteristics needed and not all models have

291

1 that.

2 But there are some genetically engineered

3 models for pediatric cancers that may have these

4 characteristics and be suitable for drug testing.

5 For example, the MYCN model for neuroblastoma may

6 be appropriate and we will try to use that if we

7 can.

8 The other caution is that a mouse is a

9 mouse, and mouse biology is not the same as human

10 biology. So, the lessons from the mouse genetic

11 models may not apply directly to the human setting.

12 There was an excellent review last year that really

13 documented this issue and made the point that more

14 humanized mice may more faithfully replicate human

15 cancers.

16 The preclinical testing program that we

17 have worked on over the last year or two to

18 initiate will be based on panels of xenograft lines

19 for the most common childhood cancers. It will

20 incorporate an in vitro testing component along the

21 lines that Peter Adamson outlined, particularly in

22 areas like the combination studies which may

23 provide valuable information.

24 We hope to be able to systematically test

25 10-15 agents per year, seeking to obtain agents

292

1 near the time that a commitment is made for the

2 initial evaluation in adults so that, by the time

3 the adult clinical experience is available and

4 there is evidence that this may be an agent that

5 could be studied in children, we will have

6 preclinical data to better address the question of

7 whether this is an agent that should be studied in

8 children. This will be implemented via a contract

9 mechanism with the primary contractor and the

10 potential for subcontracts for testing specific

11 cancer types.

12 The schema that Peter showed is shown

13 here. I will just make the point here that we will

14 be using panels of tumors. For example, if this is

15 a rhabdomyosarcoma, each panel is represented by 6

16 to 8 to 10 different xenografts, and then testing

17 at the MTD initially. When hits are identified,

18 activity is identified, then being able to go and

19 study the agent more intensively, look at a full

20 dose response, obtain PK data if that is not

21 already available, and do some of the molecular

22 studies if those are warranted.

23 A critical issue is addressing the

24 intellectual property issues. We have made efforts

25 over the past years to develop, in collaboration

293

1 with academic investigators and pharmaceutical

2 sponsors, a model MTA. This model MTA will be used

3 for all transfers by companies of their proprietary

4 compounds to NCI-supported investigators for

5 preclinical testing. Acceptance of the model MTA,

6 and it was included in the RFP for establishing the

7 preclinical testing program, but acceptance of the

8 model MTA is a requirement for participation in the

9 program.

10 I actually have some copies of the model

11 MTAs. There is one for transfer of the agent to

12 MCI and there is one for transfer of the agent from

13 MCI to the test sites. But Dr. Sherry Ansher is

14 the CTEP contact for those. If anyone wanted

15 copies, I would be glad to provide those to you.

16 In summary and in closing, appropriate

17 prioritization is key to future treatment advances

18 for childhood cancer. If we make good decisions in

19 terms of which agents we bring forward, and

20 particularly to the Phase III setting, we have a

21 chance for making advances. if we don't, then our

22 advances will be limited.

23 The Pediatric Preclinical Testing Program

24 may contribute to successful prioritization but

25 systematic preclinical testing of all agents

294

1 entering clinical evaluation in children should

2 become the standard of care, not because we know

3 what to do with these data now--we may have ideas

4 of what to do with these data, but because a

5 systematic approach is what we need to allow

6 validation of the panels and to optimize the

7 pediatric preclinical tumor panels. Thank you and,

8 again, thanks to the FDA for this opportunity.

9 Committee Discussion

10 DR. SANTANA: Thank you, Malcolm. We have

11 a few minutes for questions for presenters before

12 we go into the period of answering the questions.

13 Dr. Przepiorka?

14 DR. PRZEPIORKA: Thanks. Two questions,

15 one for either Malcolm or Peter. Peter had a slide

16 up there of I think it was MMI114 looking at a

17 single dose or dose schedule against a series of

18 tumors. If I recall, your conclusion was it was

19 not a very active drug because the AUC was ten

20 times greater than what one could expect to achieve

21 in humans. I was somewhat disappointed because I

22 could think of three or four drugs that we already

23 use for which we could probably have made the same

24 conclusion based on a single dose schedule being

25 tested.

295

1 So, my question for either of you is,

2 especially with the development of the new program,

3 is there an established panel of dose schedules

4 that will be used for drug testing so that you know

5 when a single high dose is going to be effective as

6 opposed to low continuous exposure before a drug is

7 thrown out?

8 DR. HOUGHTON: I think in the case of

9 NGI114 we have basically done other schedules. I

10 think what we would hope is that a fair amount of

11 optimization will have been done if we get a drug

12 from industry that is going into a clinical trial,

13 that a lot of the various schedules that have been

14 examined and information on which are the best

15 schedules will be made available at that point. I

16 think if you look at the size of the screening

17 program, if we went to doing the classic schedules

18 that you are going to use in the clinic, I don't

19 think the screening program has the capacity for

20 those; it certainly doesn't have the funding to do

21 that. So, I think for most drugs that will come

22 from industry, they may well have that information

23 already so that would at least allow us to do the

24 first cut using the optimal schedule that they have

25 and, in most cases, those have been quite accurate.

296

1 DR. PRZEPIORKA: If one has knowledge of

2 the mechanism of action and the pharmacokinetics,

3 could one potentially come up with the three best

4 guesses and so not have to do a whole bunch of

5 different dose schedules, or is that not a

6 reasonable approach?

7 DR. HOUGHTON: Again, a lot of that

8 information will be available to guide how we test

9 the drug in the screening program. I think we

10 still have to go to the MTD. I think that is

11 probably appropriate because one of the things you

12 want to do is get some idea of the tumor

13 sensitivity relative to an MTD in the mouse so

14 ultimately you want to do that with respect to

15 pharmacokinetics. So, irrespective of whether you

16 know the mechanism of the action of the drug, I

17 think the consensus was that you go for the MTD

18 even if you have a molecularly targeted drug where

19 you think it is a specific kinase inhibitor. That

20 is for two reasons. One is if you see no activity

21 that probably tells you that, you know, this is not

22 a drug that is suitable for treating certain

23 pediatric cancers. The other is that despite

24 having very strong evidence that a specific target

25 is, indeed, the target, when you go to the MTD you

297

1 may, in fact, reveal additional activities. I

2 think what we are trying to do is the minimum

3 amount of work, not because we don't like to do any

4 work but the minimum amount of work means minimum

5 utilization of resources to do a first cut to

6 identify those drugs that are worth pursuing, and

7 maybe looking at scheduling issues but to eliminate

8 those where we feel there is very little reason to

9 pursue that.

10 DR. PRZEPIORKA: If Dr. Chand Khanna is

11 still here, I have a question. I mean, GLP came

12 around because of some major issues in drug

13 development for adults and I would hate to see the

14 same problems arise in pediatrics because GLP was

15 not applied. The comment was made earlier that it

16 is too expensive for a vivarian in academia to

17 actually run under GLP but, given all the rules

18 that govern how you deal with animal care nowadays,

19 I can't imagine that it is not already running

20 under GLP. Does the Center for Comparative

21 Oncology animal housing at NCI--in your experience,

22 is that run under GLP and is it really a stretch to

23 try to get everybody who is going to be doing

24 preclinical testing to do that?

25 DR. SANTANA: Can you come to the

298

1 microphone, please?

2 DR. HIRSCHFELD: While he is coming, I

3 want to request the permission of the chair to have

4 Dr. Khanna and Dr. Meltzer take some empty seats at

5 the table and to have them join in the discussion.

6 I think there are empty seats between Dr. Weiner

7 and Ms. Haylock and there is an empty seat next to

8 Ms. Ettinger.

9 DR. KHANNA: Yes, a point of

10 clarification, at NCI we actually aren't going to

11 be managing pet animals in trials. We will be

12 managing those trials through veterinary teaching

13 hospitals that do operate under GCP guidelines in

14 many situations. So, that GCP hurdle is certainly

15 passed at many of those sites that we will be

16 working with and, in fact, it will be a requirement

17 for them to be involved in our cooperative groups.

18 DR. SANTANA: Donna, not to take this

19 discussion down a different route, but those for us

20 who are not familiar with the issues related to

21 GLP, since you hinted that there was an issue,

22 could somebody summarize what those are?

