1

DEPARTMENT OF HEALTH AND HUMAN SERVICES

FOOD AND DRUG ADMINISTRATION

CENTER FOR DRUG EVALUATION AND RESEARCH

ADVISORY COMMITTEE FOR PHARMACEUTICAL SCIENCE

CLINICAL PHARMACOLOGY SUBCOMMITTEE

Wednesday, November 3, 2004

8:05 a.m.

Hilton Washington, D.C. North

620 Perry Parkway

Gaithersburg, Maryland

C O N T E N T S

PAGE

Call to Order 3

Conflict of Interest Statement 5

Update on Previous Meeting Discussions/

Introduction to the Topics of This Meeting 7

Topic 1: Pharmacogenetics of Irinotecan

Introduction 30

Scientific and Clinical Evidence 49

Current and Future Perspectives on

Irinotecan Pharmacogenetics 76

Clinical Utilities of Genotyping 113

Committee Discussions and Recommendations 143

Open Hearing 166

Topic 2: Drug-Drug Interaction Concept

Paper: Issues Related to CYPs, Transporter-

and Induction-based Interactions and

Multiple Inhibitor Drug Interaction

Studies

Conflict of Interest Statement 206

Relevant Principles on Drug Interaction

Concept Paper 209

A Scientific Perspective 255

Induction-Based Interactions 269

Multiple Inhibitor Studies 323

Committee Discussions and Recommendations 345

P R O C E E D I N G S

[8:05 a.m.]

DR. VENITZ: Okay; good morning, everyone.

Welcome to the Clinical Pharmacology Subcommittee

meeting. We have a full agenda for today, and

before we proceed with our agenda, I would like to

go around the table and for every person sitting on

this table to introduce him or herself, please.

Gerry, do you want to go ahead?

MR. MIGLIACCIO: Gerry Migliaccio,

vice-president, global quality operations, Pfizer.

DR. BLASCHKE: Terry Blaschke, Stanford

University.

DR. BARRETT: Jeff Barrett, Childrens

Hospital, Philadelphia.

DR. CAPPARELLI: Edmund Capparelli,

University of California, San Diego.

DR. DAVIDIAN: Marie Davidian, North

Carolina State University.

DR. DERENDORF: Hartmut Derendorf,

University of Florida.

DR. GIACOMINI: Kathy Giacomini,

University of California, San Francisco.

DR. HALL: Steve Hall, Indiana University

School of Medicine.

DR. JUSKO: William Jusko, University at

Buffalo.

DR. VENITZ: Jurgen Venitz, Virginia

Commonwealth University.

MS. SCHAREN: Hilda Scharen, FDA Center

for Drugs, executive secretary.

DR. MCLEOD: Howard McLeod, Washington

University.

DR. SADEE: Wolfgang Sadee, Ohio State

University.

DR. WATKINS: Paul Watkins, University of

North Carolina.

DR. RAHMAN: Atiko Rahman, FDA.

DR. WILLIAMS: Grant Williams, oncology

drugs, FDA.

DR. PAZDUR: Richard Pazdur, oncology

drugs, FDA.

DR. LESKO: Larry Lesko, Office of

Clinical Pharmacology and Biopharmaceutics at FDA.

DR. VENITZ: Thank you, everyone.

Our next step is to review the conflict of

interest statement, and Ms. Scharen is going to do

that for us.

MS. SCHAREN: Good morning. 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 subcommittee

participants, it has been determined that all

interest in firms regulated by the Center for Drug

Evaluation and Research present no potential for an

appearance of conflict of interest with the

following exceptions: in accordance with 18 USC

208(b)(3), the following participants have been

granted waivers: Dr. Paul Watkins has been granted

a waiver for consulting with the sponsor and a

competitor on unrelated matters. He has received

less than $10,001 per year from the sponsor and

between $10,000 to $50,000 per year from the

competing firm.

Dr. Kathleen Giacomini has been granted a

waiver because her spouse is a member of the

speakers bureaus for the sponsor and a competitor.

He lectures on matters unrelated to the issues to

be discussed at this meeting. He receives less

than $10,001 per year from the sponsor and between

$10,001 and $50,000 per year from the competing

firm.

Dr. Edmund Capparelli has been granted a

waiver for unrelated consulting for the sponsor.

He receives less than $10,001 per year. A copy of

the waiver statements may be obtained by submitting

a written request to the agency's Freedom of

Information Office, Room 12-A-30 of the Parklawn

Building. In addition, Dr. William Jusko has been

recused from participating in this portion of the

meeting.

We would like to note that Dr. Paul

Fachler is participating in this meeting as

nonvoting industry representative acting on behalf

of regulated industry. Dr. Fachler's role in this

meeting is to represent industry interests in

general and not any one particular company. Dr.

Fachler is employed by Teva Pharmaceuticals.

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

interests of fairness that they address any current

or previous financial involvement with any firm

whose product they may wish to comment upon.

Thank you.

DR. VENITZ: Thank you, Hilda.

Our first agenda item is Dr. Lesko, who is

going to bring us up to date on the outcomes of our

previous meetings and who is going to set the stage

for the next day and a half.

Larry?

DR. LESKO: Thank you, Dr. Venitz and good

morning, everybody, and welcome to our fourth Clin

Pharm Subcommittee meeting of the Advisory

Committee for Pharmaceutical Sciences. And I'd

like to say that we've been looking forward to

today's meeting, and I think we have three

interesting topics that we will be looking for your

input and discussion of as we move forward with

these particular areas.

Following the first couple of meetings,

there was some interest in sort of stepping back

and reflecting and recapping on some of the topics

that have been previously presented to the

subcommittee and in particular to reflect upon the

value of the meeting in terms of what FDA has

accomplished with the input from the committee.

And what I'm going to do now is summarize the

topics that we've discussed at prior meetings along

with some of the status of the projects that we've

brought before the Committee.

Let me first say that again, we have some

new members on the Committee, so this will be very

helpful, I think, for those individuals, but this

Committee was established in May of 2002, and in

putting the Committee together, we selected

individuals with very specific expertise in broad

areas of clinical pharmacology that was

cross-applicable across many of the therapeutic

areas that clinical pharmacology deals with, and

those three areas were pharmacogenomics,

pharmacometrics and pediatrics, three broad areas,

but as we've seen with the past meetings, various

subtopics that are of interest to clinical

pharmacology. The three prior meetings occurred in

October 2002, April 2003 and last November.

I'll begin with some of the topics that we

covered during these meetings, and you'll see that

the topics were not confined to one or another of

the advisory committees. We've used our meetings

in a continuous fashion as the projects unfolded,

and the first topic that we actually brought before

the Committee was a methodology for identifying

patient subgroups at risk for toxicity. These are

the subgroups that represent specific or special

populations such as those with renal impairment.

And what we did in the early meeting was

propose a quantitative method that was based upon a

number of features that we thought would be

beneficial, not only the mean exposure of the drug

in test populations such as the renal impaired

population and reference populations like the

healthy volunteers, but also, we proposed a method

that looked at the distribution of exposure values,

and then, from that distribution and comparing

those two distribution curves, identified a

critical cutoff value at the high end of the

distribution curve based on the exposure response

relationship.

What we are trying to get at here is a

cutoff value above which the risk of toxicity was

unacceptable from a clinical perspective. In

addition, we showed how we could calculate the

probably of a clinically significant response

beyond that cutoff, and we proposed a standardized

decision tree for dosing adjustments.

The summary points of our discussion was

that we linked population PK with clinical outcomes

through examples with unresolved questions. We

discussed exposure response methodologies, using

modeling and simulations of adverse event

probabilities through drug-drug interactions. We

also discussed in some of our earlier meetings

decisional analysis based on exposure response

methods for assessing QT risk in special

populations.

These were intended to be examples of the

methodology, and inherent in those examples was

some methodological questions and issues that we

brought before the Committee for discussion.

So what did all of this lead to? Well,

the status of this project was that we've currently

implemented this methodology in our NDA reviews.

The methods we proposed to the Committee or a

variant of them as we went through the process are

routinely used in the quantitative analysis of

exposure response data for efficacy and safety.

The primary impact of the topic and the Committee's

recommendations have been for us in the office, we

use these methodologies and recommendations to

formulate our dosing adjustments that we recommend

for inclusion in the package insert or in the

product label.

So we really went to several methods then

from our discussions here, selecting each of them

on a case-by-case basis, depending on the question

and the issues. That was the identification of

patient subgroups at risk.

We also brought another methodology that

was intended to do basically the same thing, and

that was the utility function for optimizing dosing

strategies. The summary points associated with

this topic was that we had proposed the utility

function as a methodology based on the probability

of either an adverse event or the absence of

toxicity taking into account the magnitude of harm

if the adverse or toxicity occurs.

We worked on this project for some period

of time, and the status at the moment is that we've

postponed further development, not that it wasn't a

worthwhile project, but the underlying approach was

difficult for us in implementation. Underneath

this approach was assigning relative weights to the

value of the efficacy versus the value of the

toxicity, each of which can range from marginal to

significant, and thereby defining a therapeutic

index for the drug.

One of the ways you define these

endpoints, if you will, upper and lower limits of

acceptability and the relative benefit-risk is to

ask clinicians, which we did. We also searched

literature and looked for applications, and we

found that the approach for our purposes in

regulatory decision making was unsatisfactory

because of the difficulty in defining targets and

penalties for different measures of the utility

function.

That being said, the method certainly has

merit, and we have seen this in terms of drug

development. It certainly has merit in its

application to the selection of doses to be used in

clinical trials during the drug development

process, and we know of examples where this is, in

fact, done by sponsors. But for our purposes, at

this point in time, we have not been able to

implement it as a regulatory, quantitative tool for

decision making.

The next topic that we had before the

Committee was using exposure response relationships

in the pediatric decision tree, which is an

appendix to our exposure-response guidance that was

released to the public in April of 2003. What we

had is summary points from these discussions, which

cover two of our meetings, was a proposal for the

design of a pediatric database to effectively

extract new knowledge from the in-house studies.

This was a data mining exercise, so that we could

use the information to update our pediatric

decision tree, which right now, is used

conventionally across our therapeutic areas.

We asked the Committee to comment and

recommend the highest priority questions or queries

from this database assuming that we could establish

it, and some of the things we presented to the

Committee using the database at the time was a

model for pediatric clearance in order to predict

it, which took into account age, adult PK and

metabolism.

We subsequently proposed a systematic

pediatric research project that was fairly

ambitious. We wanted to evaluate the trends in

exposure response with age, using the information

that we had in house. We wanted to develop a

standardized approach for use across therapeutic

areas for population PK studies, and we wanted to

develop a computer-aided pediatric template for

study design that we can use during the IND process

in designing studies in collaboration with a

sponsor.

So the status of this project, following

our deliberations at the Committee, is that it's

ongoing. The progress on the database itself has

been limited for a variety of reasons. We had

difficulty accessing data in our files because of

the nonuniformity in the way data comes in. Some

of it is electronic; some of it is manual. It

became a laborious process to assemble this data,

and it's still in an ongoing mode.

The other issue that we found in mining

our database was the availability of standard PK

and PD information. It was heterogeneous. We

could not easily take everything that we had

received in the files and assemble it into a

database that would be consistent across the

submissions. So it was a major work effort for us

to begin, but we have begun, although it's not a

complete project by any means.

We had proposed to the Committee a

pediatric research project, as I said, and this was

funded by CDER in June. I have June 2003. It

actually is 2004. Last couple of months, we

received money from the Center to fund this

project. It's being headed by Dr. Peter Lee, and

we've just begun to get going on the project.

We've hired four scientists under a contract.

We've established a steering committee for this

research. It has commenced, and we have some

12-month milestones.

So the input, the project are all ongoing,

and we're looking forward to sharing the results of

that project with the Committee as we move forward

into next year.

The next topic that we covered in the

early meetings was the genetic polymorphism of

TMPT. The summary points from this discussion

included a presentation on the scientific and

clinical evidence that linked three different TMPT

genotypes with the incidence of myelosuppression.

