1

DEPARTMENT OF HEALTH AND HUMAN SERVICES

FOOD AND DRUG ADMINISTRATION

CENTER FOR DRUG EVALUATION AND RESEARCH

ONCOLOGIC DRUGS ADVISORY COMMITTEE

VOLUME I

Thursday, March 3, 2005

8:05 a.m.

Gaithersberg Hilton

620 Perry Parkway

Gaithersburg, Maryland

PARTICIPANTS

Silvana Martino, D.O., Acting Chair (A.M. Session)

Maha Hussain, M.D., Acting Chair (P.M. Session)

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

COMMITTEE MEMBERS

Otis W. Brawley, M.D.

Ronald M. Bukowski, M.D.

James H. Doroshow, M.D.

Antonio J. Grillo-Lopez, M.D., Industry

Representative

Pamela J. Haylock, RN, Consumer Representative

Maha H.A. Hussain, M.D.

Alexandra M. Levine, M.D.

Joanne E. Mortimer, M.D.

Michael C. Perry, M.D.

Gregory H. Reaman, M.D.

Maria Rodriguez, M.D.

CONSULTANTS (VOTING)

FOR COMBIDEX

Marco Amendola, M.D.

William Bradley, M.D., Ph.D.

Marion Couch, M.D., Ph.D.

Ralph D'Agostino, Ph.D.

Mark Dykewicz, M.D.

Armando Giuliano, M.D.

Dennis Ownby, M.D.

Dana Smetherman, M.D.

PATIENT REPRESENTATIVE (VOTING)

Eugene Kazmierczak - for Combidex and Prostate

Cancer Endpoints

CONSULTANTS

PROSTATE CANCER ENDPOINTS

Victor DeGruttola, Sc.D.

Mario Eisenberger, M.D.

Eric Klein, M.D.

Lisa McShane, Ph.D.

Derek Raghavan, M.D., Ph.D.

Howard Sandler, M.D.

Howard Scher, M.D.

PARTICIPANTS (Continued)

FDA (A.M. Session)

Zili Li, M.D., MPH

Florence Houn, M.D.

George Mills, M.D.

Sally Loewke, M.D.

PARTICIPANTS (Continued)

FDA (P.M. Session)

Peter Bross, M.D.

Patricia Keegan, M.D.

Bhupinder Mann, MBBS

Richard Pazdur, M.D.

Dan Shames, M.D.

Rajeshwari Sridhara, Ph.D.

Robert Temple, M.D.

Grant Williams, M.D.

C O N T E N T S

Page

Call to Order and Introductions

Silvana Martino, D.O. 6

Conflict of Interest Statement

Johanna Clifford, M.S., RN 9

Opening Remarks

George Mills, M.D. 11

Sponsor Presentation

Advanced Magnetics, Inc.

Combidex, Introduction and Indication

Mark C. Roessel 15

Mechanism of Action, Combidex

Appearance on MR Images

Mukesh Harisinghani, M.D. 17

Efficacy Data from Phase III Clinical Studies

William Goeckeler, Ph.D. 29

Safety Data from Clinical Trial

Gerald Faich, M.D. 39

Clinical Utility of Combidex and Various Cancers

Jelle O. Barentsz, M.D. 46

FDA Presentation

Efficacy and Safety of Combidex (NDA 21-115)

Zili Li, M.D., MPH 56

Questions from the Committee 88

Open Public Hearing 146

Committee Discussion 167

C O N T E N T S (Continued)

Page

AFTERNOON SESSION

Call to Order and Introductions

Maha Hussain, M.D. 204

Conflict of Interest Statement

Johanna Clifford, M.S., RN 207

Opening Remarks

Richard Pazdur, M.D. 210

A Regulatory Perspective of Endpoints to

Measure Safety and Efficacy or Drugs:

Hormone Refractory Prostate Cancer

Bhupinder Mann, MBBS 216

Towards a Consensus in Measuring Outcomes

in New Agents for Prostate Cancer

Derek Raghavan, M.D., Ph.D. 227

NCI Prostate Cancer Treatment Trial Portfolio

Alison Martin, M.D. 261

Toward an Endpoint for Accelerated Approval

for Clinical Trials in Castration Resistant/

Hormone Refractory Prostate Cancer

Howard Scher, M.D. 271

Design of Clinical Trials for Select Patients

With a Rising PSA Following Primary Therapy

Anthony D'Amico, M.D., Ph.D. 297

Open Public Hearing 330

Committee Discussion 333

P R O C E E D I N G S

Call to Order and Introductions

DR. MARTINO: Good morning, ladies and

gentlemen. I would like to begin the meeting, if

you would be so kind as to take your seats.

The purpose of this morning's meeting is

to consider a new drug application, the agent

Combidex from Advanced Magnetics, Incorporated, a

proposed indication for intravenous administration

as a Magnetic Resonance Imaging contrast agent to

assist in the differentiation of metastatic and

non-metastatic lymph nodes in patients with

confirmed primary cancer who are at risk for lymph

node metastases.

We will start the meeting by having the

members of the panel introduce themselves, and I

would like to begin on my left, please.

DR. LOEWKE: Sally Loewke, FDA. I am the

Deputy Division Director for the Division of

Medical Imaging and Radiopharmaceutical Drug

Products.

DR. MILLS: Good morning. I am George

Mills, FDA. I am the Division Director for Medical

Imaging.

DR. HOUN: Florence Houn, Office Director,

FDA.

DR. LI: Zili Li, Medical Team Leader,

FDA.

MR. KAZMIERCZAK: Eugene Kazmierczak,

Patient Consultant to FDA for prostate cancer.

DR. BUKOWSKI: Ron Bukowski, Medical

Oncologist, Cleveland Clinic Foundation.

DR. BRAWLEY: Otis Brawley, Medical

Oncologist and Epidemiologist, Emory University.

DR. DOROSHOW: Jim Doroshow, Division of

Cancer Treatment and Diagnosis, NCI.

DR. RODRIGUEZ: Maria Rodriguez, Medical

Oncologist, M.D. Anderson Cancer Center.

DR. REAMAN: Gregory Reaman, Pediatric

Oncologist, Children's Hospital, Washington, D.C.,

and George Washington University.

DR. MARTINO: Silvana Martino, Medical

Oncology, Cancer Institute Medical Group in Santa

Monica.

MS. CLIFFORD: Johanna Clifford, Executive

Secretary to the Oncology Drugs Advisory Committee.

DR. HUSSAIN: Maha Hussain, Medical

Oncologist, University of Michigan.

DR. PERRY: Michael Perry, Medical

Oncologist, Ellis Fischel Cancer Center, Columbia,

Missouri.

DR. MORTIMER: Joanne Mortimer, Medical

Oncologist, Moores UCSD Cancer Center.

DR. OWNBY: Dennis Ownby, Pediatric

Allergist at Medical College of Georgia.

DR. D'AGOSTINO: Ralph D'Agostino,

Biostatistician from Boston University.

DR. DYKEWICZ: Mark Dykewicz, Professor of

Internal Medicine, Allergy and Immunology, Training

Program Director, St. Louis University.

DR. GIULIANO: Armando Giuliano, Surgical

Oncologist from Los Angeles.

DR. BRADLEY: Bill Bradley. I am a Neuro

MRI guy. I am the Chairman of Radiology at UCSD.

DR. AMENDOLA: Marco Amendola, Professor

of Radiology, University of Miami.

DR. SMETHERMAN: Dana Smetherman,

Radiologist, Section Head of Breast Imaging,

Oschner Clinic.

DR. COUCH: Marion Couch, Head and Neck

Surgeon from the University of North Carolina.

DR. MARTINO: If you would all turn off

your mikes, and for those of you that are new to

the committee, please recognize that you need to

speak into the microphone, and it only works when

you have pushed it and the red light is on. Once

you are done with its use, please turn it off.

There is a reasonable amount of echo that

I still hear in this room. Can Audiovisual do

anything more to clarify our sound? Okay.

At this point, Ms. Johanna Clifford will

report on the Conflict of Interests.

Conflict of Interest Statement

MS. CLIFFORD: The following announcement

addresses the issue of conflict of interest and is

made a part of the record to preclude even the

appearance of such at this meeting.

Based on the submitted agenda and all

financial interests reported by the committee

participants, it has been determined that all

interests in firms regulated by the Center for Drug

Evaluation and Research present no potential for an

appearance of a conflict of interest.

With respect to the FDA's invited industry

representative, we would like to disclose that Dr.

Antonio Grillo-Lopez is participating in this

meeting as an acting industry representative acting

on behalf of regulated industry. Dr. Grillo-Lopez

is employed by Neoplastic and Autoimmune Disease

Research.

In the event that the discussions involve

any other products or firms not already on the

agenda for which an FDA participant has a financial

interest, the participants are aware of the need to

exclude themselves from such involvement, and their

exclusion will be noted for the record.

With respect to all other participants, we

ask in the interest of fairness that they address

any current or previous financial involvement with

any firm whose products they may wish to comment

upon.

Thank you.

DR. MARTINO: Dr. Mills, if you would

address the group.

Opening Remarks

DR. MILLS: Thank you, Dr. Martino.

Good morning, Committee. The sponsor of

the application in this morning's session, Advanced

Magnetics, requests marketing approval of Combidex

for the proposed indication of assisting in the

differentiation of metastatic and non-metastatic

lymph nodes, in patients with confirmed primary

cancer, who are at risk for lymph node metastases.

The Agency is asked to consider an

indication specifically for differentiating

metastatic from non-metastatic lymph nodes with

little restriction on the cancer type, clinical

staging, and whether the patients have been

previously treated.

The Agency is in the second review cycle

for this imaging product. The first review cycle

concluded with an approvable action and the sponsor

was asked to conduct additional studies to address

issues related to inconsistent efficacy results

among the differential trials and to provide a

clearer identification for the conditions of use

for Combidex.

In addition, the sponsors were asked to

address safety issues related to Combidex-induced

hypersensitivity reactions.

In today's presentation, the sponsor will

address these deficiency issues by using data that

were originally submitted to the Agency, along with

new information from a published study in the New

England Journal of Medicine.

The Agency's presentation today will focus

on whether the primary analyses that were based on

99 subjects from the U.S. studies and only 48

subjects from the European studies are adequate for

marketing approval based on the sponsor's proposed

indications, which reads as follows:

"Combidex is for the intravenous

administration as a contrast agent for use with

MRI. Combidex can assist in the differentiation of

metastatic and non-metastatic lymph nodes in

patients with confirmed primary cancer who are at

risk for lymph node metastases."

Today, we will be seeking comments on the

issues related to the sample size and the adequacy

of tumor type presentation. We will be presenting

the variable efficacy results by the tumor type and

the size of the lymph nodes.

We are seeking your opinion as to whether

these results suggest that the variations in

efficacy performance of Combidex are related to the

different tumor types and to different lymph node

sizes.

Today, we are seeking your advice on how

to better define the conditions for use for

Combidex, assuming the validity of the efficacy

results, so that use of Combidex can provide

benefits to patients particularly in affecting

patient's treatment decisions. This point is

particularly important given the risks of

hypersensitivity reactions associated with

Combidex.

Lastly, we will be seeking your

recommendations on what additional data are needed

if current data are found to be inadequate for the

marketing approval of Combidex at this time.

This concludes the Agency's introduction

to the morning session.

Thank you, Dr. Martino.

DR. MARTINO: Thank you.

For those of you that are new to the

committee and are consulting to the committee, the

final task that we will bring to you is answers to

certain questions that have been posed to the

committee by the FDA. Those are in a written

format and each of you should have those at your

desk.

They are titled as Discussion and

Questions, so please recognize that it is very

specifically to answer those four questions which

will be the focus of the discussion at the end of

this morning's presentations.

At this point, I would like to ask Dr.

Roessel from the company to introduce their

speakers and proceed with their presentation.

There will be an opportunity for questions

both to the sponsor, as well as to the FDA. I ask

that you hold your questions until their

presentations are completed.

Sponsor Presentation

Advanced Magnetics, Inc.

Combidex, Introduction and Indication

MR. ROESSEL: Good morning. Thank you,

Madam Chairman, members of the Advisory Committee,

FDA.

I am Mark Roessel, Vice President of

Regulatory Affairs, Advanced Magnetics.

Today is an important day for us as we

have been working since 1992 to bring Combidex to

clinicians and cancer patients. We are pleased to

be able to show you today data from controlled

clinical trials demonstrating the safety and

efficacy of Combidex and the great potential it has

for improving imaging in cancer patients.

We have a number of distinguished

consultants and speakers here today including

radiologists, surgeons, oncologists, and they are

available to answer any questions you may have at

the end of the meeting.

I want to bring your attention to the

indication. It has been read twice already. It is

for a differentiation of metastatic and

non-metastatic lymph nodes in cancer patients.

Here is the agenda we are going to have in

our presentation and the key topics. Dr. Mukesh

Harisinghani is going to show you the mechanism of

action of Combidex and how it appears on MR images.

Dr. William Goeckeler from Cytogen

Corporation, Vice President of Cytogen, who is our

marketing partner, is going to present to you data

from Phase III controlled clinical trials that were

designed in cooperation with the FDA for approval

of the agent.

Dr. Jerry Faich is going to review the

safety data available, demonstrating that Combidex

can be safely administered using dilution and

infusion.

Finally, Dr. Jelle Barentsz, a clinical

investigator with Combidex, is going to review with

you the clinical utility of Combidex in various

cancers.

Combidex is a diagnostic tool that

improves the anatomic imaging that is done every

day.

Now, I would like to have Mukesh

Harisinghani.

Mechanism of Action, Combidex

Appearance on MR Images

DR. HARISINGHANI: Good morning, Madam

Chairman, members of the Committee, ladies and

gentlemen.

What I am going to do in the next couple

of minutes is to review what are the current

limitations of lymph node imaging as we practice

radiology today, also give an overview of how

Combidex is acting and how it allows us to

differentiate benign from malignant lymph node, and

then also show you some examples of how it improves

sensitivity and specificity for nodal

characterization.

So, the question is why do we need to

image lymph nodes, and I think one needs to

accurately stage primary cancer, and in doing so,

it is very important to know what the nodal status

is.

It is very important to know this

information to appropriately treat the patients.

Just to give you an example, in prostate cancer

patients, if the nodes are found to be metastatic,

it essentially commits the patients to non-surgical

modes of therapy.

We also need to get a sense of prognosis,

and that is another factor why nodal metastases are

important. Again, to give you an example in

bladder cancer, if the patient is node-positive,

the five-year survival is way lower than if the

patient is node-negative.

The risk of death also increases 20

percent with each additional node being positive.

The current lymph node staging as is

performed today involves non-invasive imaging

techniques, which essentially incorporates the

cross-sectional modalities like CT and MR, and the

other is the invasive modes, which is essentially

surgery, which are considered to be the gold

standard today.

When one talks of the non-invasive

cross-sectional modalities for staging lymph nodes,

the predominant yardstick by which we differentiate

benign from malignant lymph nodes is the size

criterion, and this is what we use.

If the node is oval and less than 10 mm in

size, or if it is rounded and less than 8 mm in

size, we label the node as benign.

In contrast, if the node is oval and

greater than 10 mm, or is rounded and greater than

8 mm, we label the node as malignant.

So, let's apply the size criterion to

these two individuals. These are two different

patients, both have obtained a CT scan for staging

purposes.

The example on your left is an enlarged

node in the pelvis, which measures 18 mm and is

rounded. No matter which size criterion you use,

you would label this node as malignant.

The example on your right is a different

patient, again a patient with a primary pelvic

tumor. There is a small node in the pelvis, which

measures 5 mm. Again, no matter which size

criterion you use, you would label this node as

benign.

But at surgery, it was exactly the

opposite. Thus, you can see that size criterion is

an inaccurate yardstick by which we categorize

nodes today.

Morphology has been to a certain extent

used in conjunction with size criteria

occasionally, and one of the important morphologic

features we rely on is presence of fatty hilum, as

you are seeing here.

It is said that if the node has a central

fatty hilum, that is a sign of benignity, however,

we have seen from our experience that even small

nodes, as the case here, with the fatty hilum in

this patient with bladder cancer, was biopsy proven

to be positive and having malignant cells.

Thus, morphology, too, has its drawbacks

and when used with size criterion, can be a

problem.

Central necrosis is the other morphologic

feature which has occasionally been said to be a

very useful way to allow for diagnosing malignant

nodes, but it is important to realize that when

nodes become necrotic, they are enlarged beyond a

cm, and by size criterion, you would still call

them positive.

Well, what about surgery, which is

considered to be the gold standard, and I am going

to use prostate cancer as an example, but I think

the underlying principle can be applied or

extrapolated to other tumors, as well.

In prostate cancer, pelvic lymph node

dissection accompanied by frozen section path

examination is considered to be the gold standard.

However, the way lymph nodes are sampled today, at

the time of surgery in intermediate to high risk

prostate cancer patients, the standard pelvic

lymphadenectomy is limited. This is because the

surgeon only resects the low external iliac and the

obturator group of lymph node.

In the recent or not too recent, in an

April 2000 study published in the Journal of

Urology, it was shown that if the surgeon extends

the lymphadenectomy and takes out the high external

iliac and the internal iliac nodes, keeping all

other risk factors the same, the incidence of lymph

node metastases jumps from 10 to 26 percent, so you

can see that a potential of 16 percent miss rate if

one just follows the standard pelvic

lymphadenectomy.

So, that begs that question why don't we

do that in all the cases, because there is a

significant morbidity that comes with that

procedure. Moreover, it is also important to

realize that the frozen section analysis can also

have a false negative rate of 30 to 40 percent, so

all these factors show us the limitations of how

even when surgery is performed and nodes are

sampled, there are some limitations.

Here is an example of a patient who had

underwent radical prostatectomy, and you can see

clips where the surgeon has taken out the lymph

nodes, and as I said earlier, this is what standard

lymphadenectomy involves, is the low external iliac

group of lymph nodes.

There was a small nod posteriorly in the

pelvis that was not sampled, and the patient was

labeled as cured. Eight months later, the patient

shows back with that node mushrooming into a

full-blown metastases, and this is a good example

of how surgical sampling can sometimes be limited

by what the surgeon can see and samples.

Thus, there is a current need for a

non-invasive technique that not only detects, but

also characterizes lymph nodes with a high level of

accuracy, not compromising sensitivity for

specificity.

It also provides a broad anatomy coverage

which means you not only look at lymph nodes right

next to the primary cancer, but also can look at

lymph nodes in a broad anatomic area beyond the

confines of the regional distribution.

That is where I think Combidex, or the

pharmacologic name ferumoxtran-10, is an excellent

contrast tool that can be utilized with MR. This

is an iron oxide based nanoparticle with a central

iron oxide coat and a surrounding dextran coating.

This slide shows how the contrast acts.

After intravenous injection, the contrast lingers

in the blood vessels for a long time, has a long

blood half-life. It gradually leaks out and then

is transported to the lymph nodes where it binds to

the scavenger on macrophages. Thus, the mechanism

of action of uptake in the normal nodes is via

macrophages. So, if the node is functioning

normally and has its normal complement of

macrophages, the contrast would then localize to

the nodes and turn the normal area of the node

dark.

I would like to emphasize at this point,

two points in the mechanism of action. One is the

contrast is targeting the normal lymph node and

black is benign, so it is the normal part of the

node that is turning dark.

If you have an area of tumor deposited in

the node, then, that area of the node is devoid of

normal functioning macrophages and that area would

show lack of uptake and continue to stay bright.

Another important point to remember is

that this mechanism of action is independent of

which primary cancer affects the node, and, hence,

the lack of uptake would be present no matter which

tumor deposit is present within the lymph node.

This slide is just to show the technique

that we use. Any conventional 1.5 MR system that

exists today in the community, independent of

vendor platform, can be used for imaging the MR

with Combidex, and these are the sequences, again

nothing fancy, just regular bread and butter

sequences.

We can do post-processing, which can

provide for elegant ways of communicating the

information, but these are not essential for making

the diagnosis.

So, let me show you an example of how the

Combidex acts in real life. This is a patient who

has a known pelvic malignancy. There are two lymph

nodes in the groin. Both are hyper-intense or

bright on the pre-contrast.

Twenty-four hours after injection of

Combidex, you can see the medial node is turning

homogeneously dark, and that is the node that is

benign. The node to the right shows lack of

uptake, and that means that it's infiltrated with

cancer and, hence, it is not taking up the

Combidex.

Let me show you some examples of how

Combidex scanning improves sensitivity in detecting

metastases in small lymph nodes.

This is a patient with prostate cancer

undergoing staging. The yellow arrows point to two

very small nodes next to the external iliac vein.

Again, by size criterion, you would never call

these nodes positive.

On the pre-Combidex scan, you can see

these two nodes are hyper-intense, and 24 hours

later after Combidex, the inferior one is turning

homogeneously dark. It means that that is benign.

The one which is pointed by the red arrow shows

lack of uptake, and that is the one which is

malignant, which was proven at the time of surgery.

This is a patient with breast cancer.

Again, the patient is lying prone. Here is the

lung, the breast of the patient, and we are looking

at the axilla. Again, there are two very small

nodes in the axilla pointed by the yellow and the

red arrow, measuring between 3 to 4 mm.

After giving Combidex, the superior one is

turning dark as outlined by the yellow arrow, the

inferior one, which is the red arrow, shows lack of

update, indicating it's malignant and again proven

with surgery.

So, I have shown you how Combidex improves

sensitivity in different types of primary cancers.

It is equally important to have enhanced

specificity, which means if the node is enlarged,

you need to accurately diagnose it as benign or

malignant.

So, here is a patient with bladder cancer.

You have an enlarged node measuring 20 mm, and this

was labeled as malignant on the contrast-enhanced

CT. On the pre-contrast MR, it is hyper-intense.

Post-Combidex, it turns homogeneously dark

indicating it's benign and was proven so on biopsy.

Another example of enhanced specificity,

again a patient with prostate cancer. The two

yellow arrows point to enlarged obturator nodes,

again labeled malignant based on the size

criterion, but post-Combidex, you can see it is

turning homogeneously dark, and these turned out to

be reactive enlarged nodes or reactive benign nodes

in the pelvis.

As you can see, by improving the

sensitivity and specificity in these patients, one

can provide for improved clinical staging, and then

also provide for better surgical planning and

better radiation therapy and image-guided

intervention planning. Some of these points will

be highlighted later by my colleague, Dr. Jelle

Barentsz.

Thank you

Efficacy Data from Phase III Clinical Studies

DR. GOECKELER: Good morning. I am going

to review in the next few minutes the efficacy data

in support of the proposed indication. The studies

I will be discussing were designed to evaluate the

ability of Combidex to improve the differentiation

of metastatic from non-metastatic lymph nodes,

particularly in the post-contrast setting.

To do this, we compare the parameters of

sensitivity and specificity in both the pre- and

post-contrast image sets. The study's design,

which was conducted in cooperation with the FDA,

provided for multiple primary tumor types and

independent blinded evaluations of image sets with

histopathologic confirmation of the imaging data.

I think it is worth taking just a step

back to say that all the imaging data that you will

be presented this morning by the sponsor involves

histopathologic confirmation at the individual node

level, which is a significant undertaking.

So, in reviewing the efficacy data, I will

first go over quickly the blind read procedures

that were used in conducting the analysis of this

data, review the data from EU and U.S. Phase III

studies, talk a little bit about data from

publication in the New England Journal of Medicine

that investigated the agent in this application,

and finally, close by looking at how this

improvement in differentiation at the nodal level

impacts clinical nodal staging.

So, first, the blinded read procedure, and

there are a number of blinded reads that were

carried out in each of the clinical studies, so I

will try to explain the terminology and the

sequence in which they were conducted.

All the blinded reads were carried out

with the readers blinded to clinical, demographic,

and pathologic information, and the cases were

presented in random order.

The readers were first presented with the

pre-contrast images, and based on the pre-contrast

images alone, made an assessment on size based.

You will also see that in some of the

slides called an MRI-based diagnosis, and then the

reader made a second assessment based solely on the

pre-contrast image, which was based on the reader's

skill. In that subjective evaluation, the reader

was allowed to use any criteria they thought was

appropriate in differentiating metastatic from

non-metastatic lymph nodes.

Following those readings, the readers were

presented with the post-contrast images and carried

out an evaluation of the post-contrast side by side

with the pre-contrast images. This is a so-called

paired evaluation. The prospective primary

endpoint in each of the Phase III studies was a

comparison of the paired evaluation with the

pre-contrast size-based evaluation at the nodal

level.

Next, a period of about two weeks to

eliminate a recall bias was allowed, and then the

readers were presented, again in random order, with

the post-contrast only images, and then made an

assessment based only on the post-contrast image,

which is called the post-contrast evaluation.

Post-contrast images, there were reading

guidelines developed to assist the reader in

evaluating the nodal post-contrast images. They

were prospectively developed and finalized before

the blinded read. Thus, the Phase III blind read

of images is a valid assessment of nodal images

across a wide range of cancers.

This is the study population in the three

studies that I will be talking about - the U.S.

Phase III, the EU Phase III, and the New England

Journal. The number of patients dosed and the

number of patients with histopathology is not

always the same since eventually, not all patients

go to surgery for things that happen in the

intervening time between the imaging session and

the treatment of the patients.

This outlines the number of lymph nodes

that were evaluated in the various studies both

pre- and post-contrast and a breakdown of where

those lymph nodes resided by anatomic region in the

various cancers.

So, right into the Phase III study, in the

EU Phase III study, what we see is that in the

pre-contrast evaluations, both the size and the

subjective base, we see a high pre-contrast

sensitivity and a low pre-contrast specificity,

whereas, in the post-contrast evaluation, the

paired evaluation, what we see is sensitivity

remains high at 96 percent, but specificity is

significantly improved, and the improvement in

specificity was statistically significant over both

of the pre-contrast reads and for both of the

blinded readers.

We look at the data from the U.S. Phase

III study. It's a little bit different situation.

In the pre-contrast size-based analysis, in the

pre-contrast analysis, sensitivity was low and

specificity was high, so sort of just the opposite

of what was seen in the EU study.

In the subjective evaluation, we see that

the subjective reader's assessment resulted in a

very high sensitivity, but the tradeoff for that

increase in sensitivity was a large decrease in

specificity.

So, the pre-contrast reads had either high

sensitivity or high specificity, but not both. In

the post-contrast reads, you will see that

sensitivity was high and specificity was high, so

we had a combination of high sensitivity and high

specificity.

You will also note that in the post-only

read, in which the only image that was available

was the post-contrast image, resulted in the

highest level of imaging performance and the

greatest level of consistency.

If we take a look for just a minute at

this discrepancy between the two pre-contrast

reads, where one had high sensitivity and low

specificity, and the other was the opposite, if we

look at the false diagnoses that occurred in these

various blinded readings, and we look at false

diagnoses as a percentage of the total, we see that

the percentage of false diagnoses for both of the

pre-contrast reads is relatively the same.

What we see is that in the subjective

readers' diagnosis with the readers subjectively

overreading to try to account for the known low

sensitivity of the size-based analysis, we see a

very large percentage of false positive reads that

occur in the subjective readings, whereas, in the

post-contrast reads, we see a decreased percentage

of false reads with the lowest and most consistent

data again in the post-only read.

This is the data broken out by body

region, and you can see that in the head and neck

and breast, we saw large increases in sensitivity

when we compare the pre- to the post-contrast read,

maintaining specificity which overall resulted in

the increase in accuracy.

In the pelvis and abdomen, we had more

moderate levels of increase in both sensitivity and

specificity, the net effect of which is that the

increase in accuracy in the pelvis and abdomen is

virtually identical to what one sees in both the

head and neck and the breast.

One region that was a little bit different

was in the lung. In the lung, we see more

moderate, small increases in both sensitivity and

specificity, and we believe this has to do with

limitations of anatomic imaging in this particular

body region, and not differential uptake or

performance of the contrast agent.

So, turning now to the data published in

the New England Journal of Medicine, and I think

this data is important supplemental data that can

help us understand better some of the differences

that were seen particularly in the pre-contrast

reads in the Phase III studies and also can help us

learn a little bit more about the performance of

the agent in different size nodes.

So, this is a study carried out in

prostate cancer patients at two centers, one in the

U.S., one in the EU, 40 patients from each site.

There was a centralized independent blinded read

with histopathologic confirmation of data.

So, to address some of the issues that I

just mentioned, I am going to go through the data

in a little bit of a sequential order.

First, with regard to the issue of the

discrepancies in the pre-contrast evaluations and

also to look at the issue of the effect of nodal

size on the performance of the contrast agent, what

you see is as you move across these three studies,

the distribution of nodes categorized as either

greater than or less than 10 mm, and that is an

appropriate cut point because as Dr. Harisinghani

said earlier, that is the point at which we

differentiate a malignant from a non-malignant

node.

We see that as we move from the EU to the

U.S. to the New England Journal study, the

proportion of large nodes are greater than 10 mm in

the yellow, goes from about three-quarters to about

a third to only 7 percent in the New England

Journal study.

We see in the pre-contrast size-based

sensitivities and specificities, we see that the

sensitivities and specificities largely track with

the nodal size. That is, in studies where there

was a high proportion of large nodes, we see a high

sensitivity in the pre-contrast evaluation in the

green bars, which decreases as the proportion of

large nodes in the study decreases.

Conversely, as in the purple bars, we see

that as the percentage of small nodes increases,

then, the specificity increases also.

So, finally, in the post-contrast data,

what we see is that we see a lack of dependence of

the performance of the agent on the size of

distribution of the nodes in the study. We have

high sensitivity and specificity regardless of the

distribution of the lymph node sizes that were in

those studies.

Finally, just a word about clinical nodal

staging in the U.S. Phase III study, we looked at

clinical nodal staging where we could collapse the

nodal stage in its simplest form to where patients

were either node positive, node negative, or

indeterminate.

What we see here is a comparison of the

clinical nodal stage that was assigned based on the

images compared to the eventual pathologic stage,

and we can see as we go from the pre- to the

post-paired to the post, the percent where the

agreement was correct increases, the percent where

it's incorrect decreases, and the percentage that

could not be staged also decreases.

So, to sum up, there are two prospective

Phase III studies. The pre-contrast evaluations in

these studies show a characteristic tradeoff of

sensitivity for specificity. Post-contrast

evaluations show high sensitivity and high

specificity, which results in an overall

improvement in accuracy.

The improved lymph node differentiation

improved clinical staging. The supporting data

from the New England Journal publication showed

high sensitivity and specificity in a population of

largely small lymph nodes.

Finally, these data collectively

demonstrate the efficacy of Combidex in

differentiating metastatic from non-metastatic

lymph nodes.

Thank you. Now, Dr. Faich will review the

safety data.

Safety Data from Clinical Trial

DR. FAICH: I am Jerry Faich. Good

morning, members of the panel, Chairman, and FDA.

What I would like to do rather briefly is

review the amount of exposure data that has been

obtained for Combidex, discuss and show you the

pattern of adverse events that have occurred, make

a few comparisons with other agents, and then

discuss the proposed risk management plan for the

product.

In total, 2,061 subjects have been dosed

with Combidex. Of these, and I would like to

emphasize this and explain it, 131 received bolus

injection. This was in the process of developing

or exploring the utility of the product for liver

scanning, which required a bolus injection. That

indication and mode of administration has been

dropped.

The remaining patients, the remaining

1,930 patients were dosed with dilution and

infusion either in 50 ml or 100 ml saline, and

within those, there were 1,566 cases at all doses

who got the 100 ml dilution.

For the proposed indication and mode of

distribution, there were 1,236 patients in the NDA

receiving 2.6 mg of iron/per kg at the 100 ml

dilution over 30 minutes.

This shows you on the left-hand side the

rate of adverse events in the bolus injection 30

percent, in the middle 17 percent for 50 ml

dilution, and 14 percent on the right-hand side for

100 ml dilution showing a clear dose-response

relationship in terms of adverse events, and this

is indeed why the 100 ml dilution has been focused

on.

It needs to be said that during the bolus

injection studies, there was one anaphylactic death

that occurred immediately. That and the need to

use bolus injection for liver scanning is what led

to dropping the pursuit of that indication.

This shows you in the 1,236 patients the

pattern and rates of adverse events, you can see

going from vasodilation at 3.4 percent, rash, back

pain, pruritus, urticaria, et cetera, overall

totaling these 15.8 percent.

I would simply like to emphasize that

nearly all of these were mild, transient, and

self-limited.

Within the 1,236 core patients, 5.6

percent had adverse events from that prior list

that could be called hypersensitivity events.

Mainly these were vasodilation. It included 24

patients, however, who had more than one symptom

from that list.

Only 4 of the 1,236 patients, or 3 per

1,000, had a serious adverse event. The serious

adverse event rate is no greater than that found in

labeling for nonionic iodinated contrast media,

which ranges from 0.6 to 1.5 percent, and I will

show you that in a moment.

There were no life-threatening

anaphylactic/anaphylactoid reactions at the

proposed dose and method of administration.

In terms of immediate adverse events,

immediate hypersensitivity adverse events can, of

course, be controlled in large part by stopping the

infusion. The most common reaction, as I noted,

was flushing.

Thirty-six patients had infusion stopped

and restarted, that is, these patients were

rechallenged. Only two of them could not tolerate

the rechallenge and were discontinued. The

remaining 36 went on to complete their procedure.

Put a slightly different way, 94 percent

of all immediate hypersensitivity reactions

occurred within the first 5 minutes after dosing.

Most hypersensitivity reactions, as I indicated,

were mild to moderate in intensity.

At the proposed dose and method of

administration, out of the 4 serious AEs, 2 were

classified as immediate hypersensitivity reactions

using the FDA definition. That translates to a

rate of 1.6 per 1,000.

In terms of anaphylactoid reactions, again

using an FDA definition of affecting two body

systems, there were 12 such patients at the

proposed dose and method of administration. Two of

those were considered serious.

Four of the 12 were in the group that had

infusion stopped and then were rechallenged without

subsequent problems. The majority of these 12 had

dyspnea and flushing. There were no serious

hypotension or respiratory compromise seen in those

12 patients.

I don't mean to make much of this, but I

do show it, and it is always hazardous, and one has

to interpret data carefully when you compare one

set of data from one set of studies and labels to

another, but what I would like to do here is call

your attention to the Combidex data across the top.

The overall AE rate was 15.8 percent, the

serious AE rate was 3 per 1,000. That is those 4

cases I mentioned. If you look down in the

right-hand column just at serious AEs and compare

it to other iodinated contrast agents, both from

data in their labels and published studies, you

will see for Ultravist, that serious AE rate is 1.1

percent.

For comparators in studies done with

Ultravist, it was 0.6 percent, for Oxilan it was

1.5 percent, and for comparators to Oxilan and

studies done with it were 1.1 percent. So, this is

a basis or my basis for concluding there is not

evidence that there is increased risk of serious

adverse events comparing this drug to commonly used

iodinated contrast agents.

There is not much in the literature about

anaphylaxis in contrast agents. Here are 2 recent

studies that have been published. This is Neugut

in the Archives of Internal Medicine. His

published anaphylaxis rate done from his own

studies and across the literature was 2 per 1,000

to 10 per 1,000 or 0.22 to 1 percent. He noted

that it might be lower and most people are taking a

rate of about half that for low osmolality contrast

agents.

David Kaufman, at the Center for

Epidemiology in Boston, published this paper in

2003, and for contrast agents, this was an

international study of anaphylaxis, the observed

rate was 7 per 10,000. For nonionics, again, as I

said, 50 percent of that, about 3.5 percent, and

there was a range as you see here.

Combidex falls within or at the lower end

within that range of values.

In terms of a risk management plan for

this product, it is largely in keeping with

existing guidelines and calls for physician

education, emphasizing the need for dilution and

slow infusion obviously as a means to be able to

intervene if a reaction is occurring. The labeling

will be consistent with that, and the proposal is

to conduct targeted surveillance to gather further

data to reinforce the safety data that I have shown

you.

To summarize, then, there has been

considerable clinical exposure in the development

program. Hypersensitivity is relatively infrequent

and comparable to that of other contrast agents,

and the risk management program that I just

described is in accordance with existing

guidelines. Thank you.

Dr. Barentsz, please.

Clinical Utility of Combidex in Various Centers

DR. BARENTSZ: Madam Chairman, members of

the Committee, members of the FDA, I am an

oncologic radiologist and I have been using

Combidex MRI in more than 500 patients, and I am in

frequent contact with investigators in both the

U.S. and in Europe.

From the previous data, you have clearly

shown that this contrast agent works. A black

lymph node is normal, and a white lymph node is

abnormal. That is despite the tumors type.

Nonetheless, evaluating its clinical

utility is a lot more difficult, and for that you

need personal experience, as well as post-Phase III

studies. Based on these two, I am going to try to

show you the clinical utility and some cancer

types.

The reviewed publications were all in top

ranking journals. It was blinded post-contrast

image evaluation with gold standard histopathology,

and all those papers described a potential impact

on treatment planning.

The areas being defined where Combidex MRI

provides a significant clinical benefit were

prostate, bladder, head and neck, and breast, and I

want to address those issues with you in the next

10 minutes.

As you can see, data were collected from

almost 200 patients and almost 2,000 lymph nodes.

These are the data on sensitivity and specificity

and accuracy.

You can see that the data are highly

consistent, showing a high sensitivity,

specificity, and accuracy for all the cancers.

Now, let's start with the clinical utility

in prostate cancer. First of all, you have to

define the current strategies. Current imaging has

an insufficient sensitivity for lymph node staging,

and therefore, urologists are performing an

invasive operative surgical lymph node sampling to

detect the lymph nodes.

These techniques have limitations, only a

limited area sampled, and therefore, up to 31

percent of the positive lymph nodes are outside of

the surgical area, which have been shown by some

data recently published in the urology journals.

Furthermore, surgical sampling has a

complication rate reported to be 22 percent for the

open dissection and 5 percent for laparoscopic

dissection, including lymphocele, lymphedema, deep

venous thrombosis, pulmonary embolism, nerve

damage, and blood loss.

Because of the limitations of current

imaging technique and current staging techniques

for the lymph node dissection, these urologists are

advocating at this moment now an extended lymph

node dissection. They state that they will detect

those lymph nodes, however, this significantly

increases morbidity. The question is are the less

invasive way techniques to solve this problem.

As you can see, using the post-contrast

studies of Combidex, there is a dramatic decrease

of the number of false positives, as well as the

number of false negatives, but what is even more

important is that in our study in the New England

Journal of Medicine, in 6 percent of all the

patients, we found a small non-enlarged lymph node

which we could biopsy, and in all those patients,

we could confirm the diagnosis by image-guided

biopsy, and these patients did not undergo any

surgical dissection.

Furthermore, in 11 percent, we found lymph

nodes which were outside of the surgical field, so

they will be missed with regular surgery.

All these findings were confirmed by the

surgery because before the operation, we told the

urologists where the lymph node was, and they could

then find them.

I would like to show you two

representative cases. Here, you see a white lymph

node, metastatic, of only 7 mm in size. It is very

close to the internal iliac artery, which is

outside of the normal surgical field. In this

lymph node, we performed an image-guided biopsy

which was positive, and in this way a correct

diagnosis was being evaluated in a less invasive

manner, and this avoided inappropriate treatment.

This patient had, instead of a prostatectomy, an

androgen ablation.

In another patient, you see a lymph node

over there with a tiny white structure. You can

see it over there. This was also a lymph node

outside of the surgical field. We told our

urologist where this lymph node was located. It

was found and it was confirmed histopathologically

that this lymph node had a 1-mm metastasis.

What about bladder cancer? It is actually

the same story. In 24 percent of positive lymph

nodes, there are positive lymph nodes in 24 percent

despite negative pre-operative imaging techniques.

The presence of lymph nodes radically

changes the treatment option especially if there is

N2 and 3 node, or if there are more than 4 nodes,

so finding these lymph nodes also here is very

important.

If you perform an extended lymph node

dissection, you detect more lymph node, it will

increase survival for minimal disease, however,

also in this extended lymph node dissections, not

all lymph nodes have been sampled. Furthermore,

this increases morbidity.

These are the data in 172 lymph nodes in

58 patients from a Radiology paper, and it has been

shown that in normal-sized lymph nodes, 10 out of

12 were detected using Combidex MRI, and this

information was crucial for the surgeon to find

these lymph nodes, and they were removed.

Most important areas, also head and neck.

The survival rates depends on whether the tumor has

metastasis in lymph nodes or not. Therefore, the

status of cervical lymph nodes is vital for the

choice of therapy.

Twenty-five percent of positive lymph

nodes are found despite negative preoperative

imaging techniques like contrast CT or

ultrasound-guided biopsy. Why? Because these

lymph nodes are below normal size criteria. They

are only 5 to 10 mm in size.

Because of the fact that these lymph nodes

do not show up with imaging, head and neck surgeons

perform commonly a radical neck dissection, which

causes a very severe cosmetic deformity and has a

very high complication rate, in literature reported

up to 54 percent.

The data from Mack, et al. in Radiology

show a very high sensitivity and negative

predictive value, and furthermore, what is more

important, if you look on a patient level, they

were able to make an accurate diagnosis in 26 out

of 27 patients, and what is the most important

thing is that this information would have resulted

in reduced extent of surgery in 26 percent of these

patients, so avoiding an aggressive neck

dissection.

One representative image. This was a

patient with, on the CT scan, an enlarged 12 mm

lymph nodes, however, on the post-Combidex MRI, you

see the lymph nodes are black. This was the 12 mm

one, this was the 10 mm one, and they were normal.

In this patient, a neck dissection could have been

avoided.

Finally, breast cancer. The commonly used

staging procedure at this moment is the sentinel

lymph node staging, which has false negative

numbers of 3 to 10 percent, and is an invasive

technique, but what is even more important is that

recent data have shown that the sentinel lymph node

is the only positive lymph node in 61 percent in

patients with positive lymph nodes.

Nonetheless, these patients all undergo an

axillary lymph node dissection, and this has a high

rate of clinically significant complications.

A technique with a high negative

predictive value performed in an adjunct to the

sentinel lymph node procedure in patients with one

positive sentinel lymph node may reduce the number

of axillary lymph node dissections.

These are the published data in almost 300

patients by Michel in Switzerland, and you can see

that this technique has a high negative predictive

value.

I would like to show you one

representative case from our institution. This is

a very, very tiny primary tumor, and this was the

positive sample on lymph nodes. This lymph node is

white on Combidex, so that means metastatic, and

you can see that the second and third station lymph

nodes, that they are black, so in this patient, all

the other lymph nodes were black, which in this

case was confirmed by histopathology.

