1

DEPARTMENT OF HEALTH AND HUMAN SERVICES

FOOD AND DRUG ADMINISTRATION

CENTER FOR DRUG EVALUATION AND RESEARCH

ONCOLOGIC DRUGS ADVISORY COMMITTEE

Tuesday, May 4, 2004

7:58 a.m.

Hilton Washington

620 Perry Parkway

Gaithersburg, Maryland

P A R T I C I P A N T S

Committee Participants:

Bruce D. Cheson, M.D., Acting Chairman

[a.m. session]

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

Secretary

Otis W. Brawley, M.D.

John T. Carpenter, Jr., M.D.

James H. Doroshow, M.D.

Stephen L. George, Ph.D.

Antonio J. Grillo-Lopez, M.D.

Pamela J. Haylock, RN

Silvana Martino, D.O.

Gregory H. Reaman, M.D.

Bruce G. Redman, D.O.

Maria Rodriguez, M.D.

Sarah A. Taylor, M.D.

Consultants (voting)

For Procrit:

Kenneth Bauer, M.D.

Laurie Feldman, Ph.D.

For CRC Endpoints:

Ronelle DuBrow, M.D.

David Kelsen, M.D., Guest Chair [p.m. session]

Michael J. O'Connell, M.D.

Daniel Sargent, M.D.

Patient Representatives (voting):

Musa Mayer, New York, New York - For Procrit

Nancy Roach, Hood River, Oregon - For CRC Endpoints

FDA Participants

Clare Gnecco, Ph.D.

Harvey Luksenburg, M.D.

Patricia Keegan, M.D.

Karen Weiss, M.D.

Amna Ibrahim, M.D.

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

Grant Williams, M.D.

Richard Pazdur, M.D.

C O N T E N T S

Call to Order, Introduction of Committee - Bruce

Cheson, M.D., Acting Chair, ODAC 5

Conflict of Interest Statement - Johanna Clifford,

M.S., RN, Executive Secretary, ODAC 8

Opening Remarks - Patricia Keegan, M.D., Director,

Division of Therapeutic Biological Oncology

Products, FDA 11

Sponsor Presentations

NeoRecormon (epoetin beta) - Hoffman-LaRoche, Ltd.,

Marty Huber 13

Johnson & Johnson, Ltd.

- Introduction - Robert DeLap, M.D., Ph.D., Vice

President, Global Regulatory Affairs 26

- Evaluation of Studies - Peter Bowers, M.D.,

Senior Director, Clinical Team Leader, EPO, Drug

Development 33

- Future Clinical Data - Martine George, M.D.,

Vice President, Hematology and Oncology,

Clinical Research and Global Development 47

Amgen, Inc.

- Introduction - Dawn Viveash, M.D., Vice

President, Regulatory Affairs and Safety 53

- Aranesp Properties, Preclinical Observations

and EPO Receptor Biology - Harvey Lodish, Ph.D.,

Professor of Biology and Bioengineering, MIT 58

- Aranesp Clinical Observations and Pharmaco-

vigilance Program Conclusions - David Parkinson,

M.D., Vice President, Oncology Clinical

Development 64

FDA Presentation

Harvey Luksenburg, M.D, Medical Officer, Division

of Therapeutic Biological Oncology Products, FDA 84

Open Public Hearing - [NONE]

Committee Discussion 145

Lunch 213

C O N T E N T S (Continued)

Introduction of Committee - David Kelsen, M.D.,

Acting Chair, ODAC 214

Conflict of Interest Statement - Johanna Clifford,

M.S., RN, Executive Secretary, ODAC 216

Opening Remarks - Richard Pazdur, M.D., Director,

Division of Oncology Drug Products, FDA 218

Regulatory Background and Past FDA Approvals in

Colorectal Cancer - Amna Ibrahim, M.D., Medical

Officer, Division of Oncology Drug Products, FDA 225

Synopsis of FDA Colorectal Cancer Endpoints

Workshop - Michael O'Connell, M.D., Director,

Division of Medical Oncology, Allegheny General

Hospital, Pittsburgh, PA 242

Disease-Free Survival (DFS) vs. Overall Survival

(OS) as a Primary Endpoint for Adjuvant Colon

Cancer Studies - Daniel Sargent Ph.D., Director,

Cancer Center Statistics, Mayo Clinic Cancer

Center, Rochester, MN 260

Open Public Hearing 311

Committee Discussion 328

Adjourn 403

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

2 DR. CHESON: Good morning. Welcome to the

3 Oncologic Drug Advisory Committee, May 4th. I'm

4 Bruce Cheson from the Lombardi Comprehensive Cancer

5 Center. I am the Acting Chair of the ODAC for

6 today's session. I do not work for, very clearly,

7 the FDA in any way, shape, or form. I do this on a

8 voluntary basis. And I am delighted to have some

9 excellent colleagues of mine on this committee

10 today, and I would like to start off today's

11 session by having everybody at the table introduce

12 themselves, starting with my friend Antonio

13 Grillo-Lopez.

14 DR. GRILLO-LOPEZ: Thank you, Mr. Acting

15 Chairman. My name is Antonio Grillo-Lopez. I am a

16 hematologist/oncologist with the Neoplastic and

17 Autoimmune Diseases Research Institute.

18 MS. MAYER: I am Musa Mayer. I am the

19 patient rep for this morning's session, and I'm a

20 15-year breast cancer survivor from New York City.

21 DR. BRAWLEY: I'm Otis Brawley. I'm a

22 medical oncologist and epidemiologist, and I'm a

1 professor at Emory University.

2 DR. MARTINO: Silvana Martino, medical

3 oncology, from the John Wayne Cancer Institute.

4 DR. TAYLOR: Sarah Taylor, medical

5 oncology, palliative care, University of Kansas.

6 DR. REAMAN: Gregory Reaman, pediatric

7 oncologist at the George Washington University and

8 Children's National Medical Center.

9 DR. REDMAN: Bruce Redman, medical

10 oncologist, University of Michigan.

11 MS. CLIFFORD: Johanna Clifford, FDA,

12 Executive Secretary to this meeting.

13 DR. DOROSHOW: Jim Doroshow, medical

14 oncologist, Director, Division of Cancer Treatment

15 and Diagnosis, NCI.

16 DR. GEORGE: Stephen George, Biostatistics, Duke

17 University.

18 MS. HAYLOCK: I'm Pamela Haylock. I'm an

19 oncology nurse and doctoral student at the

20 University of Texas, Medical Branch in Galveston,

21 and I'm the consumer representative.

22 DR. FELDMAN: Laurie Feldman. I'm a

1 research scientist at the Beth Israel Deaconess

2 Medical Center in Boston.

3 DR. GNECCO: Clare Gnecco. I am the

4 statistical reviewer for several of the epoetin

5 products.

6 DR. LUKSENBURG: Harvey Luksenburg. I'm a

7 medical reviewer at the Food and Drug

8 Administration.

9 DR. KEEGAN: Patricia Keegan, Division

10 Director, Division of Therapeutic Biological

11 Oncology Products.

12 DR. WEISS: I'm Karen Weiss, Office of

13 Drug Evaluation VI, CDER, FDA.

14 DR. CHESON: Thank you.

15 Today we have an interesting series of

16 discussion, the morning of which will be a series

17 of presentations and discussions concerning safety

18 concerns associated with Aranesp from Amgen and

19 Procrit from Johnson & Johnson, both of which are

20 indicated for the treatment of anemia associated

21 with cancer chemotherapy. I was approached earlier

22 by someone from the press who said, "How come there

1 has been no buzz about this?" I think this is

2 sufficient evidence that there is buzz about this,

3 and I look forward to an interesting series of

4 discussions.

5 We'll start off with opening remarks from

6 Dr. Keegan.

7 MS. CLIFFORD: Well, actually, me.

8 DR. CHESON: Oh, excuse me. From Johanna

9 first. Johanna Clifford, the conflict of interest

10 statements.

11 MS. CLIFFORD: Thank you.

12 The following announcement addresses the

13 issue of conflict of interest with respect to this

14 meeting and is made a part of the record to

15 preclude even the appearance of such at this

16 meeting.

17 Based on the submitted agenda and

18 information provided by the participants, the

19 agency has determined that all reported interests

20 in firms regulated by the Center for Drug

21 Evaluation and Research present no potential for a

22 conflict of interest at this meeting with the

1 following exceptions:

2 Dr. Maria Rodriguez has been recused from

3 participating in all matters related to the

4 discussions of safety issues associated with

5 Aranesp and Procrit.

6 Dr. Kenneth Bauer has been granted a

7 waiver under 18 U.S.C. 208(b)(3) and 21 U.S.C.

8 505(n) for owning stock in the parent company of

9 the sponsor. The stock is valued from $5,001 to

10 $25,000.

11 Dr. John Carpenter has been granted a

12 waiver under 18 U.S.C. 208(b)(3) for lecturing on

13 an unrelated matter for the sponsor of Aranesp. He

14 is awaiting final payment of his fee that is less

15 than $5,000.

16 Dr. Otis Brawley has been granted a

17 limited waiver under 18 U.S.C. 208(b)(3) because

18 his employer has a contract with the sponsor to

19 study Aranesp. The contract is less than $100,000

20 a year. Under the terms of the limited waiver, Dr.

21 Brawley will be permitted to participate in the

22 committee's discussions; however, he will be

1 excluded from voting.

2 A copy of these waiver statements may be

3 obtained by submitting a written request to the

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

5 of the Parklawn Building.

6 Lastly, we would also like to note for the

7 record that Dr. Antonio Grillo-Lopez, Chairman,

8 Neoplastic and Autoimmune Diseases Research

9 Institute, is participating in this meeting as an

10 industry representative, acting on behalf of

11 regulated industry. He would like to disclose that

12 he is a scientific adviser to Chiron and receives

13 speaker fees from Wersch(ph).

14 In the event that the discussions involve

15 any other products or firms not already on the

16 agenda for which FDA participants have a financial

17 interest, the participants are aware of the need to

18 exclude themselves from such involvement, and their

19 exclusion will be noted for the record.

20 With respect to all other participants, we

21 ask in the interest of fairness that they address

22 any current or previous financial involvement with

1 any firm whose product they wish to comment upon.

2 Thank you.

3 DR. CHESON: Hearing no other comments,

4 now we'll go to Dr. Keegan.

5 DR. KEEGAN: Thank you. I want to thank

6 the committee and the companies who have come

7 forward to present information about the

8 erythropoietin products, both those licensed in the

9 United States and two that are not. The purpose of

10 this is to review information based on the results

11 of in the context of recent findings from two

12 studies from Europe that suggested that there are

13 certain practices in the administration of

14 erythropoietin products which may raise concerns

15 for safety of the products.

16 I want to remind everyone that the

17 erythropoietin products that were approved in the

18 United States were approved as a means of treatment

19 of anemia in a variety of settings that, over the

20 period since original approval, there have been

21 investigations into alternative uses of these

22 products, looking at other benefits such as impact

1 on survival.

2 It is in that arena that two studies

3 recently conducted in Europe identified the

4 potential for some safety concerns with those

5 particular strategies. And we felt that it was

6 important at this time to review the available data

7 that both supported the original approval of

8 Aranesp and Procrit for treatment of anemia

9 associated with cancer, to review the clinical

10 trials in question conducted in Europe, and to

11 consider what additional information should be

12 obtained at this point in time to determine whether

13 or not an issue would exist with Procrit or Aranesp

14 for the treatment of anemia associated with cancer

15 and what the design of those studies should look

16 like or to hopefully rule out any problems at the

17 labeled and recommended doses for those two

18 products. So I would ask that the committee

19 carefully consider the data presented and provide

20 us with some guidance in the approach of these

21 additional studies.

22 I would like to draw your attention to the

1 fact that there are some errors in the FDA briefing

2 document, and we have provided an errata sheet that

3 will provide corrections to those errors. In

4 addition, we have revised Question 1 of the

5 questions to the committee in the first sentence,

6 and the modified questions are available as an

7 errata sheet at the table outside of this room.

8 DR. CHESON: Thank you, Dr. Keegan.

9 Since we went around the table, we've been

10 joined by another member. If you could please

11 identify yourself and your affiliation? Turn on

12 the microphone, please. Hit the button.

13 DR. BAUER: Ken Bauer from Harvard, from

14 the VA Medical Center and Beth Israel Deaconess in

15 Boston.

16 DR. CHESON: Thank you.

17 Okay. The first presentation from a

18 sponsor will be about NeoRecormon, or epoetin beta,

19 from Hoffman-LaRoche, Ltd. Since I don't have your

20 name here, if you could also please introduce

21 yourself.

22 DR. HUBER: Good morning. I'm Marty

1 Huber, an oncologist with Hoffman-LaRoche.

2 Given the Advisory Committee's discussion

3 today of the safety of erythropoiesis-stimulating

4 agents in the treatment of cancer patients,

5 Hoffman-LaRoche volunteered to provide data from a

6 study that was recently published in The Lancet,

7 which we'll subsequently refer to as MF4449.

8 Additionally, we'd like to provide some context for

9 these findings, reviewing some other clinical

10 trials that have been conducted with epoetin beta.

11 Just a quick background. NeoRecormon is

12 the trade name for epoetin beta. It is a

13 recombinant human erythropoietin with a

14 well-established benefit/risk profile with more

15 than one million years of patient experience. It

16 has been available outside the United States since

17 1990. We did not apply in the United States for

18 approval based on patent issues. There were no

19 safety issues which prevented it from being brought

20 into the United States. It was not reviewed by the

21 FDA. It is approved for patients with renal anemia

22 as well as oncologic indications in most of these

1 countries.

2 For the presentation today, we'd like to

3 review MF4449 focusing initially on the primary

4 study results as published in The Lancet. We will

5 also show additional analyses that were performed

6 on this study. We did a meta-analysis of the

7 clinical trial data with epoetin beta, and,

8 finally, we'll look at one of our large randomized

9 studies in which we have a long-term survival

10 follow-up.

11 MF4449 was a study which was looking at an

12 investigational use of epoetin beta. It was

13 looking at, Would increasing the hemoglobin with

14 epoetin beta lead to better efficacy of

15 radiotherapy? This was trying to invoke

16 radiosensitization, and could that lead to improved

17 progression-free survival in cancer patients? The

18 primary endpoint was local progression-free

19 survival. For the rest of the study, I will refer

20 to this as PFS, or progression-free survival.

21 This is an overview of the study design.

22 Patients with head and neck cancer--and it was

1 males with a hemoglobin less than 13, females less

2 than 12--were randomized to receive either epoetin

3 beta, 300 international units per kilogram sub-cu

4 three times weekly, or placebo in combination with

5 their radiotherapy. Then they were followed up

6 until progression or another endpoint.

7 The idea was to start them two weeks

8 before the radiotherapy, but this was not done in

9 all cases. Therefore, patients received a total of

10 either seven to nine weeks of epoetin beta maximum.

11 Epoetin beta was not continued in the follow-up

12 period.

13 An important factor in this study was how

14 the patients were stratified. As you know, head

15 and neck cancer is a very heterogeneous disease.

16 Therefore, we stratified them on the basis of tumor

17 TNM Stage IV versus III. In addition, they were

18 stratified by resection status. Stratum 1 here was

19 patients who had had a complete resection. Stratum

20 2 was patients who had residual tumor after

21 resection. And Stratum 3 was, finally, patients

22 who received no attempt at resection and were

1 essentially treated with radiotherapy as their

2 primary therapy.

3 With regard to the population characs, the

4 details are in your briefing document, and they

5 were overall very well balanced. There were a

6 couple of exceptions we'd like to point.

7 First was smoking status. This was not

8 have a history of smoking but were they smoking at

9 the time. We believe this is relevant because we

10 know there is an interaction between active

11 cigarette smoking and radiotherapy which may

12 diminish the efficacy of radiotherapy. At

13 baseline, 53 percent of patients on placebo were

14 smoking; 66 percent in the epoetin beta group.

15 Furthermore, because the patients had had

16 surgery and then were randomized, there were

17 patients who had relapsed, even prior to

18 randomization. This was in balance, with 10

19 percent in the epoetin beta group, 7.6 percent on

20 placebo.

21 And, finally, for Stage IV TNM status,

22 there was a minimal imbalance at baseline, 72

1 percent versus 75 percent. But what you will see

2 is, as we start looking at subgroups, this

3 imbalance is magnified in an important subgroup.

4 These are the data that were shown in The

5 Lancet showing that there was a progression-free

6 survival advantage for placebo over epoetin beta.

7 This is follow-up from--this is month six. An

8 important point here is during the first five to

9 six months, there was no difference in

10 progression-free survival. This will contrast with

11 some of the other data that you will review later.

12 We had conducted a series of secondary

13 analyses which were prospectively planned. The

14 intent of these analyses--we looked at the

15 robustness of the data--was: Were the findings

16 robust throughout? And, also, was there

17 heterogeneity in the important subgroups?

18 Furthermore, when we looked at the

19 outcome, this inferiority of epoetin beta was very

20 much unanticipated. So this was in contrast to all

21 other clinical experience with epoetin beta. So

22 based on that, we did further additional analyses.

1 These were the planned secondary analyses to look

2 at the population robustness. What I'm showing

3 here are the Kaplan-Meiers for three populations:

4 intent to treat, radiotherapy correct, and,

5 finally, per protocol.

6 The differences between these groups are:

7 In the radiotherapy correct population, these are

8 the patients who received the radiotherapy as

9 specified in the protocol. The per protocol

10 population on the far right is not only did they

11 get the right radiotherapy, but they also got the

12 right treatment with regards to epoetin

13 beta/placebo according to dose and schedule in the

14 protocol. The n's on this, this is approximately

15 350, this is approximately 260, and this is around

16 220.

17 What's important to notice is that as you

18 get to the purer population, the treatment effect

19 actually diminishes. This is contrary to what you

20 would expect. Normally when we do these studies

21 for robustness, we are looking to see the treatment

22 effect getting larger in the population that's

1 treated who are in per protocol. So this indicated

2 to us some lack of robustness in the data.

3 We did subgroup analysis. This is a

4 forest plot. I just oriented this slide. This is

5 the categories, and these were categories we

6 normally look at in head and neck trial: stratum,

7 location, staging, age, gender, smoking status, and

8 baseline hemoglobin.

9 What we looked at is, to the left is

10 outcomes better with epoetin beta, and to right is

11 better with placebo. As you can see here, there is

12 a divergence of findings on both sides of one.

13 What we'd like to look at today is look at a couple

14 of these subgroups in which there was the highest

15 relative risk, specifically Stratum 2 and they

16 hypopharynx.

17 Looking at the progression-free survival

18 by stratum, this is Stratum 1, which were the

19 patients who were completely resected. This is

20 Stratum 2, which were the patients who had residual

21 tumor. One of the things that we found was the

22 actual progression-free survival in Stratum 2

1 placebo was better than placebo with completely

2 resected patients. This goes contrary to the

3 natural history of these tumors and numerous other

4 publications. We would clearly expect that this

5 curve should be better than this. So what we feel

6 is there is obviously some evidence of something

7 odd about this placebo group.

8 Furthermore, when we looked into the tumor

9 site, if you look at the hypopharynx location,

10 there is a wide difference; there's a major

11 treatment effect. This is placebo, epoetin beta.

12 However, all other locations there was no

13 difference in progression-free survival. So when

14 we do the subgroup analysis, the effect is

15 restricted to the hypopharyngeal population.

16 We looked further in this population, and

17 what we found was that we did have an imbalance

18 with regard to Stratum 3--30 percent in placebo, 45

19 percent epoetin beta--within this subgroup. These

20 are the patients who did not have resection or

21 attempts at resection and were radiotherapy only.

22 Furthermore, we had an imbalance in the number of

1 patients who were in Stage IV.

2 With regards to safety, I apologize for

3 this slide. This is the non-cancer-related adverse

4 events, but essentially they were balanced overall:

5 65 percent placebo, 68 percent epoetin beta.

6 I would like to point out one piece of

7 data here. In your briefing document, there's a

8 reference to placebo 5 percent, epoetin beta 11

9 percent for vascular disorders. In this

10 terminology, vascular disorders includes

11 hypertension. What we have historically done when

12 looking at these issues, we've used the definition

13 of thromboembolic events. It does not include

14 hypertension. So if you see some differences in

15 numbers, this is what accounts for it

16 When we looked at thromboembolic events,

17 we saw placebo 3.5 percent, epoetin beta 5.6

18 percent, with some--sort of slight imbalances, with

19 more on the epoetin beta treatment group.

20 Furthermore, one of the things you may

21 have noticed in the briefing document, there was an

22 imbalance in cardiovascular deaths: 10 deaths on

1 the epoetin beta group versus 5 on placebo in the

2 cardiovascular category. Given the concerns about

3 thrombovascular events, what's important to note is

4 one epoetin beta and one placebo occurred around

5 day 50. The remaining deaths occurred after day

6 100. Remember, treatment was only for a maximum of

7 seven weeks, so these events are occurring well

8 after cessation of epoetin beta treatment.

9 In summary, we believe that there was a

10 heterogeneity of treatment effect across various

11 subgroups such as stratum, baseline hemoglobin,

12 age, gender, disease location, and that there were

13 also imbalances in important baseline

14 characteristics, smoking for the overall

15 population, as well as stage and resection status

16 for patients with tumors in the hypopharyngeal

17 location.

18 With regards to meta-analysis, this was

19 pooled results from nine controlled clinical

20 trials, a total of 1,409 patients, with both solid

21 and hematologic tumors. We looked at tumor

22 progression, overall survival, and thromboembolic

1 events.

2 Once again this is a forest plot. What we

3 look at is better with epoetin beta, better with

4 placebo. This is the total population. These are

5 the individual studies. And then this is solid

6 versus hematologic.

7 What we saw was actually a reduction in

8 risk of progression with epoetin beta, 0.79, with a

9 difference approaching significance. The remaining

10 studies are relatively consistent in that most of

11 them are less than 1, with a couple of exceptions,

12 but they're very close. Also, it's a consistent

13 finding for solid and hematologic tumors. In all

14 of these we saw a reduced risk of progression.

15 For survival, we saw a risk of 0.97, so

16 it's essentially the same for epoetin beta and

17 placebo. And, once again, these studies are around

18 1. This one study, which is a higher one of 3.39,

19 if you notice, due to the wide confident intervals.

20 Very few deaths were noted in this study.

21 We also looked at thromboembolic events in

22 this study, in this pooled study, and the control,

1 of 609 patients, 4 percent, epoetin beta 6 percent.

2 This was actually quite consistent with the

3 findings I presented from MF4449.

4 So, in summary, there was no evidence of

5 increased tumor progression in patients treated

6 with epoetin beta. There was no evidence of

7 decreased overall survival. There was a small

8 increase in the incidence of thromboembolic events:

9 6 percent of epoetin beta versus 4 percent on

10 placebo. But what I'd like to note is when we

11 looked at patient years of observation and

12 corrected for that, this difference disappeared.

13 The limitation of this meta-analysis is

14 most of these studies were relatively short in

15 duration because they were looking at endpoints

16 such transfusion or hemoglobin. Therefore, we

17 looked at MF4467 to see what there a long-term

18 effect on survival. This was a double-blind,

19 placebo-controlled study of epoetin beta in

20 patients with lymphoid malignancies. The primary

21 endpoint was transfusion-free survival, and as you

22 can see, there was a robust effect on that

1 endpoint.

2 What we did was an overall survival on

3 over 340 patients in this study. This is the

4 Kaplan-Meier and, as you can see, there's no

5 difference in overall survival between placebo and

6 epoetin beta.

7 In conclusion, the MF4449 study results

8 are inconsistent with the other epoetin beta

9 studies in oncology. We believe the most likely

10 explanation for the adverse outcomes observed in

11 MF4449 are factors independent of epoetin beta.

12 The large majority of existing data shows that

13 epoetin beta does not adversely affect tumor

14 progression or survival in cancer patients.

15 Thank you.

16 DR. CHESON: Thank you.

17 We're going to reserve questions until

18 after the FDA makes its presentation.

19 Next, Dr. DeLap from Johnson & Johnson.

20 DR. DeLAP: Dr. Cheson, members of the

21 panel, and guests, good morning. I'm Dr. Robert

22 DeLap. I'm Vice President for Regulatory Affairs

1 at Johnson & Johnson Pharmaceutical Research and

2 Development, and I will be providing a brief

3 introduction to our presentation.

4 We are pleased to be able to be here today

5 to participate in this discussion of the safety of

6 erythropoietin products in patients with cancer and

7 to present our data in support of this discussion.

8 We will not have time to summarize all of the

9 information that's been generated over the years in

10 our extensive research programs, so our

11 presentation will focus on the information that we

12 deem most relevant to today's discussion. Of

13 course, we will be pleased to elaborate further on

14 any specific points of interest.

15 Erythropoietin products are approved for

16 the treatment of anemia associated with

17 chemotherapy. Chemotherapy-associated anemia is a

18 common problem for patients with cancer, and this

19 anemia can be associated with debilitating symptoms

20 and may require transfusions of red blood cells.

21 Erythropoietic products have substantial value in

22 treating anemia and its symptoms and can

1 significantly reduce the need for transfusions.

2 This benefits individual patients and also means

3 that the units of red blood cells that are

4 collected by blood banks can serve the needs of

5 additional patients.

6 The safety profile of erythropoietin

7 products has been well established in years of

8 clinical use, both in chemotherapy-induced anemia

9 and in other illnesses where anemia may occur.

10 Epoetin alfa products have been the subject of may

11 clinical studies and have been used worldwide to

12 treat more than two million patients for this

13 indication.

14 In the U.S., there are two products that

15 are labeled for treatment of patients with cancer

16 chemotherapy-induced anemia. These are Procrit,

17 marketed by Ortho Biotech, a J&J company, and

18 Aranesp, marketed by Amgen. Procrit became

19 available for this indication in 1993, and Aranesp

20 became available for this indication in 2002.

21 Products available outside of the U.S.

22 include EPREX, an epoetin alfa product that is also

1 marketed by J&J companies, and NeoRecormon and

2 Aranesp. All of these products share extensive

3 homology with naturally occurring human

4 erythropoietin, and all act by binding to the

5 erythropoietin receptor with activation of

6 downstream pathways leading to red blood cell

7 production.

8 Our presentation will describe a number of

9 studies that have been done in our extensive

10 clinical research program, and we will be talking

11 about two different types of studies. Studies in

12 supportive anemia care are the studies that were

13 used to establish the existing indication for use

14 of these products in patients with cancer--that is,

15 the treatment of anemia associated with cancer

16 chemotherapy. In this use, anemic patients are

17 typically treated with a goal to obtain at least 1

18 gram per deciliter rise in hemoglobin level, to

19 raise the patient's hemoglobin to a target range

20 that is still below normally, typically, but is

21 sufficient to reduce the likelihood of a

22 transfusion.

1 Beyond correction of anemia is the term

2 that we will be using today to describe

3 investigational uses that have evaluated the use

4 erythropoietin products to treat patients to higher

5 hemoglobin target levels. Recent studies

6 evaluating the effect or erythropoietic agents on

7 cancer treatment outcomes have often utilized this

8 design.

9 It was hypothesized that any beneficial

10 effects of treatment with erythropoietic agents on

11 cancer treatment outcomes might be magnified with

12 treatment to higher hemoglobin target levels.

13 However, some of these studies have suggested

14 unexpected risks, including decreased survival.

15 This has led to extensive work that is

16 continuing at our company to better understand the

17 observations from these studies and to ensure that

18 patients and prescribers will continue to have all

19 of the information necessary to support the safe

20 and effective use of our erythropoietin alfa

21 products.

22 Safety data we will be presenting data are

1 as follows: We will first summarize data obtained

2 in our clinical studies of epoetin alfa in

3 supportive anemia care, which, together with the

4 extensive clinical experience over more than a

5 decade, support the favorable risk/benefit ratio

6 for epoetin alfa for the existing indication.

7 Second, we will summarize data from a

8 number of investigational studies that have

9 involved treatment of patients beyond correction of

10 anemia, including indications of increased risks

11 that have arisen in some of these studies using

12 that treatment approach. We remain interested in

13 studying the effects of epoetin alfa on cancer

14 treatment outcomes, but we have modified the

15 hemoglobin target levels that we are using in that

16 research.

17 Finally, we will describe additional data

18 that we are collecting and further research that we

19 have currently under consideration.

20 We look forward to the advice of the

21 Advisory Committee today as we work to do the best

22 possible job of planning our future activities in

1 this area.

2 Our agenda for our presentation is as

3 follows: Dr. Peter Bowers, who leads our clinical

4 programs with Procrit, will summarize our data from

5 epoetin alfa studies that have been done for

6 supportive anemia care and investigational studies

7 that have involved treatment beyond the correction

8 of anemia. Dr. Martine George, who heads our

9 entire hematology/oncology clinical development

10 program, will then describe future clinical data

11 relevant to this subject that we expect to have

12 from our currently ongoing studies and an

13 additional clinical study that we are considering

14 to fill knowledge gaps in this area. Finally, Dr.

15 George will conclude our presentation.

16 We have with us today several advisors to

17 help facilitate the discussion, as noted on this

18 slide, including Drs. Jesse Berlin, Kimberly

19 Blackwell, Roger Cohen, George Demitri, Mark

20 Levine, and Brian Leyland-Jones.

21 Now I would like to introduce Dr. Peter

22 Bowers for his summary of information from our

1 clinical study database. Thank you.

2 DR. BOWERS: Dr. Cheson, committee

3 members, during the next minutes I will present a

4 summary of safety information available from

5 studies of epoetin alfa conducted in two settings:

6 supportive anemia care, our labeled indication, and

7 studies beyond correction of anemia.

8 We undertook a combined analysis of ten

9 completed randomized, double-blind,

10 placebo-controlled studies evaluating the use of

11 epoetin alfa, EPREX and/or Procrit, for supportive

12 anemia care. These data from 1,976 patients

13 represent all controlled studies in this setting

14 for which we have full patient level data regarding

15 survival available. We examined mortality hazard

16 ratios for deaths during the double-blind phase

17 plus 30 days, and also tumor response and disease

18 progression information, the latter available in

19 five of the ten studies. Thrombotic vascular

20 event, or TVE, data from the combined analysis will

21 also be presented.

22 Some points should be kept in mind

1 regarding these analyses. The studies represent a

2 variety of tumors, and many include mixed tumor

3 types. The studies were designed and conducted to

4 assess the impact of epoetin alfa on reducing

5 transfusion and correcting anemia. Thus, data

6 regarding survival and tumor response or disease

7 progression were collected as secondary endpoints

8 and/or for safety purposes. Additionally, the

9 study drug treatment period ranges from 12 to 24

10 weeks, plus 4 weeks follow-up.

11 These are the results from the combined

12 analysis for mortality. The chart in the center of

13 the slide displays the point estimates, the red

14 dots, and the 95-percent confidence intervals, the

15 white horizontal bars. Unity is the dashed

16 vertical line. A point estimate less than one

17 suggests lower mortality among epoetin-treated

18 patients, and greater than one, higher mortality. This side

19 of the chart would favor epoetin alfa;

20 this side favors placebo.

21 Please note for the combined analysis the

22 point estimate for mortality is 0.99, shown at the

1 bottom, with a confidence interval 0.76 to 1.28.

2 This means mortality among epoetin alfa-treated

3 patients was the same as placebo patients in these

4 studies.

5 We reviewed tumor response and disease

6 progression data from the five studies where this

7 information was collected. As you can see,

8 response rates were similar between treatment

9 groups, and also as you see, disease progression

10 assessed in four studies was also similar between

11 treatment groups.

12 To summarize, the established benefits of

13 epoetin alfa for supportive anemia care--that is,

14 anemia related to cancer chemotherapy--include

15 transfusion reduction and amelioration of the

16 debilitating symptoms of anemia. An evaluation of

17 the studies in the approved indication showed no

18 signal of reduced survival and no indication of an

19 adverse impact on tumor response or disease

20 progression. Thus, the benefits of epoetin alfa

21 therapy continue to be supported by a well-defined

22 and acceptable risk profile when used for the

1 approved indication of anemia in patients receiving

2 cancer chemotherapy.

3 Now I'm going to turn to studies from

4 epoetin alfa used in settings beyond correction of

5 anemia, and before presenting the clinical data,

6 I'd like to review very briefly some key

7 preclinical findings.

8 The preclinical literature suggests a

9 potential benefit of erythropoietins on tumor

10 growth. However, there are also reports that

11 suggest the possibility of a deleterious effect.

12 Many tissues, including tumor cell lines, express

13 erythropoietin receptors. In experiments by

14 Johnson & Johnson and external groups, involving

15 more than 25 different tumor cell lines, including

16 cell lines known to express erythropoietin

17 receptor, erythropoietin did not cause tumor cell

18 proliferation. Similarly, systemic administration

19 of epoetin at doses of 20 to 2,000 international

20 units per kilogram three times per week in in vivo

21 models of breast, lung, and ovarian cancer in vivo

22 did not increase tumor volume. Moreover, a

1 positive effect on tumor growth delay has been

2 observed in animal models of concurrent

3 administration of erythropoietins in chemotherapy

4 or radiation therapy.

5 There are conflicting reports regarding

6 the impact of erythropoietin on tumor cell growth.

7 Some experiments in vitro indicate increased tumor

8 cell proliferation at erythropoietin concentrations

9 5- to 100-fold greater than those achieved

10 clinically using a dose of 40,000 international

11 units.

12 Based on the balance of positive

13 preclinical data and results from Study INT-10,

14 published by Dr. Timothy Littlewood in the Journal

15 of Clinical Oncology 2001, which suggested a

16 potential positive survival impact, the company

17 conducted Study INT-76. Details of this trial are

18 summarized in your background briefing materials.

19 INT-76 is a large study, 939 women

20 receiving first-line chemotherapy for metastatic

21 breast cancer, with a simple design. EPREX or

22 placebo was administered weekly and continued for

1 12 months, regardless of chemotherapy changes or

2 disease progression.

3 Study drug was initiated at a hemoglobin

4 of 13 or below and titrated to maintain hemoglobin

5 in the range 12 to 14. The primary endpoint of the

6 study was survival at 12 months. Objective

7 confirmation of investigator-reported secondary

8 endpoints, including disease progression and tumor

9 response, were not require. The primary--excuse

10 me. Study drug treatment was discontinued at the

11 recommendation of the DSMB for the study, and at

12 that time 88 percent of the subjects had completed

13 planned study drug treatment or had been withdrawn

14 from the study. The shortest duration of treatment

15 was nine months. Blinded follow-up continued out

16 to the 12-month endpoint. Groups were generally

17 balanced with regard to prognostic factors.

18 This slide shows the Kaplan-Meier plot for

19 survival. The vertical axis is probability of

20 survival, and the horizontal axis, time in months.

21 Below the horizontal axis are the numbers of

22 patients represented at each time point. White is

1 placebo, blue represents epoetin alfa. Please

2 observe the survival curves begin to diverge

3 relatively early in the course of follow-up such

4 that by month 4 the separation was near maximal,

5 and the curves continued parallel out through month

6 12.

7 The primary endpoint, survival at 12

8 months, was 24 percent survival--excuse me, deaths

9 in the placebo group, and 30 percent deaths of

10 patients in the epoetin alfa group. This

11 difference has a p value of 0.012. The hazard

12 ratio for mortality at the 12-month time point was

13 1.37, the confidence interval 1.07 to 1.74.

14 In light of these unexpected results,

15 extensive analyses were undertaken by the company.

16 Post hoc analyses, including subgroup and Cox

17 modeling, were undertaken, and results of these

18 analyses should be considered exploratory and

19 interpreted cautiously. No particular subgroup was

20 identified as accounting disproportionately for

21 most of the mortality difference.

22 Additional data were collected in a

1 retrospective blinded chart review of the medical

2 records of all subjects in the study. While not

3 conclusive, the analyses in chart review, together

4 with data from other trials, provide some

5 hypotheses that might explain the observed survival

6 difference. An adverse impact of epoetin alfa on

7 tumor proliferation is one hypothesis. Another is

8 imbalance in fatal thrombotic vascular events. And

9 we'll look at those a little further momentarily.

10 Now, looking in detail at the cause of

11 death data we have from INT-76, investigators

12 captured cause of death on a case report form page

13 with check boxes for either disease progression or

14 other. We looked at causes of deaths at 4 months,

15 since most of the difference in mortality had been

16 seen by that time point. Investigators attributed

17 most deaths to disease progression with a

18 difference between the groups, as you can see on

19 the slide.

20 In the other category, investigators

21 listed thrombotic vascular events, chemotherapy

22 toxicity, again, with differences as shown.

1 The blinded chart review suggested a

2 somewhat higher rate of thrombotic vascular events

3 than was reported by investigators, as you see on

4 the bottom of the slide: two among placebo group

5 patients, 11 among the epoetin alfa group patients,

6 at the 4-month time point.

7 This suggests the possibility that

8 thrombotic vascular events may have been underdiagnosed or

9 -reported as a cause of death in this

10 study and may have accounted for more of the excess

11 deaths in the epoetin alfa arm than was

12 appreciated.

13 The high number of deaths within the first

14 4 months, more so in the epoetin alfa group, may

15 indicate that a more sick patient population than

16 usual for a first-line metastatic breast cancer

17 study had been enrolled. As you can see, a greater

18 number of deaths--as you have seen, rather, a

19 greater number of deaths was attributed to disease

20 progression by investigators.

21 Further supporting the observation that

22 the observed early differences in mortality may

1 have resulted in substantial part from causes other

2 than tumor proliferation, the time to disease

3 progression curves shown here--placebo, again,

4 white; epoetin alfa, blue--are superimposed.

5 Response rates for the groups are similar: 46

6 percent and 45 percent. Thirty-eight percent of

7 patients in the placebo group developed new

8 lesions, whereas 30 percent of epoetin alfa

9 patients did. These results are not consistent

10 with an adverse impact of epoetin alfa on tumor

11 growth.

12 Given that this is a large, randomized,

13 double-blind study with unbiased, if incomplete,

14 collection of tumor progression data, these results

15 should be considered carefully.

16 To summarize, in INT-76, an early survival

17 disadvantage was observed in the treatment group.

18 Deaths were attributed to investigators in

19 significant part to disease progression. However,

20 investigator-reported disease progression and

21 response rates were similar. Given these

22 inconsistencies, other potential explanations for

1 the outcome merit consideration as well and, in

2 particular, thrombotic vascular events may have

3 been underdiagnosed as a cause of death in this

4 study.

5 Now, I'd like to turn to data from other

6 studies using epoetin alfa in settings also beyond

7 correction of anemia. Here we see summarized

8 several other studies that evaluated epoetin alfa

9 use in these settings. These studies are grouped

10 to reflect status, either completed or in follow-up

11 at the top of the chart, or discontinued in the

12 group at the bottom of the chart. INT-76 is

13 included at the top for reference.

14 As you see, the table summarizes some key

15 details of the studies. In general, these studies

16 have used epoetin alfa in settings where patients

17 are not anemic or are treated to hemoglobin levels

18 that are somewhat or substantially higher than are

19 needed for correction of anemia.

20 The mortality experience is shown here.

21 For the completed or in follow-up study, with the

22 exception of Study INT-76, mortality is not

1 significantly different. The five discontinued

2 studies represent studies stopped as a result of

3 unplanned interim analyses of safety conducted at

4 the company's request. Following this review, more

5 than 15 studies continued, some with modifications

6 to reduce target hemoglobins.

7 All five studies were stopped based on an

8 unplanned analysis, and, thus, it's not possible to

9 draw definitive conclusions other than to note

10 unfavorable survival trends for epoetin

11 alfa-treated patients in some of the stopped

12 studies. Follow-up data collection for these five

13 studies is continuing to further understand the

14 results.

15 Now, let's consider the data relevant to

16 tumor proliferation or disease response, as

17 indicated by the endpoints shown on the slide:

18 response rates, time to disease progression,

19 disease-free survival, and so forth.

20 Looking at the column on the right, the

21 differences in outcomes related to tumor response

22 or disease progression tend to be small. These

1 data show no signal that epoetin alfa is associated

2 with an adverse impact on adverse impact on tumor

3 growth.

4 Turning to clinically relevant thrombotic

5 vascular events in this same group of studies,

6 clinically relevant thrombotic vascular events, or

7 TVEs, are those which would be regarded by

8 clinicians as significant and include both the

9 venous and arterial events, but exclude such

10 occurrences as superficial venous thrombophlebitis

11 or catheter-related thromboses.

12 Here I've ordered the studies by frequency

13 of clinically relevant TVEs in the epoetin

14 alfa-treated patients: 31 percent to 1 percent.

15 Please note the substantial differences in the

16 frequency of clinically relevant TVEs.

17 Study 1015 with the greatest difference in

18 TVE rates, 27 percent, is among the studies with

19 the highest target hemoglobin level.

20 In contrast to this is the frequency of

21 TVEs in the ten studies of supportive care of

22 anemia. The studies are ordered by TVE frequency

1 in the epoetin alfa group, high to low, 9 percent

2 or lower. In general, the absolute frequency of

3 TVEs is substantially lower than is seen in the

4 group of studies beyond correction of anemia.

5 Differences between the groups are also smaller,

6 with a negative number indicating more TVEs in

7 placebo group patients.

8 Overall, the odds ratio shown at the

9 bottom of the slide is 1.55, indicating a modestly

10 increase risk of clinically relevant TVEs in the

11 epoetin alfa-treated patients, the confidence

12 interval 0.96 to 2.5.

13 In conclusion, our data indicate a

14 favorable benefit/risk profile for epoetin alfa

15 with no signal of tumor proliferation or adverse

16 survival impact in settings of supportive anemia

17 care. In study settings using epoetin alfa beyond

18 correction of anemia, adverse outcomes have been

19 seen. However, there is no clear signal suggesting

20 an adverse effect on tumor proliferation. There is

21 an indication that thrombotic vascular events are

22 more frequent in studies with higher target

1 hemoglobin levels. This may account for some,

2 possibly much, of the observed survival signal.

3 Additional data are being collected, and a

4 new trial is under consideration. Dr. Martine

5 George, therapeutic area head of oncology and

6 hematology at Johnson & Johnson PRD, will share

7 further details with you.

T1B DR. M. GEORGE: Thank you. 8

9 Johnson & Johnson has been studying the

10 potential benefit of epoetin alfa in the setting of

11 beyond correction of anemia since 1999, and our

12 work in this area continues. First, I will present

13 a clinical trial design for a study considering the

14 FDA guidance. Then I will review with you how

15 populated and ongoing trials could be used to

16 address the safety questions raised.

17 We considered several clinical trial

18 designs according to the agency requests, and after

19 critical analysis, we decided to select advanced

20 breast cancer. Our proposed clinical trial will

21 focus on breast cancer based on the signal observed

22 in INT-76, on the EPO receptor presence on breast

1 tumor, which is well known, on the high incidence

2 of the disease in the population, and also based on

3 the need for homogeneity in terms of patient

4 population and chemotherapy.

5 Furthermore, early clinical trials in

6 anemic patients have suggested a favorable outcome

7 in patients with anemia treated with erythropoietin. The

8 unfavorable outcome of INT-76 doesn't

9 preclude a potential benefit in anemic patients.

10 We are assuming a potential benefit, but

11 the trial will have to be powered to exclude a

12 negative effect, as requested by the agency.

13 The objective of the trial is simple.

14 It's to evaluate the effects of EPO alfa on cancer

15 outcomes in patients with metastatic breast cancer

16 receiving first-line chemotherapy.

17 The proposed clinical trial will be

18 double-blind, randomized, placebo-controlled, and

19 will enroll patients with advanced breast cancer

20 receiving first-line chemotherapy, including taxane

21 and/or anthracyclines. Patients will be anemic at

22 entry with hemoglobin at baseline equal to or less

1 than 11 grams per deciliter before their third

2 cycle of chemotherapy. Patients will receive EPREX

3 or placebo until tumor progression, end of

4 chemotherapy, or death. The target hemoglobin

5 level in the study will be 12 grams per deciliter,

6 and we'll hold the drug if the hemoglobin goes over

7 13 grams per deciliter.

8 The endpoints of the clinical trial will

9 be as follows: The primary endpoint will be

10 progression-free survival, and because of lack of

11 time, I won't expand on how we are going to assess

12 progression-free survival. Secondary endpoints

13 will include overall survival, thrombotic vascular

14 events, response rate, and TTP.

15 Statistical methods will include a

16 non-inferiority comparison, possibly followed by a

17 superiority test. Two thousand patients will

18 provide 80-percent power to exclude a 15-percent

19 reduction in progression-free survival, assuming no

20 difference. If non-inferiority is demonstrated, a

21 superiority test will be done. There will be

22 80-percent power to detect a 15-percent gain in

1 progression-free survival.

2 There are some considerations when

3 designing the trial in which we will particularly

4 welcome your feedback. The first challenge is to

5 run a placebo-controlled trial when anemic patients

6 receive drug treatment as a standard of care.

7 Crossover of placebo patients following the

8 double-blind phase could obscure the assessment of

9 overall survival.

10 Second, functionality of the EPO receptor

11 is best addressed in fresh frozen samples.