What we discussed was a general framework for

consideration of analytical validation, clinical

validity and clinical utility for improving

benefit-risk and pharmacogenomics.

The third summary point was the discussion

that we had in front of this Committee related to

the revision of the label of 6-mercaptopurine that

would include dosing adjustments based on genotype

and the more rich information on what we know to be

the case with regard to polymorphism of TPMT.

The status of this project is that with

the input of this Committee and our Pediatric

Oncology Subcommittee, the project is in essence

complete. Both committees, if you recall some of

the discussion that we had at this Committee,

recommend a revision of the label of 6-mp to

include TPMT information in various sections of the

label. Negotiations with the sponsor of these

products are basically complete, and the updated

label for both of the thiopurines will be available

in early 2005.

The next topic that was really a new topic

back in April of 2003 was our evaluation and

labelling of drug interactions of NMEs, an

important topic because we were just beginning the

initiation of the revision of our in vitro and in

vivo guidances for industry on drug interactions.

And a summary of what we presented at the Committee

was an in vitro drug interaction decision tree for

CYP enzymes and associated label language that

would go with that decision tree. We discussed

some of the scientific basis for policy decisions

related to NDA review, label language and class

distinctions for drug interactions, and we

discussed some specific drug-drug interaction

studies involving transporters, specifically PGP

and, by extension, some of the other transporters

that are sort of on the front edge of the drug

interaction area.

The status of this topic is that it's

complete in many ways, although we have a little

bit more work to do, but the revision of the

guidance, which was the process that was behind the

topic we brought before the Committee is nearly

complete. The working group has been working on

this for some time, and we're getting close to

finalizing that guidance, which would be an update

of our current in vitro and in vivo drug

interaction guidances.

Furthermore, the topic that we've

discussed here has been included as a topic and

discussion point in the office's GRP drug-drug

interaction map and cross-labeling map, so again,

we try to transfer the knowledge and information

that we've learned through this Committee to

day-to-day practice in terms of IND and NDA

reviews.

A year ago, we introduced another new

topic. It was the end of phase 2-A meetings, and

we had a very useful discussion on the topic. We

had as a background, if you remember, the concept

paper on the end of phase 2-A meeting, and what we

presented was the principles of the concept, and we

received again a significant input on the goals,

the process, the obstacles and the metrics of

success of the end of phase 2-A meeting.

With regard to the concept paper, we have

worked on the development of a guidance for

industry on the end of phase 2-A guidance. We

anticipate this guidance will be a final guidance,

in that it's not necessarily a controversial

guidance. We like to get it out fairly soon.

However, the status is ongoing, and over

the past year, we've had at least four significant

end of phase 2-A meetings. These had to do with

the questions that we had in the concept paper.

They involved a fair amount of modeling and

simulation. In one case, we have a disease state

model that came out of the meeting that was very

useful for simulating phase 2-B and 3 trials. And

by all indications, these meetings have been a

success, both by comments we've received from the

sponsors and by comments we've received from the

medical divisions with whom we coordinate these

meetings.

So we're very optimistic about this

process as a so-called critical path activity that

has the potential to impact the efficiency, the

informational content of the drug development

process.

As I say, the deliberations were very

helpful to us in writing a draft guidance for

industry on the end of phase 2-A meeting. It's

undergone internal review, and for all practical

purposes, it's complete. There is a process to

release a guidance, and that would probably take us

through the first quarter of 2005, when we make

that guidance public.

Another topic we discussed before the

Committee was the quantitative analysis of QT. The

summary points that we presented to the Committee

was some approaches using modeling and simulation

and also metrics for assessing QTC interval

prolongation. If you recall, the metrics that we

talked about were pros and cons of maximal change

from baseline area under the QTC time curve, et

cetera. And we asked for input from the Committee

on these methodologies that we could begin to apply

in the review of QT studies within the NDA

database.

Status of this is still ongoing. There's

a lot of current discussion on standardization of

both study design and data analysis of these kinds

of studies. We've made recommendations and

presentations that have stemmed from our discussion

here at the Advisory Committee to the CDER QT

working group, who was favorably impressed by what

we delivered in terms of a quantitative approach to

assessing the risk of QT.

We also discussed drug interactions

involving somewhat unrecognized and

underappreciated potential drug interactions

involving CYP2B6 and 2C8. The summary of our

presentation and discussion here at the Committee

was that we sort of took an inventory of our

current understanding of inhibition reactions in

particular that are based mechanistically on the

2C8 and 2B6 pathways.

We discussed some of the reliability of

the in vitro and vivo associations of these drug

interactions, similar to what we do for the more

common CYP enzymes to see to what degree these in

vitro studies can be a guide to the need to do

clinical studies, and if we do clinical studies,

what are the model drugs? So we did present some

examples to the Committee with model drugs, and

I've listed a few of them there asking for comment

on the methodology and the use of the information.

So the status of this project is ongoing.

The input was seriously considered in the context

of our CDER working group on drug-drug

interactions, and there's probably a good chance

we'll be discussing more of this in subsequent

Committee meetings.

So anyway, that in a nutshell is what

we've brought before the Committee as topics. I

think you can see how they fit into those three

broad areas of pharmacogenetics, pharmacometrics

and pediatrics. We've branched out into the

drug-drug interaction area as the need arose for us

to bring this to a public discussion.

So my reflections on the first three

meetings as we move into our next one is that the

topics we've brought before you as an advisory

committee have been challenging. We recognize they

have been diverse. They've been as diverse as the

expertise of the membership.

Just so you appreciate how we bring topics

to the Committee, we try to select topics that are

relatively new and important to NDA reviews such as

the quantitative methods. You've noticed that they

are not usually drug-specific, because we bring

general topics that are cross-applicable across

many therapeutic areas. We've brought topics to

the Committee that I think are cutting-edge

science, the drug interaction area in particular

with transporters and some of the new CYP enzymes

are areas that we have a lot of issues to resolve

in terms of what we would recommend to drug

sponsors and their drug development programs.

And finally, some of the topics we've

brought here had an element of controversy, because

they were new, and I would say in our

pharmacogenetic area, we've had a lot of good

discussion and clarity about the integration of

pharmacogenetics into product labels and into the

assessment of benefit risks.

So in short, a compliment to the

Committee. The value of this Committee has been

tremendous. I think it's the only committee that

has dealt with those topics. It has given us

significant guidance on decisions we have to make

in terms of the specific areas that we've brought

forward, and it has had a very significant

influence on our clinical pharmacology program at

FDA.

And finally, many committees do vote.

Usually, this is characteristic of committees in

which specific drugs are brought forward for voting

on one issue or another associated with that. We

haven't done that very much in this Committee. The

nature of our topics really haven't lent themselves

to voting, because they are more general. We

anticipate we will be doing more of that, including

some of the topics that we will bring before you

today, but the primary benefit is not the voting,

necessarily; the primary benefit that we've

received is the copious notes that we've been able

to take and the benefit that we've had from the

discussion of the Committee.

So for all of this, I would thank you for

your service to the FDA and service to the public

as members of this Committee and, frankly, look

forward to further very interesting discussions

with you all.

I will pause at this point, and it looks

like if there's any questions, I'd be happy to

answer those. I'll turn it back to the Chair

before I move on to a specific topic.

DR. VENITZ: Thank you, Larry.

Any comments or questions by Committee

members?

I'm interested in the end of phase 2

status. What are the metrics of success that

you're considering right now to evaluate whether

this program is going to be a success?

DR. LESKO: Yes, the metrics of success

really have been a questionnaire that we're

preparing to send to the company. We also call the

company to try to get feedback on what we did as a

process and what recommendations we gave them in

terms of value. We interact both with the clinical

group at sponsors. We work with the

biostatisticians and the clin pharm folks as well

as the regulatory folks at the various companies.

We also survey the medical division that

we coordinate the meeting with to see if what we

brought to the table in terms of the quantitative

methods was perceived to have value, and then,

finally, we have, in one form or another, a

debriefing of the internal team that worked on that

preparation for the end of phase 2-A meeting to get

back into lessons learned and see what we can do

better.

What has it all meant? I think we need to

have more experience with it, but as I said in my

remarks, I think the feedback we've gotten both

from the company and the internal participants has

been very encouraging for us to go forward. These

have not been small efforts. We're somewhat

overwhelmed by the effort that goes into preparing

for these meetings in the short time frame that we

have, and it is very resource-intensive, so it's

important to us to actually get good metrics of

success, and we're going to be collecting those,

and I'd like to share that, maybe, with the

Committee at some point in time.

DR. VENITZ: And I was exactly going to

encourage you to do that, because as you remember,

the discussion that we had, even though the

Committee was very much in favor of this

initiative, there was concern about resource

allocation and so on, so you really have to

demonstrate that there is a value in doing this.

DR. LESKO: Exactly, and I remember those

comments exactly. And there was no rebuttal to

those. We just had to get into it and try it, and

indeed, we have gotten into it and tried it. But,

you know, aside from what the impact was on drug

development, the impact on us in FDA being able to

work together in a quantitative way to discuss drug

development and benefit risk in a way that you can

put on the table in terms of a model and do a lot

of what-if scenarios has really been good for us,

and I think it's, again, made us a stronger group

within the agency aside from whatever impact it had

on drug development.

It's oftentimes spoken about in the

context of critical path now, which came out this

past March of one of the leading initiatives of the

critical path project that has the potential to

influence in a positive way the drug development

process so--

DR. VENITZ: Thank you. Any other

comments, questions?

[No response.]

DR. VENITIZE: Then why don't you proceed

with the topics for the next day and a half?

DR. LESKO: Okay; so, now, switching to

the topics for the next day and a half, we're

bringing three projects, topics, to the Committee.

The first of those this morning is going to be in

the area of pharmacogenetics. Specifically, we're

going to be discussing the scientific and clinical

evidence that surrounds the UGT1A1 polymorphism and

its relationship to the pharmacogenetics of

Irinotecan.

This afternoon, we're going to bring to

the Committee a topic in the area of drug-drug

interactions. We'll be talking about metabolism

and transporter-based interactions, and again, this

is a relative topic to the revised guidance for

industry on the on-drug interactions. And thirdly,

we're going to bring to the Committee a topic from

the world of pharmacometrics, but it also is from

the world of the critical path.

We're beginning to focus on specific

critical path activities that we would like to

advance, and tomorrow, we'll be talking about one

that deals with the greater use of biomarkers

within the context of drug development and their

systematic progression to potential surrogate

markers. We'll primarily be talking about the

project and the project plans in the latter area

but not necessarily on any specific biomarker in

any given therapeutic area but a more general plan

that we hope to get input on as we move forward

with it.

So those are the three topics, and I think

what I'll do now is really launch into the first

topic, but I want to pull up the slides for that.

So the first topic of the morning is the

pharmacogenetics of Irinotecan. And we'll be

talking about the scientific and clinical impact of

UGT polymorphism. And my role up here right now is

to present a background to the topic and then turn

it over to the individual presenters that we've

scheduled for the morning prior to the discussion

with the Committee.

I'll start out with the labeling

regulations that apply to both new and to

previously-approved drugs, and the labeling

regulations are that if evidence is available to

support the safety and effectiveness of the drug

only in selected subgroups of the larger population

with the disease, the labeling shall describe the

evidence and identify specific tests needed for

selection and monitoring of patients who need the

drug. Obviously, this is not pharmacogenetic

specific but certainly I think encompasses

pharmacogenetic testing and information.

Pharmacogenetic information on drug labels

in general is not anything new. There's no current

barriers to including this information in product

labels. Many of you, I'm sure, are familiar with

examples of Herceptin, which is probably one of the

most well-known examples of pharmacogenetic

information on product labels. That is one example

of where a test and a drug therapy are used in

conjunction with one another.

If one were to survey the PDR and look at

package inserts over the years, about 35 percent of

the approved drugs have pharmacogenetic information

in the label. That doesn't necessarily mean that

that information is clinically important or

clinically relevant. Much of it is descriptive,

and much of it doesn't really translate into what

physicians would do in clinical practice having

that information in hand.