23 DR. HIRSCHFELD: I think we have a lot of

24 experts in the room but, in brief, the

25 International Conference on Harmonization, as well

299

1 as the FDA, have adopted standards under which

2 animal studies are conducted. These standards

3 collectively are referred to as Good Laboratory

4 Practice. Is Dr. Hastings still here? Do you want

5 to add anything to that?

6 DR. HASTINGS: [Not at microphone;

7 inaudible]

8 DR. SANTANA: Please us the microphone

9 because we really need to listen to the discussion

10 and sometimes it is difficult, and also record it

11 for the record. You can take the podium; that

12 would be fine.

13 FDA PARTICIPANT: GLP has many components

14 to it. It includes the test article and how stable

15 it is. The composition has to do with the people

16 that are involved with the research, like their CV

17 being on line, how they have been trained. It has

18 to do with the instrumentation and how they are

19 calibrated or if they are appropriate for the

20 testing that is being done. It has to do with the

21 animal husbandry, and how they are kept, the room

22 and the building, and many other components to it.

23 DR. SANTANA: Donna, did you want to

24 elaborate on that?

25 DR. PRZEPIORKA: Yes, I think from all the

300

1 talks that I have sat through, all the way back to

2 orientation, I believe GLP came around as a result

3 of some issues regarding fraud and poor science in

4 the late '60s, early '70s. I was just looking to

5 see if the poster was still up because I think I

6 remember the poster being up during orientation.

7 So, GLP came around as a result of a lot of

8 problems with scientific integrity in the initial

9 preclinical work that was handed in with drug

10 trials supporting FDA approval, and to have that

11 happen in the pediatric setting right now would

12 probably be a huge step backwards for pediatric

13 drug development.

14 DR. ADAMSON: I just want to clarify that

15 there is a difference between GLP and GCP. Without

16 question, pediatric trials are according to GCP.

17 What you are saying is the animal clinical trials

18 are going to be conducted according to GCP. That

19 is a different level of work but that is the

20 standard in pediatric drug development trials.

21 GLP, as we have just heard--there are very

22 few academic laboratories, adult or pediatric, that

23 do work according to GLP. That is the reality of

24 academic laboratories. There are very few that do

25 it according to GLP because the costs become

301

1 prohibitive. There are some laboratories that can

2 do it but I think they are a distinct minority.

3 Without question, would every place like to do it

4 according to GLP? Yes, but the funding is simply

5 not there to meet those costs.

6 DR. WILLIAMS: I must say that working

7 with our pharm tox colleagues we do not demand GLP

8 when we see a new IND, but we do demand that they

9 analyze where it differs from GLP and justify those

10 differences.

11 DR. SANTANA: Malcolm, were you going to

12 make an additional comment?

13 DR. SMITH: In the RFP for the preclinical

14 contract we did not specify GLP. That was at the

15 recommendation of colleagues in the Developmental

16 Therapeutics Program. You know, basically it is

17 what Peter was saying, that it would limit the pool

18 of researchers who could do that work. We will

19 have appropriate procedures in place so the

20 credibility of the results will, we hope, be above

21 question but we have not required that they meet

22 the GLP requirements in the RFP.

23 DR. SANTANA: Dr. Hirschfeld?

24 DR. HIRSCHFELD: I will make one further

25 comment and then maybe we can go to the questions.

302

1 GLP I think is more precise than GCP. GCP is very

2 open to interpretation and that was one of the

3 rationales for having our discussion this morning,

4 and it is a continuing source of guidances and

5 directives and other documents attempting to decide

6 how GCP can be applied to any particular study,

7 whereas GLP tends to be more explicit.

8 DR. SANTANA: Good. Any other questions

9 to the panel members or discussants?

10 DR. REYNOLDS: I wanted to tie a little

11 bit of what Eric was saying this morning to

12 comments made by Peter and particularly by Malcolm

13 where you suggested a standard of care would be

14 preclinical testing if we are engaging in human

15 studies in pediatrics. I think it would seem that

16 given what Eric was saying--this was really not a

17 point of discussion in the morning when we were

18 talking about monitoring but he did point out the

19 sort of ethics dilemma involved in facing a Phase I

20 study where you are looking at having to deliver

21 some prospect of benefit to a patient in the

22 context of doing the study. I think I would like

23 to suggest that we incorporate or think about some

24 sort of way that the agency might incorporate

25 Malcolm's suggestion of a standard of care, of

303

1 having some sort of preclinical data in the

2 pediatric tumor setting before engaging in testing

3 these agents in the pediatric setting.

4 DR. SANTANA: Peter?

5 DR. ADAMSON: I would actually put out a

6 caveat that that would be a goal to try to realize

7 perhaps within the next five or ten years. The

8 large majority of agents today that are active

9 drugs for children with cancer have not gone

10 through preclinical testing. There is a lot of

11 inactivity in industry with very important drugs.

12 So, it is an ideal we would like to move towards

13 but I think we are many steps away before saying

14 that that is the standard of care.

15 DR. SANTANA: Dr. Smith?

16 DR. SMITH: The other caveat would be that

17 in the future it could be a standard of care

18 because we have predictive models that we are

19 confident of and it makes sense to act on our

20 knowledge of these predictive models. That is in

21 the future, why it should be the standard of care.

22 Why it should be the standard of care now is

23 because if we don't do it systematically and obtain

24 the experience, then we won't ever get to that

25 future. So, for now the standard of care is

304

1 because only by systematically approaching this

2 problem can we develop the data that gets us to the

3 point where we are confident making decisions based

4 on these data.

5 DR. REYNOLDS: Absolutely, but I think

6 what Peter said in one of his slides is quite true,

7 and that is under the ideal circumstances of a good

8 laboratory model, if you can't get good responses

9 in your disease type you are probably unlikely,

10 almost assuredly unlikely to get those responses in

11 the children. So, I think that doing some testing

12 to exclude agents that we then would not be

13 exposing children to when they have no prospect of

14 benefit based upon what is probably a predictive

15 model--that is, if you don't get any activity in

16 the lab you are probably not going to get it in the

17 clinic--should be at least a consideration.

18 DR. SANTANA: Dr. Houghton, I think you

19 had a comment?

20 DR. HOUGHTON: Only to add that I think in

21 five years time we will have a much better idea of

22 whether this is correct or not. I think the one

23 thing that perhaps didn't come out strongly enough

24 from maybe the three of us is that what we are

25 proposing in terms of PPTP is an experiment and we

305

1 don't know how accurate the models are going to be.

2 We don't know what the flaws or the limitations

3 are.

4 So, in a way, I think it would be also

5 inappropriate if you had no activity in the model

6 not to pursue that at a clinical level because, in

7 fact, they may be very important experiments that

8 will reveal the fact that the models have

9 limitations. What we want to know at the end of

10 the day is with we are on the right track or the

11 wrong track, and if there are limitations to try

12 and address those in the next generation of models.

13 One of the biggest problems I see in the

14 development of models in preclinical development,

15 in the thirty years I have been playing this game,

16 is that we have these transitions, we transitioned

17 from syngeneic rodent models to xenografts, to in

18 vitro systems to xenografts, to perhaps transgenics

19 and nobody has taken the time to look back to see

20 what the problems were of the previous model that

21 would then allow us to develop a better model. So,

22 the next five years may be very revealing in terms

23 of the current models we have and their limitations

24 but give us the information that the next

25 generation of models won't make the same mistakes

306

1 as the previous models.

2 Questions for Discussion

3 DR. SANTANA: With those words of wisdom

4 and advice to all of us, let's go ahead and try to

5 tackle the questions for discussion. FDA is

6 requesting that we comment on three issues and, for

7 the record, I will go ahead and read the

8 introduction and the questions.

9 Because of the limited number of pediatric

10 oncology patients and because of the problems

11 unique to pediatric drug development, it may not

12 always be feasible to evaluate all aspects of

13 efficacy and safety in clinical studies. In some

14 settings, extrapolation of results from nonclinical

15 studies may be appropriate.

16 The first question is what types of

17 questions that are of potential clinical relevance

18 but are not feasible or acceptable to answer in a

19 clinical study could be addressed by nonclinical

20 studies? Then various examples are given after the

21 question that potentially could fit the answer that

22 we are being asked to provide.