12 Collecting samples may significantly slow down

13 patient enrollment into the trial and would delay

14 study completion. However, more preclinical

15 studies to assess ligand affinity, signal

16 transduction, and gene expression are warranted to

17 better understand the receptor and its

18 functionality.

19 Providing patients with a homogenous

20 chemotherapy regimen is complicated, but at least

21 three elements: the previous adjuvant

22 chemotherapy, the wide range of available

1 therapies, and constant innovation in therapy.

2 And, finally, this clinical trial should

3 provide an opportunity to better understand and

4 control the causes of thrombotic events.

5 In the next two to three years, as

6 depicted on the slide, we will have considerably

7 more information in the areas of tumor control and

8 survival from the tumor types where we have

9 observed a survival signal: breast cancer, head

10 and neck cancer, lung cancer, as well as some more

11 data in carcinoma of the cervix, all in studies

12 beyond the correction of anemia.

13 In summary, we will have a significant

14 amount of additional data in the next two to three

15 years from those recently completed studies and

16 ongoing studies. This data will provide

17 significant information in various tumor types.

18 We welcome your advice and opinions on the

19 timing, design, and challenges of the proposed

20 study.

21 And now I would like to conclude the

22 Johnson & Johnson presentation. As you have read,

1 seen, and heard, in the supportive care of anemia

2 we have extensive clinical experience which

3 supports the favorable benefit/risk profile of

4 Procrit. We take very seriously the survival

5 signal observed in metastatic breast cancer and

6 head and neck cancer that occurred in studies

7 assessing the benefit beyond the correction of

8 anemia with two different products: EPREX and

9 NeoRecormon. We have looked for and found no clear

10 tumor proliferation signal as assessed by response

11 rate and tumor progression.

12 We note that TVEs account for some,

13 potentially much, of the negative signal we have

14 observed in those trials. In contrast, some

15 studies in supportive anemia suggest a potential

16 benefit in cancer outcome, and future clinical

17 evaluation in that setting may provide the answer

18 to that question.

19 In summary, Procrit provides important

20 benefits for patients with cancer by decreasing

21 transfusion and alleviating anemia symptoms. We

22 are committed to maximizing those benefits and

1 minimizing the risks associated with its use.

2 We look forward to working with ODAC and

3 FDA to optimize our current and future development

4 programs.

5 Thank you very much for your attention.

6 DR. CHESON: Now we will move on to the

7 Amgen presentations, their partners for the day.

8 Dawn Viveash will do the introductions.

9 DR. VIVEASH: Good morning, members of the

10 committee, FDA participants, ladies and gentlemen.

11 Amgen is pleased to be here today to present data

12 regarding the benefit and safety of Aranesp in the

13 treatment of patients with chemotherapy-induced

14 anemia.

15 We have with us today a number of

16 distinguished guests: Dr. Jeffrey Crawford, Dr.

17 David DeMets, Dr. John Glaspy, Dr. Harvey Lodish,

18 Dr. Douglas Losordo, Dr. Marc Pfeffer, and Dr.

19 Joseph Eschbach.

20 In addition, we have a number of

21 independent investigators who are currently

22 conducting oncology studies with Aranesp. These

1 investigators are Dr. Overgaard, representing the

2 Danish Head and Neck Cancer Study Group; Directors.

3 Delarue and Bosley, representing the GELA Lymphoma

4 Study Group; Dr. Nitz, representing the West German

5 study; and Dr. Kahlert, representing the German

6 Gynecological Oncology Study Group.

7 I will open the presentation with a brief

8 overview on preclinical and clinical properties of

9 Aranesp. There has been a change on our agenda.

10 As you'll see, we have a different cast of

11 presenters than is shown on the published agenda.

12 We will have Dr. Harvey Lodish discuss considerations

13 regarding the epoetin receptor. His lab was

14 the first to clone the EPO receptor. He is

15 professor of biology and bioengineering at MIT and

16 is a member of the National Academy of Science.

17 Dr. David Parkinson will describe the clinical

18 observations with Aranesp, and he will also provide

19 an overview of our clinical trial program.

20 Aranesp is a distinct erythropoietic

21 molecule. The development of Aranesp represents

22 the combination of over ten years of research

1 during which time more than 450 molecules were

2 characterized. Aranesp is unique as a result of

3 its novel amino acid sequence, which allows for two

4 additional carbohydrate chains, leading to an

5 increased negative charge and increase in molecular

6 weight. The terminal half-life of Aranesp is

7 three-fold greater than epoetin, and because of its

8 longer half-life less frequent dosing can be

9 utilized compared to erythropoietin.

10 Aranesp was initially approved in 2001 for

11 the treatment of anemia associated with chronic

12 renal failure in both dialysis and non-dialysis

13 patients. It was subsequently approved in July of

14 2002 for chemotherapy-induced anemia.

15 I'd like to highlight some relevant safety

16 information from the package insert. The warnings

17 section represents prior observations from the

18 Normal Hematocrit Study which was conducted with

19 EPOGEN. This was conducted in dialysis patients

20 with pre-existing cardiovascular disease. This

21 section also addresses high hemoglobin, rate of

22 rise, and mortality.

1 The dosing guidance recommends a

2 hemoglobin target of 12 and provides instructions

3 for dosage adjustment to avoid excessive rate of

4 rise of hemoglobin.

5 The precautions section includes a

6 statement regarding the theoretical concern of

7 growth factor potential, and the adverse reactions

8 section describes the thrombovascular events.

9 You are now well aware of the findings

10 from studies with epoetin alfa and epoetin beta and

11 their observations regarding survival, tumor

12 progression, and thrombotic events. When Amgen

13 became aware of these findings, we conducted a

14 comprehensive review of preclinical and clinical

15 data.

16 The preclinical data with respect to

17 Aranesp does not support the contention that this

18 agent stimulates tumor growth. Aranesp is not

19 genotoxic. There were not proliferative or

20 hyperplastic signals in six-month toxicology

21 studies. In addition, there was no off-target

22 binding of Aranesp, and no off-target effects were

1 seen with Aranesp or erythropoietin in toxicology

2 studies.

3 In studies of tumor xenografts, one of

4 which was performed by Dr. Blackwell from Duke

5 University, who is present here today, there was no

6 stimulation of tumor proliferation. In fact, to

7 the contrary, there was a potential beneficial

8 effect observed when Aranesp was administered in

9 combination with radiotherapy in some models.

10 The clinical review includes

11 epidemiological analysis of thrombotic events and a

12 review of completed and ongoing Aranesp trials and

13 also an assessment of post-marketing experience.

14 Dr. Parkinson will review our observations from the

15 clinical data.

16 Based on this comprehensive review of

17 oncology data, we did not identify any adverse

18 survival or tumor progression signal with Aranesp.

19 The thrombotic event rate remains consistent with

20 that represented in the product label.

21 One of the hypotheses that has been put

22 forward from the signals observed in the BEST and

1 Enhanced studies relates to the role of the EPO

2 receptor in tumor progression. I would like to ask

3 Dr. Lodish to address the potential relevance of

4 the EPO receptor on tumors and the utility of

5 current methods to detect the receptor.

6 Thank you, Dr. Lodish.

7 DR. LODISH: Thank you.

8 To begin, I'd like to emphasize that mere

9 detection of the EPO receptor on tumor cells--or

10 normal cells, for that matter--does not mean that

11 erythropoietic agents drive the oncogenic process.

12 The EPO receptor is present at very low levels on

13 many normal and tumor cells, but the EPO receptor

14 does not possess any of the characteristics of an

15 oncogenic receptor.

16 For example, as you know, established

17 oncogenic tyrosine kinase receptors, such as HER2

18 or the epidermal growth factor receptor, are

19 amplified and mutated in many types of human

20 tumors. Receptors can be overexpressed as many as

21 100,000 or a million copies per cell in certain

22 cancers. In other cases, mutation leads to

1 constituitive--that is, hormone

2 independent--activation. Both cases are

3 transforming, are prognostic markers, and are

4 established therapeutic targets.

5 The situation is quite different for the

6 EPO receptor. With the sole exception of erythroleukemia,

7 where EPO gene amplification has been

8 recognized, EPO receptor amplification has not been

9 seen in human tumors. The presence of gene

10 amplification into erythroleukemic cell lines

11 illustrates that the failure to detect involvement

12 of the EPO receptor in the vast majority of cancer

13 samples is genuine and not simply a false negative

14 result. And it's my understanding that Aranesp

15 treatment of erythroleukemia is not recommended.

16 Importantly, there are no constituitive

17 reactive--that is, hormone independent--EPO

18 receptor mutants in any human or animal tumors.

19 The one case of humans with mutations in the EPO

20 receptor involve truncations of the cytoplasmic

21 domain that render the receptors hypersensitive to

22 erythropoietin. These individuals develop

1 polycythemia but have no increased tumor incidence.

2 And, in conclusion, then, the EPO receptor

3 is not known to initiate tumorigenicity or cause

4 primary solid tumors to proliferate. There are no

5 known correlations of EPO receptor expression or

6 mutation with any aspect of oncogenicity.

7 I've also been asked to comment on

8 methodological aspects of existing and potential

9 assays for functional EPO receptors on primary

10 solid tumors. And before doing that, I'd like to

11 point out several important aspects of EPO receptor

12 expression on erythroid cells.

13 First of all, over 90 percent, well over

14 90 percent of the EPO receptors in erythroid cells

15 are not on the cell surface. They're in the

16 cytoplasm on various membranes. Erythroid cells

17 have only 1,000 to 2,000 receptors on their

18 surface. Non-erythroid cells are transformed or

19 otherwise generally have much less. And,

20 importantly, surface expression of the receptor

21 requires expression of the JAK-2 protein tyrosine

22 kinase and possibly other accessory proteins.

1 Finally, the high-affinity receptor that

2 is seen on erythroid cells, the signaling receptor,

3 forms a one-erythropoietin, 2-receptor complex that

4 initiates downstream signaling. The low-affinity

5 receptors that are seen on the vast majority of

6 normal and tumor cells are low-affinity, as I said,

7 and likely are forming a 1-erythropoietin,

8 1-erythropoietin complex and are not signaling.

9 Concerning the assays that one might think

10 of for erythropoietin receptor detection in primary

11 tumors, I'd like to point out several points.

12 First of all, numerous publications discuss EPO

13 receptor expression and function in tumor cell

14 lines, but it's not clear that these translate to

15 primary tumor samples in a clinical setting. And,

16 importantly, only cell surface receptors are

17 clinically and biologically relevant. Only these

18 receptors can bind to erythropoietin and send

19 signals to the inside of the cell.

20 It's important to note that there are no

21 measurements for functional epoetin receptors

22 possible in fixed or frozen tissues. Reverse

1 transcriptase polymerase chain reaction, RT-PCR,

2 measures RNA copies or transcripts of the EPO

3 receptor gene. That does not necessarily measure

4 functional EPO receptor message and does not

5 measure EPO receptor protein, and certainly not

6 functional receptor. And, importantly, these

7 studies would require separation of the tumor cells

8 from the other cells in the tumor.

9 Immunohistochemistry measures erythropoietin

10 receptors in the cytoplasm and is too

11 insensitive to detect the minute numbers that might

12 be expected on the surface of cells. And,

13 importantly, the existing antibodies, commercial or

14 otherwise, are simply not sufficiently specific to

15 detect EPO receptors among other background

16 proteins.

17 There are ways of detecting functional EPO

18 receptors in fresh tumor biopsies, but they also

19 present many problems. First of all, these

20 measurements would require fresh samples of cells

21 and samples in which the tumor cells have been

22 separate from the non-tumor cells. Binding with

1 radiolabeled EPO to cell surface receptors is

2 possible, but it is very difficult to detect the

3 low numbers of low-affinity receptors--and by low

4 numbers, I mean under 1,000 receptors--present in

5 cells. And it's difficult to resolve the specific

6 saturable binding to cell surface EPO receptors

7 from the non-specific, non-saturable binding to

8 other cell surface components.

9 Proliferation of tumor cells in culture

10 and response to EPO is also not practical for the

11 simple reason that, as you know, fresh tumor cells

12 generally are not viable in culture. In my view,

13 the only assay that would detect functional EPO

14 receptors in tumor cells--or, for that matter,

15 other types of cells--involve EPO-induced

16 activation of downstream signaling proteins as

17 measured by, say, phosphorylation of the

18 erythropoietin receptor, the JAK-2 kinase, other

19 signaling proteins. These are complicated assays

20 that require, as do the others, on the order of ten

21 million cells per assay. The cells, again, must

22 have been purified from other cells, and in

1 non-erythroid cells, these immuno-precipitation

2 Western blot analyses are quite insensitive and

3 have a very low signal-to-background ratio.

4 So, in conclusion, there are no presently

5 available assays suitable for routine measurement

6 of functional erythropoietin receptors on primary

7 solid human tumors. Development of such assays

8 will take years, and it's unclear to me what form

9 these assays might ultimately take.

10 I now turn the podium over to Dr.

11 Parkinson, who will discuss the clinical

12 observations.

13 DR. PARKINSON: Good morning. Thank you,

14 Dr. Lodish.

15 Outlined are the clinical observations

16 which I will discuss relevant to this morning's

17 meeting. After briefly reviewing some of the

18 benefits associated with the treatment of anemia,

19 I'll present the results of Amgen's studies of the

20 risk of thrombotic events in association with

21 erythropoietins. Next I'll present the analysis of

22 survival in completed clinical trials. And,

1 finally, I'll outline a program of ongoing trials

2 involving Aranesp in different tumor treatment

3 settings.

4 Together, these trials have power to

5 detect a safety signal far smaller than those which

6 have been discussed already this morning. We

7 believe this represents a responsible and credible

8 approach to definitively resolving the questions

9 raise in this morning's meeting.

10 With regard to the cancer indication,

11 today we're here primarily to consider risks. But

12 no meaningful discussion of risk can occur in the

13 absence of a consideration of benefit. Anemia,

14 which translates in patients with cancer into the

15 important symptom of fatigue, is a highly prevalent

16 comorbidity which significantly affects the quality

17 of life in patients with cancer. Without

18 erythropoietic protein therapy, 90 percent of

19 cancer patients undergoing chemotherapy will have

20 some level of anemia, and some 40 to 60 percent of

21 those patients will require transfusions.

22 Historically, chemotherapy-related anemia

1 has been treated with transfusion, with its

2 attendant inconveniences and risks. Not only is

3 fatigue common in cancer patients, but fatigue as a

4 symptom is rated by the majority of patients to be

5 more important even than pain.

6 The left side of this panel shows the

7 hematopoietic response indication correction of

8 anemia by Aranesp therapy. Portrayed to the right

9 is the significant decrease in the rate of

10 transfusion with Aranesp therapy utilizing dosing

11 intervals extending as far as three weeks.

12 Extensive literature suggests the

13 association of this anemia correction with improved

14 fatigue and other quality-of-life scores.

15 Recognition by the oncology community of the

16 importance of anemia and the benefits of its

17 treatment with erythropoietic proteins have led to

18 the production of independent, evidence-based

19 treatment guidelines. These include treatment

20 algorithms and desirable upper levels for

21 hemoglobin.

22 These evidence-based guidelines have been

1 incorporated by Amgen into our current trials and

2 analyses. Furthermore, treatment recommendations

3 in the product label are consistent with these

4 guidelines.

5 We'll now present the results of our

6 evaluation of thrombotic events in patients with

7 cancer. First of all, it's well established that

8 patients with cancer have a higher background rate

9 of thrombotic events. A full description of the

10 epidemiology of these events in patients with

11 cancer is outlined in our briefing document. We

12 have extensively reviewed that.

13 The increased risk of thrombotic events

14 with Aranesp therapy is represented in the adverse

15 events section of the Aranesp label, as has already

16 been discussed by Dr. Viveash. But we proactively

17 initiated a reevaluation of thrombotic event

18 experience within Aranesp clinical trials--these

19 are 11 completed trials as of late last

20 year--involving more than 1,800 Aranesp-treated

21 subjects relative to more than 400 placebo-treated

22 subjects.

1 On this slide, we see that our own Amgen

2 analysis of the Medstat Claims database reflecting

3 patients treated primarily with erythropoietin alfa

4 also shows an increased risk of thrombotic events

5 with epoetin alfa therapy. This analysis is

6 consistent with the Cochran meta-analysis involving

7 cancer patients receiving either erythropoietin

8 alfa or beta, presented by Bohlius, et al., at the

9 December American Society of Hematology meeting,

10 the relative risks of thrombotic events in our

11 study and the Bohlius study being 1.4 and 1.55,

12 respectively.

13 We'll now show you our analysis of

14 survival in completed clinical trials.

15 We identified four suitable randomized,

16 double-blind, placebo-controlled trials. Two of

17 these, involving more than 600 patients, had

18 long-term follow-up and with 360 events allow us to

19 carefully evaluate Aranesp's effect on survival.

20 One trial was conducted in lung cancer and included

21 anemic patients beginning platinum-based

22 chemotherapy. A second trial involved patients

1 with five different lymphoid malignancies. In this

2 trial, Aranesp therapy was initiated when patients

3 became anemic. Finally, Amgen conducted a pooled

4 analysis involving these two trials and two

5 additional controlled trials comprising more

6 heterogeneous patient populations.

7 The first of the studies, in lung cancer,

8 is represented on this slide. More than 300

9 patients with either small-cell or non-small-cell

10 lung cancer beginning platinum-based chemotherapy

11 were randomized to weekly Aranesp or placebo. The

12 relatively homogeneous patient population, the fact

13 that most patients were beginning chemotherapy, and

14 the long-term follow-up make the study very

15 appropriate for survival analysis. Seventy percent

16 of these patients have been followed until death.

17 On this slide, we see the results of this

18 study in lung cancer. There is no evidence of any

19 decrease in progression-free survival with Aranesp.

20 In the Amgen briefing document, we've provided a

21 breakdown of small-cell and non-small-cell lung

22 cancer subjects. These subsets behave similarly.

1 This slide shows similar results for

2 overall survival. The sample size of the trial and

3 the number of observed deaths were appropriate to

4 detect reduced survival of the magnitude seen in

5 the BEST and Enhanced or Henke trials. Yet there

6 is evidence for any negative survival influence

7 with Aranesp therapy.

8 Trial 161, this lymphoid malignancy trial,

9 differs from the lung cancer trial, as I've

10 indicated, since patients with multiple lymphoid

11 tumor types were eligible, and these patients could

12 be randomized anytime during the course of

13 chemotherapy. In this study, 344 patients with one

14 of five different lymphoid malignancies with

15 chemotherapy-induced anemia were randomized to

16 receive either weekly Aranesp or placebo. The

17 distribution of the different malignancies is

18 outlined here.

19 The slide illustrates the baseline

20 characteristics of the patients in the lymphoid

21 malignancy trial. The study, while it did include

22 long-term follow-up, was again designed to study

1 anemia. As a consequence, patients were not

2 stratified for malignancy-specific prognostic

3 factors. This led by chance, as you can see, to

4 patients with the worse prognosis for both

5 non-Hodgkin's lymphoma and chronic lymphocytic

6 leukemia to be assigned to the Aranesp arm.

7 This slide indicates the trial result. We

8 see on this slide no evidence for a significant

9 decrease in progression-free survival. The hazard

10 ratio, which is adjusted for disease type, stage,

11 and IPI score, is greater than 1 but the confidence

12 interval extends below 1. We continue to follow

13 these patients.

14 On this slide, we observe no convincing

15 evidence for a significant decrease in overall

16 survival in association with Aranesp therapy.

17 Again, the hazard ratio is above 1, but the

18 confidence interval extends below 1. We've

19 presented data on individual lymphoid malignancy

20 subset in the briefing document.

21 I will now review the pooled analyses for

22 these completed trials.

1 As previously noted, two other randomized,

2 double-blind, placebo-controlled short trials with

3 short-term follow-up were considered to be

4 appropriate for the pooled analysis and to

5 contribute particularly to the study of the early

6 part of the survival curve which seemed to be so

7 important in the BEST trial results, as you've

8 heard.

9 On this slide are demonstrated the number

10 of patients and the breakdown by tumor type of the

11 patients contributing to this pooled analysis with

12 cumulative follow-up involved. Combined, these

13 trials provide more than a 80-percent power to

14 detect an effect on survival of the magnitude seen

15 in the BEST and Enhanced trials.

16 I'll now review results starting with

17 progression-free survival.

18 Portrayed here is the progression-free

19 survival in the overall pooled analysis. Note here

20 that the time scale extends to 16 weeks and that

21 the progression-free survival percent extends from

22 80 to 100. We've magnified the scale. The hazard

1 ratio is close to 1, and there is no evidence of an

2 effect of Aranesp on progression-free survival

3 during this period.

4 On this slide, we again see no evidence

5 for a negative overall survival influence in

6 association with Aranesp therapy. In addition, as

7 shown in our briefing document, the long-term

8 follow-up from this pooled data set is a hazard

9 ratio of approximately 1. The confidence interval

10 for that analysis extends from 0.8 to 1.2, which

11 excludes an effect of the size seen in the BEST and

12 Enhanced trials.

13 I will now review the analysis by tumor

14 type.

15 On this slide, I portray the

16 progression-free survival results of the pooled

17 analysis by tumor type. No clear association is

18 observed between progression-free survival and

19 tumor type. Results are similar with respect to

20 overall survival.

21 Here we find an association with improved

22 progression-free survival and overall survival is

1 observed with respect to achieving an on-study rise

2 in hemoglobin of 1 gram per deciliter or more over

3 14 days. These hazard ratios are 0.51 and 0.43,

4 respectively, with the indicated confidence

5 intervals.

6 Note that a similar association is found

7 with improved progression-free survival and overall

8 survival with respect to achieving an on-study

9 hemoglobin of greater than or equal to 13 grams per

10 deciliter.

11 In summary, our more recent analyses have

12 confirmed the appropriateness of the Aranesp

13 prescribing information with respect to thrombotic

14 event rate. In an evaluation of data from over

15 1,100 patients randomized to placebo-controlled

16 oncology trials with Aranesp, we found nearly

17 identical survival and progression-free survival

18 with Aranesp and placebo. We believe that our

19 detailed examination confirms the safety profile of

20 Aranesp and that the benefit/risk ratio remains

21 favorable and warrants continued examination of

22 potential beneficial effects on survival.

1 I will now review a program of ongoing

2 trials involving Aranesp in different tumor

3 treatment settings. We believe this group of

4 trials represents a robust approach to ultimately

5 resolving the questions raised in this meeting.

6 The trials to be described were initiated, I should

7 point out, because of evidence regarding the

8 positive potential benefits of anemia treatment on

9 patient survival. Outlined here are the relevant

10 preclinical and clinical observations providing the

11 rationale for these trials.

12 On particular note at the bottom is the

13 Cochran meta-analysis with a favorable relative

14 risk and a conclusion by the authors that more

15 trials to explore this finding were merited.

16 On the next several slides are outlined

17 the Amgen-sponsored and the four independent

18 investigator-initiated and -conducted studies.

19 The Amgen response to the information from the BEST

20 and Enhanced trials has already been described by

21 Dr. Viveash, including our formal review of all

22 ongoing clinical trials involving Aranesp being

1 conducted worldwide.

2 One of our goals in this review was to

3 identify clinical trials in which the design, the

4 size, and the patient population would be

5 particularly informative with respect to answering

6 the kinds of questions that we're dealing with

7 today. We identified five such trials--one

8 Amgen-sponsored and four utilizing Aranesp but

9 being conducted by independent investigators. All

10 of these studies are randomized and controlled.

11 One trial is itself double-blind and

12 placebo-controlled. The other four clinical trials

13 involve randomization to Aranesp or no epoetin. In

14 these trials, Aranesp treatment is administered

15 proximate to the time of chemotherapy and not for

16 the full duration of follow-up. These studies

17 include long-term follow-up with collection of

18 predefined progression and survival endpoints. In

19 addition, of course, the studies will capture

20 thrombotic and cardiovascular events. Each study

21 includes homogeneous populations with

22 stratification for disease-specific prognostic

1 variables.

2 One question posed by the FDA relates to

3 the feasibility and appropriateness of conducting

4 placebo-controlled studies. You will note that, as

5 I've indicated, one of our studies includes

6 placebo-controlled design. While these studies are

7 currently ongoing in Europe, we can report that we

8 are successfully accruing patients to a

9 placebo-controlled trial of Aranesp in

10 chemotherapy-induced anemia in the United States if

11 that's relevant to your deliberations.

12 In fact, it is our opinion that controlled

13 studies are essential in certain situations and

14 that it is feasible to conduct such studies in the

15 United States.

16 On this slide, we also indicate that the

17 number of patients for each tumor type and the

18 total number of patients for these five trials

19 being over 3,500. We believe that there is

20 particular value to an approach which incorporates

21 a range of tumors with robust numbers of patients

22 in both breast cancer and head and neck cancer. I

1 will now review each study design in detail.

2 Portrayed here is the Amgen-sponsored,

3 double-blind, placebo-controlled study. Six

4 hundred patients with newly diagnosed extensive

5 small-cell lung cancer will be randomized to

6 combination chemotherapy with Aranesp or placebo.

7 As you can see, endpoints include survival, and

8 this trial has accrued more than 200 patients to

9 date. I'd like to point out again that this trial

10 is placebo-controlled.

11 The first independent

12 investigator-conducted trial which I will discuss

13 is the neoadjuvant breast cancer trial being

14 conducted by the German Gynecologic Oncology Group.

15 Seven hundred patients with diagnosed breast cancer

16 will be randomized to dose-intense or standard

17 chemotherapy with a secondary randomization to

18 Aranesp or observation. Following induction

19 chemotherapy, surgery will be conducted. Endpoints

20 are as listed; follow-up is long term.

21 By the nature of this patient population

22 and by the nature of the study design and

1 investigator intent with Amgen support, tumor

2 tissue is being collected and stored. The trial

3 has accrued more than 400 patients, half of the

4 projected total accrual. An interim analysis of

5 the experience in the first 200 patients will take

6 place in the next several weeks.

7 The second investigator-initiated study is

8 the adjuvant breast cancer study being conducted by

9 the West German Study Group. After definitive

10 surgery, the projected 1,000 patients will be

11 randomized to center-specific adjuvant chemotherapy

12 with or without Aranesp. Endpoints are as listed,

13 and this trial has recently initiated accrual.

14 The diffuse large-cell lymphoma study

15 conducted by the French, Belgian, and Swiss GELA,

16 is outlined here. More than 600 patients will be

17 randomized to 14- or 21-day monoclonal antibody

18 CHOP(?) chemotherapy treatment regimens. These

19 patients are secondarily randomized to Aranesp or

20 supportive transfusion. Endpoints are as listed;

21 long-term follow-up is involved. This trial has

22 recently initiated accrual.

1 The head and neck cancer study being

2 conducted by the Danish Head and Neck Cancer Study

3 Group is outlined here to test the hypothesis that

4 anemia contributes to radiotherapy failure. A

5 projected 600 patients with head and neck cancer

6 are randomized to radiotherapy alone or to Aranesp

7 with long-term follow-up. The principal

8 investigator is Professor Overgaard, a

9 well-recognized authority in the field of tumor

10 oxygenation and radiation therapy. More than 260

11 patients have already been accrued to this trial.

12 In response to the Henke and Enhanced

13 trial results, the investigators have conducted an

14 interim analysis for safety. We are informed that

15 this trial is proceeding.

16 On this slide, the five clinical trials

17 are outlined with respect to the tumor types

18 involves, projected and current accrual, and the

19 detectable differences from the expected control

20 arm results. Individually, these trials will

21 accrue between 600 and 1,000 patients and have

22 power to detect absolute differences in survival

1 between 7 and 11 percent. Note that these studies

2 are ongoing outside of the United States, but we

3 believe the findings should absolutely be

4 applicable to United States practice.

5 This slide shows the statistical power of

6 the individual trials to detect an increase in the

7 risk of death. Each of these trials has reasonable

8 power to detect a hazard ratio of 1.4 or 1.5. Even

9 if the true hazard ratio is as low as 1.2, there is

10 a greater than 85-percent chance that at least one

11 of these trials will result in a statistically

12 significant difference.

13 On this slide is outlined the projected

14 accrual over time to these trials and the expected

15 cumulative patient years of follow-up. Including

16 all five ongoing studies, more than 3,500 patients

17 will be randomized in trial settings in which the

18 influence of Aranesp on survival can be compared.

19 This slide shows the power of a

20 meta-analysis illustrated in yellow of all five

21 trials. This analysis will have high power to

22 detect a true hazard ratio as small as 1.15, which

1 is far smaller than that observed in the BEST and

2 Enhanced trials.

3 Also shown on this graph in the purple is

4 the power of the meta-analysis of the neoadjuvant

5 and adjuvant breast cancer studies, a total of

6 1,700 breast cancer patients. This analysis will

7 have 80-percent power to detect a true hazard ratio

8 as small as 1.32.

9 So on this slide, I've summarized the

10 strengths of the ongoing clinical trials

11 activities. As I've discussed, these include

12 design elements which involve either double-blind,

13 placebo-controlled, or Aranesp versus epoetin

14 elements, with predefined survival or tumor

15 progression endpoints. I'd like to emphasize this

16 in view of the agency's first question.

17 While it is true that these trials are all

18 being conducted ex-U.S., we would point out that it

19 is entirely possible to conduct placebo-controlled

20 trials in the United States. These ongoing trials

21 cross multiple tumor types with approximately 1,700

22 breast cancer patients and 600 head and neck cancer

1 patients. The cumulative meta-analyses of 3,500

2 patients will provide robust power for assessment

3 of survival outcomes in this program.

4 Of note, these studies have already

5 accrued close to 900 patients. These studies

6 include careful safety monitoring, and the AGO

7 breast cancer trial incorporates tissue collection

8 to enable appropriate correlative biological

9 studies.

10 In conclusion, we've outlined the known

11 and potential benefits of therapy with Aranesp. We

12 have found no adverse effects on tumor progression

13 or survival to date in our Aranesp clinical trials.

14 To the contrary, evidence exists for potential

15 benefit from erythropoietic protein therapy, both

16 in the settings of cancer and other conditions.

17 It is our position that this potential

18 benefit should be studies, but that such studies

19 must be carried out responsibly, with carefully

20 designed and executed trials.

21 Thank you very much.

T2A DR. CHESON: I would like to thank the

1 sponsors for their very clear and on-time

2 presentations.

3 And now I'd like to turn to the FDA

4 presentation, Dr. Harvey Luksenburg--who is going

5 out the door.

6 [Laughter.]

7 DR. CHESON: Harvey, come back, please.

8 And for those of you who are standing against the

9 side wall, if you would please, for fire safety

10 reasons, stand in the back or you'll have to be

11 asked to leave the room.

12 DR. LUKSENBURG: Dr. Cheson, members of

13 the committee, ladies and gentlemen, I'm Harvey

14 Luksenburg. I'm a clinical reviewer at the Food

15 and Drug Administration, and I would just like to

16 start off by noting that I am but a member of a

17 team of very talented individuals who put in a

18 tremendous amount of work in putting together the

19 data which we'll be presenting today.

20 Now, two large randomized studies in

21 cancer patients on chemotherapy plus or minus EPO

22 have shown shorter overall survival, shorter

1 progression-free survival, and an increased

2 incidence of thrombotic and cardiovascular events

3 in the groups assigned to receive erythropoietins.

4 The erythropoietin products used in these

5 two studies are not licensed in the U.S. They are

6 NeoRecormon, epoetin beta, manufactured by

7 Hoffman-LaRoche, and EPREX, epoetin alfa, would is

8 manufactured by Ortho Biologics. Both of these

9 studies used a treatment strategy to achieve a

10 hemoglobin greater than 12 grams per deciliter,

11 which is higher than that recommended in the

12 labeling for U.S.-licensed products.

13 The clinical trials for U.S.-licensed EPO

14 products were not designed to assess the impact on

15 response rate, with one exception--the N93 study,

16 which I'll describe momentarily; they were not

17 designed to look at in a systematic way time to

18 progression or progression-free survival; and they

19 were not designed to look at overall survival.

20 Now, the goals of my talk are four-fold.

21 First of all, I'll try to give some justification

22 of why the FDA feels that the safety issues

1 observed with EPREX and NeoRecormon, the

2 non-U.S.-licensed EPOs, may also apply to

3 U.S.-licensed products. In addition, I will review

4 results of trials with EPREX and NeoRecormon, the

5 non-U.S.-licensed products, regarding the safety

6 concerns. Thirdly, I will review data available

7 regarding safety from trials of EPOGEN/Procrit and

8 Aranesp, the U.S.-licensed trials, and finally will

9 try to come agreement on the design of future

10 studies regarding these safety issues.

11 Now, the three safety issues which I'm

12 going to be discussing are, first of all, an

13 increased risk of thrombotic and cardiovascular

14 adverse events, an increased risk of tumor

15 progression in patients receiving EPO products, and

16 poorer survival in groups of patients receiving EPO

17 products.

18 Just the cast of characters. Recombinant

19 EPO products which are currently U.S.-licensed are

20 epoetin alfa manufactured by Amgen and marketed

21 under the name of EPOGEN; the same drug

22 manufactured by Amgen and marketed as Procrit by

1 Ortho Biotech; and darbepoetin alfa, or Aranesp,

2 manufactured and marketed by Amgen.

3 The EPO products which are not licensed in

4 the U.S. are epoetin alfa, or EPREX, manufactured

5 by Ortho Biologics; Epoetin beta, NeoRecormon,

6 manufactured by Hoffman-LaRoche.

7 Now, the FDA considers all these products

8 members of the same product class, and, thus, these

9 evolving safety issues are assumed to apply to all

10 products unless adequate and well-controlled trials

11 demonstrate otherwise.

12 The differences between these products are

13 as follows: epoetin alfa and beta have the same

14 amino acid sequence, but they differ in

15 glycosylation. Aranesp differs in the amino acid

16 sequence (5) and in the degree of glycosylation.

17 The similarities are meaningful. All

18 these exert their principal clinical effect by

19 binding to the erythropoietin receptor. All these

20 products have similar pharmacodynamic effects when

21 they're used at recommended dosages. And there's a

22 similar toxicity profile across all of these

1 products with the exception of pure red cell

2 aplasia, which has been seen thus far only in

3 EPREX.

4 Now, target hemoglobin, the labels for

5 EPOGEN/Procrit and Aranesp have dosage guidelines

6 based on safety data from registration studies

7 performed in patients with chronic renal failure.

8 Just to quote what is written on the current

9 labels, for EPOGEN/Procrit, "The suggested target

10 hematocrit range is between 30 and 36 percent."

11 For Aranesp, "The dose should be adjusted for each

12 patient to achieve and maintain a target hemoglobin

13 not to exceed 12 g/dL."

14 In addition, for rapid increase in

15 hemoglobin greater than 1 gm per deciliter, or four

16 points in hematocrit, in any two-week period, the

17 dose should be reduced. And the product should be

18 held if the hemoglobin is greater than 13 until the

19 hemoglobin falls less than or equal to 12 grams per

20 deciliter and re-start the dose at 25 percent below

21 the previous dose.

22 Now, the first safety issue which I'd like

1 to discuss is that of an increased incidence of

2 thrombotic and cardiovascular adverse events. This

3 is a road map, and I'll show this slide several

4 more times, and for each safety issue--thrombotic

5 events, tumor progression, overall survival--I'm

6 going to discuss only one study done in renal

7 patients, the Normal Hematocrit Study. These in

8 yellow are the studies done in non-U.S.-licensed

9 EPO, and the studies in pink are the studies done

10 in U.S.-licensed EPO products. An "x" means that

11 there's data available for evaluation for each of

12 these safety concerns.

13 Now, the licensing studies for

14 EPOGEN/Procrit and Aranesp demonstrated that

15 there's a baseline risk of thrombotic and

16 cardiovascular adverse events at their labeled

17 target hemoglobin, that is, between 10 and 12 grams

18 per deciliter.

19 A study which dramatically showed the

20 potential adverse effects of increasing the

21 hemoglobin was the so-called Normal Hematocrit

22 Study, first author Besarab, published in the New

1 England Journal in 1998. The idea behind this

2 study was that patients with chronic renal failure

3 on dialysis who had clinical evidence of cardiac

4 disease could do better clinically if they had

5 their hemoglobin raised from the nominal low 30

6 range to a higher hematocrit, around 40. And so

7 1,200 patients with chronic renal failure on

8 dialysis with clinical evidence of congestive heart

9 failure or ischemic heart disease, they were all on

10 EPOGEN at baseline and maintaining a hematocrit of

11 between 27 and 33 percent.

12 Now, both arms received EPOGEN, but they

13 were randomized to different treatment strategies.

14 One was randomized to achieve a higher hematocrit,

15 around 42, plus or minus 3. This was called the

16 so-called normal hematocrit group. The other arm

17 maintained the lower hematocrit group, as was

18 customary in practice, around 30 percent. This was

19 called the low hematocrit group.

20 This study had a composite primary

21 endpoint of either death or non-fatal myocardial

22 infarction, and here are the results. In the

1 normal hematocrit group, there's an increased

2 incidence of death, 30 percent, versus 34 percent

3 in the low hematocrit group. There's an increased

4 risk of non-fatal myocardial infarction, 3.1

5 percent in the normal hematocrit group, versus 2.3

6 percent in the low hematocrit group. And there was

7 an increased risk of vascular access thrombosis, 39

8 percent in the normal hematocrit group versus 29

9 percent in the low hematocrit group.

10 Here's a graph showing the increased

11 probability of death in the normal hematocrit

12 group, death or myocardial infarction in the normal

13 hematocrit group, and in the low hematocrit group.

14 This goes out to about 30 months.

15 Now, when I talk about target hemoglobin,

16 a target hemoglobin is only a target, and many

17 patients don't achieve that target. However--and

18 this has been seen in both the renal studies and in

19 the oncology studies--it's the dosing strategy, it

20 is the idea of pushing the dose of the

21 erythropoietin to a higher level in order to try to

22 attain the target hemoglobin. However, we've seen

1 in all these studies that the adverse event signals

2 seem to occur in the group assigned to the dosage

3 strategy aimed at the target hemoglobin, despite

4 whether they attained that hemoglobin or not.

5 Now, the next studies I want to discuss

6 are the BEST and the Henke studies. These are the

7 studies done in oncology patients using

8 non-U.S.-licensed erythropoietins. And, again, I'm

9 just talking about thrombotic events.

10 The Breast Cancer Erythropoietin Trial, or

11 the BEST Trial, used EPREX. This was a randomized,

12 double-blind, placebo-controlled trial in 939

13 patients with metastatic breast cancer who were

14 receiving first-line therapy. They received EPREX

15 or placebo for 12 months, and the therapy was not

16 started until the hemoglobin was less than 13.

17 The primary objective of this study was to

18 demonstrate superior survival at 12 months. The

19 target hemoglobin, again, was higher than what is

20 on the label, between 12 and 14, and this study was

21 stopped by an Independent Data Monitoring Committee

22 based on the first four months of safety data.

1 At four months, there was an increase

2 incidence of fatal thrombotic and cardiovascular

3 events. In the EPREX arm, it was 2.3 percent; in

4 the placebo arm, it was 0.4 percent.

5 The next trial that got our attention was

6 published in The Lancet last October by Henke and

7 his colleagues, and it used NeoRecormon, or epoetin

8 beta. This was a randomized, double-blind,

9 placebo-controlled trial in 351 patients with head

10 and neck cancer who were receiving concurrent

11 radiation therapy. All these patients were anemic,

12 less than 12 grams per deciliter in women, less

13 than 13 grams per deciliter in men.

14 The primary objective in this trial was to

15 demonstrate superior locoregional progression-free

16 survival. The target hemoglobin was less than or

17 equal to 14 in women and less than or equal to 15

18 in men.

19 Now, the incidence of cardiovascular and

20 thrombotic events was higher in the epoetin beta

21 arm, 11 percent, versus placebo--this included

22 hypertension, hemorrhage, venous thrombosis,

1 pulmonary embolism, and stroke. In addition, the

2 incidence of patients who died of cardiac disorders

3 not otherwise specified was 5 percent in the

4 epoetin beta group versus 3 percent in the placebo

5 group.

6 Next, still in the thrombotic events

7 column, I'm going to discuss the studies we have

8 available to us on the U.S.-licensed epoetin

9 products.

10 The registration studies for Procrit

11 consisted of pooled analyses of six multicenter,

12 randomized, double-blind, placebo-controlled

13 studies constituting a total of 131 patients. They

14 had various primary cancers. Three of these

15 studies consisted of patients receiving

16 platinum-containing chemotherapy and three of them

17 consisted of patients receiving

18 non-platinum-containing chemotherapy. All these

19 patients were anemic, and the primary endpoint was

20 proportion of patients transfused. There were no

21 progression-free survival or survival endpoints

22 incorporated in these studies.

1 The incidence of thrombotic and

2 cardiovascular events in the pooled data was 12

3 percent in the placebo group and 3 percent in the

4 Procrit group.

5 A post-marketing commitment study done

6 after the approval of EPOGEN/Procrit for the

7 oncology indication asked the question whether

8 giving Procrit along with chemotherapy for

9 small-cell carcinoma of the lung would have a

10 potential adverse effect on the tumor's response to

11 chemotherapy. This was a randomized, double-blind,

12 placebo-controlled, non-inferiority study which was

13 intended to enroll 400 patients with small-cell

14 carcinoma of the lung who were receiving first-line

15 therapy and their baseline hemoglobin was less than

16 14. So these patients did not necessarily have to

17 be anemic.

18 The primary endpoint, as I mentioned, was

19 the objective response rate, CR plus PR, after

20 three cycles of chemotherapy to rule out a

21 decrement of 15 percent in the overall response

22 rate with Procrit. There was no target hemoglobin;

1 however, the Procrit dose was reduced if the

2 hemoglobin exceeded 16 grams per deciliter. The

3 study, however, was terminated because of poor

4 accrual at 224 patients.

5 Now, the incidence of thrombotic and

6 vascular events in this study--we did review the

7 data after 224 patients--in the Procrit group was

8 22 percent and in the placebo group was 23 percent.

9 However, the definition of thrombotic and vascular

10 events included chest pain, not otherwise

11 specified, as well as all the other well-known

12 clinical entities. So we subtracted chest pain and

13 came up with these figures: for the Procrit group,

14 the incidence of thrombotic/vascular events went to

15 14 percent, and in the placebo group, it was 9.5

16 percent.

17 The Aranesp Oncology Registration Study

18 was a randomized, double-blind, placebo-controlled

19 study in 320 patients with both small-cell and

20 non-small-cell lung cancer, all of who were

21 receiving platinum-containing chemotherapy. All

22 these patients were anemic.

1 The primary endpoint, again, was a

2 transfusion endpoint, the proportion of patients

3 transfused between week 5 and week 12 or the end of

4 the treatment period. The dosage guidelines were

5 that Aranesp was to be held for hemoglobin of

6 greater than or equal to 14 in women and for

7 greater than or equal to 15 in men.

8 The incidence of thrombotic events in this

9 study was 5 percent in the Aranesp group and 3

10 percent in the placebo group.

11 So, to summarize the studies for the

12 thrombotic/cardiovascular events so far, in the

13 studies in which a signal was detected, the Normal

14 Hematocrit Study done in patients with chronic

15 renal failure, the incidence of non-fatal

16 myocardial infarction, 3.1 percent in the normal

17 hematocrit group versus 2.3 percent in the low

18 hematocrit group. An increased incidence of

19 vascular access thrombosis, 39 percent in the

20 normal hematocrit group versus 29 percent in the

21 low hematocrit group. In the BEST Study, done in

22 939 patients with metastatic breast cancer, there

1 was an increased risk of fatal thrombotic events in

2 the arm randomized to receive EPREX, 2.3 percent,

3 versus 0.4 percent in the placebo arm.

4 In the Henke Study in head and neck cancer

5 and the patients were randomized to receiving

6 epoetin beta, or NeoRecormon, or placebo, there was

7 also an increased risk of cardiovascular and

8 thrombotic events, 11 percent in the epoetin beta

9 group versus 5 percent in the placebo group.

10 In the thrombotic and vascular events

11 studies that didn't have a signal, the Procrit

12 pooled studies, 3 percent in the Procrit group

13 versus 12 percent in the placebo group. The N93

14 study in small-cell carcinoma of the lung, 22

15 percent Procrit versus 23 percent placebo. We put

16 an asterisk next to this because after we

17 subtracted the non-specific chest pain, we did find

18 that there was an increased risk of

19 thrombotic/vascular events in the Procrit group.

20 And, finally, the Aranesp Oncology Registration

21 Study, 5 percent incidence in the Aranesp group

22 versus 3 percent in the placebo group.

1 Now, in September 2003, three

2 placebo-controlled clinical trials in oncology

3 patients in which one arm received EPO to target a

4 higher hemoglobin were terminated because of

5 unexpected rates of thrombotic events in the EPO

6 arm.