On the other hand, this is reflecting many

years of pharmacogenetics in terms of the

well-known biomarkers of cytochrome enzymes, and

it's only now that I think we're beginning to see

evidence that some of these enzymes are clinically

important and ought to be considered more seriously

in dosing of approved drugs.

A couple of other examples: Thioridazine

is a previously-approved drug, and if you look at

the package insert for that, there's a black box

warning in there that warns physicians and patients

that they ought to avoid this drug in 2D6 poor

metabolizers because of the toxicity risk

associated with poor metabolism, and Thioridazine,

of course, is a 2D6 substrate.

Atomoxetine is an example of a relatively

new drug for attention deficit disorder. There was

information from the clinical development plan

about the relationship between pharmacogenetics,

specifically poor metabolizers and extensive

metabolizers and clinical outcomes in terms of

efficacy and safety. This was information that

both the sponsor and the agency agreed is

worthwhile to put in the label, and in fact, we

included it in the label in seven different

sections, ranging from the laboratory test section

to the clinical pharmacology section. However,

there was no reason to require or suggest that a

test would have to be done prior to using the drug.

Cetuximab is a drug that was approved that

includes genomic information and particularly tumor

genomics about receptor positivity. This drug is

an EGFR inhibitor, and there's some general in

there about the relationship between the drug and

its pharmacology and its receptor positivity from a

mechanism standpoint. Again, there was no

recommendation to require a test prior to

prescribing the drug.

Finally, as I mentioned in my opening

remarks, we have discussed before the Committee

6-mercaptopurine and azathioprine, and these

labels, for all intents and purposes have been

updated to include the polymorphism information in

multiple sections. That was deemed important.

With regard to the outcome of the 6MP and

TMPT polymorphism, in many ways, it was a step in a

general framework for assessing the

pharmacogenetics of approved drugs. We are trying

to create a structure for that type of discussion

as we look at approved drugs that may benefit in

terms of the inclusion of pharmacogenomic

information, so we start out in this case with the

absence of information in the label, which was then

discussed at the Clin Pharm Subcommittee as well as

the Pediatric Oncology Subcommittee.

The new labeling, as I said, has been

revised in consultation with FDA and includes, now,

data on increased risk of severe myelosuppression

associated with genotypes of TMPT. So we think the

label has been updated with useful information for

clinicians and patients as they weigh their options

of using TMPT testing to guide along with other

adjunctive tools and clinical information to guide

the treatment of 6MP.

The general process that we've tried to

create as a general framework for assessing

pharmacogenetics, particularly of

previously-approved drugs, where we don't have the

benefit of new drug development plans and

prospective trials, is to think about the general

process of approaching the assessment of

pharmacogenetics and its value in inclusion on

label.

In this meeting, we wouldn't mind having

comments on this. It's not one of the specific

questions, but I think we need to think about a

general process where we develop the appropriate

questions. What is the question we're asking of a

pharmacogenetic test or a piece of information? We

try to capture the relevant evidence. Oftentimes,

that comes from the published literature. You'll

see today how we try to abstract and summarize in a

sort of meta-analysis the scientific and clinical

evidence that allows us to go forward and make a

decision.

We think it's important to evaluate the

quality of the studies, and that's what we've done

in this case of 6MP as well as the Irinotecan we'll

talk about. We look at the overall strength of the

evidence from the individual studies, and we

consider other factors in the relabeling decision

that has to do with test availability, test

performance and things of that sort, and then,

finally, we move to the specific language for the

label.

So it's a general framework that we walk

through as we think about this drug or that drug or

the next drug where pharmacogenetic information may

possibly be pertinent to the improvement of

benefit-risk or dosing.

Now, today, you're going to hear about the

current understanding of pharmacogenetics and

neutropenia with regard to Irinotecan. What I've

summarized in this table is on the left-hand side,

groups that begin with all of the patients and then

moving over from left to right the prevalence of

genotypes and the risk of toxicity that we

currently feel is the case. And this comes from

the data of one paper that's been published by

Innocenti in 2004, and what this shows basically is

that all patients have a relative risk of 10

percent of developing grade three/grade four

neutropenia.

Breaking down the general population, we

have a subgroup that represents the 7/7 genotype

for UGT. Prevalence of that genotype is

approximately 10 percent. The relative risk of

toxicity, that is to say, the penetrance of

toxicity is 50 percent. Patients that are

heterozygous, prevalence of 40 percent with a

relative risk of toxicity of 12.5 percent, and at

least in this study, patients that were 6-6 had a

prevalence of 50 percent and a relative risk of

toxicity of zero percent.

So you see more of that data, but this is

the compelling data I think that facilitated

bringing this before the Committee. So we see the

potential for a test for UGT testing in this figure

that oftentimes is used to represent the value of

pharmacogenetics. And basically, we're looking at

patients with the same diagnoses that require this

drug for its approved indication and an overall

risk of 10 percent for neutropenia.

We know, however, in this general

population is a mixture of people with different

genotypes, and the value of the tests and the way

we'd like to consider this is what would this test

bring to the table as an adjunct piece of

information to use this drug in the most optimal

way? We have on the top the profile for the high

risk patient that we could identify with a genomic

test. Not all of those patients identified, as

you'll see, will develop frame toxicity. We have a

middle group that represents those at moderate

risk, and we have the bottom group based on the

data I've showed you that has a relatively low

risk.

So I think the goal of pharmacogenetics in

general and in this case specifically is to try to

differentiate and discern the differences between

patients who otherwise would be perceived to be

those patients with the same disease and the same

indication for the drug and bring that

differentiation as a tool to the clinical practice

for improving benefit-risk and drug and dose

selection.

So that's the context for what we want to

do this morning. So we'll begin with a discussion

of the scientific and clinical evidence that links

the UGT polymorphism with severe neutropenia.

We'll discuss, then, the role that testing can play

in identifying patients predisposed to severe

toxicity.

Now, I want to share with you that we have

had very positive interactions with the sponsor on

this topic. The sponsor is committed to providing

informative and understandable labeling as is the

FDA for all its drug products. Both agree that

information on UGT polymorphism and the risk of

toxicity on the label is of great importance. Both

sponsor and FDA agree to update the label to fully

inform prescribers and patients about

pharmacogenetics.

However, in this discussion today, we're

not going to get into what the label will say or

what the specific wording will say. We'll be

talking primarily about the scientific and clinical

evidence and the strength of evidence and the

questions that we've brought before the Committee.

I'll say this once now, just to sort of

create a framework for what you're going to hear.

We'll go back to the questions, obviously, near the

end of the morning. But basically, we're going to

ask is the clinical and scientific evidence that

you'll hear sufficient to demonstrate that

homozygous genotypes, the 7/7 genotypes, are at

significantly greater risks for developing

neutropenia and/or acute and delayed diarrhea from

therapy.

We're going to ask based upon what is

known on this relationship between Irinotecan

containing regimens and toxicity three

subquestions: do we know enough to recommend the

starting dose of the drug in the single agent and

combination therapy? What would be the risks and

benefits of the recommended starting dose? And if

we need more information, what is the appropriate

study to evaluate dosing in these types of

patients?

We'll ask how information about genotype

can be used in combination with bilirubin. This

is, like a lot of genetic tests, not the end all

and be all, but it's an adjunct piece of

information that, used in conjunction with other

indicators like bilirubin, can be used to improve

clinical decision making.

And finally, you'll see some information

where we're going to ask is the measurement of the

genotype sufficiently robust in terms of its

clinical sensitivity and specificity to be used as

a response predictor test for Irinotecan dosing?

So those are the questions and my

introduction. I could pause and turn it over to

the chair, or we can launch into the next

presentation.

DR. VENITZ: Any quick questions or

comments by the Committee before we start off on

the first topic?

DR. SINGPURWALLA: I hesitate to raise

these questions because I'm new here to this

Committee, but I'd like to draw your attention--I'd

like to go to that particular chart, where you had

prevalence and risk of toxicity. If I were to

understand correctly, the purpose of this chart is

to show that patients who are 7/7, and I don't know

what that means, have a 50 percent risk of

toxicity.

DR. LESKO: That is correct.

DR. SINGPURWALLA: There are two concerns

I have. One is why is the 10 percent prevalence

important? Because if you know the patient is 7/7,

then, the risk of toxicity is 50 percent. So the

fact that there are only 10 percent of individuals

who are of that particular category is not

relevant, unless you are giving the drug blindly,

without taking into consideration the

characteristics of them. So that's the first

question I want to ask.

DR. LESKO: Okay.

DR. SINGPURWALLA: Then, I have a comment

to make, and I'm not sure if this is appropriate,

but there's something called Simpson's Paradox that

arises in these particular contexts. And I'm

wondering if that has been taken into account.

DR. LESKO: Well, let's go to the first

question. I think in the first question, what I've

showed is that in the overall population, and you

use that term blinded. If you're blinded, and we

were all sitting around the room as potential

patients, the overall risk of toxicity is 10

percent. The question is which of the one out of

10 people are going to be most at risk for that

toxicity?

Without a test, you wouldn't know that.

You could obviously use other information that's

available on the drug and currently is in the label

regarding age or bilirubin levels. What the

genomic tests would do is to begin to differentiate

those people in the room, those 10 people around

the table that would be more at risk than the 10

percent risk for the general population, because

you're teasing out, obviously, the people that

would have a very low risk or no risk at all and

identifying those with a higher risk.

So if we took the general population,

subdivided it by genotype, then, the overall risk

would be 10 percent. It's only a reference mark.

I didn't mean to say more about it than that.

However, in this group of patients identified by a

genotype, by a test, the risk is much higher. It's

about 50 percent.

And what you'll hear in some of the

presentations is the likelihood ratio and relative

risk of developing this adverse event in that

defined subset by the genomic test, so that's the

relevance of the test and comparing it to the

current situation without and the future situation

with a test.

DR. SINGPURWALLA: Yes, but I'm still not

clear. If there is a 7/7 patient, I'm sorry; you

wouldn't blindly give the drug to anybody, any

person who shows up at random. You'd find out if

that is a 7/7 and then act accordingly; isn't that

correct?

DR. LESKO: Well, clinically, I'd think

you'd want to do that, because that would have a

role in making your decision to give the drug, what

dose to give and to consider other choices, but you

need something to identify that patient, and

currently, that information is not contained in the

package insert to guide the physician to make that

decision.

DR. SINGPURWALLA: I see. So now, given

that that is the case, then, I would strongly

encourage you to look at this notion, this concept

of Simpson's Paradox before you arrive at a 10

percent risk for patients overall.

DR. LESKO: Can you tell me what Simpson's

Paradox is?

[Laughter.]

DR. SINGPURWALLA: Well, yes, what is good

for--a drug is good for men, and a drug is good for

women, but a drug is not good for the population as

a whole. That's the paradox. Okay; some of your

statisticians will help you with that.

DR. LESKO: Okay.

DR. SINGPURWALLA: If not, I'll charge you

a fantastic consulting fee and help you.

DR. LESKO: Sounds good. I don't know if

you'd be on our advisory committee.

[Laughter.]

DR. SADEE: Larry, I have a comment, maybe

a different type of question that's brought up by

this polymorphism for Irinotecan.

The optimal dose, unquote, optimal dose

has been derived before these considerations, and

that includes a patient population that's really

inappropriate to calculate the optimal dose. So

what would be the position of the FDA to--in new

trials, where you combine Irinotecan? Do you go by

the old optimal dose, unquote, that included a

patient population that really shouldn't have been

included?

So the question would be how do we define

optimal dose in future trials to the normal

population, which is the inverse of saying what is

the optimal dose for those people who are at high

risk?

DR. LESKO: Right; yes, I don't think we

have the data to make that call. We'd like to know

the answer to that question, but I don't think we

have enough information to say what the optimal

dose would be for these subtypes. I'm not sure we

have enough information to talk about the optimal

dose for any of the patient groups.

However, we're going to discuss that as we

get further into the morning. In fact, I think

that's one of the questions we'll talk about: in

the absence of credible information to discern the

dose, what type of study and what type of study

design would be appropriate? And I think you'll

hear about some of the ongoing studies that will be

along those lines. So maybe we can leave that as a

partial answer and wait for the rest of the

presentations.