23 I want to comment that one of the things

24 that I gathered from some of the discussion and

25 presentation this afternoon is that some of the

307

1 field is moving to molecularly targeted therapies,

2 whatever that means, and we may have limitations in

3 our patients in being able to correctly or early on

4 assess the correct target or do multiple biopsy

5 samples, etc., to see whether relevant targets are

6 being affected. I think in that setting, in which

7 the ethical issue of providing multiple biopsies in

8 a patient may be relevant or may not make the

9 clinical studies feasible, these models could be

10 used to address those very early on so that when we

11 get to the stage of testing these drugs in

12 patients, then sampling strategy may be very

13 limited or may be focused to such a degree that

14 ethically it doesn't become a constraint for the

15 study. So, that is one setting where I think some

16 of these preclinical models potentially could help

17 us in terms of limiting the ethical barriers we may

18 have when we introduce these molecularly targeted

19 drugs to our trials. That is one example that I

20 think would be relevant. Dr. Reynolds?

21 DR. REYNOLDS: I would like to suggest

22 another example. If one is dealing with agents,

23 two new molecular entities or new agents of which

24 one may have some modest activity and the other, as

25 a single agent, may have very little activity but

308

1 in combination in preclinical studies have striking

2 synergy, requiring that you demonstrate activity

3 for each individual agent in a patient, whereas if

4 you went in with the combination you might get

5 striking activity, and using the preclinical data

6 or nonclinical data, however you want to describe

7 it, to justify the approval of the agent as a

8 combination I think would make some sense, and

9 would spare children the ethical dilemma of being

10 treated potentially with an agent that is predicted

11 by preclinical data to be fairly non-effective, yet

12 might contribute to the overall response of the two

13 agents in combination.

14 DR. SANTANA: Dr. Adamson?

15 DR. ADAMSON: In looking at the examples,

16 Steve, that you have here, almost all of them are

17 looking at host and not tumor. I think that is

18 fine and helps us think about what you are after.

19 What I would caution is that we don't know, even as

20 far as host response or, you know, developing

21 animal models, how predictive they really are, and

22 the experience with the fluoroquinolones I think is

23 a good one. We are using them and we are still

24 learning what the real risk is. We should not

25 delay the initiation of pediatric testing of

309

1 anti-cancer agents for the results of these types

2 of studies because in the balance, of course, are

3 diseases that carry a far more certain outcome for

4 certain subpopulations of patients.

5 So, yes, we need to embark on some of

6 these. We need to realize the limitations as far

7 as predictiveness, and we should not mandate that

8 they become requirements to being the human testing

9 of anti-cancer agents.

10 DR. SANTANA: Susan?

11 DR. WEINER: One of the things that

12 occurred to me was that at least some nonclinical

13 data could be very relevant, obviously, to patient

14 selection for trials.

15 DR. SANTANA: Other comments or issues

16 related to this question? What I heard, Steve and

17 the rest of the FDA, was that these examples you

18 gave are relevant and, obviously, they are

19 dependent on what you are really after so you can't

20 put them all in one box for each drug. I think you

21 have to consider them based on each individual

22 agent which is more important in terms of what you

23 want in terms of using preclinical data.

24 You heard my comment about molecularly

25 targeted therapies and potentially how that could

310

1 be an area where some of these models could be

2 used.

3 You heard a little bit also that some

4 agents which potentially may not be totally active

5 but in combination, if you could do that

6 preclinically, you could demonstrate some

7 additional activity before you actually take it to

8 patients.

9 Then I heard comments related to

10 potentially how this could be used to identify

11 potential populations if you could do the

12 preclinical work in animals, looking at some

13 markers that potentially could select the

14 populations that would most benefit once you decide

15 to do the trials.

16 Then the last comment I think came from

17 Peter Adamson that while we do all this, this

18 should not hinder our ability to get the initial

19 clinical pediatric trials started but that they

20 should occur either in parallel or maybe a little

21 bit earlier, or wherever in time, but certainly not

22 to hinder the development even if this data does

23 not exist because, actually, a lot of the questions

24 may come after you do the initial Phase I, some

25 early Phase II studies, and you want to go back to

311

1 certain models and ask the questions that may be

2 relevant by scheduling--are you hitting the right

3 systemic exposure, and things like that. I think

4 the beauty of this system is that it has to feed

5 back to what you knew from before. Hopefully, that

6 is something that we will get from this experiment

7 that will be ongoing in the next few years, that

8 information will be used to go back and then ask

9 the relevant questions about why it didn't work so

10 that then, for the next series of experiments, we

11 can potentially address that. Dr. Helman?

12 DR. HELMAN: Victor, I want to reiterate I

13 absolutely support what you say, but also just to

14 reiterate what I think both Malcolm and Peter

15 Houghton said which is that, you know, in point of

16 fact this is an experiment. Many of us have spent

17 our lifetime trying to find better ways to identify

18 screening ways to pick winners for kids and for

19 treating our patients but we don't know.

20 Just as an example and, again, to support

21 what you said and what Peter Adamson said about not

22 mandating or requiring that, I think the GI stromal

23 tumors is a very good case in point. All we knew

24 is that GI stromal tumors were defined by their

25 mutation in the C KiT receptor. That was how the

312

1 entity was defined by a group of investigators in

2 Japan, and it allowed us to separate them from what

3 was called up until then GI leiomyosarcomas. All

4 we knew was that a drug that was active in CML had

5 in vitro activity against C KiT and that was the

6 extent of all the modeling of the data, period,

7 before it was given to a patient with a GI stromal

8 tumor. The rest is history. There was no

9 preclinical data. There were simply two

10 observations, GI stromal tumors had mutations in C

11 KiT and the STI571 AK gleevec had activity in vitro

12 against inhibiting that kinase. Everything else

13 came later. So, you know, we were lucky and I will

14 take luck over anything else any day. So, I think,

15 you know, maybe we will be lucky again.

16 In retrospect, you know, Paul had this

17 data to say that by profiling he could predict, and

18 I would like to hope that in preclinical models we

19 could say that it was absolutely clear that this

20 would have been a winner but we don't know that

21 yet.

22 DR. SANTANA: Paul?

23 DR. MELTZER: I just want to make one

24 comment somewhat in the same vein. I think in

25 pediatric oncology it is extremely important to

313

1 always bear in mind the very large spectrum and

2 number of rare cancers that we encounter and I

3 would not like to see those diseases orphaned from

4 the hope of developing good treatment because we

5 mandate the need for a preclinical model which will

6 never be practical to develop.

7 DR. SANTANA: Very good point. I think

8 the practicality of the issues that we have to deal

9 with in tumor systems in pediatric oncology is very

10 relevant to the discussion.

11 DR. DAGHER: Steven can also address this.

12 I don't think the intent of the question was to

13 imply examples where there would be additional

14 mandates. I think it was actually in response to

15 issues that have been raised by the cooperative

16 groups themselves and the Phase I Consortium about

17 those kinds of hurdles. It probably wasn't the

18 intent to ask for additional mandates, although

19 often when FDA asks a question, that is usually

20 what the fear is, that we are thinking about

21 additional mandates. That wasn't the intent.

22 DR. HIRSCHFELD: Just to add to Dr.

23 Dagher's precisely right answer, the intent was how

24 can we better inform the data we have? So, that

25 was the rationale for the entire discussion this

314

1 afternoon, how can we use nonclinical data so that

2 we can improve our conclusions and improve our

3 designs and use our resources most effectively?

4 DR. SANTANA: Good. Let's move on to

5 question number two, and I think I am going to ask

6 the FDA to clarify this question a little bit for

7 me, but the question relates to what types of

8 evidence and data would be recommended in each of

9 the following domains to allow extrapolation from

10 nonclinical data and be informative for a clinical

11 condition. There is pharmacology and

12 pharmacokinetics; safety; efficacy; behavior;

13 long-term effects; developmental aspects; and then,

14 question mark, other domains.

15 Maybe I would like the agency to clarify

16 for me what do they mean by types of data or types

17 of evidence so that we can address this

18 appropriately?