7 Briefly, to summarize these studies, in

8 one, the primary cancer was small-cell carcinoma of

9 the lung; the target hemoglobin was between 14 and

10 16; the incidence of thrombovascular events, TVE,

11 was 34 percent in the EPREX group versus 6 percent

12 in the placebo group. The second study, patients

13 who had cervical cancer, the target hemoglobin was

14 between 13 and 14; the incidence of TVE, 16 percent

15 in the Procrit group, versus 5 percent in the

16 placebo group. And the third study, gastric or

17 rectal carcinoma, target hemoglobin 14 or 15; the

18 incidence of TVE, 24 percent in the Procrit group

19 versus 6 percent in the placebo group.

20 Now, the next safety issue I'd like to

21 discuss is that of tumor progression. There are a

22 number of preclinical studies which have been

100

1 reviewed, but our selective take under the

2 literature is that there are EPO receptors which

3 are present on some tumor cell lines and on tumor

4 vasculature, meaning endothelial cells.

5 EPO has been reported in some studies to

6 inhibit apoptosis, stimulate angiogenesis,

7 stimulate endothelial cell growth, migration, and

8 proliferation, and reduce survival in some tumor

9 xenograft models.

10 Now, studies supporting the approval of

11 Procrit and Aranesp for the treatment of anemia in

12 cancer patients on chemotherapy were not designed

13 to assess the impact on tumor response, tumor

14 progression, or survival. So there's a big lacunae

15 in the information that we have for the

16 U.S.-registered EPO products. And, again, I'm

17 going to go through the two studies that utilized

18 non-U.S.-licensed EPO products and then two studies

19 which we have that have data that's useful for

20 looking at tumor progression in the U.S.-licensed

21 EPO products.

22 Again, just to remind you that the BEST

101

1 Study using EPREX, randomized, double-blind,

2 placebo-controlled, 939 patients with metastatic

3 breast cancer, first-line therapy, randomized to

4 receive EPREX or placebo for 12 months, therapy

5 started at less than 13.

6 The primary objective of this study was to

7 demonstrate superior survival at 12 months. The

8 target hemoglobin was between 12 and 14, and this

9 study, again, was stopped by the Data Monitoring

10 Committee based on the first four months of safety

11 data.

12 At four months, there was a twofold

13 increase in the incidence of disease progression.

14 It was 6 percent in the EPREX group and 3 percent

15 in the placebo group.

16 At four months, there was 2.5-fold

17 increase in early mortality. It was 8.7 percent in

18 the EPREX group versus 3.4 percent in the placebo

19 group.

20 In the Henke trial, again, randomized,

21 double-blind study in 351 patients with head and

22 neck cancer receiving concurrent chemotherapy,

102

1 these patients were entered if women had a

2 hemoglobin of less than 12 and men less than 13.

3 The primary objective was to demonstrate superior

4 locoregional progression-free survival. The target

5 hemoglobin was less than or equal to 14 in women or

6 less than or equal to 15 in men.

7 For locoregional progression-free survival

8 as the primary endpoint, the relative risk was 1.62

9 favoring placebo, and the lower bound or the

10 95-percent confidence interval was greater than 1,

11 with a highly significant p value.

12 For locoregional progression, again, the

13 relative risk was 1.69 favoring placebo and the

14 lower bound of the 95-percent confidence interval

15 was greater than 1, with a significant p value.

16 Study N93, the post-marketing study which

17 looked at small-cell carcinoma, this was a

18 randomized, double-blind, non-inferiority study

19 which was intended to enroll 400 patients who were

20 receiving first-line therapy.

21 The primary endpoint, again, was objective

22 response rate after three cycles of chemotherapy to

103

1 rule out a 15-percent decrement in the overall

2 response rate in the Procrit arm. No target

3 hemoglobin was determined. The Procrit dose was

4 reduced for hemoglobins greater than or equal to

5 16, and the study was terminated at 225 patients

6 out of a projected 400 for poor enrollment.

7 This study was not designed to assess the

8 impact on time to progression, and survival was a

9 secondary endpoint, and there was no formal

10 hypothesis testing.

11 The results showed that for the placebo

12 group the overall response rate was 67 percent; for

13 the Procrit group it was 72 percent. The

14 95-percent confidence interval around the observed

15 difference had a lower bound of minus 6 percent.

16 So even though this study met its intended

17 objective despite the early termination, it was

18 able to exclude a difference of greater than 15

19 percent.

20 The Aranesp Oncology Registration Study, a

21 randomized, double-blind, 320 patients with

22 non-small-cell and small-cell lung cancer all

104

1 receiving platinum chemotherapy and all of whom

2 were anemic.

3 The primary endpoint was a transfusion

4 endpoint. The Aranesp was held for hemoglobins

5 greater than 14 in women and 15 in men.

6 The median progression-free survival was

7 five months in the Aranesp group and four months in

8 the placebo group. This study, again, was not

9 designed to assess the impact on progression-free

10 survival.

11 And here are the curves. This is the

12 placebo group here. Here is the Aranesp group.

13 Here is a year, two years.

14 So, just to summarize, the data we have on

15 tumor stimulation, first the studies in which a

16 signal was detected. The BEST Study, EPREX,

17 metastatic breast cancer, at four months an

18 increased risk of deaths due to disease progression

19 being 6 percent in the EPREX group versus 3 percent

20 in the placebo group. In the Henke Study, head and

21 neck carcinoma using NeoRecormon, EPO B, the

22 relative risk for locoregoinal progression-free

105

1 survival favored placebo, 1.62.

2 The tumor stimulation studies without a

3 signal, the Procrit group, the post-marketing

4 commitment in small-cell carcinoma of the lung, the

5 overall response rate was 72 percent in the Procrit

6 group versus 67 percent in the placebo group. The

7 Aranesp Oncology Registration trial, the median

8 progression-free survival, four months for Aranesp,

9 five months for placebo.

10 And, finally, I'd like to discuss the data

11 we have concerning poorer survival in patients

12 randomized to receiving erythropoietins.

13 Again, I'll be discussed the data we have

14 on the BEST trial and the Henke trial as well as

15 the U.S.-licensed erythropoietins.

16 Just to remind you once again, the breast

17 cancer study, 939 patients with metastatic breast

18 cancer, randomized to receive EPO or--EPREX or

19 placebo for 12 months, and the primary objective of

20 this trial was to demonstrate superior survival at

21 12 months. The target hemoglobin was between 12

22 and 14, and this study was stopped by the

106

1 Independent Data Monitoring Committee based on four

2 months safety data.

3 The estimated 12-month survival was 70

4 percent in the EPO group and 76 percent in the

5 placebo group. The relative risk of death was 1.4

6 favoring the placebo group, and the lower bound of

7 the 95-percent confidence interval was greater than

8 1, with a p value of 0.12.

9 Here are the curves for the first 12

10 months, which was the primary endpoint. This is

11 the placebo group on top, and here is the EPREX

12 group.

13 In the Henke Study, again, 351 patients

14 with head and neck cancer getting radiation

15 therapy. The erythropoietin product used was

16 NeoRecormon.

17 The relative risk of death was 1.4

18 favoring placebo; the lower bound of the 95-percent

19 confidence interval was greater than 1. The median

20 overall survival was not different, but there's a

21 trend toward poorer survival in the NeoRecormon

22 group--was 605 days in the NeoRecormon group versus

107

1 928 days in the placebo group.

2 Study N93, the post-marketing commitment

3 done in patients with small-cell carcinoma of the

4 lung, again, this study was not designed to assess

5 an impact on survival. The median survival was

6 10.5 months in the Procrit group and 10.4 months in

7 the placebo group. The overall mortality rate was

8 92 percent in the Procrit group versus 88 percent

9 in the placebo group.

10 And here are the curves. The dotted line

11 is the placebo group. The sold line is the Procrit

12 group.

13 The Aranesp Oncology Registration trial,

14 320 patients with lung cancer receiving

15 platinum-containing chemotherapy. This study was

16 not designed to assess the impact on survival.

17 The median overall survival was ten months

18 in the Aranesp group and eight months in the

19 placebo group. The overall mortality rate, 14

20 percent in the Aranesp group, and 12 percent in the

21 placebo group.

22 And this is the placebo arm here, and here

108

1 is the Aranesp arm. This is one year, two years.

2 And so, just to summarize the studies we

3 had in which there was a survival signal, the BEST

4 Study, metastatic breast cancer, the 12-month

5 survival rate, the primary endpoint, poorer

6 survival in the EPREX group, 70 percent, versus 76

7 percent in the placebo group, p value of 0.12. In

8 the Henke Study using NeoRecormon, the median

9 overall survival not significant but a trend, 605

10 days for NeoRecormon versus 928 days with placebo.

11 The studies that we have without a

12 survival signal, the N93 Study, post-marketing

13 study in small-cell carcinoma of the lung, 10.5

14 months in the Procrit group versus 10.4 months in

15 the placebo group. The Aranesp Oncology

16 Registration Study, ten months in the Aranesp group

17 versus eight months in the placebo group.

18 So, to summarize, two large, multicenter

19 studies--the BEST Study and the Henke Study--which

20 were designed to show superior survival or

21 progression-free survival, instead demonstrated an

22 increased risk of thrombotic and cardiovascular

109

1 events, a shorter progression-free survival, and a

2 shorter overall survival. Both of these studies

3 used a treatment strategy to achieve hemoglobin

4 levels greater than or equal to 12.

5 The multicenter, placebo-controlled trials

6 using Procrit and Aranesp, the U.S.-licensed

7 erythropoietins, were smaller in size; they were

8 not designed to assess the impact on

9 progression-free survival or overall survival.

10 Their treatment strategy varied: Procrit was held

11 in the N93 Study for hemoglobin greater than

12 14--the label recommends 12--and in the Aranesp

13 study it was held for greater than 14 in women or

14 greater than 15 in men.

15 So, to conclude, we have these evolving

16 safety concerns. They cannot be dismissed. The

17 current dosing recommendations we feel are adequate

18 to minimize the risk of thrombotic events.

19 However, there is insufficient information

20 concerning overall survival and progression-free

21 survival for U.S.-licensed products at approved

22 doses to assess these risks. Amgen, Ortho Biotech,

110

1 and the FDA have agreed on the need for further

2 studies to investigate these safety issues.

3 Now, the FDA recommends certain elements

4 that should be components of all current and future

5 studies which will be done to investigate these

6 safety issues. First of all, there should be a

7 homogeneous primary tumor type. There should be

8 homogeneous chemotherapy or radiotherapy regimes.

9 The studies should be designed to detect clinically

10 meaningful decrements in response rate,

11 progression-free survival, and survival. There

12 should be prespecified definitions of

13 cardiovascular and thrombotic events. And there

14 should be Data Safety Monitoring Committee

15 oversight.

16 We also recommend the determination of

17 expression and ligand affinity of EPO receptor on

18 specific primary tumor types, preferably through

19 the analysis of clinical tissue specimens or

20 through pre-existing tissue repositories

21 representing common tumor types.

22 And I think that is the end of my

111

1 presentation.

2 DR. CHESON: Thank you, Dr. Luksenburg.

3 It's now time for questions from the

4 committee to either the sponsor or Dr. Luksenburg.

5 I'd like to start, while all the people are coming

6 up, with questions for Dr. Luksenburg. On your

7 various slides, Harvey, when you're talking about

8 studies with signals, you mean with negative

9 signals, since there are a number of studies with

10 positive signals, including one of the ones on your

11 slide, 98-0297, with the ten- versus eight-month

12 survival in favor of the erythropoietin compound,

13 right? So when you say with signal, you're

14 referring to negative signal in your slides.

15 DR. LUKSENBURG: Yes.

16 DR. CHESON: Okay.

17 DR. KEEGAN: I would point out that the

18 one that you're referring to as having the positive

19 signal is actually not significantly different.

20 DR. CHESON: I know, but neither are some

21 of the others.

22 Any other questions from the committee?

112

1 Any comments from the committee? Dr. Martino?

2 DR. MARTINO: I'm reminded of a quote from

3 Enrico Fermi, which goes as follows: "Before I

4 came here, I was confused on this topic. Now I'm

5 still confused, but at a somewhat higher level."

6 [Laughter.]

7 DR. MARTINO: And I'm not sure who I want

8 to sort of address this to, but whoever of you

9 thinks you have an answer, I'd appreciate it.

10 It occurs to me that looking at the tumor

11 tissue itself to see if it has receptors certainly

12 is reasonable if it's doable. Simultaneous to

13 that, it is likely that the mechanism, if there is

14 any by which tumors grow, may not be by direct

15 involvement of the tumor cell itself, but may be

16 through some other mechanism. One of those, you

17 know, is what it might do to the vascular system

18 and neovascularization.

19 Is there some way to look at that

20 parameter? Because some of us think that that may

21 be the more likely mechanism by which tumor cell

22 growth may occur, if, in fact, it does.

113

1 DR. CHESON: Dr. DeLap?

2 DR. DeLAP: Yes, I'd like to ask Dr.

3 Francis Farrell to address that question. Dr.

4 Farrell is head of our preclinical program for this

5 area.

6 DR. FARRELL: Thanks for the excellent

7 question. Francis Farrell, Johnson & Johnson.

8 We feel that your idea does have credence.

9 Although we don't feel that the receptor on tumor

10 cells is functional, there is enough preclinical

11 data to show that EPO does have an effect on

12 endothelial cell function, including some papers

13 showing that EPO binds to endothelial cells. There

14 have been some studies showing some chemotaxis with

15 EPO on endothelial cells. There's also been some

16 data that aortic ring formation can be formed.

17 The only caveat with these experiments,

18 though, are that high doses of EPO are actually

19 used to see this effect. And in one publication,

20 the dose used was actually 50 units per ml, which

21 would be very high compared to what the clinical

22 maximal serum dose a patient would get with 40 IUs

114

1 per kg dose, which is approximately two units per

2 ml.

3 So to answer your question, though, I

4 think better preclinical modeling and xenograft

5 models where you could actually look at vascular

6 density, micro-vessel formation, I think are

7 warranted, and that would be the direction that we

8 would go in.

9 DR. DeLAP: If I could ask your

10 indulgence, we also have Dr. Kimberly Blackwell

11 here who could also contribute to this point, I

12 think, as a consultant, if we have a minute.

13 DR. CHESON: Please. That would be fine.

14 DR. BLACKWELL: Hi. I'm Kim Blackwell

15 from Duke University.

16 I, like the questioner, had some interest

17 in was this tumor effect, was it endothelial cell

18 effect, and we've embarked on a number of

19 preclinical modeling, now with well over 500

20 animals that we've looked at, both in R3230, which

21 is an ER-positive mammary carcinoma line. So it's

22 as close as you can get to a rodent model to human

115

1 model. We've also looked at CT26, which is a

2 colorectal model.

3 So, very briefly, our experiments have

4 looked at tumor proliferation using Key 67, tumor

5 growth using biodimensional tumor volume. We've

6 also looked at micro-vessel density, and I think

7 the best experiment is we've actually looked at in

8 vivo angiogenesis using a dorsal window fold where

9 you can actually measure vascular development in

10 the mammary carcinoma model. And I will say that

11 we've looked at erythropoietin in close to 16

12 mammary carcinomas and have failed to see any

13 effect on tumor growth, tumor proliferation, or

14 tumor angiogenesis. Obviously the in vivo

15 angiogenesis models involve a small number, about

16 25 animals, because those are difficult experiments

17 to do.

18 We've also looked at darbepoetin using

19 similar models in both R3230 and CT26 that was

20 alluded to the Aranesp presentation, and using

21 biodimensional models in over 200 animals with

22 R3230 tumors have failed to see effect on tumor

116

1 growth, tumor proliferation, and angiogenesis

2 measured by micro-vessel density.

3 So I agree with Dr. Farrell that this

4 really needs to be studied further in in vivo tumor

5 models because the interaction between tumor

6 endothelial cells, that's really the only way to

7 study it as opposed to studying endothelial cells

8 or tumor cells separately in cell culture models.

9 DR. VIVEASH: I'd like to ask Dr. Losordo

10 to make some comments relating to this issue.

11 DR. CHESON: Please.

12 DR. LOSORDO: I'm Dr. Losordo from Tufts

13 University and St. Elizabeth's Medical Center in

14 Boston. My expertise is actually in cardiovascular

15 where we've been studying actually the stimulation

16 of angiogenesis for various ischemic disorders.

17 And that experience I think has bearing here

18 because the patient population that we study, which

19 is generally aged and, therefore, it is somewhat

20 higher risk for cancer than the general population,

21 forces us to analyze the potential risk of

22 stimulating angiogenesis in those patients in

117

1 various in vivo models. And so as a result of our

2 work primarily using VEG-F to stimulate

3 neovascularization of ischemic tissue, we've also

4 conducted studies analyzing the impact of

5 stimulating angiogenesis in that context on tumor

6 vascularization and tumor progression by implanting

7 tumors into animals and then stimulating

8 angiogenesis by exogenous administration of

9 angiogenic cytokines and have found, in fact,

10 interestingly, that the angiogenesis that's

11 stimulated is very context-dependent, meaning that

12 in the region where angiogenesis seems to be

13 deficient, for example, in the myocardium or the

14 lower extremity where we've induced ischemia, the

15 exogenous cytokine can stimulate and improve

16 perfusion of that tissue. While the tumor itself

17 regresses under the influence of chemotherapy, the

18 vascularity of the tumor does not change at all.

19 And so what we've learned in a number of

20 studies, and that would now include also studies in

21 which we're using progenitor cells from the bone

22 marrow or peripheral circulation, to also augment

118

1 neovascularization of ischemic tissue, and in those

2 instances either stimulating the release of those

3 progenitor cells from the marrow or directly

4 implanting them into ischemic tissue also does not

5 influence tumor progression.

6 So I would say that at the same time the

7 study of these things is of great interest and

8 something that we'll likely do and continue to do

9 in the context of generating safety data for

10 ongoing clinical studies. However, it also seems

11 to me that all those preclinical studies, while

12 generating interesting science, will not trump the

13 sort of clinical trial data that's being generated

14 and continuing to be generated, which I think will

15 influence patients and clinicians to a far greater

16 degree.

17 DR. CHESON: Thank you.

18 Are there any other investigators who

19 would like to comment on this particular topic?

20 [No response.]

21 DR. CHESON: Okay. We can move on then.

22 Other questions from the panel? Dr. George,

119

1 please.

2 DR. GEORGE: I have a question for Dr.

3 Luksenburg. That was a very thorough presentation,

4 but I was a little puzzled by the way it was

5 presented with respect to studies that showed a

6 signal, those that didn't show a signal, and I was

7 left trying to do my own mental meta-analysis of

8 things to try to get some bottom line there.

9 Did you do such things? Or can you help

10 us out in that way?

11 DR. LUKSENBURG: No, we didn't. We

12 obviously reviewed data which had come in over a

13 number of years, and much of this data was from

14 registration studies which were a few years old,

15 and we looked, as did the sponsors, for evidence

16 of--we looked at the data that was there for

17 overall survival and progression-free survival.

18 But since the studies were not designed to look at

19 that, we, you know, just--we took the data as it

20 was. We did not do any meta-analyses.

21 In general, our stance is that the studies

22 that are valuable are studies--except for

120

1 thrombotic/cardiovascular disease, the studies that

2 will provide the best quality data for overall

3 survival, progression-free survival, time to tumor

4 progression, are those with homogeneous tumor

5 populations. And it's really difficult to do

6 meta-analyses with variegated tumor populations.

7 DR. CHESON: Dr. Keegan, did you want to

8 make a comment?

9 DR. KEEGAN: Yes. Actually, that was one

10 of our concerns with several of the meta-analyses

11 presented, that it's trying to put the data in

12 there in a way that--and take studies that weren't

13 intended to look at these events and provide

14 information. And I think the quality of many of

15 the studies included in the meta-analysis are not

16 the same in terms of what information they can give

17 you on progression-free survival or on overall

18 survival simply because of the heterogeneity and

19 the lack of control. So that, you know, I think if

20 we were to choose to select the studies, we would

21 try and find studies that were actually designed to

22 look at these endpoints and have the qualities that

121

1 we are recommending further.

2 DR. GEORGE: Just a quick follow-up. I

3 certainly agree with respect to some of those

4 endpoints, but survival should be a clear one.

5 DR. KEEGAN: I think when you look at some

6 of those studies--and many of them are fairly small

7 studies, and they enrolled any patient with any

8 tumor at any stage in their treatment. It might

9 tell us something about transfusion rates. That's

10 what they were intended to do. But they weren't

11 really intended to give us a good comparison of

12 impact on tumors. These studies were really done

13 in a manner not well designed to assess impact on

14 tumor, just given all the incredible variables so

15 much more important in terms of impact on survival

16 and time to progression.

17 Presumably, if there had been thousands of

18 patients, all of those variables would probably

19 have been evened out. But most of the studies, as

20 you look at them, are not particularly large, with

21 the exception of the ones that we tried to

22 highlight.

122

1 DR. CHESON: Are you satisfied with that

2 answer, Dr. George?

3 DR. GEORGE: Yes.

4 DR. CHESON: Okay. Ms. Mayer?

5 MS. MAYER: As I understand it, FDA is

6 coming to ODAC not to ask us to assess if there is

7 any level of risk associated with these products,

8 but given that there may be a level of risk, to

9 look at what kinds of clinical trials need to be

10 done. And I'm wondering since the data doesn't

11 seem to be conclusive, since there are different

12 perspectives, if it's useful for us to continue to

13 try to assess what we know already from the trials.

14 It's just a question, I guess a clarification of

15 what our task is.

16 DR. KEEGAN: I think you're right in

17 saying that if we thought we knew the answer, we

18 wouldn't be asking you to reinterpret the data for

19 us. I think we're saying that we don't think it's

20 been definitively assessed and could we seek some

21 guidance on how to really address this question.

22 DR. CHESON: And the way I see it is we're

123

1 being asked to do one of several things: one,

2 decide if the data are of sufficient concern; two,

3 if they are of sufficient concern, are additional

4 studies warranted; and, three, if additional

5 studies are warranted, are those the studies that

6 are already ongoing, as clearly elucidated by Dr.

7 Parkinson and his colleagues.

8 Dr. Bauer, please?

9 DR. BAUER: Yes, maybe I could just follow

10 up on that point, because some of the studies we've

11 heard presented clearly are driven by safety

12 concerns in terms of showing safety, but, you know,

13 as I understand the studies that are being

14 proposed, there's really a desire to show improved

15 survival. And I guess we haven't heard a great

16 deal about the rationale really in terms of showing

17 survival. I think we know about effects on

18 radiotherapy and tumor oxygenation. We also know

19 some of the high hematocrits targeted there clearly

20 are detrimental and a desire in all the studies

21 going forward to keep the hematocrit below certain

22 specified levels. I guess I would like to hear

124

1 more about really the rationale for really at this

2 point believing that there really will be improved

3 progression-free survival with the use of some of

4 these erythropoietic stimulating agents, or

5 survival overall, especially given the clear

6 detrimental effect, albeit it small, in terms of

7 thrombosis.

8 DR. CHESON: I think that most of these

9 are probably non-inferiority trials, if I'm not

10 mistaken. They just don't want to show that there

11 is a negative effect.

12 Dr. Parkinson, since you were reviewing

13 all those articles, would you like to comment on

14 that, please?

15 DR. PARKINSON: Dr. Bauer, you're correct

16 in that we did not spend a lot of time talking

17 about the rationales. The time was short.

18 Sponsors were many.

19 There is a wealth of preclinical evidence

20 which I think there are a number of people who

21 could discuss in more detail. There is a

22 significant amount of clinical evidence. I

125

1 referred to the Cochran meta-analysis, independent

2 analysis conducted, as you're aware, by the Cochran

3 group, which was considered to be suggestive

4 enough--not definitive, but suggestive enough to

5 warrant further trials. I mention that because I

6 think it's important. It's dissociated from any

7 product-related.

8 We've shown you and you've seen from other

9 sponsors quite interesting suggestions of patient

10 benefit in a number of defined settings, both of

11 radiotherapy and chemotherapy. Additionally, the

12 trials that I described which were not

13 Amgen-sponsored were initiated by independent

14 investigators based on their own independent

15 assessment of preclinical and clinical data

16 designed to test particular hypotheses, which are

17 actually superiority hypotheses. These were not

18 trials designed to look for negative survival

19 signals with erythropoietins. These were trials

20 designed to look for benefit based on--we won't

21 give you our assessment of the literature--their

22 assessment of the literature and what they believed

126

1 were important therapeutic questions to ask.

2 You know, we can go into as much

3 detail--there are actually investigators here from

4 each of those particular clinical groups. There

5 are preclinical investigators here from at least

6 two companies. There's a wealth of evidence to

7 support this kind of investigation. What we see

8 here today are two signals from two trials which

9 you've heard described and analyzed in great

10 detail. You can make your own judgment as to what

11 the value of those signals is.

12 DR. CHESON: Dr. DeLap?

13 DR. DeLAP: Since we've also

14 done--obviously we've done a number of trials in

15 this area. We clearly have a rationale for

16 proceeding in this area. I'd like Dr. Adrian

17 Thomas to address our thoughts in this area.

18 DR. THOMAS: Good morning and thank you.

19 Adrian Thomas, Vice President, Drug Safety, Johnson

20 & Johnson.

21 I think our view and position is entirely

22 consistent with Dr. Parkinson. It's entirely

127

1 reasonable to look for survival benefits with these

2 products, and we indeed embarked on INT-76 as a

3 result of results from INT-10, in which we

4 demonstrated as a secondary endpoint of survival

5 advantage and, more particularly, when we looked at

6 the subgroup of patients with breast cancer, they

7 seemed to benefit the most.

8 So I think the rationale for pursuing a

9 survival advantage is there. It's clearly in the

10 context of what the risks might be in terms of,

11 from our perspective, thrombotic vascular events

12 and the appropriate targeting of hemoglobin levels.

13 DR. CHESON: Dr. Weiss?

14 DR. WEISS: Just to, I guess, reiterate

15 what has been said, there is a wealth of data,

16 there's a lot of information, lots of variability

17 in terms of the quality of the different studies,

18 and I know it's a difficult question to try to sort

19 through it all. I think we all agree, though, that

20 there's some provocative and interesting

21 information that might suggest some benefits other

22 than just minimizing or avoiding transfusions with

128

1 erythropoietin products, and I think we'd all like

2 to be able to document that and have that well

3 established. I think there's maybe a belief system

4 that erythropoietins are benign, with the exception

5 perhaps of some slight increased risk in thrombotic

6 events.

7 So I think the question here is--and we've

8 certainly had lots of discussions with both Amgen

9 and J&J. I think we all agree that there is room

10 for further studies and further exploration, and

11 the best way to try to show a survival benefit or

12 disprove some type of disadvantage is to do it in

13 the context of very good, well-designed clinical

14 trials. And I really think that's really the focus

15 of this particular meeting.

16 DR. CHESON: Dr. Parkinson, pleaseS?

17 DR. PARKINSON: Just to say we totally

18 agree and that, although I indicated that these

19 trials were designed to look for superiority in

20 terms of the therapeutic beneficial effects of

21 Aranesp in our case, I just had a little note from

22 my statistical colleagues that, you know, just

129

1 because they're designed to show superiority

2 doesn't mean that they can't have a huge value in

3 looking for negative survival signals. So to keep

4 myself statistically in good company, I wanted to

5 point that out.

6 DR. CHESON: Thank you.

7 Dr. Martino?

8 DR. MARTINO: I just need to be sure I'm

9 clear before I say anything semi-intelligent here.

10 I need to be sure that I'm understanding the

11 following: It occurs to me that there really are

12 two trials that have shown tumor-specific bad

13 qualities, and those trials share at least one

14 thing in common, which is that they've aimed for a

15 hemoglobin above and beyond what most of us

16 considered usual and appropriate and normal and the

17 aim, at least within this country.

18 And so as we think about what questions we

19 need to answer, it occurs to me that that's a key

20 point as to are we trying to show that something

21 bad happens, are we trying to show that something

22 good happens, but the question has to be framed

130

1 within those two hemoglobin objectives, as I

2 understand it.

3 Am I clear in my thinking on the evidence

4 that exists?

5 DR. KEEGAN: I think you express very well

6 the same impression we have of the two trials that

7 showed a negative effect and the lack of

8 information we have in the other areas of

9 definitive information on the safety. If the

10 companies want to show that there's a survival

11 advantage associated with their products, we have

12 no problem with that. Our issue is really that we

13 would like for them to definitively address whether

14 or not there could be an adverse effect.

15 DR. CHESON: Dr. Brawley? Oh, excuse me.

16 Dr. Brawley will defer for the moment.

17 DR. M. GEORGE: I just wanted to follow up

18 on the previous question and reiterate that the

19 clinical trial we are proposing is to assess the

20 activity on one single tumor trial using epoetin

21 alfa within the label, so in the anemic patient

22 population; and, lastly, that we're proposing a

131

1 non-inferiority trial, which explains why the trial

2 is so large.

3 DR. CHESON: Thank you.

4 Dr. Brawley--or Dr. Parkinson first, and

5 then Dr. Brawley. Sorry, Otis.

6 DR. PARKINSON: Just relevant to that is

7 that most of the trial results that I presented

8 here today were done with clinical trials during

9 the development of Aranesp prior to development of

10 the actual label. Our current recommendations are

11 consistent, as I tried to make clear in the talk,

12 with the evidence-based guidelines, the

13 recommendations from ASH, from ASCO guidelines, and

14 from the NCCM guidelines and reflect current

15 practice.

16 Investigation of anything beyond that is a

17 matter for clinical trials and careful clinical

18 monitoring with carefully designed scientific

19 hypotheses. We would completely support that.

20 DR. CHESON: Thank you. And now, Dr.

21 Brawley?

22 DR. BRAWLEY: Actually, this is sort of in

132

1 follow-up to what Dr. Parkinson just said. My

2 understanding is that the current indication for

3 these drugs is for supportive anemia that is due to

4 either renal failure or due to chemotherapy. There

5 is no claim in the package insert that these drugs

6 improve survival in any disease. Correct?

7 DR. KEEGAN: That's correct.

8 DR. CHESON: Dr. George?

9 DR. GEORGE: I have a question. It's in

10 the Procrit area, I guess either for Dr. Bowers or

11 Dr. George--a different Dr. George. That has to do

12 with the endpoint chosen in the new study in that

13 you chose progression-free survival even though in

14 the study on which this was based, I guess, the

15 indication was--the problem seemed to be a

16 decrement in overall survival at 12 months and no

17 indication of any progression-free survival

18 problems.

19 DR. M. GEORGE: Overall survival is going

20 be a secondary endpoint in the trial, and the

21 reason why we chose progression-free survival is as

22 follows: First, progression-free survival is the

133

1 best way to assess if there's any effect--if there

2 is effect, if any, on the tumor. Second, we are

3 talking about a placebo-controlled trial so there

4 might be significant crossover if the tumor

5 progresses, if things change. So that's one reason

6 after the crossover. The second reason, which may

7 even more obscure the survival endpoint, is if the

8 patients fail the first-line chemotherapy, the

9 patients are going to cross over to another

10 regimen, and that may also obscure the survival

11 endpoint.

12 So we thought that carefully designed

13 progression-free survival endpoint--and, again, I

14 didn't go through the detail on how we are going to

15 assess it, how meticulously it's going to be

16 assessed, review by a blinded independent panel

17 will give us better enterprises. And I'm just

18 reminded that we will have 80-percent power for a

19 non-inferiority trial in survival.

20 DR. CHESON: Dr. Feldman?

21 DR. FELDMAN: I'm just wondering, are

22 there any data available or are there any trials

134

1 planned to address the issue of the use of

2 erythropoietin products in cancer patients not

3 receiving additional treatment?

4 DR. M. GEORGE: I'm going to also--on

5 behalf of Johnson & Johnson, yes, we currently have

6 an ongoing trial comparing placebo to Procrit in

7 patients who have cancer and are anemic and not

8 receiving chemotherapy. The study is ongoing.

9 Survival is going to be assessed in that trial as

10 well as progression-related endpoints.

11 DR. CHESON: Dr. Parkinson, you should

12 just probably stand there.

13 [Laughter.]

14 DR. PARKINSON: Thank you, Bruce. Yes, we

15 have ongoing trials in anemia of cancer patients

16 not actively receiving chemotherapy. Again, very

17 careful monitoring, data monitoring committees that

18 are independent, all of that.

19 I'd like to point out also the particular

20 design of the AGO Study that I mentioned earlier.

21 Those are neoadjuvant patients. They are biopsied,

22 therefore, prior to initiating chemotherapy, and

135

1 Aranesp therapy or not. And a major endpoint of

2 the trial is actually pathological endpoint at

3 surgery. So, in addition to the status of the

4 tumor, there will be the opportunity to examine

5 carefully for any evidence whatsoever of

6 angiogenesis differentials between Aranesp and not.

7 I think it's a very powerful design.

8 Investigators are very sophisticated and very aware

9 of the importance of the biological results in

10 addition to the clinical results in this trial.

11 DR. CHESON: Don't go away. What are the

12 endpoints on the previous trials that you mentioned

13 in the non-treated patients?

14 DR. PARKINSON: In the anemia of cancer

15 trials, endpoints are predominantly anemia related,

16 but follow-up is long term. Those trials were

17 designed prior to any of these discussions, and--

18 DR. CHESON: Is it possible to update the

19 statistics on those to look for survival?

20 DR. PARKINSON: Absolutely. I think that

21 there are a number of things that one may want to

22 do at the end of the committee's deliberations and

136

1 recommendations. Absolutely.

2 DR. CHESON: Thank you.

3 Dr. Carpenter?

4 DR. CARPENTER: I just wanted to comment

5 on the survival endpoint in the previous study.

6 It's very hard to show a survival difference in

7 advanced breast cancer with any treatment. Even

8 though many people think certain things may confer

9 survival benefit, it's hard to do a study large

10 enough and pure enough to find that because of the

11 large number of chemotherapy, hormonal, other, and

12 now biological agents that are available. So I

13 appreciate the company's diligence in trying to

14 sort that out, but I think their use of

15 disease-free of progression-free survival as a

16 primary endpoint is going to be a lot easier to

17 interpret and is going to be available a lot sooner

18 than trying to sort out what's going to be a

19 complicated bunch of information later.

20 DR. CHESON: I think what Dr. George is

21 getting to is some consistency among trials with

22 enterprises which, looking from the various trials,

137

1 there was some lack thereof.

2 DR. CARPENTER: Yes, but since it's going

3 to be done in breast cancer, that's going to be a

4 particularly hard thing to do.

5 DR. CHESON: Understood.

6 DR. CARPENTER: Where if it were done in

7 some other tumor where there were many fewer

8 options for treatment later--and this is going to

9 be done with first-line chemotherapy. It's going

10 to be a complex situation that might not be there

11 in other tumors.

12 DR. CHESON: Thank you.

13 Dr. Redman?

14 DR. REDMAN: Just to follow up on the

15 issue of the tumor-specific nature of the trials in

16 which there were negative signals, do the sponsors

17 have any plans to evaluate these agents in

18 non-solid tumors, in hematologic malignancies,

19 other than erythroleukemia?

20 DR. CHESON: Dr. Parkinson?

21 DR. PARKINSON: Yes, the GELA trial, a

22 very large trial, 600 patients with aggressive

138

1 non-Hodgkin's lymphomas by a well-respected, very

2 accomplished, cooperative group in France, Belgium,

3 and Switzerland.

4 DR. CHESON: Dr. Grillo-Lopez?

5 DR. GRILLO-LOPEZ: I wanted to add to what

6 Dr. Carpenter said, that an additional set of

7 confounding factors would be that in any randomized

8 trial where you have epoetin in one arm and not on

9 the other arm, you are controlling for that during

10 the course of the study. However, at some point

11 when those patients have a relapse and go on to

12 other chemotherapy, are you going to still require

13 that they do not receive epoetin ever during the

14 course of the remainder of their survival? I think

15 that would be very difficult to require, very

16 difficult to control and enforce. So at some point

17 on both arms, patients will be getting some of

18 these products, and I think that that's an

19 additional reason why overall survival is probably

20 not an appropriate endpoint for these studies.

21 DR. CHESON: Good point.

22 Dr. Martino, did you have a question?

139

1 DR. MARTINO: A question to anyone from

2 industry. Are there any known or presumed clinical

3 parameters for which a hemoglobin above 12 is known

4 or felt to be of value?

5 DR. DeLAP: Dr. Adrian Thomas of our Drug

6 Safety Group will address that question for us.

7 DR. THOMAS: Adrian Thomas, Vice

8 President, Drug Safety Johnson & Johnson. I think

9 in addressing that question, the benefits that have

10 been seen in the chronic renal failure population

11 with erythropoietin therapy have generally been

12 seen at levels of hemoglobin less than or equal to

13 12. And one can postulate that by increasing the

14 hemoglobin level by whatever mechanism, by having

15 an effect on--a positive effect on the tumor, but I

16 think we've seen indications of positive effects on

17 tumor outcomes in some of our earlier studies at

18 hemoglobin levels more typically within the anemic

19 range that we would treat patients in clinical

20 practice. I don't think that we need to pursue

21 high-target hemoglobins to look for aggressive

22 outcomes.

140

1 I also want to make a point around the

2 concept of tumor-specific outcomes. I think what

3 we've seen today is three very large meta-analyses

4 of lots of tumors, and that, in fact, I'd challenge

5 the word "tumor-specific." What we have, in fact,

6 seen in terms of a biological signal is something

7 that isn't consistent with tumors. We've seen no

8 effect on tumor response. We've seen no effect on

9 tumor response. We've seen no effect on tumor

10 progression. We have seen no effect in our studies

11 of new target lesions. What we've seen is a

12 consistent signal both within oncology and from the

13 Besarab study of fatal outcomes linked to

14 high-target hemoglobins --[microphone off]-- need

15 to be considered as a pharmacologically plausible

16 mechanism.

17 DR. CHESON: Dr. Demetri?

18 DR. DEMETRI: I'd like to make one comment

19 as a clinician who has done some of the studies on

20 also patient-reported quality of life where

21 patients have given data to support the benefits of

22 how they feel in terms of better hemoglobin levels

141

1 beyond 12, interestingly, as well as some of the

2 preclinical evidence that might support better

3 oxygenation at higher levels. Now, the latter is

4 more theoretical in terms of clinical outcomes, but

5 that was part of the rationale for the beyond

6 correction of anemia studies. And I think that is

7 one key element to those investigational

8 strategies. But there are data in the other

9 studies for supportive care for benefits at higher

10 levels.

11 DR. MARTINO: So are you saying, then,

12 that we know from patient reports that self-reports

13 of quality of life is somewhat better when a

14 hemoglobin above 12 is maintained? I just want to

15 be sure I'm understanding you.

16 DR. DEMETRI: I would say that is correct

17 from the non-randomized large-scale studies that

18 I've conducted, my colleague Dr. Glaspy has

19 conducted, as well as others, yes.

20 DR. VIVEASH: Yes, I'd just like to

21 comment that there's associative data in a number

22 of disease settings, not necessarily in oncology,

142

1 that suggest higher hemoglobins are associated with

2 better outcomes. And, in fact, we are going to be

3 doing some work in patients with chronic renal

4 insufficiency. I'd like to ask Dr. Pfeffer to just

5 talk briefly about that program.

6 DR. CHESON: Briefly.

7 DR. PFEFFER: Thank you. So outside of

8 oncology, there are some indications that there is

9 real equipoise here and we need to do more

10 research. And as a matter of fact, with the

11 burgeoning problems with diabetes and renal

12 insufficiency prior to dialysis, anemia is becoming

13 a big factor, and the epidemiology suggests that

14 this is a comorbidity and co-risk. So we're

15 undertaking, if I could just have one slide just to

16 show you the magnitude of the effort--no, that's

17 not--we're undertaking a 4,000-patient study of

18 people who are anemic, have diabetes, and who have

19 renal insufficiency, not in dialysis,. with very

20 hard cardiovasculars to determine if we can improve

21 their outcome. Obviously, with a trial of 4,000

22 patients and over two years of follow-up, we'll

143

1 have a great deal of patient experience,

2 randomizing to a strategy to maintain the

3 hemoglobin to 13 or leave it where it is under 11.

4 So that's a strategy that's going forward, so more

5 information is going to be forthcoming, and there

6 still is equipoise in the cardiovascular community.

7 DR. CHESON: Odd name for a trial when

8 only half the patients actually get treated.

9 Ms. Mayer?

10 MS. MAYER: Just a comment on the form of

11 reference to patient-reported responses to having

12 their hemoglobin level at a higher level. I wonder

13 how those patients might respond if they knew that

14 by doing so they might be increasing their risk for

15 thrombotic events. That might color patient

16 perception.

17 DR. CHESON: Thus the need for clear and

18 accurate informed consent.

19 MS. MAYER: Absolutely. I would be

20 interested to know if in the informed consent in

21 that trial that was an issue that was explained to

22 patients, or if those findings of a higher risk for

144

1 those adverse events actually came from that trial.

2 DR. CHESON: Well, those are also in the

3 package insert, but I would assume that they are in

4 the informed consent.

5 One last comment from Dr. Keegan.

6 DR. KEEGAN: Actually, I did want to put

7 into context the studies that Dr. Glaspy referred

8 on quality of life were uncontrolled studies. So

9 there was no way to put information in context.

10 The second is that he referred to patients

11 who achieve hemoglobins above 12, and we would look

12 at that as something of a responder analysis. You

13 know, patients who do well do well. I think that

14 one should consider those single-arm studies with a

15 great deal of skepticism and caution given the

16 amount of missing information that's generally not

17 there from patients who were not doing well.

18 DR. CHESON: We will have additional time

19 for discussion during the discussion period. Right

20 now why don't we take a ten-minute break and

21 reconvene here at about 12 minutes of.

22 [Recess.]

145

1 DR. CHESON: We are ready to get started.

2 Now you can sit down, Dr. Parkinson.

3 [Laughter.]

4 DR. CHESON: The next part of this session

5 involves the open public hearing. No one has

6 approached us prior to this meeting to express an

7 interest in presenting. At this point in time, is

8 there anybody who has shown up for this purpose and

9 has not talked to us?

10 [No response.]

11 DR. CHESON: If not, then we will move

12 into the committee discussion. I just want to

13 reinforce, for those of you who are new to this,

14 that this is an Advisory Committee to the Food and

15 Drug Administration. We clearly do not work for

16 them, but hopefully we work well with them.

17 We have been given a number of questions,

18 which are on a piece of paper that most of you got.

19 Dr. Keegan has modified this to a minor extent.

20 Dr. Keegan, would you just like to mention what

21 your modifications are?

22 DR. KEEGAN: It was just a clarification

146

1 of the concern that arose about placebo-controlled

2 trials, and I think the earlier wording might have

3 led the committee and others to believe that it was

4 the companies who felt that it was not feasible.

5 But our understanding is that it isn't the

6 companies but physician investigators who have

7 raised feasibility concerns. So we just reworded

8 that.

9 DR. CHESON: Clearly, from what we heard

10 from the companies earlier, they feel it is

11 feasible.

12 We have a series of questions before us,

13 some of which are more compelling and some of which

14 are less compelling, if the previous question is a

15 negative one.

16 We've already talked about the possible

17 reluctance of physicians to conduct and enroll

18 patients in placebo-controlled trials. Do we have

19 some sentiment around the table here as to whether

20 this is a possibility? Dr. Martino?

21 DR. MARTINO: I was just kind of

22 pondering. I wasn't ready to answer. But since

147

1 you've asked, I think there will be physicians for

2 whom it will be an issue. You watch a hemoglobin

3 going down, and you worry about--and I think to

4 some degree physicians will be able to tell which

5 patients are in active therapy and which are not.

6 So a placebo in this context is a relative placebo.

7 It is not a placebo in the sense that there are no

8 clues of who is getting what. You can't always

9 anticipate, you know, what that hemoglobin going

10 down or up is from. But, you know, to a reasonable

11 degree I think there will be at least the

12 assumption that one knows what one's patient is on.

13 That being said, do I think that there

14 will be physicians who will be willing and

15 unwilling to enroll in a placebo-controlled trial

16 asking these kinds of questions? I think there

17 will be physicians in both of those camps, but I

18 think there will be enough who will recognize the

19 importance of the question, assuming the question

20 is properly framed. And I'm not entirely

21 comfortable that I know that the questions have

22 been properly framed in the studies proposed. So I

148

1 have more of an issue with are the studies asking

2 the questions that I consider of importance. I'm

3 reasonably comfortable that there will be

4 physicians who will randomize.

5 DR. CHESON: Dr. Taylor?

6 DR. TAYLOR: I would agree. I think that

7 they're going to be looking at the other risks that

8 we're trying to elucidate, and they're going to be

9 willing to take those. And I think some patients

10 will be more willing to take blood than Aranesp.

11 DR. CHESON: Dr. George?

12 DR. GEORGE: I don't have anything to add

13 on whether it's possible or not, except to state

14 that it would be desirable to do this if it is

15 practical.

16 DR. CHESON: Anybody else have any

17 comments on this? Dr. DeLap?

18 DR. DeLAP: I think our major concern is

19 just we want to get studies done, and particularly

20 when we're looking at a 2,000-patient study, even

21 relative difficulties in accruing patients can be

22 an issue, particularly in a disease like breast

149

1 cancer where the therapeutic regimens can change

2 over time.