DR. VENITZ: Thank you, Dr. Lesko.

Then, our first speaker on that topic is

Dr. Atik Rahman. He's going to review for us the

clinical evidence of the role of UGT1A1 in

Irinotecan.

DR. RAHMAN: Good morning. I am Atik

Rahman, the acting deputy director of the Division

of Pharmaceutical Evaluation I, the Office of

Clinical Pharmacology and Biopharmaceutics. I'm

also the chair of the OCPB Pharmacogenetic Working

Group.

This morning, you have already heard from

Dr. Lesko regarding the approach that we took with

6MP and thiopurine methyltransferase enzyme. The

pharmacogenetics of thiopurine methyltransferase,

TMPT enzyme, and 6MP toxicity was discussed by the

Advisory Committee of the Pharmaceutical Sciences

in November of 2001. Subsequently, the topic was

also discussed by this Committee on April 23 and

also by the Oncology Drug Advisory Committee,

Pediatric Subcommittee on July 15 of 2003.

I will briefly discuss or update the

Committee on 6MP label modifications that resulted

from the Committee's deliberations and the FDA

interaction with the manufacturer of Purenithol. I

will follow that with the scientific and the

clinical evidence from the literature that we have

in the agency to demonstrate a relationship between

UGT1A enzyme polymorphism and its association with

Irinotecan toxicity.

6MP is inactivated by TPMT. TPMT is a

polymorphic enzyme. Ninety percent of the

caucasian and African-American population have

normal gene and normal enzyme activity. Ten

percent of the population have intermediate enzyme

activity, resulting from one deficient TPMT allele,

and one in 300 has low or no TPMT activity because

of two deficient allele in their gene.

There is a strong correlation between

genotype and phenotype, which is expressed as

either TPMT enzyme activity or as the levels of

6-thioguanine nucleotides in red blood cells.

A clinical study showed that 100 percent

of the homozygous patients required 6MP dose

reduction to prevent toxicity, compared to 35

percent of heterozygous and 7 percent wild type

patients. Currently, prospective trials are

ongoing to evaluate appropriate dose of 6MP in

acute lymphoblastic leukemia patients.

6MP is given with other myelosuppressive

therapy in the treatment of acute lymphoblastic

leukemia or ALL. Literature information indicated

a potential benefit of reducing the dose of 6MP in

patients with low to intermediate TPMT enzyme

activity. This reduction of approximately 50

percent 6MP dose in heterozygous and approximately

80 percent 6MP dose reduction in homozygous

patients allowed other myelosuppressive agents to

be given in full dose with 6MP during the entire

course of therapy.

Most of ALL protocol now avoid radiation

with 6MP because of the higher incidences of brain

tumors observed in TPMT-deficient patients in

previous trials.

Based on the advice of the advisory

committees and the manufacturer of Purinethol, Tiva

collaborated to include the information on TPMT

polymorphism and its relationship with 6MP toxicity

in the package insert of Purinethol. A new

subsection is included in the clinical pharmacology

section to describe the metabolism of 6MP and TPMT

polymorphism.

In the warning section, bone marrow

toxicity subsection includes a warning for

substantial dose reduction for homozygous TPMT

deficient patients. Information on the

availability of genetic tests is indicated in the

precaution section.

The availability of the test is mentioned

in the dosage and administration section of the

label, and substantial dose reduction is indicated

for patients with TPMT deficiency. In the future,

we hope to provide specific dosing recommendations

for both the homozygous and the heterozygous TPMT

deficient patients for 6MP therapy.

Now, I al provide you with the scientific

and clinical evidence that relates Irinotecan

pharmacogenetics with toxicity. Irinotecan is

indicated as a first line therapy in combination

with 5-fluorouracil and leucovorin for colorectal

cancer patients. The drug is also indicated as a

single agent for patients with metastatic

colorectal carcinoma, whose disease has recurred or

progressed after initial 5-fluorouracil-based

therapy.

Two phased randomized controlled

multinational clinical trials show that Irinotecan

in combination with 5-fluorouracil and leucovorin

increased the survival in first line colorectal

cancer patients compared to Irinotecan alone or

5-fluorouracil alone. Two multicenter randomized

clinical trials show significant increase in

survival for colorectal cancer patients whose

disease has recurred or progressed after prior

5-fluorouracil therapy.

Irinotecan is metabolized by

carboxyesterases to SN38, a metabolite which is

1,000 times more potent than the parent drug. SN38

is glucuronidated in the liver by UDP glucuronol

transferase family of enzymes, predominantly by

UGT1A1, and eliminated via biliary route.

Deficiency of UGT1A1 results in increased SN38

levels in plasma and in bone marrow cells, causing

hematologic and nonhematologic toxicities. These

toxicities result in dose delay, dose reduction and

hospitalization and even sometimes in deaths.

UGT1 gene is located on chromosome 2 and

contains at least 13 different promoter axons which

are spliced to common axons 2 through 5. UGT1A1 is

an isoform that is associated with bilirubin

glucuronidation. The isoenzyme has more than 30

variant alleles. UGT1A1*28 is a variant allele

that contains seven TA repeats in that TATA box of

the promoter region instead of six TA repeats.

Today, we will focus on UGT1A1*28 variant

only, because the agency believes that we have the

most mature data on this variant's association with

Irinotecan toxicity. I will use the term 7/7

genotype to refer to UGT1A1*28 variant in my

presentation.

Fischer et al studied the relationship

between UGT1A1*28 genotype and estradiol

glucuronidation mediated by UGT1A1 enzyme. Liver,

kidney, lung and intestinal tissues were tested for

UGT1A1, 1A6 and 2B7 isoenzymes. In the 7/7

genotype liver samples, the apparent Micholas

Mentin constant KM was not altered, but the V-max

was altered, compared to 6/6 wild type liver

samples. Liver samples with 7/7 genotype had a

fourfold lower activity of the enzyme compared to

the samples with normal gene expressions, as shown

in this bar plot.

As you have already heard, that in the

caucasian population, the frequency of homozygous

deficient 7/7 genotype is approximately 10 percent,

with a range from 5 to 15 percent. The

heterozygous 6/7 genotype is approximately 40

percent.

This slide illustrates the relationship

between the risk of severe neutropenia and diarrhea

and SN38 exposure. These data are from the phase

2-3 studies of Irinotecan in which weekly doses of

100 to 150 milligrams per meter square were

administered to patients with colorectal cancer.

In absence of individual PK data in these studies,

mean AUC data from earlier studies were used.

Despite the limited data, logistic regression

analysis suggested that the risk of severe

neutropenia and diarrhea increases with SN38

exposure.

The first article that I'd like to present

today is a clinical study that was conducted at the

University of Chicago by Dr. Mark Ratain and his

group. The article was published in the Journal of

Clinical Oncology this year. This was a

prospective study in 66 solid tumor or lymphoma

patients. The study evaluated the association

between the prevalence of severe toxicity and

UGT1A1 genetic variation.

The patients received 350 milligram per

meter square dose of Irinotecan every three weeks.

This is an approved dosing regimen for single agent

Irinotecan therapy. Toxicity was assessed during

cycle 1. Fischer's exact test was used to relate

genotype with pharmacokinetic parameters,

pretreatment bilirubin and absolute neutrophil

count.

I'd like to mention some of the highlights

of this study. The study has certain unique

features compared to the other studies that I'm

going to present subsequently. This is a

prospective trial with an adequate number of

patients who had 7/7 genotype. The study is clean

in terms of not having any contribution in toxicity

from other agents. Sometimes, it's hard to pin

down the culprit for toxicity in combination

regimen chemotherapy trials.

The onset of toxicity was rapid with the

first cycle of therapy. The PK assessment was

reliable, being conducted in a lab that has

pioneered an assay for this complex and unstable

molecule.

There was a significant difference in the

dose normalized AUC exposure between 7/7 genotype

and 6/6 genotype patients. A significant

difference was also noted between 6/7 and 6/6

genotype patients. This is a combined data set,

including 66 patients from the Innocenti study and

20 patients from another phase one study conducted

in the same institute, using 300 milligram per

meter square dose of Irinotecan.

This slide shows the relationship between

the maximal decrease in absolute neutrophil count

ANC as a function of SN38 exposure. Patients with

6/6 genotype are shown in blue. Those with 6/7

genotype are shown in green. And patients with the

7/7 genotype are shown in red. The square symbols

are the mean ANC nadirs and SN38 AUCs for the three

subgroups. The data were log-transformed and fit

using linear regression models. The blue line

shows the predicted curve for 6/6 and 6/7 genotypic

groups, and the red line shows the predicted curve

for the 7/7 patients.

The 7/7 genotype has a greater effect on

the ANC nadir versus SN38 AUC relationship compared

to the 6/6 and 6/7 genotypes. For the same AUC,

the 7/7 genotypes show a lower ANC nadir.

Overall, the study showed 50 percent of

the 7/7 genotype patients had grade four

neutropenia compared to 12.5 percent heterozygous

patients, and no wild type patients had grade four

neutropenia. There is a significant difference in

the exposure to SN38 between the

homozygous-deficient 7/7 patients and the 6/6

genotype patients as shown in the previous slide.

Also, the pretreatment bilirubin levels between the

7/7 and the combined 6/7 and 6/6 genotype patients

was significant.

The prevalence of grade four neutropenia

and grade three diarrhea in the overall

population--sorry, the prevalence of grade four

neutropenia was 9.5 percent, and the grade four--or

the severe diarrhea was 5 percent in this study.

Notable in this study, one patient died of

neutropenia-related sepsis who had 7/7 genotype and

had the highest total bilirubin level.

In the study, the grade four neutropenia

was significantly higher in 7/7 patients compared

with 6/7 and 6/6 patients. The relative risk for

grade four neutropenia for 7/7 patients was 9.3.

Only three patients in this study had grade three

diarrhea. One was a 7/7 patient, and the two

others were patients with 6/7 genotype. None of

the patients with 6/6 genotype had severe diarrhea.

The study conclusively established an association

between genotype and neutropenia.

This is a prospective phase two study

designed to evaluate the influence of UGT1A1

polymorphism on the toxicity profile, on the

response rate and on the overall survival in 95

colorectal cancer patients treated with four

Irinotecan containing regimens. Irinotecan

regimens were 350 milligram per meter square every

three weeks; 80 milligram per meter square weekly;

or 180 milligram per meter square biweekly.

Toxicity was evaluated during the entire duration

of treatment. No PK samples were collected.

Various statistical tests were applied to assess

the differences between the categorical variables

and between the related or unrelated continuous

variables.

Logistic regression was used as a

multivariate method to assess of genotype

independently predicted toxicities. I will not

present any efficacy data from this trial.

Neutropenia and diarrhea in this table

includes both grade three and grade four

toxicities. Forty percent, four out of 10 7/7

genotype patients had grade three/four neutropenia

compared to 15 percent patients with normal

alleles. Seventy percent of the patients with 7/7

genotype had severe diarrhea compared to 17 percent

patients with normal alleles. The cumulative dose

of Irinotecan received by 7/7 genotype patients was

1,398 milligrams per meter square, compared with

1,725 milligrams per meter square received by

patients with 6/6 allele.

The prevalence of grade 3/4 neutropenia

and diarrhea in the overall population was 21

percent and 31 percent respectively. Notable is

the incidence of diarrhea, which was higher in this

trial compared to what we have seen with other

Irinotecan-based trials.

Both univariate and multivariate analysis

showed statistically significant association

between appearance of diarrhea and 7/7 genotype

compared with 6/6 genotype. Hematologic toxicities

increased from 6/6 patients to 7/7 homozygous

patients from 15 to 40 percent but didn't achieve

statistical significance. Cumulative dose of

Irinotecan received by 7/7 patients were lower than

the dose received by 6/6 patients because of the

dose reduction that was necessary for the

appearance of severe diarrhea. This study

demonstrated a significant relationship between

genotype and severe diarrhea. The statement by the

author of this article shows a need for

genetic-based chemotherapy treatments for cancer

patients.