19 DR. HIRSCHFELD: This is a rather

20 theoretical question but it should be grounded in

21 the limitations of models and should be grounded in

22 data, but there are circumstances where one has

23 information in a domain and would like it to be

24 predictive, or at least informative, for some other

25 domain. So, in some cases formal rules or formal

315

1 mechanisms have been identified. As an example,

2 for the conversion from a laboratory measurement

3 from a biomarker to what could be called a

4 surrogate, where the surrogate is for clinical

5 benefit, the NCI and others have made specific

6 recommendations on what type of evidence one would

7 like to see. Going back about 160 years, there

8 were initially observations which were formulated

9 by Profs. Koch and Henley that there are some

10 conditions that would be met between the

11 identification of a microorganism and its causative

12 role in a disease.

13 So, we don't expect that for all the

14 various domains of clinical interest there are

15 formal rules to be identified, but what we would

16 like to have is some commentary on the type of

17 evidence and the strength of evidence so that if

18 someone is proposing a nonclinical approach we

19 could get some advice on whether we would consider

20 the data that are being offered as valid data, as

21 informative data. If you want further elaboration

22 we could try, but I think that is the general

23 concept.

24 DR. SANTANA: If I understood you

25 correctly, I am going to try and see if I follow

316

1 you to contribute to (a). I think we heard this

2 afternoon how systemic exposures or AUCs of certain

3 drugs can potentially, in certain animal models,

4 predict reduction in tumor volume--not cures but

5 reduction in tumor volumes at the appropriate MTD

6 that are clinically relevant. So, I think that

7 would be a good example that, if there was good

8 systemic exposure data at the MTD that was

9 clinically relevant in adults that then was going

10 to potentially begin the pediatric studies at that

11 MTD or near that MTD, and there was good response

12 data in animals at that MTD, to me, that would be

13 information that would be relevant to addressing

14 the issue of how pharmacokinetic data could be used

15 in a nonclinical setting in a preclinical kind of

16 model.

17 DR. HIRSCHFELD: So, if I may paraphrase,

18 and have that then inform the answers to the

19 others, and they may not be the same types of

20 answers, but one has a set of techniques that are

21 available in the nonclinical model that are also

22 available in the clinical model so that one can

23 make direct correlations because the technique for

24 determining AUC is the same in the nonclinical

25 model as in the clinical model and then you are

317

1 relating the readout, applying that technique and

2 then making a direct correlation. That would be

3 paraphrasing it, but the concept there is that you

4 have techniques which are identical or potentially

5 could map onto each other, and having that assay

6 availability is what lets you make the

7 extrapolation.

8 DR. SANTANA: Peter and then Donna.

9 DR. ADAMSON: I think other examples, and

10 it comes back to the need to do tumor biopsies or

11 repetitive tumor biopsies--I think if you can

12 demonstrate in an animal model or, preferentially

13 in animal models, that you have a surrogate that is

14 reasonably predictive of what is happening in the

15 tumor, that should weigh in when looking at the

16 effect in a patient. So, if you are

17 down-regulating expression of a target in a tumor

18 but you also see it in a lymphocyte and you have a

19 pretty strong correlation in your animal model, it

20 is a lot easier to get lymphocytes from children

21 than it is to get tumors from children. So, I

22 think that should weigh in as part of proof of

23 principle that you are hitting a target when you

24 actually don't have repeat access to that target.

25 DR. SANTANA: Donna?

318

1 DR. PRZEPIORKA: Actually, I would like to

2 ask for additional clarification on this question

3 because I recall one of your first slides in your

4 prior talk was, I believe, that the rule is that

5 you need at least one clinical trial as supportive

6 evidence. My question is regarding strength of

7 evidence. Do you want us to be considering

8 sufficient strength of evidence to be the sole

9 supporting data for that one clinical trial because

10 pediatric cancer is an orphan disease and you may

11 not get the chance to do anymore clinical studies?

12 DR. HIRSCHFELD: Well, if I understood,

13 and we can try to clarify this to be sure we are

14 both addressing the same issue, yes, it is most

15 likely that many pediatric malignancies, for

16 reasons that Dr. Meltzer mentioned, because they

17 are quite rare, will only have one study being

18 done. Dr. Smith elaborated just on the resources

19 of that too. So, if we are only going to get one

20 study, there are ways that we can improve our

21 interpretive ability of whatever the clinical

22 outcome may be, either safety or efficacy or

23 long-term effects or something, by using

24 nonclinical data.

25 DR. SANTANA: Malcolm?

319

1 DR. SMITH: If there is the one pediatric

2 trial, the one Phase III trial that shows a p value

3 that is favorable and you are looking for something

4 else to help you justify that this is approvable,

5 then looking at a robust preclinical data set that

6 shows the same kind of responses or anti-tumor

7 activity in the preclinical models would seem to be

8 supportive at least and provide you some additional

9 confidence that the agent was going to behave in

10 larger groups of patients as it had in the trial.

11 DR. HIRSCHFELD: Let me turn it around a

12 little bit. I guess initially all of you sitting

13 on that side of the room--and since this is an

14 audio recording, it would be Drs. Smith, Helman,

15 Adamson and Anderson--you are starting a fairly

16 extensive program which you acknowledge is an

17 experiment. So, one way of helping us would be how

18 are you going to know at the end of five years that

19 you have had a successful or an unsuccessful

20 experiment? And, what are you measuring that is

21 going to determine that? We would be interested in

22 getting an answer from each of you.

23 DR. SMITH: I will say something and then

24 let Peter chime in as well. You know, some of the

25 testing that we do will be to go back and take

320

1 agents that are already being used, for which there

2 is some background response data from the clinical

3 setting, and look at the operating characteristics

4 of the various tumor panels against those agents.

5 So, there will be kind of building of a baseline

6 for agents that we already have activity data for.

7 The others will then be looking ahead

8 prospectively. If we have agents that have been

9 tested and moved from the preclinical to the

10 clinical setting, is the activity observed

11 preclinically replicated in the clinical setting?

12 DR. ADAMSON: The clinical endpoints are

13 going to be Phase II endpoints for this experiment,

14 and you have probably heard the reasons why from

15 Malcolm's talk as far as our ability to do Phase

16 IIIs. But some of those Phase II endpoints are

17 going to be traditional objective response rates or

18 time to progression and I think in part may depend

19 on the agent and our ability to monitor those

20 endpoints.

21 But I should point out also that even in

22 the ideal setting in the next five years where

23 every drug that we potentially want to study will

24 be put through this system, this is not going to be

25 the only path to doing a clinical trial in children

321

1 with cancer. I can think of a number of

2 circumstances where almost independent of what we

3 see in our model system we are going to be doing

4 clinical studies. The obvious examples are agents

5 that have remarkable activity in adult cancers. We

6 are going to look at them in pediatric cancers like

7 we have historically looked in pediatric cancers.

8 And, part of the experiment will be if, in

9 fact, the model predicts lack of activity and we go

10 ahead because of other justifications and find the

11 lack of activity, that is going to also help the

12 negative predictive side of things. The positive

13 predictive side of things, whether we look at

14 relative response rates of simple yes/no, it met

15 activity thresholds or not, I think will depend

16 upon how many patients and how quickly we can get

17 Phase II trials going. But there will always be

18 more than one path to get a trial into children

19 with cancer. The goal, however, will be to put

20 everything that, for whatever reason, has got to a

21 clinical trial through our model system so we can

22 learn both positive and negative predictive values

23 using Phase II as the endpoint. We would like one

24 day then to start building in toxicity information

25 but right now that is a primary goal of this

322

1 program.

2 DR. SANTANA: So, if I understood, I think

3 you guys are going to try to address (a) and (c),

4 the pharmacology and pharmacokinetics and efficacy

5 in your models and use that data to decide whether

6 you move on to different model systems or whether

7 you start to introduce other domains, like looking

8 at toxicity and things like that.

9 DR. HIRSCHFELD: Were there other

10 comments?