Now, to the final conclusion. I have

tried to show you some areas of clinical utility of

this contrast agent, and as soon as we get more

experience, there will be a lot more areas.

To summarize, the current techniques to

detect positive lymph nodes in prostate, bladder,

head and neck, and breast cancer have significant

limitations.

Combidex MRI shows high sensitivity and

specificity not only on the nodal basis, but also

on the patient-to-patient basis, which for a

clinician is even more important.

Therefore, Combidex MRI may reduce the

extent of surgery and morbidity, and finally,

Combidex MRI identifies additional positive lymph

nodes for biopsy or image-guided extended lymph

node dissection in this way improving the staging

of the surgeon.

Thank you.

MR. ROESSEL: Thank you. That concludes

our presentation.

Our clinical data and the clinicians I

think have shown you that Combidex is an important

diagnostic imaging tool that improves the current

practice.

Thank you. We are available for any

questions you have.

DR. MARTINO: Thank you.

At this time, I am going to ask Dr. Li to

present his view of this data, and once that is

done, we then will take questions for both the

sponsor and the FDA.

FDA Presentation

Efficacy and Safety of Combidex (NDA 21-115)

DR. LI: Dr. Martino, members of panel,

ladies and gentlemen, good morning. My name is

Zili Li. I am a medical team leader with the

Division of Medical Imaging and Radiopharmaceutical

Drug Products at FDA. I am a board-certified

physician in preventive medicine with special

training in epidemiology.

Today, I would like to share with you our

review of findings of NDA Application 21-115

Combidex.

I would like to start off by noting that

this presentation represents a collaborative effort

by a group of highly dedicated reviewers at FDA

whose names are on this list.

Combidex is an MR contrast agent. The

proposed clinical dose is 2.6 milligram iron per

kilo of body weight.

Of three methods of administration which

has been used in the clinical development program,

the sponsor select the dilution in 100 cc with the

slow infusion over 30 minutes of a standard measure

of administration.

The other two methods, particularly the

direct injection, is no longer being proposed.

This slide summarized the indication that

had been proposed by the sponsor--I will go over

one more time--that Combidex can assist in the

differentiation of metastatic and non-metastatic

lymph nodes in patients with confirmed primary

cancer who are at risk for lymph node metastases.

I would like to draw your attention to the

fact that this is a broad indication. If granted,

this agent can be used for almost all cancers

regardless of type, size, clinical stage, whether

patient has been previously treated with drug,

biologic, radiation, or surgery.

One objective of today's presentation is

to show you why the Agency has concerns for such a

wide or broad indication given the level of

efficacy and safety observed from clinical trials.

To support this indication, the sponsor

submit one U.S. and three European Phase III

studies. In addition, sponsor also ask Agency to

consider data from a published article in the New

England Journal of Medicine.

For the safety, the sponsor submitted a

safety data adverse event profile in particular

from approximately 2,000 individuals who received

Combidex from multiple clinical studies.

I would like to make a remark on this New

England Journal of Medicine article. This study is

pooled analysis from two ongoing clinical studies.

One is U.S. IND study, is under sponsor's IND. The

other study is non-IND study and in Europe.

The clinical investigators themselves took

initiative to combine 40 cancer patients from each

original study to form the basis for this New

England Journal of Medicine study. At this time,

however, it is unclear to us how those 80 patients

were selected, and more important, after repeat

requests, the sponsor is not able to provide us the

original source document which included pre-defined

statistical plan, blind reader evaluation manual,

and original copy of blind readers' evaluation of

the medical imaging.

For that reason, the Agency cannot

conclude this study was conducted in compliance

with the Federal regulations pertaining new drug

application. For that reason, we are not able to

consider this study as adequate and well-controlled

study.

However, the Agency do agree that the

cases present in this article may demonstrate some

potential the benefit of the use of Combidex in a

clinical setting.

I also would like to draw your attention,

say a few words about this U.S. IND study. We just

got update from sponsor yesterday. This study is

closed at this time. Roughly, they have 220

patients enrolled including 91 prostate cancer and

34 bladder cancer patients.

Although the original protocol require all

the pathology confirmation and MR imaging for all

the patients, at this time it is not clear to us

how many patients for this study will have both

information available for a meaningful analysis for

efficacy if such analysis is needed.

Now, I would like to first highlight the

differences between sponsor and the Agency's final

conclusion regarding efficacy and for safety.

As far as for the efficacy, the sponsor

believes the non-contrast MR agent only offer high

sensitivity or high specificity, but not both. The

advantage of this Combidex is its ability to offer

both high sensitivity and specificity consistently

regardless type of cancer or size of the lymph

node.

At this time, the Agency is not able to

draw such a conclusion because of the

generalizability and validity issues we are going

to show you in the later presentation, and also in

the later presentation, we are going to show some

preliminary evidence which may suggest the

performance of Combidex may vary by size or type of

cancer.

For the safety, sponsor acknowledge that

Combidex is associated with hypersensitivity

reaction, however, their emphasis is that no death

or life-threatening AEs are associated with the

proposed clinical method of administration. That

is the dilution with the slow infusion.

Also, I just noticed in the sponsor's

presentation is new to us that they make a claim

that this agent's safety profile is equivalent to

the iodinated contrast agent. I believe in your

briefing document, they also made a claim that

serious adverse event with the Combidex is only

one-third of that iodinated contrast agent.

Our position is that dilution and slow

infusion are not entirely free, and also we

disagree that the Combidex, the safety profile

resemble that of iodinated contrast agent.

This slide highlights the issues we are

going to bring to the panel today. For the

efficacy, we are going to talk about sample size.

We are going to talk about representation of

different tumor types in the clinical study.

We are also going to talk about impact of

study inclusion/exclusion criteria. Later, the

last one, we are going to talk about develop use of

Combidex imaging guidance, which was the major

issue in our briefing documentation to you.

For safety, we are going to talk about the

hypersensitivity reaction. We are also going to

make a comparison with iodinated contrast agent.

Then, we are going to follow up with the

discussion of risk-benefit ratio, including the

sponsor's proposed risk management plan and our

emphasis on the need to understand, to define the

conditions of use for this product.

From the sponsor's presentation, it was

stated that total 152 U.S. patients and 181

patients from a European study received Combidex

injection, however, what was not apparent on their

slide was the number of patients who were actually

included in the primary analysis. What we are

showing you is, because there are two different

blind readers, so they may see the different people

different, so the number may vary slightly.

For the U.S. study, there is only 64

percent of original total population were actually

involved in the final analysis. For the European

studies, the number varies from zero, 16 percent,

roughly 20 percent to 41 percent. It only

represent a small proportion of the patients who

originally received the Combidex.

I need to make a clarification for the

study with zero participation. This is a breast

cancer study. You probably read our briefing

document. The original statistical plan for the

European study is on the patient basis. It is

totally different from what they did here. So, for

that reason, the individual nodal level analysis

was never performed, so those people cannot include

in their primary analysis and consistent with U.S.

statistical plan.

The small number of patients or small

proportion of patients included in the primary

analysis create two dilemmas for us. The first, we

need to understand whether the estimate we got from

this population is applicable to entire population.

The second one is because of the small

number of patients, we want to ensure that the

patients included in the analysis more represent

the cancer patient distribution in the United

States.

This is the second issue we would like to

bring to your attention.

Based on the statistic provided by

American Cancer Society, it is estimated this year,

2005, there is going to be 1.4 million new cancer

diagnosed. The left two column showed you the rank

of the top 10 cancers and also showed their

percentage distribution in the United States. I

need to mention that lymphoma or leukemia are not

included in this table.

On the right two columns show the number

of patients and their distribution for each type of

cancer included in the primary analysis. I would

like to bring your attention to the fact they have

two readers. In this slide, we pick the highest

number in this table.

You probably noticed that the majority of

patients come from head and neck, which is ranked

roughly number 6 in the frequency distribution, and

also you probably noticed that prostate cancer

being the number one in the United States. There

is only 5 patients from the United States and 5

patients from Europe was included in the primary

analysis, and the highest number each category is

only in here is 37.

Also, I need to remind you that for

European study, the sponsor showed you the majority

nodes are larger than 10 mm. Actually, in reality,

all 37 patients have a node larger than 10 mm, so

there is no nodes like the 10 mm for the European

study for this population, particularly this head

and neck what I referred to.

So, you probably will ask why that so many

patients are not included in the primary analysis.

I would like to bring your attention to the fact

the primary analysis was conduct at the nodal

level, so the target lymph nodes, which should be

included in the analysis, is represented here, the

large circle here, is all the lymph nodes

visualized by site investigators.

When patient enrolled, when they take MR,

site investigator looked at the MR to circle the

node they see on those MR images. That should form

the basis for primary analysis. However, not all

the nodes was able to match with pathology, so you

drop some nodes right over there.

Then, when you present the same images,

the unmarked images to blinded reader, the blinded

reader may not pick up the same nodes the original

investigator picked in the first place, so you drop

some nodes over there.

Then, for the comparison purpose, because

they want to compare the post-images with the

pre-images, you can only do analysis on the nodes

identified on both end, so for that reason, you

have a few nodes drop again, so by the end, the

nodes included in the analysis is much smaller than

the nodes originally seen by site investigator

initially.

This table actually show you the

deposition of how the nodes got lost with each

process. In the U.S. study, this is the number of

patients. The first row showed you number of nodes

originally visualized by the site investigator,

which should form the basis for primary analysis -

371, 834, 333, and 234.

This row showed you what percentage of

those nodes have matched pathology, and this row,

the final one, showed you what number, how many

nodes were actually included in the primary

analysis. You can see it is roughly from 3

percent, 6 percent, to 45 percent of nodes was

originally seen is included in the primary

analysis.

The fundamental assumption for this

clinical development program is that the

performance of Combidex should be independent from

the type of cancer and the size of lymph nodes.

That was why originally that was allowed for

different cancer patients included in the one

study.

However, if you look at this performance

of Combidex, by different type of cancer, you will

see, first, this is the sensitivity slide. You

will see in the U.S. trial, the variation from 76

to 100 depending on the site of primary cancer, and

the 95 percent of the lower boundary could go as

low as 55 percent.

Only if you are willing to accept

assumption that Combidex performance is independent

of sites, you get 83 percent performance with the

lower boundary 73. That is exactly the reason why

the Agency was so worried about small lymph nodes,

small size, because from this table we really don't

know whether it's a variation because of the random

event, or if it truly reflects the different

performance of Combidex among the different type of

cancers.

This is the same table for the

specificity, which again challenge assumption

whether the Combidex, the performance should be

considered or accepted independent from the type of

cancers.

You notice depending on the different

sites, the specificity vary from 44 to 91, and with

the lower bound, can go as low as 21 percent. The

significance of the two slides is that with dose

variation we will have a very hard time to

understand what is appropriate performance

characteristic of this Combidex-enhanced MR

contrast agent, and if indeed the performance are

different, if this drug is approved for all the

cancers, this information may be misused by the

clinician to make their clinical judgment.

The next issue is about study

inclusion/exclusion criteria. I will go very fast.

Basically, for this study, the people who received

treatment, chemotherapy or radiation therapy in the

past 6 months was excluded.

Actually, in reality, when you look at the

people included in the primary analysis, I don't

think any of them had any prior treatment, so

mainly this database, we believe, if valid, only

applied to people who are newly diagnosed patients.

This is issue about development of a

clinical MR imaging guidance. Why is this imaging

guidance so important? It is because for the

radiation to use this contrast agent, you need to

have a standard way to interpret imaging. So, we

work with sponsor to ask them to come with the

guidance.

So, this actually, the clinical trial is

actually to validate the guidance for this validity

and usefulness, however, originally, from the NDA

submission, it appeared to suggest this guidance

was developed and validated from the same database.

That is the U.S. database. That was a big concern

for us because basically, if that is true, that

destroyed independence of this guidance themself.

Later on when we spoke to sponsor, they

provided us a revised statement. Basically, the

guidance was developed by use of Phase II images,

it is not Phase III.

Sponsor's consultant, when she developed

this guidance, she did look at the 16 cases from

Phase III trials, however, no pathology was

provided, and also, there was a statement that

there is no more changes for the guidance after

review of Phase III data.

To support their statement, sponsor did

submit original soft document to FDA for our

verification. We also had extensive discussion

with their consultant to recall what happening on

that day for the development of a Combidex imaging

guidance.

All we conclude at this time is that,

first, we do not have definitive evidence to

absolutely exclude the probability that Phase III

data has no impact in this guidance development,

however, the evidence provided by the sponsor is

consistent with this revised statement, therefore,

at this time, we decided not to pursue this issue

any further unless there is new evidence emerge.

The second issue we are having, which I

will present was included in our briefing document,

is in the European study, this guidance, the core

instrument actually was not used by the blinded

reader. The blinded reader was using a different

guidance to make their diagnosis.

At this time, the sponsor is not able to

provide any documentation for us to understand

which method or who actually do the translation

from this guidance and to this one. Actually, the

question we are having for the committee,

especially for people expert in MR imaging, is

whether the similarity or correlation between these

two guidance is so great, the Agency should not

worry about who did it and with all this

documentation.

Now, I would like to switch to the safety

side of Combidex evaluation. I will focus my

presentation in Combidex-induced hypersensitivity

reaction.

There is one case hypersensitivity-related

death in a clinical development program. This is a

70-year-old male with history of allergy to

contrast, who received undiluted direct injection

and developed hypersensitivity reaction immediately

after injection and become unresponsive.

At the clinical site, however, there were

no appropriate personnel or emergency response

available, so they have to call 911. When the EMT

arrived, they delivered CPR and epinephrine. When

the patient get to the hospital, patient was

pronounced dead approximately 35 minutes after this

injection. An autopsy revealed no MI or PE, and

they conclude this is a Combidex-related

anaphylactic shock.

I would like to make two points here.

This injection is no longer being used. The second

one, we are really concerned about the lack of

appropriate personnel for emergency situations

especially if this drug is found to be valid, safe,

effective, there is many free-standing clinical

imaging centers around the country, so we need to

have a way to ensure this drug to be used

appropriately. That is with assumption that if

this study is valid and the drug is safe.

This table shows the distribution of the

safety database or number of patients by

administration and by the dose. There are a total

of 2,061 patients exposed to Combidex, 1,236

patients received proposed clinical dose, 131

patients received bolus injection. Those three

groups will form the comparison for our next few

slides.

This slide shows the rate and severe

hypersensitivity reactions by the three different

subgroups I just mentioned to you. For the

clinical proposed dose, the rate of

hypersensitivity reaction is 5.3. For direct

injection, it is 6.1.

I would like to let you know that in your

briefing document, this number is slightly higher

because we just discovered some computer error, so

made correction on this slide.

People may define the severity

differently, so we use few indicators to give you a

range of severity, so you can pick which one is

appropriate for you. The first one is death. The

second one is serious events, which was the event

that meet the regulatory definition for serious

adverse event.

The next one is hypersensitivity involve

at least two body systems. The next one is the

patient was treated with antihistamine. The last

one is the patient treated with steroid. Most of

them are IV steroid.

If you look at this population, there is

no deaths. There is two cases the sponsor point to

you meet the definition of serious event. There is

13 cases that involve two body systems, 27, or 2.4

percent, of people treated with antihistamine, and

1.5 percent of people need IV steroids.

This slide outline the presenting symptoms

of hypersensitivity reactions. We work extensively

with our internal expert at FDA. We define

hypersensitivity reaction with the following three

groups of symptoms.

First, is skin reaction. The second group

with the respiratory difficulty with cardiovascular

symptoms together. The third one with the facial,

laryngeal, and general edema. This table show the

distribution of the patient presentation.

You will notice the majority of patients

present with skin symptoms, however, this slide

does show that direct injection, they may associate

with a high percentage of people with more severe

symptoms.

This is a slide I would like to bring to

your attention with a comparison with iodinated

contrast agent. The sponsor told you that there

were 4 cases serious AE happened in the clinical

program. That was an incorrect statement. In

reality, there was 29 serious events happened in

the clinical program.

The reason for include there, because the

25 cases, the Agency do not consider is drug

related, therefore, we didn't include it in our

analysis.

In the comparator, iodinated contrast

agents in their Table 9 safety presentation, they

are including all SAEs regardless whether drug

related, so that is we believe incorrect

comparison. So, that is why the number of events

in Combidex group is smaller than the iodinated

group.

This table, we focus on the

hypersensitivity reaction between Combidex and the

iodinated contrast agent. If you read the labels,

three labels which have clinical data for iodinated

contrast agents, totaled together there are 4,545

patients received iodinated contrast agent. There

is no death happening. For Combidex, there is 1

death of all the people receive Combidex. There is

zero out of 1,000 who has clinical dose.

For the serious AE, which is associated

with the Combidex, this is zero over here, and you

have 6 cases out of 2,000 for all doses, you have 2

cases for the clinical proposed dose.

Also, the last one, the column, we show

the percent distribution of those symptoms suggests

hypersensitivity reaction, you can see the rate is

quite different, the relative risk is quite

different. We do not want to draw definite

conclusion over here because we understand the

population are different, but at least this table

do not support this two rate are comparable.

When you talk about whether the drug is

appropriate for populations, you basically talk

about the risk-benefit ratios. From the sponsor's

presentation, they believe the best way to manage

to get a best ratio is to focus on the risks. I

will show you their risk management slides later.

From our end, we believe from the safety

data we have at this time, this drug is definitely

associated with hypersensitivity reaction.

Although we have not observed serious event, more

serious event including death in the proposed

clinical dose, our level of assurance is limited by

the number of patients involved in that group of

patients who received the clinical dose.

At this time, we are only able to say that

the death-related hypersensitivity reaction

probably will now be higher than 1 out of 400 or

500 people based on data. Anything beyond that,

that is purely speculation without any data.

Sponsor present to you their risk

management program. I rearranged our slides.

Basically, they say if we provided dilution and

slow infusion, and educate physicians to the

labeling and to the targeting academic center, they

should be able to adequately address the safety

issue.

We believe this is item we need to discuss

to implement, and also we believe that with

uncertainty with those severe events with this

Combidex administration, when you focus on the

issues, enhance the benefit of this drug to the

appropriate population.

We need to better understand actually the

performance of Combidex by different type of tumor

and the nodal size, because we have preliminary

evidence those performance may vary. Also, we need

to define appropriate patient population or

condition for use, that the use of Combidex, the

benefit will outweigh the risk, potential risk.

This is a table to support our preliminary

conclusion that performance of Combidex may vary by

type, by size of nodes, in addition of the type of

cancer. This analysis actually was conducted by

sponsor. We didn't make any modification to their

slides. We just presented their slides, their

result to you.

On the top is for the nodes less than 10

mm, the bottom row is for nodes larger than 10 mm.

You can see for the nodes less than 10 mm, the

sensitivity from their clinical database is between

67, 66 percent, and the specificity is 80 to 78

percent.

For the nodes larger than 10, the

sensitivity is 93, 98 for different readers, and 56

and 71. This, I would remind you, this is just a

point estimator. We have not put 95 percent lower

boundary yet.

If we put in the boundary, this number

could even be lower. We also don't know whether

there is interaction between size and type of tumor

because so small nodes that was included in the

primary analysis would not allow us to do a further

analysis.

This table showed you the prevalence of

nodes being positive by size of lymph nodes. Why

this information is important is because the

sponsor showed you the positive predictive value

and the negative predictive value in their

presentation.

To better understand that positive and

negative predictive value, you not only need to

understand the performance, that is, sensitivity

and specificity of agent, you also need to know the

prior probability that the prevalence of this node

being positive before you give a drug.

This data collected from their studies,

and for nodes less than 10, because we don't have

the MR imaging measurement, so we have to use the

pathology measurement as a surrogate over here.

For nodes less than 10, the prevalence range from

10 to 21 percent, which means if you see nodes less

than 10 mm, the probability that the nodes be

cancer-positive range from 10 to 20 percent from

this data.

If the nodes are more than 10 mm, then,

the probability from 34 to 60 percent depending on

different study. We still don't know why there is

variations.

Also, you probably reviewed the New

England Journal of Medicine. From their study, the

percentage is even higher. They got 75 percent of

people for the nodes larger than 10 has a cancer.

So, how are we going to put all this

information together to understand or to help us to

understand the value of Combidex to help physicians

in their patient care decisionmaking, or for any

other benefit that they believe is good for

patients?

I will present to you the predictive

values of a positive or negative Combidex test. I

will go over slowly with you. For the lymph nodes

less than 10 mm, the sensitivity is 68, the

specificity is 80. We make this assumption. This

has not been demonstrated by data yet, because the

lymph nodes, the number are too small, but we

assume if this is what we observed.

The prevalence tell you what is the

probability the nodes is cancer, whether they are

cancer-positive nodes before you give Combidex.

The positive predictive value really tell you after

you give Combidex, and if you get a positive

result, what is the probability that node is

metastatic at that time.

The negative predictive value tell you if

you gave Combidex, and the result is negative, what

is the probability that node is negative.

We look at different scenarios. If the

prevalence is 1, based on data or based on your

suspicion, the clinical knowledge, if you are

thinking the node, the probability of metastasis is

only 1 percent, based on this performance, even

Combidex is positive, the probability that nodes

being positive is only 3 percent, so the people

should make their own judgment this kind of

improvement where they have clinical implication or

values to help you to make decision to the patient

care.

When the prevalence get into 10, 25

percent, you see big changes here in the

probability, and this probably will getting higher

if sensitivity and specificity get improved, which

means that after you get a Combidex test, these

nodes more likely become cancer. You may go ahead

to biopsy that one to confirm your suspicion.

However, the positive predictive value is

not that high enough, so we believe with this

probability or likelihood, you will never make

final diagnosis based on the Combidex positive

result only, so most likely you will go to biopsy

to confirm it.

So, we do believe for nodes less than 10,

there might be potential values for Combidex if

performance is constantly demonstrated to help

physicians to select nodes for further evaluation,

to help patients to make some decision.

Let's look at nodes more than 10 mm. You

already heard from sponsor for those nodes, most

physicians will already consider is metastatic

cancer, so for those nodes more than 10, most

likely you will proceed with biopsy anyway without

Combidex.

The question you probably can ask yourself

in that scenario is if I get negative results from

Combidex, is that going to prevent me from going to

a biopsy. Here is the result. As I showed you,

the answer can vary depending on what is the

pre-probability, how likely that nodes being

positive before you give Combidex.

Before Combidex, if the probabilities are

low, then, you get a pretty high assurance if you

get an accurate result, it is going to be a true

and accurate result, however, as you will see, in

my previous presentation, the probability already

got up to 75 percent or 60 percent. In that range,

if you get a negative result, you only get 80

percent assurance that the node is negative. You

still have 20 percent probability the nodes become

positive, so maybe in that scenario, most

physicians probably would still go ahead to do a

biopsy for nodes even Combidex is negative.

So, for that reason, we are seeking your

advice to see how we can understand the values of

Combidex for nodes more than 10 mm for helping

patients.

Also, where you would emphasize what my

assumption here is based on the performance and

which we believe has not constantly demonstrated

from a clinical development program.

So, based on everything I present today is

we believe or the data seem to suggest that

Combidex may not have a value for people with a low

risk, that patients with lymph nodes larger than

10, the value may be limited, and also this cannot

be substituted for the confirmation. Also, we

believe there probably is not a good surveillance

of the recurrence of cancer, because that

population was not studied.

This list and go on and on, and very long,

so that is why we are really concerned with the

general indication. So, the key question we ask

ourself, we are seeking your advice is how the

Combidex result will really benefit to patients.

We don't want to leave you a wrong

impression that FDA do not care about knowing the

nodes, whether positive or not, we care greatly,

however, there is non-contrast agent available. We

try to understand what is additional value with

Combidex to bring it to the table in addition to

the non-contrast agent.

We also understand this test cannot be

used as confirmatory test, so we try to understand

what role this will play to help a physician help

their patients.

We also understand this drug may associate

with the potential, the risk, so we want to make

clear the use of this drug in appropriate

populations, the benefit with risk.

In the later discussion with the sponsor,

sponsor proposes four types of cancer which might

benefit, that Combidex may have a beneficial effect

to the patient, and they also presented those

cancers in their presentation.

For the prostate cancer first, I said

earlier the Agency do believe for nodes less than

10, Combidex may have a potential value, however,

we are struggling with the fact there is only 5

patients from U.S., 5 patients from the European

study included in the primary analysis, and the

estimate is so unstable from the data I just showed

you, we just have no clear understanding what is

the true performance of the Combidex for that

population.

Also, the same concern applied to bladder

cancer, breast cancer, and in less degree to head

and neck cancer, because they have more patients,

but I would like to bring your attention again for

head and neck cancer, most of nodes in European

trial, actually, all the nodes in European trial is

more than 10.

So, with that, I will conclude my

presentation. Thank you very much for your

attention. We are looking forward for your

guidance to help us to determine the efficacy and

safety of this product.

DR. MARTINO: Thank you, Dr. Li.

Questions from the Committee

DR. MARTINO: At this point, I will turn

to the committee and give you the opportunity to

ask questions both of the sponsor, as well as of

the FDA. As you do that, please raise your hand.

Your name will be taken down, and I will call on

you as we go around, so please don't yell out, we

will acknowledge you in turn.

I would like to ask the first question. I

would like the sponsor to make it clearer to me how

they actually looked at the MRIs. I am still not

entirely clear what they did first, what they did

second, and who, in fact, were the radiologists,

were they a specific group of radiologists, were

there any radiologists, please clarify those issues

for me.

DR. GOECKELER: Let me start by saying the

question with regard to who made the diagnoses, the

order in which that was done was shown in the

slides, so that the pre-contrasts were done first,

and those diagnoses were committed to. Then, there

was the paired, and then after some time there was

the post-only.

In terms of who did that, are you

referring to the specific specialty of the

radiologist involved?

DR. MARTINO: No, I am trying to figure

out did you have two radiologists that looked at

all of the films, did you have 100 radiologists? I

am trying to understand that element.

DR. GOECKELER: I will address that, thank

you.

For the U.S. Phase III trial, there were

two blinded radiologists each independently, and

the data has been reported both for each individual

reader or, as reported today, is the average of the

two readers.

DR. MARTINO: Can you also clarify to me

what the task of the radiologist was? I know you

have shown it, but I need it clear in my own mind

what was the charge given to them at each of these

interventions?

DR. GOECKELER: I am going to ask Mark

Roessel to speak to that issue a little bit in

terms of how the radiologists, what they were

actually asked to do on each of the blinded reads.

MR. ROESSEL: The blinded readers were

given training and given the guidelines to evaluate

lymph nodes, but they weren't given any direction.

The nodes were not marked on the images, so they

saw the pre-contrast images and any nodes they

identified, they circled, and they made a

diagnosis.

Then, on the paired evaluation, they did

the same thing. They circled the nodes. But the

nodes were not pre-identified on the images. The

FDA, when we designed the blind read, told us that

if we circled the nodes that we had pathology on,

that that would bias the readers, so the images

weren't marked, and then they did the same with the

post alone, they circled the nodes, put an arrow,

and gave their diagnosis.

Does that answer the question?

DR. MARTINO: It does. What constituted

the denominator for pathology, then, it was the

node as seen post-contrast?

DR. GOECKELER: Well, as Dr. Li indicated

on his slide, one of the reasons that these

patients and nodes drop out along the way is that

the two readings were done on unmarked images, and

then the nodes were also taken out just according

to standard surgical procedures.

So, then, after all those readings were

done, and then the readings had to be matched to

the pathology, so in order to be evaluable at the

end of all that, the node had to be read on both

the pre-contrast image and then identified and read

on the post-contrast image, and then it had to have

pathology.

So, when you impose those sequential

conditions for unmarked images, that is why some of

the nodes fall out along the way.

DR. MARTINO: So, then, it was, in fact,

the same node. The node had to have been seen on

non-contrast, also seen on contrast, and pathology

done. That, then, constituted the denominator. Am

I clear on that?

DR. GOECKELER: Yes, ma'am.

DR. MARTINO: Dr. D'Agostino.

DR. D'AGOSTINO: I have a couple of

questions, first, of the sponsor, and then Dr. Li.

If you look at Slide 9 on the sponsor's

presentation, this is page 5 of the handout.

DR. GOECKELER: Is it possible to get that

slide?

MR. ROESSEL: Yes.

DR. D'AGOSTINO: It was the sponsor's

presentation, I am sorry, the efficacy analysis.

DR. GOECKELER: Could you help us with the

title, what it says on the slide?

DR. D'AGOSTINO: Slide 9 is Nodal

Analysis, U.S., Phase III.

DR. GOECKELER: Is this the slide you are

referring to?

DR. D'AGOSTINO: Yes. I guess I was

surprised that there were no confidence intervals

given as the presentation was made. Later on, the

FDA presentation did have some confidence

intervals.

What I am interested in, in this here, is

how big were these confidence intervals if you

looked at, say, the post-contrasts and compared

them with the pre-contrasts for the paired, I mean

certainly the sensitivity doesn't change or they

would overlap.

Is there a real differentiation between

the specificity or are the confidence intervals so

large that it gets blurred?

DR. GOECKELER: I believe we have a slide

that has the data with the confidence--if not, I

can obtain it, and if someone could pull that data

for me, I can provide it to you. I don't have it

sitting right here this minute. I believe it was

in either the briefing book or if someone could

pull the data.

If you give me just a minute, I can

provide you the answer to that question. Perhaps

we could take another one.

DR. D'AGOSTINO: The other question is,

you know, the second question that follows is, as

you go to the body regions, which is Slide 11 in

this sheet here, how do you make a statement or

what kind of statement can be made from the

statistics point of view, and then hopefully from a

substantive point of view, that it makes sense to

pool these different body regions, because it seems

to me in terms of the questions that are asked

later on, if we go to particular body regions, it

has to be such a small number of nodes involved,

and such a small number of subjects, that the

inferences are really going to be almost

impossible.

So, is there an argument, and I haven't

heard it, that says you can, in fact, combine these

body regions?

DR. GOECKELER: I am going to ask a couple

of the clinicians that routinely image these

patients, but, first of all, you will recall from

Dr. Harisinghani's talk in the beginning that the

mechanism of action of the drug depends on, not a

primary tumor, but a physical process of

displacement of macrophages within a lymph node.

So, the study was designed with a variety

of primary tumors based on the way the imaging

agent acts in terms of imaging lymph nodes.

Mukesh, would you like to comment on that

further?

Well, with regard to the specific body

regions, then, the study obviously was carried out

in a mixed populations of patients, and I think

that obviously, if you start splitting out a large

number of subgroups, the confidence intervals for

any given subgroup increase.

I think that looking at the study as a

whole, which was designed to evaluate the premise

of differentiation of lymph nodes, obviously, that

occurred. With regard to the subgroups, I think

what is important is that there are consistent

trends amongst those subgroups based on the

mechanism of action of the drug.

DR. D'AGOSTINO: Moving on, I have just a

couple more questions, I obviously don't want to

tie up everything here.

In terms of the post-contrast, we were

told in the last presentation that not all the

nodes were actually used because you want to have a

pre- and a post, but there were nodes that were

there.

Was any analysis done on the nodes that

didn't enter into the post?

DR. GOECKELER: Yes, there was a separate

analysis that was done called the "blinded

overread." It is not one of the ones that I

described to you, but it involved a much higher

percentage of the total nodes.

So, it was again a blinded reading of the

nodes, and there was histopathologic correlation of

the data at the nodal level for each of the

readings, and I can show you--

DR. D'AGOSTINO: Yes, it would be nice to

see what the sensitivity and specificity was.

DR. GOECKELER: --what happened in those.

Can you first show the data in terms of

the numbers of patients that were evaluated both in

the unmarked images and in the blinded overread?

These are the numbers that were evaluated

by each reader in the blinded overread, and you can

see, based on the various reads, the number of

nodes that were read and for which there was

histopathologic confirmation for each reader and in

each diagnosis.

DR. D'AGOSTINO: Do you have the

sensitivity and specificity?

DR. GOECKELER: Can you show me the data

on false diagnoses in this, because that

essentially relates to, and we can go back then?

If you have a slide on sensitivity and specificity,

I think you do.

This is the data on the false diagnoses

that occurred in the larger reading population.

You can see the trends are largely the same as we

saw before, about 15 percent with the post-contrast

reads, and 25 percent are slightly higher.

We did see a higher variability between

blinded readers and the blinded overread for the

individual readers.

DR. D'AGOSTINO: It would be nice to see

the sensitivity and the specificity and the

confidence intervals.

DR. GOECKELER: Do you have the

sensitivity and specificity? Get me the numbers,

so that I can just provide them.

DR. D'AGOSTINO: Again, maybe we can come

back to it.

DR. GOECKELER: I can give you the

numbers, and I can tell you that the trends are

very--

DR. D'AGOSTINO: I think it would be very

helpful, but I don't want to tie it up here.

My last question is that you did a lymph

node as the unit of analysis. There is still the

subject, and sometimes in other activities, I don't

know about the nodes, but in other activities, when

you are looking at the same subject, and you are

taking different specimens, and so forth, they tend

to be correlated.

So, if you did a person analysis, what

would you do with the person, what would you say

about the person? Your sample size is greatly

reduced. Are there still your inferences?

DR. GOECKELER: Yes, the analyses were

also carried out at the patient level, so we have

the same data for each of the analyses pre- and

post-contrast at the patient level. I am going to

ask for a slide one more time.

100

DR. D'AGOSTINO: Maybe they can produce it

later on, the confidence intervals around some of

these things I am talking about.

DR. GOECKELER: No, actually, I think they

have it. I will tell you and then the slide will

be up here in just a second, that the trends we saw

in sensitivity and specificity at the nodal level

translated through to the patient level also.

Here we go. But this is nodes less than

or greater.

DR. D'AGOSTINO: It is really not only the

point estimates, but the confidence intervals, what

are you actually saying about the individual, how

much confidence you have.

DR. MARTINO: Dr. Hussain.

DR. HUSSAIN: I have a question to the

sponsor, and it strictly relates to the study

design, because I am still not clear about really

what the design was, so starting with the

eligibility criteria, how were the patients

characterized, were there standardized surgery, and

was the surgery required each time if it was

101

prostate or breast or bladder or head and neck, to

actually do the same template or do beyond what is

normally needed?

And understanding that my specialty, and I

am a gyn-oncologist, that there are certain

prognostic features that will make you feel or

believe that the patient has a high probability of

a lymph node positivity, say, in prostate cancer if

a guy comes in with a T2 disease, PSA of 50, and a

Gleason score, say, of 9, was that accounted for,

because in this patient you would think, based on

clinical criteria only, without even imaging, that

those are very high odds of having this patient

have lymph node positivity.

So, with all that taken into account, and

if it's not, why not, and what is wrong with having

done the appropriate studies, which is accounting

for the subpopulations as having adequate head and

neck patients, adequate breast patients, adequate

lung patients, and so on, to try to make some

conclusions from that?

And final question, and maybe I didn't see

102

it, but what actually was the Phase III trial, what

was compared to what?

DR. GOECKELER: Let me take a couple of

those and then refer some of those to other people

who are more directly involved.

With regard to the comparison, the primary

comparator was the paired evaluation as compared to

the size-based evaluation on pre-contrast. So,

those were the prospectively designed endpoints for

the Phase III studies.

With regard to the treatment of the

patients and how it was decided which nodes would

be sampled, I am going to ask Mark to comment on

that. That varied a little bit as Dr. Barentsz

said between the Phase III studies and what Dr.

Barentsz presented in the post-Phase III studies.

So, Mark.

MR. ROESSEL: In the Phase III studies,

the entry criteria were patients who had a known

primary, who were scheduled for either surgery or

biopsy, and who had suspicion of metastatic disease

spread to lymph nodes.

103

There was no direction as to what the

surgery or biopsy procedures would be. It was just

based on the clinical investigator.

DR. GOECKELER: The standard of practice

at the institution.

MR. ROESSEL: Does that answer the

question?

DR. HUSSAIN: I guess what I am asking is

was it the sense of the treating physicians, or

were there guidelines that said if you had this

size tumor, this kind of risk?

MR. ROESSEL: No, there were no--

DR. HUSSAIN: So, this was left random to

the person enrolling the patient based on their gut

feeling whether the patient have--

MR. ROESSEL: There were no guidelines

given. The entry criteria were just that, patients

with a known primary who were scheduled to have

surgery or biopsy, so that we could get

pathological confirmation of nodal status.

DR. GOECKELER: Did you have another

question, Dr. Hussain, about risk stratification

104

and predictive of--I am going to ask Dr. Roach to

speak to that with regard to relative risk and some

of the models and selection of patients who might

be most appropriate for treatment.

DR. ROACH: In the sponsor's indication,

it specified that patients who were at risk for

nodal involvement, so the clinical use for this

agent in patients with prostate cancer would be

patients at intermediate and high risk disease for

whom we have data from randomized trials that

demonstrates that treating the nodes is beneficial,

and that, in fact, it is important to treat as many

of the nodes as possible.

So, this agent would be useful for

identifying where the nodes are located and allow

us to reduce the morbidity of giving radiotherapy

in patients with prostate cancer.

DR. MARTINO: Dr. Levine.

DR. LEVINE: I have several questions.

First of all, for the sponsor, are you asking that

the individual, that the patient would have two

different MRI scans, in other words, your

105

indication is based on the post-read, so that means

that you are asking that patients are now going to

have a pre- and a post-MRI? So, that was one

question.

My second question, what was in those

benign nodes? You know, there are infiltrative

diseases of nodes, TB, MAC, et cetera. What were

those benign nodes, and what kinds of benign

conditions, in fact, fulfill your requirements for

benign?

Number 3. This is kind of a funny one,

but how did you know that the correct node was

actually taken out? Did you do an MRI scan after

surgery to know that you really took the right node

out?

DR. GOECKELER: Let me ask, in terms of

the matching, since Dr. Harisinghani has been

involved in a number of these studies, how that is

done.

The first part of the question dealt

with--I am sorry?

DR. LEVINE: Is the company requesting

106

that the patient have two different--no, not two

different reads--two different MRI scans?

DR. GOECKELER: Two different images,

yeah.

DR. LEVINE: And who pays for that?

DR. GOECKELER: In the conduct of the

clinical studies, that was required, because the

primary endpoint was the comparison of a

pre-contrast and a post-contrast read, and I am

going to let the radiologists comment upon how they

read these scans and how they match the nodes in

the clinical studies.

DR. LEVINE: That actually wasn't the

question. The question is if this compound is

licensed, are you asking that the patient be sent

to MRI scan twice?

DR. HARISINGHANI: And the answer is yes,

the patient will require two scans pre- and after

contrast administration, and in terms of being able

to correlate the nodes specifically to the areas on

how we know that surgically, we are right, it is an

arduous and a difficult task, and for that reason,

107

we have developed exquisite anatomic maps to which

we map the nodes when we read these out, and the

surgeons then correlate them to fix the anatomic

landmarks, which could be the vessels or bony

landmarks, and that is how they figure out where

the nodes lie.

DR. LEVINE: All right. Another question

was the character of the reactive lymph nodes, what

were they?

DR. HARISINGHANI: The benign enlarged

lymph nodes ranged in etiology. Most of them are

reactive nodes, not pointing to any specific

etiology for the so-called reactive lymph nodes,

but we had occasional cases of sarcoidosis.

I must say there were no caseating

tuberculosis at least in the trials that I have

been involved. I am not sure of the general trend,

but the benign nodes mainly were reactive and

enlarged.

DR. LEVINE: And the sarcoid case

fulfilled your criteria as benign, as well?

DR. HARISINGHANI: Yes, that was the case

108

I showed earlier in the presentation where it

behaved like a reactive lymph node.

DR. LEVINE: Have you guys done a cost

analysis of the efficacy of this approach given the

fact that you are going to do two MRI scans, is

there a cost analysis perhaps?

DR. HARISINGHANI: We have not formally

studied this in the States, but Dr. Barentsz's

group in the Netherlands has actually published

their results on cost saving.

Do you want to comment on that?

DR. GOECKELER: Also, just let me comment

that although two separate imaging sessions were

required in the clinical trials, because of the way

that clinical trials were conducted, different

investigators in the post-Phase III setting

interpret pre and post different ways, and Dr.

Barentsz can comment on that also.

DR. BARENTSZ: I would like to comment on

the first question first, about cost. We recently

published a paper in European Radiology in which

we, based on the sensitivity and specificity data,

109

did do a calculation and analysis on the health

care perspective.

If you are including this technique, it

will save, in Europe, 2,000 euros per patient, but

that is I think not the most important thing. The

most important thing, it saves also morbidity.

That was not taken in account in that study.

To reflect on the pre- and post-contrast,

as among radiologists there are some discussions

going on, at this moment, with some newer

techniques, you are able to make a sequence which

is insensitive to iron, so you can tell the machine

"Iron Off," and you can tell immediately after

that, "Switch on Iron," and that will substitute

for the pre-contrast examination.

Nonetheless, to start in the initial phase

for new readers to get some experience, it is

advised to use both of those examinations pre and

post. I am performing now and studying in the

Netherlands, in foreign patients in prostate

cancer, a multi-sound study only doing the post

just by having insensitive and sensitive sequence.

110

Also, if you have looked at the data of

the sponsor, you can see that if you do the

post-read only, it gives a very good result.

Perhaps you can comment on that also, Mukesh.

DR. HARISINGHANI: I think, as Dr.

Barentsz alluded to, for initial training purposes

you need both scans. Once the individual is

trained, then, yes, with the existing technology,

we can then, as he said, switch on and switch off.

Then, it would be possible that you could just do

the post-contrast study.

MR. ROESSEL: If I might add, because we

need to be clear about labeling for this, as the

sponsor, the proposed labeling, the proposed

package insert does not specify that you have to do

a pre-contrast image and a post-contrast image.

DR. MARTINO: Dr. Mortimer, you are next.

DR. MORTIMER: I wonder if the sponsor

could clarify the management of the lymph nodes.

Were the lymph nodes just handled in a routine

fashion? Were those nodes that were suspicious

handled in any different manner to ensure micro

111

metastatic disease?

DR. GOECKELER: Let me make sure I

understand. In terms of obtaining them in surgery

or--

DR. MORTIMER: Actually, reviewing them

histologically, so to make an analogy of sentinel

node mapping, the sentinel node is immunostained.

DR. GOECKELER: I think I understand. The

histology was reviewed without knowledge of the

image findings. So, they didn't analyze those

particular nodes any different than they did any

other nodes that were in the study.

DR. MORTIMER: And it was just H and E

slicing and--

DR. GOECKELER: Right.

DR. MARTINO: Dr. Perry.