This is the third study that I'm going to

talk in detail about. This is a retrospective

study of 75 metastatic colorectal cancer patients

receiving two common Irinotecan containing

combination regimens. Irifufol regimen contains 85

milligram per meter square weekly Irinotecan, given

with 1,200 milligram per meter square weekly

infusional 5-fluorouracil plus 100 milligram per

meter square bolus leucovorin. Folfiri regimen

contained 180 milligram per meter square biweekly

Irinotecan given with 2,500 milligram per meter

square infusion of 5-fluorouracil and 400 milligram

per meter square leucovorin. No PK samples were

collected in the study. Kruskal-Wallis test was

used to assess the statistical difference among the

three populations.

Seventy-one percent of the 7/7 patients,

five out of seven, compared with 10 percent 6/6

patients, three out of 31, had grade 3/4

neutropenia. Sixty percent of the 7/7 patients had

neutropenic fever compared to no 6/6 patients

suffering from neutropenic fever. Neutropenic

fever was associated only with patients who carried

at least one deficient allele of UGT1A1.

Irinotecan courses had to be delayed in

five out of seven patients in the 7/7 group

compared with 21 out of 35 in 6/7 and 10 out of 31

in 6/6 group. 100 percent of the 7/7 patients

whose therapy had to be delayed for toxicity had to

be hospitalized compared with no 6/6 patients with

delayed therapy requiring hospitalization.

The prevalence of grade 3/4 neutropenia

and diarrhea in the overall population was 30

percent and 7 percent respectively. There was no

association between genotype and diarrhea because

of the low frequency of diarrhea in this trial.

There was a strong correlation between genotype and

grade 3/4 neutropenia. 100 percent of the 7/7

patients who had severe neutropenia needed delayed

therapy and hospitalization compared to none of the

6/6 patients who had neutropenia and/or diarrhea.

The authors mentioned in this article that

hematologic and digestive toxic events were not due

to 5-fluorouracil because all of the patients had

5-fluorouracil dose adjusted individually to avoid

severe 5-fluorouracil toxicity.

Literature includes other adequately-sized

studies that I'd like to summarize for the

Committee. These are the two PK studies that

evaluated the effect of variant alleles on

Irinotecan disposition. Mathijssen's study

evaluated a number of genes associated with the

metabolism, transport, and disposition of

Irinotecan. UGT1A1 genotype did not correlate with

Irinotecan disposition. Notable, there were only

two UGT1A1*28 patients in this study.

On the other hand, Paoluzzi's study showed

a significant decrease in the exposure ration of

the SN38 glucuronide to SN38, indicating a

reduction in the formation of the SN38 glucuronide

in 7/7 patients.

Font, et al., published a phase two study

evaluating the activity of docetaxel and Irinotecan

in 51 non small cell lung cancer patients.

Irinotecan 70 milligram per meter square was

administered with 25 milligram per meter square

docetaxel. The study did not see any correlation

between genotype and toxicity. The overall

incidence of grade 3/4 neutropenia and grade 3

diarrhea in this study was low. Also, the dose of

Irinotecan used in the study was 70 milligrams,

compared to 100 to 125 milligram per meter square

dose used in combination studies of Irinotecan.

A retrospective analysis of 118 Japanese

patients who received Irinotecan containing regimen

showed that UGT1A1*28 genotype was a significant

predictor of severe toxicity. In this analysis, 55

percent of the patients, irrespective of genotype,

who had grade 4 neutropenia also had grade 3/4

diarrhea, and 73 percent of the patients who had

grade 3/4 diarrhea also had grade 3/4 neutropenia.

Sai et al in Japan conducted a PK study to

evaluate the relationship between SN38 PK and

UGT1A1 haplotype. UGT1A1*28 was associated with

reduced SN38 glucuronide to SN38 area under the

curve ratio and increased total bilirubin.

Iyer, et al., published a prospective PK

study in 20 patients that related genotype with

reduced SN38 glucuronidation rates and lower

absolute neutrophil counts in 7/7 patients compared

to 6/6 genotype patients.

In the evaluation of the clinical data

provided to the agency for the approval of this

drug, certain predictive factors were related to

increased toxicities. These are observations only

and not statistically powered data that allowed the

agency to recommend a reduced starting dose of

camptazar for patients equal to or older than 65

years; patients who received prior pelvic or

abdominal radiotherapy, patients whose performance

status was two, and patients with increased

bilirubin levels.

The reduction was by only one level. That

is for the 350 milligram per meter square every

three weeks regimen, the starting dose will be 300

milligram per meter square. Similarly, if the

normal starting dose is 125 milligrams per meter

square weekly, the predictive factors will

recommend a dose of 100 milligrams per meter

square. If the patients tolerated the reduced

starting dose, the dose in the next cycle is

increased to the standard dose.

In the geriatric use section of the label

we have the statement the starting dose of

camptizar in patients 70 years and older for once

every three week dosage schedule should be 300

milligram per meter square, a 50 milligram per

meter square reduction from the standard dose.

Based on the scientific and clinical evidence

available in the current literature, the agency

believes that genotype is a predictive factor for

Irinotecan dose limiting toxicity. The agency also

believes that SN38 level is a likely predictive

factor for toxicity.

Based on Innocenti's article, UGT1A1

polymorphism information will help to reduce the

overall incidence of grade four neutropenia from 10

percent to 5.7 percent, almost a 50 percent

reduction in the incidence of grade four

neutropenia. Irinotecan can be given as a weekly,

biweekly or every three weeks regimen. One of the

regimens may be more appropriate for the 7/7

genotype patients. Genotype testing, combined with

bilirubin levels and other predictive factors shown

in the previous slide, will allow the physicians to

select Irinotecan therapy more judiciously in the

high risk patients. Alternate therapy, either in

the first line or in the second line setting, may

be a choice for the 7/7 genotype patients.

I'd like to thank my colleagues from the

Office of Clinical Pharmacology and

Biopharmaceutics, Dr. Larry Lesko, Dr. Shiew-Mei

Huang, and Dr. Felix Frueh for helping me out with

this presentation. I'd like to thank my colleagues

from the Division of Pharmaceutical Evaluation I

for helping me out with or for providing the PK

analysis of the Innocenti's and the Phase I data.

They are Dr. Roshni Ramchandani, Dr. Yanning Wang,

Dr. Brian Booth, and Dr. Joga Gobburu. I'd like to

thank my boss, Dr. Mehul Mehta, for giving me the

time to prepare for this meeting, and last but not

least, I'd like to thank my colleagues from the

Division of Oncology Drug Products, Dr. Grant

Williams and Dr. Richard Pazdur, for challenging me

to translate the principles of clinical

pharmacology, especially pharmacogenetics, to

clinical practice and clinical use.

Thank you.

DR. VENITZ: Thank you, Atik.

Any questions or comments by Committee

members?

And what I'd like to do is after each of

the presentations, give everybody an opportunity to

ask questions for clarification. We have a whole

hour reserved from 11:00 to 12:00 to discuss the

specific questions that Dr. Lesko wants us to

address.

So, any specific questions?

DR. GIACOMINI: Yes, I had a couple of

questions.

Was race specifically or ethnicity

specifically looked at in any of the studies that

you reviewed?

DR. RAHMAN: I would say no, because the

Japanese study included only the Japanese

population, and the other studies in Europe and the

United States included only the caucasian

population. So we at this time don't know the

prevalence in blacks, in the African-American

population, or in other populations. I think the

polymorphism is less prevalent in the Asian

population. Dr. Howard McLeod might correct me on

that.

DR. MCLEOD: There is ethnic variation for

the frequency both of the 6 and 7 allele but also

the presence of either five repeats or eight

repeats seem to be more common in the, I think, the

racial minority groups found in the United States.

The impact of those other alleles is not completely

clear, although Dr. Ratain may address that in his

presentation.

DR. GIACOMINI: Okay; and as a followup to

that, then, maybe you can explain it to me: what

is the difference? I mean, you've called it the

star--I think, is it 28 haplotype versus 7/7. What

is in the haplotype besides the promoters? Are

there some other snips in the haplotype, and what

are they? And are they functional in any way?

DR. RAHMAN: Dr. Howard McLeod?

DR. MCLEOD: It's a genotype, Kathy, not

haplotype. Star-28 is the name that was stuck on

the 7/7 repeat, so folks that are homozygous for

the 7/7 genotype are called star-28 by the powers

that be in determining a nomenclature for UGT1A1.

DR. GIACOMINI: So it's not like there's

any other snip in there.

DR. MCLEOD: Correct.

DR. GIACOMINI: It's just simply not.

DR. RAHMAN: Can I clarify that a little

bit? Actually, that was not an appropriate term,

but there was a haplotype study that has been

published by Dr. Margaret Eng's group that are

associating star-28 with other variant alleles,

star-60 and others, and there is a paper out on

that. So I kind of was alluding to that, that in

that study, the number of star-28 patients,

homozygous patients, were low, and they were

associated with other risk factors.

DR. GIACOMINI: Okay; but we can assume

when somebody says star-28, they mean the 7/7;

they're homozygous, they mean the 7/7 genotype.

DR. RAHMAN: Seven, yes.

DR. GIACOMINI: Okay; one other question:

in terms of mechanism, I didn't see you quote any,

you know, like a reporter assay or something in

which we're seeing a transcription rate difference

between the 7, some biochemical mechanism which

supports what you're seeing clinically in terms of

that. Are there biochemical data like that?

Mark's saying yes. You can't talk, Mark.

DR. MCLEOD: In the--I don't know if

everybody received--there was a packet that we

received of light reading. Within that is a number

of papers addressing those issues, including a

paper from Dr. Boiler Scripps that looked at both

the racial issue as well as the promoter variance

in terms of luciferase assays, seeing this stepwise

inverse relationship between length of the TA

repeat and the amount of transcript that's

produced.

DR. GIACOMINI: Okay; thank you.

DR. VENITZ: Go ahead.

DR. HALL: Can you give some idea as your

part of the agency as to your views on why, you

know, you know, several or a significant number of

the 6/6s get the toxicity, and a significant number

of the 7/7s don't get any toxicity? Do you have

some kind of rationalization for this?

DR. RAHMAN: As I've shown in this, in my

presentation that the toxicity in the 6/6s

definitely lower than 7/7 patients. As you can

see, 71 percent versus 10 percent. In one of the

studies, none of the 6/6 patients had toxicity. So

I'm not sure if there is a significant--the

statement that you made, there is a significant

toxicity in 6/6 patients is really true. But there

is toxicity in the 6/6 patients also, and that

could be attributed to not only Irinotecan

depending on the studies if there were a

combination of regimens used. It could be also

attributed to the others. But it also could be

attributed to Irinotecan.

DR. HALL: So have you considered other

mechanisms other than this 1A1 polymorphism as a

contributor to the toxicities?

DR. RAHMAN: Yes; as I've shown that other

predictive factors have been associated for

predicting the toxicity, like the bilirubin levels

is one of the predictors that has been kind of

alluded to being related to toxicity

DR. HALL: But bilirubin would be

metabolized by the enzymes. So they would be

somewhat correlated.

DR. RAHMAN: Yes.

DR. HALL: Other enzymes? Other genes,

perhaps?

DR. RAHMAN: There are papers out in the

public domain which are trying to associate other

UGT1A1 enzyme, and the factors that could also

contribute is the sensitivity of the individual

patients to specific neutropenia or to severe

diarrhea, and that is something I am not aware of

the magnitude of. But there is an understanding

that some of the patients might be overly sensitive

to a certain kind of regimen compared to the

others, even that don't carry any of the homozygous

deficient alleles.

DR. VENITZ: Jeff?

DR. BARRETT: I had a question about the

prevalence rates in your responses both in the

grade 3 diarrhea and in the neutrophil count. It

seems that there's quite a bit of interstudy

variability, and I know in your pooled analysis,

you were basing this on mean data, but is there

any--and now, with the Innocenti PK information

available, you really could use some of that

information to back-project some of the individual

variance in those models. Is that going to happen?

DR. RAHMAN: So, first of all, I'm trying

to address the question about finding differences

in the neutropenia and diarrhea in different

studies. What happens is that in certain trials,

the patients were allowed to take premedication for

diarrhea, so that might have helped. It is

approved in the label for using of loperamide and

other agents for controlling of diarrhea.