11 DR. HELMAN: Well, again, I maybe would

12 rather address not necessarily the predictive value

13 of the models but the biologic importance of

14 gaining more information. For example, you know,

15 you heard Malcolm briefly discuss the hope that we

16 can have both some protein profiles, RNA profiles,

17 and if there are subsets--I mean, we are going to

18 use six to ten models so it may be--I have yet to

19 do even a mouse experiment where I consistently

20 cure 100 percent of the mice. Usually it is 90

21 percent in really good experiments, and sometimes

22 60 percent. So, if we can identify correlates of

23 response, things that Paul Meltzer talked about,

24 and then find that these are, in fact, important

25 biologic discriminators between people likely to

323

1 respond, for reasons we may have no idea, and just

2 generate hypotheses and if that correlates at all

3 with somehow what we then can use in the clinical

4 study, I think we will make some important steps

5 forward.

6 I would just make the comment that it is

7 something we try to hold ourselves to now because,

8 you know, I think although we all like to think

9 that there are ten more gleevecs out there, the

10 likelihood of hitting a grand slam when we do

11 clinical studies is extraordinarily small. So, if

12 we do a clinical study with a therapeutic endpoint

13 and the therapeutic endpoint is negative but we

14 learn an important biologic principle, we will

15 continue to make progress. If the only thing we

16 learn is that this is inactive, we have put a lot

17 of patients into a study that we come out not

18 knowing anything more, other than that this thing

19 is not active.

20 DR. HIRSCHFELD: Right. If I may just

21 follow that up, that is exactly the direction where

22 we would like to get some more advice on and

23 thinking. So, could you elaborate on what you

24 would mean by an important biologic observation

25 even if the clinical result is disappointing?

324

1 DR. HELMAN: Well, the easiest thing would

2 be we have a kinase that we think is important for

3 the biology of the tumor. We give a drug. It

4 inhibits the kinase and all the patients progress.

5 In the end we have learned a very important point

6 which is that that enzyme is irrelevant for the

7 progression of this disease in a patient. I think

8 that is an incredibly important observation to

9 make.

10 DR. SANTANA: Dr. Reynolds?

11 DR. REYNOLDS: I think that one of the

12 things we have to keep in mind when we are talking

13 about these kinds of transitions that you are

14 talking about, Lee and Peter as well, is that the

15 clinical experience in your Phase IIs will be

16 pretty much in patients that are refractory to

17 existing agents. In some diseases one can imagine

18 that is sort of like up-front patients but for the

19 most part that is patients who have gone through

20 therapy and maybe years out from therapy and it

21 recurred.

22 So, I think in the context of that and

23 thinking about the way the FDA looks at things

24 where they generally approve an agent for a

25 specific indication, like for second-line therapy

325

1 in disease X, we have to keep that in context in

2 the preclinical modeling and we have to make sure

3 that the preclinical modeling doesn't just reflect

4 up-front patients but that it also reflects this

5 refractory population so that we can make those

6 correlations. For example, what you were talking

7 about, Lee, where you hit your molecular target and

8 you get zero responses, that doesn't mean that the

9 agent wouldn't necessarily work in up-front

10 patients and be an effective agent, and maybe your

11 preclinical models would have said that it worked

12 but then they all developed drug resistance that

13 got around it.

14 So, all those are very complex issues and

15 I think we are going to have to spend a lot of time

16 thinking about these but, more particularly, spend

17 time developing the models so that they reflect the

18 clinical setting as much as we can.

19 DR. SANTANA: Steve, did you get what you

20 wanted from the panel?

21 DR. HIRSCHFELD: If I may summarize at

22 least what I heard, and then I will let you, of

23 course, do the more formal summary as we pursue it

24 just a little more because I think this is an

25 important discussion, the context would be that

326

1 people are very interested in nonclinical models.

2 The question is how informative are those data.

3 So, what we have heard so far is that if you have

4 the same technique to measure something, whatever

5 that may be, in the nonclinical model and the

6 clinical model you can do a direct correlation.

7 If you have surrogates in the clinical

8 model that could map onto the nonclinical model,

9 without defining how those surrogates are validated

10 but we will presume that there is a validation

11 process in effect, that could also be used as a

12 mechanism to inform.

13 We also have an approach, to go back to

14 something Dr. Meltzer referred to, training, that

15 we have historical clinical data which then can be

16 used to validate a nonclinical model by using the

17 same types of agents in the nonclinical model and

18 seeing if it correlates to the historical record.

19 So, that is yet another approach.

20 Then we have prospective testing as an

21 approach where we would ask a question of the

22 nonclinical model and ask either the same or what

23 we think is a related question to the clinical

24 model and see if the answer comes out in a way that

25 it is either identical or can be mapped.

327

1 Then, lastly, we have biologic correlates

2 where we are not asking a specific outcome

3 mechanism of the clinical circumstance but we are

4 just trying to pick up information to help

5 mechanistically understand, and then go back to the

6 nonclinical model and use that as some form of

7 evidence.

8 So, that is what we have heard so far, and

9 I think that is all highly useful but, since this

10 is a new area, we want to take the opportunity

11 while we have the expertise available and these

12 presentations fresh in mind to see if there are

13 other aspects that ought to be probed because in

14 some ways we can, hopefully, at least inform if not

15 partially drive a research agenda to improve the

16 validation process.

17 DR. SANTANA: Dr. Reynolds?

18 DR. REYNOLDS? Steve, in general what we

19 have been thinking about in terms of when you think

20 about labeling indications and looking for a

21 positive result is to say, okay, this has efficacy

22 in a particular tumor type. What about the

23 negative condition? For example, if an agent was

24 to go through clinical trials and show activity and

25 have a registered indication for a pediatric tumor

328

1 but preclinical studies showed that there was a

2 subset of that very disease that was very unlikely

3 to respond to it and there were some limited

4 clinical correlations that showed that was the

5 case, could that be incorporated in the label and

6 used as informative information for pediatric

7 oncologists? How would the negative side be

8 approached?

9 DR. HIRSCHFELD: Well, that is exactly one

10 of the scenarios we have been anticipating. I will

11 give a very brief comment on the aspects of that.

12 First, the question is not restricted just to

13 product labeling. We are in a position of

14 attempting to advise people on a continuing basis,

15 primarily the pharmaceutical industry but also

16 investigators, saying what type of studies would

17 you like to see? This is a question that is asked

18 essentially on a daily basis, and all of us spend

19 probably at least 40 percent of our time meeting

20 with people and attempting to answer their

21 questions in this regard. So, I would view it as

22 the spectrum, and that includes our colleagues

23 whose focus is the domain of nonclinical data. So,

24 I would view this as a spectrum of how to best

25 utilize resources all throughout the developmental

329

1 cycle of any product and not restrict it just to

2 the labeling.

3 Now, the other aspect is how can we use

4 negative information? We have used that clinically

5 but I think what you are asking, and this is

6 something that we discussed in April, 2001

7 previously, and that is should negative data inform

8 us to not invest the resources nor expose patients

9 to risk for a given agent? Now, three years later

10 almost, we would like to ask the question--we are

11 very interested in that because of the potential

12 savings, but what kind of evidence should we use to

13 have confidence in those negative data?

14 DR. REYNOLDS: If I could just ask Peter,

15 your point being, well, if the agent has some

16 activity some place it should be tested in

17 pediatrics, where could the interface between

18 preclinical model testing that shows it is probably

19 not going to work and limited clinical data in the

20 pediatric setting come together to diminish the

21 number of patients exposed to a potentially

22 ineffective agent?

23 DR. ADAMSON: As Peter Houghton said, I

24 think until we do this systematically we are not

25 going to be able to answer this question because we

330

1 are just going to have biased data. So, if we can

2 do it systematically and we can build an experience

3 as far as what these models' positive and negative

4 predictive values are, then I think we really can

5 start making informed decisions when we see

6 negative data that we shouldn't pursue it.

7 Given the limitation of resources, even

8 before we have that data we are likely to apply

9 some of this on an assumption that they are going

10 to be predictive. But historically, as well as in

11 the current environment, when an agent comes on

12 market for an adult indication it will almost

13 invariably be used by physicians of children who

14 have refractory cancer. That is the reality. So,

15 we might as well, for agents that are clearly

16 active and as long as it is not beyond the realm of

17 scientific plausibility--I mean, we are not

18 studying estrogen receptor--well, I shouldn't say

19 that; probably people are--

20 [Laughter]

21 --someone should be able to come up with

22 an example of what wouldn't be used in a child.