DR. PERRY: A comment for Dr. Li. Your

point number 2 about inadequate representation of

tumor types, I don't think the sponsor ever

attempted to try to do all sorts of tumor types.

For many kinds of cancers, this methodology is not

necessary. For melanoma, as an example, we have

112

other staging systems or imaging systems that are

quite sufficient.

So, I think it is an unfair criticism to

say, when they set out to study four tumor types,

that they didn't do all the tumor types. I don't

think that is--that is a cheap shot in my opinion,

and I don't think that is an appropriate criticism

of the sponsor.

For the sponsor, when it comes to

education should this product be approved, I think

you are focusing on the wrong market. I think if

you put the emphasis on physician education, you

are really going to miss the mark by a long shot.

It is really the tech who gives the medicine, it's

not the physician.

I don't know any physician that I have

ever seen administer a contrast agent. Perhaps

it's different in Europe or in other locations, but

if it is, I would like to know that, but it seems

to me it is the techs who are going to need to be

educated and make sure that they give it the right

way, and if you focus on the physicians, you are

113

going to have problems.

DR. MARTINO: Dr. Brawley.

DR. BRAWLEY: There are a couple of

statements that were made in the FDA presentation

that I would like to get the sponsor's response to

them.

The first is of 152 and 181 patients who

received Combidex in the U.S. and the European

studies, a third of patients were censored from the

U.S. study, and two-thirds of patients were

censored from the European study, and not included

in the primary analysis.

I would like your response to that, and

then I have a couple others.

DR. GOECKELER: Yes, sir. First of all,

with regard to the European studies, as I think

someone indicated at the beginning, the European

studies themselves were initially carried out by

the European sponsor with different endpoints, so

they were analyzing patients at the patient and

group and nodal level.

So, in those studies initially, there was

114

nodal matching predominantly only amongst the large

nodes because it was felt at the time, and you have

to recall that these studies were all done seven or

eight years ago now, it was felt that the matching

could be better done on those large nodes, and I

think that is why there is a disproportionate

number of large nodes in the European studies.

After the studies were done, the sponsor

met with the FDA and agreed that they could take

data that was acquired at the individual node level

in those studies and analyze it in a blinded read

through the same sort of matching procedures, using

the same sort of analyses that were carried out for

the U.S. study.

So, one of the consequences of that is

that there were a large number of nodes removed

from those patients that weren't matched on a

node-by-node level. So, if you look at the gross

number of nodes, and the numbers that were

originally--and then the ones that were eventually

matched up by two blinded readers and then had

pathology, it's a smaller percentage in the

115

European studies.

DR. BRAWLEY: A couple more follow-up

questions.

I am told that there are only 5 prostate

cancer patients from the U.S. and 5 from Europe in

the primary analysis. Is that true?

DR. GOECKELER: Yes, that's true, and one

of the reasons, if you look at both the U.S. and EU

Phase III studies, the purpose of the studies was

to investigate the ability of the agent to

differentiate nodes, malignant from non-malignant.

I think that when you move on to--and

obviously, you can subset that a lot of different

ways, either by body region or individual tumor, or

any number of other ways, and if you do that,

certain categories will be large or small, and the

confidence intervals will react accordingly.

I think that that is why, when we turn to

the issue--and I think those studies did show that

Combidex improved the ability to differentiate

malignant from non-malignant lymph nodes.

I think that as Dr. Li indicated and as we

116

indicated, when you move on to the question of

where does that provide a clinical benefit, the

tumors that we presented on were ones where not

only we believe there is a clinical benefit, but

also that there was supplemental data post-Phase

III, not only on imaging performance, which you saw

in the slides that Dr. Barentsz provided, but also

on how that imaging performance impacted on

clinical utility.

DR. BRAWLEY: So, you are trying to

convince the committee that this drug is safe,

effective, and efficacious in prostate cancer with

a series of 10 prostate cancer patients.

DR. GOECKELER: Well, I wouldn't make the

argument about the risk-benefit solely on those 10.

I think we have to look at some of the additional

supplemental data that is available from other

places, such as the publications in the New England

Journal and other places.

DR. BRAWLEY: I have also heard that

certain source documents, including a pre-defined

statistical plan, blinded reader manual, the

117

original copy of the blinded reader efficacy

evaluation, were not available to the Food and Drug

Administration.

I would like you to respond to that

allegation.

DR. GOECKELER: Well, I think that there

have been some questions raised about the exact

sequences of events in which the nodal imaging

guidelines were developed and finalized, and I

addressed that on one of the slides that I

presented from the sponsor's perspective. The

guidelines were finalized prior to any blind

reader, availability of blind read data. Mark, if

you would like to expand on that.

MR. ROESSEL: I am sorry, I think you are

answering a different question. I think the

question was about the prospective plan being

available for the New England Journal of Medicine

article. Is that correct?

DR. BRAWLEY: That's correct.

MR. ROESSEL: The material that was

published in the New England Journal of Medicine

118

article, as Dr. Li really nicely showed, was done

independently of the sponsor. Two clinical

investigators, one in Europe and one in the U.S.,

got together and took 40 patients from trials that

they were conducting and did a blinded read.

We don't have, as the sponsor, again, it

was done independent of us, on their own

initiative, I think is the way Dr. Li put it, we

don't have from them a prospective statistical plan

or prospective plan for conducting that blind read.

We do have that for our Phase III studies,

of course, for our clinical studies.

DR. BRAWLEY: Let me just say

parenthetically that that is an acceptable answer,

I understand that answer, but I need, and I don't

want to criticize this company, Advanced Magnetics

at all, I definitely don't want to impugn Advanced

Magnetics, and I do want the news media to listen

to this.

In my last four years here, I have seen

some companies come before this committee, and some

companies submit data to the FDA, and what is done

119

is sort of slight of hand, with selection biases in

terms of choosing patients, to try to make one's

point that a particular drug or a particular agent

works, and we have to be very, very careful

whenever we look at data to understand exactly what

the source of the data is and the validity of the

data, and most importantly, the selection biases of

the patients going into the data before we can make

a decision.

That is a point that has been missed

repeatedly in a number of newspaper editorials

about drug approval recently, so that is the basis

for my question. You, sir, you did give me an

acceptable answer, and again I want to state I

don't want to at all impugn your company.

Last question. I heard that a patient

died getting this contrast agent. I thought I

heard that the patient got the contrast agent in a

facility that was not able to treat an allergic

reaction.

Is that true?

DR. GOECKELER: Mark, you can comment on

120

the facility, and I am going to ask Dr. Bettmann to

comment on sort of the guidelines and regulations

regarding what those sorts of facilities are

required to have.

MR. ROESSEL: The facility in question was

a free-standing MRI unit. We made sure in our site

qualifications for doing clinical trials that

equipment was available to treat any reactions that

occurred. They did have emergency equipment, which

I think is what you asked me, they did have it

available. Apparently, they didn't choose to use

it.

DR. BRAWLEY: That, too, is an acceptable

answer, I just want to go on the record as saying.

DR. MARTINO: Dr. Houn, did you want to

make a comment?

DR. HOUN: Yes, just to clarify when a

sponsor obtains right of reference to studies to

support their application, they have to be able to

provide to FDA access to underlying data to provide

the basis of the report of the investigation.

This did not happen with the New England

121

Journal study, and also just as a reference to the

committee, FDA didn't mean to give a cheap shot in

terms of the numbers of people enrolled, just in

previous approvals for ProstaScint, prostate cancer

only imaging drug, there were 152 people entered

into the analysis only with prostate cancer, and

there were 183 that were followed for the open

label efficacy study.

When we did NeoTec, a lung cancer

detection for non-small cell lung cancer, there

were 228 entered into the analyses. When we

approved PET-FDG, that got a broad indication for

all kinds of cancers. There were 1,311 people

entered into the analyses.

DR. MARTINO: Dr. Reaman.

DR. REAMAN: Just a question again about

the eligibility criteria, and I guess to somewhat

follow up on the issue of selection bias.

You stated that any patient with cancer

who was at risk for developing lymph node

metastases were eligible for this study, and they

were eligible based on whether or not they were

122

going to then have either a biopsy or a surgical

procedure.

So, how was the decision as to whether

they were going to have surgery or a biopsy

procedure made, by equivocal or positive

radiographic studies before they were entered on

this study, or did they have palpable adenopathy?

Other than the breast cancer patients in the

sentinel node biopsy, I am still not satisfied that

this isn't a selected population.

DR. GOECKELER: I will ask Mark to expand

on that, but I believe it's the case, and Mark can

verify, that the image findings, the post-contrast

image findings could not play a role, and were not

available to the physicians in making those

assessments.

So, the physicians did not have any

post-contrast image findings on which to base that

assessment of whether the patient then went on to

surgery or biopsy. It was done based on the normal

clinical information that would be available to

make that decision for every other patient.

123

DR. REAMAN: So, radiographic studies

weren't part of the clinical information?

DR. GOECKELER: Well, I think that the

pre-contrast, you know, you could have a CT or an

MRI pre-contrast, but no post-contrast image

findings.

DR. MARTINO: Dr. Bradley.

DR. BRADLEY: I have a couple of questions

maybe for the authors of the New England Journal

article, following up on a question by Dr. Li.

How did you select those 40 and 40

patients from a group that was 3 times larger? I

mean selection bias kind of comes to mind, but what

selection criteria did you use?

DR. HARISINGHANI: It is 3 times larger

now, but it wasn't then. The selection was

consecutive patients who were scheduled to undergo

radical prostatectomy both at the U.S. and at the

European site.

They were of the intermediate and

high-risk category, I must admit to that in terms

of the patient selection.

124

DR. BRADLEY: And then a follow-up

question. You showed some very nice images of very

small nodes, one of you, or positive nodes. With

5-mm cuts, and no way of guaranteeing that you are

in the same place for the second scan, how do you

know you are comparing the same nodes pre and post,

particularly not for you, but for the chest where

you have respiratory artifact?

DR. BARENTSZ: In our New England Journal

paper, we used 3-mm cuts in the obturator plane,

and we used 5-mm cuts in the axial plane. We

performed a combination of sequences which

visualized the anatomy and also a sequence which

visualizes the iron, and based on also a 3D

sequence which we performed, we were able to

compare the pre and post and exactly locate the

lymph nodes where they were, so we could make a

very accurate match on the 3-mm and 5-mm images.

Also, we located the nodes in relation to

the vessels. So, I agree with you that

localization and the location of lymph nodes is

very important.

125

DR. BRADLEY: So, the slice location of

3-mm slices was accurate, looking at the other

anatomy?

DR. BARENTSZ: Absolutely.

DR. BRADLEY: A follow-up question. On

the 15 percent--this may not be for you guys--but

15 percent false positive and false negative, we

have talked a little bit about what might cause a

false positive. What about false negative, any

thoughts, did you do an analysis of why they were

false negative?

DR. HARISINGHANI: I think there are two

issues here at least from our study. I would let

Bill answer for the general part, but the false

negatives are mainly as we are talking of nodes

which are smaller than 5 mm, then, the current

resolution of our scanner only enables us to be

confident at a certain level, and that could

account for the false negative reads.

DR. BRADLEY: Then, one final question for

the sponsor. Why did you choose a 0.2T Hitachi

when this is clearly a magnetic susceptibility

126

agent? Is it so sensitive that a gradient echo at

0.2 shows you what you see at 1.5? Also, I suspect,

having read all of this, that that was also where

you had your single death, is that correct?

DR. GOECKELER: I am going to have to ask

Mark or Paula to comment on the specific imaging

equipment. Please recall that the death was in a

liver imaging study, not in a lymph node imaging

study.

DR. BRADLEY: Right. I saw the physician

of record on that, who happens to own a bunch of

low-field magnets in Ohio. I am just wondering if

it is the same case. But why include a 0.2 at all?

MR. ROESSEL: We tried to include in the

Phase III clinical studies, we didn't specify the

imager to be used. There was no requirement for it

to be a 1.5T or 0.2T. The fact is we provided the

Agency with the information on the types of imaging

equipment used, and I think most of them were 1.5T,

the vast majority. It was a very, very small, I

think one or two that used 0.2T in the studies.

DR. BRADLEY: Just to follow up, was the

127

0.2 Hitachi also where the death occurred?

MR. ROESSEL: That, I don't know.

DR. MARTINO: Ladies and gentlemen, we are

running short of our allotted time, but I

appreciate these questions as important, and that

is why I am giving you a little more time in this

part of the meeting.

That being said, I would ask those of you

asking the subsequent questions, please be sure

that your questions are necessary to your thinking

about the efficacy and the approval of this agent,

and are not just purely for your perhaps

intellectual curiosity.

Dr. Giuliano.

DR. GIULIANO: I am a surgeon, Dr.

Martino. We have limited intellectual curiosity,

so my--

DR. MARTINO: I know.

[Laughter.]

DR. GIULIANO: Therefore, my questions

will be brief. But I am struggling as a surgeon

through these documents. We say the surgical

128

procedure was not altered, the post-enhancement

images were not available.

How did you instruct the surgeon to remove

the Combidex abnormal enhanced lymph node? He or

she had to know what that node was, where it was.

It had to be labeled as such. So, on a

node-by-node analysis, I think that introduces a

surgical bias because as any surgeon knows, it is

easier to find a positive node than a negative

node.

In addition, using the node-by-node

analysis, what happens with nodes not seen on MR

that are removed? For example, if this agent did

not alter your surgical operation, the patient with

a prostatectomy may have had a pelvic lymph node

dissection, and there was one node that had been

identified on your preoperative images or an

axillary dissection for breast cancer, and there

are one or two nodes, and 15 or 20 nodes were

removed.

If you look at the 1 or 2 nodes, which had

to be seen on the image, had to evaluated

129

histopathologically, and they correlated, let's say

they were both negative, what if all of the

remaining nodes were positive or one of the

remaining nodes was positive, how was that dealt

with statistically or in your presentation? I

could not understand that.

DR. GOECKELER: I will ask Dr. Anzai to

talk about the nodal matching and how those nodes

were identified, and how imaging was or wasn't used

in the identification of those nodes.

DR. ANZAI: I am the radiologist involved

in Phase II and III clinical trials. Your comment

is absolutely right. This was the hardest trial

that we ever had in Radiology, that I personally

have to have images going to OR when the patient is

in operating site, and we have to ask a surgeon to

make stitches on a certain anatomical level.

For example, a head and neck radiology, I

have to ask the surgeon to make stitches on the

submandibular--this is the jugular vein, so in

between this lymph node is the lymph node that I am

seeing in imaging, and it was very labor intensive.

130

Many of the radiologists have to be in the

OR with this graph, and the surgeon to identify,

correctly identify those lymph nodes on imaging, or

lymph node in a patient, so the pathologist would

identify this is the exact lymph node that we saw

in imaging.

That is why the sample size was so small,

because we have to have a certain confidence that

the imaging on the lymph node is matched with final

pathology. That is why the size of the lymph node

that is seen in all the cancer patients are small,

but this is such a labor intensive study, but we

did as much as possible to correlate imaging on a

lymph node with surgical pathology by being in the

OR.

The second question for statistics, maybe

Mark can comment.

DR. GOECKELER: I think that the issues

that have just been identified by Dr. Anzai and

others are the ones that account for the analysis

that Dr. Li showed, where you start out with a

large number of nodes and then if you are going to

131

require evaluation on unmarked images to avoid bias

in the reading of the data, then, you lose some

nodes along the way, because the readers don't all

identify the same nodes every time they read.

That is why you see some of the nodes or

the numbers dropping off at every level. We tried

to address that in part by looking at another read

that involved the blinded overread, which are a

much larger percentage of the nodes.

DR. GIULIANO: Maybe I wasn't very clear

about that. My question is if the labeled node

from the operating room is the one identified on

the MR, and histologically evaluated, and is

positive or negative or whatever the correlation

is, but other nodes that were not seen are

positive, was that counted as a false negative or

was that not counted because the other nodes were

not seen on MR?

DR. GOECKELER: No, the primary analysis

was at the nodal level, so those numbers that were

presented were at the nodal level. There were

other analyses the data tracked very closely at the

132

patient level where you can look at the patient

level also.

DR. GIULIANO: Thank you.

DR. MARTINO: Does that answer your

question, Dr. Giuliano, because I am not sure that

it did.

DR. ANZAI: Let me add one thing. I think

your question that the lymph node that not

identified on the MRI, how do we handle that. I

think a nodal level correlation, we didn't look at

those lymph nodes were pretty not pre-identified by

imaging, but a patient level analysis, if, for

example, MRI showed all the normal lymph node, but

pathology somehow find one positive lymph node that

not identified MRI, I think that was considered to

be false negative.

DR. GIULIANO: Perhaps you could share

that patient analysis, would that be appropriate,

Dr. Martino?

DR. MARTINO: Well, to be honest with you,

I think at this point you are going to have to make

your decision realizing that the data that you need

133

perhaps are not presented to you right now. I

think that may be one of the issues.

Dr. Bukowski.

DR. BUKOWSKI: I am trying to understand

the efficacy and benefits of this approach, and

there was a statement made that there is a decrease

in morbidity when you apply this particular

product.

Can you help me understand what the

implications are? Are you implying that there will

not be a need for surgery if there is an identified

positive node, or that there will be then a

percutaneous biopsy done, and, if so, what is the

likelihood of being able to biopsy the small nodes

that you are referring to, less than 10 mm, using

techniques not only at academic centers, but

centers elsewhere?

DR. BARENTSZ: You raise a very good

point, and I would like to address a little bit to

our New England Journal paper, which is different

from the Phase III study in that way, that in the

New England Journal paper, we were able to--we were

134

allowed to include data which were obtained from

the Combidex MRI into clinical practice.

So, that paper shows better the real

clinical effect of what this contrast agent can do.

So, if we found an extra node, we were allowed to

tell to the surgeon, and I again agree with you,

communication with the surgeon where the node is,

is very important.

Mukesh and I, we started by making some

nice schemes, which have been used by the surgeon,

and sometimes we, well, we went to the surgery

room. So, we added the information of the MRI for

the surgeon, and we asked our surgeon how this

scan, how did this really change his management,

did that decrease the extent of surgery.

Actually, the black nodes, they are

normal, and if you have a high sensitivity and a

high negative predictive value, but if you have

both very high, as what we obtained in our paper in

the New England Journal, both on the patient and as

on the nodal level, that means that the risk after

an MRI, that the patient has a negative lymph node

135

is extremely high.

That means the number you are missing is

extremely low, and that current threshold, our

urologist advises, but I would like also to have

one of the urologists to speak on that. That is

very important clinical information which may

actually decrease the number of lymph node

dissections.

If you have a positive lymph node, it

always must be confirmed histopathologically. If

it's large, 7 mm, 6 mm, or 10 mm, you can do that

by image-guided biopsy. If it's smaller, you have

to tell the urologist the node is down there, and

he can remove it.

Perhaps the urologists can make also some

clinical remark on that. Comment about the

clinical use, how this technique can be applied,

what will you do if you have a negative MR

Combidex, what will you do if I am saying it's a

positive lymph node.

DR. KALINER: Well, first of all, any

information that I give as a clinician, first of

136

all, I am a urologist for the last 16 years at

George Washington University, and recently joined

Cytogen as the Vice President of Medical Affairs,

so I have a lot of experience in surgery and

urology.

Any information I can get that helps me

identify whether there is more extensive disease or

not is extremely important with these patients.

So, in the case, if I have a negative Combidex

scan, first of all, I wouldn't do a Combidex scan

unless it is somebody that is intermediate to high

risk, as many of these patients were, so they are

stratified by risk to begin with.

So, this is somebody that has a negative

Combidex scan, we still would perform the lymph

node dissection, but if there was a reason to look

in an extended area, which we know pathologically

does occur, then, that scan can help guide us to do

that.

On the other hand, if we did find

something ahead of time, we may be able to

eliminate doing an invasive procedure by performing

137

a biopsy or perhaps a laparoscopic lymph node

dissection as opposed to an open procedure. There

are a variety of ways to look at doing that.

Any way that I can get more information to

help prevent an invasive procedure when it is not

necessary is extremely important.

DR. MARTINO: Dr. Dykewicz.

DR. DYKEWICZ: I have two questions

regarding safety and adverse events. The first is

whether slowing the rate of the infusion as

proposed will really reduce the risk of

hypersensitivity reactions.

In the sponsor's presentation, there was

data presented showing that the number of adverse

events were reduced with the use of that

administration method, but, of course, adverse

events could include both hypersensitivity and

non-hypersensitivity events.

Hypersensitivity events are the ones that

are potentially going to lead to fatalities, so

that is where I have my greatest concern.

The FDA analysis was that the overall risk

138

in severity of hypersensitivity reactions was

actually not reduced, and they presented one data

on Slide 21, Presenting Symptoms of

Hypersensitivity Reactions, that showed that at

least in terms of urticaria, the rate even

increased with slowing the infusion rate from 63

percent with the bolus to 85 percent.

Some of this I think is probably just a

result of the signal of having a relatively smaller

population with the bolus group, but from the

standpoint of the sponsor, are you of the belief

that the slower infusion rate will significantly

reduce the risk of hypersensitivity reactions?

DR. GOECKELER: I think the issues are

related to risk and management, and I am going to

ask Dr. Page to speak to that, please.

DR. PAGE: The most telling data about

this are to look, not at all hypersensitivity

reactions, which again tended to be--this is an

iron product, so that the notion is that any

exposure in the bloodstream is likely to cause some

activation of mediators, so you are going to see

139

some flush.

So I would contend that the notion of

hypersensitivity is probably too broad. That is

what we are looking at, it is a hypersensitivity

reaction, and in that sense, I agree with the

statement that it is not clear that dilution will

reduce rates of hypersensitivity, but I believe the

data show convincingly that they will reduce severe

both all AEs, as well as hypersensitivity AEs.

In the case of bolus, there were 3 serious

adverse events out of 131 patients. That is a rate

of 2.5 percent. In the case of diluted, there was,

in fact, only 4 out of 1,200, and, of course, that

is a rate on the order of 0.3, so there is a log

order difference in the rate of severe adverse

events. That is one piece of information.

The other is we know that in patients who

are having an immediate hypersensitivity reaction,

you can turn off the infusion, the reaction goes

away, and you can restart the infusion. So, it is

not only the accrued rate of all the reactions.

The real question is severe, and the reason is can

140

you intervene.

DR. DYKEWICZ: The second question, which

actually dovetails with that, and a question that

Dr. Brawley had asked about earlier, is the acute

treatment of the serious hypersensitivity

reactions.

Were any of these patients given

epinephrine?

DR. PAGE: I believe none were. Mark,

correct me if I am wrong there. Some were given

steroids, of course, some were given albuterol in

one case. As far as I recall, there was no

epinephrine given.

DR. DYKEWICZ: Well, this is no indictment

specifically of the sponsor, but for discussion

later, I would raise the point that the treatment

of choice for a serious hypersensitivity reaction

would be epinephrine.

DR. PAGE: And would you say that is true

if there was no hypotension and on cessation of

infusion, and there is no acute respiratory

compromise?

141

DR. DYKEWICZ: Potentially, yes. Studies

have shown that in anaphylaxis, delay in the

administration of emerging anaphylaxis is

associated with an increased fatality rate.

Obviously, this requires some clinical

judgment depending upon the clinical presentation

of the patient, but I would say that, in general,

if you have patients with serious hypersensitivity

reactions, that none have received any epinephrine,

that is sad in my opinion as an allergist.

But again, this is nothing specific for

the sponsor of this agent. I think it is

reflective of the standard of care generally.

DR. GOECKELER: Dr Bettmann.

DR. BETTMANN: I wanted to comment as a

clinical radiologist. I think your point is very

well taken. My recollection of the data are that

the only patient that was given epinephrine was the

one patient who died, and that patient was given in

a very delayed fashion, so it was inappropriate.

Again speaking as a clinical radiologist,

it gets to the point of who treats these reactions

142

and how, and how are they trained, and that gets

back to what Dr. Brawley touched on about why was

the study done, that one fatality, in a place where

the reaction couldn't be treated appropriately.

I think the answer is simply that there

are, the American College of Radiology has very

clearly stated that contrast should not be injected

where there isn't equipment to treat reactions that

are potentially fatal and where there aren't people

who are ACLS trained.

So, you started by saying it's not an

indictment against the sponsor, I think perhaps

it's an indictment against clinical radiology.

There is no question that patients should be

treated appropriately, there is no question that

the appropriate treatment is known. It is a matter

of linking those two.

I think that is a question that sort of is

unfortunately way beyond Combidex.

DR. MARTINO: Thank you.

Dr. Rodriguez. For the rest of you, there

is only three of you. Please be brief and

143

succinct.

DR. RODRIGUEZ: I just want to be very

clear about one issue. One of the committee

members previously said that the company obviously

did not intend this product to be used in all

malignancies.

As I read the application or in this

proposed indication, however, it is worded exactly

the same in both the FDA presentation and the

sponsor, and it states that it is to assist in the

differentiation of metastatic and non-metastatic

lymph nodes in patients with confirmed primary

cancer who are at risk for lymph node metastases.

So, to the sponsor, are you, in fact,

requesting that the FDA approve this product for

broad application in all malignancies?

DR. MARTINO: I will take a yes or no

answer to that. That is all that is necessary in

my mind.

DR. RODRIGUEZ: That is all I need.

DR. GOECKELER: The indication was based

on the Phase III clinical trials. I think the FDA

144

and the sponsor --well, that is the indication that

is being sought, yes.

DR. MARTINO: Thank you.

DR. D'Agostino. Succinct and brief.

DR. D'AGOSTINO: I will be very brief.

Just to go back to some of the questions I raised

earlier in here, it seems to me, and the sponsor

can say yes or no, that what we are dealing with is

trying to evaluate efficacy based on not all the

subjects available, not all the nodes available, if

there is differences between the pre and post in

terms of sensitivity and specificity, it is

basically on a per-node basis. It is not based on

per type, body region, and it is not based on a

per-person basis.

I don't see any justification for

combining the body regions by statistical criteria.

I didn't see anything on what happened to the nodes

that weren't in the paired analysis, and I think on

the per-patient basis, you have such a small number

of patients, that we probably don't have any

significance on sensitivity, specificity, and

145

disposition of the patient.

A yes or no from the sponsor would be

interesting.

DR. GOECKELER: There were a lot of

questions. First of all, with regard to the body

regions, those weren't combined. The data sets

were for the entire populations. They were

subgrouped out after the fact.

So, the primary analysis was for

differentiation of metastatic from non-metastatic

lymph nodes based on the entire population. That

is why the indication that is being sought is

written the way it is.

With regard to the question of where there

is a clinical benefit to that, I think that is why

we presented additional data from additional

studies in specific cancers.

DR. MARTINO: Mr. Kazmierczak, the last

question.

MR. KAZMIERCZAK: Thank you. My one

question on generalization was already asked and

answered. In the FDA's presentation, they

146

indicated that certain patients were excluded on

the basis of pretreatment with radiation or

androgen ablation.

I would like to have the sponsor comment

on whether the FDA statement that Combidex should

be used for newly diagnosed patients as a

restriction is reasonable.

DR. GOECKELER: I think it is the

population that has been studied in clinical

trials, yes.

DR. MARTINO: Thank you, ladies and

gentlemen. I will give you a five-minute break

only. I will start without you.

[Break.]

Open Public Hearing

DR. MARTINO: The next portion of this

meeting is the open public hearing. Those of you

who have requested permission to speak at this

portion of the program, I will remind you that you

have five minutes only. Please identify

yourselves, and there is a microphone in the middle

of the room, which is the one that you will be

147

using.

I need to read a statement, so that you

all understand the purpose of this portion.

Both the Food and Drug Administration and

the public believe in a transparent process for

information gathering and decisionmaking. To

ensure such transparency at the open public hearing

session of the Advisory Committee meeting, the FDA

believes that it is important to understand the

context of an individual's presentation.

For this reason, FDA encourages you, the

open public hearing speaker, at the beginning of

your written or oral statement to advise the

committee of any financial relationship that you

may have with the sponsor, its product, and, if

known, its direct competitors.

For example, this financial information

may include the sponsor's payment of your travel,

your lodging, or other expenses in connection with

your attendance at the meeting.

Likewise, the FDA encourages you at the

beginning of your statement to advise the committee

148

if you do not have any such financial relationship.

If you choose not to address this issue of

financial relationship at the beginning of your

statement, it will not preclude you from speaking.

MS. CLIFFORD: Our first speaker is Mr.

Curtis Holladay.

MR. HOLLADAY: I do not have any financial

affiliation with the sponsor, however, my travel

and lodging is being paid for.

I am Curtis Holladay, a 73-year-old

prostate cancer survivor of seven years, here this

morning to tell you about my recent experience with

the Combidex test. As you will see, this

diagnostic tool was crucial to understanding the

stage and disposition of my prostate cancer thereby

allowing the opportunity for experts to prescribe

the appropriate therapy.

Diagnosed in 1997, I subsequently

underwent radiation therapy, both seed-implant and

external beam. After it had become evident that the

radiation therapy had failed, a hormonal therapy

was employed, but discontinued after a year due to

149

liver toxicity.

During this time, four bone scans and two

computed tomographies revealed no evidence of

metastasis. In order to determine my eligibility

for local treatment, the question of metastasis to

the lymph nodes had to be answered.

Internet searches and consultation with

Dr. Stephen Strum led to the Combidex technology as

offering the most reliable test. Although one of

the Phase III clinical trials was run in the U.S.,

it was still not available here, but it was

available at UMC St. Radboud at Nijmegen in the

Netherlands. It was under the direction of Dr.

Jelle Barentsz whose work was reported to be

outstanding. The importance of the information to

be gained left me no choice but to travel to the

Netherlands at my own expense.

Our party arrived at the Amsterdam Airport

early morning and drove to Nijmegen for me to

receive the Combidex contrast injection that

afternoon. The MR scan was performed the next day

allowing the required 24-hour waiting period.

150

There was no pain or unpleasant effect from the

Combidex injection.

Dr. Barentsz reviewed with me the MR scan

images. He pointed out images of lymph nodes on my

left side were white or illuminated, indicating

metastasis. Images of the lymph nodes on my right

side were dark or black, indicating that they were

normal, free of metastasis.

Although I had hoped for a better outcome,

it was better to know than not to know. The

Combidex test made it clear that a local therapy

was no longer an option and that a chemo-based

therapy would be necessary to check the metastasis.

I would hope my personal testimony helps

persuade the FDA to approve the Combidex test for

use in our country as it becomes more evident every

day that we need to bring available tools and

resources to bear on this unrelenting disease.

Thank you for the opportunity to make this

statement.

MS. CLIFFORD: Thank you, Mr. Holladay.

Our next speaker is Barbara Lestage.

151

MS. LESTAGE: Good morning. I am Barbara

Lestage. I am a 9-year breast cancer survivor from

Wrentham, Massachusetts. I am currently Chair of

the American College of Radiology Imaging Network's

Patient Advocacy Committee. I served for two years

on NCI's Central IRB, and also was Chair of NCI's

Director's Consumer Liaison Group for three years.

I was invited to speak today by Advanced

Magnetics, which is covering my expenses.

I don't know how many of you have been

personally diagnosed with cancer and understand,

not only what a frightening time it is or the

confusing one it is, as well. Living in the Boston

area, I was fortunate to have three world renowned

physicians to advise me, but unfortunately, they

did not agree on what my treatment should be.

I learned during this very difficult time

that in spite of all the progress which has been

made, treating cancer is often as much an art as it

is a science, because there is still so much that

we do not know.

Obviously, for each individual patient,

152

the goal is to gather as much information as

possible, so that the treatment can be tailored to

their particular cancer with the goal of neither

undertreating nor overtreating the patient, which

can lead to unnecessary side or late effects and

adversely affect the quality of life.

In my case, two of my physicians wanted me

to have a nodal dissection, but my surgeon thought

that because my primary tumor was so small, it was

non-high grade, and there was no lymphatic nor

vascular invasion, that the morbidity, which could

be caused by a nodal dissection, would outweigh any

information which might be gained by doing one.

I can't tell you how many agonizing hours

and days I spent going over, not only the

conflicting opinions, but the literature before

finally deciding against a nodal dissection.

Now, nine years later, it seems pretty

clear that the decision I made was the correct one.

I spent many years wondering and worrying if I had

made the right decision.

When I heard about the trial using MRI and

153

Combidex, I thought to myself how wonderful it

would have been to have been able to have such a

scan. While in my case, it would not have made a

difference in my treatment, it would have given me

enormous peace of mind.

Obviously, for many patients, it would

help determine, not only the extent of their

treatment, but the type of treatment that they

would have.

I suppose the question could be asked why

do we need a new way of determining nodal status

when we already have several, but I think there are

three reasons why we need one.

First, is that the current method of

determining that based simply on lymph node size

alone has an accuracy rate of only 68 percent,

while the stated accuracy rate for MRI and Combidex

is 85 percent.

Second, is that for many patients, a nodal

dissection requires a second incision, which can

sometimes leave the site numb for years with

prickling, tingling, pain, burning, and often

154

leaves the muscles weak.

The third is the risk of lymphedema, which

for breast cancer patients is about 15 percent of

those with a total nodal dissection and is severe

in 1 to 2 percent of those women. Women who have

had a nodal dissection for the rest of their lives

have to avoid anything which might cause lymphedema

to develop.

This means they must constantly remember

to avoid hot baths or showers, sunburns, harsh

soaps, insect bites, tight sleeves, or even playing

with a beloved cat or dog. More importantly, they

must avoid having their blood pressure tested or

receive any sort of injection or blood draw in the

arm on the side where they had their nodal

dissection.

A friend of mine was diagnosed with

cervical cancer in 2001. She was given a radical

hysterectomy and had 35 nodes removed, all of which

turned out to be negative. She didn't have any

problems at first, but then her left leg became

infected, which has led to chronic lymphedema.

155

Each day she must spend an hour with her

legs in the air, massaging them to try and get the

fluid out. Then, she must wear compression hose for

the rest of the day, and she must bandage her legs

every night before bed.

Flying is possible only if she can stand

and walk for most of the flight, and she must

constantly carry antibiotics with her in case of

infection. She used to wind surf and hike, but now

because of the risk of a scratch or poison ivy,

those and many other activities are no longer

possible.

Because of the medical insurance she has,

the physical therapy and the compression bandages

often have to be paid for out of pocket.

Because of my two years on NCI's Central

IRB, I understand the difficulty of balancing the

risks and benefits of new drugs while trying to

provide the best possible treatment to cancer

patients.

We talked this morning about the value of

physician education and technician education, and I

156

would suggest to you that equally important is

patient education. I feel very clearly that as long

as the risks are explained to a patient, they

should have the opportunity to have a new drug if

they feel that the potential benefits outweigh the

potential risks of doing so.

I understand that Combidex and MRI is not

risk-free, but I believe the risks to be

reasonable, and that for many patients, they are

clearly outweighed by the benefits of a new, more

accurate, non-invasive way of determining nodal

status.

Thank you.

MS. CLIFFORD: Thank you, Ms. Lestage.

Our next speaker is Mr. Mendinger.

MR. MENDINGER: My name is Larry

Mendinger. I am a home builder from Ashland,

Oregon, and Combidex paid for me to fly here,

however, I can tell you that is a negative

investment for me, because I am missing three days

of work.

I have prostate cancer. I was diagnosed,

157

oh, three years ago or something like that. It

should have been four or five, but my doctor didn't

happen to notice what my PSA was doing.

I went through some treatments, which

seemed to stave off the growth of the tumors, and

my PSA kept bouncing around for some time. In the

last eight or nine months, I have had--well, I

should say before I had Combidex, myself and my

insurance company probably spent $18,000 on

everything from ProstaScint to CT scans and PET

scans, and all that stuff.

It all showed, well, we really don't think

so, that you really have anything to worry about,

we can't seem to see it. So, when I finally--I was

feeling very uncomfortable and my PSA was going up

drastically, last summer my doctor heard about the

Combidex, and he sent me to Dr. Barentsz's place in

Nijmegen--did I say that right, Nijmegen, thank

you--beautiful place, and very enjoyable trip.

I had the Combidex and I sat down in my

shirt and kind of half-naked, but afterwards, and

looked at the scanner with the doctor, and there

158

was absolutely completely, black and white, exactly

what was wrong with me. I have to say I can't tell

you as a patient what that means when you actually

know what is going on in your own body,when all

these other people, with all this money spent,

can't tell you.

The other thing I want to say is I did not

go for a surgical procedure to begin with, because

my urologist said, well, you need to have surgery

right away when I first had my diagnosis, and I

went home and I downloaded--I finally found a

procedure diagrammed on Johns Hopkins University

website, and I looked at that and I said, you know,

I am not a surgeon, but that looks like brain

surgery to me, no thanks.

So, I have been looking for a way to

remain intact as a man, and this was really

important. You guys need to approve this.

MS. CLIFFORD: Thank you, Mr. Mendinger.

Our next speaker is Ann B'rells.

MS. B'RELLS: I am Ann B'rells from

Schenectady, New York, and I want to thank you for

159

having me. I have no financial interest in the

company and paid my own way to the meeting. The

only consideration I am taking is ground

transportation back and forth to the airport and

maybe lunch.

I come to this hearing as a breast cancer

survivor for three years. Three years ago, I was

diagnosed with breast cancer during a routine

mammography, and ultrasound proved it, an

aspiration revealed cancer, which was small and

fairly well defined, and I was told at that point

that a lumpectomy was in order.

In order to find out how the cancer was

spread, it was recommended that I have a sentinel

node biopsy also at the same time. No other way of

identifying the lymph nodes was suggested to me

because of the comments that you have heard earlier

today.

After the surgery, I was lucky and the

sentinel was clear of cancer. Unfortunately, so

was all the other material they had taken, and

ultrasound showed that they had missed the lump,

160

and they had to go back in and get it.

I was recommended for radiation and

Tamoxifen or Arimidex, and I chose Arimidex.

The reason I am here talking to you today

is that at the point of after the second surgery,

when I had to make a decision about treatment, it

was quite clear that there was no way, a

non-invasive way--and you have just heard all about

the problems of taking all the lymph

nodes--available to me even though there was a

several month delay between the operations.

There was no way I was going to have

general chemotherapy as opposed to Arimidex or

Tamoxifen because of the side effects and possible

mortality from that.

At this point, the only diagnostic tools I

have are the usual physical exams, mammograms,

breast ultrasounds, and uterine ultrasounds. I

have had a couple of scares as everybody has, and I

can't repeat often enough the emotional and other

physical effects from just the fear.

To be able to have known after the second

161

lumpectomy, to have had a test that my doctor would

have recommended, and I think that he would have

recommended this one, would have been wonderful.

I just want the committee to understand

that even though I was treated only three years

ago, that there are always complications that come

up, and that the ability to understand what is

going on with lymph nodes without actually taking

them out would be wonderful.

The second comment I have is that although

sentinel node removal is a much milder activity

than taking more of them, it still carries a small

risk, and that risk leads you to the same

preventative activities of only having one arm to

give blood, et cetera. So, that is another reason

that it would be wonderful if the sentinel, which

also misses, what is it, 15 percent of the active

cancers, could be eliminated.

So, I thank you very much for your

attention.

MS. CLIFFORD: Thank you for your

comments, Ms. B'Rells.

162

Our next speaker is Tom Brady.

DR. BRADY: I am Tom Brady. I am from

Boston. I am a Patriot, but I am not the

quarterback. It is a problem I have periodically.

I am here with no financial interactions

with the company. They have never supported my

research. I am actually the Director of Radiology

Research at the Massachusetts General Hospital and

Professor of Radiology at Harvard Medical School.

I came here for the first time in my life

to address the FDA, because I felt that this was

important enough to take a day off from work--I

appreciate the prior speaker saying you can't pay

for a day off of work--and come here to say a few

things.

The first is there is no perfect

pharmaceutical or contrast agent. The FDA, in its

wisdom, 40 or 50 years ago, did not approve a drug

called thalidomide, which saved thousands of lives

and deformities. That drug is currently I believe

approved for a number of applications around the

world including leprosy and other vascular

163

problems.

I was really impressed by the data that

was generated in Europe and at the MGH and

presented in the New England Journal of Medicine

primarily because, as a radiologist, there is

really no way to evaluate small lymph nodes,

whether they are benign or malignant.

2-deoxyglucose, which is a PET agent,

which was not commented on here today, is extremely

good especially for looking at larger lesions, but

the ability to identify with high accuracy disease

in small lymph nodes can significantly change the

management of patients.

I concur with the studies from Europe on

the cost efficacy. We will see more of those

studies from the MGH coming out soon, and we

believe that it will, in fact, demonstrate that at

a high degree of efficacy.

So, in summary, I thank the committee for

this opportunity. I don't want to take additional

time, but I think that this agent should be

approved. Thank you.

164

MS. CLIFFORD: Thank you, Dr. Brady.

DR. MARTINO: The Committee would like to

thank all the public speakers and all those of you

who are in the audience who perhaps would care to

speak, but have chosen not to do so. We do

appreciate your being here.

I will tell you as a clinician,

particularly those of you who are patients, and who

understand these things from a very personal

perspective, that the Committee welcomes your being

here and appreciates you putting things in a

certain perspective for us. So, please know that

we value your contribution.

We are now going to turn to the discussion

portion of the meeting, and this will end with

ultimately an actual vote that will be taken. So,

realize that the vote will be the last part of what

we are going to do this morning. You will have

opportunities to discuss this before we actually

request a vote of you.

Dr. Ownby, I have been told that you had

some burning question that I somehow ignored. If

165

it is still burning in your heart, I will allow you

to ask it before we proceed.

DR. OWNBY: It was answered previously.

DR. MARTINO: Thank you.

There are a series of questions which have

been provided to each of the committee members. We

are going to focus, however, on truly the very last

one, because I think the other three are somewhat

encompassed within the final question.

Before I do that, I just want to remind

this committee of what it is that is being sought

here today from the maker of this agent. It is an

indication for intravenous administration of this

agent in differentiating metastatic from

non-metastatic lymph nodes in patients with

confirmed primary cancer who are at risk for lymph

node metastases.

I do want you to recognize the nature of

those words. They are not asking for a particular

tumor, nor for any particular size of lymph node.

We have to deal with the question and the request

as they have posed it to us. Please keep that in

166

mind as you go through the next deliberations.

The question that the FDA wants us to

answer for them, and for those of you that are

guests to this committee, realize that this

committee is advisory to the FDA. We give them our

opinion. They then take that into consideration as

they make final decisions.

Most of the time I think they take us

quite seriously, however, so there is weight to

your thoughts and to your vote.

Question No. 5. Do the data demonstrate

that Combidex is safe and effective for marketing

approval based on the sponsor's proposed

indication?