3 The studies need to be done; they need to

4 be done efficiently and relatively quickly. And so

5 I think we do still have some concerns in this

6 area, although I would agree it's not a complete

7 bar, but it is certainly an issue that has to be

8 considered in the design of these studies.

9 Could I just ask Dr. Cohen to speak on

10 this briefly?

11 DR. COHEN: I just want to speak as a

12 clinical investigator. I would not underestimate

13 the challenges of conducting placebo-controlled

14 trials. I would conduct them in Europe and the

15 United States because, to echo one of the committee

16 members' comments, there will be investigators

17 falling into both camps.

18 Also, the way that the question is framed

19 is absolutely critical. In order to get patients

20 to agree to be enrolled in these trials, I think we

21 have to postulate that there is a survival benefit

22 in using these drugs. Of course, the trials are

150

1 also powered to exclude a meaningful decrement in

2 survival.

3 And I think the trials will need to be

4 conducted by very mature clinical investigators

5 with meticulously written informed consents to

6 portray the issues accurately to the patients. But

7 they are feasible. There are unmistakable

8 challenges that will require a very prolonged and

9 multinational approach in order to get the job

10 done.

11 DR. CHESON: Dr. Martino?

12 DR. MARTINO: Patients are already aware

13 of the two trials that have brought this issue

14 forward, and those of us that practice oncology

15 have lots of patients who have called and written.

16 And so it's not entirely an issue of placebo. It

17 is also an issue of patients knowing this newer

18 data who may not want to be randomized to the

19 treatment portion of this. So, you know, it isn't

20 exclusively a placebo issue in my mind.

21 DR. COHEN: And I think in that regard we

22 need to explain carefully to the patients what it

151

1 is we're trying to do. We are trying to treat

2 anemia. We are not going into correction beyond

3 anemia. If we explain the issues carefully to the

4 patients, they will be less afraid and more willing

5 to participate.

6 DR. MARTINO: I need to pursue this a

7 little bit more if you'll allow me. Probably the

8 thing of greatest concern to me right now is it

9 appears to me that perhaps the real issue at gut

10 here is, in fact, the level of hemoglobin. That to

11 me is a reasonable explanation to the discrepancy

12 in the data. And with the exception of the

13 diabetic trial that was presented a few moments

14 ago, I have yet to see--perhaps I've missed it, but

15 I have yet to see a trial in cancer that addresses

16 what I think may be the issue.

17 DR. CHESON: From the sponsors, is there

18 such a trial that Dr. Martino is looking for that

19 has been--

20 DR. DeLAP: Let me clarify the issues that

21 you're trying to address here, whether there is an

22 expectation of a benefit in the anemic population

152

1 or--

2 DR. MARTINO: No. My question is: The

3 two trials that have shown a tumor-specific

4 negative effect, okay? Both of them dealt with

5 aiming for a hemoglobin above, you know, the usual

6 12 or so, okay? That may be exactly the issue.

7 That may be exactly the issue. And if that is

8 exactly the issue, then the proposed trials aren't

9 addressing that. And you could be doing all kinds

10 of things and never getting at the issue.

11 DR. DeLAP: You're correct in that there

12 is the one trial that addresses the target

13 hemoglobin level prospectively. That was the

14 Normal Hematocrit Trial, which was in renal

15 patients, not in cancer patients. We do not have a

16 trial randomizing patients to different target

17 hemoglobin levels.

18 We do have--actually, if I could just call

19 up our slide DE3, I think it is. We do have one

20 experience in our clinical trials program that I

21 think speaks to this, which is one of trials where

22 we have the biggest issue with these TVE events and

153

1 possibly some survival impact is this small-cell

2 lung cancer trial which was terminated prematurely,

3 again, for TVEs.

4 Now, the interesting thing about this

5 trial is that, as originally designed, the patients

6 were treated to a target hemoglobin of 14 to 16.

7 In October 2002, for reasons unrelated to looking

8 at any data but just that that seemed to be too

9 high of a target, the target was modified to 12 to

10 14. Patients were randomized both before and after

11 the amendment. Patients were treated for similar

12 durations with erythropoietin therapy both before

13 and after the amendment. And yet you can see that

14 in the pre-amendment group, although the numbers

15 are small, in the pre-amendment group 42 percent of

16 the patients in the erythropoietin alfa arm had

17 these TVE events. And in the post-amendment group

18 treating to the lower target--which is still a

19 higher target than we might like to use now, but,

20 clearly, at the post-amendment point, it was 10.5

21 percent. So that's suggestive evidence, at least

22 in a cancer population, that we're following the

154

1 right path by treating to a lower hemoglobin

2 target.

3 DR. CHESON: Do you have any information

4 on what the median hemoglobin was that was attained

5 in the two arms--not in the two arms but in the two

6 patient populations?

7 DR. DeLAP: Let me refer that to Dr.

8 Adrian Thomas, who has more details about that

9 study.

10 DR. THOMAS: I think this is certainly an

11 interesting question. What we observed

12 pre-amendment is, in fact, that the hemoglobins in

13 the patients who developed TVEs were around the 15

14 level, and following the amendment the hemoglobins

15 were around the 12 to 13 level. And so I think we

16 can see, although not pre-defined, we can see some

17 empiric evidence of the effect of changing the

18 target hemoglobin level.

19 DR. CHESON: Dr. Parkinson?

20 DR. PARKINSON: Just a comment that, as I

21 indicated earlier, the clinical trial results that

22 I demonstrated were, in fact, conducted with trials

155

1 that took patients to target hemoglobins of 13.

2 That doesn't reflect the current practice, but

3 that's, in fact, what was the operative practice

4 during the conduct of those trials. There are

5 other reasons to go higher, which we could get

6 into. Professor Overgaard is here, and he spoke to

7 me at the break about the rationales for doing

8 that. But I think that's not where you're coming

9 from, Dr. Martino. Is that correct?

10 DR. MARTINO: I'm trying to figure out

11 what the real question is that we want to answer

12 here, and I guess one of the questions is: If you

13 aim for 12, or thereabouts, is there an effect on

14 tumor biology, survival, whatever endpoint you want

15 to look at? And that's a very worthwhile question.

16 DR. PARKINSON: Okay. In that case, could

17 I call upon Professor Overgaard, who is here from

18 the Danish Head and Neck Cancer Study Group

19 someplace? He told me he--oh, there he is. How

20 could I miss you?

21 DR. OVERGAARD: My name is Jens Overgaard.

22 I come from Denmark. The reason for this was that

156

1 earlier, before the break, there was a question as

2 to what are the rationales for having a survival

3 benefit of these trials here, and it was said that

4 there were plenty. But it is fairly simple,

5 basically, because this is a matter of oxygen

6 delivery. And we must assume that what we really

7 would like to have is more oxygen brought forward

8 by more hemoglobin into the tumor. And that oxygen

9 should do something in benefit for outcome. That

10 means it should interact with the treatment, which

11 will be better in one way or another if there is

12 more oxygen delivered. And what you said, people

13 know that might be very well the case because

14 hypoxia is a key issue in the response to

15 radiotherapy.

16 Now, if that treatment will be better

17 issued in turn also influence the survival in the

18 (?) , these are the fundamental simplicity of the

19 design of the rationale. In such studies and the

20 one we are doing in radiotherapy, it is a matter of

21 lifting oxygen delivery from one level to a higher

22 level. It is not a matter of lifting to some

157

1 specific level of 12 or whatever. It's just a

2 matter of having a differential. And the only

3 thing that puts a limit on that differential is the

4 ceiling. So what we have to discuss here is more

5 where is the ceiling, where is it halfway up to the

6 ceiling, because we need to have the room for

7 excess oxygen delivery if we have to do survival

8 benefit trials.

9 DR. CHESON: Dr. Keegan?

10 DR. KEEGAN: Dr. Martino, is your

11 question--and I think it's our question, too--that

12 if studies are done using a higher hemoglobin,

13 permitting or encouraging a higher hemoglobin

14 target and they show that there is, in fact, a

15 detrimental effect, we will have no information on

16 whether or not at the approved dose and for the

17 intended and licensed indication, which is

18 avoidance of blood transfusion, whether or not

19 these are safe? And so by not starting first at

20 the approved dose and schedule and in the currently

21 indicated population we may be actually prolonging

22 our time to getting an answer?

158

1 DR. MARTINO: In the ideal world, what I'd

2 like to see is both of these hemoglobin dose levels

3 addressed and in each of those, the same question

4 asked: Is it good? Is it bad?

5 Now, that's what I'd like in the ideal

6 world. I do recognize that there's another issue

7 here, which is a trade-off in the sense that it

8 already is fairly apparent that there are more

9 complications as you increase the level. So you

10 get to this issue of, you know, relative good and

11 relative bad. But it really is each of those

12 levels which are of concern to me.

13 DR. VIVEASH: Could I just comment? I

14 decided I'd give Dr. Parkinson a break for a

15 moment.

16 He presented a number of studies,

17 forward-looking studies, some of which are ongoing.

18 The vast majority of those are actually using the

19 current label target hemoglobin so we'll address

20 the one question. The DAHANCA Study actually goes

21 to a higher hemoglobin, and we feel both approaches

22 are valid as long the studies are appropriately

159

1 conducted with the appropriate endpoints and

2 appropriate safety monitoring.

3 DR. CHESON: Thank you.

4 Dr. Redman?

5 DR. REDMAN: This is regarding

6 randomization, especially to Amgen and the European

7 studies. Your target hemoglobin is somewhere

8 between 13 and 14, it sounds, in most of the

9 studies--the small-cell, the breast. I don't know

10 what the policy is or what the benefit is in

11 Europe. In the United States, the blood bank will

12 not release blood for a hemoglobin of 13 unless the

13 patient is actively bleeding. So how are you

14 controlling for the transfusions in those?

15 DR. PARKINSON: These are not

16 hemoglobin-controlled trials. These are trials of

17 Aranesp to specified levels versus transfusions as

18 used in regular clinical medicine. That will

19 differ in different settings.

20 DR. CHESON: Are there any more comments?

21 Ms. Mayer?

22 MS. MAYER: I'd like to return to the

160

1 issue of accrual to randomized trials. I have some

2 concerns that patients may have difficulty

3 submitting themselves to randomization, whether or

4 not they're randomized to either arm, actually,

5 because I think patients tend to come to their own

6 conclusions in situations where it's really unclear

7 and where there are complex risk/benefit ratios

8 like the ones we're discussing. And in those

9 situations, I think patients like to have choice.

10 You know, given that this is on the market, some

11 may choose to have transfusions and to avoid EPO

12 until these issues are resolved, while others may

13 decide it's a reasonable risk to take.

14 But the real question is: Will they be

15 willing to be randomly assigned? And I think that

16 will be also mediated by the kind of media coverage

17 this gets and how it's presented to them. It's a

18 really problematic issue because the drug is out

19 there.

20 DR. CHESON: Okay. If we could summarize

21 the first question, which boils down to: Is it

22 reasonable to request that placebo-controlled

161

1 trials be conducted to assess the risks of or rule

2 out a negative effect of EPO on time to progression

3 and survival? And my feeling is that we all feel

4 that it's not only reasonable but it's probably

5 essential. Is that the sense of the committee?

6 VOICES: Yes.

7 DR. CHESON: Antonio?

8 DR. GRILLO-LOPEZ: Perhaps with the

9 exception noted earlier in our discussion that

10 overall survival may not be the best endpoint, but

11 time to progression or progression-free survival

12 could do it.

13 DR. CHESON: Okay. Dr. Redman?

14 DR. REDMAN: I'm sorry. The question

15 between time to progression versus survival, is

16 that what you're asking? Or just--

17 DR. CHESON: No, I was asking the concept

18 of--I wasn't asking. I was summarizing the concept

19 of doing the randomized controlled trial, whatever

20 the endpoint we decide to be, is not just

21 reasonable but is necessary.

22 DR. REDMAN: Yes, okay.

162

1 DR. CHESON: Now, as far as the endpoint,

2 as we are going to be discussing this afternoon and

3 have discussed in the past, this endpoint may

4 differ from tumor type to tumor type. And there

5 are pros and cons, as we've heard passionately from

6 Dr. Grillo-Lopez, about one endpoint versus

7 another, PFS versus overall survival. And Dr.

8 Carpenter made that point also in breast cancer.

9 It sounds like progression-free survival is

10 probably a better endpoint.

11 Can we do this trial in the U.S.? The

12 second part of this. And I think we heard from our

13 patient advocate that it might be difficult, but I

14 think we also heard from the sponsors that these

15 trials are accruing, and hopefully they will

16 succeed. So I think the answer to that one is

17 probably also the--

18 DR. DeLAP: Could I ask Dr. George to --

19 [off microphone].

20 DR. CHESON: Please.

21 DR. M. GEORGE: If I have the chance to

22 comment on accruing patients in the U.S., if I

163

1 think it's feasible, it's feasible. How feasible

2 it is, that's the real question, because we want to

3 have the answer to the question really, really fast

4 and not wait. So we can have a trial up and

5 enrolling patients in a placebo-controlled trial

6 (?) period of time and wait for the answer or do

7 it in a different way and including patients

8 outside of the U.S. So the primary reason to the

9 trial outside of the U.S. is speed.

10 DR. CHESON: Ms. Mayer?

11 MS. MAYER: I have some concerns, I guess,

12 about our making use of patient populations outside

13 the United States to avoid the ethical issues that

14 may arise in doing trials here. It has to do with

15 disclosure, I suppose, and how patients interact

16 with their health care systems. I don't think it's

17 a simple issue that we should just glide right over

18 and say, yes, do the trials abroad.

19 DR. DeLAP: We're very sensitive to these

20 ethical issues and, in fact, it's certainly the

21 company's position--I'm sure it's also the position

22 of Amgen--we will not pursue a study in a

164

1 particular region because, you know, there are

2 ethical questions. It has to be a fully ethical

3 and well-justified trial wherever it's done.

4 DR. CHESON: Dr. George--

5 MS. MAYER: Can I do a follow-up on that?

6 I just want to point out that in countries where

7 the blood supply is not as safe as it is in the

8 United States, this may be a particular issue.

9 DR. PARKINSON: Just a comment. We work

10 globally. Cancer is a global problem. It is

11 solved by global cooperation. We work under the

12 same rules globally, just as Dr. DeLap emphasized,

13 same kinds of informed consent, same practice. We

14 would not work in a place where those kinds of

15 parameters were not in equipoise.

16 And with respect to the ability to work in

17 the United States, as we heard, with mature

18 investigators who can ask questions responsibly,

19 even in the placebo-controlled setting, I just

20 wanted to re-emphasize we've just accrued 145

21 patients over three weeks to a placebo-controlled

22 study of Aranesp in chemotherapy-induced anemia.

165

1 Is it as--no, it isn't as easy, of course. It's

2 never as easy in a randomized trial as it is in a

3 single-arm trial. And it's never as easy in a

4 randomized trial when there's a placebo control.

5 But sometimes it's actually necessary to

6 adequately--and we believe our responsibility is to

7 answer these questions definitively. We believe

8 patients have been confused by the reports of these

9 studies, and we believe that it's our

10 responsibility to them to answer this.

11 DR. DeLAP: To the extent that we're

12 looking at a more homogeneous population, we've

13 certainly done a lot of placebo-controlled trials

14 in chemotherapy-induced anemia or in

15 non-chemotherapy-induced anemia in cancer patients,

16 we're studying that also in the U.S. in

17 placebo-controlled trials. But when you start

18 focusing in on a specific population with a lot of

19 criteria to get as homogeneous a population as

20 possible, you can't cast as broad a net as you can

21 for chemotherapy-induced anemia.

22 Again, we're just saying--I think we're

166

1 all saying the same thing, but practicality demands

2 that if we're going to do this kind of work

3 efficiently, it has to be global.

4 DR. CHESON: Dr. George?

5 DR. GEORGE: I just wanted to clarify one

6 thing. The FDA can correct me if I'm wrong, but

7 there is nothing in the regulations or guidelines

8 that prohibit exclusive use of, in fact, foreign

9 data, if you want to call it that, in proving

10 things, right? That's one point.

11 DR. KEEGAN: Yes, that's correct.

12 DR. GEORGE: But with respect to that

13 little broader issue, it's certainly the case that

14 medical practices and cultures differ in countries

15 that would make it possible or more likely that you

16 would enter more patients from one country than

17 another. That's inevitable. And I don't think

18 that has anything to do with ethics unless you

19 believe there's some kind of universal ethics that

20 doesn't--you know, that applies to all countries,

21 which I think, you know, with respect to equipoise,

22 is not really true. That is, it's been shown in

167

1 other studies that the Europeans are more skeptical

2 of certain kinds of things that maybe we do, and

3 vice versa. So it doesn't bother me, as long as,

4 as Dr. Parkinson says, you're dealing in an area

5 that's accepted all the usual rules and

6 regulations.

7 DR. WEISS: If I can just add, it's also

8 just an issue of whether or not you can generalize,

9 as we've had some discussions, the results

10 across--you know, overseas to U.S. populations and

11 whether or not there are significant differences in

12 practices that would make those results somehow not

13 applicable.

14 DR. CHESON: Well, obviously, there are

15 some agents which will require some pharmacogenomic

16 differences, as we heard back last year with one

17 particular drug that was more effective in one

18 country than in another. But, in general, this

19 shouldn't be that big a problem.

20 Dr. Martino?

21 DR. MARTINO: I want to deal a little bit

22 with the issue of whether data which is generated

168

1 in Europe or elsewhere is accepted in this country.

2 You know, we have a very mixed history in this

3 country of what we accept that isn't generated

4 here.

5 Now, there are things that the companies

6 themselves can do to either enhance this separation

7 or to not allow the separation. And so I just want

8 to remind them that when their data is summarized

9 and presented, whatever the results are, it becomes

10 critical to present the data as it is meant, which

11 is one study done internationally. Oftentimes with

12 large studies, especially when the results aren't

13 exactly what you had hoped for, there's a tendency

14 to then separate the American group and its

15 results, the European group and its results, and

16 they're not always concordant.

17 And so there are ways to actually either

18 accentuate the American desire to not accept

19 non-American results, depending on how one handles

20 the presentation of these data.

21 DR. CHESON: Point well taken.

22 Dr. Keegan?

169

1 DR. KEEGAN: Just one last comment about

2 the European data. Dr. Weiss has summarized pretty

3 much our position on that. But one nuance here is

4 that Procrit is not approved or not marketed in

5 Europe. So if there was a study either that had

6 European sites or a separate European study, it

7 might be conducted with EPREX rather than Procrit.

8 We've already taken the position that these are

9 class effects, to some extent, and we think it

10 would at least address questions in the class.

11 Would the conduct of a study in which certain

12 patients received a related product in the class

13 different in Europe than in the U.S. pose problems,

14 or for instance, if the study was conducted

15 entirely in Europe, the data were obtained with

16 EPREX, would that be problematic, do you think, to

17 the committee in looking at--

18 DR. CHESON: Well, the reason we're here

19 is two studies conducted with a different product

20 that's caused a flurry in this country. So I think

21 the opposite would probably hold true. If those

22 studies had not shown this potential problem, we

170

1 wouldn't be sitting here today.

2 So I personally would find these very

3 similar compounds, the data from them to be

4 applicable on both sides of the puddle.

5 Dr. Feldman?

6 DR. FELDMAN: Just one very brief comment,

7 made perhaps out of the naivete of a non-clinician,

8 but I was a little bit concerned by the comments of

9 using European studies because it could be done

10 faster. I don't think the speed of getting the

11 answer should really be an issue here. I think the

12 idea is to get the best answer and most complete

13 answer to the questions.

14 DR. DeLAP: At J&J we agree with that,

15 that speed per se is not the issue. Getting the

16 best answer in a reasonable period of time is. But

17 there is some interrelationship because there are

18 changes, for example, in breast cancer them

19 regimens. So if you're trying to have a study that

20 has a relatively homogeneous approach to the

21 treatment of patients, and then that study turns

22 out extending out for six or seven years, it may

171

1 impact the quality of the study. Also, I think

2 there is a strong desire to address this question

3 as promptly as we can. Again, speed at all costs,

4 no. But speed to get a good answer, yes.

5 DR. CHESON: Dr. Redman?

6 DR. REDMAN: I just want to make a comment

7 about speed. As a clinician, I think if the trial

8 is adhered to, the quicker the accrual goes to the

9 trial, it is best. Speed does not imply a bad

10 trial.

11 DR. CHESON: Dr. Reaman?

12 DR. REAMAN: I would echo the comment on

13 speed. But for clarification, are we talking that

14 trials will either be done in Europe or in the

15 United States? Or if this really is a global

16 initiative, would they be international trials?

17 DR. CHESON: To finish my summary of this

18 before moving on to the next question, yes, these

19 trials are necessary, and I think the studies are

20 ongoing on both sides of the ocean, and that's how

21 they should be done. And hopefully they will be

22 completed alacrity in both situations.

172

1 Now, I think we already got to the other

2 question, that if there are so many variables

3 affecting response rate, survival, and safety, the

4 tumor type, et cetera, et cetera, if you had one

5 large trial that we all agreed on the endpoints,

6 that was conclusive in one disease, since we are

7 all here considering homogeneity of patients, would

8 that one trial answer the question for all

9 diseases? Or do we require now multiple trials in

10 different tumor types?

11 I personally feel that if we can answer it

12 in one very nice, well-done study in a common solid

13 tumor, that would answer it for me. Dr. Martino?

14 DR. MARTINO: But it seems to me that the

15 companies have already made these decisions, that,

16 in fact, they are doing several trials in different

17 tumors, and I have to say enough time has passed

18 that I only remember a few of the studies.

19 And so I would be really happy if someone

20 would succinctly review those trials because I

21 thought what was wanted from us was our thoughts as

22 to whether these trials were good, bad, or

173

1 indifferent.

2 Now, granted that they're already in

3 progress, perhaps my views on any of them are

4 irrelevant.

5 DR. CHESON: Could you state your name and

6 affiliation, please?

7 [Laughter.]

8 DR. PARKINSON: Well, just to remind you

9 about the clinical trials that we discussed, the

10 first is an Amgen-sponsored trial in small-cell

11 lung cancer, and 213 patients of the anticipated

12 600 patients have been accrued. That's

13 placebo-controlled.

14 The next trial is the AGO trial that I

15 talked about. That's a neoadjuvant breast cancer

16 trial being conducted by the German Gynecologic

17 Oncology Study Group. That's a study with an

18 anticipated enrollment of around 700 patients, a

19 little more than that, of whom 400 patients have

20 already been accrued. I indicated that that

21 interim analysis, which will include pathology

22 endpoints, will be looked at by the Data Monitoring

174

1 and Safety Committee in approximately five or six

2 weeks.

3 The third study is a study by the Western

4 German Study Group, and they'll be studying

5 adjuvant breast cancer patients, and that trial has

6 just started accrual.

7 The fourth study, the GELA Study, is a

8 study in aggressive non-Hodgkin's lymphoma--oh,

9 there it is. Thank you. I was doing this by

10 memory. This is so much easier, actually. The

11 GELA group is studying patients with non-Hodgkin's

12 lymphoma randomized to either dose stance or

13 standard chemotherapy plus or minus Aranesp or no

14 epoetin.

15 And then the final study is being

16 conducted by the Danish Head and Neck Cancer Study

17 Group. You've heard already this afternoon from

18 Dr. Overgaard about the rationale for that study

19 and the fact I've indicated here that 260 of the

20 600 patients have been already accrued, with a

21 safety interim analysis already completed and the

22 study continuing.

175

1 DR. DeLAP: If we can just quickly come

2 back to the slide from Dr. George's presentation.

3 DR. M. GEORGE: Thank you. This is a

4 slide I showed you earlier in four separate tumor

5 types where we have ongoing or completed clinical

6 trials where a tumor-relevant endpoint is the

7 endpoint.

8 We have many other trials, some very, very

9 large, including thousands of patients, in tumor

10 types like breast cancer, adjuvant breast cancer,

11 or Hodgkin's disease. Those trials enrolled 1,000

12 patients, but are not geared toward survival but

13 assessing correcting anemia and quality of life.

14 So I'm not going to present any of those trials,

15 but the list very, very lengthy.

16 In the tumor type, where we have relevant

17 endpoints are head and neck cancer. The (?) of

18 the seven(?) trials has enrolled, is completed, has

19 enrolled 301 patients, is currently under

20 follow-up, and the primary endpoint is disease-free

21 survival at two years. We will have those data

22 very, very shortly.

176

1 The RTOG study is a study in patients also

2 receiving radiation therapy for advanced head and

3 neck cancer. The study started with radiation

4 therapy alone, then was amended to include

5 chemotherapy also.

6 As mentioned earlier in Dr. Bowers'

7 presentation, the study was stopped to accrual

8 because of the increased incidence of TVE. And

9 when the Data Safety Monitoring Board of the study

10 reviewed the interim data, they thought that in

11 their trial there was no possibility of showing a

12 benefit. Those patients are in follow-up, and we

13 will have the data shortly

14 In non-small-cell lung cancer, there is a

15 pretty large study in Germany called GER-22, which

16 is planned to enroll 612 patients. Current

17 enrollment is around 250 patients, and the study is

18 ongoing. The last Data Safety Monitoring Board

19 meeting was a few weeks ago, and the trial is still

20 ongoing. The patients received chemotherapy first,

21 followed, after three cycles of chemotherapy, by

22 radiation therapy. The patients who have locally

177

1 advanced Stage III untreated non-small-cell lung

2 cancer--

3 DR. CHESON: We're going to need to limit

4 the details on this, because the point was: Do we

5 need more than one trial? And it's quite obvious

6 from both of these slides that we already have

7 many, many trials going on.

8 DR. M. GEORGE: And we're proposing a

9 large trial on top of all those trials.

10 DR. CHESON: Thank you.

11 DR. KEEGAN: Dr. Martino, just to clarify

12 the sequence of events, as we became aware of this

13 data, we contacted the companies to determine what

14 studies they had available or planned that might

15 speak to this question. But the purpose of this

16 committee is to comment on the qualities of such

17 trials that you think should be incorporated to

18 provide convincing data. So that if, in fact,

19 although they have many trials that are in the

20 works or ongoing, if you find that they are lacking

21 critical elements, we would like to hear that so

22 that we can negotiate with the companies the

178

1 appropriate trial to get the data. So that if you

2 see that there are critical elements missing,

3 that's really why we need to talk about this. They

4 may have studies going, they weren't intended for

5 the purpose we are here to discuss today, but they

6 may fit the bill. If they don't, we would like you

7 to say so.

8 DR. CHESON: Clearly, we haven't seen the

9 protocols, but based on what we've seen in the way

10 of presentations, there are quite a number of

11 studies for which the primary endpoint are those

12 that the FDA is looking for. They are accruing

13 patients, and so from my perspective hopefully at

14 least several of these out of the very large number

15 will be addressing the important issues that have

16 brought us here today.

17 First will be Dr. Redman.

18 DR. REDMAN: I agree with Dr. Cheson. The

19 studies are ongoing. In order to analyze a study

20 based on one slide is next to impossible. It lacks

21 a lot of information. But I think looking at the

22 companies that are doing those trials and the

179

1 investigators that are doing them, I certainly

2 don't have a problem with what's been going on.

3 DR. CHESON: If you'd like us to look at

4 some of these protocols with you and make sure they

5 have the appropriate elements, we'd be glad to do

6 that in our advisory capacity.

7 DR. WEISS: I'm just wondering if--Dr.

8 Martino started this discussion earlier on about

9 the issue of whether or not you should study a

10 population--everybody agrees, I think, that a

11 homogeneous population is important, but whether

12 you should try to address the issue if you have a

13 target hemoglobin of sort of the standard range

14 that's in the label, which is approximately 12 or

15 so, versus a strategy of pushing to the higher

16 hemoglobins. It seems like there's some diversity

17 of opinion around the table about what are the

18 important questions or how the study is to be

19 designed, what should be the strategy in terms of

20 the targets. And just looking at the slides that

21 Dr. Parkinson presented where they summarized on

22 the slides what the targets were, there were a

180

1 number of them that basically strove for a target

2 of about 14. Obviously, it's a little bit

3 different, I guess, if you're talking about men

4 versus women and what trials. But, in general,

5 you're talking about 14 except for the one Danish

6 trial which would be achieving a target of 15 or

7 pushing to try to, I guess, taper or stop the dose

8 if the hemoglobin goes to 15.

9 We didn't really hear from Dr. George

10 about what the targets where in those numbers of

11 different ongoing trials. So I'm just wondering if

12 those--except that the one that they're proposing

13 to do in breast cancer, which is actually designed

14 to target a hemoglobin that actually is at the

15 recommended label of hemoglobin, which is about 12

16 or so.

17 So I'm just wondering--I mean, it seems

18 like there's a smattering of many different trials

19 and many different tumor types, some looking at a

20 target of one versus a higher target. There's sort

21 of a whole hodgepodge of things, but is there a

22 particular issue with respect to target that--I

181

1 guess I'd like to hear from the committee whether

2 or not there's a particular target that we should

3 really be asking the companies to look at in terms

4 of the strategy in terms of trying to assess

5 benefits and risks.

6 DR. CHESON: Dr. Parkinson, a quick

7 answer.

8 DR. PARKINSON: Just one quick

9 clarification, for the five trials our target of

10 13, okay? Hemoglobin, withheld if above that at

11 14. That's European label, so that's guidelines in

12 Europe. The fifth trial is the Danish trial. Just

13 a clarification.

14 DR. THOMAS: As a further point of

15 clarification, we have amended all ongoing

16 protocols in this area to reduce the target

17 hemoglobin levels to a uniform level, and our view

18 would be that to do one at a higher level is to do

19 a study on TVE, not on benefits in terms of tumor

20 responses.

21 DR. WEISS: I guess it goes back to

22 something that Dr. Martino raised earlier, which is

182

1 that many of these studies, with the exception of

2 just a few minor ones, are actually now tapered

3 down to looking at outcomes within the recommended

4 label target. Whether or not that--you know,

5 whether or not you have comments about that,

6 because I know you raised that issue earlier,

7 whether that's something that should actually be on

8 the table to consider. If there are no issues at

9 those targets, is it appropriate to try to push to

10 higher targets to evaluate potentially other

11 benefits in terms of better survival and other

12 outcomes?

13 DR. CHESON: Dr. George was next.

14 DR. GEORGE: Well, my comments weren't

15 directly related to that.

16 DR. CHESON: We want to finish this first.

17 Do you have a comment related to this?

18 DR. CARPENTER: It would seem at least

19 logical to me to approach the question about tumor

20 benefit or risk at the levels currently targeted

21 now. If we then find either some beneficial effect

22 or at least exclude a detrimental effect, that

183

1 would leave the table open to go to a separate

2 study that compares two levels. But it seems to me

3 that if we try to answer the two questions at the

4 same time, we're going to get numbers problems.

5 We're going to get problems with speed of accrual

6 that are going to make it harder to get a timely

7 answer to the first question.

8 DR. CHESON: I agree with you. It's my

9 feeling that we would first like to have our level

10 of comfort at the indicated dose of the drug that

11 it was safe and efficacious. If there are

12 questions, such as the head and neck study we've

13 heard about, of higher doses, then those are

14 investigational doses that can be explored

15 separately. But I'd be willing to hear from my

16 colleagues if they disagree or agree.

17 Dr. Reaman?

18 DR. REAMAN: I absolutely agree, and I

19 think if we're going to be looking at safety at the

20 currently recommended indication, then we really

21 ought to consider excluding those trials in which

22 there is a target higher than what is the indicated

184

1 target in the package insert.

2 DR. CHESON: Any other comments on this

3 particular point?

4 [No response.]

5 DR. CHESON: Then we go to Dr. George.

6 DR. GEORGE: Well, the point I was going

7 to make was in reference to these issues of the

8 design. And it has to do with making sure that

9 what we're really trying to do is eliminate a

10 detrimental effect of some magnitude, I think. And

11 that can be done in studies; even when it's some

12 kind of superiority design, you can still make sure

13 you're looking at things that--because some of the

14 trials that even were presented showed--they were

15 presented as if there was no difference. But, in

16 fact, if you look at things just as simple as the

17 confidence intervals on certain things like hazard

18 ratios, they didn't exclude something that might

19 have been pretty detrimental, even though the

20 curves were superimposable and looked exactly the

21 same. That's the problem with these

22 non-inferiority kinds of designs.

185

1 But I think we have to keep that in mind,

2 that that's really what we're after here. I mean,

3 if it works, great. But what we're trying to do is

4 make sure it's not something bad.

5 DR. CHESON: Ms. Mayer?

6 MS. MAYER: Perhaps I missed one of the

7 trials, but the BEST Trial was done in first-line

8 metastatic breast cancer. The proposed two breast

9 cancer trials I believe are both adjuvant trials.

10 My question is: Is there a concern that

11 you might not in adjuvant trials capture the same

12 effect that appeared perhaps in the metastatic

13 trial?

14 DR. M. GEORGE: If I may try to answer

15 your question, the proposed trial is not a trial in

16 adjuvant breast cancer, but it's to treat patients

17 who have metastatic disease. There are some major

18 differences between the BEST Trial and the proposed

19 trial. The first one is the patients are anemic at

20 entry. The second is how we are going to assess

21 endpoint. The third one is the duration of therapy

22 with erythropoietin. In the BEST Trial, the

186

1 duration of therapy was one year targeting high

2 hemoglobin level.

3 DR. CHESON: Thank you.

4 Dr. Taylor?

5 DR. TAYLOR: I want to support the idea

6 that we are looking at more than one tumor because

7 I don't think we do know exactly why people may do

8 worse. I don't think we have an etiology or a

9 mechanism for which the erythropoietin product may

10 be adversely affecting people. So I think that

11 looking at different populations is not a bad idea.

12 The other reason is that extrapolations

13 are already made in that a lot of women receiving

14 adjuvant chemotherapy are on erythropoietin

15 products, and we need to know is it just the fact

16 that a woman has metastatic disease, is sicker, and

17 has other predisposing factors, or is it

18 erythropoietin?

19 DR. CHESON: Okay. Well, that question

20 has answered itself in that we have so many trials

21 going on in so many diseases.

22 Dr. Bauer?

187

1 DR. BAUER: I think there's a fairly

2 narrow margin here, especially from the safety. I

3 think the clear thread is that, you know, you drive

4 hematocrits up higher and you get more thrombogenicity. And

5 so I think, you know, the trials, I

6 guess, we're all talking better built in with

7 hemoglobin limits which are lower than those that

8 might have been desired, say in some of the

9 radiotherapy trials to improve oxygenation,

10 especially when you're talking about trials where

11 you're entering people who will probably have

12 normal hematocrits who would normally not be

13 candidates for erythropoietic growth factor. Your

14 margin in terms of driving up hematocrit is not all

15 that great.

16 But I think in response to the query, I

17 think you have to have clear limits in terms of

18 keeping hemoglobin either below 14 or certainly

19 target 13 and stop, or something, for patient

20 safety, for protection, because you just don't want

21 hematocrits to go uncontrollably high.

22 DR. CHESON: Okay. So, to summarize the

188

1 answer to this question, there really is no need to

2 summarize this because we already have multiple

3 trials going on. And if you have concerns about

4 the specifics of the trials, I'm sure you have the

5 protocols, and if you need some of us to go over

6 them in our particular areas of expertise, I'm sure

7 my colleagues would be glad to do that.

8 The next point I think we discussed a

9 little bit earlier, but maybe not conclusively, and

10 that is, the FDA has recommended that trials be

11 conducted in primary tumors where the EPO-R status,

12 whether it be expression, ligand, affinity, and

13 functionality of malignant cells in tumor

14 vasculature is known.

15 That's going to be tough. I think we

16 heard some very eloquent information this morning

17 that, A, it's going to be difficult and, B, it may

18 not be totally relevant, but I'm opening the floor

19 to additional discussion as to whether this

20 is--it's a nice idea, but is it doable in a variety

21 of circumstances? Feasibility, technicality, and

22 is it really relevant? Dr. Doroshow?

189

1 DR. DOROSHOW: Yes, I think that although

2 I'm usually a proponent of obtaining fresh frozen

3 materials for correlative studies, I think that

4 this is not feasible other than in the neoadjuvant

5 setting. I think the trials are of such size that

6 at most you will get a very small fraction that

7 will be potentially not reflective of the outcomes

8 you're trying to study. That's even irrespective

9 of the elegant data that we were presented with

10 earlier about lack of relevance.

11 DR. CHESON: Okay. So does anybody else

12 want to comment on--could you please identify

13 yourself?

14 DR. ROSENBERG: Yes, I'm Amy Rosenberg,

15 the Director of the Division of Therapeutic

16 Proteins. And while I agree I think it would be

17 difficult to characterize these receptors,

18 especially functionally, I don't think it's

19 impossible. Techniques of laser capture and

20 micro-dissection and protein arrays that can assess

21 via antibodies--antibody arrays, phosphorylated

22 proteins, are available. I think rather than

190

1 conclude that it's impossible, perhaps it would be

2 instructive to find out whether using more novel,

3 new techniques, it's possible to look at this.

4 Because, otherwise, we're not going to know

5 anything about the biology. We're not going to

6 know--you get a clinical result; you're not going

7 to know how to correlate that with functional

8 effects, especially for tumor activity.

9 So I think it's actually a critical point.

10 I think we'll learn very little except a clinical

11 outcome if we don't try and pursue it. And I think

12 there are ways to pursue it, and I think that those

13 should be looked into.

14 DR. CHESON: My concern is that these are

15 multicenter and perhaps multinational studies

16 where--I guess our colleague who is the expert over

17 there might want to comment again about whether

18 there are enough reference laboratories that could

19 do these or the samples could be shipped in, what

20 the feasibility is for shipped samples versus

21 on-site samples, et cetera, et cetera.

22 DR. LODISH: Well, as I tried to indicate

191

1 in my earlier presentation, we're at a level of

2 research rather than a robotized, commercialized

3 assay that could be done reproducibly. I think

4 things like laser capture to isolate tumor cells,

5 arrays at an ultra-micro-level clearly are the wave

6 of the future, but they're not practical now. And

7 certainly in a clinical setting I couldn't advocate

8 for them at all. And--well, let's end it there.

9 Ten years from now, we may revisit the system.

10 DR. CHESON: Thank you.

11 DR. DeLAP: A comment?

12 DR. CHESON: Okay.

13 DR. LEVINE: Mark Levine, McMaster. The

14 proposed trial is in women with metastatic breast

15 cancer. It's very difficult to get fresh tissue in

16 those patients. If there's chest wall disease and

17 so on, it's possible, but our experience, many of

18 us in the room, of doing trials in metastatic

19 breast cancer, it's not like adjuvant. It's not

20 early-stage breast cancer. So I don't think it's

21 feasible.

22 DR. CHESON: I think the only setting

192

1 where it may be possible is in a very limited

2 single institution or maybe a couple of

3 institutions who are part of the large cooperative

4 arrangement doing it on a very pilot, very

5 experimental basis as an exploratory issue. But to

6 do it as part of this sort of trial, with these

7 sorts of trials, and which diseases would you do

8 them in, again, if you talk about micro arrays and

9 those things, they're going to differ from disease

10 to disease, stage to stage. If you look at

11 lymphomas, even within histology of multiple

12 different array patterns, I'm not sure that we are

13 quite there yet.

14 Dr. Keegan, did you want to say something?

15 DR. KEEGAN: I think that the sentiment

16 behind this question was really one of

17 generalizability. If we do a study and we see a

18 negative outcome, an adverse outcome, do we

19 generalize it to all tumors? Similarly, if we see

20 no evidence of an effect, do we generalize it? And

21 this was one attempt to try and look at possible

22 mechanisms by which this might be affecting it.

193

1 If there's agreement that it can't be done

2 in the clinical trials due to lack of technology, I

3 think we have left open the possibility of trying

4 at least to characterize different tumor types

5 through tissue banks or other means so that we can

6 put the results of different trials in context,

7 particularly if we get answers that are not

8 consistent between different tumor types. I think

9 that was our concern also in terms of how one

10 chooses to select the tumor types to begin with.

11 DR. CHESON: I think this is an excellent

12 target for FDA-directed funded research.

13 [Laughter.]

14 DR. CHESON: Dr. Brawley?

15 DR. BRAWLEY: You know, I was wondering,

16 are we going to start doing a number of biopsies on

17 perhaps thousands of patients that are unnecessary?

18 If you have to go get the tissue, you know, for

19 clinical reasons, that's one thing. But to just do

20 the biopsy for the purposes of doing the biopsy for

21 science, that's another issue beyond the logistics

22 of how the tissue is going to be handled, you know,

194

1 minus 70 or liquid nitrogen and so forth, and the

2 transport and logistical issues. I think we have

3 to worry about that ethical issue.

4 DR. CHESON: Well, a lot of those

5 questions, such as transport, et cetera, can be

6 done on non-human tumor samples.

7 DR. BRAWLEY: Right.

8 DR. CHESON: But, for example, doing

9 biopsies, the CLGB is going to be conducting a

10 lymphoma trial looking at micro arrays and diffuse

11 large B-cell lymphoma prospectively, and it will

12 require needle biopsies of patients who have

13 already been biopsied. And we expect it's going to

14 hamper accrual because a lot of people won't want

15 to be re-biopsied, but to some the information will

16 be of sufficient importance that it may be of

17 interest and whatever.

18 Dr. Feldman?

19 DR. FELDMAN: Yes, I'd like to separate

20 out the issue of feasibility with that of

21 relevance, and I agree that it probably is not a

22 very feasible thing to do, whether it's in a

195

1 clinical trial or samples shipped to research

2 laboratories.

3 But I would disagree with those who think

4 that this may not be relevant, and I think it would

5 be very important, whether it be by FDA-directed

6 research or some other preclinical way, to find out

7 the precise relevance of these EPO receptors on

8 tumors.

9 DR. CHESON: Absolutely. We're not

10 disagreeing with that.

11 DR. FELDMAN: Particularly in those

12 tissues where normal tissue does not have EPO

13 receptor and tumor does, which includes the breast.

14 DR. CHESON: We agree that is a scientific

15 question of importance. It's just the feasibility

16 in a large-scale trial that we were considering at

17 the moment.

18 Dr. George?

19 DR. GEORGE: Just to follow up on

20 something Dr. Keegan said, just a warning. I

21 predict that you will get results that are going to

22 be hard to interpret. They're not going to be

196

1 entirely consistent. So it may be how you think

2 about this ahead of time. It may be a good time to

3 do that, how you're going to put all this together.

4 DR. WEISS: We'll come back to you at this

5 committee when those results are all there.

6 DR. CHESON: We will look forward to it.

7 [Laughter.]

8 DR. KEEGAN: I think to go to your point

9 and to correct something that you said, Dr. Cheson,

10 in fact, we have only begun to look at this

11 information, and we have not even reviewed the

12 protocols because most of these were actually not

13 company-sponsored trials. We really haven't gotten

14 to that yet. So we have not--I mean, this is very

15 timely advice for us in looking and giving guidance

16 and the necessity for getting additional

17 information. So we will look at it with an eye to

18 getting inconsistent results across products,

19 across trials, and try and build into that, and we

20 may need to come back to you on some of those

21 issues.

22 DR. CHESON: Now would be a good time to

197

1 get the protocols to ensure the consistency, so

2 that you get the answer, and you find out in five

3 years that someone didn't do what needed to be

4 done.

5 Who was over there? Dr. Bauer?

6 DR. BAUER: To reiterate a theme here, you

7 know, that any results that are obtained in one

8 tumor type I think are -- [microphone off] -- it's

9 not generic anemia. This is something that's

10 tumor-specific, potentially treatment-specific,

11 given the design of the trials, chemotherapy in

12 some and radiation in others--I'm sorry--will be

13 lack of generalizability from one tumor to another.

14 DR. CHESON: I think we have many, many of

15 the important tumor types included here, so we will

16 get generalized information. So that's our sense

17 on this particular question. Do we need to

18 summarize it any more, or have we got it? Okay.

19 The next one, clinical conditions

20 comprising thrombotic and cardiovascular events

21 vary from study to study. What are those specific

22 events that are clinically important?

198

1 I think we kind of all agreed on most of

2 them--didn't we?--that were in the protocols.

3 There was one where Dr. Luksenburg excised chest

4 pain, and a number of us were discussing this, and

5 we were wondering if you had actually gone into the

6 study data to find out what that chest pain really

7 was and whether it might not have been really

8 relevant to the trial, and not before these were

9 just tossed out.

10 DR. LUKSENBURG: I don't think we have

11 that specificity of attribution, just as chest

12 pain.

13 DR. CHESON: Okay. So it's kind of hard

14 to arbitrarily just yank them all.

15 Dr. Keegan, did you have something?

16 DR. KEEGAN: That's what I was going to

17 say. That's actually the problem with a lot of

18 safety data that we collect, that if you don't

19 target in advance what you want, you get things

20 that are coded in ways that make it difficult to

21 determine what it is you're looking at. In the

22 particular study that Harvey was alluding to, that

199

1 was actually in lung cancer. And so we thought it

2 would be particularly difficult there to determine

3 what the chest pain was attributed to. But what we

4 need are specific items to make sure that we're

5 capturing the relevant and important--

6 DR. CHESON: Now, in--I forget which of

7 the two partners here had a slide of eligibility

8 and toxicity and listing what were considered

9 cardiovascular problems. If someone would just put

10 it up here real quick so we can say yea or nay and

11 come to agreement?