Now, there are two components of diarrhea.

One is the early phase diarrhea, and the other is

the late phase diarrhea. I have kind of focused on

Innocenti's article, because that truly was trying

to address these toxicity issues. And their study

was trying to look for diarrhea as well. However,

the incidence was pretty low.

Now, we are exploring the pharmacogenetics

of this drug. However, we're trying to see if SN38

could be a good predictor or could add on to to

come to a kind of dose for 6/7 and 7/7 patients,

but this is still in the earliest stage, and I

think we need some more solid data to show the

relationship and then can make a difference.

DR. SADEE: The star-28 genotype is

associated with Jorbert's syndrome.

DR. RAHMAN: Right.

DR. SADEE: And so my question is how

often is this diagnosed, and what is the

correlation between having a patient diagnosed with

Jorbert's syndrome and toxicity? So, in other

words, could we substitute a genotyping with a

diagnosis for Jorbert's Syndrome?

DR. RAHMAN: I haven't come across any

article to address that question.

DR. VENITZ: Larry?

DR. LESKO: Yes, just on the last

question, you know, another way to think about this

is as an adjunct test. You could actually think of

tests being done in parallel. If you think of a

screening test, you're screening an entire

population with or without an elevated bilirubin.

You could increase, I think, the sensitivity of a

UGT test by maybe screening people that signal by

their high bilirubin that they may be at potential

risk for the genotype.

So either in parallel or in serial, I

think you could enhance the value of the test by

thinking of it as an adjunct to the current

information that somebody would have. When you

look at--getting back to the prevalence issue, when

you look at prevalence of Jolbert's syndrome as a

function of the ethnic or racial group and the

prevalence of the UGT, there are some parallelisms

there.

In a review article that appeared last

year, the range of prevalence of UGT was anywhere

from 2 to 3 percent in Asian populations; up to 23

percent in blacks and Africans. So that's sort of

the 2 to 23 percent range that people have reported

for the prevalence of the polymorphism.

I was going to maybe add this other

information, because Atik showed the slide, and it

was shown as average values, but we were looking at

the question Steven Hall raised about the 6/6s

becoming toxic and the 7/7s not, but I think what

we have here is a probabilistic issue that if you

look at the variability in SN38 area under curve in

each of the genotypes, there's clear distinctions

based on mean values. However, the low end of the

6/6 area under curve and the high end of the SN38

area under curve for the 7/7s does overlap.

So I think what we're seeing, then, is

some risk in the 6/6 homozygous and some lack of

risk in the other people as one possible

explanation. I think another explanation is that

in most of these cases, more than the target allele

was not looked at, so there could be other alleles

that would be predisposing individuals to risks

that weren't measured, perhaps because of, you

know, the ethnic or racial background. But these

are some of the possible reasons anyway, but

certainly, the pharmacokinetic explanation seems to

make sense based on what we know about area under

curves related to nadirs of neutrophils.

DR. MCLEOD: Back to Dr. Sadee's question:

in the prospective study in the GI intergroup

throughout North America, the N9741 study,

Jolbert's syndrome was one of the flags in the

inclusion criteria or exclusion criteria. Yet, we

had an 8 percent frequency of the 7/7 genotype.

But just kind of highlights the lack of diagnosis

of Jolbert's syndrome, because it is a subclinical

benign hyperbilirubinemia syndrome.

It's out there, and so, plenty of people

are getting this drug without that diagnosis,

because it's not something that's really evaluated

in common practice.

DR. WATKINS: I was the only--as the only

hepatologist here, I'd just reiterate that it's a

subclinical diagnosis that can be brought out by

fasting and certain other conditions, like certain

protease inhibitors.

But in the studies that have looked at, I

believe, even the majority of 7/7s have bilirubins

within normal limits, so it would not be a

surrogate. I guess one question is how much does

the genotyping add in a multiple regression if you

include in serum bilirubin, which I haven't heard,

but I'm sure someone is going to address it.

DR. SINGPURWALLA: Your slide 15,

retrospective analysis, you showed two pictures,

one on the left, my left, and one on the right.

You have predicted versus observed. How did you

get the predicted?

DR. RAHMAN: Okay; the predicted line came

from the regression analysis.

DR. SINGPURWALLA: So the predicted is

based on the observed data.

DR. RAHMAN: Right.

DR. SINGPURWALLA: I'm curious why you

didn't fit a straight line. I know if you fit a

straight line--

DR. RAHMAN: Right.

DR. SINGPURWALLA: --it would go out, but

why did you choose that particular form?

DR. RAHMAN: It was done by our

pharmacometric folks, and they have got a better

understanding of the modeling that they did. And I

think they thought that this was the appropriate

regression to use rather than a linear regression.

DR. SINGPURWALLA: I understand, but there

were three points, and you can draw all kinds of

curves.

DR. RAHMAN: Yes, that's true. Actually,

the intent was to show that there is a relationship

that we have seen, but it's very soft at this time,

as I've said, because these are all the mean values

that we are lumping together; it's not the

individual ones, which would have given us a much

better fit. And also, in large clinical trials,

the PK is not collected.

So we kind of lumped them together and had

only three reliable mean values that we could do

something with predictions.

DR. SINGPURWALLA: So rather than saying

predicted, you should really say fitted.

DR. RAHMAN: Fitted.

DR. SINGPURWALLA: Because when you say

predicted, I'm thinking of some theory that tells

you what's the probability.

DR. RAHMAN: Right, right.

DR. SINGPURWALLA: As the dose increases.

DR. RAHMAN: Right.

DR. VENITZ: Ed?

DR. CAPPARELLI: Yes, I just wanted to

echo a little bit of what Dr. Watkins was saying,

and one of the studies you mentioned that there was

an evaluation of bilirubin levels as well as

genotype, and you'd expect them to be, you know,

highly correlated, and one of the questions is

what's the independent component that the genotype

is going to give in conjunction with the fact that

there is some induceability of this enzyme, and so,

genotype may not fully predict especially, and a

single genotype differential may not predict.

The other sort of related question that I

have is has there been any look at inducers of

CYP3A? Because the APC metabolite actually

represents a larger portion of the compound that

ends up in urine and feces. And so, if that goes

by 3A4, again, you may have some differential there

as well as if you have some inducers, you may have

some confounding of the genotype.

DR. RAHMAN: One thing I can tell you is

that there are studies going on, I think, which

Pfizer will present showing that they are looking

at, besides UGT1A1*28, other genetic factors and

other 3A4, 3A5 to show if--to see if they have any

association with the toxicity. So there are

studies going on. That's how far I know. But I

have not reviewed or looked at any articles

focusing on that.

And also, Dr. Ratain's and Dr. Howard's

group are looking at all these various factors in

association with toxicity for Irinotecan.

DR. DAVIDIAN: My comment just pertained

to Nozer's comment a minute ago. This is a

logistic regression, right? Is that what was done?

DR. RAHMAN: Yes.

DR. DAVIDIAN: I think it was a logistic

regression.

DR. RAHMAN: Right.

DR. DAVIDIAN: So this is just the fitted

probability curve.

DR. RAHMAN: Right.

DR. DAVIDIAN: Was this based on these

three mean values? Or was it based on--or are you

just showing the mean values on the plots?

DR. RAHMAN: It is based on the three mean

values.

DR. DAVIDIAN: And that's all.

DR. RAHMAN: Right.

DR. DAVIDIAN: Oh, okay.

DR. VENITZ: Okay; Atik, let me ask you

one final question for my clarification: the

Innocenti study is the only one that's a single

agent study. All the other studies are

multiagents.

DR. RAHMAN: That is correct.

DR. VENITZ: So that is really the only

study that allows us to look at Irinotecan as

opposed to the contribution that other

chemotherapeutic agents might play--

DR. RAHMAN: That is true.

DR. VENITZ: --in toxicity.

DR. RAHMAN: As far as I know, that is the

only prospective study that was actually conducted

to address this association of genotype with

Irinotecan toxicity. So this was kind of--this was

a focused study looking at these specific issues,

which was based on another phase one trial which

they conducted early on with 20 patients, so they

kind of, you know, expanded on that and moved on to

do this prospective trial to address the issue.

DR. VENITZ: But all the other studies

that you reviewed either prospective or

retrospective in nature, they involved other agents

as well.

DR. RAHMAN: Yes, like in the four

different regimens in the second trial that I

talked about, Markelos' trial, I think they had one

arm with a few patients who received a single

agent, Irinotecan, and then the other arms. So

there are blips of single agent here and there, but

the other studies definitely had other components

besides single agent trials.

DR. VENITZ: Okay; if there are no more

questions, then I thank you.

Our next presenter is Dr. Parodi. He is

the director of clinical pharmacogenomics at

Pfizer, and he's going to give us the Pfizer

perspective.

DR. PARODI: Good morning.

I would like to thank Dr. Lesko, Dr.

Rahman for inviting Pfizer to participate in this

meeting. I would like also to acknowledge my

colleagues Dr. Mark Morrison and Dr. Akitunde Belo,

who are here to answer any questions that may be in

regards to clinical or pharmacokinetic issues

during the meeting.

During this presentation, we would like to

reiterate the commitment of Pfizer to the safety of

all of our products. In addition, we would like to

talk about how we are applying pharmacogenetics at

Pfizer. At Pfizer, pharmacogenetics is getting

important information during the whole drug

development process, from early discovery research

through development and through the safety and

efficacy of our marketed products.

Today's discussion is around Irinotecan, a

cytotoxic agent that has been proven to be an

effective therapeutic choice for patients with

metastatic colorectal cancer and colorectal cancer

in general. Since the late nineties, there have

been several publications reporting a relationship

between Irinotecan's safety and genotype. We have

kept abreast of these publications, and we have

provided for the Advisory Committee a summary of

those published reports in the background document.

We will review this data, and we will

present to a highlight of what Pfizer is doing to

continue to expand the database regarding the

Irinotecan pharmacogenomics. We will also talk

about how we're working in collaboration with the

FDA to provide useful information in the label that

may refer to this particular genotype.

We believe this forum is going to be an

important meeting, where we can share our views and

our ideas about the subject and present activities

that we are currently undertaking to address these

issues. I would like to outline how

pharmacogenomics is being used at Pfizer. At

Pfizer, we use pharmacogenomics as a generic term

which also encompasses what may be traditionally

called pharmacogenetics. We use disease genetics

to select targets. We use the knowledge of the

variation of our targets to improve the safety and

efficacy profile of our drug candidates. And in

some cases, we are using the genetic variation to

define subpopulations for conducting proof of

concept studies.

For our marketed products, we are also

looking into the effect that genetic variation has

on the safety and efficacy of our drugs. Today,

we're focusing on the effect of genetic variation

on the safety and efficacy of Irinotecan.

We have been interested in this area since

the first studies that were reported reporting the

relationship between Irinotecan safety and

genotype. Later, Dr. Ratain, who has been a leader

in this effort, will address and present his ideas

on the subject.

During recent years, we have supported and

sponsored many clinical trials that contain a

pharmacogenomics component. We will go through a

detailed list of the projects we are supporting or

sponsoring later in the presentation. Most

recently, we have engaged in a collaboration with a

company in Germany called Epidaurus to explore the

significant variation of transporters in

metabolizing genes in pharmacogenomics.

Today, we will focus on our knowledge of

the variation in the genes involved in the

Irinotecan disposition and metabolism. This is a

representation of the genes involved in Irinotecan

metabolism, taking into account only what's going

on in the gastrointestinal lumen, where we have

more information. We have very little information

regarding the genes involved in the disposition of

both Irinotecan and its metabolites, both at bone

marrow level or at tumor level.

In general, we can say that the mechanism

for disposition and metabolism of Irinotecan is

complex, and it involves several genes. We want to

also note that all of these genes are very

polymorphic, and those polymorphisms are known to

be functional. As pointed out earlier, there is a

great deal of variability in the frequency of the

different alleles of these genes in the different

ethnic groups, which makes extrapolations from one

ethnic group to another very difficult.