If yes, are there post-marketing studies

you would recommend to them? If no, do the data

demonstrate that Combidex is safe and effective for

marketing approval for any other indications?

If yes, please describe the patient

population and clinical setting for which Combidex

would be indicated, and, if no indication is

supported by the current data, please recommend

167

what additional studies or data are needed.

It is on these questions and their nuances

that I would now like to invite you to give us your

thoughts.

As we did before, please raise your hand.

I will recognize you in turn.

Who wants to start? Dr. Brawley, you are

always a good one to get us going, so I think I am

going to turn to you.

Committee Discussion

DR. BRAWLEY: I guess I will start out. I

just wrote a couple of things while I was hearing

some of the public comment. My concern is the guy

with prostate cancer who is told that he has

positive nodes by the scan, but in reality, the

nodes are negative, and he does not get a radical

prostatectomy because the scan was wrong.

If you go through the mathematics that we

just had, and this incredible mathematical thing,

talking about epidemiologic terms, such as

sensitivity, specificity, positive predictive

value, negative predictive value, and another thing

168

called accuracy, which is a different kind of

accuracy from what the lay people talked about,

that is going to happen.

You are going to have a guy who has

negative nodes, who gets this test, and he is told

you have nodal positive prostate cancer. The guy

does not get a radical prostatectomy which could

save his life maybe, but it would be the end of

prostate cancer for this man if he got that

operation. I can guarantee you if we test 10,000

people, there is actually going to be a handful,

more than 20 or 30 men, maybe over 100, who will be

robbed of radical prostatectomy because of that.

The inverse, I am very worried about a

woman who gets this test for breast cancer and is

told you do not have node-positive breast cancer,

and, in reality, she does, and she ends up

relapsing and dying from her breast cancer in 5 or

6 years from now.

With the mathematics that was presented

here, I can guarantee you that is going to happen.

I just want to say that and I want to say,

169

yeah, we definitely need something that helps us to

discern node positivity from node negativity. I

like what I have seen here, but I think we need

like 10 times as many patients as we currently

have.

The next question I have for the FDA is am

I allowed to consider the New England Journal data,

which is not auditable and which the company has

not turned over to you?

DR. MARTINO: Could we get an answer from

the FDA on that?

DR. HOUN: It was submitted to the

application with the right of reference, however,

we have not been able to get any source documents,

so we do not consider it a study that would support

marketing.

You can give us your opinion of it, but it

does not meet Federal requirements for a study.

DR. BRAWLEY: Thank you.

DR. MARTINO: Dr. Smetherman.

DR. SMETHERMAN: With respect to breast

cancer, and Dr. Giuliano can probably speak to this

170

with even more authority than I, I think we have

kind of almost moved past this level with sentinel

lymph node.

We are not really looking at the sentinel

nodes with just H and E staining, we are looking at

them with immunohistochemistry. It is certainly

as, you know, the sponsor pointed out, 61 percent

of patients with a sentinel node only positive will

have additional positive nodes, but I don't think

they are suggesting that having this test, even if

it were negative, would obviate the need for them

to have the sentinel lymph node dissection anyway.

So, I think at least in what we are

commonly doing on a day-to-day basis in breast

imaging and breast surgery, this probably wouldn't

really be that relevant.

DR. MARTINO: Yes.

DR. BUKOWSKI: I listened to Dr. Brawley

and I must say I agree that the data we heard today

just there is not enough information on the various

patient groups to be convincing that, in prostate

cancer, for example, this will be a useful test in

171

terms of the auditable data that were reviewed.

I am concerned that let's say we approve

the application, as this material enters the use by

individuals, there will not be proof of a positive

node or a negative node, one will just accept the

radiologic view of that saying it is positive or

negative as we have heard.

The specificity and the accuracy doesn't

sound like that would be supported by what we have

heard today, so I am somewhat concerned by the

number of patients that were included in the small

subset. I just don't think there are enough

prostate cancer patients, for example, to support

utility in that particular setting.

DR. MARTINO: The problem that I have

really are many with this. Do I think that this

identifies certain lymph nodes that are not

appreciated in other ways? I think they have

convinced at least me that yep, that's true.

Is the value of this in people who have a

node that is greater than 10 mm, where others would

already have identified it? I am not sure that

172

that is the right place.

Is it really something that is of value in

someone whose lymph node measures less than that,

where other modalities might miss it? I am not

sure they have convinced me how good they are in

that setting.

So, I am actually very hopeful of this

modality. I have to say that it does have some

value in my mind. I am just struggling with am I

sure enough of what its value is, to what degree

can I trust the information that comes from it, and

in whom can I trust it, where does it do nothing

other than just confirm something I already knew,

where does it allow me to avoid doing a surgery,

where does it guide me to a lymph node that maybe I

should do a surgery on.

There are just so many questions that I

just, in my mind, cannot answer from the amount of

data that has been presented, yet, I am intrigued

that there is something here if only I could be

sure of what that something was.

So, I am struggling with this whole

173

concept as to how trustworthy is the data at this

point in time and how do I really use it

clinically, because ultimately, if it can't be used

clinically, in a manner that I understand, my guess

is that everyone else will have the same problem.

The charge that this committee has is not

just to sort of judge whether something is

interesting. Lots of things are interesting, lots

of things have some value. What this committee is

charged with is giving an opinion as to whether,

with the data that exists now, we are ready to

basically say anyone out there should have this

test available to them and the results of it should

be then used for clinical judgment.

I am struggling with that major leap of

faith, but I can't sort of lose track of what our

real job is here.

Dr. Amendola, you are up next.

DR. AMENDOLA: Let me tell you I am a

practicing radiologist with a special interest in

GU radiology. Prostate cancer, as you probably

know, is one of the most controversial cancers

174

regarding therapy today. One of the key reasons

for this is that we don't have a good method of

staging this tumor especially one of the problems

is staging a lymph node, which is a key element for

management of these patients.

There is another agent which is being

used, which is another imaging modality is PET

scanning, which happens to be not as good in the

pelvis as in other areas of the body because of

technical reasons.

I agree that the data that was presented

was not completely convincing from a statistical

standpoint, but I think that given the status of

our poor accuracy with the current imaging methods

that we have to image lymph nodes that are diseased

in patients with prostate cancer, I think that

taking the risk-benefit ratio, there is a group of

patients with prostate cancer would be highly

beneficial to use this modality.

If we could save some patients from

unnecessary surgery or radical radiation, I think

that this would be a very good thing to do. Thank

175

you.

DR. MARTINO: Dr. Levine.

DR. LEVINE: It seems to me that there is

real potential for the agent, and my problem is

that the indication that is being requested is not

really based upon the data that would allow me to

do that.

So, number one, the indication says all

tumors, all comers basically, but the presentation

is not dealing with all tumors, and somewhere in

your documentation it excludes lymphoma as an

example, but that is not stated on your indication,

so the indication is too broad based upon the data

presented.

Even the issue of newly diagnosed versus

status post-radiation, you know, I see that it has

been used in people who have had radiation before,

and maybe that is valuable, but I don't know, and

the indication doesn't state that or doesn't

qualify that, so that would be another area that

needs to be evaluated more carefully, studied more

carefully.

176

The other indication, it seems to me, is

in those tumors or those lymph nodes that are

small, less than 5 mm, and if you now break down

the data that exist into that group, you know,

whatever the specific cancer is, and the newly

diagnosed, and now less than 5 mm, there is so

little data here that it is very frustrating.

I guess I would ask you to think of that

and come back. Thank you.

DR. MARTINO: Dr. Hussain.

DR. HUSSAIN: So, I had the chance to hear

the scientific presentation previously by the

doctor from Mass. General who presented, and I

think he did an excellent presentation again here

today, and I guess in my mind, this technology is

quite potentially promising, but I would underline

potentially.

I don't believe the trials that have been

shared with us, and the results of them, and

certainly I think the designs were very flawed, I

have to tell you that. I came in with more

enthusiasm, and as I sat and listened more, my

177

enthusiasm went down.

I think the comments that were made about

sparing people surgery or added treatment is a

premature statement to be made. Staging gives you

information. You use that information to make a

decision.

I would point out that in today's

standard, if there is a microscopic lymph node

positivity, which is what you are talking about

here, there are patients who are being operated on

and offered additional therapy, so I don't think we

need to play on the angle that if you have one node

by the scan, that means you basically are to be

doomed to no treatment or some hormone treatment

that is not going to cure you. I think that is

really the wrong strategy here.

The one thing about this from my

perspective, I think what I would have liked to see

is a well-characterized patient population where

clinical and other predictors of outcome are

incorporated and how this thing actually played in.

The other thing that I would point out is

178

the way you are asking for it would apply for

people who have seen therapy and failed for

assessment, so a guy who has had a radical

prostatectomy or radiation therapy, and comes back

with a rising PSA, is also covered under this

umbrella, and I don't believe you showed us any

data to say that this would be a reasonable thing

to do.

I am not going to comment much about

breast cancer, but I think the same thing applies.

I would have loved to see a well done trial, a well

characterized population, all the information out

there, and the statistical assumptions to start

with, what you are looking for, what did you

expect, and I do reiterate that the template for

the lymph node dissection is to me--I am an

oncologist, not a surgeon--but it is important.

The questions that were asked from the

surgeon before were very, very relevant, I think,

and I think not having that information is a major

flaw.

DR. MARTINO: Dr. Couch.

179

DR. COUCH: I routinely read my own scans

before we decide in the tumor board what to do with

my patients, and I view this as another source of

information. I am going to scan all my patients to

decide what their stage is and what the best

treatment is.

The lack of evidence from that is always

astounding to all of us. This, to me, seems a

reasonably safe agent that could give you

potentially more information. Would it determine

alone what I do with our patients? No, it might

make you think you would do a further testing or

consider a node biopsy or a fine needle aspiration,

but it is more information.

It is important even when we have people

with disease to find out what the radiation reports

will be, whether they are at high risk and

therefore would qualify for chemotherapy. To me,

the data was supportive of use in head and neck

cancer patients.

These studies are difficult to do. I

think they have done a good job, to understand that

180

the radiologists went to the operating room and

looked at what nodal basins were being removed and

analyzed was very reassuring to me.

So, I actually think that this is quite

promising and the data, to me, is enough to approve

this broad indication. I am a little confused, and

I asked this of the FDA for the following reason.

We order tests and we understand it is part of the

treatment plan for the patient. It is not going to

determine alone what I do with the patient.

You are asking this company to say we will

approve this for a certain patient subtype that

hadn't had radiation and chemotherapy, and then you

want the company to come back again and again for

each sub-subcategory?

DR. MARTINO: The FDA needs to answer

that, please.

DR. HOUN: I think it depends on the drug

and the indication and the disease being studied.

A disease like ulcers, they get a treatment

indication for acute ulcers. If you want to say

you can maintain ulcer quiescence, you have to do

181

another study, a year-long study to demonstrate

that.

So, there are disease conditions where to

treat the level, the need for data on different

stages of the disease, prevention of ulcers is

totally different with NSAIDs. We do ask for

different studies, and for diagnostic agents, we do

have different types of indications.

They are seeking an indication for disease

detection. They are not seeking an indication for

patient management like to help you better stage.

If that was the case, then, we would compare

regular staging to this, and that would be the

clinical trial.

So, they are asking for disease detection,

and they are looking for cancer detection.

DR. COUCH: I think that is

extraordinarily difficult. For instance, I am glad

to see they excluded in their studies patients with

head and neck cancer that had had previous

radiation and chemotherapy. What happens to those

lymph nodes is unknown, and we are having trouble

182

even with PET scans, which we think is probably the

best imaging modality to understand which patients

have residual disease in their lymph nodes.

So, I think it is a little bit different,

and I think that that is decided upon with the

understanding of the specificity and sensitivity

and the clinical judgment.

DR. HOUN: If you give us advice that this

is good for primary presentation and that further

studies are needed for other presentations, we

would like to hear that, or if like all comers are

fine, we would like to hear that, as well.

DR. MARTINO: Dr. Mortimer.

DR. MORTIMER: As I think about

decisionmaking in this process, I think about

whether this test actually provides us information

that will make me change therapy, and I guess I

would reiterate what Dr. Hussain said in the

prostate cancer setting. Given the sensitivity of

PSA and the value of node dissection, I am not sure

that it actually fulfills that criterion.

However, I would like to make a plea that

183

the more interesting data really isn't the head and

neck population here who are underrepresented in

the advocacy group for a variety of obvious

reasons, and I think that data was actually the

most interesting.

DR. MARTINO: Dr. Ownby.

DR. OWNBY: I have two interrelated

concerns. One, as I understand the indication, and

the FDA experts can correct me, this would be

approved for all ages, and not a single group, and

that would include children, and yet there is no

child data in this.

My related concern is if you look at

anaphylaxis and anaphylactoid reactions in large

populations, young adults and teenagers seem to be

at particularly higher risk, and the only age

stratification of this data is the 65 and over, and

I would certainly like to see some further

stratification before considering that a very low

incidence procedure while clinicians are clearly

going to use it more in an advanced age population,

I think this is a very broad approval request.

184

DR. MARTINO: Dr. Reaman.

DR. REAMAN: I would just follow with

that. I think it is a very broad approval request

and I think some of the comments that I have heard

here, the real operative word in the review of this

is promising. I think this is probably one of the

most exciting agents we have had the opportunity to

actually review in this committee, but

unfortunately, the data presented to us was

probably some of the least satisfactory from the

standpoint of study design and conduct.

The FDA was criticized for making a low

blow because of the mix of patients and the broad

application. I would like to defend the FDA and I

think including 10 patients or 15 or 20 patients

with the three most common malignancies, and then

asking for a broad indication is really

inexcusable, whether it is in a primary diagnosis

setting or in a previously treated setting.

I am concerned that if this were to be

approved, that it would be used widely with no

experience in the previously treated setting, in

185

the setting of follow-up, and in the setting of

pediatric cancer.

Most patients or most children with cancer

are diagnosed with disseminated disease, and the

question of nodal metastasis is a very common

issue. I think the fact that this hasn't been

tested and the likely incidence of hypersensitivity

reactions is a major concern.

DR. MARTINO: Dr. Brawley.

DR. BRAWLEY: Thank you. Dr. Reaman, I so

agree with everything you just said. I really

wanted to vote for this drug today, however, it is

just not proven, it is just not proven with the

data in front of us.

I don't want to get into a lecture on

screening and epidemiology for the clinicians who

don't normally get involved in this, but I will say

for a diagnostic procedure, you typically want very

high specificity, 95 percent or higher. This is

specificity which is much lower than that, and that

lower specificity means that the decisions that you

make, you really have very little confidence in the

186

decisions that you are going to be making with that

low specificity.

Now, one way that you can increase

specificity is to enrich the patient population

that actually has the disease, to use other

clinical indicators, such that you are only using

the test on people who are very highly likely to

have the disease.

They tried to do that, and it is very

fair. That is what I would call a fair selection

bias. It would be not appropriate to do this test

on somebody who you didn't know have cancer

already, for example.

So, using clinical methods to increase or

enrich the odds that you are going to find disease

is totally fair, but even when they did that, the

specificity is less than 90 percent in most

instances.

I would concur that where we actually do

have the best evidence of efficacy--and I actually

split out efficacy versus effectiveness, they are

different things--is in head and neck cancer.

187

DR. MARTINO: Dr. Bradley.

DR. BRADLEY: I am a practicing

radiologist that has been doing MRI for 26 years.

We often make decisions based on imperfect data,

any of the clinicians in the room know that. What

I have seen, I believe in. If it is approved for

one set of cancers, the clinicians in this room

will probably use it for all sets of cancers even

though it is not in the package insert. We do that

all the time.

But I would like to speak to the

specificity and specifically to reducing the false

negative. If they get a false positive, they get a

biopsy. If they get a false negative, they die of

their cancer.

There are technologies coming down the

pike, in fact, many of them are on their way right

now that are definitely going to increase the

specificity of this agent. This is a magnetic

susceptibility agent, turns things dark in a higher

magnetic field.

I spoke earlier about why did you include

188

a 0.2T, well, the major market right now is 3

Tessler. Standard high field has been 1.5. We

just ordered eight 3T's at UCSD. There is a larger

market for 3 Tessler MR than there is for all of

the low-field magnets now. 3 Tessler will be much

more sensitive than 1.5 Tessler using the same

technique for the same concentration. So, I would

imagine the false negatives would reduce on that

basis.

I also mentioned, I asked the authors of

the New England Journal article how did they get

exactly the same slice thickness. Well, there is a

technology that is available in the brain called

auto-align which gives you exactly the same

position in the brain.

When I spoke to the inventor of that

technique, could it be applied to the body, he said

yes, it hasn't been yet, but it could be. So, now

you have got exactly the same node pre and post, in

exactly the same position, on a higher field

scanner, using more sensitive techniques, I am sure

that the false negative rate will be reduced and

189

the specificity will increase.

DR. BRAWLEY: Can I just say if I were

presented that data, I would happily vote for this

drug to be approved, but I haven't been presented

with that data.

DR. MARTINO: Dr. Perry.

DR. PERRY: First, let me apologize to Dr.

Li. I think I misinterpreted the sponsor's

indication and I apologize to you and the FDA if my

comments offended anyone. It was certainly not

intended. I intend to rouse some rabble, but not

unnecessarily.

Secondly, I am impressed by the safety of

the agent. I don't have any particular concerns

about that. I don't think there is anything you

can inject into somebody intravenously that doesn't

have some problems, and I think with the

appropriate premedication and precautions, the drug

is safe.

I am not yet convinced that it's effective

and I am not yet convinced that it's effective for

all the tumor types that it would conceivably be

190

used for, and at the moment I am inclined not to

vote for it.

DR. MARTINO: Mr. Kazmierczak, please.

MR. KAZMIERCZAK: Yes. As I pointed out,

I am a patient consultant to the FDA for prostate

cancer. I was diagnosed back in '98 or '99, and at

the time I had an MRI, which was negative, and I am

not sure if I had had the Combidex-enhanced MRI

that it would have changed the fact that I had a

radical prostatectomy, I am not convinced it

wouldn't have showed a negative result. I am not

sure that the accuracy is such that it would have

changed the therapy that I eventually elected.

I do agree with some of my friends up here

that the more information you have on risk and

benefit, the better the patient feels about the

decisions that he makes. I found out a long time

ago that I don't let doctors make decisions for me

anymore, I try to work with them to make the

decision.

It turns out even after my radical, my

cancer was not confined to the prostate, it had

191

seeped into the seminal vesicle, and I am not sure

that Combidex would have found that out. So,

essentially, I have had a rising PSA, went on to

adjuvant treatment with radiation.

I still have a rising PSA, so I got myself

a Viadur implant, and I am not sure that any of

these therapies that I went through, and I probably

have a disease that really attacks me very badly,

so I am probably going to die of this disease, and

I am not sure there is anything available other

than one of these wonderful clinical trials for me

at this point.

That said, when I read this information, I

was really hoping I could vote for this, but the

more I thought about it, the more I wondered

whether or not it would have made any difference to

me in terms of the decisions that were made. That

is my perspective. Thank you.

DR. MARTINO: Dr. Reaman.

DR. REAMAN: It was answered, thanks.

DR. MARTINO: Please.

DR. DYKEWICZ: To address a few issues

192

about safety, I do agree that the safety of the

agent is within the realm of consideration for

standard clinical practice. It is a question, of

course, of always risk versus benefit.

The risk of this agent, I think is

probably not that significantly greater than

radiocontrast media, although it may be. We still

are looking at relatively low numbers of patients.

I would say looking at it from an allergy

perspective, although the radiocontrast media is

certainly a relevant analogous situation, iron

dextran may be a more direct analogous comparison,

and there, of course, the reaction rate is somewhat

higher than with radiocontrast media.

Unfortunately, if we look at strategies to

reduce iron dextran reactions, nothing has really

been held to large-scale trials. There are case

reports about medication pretreatment as used in

radiocontrast media to reduce the risk, but I am

not clear that that would necessarily enhance the

safety.

That being said, we do know from

193

radiocontrast media, which is analogous in the

sense that this is most likely a non-IGE-mediated

anaphylactoid reaction, we do know that medication

pretreatment can significantly reduce the reaction

rate.

Now, this also, though, gets at the point

about what the type of medication pretreatment

should be. When you look at radiocontrast

pretreatment regimens that have been used, the best

data for protection is where corticosteroids are

given well in advance of the administration of the

agent, for instance, a regimen that would give

steroids 13 hours before or 7 hours before or 1

hour before, and not just, if you will, on call to

the Radiology suite.

I am kind of really troubled by the fact

of looking at, if you will, the standard of care

for treatment of patients in Radiology

administration areas in terms of what is done to,

number one, pre-treat patients who may be at

increased risk, and, number two, how to treat it.

I am not sure if there is good recognition

194

out there that you need to give steroids well in

advance of administration in order to significantly

reduce the risk.

I am pretty sure that there is a lack of

awareness about when epinephrine should be

appropriately used. I think this is a situation

where evidence-based medicine and the standard of

care are not meeting.

We now know that, for instance, with

epinephrine, it should be given IM rather that

sub-Q to try to get more rapid administration, and

to summarize all these musings, if you will, I

would say that it would be reasonable to try to

reduce the risk of a reaction by using a medication

pretreatment regimen that has been demonstrated to

be effective in radiocontrast media.

Whether the company would be held to do

that as part of a label indication, I think depends

on whether you demonstrate good efficacy. I think

my sense about this is head and neck cancer has

been demonstrated to be a scenario in which this

agent would be of value, and if you are looking at

195

a risk-benefit assessment, you could make a case

for approving the drug.

But if we are looking at in general, the

broader area of oncology, and having a very general

label for all types of cancer, with an agent that

maybe has a significant reaction rate risk of 1 to

2 percent, I think that gives me real pause for

concern.

DR. MARTINO: Dr. Amendola, did you have a

question?

DR. AMENDOLA: Regarding the question of

the value of pre-medication, this is well

recommended in the literature that you can decrease

the rate of reactions by giving them the patient

pre-medication with steroids especially for

iodinated contrasts. I am not aware of any

literature regarding this type of contrast. I have

another comment.

There is currently an FDA-approved MR

contrast material which is very similar to this

one. It is called Feldex [ph], which is also an

ultra-small, USPI, it is called. To my knowledge,

196

we use this fairly often, and we have not had any

serious reaction and, to my knowledge, there has

been no deaths related to Feldex, but maybe some

other members of the panel have more experience

with this.

DR. MARTINO: Dr. Hussain.

DR. HUSSAIN: I have a question to the

FDA, Dr. Li, or any of the team. When the sponsor,

I believe, or yourself mentioned that they sat down

to talk to you about the design of the trial, what

was the advice, and what was the spirit in which

the trial was designed, was that designed for an

indication approval?

DR. HOUN: I think that it has been a

course over the years that we have worked with the

sponsor, and I do have to say that their attempt to

get the correlation between images and pathology,

as you can tell as Dr. Anzai described it, is very

difficult, and they did a very good attempt to try

to do that.

So, we did look at their proposals, we

provided comments. They revised according to our

197

comments. I think our goal was to ensure that

pathology was obtained for these nodes.

I think your comment and the committee's

comments what were other factors that might

influence the actual surgical field, and were they

well described, unless the sponsor has more

information, I don't believe we were discussing

those specific criteria.

DR. HUSSAIN: Did you tell them, for

example, that you needed to have that many patients

of that tumor because one of the critique regarding

their request for the indication, that this is a

blanket, and they did not address different tumor

types, was that discussed with them in advance,

that unless you come in with that number of

patients with that tumor type, I mean in all

fairness to them, if they listened to what you told

them, and now it's not fair to them because they

did exactly what they were told, and they come back

and now they are told that is not good enough.

I guess what I am trying to find out what

was the advice of the FDA in the first place.

198

DR. LI: I will try to give my answers.

The interaction has been going on for six years. A

lot of people give initial comment, may not be here

anymore, but when we look at the record, you look

at the original patient population is 181 and 162,

and I don't think the sponsor anticipated, as we

never anticipated, that so many patients was

dropped from the study, was not able to do a

primary analysis.

So, if all the patients was included, that

may provide some--I mean I couldn't speak right now

what the data might be, but will probably provide

more assurance for us. This is one thing I don't

think the sponsor realized at the design stage that

so many patients--we didn't realize that either.

Also, I just make a point that there is

another issue that both sides never realized is

that pre-contrast MR sensitivity and specificity is

a moving target. You see from a U.S. trial they

made it from primary analysis, but from European

trial, they never it. This is an issue that the

sponsor and us never realized at the beginning, but

199

what I want to say is Agency really showed the

maximum, at least showed the maximum flexibility,

say if you really stay on the original design, the

original design said you have to have two trials.

Each trial, your sensitivity post have to

beat pre. That is basically statistical design.

In the U.S. trial, they meet that design, they are

able to show improved sensitivity, but in the

European trial, they failed the sensitivity.

So, if you just take at face value, it's a

failed trial, however, when we realized the reason

they failed the trial, it is because they only

included large size in the European trial. That

makes the sensitivity so high, no way they can beat

it.

So, we say let's go back, let's come back,

look at more evidence, look at what's really the

clinical question whether we can take a look at

data to see whether it's clinical value.

So, that's why you see the analysis by

subgroup. That was not original plan, that's true.

The sponsor, what they said that's true, that's not

200

original plan, but when first primary analysis

failed by the face value, when we started looking,

by its group, by those things of value there, then,

you see that this group is seen by size, by tumor,

that makes people starting to realize wait a

minute, what assumption we are having right now and

whether we should approve it for broad indication.

That is why we come here, ask for your

advice, to guide us how to handle the situation

over here. I hope I answered your question.

DR. MARTINO: Dr. Bradley.

DR. BRADLEY: I have a couple of questions

related to our relative lack of data, particularly

for the wider indication. As a radiologist, I

assume that all lymph nodes filter lymph and, in

this case, Combidex, the same way.

Does anybody know that lymphoma, for

example, would invade a lymph node in a different

way? I know that they didn't get lymph node data,

but they said it was because they had trouble

getting pathology, so that is one question.

Another question for Gene, they missed

201

your seminal vesicles. Did you have an optimal MRI

with intrarectal coil and the whole bit?

MR. KAZMIERCZAK: I believe I did.

DR. BRADLEY: You would know.

MR. KAZMIERCZAK: I believe I did, but I

am like the statisticians, I am 85 to 90 percent

sure.

DR. MARTINO: At this point, are there any

final comments? Otherwise, I will bring the

question to a vote.

Seeing no other hands raised, i will now

call you to a vote, and we will start on my right

with Dr. Couch, and as you state your vote, which

is a yes or a no, I need you to state your name

first for the record, please.

The question is do the data demonstrate

that Combidex is safe and effective for marketing

approval based on the sponsor's proposed

indication.

DR. COUCH: Marion Couch. Yes.

DR. SMETHERMAN: Dana Smetherman. No.

DR. AMENDOLA: Marco Amendola. Yes.

202

DR. BRADLEY: Bill Bradley. Yes.

DR. GIULIANO: Armando Giuliano. No.

DR. DYKEWICZ: Dykewicz. No.

DR. OWNBY: Ownby. No.

DR. MORTIMER: Mortimer. No.

DR. PERRY: Michael Perry. No.

DR. HUSSAIN: Hussain. No.

DR. MARTINO: Martino. No.

DR. REAMAN: Reaman. No.

DR. RODRIGUEZ: Rodriguez. No.

DR. LEVINE: Levine. No.

MS. HAYLOCK: Haylock. No.

DR. DOROSHOW: Doroshow. Yes.

DR. BRAWLEY: Brawley. No.

DR. BUKOWSKI: Bukowski. No.

MR. KAZMIERCZAK: Kazmierczak. No.

DR. MARTINO: And our tally? There are 15

No's and 4 Yes's.

That is the end of our meeting. I thank

you all for participating. Are there any

additional questions from the FDA before I release

the group?

203

Okay. There are no questions for the FDA.

At this point, I will remind you that the next

meeting begins at exactly 12:45 in this room.

Thank you.

[Whereupon, at 11:58 a.m., the proceedings

were recessed, to be resumed at 12:45 p.m.]

204

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

[1:03 p.m.]

Call to Order and Introductions

DR. HUSSAIN: Ladies and gentlemen, if you

don't mind taking your seats, please. We are going

to try to start the afternoon session.

My name is Maha Hussain from the

University of Michigan. I want to welcome you all

to the afternoon session. The session will

specifically deal with potential alternative

endpoints to design trials for prostate cancer

specifically with the intent of expediting the

approval process of agents in this particular

disease.

We will start with the introductions. I

will begin with the FDA on my left. Dr. Williams.

DR. WILLIAMS: Grant Williams, FDA.

DR. KEEGAN: Patricia Keegan, FDA.

DR. SRIDHARA: Rajeshwari Sridhara, FDA.

DR. SHAMES: Dan Shames, FDA.

DR. BROSS: Peter Bross, FDA.

MR. MANN: Bhupinder Mann, FDA.

205

MR. KAZMIERCZAK: Eugene Kazmierczak,

Patient Consultant, Prostate Cancer.

DR. BUKOWSKI: Ron Bukowski, Cleveland

Clinic.

DR. BRAWLEY: Otis Brawley, Emory

University.

DR. DOROSHOW: Jim Doroshow, NCI.

MS. HAYLOCK: Pam Haylock, Consumer

Representative.

DR. LEVINE: Alexandra Levine, University

of Southern California.

DR. RODRIGUEZ: Maria Rodriguez, M.D.

Anderson Cancer Center.

DR. REAMAN: Gregory Reaman, George

Washington University.

DR. HUSSAIN: Again, Maha Hussain,

University of Michigan.

MS. CLIFFORD: Johanna Clifford, Executive

Secretary to the ODAC.

DR. MARTINO: Silvana Martino, Medical

Oncology, Cancer Institute Medical Group, Santa

Monica.

206

DR. PERRY: Michael Perry, Ellis Fischel

Cancer Institute, University of Missouri.

DR. MORTIMER: Joanne Mortimer, Moores

UCSD Cancer Center.

DR. GRILLO-LOPEZ: Antonio Grillo-Lopez.

I am a hematologist/oncologist, a five-year cancer

survivor this month, and the industry

representative.

DR. SCHER: Howard Scher, Memorial

Sloan-Kettering Cancer Center.

DR. D'AGOSTINO: Ralph D'Agostino, Boston

University.

DR. D'AMICO: Anthony D'Amico, Dana Farber

Cancer Institute.

DR. McSHANE: Lisa McShane, NCI.

DR. SANDLER: Howard Sandler, Radiation

Oncology, University of Michigan.

DR. KLEIN: Eric Klein, Cleveland Clinic.

DR. DeGRUTTOLA: Victor DeGruttola,

Harvard School of Public Health.

DR. ANDRIOLE: Jerry Andriole, Washington

University in St. Louis.

207

DR. EISENBERGER: Mario Eisenberger,

Medical Oncology, Johns Hopkins.

DR. RAGHAVAN: Derek Raghavan, Cleveland

Clinic, Taussig Cancer Center.

DR. PAZDUR: Richard Pazdur, FDA.

DR. HUSSAIN: I would like to introduce

Johanna Clifford to read the Conflict of Interest

statement.

Conflict of Interest Statement

MS. CLIFFORD: The following announcement

addresses the issue of conflict of interest with

respect to this meeting and is made part of the

record to preclude even the appearance of such.

Based on the agenda it has been determined

that the topics of today's meeting are issues of

broad applicability and there are no products being

approved.

Unlike issues before a committee in which

a particular product is discussed, issues of

broader applicability involve many industrial

sponsors and academic institutions.

208

All special government employees have been

screened for their financial interests as they may

apply to the general topics at hand to determine if

any conflict of interest existed. The Agency has

reviewed the agenda and all relevant financial

interests reported by the meeting participants.

The Food and Drug Administration has

granted matters waivers to the special government

employees participating in this meeting who require

a waiver under Title 18, United States Code Section

208.

A copy of the waiver statements may be

obtained by submitting a written request to the

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

of the Parklawn Building.

Because general topics may impact so many

entities, it is all not practical to recite all

potential conflicts of interest as they apply to

each member, consultant, and guest speaker.

The FDA acknowledges that there may be

potential conflicts of interest, but because of the

general nature of the discussions before the

209

committee, these potential conflicts are mitigated.

With respect to the FDA's invited industry

representative, we would like to disclose that Dr.

Antonio Grillo-Lopez is participating in this

meeting as an acting industry representative acting

on behalf of regulated industry. Dr. Grillo-Lopez

is employed by Neoplastic and Autoimmune Disease

Research.

In the event that the discussions involve

any other products or firms not already on the

agenda for which FDA participants have a financial

interest, the participant involvement and their

exclusion will be noted for the record.

With respect to all other participants, we

ask in the interest of fairness that they address

any current or previous financial involvement with

any firm whose products they may wish to comment

upon.

Thank you.

DR. HUSSAIN: Thank you, Johanna.

I would like to make a couple of comments

210

before Dr. Pazdur begins his presentation. We have

several invited experts to discuss different

aspects of PSA and other endpoints as it relates to

prostate cancer.

The intent with this afternoon's session

is not so much to come up with a specific endpoint

to address necessarily, but rather a task list of

what perhaps is needed to begin to develop

different endpoints for evaluation of drugs in an

expedited manner in prostate cancer, so I would

like us to, as much as possible, focus on that

issue, and not get hung up on little details that

may not be serving the purpose as a whole.

Without further delay, I would like to

introduce Dr. Richard Pazdur, the Director of

Division of Oncology Drug Products, as the first

speaker.

Opening Remarks

DR. PAZDUR: Thanks, Maha.

I just have some introductory comments to

go over the process that we are addressing here

today. As you know, part of one of the big

211

initiatives that the Division of Oncology and

Oncology, in general, at the FDA, has been to

perform a review of different endpoints diseases,

and these endpoints are the approval endpoints that

we would use for approval of new molecular entities

and also supplemental NDAs.

This is a process that we have tried to

integrate into the greater oncology community and,

hence, before we have ODAC meetings on specific

diseases to discuss endpoints, we usually have a

workshop, and we had a workshop on prostate cancer.

It has been almost about a year ago, I think, where

we had a workshop in Bethesda, held in conjunction

with ASCO and AECR.

The purpose of this meeting was really to

explore areas and controversies of endpoints, and I

think one of the most controversial area obviously

was the optimal use of PSA, how to use it, when to

use it, where to use it, and I think a lot of

discussion that we will have today will center on

the PSA issue.

As Maha stated, I think one of our goals

212

is not to achieve direct consensus here, but to

raise issues. If there is consensus, let us hear

it, and we will be happy to take it into

consideration.

One of the things that I was most

impressed about as far as attending the workshop

that we held was the controversial nature of PSA

and endpoint for prostate drug development, and I

think we should be aware that there can be

agreements or disagreements, but ultimately, at one

time or another we are going to have to come to

some decision on the use of biomarkers in prostate

cancer.

So, I am not trying to discourage

discussion on this or lack of consensus, but I

think we have to be realistic that there are many

controversies that exist today in the use of PSA.

If I take a look at the other diseases

that we have held workshops on, for example, colon

cancer and lung cancer, I would have to say that

this has been the most difficult area to review,

and the battling of different people and different

213

ideas regarding PSA has been one that has

necessitated the Division actually to have separate

meetings and separate discussions with people in

the academic community after the ODAC meeting.

Here again, I think the one thing that was

clear is that this is a controversial area that we

need to bring some plans on how to further develop

this.

Hence, therefore, we have the program that

is outlined in your sheet that has been provided to

you. The first talk will be by Dr. Bhupinder Mann,

who will go over our past regulatory approvals for

drugs. That is what we have done.

Then, we have asked Dr. Derek Raghavan,

who was one of the Co-Chairs of the ASCO-AACR

meeting, to basically try to summarize the

highlights of that ASCO-AACR meeting, and his

talked is entitled "Towards a Consensus in

Measuring Outcomes in New Agents for Prostate

Cancer.

The third talk is one that will be given

by the NCI, and it is the NCI Prostate Cancer

214

Treatment Trial Portfolio. I was interested in

bringing this issue up because I think if we do

discuss endpoints, there has to be a discussion of

prospective evaluation of these endpoints, and we

have to have an idea what the NCI is doing as far

as supporting prostate cancer research and how we

can utilize those trials to embed endpoints and to

look at them in a prospective fashion. Hence, we

have asked the NCI to please provide data, not

data, but a description of ongoing research that

they have.

The last two talks, given by Dr. Howard

Scher and Anthony D'Amico, basically stemmed out of

our AACR and ASCO symposium, and is somewhat an

exploration of issues that were explored during

those workshops.

Howard will give a talk entitled "Toward

an Endpoint for Accelerated Approval for Clinical

Trials in Hormone Refractory Prostate Cancer," and

then Dr. D'Amico, Anthony D'Amico will end by

discussing clinical trial designs for selected

patients with a rising PSA following primary

215

therapy.

I want to emphasize we are not here to

bury PSA, we are not here to praise PSA. We are

here basically to have a discussion of the existing

data that supports its use. Any endpoint that we

have for drug approval has to be credible.

I am not asking for perfection here with

any endpoint, and as you realize, we have used many

endpoints that are not true surrogates. For

example, in our accelerated approval program, we

ask for surrogates that are reasonably likely to

predict clinical benefit, but we have to have some

basic comfort, some basic understanding of that

endpoint.

That endpoint has to have credibility and

some acceptance, not only by the FDA, but by the

greater oncology world, and that includes you

people as investigators and also patients.

With that ado, I will turn over the

program to Dr. Hussain.

DR. HUSSAIN: Thank you, Dr. Pazdur.

Dr. Temple, would you please for the

216

record state your name.

DR. TEMPLE: Robert Temple. I am the OD-I

Director, Office Director.

DR. HUSSAIN: Thank you.

Our next speaker is Dr. Bhupinder Mann,

who is a medical officer, Division of Oncology Drug

Products, and the discussion will be regarding a

regulatory perspective of endpoints to measure

safety and efficacy of drugs in the setting of

hormone refractory prostate cancer.

A Regulatory Perspective of Endpoints to Measure

Safety and Efficacy of Drugs:

Hormone Refractory Prostate Cancer

DR. MANN: Good afternoon. I am going to

first present a review of the endpoints which have

been used during the past few years in approval of

drugs for treatment of advanced hormone refractive

prostate cancer.

I have focused specifically on these

approvals as these are illustrative of the

underlying regulations. Later, I will also

summarize some of the difficulties which are

217

encountered in reliably measuring safety and

efficacy of treatments in prostate cancer, reviews

of both the traditional and the innovative

endpoints.

Approval of a new drug requires

substantial evidence of effectiveness derived from

adequate and well-controlled clinical

investigations.

Before 1992, endpoints used for drug

approval were required to represent clinical

benefit. Some of the endpoints were direct

measures of clinical benefit, for example,

improvement in survival or improvement of disease

symptoms. Others were accepted surrogates for

clinical benefit, for example, durable complete

responses in acute leukemia.

Since 1992, accelerated approval

regulations have allowed the use of surrogate

endpoints that are reasonably likely to predict

clinical benefit. Accelerated approval may be used

when a new drug would provide benefit over

available therapy. Accelerated approval also comes

218

with a requirement to do post-approval studies to

confirm that the drug does provide clinical

benefit.

During the last 10 years, three drugs were

approved for treatment of advanced hormone

refractory prostate cancer. Each of these three

approvals was based on clinical benefit endpoints.

None of these drugs was approved under the

accelerated approval regulations.

Most recent of these approvals, that of

docetaxel in 2004, was based on demonstration of

improvement in the overall survival. Overall

survival remains one of the most meaningful

endpoints in controlled clinical trials in cancer.

It reflects both the safety and efficacy of a

treatment.

It is an obvious direct measure of

efficacy and a longer overall survival, also

provides a reassuring measure of safety. A therapy

with significant toxicity and possible mortality of

its own is unlikely to result in a net survival

benefit.

219

Efficacy of docetaxel, in combination with

prednisone was demonstrated in a well-controlled

clinical trial by a significant prolongation in

overall survival.

In the pivotal trial TAX-327, two

different schedules of docetaxel administration, 3

weekly and weekly, were compared to a control arm

of mitoxantrone. Each of the 3 arms included

prednisone.

Cumulative dose of docetaxel

administration in the 2 study arms was the same at

750 mg/sqM.

1,006 patients are enrolled in this trial.

Primary efficacy endpoint was overall survival.

This was defined as time from randomization to

death from any cause.

Overall survival was significantly

superior to the docetaxel given every 3 week arm

compared to the control arm mitoxantrone, and the

results of every 3 week comparative arms are

summarized in this table.

Median overall survival was 18.9 months

220

for docetaxel and 16.5 months for mitoxantrone.

This was statistically significant.

FDA has accepted endpoints based on

measures of patient symptoms and other non-survival

indices of disease morbidity. Marketing approvals

for mitoxantrone and zoledronic acid were based on

non-survival endpoints.

Mitoxantrone, in combination with

prednisone, was approved in November of 1996. It

was approved for initial chemotherapy for treatment

of patients with pain-related to advanced hormone

refractory prostate cancer.

Its efficacy was shown in an open label,

Phase III controlled clinical trial, 161

symptomatic patients are enrolled. Endpoint used

was palliative response. This endpoint was

prospectively defined. It consisted of a 2-point

improvement on a 6-point pain intensity scale,

accompanied by a stable analgesic score and

duration of improvement lasting at least 6 weeks.

A palliative response was seen in 29

percent of the patients who received mitoxantrone

221

compared to 12 percent in the control arm. Median

duration of this palliative response was longer for

mitoxantrone at 229 days compared to 53. Median

time to disease progression was significantly

longer, 301 days compared to 133 days, however,

this trial of 161 patients did not demonstrate a

statistically significant difference in survival

between the two arms. PSA decrease of 75 percent

or greater was seen in a significantly high number

of patients in the mitoxantrone arm.

Zoledronic acid is a bisphosphonate. In

2003, it was approved for treatment of patients

with progressive bone metastases from prostate

cancer.

Endpoints used in that trial was a

composite endpoint of several skeletal-related

events. A composite endpoint can be useful when

disease manifestations are diverse. It can

increase the power of a study.

Previously, this endpoint had been used to

measure efficacy of pamidronate for lytic bone

disease in multiple myeloma and breast cancer.

222

Several prospectively defined skeletal

events were included in this composite endpoint:

pathological bone fractures, spinal cord

compression, and surgery or radiation therapy to

bones to treat a fracture, to stabilize an

impending fracture, to prevent or treat a spinal

cord compression, or for pain relief.