12 DR. DeLAP: I think we can do that. I

13 think alternatively we could--we have Dr. Mark

14 Levine with us, and he could speak to what's

15 necessary or--well, okay. This is a broad

16 definition. Obviously, there are a lot of subcategories

17 that feed into these major categories.

18 But there are venous and arterial, so there are

19 deep venous thrombosis, pulmonary embolism,

20 arterial thrombosis, myocardial infarction,

21 cerebral vascular accident.

22 I think what I'd come back to, though, is

200

1 what I think Dr. Keegan was saying, that if we're

2 going to study this, we need to study these as

3 endpoints rather than as serious AEs, such as kind

4 of get collected in a study. So I was just going

5 to ask Dr. Levine if he had any comments about how

6 we should do this as study endpoints.

7 DR. LEVINE: I'll just be brief. I think

8 the agency--and Ken Bauer well knows that in the

9 thrombosis trials that the agency looks at, there

10 are standard definitions for objectively documented

11 pulmonary embolus, deep vein thrombosis; on the

12 arterial side, myocardial infarction and stroke.

13 They're well defined in the literature, and that's

14 what should be defined prospectively, and I think

15 that would advance the field much more than just

16 looking at AE forms.

17 DR. CHESON: Now, are you suggesting that

18 there be some sort of ongoing screening for these

19 events or that there just be a heightened awareness

20 of their clinical presence?

21 DR. DeLAP: I think that we're agreeing

22 that this is an issue that needs further research,

201

1 and I think that there needs to be structured

2 research in future protocols so that we're all

3 talking about the same thing and we're actually

4 assessing these in a precise way so that we can get

5 answers and actually start making comparisons

6 across trials and those kinds of things.

7 DR. CHESON: Dr. Bauer?

8 DR. BAUER: It's pretty clear. We're

9 talking about symptomatic endpoints. We're not

10 talking about routine screening. So we're talking

11 about clinically symptomatic relevant events, which

12 are different for other FDA-approved indications

13 for prophylaxis. We're talking about patients who

14 present--

15 DR. CHESON: So you will miss things, but

16 they're probably not clinically relevant.

17 DR. BAUER: Well, the important thing is

18 that the symptomatic events then be objectively

19 documented by appropriate radiographic studies, and

20 that's what sometimes is lacking in AEs, that

21 patients are--if they're suspected of having a

22 thrombotic event that they're objectively diagnosed

202

1 by appropriate imaging studies, and that's what we

2 really need.

3 DR. CARPENTER: And the other thing that

4 will turn out to be important is to ask--since

5 these are going to be symptomatic things, to ask at

6 prespecified intervals so that if there are

7 differences which might occur, they can be picked

8 up regularly. We won't bias the ascertainment.

9 DR. CHESON: And I think we need to also

10 have careful histories for pre-existing conditions

11 and that these things be evaluated such as

12 with--I'm surprised you didn't say that--looking

13 for hypo-coagulable conditions, the factor

14 deficiencies and what have you, protein

15 deficiencies.

16 DR. KEEGAN: Are you suggesting that as

17 eligibility criteria, Dr. Cheson?

18 DR. CHESON: No. I'm suggesting that you

19 take a careful history as part of the entrance onto

20 the study, that we know whether there is a family

21 history or personal history of prior DVTs,

22 cardiovascular disease, et cetera.

203

1 DR. KEEGAN: Right. But in the absence of

2 specifically developed case report forms, I think

3 the likelihood of getting good, quality data on

4 that might be difficult. I'm not sure if the

5 ongoing--I mean, remember, we're playing catch-up

6 here--these ongoing trials are specifically

7 capturing that information.

8 DR. CHESON: Just a suggestion.

9 Dr. Reaman?

10 DR. REAMAN: I was going to actually ask

11 how--I certainly applaud the use of these as study

12 endpoints instead of just AE findings. But how is

13 that going to impact on the ongoing trials that are

14 actually going to be used to answer these

15 questions. Are there plans to amend studies

16 looking at these at study endpoints?

17 DR. PARKINSON: Just a couple of comments.

18 First is we've generally not used these as

19 endpoints but, rather, as prespecified points of

20 interest, and that's probably a good way to go, as

21 I think you just suggested, Dr. Keegan.

22 With respect to the ongoing trials, we're

204

1 very much interested in the committee's

2 recommendations in this regard. Clearly, they need

3 to be followed--they need to be followed as

4 prespecified points of--pieces--what do you call

5 that? Events of interest. I was thinking of

6 points of light there for a second.

7 But one of the things, I think, that might

8 be very interesting--and I'd ask the committee for

9 their advice--is about using common prespecified

10 events of interest to allow comparability in

11 different clinical trial settings, because clearly

12 this is complicated. We have analyzed thrombotic

13 events every which way but loose since these trials

14 were--not published, but the results became

15 available. And so you find an association which is

16 rather weak with the use of epoetins. The highest

17 association is a history of prior thrombotic event.

18 Another association, which is independent,

19 relates to performance status. We've never

20 excluded patients with prior thrombotic event from

21 our trials. That may not always be the case. So

22 there are a number of parameters and a number of

205

1 collection parameters that would be nice to be

2 standardized in the interest of trying to get

3 closer, as we all I think are interested in here

4 today, closer to the real answers.

5 DR. DeLAP: If I could just add, we've

6 started to collect that kind of information in our

7 latest clinical trials, you know, more

8 prospectively, but obviously it will be very

9 helpful, as Dr. Parkinson says, if we have a

10 uniform way of doing it so that we can, you know,

11 compare notes, as it were, and get meaningful

12 interpretations in multiple trials.

13 DR. CHESON: Thank you.

14 Getting just to the last--yes?

15 DR. WEISS: Just a real quick

16 clarification. So you talked about getting a good,

17 careful history, family history, prior histories,

18 et cetera, maybe detecting undisclosed

19 hypo-coagulability states. Are people thinking,

20 though, something that Dr. Parkinson alluded to,

21 that those people should be excluded from trials or

22 just carefully document it so you can try to

206

1 evaluate what their risks are relative to other

2 populations and trials?

3 DR. PARKINSON: You know, I think we're

4 all interested in real-world answers because these

5 drugs are really used, generally. Because we have

6 not excluded patients with prior thrombotic events,

7 our rates actually reflect real-world use of the

8 drug in patients even with a prior history of

9 thrombotic event.

10 That would be our feeling, but we'd be

11 interested in the committee's discussion on this.

12 DR. CHESON: I'm all for the real world.

13 [Laughter.]

14 DR. CHESON: I knew you'd like that.

15 Have you all gotten from the committee

16 what you need on this particular question?

17 DR. KEEGAN: I think we've gone a long

18 way. I just would ask if Dr. Bauer would comment a

19 little bit on the type of documentation that you

20 would like for these sorts of events so that while

21 we're still in the public forum here you could

22 comment on how these should be documented for

207

1 purposes of data collection.

2 DR. BAUER: You know, for the five

3 entities that Mark Levine put up, there's strict

4 criteria, be it, you know, CT, angiography, or

5 ultrasound for a leg DVT and stroke and so forth,

6 and myocardial infarction, standard criteria. So I

7 think that suffices.

8 Let me just go back to this issue of

9 screening regarding eligibility. I think it should

10 be simple, and the strongest issue you'd want to

11 know about is really personal history of prior

12 thrombosis. And I guess at a minimum for this kind

13 of trial, I guess I would be uneasy about enrolling

14 people with prior history of thrombosis in this

15 trial as the only real exclusion, other than a

16 known thrombophilic disorder. I'm not advocating

17 it by any which way routine screening. But the

18 issue, I think, of enrolling people who have had

19 documented prior thrombosis, you know, I think is

20 an issue for the FDA and trial design, since I

21 gather they are allowing people to enroll who have

22 had prior thrombosis.

208

1 DR. PARKINSON: We didn't actively seek to

2 enroll them. Let me make that clear.

3 [Laughter.]

4 DR. PARKINSON: We did not exclude them,

5 and we recorded that information, which is why we

6 can present to you do our analysis of this. And

7 that is, I think, what we would advocate.

8 DR. CHESON: But the other issue is that

9 patients with cancer are already at the increased

10 risk of thrombotic events.

11 DR. PARKINSON: TE-25--oh, sorry.

12 DR. CHESON: What?

13 DR. PARKINSON: That analysis is actually

14 quite interesting. This is the pooled oncology

15 trials analysis looking at this potential

16 interaction between this history of prior

17 thrombotic event and treatment. It's interesting.

18 I'll leave it to you to interpret.

19 DR. DeLAP: Our data also says that the

20 biggest predictor of whether a patient is going to

21 have a thrombotic episode on the trial is if they

22 had one before, both in the placebo group and in

209

1 the treatment group. And, in fact, the other thing

2 I would just add is that in looking at the

3 different subsets of numbers of risk factors for

4 thrombotic events, it looks like there is some

5 added risk with erythropoietic therapy at any given

6 baseline risk. But it's not something that gets

7 profoundly worse at the higher baseline risk. So I

8 wouldn't--you know, I think it's better, as was

9 said before, to include as broad a population of

10 patients as you can and see what the answer is.

11 And the data that we have suggests that you can

12 actually enroll patients with a fairly significant

13 underlying risk of thrombotic events, and you may

14 see some additional risk, but it doesn't look like

15 it's a profound additional risk on top of the

16 underlying risk.

17 DR. CHESON: Dr. Bauer?

18 DR. BAUER: I dare say, with the consent

19 form pretty prominently in, you know, risk of

20 thrombosis is one of the adverse effects, in the

21 real world you're going to get very few of these

22 people into these trials.

210

1 DR. CHESON: I think that gets to the last

2 question, and that is, Should we have special

3 trials risk for high risk and low risk? And I

4 think that would be a difficult set of trials to do

5 because they all become at high risk when they have

6 cancer; it's just that some are higher than others.

7 And I'll repeat my question of the agency.

8 Are there any other issues that we have not

9 addressed to your satisfaction this morning?

10 DR. KEEGAN: I just want to make a comment

11 about an issue that was raised that I don't think

12 was fully resolved, and that was the concern about

13 looking at impact on overall survival because of

14 the difficulties with interpretation of data

15 following completion of the treatment. And I would

16 like to make it clear that our feeling is that

17 there may be difficulties in interpretation, but we

18 don't think that that difficulty should preclude

19 our attempts to determine if there are survival

20 impacts. So that while progression-free survival

21 is an important endpoint to look at, we should also

22 attempt to address the question on survival.

211

1 DR. CHESON: I agree. Whereas

2 progression-free survival to many of us is the

3 preferred primary endpoint in this setting, the

4 trials should be powered to adequately detect

5 survival differences as well as secondary

6 endpoints.

7 Are there any--overall survival,

8 right--additional comments from the committee? I

9 see two hands up. Ladies first. Dr. Taylor?

10 DR. TAYLOR: Well, I would agree, you have

11 to look at overall survival, because I still go

12 back that we don't know what the mechanism is for

13 erythropoietin effect on survival. And to just

14 look at progression-free survival is not going to

15 answer that question. And, yes, there will be

16 difficulties, but we have to know what that is.

17 DR. CHESON: Dr. Grillo-Lopez?

18 DR. GRILLO-LOPEZ: I believe that these

19 studies are a real challenge. They are difficult

20 conduct and difficult to interpret at the end. And

21 one additional factor that we haven't mentioned is

22 the use of concomitant medications which might be

212

1 anti-coagulant or pro-coagulant in nature.

2 And, again, we haven't seen the protocols,

3 as you have said, but I would assume that the

4 sponsors are collecting data on concomitant

5 medication because it's fairly standard. However,

6 it's important also to understand that the severity

7 of an adverse event is also going to be impacted by

8 how rapidly therapy is instituted, what kind of

9 therapy, and then the duration of that event is

10 also impacted by those considerations.

11 So it's just additional data that needs to

12 be collected in order to make sense of the results

13 at the end.

14 DR. CHESON: Very good point.

15 Any other comments or questions?

16 [No response.]

17 DR. CHESON: If not, I would like to thank

18 the sponsors for their excellent presentations,

19 carefully prepared, full of interesting data, and

20 to my colleagues in the agency and on the panel for

21 a very lively, interactive, and hopefully

22 productive discussion.

213

1 DR. WEISS: We second that. Thank you for

2 all your input.

3 DR. CHESON: We'll be back here at

4 12--make it 1 o'clock. We'll give an extra five

5 minutes. Thank you.

6 [Luncheon recess.]

214

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

2 [1:00 p.m.]

3 DR. KELSEN: Good afternoon. My name is

4 David Kelsen, and I'm a former member of ODAC, and

5 I've been asked to serve as Acting Chairman this

6 afternoon. And I'd like to welcome you this

7 afternoon to a session which will discuss endpoints

8 for colorectal cancer, regulatory approval.

9 Before we begin the session, I'd like to

10 ask the members of the committee to introduce

11 themselves, and we'll start with Dr. Grillo-Lopez.

12 DR. GRILLO-LOPEZ: I'm Antonio

13 Grillo-Lopez. I am a hematologist/oncologist with

14 the Neoplastic and Autoimmune Diseases Research

15 Institute.

16 MS. ROACH Nancy Roach from the Marti

17 Nelson Cancer Foundation, and I'm the patient rep

18 for this session.

19 DR. SARGENT: Dan Sargent, biostatistician

20 from the Mayo Clinic.

21 DR. O'CONNELL: Michael O'Connell, medical

22 oncologist and Director of Allegheny Cancer Center

215

1 and Associate Chair of the NSABP.

2 DR. BRAWLEY: Otis Brawley. I'm a medical

3 oncologist and epidemiologist from Emory

4 University.

5 DR. MARTINO: Silvana Martino, medical

6 oncology, from the John Wayne Cancer Institute.

7 DR. TAYLOR: Sarah Taylor, medical

8 oncology, palliative care, University of Kansas.

9 DR. REAMAN: Gregory Reaman, pediatric

10 oncologist, the George Washington University and

11 Children's National Medical Center.

12 DR. REDMAN: Bruce Redman, medical

13 oncologist, University of Michigan.

14 MS. CLIFFORD: Johanna Clifford, FDA,

15 Executive Secretary to this meeting.

16 DR. CHESON: Bruce Cheson, Georgetown

17 University, Lombardi Comprehensive Cancer Center.

18 DR. GEORGE: Stephen George, Biostatistics, Duke

19 University.

20 MS. HAYLOCK: Pamela Haylock, oncology

21 nurse, and I'm the consumer representative.

22 DR. CARPENTER: John Carpenter, medical

216

1 oncologist, University of Alabama at Birmingham.

2 DR. RODRIGUEZ: Maria Rodriguez, medical

3 oncologist, M.D. Anderson Cancer Center in Houston.

4 DR. DuBROW: Ronnie DuBrow. I'm a

5 radiologist at M.D. Anderson Cancer Center in

6 Houston also.

7 DR. IBRAHIM: Amna Ibrahim, medical

8 officer, Division of Oncology Drug Products.

9 DR. HIRSCHFELD: Steven Hirschfeld,

10 pediatric oncologist, Center for Biologics, FDA.

11 DR. WILLIAMS: Grant Williams, Deputy

12 Director, Division of Oncology Drug Products.

13 DR. KEEGAN: Patricia Keegan, Division

14 Director, Oncology Biologic Products.

15 DR. PAZDUR: Richard Pazdur, Division

16 Director, Oncology Drug Products, FDA.

17 DR. KELSEN: Thank you. I'll ask Ms.

18 Clifford to read a statement about conflict of

19 interest.

20 MS. CLIFFORD: Thank you. The following

21 announcement addresses conflict of interest issues

22 with respect to this meeting and is made a part of

217

1 the record to preclude even the appearance of

2 impropriety at this meeting.

3 The topics to be discussed this afternoon

4 will not focus on any particular product or company

5 but, rather, may affect all manufacturers of

6 products to treat colorectal cancer. The conflict

7 of interest statutes prohibit special government

8 employees from participating in matters that could

9 affect their own or their employer's financial

10 interests. All participants have been screened for

11 interests in the products and companies that could

12 be affected by today's discussions.

13 In accordance with 18 U.S.C, Section

14 208(b)(3), the Food and Drug Administration has

15 granted waivers to Dr. David Kelsen and Dr. Daniel

16 Sargent because it has determined that the need for

17 their services outweighs the potential for a

18 conflict of interest. A copy of the waiver

19 statements may be obtained by submitting a written

20 request to the agency's Freedom of Information

21 Office, Room 12A-30 of the Parklawn Building.

22 We would also like to note that Dr.

218

1 Antonio Grillo-Lopez, Chairman, Neoplastic and

2 Autoimmune Diseases Research Institute, is

3 participating in this meeting as an industry

4 representative, acting on behalf of regulated

5 industry.

6 In the event the discussions involve

7 products or firms not on the agenda for which an

8 FDA participant has a financial interest, the

9 participants are aware of the need to exclude

10 themselves from such involvement, and their

11 exclusion will be noted for the record.

12 With respect to all other participants, we

13 ask in the interest of fairness that they address

14 any current or previous financial involvement with

15 any firm whose product they may wish to comment

16 upon.

17 Thank you.

18 DR. KELSEN: Thank you. We'll open this

19 afternoon's session with opening remarks from Dr.

20 Pazdur.

21 DR. PAZDUR: I have to take a look at the

22 audience, and I noticed that it's really dropped

219

1 down in attendance, and perhaps it reflects the

2 departure of the stock analysts since we're not

3 talking about any product-specific application

4 here.

5 I began the comments on Monday, and the

6 presentation that I'm going to give I think is very

7 similar to what I gave in my introductory remarks

8 before we discussed the two drugs on Monday. And I

9 just want to go over some of these points because I

10 think that these points are germane not only to the

11 discussions that we had on Monday, but also are

12 germane to a discussion on colorectal cancer

13 endpoints, and these are recurring themes over and

14 over and over again.

15 The agency is open. That's why we're

16 having this discussion with you. We want

17 transparency of process. We want to make sure that

18 the endpoints that we select and discuss with our

19 regulated industry are ones that really measure the

20 true efficacy of the drug, are really going to give

21 us a determination of why we should approve a drug.

22 As you are all aware of, we have

220

1 traditionally held to the standard of the

2 demonstration of a survival advantage for the

3 approval, the regular approval of a drug in the

4 first-line setting, and also in the adjuvant

5 setting of colorectal carcinoma. And as I

6 mentioned and we discussed throughout these

7 proceedings, we've looked at survival as an

8 unambiguous endpoint. It's measured on a daily

9 basis. We feel that given an accepted safety

10 profile that it is the ultimate in clinical

11 benefit.

12 But, nevertheless, as I stated in my

13 opening comments on Monday, we realized that there

14 can be shortcomings of a survival analysis,

15 depending on what setting one is looking at.

16 Obviously, survival analysis requires large numbers

17 of patients. This may or may not be a problem in

18 colorectal carcinoma. Obviously, it's not as big a

19 problem in colorectal carcinoma as it would be, for

20 example, in hairy cell leukemia or in lymphoma.

21 There are problems perhaps with long

22 patient follow-ups, which generally is not that big

221

1 of a problem in metastatic colorectal carcinoma,

2 but with improvements in survival, fortunately, we

3 are seeing that patients with colorectal carcinoma

4 live longer.

5 Perhaps one of the areas that we're most

6 concerned about is this issue of crossover, and

7 crossover can go two ways, and we've seen this in

8 discussions of applications. Obviously, it can

9 obscure a survival advantage in a randomized study,

10 but if there is an unequal crossover going in one

11 direction, it may actually provide you the

12 suggestion, at least of an erroneous conclusion

13 based on survival.

14 So, by all means, all of our endpoints

15 that we have are far from perfect, and I think

16 people have addressed this throughout the day.

17 We're here to get your consensus and your feeling

18 on where we should go with our discussion with

19 industry in the future.

20 I think when we talk about the specifics

21 here, let's go over endpoints, and I think issues

22 that we need to focus on--and we will be focusing

222

1 on these during the presentations, but also in the

2 discussion. When we take a look at the adjuvant

3 setting and look at disease-free survival, I think

4 we have important questions that we must address.

5 If we accept this as a regulatory endpoint, are we

6 saying that disease-free survival is a surrogate

7 for survival, overall survival? Is it reasonably a

8 likely surrogate for overall survival? And those

9 are the key words for accelerated approval. Is it

10 a surrogate for an improved life because one delays

11 the uncertainties of the diagnosis of cancer being

12 made in an adjuvant setting at the time of relapse?

13 So there are issues here that I think we need to

14 address when we talk about disease-free survival.

15 When we talk about in the advanced disease

16 setting, when we're talking about time to

17 progression or progression-free survival, again,

18 are we saying it is a surrogate for survival or is

19 it a really true endpoint of clinical

20 meaningfulness in itself? And those are some of

21 the questions that we will be posing to you.

22 In the two applications that we saw

223

1 yesterday, especially with the first, I think there

2 were important questions that we discussed

3 regarding the rigor of measurement of a time to

4 event endpoint such as progression-free survival.

5 One has to account for missing visits, asymmetry of

6 follow-ups. What is the role of an external

7 radiology committee vis-a-vis the response or the

8 progression determination, I should say, of the

9 actual investigators that are seeing these

10 patients? And could this vary from disease to

11 disease? For example, in colon cancer where most

12 of the progression is picked up on CT scan, are

13 investigators' determinations in a randomized study

14 which might balance out clinical findings of

15 progression really going to be that important when

16 we have a discussion of what is the role of a

17 radiology committee?

18 We've had a tendency for recent

19 applications--and you've seen these because they've

20 come to you--to be asked to make decisions on the

21 basis of one trial. Should we require a greater

22 degree of statistical significance if we take a

224

1 look at one trial versus two trials? I'll let you

2 know as a caveat that there are many divisions in

3 the FDA that do look at a higher degree of

4 statistical persuasiveness when they examine one

5 trial. If we move away from survival, would this

6 be especially important to require a higher degree

7 of statistical significance?

8 As we had in our discussion of the morning

9 application on Monday, what is the value of a small

10 increment in progression? How does one define that

11 as one enters the trial prospectively with the

12 company?

13 So these are just some of the questions

14 that I want to pose to you. Here, again, our whole

15 purpose in looking at this is a degree of

16 transparency. We're open. We want to make sure

17 that we're giving the correct advice to patients.

18 I always say there are sins of omission and sins of

19 commission when we're in drug development and

20 making regulatory decisions. Many times the

21 marketplace itself will address a bad drug that's

22 out there. People simply won't use it. However,

225

1 if drugs don't get out there, the marketplace and

2 market forces cannot answer that question. And,

3 obviously, there is always a degree of balance.

4 What we're looking for, however, when we have an

5 ODAC meeting such as this is what is the rigor,

6 what is the science that would go into making these

7 decisions.

8 Thank you very much.

9 DR. KELSEN: Thank you, Dr. Pazdur.

10 I'll ask Dr. Ibrahim now to discuss

11 regulatory background and past approvals.

12 DR. IBRAHIM: Good afternoon. I will be

13 discussing the regulatory background and past FDA

14 approvals in colorectal cancer.

15 First, the presentation outline will be as

16 follows: I will discuss a background of regulatory

17 requirements for drug approval, endpoints for

18 regular and accelerated approval, agents approved

19 for adjuvant, first-line, and recurrent therapy of

20 colorectal cancer, and endpoints used for them will

21 be presented. I will review briefly the major

22 trials that led to drug approval, first for drugs

226

1 for adjuvant therapy and then for first-line and

2 second-line therapy. Finally, I will conclude with

3 the endpoints that have supported approval of drugs

4 for colorectal cancer in these three treatment

5 settings.

6 Drug approval requires adequate and

7 well-controlled studies demonstrating that the drug

8 is safe and is effective for the approved

9 indication. The safety requirement comes from the

10 Federal Food, Drug, and Cosmetic Act of 1938. The

11 efficacy requirement is from a 1962 amendment to

12 the Act.

13 There are two routes to a new drug

14 approval. The traditional route is a regular

15 approval. Sometimes it is referred to as full

16 approval. It requires the demonstration of

17 clinical benefit or an effect on an established

18 surrogate for clinical benefit. Clinical benefit

19 is usually considered to be tangible benefit of

20 obvious worth to the patient, such as prolongation

21 of survival or relief of pain.

22 Sometimes FDA has accepted surrogates as

227

1 the basis for regular approval, usually after much

2 clinical experience with the surrogate and

3 widespread acceptance by patients and physicians.

4 Examples are lowering blood pressure and lowering

5 cholesterol. On occasion, these assumptions of

6 obvious benefit have been proven wrong, such as the

7 benefit of suppressing some arrhythmias.

8 Another mode of approval is accelerated

9 approval, approval which can be based on a

10 surrogate endpoint considered to be reasonably

11 likely to predict clinical benefit. I will discuss

12 accelerated approval in a later slide.

13 One of the central questions we address at

14 the end of Phase II meetings is: How many trials

15 are needed for approval? The usual answer is more

16 than one, and this is based on the definition of

17 substantial evidence of effectiveness in the

18 amended Food, Drug, and Cosmetic Act and the fact

19 that the word "trials" is plural in that

20 definition. Reasons for needing additional

21 evidence are the possibility of unrecognized trial

22 bias and also just chance alone.

228

1 However, FDA has recognized that sometimes

2 results from a single trial may suffice. Although

3 approvals based on a single trial have been granted

4 on occasion for many years, this practice was

5 written into law by the FDA Modernization Act, or

6 FDAMA, in 1997. The possible use of only one trial

7 was also detailed in the FDA Effectiveness

8 Guidance, finalized in 1998. As worded in that

9 guidance, a single trial may suffice, but generally

10 only in cases in which a single multicenter study

11 of excellent design provided highly reliable and

12 statistically strong evidence of an important

13 clinical benefit, such as an effect on survival and

14 a confirmatory study would have been difficult to

15 conduct on ethical grounds.

16 Regular approval requires evidence of

17 clinical benefit or improvement in an established

18 surrogate of benefit. In oncology, survival is

19 obviously the gold standard for clinical benefit.

20 But the FDA has accepted other endpoints for cancer

21 drug approval.

22 In the 1970s, FDA usually approved cancer

229

1 drugs based on objective response rates. In the

2 early 1980s, after discussion with ODAC, FDA

3 determined that response rate was generally not

4 sufficient evidence for approval. Given the

5 toxicity of cancer drugs, approval needed evidence

6 of improvement in survival or in a patient's

7 quality of life. Example: improved physical

8 functioning or improved tumor-related symptoms.

9 There have been recent examples of

10 endpoints that were accepted as established

11 surrogates of clinical benefit in specific cancer

12 settings. These endpoints supported regular

13 approval. Disease-free survival has been accepted

14 as an adequate endpoint in the setting of adjuvant

15 treatment of breast cancer based on the belief that

16 a large proportion of the recurrence were

17 symptomatic.

18 Durable complete response was considered

19 an acceptable endpoint in testicular cancer and

20 acute leukemia because the untreated conditions

21 were quickly lethal, or even in some chronic

22 leukemias and lymphomas when it was clear that

230

1 remission would lead to less infection, bleeding,

2 and blood product support.

3 Even with solid tumors, the FDA has

4 suggested that tumor response may sometimes support

5 approval, but that this judgment needs also to

6 consider additional evidence such as response

7 duration, relief of tumor-related symptoms, and

8 drug toxicity.

9 As discussed in the following sections,

10 response rate with adequate response duration has

11 sometimes supported either regular approval or

12 accelerated approval, especially in patients with

13 heavily pre-treated or refractory disease, and

14 especially with less toxic therapies, such as

15 hormone treatment of breast cancer.

16 Recently, the Division of Oncology Drug

17 Products evaluated the basis of approvals since

18 1990 for drugs in our division. As shown on this

19 slide, survival was the approval endpoint in the

20 minority of approvals: 73 percent of all approvals

21 were not based on survival, and if you exclude

22 accelerated approvals, 67 percent of approvals were

231

1 not based on survival.

2 Let's turn to accelerated approval. This

3 slide lists the major issues. The accelerated

4 approval regulations are for diseases that are

5 serious or life-threatening, where the new drug

6 appears to provide benefit over available therapy.

7 The key point for our consideration is that

8 accelerated approval can be granted on the basis of

9 a surrogate endpoint that is reasonably like to

10 predict clinical benefit.

11 After accelerated approval, the applicant

12 is required to perform a post-marketing study to

13 demonstrate that the treatment with the drug is

14 indeed associated with clinical benefit. If the

15 post-marketing study fails to demonstrate clinical

16 benefit or if the applicant does not show due

17 diligence in conducting the required study, the

18 regulations describe a process for rapidly removing

19 the drug from the market.

20 The approved agents in the table are

21 listed according to the treatment setting with the

22 drugs for adjuvant use in the left column, for

232

1 first-line use in the middle column, and those for

2 the recurring cancer in the column on the right. I

3 will present them to you in chronological order.

4 5FU was the first drug approved for colon

5 cancer in 1962. We will not discuss this further

6 since 5FU approval predated the era of controlled

7 clinical trials in oncology.

8 After a long gap, levamisole was approved

9 in combination with 5FU in 1990 for adjuvant use.

10 Although reports in the literature have been

11 described regarding results supporting the use of

12 5FU Leucovorin for adjuvant therapy, the FDA has

13 not received an NDA submission supporting this

14 indication. Leucovorin was approved in 1991 in

15 combination with 5FU for first-line therapy.

16 Irinotecan initially received accelerated

17 approval for recurrent colorectal cancer in 1996,

18 followed by a regular approval for the same

19 indication. Subsequently, in 2000 it was approved

20 for first-line use.

21 Capecitabine is the only agent approved

22 for first-line setting based on non-inferiority

233

1 analysis.

2 Oxaliplatin in combination with

3 5FU/leucovorin received an accelerated approval for

4 recurrent colorectal cancer, which was converted to

5 a regular approval, and then it was also approved

6 for first-line therapy earlier this year.

7 Bevacizumab and cetuximab in 2004 are the

8 first biologic agents to have received approvals

9 for first-line and recurrent colorectal cancer,

10 respectively. The approval for bevacizumab was

11 regular, and for cetuximab it was accelerated.

12 As you will see, survival was the endpoint

13 supporting all regular approvals. Randomized

14 trials demonstrating superiority led to all but one

15 of the regular approvals. For one drug,

16 capecitabine, non-inferiority in overall survival

17 supported regular approval. Three drugs received

18 accelerated approval in previously treated

19 populations. Two were supported by a response rate

20 in single-arm trials and one by a response rate and

21 time to tumor progression superiority shown in

22 interim analysis of a randomized trial.

234

1 Now, agents for adjuvant therapy.

2 Levamisole was approved in combination

3 with 5FU in 1990 based on the results of two

4 trials. After surgery, patients were randomized to

5 no further therapy, levamisole alone, or 5FU plus

6 levamisole. Levamisole plus 5FU demonstrated a

7 reduction in death rate by about 30 percent. The

8 follow-up period was two to five years for these

9 studies. Although the contribution of levamisole

10 to this regimen was not demonstrated in clinical

11 trials, this was the first adjuvant regimen to show

12 a survival benefit. And levamisole was approved

13 based on these results.

14 Agents for first-line therapy.

15 The combination of 5FU/leucovorin was

16 approved for treatment of advanced disease in 1991.

17 Study 1 is a five-arm study, but for simplicity

18 only three arms are shown in this table. A

19 randomized study demonstrated improvement in

20 response rate, time to tumor progression, and

21 overall survival of high- or low-dose leucovorin

22 combined with 5FU. These two arms of the study

235

1 were extended along with sequential methotrexate,

2 5FU/leucovorin arm from the same study. This is

3 the study seen in the table. The results remain

4 consistent with the initial study. Overall

5 survival was about 12.5 months in the initial as

6 well as the extension of the study.

7 In 2000, irinotecan was approved for

8 first-line therapy following its initial

9 accelerated approval for refractory colon cancer.

10 Two randomized, multicenter trials compared

11 infusion of 5FU/leucovorin plus or minus irinotecan

12 in untreated patients. Each trial had over 300

13 patients and demonstrated an improvement in

14 response rate, time to tumor progression, and

15 overall survival.

16 These differences in survival were

17 observed in spite of second-line therapy in a large

18 number of patients on both arms, including

19 crossover to irinotecan-containing regimens in the

20 control arm.

21 Capecitabine is the only colon cancer drug

22 approved based on non-inferiority analysis. The

236

1 combined survival data from two open-label,

2 randomized trials of capecitabine versus

3 5FU/leucovorin formed the basis of approval.

4 Sufficient historical data existed to allow a

5 reasonably precise estimate of the effect of

6 5FU/leucovorin on survival. The non-inferiority

7 analysis showed that at least 50 percent of the

8 5FU/leucovorin effect was retained by capecitabine.

9 This drug was approved for a restricted first-line

10 indication, "for patients when treatment with

11 thoropyrimidine(?) therapy alone is preferred."

12 One drug, oxaliplatin, and one biologic

13 agent, bevacizumab, were approved for first-line

14 use in colon cancer in 2004. In one randomized

15 trial, a combination of oxaliplatin with

16 5FU/leucovorin, known as the FOLFOX4 regimen,

17 demonstrated superiority in overall survival when

18 compared with the control regimen of IFL. The

19 study design was complicated, and there was an

20 unequal crossover. Twenty-four percent of the

21 patients in the IFL arm received oxaliplatin in

22 their second-line therapy; whereas, only 8 percent

237

1 of patients in the oxaliplatin combination arm

2 received irinotecan. An improved time to tumor

3 progression supported the improved survival

4 observed in the FOLFOX4 arm. Additionally, it

5 could be inferred that oxaliplatin plus

6 5FU/leucovorin administered sequentially with IFL

7 is better than irinotecan plus 5FU/leucovorin

8 without oxaliplatin.

9 The safety and efficacy of bevacizumab in

10 the initial treatment of patients with metastatic

11 carcinoma of the colon and rectum were studied in

12 two randomized, controlled clinical trials in

13 combination with intravenous 5FU-based

14 chemotherapy. The results of the larger trial with

15 over 800 patients demonstrated a superiority in

16 overall survival by about five months. In the

17 smaller trial, a randomized, exploratory Phase II

18 trial with just over 100 patients, statistical

19 significance was observed only for progression-free

20 survival in the 5FU/leucovorin plus 5 milligrams of

21 bevacizumab. There was a trend for improved

22 survival.

238

1 Agents for refractory cancer.

2 In 1996, irinotecan was the first

3 chemotherapy agent since 5FU to receive approval

4 for treatment of pretreated, advanced colorectal

5 cancer. Three single-arm studies with response

6 rate ranging from 14 to 21 percent and response

7 duration of 5.8 months led to accelerated approval

8 for second-line therapy. A survival benefits was

9 subsequently demonstrated in two randomized trials

10 shown in the next slide.

11 These randomized trials demonstrated

12 superiority in survival by 2 to 2.5 months against

13 best supportive care, and 5FU-based regimens led to

14 regular approval in the second-line setting.

15 Interestingly, these trials were not part of the

16 original regulatory plan to convert the accelerated

17 approval to regular approval. While the single-arm

18 trials were being reviewed by FDA, these

19 confirmatory trials were being conduct in Europe.

20 The initial agreement between the sponsor and FDA

21 was that the trials in the first-line setting were

22 to provide initial proof of clinical benefit.

239

1 Oxaliplatin in combination with

2 5FU/leucovorin received accelerated approval based

3 on improved response rate and time to tumor

4 progression, shown at an interim analysis of a

5 three-arm randomized trials. Patients in this

6 trial had disease which progressed on or recurred

7 within six months of treatment with the IFL

8 regimen. The oxaliplatin combination arm had a

9 response rate of 9 percent versus 0 to 1 percent in

10 the single agent oxaliplatin arm and 5FU/leucovorin

11 control. The time to tumor progression was

12 increased by two to three months compared to the

13 other two arms.

14 There were some important observations.

15 Because of the inclusion of the single agent

16 oxaliplatin arm, the contribution of 5FU/leucovorin

17 to the combination regimen was shown definitively.

18 It also demonstrated that oxaliplatin should not be

19 used alone in the pretreated population. Follow-up

20 of this study did not demonstrate a survival

21 advantage for the oxaliplatin regimen.

22 Cetuximab used in combination with

240

1 irinotecan received accelerated approval in 2004

2 for the treatment of EGFR-expressing metastatic

3 colorectal carcinoma in patients who are refractory

4 to irinotecan-based chemotherapy. An accelerated

5 approval was also granted for cetuximab as a single

6 agent for the treatment of EGFR-expressing

7 metastatic colorectal carcinoma in patients who are

8 intolerant to irinotecan-based chemotherapy.

9 This regular approval is based on one

10 two-armed randomized trial and two single-arm

11 studies. The multicenter, randomized, controlled

12 clinical trial was conducted in over 300 patients

13 randomized to receive either cetuximab plus

14 irinotecan or cetuximab monotherapy. Cetuximab

15 plus irinotecan improved time to tumor progression

16 by about 2.5 months compared to the single agent

17 cetuximab.

18 In the single-arm trials, the overall

19 response rate was 9 to 15 percent for single agent

20 cetuximab and in combination with irinotecan. The

21 median durations of response were approximately 6.5

22 to 4.2, respectively.

241

1 In summary, the FDA requirements for drug

2 approval were reviewed, including the need for

3 evidence from well-conducted, well-controlled

4 clinical trials, or sometimes from a single trial

5 plus confirmatory evidence; and regular approval

6 which needs evidence showing clinical benefit or an

7 accepted surrogate for clinical benefit; and

8 accelerated approval which must show an advantage

9 with respect to available therapy and may use an

10 endpoint that is only reasonably likely to predict

11 benefit. We reviewed the approval endpoints that

12 FDA has accepted over the past several years.

13 I will conclude my presentation with this

14 slide, which gives an overview of the basis of

15 approval in colorectal cancer. Levamisole with 5FU

16 is the only drug approved for adjuvant therapy

17 after demonstration of superiority in survival.

18 Five drugs are approved for first-line therapy, and

19 they are 5FU, irinotecan, oxaliplatin, bevacizumab,

20 and capecitabine. Superiority analysis for first

21 (?) and non-inferiority for capecitabine for

22 survival led to the approval. Irinotecan,

242

1 oxaliplatin, and cetuximab are the three drugs that

2 have received accelerated approval for recurrent

3 disease based on response rate and on time to tumor

4 progression.

5 For irinotecan, clinical benefit, that is,

6 survival, was demonstrated later in two randomized

7 studies, leading to full approval in the recurrent

8 disease setting. Oxaliplatin's accelerated

9 approval was converted to a regular approval on the

10 basis of a large randomized trial in previously

11 untreated patients.

12 Thank you.

13 DR. KELSEN: Thank you. We're going to

14 hold questions until after the two additional

15 presentations, and I'll now ask Dr. O'Connell if

16 he'll give his synopsis on the FDA Endpoints

17 Workshop.

18 DR. O'CONNELL: Thanks very much, David.

19 My task today is to review for you the

20 results of the workshop sponsored by the FDA and

21 held on November 12th of last year. I had the

22 privilege of co-chairing this meeting along with

243

1 Dr. Pazdur. Dr. Williams and Dr. Ibrahim also gave

2 presentations. Dr. Kelsen was a panelist. Dr.

3 Sargent also gave a presentation there.

4 The purpose of that workshop was to

5 discuss both the positive and the negative aspects

6 of various endpoints for approval of new drugs for

7 colorectal cancer. Specifically, it was not reach

8 a consensus or to give FDA any advice. That's the

9 job of this committee today.

10 Secondly, we were to identify areas for

11 further research that might help identify more

12 effective endpoints for colorectal cancer drug

13 approval.

14 And then third was to provide information

15 to you so that you could give whatever

16 recommendations you think appropriate to the FDA

17 based upon this discussion.

18 The workshop consisted of a series of

19 presentations, very similar to the one you just

20 heard, regarding the regulatory background and the

21 summary of previous approvals. There were five

22 presentations given by different panelists at this

244

1 meeting, and I'll briefly summarize these

2 presentations for you.

3 I said there was a very lively,

4 interactive discussion between the speakers and the

5 multidisciplinary panel. There certainly was a

6 free range of expression of opinions on the various

7 endpoints. And then we concluded with some

8 discussion of questions that were posed by the FDA.

9 My goal in this presentation today is to

10 give you a very brief capsule summary of the

11 presentations and the main points of discussion

12 without going through a litany of all of the

13 discussions that occurred over that six-hour

14 period.

15 The focus is really to provide some

16 information regarding new endpoints for regulatory

17 approval of drugs for colorectal cancer and, in

18 particular, there are three endpoints that are of

19 particular interest that I'll emphasize during my

20 presentation: time to progression as a regulatory

21 endpoint for first-line metastatic colorectal

22 cancer; three-year disease-free survival as an

245

1 endpoint for regulatory approval in the colon

2 adjuvant situation; and three-year local control as

3 an endpoint in rectal cancer adjuvant studies.

4 So let's briefly go through the

5 presentations. One of these presentations was

6 given by Dr. Charles Blanke from the University of

7 Oregon. His topic was the use of biomarkers or

8 quality of life as regulatory endpoints for

9 patients with colorectal cancer. There was a fair

10 amount of discussion regarding the use of the

11 carcino-embryonic antigen, or CEA, and it was the

12 point of view of the speaker and the panel that it

13 really wasn't possible to consistently predict

14 clinical benefit based on fluctuations of CEA.

15 Further, the ASCO guidelines do not

16 recommend other biomarkers for colorectal cancer,

17 including the variety of molecular markers that

18 have more recently been described in the

19 literature.

20 Dr. Blanke then went on to discuss the

21 pros and cons of quality-of-life analysis in

22 colorectal cancer patients. Of course, he pointed

246

1 out that there are multiple methodologic issues

2 involved with quality-of-life measurements,

3 problems with missing data, problems with perhaps

4 not asking the correct question in the

5 quality-of-life questionnaire or instrument that's

6 pertinent to the particular disease process under

7 question.

8 He pointed out that it really isn't know

9 whether there are significant changes in

10 quality-of-life parameters in regimens known to be

11 effective in colorectal cancer and also pointed out

12 you really can't discriminate between safety and

13 efficacy based on quality-of-life endpoints.

14 Perhaps one of the most important issues

15 was that many patients with metastatic colorectal

16 really don't have significant tumor-related

17 symptoms. They don't have severe pain. They don't

18 have a significant decrease in performance status.

19 Many of these patients are asymptomatic or have

20 minimal symptoms, questioning the use of resources

21 in measuring this parameter in a population that

22 frequently does not demonstrate significant

247

1 symptoms.

2 Dr. Blanke then went on to discuss the

3 clinical benefit response, which, of course, was

4 used as a regulatory endpoint for approval of drugs

5 in pancreatic cancer where most patients with

6 pancreatic cancer do have pain, do have significant

7 reduction in their performance status, and

8 frequently have significant weight loss. Again,

9 the issue with metastatic colorectal cancer, these

10 parameters are frequently not present.

11 Further, those three specific symptoms do

12 not really adequately encompass the variety of

13 symptoms that patients with metastatic colorectal

14 cancer might experience related to bowel

15 obstruction, the development of ascites, liver

16 dysfunction and so on.

17 And, again, this type of clinical benefit

18 response was felt perhaps to be of most benefit in

19 patients that were likely to be very symptomatic,

20 which would include patients with rectal cancer to

21 a much greater degree since patients with rectal

22 cancer frequently experience local tumor recurrence

248

1 which can result in severe pain, ureteral

2 obstruction, and other clinical problems.

3 Dr. Meg Mooney from the National Cancer

4 Institute then provided a discussion of endpoints

5 for neoadjuvant and adjuvant therapy of rectal

6 cancer. Here, in contradistinction to colon

7 cancer, the failures are frequently very

8 symptomatic, and there's a higher proportion of

9 locoregional failures as a component of the

10 failures in patients undergoing potentially

11 curative surgery. In fact, one of the panel

12 members was very precise in stating that local

13 tumor control at three years is an appropriate

14 endpoint for full approval, and there were several

15 other members of the panel that had the same point

16 of view and no dissent.

17 Pathologic complete response engendered

18 more discussion. If you have a patient receiving

19 preoperative radiation and chemotherapy, one

20 measure of efficacy is to determine whether in the

21 resected specimen there's any histologic evidence

22 of residual tumor. And although it was felt to

249

1 definitely relate to biological activity, there

2 were quality control issues raised: evaluation of

3 the radium margin(?), quality control issues in

4 determining whether or not there truly was

5 microscopic residual disease. So the general

6 feeling there was that pathologic complete response

7 might be premature as a regulatory endpoint at this

8 point.

9 Colostomy-free survival is the endpoint in

10 the management of anal carcinoma, the clinically

11 relevant endpoint, and, in fact, would also apply

12 to a certain subset of patients with rectal cancer,

13 but only to patients that have very low-lying

14 tumors. And so for patients with colon cancer or

15 for rectal cancer above the very distal several

16 centimeters, this was not felt to be a helpful

17 endpoint in colorectal cancer.