The UGT1A1 polymorphisms are probably the

best known, but information is constantly being

accumulated about the polymorphisms in other genes.

As recently as a couple of weeks ago, there has

been a couple of reports reporting on novel

polymorphisms in the carboxylesterase genes.

Our discussion today will focus on one of

these genes, UGT1A1, in particular, one snip in one

of these genes, and we would like to address the

relevance of the published data associating that

polymorphism, specifically what has been called the

7/7 or star-28 polymorphism, in regards to

neutropenia and diarrhea.

By now, you are familiar with these

studies. We have basically conducted an extensive

review of the published literature and have

selected these studies which are full papers,

because they provide the frequencies of diarrhea

and neutropenia as well as the frequencies of

genotypes for the UGT1A1 gene.

Something we would like to note is that

these studies contain a rather small number of

individuals included in the studies. Given the low

frequency of the 7/7 genotype, very few patients in

this group have been actually included in these

studies.

Again, all studies don't include the same

type of cancer patients. Two of these studies

include only colorectal cancer patients, while the

other three include primarily lung cancer patients.

Probably more significantly is the issue that these

studies all use different dosing regimens, both in

the intensity of the dose of Irinotecan and the

schedule, and more significantly yet, the inclusion

of 5-fluorouracil, a known agent that causes

neutropenia.

Although there are many differences in

these studies, we have attempted to look at the

data in a comprehensive way. So we realize that

this data can be interpreted from multiple

perspectives, so what I'm going to provide for you

in the next few slides is a statistical analysis of

the data. All of the analysis has been done based

on the raw data presented in those papers and are

unadjusted for any known factors.

First, we would like to look at the rate

of neutropenia in the UGT1A1 7/7 patients compared

to the rate of neutropenia in the group containing

the 6/6 plus the 6/7 patients. A simple look at

this table tells us that the frequency of

neutropenia in the 7/7 patients is higher than the

frequency of neutropenia in this other group. In

order to establish a comparison between the

different rates recorded in the studies, we

calculated the odds ratios and the 95 percent

confidence intervals as represented by this

statistic to quantify the association between

genotype and neutropenia.

The odds ratios vary from study to study

and have very wide confidence intervals. Based on

the 95 percent confidence intervals, the odds ratio

was statistically significant in three out of the

four studies.

Without adjusting for known risk factors,

this univariate analysis shows a statistically

significant association between UGT1A1 7/7 genotype

and neutropenia, although we note that the

association varies among studies with odds ratios

between 2.5 and 16.7. This variability could be

due to the small sample size, differences in the

dosing schedules and the contribution of

5-fluorouracil to neutropenia; the fact that we are

not controlling for known factors such as baseline

bilirubin levels, age, performance status, and

prior pelvic radiation; and indeed, differences in

the population's treatment, both from the ethnic

perspective and also from the tumor type.

In a similar fashion, we have done the

same analysis for diarrhea, grade three plus. In

this case, we have included the Font study, because

the Font study reports the rates of diarrhea for

the different genotypes. It was not included in

the analysis of neutropenia because Font did not

provide the data for neutropenia separately in his

publication.

Again, we would like to note that this

analysis has been done without adjusting for known

factors, and if we look at the diarrhea rate

between the 7/7 genotype and the group of 6/6 plus

6/7, simple inspection of the rates shows that it's

difficult to draw a general conclusion. Here

again, we calculate the odds ratios and 95 percent

confidence intervals as a representative statistic

to quantify the association between genotype and

severe diarrhea.

The odds ratios vary from study to study

and have wide confidence intervals, and based on

the 95 percent confidence intervals, we can say

that two out of five studies were statistically

significant.

In summary, we have performed a

comprehensive review of the published literature

and selected publications that provided genotypes

and rates for neutropenia and diarrhea in

Irinotecan-treated patients. Although there are

significant differences among studies, we analyzed

the data without adjustments using odds ratios and

confidence intervals as a representative statistic.

We conclude that there is a statistically

significant association of UGT1A1 genotype in the

development of neutropenia. The association of

genotype in diarrhea is not as consistent among

studies.

Now, if we want to translate this

association data to a predictive performance of a

test, we need to assess multiple parameters. We

have used the same published rates for neutropenia

in genotypes used for the association analysis to

calculate the test performance parameters.

The following analysis assumes that the

genotyping test is 100 percent accurate for the

detection of UGT1A1 7/7, 6/6 and 6/7 genotype. We

have calculated the performance parameters based on

the reported rates in the publications that were

examined previously for neutropenia.

First, we look at the clinical

sensitivity. The clinical sensitivity can be

interpreted as the probability that those patients

that have neutropenia will have the 7/7 genotype.

We note that the clinical sensitivity varies from

15 percent or 0.15 to 50 percent.

Probably for our discussion, it's more

important to look at the predicted values.

Ideally, we would like to have a test with high

predicted values, maybe approaching 100 percent

possibly. Under the assumption that we know that a

patient has the 7/7 UGT1A1 genotype, the overall

probability that the patient will develop

neutropenia will be about 50 percent.

These values are not highly predictive for

developing neutropenia. Given that we cannot

accurately predict the development of neutropenia,

we have to be cautious when balancing the risk for

neutropenia and the benefit of treatment.

Neutropenia is generally manageable, and dosing

reductions for all UGT1A1 patients would result in

unnecessary reductions for 50 percent of the

patients, and the outcome is unknown.

We think that this data furnishes a

provocative signal hinting at the biology and

provides guidance for additional ongoing research

in this area. We recognize the importance of this

data that has been collected so far, but we also

feel that more research is necessary. As mentioned

earlier, we have many ongoing sponsor and supported

trials that investigate the UGT1A1 and other

genetic factors and their association with severe

neutropenia and diarrhea. In parallel, we have

ongoing discussions with the FDA to understand the

implications of the published data and what may be

an appropriate use of this data in the Camptosar

label.

These are the sponsored or supported

studies that include Irinotecan-treated patients

and that have a pharmacogenomics component. These

studies may address some of the limitations of

previous published studies, in particular, sample

size, the analysis of multiple genetic factors, the

possibility for controlling for known factors, and

the inclusion of current standard of care regimens

for first line metastatic cancer or colorectal

cancer patients FOLFIRI and FOLFOX.

Although most of these studies are

ongoing, we would like to highlight that study

N9741 is finished, and the NCCTG has almost

completed an analysis of pharmacogenomics data for

15 polymorphic markers on 10 genes. A publication

is planned in the near future by the NCCTG.

The future looks very promising. The data

from these studies will provide important new

information in addition to other efforts and other

studies that are being conducted by other

investigators. We hope that from these studies, we

can better define the magnitude and strength of the

association between UGT1A1 and safety; we can also

identify other potential covariants of severe

neutropenia and diarrhea, and as the data matures

from the ongoing studies, we look forward to

providing additional information for health care

providers and patients to aid their treatment

decisions.

I would like to acknowledge a large team

of Pfizer colleagues who have worked together to

provide this presentation this morning. Thank you.

DR. VENITZ: Thank you.

Any questions, comments by Committee

members?

Can I ask you to go back to your slide

100

where you discuss the performance of the test? You

focus on the sensitivity. Would you care to

discuss the specificity and the negative predictive

value?

DR. PARODI: The clinical specificity

basically gives the overall probability that given

that a patient has neutropenia, does not have

neutropenia, will not have a 7/7 positive test. In

general, the clinical specificity seems quite high.

The negative predicted value gives a probability

that given that the test is negative for 7/7, what

is the probability that that patient will not

develop neutropenia?

The overall values that you see there are

averages. They are not weighted averages. We feel

that a negative predicted value is relatively high

and much better than the positive predicted value.

DR. VENITZ: So would it be fair, based on

this analysis, then, to say if you did the test on

a large number of patients, you may not necessarily

predict neutropenia with a 50 percent sensitivity,

but if your star-28 is negative, you have a very

101

small chance of developing neutropenia?

DR. PARODI: Certainly, if you get a

negative value--a negative test for 7/7, you have a

high probability of not developing neutropenia than

if you had a positive value for 7/7; basically, a

50 percent chance of developing neutropenia is

equivalent to a toss-up.

DR. VENITZ: Right, but that's from your

perspective the bad thing, but the good thing is,

on the other hand, if you have a negative test,

you're unlikely, very unlikely, to develop

neutropenia; is that correct?

DR. PARODI: That is correct, but overall,

given the incidence of neutropenia, the likelihood

that you will develop neutropenia anyways is low.

DR. VENITZ: Any other questions?

Yes?

DR. SINGPURWALLA: Yes, I was just looking

at your first slide where you had the odds ratios

and the confidence limits around the odds ratios.

DR. PARODI: Which one? For neutropenia

or diarrhea?

102

DR. SINGPURWALLA: Well, let's just take

that one.

DR. PARODI: Okay.

DR. SINGPURWALLA: I can't pronounce all

these things.

DR. PARODI: Oh, me neither.

DR. SINGPURWALLA: So let me try and

understand the objective of this slide from a

layperson's point of view.

DR. PARODI: Right.

DR. SINGPURWALLA: If I was 7-7, and if I

took this medication, then, it appears that there

is a 50 percent chance that I'll get an adverse

reaction; is that correct?

DR. PARODI: That is correct. You have a

higher probability of having an adverse reaction.

DR. SINGPURWALLA: And if I was either a

6/6 or a 6/7, I have a lower probability.

DR. PARODI: That is correct.

DR. SINGPURWALLA: If I pool all those

numbers, I find the answers to be 50 percent

probability and 25 percent probability roughly, if

103

I just add everything now.

DR. PARODI: Yes, yes.

DR. SINGPURWALLA: So I'd still be scared

if I had a 25 percent chance of an adverse

reaction.

DR. PARODI: Yes, and you should be.

DR. SINGPURWALLA: Is that the point

you're making?

DR. PARODI: The point we're trying to

make is we try to really provide--we'd have liked

to have done a meta-analysis of this data. Given

the differences, significant differences in these

studies, it's really unfair to pool all this data

together in a meta-analysis exercise. So we are

basically presenting this data in a tabular form,

using a calculated statistic as a comparator

between studies, because not all studies reported

the same statistic, as Dr. Rahman indicated. I

mean, some people used one statistic; some the

other.

So it made the tabulation and the

comparison between studies a little bit difficult

104

but--

DR. SINGPURWALLA: I see your point, but

look at the confidence limits. They're so wild.

DR. PARODI: The confidence limits are

very wide, and that is primarily due possibly to

the fact that these are very small sample sizes.

DR. SINGPURWALLA: That's right.

Therefore, it makes sense to pool them.

DR. PARODI: It makes sense to pool them,

but, I mean, we're doing this in a highly

abstracted way, because I think pooling the data is

really not warranted. This is basically an

exercise, and also, we have not adjusted for known

factors, because it is difficult to extract from

the policy literature what was the performance

status of the patient, what was the baseline

bilirubin. All of these adjustments will have to

be made, since these are known covariants in the

incidence of neutropenia.

DR. SINGPURWALLA: Can I suggest that you

consider the use of prior odds and the posterior

odds in these kinds of studies?

105

DR. PARODI: Can you be more explicit?

DR. SINGPURWALLA: Well, prior odds are

you put prior distributions on these ratios, and

you compute the aposterior using these.

DR. PARODI: Okay; prior probabilities

and--

DR. SINGPURWALLA: Right.

DR. PARODI: We could do something like

that.

DR. PAZDUR: I wanted to bring up some

clinical issues here, and perhaps I realize that

the company Pfizer kind of got this drug from

Pharmacia, who did most of the development on this,

and feel free, obviously, to discuss these with

your clinical colleagues--

DR. PARODI: Right.

DR. PAZDUR: --if they have an issue with

this. But I think it's very important for the

Committee here to understand the clinical

development of this drug, and obviously, we'll be

talking about an effect on dose reduction and

potentially a potential reduction in efficacy. And

106

I wanted to give the Committee some idea about how

the dose of this drug was selected on either

schedule. There's two schedules on the product

label: a weekly schedule times four and then an

every three weeks schedule.