A change in the antineoplastic therapy due

to increased pain was an added event specifically

for this prostate cancer trial.

Efficacy was demonstrated in a

placebo-controlled, double-blind trial of 643

patients. There was an 11 person absolute decrease

in the proportion of the patients with at least 1

SRE favoring zoledronic acid. Another measure of

efficacy was an increase in the median time to

first skeletally-related event. This was 321 days

for the control and this had not been reached for

the zoledronic acid arm.

Now, I will briefly present the

difficulties encountered in evaluating treatments

of prostate cancer. These issues will be covered in

223

depth by Dr. Raghavan.

In general, difficulties encountered in

evaluating treatments of prostate cancer stem from

several factors. These relate to the

characteristics of the disease itself,

characteristics of the patient population, and the

prevalent clinical practices.

One disease factor that makes it difficult

to evaluate treatment is the heterogeneous natural

history of both the advanced and the early stage

prostate cancer. Disease course is highly variable

with diverse clinical manifestations.

At least until recently, that is to say

until docetaxel approval, use of traditional

endpoints, for example, overall survival in

evaluation of treatment efficacy had been of very

limited utility.

In this disease, on one extreme in many

patients a rising PSA may be the only sign of the

advanced disease. These patients do not have any

disease-related symptoms, their bone scans are

negative, performance status and quality of life

224

are well preserved. Although the survival

experience of these patients can vary, the vast

majority have a relatively long survival.

On the other extreme, there are patients

who have rapidly progressive disease, they have

disease-related symptoms, their performance status

is affected by their disease, quality of life is

impaired, and their survival is shortened.

Clinical benefit of treatment is now well

established for these patients.

Two characteristics of the patient

population which make it difficult to evaluate

treatments for advanced prostate cancer are the

advanced patient age and comorbid conditions.

Whenever you measure survival, and a large number

of trial participants have a disease with a long

natural history, the observed results are

confounded by competing causes of mortality, and

interpretation of the observed results can become

difficult.

Thus, the advanced age of the vast

majority of the patients with prostate cancer and

225

comorbid conditions they may have at that age, they

can make conduct of a clinical trial and

interpretation of the results difficult.

These patient characteristics are often

cited as an explanation for the inability to show

clinical benefit in terms of overall survival

prolongation both in prostate cancer and other

advanced cancers.

However, in 2004, investigators were able

to show overall survival advantage for docetaxel in

two different clinical trials even though they used

slightly different regimens. One can argue that it

was the lack of drugs with enough activity that it

was difficult to demonstrate clinical benefit in

terms of an improvement in overall survival.

Finally, prevalent clinical practices are

a factor which lead to difficulties in evaluation

of treatments for prostate cancer. Currently, in

clinical practice, as well as during the conduct of

clinical trials, treatment changes are frequently

driven by changes in the PSA level, thus, many

patients can go off study before any clinical

226

endpoint of disease progression is reached.

Subsequently, data on other endpoints of

interest may not be collected at all. Collection

of such data is necessary to eventually define

clinical benefit from a treatment as well as to

confirm the validity of a surrogate endpoint.

PSA-based endpoint may be acceptable

surrogates for anti-tumor activity of a drug, for

example, in a Phase II clinical trial. However,

reliable use of PSA-based endpoints as surrogates

for clinical benefit in Phase III controlled

clinical trials when two treatments are being

compared, it remains to be defined.

This needs to be explored further. A

surrogate endpoint that is reasonably likely to

predict a clinical benefit can be the basis of an

accelerated approval. However, the new drug should

provide an advantage over available therapy, and

the clinical benefit needs to be confirmed

subsequently.

Thanks for your attention and I would like

to acknowledge the contribution of all these

227

individuals for this.

DR. HUSSAIN: Thank you, Dr. Mann. We

will hold all questions until the final speaker,

Dr. D'Amico, and then we will open up the floor for

discussion.

Our next speaker is Dr. Derek Raghavan,

who is Chairman of the Department of Hematology and

Medical Oncology at the Cleveland Clinic Taussig

Cancer Center. He will discuss Towards a Consensus

in Measuring Outcomes in New Agents for Prostate

Cancer.

Towards a Consensus in Measuring Outcomes in

New Agents for Prostate Cancer

DR. RAGHAVAN: It is always a pleasure to

follow Dr. Pazdur's introduction because, as you

know, the FDA is characterized by scholars of

Shakespeare, and I personally was relieved that he

chose to quote from Julius Caesar rather than from

Hamlet, because I thought he would have probably

gone for the cheap shot of asking the question to

pee or not to pee, but fortunately, he didn't.

So, my task is to discuss some of the

228

complexities that came out of the meeting in which

our original plan was to try to achieve a consensus

about what should be the new era of evaluating

prostate cancer studies, and as Dr. Mann has said

very elegantly, there are a number of confounding

variables that make it difficult.

Probably the hardest thing is that

prostate cancer spans such a broad spectrum, and it

goes from a disease that unfortunately can kill

people in less than a year to a disease that can be

metastatic and which can co-exist in a patient for

more than 10 years, and the key is to try to

identify which variant of the disease one is

dealing with.

As Dr. Mann mentioned, there are the

additional confounding variables of the advanced

age of the patients and the many symptoms of aging

that go with them, and Dr. Eisenberger and I were

just commiserating with each other that having

worked together in this field for 30 years, we now

have most of the symptoms that our patients have

acquired, and it's a sad thing, and the point that

229

I make is to the best of my knowledge, both of us

have normal PSAs, and yet we have aches and pains

and sometimes failure to thrive and fatigue, so it

can be really quite difficult to identify the

specifics of the disease against the background of

a well older patient.

Then, there is the phenomenon of death

from competing risk, which happens in any of the

studies that are relatively long.

Howard Scher made I think an important

contribution and spent some time talking about this

at our series of meetings of what he terms the

states model and what many of us would simply

identify as the staging approach to prostate

cancer, and I think correctly Howard has made the

point that there are many different scenarios that

the FDA will need to address in quantifying drugs

that are presented here.

There is the sort of conventional testing

ground for new medications in prostate cancer, the

patient with advanced conventional metastatic

disease. When some of us started practicing the

230

management of prostate cancer, the typical patient

would have a large volume of disease with

narcotic-dependent pain, potentially pathological

fractures, and that has changed over a period of

time.

There is the question of whether one has

had effective hormonal therapy, and with some of

the newer regimens that are around, there are data

that suggest that some of the newer drugs don't as

effectively suppress hormones as some of the

standards of care, and there are compliance issues

and issues that relate to drug uptake.

Then, we are looking at earlier stage of

disease when we are faced with using more advanced

treatments. With the increasing microscope that is

focused on prostate cancer and pressure from the

community to deliver the goods, patients are

looking to find relapse at an earlier stage, and

physicians are being faced with the problem of

sometimes treating disease that they can only

measure biochemically, which certainly will have

changed the situation, and that leads us to the

231

phenomenon of stage migration, which I will return

to through this presentation - in brief, changes in

imaging, changes in the use of PSA and other

markers, and even a functional migration in that we

are now tending to use quality of life parameters

as another measure of outcome, so that the

goalposts in a way have widened.

Normally, when I steal Howard Scher's

slides, I apologize and act embarrassed, but my

role is, in fact, to summarize discussions that we

were involved in, so I steal these slides with

absolutely no apology at all. There are only two,

Howard, of yours, so I am not actually giving your

talk for you, although I will do it perhaps a

little more elegantly and with larger words.

So, I think an important point that Howard

has demonstrated here is the concept of a continuum

of disease from the initial prostatic evaluation

through to advanced disease, and the reality is

that a particular product can be used at multiple

points through the course of the disease, and thus

one would anticipate different types of outcome.

232

One has even the situation for the

asymptomatic patient where one may decide to do

nothing and simply watch the patient in the context

of a slowly evolving disease, and that brings in

the biggest problem.

At the moment, there is a vogue, an

affection for stable disease as a rediscovered

category. There are now a series of static drugs

where the claim is that these drugs somehow

influence the natural history of the disease by

making it more stable than it was before the drugs

are used, and that may be a very reasonable

concept, but it is a concept that is somewhat alien

to the standard practice of oncology, and what, in

fact, is a cause of concern is that there may be

the potential for misinterpretation of data when

one has the phenomenon of stage migration, such

that one is looking at stability of disease at an

earlier phase in the natural history of the

disease.

So, that is where the question of PSA as

an initial endpoint and quality of life measurement

233

will come in.

Just to remind us all of the scenario and

remembering that we have the world's expert on AIDS

here, Dr. Alexandra Levine, who shared prostate

cancer patients with me at USC, but just to remind

her of the history that antedated her involvement

into oncology, I would just like to remind you that

the Nobel Prize for Medicine was awarded to Charles

Huggins and Clarence Hodges, I think Hodges being

the only urologist ever to win a Nobel Prize, and

that was given for the demonstration of an ability

to suppress the growth of prostate cancer in dogs.

The models that we used in the pre-1960s

era were essentially much cruder, but were still a

good reflection of the disease as we know it today.

They just reflected a more advanced variant of the

disease going to Howard's states model, the more

advanced end game part of management.

Human studies at that time, as I

mentioned, were characterized by patients with

large tumor cell volumes and symptoms to go with

them. Unfortunately, at the time, although they

234

were the best available, the endpoints that were

measured were imprecise, they weren't structured

ways of measuring the degree of improvement of

pain.

We maybe correctly or maybe incorrectly

said pain, yes or no. There was the acid

phosphatase measurement, which was clearly an

imprecise one that correlated occasionally, usually

Monday, Wednesday, and Friday with disease

outcomes, but Tuesdays, Thursdays, and Saturdays

didn't, and Sunday was in the eyes of the Lord.

So, the reality of the situation was that

we had markers that were unreliable and didn't

correlate directly with tumor volume. So,

ultimately, the one quantifiable endpoint came to

be survival, and that stood the test of time.

Now, as physicians spent more time dealing

with patients with varying stages of prostate

cancer, they started to look for different

surrogates of outcome, and it was in that period

that the National Prostatic Cancer Project, one of

probably the most underappreciated useful entities

235

that we have had in the U.S., actually did a lot of

very important work, were able to start to model

the concept of the variability of the different

states, they just didn't call it that.

So, they identified the category of stable

disease within prostate cancer, identified that

there was a variant that evolved slowly, and then

set about trying to structure what constituted

stability and were there different levels of

stability and could you influence stability in a

meaningful way.

In other words, if a patient had absence

of progression for 6 months, was that less good

than absence of progression for 12 months, and the

logical answer to that would be sure, provided the

progression was being measured in an accurate way.

The whole situation became a little more

complex with the very, very important

identification by Ming Chu and his colleagues at

Roswell Park of the entity prostate specific

antigen, which has totally revolutionized the way

we think about prostate cancer. The truth of the

236

matter is that it has allowed us to start to look

at this disease in a subclinical way.

The problem is that this has had its own

complexities, and as we have used PSA more and

more, we have come to understand that there is the

phenomenon of release, so sometimes PSA going up is

good, and sometimes PSA going up is bad, and the

problem has been that with the passage of time, our

ability to quantify outcomes has been obfuscated by

a lack of understanding of this molecule and its

production.

Once again, in the 1970s to the 1990s, the

availability of PSA led to stage migration and

because it was being used for screening purposes,

resulted in functional terms in a much higher level

of awareness of the public of the entity of

prostate cancer which heretofore had not really

been a very well-known entity at all.

Bhupinder Mann has shown you this snapshot

of the approvals, and this is simply to remind us

of the parameters that we used for approval in the

past.

237

Now, currently, there are a number of

situations that bring pressure on all of us to try

to come up with the goods, and which certainly led

to some extent to the development of a series of

meetings to try to structure our assessment of

outcome of novel products, be they cytotoxic or

cytostatic.

There is question that the microscope is

on the community of patients and physicians who are

involved with prostate cancer. There is a

requirement for us to do better than we have done.

This is a common entity, it is being diagnosed more

frequently. It may even be developing into

epidemic proportions.

It is not absolutely clear whether that is

reservoir effect or a real finding, but what is

absolutely clear is that in contrast to the United

States, if we look at the Far East, in Singapore,

in Hong Kong, in China, there is clearly an

epidemic of prostate cancer and no one knows why.

It is clearly more than just doing PSA

screening. It may have to do with lifestyle and

238

diet, it could be a whole bunch of things, but it

is quite clear that prostate cancer incidence rates

are increasing rapidly, so we are going to be faced

worldwide with an epidemic of this disease and need

to be ready for it.

Currently, there is a new era of stage

migration. We now have the PET scan, which is being

rationalized as useful for the diagnosis and

management of advanced prostate cancer.

At the symposium that we held some months

ago, Dr. Steve Larson from Memorial Sloan-Kettering

gave a very erudite discussion of the new

strategies of quantifying response using

radionuclide bone scans, tomography, and the new

tools, so this is again allowing us to look at both

outcome migration and stage migration in a

completely different way.

As you have heard mentioned, as I am going

to talk about, and I am sure Howard will, as well,

there is a refinement in the understanding of PSA

response. So, at the present time, new endpoints

are being presented.

239

Clearly, there is an increased refinement

of measurement of quality of life, and I would like

to talk about that, because I, in fact, am not a

great believer at least in using the refinements of

quality of life measurement as an index of

acceptance today. I don't think we are ready for

that.

The issue of absence of progression for

some of the cytostatic agents I think is going to

be perhaps the most controversial item that we will

need to face today, and then the issue of having

PSA, prostate specific membrane antigen, PSMA, and

the whole concept of time-dependent fluxes.

There was a time when we simply said if it

goes down, that's good, now we are starting to look

at time points and trying to interpret what is a

significant time point - is a 50 percent reduction

at 3 months better than a 50 percent reduction at 2

months, and, if so, how much better and what does

it mean.

So, ultimately, we have a whole series of

different endpoints, and the key question I think

240

that we need to address today is should survival

still be regarded as the standard, and if it isn't

the ultimate test, because it is confounded by

death from other causes, because it may be

confounded by a series of salvage therapies, in

other words, a new drug may work for a time and

then depending on the pathway that the patient

follows, again going back to the state's model, you

may end up having different follow-on pathways of

treatment.

If survival isn't an ultimate test, and we

decide to bring in quality of life with some of the

surrogates, will they lead us to new treatments

that actually alter outcome.

The big concern about the screening debate

today has been we are not still sure after many

years of PSA screening, are we actually saving

lives or are we just moving the diagnostic point.

So, one of the things that I think is a

concept that most people who treat prostate cancer,

be they surgeons, radiation oncologists, medical

oncologists, or palliative care physicians,

241

whatever point in the disease, I think we would all

recognize that for the different states of the

disease, the aims are going to be different, and

this is essentially taken again from Howard's

presentation although it is not his slide, just

identifying the different aims and outcomes that

relate to each of the stages of the disease.

Clearly, the focus will change from when

there is very little disease, trying to stop it

from evolving into something that is

life-threatening versus when there is advance to

hormone refractory disease, then, actually feeling

that you are playing what might be an end game and

trying to prolong that for as long as possible.

So, clearly, acceptance of drug X for the

patient who has PSA-only disease with no bone scan,

no physical findings, no symptoms, the nature of

what will influence the acceptance of that entity,

the force must be different from what will

influence the acceptance of an entity on your

right, in other words, advanced hormone refractory

disease that is symptomatic and which has the

242

potential to kill a patient in three to six months.

Again, for those of you who aren't

familiar with prostate cancer, this gives you a

sense of what a protein disease it is, and even

today, in a high end clinical practice such as at

Memorial Sloan-Kettering or M.D. Anderson or the

Cleveland Clinic, or any of the places, Hopkins,

that have major prostate cancer programs, people

every day of the week will see patients who happen

to come in, not knowing about prostate cancer and

therefore having allowed the disease to get totally

out of control with a whole series of

constitutional features that can cause hemorrhage,

that can cause pruritus, it can cause weight loss,

it can cause symptoms related to the sites of

metastatic involvement, back down to the patient

who will come up after a radical prostatectomy or

radical radiotherapy with a PSA that has gone from

0.05 to 0.1, so it is very difficult for the FDA to

look at this, in my opinion, as a unit entity.

So, one of our tasks will be to try to

give advice to ODAC about how to structure the way

243

of presenting a framing reference.

This is some work from Don Newling from

the United Kingdom, and it is just illustrative of

just how big an impact stage of presentation can

have, so this was looking at a series of his Phase

II trials where he looked at simple parameters that

resulted in the presentation of patients, and as

you can see, the median time from progression to

death, for example, for just a PSA increase was

dramatically different from the time frame for a

patient who presented with a liver metastases.

Now, today, there is a new nuance that we

understand, and that is that many patients who

present with liver metastases don't actually have

classical adenocarcinoma of the prostate.

Today, if you summarize the folks around

this table who see prostate cancer and treat it,

and ask the question what do you think about when a

patient presents with liver metastases in

isolation, everyone will tell you I think about

neuroendocrine small cell variant carcinoma.

It may not be that, but it is almost

244

certain that many of the cases over the last 20

years, that have shown response in the liver with

prostate cancer, have probably been of that

variant. So, that is a novel entity and again

relates to a histological migration with time.

So, one of the things that we need to deal

with is the impact of earlier intervention and what

it does to survival curves, so, for example, we

have heard mention briefly of two pivotal studies

that were reported in the New England Journal of

Medicine earlier this year, one led by Dr.

Eisenberger, who is here, and one from the

Southwest Oncology Group and its friends, and the

principal investigator of that was Daniel Petrylak,

and I was involved in that publication myself.

So, this was a survival curve that was

yawned at by the press. They looked at the

figures. They said p value of 0.01, Taxotere

better than mitoxantrone, big deal, and if you look

at that survival curve, I think you have to accept

that this is not a home run. It was the first or

one of the first two trials that showed a survival

245

benefit for one drug over another, and that is

important.

It was very similar if we go back 15 years

in the history of breast cancer to the sort of

figures that we saw in advanced breast cancer, that

went on to help us develop therapeutic strategies

of adjuvant care.

In fact, this type of study has led to the

development, for example, of SWOG-9921, a

randomized trial that looks at hormones plus or

minus chemotherapy for patients with locally

advanced prostate cancer, but accepting that it's

an interesting paradigm, those curves are not very

impressive to look at.

Just note that the number of total cases

is 670 and keep that in mind.

Now, if you want to consider the surrogate

outcomes, that is way more attractive, and this

relates to the 50 percent PSA reduction that was

identified, and blind Freddie could identify the

difference on the left of docetaxel versus, on the

right, mitoxantrone, and that is the stuff that

246

headlines are made of.

So, we decided that we would do some

modeling in the Southwest Oncology Group, and this

is work that was done by Catherine Tangen [ph], who

is a lead biostatistician in the GU Committee,

Genitourinary Cancer Committee, and she did some

very interesting modeling where she looked at

survival by a surrogate, which was 50 percent PSA

reduction at 3 months, and that actually is quite

impressive.

If you had treatment A versus treatment B,

you would say home run, that's a really big

difference. So, here, what we are identifying is

that a surrogate outcome is actually reflective of

an important endpoint, and if we make it a little

more interesting, and we then put in the responses

broken down for the type of treatment, you will see

that again the key difference relates to surrogacy,

but here is the problem.

Let's go back for a minute and add the

numbers of risk, and what you will notice is that

the number is 520 patients would had serial PSA

247

values, 520 out of nearly 700 cases, so what that

means is that we have lost from the denominator a

large number of cases.

The reason that is important, if you go

back to here, is look at the number of deaths in

that study, and the number of deaths, while clearly

important, in absolute terms is not all that

dramatic.

So, the point that I wanted to make in

taking you through this circuitous argument is that

we need to be extraordinarily careful when we leap

to a new surrogate, if we don't set the framing

reference of did we lose patients by using that

surrogate and what happened to the patients that we

lost that might have influenced the outcome, in

that situation we need to be very careful before we

set new standards.

So, my plea today is that we shouldn't be

setting new standards. I think we should be

identifying endpoints that require further study

and that the FDA might be able to require in the

trials that are presented to them. I think the FDA

248

has the potential to influence medical history here

by making certain demands.

So, my view is we are not ready for prime

time changes to outcome, but I think we are ready

to look for new indicators of outcome.

Now, measurement of quality of life has

been particular popular, and unfortunately, somehow

one is cast in the role of being anti-patient if

one says that one doesn't like quality of life

measures as a finite indicator.

I hope that my clinical career hasn't

suggested that I am anti-patient, because I see

myself as a substantial patient advocate. I just

don't happen to think that this set of measurement

tools is ready for interpretation yet, and the

reason, I have summarized here. There is difficulty

of assessing response.

Within the stable category, we have a

widening of the goalposts and the problem is that

measures of quality of life, as I mentioned, as Dr.

Eisenberger and I creak through our coffee and

biscuits that he was kind enough to bring to my end

249

of the table, those measures of quality of life are

confounded by the age and intercurrent problems of

the patients.

They can relate to age, they can relate to

therapy, they can relate to a whole bunch of

things, and as we look at the data that are

available to us, there is clearly a dichotomy

between objective measurement, subjective

measurement, and whatever in that frame of

reference PSA constitutes, which is somewhere I

guess in the middle, but closer to objective. I

think the key problem is that the optimal

technology has not yet been defined.

Now, what is good is that we have, in

fact, begun to rationalize our approach to this.

So, again, I want to be very clear that I am not

opposed to developing the methodology. I just see

it as still work in progress.

These are some of the patient reporting

domains that will come up again and again in the

different quality of life assessment schemes, and I

am not going to read them, they are all provided in

250

the handouts that are available, but they relate to

different ways of looking at a patient and asking

the question how do you feel.

There are structured scores like the

McGill Melzack, which involves looking at present

pain intensity, and there it is an attempt to

mathematize the assessment of outcome. The problem

is it is not yet clear what is the best way of

using this tool.

It allows for a 2-point reduction, which

is the best type of reduction. Ideally, if you

have a 2-point reduction on the McGill Melzack

scale from 2 down to zero, that is a big win. But

what is the impact if you happen to have a tough

Anglo Saxon dockworker who has a high pain

threshold and claims only to have one level of pain

at the beginning and he goes to zero. Is that

somehow less important, and the answer is we don't

know.

What if there is no pain, but there are a

whole series of bone metastases that are present,

what is the impact of having no change in pain?

251

So, my point is simply that all the models that I

have summarized up there address that dichotomy

very poorly.

The problems include the impact of

baseline variables, as I have said, we don't really

know how to score them, we have got no good model

for dealing with missing data, in other words, the

patient who doesn't fill in one of these scales, is

he too sick to fill it in, is he so well that he

couldn't be bothered, is he not bright enough to,

is he too busy? In other words, we don't know how

to integrate that into our assessment of this

endpoint.

The statistical analysis is another

problem, do we look at absolutes, do we try to

construct an area under the curve for the number of

days spent in agony versus the number of days spent

doing wonderful things.

We don't have a good mechanism, we don't

even have a model, such as looking at receiver

operating characteristic curves, which we use

sometimes when we are not quite sure where a cut

252

point appears because we don't know how to define

the cut points properly.

So, our confounding variables, to add up

to that, is what the patient knows. We have

tremendously educated patients, so you can

have--and I am sure the physicians on the panel

have seen this--a patient who comes in and you say

how are you feeling, and he says great, and you

say, well, I am glad because I have some bad news,

your PSA just went up 50 points, and they walk out

feeling horrible.

It is not that there is anything foolish

in that, it's that knowledge of PSA is integrated

into the model, and so it confounds our ability to

assess it.

There are clearly differences in the way

different racial groups and different societal

groups address pain, death, dying, cancer, and our

models don't allow for those differences of

perception.

So, then the question is what does that

leave us with, and I thought I would use an

253

illustration which was the one that Dr. Mann

mentioned, of how mitoxantrone was approved.

Studies of mitoxantrone in the Phase II

fashion dated back to about 1982-83. I think in

Australia we did one of the very early ones where

the assessment of quality of life was whether you

could drink a beer. That didn't translate to the

USA, but it was a pretty good endpoint as far as I

was concerned.

More recently, Ian Tannock, who has one of

the leaders in assessment of quality of life with

the Canadians, did a randomized trial comparing

mitoxantrone plus prednisone versus prednisone

alone, a small number of patients, and the primary

endpoint was palliation with a secondary endpoint

being survival.

Now, this survival curve has been

interpreted universally to show that mitoxantrone

doesn't improve survival, and that is a fundamental

misunderstanding of the design of the study,

because in truth, this was a relatively small

study, the p value, in fact, reflected a trend in

254

favor of mitoxantrone for survival, and the key

issue was this study allowed crossover, and so it

allowed a patient who was on prednisone, if he

progressed, to cross over to mitoxantrone.

My interpretation of this study is

that--and it has influenced my practice heavily--is

there isn't necessarily a rush to run to

chemotherapy in a patient if you have a relatively

indolent pace of disease.

But what influenced the FDA, I think

correctly, was this chart, which was an attempt to

look at area under the curve for quality of life,

and what it showed is that despite the toxicity and

cost of mitoxantrone, the patients who received

mitoxantrone front line had a better quality of

life.

They did a series of other assessments

that related to cost economics and identified that

it was cheaper for the Canadian community, that

there were patients going back to paying taxes

sooner, they were spending less time dying in

hospital, so this was a drug that actually did

255

influence outcome.

Dowling and colleagues in the Annals of

Oncology, looked in a little more detail at the

whole issue of studying that trial, so this was a

retrospective analysis, and what it showed very

clearly was that while mitoxantrone had been quite

useful, palliative response did not predict for

survival, and there were major discontinuities

between quality of life measures, PSA response, and

ultimate survival.

So, that should make us very, very

cautious about interpreting or overinterpreting

data. I always think it is a good thing to suck up

to the Chair, so I did want to show this slide that

I stole without apology from Dr. Maha Hussain at a

previous time. Maha, thank you for providing the

slide. I have jazzed it up a little bit.

The point of this slide is to demonstrate

simply that the issue of variability of quality of

life is not an inherent characteristic of the agent

mitoxantrone. These are a series of drugs that

hold up a cell cycle in a fashion analogous to

256

Taxol or Taxotere, and what it simply shows, as you

look down at the columns on the right, is that

there is a variability of survival, there is a

variability of pain improvement, and there is a

variability of PSA response.

So, it simply says the Venn diagrams of

assessment of outcome do not overlap very well

irrespective of which agent is being used.

Dr. Eisenberger, at our symposium,

presented data from TAX-327, one of the two pivotal

trials that seems to have been responsible for the

approval of Taxotere for prostate cancer, and this

is important work.

I am showing this just to show that both

studies give us the same message. If we look at

the different indicators of outcome, again looking

at the denominator of cases for which data are

available, you can see that there is a really quite

dramatic heterogeneity of interpretation, and

depending on what you want to draw from this set of

data, you can draw pretty much whatever you wish.

I think there is a general consensus that

257

in each of the parameters, docetaxel won and that

the difference wasn't all that great.

This may be a good time to quote Benjamin

Disraeli, one of the former prime ministers of the

United Kingdom, who was quoted to say, "There are

lies, damn lies, and statistics," and that may well

relate to the way the confounding difficulties we

have in interpreting data from the prostate cancer

environment.

I think this is an important study to show

because it shows how the community can make serious

mistakes. Now, this was an important study

published by Tom Beer and his colleagues from the

University of Oregon, and they looked at the

combination of Taxotere and a vitamin D analogue,

and this hit the headlines in virtually every major

publication in the USA.

I was puzzled because this was a Phase I

study, and it was a Phase I study in which there

were indices that I have summarized there. They

identified the ability to achieve PSA response,

survival was not an endpoint, because the numbers

258

which are small, and it was a Phase I study, and

what was puzzling was that there was a complete

discontinuity between the different indicators of

patient-driven outcome, and yet the press heralded

this as a major breakthrough.

So, I think what it shows us is that we

have to be very careful in interpreting quality of

life data.

Finally, Tannock and his team have also

addressed in specific ways the impact of placebos

in oncology, and I think it is always good to

remind ourselves something that we know, but that

we sometimes forget in dealing with prostate

cancer, which is that the placebo effect can

certainly have an impact on quality on life.

It generally doesn't improve performance

status and it generally doesn't improve survival,

but it does alter quality of life, so that means

that we need to be looking at the quality of life

assessments very carefully and assessing them in

the context of the interpretation of the placebo

effect.

259

So, my take on patient reporting of

symptoms is that if we incorporate them into the

evaluation of new agents, it will lead to an

additional stage response migration. These will

one day be very useful tools in the assessment of

prostate cancer, but I think that the tools that

are extant at the moment are not ready for prime

time. We certainly need to be incorporating them

into our assessments, but they shouldn't be the

drivers of decisionmaking.

PSA response versus symptom response

versus toxicity lead to a disconnect, and that may

sometimes be because big trials don't allow for a

detailed structured assessment of what are the

factors causing that disconnect.

So, it leads me to feel that this area

should be regarded as work in progress by the FDA

in its formal and structured evaluations of new

products. Survival has been the standard. It is

my personal belief, supported by some data that are

not yet incontrovertible that time dependent PSA

kinetics will ultimately be a very useful surrogate

260

of outcome, but we will need to apply Howie Scher's

states model such that we acquire data for time

dependent PSA kinetics in a series of different

clinical contexts.

One size fits all simply won't work in

giving us a meaningful evaluation of new products

that come into the marketplace, and most

particularly those that are cytostatic in their

type.

We will still need to do well-powered,

large, carefully designed, structured randomized

trials, and those trials should require surrogates

to be evaluated including quality of life and

patient reporting, PSA response, PSA time dependent

kinetics, perhaps markers of bone turnover, and we

shouldn't throw out survival just because it may be

a confounded variable.

Ultimately, we haven't figured out an

optimal way of assessing the cytostatic drugs. We

spent a lot of time at the symposium discussing

those, and I am figuring Howard will probably talk

a little about that. I feel personally that this

261

is work in progress, so I have stayed away from

putting up the assessment of cytostatic drugs in a

structured fashion, because I think we are still

learning how to do that.

Ultimately, I think we are not ready for

definition of a new era, but I think the FDA is

very well positioned to demand certain things of

the companies and the agencies that produce new

medications to allow us to finally define what is

the new era in prostate cancer treatment.

Thank you.

DR. HUSSAIN: Thank you, Dr. Raghavan.

Our next speaker is Dr. Alison Martin from

the NCI. She will be addressing the NCI's

Portfolio of Prostate Cancer Treatment Trials.

NCI Prostate Cancer Treatment Trial Portfolio

DR. MARTIN: Good afternoon, Madam

Chairman, members of the panel, Dr. Pazdur. Thank

you for the invitation to present.

I was considering how to be useful to

these proceedings since many of the investigators

that my program, the Cancer Therapy Evaluation

262

program, funds through the cooperative groups are

here, have reported their trials in prostate

cancer, including with surrogate endpoints and

including today.

So, I decided to step back and focus on

our program as a capacity to further address the

questions that come up in these proceedings and at

other times. I have talked myself into the

possibility that our program is at a crossroads in

the sense that we have approved more concepts this

year for prostate cancer treatment than any other

year in the past decade, and that we have seen more

hypothesis generating for surrogates than at any

other time.

So, I would like to encourage us all to

think about how we can maximize the capacity across

all of these trials.

Currently, I think we are standing from a

position of strength and weaknesses. With regard

to some of the strengths, there are a number of

randomized treatment trials that are mature, which

provide us with well-defined cohorts, high quality

263

and long term follow-up and defined treatments.

We heard about some of them a year ago at

the PSA Workshop RTOG 92-02, which Dr. Sandler has

reported on, and we will hear again from Dr.

D'Amico, which looks at PSA doubling time as a

surrogate for survival in high-risk, early stage

patients.

You have already heard and will hear again

later this afternoon about the two trials, SWOG's

9916 and Aventis-sponsored TAX trial that led to

the approval for docetaxel, and by finally

identifying a treatment which had an impact on

survival in the randomized setting, it provided an

opportunity to look at surrogates across arms and

across trials.

Separate from the randomized trials, there

are significant longitudinal databases from certain

cancer centers with large cohorts and CaPSURE/CPDR.

There are limitations also. You have

heard from Dr. Raghavan quite eloquently about the

population issues and the heterogeneity, coupled

with stage and assay migration. There are also

264

design issues, which is that most of the randomized

trials weren't prospectively designed to ask a

surrogate question or consider the power associated

with that.

Furthermore, the schedules for the

collection of PSA or other intermediate markers,

such as bone, may not have been sufficiently

specific. Even if it were within one trial, it may

have differed across the trials.

There have been treatment issues limiting

us in our questions and answers. One, the fact

that some treatments, for instance, hormones can

interact with the surrogate of interest, or that

there has been a lack of effective treatments to

allow validation of the surrogate's association

with survival.

Now, I would like to move from separate

from limitations of individual databases. Once we

have identified a database that may be

contributory, there are difficulties we have

experienced in terms of analyzing those databases

in a timely fashion using the same surrogates of

265

interest.

It is rare that people turn over their

databases to someone else of whom it is a priority,

so it requires a collaboration which is sometimes

difficult to arrange and perhaps best thought of

prospectively.

Other trial design issues, whether we are

looking at data mining existing databases or

looking forward to how we should consider the

designs that are coming up this next year, or

whether we want to ask questions about PSA as a

prognostic factor, as an eligibility criterion to

make our cohort more homogeneous, or to choose a

high-risk cohort to allow us to arrive at an answer

sooner.

Do we want to use PSA as an outcome

measure, and, if so, which outcome? Do we want to

use it as an indicator itself of cure, for

instance, in an initially diagnosed patient treated

with a radical prostatectomy who either did not

nadir or has a return of the PSA to a certain

level, is that sufficient to tell us that the

266

physician prescribed treatment, which was intended

to be curative, had failed, or do we want to ask

whether even though they were not cured, we need to

know how this correlates to survival, does that

depend on the adjuvant or neoadjuvant treatment

given, and the risk classification.

Are we interested in PSA as it correlates

to some other measure of clinical relevance, such

as those listed on the slide, and then which PSA

parameter are we to use? Once we choose, what is

the magnitude of change that we think will be

relevant and the strength of association with the

outcome of real interest.

Potential opportunities in the near

future. We have approved 6 and there perhaps will

be a 7th later this year, treatment concepts, and

we expect actually they may open in the same year

that they are approved due to a number of new

processes, one, the collaboration with the FDA at

the time of concept approval, and also our

collaboration with investigators and the generation

of the protocol. Rather than holding our review to

267

the end of the process, we are integrated into the

process.

Of these approved concepts, they will be

accruing in each of the clinical states that the

previous speakers have mentioned. I have taken the

liberty of borrowing the clinical state slide from

both Drs. Raghavan and Scher, and inserted the

pending trial next to it for ease of reference.

The goals are as previously listed with

some of the goals added for the cohorts that have

clinical metastases, either non-castrate or

castrate.

Although I didn't list survival of

prostate cancer, specific mortality is a goal in

the first two boxes, and they weren't previously

either, it should be stated that, of course,

survival is important when any intervention is

given. It is just that there are also

comorbidities and competing causes of death and

nearer term outcomes that may be relevant also.

In localized disease, there will be a

trial with hormone therapy coupled with

268

docetaxel-based regimen in two cooperative groups.

In hormone-resistant rising PSA state, there will

be a vaccine trial, and as currently planned, while

survival will be collected as a secondary endpoint,

the primary endpoint is the incidence of clinical

metastasis.

The other trials are androgen deprivation

therapy with a backbone of docetaxel, and lastly,

in the population that showed the docetaxel had a

survival benefit, the addition of either

bevacizumab or Atrasentan.

In conclusion, these are some of the

strategies we have thought of and we would welcome

other comments and suggestions on how to maximize

the return from these trials.

Number one, of course, nesting a surrogate

question into the therapeutic trials. There are

probably still databases, well, I know there are

databases that could be mined for hypothesis

generation of surrogate endpoints, but at any rate,

can we prioritize the most important, if it's PSA

response by 50 percent at 3 months, so be it.

269

Are there others, how many can we

incorporate prospectively without suffering from

multiple comparisons, one, two, three?

Can we, while we are looking at surrogacy,

compare it to survival, but, as well, some other

intermediate endpoints of interest?

Separate from embedding a surrogate, can

we systematically create a comprehensive database

for subsequent interrogation, so that if our PSA

definitions change or we are interested in some

other question, we can interrogate the database

across trials, not just within trials?

To the extent possible, can we harmonize

the amount of PSA data collected prior to treatment

to look at new risk classifiers? Can we

standardize when they are collected, when bone

scans are collected, so that we know when there is

time to clinical metastases in a more rigorous way?

Do we want to know when a patient becomes

castrated, if they have been treated with hormones?

There will no doubt be, in the future,

more informative markers, although we may not know

270

exactly which ones they are now, and to the extent

possible, we would like to encourage specimen

banking depending on the stage of disease, blood or

tumor.

Lastly, two other partners that would be

helpful to our efforts right now are to

prospectively identify what industry trials are

relevant to the clinical states, and try to

harmonize our schedule of collection of data.

We are opening six or seven trials this

year. That certainly does not represent very many

in each clinical state, and we would like to work

with our industry partners and the FDA to encourage

industry.

Lastly, the Cancer Diagnosis Program has

an initiative PACCT, the Program for Assessment of

Clinical Cancer Tests. They have worked with

breast cancer field and color cancer to identify

new risk classifiers, and they have made a

commitment this year to convene a strategy working

group to further identify trial designs and

questions with PSAs and other markers.

271

With that, I will conclude. Thank you.

DR. HUSSAIN: Thank you, Dr. Alison.

Our next speaker is Dr. Howard Scher from

Memorial Sloan-Kettering. He will discuss similar

endpoints dealing with accelerated approval for

clinical trials in castration resistant/hormone

refractory prostate cancer.

Toward an Endpoint for Accelerated Approval

for Clinical Trials in Castration Resistant/

Hormone Refractory Prostate Cancer

DR. SCHER: Thank you very much.

I won't try to mimic Derek's accent, but I

will echo some of the same themes.

What I think is becoming apparent from the

previous presentations is that we are, in fact, in

a position to ask the questions which will allow us

to better understand different intermediate

endpoints, because for the first time, we are

actually conducting trials that are large enough

and enroll a sufficient number of patients to

address meaningful questions.

So, just briefly to summarize where we

272

have been in terms of outcomes assessment, we all

recognize that the manifestations of prostate

cancer are very, very difficult to assess. The

clinical realities are that PSA levels guide what

we do in clinical practice.

We are now faced with the challenges, PSA

response outcome or progression measure which is

reasonably likely to predict clinical benefit and

form the basis of an accelerated approval.

I would like to argue that these trials

can be designed, but before we can actually say

anything about the rule of PSA in outcome

assessment, we actually need to prospectively

design the trial, as you have heard from previous

speakers, in which the endpoint, based on the

marker, is embedded.

So, I would like to think a little bit

more in terms of the disconnect that has been

discussed earlier in terms of PSA response, symptom

assessment, and effects on survival.

All of these can be important clinical

endpoints, and if we start thinking about treatment

273

objectives across clinical states, we can really

divide them into two categories, which I will call

eliminate/relieve versus prevent or inhibit

progression.

If we start thinking about the patients

who have progressed post-hormonal therapy,

so-called castration resistant or hormonal

refractory state, we are really now dealing with

two discrete populations, and they represent

patients who have received hormonal therapy without

any evidence of clinical metastasis on physical

examination or on an imaging study, the so-called

rising PSA castrate state, and those patients who

first received hormonal therapy at the time of

objective detectable disease on an imaging study or

physical signs or symptoms of disease, which we

have called the clinical metastasis castrate group.

I will be focusing most of the discussions

on those patients who have overt metastasis at the

time hormonal therapy was initiated, although

certainly the discussions will hold for patients

with a rising PSA.

274

What we are dealing with again is the

battle or race between death from other causes,

which is inevitable, versus death from disease,

which is what we are trying to prevent.

So, if we think about patients really in

two categories, if there are manifestations

present, we will use those manifestations to assess

a response measure designed to either eliminate a

symptom, relieve or control it.

If it is not present, we can think in

terms of how do we prevent it from occurring in the

future, and here the risk assessment models are

very important in terms of how do we know that the

patient is likely to need therapy for a specific

event, and we have a very unique opportunity to

data mine some existing databases with regards to

eligibility for trials.

If we think about what the outcomes are as

you are sitting with the patient or explaining a

trial to your colleagues, you would like to be able

to say that what you have assessed is clinically

relevant and of tangible and concrete benefit, and

275

obviously, we will factor in the risk/reward ratio

before we think about therapy.

Looking back at the approved drugs, you

can see how this eliminate/relieve or prevent

objectives has been played out. The

bisphosphonates, radiopharmaceuticals,

chemotherapy, the original approval of mitoxantrone

and prednisone were based on response measures that

showed the elimination or relief of symptoms.

We can think of delaying symptoms or

change in therapy, skeletal-related events in terms

of a progression endpoint, and even death from

disease is a progression endpoint because you are

preventing death from cancer, but none of these

approvals were based on measure of tumor

progression, and none of them were based on a

post-therapy change in PSA.

So, we think back now in terms of

eliminate/relieve. We are thinking about the

manifestations of disease that are present, how we

relieve those manifestations, a response algorithm,

and figure out what they mean.

276

It was interesting looking across our own

series of patients at MSKCC treated with

chemotherapy versus the two recently reported

SWOG-9916 and TAX-327 in terms of the frequency of

the different manifestations of prostate cancer.

As you can see, the frequency of

measurable disease is on the order of 20 percent.

There is a component of patients with visceral

metastases. Arguably, these have a worst

prognosis. Many of the so-called nodal sites we

are looking at are actually very small, and one can

argue what their clinical significance is,

particularly when you are looking at the changes in

size.

The dominating theme in this patient

population is osseous metastasis and a rising PSA,

and symptoms are variably reported, and Dr.

Raghavan gave a very elegant discussion of the

issues surrounding quality of life, but about 35 to

40 percent of patients will have some symptoms

which are recorded as significant, but again the

dominating symptom complex that we are treating,

277

trying to relieve or prevent from recurring relate

to complications of bone disease.

What are the outcome measures? If you are

looking at disease in the primary site, there are

no defined criteria. For soft tissue disease, we

have been mandated to use RECIST, which has

problems because it relates only to your relatively

unique proportion of the symptoms of prostate

cancer. It does not address issues related to bone

metastasis or PSA.