18 Dr. Tom Fleming from the University of

19 Washington then gave a very articulate and, in

20 fact, I would say, impassioned presentation

21 regarding surrogate endpoints and non-inferiority

22 trials. He pointed out, as you've already heard,

250

1 that primary endpoints for drug regulation need to

2 be sensitive, measurable, and clinically relevant,

3 with the accepted endpoints or measures of clinical

4 benefit being improvement in survival or decrease

5 in tumor-related symptoms.

6 He then went on to discuss surrogate

7 endpoints, pointed out that biological activity

8 might be reflected in a surrogate endpoint, but

9 that might not establish clinical benefit for

10 patients. He gave a couple of examples from the

11 cardiovascular literature where flecainide and

12 other agents were used as antiarrhythmics and can

13 prevent ventricular tachycardia. The surrogate

14 endpoint was improved but, unfortunately, was

15 associated with a high risk of sudden death and so

16 that the overall benefit of these agents was

17 abrogated by the delayed and unexpected toxic

18 effects.

19 He stated that meta-analyses were really

20 required to adequately validate a surrogate

21 endpoint. That is, even if you had had a study or

22 two where there was a significant association

251

1 between a surrogate--progression-free survival, for

2 example, and survival--that you really needed to

3 have a cadre of studies to evaluate that

4 relationship in several different venues to be

5 certain that the surrogate was truly predicting

6 clinical benefit. And he pointed out that such

7 surrogate markers that are adequately validated are

8 distinctly rare in clinical medicine, but I believe

9 it was Dr. Williams that pointed out that the FDA

10 has granted approval using surrogate endpoints that

11 haven't been formally validated, and we saw some

12 examples just a few moment ago in Dr. Ibrahim's

13 presentation.

14 Tom then went on to discuss

15 non-inferiority trials, and I won't belabor this

16 point except to say that there are very important

17 methodological factors that need to be taken into

18 consideration, that it's not enough for the curves

19 to overlap, you need to be certain that you're not

20 allowing a significant decrease in therapeutic

21 effect in these non-inferiority trials. And there

22 was some lack of enthusiasm in general on the part

252

1 of the panelists that these studies might not truly

2 move the field forward. And the main area of

3 interest for non-inferiority trials is if there was

4 a treatment that was substantially less toxic than

5 the current standard.

6 Then we move on to the last two

7 presentations, which focused on time to tumor

8 progression and disease-free survival for the

9 adjuvant situation, respectively. This

10 presentation was given by Dr. Langdon Miller. He

11 discussed clinical benefit or the time to tumor

12 progression as a clinical benefit endpoint for

13 first-line metastatic colorectal cancer. And Dr.

14 Miller took the point of view of making a very

15 strong case for time to tumor progression.

16 He pointed out that in colorectal cancer

17 now, in contradistinction to years gone by, we have

18 multiple therapies that do have benefit in this

19 disease and that it has, therefore, become more

20 difficult to assess the impact on survival because

21 of these effectiveness therapies that can have an

22 impact on second-line treatment. Second-line

253

1 treatment can be effective and thereby obscure the

2 relationship between the initial treatment and the

3 ultimate survival of the patient.

4 He argued against the use of symptomatic

5 progression or time to symptom progression because

6 these patients frequently aren't symptomatic to

7 start out with. It's very subjective and difficult

8 to measure.

9 He stated from his point of view that time

10 to tumor progression should be a valid endpoint for

11 full approval in first-line colorectal cancer

12 because this endpoint directly evaluates changes in

13 the disease burden, that is, regression of tumor or

14 lack of progression of tumor; correlates with other

15 outcomes and, in particular, survival, and I'll

16 show you some data on this point in just a moment.

17 It has the big advantage that it's not confounded

18 by subsequent therapies. Second-line treatment

19 won't affect the time to tumor progression, and he

20 made that point that it offers utility as an

21 endpoint in non-inferiority trials because the

22 sample sizes that would be required would be much

254

1 smaller.

2 He made the point that the endpoint can be

3 objectively quantified, reviewed, and audited by

4 external panels; if you're doing radiographic

5 procedures to document the lack of progression at a

6 particular point in time, offered clear

7 interpretation, straightforward analysis, and

8 certainly would conserve patient resources and

9 hasten drug development.

10 The data that he presented correlating

11 time to tumor progression and survival in

12 first-line metastatic colorectal came from two

13 clinical trials involving 1,000 patients treated

14 with irinotecan-based chemotherapy. And so he had

15 primary patient data for these 1,000 patients and

16 found a very strong correlation between time to

17 tumor progression and overall survival among these

18 1,000 patients. A Cox analysis was performed,

19 plugging in all of the important prognostic

20 discriminants, and time to tumor progression was

21 still strongly associated with survival.

22 Secondly, in the questioning session, we

255

1 asked whether any meta-analyses had been done of

2 this surrogate endpoint, and the response was that

3 one meta-analysis has been performed involving the

4 published summary results of 29 trials involving

5 some 13,000 patients, where there was, again, a

6 highly significant correlation between time to

7 tumor progression and survival, but this was not

8 with the primary individual patient data.

9 This presentation generated a lot of

10 discussion among the panelists. There was some

11 concern expressed that there really needs to be a

12 very objective and reliable methodology for

13 assessing time to tumor progression. It's not

14 nearly as definitive as patient survival. But it

15 was felt that with modern radiologic techniques and

16 external review committees and properly written

17 protocols, that particular barrier could be

18 addressed.

19 There was a lot of discussion regarding

20 whether time to tumor progression reflects clinical

21 benefit in its own right per se, and here I'd say

22 that there was a big disagreement. There was not

256

1 consensus on the part of the panel.

2 For example, if a patient is asymptomatic

3 and has metastatic disease, his time to tumor

4 progression is perhaps prolonged by a month or two,

5 but he experiences very severe toxicity as a result

6 of the chemotherapy, and the overall survival is

7 not really changed. How much of a benefit is that

8 to the patient? And there was, therefore, not a

9 consensus on that particular point.

10 Is time to tumor progression reasonably

11 likely to predict clinical benefit based upon the

12 association between TTP and survival? I think the

13 majority of the panel would feel that would be the

14 case, but also that a more complete gestalt

15 regarding the patient and the clinical circumstance

16 need to be taken into consideration. We heard that

17 comment earlier today. We're interested in the

18 response rate, the survival, the toxicity pattern,

19 in addition to time to tumor progression to really

20 make an overall assessment of the benefit of the

21 treatment for the patient.

22 Then, finally, Dr. Dan Sargent, who is

257

1 here today, gave a preliminary analysis of the

2 correlation between three-year disease-free

3 survival and overall survival as endpoints in

4 evaluating adjuvant therapy for colon cancer. Dan

5 presented the results involving some 12

6 prospectively randomized, Phase II clinical trials

7 in patients with resectable colon cancer. There

8 are some 38 treatment arms involved in these 12

9 trials and more than 10,000 patients involved. And

10 he did have primary data.

11 The preliminary conclusions of this

12 presentation showed a rather striking correlation

13 between three-year disease-free survival and

14 five-year overall survival. The event rates, that

15 is, the number of relapses in three years or the

16 number of deaths within five years, was virtually

17 identical so that whether you used one endpoint or

18 the other, this did not have a significant impact

19 on the sample size. He found that three-year

20 disease-free survival may slightly overestimate

21 differences in five-year overall survival,

22 particularly in the experimental arms of these

258

1 randomized trials. And there were three of the

2 studies where there was a statistically significant

3 difference in three-year disease-free survival at

4 borderline p values, in the range of 0.03 to 0.04,

5 but no significant difference in five-year overall

6 survival. But the point was made that some of

7 these trials were not adequately powered to detect

8 differences in overall survival. And it was also

9 pointed out, I believe by Dr. Fleming, that this

10 was not a formally validated surrogate.

11 So Dan presented those results as a work

12 in progress and in just a moment will be presenting

13 an update of this analysis where he's done

14 considerable additional work since that time.

15 There was discussion among the panel as to

16 whether three-year disease-free survival

17 represented clinical benefit per se in its own

18 right, and there were a number of individuals there

19 that felt this was the case, independent of

20 survival effect.

21 I guess I would express a personal concern

22 that survival also would need to be evaluated in

259

1 these studies to be certain that there wasn't a

2 delayed adverse impact on survival that wouldn't be

3 seen until some delayed point in time, so you

4 wouldn't want to trade a significant benefit in

5 three-year disease-free survival for a significant

6 decrement in long-term survival. And it was also

7 pointed out that disease-free survival is used for

8 full approval in breast cancer adjuvant therapy.

9 Why not in colon cancer?

10 And so, then, to conclude, we bring these

11 questions for your consideration today. Should the

12 following endpoints be recommended to the FDA for

13 new drugs in colorectal cancer? And if so, should

14 they be for full or for accelerated approval?

15 In the colon adjuvant setting, is

16 three-year disease-free survival an appropriate

17 regulatory endpoint? There was considerable

18 feeling expressed at the workshop that this would

19 be the case.

20 For first-line metastatic colorectal

21 cancer, is time to tumor progression or

22 progression-free survival an appropriate endpoint?

260

1 And there was considerable feeling that it was

2 reasonably likely to correlate with clinical

3 benefit.

4 And in the rectal adjuvant setting, should

5 three-year local control, preventing the

6 devastating symptoms from local tumor recurrence be

7 a regulatory endpoint for new drugs being studied

8 in the rectal adjuvant setting?

9 Thank you very much.

10 DR. KELSEN: Thank you, Mike.

11 I think this is a very good time to go to

12 Dan Sargent and hear the update on his analysis.

13 DR. SARGENT: Thank you very much. I

14 appreciate the opportunity to present updated data

15 today from a meta-analysis exploring the question

16 of disease-free versus overall survival as an

17 endpoint for adjuvant colon cancer studies.

18 In the setting of colon cancer, it is

19 clear to impact and improve the chance of cure, we

20 must decrease the rate of relapse. Eighty-five

21 percent of deaths within eight years of diagnosis

22 are following a recurrence of the cancer, so

261

1 recurrent colon cancer is certainly the primary

2 cause of death in patients who are initially

3 thought to be able to be surgically cured.

4 In addition, due to the devastating

5 consequences of recurrence of disease, prolonging a

6 patient's time without disease certainly should

7 have beneficial impacts on their quality of life.

8 This led us to explore the following

9 hypothesis: that disease-free survival assessed

10 after three years is an appropriate endpoint to

11 replace overall survival in adjuvant colon cancer

12 trials. The benefits of such a change would be

13 clear. This would allow the more rapid completion

14 and the reporting of clinical trials and, if it

15 held true, would allow promising agents to benefit

16 patients more quickly.

17 In order to assess this question, we have

18 gathered data from multiple large, randomized

19 trials. We have individual patient data from every

20 trial, and the analyses started out, at least,

21 simple, comparing disease-free survival and overall

22 survival for study arms.

262

1 We've chosen landmark time points of three

2 years for disease-free survival and five years as

3 an endpoint for overall survival, and you'll see

4 why.

5 In addition to looking on an arm-by-arm

6 basis, we have looked within trials, looking at the

7 difference between the control arm and the

8 experimental arm of each trial to determine if

9 differences that are present on one endpoint are

10 translated over into the other endpoint. We feel

11 that the most important comparison is the

12 comparison of hazard ratio. That is, what is the

13 hazard ratio between a control and experimental arm

14 for disease-free survival? What is the hazard

15 ratio comparing control to experimental arms for

16 overall survival?

17 To make sure everyone is clear, we used

18 the following definitions: Overall survival is the

19 time from randomization to death due to any cause.

20 Disease-free survival is the time from

21 randomization to the first occurrence of either a

22 recurrent event or death. And we do note that

263

1 second primaries were not included as events in our

2 disease-free survival. Having said that, the rate

3 of second primaries is very low, and including them

4 has almost no impact on these analyses.

5 With respect to the validation of

6 surrogate endpoints, many methods have been

7 proposed, and there is no agreed-upon standard of

8 practice in the statistical community. Therefore,

9 we have chosen to examine multiple approaches

10 ranging from simple to complex.

11 The simple approach is to use a weighted

12 linear regression of one endpoint on the other,

13 weighting by the sample size of that trial.

14 Another approach--and I will explain each

15 of these approaches as I present them--is the

16 Prentice and Freedman approach looking at a

17 quantity known as the proportion explained. Two

18 other sets of authors--Begg and Leung, and

19 Burzykowski and colleagues--have proposed other

20 methods in Journal of Royal Statistical Society

21 recently, and I will explain those as they are

22 presented.

264

1 On this slide, the trials that are

2 included in the analysis are listed. We see that

3 they range from first accrual in 1977 all the way

4 down to 1994, so we do span a considerable amount

5 of time. Many of these trials had surgery-alone

6 control arms, but some of the later trials had

7 5FU-based treatments in all arms. Sample size

8 ranged from approximately 250 up to about 2,200;

9 total sample size, close to 13,000 patients, and a

10 total of 33 different treatment arms.

11 Among those 33 arms, there were nine that

12 were surgery-alone control arms and 24 that were

13 considered active treatments in that they were

14 5FU-based.

15 The median follow-up on these patients is

16 eight years, and we have complete data to five

17 years on 93 percent of patients. And there was

18 some inconsistency among these studies in long-term

19 follow-up, and, therefore, we have censored all

20 analyses at eight years because that was consistent

21 follow-up through eight years in these studies.

22 Just looking at the patient

265

1 characteristics, we can see that most of the

2 patients were between the ages of 50 and 70. About

3 15 percent were over the age of 70, so consistent

4 with the age distribution on clinical trials, but

5 probably skewed younger than the distribution in

6 the overall population.

7 We had about a 50/50 split on gender; 20

8 percent or so were treated with surgery alone, and

9 the majority of patients, 62 percent, were Stage

10 III. Stage I patients were included in one single

11 trial.

12 Turning to some data, here we've got the

13 recurrence rate by six-month intervals from the

14 date of randomization. If we add up adjacent

15 number of figures, we get recurrence rates by year.

16 So we can see that in the first year following

17 randomization, approximately 10.5 percent of

18 patients recur; in the second year, it's actually

19 the highest recurrence rates; adding these numbers,

20 you get 12.5 percent; and about 7 percent in the

21 third year. After that, the rate of recurrence

22 falls off rather steeply.

266

1 So, really, the dominant force in

2 recurrent disease happens in the first three years

3 following randomization.

4 We then looked at the time from occurrence

5 to death, and consistent with data that we've known

6 for a long time on advanced colon cancer, we have a

7 median time from occurrence to death of about a

8 year. And so patients that recur by three years

9 very likely will have died by five years.

10 We then looked at the rate of agreement on

11 a per patient basis for these two endpoints. So

12 what's shown here--and notice the scale does not go

13 to zero; it's magnified to show additional

14 detail--is the concordance rate between a

15 disease-free survival endpoint at x years, where x

16 ranges from one up to five, and the overall status

17 at five years.

18 So what does this mean? If we look at the

19 three-year time point, we see approximately 90

20 percent agreement between your disease-free

21 survival status at three years and your overall

22 survival status at five years. And we can see that

267

1 this curve climbs for the first year or two

2 following randomization, but that it really

3 plateaus at about the three-year time point.

4 So that suggested that three years is an

5 appropriate time point to look, and here shown

6 graphically is the simple rate of disease-free

7 survival at three years compared to overall

8 survival at five years. And, again, notice that

9 these scales are magnified to show additional

10 detail. They do not go from zero to one. If they

11 did, you'd just see this little crowd in the

12 middle. So we blew them up.

13 Spearman correlation is 0.89, R-squared

14 from our regression is 0.86, both measures

15 indicating significant concordance between these

16 two effects.

17 Our regression equation result was that

18 overall survival is, in essence, zero plus one

19 times three-year disease-free survival. Looking at

20 the p values, we see that the intercept is not

21 significantly different from zero. The slope is

22 significantly different from zero, but it's not

268

1 significantly different from one. And so

2 statistically we cannot reject the simple equation

3 that five-year overall survival equals three-year

4 disease-free survival.

5 Within each of the study arms, 33

6 different study arms, the largest difference in

7 absolute numbers was 6 percent between disease-free

8 survival and overall survival. In 27 of the 33

9 arms, the difference between these two endpoints

10 was 3 percent or smaller.

11 So the first set of conclusions is that on

12 a patient-by-patient basis, three years does seem a

13 reasonable time point to look. The recurrence rate

14 is higher in the first three years and then falls

15 off. The survival following recurrence is about a

16 year. And the per patient concordance reaches its

17 peak at about three years and then plateaus. And

18 on an arm-by-arm basis, three-year disease-free

19 survival from regression modeling is an excellent

20 predictor of five-year overall survival.

21 Perhaps more importantly, we're interested

22 in the question of: Does a comparison of study

269

1 arms using disease-free survival reach the same

2 conclusion as if we used overall survival? Because

3 that's what we really want to know: Does a new

4 treatment do better than an old treatment?

5 In order to do this, we attempted to

6 actually mimic the conduct of the clinical trial

7 because at three years of minimum follow-up--some

8 patients have been on the study for longer than

9 that. Studies take two or three years to accrue,

10 and so three years after the last patient is

11 registered, some patients will have been followed

12 for four years or five years. And so because we

13 had the individual patient data, we attempted to

14 replicate the analysis that would have been

15 completed at the three- and five-year time points

16 to try to answer the question: What if we did the

17 analysis at the time that the analysis would have

18 been done, not retrospectively?

19 We also have started to perform analyses

20 at three years median follow-up as opposed to

21 minimum follow-up, and the conclusions we're

22 reaching are very similar on those endpoints. But

270

1 that is not a completed work.

2 Perhaps the most important slide, at least

3 in my opinion, this plots the hazard ratios

4 comparing control arm to the experimental arm in

5 each study. On the x axis is the disease-free

6 survival hazard ratio compared to the overall

7 survival hazard ratio. We see a very tight

8 concordance. Again, we have a Spearman rank

9 correlation of 89 percent, and the R-squared from

10 our regression is 0.87, indicating a tight and

11 consistent relationship between hazard ratios for

12 disease-free and overall survival.

13 The regression equation is that the

14 overall survival hazard ratio is 0.09 plus 93

15 percent times the disease-free survival hazard

16 ratio. If we look at the parameter estimates, we

17 will again see the intercept is not significantly

18 different from zero. The slope is significantly

19 different from zero, but not significantly

20 different from one. So, again, we cannot reject

21 that the hazard ratio for overall survival equals

22 the hazard ratio for disease-free survival.

271

1 To translate this into some real numbers,

2 I have the panel on the right, where what's given

3 is suppose we see a hazard ratio for disease-free

4 survival of 0.6. What does that suggest for a

5 hazard ratio for overall survival? And the

6 translation is 0.65, and it ranges across the

7 values of disease-free survival hazard ratios that

8 we might see. And what we see is the regression

9 equation suggests a slight attenuation of hazard

10 ratios from the disease-free survival to the

11 overall survival towards one. But it's a slight

12 attenuation, on the order of about 10 to 15

13 percent. And I'll describe this grade in more

14 detail.

15 For the statisticians in the audience, now

16 I get to have some fun, if that's your idea of fun.

17 Looking at some formal measures of

18 surrogacy, the proportion explained was proposed by

19 Freedman in 1992 as follow-up work to work by

20 Prentice in 1989. In essence, this approach fits

21 two Cox regression survival models--one without the

22 surrogate endpoint included, one with the surrogate

272

1 endpoint included.

2 If the surrogate is truly related to the

3 outcome of interest, the surrogate should explain

4 most of the variability in that model. And so what

5 they propose to do is look at the proportion of

6 treatment effect explained by the surrogate, and if

7 the surrogate explains close to 100 percent of the

8 variability, if they presume the surrogate, that

9 would imply that it is a good surrogate.

10 This measure has been criticized by

11 several authors for many reasons, one of which is

12 that it's not actually a true proportion and it's

13 not bounded between zero and one. Nonetheless,

14 this is probably the most common method used, and

15 so we fit that to this data set.

16 Here are the results from the two models.

17 Looking first without disease-free survival as an

18 endpoint, as a surrogate in the model for overall

19 survival, this is the log hazard ratio indicating a

20 very significant benefit for treatment when

21 disease-free survival is not included a parameter

22 in the model. When disease-free survival is

273

1 included as a parameter in the model, the p value

2 becomes non-significant, indicating that the

3 disease-free survival is explaining almost all of

4 the variability in the endpoint.

5 When you calculate the proportion, you

6 actually come up with 138 percent, validating the

7 criticism of this measure that it's not a true

8 proportion. Nonetheless, this does imply that

9 disease-free survival may be a good surrogate for

10 overall survival.

11 A more sophisticated approach was

12 recommended by Burzykowski and colleagues in 2001,

13 where they fit a bivariate copula survivor model,

14 which, in essence, fits the survival model

15 using--examines the effect of a set of covariates

16 on both endpoints of interest. And if the effect

17 of the covariates on both endpoints is similar,

18 that suggests that the two endpoints themselves are

19 similar. And so they defined two measures, a trial

20 level R-squared to look at the concordance between

21 endpoints on a trial-by-trial level, and an

22 individual level to look at the per patient

274

1 concordance. And values close to one for both

2 measures indicate surrogacy.

3 The results applied to this data set have

4 an R-squared value of 0.85, confidence interval of

5 0.72 to 0.99, and at an individual level we have a

6 concordance measure of 0.9.

7 How to interpret these results. In the

8 paper that Burzykowski and colleagues published,

9 they had an example from ovarian cancer, and they

10 actually had values of R-squared and TAO very close

11 to these values in their example. And their

12 conclusion was that it seems plausible to conclude

13 that this is a valid surrogate given the values

14 that we see here.

15 This is a graphical representation of that

16 method looking at the impact on disease-free

17 survival time compared to the impact on overall

18 survival time. These are log hazard ratios. The

19 size of the circle is proportional to the sample

20 size of the trial. Again, we see high, tight

21 concordance between the two measures using this

22 sophisticated model.

275

1 Finally, an approach I really tend to

2 prefer, Begg and Leung gave a very simple measure.

3 The validity of a surrogate endpoint should be

4 judged by the probability that the trial results

5 based on the surrogate endpoint alone are

6 concordant with the trial results that would be

7 obtained if the true endpoint were observed and

8 used. Simple, straightforward, do they give the

9 same conclusion? Who needs fancy statistics?

10 Of 18 total within-trial comparisons, we

11 compared the two arms using the endpoint of

12 disease-free survival, using the endpoint of

13 overall survival, log rank testing. Straightforward, as

14 simple as we can get.

15 Of the 18, 16 gave the same conclusion

16 regardless of which endpoint you used. Eleven

17 trials had no difference between the two arms for

18 either endpoint; five had significant differences

19 between arms for both endpoints.

20 There were two trials that were

21 significant only for disease-free survival, but

22 both of these had p values of 0.03, so only

276

1 marginal significance for disease-free survival.

2 That is shown graphically on this slide

3 where for each of the 18 trials we have plotted in

4 yellow the disease-free survival estimate and

5 confidence interval, and in blue the overall

6 survival confidence interval and estimate. And as

7 you go in sets of two, you'll notice how similar

8 within a trial the confidence interval and the

9 estimates are for these two endpoints.

10 If you look more closely, you will see

11 that most of the time the blue dot is a little

12 closer to one than the yellow dot. What does that

13 mean? It means that we have a slight attenuation

14 of the effect, that the hazard ratio for

15 disease-free survival is a little bit farther away

16 from one than the hazard ratio for overall

17 survival. But, again, this attenuation is very

18 slight, and that's consistent as you go down the

19 plot.

20 Focusing on two trials in particular, one

21 of the comparison within the trial, NSABPC-04, and

22 the other was an NCCTG trial from 1978, these were

277

1 the two trials where the disease-free survival

2 hazard ratio, confidence interval, you can see that

3 it excluded one, and it was significant. And for

4 the overall survival, it included one, thus was not

5 significant. And that's the same here. But you

6 can see in both cases the disease-free survival

7 hazard ratio got very close to one, and the overall

8 survival hazard ratio hardly excluded one. So the

9 results are really consistent with each other, and

10 what we ran into was just a little bore edge effect

11 there.

12 So the second set of conclusions. As an

13 endpoint for comparison, the hazard ratio for

14 disease-free survival is an excellent predictor of

15 the hazard ratio for the overall survival with a

16 slight attenuation. Marginally significant

17 improvements in disease-free survival may not

18 translate into overall survival. The formal

19 measures that we have assessed do suggest surrogacy

20 is appropriate for these two endpoints.

21 How to translate this into something that

22 can be helpful and useful to the practicing

278

1 clinician--at least I hope. Let's suppose in a

2 trial of 2,000 patients we observed a disease-free

3 survival hazard ratio of 0.8. Using our model, you

4 can translate this into a predicted hazard ratio

5 for overall survival of 0.84. So we see that

6 slight attenuation.

7 In addition, we can compute a 95-percent

8 predicted interval for the hazard ratio for overall

9 survival. In this case, it would go very 0.77 to

10 0.91. So in this case of a trial of 2,000, if you

11 observe 0.8, you're 95-percent prediction interval

12 for overall survival would exclude one.

13 We can do this not only for a value of

14 0.8, but we can do it for any observed value of

15 disease-free survival and calculate bounds like

16 this. Now, of course, that depends on the sample

17 size from your trial. This example used 2,000

18 patients. If we instead use 1,000 patients--and

19 these red lines are a little bit hard to see, but

20 we can see that the lines fall outside the lines

21 for 1,000. They get wider. The prediction

22 interval is wider. And if you have a trial for

279

1 3,000 patients, the blue bands get narrower. So

2 based on the sample size, we can see how sure we

3 can be about our prediction.

4 Now, suppose we want to be sure--and I

5 apologize, these red lines show up much better on

6 my screen than they do here. Maybe I'll just go to

7 the yellow line. Okay.

8 Suppose that we want to ensure that our

9 overall survival hazard ratio--that the prediction

10 interval for our overall survival hazard ratio

11 excludes one. What we can do is go across the line

12 and come down, and notice that if our observed

13 disease-free survival hazard ratio is less than

14 0.90, our predicted interval for our overall

15 survival hazard ratio will exclude one. And, also,

16 for the case of 3,000, you can just calibrate it as

17 you see fit.

18 In order to test the validity of this

19 model, we have performed leave-one-out

20 cross-validation. What does that mean? It means

21 of the 18 comparisons, we took one out at a time,

22 fit the model to the 17 that remained, used the

280

1 data from those 17 to predict what happens in the

2 trial that we did not include in our model, and

3 then see if the model based on the 17 predicts well

4 on the one that we did not include their data.

5 Shown here in the blue dots are the

6 predicted hazard ratio for overall survival based

7 on disease-free survival. In the red are the

8 actual results. And we can see that for 17 of the

9 18 trials, the actual result fell with the

10 95-percent prediction intervals. This is exactly

11 what we would expect. If we have 18 and we're

12 computing 95-percent confidence intervals, one of

13 them should fall outside. Here it is. It fell

14 outside, but just by a little bit. Also notice

15 that sometimes the actual, which are the red, are

16 above the blues; sometimes they're below the blues.

17 So this indicates that the model is calibrated well

18 and is predicting accurately.

19 Turning to a few points for discussion, we

20 did have individual patient data from all of our

21 trials. All of these trials used 5FU-based

22 regimens. They did includes a mixture of Stage III

281

1 and Stage II patients. Our preliminary work

2 suggests that the concordance is somewhat stronger

3 for Stage III than it is for Stage II. That's not

4 surprising because recurrences would happen more

5 quickly in Stage III than in Stage II. But we're

6 doing further analyses on that point, but we do

7 feel that for trials similar to those that were

8 included here, which included a mixture of Stage II

9 and Stage III, these results should be relevant.

10 I think open for discussion is issues

11 about how relevant this is to the current practice.

12 For example, we now have more advanced--more

13 effective therapies available in the advanced

14 disease setting. We've improved the median

15 survival from 12 months to 18 to 20 months. Having

16 said that, most people who recur by year three

17 still die by year five.

18 In addition, we have improved methods for

19 detection of occurrence with improved imaging

20 techniques, so perhaps recurrences are being

21 detected earlier in a more curative state.

22 I think it's also open for discussion what

282

1 about non-cytotoxic or targeted agents. Perhaps

2 instead of preventing a recurrence, perhaps just

3 delay a recurrence. And if such agents are

4 available, the concordance between these endpoints

5 could decrease. Or maybe we just need to look at

6 different time points.

7 Conclusions are that for the studies we've

8 examined, disease-free survival is an excellent

9 predictor of overall survival. It meets most

10 formal definitions of surrogacy. There is a modest

11 attenuation of treatment effect between these two

12 endpoints on the order of 10 to 15 percent. And

13 the model allows prediction of the benefit on

14 overall survival based on what we observe for

15 disease-free survival.

16 I close by acknowledging my many

17 collaborators from around the world on this project

18 and put in a plug. This is still not the final

19 analysis that will be done. We have recently in

20 the last few weeks gained data from three

21 additional large trials. It will be included in

22 the analysis to be presented at ASCO in June of

283

1 this year, and I did want to note that I did

2 receive permission from ASCO to present this

3 material at this meeting today.

4 Thank you very much.

5 DR. KELSEN: Thank you, Dr. Sargent.

6 So at this point we'll open the floor for

7 questions to any of the four presenters, questions

8 from the panel.

9 DR. GEORGE: You started off giving a nice

10 presentation of why three-year disease-free

11 survival, how it relates to five-year overall

12 survival, and then you seemed to kind of drop that

13 as you got further into it, talking about hazard

14 ratios, which I assume were based on estimates from

15 the whole data, not just restricted to three years

16 and five years. Is that true?

17 DR. SARGENT: No, they were--the hazard

18 ratios for disease-free survival were calculated

19 using only the data from the first three years.

20 DR. GEORGE: And despite your elegant

21 arguments for why that seems to work well, why

22 would you not use all the data? Is this just a

284

1 pragmatic thing that you want to be able to predict

2 it earlier?

3 DR. SARGENT: The question was not so much

4 from an academic standpoint as it was--we have

5 to--we took a pragmatic approach. We have to do an

6 analysis at a certain time point. You analyze a

7 trial, and the question is when can we analyze a

8 trial. And really, the goal is: Can we analyze a

9 trial more quickly?

10 And so if we're able to analyze a trial

11 earlier and reach the same conclusions if we

12 analyzed it later, that was something that we

13 sought to show.

14 DR. GEORGE: But the thing that's still

15 puzzling me about this is the three-year

16 disease-free survival is sort of--it's like a per

17 patient analysis, is it not? That is, you're

18 really looking at, on each patient, whether--what's

19 the chance of making it to the three years

20 disease-free survival. It's not three years

21 calendar time from the time you start the study,

22 because they're two different things.

285

1 DR. SARGENT: That is correct. For the

2 first set of analyses, which were on the per

3 patient basis, it was to establish is three years a

4 reasonable time point to look within a patient.

5 Then we turned our attention to what happens when

6 you actually analyze the trial, and you have to

7 analyze the trial at a certain time point, and we

8 chose what if we analyzed the trial at the

9 three-year time point using data from all the

10 patients, and if the patient had four years because

11 they were entered earlier, taking advantage of that

12 data.

13 So the first set of analyses was really to

14 establish that three years is a sensible time point

15 to look on a per patient basis, and then that was

16 supported then later by is three years a sensible

17 time point to look on a per trial basis.

18 DR. KELSEN: Dr. Brawley?

19 DR. BRAWLEY: Dr. Sargent, I want to

20 congratulate you. I just sat through a statistics

21 lecture, and I actually think I understand it.

22 Maybe I need a head CT.

286

1 [Laughter.]

2 DR. BRAWLEY: The question is: You make a

3 very strong argument for use of disease-free

4 survival as a surrogate for overall survival when

5 using anti-neoplastic agents. Can you speculate on

6 how well this model would translate if we were to

7 start looking at things like growth factor

8 inhibitors, where instead of looking at

9 disease-free survival we would be looking at things

10 like progression-free survival?

11 DR. SARGENT: I really don't feel

12 comfortable extrapolating beyond the range of the

13 data and the trials that we included in the

14 analysis. Thank you.

15 DR. KELSEN: Dr. Redman?

16 DR. REDMAN: Something along similar lines

17 to that. With the model that you have set up, if

18 we now go out five years from now--and I don't know

19 what's going to be happening, but if we now know

20 that the median survival for advanced or recurrent

21 colorectal cancer goes out to 24 or 30 months, do

22 you think this model will still hold? Or the other

287

1 question also is if we then push the time to

2 relapse out with therapies.

3 DR. SARGENT: Regarding the first

4 question, I think the model will be less sensitive

5 to that because the magnitude--the advances that

6 have been made in advanced colorectal cancer are

7 wonderful. In absolute magnitude, they're still

8 modest. And so if we increased the median survival

9 from one year to two years, the reality is everyone

10 who recurs year one, everyone who recurs year two,

11 and most of the people who still recur in year

12 three will still have an unfortunate outcome of

13 death by year five. And so I think we need to have

14 a pretty profound impact on survival in the

15 advanced disease setting to translate into this

16 model.

17 Having said that, it may be that we have

18 to look at a later time point, and I think we have

19 the opportunity with the collaboration that we've

20 established and the data that we have, we've

21 already been pledged to have data from some of the

22 new trials when they become available. I think, of

288

1 course, the challenge is the five-year data is not

2 available from those trials.

3 With respect to the second question, I

4 think that's more up in the air. I think if

5 recurrences are simply delayed as opposed to

6 prevented, then I--and recurrences start happening

7 more frequently out in the fourth year and the

8 fifth year, then I think that could have some

9 pretty serious consequences to the validity of this

10 model, and it would need to be assessed again for

11 those different agents.

12 I think Ross Prentice in his seminal work

13 on this topic has made the point that a surrogate

14 is really relevant and related to the treatments

15 that are being used, and if treatments are used

16 that have different mechanisms of action or

17 influence the endpoints in different ways, then the

18 surrogate endpoint that had been previously

19 validated may not be considered valid anymore and

20 would need to be re-validated for that new set of

21 agents.

22 DR. KELSEN: Dr. Martino?

289

1 DR. MARTINO: Two questions, I think both

2 to Dr. Sargent, but the rest of you may chime in.

3 First of all, I'm not sure that I now

4 understand when you say three years, what we're

5 counting from. So explain that to me first.

6 DR. SARGENT: Okay. We've done two sets

7 of analysis. One is on a per patient basis, and

8 that is three years from the date that they are

9 enrolled in the trial

10 DR. MARTINO: Okay.

11 DR. SARGENT: The second set of analysis

12 is on the trial-by-trial basis, and that is doing

13 an analysis--presuming that we perform our primary

14 analysis at the time point three years after the

15 first--excuse me, after the last patient is

16 enrolled. So we have three years minimum follow-up

17 on all patients, but some patients may have four

18 years, some patients may have five years, because

19 they entered the trial earlier.

20 DR. MARTINO: And so which of those are

21 you advising to this group? That's what I'm not

22 clear on.

290

1 DR. SARGENT: Okay. I think what's

2 relevant primarily for this group--I'm not advising

3 anybody. I'm just presenting data. But I think

4 from this committee's perspective, if I was sitting

5 on the committee, you see data presented by a

6 sponsor that compares two trials arms--a control

7 arm and an experimental arm--and that's an analysis

8 that's done with a specific endpoint at a specific

9 time point. And I would suggest, based on this

10 data, that for the type of agents that have been

11 explored in this analysis, an analysis that's

12 presented on disease-free survival three years

13 following the entry of the last patient on study is

14 an excellent predictor of an analysis that may be

15 subsequently presented to this committee at a time

16 point five years after the last patient is entered

17 and on an endpoint of overall survival.

18 DR. MARTINO: Now, the other way that I've

19 seen this type of data presented, predominantly in

20 breast cancer, is that one actually specifies how

21 many events you want to see, and then when those

22 have occurred, you use that as the time point at

291

1 which you compare two arms. I need your thoughts

2 on that way of doing things.

3 DR. SARGENT: Okay. That's an excellent

4 point.

5 The time points are all related to the

6 minimum duration of follow-up, and what this, in

7 essence, presumes, if I was a sponsor organizing a

8 trial, I would design my trial and design my

9 hypothesis tests so that the number of events

10 necessary to provide my power became available at

11 the time we projected the last patient would have

12 been followed for three years. And so this is all

13 presuming that we have enough events to power our

14 trial appropriately.

15 DR. MARTINO: Okay. And that gets me to

16 my final question. As has happened in breast

17 cancer in the adjuvant setting, I anticipate a

18 similar behavior will occur in colon cancer, which

19 is that as you have a few more agents that appear

20 to work in the metastatic setting, you now to start

21 to ask adjuvant questions in patients with lesser

22 and lesser disease. And so presumably now a

292

1 three-year endpoint, however one defines that

2 three-year, might encompass most of the

3 recurrences. But as you start to look at lesser

4 and lesser disease, that three-year won't quite be

5 the same. You may have to wait for five years for

6 patients with little disease to recur for you to

7 then capture 80 percent or whatever percent of them

8 you want.

9 So if I'm understanding that correctly,

10 then whatever decision is made today may be less

11 applicable with the passage of time, and that time

12 might be even two years from now.

13 Do you understand my question? Am I

14 making sense?

15 DR. SARGENT: Yes, absolutely. So to

16 comment on that, I think that the results that I've

17 presented today are relevant to trials that would

18 be conducted with a similar patient population as

19 were included in these trials. And these trials

20 included a mixture of Stage II and Stage III

21 patients. It was actually quite consistent, about

22 a 60/40 to 50/50 split between Stage II and Stage

293

1 III's. I think in a study of just Stage II

2 patients, these particular data may be less

3 relevant. However, we have individual patient

4 data. We are performing the analyses in just the

5 Stage II patients and in just the Stage III

6 patients to see if the concordance is as strong in

7 each group. And as I stated, our preliminary

8 results are that the concordance is stronger in the

9 Stage III patients than it is in the Stage II

10 patients. But having said that, if a trial has

11 about this mix of patients, I think these results

12 would hold valid.

13 DR. KELSEN: Ms. Roach?

14 MS. ROACH: First of all, I saw this

15 presentation in November, and it was fascinating to

16 see the work that was done since then. So thank

17 you.

18 I have two questions, and both of them are

19 pretty straightforward. If and when someone comes

20 forward with a proposal for a trial for, say,

21 Avastin and a 5FU-based regime for Stage III

22 patients to delay and/or prevent recurrence, then

294

1 are you saying that this--they might be able to

2 think about disease-free survival, but really,

3 overall survival would need to be the endpoint on

4 that because of the use of a biologic.

5 DR. SARGENT: Well, I think I have a

6 comment and a response.

7 The comment is that I think it's up for

8 this committee to decide two questions. One is:

9 Is disease-free survival a surrogate for overall

10 survival? But I think the other question is: Is

11 disease-free survival an important endpoint on its

12 own regard, irregardless of its relationship with

13 overall survival?

14 And so if this committee feels that

15 disease-free survival is an important endpoint on

16 its own, then I think the question becomes less

17 relevant.

18 With respect to, though, if the endpoint

19 of disease-free survival is only felt to be valid

20 due to its surrogacy or due to its relationship

21 with five-year overall survival, then I do not

22 believe this data would provide a support for that

295

1 surrogacy to hold with this different class of

2 agents.

3 MS. ROACH: Okay. And then my second

4 question is: Can you keep this going? You've

5 started something really (?) here, and so how do

6 you keep it so that in ten years it's not

7 completely useless?

8 DR. SARGENT: Well, we've already

9 established collaborations with many investigators,

10 including the new trials that have been done with

11 irinotecan and oxaliplatin in the adjuvant setting.

12 Both of those investigative groups have agreed to

13 participate in this analysis, and so we'll be able

14 to update our analysis there. And at that time,

15 that's all that there is out there. The biologics

16 are just entering the adjuvant trials, and so it

17 will be, you know, eight years really until that

18 data is available, presuming they accrue for three

19 years and have five years additional follow-up.

20 So I think that those questions are very

21 relevant; however, I don't anticipate this

22 committee would be seeing any such data for quite

296

1 some time.

2 DR. KELSEN: Dr. Brawley had a follow-up

3 question.

4 DR. BRAWLEY: Yes, part of which has been

5 answered. Dr. Sargent would you agree with the

6 point that the correlation between disease-free

7 survival and overall survival is a much tighter

8 correlation than, say, as you apply years to it,

9 especially when you look at the stage issue?

10 What I'm trying to say, in short, in as

11 few words as possible, is as stage goes down, maybe

12 disease-free survival needs to go up. But it can

13 still maintain a good correlation with overall

14 survival.

15 DR. SARGENT: I think that there are two

16 factors that relate to that. One is, as the stage

17 goes down, the time to recurrence probably goes up.

18 The second is that, as the stage goes down, fewer

19 of the deaths are due to the cancer and more are

20 due to other causes. And so I think, A, the time

21 point may differ for earlier-stage cancers, that we

22 may have to look at three or four or five years

297

1 because recurrences are later. And, second, my

2 expectation--and I don't have data to support

3 this--is that the attenuation of the effect may be

4 larger due to a greater proportion of deaths due to

5 competing causes.

6 DR. KELSEN: It may be as stage goes down

7 that it's not time to recurrence changes; absolute

8 cure rate is higher, and the model for breast

9 cancer may not be 100-percent valid. And so time

10 to a non-cancer-related event may be much more of

11 an issue.

12 Other questions? Yes, Dr. Rodriguez?

13 DR. RODRIGUEZ: I know that we were

14 focusing on the analysis of correlation of

15 disease-free survival with overall survival, but I

16 also noticed that, you know, this covers a wide

17 range of time frame for the studies. And I noticed

18 that the design of the studies keeps shifting from

19 initially the control arm being surgery only, now

20 to arms using 5FU. So are you seeing a trend in

21 this meta-analysis for longer disease-free

22 survival, even as the complexity of the adjuvant

298

1 treatments increases? Is that true or not?

2 DR. SARGENT: Well, we have--two responses

3 to that, I guess.

4 First is that we have explored the

5 validity of the relationship over time. And has

6 the relationship between disease-free survival and

7 overall survival changed over time? The answer to

8 that one is no. That has been very consistent.

9 And, in fact, if you so desired, I could go and

10 show that on a slide, the slide of the hazard

11 ratios--maybe I don't have that. I think I do have

12 that slide in there, actually.

13 But related to your specific question, I

14 think, is have we seen over time the absolute

15 benefit, and we've actually tried to stay away from

16 such an analysis because that involves comparisons

17 of non-randomized arms to each other. Nonetheless,

18 that is something that we have observed, that the

19 survival rates for either disease-free or overall

20 survival from the trials performed in the early

21 1980s compared to the trials that have become

22 mature in the late 1990s, though the overall

299

1 survival and disease-free survival rates certainly

2 have improved over time, is that due to better

3 treatments? Is that due to better staging? Is

4 that due to better supportive care? Is that due to

5 better surgery? We don't know. And so we have

6 stayed away, actually, from making those sorts of

7 comparisons of absolute treatment effects over

8 time--excuse me, of absolute survivals over time.

9 What we've focused on is the treatment effect over

10 time, and is the treatment effect, comparing the

11 treatment arm to the control arm, consistent--which

12 it is.

13 I hope that answered the question. Thank

14 you.

15 DR. KELSEN: Dr. Cheson?

16 DR. CHESON: To ask a somewhat naive

17 question falling under the category of "we should

18 be so lucky," but if we were to develop a much more

19 effective treatment for relapsed patients, how

20 would that impact on this? And how do you take

21 into account the fact that this new therapy may

22 have some sort of interaction with the initial

300

1 therapy, either positive or negative? Meaning it's

2 going to work if you had x, but it's not going to

3 work if you had y.

4 DR. SARGENT: Well, I think it is

5 important to note that if we could triple the

6 survival following recurrence, I think that would

7 have an impact. I think, you know, in terms of the

8 time points we're looking at, we're looking at a

9 two-year window between the three-year time point

10 and the five-year time point.

11 Once we start pushing the median survival

12 following occurrence out past that two-year window,

13 then I think it could really have a bigger impact

14 on these results. We haven't seen that yet.

15 Hopefully we will.

16 With respect to interaction between the

17 treatment they received first and the treatment

18 they received subsequently, we only have data on

19 patients who were treated with a 5FU-based sort of

20 thing initially. And I guess I really can't

21 speculate as to if patients are treated with some

22 other sort of agent up front.

301

1 DR. KELSEN: I have a question for FDA,

2 for Dr. Ibrahim, related to using other tumors as a

3 model. In breast cancer, three-year disease-free

4 survival is recognized for approval of a new agent

5 in the adjuvant setting as opposed to colon cancer,

6 talking about today, and I think that you said it

7 was because breast cancer recurrences were

8 symptomatic. Does that apply to both hormonal

9 therapy as well as cytotoxic therapy? And is it

10 correct that the rationale for using three-year

11 disease-free survival in breast cancer was based on

12 the fact that women would be more likely to be

13 symptomatic from a recurrence than, say, a man or a

14 woman who has colon cancer? And is that still true

15 with modern imaging today?