23 These drugs are going to be used until we have

24 convincing evidence our models have both positive

25 and negative predictive values. As Peter said,

331

1 hopefully, in five years we will be able to give

2 you a better answer to that question.

3 DR. HIRSCHFELD: True enough. I will just

4 state that we have labeled products that do not

5 have what we consider to be activity in children on

6 the basis of clinical data, sometimes using up to

7 100 children with no evidence of efficacy, at least

8 in a particular disease or particular dose. We

9 have labeled these things, that they should not be

10 used in children and we are very interested in

11 making sure that there is not inappropriate

12 exposure.

13 DR. SANTANA: Kind of following that

14 discussion, I think the issue of negative data--you

15 know, it depends on whether you can explain why the

16 data is negative. That is the critical issue. It

17 is not that it is negative data because negative

18 data can be very good data. It is can you explain

19 why it is negative, why it failed? If you can find

20 the reasons why in your particular experiment it

21 didn't work, to me, that is very informative data

22 and it should not go out with the baby. You know

23 what I am saying?

24 So, it is a very theoretical discussion of

25 this issue because if you don't do the experiment

332

1 correctly you wind up with negative data, but if

2 you do the experiment correctly and you wind up

3 with negative data and explain why it was negative,

4 to me, that is an advance and I think that should

5 not be thrown out. Donna?

6 DR. PRZEPIORKA: Actually, just thinking

7 about Eric's slides from this morning indicating

8 that in the pediatric setting at least we are

9 looking more towards beneficence and doing good for

10 the patient, and having sat on an IRB, I was jut

11 wondering under what circumstances would I get a

12 protocol for a pediatric study that says there is

13 no evidence that this drug is effective in tumors

14 that kids have but we are going to do a Phase II

15 study? That would be a very difficult protocol to

16 pass through an IRB.

17 DR. ADAMSON: I agree but there are a lot

18 of protocols that come where there is no data in

19 children. It is a cytotoxic and there is no data

20 in pediatric models and we do those studies because

21 we accept that cytotoxic agents likely do have

22 activity in pediatric malignancies as a class. It

23 is a horrendous problem when you think about how

24 little data we base it on. There has to be some

25 scientific plausibility that the drug is going to

333

1 work.

2 Related to that, I can almost guarantee

3 that gleevec has been tried in every pediatric

4 malignancy to some extent. What we would much

5 rather do is say let's study it where we think

6 there is scientific plausibility, and we are doing

7 that now on very limited data, basically which

8 tumors do we think express kinases that gleevec

9 might inhibit? At least that gives us scientific

10 rationale and will give an answer. If it is

11 negative, I think that is important information

12 because then at least we have the data, we put it

13 out there and people aren't exposing children to

14 gleevec simply because it is the most active agent

15 in CML. The same is true for adult malignancies as

16 well. I bet gleevec has been used in virtually

17 every adult cancer that exists by someone.

18 DR. KHANNA: It is also used in almost

19 every veterinary cancer--

20 [Laughter]

21 -- but one thought I wanted to follow-up

22 with on Peter's comments was that the models are

23 validated or found to be predictive within the

24 context of the agent that was assessed so that

25 agent X with model Y, if there is activity, doesn't

334

1 say that that model is a predictive model for a

2 cancer in general. So, I think there is a

3 complexity there that has to be incorporated in the

4 next step of the analysis.

5 DR. SANTANA: Dr. Grillo?

6 DR GRILLO-LOPEZ: If I may, I would like

7 to focus on the issues at hand in a little bit

8 different way. Clearly, the medical need that we

9 are discussing is a need to make new effective

10 therapeutic agents available to children as soon as

11 possible. Now, in the setting of the interaction

12 between the agency and a pharmaceutical company you

13 might look at two extremes. One extreme might be

14 where an agent is to be developed exclusively for a

15 pediatric malignancy and may not have any

16 applicability in adult malignancy, and those may be

17 very few and far between. But in that situation I

18 guess the agency has to be more rigorous about the

19 clinical data that needs to be submitted and

20 supported by preclinical data than the other

21 extreme, an agent that is clearly active in adult

22 malignancies and where you could make the

23 extrapolation that it should be active in pediatric

24 malignancies.

25 Most of those agents are the agents that

335

1 we have today in our armamentarium, and most of

2 them have been approved with very little pediatric

3 experience, if any in some cases. One of the

4 questions is if you do have an agent that is very

5 active and that deserves to be approved for an

6 adult malignancy whose responsibility is it to do

7 the studies to show whether or not it applies in

8 childhood malignancy? On the one hand, there is

9 the need to find out; on the other hand, there are

10 all of the obstacles that we have discussed today

11 and the fact that there are not enough patients of

12 pediatric age to go around. We can't do the

13 studies in all of the available agents even today

14 and on the other hand there is the need that we

15 have. So, as a medical community interested in the

16 cancer patient, we need to find out whose

17 responsibility it is to do those studies.

18 DR. SANTANA: I think it is all of our

19 responsibility, everybody in this room.

20 DR. GRILLO-LOPEZ: I think that is the

21 answer.

22 DR. SANTANA: That is why we are here and

23 we have been here for a long time.

24 DR. GRILLO-LOPEZ: Let me go further, that

25 answer says that it is not the exclusive

336

1 responsibility of a pharmaceutical company and,

2 therefore, should not be a requirement for approval

3 of an agent that is shown to be active in adult

4 malignancy. However, how do we approach the issue?

5 The issue can be approached in a variety

6 of ways with the support of the nonclinical data

7 that we have discussed here today, and the simplest

8 way might be to produce clinical and nonclinical

9 evidence that the pharmacology and the

10 pharmacokinetics are similar to those of adults and

11 that the safety profile is similar. That could go

12 into a package insert without requiring that it be

13 an indication. Another more stringent way would be

14 to have it go into the package insert of an

15 indication, and there you would require at least

16 Phase II trials as a minimum.

17 DR. HIRSCHFELD: Rather than addressing

18 the specifics of what goes in product labels and

19 what does not, I would like to summarize by saying

20 it seems that for all of the nonclinical models as

21 they may apply to pediatric oncology we have

22 question marks. So, I think collectively we should

23 encourage validation and we should encourage

24 multiple approaches to the models so that we can

25 gain confidence in the models and, by gaining

337

1 confidence in the models we can begin to move

2 toward the scenario where the models and the

3 clinical data can be weighted in such a way that we

4 can have a better understanding of what we are

5 looking at.

6 DR. SANTANA: Yes, I think you said it

7 well. I think while we move towards perfection, if

8 we could ever reach perfection, the systems that we

9 have at hand have served us to some degree and we

10 should not hinder development of any pediatric

11 studies until those models are truly validation and

12 we have the answers to all the questions. I think

13 what we have done up to today has served us to some

14 degree and I think the agency needs to recognize

15 that and deal with each one of the drugs or the

16 compounds or the issues at hand on a case-by-case

17 basis, obviously trying to formalize things in such

18 a way so that everybody kind of does it in the same

19 way until we reach that point of perfection. I

20 think you heard earlier today that it should not

21 hinder our progress until we can validate all these

22 domains and models and come back to you and say

23 this is the best way of doing it. I don't think we

24 are there yet, and I think that is the difficulty

25 of why we struggle with this question.

338

1 DR. HIRSCHFELD: Right.

2 DR. SANTANA: It is very theoretical but

3 we are not there yet. We can give you some

4 examples but we can't give you the whole universe.

5 DR. HIRSCHFELD: Clearly. So, thank you

6 for those examples. Maybe, in the remaining

7 minutes, we could try to touch on the last

8 question.

9 DR. SANTANA: Exactly where I was heading.

10 The last question is are there additional

11 recommendations for the effective use of

12 nonclinical data? For example, will open

13 literature reports be generally acceptable? Is

14 documentation of compliance with Good Laboratory

15 Practice necessary to evaluate animal data? Should

16 nonclinical data be submitted as an independent

17 report with a presentation of primary data

18 sufficient for verification and review?

19 I am going to try to skip to the last one

20 and ask the agency how they would use this

21 verification and review when this preclinical data

22 is being presented. How are you going to judge

23 that data? It is not just that the data is

24 submitted to you, but what tools and what processes

25 will you use to verify and to review the data?