For bone metastasis, there is no standard

criteria for response, and I will go through some

of the post-therapy PSA change metrics. In terms

of assessing quality of life, we always feel better

if there is pain relief, but we also like to see

what are corroborating domains, that is, the

patient was more mobile, more active, slept better,

less constipated because of analgesic uses, and we

have all pretty much agreed in the community, if

you will, that the group categorizations of CR, PR,

and stable disease are really of little value when

it relates to clinical trials.

278

So, thinking about the post-therapy PSA

endpoints, one can look at decline, no rise or

fall, undetectable, normalization. Some of these

are relatively infrequent occurrences unfortunately

with our available therapies, so most of our

reports have focused on either a decline by a fixed

degree, most reporting 50 percent, or more

recently, no rise or no fall at a fixed time point,

but whatever decision rule one is looking at in the

Phase II setting will vary depending on what type

of drug you are studying.

The differentiating agent, for example,

may make the PSA go up before it goes down. The

cytotoxic drug may arguably make the PSA go down.

Otherwise, it is likely to be ineffective, but

whatever response measure is used in most criteria,

the change that you see is required to be detected

over a period of time.

There was a consensus meeting in the late

1990s. A consensus was described for a PSA

response, which required a 50 percent decline from

baseline, and as you have seen here, in this

279

particular illustration, the decline was confirmed

on multiple occasions.

The reporting standard has become 50

percent or greater decline as a "PSA partial

response," which is confirmed by a second value

four weeks or more apart, but even within these

criteria, there was recognition that there are

other issues of relevance, as stated, different

endpoints can also be reported.

Time to PSA progression and index of the

durability of the response was of interest, and in

order to be considered in a response category,

there could be no evidence of clinical or

radiographic progression, again arguing that other

manifestations of disease must still be monitored.

Looking for associations of PSA decline

and survival, again, a 50 percent decline versus no

50 percent decline. These particular analyses were

done using a landmark method, that is, the patients

had to live a period of time before survival

distributions were analyzed, and these results were

analyzed on an independent data set, but in both

280

situations, where there was a 50 percent decline or

no 50 percent decline, no rise versus rise, again

at 12 weeks, there did appear to be a survival

benefit for the patients who achieved this

endpoint, and, as illustrated, several groups have

shown this.

More recently, other measures are being

considered, a variety of metrics. In this case, as

I am sure Dr. D'Amico will discuss further, the

ratio of the post- versus pre-therapy PSA slope,

but with the consistent theme that one sees that

regardless of the metric used, these trials are

reporting a difference in survival based on the

outcome measure.

So, clearly, we are at the point now where

the associations between a PSA decline have been

demonstrated. This makes sense. If you are

studying a cytotoxic drug, you kill cells, PSA

should go down.

This may not apply, as Dr. Raghavan

mentioned earlier, to non-cytotoxic agents or, for

example, a drug directed at a component of the

281

metastatic process, for example, an angiogenesis

inhibitor or a specific bone targeting agent that

may not necessarily kill cells.

But missing in all these analyses were

positive randomized trials to explore the surrogacy

questions.

I won't detail these trials, these have

been reported before, and we are all familiar with

them. Suffice as to say that in 2004, there were

two trials reported which did show a survival

benefit, which allowed the exploration of whether a

specific PSA outcome measure was associated with

survival.

Again using various criteria for

surrogacy, in this case the Prentice criteria, Dr.

Petrylak and his colleagues asked the question

whether achieving any PSA value--it could be a

single value--below 50 percent of baseline was

associated with survival, this performed in the

context of the SWOG-9916 trial.

Again, as shown earlier by Dr. Raghavan,

there was a first qualification required that there

282

be a survival benefit for therapy. Looking again

at the association between the 50 percent decline

and no 50 percent decline, a significant difference

of 50 percent improvement if one is looking at the

survival distributions.

When one accounts for the 50 percent

decline, the treatment effect disappears. So, this

would appear to satisfy the Prentice criteria, but

what has been misinterpreted is whether or not

these results can, in fact, be extrapolated to

other trials, and the answer is no, this would

apply only to this trial, and it may not

necessarily be applicable to other therapies.

But it did suggest at least for this

specific treatment that a 50 percent decline from

baseline could be used as a surrogate for survival,

but again, we do not have multiple trials in which

to address this particular question, and at this

point it could only be listed as a hypothesis.

So, TAX-327 was like was reported, showing

a similar PSA response rate as we discussed, and in

this particular trial, although the PSA response

283

rate as reported was identical, there was only a

survival benefit demonstrated for the Q3 week arm.

For patients who received weekly therapy, there was

no difference in overall survival. So, this raises

the question as how much survival is explained by a

post-therapy PSA change. In order to be a true

surrogate, you like all of the survival to be

explained by the post-therapy PSA change.

This led us to look at what is the

association between time dependent changes in PSA

and relative risk of death. This was again a

retrospective analysis of patients treated in Phase

II trials.

You see the risk of death for a very low

PSA appears to be higher than patients with a

moderate level PSA, as illustrated by the dip in

the curve, and as the PSA levels go up, associated

with much higher tumor burdens, the risk of death

increases, but the amount of survival that was

explained in this analysis was only about 17

percent, and as my statistical colleague, Dr.

Halabi reminds me repeatedly this is not enough to

284

base treatment decisions.

So, looking back at some of the other

trials that have been reported, it was of interest

in Dr. Crawford's presentation, looking at the

construct of a 50 percent decline in the context of

SWOG-9916, association with survival was about 22

percent.

Using a metric of change in PSA velocity,

16 percent, Dr. D'Amico's slope changed 22 percent,

similar to ours, so again there is a significant

amount of survival which does not appear to be

explained on the basis of PSA decline.

What about palliative response? Again, to

show how Dr. Raghavan and I are thinking in a

similar fashion, which is scary to some degree,

there has clearly been a disconnect between the

observation of a palliative response and a PSA

response.

This was the work of Dr. Tannock cited

earlier, of looking at mitoxantrone/prednisone

trial, which did lead to the approval of

mitoxantrone and prednisone, and to my view

285

established a very important principle that

systemic chemotherapy could provide palliation of

symptoms of the disease.

Looking at the PSA response rates, the

palliative response rates appear to be similar, but

in the proportion of patients who achieved, looking

at PSA response relative to palliative response,

only 60 percent of patients who achieved a

palliative response had a decline in PSA.

This was very dramatic in terms of the

prednisone arm where only one patient showed a

significant decline in PSA, although a proportion

did show a palliative response.

So, where does this leave us in terms of

PSA change and survival? Trial 9916 showed that

there was an association of PSA decline and the

treatment effect was eliminated when adjusting for

the intermediate, did not see the same effect in

both arms of the TAX-327 study. The Q3 week arm

was the only arm to show a survival difference.

Although we have used different metrics in

the construct, and looking at retrospective Phase

286

II data, and post-trial analyses of randomized

comparisons, the amount of survival that is

explained appears to be very similar, about 20

percent.

Does this make sense? Yes, it does make

sense, because if you think about what does PSA do

in terms of prostate cancer progression, it is

really not known. There has been some speculation

as to its modulation of growth factor effects, but

one could understand that PSA alone does not

necessarily drive a prostate cancer cell.

We still have to remember in terms of

clinical benefit that there is this association of

a PSA response and a palliative response, which

reminds us that we must continue to monitor the

other manifestations of the disease, and we all

know based on pathologic studies that not all cells

within a tumor in fact express PSA, so we may be

dealing with a component of clonal selections.

But a limitation of all of these analyses

is that they were retrospective and they were not

the results of prospectively designed trials

287

looking at a question around the marker.

So, maybe if we have so much difficulty

with response, maybe we should think about the

failure to progress or looking at a non-progression

endpoint.

If one considers the importance of

following patients using different measures, both

physical assessments, symptom assessment, PSA, and

imaging studies, perhaps we can start getting a

better index of whether or not we are changing the

disease particularly if we are enriching the

population that we are treating for high risk of a

clinical event.

If you are thinking exclusively about

overall progression of disease, you don't really

have to worry about surrogate, you have defined it

on a clinical endpoint, and it is really going to

be a measure that will be drug mechanism

independent depending on the question that you are

asking.

So, if we think about preventing

progression of disease, we do have criteria for

288

some of the manifestations. For measurable disease,

we do have RECIST. We do have a problem in that we

do not have scan criteria which have been

standardized to assess serial changes in bone scan.

We do know that in about 70 to 80 percent

of cases, however, that PSA elevations do precede

other measures of progression, so this may be

sufficient and certainly a point of discussion of

whether this type of endpoint could be considered

in the context of a prospective study.

For quality of life measures, again, there

are validated instruments. These are not 100

percent concordant with PSA, and death from disease

is clearly an endpoint that will not be debated.

There has likewise been as a result of

collaborations in the academic community,

standardization of reporting and definitions of

progressions that we accept.

This is an illustration from the JCO

publication in 1999 showing a definition of

progression by PSA, which includes a 25 percent

rise from the nadir as the time point, but again

289

keep in mind, as shown earlier, that we can see

benefits which are clinically significant or at

least lead to drug approval without an effective

PSA, which is clearly illustrated by the endpoint

used for the approval of zoledronic acid, which was

a reduction in skeletal-related events at 15 months

in a patient population at risk.

So, we have been asked to put up a bar,

and I have been debating with many people what this

bar actually means, because what we have been

challenged to do is to come up with a measure that

is reasonably likely to predict clinical benefit.

The regulations for accelerated approval

are very clear. They require substantial evidence

from well-controlled trials regarding a surrogate

endpoint. The problem that we have had in prostate

cancer clinical trials, too few studies, too little

participation by both patients, physicians, and

overall community at large in these studies.

Until recently, the trials were

underpowered and undersized. As shown by Dr.

Raghavan earlier, the response observed with

290

estramustine and vinblastine in the early 1990s was

not dissimilar to what we are seeing now, yet the

Phase III trials that were designed were not of

sufficient size to actually address the survival

question.

Although we have looked at various

associations between PSA outcome measures and

survival, these are all retrospective analyses.

They were not derived from trials prospectively

designed to test the value of the surrogate

measure.

So, as we look forward, we do have several

challenges. We have to balance the clinical

realities of practice, that treatment is rarely

continued if the PSA is going up, and this is one

of the problems I have in terms of slope

modulation.

The patients comes in with a graph, it is

going up, they are not happy. If the treatment is

going down, it is very hard to stop treatment.

That is reasonable, although in many cases, there

may be other measures suggesting that the treatment

291

is no longer working.

We have seen in clinical trials that there

are specific protocol-mandated definitions of

progression. That can lead to premature

discontinuation of a drug. This will relate

primarily to some of the definitions that have been

applied to the use of bone scanning agents.

One or two new lesions dictates

progression, and I will illustrate a couple of

situations where that, in fact, may not be the

case. What we really need is clear evidence of

progression before treatment is continued.

It is not as if we are withholding

tremendous options, so an approach, when I am

discussing treatment with a patient is trying to

really make sure it is either working or not

working before you abandon it, because you don't

necessarily know what will be next, and you don't

want to abandon a treatment that may, in fact, be

helping an individual.

So, here is an example of a patient.

Actually, this data was generated yesterday, so

292

it's contemporary. Here is a patient who is

progressing after previous microtubular targeting

therapy.

His PSA went up to the low 300s. The

date, which you may not be able to see, is early

October of 2004. His PSA after this next

chemotherapy has been going down. He is

asymptomatic, his pain is resolved. His bone scans

are stable.

He would not meet the criteria for a PSA

response, and arguably, this is a patient who is

benefiting, and even though he has shown a degree

of myelosuppression, he religiously comes in for

his treatment. So, this patient would be missed as

a responder or a patient who is benefiting from

therapy if we were stuck with a 50 percent decline.

Here is another illustration. If you look

at the patient's baseline bone scan on the upper

left, there are some lesions visible in the

skeleton and in the manubrium.

At the three-month scan, there were two

lesions that appeared, one in the rib and one in

293

the vertebra. By some protocol criteria, this

would be considered progression. The patient was

asymptomatic. His PSA kinetic curve is on the

right. You can see the PSA is going down.

Treatment was continued. A bone scan was

done a six months. It remained stable. Patient

remained asymptomatic and subsequently, there was

an improvement in these lesions.

So, this mandates very cautious

interpretation of bone scans, something we have to

consider as we are designing trials going forward.

So, what might a prospective trial look

like which is powered on survival, which has an

intermediate endpoint embedded, which might be

considered for interim approval?

The first question one might ask, and this

is an example of powering a trial on survival, does

Treatment A prolong life relative to Treatment B?

In the first line setting, this could be patients

with no prior chemotherapy, obviously, this would

be going against a standard of Taxotere, or in the

second line setting with one prior therapy, one

294

could power on trial on survival, for example, a 25

percent improvement, and secondary endpoints might

include a PSA response definition using the

consensus criteria, for example, a 50 percent

decline, a PSA progression criteria. Again, there

are consensus criteria for same, or a composite

endpoint that includes PSA.

It is obviously in yellow, which is where

my bias happens to be. One could certainly

consider an accelerated approval based on an

interim evaluation assuming the trial endpoint was

met, with the proviso that the trial accrual and

monitoring continue until accrual was complete, the

analysis complete, to assess the primary endpoint,

which in this case would be survival.

As mentioned earlier, it becomes critical

in these trials not to stop following patients at

the first sign of progression. They need to be

followed and monitored at fixed intervals after

treatment in order to better define the clinical

course if we are going to validate some of these

endpoints.

295

The CLGB has designed one such study, and

Dr. Halabi was kind enough to allow me to present

this. The PI will be Dr. Kelly at our institution.

They are studying whether the addition of an

anti-androgenesis agent Avastin will improve the

outcomes to standard first line chemotherapy.

The primary endpoint is looking for

prolongation of life. The secondary endpoint will

look at a progression-free survival endpoint

comparatively between the two regimens.

Eligibility is risk based, based on nomograms and

risk of mortality with stratifications based on a

nomogram that was developed by Dr. Halabi, and all

symptoms of disease and manifestations will be

recorded.

The primary endpoint is to look for a

reduction in the hazard ratio of death of 25

percent using a two-sided analysis, and they will

explore associations between progression-free

survival. This is not intended as an approval

study.

Another example might be in the second

296

line setting for a cytotoxic drug - does this new

cytotoxic drug (a) prolong life relative to

mitoxantrone and prednisone, for example, we can

discuss what the comparator might be, in patients

who have received one prior chemotherapy.

The secondary endpoint might be to compare

the PSA or overall progression-free survival of the

two regimens. Again, the trial would be powered on

survival, and consider PSA progression or a

composite that includes PSA for a potential for

accelerated improvement as enrollment on the trial

continues to reach the primary endpoint.

So, where we are now? Clearly, we still

recognize that this is not a straightforward

disease to manage. There are clear difficulties in

assessing response and outcomes. We must address

within our trials the clinical realities that PSA

levels and changes in those levels do drive

treatment, and the question remains for us to prove

prospectively whether there is a PSA response or

progression construct that can predict for true

clinical benefit and form the basis for an

297

accelerated approval.

But I clearly believe we are in a position

to do those trials, and there has been a

demonstrated commitment to complete the trials of

adequate size and power, so that we can actually

address these questions going forward.

Thank you very much.

DR. HUSSAIN: Thank you, Dr. Scher.

Our final speaker is Dr. Anthony D'Amico

from the Harvard Medical School, who will discuss

the design of clinical trials for select patients

with a rising PSA following primary therapy.

Design of Clinical Trials for Select Patients

With a Rising PSA Following Primary Therapy

DR. D'AMICO: While we get the screen up,

I want to thank Dr. Pazdur for letting me part of

this experience. Actually, it has been a wonderful

thing to put this set of data together, and it has

been a lot of fun.

I also want to thank Johanna Clifford and

Diane Spielman for all the logistical support that

you helped me with during the course of getting

298

here.

I never used to put humor into my talks

until I met Dr. Raghavan. There was a talk that he

was giving once at the course we have in Boston

every other year, and I was very impressed with his

delivery, in addition to the information that he

gave.

He said to me, though, today on the way in

that you are not supposed to have a joke prepared,

you are supposed to do it on the fly. I was

thinking to myself, well, maybe when I reach his

age, I will be able to do that, or maybe if I reach

his age, I will be able to do that.

What I would like to talk about here is a

very specific disease state, the rising PSA after

surgery and radiation in a very well-defined

population, people who have, in some people's data

sets, achieved "surrogate for cancer death," with a

very specific endpoint that involves the standard

endpoints - death due to prostate cancer and

metastatic disease predates that, and then also

consider some questions that we could raise about a

299

PSA construct.

Now, I am a believer that it is more

important that the information that you give is

concise and important more than it being excess

volume, so contrary to Dr. Raghavan's suggestion, I

am going to tell you a little thing I had planned,

because it sets the stage for this talk.

I enjoy martial arts, it is something I

have been doing since I am a child, and this is a

story that I heard once that I really found very

interesting.

There is a young gentleman who wants to

enter a Buddhist monastery, and he is told at the

age of 12, "Well, listen, you know, this is a

strict place, there is vows you have to take,

something called chastity, poverty, silence." He

says, "In fact you only get to speak two words

every five years." He says, "I want to do it." So

he goes into the monastery and does his first five

years, and when he comes out, okay, "You have got

two words, you're 17 now, what are they, and he

says, "Bed hard."

300

Okay, fine. Go back on in. Comes out at

the age of 22 after 10 years, he gets two more

words, and he says, "Food cold." They look at him,

sort of a seniors look back and forth and shake

their head. "We will give him one more try."

After the 15-year stint at the age of 27, he comes

out, and he says, "I quit." They said, "Fine, you

have done nothing but complain since you got here."

So, in terms of this particular construct,

I am going to start designing a clinical trial from

the first slide, and the first thing we need to

design in a clinical trial setting is patient

selection. Let me focus you again on the disease

state that we will be talking about, is the rising

PSA following surgery or following radiation.

In my mind, and there may be some dispute

about this, if one really wants to have a

"alternative" endpoint to the standard endpoints,

the place where it is needed most in my mind is the

earlier states of bad disease to come, and not the

endpoint of the disease where they have only got an

average 18 months to live.

301

I think that, you know, the TAX-327 and

the SWOG-9916 study from accrual to publication was

four years, which isn't bad. I mean that

survival-based studies in hormone refractory

metastatic disease are not unreasonable, but in

locally advanced prostate cancer, the bolus study,

the ORTC, the RTOG studies, 92-02, a study we ran

in localized high-risk prostate cancer, radiation

plus or minus hormones from start of accrual to

publication was 10 years.

This is where, if anything, we need help

in defining endpoints that are clinically

meaningful and earlier. So, with that said, we

have a huge amount of information, and as I go

through each of the centers or cooperative groups

that have contributed, I will recognize them.

There has really been a national effort

that has been designed at exploring PSA doubling

time following radiation or following surgery, and

I will take you through all the information that

has been published to date or soon to be presented,

and I have gotten permission from the investigators

302

in ASCO where some of this will be presented to

show some summary slides.

But what we have learned is that the

doubling time following radiation or surgery is

significantly associated with time to

cancer-specific death following the institution of

PSA failure on which the doubling time calculation

is based.

This data comes from a series of

multi-institutional and single institutional

studies, and I am going to highlight four of them

because each one has a unique characteristic.

The first one is RTOG 92-02 where patients

managed with radiation were randomized to short- or

long-term hormones. The next one is a

multi-institutional database, 44 institutions

around the country, CaPSURE, which is run through

Peter Kal [ph] on the West Coast, and CPDR, which

is run through Jud Mool [ph] and Dave McCloud here

at Walter Reed, and then two single institution

studies of importance, Johns Hopkins and Barnes

Jewish, Johns Hopkins because this was a place

303

where no one got hormonal therapy for a rising PSA

until the bone scan was positive.

That is a unique data set, which tells us

something about the natural history of a rising PSA

patient following failure after surgery; and then

the Barnes Jewish, Bill Catalona's database, which

I will show you some results from, and will be

published later in the year, from a group of men

who were prospectively screened.

Everybody had serial PSAs, so this is the

stage migration issue that Derek Raghavan was

talking about. We will look to see what doubling

time does in that particular group, and how

significant or lack of significance is it, so let's

go through it.

This is from Dr. Valacenti and Dr. Howard

Sandler. This is the schema for RTOG 92-02, and

this study has been published in the Journal of

Clinical Oncology, but what is soon to come is the

slide that follows.

The two arms are shown, locally advanced

prostate cancer T2c-T4, Pretreatment PSA is under

304

150, 1,500 patients or so randomized to radiation

with 4 months of hormonal therapy or 2 years and 4

months.

This is what they found. They applied the

full Prentice criteria to the model. I am not

going to present that and I will get to why later.

But the one point I am going to bring out is that

when they looked at PSA doubling time, and

specifically this is for a break point at 1 year,

they found a 6-fold increase in cancer-specific

death. The confidence interval is pretty tight, 4

to 9, as shown.

For this particular parameter, for the

first time really in a group of men managed with

radiation and hormonal therapy, it hasn't been done

before. It has been done for radiation, it has

been done for surgery. These are guys getting

radiation and short- or long-term hormonal therapy,

so this is new information.

There is the cancer-specific survival plot

or 1 minus the cumulative incidence of cancer

death, stratified by the doubling time, a

305

significant difference.

Of note, I will just point out that if you

look at the guys with the doubling times less than

12 months, which is the dotted curve at the bottom,

by the time you get out about 5 years, already

about 25 percent of these people have died of

prostate cancer, and it is a doubling time less

than 12 months. That story is going to evolve.

Here is the study that was written up by

CaPSURE and CPDR databases, that

multi-institutional database several years ago now

in JNCI, and what was shown here is that in a

select group of people with a doubling time less

than 3 months, which are the green and black curves

at the bottom for radiation and surgically managed

patients respectively, that the median survival was

only 6 years.

This stood in contradistinction to the

Pound's paper from Hopkins, which said the median

survival for guys with a rising PSA is 13 years,

until you put the things together and realize that

the Pound data incorporated everybody on the plot,

306

and we get to the very important point that this

makes, which is in the rising PSA cohort, all

patients are different, it's not one group.

In that one state of disease, you have got

a multitude of biology, from the very worst to the

very best, and the very worst can be characterized

by the short doubling times, the very best by the

longer ones, and there is the basis for patient

selection for a clinical trial.

Let's go to the rest of the data. The

hazard ratio from that data set was 20, a 20-fold

increased risk of cancer-specific death if your

doubling time was less than three months as opposed

to three months or more in contrast to the value of

6 when the doubling time break point was 12.

Now, here is an interesting slide that

hasn't been shown yet. This is Dr. Catalona's

screened database, all these men, 8,000 or so of

them have had a screened PSA each year. Their

median PSA of diagnosis is 4.2, so they are very

early, but the fascinating thing to me is that that

red and blue curve at the top, red is overall

307

death, Kaplan-Meier incidence of death, and blue is

cumulative incidence.

If your doubling time is less than 3

months, even though you were screened, you still

got a very high death rate, and the point I want to

make is if you look out five years, where the

numbers at risk are still reasonable, overall death

and cancer death are the same.

That goes right along with this doubling

time less than 3 months being very highly

correlated, if you will, a surrogate for cancer

death, even in a screened population.

In the population of them at longer

doubling times, cancer death and overall death are

about 50 percent of one another. You can see if

you work it out, half of death is due to other

causes, half of death is due to cancer in the

orange and the green curves below.

But the striking thing that I find here,

as you look at the numbers at risk at time zero,

the percent of patients who have a doubling time

less than 3 months in a screened cohort is 7

308

percent, and in the CaPSURE or CPDR, what happens

in the community where some people get screened and

some people don't, it is 20 percent.

So, it is very interesting to me that the

proportion of men with bad biology at the time of

recurrence in the screened group is much less as

you would probably expect than it is in a

relatively normal community population.

It gives us some estimate, here is 0.2, at

the size of the population we could enroll into a

study where the patient selection is based on

doubling time, and I will tell you what the study

is in a moment. Those are treated patients with

surgery.

Now, this is the slide from Dr. Parton,

Dr. Eisenberger, and Dr. Friedland at Johns

Hopkins, and this, too, is to come, but it is a

fascinating description in my mind where they have

broken out surgically managed patients, doubling

time now not as a categorical, but as a continuous

variable, and the adjusted hazard ratio of 0.86

with tight confidence intervals basically says that

309

as your doubling time goes up, your risk of cancer

death goes down by about 14 percent per unit

increase in doubling time.

But look at the plot here of

cancer-specific survival, the largest doubling time

is at the top, the less than 3 months at the

bottom. Again, you are getting that 5- or 6-year

median survival in that doubling time less than

3-month group. The same number we are seeing it

over and over again, multiple databases showing

that that doubling time less than 3 month group has

got about 5 or 6 years to go.

But it is nice to see that there is a

stratification in survival that goes from the worst

doubling times to the best or the longest

illustrated in this particular database.

The other thing that is interesting here

in this well-selected group of patients is they

have exactly 7 percent of men on this plot with a

doubling time less than 3 months exactly the same

as Bill Catalona's.

It sort of shows you that as you go from a

310

community database, where all comers come in to a

very select institutional database, the proportion

of the most unfavorable go down, but nonetheless,

to my mind, it is validation that that group does

poorly, whether they were screened and they end up

there, or that they weren't and they end up there.

So, in summary, in terms of patient

selection, what we have here are data from

cooperative groups, the RTOG, multi-institutional

databases, CaPSURE/CPDR, and Centers of Excellence

- Hopkins and the Barnes Jewish where the screening

studies were started, showing that doubling time is

significantly associated with cancer-specific

mortality.

I am staying away purposely from surrogate

for the following reason that I will now state. I

have discussed surrogacy with many different

statisticians. Dr. Rubin is the one who is closest

to me who runs statistics at Harvard University.

He has pointed out all of the issues, the

difference between a clinical surrogate and a

statistical surrogate.

311

I would submit that even if you run a

randomized study and you apply Prentice's criteria,

and you show it works, it still may not work

clinically, and the way that one would get around

that is by having multiple measures of surrogacy,

things like proportion of treatment effect

explained, the PTE model, and multiple studies all

showing the same thing, like I just showed for

doubling time, that would get us to the point where

we need to be.

So, I am staying away from surrogacy, I am

saying with associations or prognostic factors for

the time being, and the conclusion I would make

from the data I just showed you is that the

doubling time itself is significantly associated

with cancer death whether you have had surgery,

radiation, radiation and short-term hormones, or

radiation and long-term hormones, and that is just

about every treatment you can offer to a man who

presents upfront.

So, it covers all the treatment domains,

and doubling time less than 3-month group is a

312

particularly poor prognostic group and represents

about 20 percent of men who come from the community

where screening is not practiced necessarily, and

about 6 to 7 percent of men who come from a

screened group.

But the point I am going to make is it

doesn't matter how you get there, once you are

there, you do poorly whether you were screened or

not, because I think that that very short doubling

time is reflecting biologic behavior.

So, now we have identified some patients

for study. Now, we need some issues from clinical

practice and what has been done in this country to

decide what the arms of this study are going to be.

So, in the United States for patients with

a rising PSA, as Dr. Scher and everybody has said,

PSA dictates management, the rate of rise of PSA

has been shown to influence when hormonal therapy

is used.

Peter Carroll from the CaPSURE database

has shown this quite nicely the PSA doubling time

or velocity or how quickly the PSA rises is

313

directly associated with the timing of hormonal

therapy. The doctor looks at the PSA going up

quickly, the patient looks at it, something is

done, and in the community, that something is

hormonal therapy.

In academic centers, it can be anything

from vaccines to Celebrex, et cetera, on studies,

but in the community, which is where we are aiming

this, the big picture of what we do in this

country, it's hormonal therapy.

Then, a very important piece of

information from the Hopkins database where men

didn't get hormonal therapy until their bone scan

was positive. What is the median time to a

positive bone scan following PSA failure in a guy

with a very short doubling time - 18 months from

the one database that could actually measure it,

where hormonal therapy was withheld until the bone

scan was positive.

So, there is your next piece of

information, and that is what sort of drives

people's thinking in the community to start

314

hormonal therapy. So, the bottom line is that

patients with a doubling time less than 3 months

are offered hormonal therapy. Whether it has been

proven to improve survival or not is not the case

here, it is what is done, so I would submit that

that is a reasonable control arm.

So, here is the study. The treatment arms

would be hormonal therapy plus or minus some

systemic therapy in the setting of a doubling time

less than 3 months.

Now, what systemic therapy are we going to

choose, or even more importantly said, what class

of agents are we going to choose? This is where

the talk takes another twist.

I would say that Taxotere is the leading

contender because it is the drug that has been

shown to prolong survival in men with hormone

refractory metastatic disease, and the thinking is,

well, we will backstep it into earlier states and

maybe we will even see more of a benefit.

Maybe we won't see any at all, but that is

what studies are for. So, that would be my number

315

one choice would be docetaxel or Taxotere, but

there could be a number of other agents used, but

let's be careful here.

I would not recommend an agent that isn't

cytocidal for the reasoning I am about to go

through with the last part of the talk, which has a

whole host of data addressing this issue.

We don't know in the cytostatic agents, or

agents that modulate PSA, whether anything I am

about to say holds, but in the cytocidal, the ones

that kill cancer, as Dr. Scher was sort of alluding

to, you kill prostate cancer, the PSA tends to go

down in the hormone refractory state, well, that is

why I would stick to cytotoxics. I put Taxotere as

number one, there could be other agents, but I

think they have to be in that class.

So, now the last part which gets to the

endpoint of this clinical trial. So, you have a

rising PSA patient. You have given me hormonal

therapy plus or minus some new cytotoxic, Taxotere

or other.

The primary endpoint, the conventional one

316

would be time to bone metastases. It's the next

clinically relevant event that comes along the

path, and your secondary endpoints would be time to

cancer death and overall death, all-cause death.

I think that is your standard approach,

and no one I don't think would argue with that, and

it is very reasonable, and this study is being

done. Dr. Scher and I have been talking about it.

I think Dr. Scher has already got it underway. So,

this study is already happening or about to happen.

But PSA, and this is where I am going to

sort of focus my last part, you know, what is the

evidence, is there any evidence to suggest an

association between the nadir level of PSA--and I

use 0.2, more than 0.2 as a detectable level,

because that is a fairly good consensus across the

country--what is the relationship between someone

who goes on hormonal therapy, rising PSA, and

doesn't get below 0.2?

Is there a relationship between that

person and time to cancer-specific death in that

setting?

317

Now, I am going to show you a series of

studies that I will argue that there is a

significant relationship statistically, and then

the last point, clinically.

Here are the databases from which these

arguments will be made or evidence will be

presented. First, the last one I will show you is

the multi-institutional database, the CaPSURE or

CPDR contingent. I will start with the single

institution databases from New York. Peter

Scardino, Bianco, and Howard Scher actually worked

on this project. Then, I will show the Harvard and

Barnes Jewish single institution experience.

So, here is the New York experience. We

had 346 men who underwent surgery. Now, this is

interesting because not all of them are bone scan

negative at the time of entry, 81 percent. I will

address that later. The endpoint they used was

time to cancer-specific death, prostate cancer

specific mortality following 8 months of hormonal

therapy, very bright, because it takes at

least--the median is 3 months, which we found and

318

others, but it can take up to 8 months before your

PSA nadirs.

So, you set your time zero at 8 months

following the institution of hormonal therapy, so

you are not biased. Everybody has had a chance to

experience a nadir or not by that point. So, your

categorical variable or continuous, however you

want to look at it, continuous or categorical, has

happened by that point.

The covariates that they looked at in the

model was PSA level at the start of hormonal

therapy, the pre-hormonal therapy PSA doubling

time, the PSA nadir that actually occurred within 8

months of hormonal therapy, and then prostatectomy,

T-category Gleason score, and bone scan status

positive or negative, and the results are shown

here.

The PSA nadir level being undetectable was

very significant, as was the PSA level at the time

of hormonal therapy, and if they had a

pre-treatment PSA doubling time greater than 3

months, they did much better than if they had one

319

less or equal to 3.

The other factors, factors related to the

prostatectomy specimen, bone scan status didn't

matter. There were 63 cancer deaths out of the 360

or so patients, and the median survival for

patients who never nadired on hormonal therapy was

about 5 years, which is again consistent with that

6-year number I gave you before, you are just a

little bit further into the picture now, it's

short.

This is the data that they have, the slide

that Dr. Bianco sent me. That dotted line at the

top, this is cancer-specific death, the dotted line

at the top is the guys who never nadired and who

had a pre-treatment PSA doubling time less than 3

months.

Now, they didn't put numbers at risk on

here, but you have essentially got 100 percent

deaths in the first decade estimated, but if you go

out 5 years, you have got 80 percent of the people

gone estimated, okay, because it is always subject

to follow up, it's a pretty bad group.

320

Whereas, the people who died of disease,

if they did nadir, is the other dotted line where,

when you go out about 5 years, you have got about

15 percent deaths. So, there are still some people

dying even if they nadired, and I want to make an

important biological or clinical point here.

This is the twist in my mind. If you

nadir on hormonal therapy, it doesn't mean you

don't have hormone refractory disease, because

there are still some people who go on to die even

if you nadir on hormonal therapy, 20 percent at 5

years, and double that by the time you get out to

10 years.

But if you don't nadir on hormonal

therapy, you damn well have hormone refractory

disease because almost everybody is dead within the

first decade, and I think that is an important

point because it is saying it's like, you know,

when we biopsy the prostate, if the biopsy is

negative, it doesn't mean they don't have prostate

cancer, but if you find it, they do.

The same concept here. I think that the

321

nadir is an important construct because when it

doesn't happen, it is very bad; when it does

happen, it is not as bad, but it still can be bad.

Now, let's go on a little.

This is the data from Harvard and from

Bill Catalona, the Barnes Jewish group. This is

doubling time less than 3 months, did they get

below 0.2 or not, the same picture as Dr. Bianco,

Dr. Scardino, Dr. Scher's data set, same picture.

A lot of death if you didn't nadir, almost 100

percent in this case by 7 years, but still some,

but not nearly as much if you do nadir.

Then, going ahead, the final study. This

is the multi-institutional study from CaPSURE and

CPDR, which included 486 men who had surgery, 261

who had radiation. At the time of hormonal

therapy, everybody who had a bone scan which was

negative.

The endpoint here is the same endpoint

that the New York group used, time to

cancer-specific mortality following 8 months of

hormonal therapy. The covariates are all the same

322

covariates I just mentioned, and the results are

exactly the same with the only exception being that

Gleason 8 to 10 came in, but everything else in

terms of PSA nadir, pre-treatment PSA doubling

time, and the PSA level at the start of hormonal

therapy are all significant.

In this study, there were 53 deaths, a

little over half of them from prostate cancer, and

the hazard ratio adjusted for all of these factors,

when you didn't nadir, there was a 20-fold increase

in cancer death.

Now, let's look at the actual plots, the

graph first. This table is important from a

statistical standpoint and a power issue if you

were going to design a study in this group. I want

you to look at where the events occur.

If you look at doubling time less than 3

months, and you look at the column that says Number

of Patients, Number of Prostate Cancer Deaths, you

will see that 21 of the cancer deaths occurred in

the guys who didn't nadir and had a doubling time

less than 3 months; 3 occurred in guys who did

323

nadir and had a doubling time less than 3 months.

Now, you go across the table and you go

from 21 to 23 to 24, you pick up two more events

and then one more event, and at the bottom, 3 to 4

to 4, you pick up one more event. What I am saying

is that the vast majority of the 28 deaths are in

that upper left-hand corner box, the doubling time

less than 3, and the PSA nadir greater than 0.2.

The reason why this is important is that

if you take a trial and you select people with

doubling time less than 9 months or 6 months, you

will still see a difference, as I am about to show

you, but the difference will be dampened by the

fact that almost all of your events are occurring

in that enriched population with the shortest

doubling times.

My point is just that for a power purpose,

as I will show in the next three slides, the

selection should be very strict if you really want

to get an endpoint quickly.

So, here is the plot now, the one I have

been showing you all along from the New York group,

324

the Harvard, Barnes Jewish group, and here is now

the multi-institutional group. This is doubling

time less than 3 months, did they nadir or not, the

same story, same picture. If they don't nadir,

they do terribly, almost everybody is estimated to

die within 7 years. If they do nadir, some still

die, but not all, not nearly as much.

The number at the bottom, 68 over 224 just

tells you the percent of patients, which is 30

percent of men whose doubling time is less than 3

months, going on to hormonal therapy don't get

below 0.2, almost a third.

Now, that's in contradistinction to what

we think, we put people on hormonal therapy, the

PSAs go right down. Well, that is because most of

them are not doubling time less than 3 months

coming in. Most of them are 6 months or 9 months

or 12 months.

So, you will see as you go to the next set

of slides, here is doubling time less than 6

months, 25 percent of them don't go down to

undetectable levels, and the survival difference

325

here is still significant, but I want you to

remember that the only thing that is driving this

big difference is the group of men with a doubling

time less than 3 months.

Almost all of the events in this doubling

time less than 6 months are coming from that very

poor group. The same thing with doubling time less

than 9 months. Now, you have got 22 percent of

people who don't nadir, all being driven again by

that worst cohort.

I don't want you to get fooled here by

looking at these big differences in 6- and 9-month

plots. You have to know where the numbers are

really coming from.

So, we are almost done, 2 slides to go.

So, the summary of what I said. In a group of men

who come in with a rising PSA that is rapid, a

short doubling time, less than 3 months, a third of

them, 30 percent of them don't nadir at least in

this multi-institutional database and the other

ones I showed you, despite hormonal therapy, and in

my mind, given how quickly and how vastly they all

326

die of cancer when you look at those cumulative

incidence plots, they have to have some component

of hormone-resistant prostate cancer in them. I

can't imagine that they don't. So, that is Point

Now we come to the study hormones plus or

minus Taxotere, and the question is if a guy

doesn't nadir to less than 0.2 on hormonal therapy

and docetaxel, what does that say? We know that

docetaxel doesn't decrease testosterone levels.

That has been shown by William Ohe and others in

studies of neoadjuvant Taxotere Phase II studies

prior to surgery.

So, it doesn't go through that mechanism,

and when PSA does go down, it has been suggested

from the hormone refractory state that at least

there is some association with that in cancer

killing or cancer death. That led to a survival

benefit, but there was a disconnect between a PSA

reduction of 50 percent and survival. Why? Well,

perhaps you don't have the ability here of zero,

the nadir.

327

See, in the hormone refractory state, you

get down to 4 or 10, you are happy, but here, you

are going to either be undetectable or not. So, if

you don't go to undetectable levels on hormonal

therapy and docetaxel, I would submit--and this is

a hypothesis, but I think it's a darn good

one--that you are hormone resistant and you are

Taxotere resistant, and in my mind, that means you

are dead from prostate cancer because there is

nothing else that we know works, so I think that is

a good endpoint.

That is my opinion, but that is a

discussion that we can have. So, the trial that I

would then project to you is that if you nadir

above 0.2, 30 percent of the time on hormonal

therapy, and you could show that that goes down to

10 percent or less on hormones and docetaxel, would

that be likely to delay your time to distant

metastasis, would that be likely to delay your time

to cancer death?

That is a question, I can't answer it. I

could guess. I think the answer probably is yes,

328

but I don't have that. That is the first question

here. Then, the second question is in the setting

of a Phase III randomized trial, if the proportion

of men who didn't nadir went from 30 percent to

less than 10 percent, would this produce a clinical

benefit, and the clinical benefits I put below, the

time to bone metastases, the time to cancer death.

Those are the accepted endpoints in this setting.

The question is, is there a connection

between this nadir construct in that trial that I

described, not all trials, not all agents, this

very specific trial, is there a connection or not?

The only way to answer that scientifically

is to do the study powered for a distant metastasis

and/or survival, and see. But are we at a point

where we already can see?

DR. HUSSAIN: Thank you, Dr. D'Amico.

I want to thank all the speakers for very

informative presentations and for sticking to time.

I am going to be slightly more lenient than Dr.

Martino earlier, and give you a 10-minute break. I

would like us to assemble at 3:10 if you don't

329

mind, so we can begin to dissect all of the

information that we heard.

Hopefully, we will have a robust, lively,

but most importantly, productive conversation where

we would come out with some plans. Thank you.

[Break.]

DR. HUSSAIN: Before the committee

discusses some of the issues that came up, I would

like to begin this session of open public hearing.

Prior to inviting members of the public to make

their statements, I would like to read this

statement.

Both the Food and Drug Administration and

the public believe in a transparent process for

information gathering and decisionmaking. To

ensure such transparency at the open public hearing

session of the Advisory Committee meeting, the FDA

believes that it is important to understand the

context of an individual's presentation.

For this reason, FDA encourages you, the

open public hearing speaker, at the beginning of

your written or oral statement to advise the

330

committee of any financial relationship that you

may have with any company or any group that is

likely to be impacted by the topic of this meeting.

For example, the financial information may

include a company's or group's payment of your

travel, lodging, or other expenses in connection

with your attendance at the meeting.

Likewise, the FDA encourages you at the

beginning of your statement to advise the committee

if you do not have any such financial

relationships. If you choose not to address this

issue of financial relationships at the beginning

of your statement, it will not preclude you from

speaking.

Also, those of you from the public who

have not signed up to speak, you will be allowed to

speak after the registered members have already

done that. Thank you.

Open Public Hearing

MS. CLIFFORD: Our first speaker is John

Willey.

MR. WILLEY: My name is John Willey. I am

331

the treasurer and board member of the National

Prostate Cancer Coalition, which is America's most

active group in the fight against prostate cancer.

I speak on behalf of many other prostate

cancer survivors, and let me back up for a second.

No one paid my way here. I have no financial

interest in any drug company unless they are owned

by a mutual fund that I am not really going to the

second layer of, but I have no financial--I did get

a free lunch today, though.

When I was diagnosed with prostate cancer

at age 47, there was a lot of problems, and one of

the problems, as a baseball fan, was how many more

seasons was I going to see. One of the people that

I have gotten to know as I have done a lot of work

for prostate cancer was Larry Lucano, and as some

of you may know, he took over the Boston Red Sox in

2002, and they came on to win the World Series

after years of frustration this last year.

I would submit to you that three years

would be a real good time for drug approval, that

that is something that we should really shoot for.

332

To get there, surrogate endpoints is the only way

to go.

We are dragging our heels on this. I look

back on the June meeting and I am wondering what

has happened from June to here. We really need

desperately to get something moving on surrogate

endpoints.

As you know, there are over 2 million men

who are now suffering from prostate cancer, and

about 1 in 6 men will be diagnosed with prostate

cancer. Vietnam veterans, such as myself, have an

added higher incidence, about twice the national

average.