16 DR. IBRAHIM: I'm not sure I can answer

17 that question. Maybe Rick or Grant--

18 DR. PAZDUR: Our opinion regarding breast

19 cancer, which occurred many, many, many years ago,

20 was based on the fact that it was believed that

21 these recurrences were symptomatic. Okay? Whether

22 one wants to believe that now or not believe it

302

1 with introductions of other imaging, closer

2 follow-up of patients, et cetera, is open to

3 discussion.

4 I don't know how much relevance that has

5 here because I would see that the vast majority of

6 recurrences from colorectal carcinoma, especially

7 as our follow-up of patients and our radiographic

8 imaging becomes better and more intense, that most

9 of these recurrences are not symptomatic. So it's

10 a little bit different situation. I don't

11 necessarily think we have to rely on that it's

12 occurred many years ago. I wouldn't use that as

13 any regulatory precedent that we use that basis,

14 because I don't even know if it would hold at this

15 time. Perhaps Silvana would like to comment on

16 recurrences and symptoms.

17 DR. MARTINO: Well, a couple of thoughts,

18 Rick, because this is one of my issues as well. My

19 impression is that sometimes the FDA has accepted

20 three-year disease-free events, but for the most

21 part, we tend to pilot things to five years, not to

22 three, when we do, you know, large, intergroup sort

303

1 of trials. So I'm not sure that three years is

2 where you've given most of the approvals in breast

3 cancer. I believe it is closer to five. Number

4 one. Okay?

5 But relative--so that's that. Okay.

6 Relative to patients becoming symptomatic,

7 I don't think that that biology has changed. When

8 a patient with breast cancer recurs, she generally

9 is symptomatic, because often what drives the

10 x-rays that you are going to do are, in fact,

11 symptoms. Very rarely is it something else.

12 I don't understand the biology of colon

13 cancer well enough--because this is what's going

14 through my mind, is I keep hearing several of you

15 who deal with colon cancer using this expression

16 that they're asymptomatic. And I'm assuming what

17 that means is they've got something in the liver,

18 for the most part, and it's not causing them a new

19 problem, though I'm not sure how you figured it out

20 that they had it in the first place. But there

21 must be some time point where they do become

22 symptomatic, and one of the things I need to

304

1 personally understand as I think about disease-free

2 survival as a valuable endpoint onto its own,

3 exclusive of survival, is if you do have this

4 asymptomatic colon in a recurrent patient, is there

5 some time span when you can say, well, within a

6 year most of them are going to be symptomatic,

7 anyway? Is there such an understanding

8 biologically?

9 DR. KELSEN: Dr. O'Connell and I will both

10 address that. I think Dan pointed out, first of

11 all, that the time to death from recurrence prior

12 to newer agents is about a year.

13 DR. MARTINO: But it's not time to death

14 I'm interested in--

15 DR. O'CONNELL: It's time to symptomatic

16 progression, and there have been studies done in

17 patients with known metastatic colorectal cancer

18 who were asymptomatic at the point of beginning the

19 observations, and this was worked on by Dr.

20 Moertel. The median time to progression is about

21 five months, and 80 percent of patients were

22 symptomatic within one year. But the median was

305

1 five months.

2 DR. MARTINO: So there is a reasonable

3 correlation there that if you recur within that

4 year, 80 percent will be symptomatic.

5 DR. O'CONNELL: Yes.

6 DR. MARTINO: So to me--and this becomes

7 the issue in terms of is overall survival the only

8 objective that we should be aiming for, because it

9 strikes me that if symptoms follow reliably to that

10 degree, that disease-free-ness is important.

11 DR. O'CONNELL: Yes.

12 DR. MARTINO: And is important all by

13 itself. The other is wonderful, but it doesn't

14 obviate that there's value in being disease-free

15 because you will become symptomatic within a

16 reasonable short period of time.

17 DR. O'CONNELL: I agree.

18 DR. KELSEN: Ms. Roach?

19 MS. ROACH: I have a follow-up question on

20 that. What is the typical timeline, in your

21 judgment, between symptomatic progression--the

22 development of symptoms and then death?

306

1 DR. KELSEN: I think Dr. O'Connell

2 commented on this. There are several trials, not a

3 large number of randomized studies of no treatment

4 versus immediate treatment, which gave the

5 symptoms, the Nordic trial and several others, and

6 the time frames were exactly what Mike said.

7 DR. O'CONNELL: Actually, Dr. Miller

8 presented some data at a workshop as well that if

9 one looked at the point in time from progression,

10 with advanced metastatic disease progressing, to

11 the time of death, it's about eight months, and

12 with salvage therapy out to 11 months or so. So,

13 again, there's a period of several months from the

14 time of developing symptoms until death. And I

15 guess it wasn't precisely symptomatic progression

16 that Langdon was talking about. It was any

17 progression. The median time was eight months from

18 the detection of any progression from metastatic

19 disease, whether symptomatic or not, and death.

20 And so presumably it would be shorter than that if

21 it was asymptomatic progression, coming back to

22 about the five- to six-month range again.

307

1 DR. KELSEN: Dr. Hirschfeld, you had a

2 question?

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

4 Sargent, and I, too, want to congratulate you on

5 the initiative of this very intriguing analysis.

6 Several of our colleagues around the table have

7 pointed out the potential limitations of the

8 analysis with regard to types of therapy, types of

9 products. We're particularly interested in

10 immunotherapies, among others. But we haven't yet

11 discussed alterations in how one measures

12 progression, and there have been a lot of

13 developments in looking at PET scans and other

14 types of imaging techniques, as well as other

15 potential techniques.

16 So to maintain the interest and to follow

17 Ms. Roach's suggestion of having this as an ongoing

18 project, what other types of analyses then would

19 you entertain or explore to look at disease-free

20 survival and overall survival, other than your

21 landmark analyses, which has been pointed out is a

22 shifting target already?

308

1 DR. SARGENT: Well, I think the trials

2 that were conducted and included in this analysis

3 were conducted in, for the most part, the pre-PET

4 era and had very protocolized follow-up. And so I

5 guess what further analysis would we conduct, I

6 think we would want to look now--once we look at

7 some new trials--at the method of assessment of the

8 recurrence and is it true that, say, PET-detected

9 recurrences are as highly correlated with survival

10 as non--as physically detected or x-ray or CT. I

11 think the new trials actually provide much richer

12 data sets than many of these older trials that

13 collected very much bare-bones sort of approaches.

14 And so I do think there will be a number of

15 additional pieces of information that we can look

16 at.

17 With respect to immunotherapies in

18 particular, I guess the jury is still out, and we

19 have to get some actual data on five-year

20 assessment with those therapies.

21 DR. KELSEN: Dr. Pazdur?

22 DR. PAZDUR: I think, you know, several

309

1 people have brought up, well, what if our

2 evaluation techniques change? What if the drugs

3 change? God knows. Okay?

4 As a discussion here, I think we have to

5 point out where we are now. Obviously, we can

6 always reassess where we're going to and what

7 changes will be impacted. But I kind of want to

8 direct the attention and the flow of the

9 discussion, because we have a lot of material to

10 cover here, toward what we have at hand. We could

11 always talk about what will be a new improvement in

12 ten years or five years, what will be the role of

13 PET scanning, what will be the role of this and

14 that. That will impact--then we as a regulatory

15 agency have to make that decision.

16 I'll just parenthetically add that, even

17 though we are allured by new mechanisms of actions

18 of drugs, many times, at least in the advanced

19 disease, we've seen very consistent effects on

20 established endpoints--Avastin, for example, having

21 a consistent effect on our ways of measuring

22 anti-tumor activities or response rate improvement

310

1 and improvement in time to progression or

2 improvement in survival.

3 So even though drugs may have a different

4 mechanism of action--and, granted, it's in the

5 advanced disease--they still may ultimately express

6 their effect on more conventional endpoints. But

7 here, again, I think our time is somewhat limited

8 here, and we could go off and hypothesize in

9 multiple different directions. But we're here,

10 we're working in 2004, and let's keep the

11 discussion to that and move forward.

12 DR. KELSEN: Any other questions of the

13 panel? Dr. Brawley?

14 DR. BRAWLEY: In follow-up to what Dr.

15 Pazdur just said, because there are some points

16 that I had, if you look back over the last 30

17 years, you've got a number of trials as technology

18 has changed over time. Lead-time bias has been

19 introduced with each introduction of each new

20 technology. Even within CT scan generations, we've

21 increased lead-time bias.

22 The randomization and the fact that the

311

1 trial is all being done at the same time has always

2 sort of equalized that, and the one thing that we

3 can do, Rick, is look backward and we can see that

4 the development of CT scan, the introduction of MRI

5 and so far the introduction of PET scan has not

6 really changed disease-free survival as a good

7 correlate for overall survival.

8 DR. KELSEN: Any other questions from the

9 panel or from FDA?

10 [No response.]

11 DR. KELSEN: If not, we'll then go to the

12 open public hearing portion, and there is one

13 speaker, I believe Mr. Carroll. Before we have Mr.

14 Carroll's comments, both the Food and Drug

15 Administration and the public believe in a

16 transparent process for information gathering and

17 decisionmaking. To ensure such transparency at the

18 open public hearing session of the Advisory Board,

19 the FDA believes it is important to understand the

20 context of an individual's presentation. For this

21 reason, FDA encourages you, the open public hearing

22 speaker, at the beginning of your written or oral

312

1 statement to advise the committee of any financial

2 relationship that you may have with any company or

3 any group that's likely to be impacted by the topic

4 of this meeting.

5 For example, the financial information may

6 include a company's or group's payment of your

7 travel, lodging, or other expenses. Likewise, FDA

8 encourages you at the beginning of your statement

9 to advise the committee if you do not have any such

10 financial relationships.

11 If you choose not to address this issue of

12 financial relationship at the beginning of your

13 statement, it will not preclude you from speaking.

14 MR. CARROLL: Thank you, Mr. Chairman, and

15 good afternoon to the committee. My name is Kevin

16 Carroll, and I'm employed by AstraZeneca in the

17 role of global statistical leader for oncology,

18 based over in the U.K. What I'd like to do for the

19 next ten minutes or so is to share with you some

20 thoughts and some data that I believe are relevant

21 to your discussions with respect to the use of

22 progression as an endpoint in colorectal cancer

313

1 studies.

2 My time is limited, and I do hope you'll

3 forgive me if I rush through these slides a little.

4 In response to the workshop in November

5 and the calls to look at progression and survival

6 data in the first-line setting, we at AstraZeneca

7 did look at our experience in this area with

8 Tomudex, and we found that the data that we have in

9 that clinical trial and program support

10 progression-free survival in the first-line setting

11 as a true surrogate for survival.

12 Furthermore, we undertook a brief review

13 of the emerging mixture in this area and found that

14 the observation made in our Tomudex program was

15 generally supported by the literature.

16 Furthermore, as we saw yesterday, there

17 were considerable concerns about using

18 progression-free survival in terms of issues

19 relating to the timing of the event and potential

20 introduction of bias. As we move through these

21 next few slides, I hope to share with you an

22 alternative analysis being an event count analysis,

314

1 which I believe provides a simple alternative to

2 the analysis of PFS time and avoids the kinds of

3 concerns that we have seen yesterday.

4 Lastly, we maintain that progression is a

5 meaningful endpoint in and of itself in first-line

6 colorectal cancer and, given the complexity of

7 crossover and the increasing number of available

8 effective therapies, should be employed as the

9 primary endpoint in the first-line setting, which

10 is a view that is common with views expressed in

11 the literature.

12 In the mid-1990s, AstraZeneca sponsored a

13 program of three Phase III randomized trials of

14 Tomudex versus 5FU in the first-line treatment of

15 advanced colorectal cancer. On this next slide, I

16 briefly show you the results of these trials,

17 primarily to indicate that there is a treatment

18 effect on both progression-free survival and

19 overall survival in these trials. And, therefore,

20 in the same way as we've just seen, we can formally

21 assess whether there's any evidence of surrogacy in

22 this setting in this data set.

315

1 When we do that, we find indeed that there

2 is evidence based on these data that PFS is a true

3 surrogate for survival in this setting. As has

4 been mentioned before and also a has been discussed

5 in the committee in the past, progression is not a

6 matter of correlation--sorry, surrogacy is not a

7 matter of correlation. What we're trying to

8 establish is whether the effect of treatment on the

9 endpoint of interest--in this case survival--is

10 mediated through an effect on an earlier endpoint.

11 In simple terms, this means that if we were to do

12 an analysis of survival and we adjusted for the

13 early effects of progression, would the treatment

14 effect on survival vanish? And, indeed, if we do

15 that analysis on this data set, what we find is

16 that a survival analysis adjusting for

17 progression-free survival is no longer significant,

18 and that suggests that progression is indeed a

19 surrogate, at least in this data set.

20 Furthermore, there are more sophisticated

21 means of assessing surrogacy, and I think we've

22 just seen some of those touched on. And if we

316

1 apply these more up-to-date techniques, we're able

2 to predict the effect of a 5FU-like treatment on

3 survival given its effect on progression. And I

4 think there's a mistake in your slides--in your

5 handout, and what we find in the Tomudex data is

6 that if progression was increased by, say, 50

7 percent, we would expect survival to be increased

8 by around 29 percent, with a confidence interval as

9 shown. And I think such predictions are going to

10 be useful if we're thinking about using progression

11 in the first-line setting.

12 The positive association between the

13 effect of treatment on survival and the effect of

14 treatment on progression-free survival is displayed

15 in this figure, which is very similar to the one

16 that you've just seen. And this is using the

17 methodology published by Buyse and Molenberg

18 recently.

19 What we see here, the circles are actually

20 regions in the trial program, in fact, distinct

21 countries that participated in the Tomudex trial

22 program. And what we see is there is a significant

317

1 correlation between the effects of treatment on PFS

2 and the effects on overall survival. And that

3 further supports the surrogacy of progression-free

4 survival in this setting.

5 Of course, the little bit of data I've

6 shown you on Tomudex in response to comments made

7 in the workshop is really only one small piece of

8 data, and I think we clearly have to look at all

9 the available data in order to place this

10 information into context. And I just placed on

11 this slide the recently emerging and published

12 information in the first-line setting where I think

13 you can see that there are large effects on

14 progression-free survival across a number of trials

15 which generally, but not always, are translating

16 into survival benefits. Clearly, the

17 interpretation of these data is made complex by

18 crossover issues, by maturity issues, and follow-up

19 issues. But, nevertheless, I think you might agree

20 that these data tend to support progression as an

21 endpoint in this setting.

22 Very similar to the previous presentation,

318

1 what we really need to do here is to apply a

2 meta-analytic approach to all the available

3 first-line data in order to truly establish once

4 and for all the relationship between progression

5 and survival in this setting, and that would be

6 something that AstraZeneca would very much support

7 as a willing participant.

8 Now, moving on briefly to talk a little

9 bit about issues in using progression--and we saw a

10 number of issues debated yesterday, and this slide,

11 in fact, was also used yesterday, where clearly

12 progression-free survival time is not known with

13 complete certainty, and that can lead to

14 overestimation and bias, and this is of great

15 concern.

16 The key question in my mind is: What can

17 you do about that? And I think there are a number

18 of very complicated, sophisticated ways of trying

19 to deal with complicated sensory mechanisms and a

20 number of assumptions for the timing of event, and

21 I'm not sure that any of those methodologies are

22 really satisfactory.

319

1 One simple approach that might be

2 considered, I think, is as an alternative, or at

3 least in support of PFS time analyses, that we

4 actually compare treatments on the basis of an

5 overall event count over the trial follow-up

6 period. This is an idea which is actually very

7 similar to the single time point approach that I

8 think was discussed both in the workshop in

9 November and also in the Advisory Committee in

10 December. And I'll show you a quick example of

11 that in a moment.

12 Essentially, if you were to employ an

13 event count analysis, the benefit that you would

14 have is that you would be comparing treatments free

15 from concerns about the timing of the event, which

16 was at least one of the issues yesterday. The

17 treatment effect--the difference between treatments

18 could be described usefully in terms of the

19 relative risk of progression over the follow-up

20 period, and, furthermore, it's relatively

21 straightforward to show that if you use this

22 alternative endpoint, there's relatively little

320

1 loss in statistical power. And, in fact, in

2 circumstances where the treatment effect is delayed

3 so the Kaplan-Meier doesn't open at the beginning

4 but opens at some later time point, it's actually

5 more powerful than the regular way we look at data

6 today. And, therefore, I would think that this

7 kind of event count analysis should at least be

8 considered as a supportive analysis when looking at

9 analyses of progression-free survival time because

10 it provides reassurance with respect to a lack of

11 bias and provides reassurance that perhaps

12 conclusions on PFS time are robust.

13 As I promised, I think it's helpful just

14 to illustrate this endpoint with an example, and

15 this slide is rather complicated so I'll just take

16 a moment to explain what's going on here.

17 What we can do is we could take a regular

18 Kaplan-Meier curve and we can break the follow-up

19 axis along the bottom as shown on this slide here.

20 The blue circles on this slide represent the hazard

21 ratio derived from the regular analysis of PFS

22 time, and the red circle represents an analysis of

321

1 an event count, ignoring the time to progression

2 and getting around some of those problems we talked

3 about yesterday.

4 So if we consider the first three months

5 of the Kaplan-Meier curve--sorry, the first six

6 weeks of the Kaplan-Meier curve, what we see is

7 that whether we do an analysis of PFS time in a

8 regular way or whether we do a simplified analysis

9 of the events that occurred over that period of

10 time, you get essentially the same answer.

11 If we extend the follow-up period to a

12 12-week follow-up and then do a PFS time on the

13 first 12 weeks and get the hazard ratio--and we

14 plot that in blue--we can also calculate the

15 relative risk just on the numbers of events. And,

16 again, you can see that the two analyses are very

17 similar and so on through follow-up.

18 What this rather complex slide shows you

19 is that there is really no difference between the

20 outcomes achieved when you use a PFS analysis and a

21 simplified event count analysis in this trial.

22 That suggests that the PFS conclusions reached

322

1 here, at least, are robust. There is no bias

2 introduced because we see results that are

3 supported by a simpler analysis of event count.

4 And I think, therefore, this provides some

5 reassurance that we can employ simpler methods of

6 the data analysis and collection in first-line

7 colorectal trials and others when looking at PFS.

8 In summary, then, I would just close by

9 saying that AstraZeneca's Phase III program data on

10 Tomudex provide evidence to support PFS as a true

11 surrogate for first-line colorectal cancer. The

12 recent literature I think is supportive of that

13 observation, that improvements in PFS are generally

14 followed by improvements in survival. Furthermore,

15 there are always concerns using progression-free

16 survival, and I think we can consider an event

17 count analysis as at least as supportive analysis

18 if not a direct replacement for the regular

19 analysis of PFS time when concerns exist about the

20 imputation of times and also asymmetric follow-up.

21 And, of course, an event count analysis can

22 accommodate and get around the issues of asymmetric

323

1 follow-up.

2 Finally, I would just say that,

3 irrespective of whether we ever formally and

4 convincingly establish surrogacy between PFS and

5 survival using rigorous statistical methodology in

6 the first-line setting, we would maintain that

7 progression-free survival is a clinically

8 meaningful endpoint in and of itself. And given

9 the issues of crossover and an increasing number of

10 therapies available as second-line treatments, PFS

11 should be employed as a primary endpoint in

12 clinical trials in the first-line setting.

13 Thank you for your time and attention.

14 DR. KELSEN: Thank you, Mr. Carroll.

15 We have time for one question. Dr.

16 O'Connell?

17 DR. O'CONNELL: Yes, I just wanted to make

18 a comment that at the workshop the one point for

19 further research that emanated from that meeting

20 was exactly what you just suggested to do. In

21 fact, a formal meta-analysis from the cooperative

22 groups in the United States to determine the

324

1 association between progression-free survival or

2 time to progression and overall survival to have a

3 more broad view than the two or three or now four

4 studies that have been discussed so far.

5 DR. KELSEN: Thank you, Dr. O'Connell.

6 At this point, we're going to take a--do

7 you have a question, Rick?

8 DR. PAZDUR: I have one. AstraZeneca did

9 three trials with--and I don't think you mentioned

10 the results. What we're obviously interested in

11 is: Does time to progression, if you measure it,

12 predict for survival, subsequent survival? And of

13 those three trials that were using time to

14 progression, how did that correlate with survival

15 in those individual studies? If you take--I think

16 it was like O11, O12, I forgot the actual numbers.

17 I don't know the data specifically--

18 MR. CARROLL: Yes, I'm very happy to talk

19 to individual trial results. I think it's a very

20 good question. I did flash up a slide very

21 briefly, but time was short so I went straight past

22 it.

325

1 What we have, there were three trials,

2 each of about the same size, and what you find is

3 that in two of those trials you can individually

4 apply the formal Prentice criteria for surrogacy,

5 and in two of those trials we see that about half

6 of the effect on survival is explained by the

7 effect on progression-free survival, which is very

8 consistent with putting all the data together,

9 which is what I've shown on this slide.

10 So the individual trials support the

11 overall result in terms of surrogacy, and if we

12 apply--the other methodology that could be applied

13 is the Buyse-Molenberg that we've seen before where

14 we try and predict the effect on survival given the

15 effect on progression. And that methodology can be

16 applied to two trials because one of the trials

17 showed a very small effect and, therefore, it was

18 kind of difficult to apply that methodology. But

19 in the two trials where we could apply this

20 alternative methodology, again, we saw that there

21 was a correlation, a significant correlation

22 between the effect on progression and effect on

326

1 survival. So the overall results I've run through

2 quickly are supported by the individual trial data.

3 And, in fact, we will be publishing this material

4 with the--

5 DR. PAZDUR: So what you're saying, if you

6 took all three of those trials, in two of them if

7 we made a decision does PFS correlate with survival

8 and improvement in survival, we would have been

9 correct in two of those trials. There was an

10 improvement in PFS. And then subsequently in that

11 trial, it was correlated with a positive effect on

12 survival. That was present in two trials, and then

13 in the third one it was not. Is that what you're

14 saying?

15 MR. CARROLL: No. I'm saying that there's

16 one trial where individually you can't apply--the

17 criteria we've talked about today require special

18 conditions to be in place for significant effects

19 and endpoints. So you couldn't, strictly speaking,

20 apply the criteria to some trials, so we don't

21 know. But the two trials we could apply the

22 criteria, we could predict survival given the

327

1 progression effects.

2 DR. PAZDUR: Okay. Thank you.

3 DR. KELSEN: If there are no further

4 questions, we're going to take a ten-minute break.

5 We'll reconvene at 3:20.

6 [Recess.]

7 DR. KELSEN: Okay. Before we start, Dr.

8 Pazdur wants to make a few comments.

9 DR. PAZDUR: In my introductory comments,

10 I forgot to make a very important comment, and that

11 deals with the process that we're going through

12 looking at the endpoints. And I'd like to express

13 the agency's personal gratitude to both ASCO, the

14 American Society of Clinical Oncology, and AECR for

15 their efforts in assisting us with the various

16 workshops we've had. They've done a terrific job.

17 The people involved have been excellent in

18 coordinating multitudes of activities that go into

19 these workshops.

20 So, again, I wanted to bring that up, and

21 I was remiss in not doing so. Thank you.

22 DR. KELSEN: Thank you, Dr. Pazdur.

328

1 If we can turn to the questions of the

2 committee: In December, the committee discussed

3 the issue of disease-free survival as a general

4 matter dealing with many tumors. And what the

5 agency would like us to talk about today is limited

6 to colon cancer, not discussing other tumors.

7 I think everyone has had a chance to read

8 the questions to the committee. I'd like to go to

9 Question No. 1. I'll read Question No. 1, and then

10 we'll open it for discussion.

11 Question 1: For colon cancer drugs, could

12 an increase in disease-free survival compared to

13 standard therapy represent clinical benefit and be

14 an adequate basis for regular drug approval?

15 We'll open that now for discussion.

16 DR. PAZDUR: One point that I'd like to

17 bring up is obviously we are assuming that there is

18 a sufficient magnitude of effect, obviously if the

19 magnitude comes into being and is the data quality

20 appropriate, et cetera, assume that that's a given.

21 We realize that that's a given.

22 DR. KELSEN: And assume that it's either a

329

1 very large adequate trial or trials.

2 Discussion from the committee? Dr.

3 George?

4 DR. GEORGE: I'll start. I think the

5 answer is yes, based on what I've heard and know,

6 but it's what we know today with the current

7 therapies and the current modalities for detection

8 and so forth, all those caveats. But that's all we

9 have to go on. I think the future may hold

10 something different, but so I'd say certainly the

11 answer is yes here.

12 DR. KELSEN: Dr. Brawley?

13 DR. BRAWLEY: I think the answer is yes,

14 and I'd actually also propose thinking about

15 something that would be a little perhaps

16 innovative. You could give a tentative approval or

17 some type of approval based on disease-free

18 survival, and then that same cohort or the same

19 study population could ultimately be studied to get

20 overall survival later on.

21 During the period of time between the

22 initial approval for disease-free survival, you

330

1 could actually--people in the United States could

2 actually use this drug, and then there would be a

3 secondary review at the time the overall survival

4 data was available.

5 DR. KELSEN: Dr. George?

6 DR. GEORGE: I think, though, what you're

7 talking about there sounds more like accelerated

8 approval. What this is talking about is

9 disease-free survival as a clinical benefit itself,

10 which would be regular approval, unless you're

11 proposing to change--

12 DR. PAZDUR: Correct. To follow up on

13 Otis' answer, basically, we normally would take a

14 look at mature survival data, with the caveat that

15 we're very interested, as Mike pointed out, that

16 there isn't any decrement in survival. That's an

17 important point, and we've done this with multiple

18 applications outside of this area.

19 DR. KELSEN: Yes, Steve?

20 DR. GEORGE: I don't want to keep jumping

21 in here, but that's sort of Question 1(d), I think.

22 Could we--

331

1 DR. KELSEN: Yes, we'll get to--

2 DR. GEORGE: Do you want to wait to come

3 to that?

4 DR. KELSEN: What we're going to do is

5 we're going to discuss and vote on the big print,

6 on the big question of regular approval. And then

7 depending on the vote of the committee, we'll then

8 look at the subcategories (a), (b), (c), and (d) as

9 they apply.

10 Ms. Roach?

11 MS. ROACH: My answer is yes, but, as long

12 as the novel treatments coming down the pike, the

13 work that's done--that's been done to show the

14 relationship is continued to keep showing that

15 relationship and the clarity of the relationship.

16 DR. KELSEN: Thank you.

17 Dr. O'Connell?

18 DR. O'CONNELL: Just to clarify what Dr.

19 Pazdur said, a regular approval for three-year

20 disease-free survival would entail examination of

21 the five-year survival to be certain that there

22 wasn't some delayed detriment, and that it wouldn't

332

1 be necessary to look at simply accelerated approval

2 at three years to assure that survival would be

3 subsequently examined. Is that right?

4 DR. PAZDUR: We would negotiate with the

5 sponsor to look at that. That would be part of the

6 agreement.

7 DR. KEEGAN: Actually, I'm concerned that

8 you don't confuse a required committee to collect

9 the data with an agreement. Regular approval would

10 be completed upon the three-year disease-free

11 survival data. So you may or may not get the

12 five-year data. We would ask for it, and it could

13 be an agreed-upon commitment. But we wouldn't have

14 the same ability to withdraw an approval based on

15 failure to complete that commitment, for instance,

16 which may be a distinction without--

17 DR. BRAWLEY: Yes, I'm accepting reality.

18 I was at the beginning of my comment expressing

19 what I wish the law would allow. I understand the

20 law does not allow that.

21 DR. KELSEN: Dr. Williams?

22 DR. WILLIAMS: I'm hearing a little bit of

333

1 confusion of comments. Dr. Martino earlier

2 mentioned the concept that there might be

3 symptomatic recurrences and, therefore,

4 disease-free survival itself was a clinical

5 benefit, I would guess regardless of the time or

6 the setting, et cetera, that delaying that

7 suffering was the endpoint. But I'm also hearing

8 comments that, well, as long as things don't

9 change, et cetera, which would suggest that it

10 primarily is the surrogacy for survival that's

11 driving you.

12 So I don't know if you want to clarify

13 whether--that if you beat the best thing out there

14 with regard to disease-free survival in any realm

15 and would that be clinical benefit, or would it

16 only be tied to this particular set of analyses

17 that have to do with surrogacy for survival?

18 DR. O'CONNELL: From my point of view,

19 there would be clinical benefit associated with an

20 improvement in three-year disease-free survival per

21 se, not as a surrogate. But I would also want to

22 know what the long-term outcome is going to be to

334

1 be certain there wasn't some unexpected deleterious

2 effect on overall survival.

3 DR. WILLIAMS: But you're not requiring

4 that it fulfill the presumed surrogacy--

5 DR. O'CONNELL: Correct.

6 DR. WILLIAMS: --just that you don't have

7 a bad outcome.

8 DR. O'CONNELL: Correct.

9 DR. HIRSCHFELD: I'm sorry. A

10 clarification. I think the question is

11 disease-free survival without a specific landmark

12 analysis attached to it, and it's not three-year

13 disease-free survival or some other prespecified--I

14 think that's--

15 DR. KELSEN: That is correct.

16 DR. HIRSCHFELD: --the point we're seeking

17 advice on.

18 DR. KELSEN: Correct. That's number (a).

19 Dr. Rodriguez?

20 DR. RODRIGUEZ: I just had--I guess it's

21 for clarification. If indeed for whatever reason

22 subsequently it was found that this combination or

335

1 this drug did cause unexpected mortality, I assume

2 the same process would follow through as is done

3 with, for example, the cardiac drugs that were

4 found to cause premature death?

5 DR. KELSEN: Dr. Pazdur?

6 DR. PAZDUR: For any approval, yes, if

7 there is an unexpected toxicity associated, we

8 would review that, bring it back to this committee,

9 and the drug could be withdrawn, that indication.

10 DR. KELSEN: Dr. Taylor?

11 [No response.]

12 DR. KELSEN: Other questions for

13 discussion?

14 [No response.]

15 DR. KELSEN: Okay. So I will read the

16 question again, and then we will vote on this

17 question. And I've been asked to make sure

18 everybody pauses a little bit after the person

19 before them so they can get all the votes down

20 correctly.

21 So we're voting on the following question:

22 For colon cancer drugs, could an increase in

336

1 disease-free survival--not yet defined--compared to

2 standard therapy represent clinical benefit and be

3 an adequate basis for regular drug approval?

4 MS. ROACH: Yes.

5 DR. SARGENT: Yes.

6 DR. O'CONNELL: Yes.

7 DR. BRAWLEY: Yes.

8 DR. MARTINO: Yes.

9 DR. TAYLOR: Yes.

10 DR. REAMAN: Yes.

11 DR. REDMAN: Yes.

12 DR. KELSEN: Yes.

13 DR. CHESON: Yes.

14 DR. GEORGE: Yes.

15 MS. HAYLOCK: Yes.

16 DR. CARPENTER: Yes.

17 DR. RODRIGUEZ: Yes.

18 DR. DuBROW: Yes.

19 DR. KELSEN: That sounds unanimous to me.

20 So the recommendation of the committee is

21 that disease-free survival be considered for

22 standard--as a clinical benefit for full approval.

337

1 And now I'll ask for brief discussion and comment,

2 if any, for 1(a), which for the audience asks the

3 question--guidance for the duration at which that

4 time point should be. Should that time point be

5 three-year disease-free survival or five-year

6 disease-free survival or presumably some other

7 point in between?

8 DR. PAZDUR: And just to follow up on

9 Grant's question so we are clear on this, what this

10 unanimous vote is saying is that you all feel that

11 this is of benefit in itself.

12 DR. KELSEN: Three-year versus five-year.

13 Comments? Discussion? Dr. Carpenter?

14 DR. CARPENTER: I think everything we've

15 heard rather careful and extensive study on is

16 three-year, and the lack of information,

17 well-documented information and careful study on

18 the other endpoint, it would seem the most sensible

19 to use the one that's been the best studied now and

20 leave it open to alternative durations.

21 DR. KELSEN: Dr. George?

22 DR. GEORGE: My comment on this is that

338

1 three years, I think, seems reasonable as it exists

2 now, but it's a three-year minimum follow-up, I

3 think is what we're talking about, because the

4 accrual period could vary widely, and we're talking

5 about a minimum of three-year follow-up on each

6 subject, or at least enough follow-up on enough

7 patients for three years to have a reliable answer.

8 So I think there's some fuzziness here in

9 whether we want to be looking at three-year--a real

10 three-year disease-free survival or we just want

11 enough follow-up on all patients so we're

12 reasonably sure to have captured a--gotten a

13 reliable answer to the question we're trying to

14 ask, and that three years was based on primarily

15 because that's where the action was, so to speak,

16 that's where the events were occurring.

17 So I don't particularly--I'm not

18 particularly sold on the idea of looking at--say

19 when you end up looking at this one point in time,

20 three-year disease-free survival.

21 DR. KELSEN: Dr. Williams?

22 DR. WILLIAMS: Again, I think it depends

339

1 on what we're talking about. The reason that you

2 would pick three years--certainly we've seen a lot

3 of good comparisons to survival. But if you take

4 the philosophical attitude that it's benefit, it

5 would seem less important for that. But, of

6 course, it is near the plateau and perhaps if you'd

7 like to get away from, you know, where most of the

8 action has occurred--I mean, do you have any

9 feelings regarding--out of the context of

10 surrogacy, why three years?

11 DR. GEORGE: You're asking me? No, I

12 think as with any disease, you'd want to be sure

13 that you have gotten to a point where you're

14 reasonably sure that most of the events have

15 occurred. If you do it too early, you're liable to

16 fool yourself. So you want to go out far enough.

17 Now, that could change with time, with

18 therapies or improvements or such, but that's why I

19 say I don't like sticking with the--I don't like

20 just saying three years versus five years. I think

21 it should be more dependent on--I mean, if you

22 enter everybody--suppose you had a trial that

340

1 accrual was so rapid that everybody entered on the

2 first day. Then in three years, you'll have most

3 of the events. But if you have another trial that

4 takes years to accrue, you're going to have those

5 early patients who will have some information, but

6 not the later ones. And so you want to go far

7 enough so you have enough information to make the

8 analysis an appropriate one.

9 DR. PAZDUR: I have a question for Dan and

10 for Mike, NSABP and NCCTG. When you're doing an

11 adjuvant study, you're follow-up, your initial

12 analysis, your three-year analysis which you

13 normally do, would it be three years following the

14 last patient, or is it a median of three years

15 follow-up? Because, remember, we're getting most

16 of our data now on many of these adjuvant protocols

17 from the cooperative groups, and I need to know

18 their understanding on this point.

19 DR. SARGENT: I'll answer first, Mike.

20 It's actually an event-driven analysis as opposed

21 to a time-driven analysis. But our general policy,

22 within NCCTG, at least, is to base our estimation

341

1 of when that would occur based on a projected event

2 rate and accrual rate to project that analysis to

3 occur at about three years after the close of

4 enrollment. And so it's really event-driven as

5 opposed to time-driven, and I'd just like to make

6 that point to emphasize what Dr. George indicated,

7 that I think it's very important to note that my

8 analysis that has been conducted did not just look

9 at the single time point of three years. It used

10 all the data from the patients up until a time

11 point three years after the close of randomization.

12 It used hazard ratio and logrank tests. It did not

13 look at a specific rate at a specific time point.

T5A DR. O'CONNELL: The NSABP trials are also

15 event-driven, and so there are several interim

16 analyses and final analysis after a given

17 proportion or a certain number of events occur,

18 same as the NCCTG.

19 DR. KELSEN: Dr. Martino?

20 DR. MARTINO: I just want to underscore

21 the importance of these last few statements that

22 were made, because we've been throwing around this

342

1 three-year thing as if we all knew what it meant,

2 and there really are at least three possible things

3 that I understand it could mean. I'm sure the

4 statisticians have more. And so this becomes--you

5 know, understanding what we mean by this to me is

6 very crucial. You know, recognizing that some of

7 these things are, in fact, driven by the

8 inter-group relationships, but there are drug

9 companies now who also run their own adjuvant

10 trials.

11 And so unless you have a clear

12 understanding, I could see me sitting here with

13 someone saying, yeah, but to us, three years

14 didn't--wasn't event-driven but, rather, was three

15 years from some date.

16 So we need to be very clear that we're

17 unanimous on this.

18 DR. KELSEN: I was actually going to ask

19 if you wanted to reformulate the question for us or

20 just leave it in this general sense back to you of

21 three-year disease-free survival.

22 DR. WILLIAMS: I think we understand.

343

1 DR. KELSEN: Dr. Taylor?

2 DR. TAYLOR: I just wanted to say what she

3 was saying. It can't be just three years from the

4 start. You have to make a definition.

5 DR. KELSEN: Any other discussion?

6 DR. SARGENT: The data that I presented is

7 three years minimum follow-up on each patient.

8 Now, having said that, this is an ongoing analysis,

9 and we've started to look at three-year median

10 follow-up, and the initial results look very

11 promising with three-year median. But a member of

12 the audience during the break said, well, what

13 about two years? What about some other time? And

14 so I think that's still a question for ongoing

15 investigation, but the data that we've looked at so

16 far does support the three-year minimum follow-up

17 time point and consider all data available up until

18 that time point.

19 DR. WILLIAMS: And I think you've made a

20 good case that three years is when you're

21 approaching the plateau of the curve, and that

22 seems like a reasonable basis. So we're hearing

344

1 three-year minimum as your recommendation at this

2 point in time.

3 DR. KELSEN: Yes, do you need us to vote

4 on this, or are you satisfied with the tenor of the

5 discussion?

6 DR. PAZDUR: We're satisfied.

7 DR. KELSEN: Okay. If there's no further

8 discussion about that, (b) and (c) sort of are

9 answered since we voted in favor of regular

10 approval as representing clinical benefit. Would

11 the agency like us to discuss (d) for guidance.

12 DR. PAZDUR: Yes.

13 DR. KELSEN: So I will briefly read (d).

14 I will summarize 1(d) for the panel.

15 Consider a study in which there is a

16 statistically significant difference in

17 disease-free survival, but after adequate follow-up

18 there's no evidence of a survival effect, there is

19 no survival trend in favor of the experimental arm.

20 Would increased disease-free survival alone be

21 adequate for approval in this setting? If so,

22 discuss the nature of the clinical benefit from the

345

1 increased disease-free survival when there's no

2 survival benefit. That is, the study's presented

3 and disease-free survival is clearly improved, but

4 you look at the curves and it doesn't look like

5 survival is going anywhere.

6 Discussion? Dr. Martino?

7 DR. MARTINO: Well, I think we actually

8 have discussed this, and I think the point that was

9 made originally was that we felt that in and of its

10 own this would be a valuable clinical endpoint.

11 The only caveat is if there had been a bad survival

12 outcome, in which case, you know, you have recourse

13 to how you handle that. But it would not--but

14 other than that, I think we've answered your

15 question. Haven't we?

16 DR. PAZDUR: I think you've answered it,

17 but what we're looking for is a little bit of

18 clarification why. Because there are some of

19 perhaps a more conservative element that would say,

20 you know, if you're just saying that you're sparing

21 people toxicity of chemotherapy for advanced

22 disease, or you're treating a far larger portion of

346

1 people with chemotherapy in the adjuvant setting.

2 So why specifically in your clinical judgment do

3 you think an improvement in disease-free survival

4 is important?

5 DR. MARTINO: Well, what you all have

6 reminded at least me of today is that when a person

7 recurs, you can sort of anticipate that within some

8 months--and those months aren't many--that, in

9 fact, they will be symptomatic. And so for me,

10 that is good enough. I'm quite satisfied that

11 preventing symptoms is valuable.

12 DR. KELSEN: Dr. George?

13 DR. GEORGE: To follow up on that a little

14 bit, two points. One is the clinical benefit is

15 in, number one, that progression follows fairly

16 shortly; and, number two, there's something we

17 haven't discussed, I think, in that there's a

18 psychological aspect that I'm willing to sort of

19 accept, that if someone says if you delay

20 progression it's a good thing, sort of face

21 validity, almost, which I'm willing to accept that.

22 Of course, I wouldn't be willing to accept it quite

347

1 as readily if there weren't this knowledge that

2 there are symptoms coming soon after.

3 Now, here's the problem with the survival

4 thing, though, that I don't know--I think we have

5 to really think this through. If you do have

6 regular approval for disease-free survival and then

7 continue to follow for survival, there's at least a

8 theoretical possibility that some new agent would

9 have some weird mechanism of action that could have

10 a nice effect on disease-free survival and have

11 some longer-term deleterious effect on survival

12 through some mechanism that we don't know about.

13 Now, in a particular study if you were to

14 look at that, what might happen? You might approve

15 it based on disease-free survival, and you say,

16 well, as long--you gave the example of having, say,

17 no effect on survival. But that implies, if you

18 had no effect on survival, that you're really not

19 ruling out an actual decrement in survival. I

20 mean, you could actually have survival look better,

21 not be significantly better. By usual statistical

22 things, approaches, you would say, well, you really

348

1 haven't ruled out a slight negative effect. And if

2 you had, say, the two survival curves lying flat on

3 top of each other, you haven't ruled out probably a

4 pretty big decrement. This has to do with the same

5 kinds of arguments that are made in non-inferiority

6 kinds of studies.

7 But that could put you in a quandary. I

8 mean, you could say--especially it would put you in

9 a quandary if survival starts looking a little

10 worse. I mean, it may not be worse, but it's--you

11 really are worried that maybe what we've done here

12 is approve something that looked good in

13 disease-free survival and, in fact, overall

14 survival could actually be worse, despite all our

15 work in looking at this as a clinical benefit in

16 itself and as a surrogate. So that's a worry?

17 DR. PAZDUR: I realize you're worried, and

18 we would be looking at this, and I think most

19 sponsors would be following patients for survival.

20 Why? Well, obviously, if they have a survival

21 benefit, they'd want to make that survival claim.

22 Now, the question that I have which we

349

1 asked for a first-line setting, but is really

2 germane here, if we moved away from survival as a

3 primary endpoint of a trial when we discuss these

4 to a disease-free survival, what should the studies

5 be powered for? Because that is a question. And,

6 remember, if we don't ask a survival question and

7 have under-powered trials, we have the potential of

8 never knowing that we have, you know, affected

9 survival, which would be very deleterious, I think,

10 to the field of oncology in general, not to really

11 have an accurate depiction of what our therapies

12 really give patients.

13 We would be happy to have a primary

14 endpoint of disease-free survival and perhaps a

15 secondary endpoint where the trial would be

16 powered. Obviously, it would have more patients, a

17 trial powered for survival. Am I correct on that?

18 DR. SARGENT: Well, the event rates for

19 disease-free survival after three years of

20 follow-up and for overall survival after five years

21 of follow-up are virtually identical.

22 DR. PAZDUR: So they're not different.

350

1 DR. SARGENT: So the power--the sample

2 size should be the same.

3 Now, recognizing that if the trial follows

4 the pattern of these, there is that slight

5 attenuation of the impact. And so if the question

6 is do we need adequate power to detect the slightly

7 attenuated effect, then you may need a somewhat

8 larger trial for overall survival, but not by very

9 much. We're talking about the order of 10 percent,

10 and the suggestion that I gave, I think, in

11 November was that if you did power it for

12 disease-free survival at, say, a hazard ratio of

13 1.4, you might consider powering it

14 for--overpowering it a little bit for, say, 1.35,

15 which would then give you the power to detect

16 overall survival at 1.4, assuming a slight

17 attenuation.

18 DR. PAZDUR: But we could, to allay Dr.

19 George's fear, in the formal statistical analysis

20 plan require an analysis and data submission as

21 part of the move away from survival, looking at it

22 as a secondary endpoint. And obviously you would

351

1 have to win on overall survival as you--I mean, you

2 would have to win on disease-free survival to look

3 at overall survival. But one would think one would

4 do that anyway, you know, the natural history of

5 the disease.

6 DR. KELSEN: Dr. Williams?

7 DR. WILLIAMS: I think embedded in this

8 question is the concern about what should you

9 expect to see with regard to the survival hazard at

10 the time you do this minimum three years'

11 follow-up. And I don't know. Certainly things do

12 change over time, also, with treatments that may

13 have an effect on survival. So I don't know if we

14 should be expecting to see a trend in survival, if

15 it's going to occur, at the time you would do this

16 analysis. Do you have any idea? Should you be

17 expecting a trend in survival?

18 DR. SARGENT: At the three-year time

19 point?

20 DR. WILLIAMS: Yes.

21 DR. SARGENT: I wouldn't count on it. We

22 haven't looked at that issue specifically, but the

352

1 rate of death at follow-up time does not have that

2 sharp spike. People continue actually to die at a

3 pretty uniform rate over the first five years, and

4 we know that because we've tested the validity of

5 some of the statistical models. For example, an

6 exponential survival model fits very well for

7 overall survival, which in essence assumes that

8 your risk of death each year is constant over time.

9 An exponential survival model does not fit for

10 disease-free survival because there's this sharp

11 spike in recurrences earlier that falls off later.