339

1 Because I think that will be critical in terms of

2 getting the acceptance of individuals to submit

3 that data--

4 DR. HIRSCHFELD: Sure.

5 DR. SANTANA: --whether independent or

6 part of the submission.

7 DR. HIRSCHFELD: In brief, if we don't

8 have a track record for pediatric oncology it is an

9 open arena so we are attempting to just gain some

10 input into what would be considered acceptable

11 levels of evidence in this regard. We have much

12 more experience in moving from preclinical to the

13 IND phase, but if we are looking for the

14 nonclinical data to supplement clinical data this

15 would be a new area for us. So, we don't have

16 precedents and we can't comment to you, for a

17 variety of reasons, about what we would like to

18 see. We are just trying to get a sense from our

19 invited experts for what you would consider to be

20 acceptable.

21 DR. SANTANA: Yes, I think the quandary we

22 get into is--

23 DR. GRILLO-LOPEZ: Clarification, please,

24 acceptable for what?

25 DR. HIRSCHFELD: Verification of clinical

340

1 findings.

2 DR. GRILLO-LOPEZ: I am sorry to insist on

3 the clarification. Although you don't want to talk

4 about labeling but it is an important issue because

5 you could be saying acceptable for labeling.

6 DR. WILLIAMS: I might elaborate just a

7 little. I think certainly we do include in our

8 labeling a lot of different pharm tox, biopharm, a

9 lot of different kinds of data and we do accept all

10 kinds of data for clinical use also. The general

11 principles are that, at least for clinical, we

12 often go out and audit but we have sometimes, in

13 circumstances where we have multiple different

14 literature references that all point to the same

15 thing accept the paper. Then, as I mentioned

16 earlier, when we get pharm tox data in we generally

17 like to have data to review and generally, if it

18 doesn't meet the GLP standards, we like people to

19 sort of specify how that differs.

20 So, I would sort of maybe even propose

21 that in general those same kinds of standards would

22 probably apply to nonclinical data, that if you

23 didn't do it according exactly to our standards you

24 certainly would support it in some way.

25 DR. SANTANA: Clarify for me, when a

341

1 sponsor comes to the agency with an NDA and there

2 is preclinical data there, that data gets reviewed

3 and you already have defined what strategies you

4 are going to use to review that data. What you are

5 implying is that those same parameters would be

6 used for some of these experiments that we are now

7 undertaking.

8 DR. WILLIAMS: I guess what Steve was

9 saying is when we are talking about a

10 pharmaceutical company that is doing everything

11 under GLP, that is one thing. It looks like in

12 this setting we might be getting different kinds of

13 data that aren't necessarily exactly as pure as

14 that. Recognizing that, I guess to what extent

15 would you go to either compromising or specifying

16 in a certain area certain rules or parameters

17 before you would accept it?

18 DR. SANTANA: Dr. Smith and then Dr.

19 Helman.

20 DR. SMITH: Certainly for the contract we

21 are involved with, if FDA has recommendations in

22 terms of reports, we would be glad to consider

23 those and to incorporate those and provide you with

24 reports if those are what the agency needed for a

25 particular consideration.

342

1 DR. HIRSCHFELD: I will just address that

2 before Lee speaks. We are asking you today for

3 recommendations because we don't have a position

4 yet. So, that is where we stand.

5 DR. SMITH: Grant described some kind of

6 characteristics that you might be looking for so we

7 would be open to considering the report formats

8 that would be easier for you to review and be more

9 informative to you.

10 DR. SANTANA: Lee?

11 DR. HELMAN: I wanted to ask a question

12 because actually I think it was Dr. Hastings who

13 mentioned this, and nobody has followed up on this

14 and I found it very intriguing, and it follows with

15 some of the information that Chand discussed, which

16 is if we use spontaneous animal models to test the

17 efficacy of a compound and we collected toxicity

18 data, would that be enough if the toxicity data was

19 of high enough quality to not then require

20 additional toxicity data in healthy animals? In

21 fact, I think there is data to suggest that

22 tumor-bearing animals have toxicity that is not

23 necessarily the same as healthy, normal small

24 mammals. I mean, it is something we haven't really

25 discussed, which is the coupling of efficacy data

343

1 in pet models and toxicity data, and would that be

2 valid enough to then not require the standard

3 beagle dog or rhesus monkey toxicology?

4 DR. HASTINGS: Well, first, this is

5 obviously a decision for the oncology division to

6 make about what would be sufficient, but depending

7 on what you knew about the toxicology of a drug to

8 start out with, yes, you might be able to have that

9 as a complete package to support both safety and

10 efficacy. I think the important issue here

11 though--and this is my own personal opinion and I

12 am not speaking for the division, but what we

13 really would like to have, what I would really like

14 to have is the raw data. Remember, GLP is

15 basically a set of bookkeeping rules to ensure the

16 integrity of the study and the validity of the

17 data. That is really what it is all about. Maybe

18 you won't have a quality assurance statement or

19 anything like that, but I think that is what we

20 would want to have in order to know whether or not

21 the safety data you acquired in a diseased animal

22 model, in fact, is valid enough to make a decision

23 about safety in that condition. But I think that,

24 yes, you can get toxicity in, as you said, a

25 spontaneous animal model that actually might be

344

1 more relevant to the actual indication than the

2 kind of toxicology data you would get in a healthy

3 animal. Does that answer your question?

4 DR. HELMAN: To me, it is really a new

5 concept.

6 DR. HIRSCHFELD: Our approach is that we

7 will be naive and just for a moment pretend there

8 was no FDA and you don't have to ask us how we want

9 it and you are just trying to make a decision. So,

10 you have no clinical data and what we are

11 anticipating is that GLP could potentially be a

12 burden on people so you are going to do something

13 less than GLP and you are going to use it yourself

14 to make decisions and to determine whether the

15 model is good or not good.

16 So, what we are asking here is, given that

17 GLP could not necessarily be the standard you could

18 practically adapt, what is the standard that you

19 are comfortable with? What would you look at; what

20 would you read that you would say, well, this is

21 valid? So, that is what we are asking.

22 DR. SANTANA: Dr. Reynolds?

23 DR. REYNOLDS: Steve, I think the issue,

24 as you hit the nail on the head, is that GLP, which

25 is a very good concept, is not necessarily

345

1 adaptable to the academic setting where limited

2 resources are brought to bear especially on

3 pediatrics where resources are limited. Whereas

4 the pharmaceutical industry has the investors to

5 spend those resources, we do not necessarily in the

6 academic laboratory have those.

7 The problem is when you say less than

8 that, what would we say is acceptable, well, I

9 think everyone in this room can think of examples

10 from the far end of the spectrum of data where you

11 would ask, well, how did that ever get published

12 all the way to data which according to the

13 regulations is not GLP but is what people would be

14 very confident in using for any purpose.

15 So, what I think we need is not for us as

16 a committee to answer your question, but actually

17 for some guidance from the experts in the FDA that

18 look at GLP issues as to what kind of standards one

19 could apply that are less than full rigor that

20 would be acceptable for the purposes that we want

21 to use these data for.

22 DR. WILLIAMS: I know that our division

23 commonly accepts things that are not GLP, but we

24 just have the applicant look at the sections of the

25 GLP and tell us how they differ and how they think

346

1 they meet the spirit of it. So, I think that is

2 doable. Maybe it could be doable in a more formal

3 setting that met your particular needs for what you

4 are dealing with whether it is tumor models or

5 whatever.

6 DR. REYNOLDS: Just to finish that, I

7 think this is really important in the concept of

8 what Lee was kind of talking about in terms of

9 using these pet animals, which are fascinating

10 models, because they are never going to make GLP

11 standards. They are actually clinical practice.

12 So, how does one interdigitate those two different

13 worlds into a process that can then be used by the

14 regulatory process?

15 DR. SANTANA: Dr. Khanna?

16 DR. KHANNA: There is a little bit of a

17 precedent that is set for drugs that are pursued

18 for the field of animal health and are approved

19 through the Center for Veterinary Medicine within

20 the FDA. The issues that we deal with there are

21 basically the availability of raw data, the

22 contemporaneous keeping of records, and the use of

23 standardized tests and measures against those

24 animals.