Prostate cancer gets about 17 percent of

the diagnosis of non-skin cancers, and yet has only

about 7 percent of the funding for research. We

desperately need surrogate markers in place to get

new drugs in place. Without the new drugs, we are

not going to have any sort of pushing back of this

disease, so that it is a chronic, treatable

disease.

I have been on a vaccine GVAX, and that is

333

only in a clinical trial and I received it twice,

in '98 and '99, but that alone has kept me going.

We need multiple of these types of drugs that can

push back prostate cancer and put it into a chronic

state, so that men can live with this and die of

other causes.

Thank you for your time.

DR. HUSSAIN: Thank you, Mr. Willey.

Are there any other members of the public

that wish to speak?

[No response.]

DR. HUSSAIN: I think that concludes our

open public session.

Committee Discussion

DR. HUSSAIN: In preparation for the

discussion, I wanted to sort of summarize some of

the points that were made by the speakers. Then, I

would like to ask the FDA for points of

clarification on some issues of approval, and then

we will go into the questions. The speakers can

correct me if my summary is not in spirit with what

they have said.

334

The first thing I think what I heard from

everyone, that survival certainly is the gold

standard, whether it is practical or not practical

to reach, but clearly that is the key.

In the era of active agents, at least in

the advanced setting, it is not an impossible goal

to get, so unlike previously, where we didn't have

good drugs, the problem is not so much the disease,

it is really not having active agents.

What you also heard that there are

multiple states of the disease to address different

endpoints for drug approvals, and that each state

would need to be addressed in a separate way.

There are some potential PSA kinetics that

might be promising, and I underline promising,

because they clearly have not been shown and

validated prospectively, but that they are

promising and, in fact, will need or may need to be

prospectively validated, and that each of these

points should be defined in light of the therapy

that has been utilized, that one cannot use a

one-size-fits-all for these endpoints.

335

That integrations of other disease-related

outcomes are important and should not be excluded,

and it should be perhaps included as part of a

composite benefit endpoint.

Is that pretty much within the spirit of

what you all said? Okay.

Now, I want to just address a few points

to the FDA, Drs. Pazdur or Temple, or any of the

group. Does an accelerated approval require a

Phase III trial? In my experience over the last

year, there have been presentations of drugs where

they have been approved based on some good results

in a large Phase II trial. I just want a

clarification on that, so that will help us in our

discussion.

DR. PAZDUR: Here again, let's distinguish

what accelerated approval is. It's an effect on a

surrogate endpoint reasonably likely to predict

clinical benefit, and it has to be an improvement

over existing therapy or available therapy I should

say.

Now, there has been a lot of I think

336

confusion in the oncology community between

accelerated approval and using a single-arm trial

for accelerated approval. When you are using a

single-arm trial generally, you have to perform the

trial in a very refractory disease population,

because the comparison is usually to a situation

where we are saying that there is no existing

therapy, hence, you could use a single-arm trial

since the control is recognized as having--there is

no control basically, there is no available

therapy, so any improvement would be considered an

improvement, or "any" in quotations.

The issue here is yes, we would be happy

to look at other stages of disease and have a

randomized trial. We have advocated doing

randomized trials and doing interim analysis

looking at surrogate endpoints of response rate of

time to progression, and granting accelerated

approval on that, and continuing the study on to

demonstrate clinical benefit of survival.

That was one of the initial trials that we

did was the initial approval of oxaliplatin, 5-FU

337

and oxaliplatin in colon cancer was a randomized

trial initially approved on response rate and time

to progression in a randomized trial, but here

again, you have to be better on that surrogate

endpoint than the control arm.

So, there is various ways of doing it.

The major issue is the surrogate has to be in a

clinical estimation reasonably likely to predict

clinical benefit, and you have to demonstrate to us

convincingly that it is an improvement over

available therapy.

We have even in some discussions looked at

improvements in terms of toxicity or safety being a

benefit rather than efficacy.

DR. HUSSAIN: Just so that I understand,

so in a third line setting, for example, if you

argue that Taxotere is first line, and mitoxantrone

for the sake of discussion is second line, someone

comes up with a 100-patient trial that shows some

composite benefit of palliation, what looks like in

the PSA activity, may be measurable to these

activities, stabilization and maybe some quality of

338

life, would that, in fact, make it for the

possibility of an accelerated approval in third

line setting pending appropriate trials to be done?

DR. TEMPLE: The trouble is you have

quoted a lot of different kinds of endpoints. The

principal endpoint that we have relied on in

single-arm studies has been tumor response, the

contention being that tumor responses are unusual,

to say the least, in the absence of therapy, so if

you see a tumor response, it probably can be

attributed to the drug.

We would not say the same thing about

palliative responses or improvements in pain. You

really do, we would say, need a control group

there. So, that is not as satisfactory. Whether

PSA convinces you, that is what you are going to

talk about.

DR. PAZDUR: The point also is that those

endpoints that you specified are truly clinical

benefit endpoints of pain benefit, so they would be

looked at potentially as full clinical benefit.

DR. EISENBERGER: I do believe that, in

339

general, for cytotoxics, a clinical trial that

would set the bar at survival is still a reasonable

thing, but I would suggest also that as we progress

with our targeted approaches, that we actually

consider paradigms that would have a clinical

meaning, such as a bone-targeted approach, for

instance.

If you delay the onset or progression of a

composite, similar to the Zometa, so that these

paradigms be considered, and these are

disease-specific, but also treatment-specific,

then, they perhaps have a different consideration.

DR. PAZDUR: I think, generally speaking,

we would not have a problem with that. Again, we

are looking always at a risk-benefit relationship

here, and if there is a more favorable toxicity

profile, I think there could be an argument made

for a delay in a certain event happening. We did

this, for example, with the bisphosphonates.

DR. EISENBERGER: For instance, an example

is a trial that would build on the efficacy or the

primers that were used for the approval of the

340

bisphosphonate, one would use a radiopharmaceutical

and add it to a bisphosphonate, and develop a trial

in that fashion, that would have nothing to do with

survival, certainly not with PSA, but with the

interference with progression of bone target

approach.

DR. TEMPLE: But one of the things that

would certainly be discussed was whether you have

made a change in the person's symptoms of some

kind, or whether you have changed a radiologic

thing. I am not taking a position, but that would

be something that you would have to discuss.

DR. EISENBERGER: Obviously, the trials--

DR. HUSSAIN: Excuse me, Dr. Eisenberger,

can I please define just some of the ground rules.

That way, we don't end up with in a duel and miss

the overall discussion here.

So, the ground rules will be that you

raise your hand. We will call on you in order to

make the point. In order to accommodate as many

people to participate, it would be very good to

have very brief and clear points, and I would like

341

us to, those who want to rebuttal a point, again to

raise their hand, and in that way we will call on

them in order.

The topics that were put for discussion,

they are listed in front of you, and the first

question, I am going to read it in general, but I

am going to try to take the Chair's prerogative and

maybe improvise the way we look at it.

The question reads or the point of

discussion reads: Regulations allow granting

regular or accelerated approval to a drug after

demonstration of safety and efficacy. Considering

these two situations, discuss the clinical states

in which PSA-based endpoints should be evaluated

for use in clinical trials to provide evidence to

support either type of drug approval.

Based on what we heard today, there is

clearly I think two general distinct states that we

probably should focus on, and not get into too many

breakdowns.

There is the early stage disease which Dr.

D'Amico was pointing to. I would like to reserve

342

that for the second part of the discussion. But

the first part would be metastatic conventional

hormone refractory state of disease i would like us

to focus the questions on.

In your comments, please phrase whether

you believe PSA by itself or some other composite

endpoint is what you think is needed.

Anybody wants to begin? Dr. Brawley, we

will call on you.

DR. BRAWLEY: Thank you. I was very

impressed with all the speakers this afternoon. I

will tell you my prejudice right now is Howard

Scher had a slide that said that PSA plus other

endpoints is one point. That might be a reasonable

thing to look at as an endpoint.

PSA by itself clearly is not a good

surrogate endpoint except for the one state of PSA

rising is a bad thing clearly.

DR. HUSSAIN: Dr. Klein.

DR. KLEIN: I would like to disagree a

little bit with Otis. I think there is substantial

evidence in the urologic literature, although not

343

all of it is as rigorous as defined by the Prentice

criteria, that PSA doubling time or another form or

another derivative of PSA kinetics really reflects

the biology of the disease, and Anthony showed a

lot of it, but there is more.

There is evidence in the pre-diagnosis,

pre-prostatectomy model that a rapid PSA doubling

time is associated with poor survival despite

aggressive therapy.

There is evidence that Anthony has fleshed

out that after treatment, that it is associated

after radiation or surgery, there is evidence in

the older literature and in the JAMA article that

was published multi-institutional study last year

by Andrew Stevenson, that in response to predicting

a response to radiation therapy, that it is

predictive, and all of those things are based on

PSA doubling time or some derivative of PSA

kinetics.

When you see a predictor like that, that

crosses the boundaries of all the different

clinical states, it says to me that it is capturing

344

the essence of the biology of the disease, and we

ought not ignore that.

I would agree with Dr. Scher's point that

he showed with those two cases, that PSA doubling

time is not going to be the perfect surrogate for

every case. You will always find exceptions. But

we are in a situation now where we have a clear

need for new drugs in the management of all these

different states in prostate cancer.

Neither pharma nor big academic centers

are going to put a lot of time, effort, and money

into looking at survival when the survival

endpoints are so far off, and it really is time now

to design the clinical trials as has been suggested

with a PSA kinetic-based endpoint to try and

validate it, and whether that will be sufficient

for accelerated approval or not, I don't know.

But if we don't do that, we are going to

be stuck, and I would just sort of add that we may

not as a group today agree on what the appropriate

PSA endpoint is, but we should go where the bulk of

the data is, which I think supports PSA doubling

345

time as an appropriate surrogate to test in

clinical trials, and then we can have some data and

say yes or no, this was the right thing to do.

DR. HUSSAIN: I just want to point out and

remind you, please, we are talking strictly right

now about metastatic hormone refractory disease, so

if you don't mind limiting your comments to that,

and then we will get to the early stage disease.

Dr. Andriole, did you have your hand up?

DR. ANDRIOLE: Yes, I did. We are talking

about the later stage of patients with hormone

response disease, and my question or thought to the

medical oncologists, which I am not, is it feasible

to do a study in which men with this stage of

disease are blinded to their PSA, and just treat

them and make your treatment decisions on the basis

of symptoms?

Number 1. The first question, is it

ethical, and number 2, were it to be considered

ethical, would it be doable, because if you could,

that would I think give us a lot to talk about.

DR. HUSSAIN: I would probably respond

346

simply by saying no, I don't think it's doable.

Ethical, we can debate it later, but I think doable

is more important than ethical.

Dr. Scher.

DR. SCHER: I think what we have seen

about PSA doubling time is that it becomes an

important prognostic factor as to who is at high

risk for a significant event, and that has to be

distinguished from a treatment predictive factor,

which is a post-intervention outcome.

What we have seen across the states is

that this has become critical to identify patients

for enrollment, but that does not tell you anything

about its role as a potential outcome measure.

DR. HUSSAIN: Dr. Martino.

DR. MARTINO: At the risk of being simple,

ladies and gentlemen, I need to ask a question, and

I would like a simple answer from somebody, because

you are all rattling on, as best as I can judge

right now. I want to focus the group on people

with metastatic disease, not early disease,

metastatic disease. I think that was the point

347

that you were taking us to.

Is a change in PSA alone, is a change in

PSA alone adequate for any of you to change

therapy? To me, that is really the question.

That's the question, and I would like an answer to

that. Is PSA alone adequate?

DR. HUSSAIN: Dr. Brawley, do you want to

take that?

DR. BRAWLEY: Yes. First off, Eric, I

agree with everything you said for localized

disease, but in the case of metastatic disease, I

do believe--well, first off, if you give Taxotere

and measure PSA several days later, you will have

an increase in PSA because of tumor dying out and

releasing PSA.

But a sustained increase in PSA, while one

is getting cytotoxic chemotherapy, to me does mean

progression of disease. A decline in PSA is not

nearly as much information to me as a rise in PSA.

DR. HUSSAIN: Dr. Scher, you wanted to

respond to that?

DR. SCHER: If the PSA is going up, and is

348

not affected in any way by a cytotoxic agent, that

is a indication that it does not work.

I would add that there are, for example,

using weekly Taxotere, you can see delays in the

decline of PSA for upwards of 6 weeks, so that is

important information to explain to a patient.

What? We agree.

DR. BRAWLEY: We are saying the same

thing.

DR. SCHER: We agree, yes. Going down,

it's helpful, but it's not the whole story.

DR. HUSSAIN: Dr. Eisenberger.

DR. EISENBERGER: I just want to also,

just for the sake of keeping in the record, when

you treat patients with Taxotere, and the PSA goes

up, it doesn't mean that it has anything to do with

Taxotere. It is the disease that is progressing.

We actually looked into that in TAX-327.

So, these are patients who are rapidly

progressing, who will take a little longer for

their PSA to go down, but that doesn't happen very

frequently. Most of the time, early rises in PSA

349

equal progression.

DR. HUSSAIN: Dr. D'Agostino.

DR. D'AGOSTINO: I am not sure I digested

all the material on the accelerated approval. If

some sponsor gets an accelerated approval based on

PSA, they still have to do a clinical study, right?

So, in terms of moving the discussion, it seems to

me like the PSA analyses that we have seen have

more been like baseline as opposed to if you have

this, you are in trouble.

The progression, I haven't heard that

really said that that leads to anything. But

studies that talk about the progression as the

Phase III in an accelerated approval, and then

followed in the Phase IV with a harder endpoint,

and the confirmation of the PSA rising, I think

would be a sort of a scenario that one could

possibly implement without running into some big

ethical problems.

But I haven't seen, just to iterate, I

haven't seen the increase, the doubling of the PSA

as being an indicator of mortality in the data that

350

I have seen presented.

DR. HUSSAIN: Dr. Raghavan.

DR. RAGHAVAN: I would like to come back

to answer Dr. Martino's question. So, the answer I

think, Silvana, is it depends on the context. I

don't think you can predicate management solely on

PSA because prostate cancer is a heterogeneous

disease.

Now, if you want to do it on averages,

that is, on average will I be accurate most of the

time, then, you can do it. If a PSA drops 75

percent, most of the time that correlates with a

good outcome. If the PSA consistently rises over a

period of 3 months, most of the time that

correlates with a bad outcome.

But there are some phenomena that

interfere with the answers that we have heard

before. For example, there are quite clear data

that show that for a number of cytotoxics, if you

are silly enough to do daily PSAs, which very few

people do, you will identify a flare-up reaction

with release of PSA in response to a cytotoxic,

351

much as you occasionally do in breast cancer with

one of the breast markers.

Many of the clinical trials that we talk

about sample PSA values at weekly or 3-weekly

intervals, so they don't actually have the data to

answer the question.

The second phenomenon is a clinical one,

which is you will see patients who have a clone

that produces PSA that disappears during

chemotherapy with a resistant clone that is silent,

sometimes neuroendocrine, sometimes not, where you

will have a patient who is actually deteriorating,

losing weight, losing performance status,

increasing pain. So, this is the disconnect between

symptoms and PSA when the PSA goes down.

So, the answer to your question is it

depends on what proportion of the time you are

prepared to accept being right or wrong.

DR. HUSSAIN: Dr. Martino.

DR. MARTINO: So, can I then conclude that

PSA alone would not be an adequate way to power a

trial, that something beyond that must be added?

352

If that is correct, then, can we move on a little?

What would be the other things that would need to

be added?

DR. HUSSAIN: Dr. Klein.

DR. KLEIN: You are correct, you are

correct. I mean I would point out again that we

are looking for a surrogate that describes or

predicts the behavior of a population, not the

individual exceptions. No surrogate is going to

perfectly predict the outcome for an individual

patient, and we need not to perseverate on that

issue. I think we need to move beyond that.

I think what you have heard today from

everybody is that there is a substantial amount of

evidence that suggests that a PSA derivative may be

a useful surrogate, but it needs to be tested in a

prospective clinical trial, using a standard

clinical endpoint, before we will accept that.

DR. HUSSAIN: Dr. Temple.

DR. TEMPLE: It just seems we are saying

that PSA isn't an absolute thing, there are a

variety of measurements that have already been

353

discussed, like doubling time or percent reduction,

or something, so that you might not be convinced

that any change means something, but you might be

convinced that some kind of change, a nadir less

than 0.2 or something means something.

Can I just say something about possible

study designs? It is always tempting to take a

look at the people who have a response, like whose

PSA goes to something very low, and then see how

they do compared to people who don't get that

response.

This has been done for years, and it

always gets the same criticism that maybe this is

true true unrelated, you might have picked out the

people with a good prognosis because they are the

ones who responded.

There is a study design that I want to

throw out, so you can tell me it's impossible, that

avoids that problem. If I understood the slides I

saw, you can expect a reasonable percentage of

whatever PSA response you are going to get in about

6 weeks. It would therefore be possible to take a

354

population, treat them all, look at what happened

at 6 weeks, and then stratify according to

response, you know, 50 percent, 40 percent,

whatever people thought was meaningful, stratify

and randomize to treatment and no treatment.

You do that, and you see a better

response, you see a better outcome on whatever it

is you are measuring, associated with a bigger PSA

response, and then you don't have to worry about

Prentice anymore, because if you saw that, that

would make it a credible surrogate for outcome, I

think.

Now, the obvious question is would anybody

let you do that trial. Everybody would be on

whatever hormonal therapy there be, but you would

have to take people who had a response that at

least some people believe in and not give them the

drug. So, it would be nicer if you could do some

scan at one day or something, and people would be

more comfortable with that, but I would be

interested in what people think about that as a

possible design.

355

It really does avoid the true true

unrelated problem.

DR. HUSSAIN: Dr. D'Agostino.

DR. D'AGOSTINO: Maybe I am not following,

but doesn't that sort of stratify by what you think

might be severity as opposed to saying PSA is

progression after you have taken the drug is going

to be useful?

DR. TEMPLE: Well, you are going to look

and see, I mean you may also stratify by the

pre-treatment doubling time or something like that,

but, no, you are taking--let's make it up.

Let's say you want people who fall to less

than 0.2, that is one stratum. Less than 0.4 is

another, no response is another. We will have 3

strata. Then, you randomize to the treatment or no

treatment, and you show presumably that people who

had no response don't get any benefit on whatever

it is you are measuring, but the people who were

knocked down to 0.2 by the treatment have a

dramatic improvement in outcome.

DR. D'AGOSTINO: I am missing. When do

356

you stratify, do you put them on treatment, wait

until they respond?

DR. TEMPLE: Everybody goes on treatment.

You look at the response and then you stratify.

You stratify by response.

DR. D'AGOSTINO: But you have the

individuals. I thought you said you looked at

doubling and then you categorized individuals, then

randomized within those categories.

DR. TEMPLE: That's right.

DR. D'AGOSTINO: Well, they don't have

treatment before you randomize.

DR. TEMPLE: They have all been treated

for 6 weeks.

DR. D'AGOSTINO: They have all been

treated for 6 weeks.

DR. TEMPLE: For 6 weeks or 4 weeks, or

whatever you think is long enough to know what

their PSA response is. You then randomize them to

treatment and no treatment. So, you have got to

hope the 4 weeks of treatment doesn't make too big

a difference. If it did, that would undermine this

357

design.

You then have groups who are stratified by

response, and you then randomize to the two

treatments. So, it is multiple randomized trials

in people with different responses. Now, whether

you can do that or not, I don't know, but I think

it does have the potential for answering the

question whether the PSA response, in fact,

predicts an effect of therapy on some other kind of

outcome, like death.

DR. D'AGOSTINO: But your outcome would be

death?

DR. TEMPLE: Well, you choose the outcome.

Time to progression, I mean I am not trying to

choose the outcome, one that you feel is a

comfortable outcome. Could be time to bone mets or

whatever you want really.

DR. HUSSAIN: Dr. Raghavan.

DR. RAGHAVAN: So, this is a composite

answer to a composite endpoint. This is Dr.

Eisenberger and myself muttering together. So, if

we understood you correctly, and the randomization

358

comes in patients who have had a PSA response, it

goes back to what Dr. Hussain said. It is not

doable in the world today, because patients are so

PSA dependent, irrespective of what oncologists

think. Most urologists, as you heard earlier,

believe in PSA, and get excited about the concept

that it correlates with the disease.

So, if you have a patient with metastatic

disease, in the early part of their PSA-associated

lives, PSA is very important as a parameter of what

is going on. It becomes less important later, but

they have been trained to be PSA responsive.

So, to say to a patient whose PSA has

disappeared, well, we are going to flip a coin, and

on the toss of a coin, you might not get that

treatment that is about to save your life, has no

chance of working. So, you will get an accrual of

zero.

DR. TEMPLE: Well, not if they can't get

the drug any other way, they won't.

DR. HUSSAIN: Dr. Perry.

DR. PERRY: I apologize. I feel like a

359

nickel among dimes here, listening to all the

experts on prostate cancer. Perhaps you could

answer a simple question for me. Everyone wants to

compare PSA against a hard endpoint, and I don't

know what that hard endpoint is.

I hear you say that survival doesn't work

because too many people die of comorbid diseases,

and it takes too long. it is the ultimate great

endpoint, but for practical purposes it isn't going

to work. Time to progression is complicated, and

bone scans don't work.

So, what are we going to compare PSA

against in these trials?

DR. HUSSAIN: Dr. Perry, I think that in

the hormone refractory setting, I think patients, 9

out of 10, of they were going to die, they are

going to die from their cancer, so that is not a

problem there.

DR. PERRY: It's going to take a long

time.

DR. HUSSAIN: Dr. Perry, the median

survival in a hormone refractory patient--and

360

perhaps that is where this whole thing seems to be

sort of, if I want to say, oxymoronish in some

ways--in the late stage disease, I don't think we

have too much of a problem of time way and beyond

any other solid tumor.

The median survival of your best patient

population that go into chemotherapy trials is a

year and a half, that is how good we are, this is

it, a year and a half. So, I guess in my mind, the

hormone refractory setting in front line, if I may

just put my two cents in there, to me, the answer

is clear. It's survival endpoints, get drugs up

there and randomize and get it done with.

Where I think--and perhaps if we can maybe

just to get focused a little bit--if we can agree,

for example, that in a front line setting, survival

should still be front line for brand-new metastatic

hormone refractory disease, that survival is the

endpoint because these trials are not difficult to

do from time points.

It is more in terms of patient accrual

into the trial, and I think we have demonstrated in

361

the last 5 to 10 years that we have really

maximized per year our ability to get these

patients in to do trials in a short period of time.

Where I think there may be room to get

drugs more into these patients is in the second and

third line setting where now that we have Taxotere

front line, but it is not exactly curing patients,

so the question is can we envision trial designs

that are short of being randomized 700-patient

trials, that would allow us to test some promising

agents in that setting and give us some expedited

drugs into the market while we prove the principle.

If I may ask that we focus on that point

perhaps, because as I am speaking, I see everyone

shaking their head that they are agreeing that,

without even a vote, that survival for front line

hormone refractory is a done deal, so let's just

move on.

The question is we have a second line or

third line setting, whatever you want to argue it,

can anyone make a recommendation for what they view

as a trial design that would be of value? Since

362

there are people who have raised their hand prior

to that, Dr. McShane, I am going to allow her to go

first.

DR. McSHANE: Some of the points I was

going to raise have already been raised, but I

would like to emphasize that to really establish

something as a surrogate, no matter what setting we

are talking about, it takes more than a single

trial.

You have to demonstrate that repeatedly,

over multiple trials, that the answer you get on

the definitive endpoint is the same as the answer

you get on the surrogate, so I think we need to

keep that in mind.

DR. HUSSAIN: So, is what you are saying

that from everything you heard, that PSA, as it

stands right now, with all the suggestions about

its correlation to outcome, is not yet a valid

endpoint to be trusted 100 percent until we

validate it?

DR. McSHANE: That would be my opinion.

DR. HUSSAIN: Just so that people know,

363

the two randomized Phase III trials that Dr. Alison

point out to, the CLGB and the SWOG trial that is

going to look at Taxotere, Atrasentan versus

Taxotere, there are built into it prospectively

criteria to validate the observations that were

made in the TAX-3 trial, and then the SWOG-9916

trial, so some validation on percent decline of PSA

is being built into these trials prospectively, and

this may, in fact, serve as a model for cytotoxic

chemotherapy for screening.

Dr. Eisenberger had his hand first.

DR. EISENBERGER: I just wanted to go back

on the PSA. I think we are trashing too much the

PSA. The PSA, in fact, is used in clinical

practice extensively. If a PSA is going down, we

know the patients are being helped, if the PSA is

going up, it's actually most likely not being

effective, and that is what we use.

We effectively use PSA to define whether a

therapeutic regimen in the Phase II setting is

going to be effective or not, and regardless of

whether we agree exactly on how much and for how

364

long the PSA declines, this was done in the

docetaxel regimens, and this is how we eventually

defined in two, Phase III trials that there is a

survival advantage.

So, I don't think there is a question that

the changes in PSA sort of tell us whether a

therapy is working or not, and here is the

difficulties. When we are actually trying to pin

this down and look at a surrogacy for any survival

or any other outcome, this is where there is a

problem.

Part of the problem is that you can't do a

trial when the PSA is going down, and then stop

therapy in a substantial proportion of these men,

and you cannot continue a trial if your PSA is

going up, if this is what you design, just to test

the PSA, I think it would be a waste or it would be

very difficult to do, and that is why I think it

would be a waste of resources.

But one of the things that I wanted to

refocus here, what we are trying to do here, is we

are trying to come up with a reasonable hypothesis

365

that need to be incorporated into Phase III trials

from now on. I don't think it's enough for us to

just do a Phase III trial and find out whether

there is a survival advantage.

I think what we need to do is we need to

come up with trials that will look at survival as

the main endpoint, but also test a certain

hypothesis, which is reasonable, and I think we

ought to focus on that here today, and provide you

with something which is clinically relevant and

testable in the context of Phase III trials. This

is what Anthony tried to do and this is what Howard

tried to do, and maybe we ought to focus on that.

DR. HUSSAIN: I will get back with you as

the first person to make a hypothesis once I get

the other individuals to speak, so get prepared.

Dr. D'Amico.

DR. D'AMICO: I just wanted to just

highlight a point that has been made, and that has

been made by several people. In the two, Phase III

randomized studies in hormone refractory metastatic

disease that we have heard about today, the SWOG

366

and the TAX-327 study, they accrued somewhere

between 700 and 1,000 patients in a year and a

half, and then they had follow-up, and they were

published four years after accrual started.

So, I want you to think about if we had a

surrogate in that setting that was based on PSA,

that you could figure out within 3 months after

treatment ended, you have a year and a half to

accrue, and then after that, another 6 months,

let's say, to do your analysis, 6 months to have it

peer reviewed and published, when you add all that

up, that's 2 1/2 years, so you will buy a year and

a half perhaps at best if everything goes exactly

perfectly in this setting with the surrogate.

That doesn't mean we shouldn't explore

that, a year and a half could be very valuable, but

I want people to understand exactly what are we

talking about when we are talking about end-stage

prostate cancer in a surrogate, we are talking

perhaps a year, year and a half sooner to report.

But maybe more importantly, with the

studies that have been designed and have this PSA

367

constructs built into it, we will learn something

about the biology. It is conceivable in terms of

study design that if these PSA constructs aren't

proven to be important, that you can start somebody

in a randomized study, they achieve a certain PSA

endpoint which you now know is important, and you

take them and put them, randomize them onto the

next study based on that construct, it is possible

that you might then be able to figure out something

sooner in the game.

But I just think that the point I want to

make is a surrogate in this setting could be of

some value, but if you are looking at the most

value for a surrogate, clearly, we will talk about

it later, in earlier disease would be where that

biggest impact could be made.

DR. HUSSAIN: Dr. Grillo-Lopez.

DR. GRILLO-LOPEZ: Thank you. I wanted to

make two comments. First, I certainly don't agree

that overall survival should be the gold standard

even for front line in the setting that we are

discussing, and if you look at one of the studies

368

that has been discussed, presented a couple of

times today, the docetaxel versus mitoxantrone

study, and you see that the median survival was

reached in 16 to 18 months, that means that half of

the patients had progressed and died before that,

so they probably had received some other therapies.

So, not only the median, but the rest of

that Kaplan-Meier curve was affected depending on

what other therapies plus a number of other

confounding factors those patients had. Overall

survival is not a good endpoint even in this

setting.

The second point I wanted to make is that,

again, searching for focus in this meeting, from

what I hear the FDA saying, and from the content of

the agenda, I think that the FDA is really looking

for recommendations for surrogate endpoints that

could be helpful to the FDA in getting their job

done, and certainly helpful to pharmaceutical

industry in getting these products approved faster.

I also hear that some of the studies, the

large randomized trials that are necessary to

369

validate the endpoints may take 4 to 10 years.

That was the comment from one of the speakers.

So, if we were to say that today, we

cannot recommend to the FDA a surrogate endpoint,

and that we have to wait 4 to 10 years, that means

that pharmaceutical companies can really not

negotiate with the FDA for another 10 years or 4

years to start a trial, which would then take

another 4 years to complete.

So, we are saying that at the earliest, if

that happens, we would not be doing trials based on

or we would not be completing trials based on

surrogate endpoints for another 8 to 20 years.

So, we need to take some risk. We around

the table today need to take some risk and say with

what we know today, which may not be perfect, which

may not be 100 percent validated, is there some

surrogate endpoint, PSA, PSADT, whatever, that can

be used today while we take those 10 years to

validate all of this with 100 percent certainty.

DR. HUSSAIN: Dr. Temple.

DR. TEMPLE: A point you have made several

370

times now is that if you take people who are

hormone refractory, we are talking about much

shorter periods of time, so don't make it 10 years

right away.

The other thing is that I think Dr.

D'Amico's data on initial therapy show

unequivocally that if you put the right people into

the trials, namely, people with short doubling

times, you can do a study very rapidly, and if you

put the wrong people into the trial, you have no

chance of ever finding anything, because there are

not going to be any deaths.

So, that was too discouraging, I think.

There are ways to do these even if mortality is the

endpoint, but we have never said that mortality is

the only endpoint, and if you look at the approvals

there have been, they used other endpoints which

occur earlier than mortality.

Can I ask Dr. D'Amico a question? Even

though people are critical of studies that show the

relationship between outcome and the results on a

test, a potential surrogate because it might be

371

confounded, that is the thing you start with. I

mean there has to be a relationship between outcome

and the putative surrogate in an after-the-fact

way, or you don't have a chance.

So, my question for you is, have you

looked at nadir, say, as a good candidate endpoint,

corrected for baseline doubling time, because in a

lot of the data you showed, the two were going

together, but one of those is a characteristic of

the tumor, has nothing to do with treatment, but

the nadir does have to do with treatment, so can

you tease out the nadir effect and relate that to

outcome?

Maybe you have already done that, because

you would expect that at a minimum, even if you

weren't entirely satisfied with that approach, it

is still what you would expect.

DR. D'AMICO: I will say it quickly

because it really doesn't apply to metastatic

disease as far as I know, because I haven't looked

at it in metastatic disease.

DR. HUSSAIN: Thank you, Dr. D'Amico.

372

DR. D'AMICO: The answer is they are

independent, because they are both significant in a

multivariable analysis.

DR. HUSSAIN: Dr. DeGruttola.

DR. DeGRUTTOLA: I wanted to return a

little bit to the topic of validating surrogates,

and I think an important point here is that the

goal of the surrogate is to know that the effect of

treatment on the surrogate predicts the effect of

treatment on the clinical endpoint, and there is a

number of ways to do that, as Dr. Temple mentioned.

The design that he proposed is an elegant one, but

obviously is only workable when there is

uncertainty about the surrogate, so that people

will accept the idea of being randomized even if

they had a surrogate response.

The other approach is just to collect

information from a number of trials and show that

you can actually predict the extent of treatment

benefit from the effect on the surrogate.

A number of people have commented on the

Prentice condition, and I think the Prentice

373

condition is conceptually very useful, the idea

that if you have a test on the surrogate, it's a

valid test of the clinical endpoint, but I think

that operationally, it may not be the best way to

try and approach the issue of surrogacy.

First of all, meeting the Prentice

condition, which is that the hazard of the clinical

endpoint, given the surrogate, is not impacted by

the treatment, in other words, once you know the

surrogate, the treatment gives no additional

information about the risk of the endpoint.

That isn't really necessary to show that

something is a good surrogate. I mean in a case of

using HIV viral load in AIDS, no one has ever

demonstrated, in fact, that the Prentice conditions

are met. Michael Hughes and colleagues work showed

that, in fact, only a relatively modest proportion

of treatment effect was explained by HIV, but it

still has turned out to be a very good surrogate,

as everyone knows from the declining death rates,

and so on.

The other thing is that it is not

374

necessary. It may also not really be sufficient.

The problem is that a lot of the analyses that are

used, are the so-called showing the proportion of

treatment effect explained is close to 1, but those

estimates tend to be highly unstable both in terms

of large confidence intervals, unless you have

really big treatment effects, and also, they are

very subject to fluctuations when you include or

don't include certain covariates, and so on.

I think that they are useful analyses to

do, I think you can learn from them, but I am not

sure that that should be the primary way of

addressing surrogacy.

The other point that Tom Fleming has made

a number of times in print with a number of

colleagues is that there is an identifiability

issue that if the treatment can have negative

effects on the outcome of interest by a different

mechanism from the positive effects, you can show

that a proportion of treatment effect is quite

large, when, in fact, the surrogate isn't

explaining most of the benefit.

375

So, I think that while the Prentice

condition is a useful way to think about things,

and the proportion of treatment effect explained,

are useful analyses, other approaches may be

preferable for establishing surrogacy.

DR. HUSSAIN: Dr. Scher.

DR. SCHER: I would just like to try to

refocus the discussion a little bit. For the

patients who progress on hormones, there are two

populations, the first line setting where the

median survival is 18 months, maybe a little longer

with the stage migration, and the second line

setting when you are in the order of 12 to 16

months depending on what you look at.

The response in patients after second line

therapy, using PSA criteria, is less than 15

percent. So, it is highly unlikely you are going

to see a significant impact on survival.

So, the question I would like to pose is,

if you are designing a trial based on survival for

the sake of argument, in which you will embed some

PSA construct with or without other measures, would

376

the Agency accept a trial which includes more than

one intermediate on which to base an accelerated

approval, or are you restricted to declaring one,

looking at others?

So, for example, if you put in a trial

which has one metric, which is PSA response, a

second which is based on a PSA progression, and a

third which is based on PSA progression plus

clinical progression, if all of those three were

proposed in a trial powered on survival, could you

do an analysis and not be penalized because you

happen to select number one, number two, or number

three, as your hypothesis?

DR. PAZDUR: You would probably have to

have some decision tree here. The answer is yes,

but you would have to prospectively adjust here.

There are many trials that have multiple secondary

endpoints.

DR. SCHER: But the question is if you are

using one of those secondary endpoints as the

embedded indication for reasonably likely to

predict, while the trial goes on to completion, do

377

you have to declare one, or conceivably could more

than one be looked at?

DR. TEMPLE: You have to preserve your

alpha, there would be a debate about it, since

those are obviously not completely independent, you

have to argue about what the correction would need

to be, and just--ask Ralph, he will tell you.

DR. D'AGOSTINO: You are powering it on

mortality, you said, right, survival, so I think

you could very comfortably run a study like this.

It may be overpowered on the surrogates, if

anything, and that's okay, but that might be what

will happen, and you can protect yourself in terms

of alphas, and what have you, because you are going

to have such a powerful study on the surrogates, it

is the question of do you list the surrogates, do

you know the surrogates, are we comfortable enough

with the surrogate that we are proposing.

I had another question I wanted to ask,

and it goes back to Lisa's in terms of pushing for

the surrogate. As a statistician, I would be the

last one to say that surrogate variable don't need

378

careful validation, and what have you, but

sometimes the hell with that, and when you have the

accelerated approval, the surrogate is reasonably

likely to predict a clinical benefit.

If we are talking about situations, second

line, and what have you, where you might be able to

put together a reasonable study with the surrogate,

the proposed surrogate, and then move on to a Phase

IV that really has an endpoint--

DR. PAZDUR: Or continuation.

DR. D'AGOSTINO: Phase III, it depends on

how long the accrual is. If the accrual is fast,

and the mortality, you know, it is going to be a

solid one, then, why talk about it at all.

But you don't want to have the study based

on the survival as, you know, sort of losing track

of the fact that if you can run it fast enough on

the survival, then, you can look at the surrogate,

and I presume everybody would say that would be a

fine study.

I am concerned with the situation when you

are talking about the survival is going to take too

379

long, trying to get these hard endpoints is going

to take too long, so can you do something with a

reasonably likely surrogate, and then put a more

careful study together where you can confirm that

surrogate variable.

DR. SCHER: The median time to progression

in the TAX-327 and 9916, was on the order of 6

months, and if you are looking at median survival

of 18 to 20 months, that is not--

DR. D'AGOSTINO: If you can do it.

DR. SCHER: A progression-based trial,

whether it is PSA or PSA response, you would still

be saving 18 months to a year, so that is

significant.

DR. HUSSAIN: Only because we have a lot

of area to cover, I want to ask you to please be

brief and make the point.

Dr. Klein, you had your hand up.

DR. KLEIN: I just wanted to add something

to what Dr. D'Amico observed about a benefit in

terms of defining a surrogate and getting the

answer 18 months early. That is one benefit for an

380

individual agent, but there is another benefit. If

we can define that surrogate, we can screen other

agents a lot more rapidly, and that 18 months is

very meaningful in that setting in assessing

alternative or new agents. So, there is both

benefits.

DR. HUSSAIN: Dr. Raghavan.

DR. RAGHAVAN: I wanted to just respond to

Tony Grillo-Lopez's comment, because I think one of

the things we haven't stated today, but is

implicit, is that there is an awful lot of work

going on at the moment with the data that we have

already acquired.

So, I know the TAX-327 team are busily

playing with numbers, as are the SWOG team and many

other people around the world. There is the tool

of meta-analysis. So, I think it's a little facile

to suggest that if we don't come up with the answer

today, we are somehow committing a crime against

mankind.

I think the reality of the situation is if

I bring a new product to the FDA tomorrow, and we

381

set up a series of parameters that I embed in my

trial, those parameters will have better data

available to help evaluate them by the time the

study is done.

So, it is not as if, as has been implied

now a couple of time, that it is a bad thing to do

this in a scientifically rational way. There will

be data. Nothing that anybody has said today is

going to come out of left field as a surprise.

We know what the current potential

surrogates are, and that is why I was making the

plea to embed them. I still think, Mike Perry,

survival is a good place to anchor this. That was

the point I was making.

Where it becomes blunted is if you don't

take the state's model into consideration, in other

words, survival for someone with early stage

disease becomes much more hard to interpret.

DR. HUSSAIN: Dr. Sridhara.

DR. SRIDHARA: I just wanted to go back to

Dr. Scher's question of having three sort of

PSA-based endpoints and how do we deal with it if

382

we want to keep all three of them as primary

endpoints, and in this case, you are powering the

study for overall survival.

I think if you can prioritize which one of

them is the first one that you are going to look

at, then, probably you don't have to pay a penalty,

in other words, if you can go, okay, this is the

first one, this is the second one, and this is the

third one.

But I think you have to carefully examine

the data that is already available, what would be

these three, and how would you prioritize. If you

think of PSA response and PSA progression,

obviously, you will be seeing PSA response before

you see the PSA progression.

So, there will be some kind of time effect

in your prioritization of how you want to look at

it, and whether you want to give higher priority

for progression, that may be something that you

want to look at, and then we can deal with it

statistically. That is not an issue.

DR. HUSSAIN: Dr. D'Agostino.

383

DR. D'AGOSTINO: If we are talking about

the setting where you can do a mortality trial and

get it done in a reasonable amount of time, then,

putting forth different variations of the surrogate

can I think easily be put in, and they probably

will have a lot of power.

You can put them in a sequence, as you

said, but probably if we are clever enough, we

could probably have a reasonably good power on all

three of them, three or four, so I think that would

be a very sensible type of design.

DR. HUSSAIN: Can we then use that

criteria for first line?

DR. SCHER: Yes.

DR. HUSSAIN: Powered for survival and use

other endpoints. Howard?

DR. SCHER: Yes. I mean I obviously have

a bias toward progression, because I think that PSA

response doesn't capture all the information that

you can learn, and there is a time factor, but I

think if people can start developing trials, and

not pay a penalty for selecting one, then, we have

384

really made significant progress.

DR. SRIDHARA: I think if you can

elaborate on how you are going to define this

progression, that would be important, like how

often are you going to measure this, and how are

you going to deal with missing values if it comes.

I think these are the issues that we come

up with progression in other solid tumors, when we

are trying to measure progression, it is a question

of how often you measure, and if you have missing

values, how are you going to deal with these

missing issues, and those have to be very specific.

DR. HUSSAIN: Let me ask you then a

question. Supposing you have drug A you are

testing against Taxotere, and drug A wins against

Taxotere for a primary endpoint of time to

progression by, say, 4 months, and the survival is

no different, is that drug not worth it?

DR. PAZDUR: Why? Why isn't the survival,

why aren't you winning that survival?

DR. HUSSAIN: If I am God, I will answer

it, but I am not.

385

DR. PAZDUR: I guess the question that I

am asking, is it crossover effect, is it inadequate

powering of the trial--

DR. SCHER: Or is it a bisphosphonate that

doesn't affect survival and affects clinical

events.

DR. HUSSAIN: If the drug is brought here,

and it has a phenomenal time to progression or

progression-free survival benefit, and not a

survival advantage, I guess that ties into my

question that I was going to ask you, have there

been drugs approved based on a progression-free

survival?

DR. PAZDUR: Oh, of course.

DR. HUSSAIN: Even though there is no

survival advantage?

DR. PAZDUR: Of course, correct.

DR. HUSSAIN: So, a setting like this

would not basically kill the drug.

DR. PAZDUR: As long as there is not a

decrement in survival.

DR. TEMPLE: We have brought this question

386

to the committee. There are at least two major

reasons why you don't see an effect on survival.

One is that people cross over when they progress.

That has got to go in the direction of not showing

an effect even though you don't know how big it is.

The second is just as a hazard ratio

matter, going from 10 to 8 is as bigger effect than

going from 20 to 18, so survival is more difficult.

It is clearly more difficult especially if it's at

some distance from progression.

So, yeah, there are a lot of drugs that

have been approved based on progression.