12 So I think to answer the question, none of

13 the data that we have analyzed would suggest that

14 there should be a clear, significant benefit for

15 overall survival at the three-year point just

16 because there is one in disease-free survival.

17 DR. WILLIAMS: And I wonder, do some of

18 the earlier studies that were using no treatment or

19 just surgery, might they have seen a little more of

20 an early survival effect, you know, than the later

21 studies that include an active adjuvant arm--I

22 mean, the control as an adjuvant arm?

353

1 DR. SARGENT: I'm trying to make sure I

2 understand the question. Could you rephrase the

3 question?

4 DR. WILLIAMS: Well, would you--when there

5 was not an active adjuvant--active control arm,

6 would you have seen a survival effect earlier,

7 perhaps, you know, so you would have seen a

8 survival trend earlier than you would now where the

9 active control arm has an adjuvant active control?

10 DR. SARGENT: It's actually been pretty

11 consistent over time that if the curves separate,

12 they separate relatively early and continue with

13 the separation, and that's been consistent both in

14 the early trials and in the later trials that we've

15 looked at.

16 DR. KELSEN: Steve?

17 DR. HIRSCHFELD: In addition to a

18 decrement in survival, there's also interest in and

19 certainly we have intentions to follow other events

20 which could be catastrophic, like second

21 malignancies, and these have shown up in some

22 circumstances or some delayed neurologic

354

1 impairment.

2 DR. KELSEN: Right, and with new biologics

3 it may be not five years, it may be seven years.

4 It could be some other time.

5 Have you had enough discussion and

6 guidance from us and we don't need to vote on that?

7 DR. PAZDUR: Yes.

8 DR. KELSEN: Okay. So at this point we

9 have voted in favor of accepting disease-free

10 survival as representing clinical benefit and

11 approval, regular approval, and we'll move to the

12 next question, Question No. 2, which I'll read,

13 which now deals with advanced patients, presumably

14 Stage IV patients. When a surrogate endpoint for

15 clinical benefit is needed in advanced colon

16 cancer, would the preferred surrogate endpoint be

17 progression-free survival or time to progression?

18 Discuss progression-free and TTP in the first-line

19 treatment setting first.

20 Discussion from the committee. Dr.

21 Sargent?

22 DR. SARGENT: My point I guess would be

355

1 that I don't think either TTP or PFS has been

2 validated as a surrogate endpoint in this setting.

3 DR. KELSEN: Other comments? Steve?

4 DR. GEORGE: We should probably have a

5 clear-cut definition of a difference in these two

6 endpoints. I had a question about this before.

7 DR. WILLIAMS: Primarily the deaths are

8 included in progression-free survival.

9 DR. GEORGE: Right. That's the

10 difference, and the question--that makes the

11 question about the time to progression where you

12 could have deaths without progression. There is

13 still a question of how those are handled. This is

14 sort of a technical point, maybe, but, you know,

15 it's a competing risk kind of problem.

16 DR. WILLIAMS: The point that Tom Fleming

17 at the workshop was that his belief was that the

18 clinical benefit endpoint should include deaths

19 because obviously it's a very important outcome.

20 Of course, there are those who believe

21 that the more pure tumor endpoint is time to

22 progression. If you're trying to measure tumor

356

1 effect, that would be it. So, you know, there are

2 two different views on this.

3 DR. KELSEN: Dr. O'Connell?

4 DR. O'CONNELL: I guess I would argue in

5 favor of including death in the parameter to be

6 assessed for a couple of reasons: one, if a

7 patient dies and you don't have any information

8 about the cause of death, these patients all have

9 proven metastatic disease and there's a higher

10 likelihood that cancer contributed to that

11 patient's mortality in the advanced disease

12 setting.

13 And, secondly, if the patient dies because

14 of toxicity related to the treatment, that's

15 awfully important to know from a clinical

16 standpoint.

17 DR. KELSEN: Dr. Redman?

18 DR. REDMAN: I tend to agree with that. I

19 think including all deaths because sometimes we

20 don't know what the relationship is between the

21 treatment that we administer and a comorbid

22 condition that exists in this population.

357

1 DR. KELSEN: Can I ask the agency to

2 comment on this? The question is asking, if we

3 chose between these two alternatives as surrogates,

4 the current regulatory stance is a survival

5 improvement. So is this question asking--

6 DR. WILLIAMS: Those are, you know, the

7 next questions. But as we go forward in our next

8 questions, are they going to be PFS or TTP? Then

9 you can answer the heavy questions.

10 DR. KELSEN: All right. Let's discuss the

11 light question first.

12 Dr. George?

13 DR. GEORGE: Well, I'll go back to that

14 definitional issue, and I think it's the

15 progression-free survival that should be used for

16 reasons both because all these things we don't know

17 about the deaths that nominally don't occur with

18 recurrence, but also just from a technical point of

19 view, it gets more difficult to do those kinds of

20 analyses. They're not as simple. And so I think

21 for both those reasons I would prefer the

22 progression-free survival.

358

1 DR. KELSEN: Ms. Roach?

2 MS. ROACH: One of the things that came up

3 very clearly in the discussion yesterday were the

4 problems with using either kind of progression

5 endpoint as a surrogate endpoint or a real

6 endpoint, such as how to deal with new lesions and

7 validating the progression of non-measurable

8 disease. How--can you formalize that process?

9 DR. WILLIAMS: Yes. We definitely are

10 working on that, and we're going to be working on

11 the guidance and have internal work on it, and

12 we're going to certainly have some external

13 discussion and comments. So we certainly think it

14 needs a lot more work.

15 DR. PAZDUR: I'd like to amplify that

16 point. I think that's an excellent point, Nancy,

17 because that was, you know, a major problem with

18 some of the applications that we have seen.

19 I think oncology in general has relegated

20 this progression-free survival kind of to this

21 nebulous area where one doesn't address and

22 approach this with rigor. I think we've outlined

359

1 some of the problems with it that will need to be

2 put forward, not only in a guidance but in a plan,

3 prospective plan that the company writes, which may

4 be different from one drug to another here. I

5 think there's pros and cons of how to handle this.

6 But it has to be prospectively managed--interval

7 between assessments, what to do if somebody misses

8 a visit, how to handle the independent radiology

9 committee that is looking at this data versus what

10 the investigator brings forward. One cannot, after

11 somebody has taken a look at the data, decide,

12 well, I'll go with the investigator or I'll go with

13 the independent review committee. Obviously this

14 inflates error rates.

15 In colon cancer, we may want to look just

16 at the radiology review since most people don't

17 have physical findings to that degree and in a

18 randomized study they'd balance out.

19 But this needs to have attention. We're

20 talking internally about how to review the x-rays,

21 how many of these x-rays to look at. We are not,

22 obviously, going to look at 10,000 x-rays at the

360

1 FDA. We're going to be auditing x-rays in that

2 regard.

3 The issue is one of--and I'm glad Dr.

4 DuBrow is here--including radiologists as

5 investigators, and I think that needs to be done

6 because it has to be--these reports that we get

7 have to have a uniform meaning to them. We can't

8 just get these vague reports that the radiologists

9 give out--"There is a suggestion of a soft-tissue

10 mass. Clinical correlation is indicated."

11 I think there's going to have to be

12 identification of a radiologist at each site,

13 adequate resources directed toward that individual,

14 measurements of the lesions prospectively by that

15 given radiologist.

16 DR. KELSEN: Dr. DuBrow?

17 DR. DuBROW: Can I just add one thing?

18 That is, in your original conception of the

19 protocol that you've built into it radiographic

20 techniques that allow you to compare one study with

21 another so that the exact same technique is used

22 each time on the same type of scanner with the same

361

1 type of intravenous contrast, et cetera.

2 Otherwise, these studies become impossible to

3 compare.

4 DR. KELSEN: Dr. Taylor?

5 DR. TAYLOR: I'm going to have to have a

6 very specific definition of what progression is

7 going to be because I think that can be very vague

8 as well. It makes a study a much more difficult

9 study for those of us who may be in Kansas and who

10 their patient comes in from Winfield to Kansas City

11 with their scans, and it's easier to always do it

12 in Winfield. That's a big deal for some patients,

13 and you may end up scanning them that day

14 emergently, and you're comparing other scans. It

15 makes it more difficult in many ways for the

16 investigator.

17 DR. PAZDUR: To follow up on--remember,

18 this criteria that we used were meant for response

19 criteria, not progression criteria, also, and so we

20 really need to revisit the whole area.

21 DR. TAYLOR: You have to define that.

22 DR. KELSEN: Dr. Redman?

362

1 DR. REDMAN: I can't avoid a political

2 statement. So you're in favor of reinstituting the

3 funding budget to the cooperative groups up to the

4 level that was approved?

5 [Laughter.]

6 DR. PAZDUR: I love all cooperative

7 groups.

8 DR. KELSEN: Other discussion?

9 [No response.]

10 DR. KELSEN: Would you like us to vote on

11 this point for you?

12 DR. PAZDUR: Yes.

13 DR. KELSEN: So I'm going to phrase the

14 question as follows: When a surrogate endpoint for

15 clinical benefit is needed in advanced colon

16 cancer, the preferred endpoint is progression-free

17 survival. Yes means yes, and no would mean that

18 you don't accept that.

19 MS. ROACH: Does yes mean yes with all the

20 caveats we've put in there?

21 DR. KELSEN: Yes always means yes with all

22 the caveats.

363

1 MS. ROACH: Yes.

2 DR. SARGENT: Yes.

3 DR. O'CONNELL: Yes.

4 DR. BRAWLEY: Yes.

5 DR. MARTINO: Yes.

6 DR. TAYLOR: Yes, but I would like to see

7 it validated in some way.

8 DR. REAMAN: Yes.

9 DR. REDMAN: Yes for PFS.

10 DR. KELSEN: Yes.

11 DR. CHESON: Yes.

12 DR. GEORGE: Yes.

13 MS. HAYLOCK: Yes.

14 DR. CARPENTER: Yes.

15 DR. RODRIGUEZ: Yes.

16 DR. DuBROW: Yes.

17 DR. KELSEN: Two unanimous votes.

18 So we will now go to--we're now

19 recommending PFS as the surrogate, and now the

20 question--Dr. Pazdur?

21 DR. PAZDUR: Before we get into Question

22 No. 3, I kind of want to lay out where our

364

1 discussions in the agency have gone, looking at

2 moving away from survival, because I think it's

3 important for people to realize that this has

4 undergone extensive discussion in the agency for

5 years. Okay? And we can't just look at this as,

6 you know, one day we got up and we just think PFS

7 is better than survival.

8 And when you're discussing these

9 questions--and I think this is particularly germane

10 in colorectal carcinoma as we have more and more

11 agents available--the results of the oxaliplatin

12 first-line trial I think is a good example of

13 this--is the effect of--confounding effects of

14 therapies.

15 In essence, when we began our discussion

16 on Monday, which many of you weren't here, we laid

17 out some principles that one reason or several

18 reasons to move away from survival might be some

19 disadvantages. These would include crossover or

20 confounding effects of other therapies, if there

21 was a particularly long follow-up in the natural

22 history of the disease, for example, in indolent

365

1 lymphomas or carcinoids where it would be almost

2 impossible to look at survival data in a

3 meaningfully expedited fashion; and, thirdly, the

4 large numbers of patients that are frequently

5 required.

6 But we have to have a reason of why we're

7 moving away. It can't just be we wake up one day

8 and, okay, we have a new committee here, the

9 committee five years ago or ten years ago voted on

10 survival, and now that there's new members here.

11 So I'd like to hear some discussion of why

12 in this particular disease setting--and perhaps

13 I've already laid it out for you--is the reason.

14 DR. KELSEN: Okay. So we'll open that for

15 discussion. Dr. O'Connell?

16 DR. O'CONNELL: I think you did lay out

17 the--

18 DR. PAZDUR: Not to lead you.

19 [Laughter.]

20 DR. O'CONNELL: Well, you don't have to

21 lead very hard because that's exactly what I think.

22 It's interesting that the treatment of colorectal

366

1 cancer is migrating--is becoming much more similar

2 to the treatment of breast cancer over the years.

3 And many of the issues that we as GI oncologists

4 never had to face before, we're suddenly confronted

5 with. And with the multiple alternative drugs that

6 are now available, it makes it very difficult to

7 use survival as a primary endpoint to evaluate the

8 initial treatment because of the effectiveness of

9 salvage therapy. I think that's the main reason to

10 consider progression-free survival as a valid

11 regulatory endpoint.

12 It's not so much to shrink the sample size

13 or decrease the cost of doing clinical trials or

14 necessarily to make them more efficient. It will

15 achieve all of those effects, but the real reason,

16 I think, is that we now have to contend with--and

17 it's a very good thing to contend with--the

18 effectiveness of salvage therapy.

19 DR. KELSEN: Other comments from the

20 committee? Dr. George?

21 DR. GEORGE: To follow up on that, Mike,

22 are you saying then that progression-free survival

367

1 is a clinical benefit? Because I think we don't

2 have the evidence for the surrogacy issue. But is

3 it a clinical benefit in the same way just by

4 simply delaying progression that is somehow in

5 itself a benefit?

6 DR. O'CONNELL: It's a much more

7 controversial point, I think, than with

8 disease-free survival in the adjuvant situation,

9 because here these patients all have metastatic

10 disease, advanced, incurable malignant disease, by

11 definition in going into the study. There's not

12 the psychological benefit or psychological

13 detriment of realizing that you have a recurrence

14 in the adjuvant situation. You know that you have

15 incurable malignant disease as you go into these

16 treatments. So you don't have that psychological

17 impact in the advanced disease setting.

18 And if one looks at a one- or two- or

19 three-month extension of progression-free survival

20 but pays the price of a 25- or 50-percent rate of

21 grade 3 and 4 toxicity, how much clinical benefit

22 has really accrued to the patient? And so I'm less

368

1 convinced that progression-free survival is of

2 clinical benefit in its own right. I think that it

3 is reasonably predictive of survival. I don't

4 think that the data is nearly as robust for

5 progression-free survival as it is for disease-free

6 survival in the adjuvant situation. But the

7 AstraZeneca data that we heard today, the two

8 trials that Dr. Miller presented, and a

9 meta-analysis that was referred to, all suggested

10 that progression-free survival did have some

11 correlation or surrogacy to overall survival.

12 Now, Dr. Sargent may have some additional

13 information that might tend to go a bit against

14 that argument, and perhaps he should share another

15 meta-analysis that I wasn't aware of.

16 DR. SARGENT: I think the data is actually

17 relatively consistent on this point, and that is

18 that there is a moderate correlation between PFS or

19 TTP and overall survival. The data that was

20 presented today had a proportion explained of about

21 50 percent. Previous analyses have also shown

22 about a 50-percent proportion explained. Some

369

1 relationship--I think it is actually pretty well

2 established that it's not a surrogate marker in

3 this case. I think further analyses are probably

4 required, but there certainly, from my opinion, is

5 not evidence of formal surrogacy in this case.

6 DR. PAZDUR: Could I make a point or ask

7 you a question? This was done from a

8 retrospective--a meta-analysis, I take it, your

9 statements?

10 DR. SARGENT: There is a publication by

11 Burzykowski and colleagues in 2001, Journal of

12 Royal Statistical Society Series C, that actually

13 did explore this exact point. It was a limited

14 analysis, and they actually concluded that there is

15 no evidence to support formal surrogacy of

16 disease--actually, I believe that was TTP and

17 overall survival. Not to say they didn't consider

18 that there was a relationship. There is a

19 relationship, there is a correlation, but it does

20 not meet formal surrogacy criteria.

21 DR. PAZDUR: The issue that I wanted to

22 bring up is if it was a meta-analysis done on

370

1 earlier trials, remember the magnitude of the

2 effect has a great deal to do with the relationship

3 between the surrogate endpoint and the eventual

4 outcome. If we took a look at response rates, for

5 example, in colon cancer, the response rates in the

6 5FU era were 15 percent, with 5FU-leucovorin, and

7 now we're approaching 45, 50 percent in some

8 trials. Again, partial responses.

9 Would you take a look at--do you think

10 that that could have had some influence on it?

11 DR. SARGENT: Absolutely. I think you can

12 only--a surrogate is only as--can only be as strong

13 as the effect is. And if there's a modest effect,

14 then the surrogate can only do so much. So I guess

15 my point is that with respect to 5FU-based

16 treatments where the analyses have really been

17 conducted, the multi-study analyses, they haven't

18 demonstrated it. It indeed become stronger with

19 respect to the new regimens, but those analyses

20 just haven't been conducted at this time point.

21 DR. KELSEN: Ms. Roach?

22 MS. ROACH: I have a question for Dr.

371

1 Hirschfeld or Dr. Keegan. Along this line, as the

2 new--and I know I'm not supposed to talk about the

3 stuff coming down the pike, but there are some

4 things in the pipeline that seem fairly close to

5 coming to FDA for evaluation. And they are much

6 less toxic, or at least that's my impression. And

7 one of the things that--one of the issues that

8 comes up with treatment on a consistent basis is as

9 you're dealing with people who are progressing, you

10 don't want to put them in--I'm sorry. It's been a

11 long day for everybody. You don't want to expose

12 them to a toxic therapy, but if the therapy isn't

13 toxic, does that change the whole endpoint

14 discussion? Does that change the framing of the

15 discussion?

16 DR. KEEGAN: I would say that for some of

17 the biologic products where there's been a

18 perception that they have a relatively modest

19 toxicity profile more in the range which is

20 observed with hormonal therapy, that that has been

21 taken into account in that the presumption is, as

22 for many of the hormonal therapies, rightly or

372

1 wrongly, that there isn't really a lot of

2 treatment-related toxic deaths, which was part of

3 the feeling behind the need to assess survival for

4 the more toxic anti-neoplastic therapies.

5 I think my concern is that we started with

6 the presumption that biologics might not be very

7 toxic, and I think what we're seeing is that what

8 they really have is a very different toxicity

9 profile. For instance, we don't see traditional

10 cytopenias and alopecia, but we see other things.

11 And that I think we don't have a lot of experience

12 weighing into whether or not that could ultimately

13 have a very negative effect both on, you know,

14 quality of life or even survival if they could do

15 that. I think that that's one of the concerns.

16 I think the other is that I'm a little

17 leery of going--I understand what Dr. Sargent said

18 about the fact that we've got a lot of data with

19 anti-neoplastics and we don't have a lot of data

20 with the biologics yet to know if the same

21 predictability, the same relationships are going to

22 hold.

373

1 So I think we're looking at two changing

2 fields at the same time, and it's a little hard to,

3 on the one hand, say, well, I'm sure that all the

4 efficacy relationships will hold but the toxicity

5 issues won't really apply, they shouldn't come into

6 play here.

7 I would rather consider if we were going

8 to treat them in a similar fashion, treat them kind

9 of similar across the board, by and large, and not

10 make a presumption before we have the data that, in

11 fact, they might have the same kind of survival

12 impact or toxicity concerns that some of the more

13 traditional products--or at least not with as

14 little information as we have.

15 DR. HIRSCHFELD: I'd like to respond also.

16 I think the biggest driver in terms of the

17 attractiveness of the therapy is not the anticipated

18 toxicity, but it's the effect size. And I

19 think with the evolution of small molecules as well

20 as the biologics and immunotherapies, we will

21 always evaluate the toxicity versus the benefit in

22 making decisions. And presumably the benefit would

374

1 always outweigh whatever toxic events or adverse

2 events may occur. But what drives the field

3 forward is the effect size, and there we will have

4 to see as these data come in. And then we can go

5 back to Dr. Sargent and ask him for a new analysis.

6 MS. ROACH: I have kind of a follow-up on

7 that or just a comment real briefly. I think this

8 discussion shows how complex this issue is, and I

9 think that transparency of process and product is

10 critically important to bring people along as we're

11 dealing with all of these different shifts in the

12 landscape. And I would urge FDA to be more

13 forthcoming during reviews and approvals. For

14 example, posting the material, the briefing

15 material for ODAC is great. That still leaves an

16 awful lot of products where that kind of material

17 isn't posted. And I think that that's the kind of

18 thing that will help bring the community along and

19 help them understand why you all are choosing to do

20 what you do.

21 DR. HIRSCHFELD: All approved products

22 have--just a point of information, all approved

375

1 products have the reviews posted now on the

2 Internet. The only reviews that are not available

3 publicly are for those products which are not

4 approved at the time they're submitted.

5 DR. KELSEN: Dr. Taylor?

6 DR. TAYLOR: I want to go back to the fact

7 that we haven't validated this. I'm a little bit

8 uncomfortable in that I understand the problems

9 with survival and I'm very accepting that we don't

10 have good ways of determining these things. But

11 we've kind of thrown out response rate for various

12 reasons, partially because we don't think it

13 necessarily correlates with survival, and now we're

14 going to be willing to accept what I would have

15 defined as stable disease on a number of Phase II

16 trials in that patients aren't progressing, they

17 have very stable disease. And I think we do have

18 to validate whether that truly means something.

19 I'm also less willing to say it is a

20 clinical benefit as I see a lot of people in the

21 palliative care setting who are not progressing but

22 have very miserable lives. And you can have a very

376

1 stable disease and have lots of symptoms, and I'm

2 not sure that it--I just really hope that we can

3 find a way to validate this or find some other

4 means.

5 DR. PAZDUR: Let me address that issue.

6 You know, when we're talking about stable disease,

7 we're usually talking about a single-arm trial.

8 Here we would be requiring a randomized study with

9 a robust finding in this, and I think that's other

10 areas that we might want to discuss, how robust,

11 how real that finding is.

12 Remember from my previous comments, we

13 have to first figure out if it's real, and then the

14 robustness of this and its relationship to toxicity

15 comes into play here, to get back to one of the

16 points Nancy was addressing.

17 The other issue that--you know, we are

18 fixated on this correlation between survival and

19 PFS, but remember, one of the other issues that has

20 been promulgated by the agency is not only is the

21 effect of a drug could be manifested by an

22 improvement in the quantity of life, but also in

377

1 the quality of one's life. And I'm not talking

2 about quality-of-life tools here. I'm talking

3 about if people would consider this a relatively

4 established surrogate if one had an improvement in

5 progression-free survival or an improvement in

6 one's quality of life, perhaps even when they're in

7 that progression-free survival zone.

8 DR. TAYLOR: I think that's harder to

9 define, though, and it certainly in a group is a

10 much harder thing to define, because as you work

11 with people and talk with them, there are some who

12 are willing to trade coming to the doctor and

13 taking chemotherapy and others who are not. So I

14 think it's a much more difficult--I'm not sure it's

15 your perfect answer.

16 DR. WILLIAMS: I just want to make the

17 observation, I think Dr. O'Connell suggested that

18 at first recurrence, most patients are

19 asymptomatic; but then they subsequently progress

20 and are symptomatic. So I would guess that at

21 least in the first-line setting, most of those

22 stable disease patients would not be predominantly

378

1 symptomatic until they progressed again. That I'm

2 just reading into your earlier comment, Dr.

3 O'Connell.

4 DR. TAYLOR: I think it depends on when

5 they get to go on the study and when they decide

6 and whether their doctor told them to wait until

7 they were symptomatic to take treatment.

8 DR. KELSEN: Dr. O'Connell?

9 DR. O'CONNELL: I wonder if I can ask Dan

10 to comment on the data that was presented at the

11 workshop where there were 1,000 patients treated

12 with irinotecan-based combination chemotherapy,

13 where there was a substantial difference in

14 treatment effect, about 50-percent objective

15 response rates with the irinotecan combination

16 treatments compared to the controls.

17 In those patients that received the

18 irinotecan-based treatments, time to tumor

19 progression--not progression-free survival in that

20 analysis but time to tumor progression was highly

21 correlated with overall survival, even when

22 corrected for various prognostic discriminants

379

1 within a Cox covariate model. It's not a formal

2 test of surrogacy, but does that data convince you

3 or make you think that time to tumor progression

4 would be a reasonable predictor of survival?

5 DR. SARGENT: It's part of the puzzle, but

6 two trials looking at a single agent I don't think

7 are sufficient evidence, at least to convince me.

8 DR. KELSEN: Other discussion from the

9 committee?

10 [No response.]

11 DR. KELSEN: So a minute ago, if we had to

12 choose a surrogate, we favored PFS. But the

13 question we're being asked now is a different

14 question, so I'll read this again before we vote on

15 it.

16 For approval of drugs for first-line

17 therapy of advanced colon cancer, presumably Stage

18 IV, could PFS/TTP, understanding our previous

19 discussion, benefit of a new drug compared to a

20 standard first-line regimen comparitor on justify

21 regular or full drug approval? And then the agency

22 has got a small comment: Assume the standard

380

1 control arm has a known small survival benefit.

2 So we're now being asked to vote upon,

3 unless we have further discussion, the issue of

4 whether we would recommend regular drug approval.

5 I don't know whether we'll then discuss it would

6 have a role in accelerated approval or whatever.

7 Other discussion before we go to a vote?

8 [No response.]

9 DR. KELSEN: Okay. If not, Ms. Roach?

10 MS. ROACH: Can you start over there this

11 time?

12 [Laughter.]

13 DR. KELSEN: Sure. Dr. DuBrow?

14 DR. DuBROW: Yes.

15 DR. RODRIGUEZ: Since I've gotten less

16 convinced as I've heard later comments, I think my

17 answer is no.

18 DR. KELSEN: John?

19 DR. CARPENTER: I'm going to abstain on

20 this. I'm not sure.

21 DR. KELSEN: Okay.

22 MS. HAYLOCK: Yes.

381

1 DR. GEORGE: No.

2 DR. CHESON: Yes.

3 DR. KELSEN: Yes.

4 DR. REDMAN: Yes, as long as we get to the

5 answers of four.

6 DR. REAMAN: Yes.

7 DR. TAYLOR: No.

8 DR. BRAWLEY: Yes.

9 DR. O'CONNELL: Yes.

10 DR. SARGENT: No.

11 MS. ROACH: No, not until we have all of

12 the above.

13 DR. PAZDUR: A relatively mixed vote, I

14 take it.

15 DR. KELSEN: It's an eight to five vote.

16 DR. PAZDUR: Eight to five. Okay. Let me

17 throw out this suggestion for you. How about we're

18 in a situation where we have a reason--we have an

19 improvement in progression-free survival or time to

20 progression, and the survival advantage is not

21 demonstrated; however, there is convincing evidence

22 that there has been crossover of therapies that

382

1 could explain why we're not seeing a survival

2 advantage. Should we accept in that situation the

3 effect on the "surrogate" of time to progression?

4 And this is a real live example of many years ago.

5 DR. KELSEN: Yes, it certainly is. Open

6 for discussion. There's a confounding variable

7 that may have affected survival.

8 DR. PAZDUR: You can postulate a reason

9 why you have not demonstrated a survival effect,

10 for example, confounding of the survival analysis

11 by crossover.

12 DR. BRAWLEY: But, Rick, by the same

13 token, a placebo would do the same thing.

14 DR. PAZDUR: No, I'm talking about if you

15 have a known--you know, if you have, say, a

16 standard therapy or some--you know, not a

17 placebo-controlled trial we're talking about. I'm

18 talking about if you have a reason to deviate from

19 your suggestion here, would there--let me ask it in

20 another way. Is there any situation where you

21 might deviate from this suggestion?

22 DR. SARGENT: I would deviate if two

383

1 circumstances were present: A, we have substantial

2 evidence of differential crossover; and, B, there

3 is a trend in survival in the appropriate

4 direction. It may not be significant, but at least

5 it's consistent with the PFS results.

6 DR. PAZDUR: Just to clarify, I realize

7 you answered the question in the affirmative. For

8 those who felt negatively about it, okay? And

9 that's who I'm addressing this question to. Would

10 there be--

11 DR. KELSEN: I think it's appropriate--

12 DR. PAZDUR: --sensitivity to not being so

13 dogmatic as saying, no, I will only accept survival

14 in those people.

15 DR. KELSEN: And this further discussion

16 is appropriate because, clearly, the magnitude of

17 the vote indicated how big the unease is and how

18 controversial this point might be.

19 Is there any other discussion? Yes, Dr.

20 George?

5B DR. GEORGE: I think my unease about it

22 was because, unlike disease-free survival in the

384

1 adjuvant setting, progression-free survival I don't

2 think has been established in the same way as the

3 surrogate, nor is it obvious to me that it's the

4 same--has inherently something in it that's a

5 clinical benefit all by itself.

6 Now, with respect to the crossover issue,

7 I've made this point before, but no one seems to

8 listen, but I'll say it again just for a general

9 point. That is, I think this falls into the

10 example of something where you would like to get

11 the answer to something but you can't get it; that

12 is, you say I have this new treatment and I'm going

13 to compare it to the standard. I'd really like to

14 know whether it prolongs survival, but I'm giving

15 this very early in the disease. I have all these

16 other things that are liable to be given at some

17 point for some reasons that I can't control. And

18 all I can do is I'm doing this randomized study,

19 and I'll observe what happens.

20 My point about this is that is the answer;

21 that is, even--no matter what you try to do to try

22 to explain it, the answer is if I start off trying

385

1 to give these two treatments, in the current

2 setting with the available therapies and the

3 real-world situation, this therapy did not prolong

4 survival. Now, you can give reasons; it may be

5 because of crossover, may be because of other

6 therapies that were given. The answer is still the

7 same. It didn't prolong survival.

8 So that's when you would definitely like

9 to have something that could give you some answer

10 that, like progression-free survival might tell you

11 something biologically and say, all right,

12 something's going on here with this therapy. But

13 in the real-world setting, it doesn't prolong the

14 survival. So that's the answer with survival. So

15 if I'm stuck with survival--

16 DR. PAZDUR: Would you buy, for example,

17 progression-free survival in that situation, or in

18 any situation, to reasonably likely predict

19 clinical--

20 DR. GEORGE: Yes, that's what I was going

21 to get to. I think that it's like an accelerated

22 approval kind of thing. I don't know if you may

386

1 want to talk about that.

2 DR. PAZDUR: Let me give you the scenario

3 here so I think you people could understand the

4 real-world situation that we face frequently.

5 Obviously, people develop drugs and they are highly

6 touted to be very effective therapies, and there's

7 great interest on the part of patients to receive

8 these therapies before they are approved. Many

9 times we're requested both in the first-line

10 setting, and even more advanced disease setting,

11 that at the time of progression people will get a

12 perceived effective therapy even though it hasn't

13 been approved. And, therefore, we can get into

14 problems when we have a survival analysis because,

15 you know, both the groups of patients that are

16 randomized eventually will get the drug.

17 We saw that, for example, in the

18 third-line setting with oxaliplatin where the vast

19 majority of patients entered on the trial in the

20 third-line setting, I'm talking about, got the

21 drug. More, I think, for the more advanced disease

22 setting, the later disease--

387

1 DR. KELSEN: Since this is now a more

2 pressing issue, let me just look at Question 4, as

3 you wrap it into your discussion, because now that

4 we've indicated by a split vote that PFS/TTP might

5 be an acceptable standard for regular drug

6 approval, the agency wants to know a little bit

7 more, wants to know--does that mean they have to

8 have a big difference between these groups? Could

9 we discuss the magnitude of that difference?

10 We frequently are talking about trial and

11 trials, so could the committee comment on point 4?

12 Dr. Redman?

13 DR. REDMAN: Yes, just for those that

14 voted no, I mean, is it an absolute, or is it a

15 degree? If you have a randomized trial, drug A

16 versus drug B, in a metastatic setting and the

17 progression-free survival of the standard is two

18 and the progression-free survival is ten months,

19 and yet it's going to take another three years to

20 find an overall survival advantage, is it an

21 absolute no, you won't accept that? Or is it just

22 a degree? Because I think that is what 4 is

388

1 asking. I mean, nobody's going to say, gee,

2 there's a three-week progression-free survival, you

3 know.

4 DR. KELSEN: Well, I think the agency is

5 asking that question. That's exactly the question

6 that--

7 [Simultaneous conversation.]

8 DR. REDMAN: I think that's dependent on

9 the drug and its side effects, and I think that's

10 what clinical medicine is. You can't make a cutoff

11 and say, gee, you know, if it's one month, six

12 weeks, you know, if we're going to do bone marrow

13 transplant, you're going to be in the hospital for

14 four weeks to get a four-week progression-free

15 survival, I mean--

16 DR. PAZDUR: --asking the question because

17 I want to get some degree of flexibility here on

18 where people stand, because people see these votes

19 and obviously can come down and say, well, ODAC

20 said this; therefore, you must adhere to this. And

21 I'm just wondering if we could have more discussion

22 on people's flexibility on this point.

389

1 DR. KELSEN: So could we have comments--

2 DR. PAZDUR: Maybe magnitude of

3 difference.

4 DR. KELSEN: Yes. Dr. Taylor, then Dr.

5 Cheson.

6 DR. TAYLOR: I think that, as Bruce has

7 said, I think you have to individualize it. I

8 certainly wouldn't--I could be flexible if I saw a

9 dramatic difference between it. But I think that

10 we are choosing something that we haven't done

11 before, and we have to be very cautious. And

12 certainly some of the drugs we looked at yesterday,

13 they would have had a--it had been on the market,

14 and I don't think that would have been appropriate.

15 DR. KELSEN: Dr. Reaman and then Dr.

16 Cheson.

17 DR. REAMAN: I would be flexible also,

18 although I voted affirmative. But I think in

19 general, the magnitude would have to be very

20 significant.

21 DR. PAZDUR: You're not talking

22 statistical significance.

390

1 DR. REAMAN: Clinically significant, not

2 statistically.

3 DR. CHESON: I think it's not only

4 quantity but it's quality, and one of the

5 discussions yesterday we were talking about was

6 there some change in performance status, was there

7 some change in symptoms. And there's a difference

8 between two months of good life and two months of,

9 as Sarah was talking about before, really poor

10 quality of life. So I think you have to be

11 flexible and individualize somewhat both on the

12 duration and what that duration means to the

13 patients. And for those sorts of studies, you

14 should encourage them to obtain that other

15 information such as functionality--not necessarily

16 formal fact quality of life and those sorts of

17 things, although it's not a bad idea, but to get

18 other measurements that would support it.

19 DR. PAZDUR: One of the things as we asked

20 in--perhaps--I don't know if you want to comment

21 about it, we would ask or we have been asking--in

22 discussing about this, asking for the trials to be

391

1 powered for survival, obviously, and to look at

2 that issue also.

3 DR. KELSEN: Dr. Carpenter?

4 DR. CARPENTER: It might be helpful if we

5 just said that I think most of us would be looking

6 in terms of months as opposed to days and weeks.

7 DR. O'CONNELL: Yes.

8 DR. CARPENTER: As far as an increase, if

9 you were to give an order of magnitude. Then if

10 you're talking about months, the other things that

11 would be critically important would be the things

12 that Dr. Cheson mentioned.

13 DR. KELSEN: Dr. Reaman?

14 DR. REAMAN: I'm just going to follow up

15 on Dr. Cheson's comment, and, Rick, I think you

16 mentioned that you were going to be preparing a

17 guidance to industry, and I think it would be very

18 important to include as part of that the

19 suggestion, if not the requirement, to do formal

20 quality-of-life questions or to address those

21 issues.

22 DR. KELSEN: Ms. Roach?

392

1 MS. ROACH: My mom would love it if I was

2 a doctor.

3 I think the problem with black-and-white

4 answers on all this, while I understand you'd like

5 certainty, is that there's always a degree of

6 judgment. And so I think looking at it in terms of

7 where we want to get and did we get there is maybe

8 more helpful. So the orders of magnitude that you

9 all are talking about are right by my perspective.

10 I also think that in terms of the

11 evidentiary requirements, there is a ton of really

12 interesting and intriguing imaging things coming

13 down the pike, with volumetric measures and

14 activity and things like that. And I think if we

15 could use some of what we do here to validate the

16 technology as well as validate the drug, it would

17 be helpful to everyone.

18 And I also want to put in a plug for

19 putting the funding back to the cooperative groups.

20 DR. KELSEN: So if I could summarize, what

21 I think we've heard is that the committee was for

22 and has added guidance about it being but a

393

1 substantial difference in PFS/TTP, and the

2 magnitude of the evidence would be quite

3 convincing.

4 Do you want us to discuss 5 as well

5 or--okay. So I'll go to the last point on the

6 agenda, which is: If one accepts PFS/TTP, what, if

7 any, survival evidence should be needed? And the

8 agency specifically wants to know whether the

9 studies should be powered to rule out a negative

10 impact on survival and whether or not they should

11 be, on the converse, powered to look for a

12 realistic improvement in survival. So if one

13 accepted TTP or PFS. Dan, if you want to make a

14 comment, or Steve?

15 DR. SARGENT: Well, my comment with

16 respect to 5(b) is a three-month improvement in TTP

17 that might translate into a three-month improvement

18 in overall survival are very different elements.

19 And requiring a trial, given the answer to No. 3

20 was yes, requiring a trial to be powered for

21 overall survival may indeed be prohibitive given a

22 modest benefit that may be expected in overall

394

1 survival.

2 DR. WILLIAMS: I think this was written

3 considering a very realistic setting, whereas the

4 competitor drug out there does have a survival

5 advantage of two years. And I think, you know--two

6 months, I'm sorry. Right. So in that setting, I

7 mean, you have to think about would you or would

8 you not be ruling out that you were inferior to the

9 other drug.

10 DR. SARGENT: I think 5(a) is very

11 reasonable. To rule out a decrement is very

12 different than having power to demonstrate an

13 improvement. And so I think my opinion on 5(a) is

14 yes; 5(b) is probably no.

15 DR. KELSEN: Dr. George?

16 DR. GEORGE: When I first saw 5(a), I

17 interpreted it a little differently. I thought you

18 were looking at some non-inferiority trial which

19 would--I would have said no because it's huge. But

20 to rule out a specific decrement, I think it's a

21 good idea to look for that to make sure that it's

22 not done and to have that prespecified is a good

395

1 idea.

2 For (b), I'm a little less sure. It

3 depends on the--you know, I guess the realistic

4 improvement, I don't know what that means exactly,

5 but--

6 DR. PAZDUR: Well, even now, obviously,

7 when we ask for powers to be--the trials to be

8 powered, there is a guesstimation of an effect.

9 One of the reasons why we are interested in this is

10 obviously we are facing increasing numbers of

11 single trials that are coming to us, and sometimes

12 trials that are underpowered, which leaves

13 everybody in a quagmire of what to do with these

14 trials. Do we have a real treatment effect? A

15 fear that if we go to a time to progression or

16 progression-free survival that would require a

17 fewer number of patients, we'll see a gradual

18 decrease in the size of patients numbers that are

19 being entered on trials.

20 Again, if we never ask a survival question

21 and power for some type of survival improvement,

22 okay--and remember, we're asking for a robust

396

1 finding in time to progression which would probably

2 translate into a smaller survival effect. One

3 should be able to see that.

4 DR. GEORGE: What I was suggesting in this

5 setting, it would be not necessarily to power to

6 detect realistic improvement, as you've stated it

7 here, but to design the study appropriately based

8 on the time to progression and then look at--then

9 address the issue carefully of what that means,

10 what kind of things you could pick up with respect

11 to survival, and when you could pick them up, and

12 make sure that--I mean, I'm just saying this in

13 sort of a subjective way. You'd have to just

14 assess whether that seems reasonable, in other

15 words, not do it in the usual way you design a

16 study where you say I'm trying to pick up this kind

17 of difference, but just say in this setting I can

18 pick up this sort of difference at this time during

19 the analysis. You know, giving plots, in other

20 words, instead of just picking a point and saying I

21 have a specified power at this alternative.

22 DR. WILLIAMS: I have to say I interpreted

397

1 the question a little bit differently. I actually

2 helped to write it. But to rule out a survival

3 decrement, I mean, one interpretation could be

4 there's another drug out here with a two-month

5 benefit and maybe I'm being compared to it. I want

6 to make sure I haven't lost some of that. So that

7 could be not very different from a non-inferiority

8 study; whereas, (b), you know, somebody can always

9 make some idea of how much survival you might be

10 detecting. So I'm not sure--do you have any

11 guidance on (a) what we should be looking for?

12 We've got a little progression advantage, perhaps

13 substantial, compared to a drug that has a

14 two-month increase in survival. Do we need in any

15 way to rule out we're losing that, or we just

16 presume that we're not since we're--

17 DR. KELSEN: One of your problems would be

18 because, as I think Mike said before--he had to

19 leave--with more and more new agents coming down

20 the line, where are you going to see where you lost

21 the survival? I mean, how will you do that? And

22 that's what I was wondering, because you'll now

398

1 have a first-line therapy, a second-line. We're

2 talking about a third-line therapy. You know, we

3 might get like breast cancer and have fourth-line

4 and fifth-line therapies. And so where was it lost

5 in this off-protocol, presumably, list of agents

6 that the patient got? And I'm not sure

7 procedurally how you'll be able to identify that

8 quite so easily, but I'd be interested in how it

9 statistically could be approached.

10 DR. GEORGE: This could be a real problem

11 if you're looking at it the way you just expressed,

12 as some kind of non-inferiority. It would be a

13 real problem in doing the studies. And I think

14 that's not what you want to do because that's--you

15 know, they would be huge trials to answer--I mean,

16 to not really address the really important

17 questions in this area.

18 So I think you have to do something kind

19 of pragmatic, is what I'm thinking here, that you

20 would specify in the design something about what

21 you're going to be looking for. But you don't

22 design the study to be definitively sure that

399

1 you're not more than some small decrement below the

2 control.

3 DR. WILLIAMS: So you're suggesting some

4 kind of--basically a safety type decrement, in

5 other words, I can rule out that, you know, I've

6 induced some sort of survival decrement.

7 DR. KELSEN: I think one of the problems,

8 you know, in this, since we're talking about a

9 disease, in this disease, not like small-cell lung

10 cancer, for example, if the patients who enter the

11 first-line study are really almost asymptomatic or

12 to a large amount asymptomatic, you have this

13 window. So it's not--if you don't get them into

14 remission on the first regimen you won't have time

15 to get to that second regimen. And so it would be

16 really hard for you to look, I think, for that you

17 lost two months somewhere in there, but you'll have

18 hopefully more than one shot with currently

19 available therapy.

20 I'm sorry. Ms. Roach, you had a question?

21 Oh, Dan?

22 DR. SARGENT: I think my proposal would be

400

1 somewhat of a confidence interval-based approach

2 where we may not see an advantage for overall

3 survival, but we have a sufficient sample size to

4 estimate our confidence interval around our

5 estimated effect on overall survival that does

6 exclude a decrement in survival. So hopefully the

7 hazard ratio, you know, may not be significant, but

8 at least is in the right direction and the

9 confidence interval is tight enough that we're sure

10 that it's not indeed a decrement.

11 DR. WILLIAMS: And I guess the $100

12 question is: What is the size of that decrement?

13 DR. SARGENT: I think it's relevant to

14 what the improvement was compared to the previous

15 standard.

16 DR. KELSEN: Have we been able to answer

17 the questions that the agency posed? Are there any

18 other questions that you'd like us to discuss or

19 any other points you'd like us to discuss today?

20 DR. PAZDUR: We did have the rectal cancer

21 question, and I don't know if that is something

22 people would entertain at this time, whether a

401

1 difference with rectal recurrence would signify

2 clinical benefit. And we're talking about probably

3 adjuncts to radiation therapy, that type of a

4 situation.

5 DR. KELSEN: And just to refresh people's

6 memory, when Dr. O'Connell gave his presentation,

7 you might remember that he said the third point

8 from the workshop, in addition to what we've

9 covered today, was a recommendation that at

10 three-year disease-free survival--sorry, three-year

11 freedom from local failure in rectal cancer was a

12 very meritorious thing to have because of the

13 symptoms, and the agency I think is asking for

14 guidance and what's the view of the committee in

15 view of that.

16 DR. CARPENTER: That's regularly

17 associated with symptoms, and it seems to be--that

18 seems to be easier because clear delay or avoidance

19 of major symptoms is just going to be a benefit, it

20 seems to me.

21 DR. KELSEN: Yes, the issue they'd face

22 is, you know, how you validate that they failed

402

1 locally but that's an imaging--doing careful

2 imaging question.

3 DR. CARPENTER: That's definable.

4 DR. KELSEN: Yes, that's definable.

5 DR. CARPENTER: You could make some

6 criteria of how you're going to do that.

7 DR. KELSEN: Is there any other discussion

8 about that?

9 [No response.]

10 DR. KELSEN: Because we're making a broad

11 recommendation in a few minutes. But it sounds

12 like there's support for the workshop's conclusion

13 that this is an important issue for the agency to

14 consider as a different way of approving an agent.

15 Any other issues you would like us to

16 discuss?

17 DR. PAZDUR: Not that I am aware of. I'm

18 cognizant of the short discussion on this. We

19 would bring it back to the committee or for

20 external discussions with our consultants before

21 we'd make any final agreements regarding the latter

22 point, because I realize we haven't had sufficient

403

1 discussion on that. And plus many of the members

2 have already left.

3 DR. KELSEN: If that's the case and

4 there's no further discussion, I want to thank the

5 members of the committee for their participation

6 today, also for the opportunity to chair the

7 session. Thank you very much.

8 [Whereupon, at 4:44 p.m., the meeting was

9 concluded.]

10 - - -