25 So, speaking only to the use of the pet

347

1 animal studies, there is a body of regulations that

2 oversees these trials and, in fact, those same

3 guidelines which are more GCP-like may be very

4 useful in studies in mice, and they are not as

5 onerous as GLP, and there are probably areas for

6 modification but they may be a good resource to

7 look at.

8 DR. SANTANA: Dr. Adamson?

9 DR. ADAMSON: Steve, I think there are

10 really two scenarios. One is that there are

11 observations made by an independent laboratory that

12 hadn't necessarily set out to generate the data

13 that was going to go to the agency but that is

14 important data. There, I think the scientific

15 method is a pretty robust one. That is, it

16 undergoes peer review and if it is important

17 someone ought to repeat it and show the same thing.

18 I would hold any of those observations to the same

19 standards. I mean, if something is not

20 reproducible by another laboratory, it is not to

21 say throw it away but it should raise some

22 questions.

23 I think what we have been spending more

24 time on is, okay, we are undertaking a program, the

25 sole objective of which is really to provide

348

1 guidance for drug development. There, I think if

2 you have a standardized approach where the

3 methodology is well described and there are

4 standard operating procedures--again, it isn't GLP

5 but it isn't the opposite of GLP but it is several

6 steps toward it--and I would love to hear Peter

7 Houghton's opinion on this--but I think it would be

8 reasonable to get access to the raw data. Because

9 usually the limitation of GLP is manpower and

10 resources, and if it is important and we can do a

11 data dump and someone else at the agency can crank

12 through it to see if we get the same results, I

13 think that is a reasonable approach.

14 DR. HIRSCHFELD: Dr. Hastings, there is a

15 seat there with a microphone and you can take that,

16 and I will just clarify the question. When we said

17 primary data, that is synonymous with raw data;

18 that is unprocessed data.

19 DR. HASTINGS: Right. I just want to make

20 one point. Actually, we have talked about the

21 pet--well, the companion animal studies. I believe

22 that under the regulations if you do an

23 experimental study in companion animals or pets you

24 have to have an IND with the Center for Veterinary

25 Medicine.

349

1 DR. KHANNA: I will just briefly respond.

2 That is not necessarily true. It depends on the

3 basis around which you are trying to pursue the

4 drug. If you are pursuing that drug for a

5 veterinary indication it needs to go through the

6 CVM. If you are not, the CVM has told us that they

7 would not want to be involved in the review of that

8 data that is going towards the human development of

9 a drug. In fact, they request from us to get

10 regulatory discretion from the human side.

11 DR. HASTINGS: So, you have already

12 discussed that with CVM?

13 DR. KHANNA: Yes.

14 DR. HIRSCHFELD: I can just verify that I

15 was specifically involved in a case or consulted

16 where it turned out to be Dr. Khanna who was

17 submitting a protocol and we were asked whether

18 this was going under an IND that existed for human

19 studies, and we were able to verify that, yes, it

20 was under an IND for human studies and that the

21 data would feed into the collective pool of data

22 for understanding the potential human application

23 and then the Center for Veterinary Medicine

24 gracefully withdrew.

25 DR. WILLIAMS: It seems like a small

350

1 working group between the Center for Veterinary

2 Medicine and FDA and oncology groups especially

3 could work out some kind of formal/informal

4 arrangement.

5 DR. SANTANA: Yes, I think that would be

6 critical because as this experiment unfolds over

7 the next few years we want to make sure that the

8 data that we are collecting, and the way we are

9 collecting the data will be acceptable to the

10 agency because, if not, we are going to be faced

11 with the issue of how do we advance drug

12 development in children if the data, for one reason

13 or another, hits a regulatory snarl and is not

14 accepted by the agency. I think Donna had a

15 question or a comment.

16 DR. PRZEPIORKA: Yes, for the record, if

17 an academic institution participates in a trial

18 that goes to the FDA, the FDA can come and audit

19 that academic institution to make sure their

20 clinical trial was done appropriately. If an

21 academic laboratory has their data used to support

22 an IND, is that laboratory open for being audited

23 by the FDA as well?

24 DR. HIRSCHFELD: The data from that study

25 would be, and we have had circumstances where there

351

1 were, let's say, perceived irregularities in data

2 from a laboratory under a number of INDs and what

3 we have done is for-cause inspections of that

4 facility. But if it is a single study that an

5 academic laboratory is doing and the data appear to

6 be internally consistent and robust, that usually

7 is not a trigger for an audit.

8 DR. SANTANA: Dr. Grillo?

9 DR. GRILLO-LOPEZ: In the setting of an

10 existing NDA, that is a product that has been

11 approved let's say for an adult malignancy, that

12 NDA is a pharmaceutical company's NDA that has been

13 submitted to the FDA and obtained that approval.

14 Usually the way the data works its way into that

15 NDA and the overall database is through the

16 pharmaceutical company. So, one way that your data

17 would get to the hands of the FDA would be through,

18 in this case, this third party pharmaceutical

19 company that then transmits it to the FDA and the

20 FDA will ask for the raw data. In fact, in

21 pharmaceutical companies we practically always

22 submit the raw data to the FDA in addition to all

23 analyses and interpretations, etc., with some

24 exceptions where a publication might be sufficient

25 for some particular purpose. So, in many

352

1 situations you would be working with the

2 pharmaceutical company to put data in a format that

3 would be acceptable to the FDA unless you held your

4 own IND, and then you would file the data to your

5 IND, if I am correct.

6 DR. HIRSCHFELD: Right, and again we are

7 not restricting it to the NDA filing final phase of

8 development but we are opening the whole discussion

9 to all aspects of product development.

10 DR. GRILLO-LOPEZ: Yes, I understand but I

11 just wanted to make the point again, as I did

12 earlier this morning, that as we are conducting

13 this discussion it is missing one leg of the stool.

14 I am the only industry representative around this

15 table, and to make this discussion more effective

16 we should have had other industry representatives

17 and presenters who are more expert than I on

18 pediatric oncology.

19 DR. SANTANA: Dr. Grillo, noted again in

20 the discussion of the afternoon of that issue.

21 Susan, you had a comment?

22 DR. WEINER: Yes, I guess I wanted to tie

23 it to the discussions earlier in the day from the

24 public's perspective, from the family's

25 perspective. That is, the most valuable resources

353

1 I think that we all have in this situation are the

2 patient resources, the number of kids who are

3 involved and their well being and time. Insofar as

4 the conduct of any preclinical or nonclinical

5 activity is done for its own sake or is done

6 without it being in direct service of advancing

7 therapies for kids, I think we have to question

8 that and be mindful of that.

9 In addition, I think it is very important

10 that the agency, when they consider what they

11 require of sponsors or what kinds of studies they

12 believe should be done on kids given what has

13 happened in the preclinical setting, the notion

14 that resources have to be conserved--that risk has

15 to outweigh benefit, to be sure--but that resource

16 have to be conserved because they make commitments

17 into the future that may not be necessary I think

18 is vital, and it is a vital selling point to the

19 families community to hear that from you and I

20 appreciate the exquisite nature with which the

21 discussion is taking place.

22 DR. SANTANA: I think that is a very good

23 concluding comment for the discussion this

24 afternoon and I couldn't have said it any better.

25 So, unless the agency requests that we provide any

354

1 further comments, I think we have attempted to give

2 you the best that we could, given what we were

3 asked to comment on. So, I want to thank everybody

4 that participated. I know it was a tough

5 discussion this afternoon because it was more

6 theoretical based rather than practice based but,

7 hopefully, in the future, once we get more data, we

8 can probably relate it to more practical issues at

9 some future point.

10 DR. HIRSCHFELD: And I want to thank all

11 of you for helping us. We could say that we are at

12 the edge and trying to push it but we don't even

13 know where the edge is, and I thank all of you for

14 helping us explore the unknown with the hope that

15 the future will be the known, and our gratitude is

16 noted too.

17 DR. SANTANA: Thank you, Dr. Hirschfeld.

18 [Whereupon, at 5:10 p.m., the proceedings

19 were adjourned.]