DR. HUSSAIN: Dr. Grillo-Lopez.

DR. GRILLO-LOPEZ: There is two points I

would like to make. My friend at the end of the

table here, I don't know that anything has been

approved on the basis of a meta-analysis as a

primary pivotal trial although it is useful in

support of certain data.

But more importantly, the word

"accelerated" means to be faster than something, in

this case, regular approvals, and the more we make

387

the accelerated approval mechanism similar to the

regular approval mechanism, the slower it gets.

So, we tend to discuss randomized trials a

lot, and, yes, those are more elegant, perhaps they

give you greater security that you are doing the

right thing, but the faster way to develop a new

agent is with a single-arm trial with the

appropriate endpoints, and that is where this

committee has to take some risk today and come up

with suggestions to the FDA on some appropriate

endpoints for those kinds of trials.

I like the situation where Dr. Temple is

the optimist and I am the pessimist, because it

allows me to make my points more strongly and gives

me hope that you are going to act faster in

approving drugs.

So, I look at prostate cancer drug

approvals, and in the past 24 years, there has been

three. It is better than nothing, but it is a

dismal record for prostate cancer patients that

only three new agents have been approved in 24

years, ladies and gentlemen.

388

The other thing that I want you to

consider is that as you look at the audience here,

this is a relatively small audience this afternoon,

and if you discount the analysts, the media people,

and if you count only the company people, the

pharmaceutical company people who are here because

they have under development a prostate cancer

agent, there are very few of them.

We need to ask ourselves why, why is there

not more interest in developing new agents for

prostate cancer, and, in part, it may be the

hurdles that they have to overcome in getting these

agents approved.

DR. TEMPLE: I really must respond. There

is no evidence that supports what you are saying.

There may just not be any drugs around. We don't

know whether it is the difficulty. I really don't

think that is fair, and I don't think you should

say it.

DR. HUSSAIN: I have to agree with Dr.

Temple.

DR. GRILLO-LOPEZ: But he didn't raise his

389

hand. I need to rebut him, he did not raise his

hand. He jumped in and he had interrupted me once

before also in the same manner without raising his

hand and asking you for a turn.

DR. HUSSAIN: Dr. Temple, please raise

your hand.

DR. TEMPLE: Shall I repeat it?

DR. HUSSAIN: You have the floor.

DR. TEMPLE: I just don't think you can

say what the reason for the lack of interest in

prostate cancer is. I certainly don't know what it

is. In fact, if you look at the approvals that

there have been, they are not particularly

burdensome, they have not required survival for the

most part, so I just don't think you can say what

you said and know that it's true.

DR. HUSSAIN: Dr. Williams.

DR. WILLIAMS: Your question about time to

progression, I think it was a bit abstract in this

setting. Yes, we have used time to progression in

other settings with solid tumors you can measure.

One of the biggest problems in prostate

390

cancer is that we don't have time to progression,

we have time to PSA mostly because people change

therapies, and therefore, we don't, in general,

have time to progression, and I think the point

could we use time to progression is a bit abstract

unless we really develop an endpoint that we can

call time to progression, believe there is time to

progression, believe it represents what time to

progression represents in other settings, and also

can measure without 40 to 60 percent missing data.

So, yes, we have done it in other

settings, but one of the biggest problems in

prostate cancer is we don't have a time to

progression endpoint.

DR. HUSSAIN: So, for this part, I am

going to take one more response from Dr. Brawley,

and then I would like us to go to the next session,

and those of you, while Dr. Brawley is speaking,

think about what you would like to be hypothesizing

to test in the context of a Phase III trial or a

Phase II trial for that matter.

Otis.

391

DR. GRILLO-LOPEZ: Can I ask for a turn,

because he has made a statement that I have no

basis for what I have said. I have to rebut that.

DR. HUSSAIN: Then, I will give you a

moment after Dr. Brawley has done his presentation.

Yes, sir.

DR. BRAWLEY: You actually may want to

rebut me, too. I have had the opportunity to do

compare and contrast between prostate cancer and

breast cancer. Why is it that a number of the very

basic fundamental questions in breast cancer, such

as does mastectomy or lumpectomy save lives? Why

do we have the answer to that, yet, in 2005, we

still have an open question is radical

prostatectomy better than watchful waiting?

Part of the answer--and it relates

directly to this validation of a surrogate endpoint

issue--so frequently over the last 30 years, men

with gray hair have just wanted to jump to a

conclusion, and not be very scientific and not

validate surrogate endpoints, and that is why it is

really important that we finally get around in this

392

disease to finally being scientific in doing it.

One of the reasons why no drugs have been

developed, and we have studied this issue, as well,

is actually the doctor community that treated

urologic diseases in the 1970s and 1980s, or

especially early '80s, were not very friendly

toward randomized clinical trials. They and your

patients already knew all the answers, so why do

the science.

One of the wonderful things over the last

15 or 20 years is you now start having a number of

very sophisticated urologists, some of whom are in

this room, like Dr. Klein, who are designing

clinical trials.

Now, some of those clinical trials, even

today, we are having trouble getting men to go into

those clinical trials, so we can finally get the

answers. All you have to do is look at all the

cooperative group clinical trials in prostate

cancer that are not filling up with patients,

unfortunately, because so many men know what the

answers are now and don't care about their sons

393

actually getting real answers applied to them as

opposed to fake answers.

DR. HUSSAIN: I am going to give you, Dr.

Grillo-Lopez, some time to respond, but if you

don't mind being brief, so that we can get into the

second part, which is what you had been advocating

for, is to hypothesize something that we ought to

test, so if you don't mind, go ahead.

DR. GRILLO-LOPEZ: Very briefly. I am

glad to see the FDA jumping in and commenting every

time I say anything, which means that what I am

saying is important enough and/or controversial

enough to merit a response from them even if they

don't raise their hands and ask for a turn.

Secondly, I am on like seven or eight

scientific advisory boards. All of them are small

companies, but they do have three to five or more

new agents that are in clinical trials.

Of all of those, there is only one agent

that is going to be studied in prostate cancer,

because for a variety of reasons, these companies

have been dissuaded from studying their promising

394

agents in prostate cancer, and that is why they are

not here today. Only one of those companies is

represented here today.

DR. HUSSAIN: Thank you.

The second point that I think we need to

discuss before we go to early stage disease is the

issue of the fact that there are several PSA-based

endpoints that are possible. We are to discuss the

approach to select the endpoints for further study

in a prospective clinical trial.

While you are thinking about that, I

thought I will summarize what I heard from the

first part of the discussion, which was very lively

and I think very informative, in that we are all

agreeing that survival is good for metastatic

hormone refractory disease front line, but we are

also willing to entertain the possibility of

designing trials with some composite, albeit

clinically meaningful, endpoints in trials that are

powered for survival--is that a fair estimate--and

that whatever PSA exploratory analyses that there

are will need to be validated in the planned

395

prospective Phase III trials.

So, those who are from industry out there,

you have your work cut out for you. We need those

trials. I had promised that I was going to call on

Mario first and then Derek next for the issue of

what PSA endpoint to look at or whatever other

composite endpoint we want to look at for

surrogacy. Mario.

DR. EISENBERGER: I just want to again

point out that only recently we had two prospective

randomized Phase III trials showing a survival

advantage. I think all of us now are very busy

looking at the databases and come up with models

that represent reasonable hypotheses.

I was gratified to hear that the Agency

may be considering approving drugs if you look at

different models and different paradigms as long as

they are clinically relevant or clinically

meaningful.

The question is if you hypothesize

something where there isn't agreement that this

could be clinically relevant, and the trial is

396

approved for survival and possible survival and is

accrued, if you reach that endpoint early on, even

though you don't have survival data, could that

constitute reason for an accelerated approval

without demonstrating that there is a survival

advantage at this point?

DR. PAZDUR: Yes, that is the while

purpose of accelerated approval.

DR. HUSSAIN: Yes, that is what we just

said.

DR. EISENBERGER: What I did not hear, at

this point we don't have a validated model that has

shown to correlate with survival, so I am talking

about--

DR. PAZDUR: But I think, you know, that

would have to be discussed and it is something that

we would have to agree with, with the Advisory

Committee, et cetera, and this is one of the

reasons why we are holding this is, is there an

endpoint that is reasonably likely to predict

clinical benefit.

Remember, we are not asking for surrogacy

397

via the Prentice criteria here. Remember our past

accelerated approvals, they have been on response

rates that are 15 percent, 10 percent, 20 percent,

and I think in the oncology world, people could

question whether these are true surrogates, but we

have accepted these as reasonably likely

surrogates.

The other point, we are talking about

these endpoints as if they existed in a vacuum, and

not having any magnitude to them. For example, if

we took an endpoint that was a PSA nadir of a

certain value that was predefined, there is a

tremendous difference between a drug that produced

a 5 percent PSA nadir versus something that had a

90 percent PSA nadir in the population.

So, I think we have to think about that

also in making regulatory decisions and also

looking at these endpoints that are still yet to be

proven. There is a magnitude here that has to be

looked at also.

DR. HUSSAIN: Just remember this is your

chance to recommend whatever your wish list is of

398

potential PSA endpoints or other potential

endpoints that are to be put forward for the test,

so that is exactly what we are talking about here,

and I would like us to not go back to what we

discussed about approvals and otherwise.

Dr. Raghavan, you are next.

DR. EISENBERGER: Can I just say on

TAX-327 at this point we are defining a progression

model, you know, censoring was initial, and we are

reformatting the database, coming up with a

definition of a progression composite that

correlates in a multivariate analysis with

survival, and that will then be a testable

hypothesis for a subsequent Phase III trial, which

will be powered for survival.

So, if you reach a reasonable test of that

hypothesis, as I understand that could be as long

as it's reasonable, the reason for accelerated

approval.

DR. PAZDUR: A lot of it depends on the

magnitude of change here that we are seeing in that

endpoint. One thing that is dangerous about these

399

composite endpoints that we are talking about, that

have PSA as one of the composites, is the whole

endpoint may be driven by the PSAs.

You know, if you are taking a look at bone

scanning plus PSA, let's face it, that whole

endpoint is going to be driven by PSA changes, and

we are kind of fooling ourselves by calling it a

composite.

DR. HUSSAIN: Dr. Raghavan.

DR. RAGHAVAN: So, I am going to

hypothesize and run. I would suggest that we begin

to explore strategically for accelerated approval

the use of the 3-month PSA 50 percent reduction,

and I like Howard Scher's idea of multiple

endpoints, so I would add to that a 75 percent

absolute PSA reduction.

I would put in the caveat, just to remind

everyone of history, Dr. Temple and Dr. Justice

were troopers here when I was on ODAC, and we had a

very controversial drug that came to us that was

fated on two bases, wonderful PSA responses,

wonderfully high level of toxicity, and a

400

pharmaceutical company that had one pivotal trial,

and were outraged when we turned them down.

The reality of the situation was that the

turndown was based on inferior survival in the test

arm, so the divorce of survival from surrogate

endpoints shouldn't be allowed to happen, because

it is a trap.

I am totally sympathetic to Dr.

Grillo-Lopez that companies can go belly-up with

new products, but the flip side of that is

companies can make a lot of money from good

products. I personally don't lose any sleep over

the fact that all the drugs that don't work in

prostate cancer haven't been approved. I totally

agree with Bob Temple, why would we prove

ineffective drugs.

So, I would hypothesize that PSA time

dependent kinetics are worth exploring, but that

that exploration should not be divorced from a

standard that we know in hormone refractory

disease, which is survival.

DR. HUSSAIN: Just to modify what you

401

said, or to ask you the question, is it important

to bring your PSA down or is it important to bring

it down and keep it down?

DR. RAGHAVAN: I would say we need to do

both. That is why I like Howard's idea of being

flexible. I would bring it down and keep it down,

and we have some data from SWOG, preliminary, that

suggests a 3-month time point, and I have proposed

a 75 percent reduction, and I am not going to fight

anyone, if they want to make it 50 percent

absolute, that's fine. I just think 75 percent is

setting the bar a little higher as we understand

it, and I figure that is a good place to start the

discussion. But I am going to leave now, so thank

you.

DR. HUSSAIN: Dr. D'Agostino, and then Dr.

Scher.

DR. D'AGOSTINO: I may be asking out of

turn, because I want to go back to the accelerated

approval. If it was a mortality study, you have

the surrogate, a likely surrogate, you give

approval, then, you tell them to continue this

402

mortality trial. You don't stop and start all over

again.

DR. TEMPLE: But, Ralph, that is if the

accelerated approval is based on the early phase of

a trial that is ongoing.

DR. D'AGOSTINO: Exactly.

DR. TEMPLE: As you know, because they

were presented to the Oncology Committee, we have

not always done that. Sometimes you have to start

a new trial, and that doesn't always happen, et

cetera, et cetera.

DR. D'AGOSTINO: No, but this fits in

nicely with the way we are talking about a

mortality trial with surrogate endpoints built in,

and then you can move on.

DR. PAZDUR: And that accelerated approval

paradigm is commonly used in AIDS with 6-month

viral load reductions going on to 12 months.

DR. HUSSAIN: Dr. Scher.

DR. SCHER: I was just, you know--

DR. HUSSAIN: Oh, you are deferring to

someone else? Dr. Temple, sir.

403

DR. TEMPLE: No, I had a fundamental

question. There are now enough data, I would have

thought, so that one could start looking among the

trials that exist already and look at various

candidate surrogates and see, you know, with all

the flaws that this after-the-fact stuff has, and

see at least whether they predict, so you would be

able to say a 50 percent reduction, no, that

doesn't tell you anything, 90 percent reduction,

that is pretty good, that does predict.

So, there ought to be some way to look at

those right now and see which ones are promising

candidates. You can tell me I am wrong, but that

is where you would usually start.

DR. HUSSAIN: In fact, this is what I was

making a comment about. In the SWOG-9916 trial,

there is PSA data that has been analyzed. It may

turn out that different cutoffs are more in line

with the Prentice prediction of a surrogate than

others, and that is what is going to go into the

prospective validation.

As we speak, the paper has been written,

404

so the information will be scrutinized and come to

publication, but after Dr. Scher speaks, I have a

question to the FDA.

When was a response ever validated as a

surrogate by all the harsh criteria that we were

asked to comply with? Has any disease where you

accept response as a measure to approve a drug

where a response actually was rigidly, you know,

scrutinized for the Prentice criteria or any other

criteria?

DR. PAZDUR: Here again, you know the data

in prostate cancer, breast cancer, colon cancer.

There is a great deal of debate regarding response

rates and their even correlation to survival, let

alone true surrogacy.

That is why we have used those generally

as reasonably likely, and here again, I think when

one takes a look at a response rate, you know, you

have a number of complete responses, duration,

where they occur, are they associated with

symptoms. You know, it is a very complicated

issue, and it is not just in a vacuum here.

405

It is a complicated thing and there isn't

a lot of data here. You know, they are still

arguing about response rate correlation in colon

cancer with survival, the area that I am familiar

with, and a lot of this has to do with our

therapies in the 5-FU era were so meager.

DR. HUSSAIN: Howard.

DR. SCHER: I don't know how specific you

would like to be, but I would like to make the

argument that there is a durability component that

is important to the response duration. Again,

looking at the median, I would put the bar at 6

months for the response, whether it's a 50 percent

decline.

I would also add the no rise versus rise,

because you do see patients who do achieve their

nadirs beyond 6 months and some patients who

benefit who never achieve a 50 percent nadir.

Those patients would be, I would argue, devastated

if they were taken off treatment when their PSAs

were going down with no other signs of progression.

I was very encouraged by Dr. Sridhara's

406

comments that we could look at multiple endpoints

at the same time and just power the trial based on

the one that might occur most distally using a 6-

or 9-month time frame.

DR. HUSSAIN: Just for the sake of my

summary here, and I don't mean to interrupt you,

what was your first proposal?

DR. SCHER: I would add no rise versus

rise in PSA.

DR. HUSSAIN: And that is the only

proposal that you made?

DR. SCHER: And I would add one that would

include as progression, objective measures

obviously to be discussed, you know, either

clinical deterioration or change in therapy. When

we looked at our patients on first line

microtubular targeting agents, 120-odd patients

were treated, about 85 went on to second line

therapy. The median time to administration of a

second chemotherapy was 6 months.

So, that is essentially where that is

coming from and arguably, the decision to change

407

chemotherapy would suggest that the patients needed

a change in treatment. This wasn't a soft

endpoint.

DR. HUSSAIN: Just so that we don't stay

all night--unless you want to, and I am happy to

stay because I am here until tomorrow--what I

wanted to do is not miss the discussion on early

stage. I have here Dr. DeGruttola who had his hand

up and then Dr. Sandler and then Dr. D'Agostino,

and then we will move to the early stage disease.

DR. DeGRUTTOLA: I just want to say

briefly in response to the question about are there

other diseases in which there was extensive

analyses of surrogate that led to changes in FDA

policy, and I think AIDS is an example of that with

the viral load measure, surrogacy analyses were

done.

In addition, it was found that with more

potent drugs, you could drive virus below levels of

suppression, at which point immunological decline

was very much slowed down, in fact, immunologic

function definitely improved, and I would think if

408

it were possible in the prostate setting that some

people could be driven to PSA levels where

progression is really quite rare, I don't know if

that is possible, but that would be sort of

comparable to the AIDS setting.

Even if that weren't possible in most

patients, but it were possible in some, and in

those patients, they didn't progress, progression

rates are really low, then, I would think that that

would be the kind of evidence that might be

developed without needing a whole large range of

studies.

DR. HUSSAIN: Dr. Sandler.

DR. SANDLER: My question is were you

going to save some time to talk about localized

disease, and you answered it, thank you.

DR. HUSSAIN: And I just said yes.

Dr. D'Agostino.

DR. D'AGOSTINO: I just want to make sure

that because we are saying you can do three or four

endpoints, the surrogate endpoints, that that makes

it an easy task. I mean one has to be very careful

409

about how you do spend your alpha through them.

You don't necessarily just power against

the one you think is least likely, so you want to

keep us statisticians in business and make sure you

visit one.

The other is that when you move to these

surrogate endpoints, then, the visit scheduling,

and so forth, that we were talking about, becomes

very, very important. I mean you are not just

asking did the person live or die, you are asking

within a month, within a week, and so forth, what

is happening, and so it is a whole different level

in following these endpoints, and that has to be

built into the studies.

DR. HUSSAIN: Thank you. So, if I were to

summarize as to what hypotheses on the table to be

tested, it is a percent decline of PSA, and it is

your wish 50, 75 less or more, at some finite

period, percent decline of PSA at 3 months. Dr.

Scher suggested no rise versus rise plus some other

objective criteria for progression.

Are there any other suggestions or ideas

410

or thoughts or hypotheses? Okay.

Then, I would like us to spend the rest of

the time--and I would want to point out there is

one population that we have not gotten to speak

about, which I think is very important because

there may be a clear answer in it, and that is the

non-metastatic androgen-independent patients where

potentially time to development of metastases would

be a good endpoint in randomized trials, and it

will not take a million years to get.

So, if I may sneak this in and put that as

a conclusion statement and move into the early

stage disease, and open the floor for that. Mario.

DR. EISENBERGER: We don't have enough

data to give you a more solid, concrete model, so I

would like to ask you the opportunity for maybe in

the next three months to provide the data from

TAX-327.

DR. PAZDUR: Yes, that would be fine, and

here again I want to emphasize one of the reasons

why we are having this workshop is an exploration

of ideas. This is not the last time that we will

411

be discussing this whole issue, believe me, and we

will be bringing this back and hearing from you and

doing other discussions with you before our next

ODAC.

DR. HUSSAIN: So, who wants to start the

discussion? Dr. Sandler.

DR. SANDLER: Thank you. I think I would

first just like to congratulate Anthony D'Amico

because I though this presentation was terrific,

and the idea of the PSA nadir of 0.2 as being a

sign of treatment progression or the hormone

refractory state, I think is fascinating.

In terms of localized disease, something

relatively uncontroversial, I hope, and that is

that for a novel local ablative technique, such as

radiation, surgery, cryotherapy, who knows, some

new novel local technique, I think that simple

biochemical failure is an adequate endpoint for a

clinical trial.

I think that has wide acceptance in the

community, something that Rick Pazdur mentioned in

his introductory remarks, so in my mind, if I am

412

thinking of a new machine or a new radiation

technique, and I define a Phase III study, while I

would love to collect data for survival, I think

that primary PSA failure is an adequate assessment

of the efficacy of the novel technique.

I don't know whether that is worth

discussing or not, but I think since some of it may

be outside of CDER, but I think it is an important

issue nonetheless, especially since it affects how

CTEP deals with a lot of prostate cancer clinical

trials.

DR. HUSSAIN: Just so that I understand

it, and clarify it, that would be applicable to an

intervention at the prostate level modality, you

are not suggesting that, for example, for radiation

and Taxotere, to have a PSA endpoint.

DR. SANDLER: Right. I am thinking of

radiation A versus radiation B. Although I think

there is data to support the ability to use

short-term hormone therapy concurrent with the

novel therapy, and still have an adequate PSA

endpoint, for example, in the radical prostatectomy

413

trials where neoadjuvant hormone therapy was used,

there is wide acceptance that there is no

difference in biochemical failure, and that has

affected the way therapy is done in the U.S.

So, I think that a biochemical endpoint,

even with short-term hormone therapy, is adequate.

DR. HUSSAIN: Dr. Klein.

DR. KLEIN: Just a follow-up question for

Dr. Sandler. There is a lot of controversy over

how you define PSA failure for therapies that leave

the prostate in situ and I am wondering if you

could suggest to us some consensus definition that

would be appropriate.

DR. SANDLER: I think that there is a lot

of controversy, as you mentioned, in terms of the

radiation community as to when you declare someone

a failure. If someone really has cancer, they will

express it by having an elevated PSA, so the

question is not whether they failed or not, but

maybe how quickly you can call the failure and how

important the failure is.

In the radiation community, we are moving

414

from the well-used ASTRO definition of three

consecutive rises in PSA to a discrete rise of 2

nanograms above the post-therapy nadir. The

advantage of that definition is that it probably

works for both patients who received hormone

therapy and radiation and those who received

radiation therapy alone.

I think in the surgery community, there is

even some controversy as to when biochemical

failure occurs, but if you really have cancer, it

doesn't really matter whether you call it a 0.3 or

0.4 or 0.5, it will show up.

DR. HUSSAIN: Dr. Bross.

DR. BROSS: Yes, I am from the Center for

Biologics. I would like to thank the gentleman

from the open public hearing. He is the only one

who has mentioned vaccines so far, and I will say

in contrast to the metastatic indication, there are

a lot of vaccine studies, and we desperately would

appreciate advice from the Advisory Committee in

terms of what is the appropriate trial design.

At the moment, we have a number of study

415

proposals that propose to include patients with a

rising PSA, usually PSA doubling time of less than

six months, say, and the question is, is this an

appropriate endpoint.

Dr. D'Amico suggested less than three

months was more associated, and also what is the

appropriate trial population, is there a population

of patients with rising PSAs and no metastatic

disease.

That would be the easiest to study because

then you just study the time to development of

metastatic disease, and also, what is the

appropriate measurement of metastatic disease. We

don't usually accept bone scans, but in this case

it may be appropriate.

So, we would very much appreciate any

advice you can give us with respect to this

population.

DR. HUSSAIN: Dr. D'Amico.

DR. D'AMICO: I think it is important that

we stay to the topic because we have gone

completely away from the rising PSA state just

416

following surgery or radiation, and, Maha, maybe it

would be good for you to phrase the questions that

you want us to answer in that particular disease

state. Maybe they are the same questions we just

did.

DR. HUSSAIN: It's the same question as to

the PSA endpoints, but dealing with early stage

disease. In early stage disease, I think there are

two settings. There is the brand-new patient where

you are trying to maximize and/or improve local

therapy for with or without systemic treatment or

some other modality, and then there is the setting

of rising PSA post-local therapy.

Howard began making comments about the

first part, and then, of course, there is the

rising PSA. If I may make a suggestion, because

the bulk of the population that we struggle with is

the rising PSA post-local therapy, if you don't

mind that we focus on that population, and Dr.

Bross raised the question regarding, for example,

vaccines in that setting and what kind of trial

design, if, Anthony, you want to take that.

417

DR. D'AMICO: I think in the talk that I

gave, I really tried to focus very carefully,

because I don't think you can explore all types of

drug classes with one study design, and so that the

design that I put forth was one in which you take a

very unfavorable population, short doubling time.

You can justify hormonal therapy as the

conventional treatment, and then you randomize them

to the plus or minus a cytocidal agent in which the

endpoint and progression is not nadiring.

So, I think that with vaccine therapy,

this trial design would not apply necessarily,

because--again I am not an expert in vaccines, but

my understanding is that whether it's cytocidal or

cytostatic is a question.

If it's cytostatic, I don't think you

should apply such a design with a nadir construct.

If it's cytocidal, you might be able to.

DR. HUSSAIN: If I may just take the

Chair's prerogative to make a comment, I am not so

sure that the rising PSA population--and I am going

to go out on a limb to make that statement--is the

418

appropriate population for drug discovery, so that

if you have vaccines that have not shown evidence

of activity in advanced disease, to me it is wrong

to bring them into the minimal disease setting

trying to prove a point.

So, to me, this population as Anthony

suggested, is perhaps a population where we would

have some drugs that have shown some evidence of

activity in advanced disease in some form of

setting, that we know they will have a chance of

working, and then bring them into the early

setting.

Howard.

DR. SCHER: I will respectfully disagree.

I think there are patients within this cohort who

have a relatively favorable prognosis, yet, who are

extremely concerned about their doubling times and

rising PSAs, who do not want to undergo medical or

surgical castration, in whom there is an

opportunity to explore vaccines.

This is actually in some cases ideal

because you have a biomarker. In designing the

419

trial, I mean I think you can do exploratory

studies in patients with modest doubling times, we

have selected a year or two as that window.

If you are looking for a patient with a

rapid doubling time, that is not the group I would

use an experimental untested vaccine, but if you

did see some efficacy or some effect on PSA,

recognizing it has problems, a design that could be

used is a discontinuation design with hormones plus

or minus the vaccine.

So, we have actually looked at effects of

hormones, and looking at long-term survival, as

suggested, there are patients with minimal tumor

burdens who may, in fact, be cured with hormones

alone, and if you want to integrate cytotoxics, you

would have to build on those results.

So, you could then look at what proportion

of patients reach a nadir that is undetectable and

stay there as one readout depending on the level of

effect that you see.

So, I think to try to do this study

looking at what is a reasonable endpoint of

420

objective metastatic regression would be very, very

difficult as the PSAs are going up. So, you would

have to show some treatment effect first.

DR. HUSSAIN: Dr. Williams.

DR. WILLIAMS: Anthony, I think to clarify

your endpoint, because I think I understand it from

our earlier discussions, it is a dichotomous

endpoint which is measured at a certain length of

time after a patient is treated, perhaps 8 months,

so at 8 months, the percentage of patients who will

have not nadired, that is the endpoint you are

suggesting, is that correct?

DR. D'AMICO: That's correct, and the 8

months is important because multiple studies have

shown that it can take up to that long for that to

occur, and the way the study would be exactly

powered is you would use from the data available

with hormonal therapy alone, you would know the

percent of patients who would achieve progression,

that endpoint, not nadiring, and then you would say

I would like to see a 10 percent or 15 percent

difference.

421

I think having said that, it would be

important that that study looked at that as an

endpoint, but that it be powered for time to

distant disease, which is your clinically

significant endpoint, you know, beyond the rising

PSA.

DR. HUSSAIN: Dr. Sandler.

DR. SANDLER: I just wanted to comment

about the rising PSA situation after local therapy.

Usually, we are talking about rising PSA after

surgery.

Now, those patients could have distant

disease, but there is a certain subset of patients

with a rising PSA after surgery who only have

localized disease, so in the design of clinical

trials testing new systemic therapy, I think we

should be careful to either mandate or allow local

therapy, such as radiation, prior to enrollment.

I mean I think that it is potentially

unethical to treat a localized prostate cancer

patient only with an untested novel systemic

therapy.

422

DR. HUSSAIN: Dr. Martino.

DR. MARTINO: Just a basic question to

those of you who deal with this disease. Given a

patient who has had local therapy only, a

newly-diagnosed patient, be it surgery or radiation

plus or minus whatever hormonal therapy you folks

like to use, and then you are watching them and

their PSA rises, do we actually know that you

intervening at a time before there is clinical

symptomatology actually alters anything, and what

is that something that is altered, is it survival

or is it time to clinical event?

DR. HUSSAIN: If I may answer that, and

that is, there is no prospective data that has

demonstrated an impact of an intervention in a

randomized manner, so that is a fact of life.

The history of hormonal therapy in this

setting is that over and over again, not

necessarily in the rising PSA, but if you look at

all the hormonal trials historically, there is an

indication potentially that hormone therapy may

delay progression.

423

If you give it adjuvantly, you can prolong

life, but no one knows whether you intervene today,

when the PSA is 5, versus when the PSA is 100, or

when you have metastatic disease, that survival is

impacted.

So, that is an unknown, and that is the

biggest shame of our community, is that that simple

question has not been answered.

DR. MARTINO: Whether it has altered time

to clinical symptomatology.

DR. HUSSAIN: Well, clinical

symptomatology is a bit late in the process. I

think we know it delays metastases. There have

been trials looking at early versus delayed

hormonal therapy for patients who have either

upfront metastatic disease or locally advanced

disease, indicating that those who have been

treated with hormones have less odds of developing

symptomatic disease, that there is more cord

compression, bladder outlet obstruction, things of

that sort.

Understanding that this is not the case in

424

this country, patients get treated early when they

have metastatic disease, and as you heard from Dr.

D'Amico, a fair number of people in the community

are treating simply by a rising PSA.

Dr. Scher.

DR. SCHER: We don't have sufficient data

in terms of overall survival, but again looking at

long-term outcomes, the difference in survival

appears to be similar to what you see in breast

cancer populations.

So, we have actually looked at the

proportion of patients who achieve a nadir of zero

post-prostatectomy, and, yes, we included radiation

patients, in relation to whether they had minimum

metastatic disease and their PSA levels, and not

surprisingly, there is an association.

So, why not, with the availability of

cytotoxic drugs, why can't we just shift the

paradigm to try to make an undetectable PSA as our

endpoint, and ideally, if you have an undetectable

PSA, and the patient is off hormones, and their

testosterone levels are normal or back to their

425

baseline, arguably, those patients may, in fact, be

cured, and I think that is where we should be

setting the bar for the patients who have

aggressive disease.

DR. HUSSAIN: Dr. Eisenberger.

DR. EISENBERGER: I hear all of the

discussions, and I think what we need is to come up

with a reasonable set of endpoints other than

survival. Survival is not possible. That is

ultimately an endpoint where we can validate some

of our hypotheses.

I just want to point out that today, we

don't know what the long-term effects other than

toxicity of early antigen deprivation is for

patients with non-metastatic prostate cancer.

The fact is, is that we probably estimated

somewhere around 7 out of 10 men today get treated

with hormonal therapy before they develop

metastatic disease, and that is a measure issue

when we talk about designing our clinical trials,

unless you blind PSA and do it.

At Hopkins, as you know, there is a

426

generation difference. In the past, very few

patients would actually receive hormonal therapy

until the development of studies, there is a

current database, about 11 percent of our patients,

and 5,000, that Anthony demonstrated, and 900 with

high PSA relapsed. Eleven percent received antigen

deprivation treatment.

As we update these data, now, about

30-some percent of these patients are now receiving

antigen deprivation treatment, so there is at an

institution where there is a conservatism in terms

of initiating hormonal therapy, and that is

changing.

So, I don't know what the endpoint should

be, but I think that that is the critical first

step before we talk about anything is what, in an

adjuvant trial, is bone scan metastasis what we

need to use as an endpoint, is it any form of

metastasis, is it initiation of therapy using

certain parameters? I don't know. I think this is

where we need to focus a lot of our discussion.

DR. HUSSAIN: Dr. Brawley and then Dr.

427

D'Amico.

DR. BRAWLEY: I just want to make a couple

of brief comments. These are all related, yet

unrelated, as we talk about PSA rise after

treatment.

Dr. Catalano recently published a paper

that suggests that 30 percent or more of his

patients, does a radical prostatectomy on have a

rising PSA within five years of the radical.

In the prostate cancer outcome study done

by the NCI, for all comers in a large community, a

large city, it was nearly 40 percent of people who

undergo a radical prostatectomy have a rising PSA

afterwards, so the number of individuals who have a

rising PSA afterward is a considerable number of

individuals.

Many of them do get hormones, which cause

osteoporosis, and I have seen patients who have

died, not of their prostate cancer, which they were

technically cured of, but of a broken hip due to

their hormonal therapy.

Also, the prostate cancer prevention trial

428

data, which Dr. Klein knows probably better than

anyone, that study screened a large number of men

in their 60s for 7 years and diagnosed 12 percent

of those men with prostate cancer due to screening

and then said the hell with screening and biopsied

everybody who had a normal PSA for 7 years and

found that 15 percent of those men had prostate

cancer.

So, of this 26, 27 percent of all men in

their 60s who have been diagnosed with cancer, the

NCI, through Rocky Foyer's [ph] data, indicates

that 3 percent of them will die from prostate

cancer.

So, with screening, we can diagnose 12

percent of men with prostate cancer, only 1 out of

every 4 for whom will ultimately die from the

disease, but I worry about the guys who get

radicals and have a rising PSA, and that rising PSA

is actually not a threat to their life.

So, a survival study using that group of

people actually might be fraught with some dangers.

The last thing I want to note is I talked

429

about the new breed of academic urologists and I

forgot to mention Dr. Andriole who is over there.

DR. HUSSAIN: Dr. D'Amico.

DR. D'AMICO: Mario's point that even at

Johns Hopkins where hormonal therapy was withheld

before a positive bone scan, now went from 11 to 30

percent of people having that, sort of documenting

what I said, and that is, even in now academic

centers, fastly rising PSAs, people go on hormonal

therapy, so I think it justifies that in a control

arm of patients with very rapid rises in PSA, i.e.,

short doubling times.

To Dr. Scher's point, this is an important

one. An undetectable PSA, this is really important

that you get this, an undetectable PSA, after

hormonal therapy following surgery or radiation,

does not mean you are in the clear. You can still

recur and die of prostate cancer. That is what

those curves showed you. Guys who nadired, 15, 20

percent of them still died of prostate cancer, but

a detectable PSA is always bad.

So, I am arguing that the endpoint should

430

not be undetectable, but detectable. It is just

the reverse. It talks about the endpoint being a

progression endpoint. A statistician would say

this is an event, not a non-event, this is an

event, don't get undetectable PSA, it's an event,

and that is always bad, so I think that that is an

important distinction.

To Otis' point, I think your point about

people getting hormonal therapy and dying of the

side effects of treatment is an important one,

osteoporosis, neurocognitive issues, QT, and so on.

That is why the long doubling time patients should

go on studies with vaccines and things that are not

involving a treatment that might impact their

longevity, and the short ones, though, don't die of

side effects, they die of prostate cancer.

But I have one more thing, Otis, that you

will find interesting. The 3 percent number,

one-third of people sustain PSA failure, a fifth in

the community have a short doubling time. That is

one-third times one-fifth is one-fifteenth. That

is 6 percent.

431

I showed you that of the guys with the

short doubling time, a third of them don't nadir.

One-fifteenth times one-third is one-45th. That is

2 percent, 2 1/2 percent. There is your 3 percent.

So, the people who die from prostate

cancer, the 3 percent number can come from PSA

failure, short doubling time, don't nadir, that is

3 percent, and I think that's it, and I think it's

that simple.

DR. BRAWLEY: I am agreeing with that.

DR. HUSSAIN: Okay, and I am glad we are

one big happy family.

Dr. Klein. I am going to probably ask

afterwards do we want to wrap it up, or do you have

any other burning questions, the FDA? No. Then, I

will summarize and then people can object to my

summary or agree with it, and we will go from

there. Dr. Klein.

DR. KLEIN: So, to focus on the

post-treatment rising PSA population due to Dr.

D'Amico's work and all the people who collaborated

with him, we have more data on PSA as a predictor,

432

as a surrogate loosely used, than in any other

disease state in prostate cancer, and we ought not

ignore that.

I would put my vote with the suggestion

that PSA doubling time be used as a stratification

or selection criteria in this population, and that

nadir versus not be used as the response criteria,

and at least as an initial pass, and we should move

on that quickly. We don't need more preliminary

data on this, it has all been done.

DR. HUSSAIN: Dr. Pazdur.

DR. PAZDUR: Here again, you are using it

as a prognostic factor. The question that I have,

one of the areas that we have looked at is

basically in order to verify a surrogate or even a

correlate, you have to have an effective therapy in

the disease, and that is what enables us now to

take a look at hormone refractory disease, because

we could look at the Taxotere studies, for example,

and we could develop new drugs in that area.

But in this PSA rising situation where you

have no therapies that have demonstrated an impact

433

on a clinical endpoint--

DR. HUSSAIN: Because they have never been

tested.

DR. PAZDUR: Okay, but you don't have

them, so how can you then verify any surrogate

endpoint here, and I guess that is a question for

Dr. D'Agostino, can you, in the absence of an

effective therapy, really--I am using the word

verify loosely, because I don't want to use the

word verify a surrogate, but even look at a

correlation even.

DR. D'AGOSTINO: Are we talking about

where we are going to do a mortality study or some

endpoint?

DR. PAZDUR: There is no tie to what you

are doing to a clinical endpoint here in this PSA

rising, is that what you said--validate it, to

validate it.

DR. HUSSAIN: This is the rising PSA, Dr.

D'Agostino, where those patients, if you take them

as a lump sum, a good number of them are not likely

to die from their disease, but as Dr. D'Amico

434

pointed out, there is a subset based on sort of

retrospective PSA data analysis, that you might be

able to predict that this is the subset that has a

shorter survival.

But I think what Dr. Pazdur is asking,

since all of that is retrospective, how are you

going to then design a trial with an endpoint that

can be reached before we all retire, and validate

at the same time whatever hypothesis you have about

a doubling time, or a nadir PSA, or any other

criteria.

DR. D'AGOSTINO: I think you would have to

go back to something like Bob was suggesting

earlier, that you try to get a reading on the

increase and then randomize things of that nature

that you have to be able to sort of set a baseline

and then move on.

DR. KLEIN: In this population, I think

you would have to use a time to progression

endpoint, recognizing the imperfect definition. No

one will ever stay on a study long enough or, as

Dr. Scher pointed out, detectable versus

435

undetectable PSA. That triggers additional therapy

in the community, and the clinical benefit is no

additional therapy.

DR. HUSSAIN: Dr. Scher, you had your hand

up a moment ago.

DR. SCHER: I reiterate that I think the

undetectable PSA with a normal testosterone is a

good start to suggest efficacy. In the minimal

disease settings and other tumor types, there is a

cure rate, so we can ask the question appropriately

using the prognostic models that have been so well

described now, and use them in trials.

DR. HUSSAIN: So, let me ask you this,

though. That would not apply if the patients have

gotten hormone treatment, and that you have to set

your clock for looking at a PSA relapse point or

undetectable point from the point in time where the

patient's testosterone recovered. That, by itself,

adds another magnitude of weight.

DR. SCHER: We have designed the study

which will enroll patients with doubling times of 6

months or less using the endpoint of an

436

undetectable PSA at 3 years, accounting for the

fact that about 20 percent of the patients will not

recover their testosterone levels, and the total

duration of hormonal exposure will be 18 months.

So, it is a 3-year endpoint for the trial,

and we will hopefully be opening shortly.

DR. HUSSAIN: The last comment is Dr.

Eisenberger.

DR. EISENBERGER: Again, I do feel that

there are data to suggest that certain PSA changes

may, in fact, be a reasonable endpoint at some

point in the future.

What we don't have at this point is any

correlation between any of these PSA data in a

prospective fashion with more conventional

endpoints, such as time to progression, for

instance. This what we need.

I think what we need is a focus here

today, is what does that constitute at this point

in time, and not what a 6-month PSA or nadir PSA is

following therapy. What are we going to validate

that against? That is what you want to know at

437

this point in time.

DR. HUSSAIN: Just because I said that,

ladies go, so that's fine, please.

DR. McSHANE: I think you raise an

excellent point and just so that we are not all

here again in five years still debating these

issues, what can we do now to get the data that we

need, so that we won't continue to debate.

I think Dr. Martin from the NCI suggested

some very good possibilities. We have several NCI

trials. They are ripe for putting in these

surrogate endpoints now or these potential

surrogate endpoints now, but if we don't decide on

what kind of schedule we are going to measure the

PSA or at least collect the specimens, so that PSA

or something else could be measured, we are going

to end up with data that we can't compare across

studies and we will still be scratching our heads

in five years.

So, I think it would be very beneficial if

this committee could--and I know we are running

short on time--could spend just a few moments

438

discussing, you know, if we could have whatever we

wanted, what would we do in the way of measuring

PSA or collecting specimens, should we collect it

on every trial, only in certain kinds of patient

groups, should we collect it every 3 months, should

we collect it every week, you know, what can se do,

so that we are not still debating this issue.

DR. HUSSAIN: Do you want to go a little

bit more or you want to wrap it up and this would

be the subject of future discussions? Dr. Pazdur,

it is 5:05.

DR. PAZDUR: I guess what I would ask is

the committee's opinion, would they rather go on

with further discussions or wrap things up, because

I know we are losing some people because of

flights.

DR. HUSSAIN: If I may suggest, I think we

have accomplished a fair amount. I do think it's a

good idea to have us all digest everything that got

said.

If I may wrap up this session dealing with

a rising PSA population and the local disease, I

439

would say the local disease, there is really no

consensus or plans at this moment. The rising PSA,

what has been put for discussion is perhaps one of

two possibilities, as was suggested, begin looking

at some validation of certain surrogacy endpoints

and the ongoing trials or the planned trials for

the rising PSA population.

The other alternative is to bite the

bullet and say we have enough data on from

retrospective series showing that certain doubling

time or some PSA kinetic is likely to predict for

poor prognosis patients and begin then targeting

those patients for clinical trials. Is that a fair

assessment?

Mario, I am going to have to cut the

discussion.

I want to thank the Committee members for

a wonderful discussion. Before I end, I want to

thank especially the public, particularly patients,

patient advocates, patients' families, those who

are interested and concerned about prostate cancer.

I want to thank you all. Please know that

440

we are all here because we have patients' interests

at heart, no other real issues, and thank you very

much.

[Whereupon, at 5:11 p.m., the meeting was

adjourned.]

- - -