How is that dose selected in the 1990s

here? And that has carried us forward here

throughout the entire clinical development, and I

was wondering if you could give us some idea: how

did you get this dose? What was it based on?

DR. PARODI: I would like to defer to one

of my colleagues to answer that particular

question.

MR. MORRISON: Okay; thank you. I'm Mark

Morrison. I'm the medical team leader for

Camptosar in the U.S. I've been with Pfizer, so I

don't have first hand experience of the development

at Pharmacia; however, the dose was arrived at by

the standard mechanism of looking at MTD and

pushing the dose up to the MTD and then backing

down to a tolerable dose just below MTD. So it was

a standard development.

107

DR. PAZDUR: Well, I guess the point that

I wanted to bring out here, when the dose was

selected, we were looking in a 5-FU refractory

population, and the dose was being looked at in

terms of response rate here, okay, which was

relatively modest. We were looking at 15 percent

response rate. And the point that I want to bring

out is what is the relationship that the company

has with dose and a clinical outcome of an impact

on survival? Because here again, if we talk about

perhaps changing the dose, you have to be cognizant

about any missing data that we have on a dose and

the ultimate clinical outcome and what is that

level of certainty that we have regarding that

dose, the package insert dose and clinical outcome?

MR. MORRISON: That's actually something

that we proposed to look at going back into the

databases. What we do know, in the first line

setting, we use a combination of bolus IFL is that

with dose reduction after cycle one and follow them

out, each group is dose-reduced, and you do see a

slight trend for a difference in efficacy; however,

108

it's not statistically significant, so that you'd

have to come to the conclusion that the overall

efficacy at the end of the day was very similar,

because both wind up being dose-reduced more.

We do need to look at that in the

single-agent setting going back to the second line

studies. One thing I'd like to point out, if we go

back to the probability slide, looking at negative

predictive value, I think a very important point to

make in looking at the label, the incidence of

grade four neutropenia would be expected if you

average the two trials, which isn't statistically

valid, because they're two different populations,

but it comes out to about 18 percent.

So the negative predicted value tells you

that you have a 17 percent chance of having the

effect or an 83 percent chance of not having the

effect, and that's what we know to begin with. So

the test actually is more indicative of the

standard population. The positive predictive value

of 50 percent gives us added information that these

patients are at increased risk over the general

109

population; however, we would like a test with a

positive predictive value of 90 or 95 percent.

So what other factors are coming into

play? Is it the carboxylesterase, for example? Is

it transporters? Camptosar itself is present in

micromolar concentrations versus SN38 in nanomolar

concentrations.

So given the difference in efficacy

between the two compounds, they're both present at

therapeutic concentrations, and UGT is important

for SN38 much more so than for Camptosar, and the

carboxylesterase may be a very important factor.

We don't know whether it is; upcoming data from our

clinical trials will hopefully give us an answer,

and likewise, transporters in the bone marrow and

in the gut and in the liver may help us unwind the

story of selectivity and look at the therapeutic

index and try to figure out what combination of

factors might give us a positive predictive value

greater than 90 percent.

So we're striving to do that, and we will

have data available in the near future from a

110

number of trials that Dr. Parodi has mentioned to

try to improve on that ratio.

DR. PAZDUR: Just one last question. We

spent a lot of time on this slide. I've seen it

put up now three or four times, and it addresses

severe neutropenia. However, you know, if you ask

medical oncologists that actually use this drug, if

you ask them what are the top 10 toxicities with

Irinotecan, one to nine would be diarrhea,

diarrhea, diarrhea, diarrhea, diarrhea, diarrhea,

diarrhea, diarrhea.

So is it really a fair--just to look at

neutropenia here, are we really missing something

by not really looking at what is the most

clinically relevant toxicity, and that is either

diarrhea alone which leads to the hospitalization

or, more importantly, diarrhea in the presence of

severe neutropenia, which generally is very

problematic and is usually associated with the

deaths that we have seen on this drug?

So I would like to make sure that the

Committee understands, you know, the clinical

111

relevance of neutropenia that we're talking about

here. Severe neutropenia in oncology circles, we

deal with on a daily basis here. The real toxicity

with this drug that we should be paying attention

to is diarrhea and severe diarrhea that will lead

to the patient's hospitalization.

One last question, just to give the

Committee an idea of kind of the softness on the

dosing on this drug. If one would take a look at

the single agent use of Irinotecan before it went

into combinations, at the labeled doses of 125

milligrams per meter squared, how many people could

actually be maintained on that full dose?

MR. MORRISON: I think the dose intensity

for the various drugs ranges from about 70 to 80

percent.

DR. PAZDUR: Okay; but how many people

would require dose reductions, I'm asking

basically?

MR. MORRISON: By cycle two, I know in the

IFL data, for example, in first line--

DR. PARODI: Single agent.

112

MR. MORRISON: Yes, single agent, I don't

have that figure.

DR. PARODI: It's usually the majority,

usually a high, high number of people.

MR. MORRISON: And if I could comment on

the diarrhea issue, this is something we're

absolutely looking at, and we're very concerned

about neutropenia occurring in the presence of

diarrhea when we've got endothelium or epithelium

is compromised. And we would like to see what

correlates with diarrhea, and I think more

importantly, we're urging the medical community to

use the infusional regimen of 5-FU, which has been

shown to cause less diarrhea and actually seems to

be more efficacious.

So we're actually advocating use of the

FOLFERI regimen, which is a two-day infusion of

5-FU preceded by a bolus of 5-FU, and the

Irinotecan and 5-FU are given once every two weeks.

So that seems to be a much more tolerable regimen;

requires less dose reduction; and appears to be

more efficacious, and we are looking at these same

113

pharmacogenomic correlates for UGT and a whole host

of other genes with that regimen in our ongoing

trials.

DR. WATKINS: Just a quick point, which is

obvious to everybody, I'm sure, on the panel here.

But as a newcomer, it obviously would be very nice

if the genomic DNA and the clinical data was

available on the patients who went through the

studies back in the early nineties, that would

greatly be to patients' benefits. And I guess

since you started off the talk talking about

Pfizer's global role in pharmacogenetics and as a

leader in R&D, is Pfizer now routinely collecting

genomic DNA and creating databases and bank so that

when such questions come up in the future for

drugs, you can very quickly go back and--

MR. MORRISON: Yes.

DR. WATKINS: --verify these rather than

doing large phase four studies?

DR. PARODI: Absolutely. We have a very

large commitment in the company to

pharmacogenomics, including systematic collection

114

of DNA samples, and we are actually developing the

right infrastructure to store and retain this

valuable asset for future investigations.

MR. MORRISON: And we're actually looking

at this not just in terms of Camptosar, but we're

looking at genes involved in a number of other

compounds as well. We're looking at genes involved

in metastasis; for example, in a protocol that

we're just getting ready to launch, we're going to

have tumor samples from the primary tumor and from

liver metastases in a neoadjutant program.

So we're looking at genes that are

involved in invasion metastasis and responsive

therapy, so we're trying to look at everything

across the gamut.

One thing I neglected to mention was we're

also looking very carefully at bilirubin levels,

and we do have in our label a statement concerning

data looking at bilirubin in the normal range,

even. In the range of 1.0 to 1.5 milligrams per

deciliter, there is a significant increase in

toxicity. And that is within the label, and it's

115

brought to clinicians' attention.

And this is statistically significant

compared to patients with bilirubins less than 1,

so we're very concerned about this. And we would

like to see how bilirubin correlates with UGT, and

maybe Luis can comment more on that, because I know

in the analysis by Dr. Innocenti, that was

addressed.

DR. VENITZ: Steve.

DR. HALL: Yes, I noticed on your initial

slide talking about the metabolism of the drug and

its metabolites that only the UGT1A1 was mentioned.

And that seems to be the theme so far.

Now, there's a growing literature, and

Kathy's just done a literature search here on the

computer that, you know, other UGTs are involved,

and in your own materials that you supplied to us,

there was a study from the Foxchase Cancer

Institute, I think, that implicated for sure the

1A7 UGT also as a contributor.

So I wondered if you had any information

on the impact of the other UGTs, and secondly, I

116

noticed in the list of genes that you plan to look

for in the studies you listed again in your

materials, there were no other UGTs mentioned, and

I wondered if that was something that in the

short-term, you would be able to get some concrete

insight into rather than in the longer-term

studies.

DR. PARODI: I think we have, as I

indicated, sponsored the N9741 study, which has

established a collection of DNA samples from

Irinotecan-treated patients. And maybe there is an

opportunity in using those samples to investigate

other candidate genes that may be associated with

outcomes.

In our earlier studies, we had not

collected a DNA sample from the earlier

registration studies, but as I indicated in another

slide, we had on the other collaborative studies

that we're conducting right now, we are collecting

samples for future analysis. So if we wanted an

answer about, well, what about UGT1A7, I think a

more immediate answer can come from maybe

117

genotyping those samples from the 9741 and getting

an answer.

Maybe Dr. McLeod would like to comment on

that.

DR. MCLEOD: In the context of several of

the GI intergroup studies which I am involved in as

well as several others in this room, Dr. Ratain,

Dr. Giacomini, we have tried to take a drug pathway

approach that does not focus on any one particular

element of the disposition of the drug. And so,

missing from this slide here is also many of the

pharmacodynamic markers that are starting to come

out of the some of the screening systems we have.

And so, taking this pseudoholistic

approach as much as our knowledge lets us, we're

trying to understand these issues. So any genes

that come out of these screens are fair game and

hopefully will complement the additional data

that's available.

DR. SADEE: I just want to bring up the

issue of how we use genotype and also as a question

on this. Obviously, and just looking at the 6 and

118

7 alleles for the gene, including the enzyme, we

will have three genotypes. One is you get both

7/7; that's homozygous for one; and then, you have

homozygous for 6/6; and then, you have the third

population that's heterozygous.

Now, you have chosen, in most of your

slides, to combine the homozygous 6/6 with the

heterozygous 6/7. And in this particular case,

it's very likely that there is additional

functional polymorphism in this gene. So you have

a much greater chance for a large variety in the

heterozygous population that you included with the

6/6 or the 6/6 and the 6/7. They're very

different. And you also have, in the 6/7

population, clinical symptoms for--suggesting that

this is truly, again, a different population.

So my question is how do we deal with, if

we make dosage recommendations, and we have one

population where it's very uncertain; we have one

population--that would be the heterozygous

population; one where it's--and the two others are

more certain, I would say, or we have better

119

predictors.

So can we actually combine them, those

populations, heterozygous and homozygous, and would

that be a basis of making decisions along the lines

of dosages?

DR. PARODI: The reason which we combined

6/6 and 6/7s was basically because we saw that

those groups, at least in the reported data,

behaved almost equivalently. So from the phenotype

that we're looking at was almost indistinguishable.

So that was the reason why we--of course, when

you--I take your point that even within one of

these genotypes, like the--even including the 7/7,

it can be genetically very heterogeneous, because

any of these groups can be genetically

heterogeneous.

But from the point of view of the

phenotype, it certainly made sense to us to combine

6/6 and 6/7s and compare that to the 7/7s.

DR. LESKO: Yes, it's really two questions

with regard to the information provided. The first

is thinking about risk factors separately versus

120

complements. You mentioned there appears to be a

relationship, albeit imperfect, between

pretreatment bilirubin levels and the level of

toxicity and neutropenia, and that's consistent

with the literature. There's been several articles

that have pointed in that direction with modest

predictive values, let's say.

However, if I were to--and this may be

something that's worth calculating--if I were to

take individuals with certain preexisting bilirubin

levels and then add to it as a complement the

genotype information, would I then increase

predictive value in terms of my risk of toxicity,

my individual risk of toxicity, to the point where

it would be higher than it is?

That would be sort of the one question,

and one could go back and look at that, I'm sure,

with the data in the files, and taken together,

those two indicators, I think, would give a pretty

good indication of an individual's risk of

developing toxicity.

The second question is with regard to

121

dosing. I mean, from our discussions, it's clear

that dosing in the label is based upon some

clinical studies, but there's also, as I recall,

some dose reductions recommended in the label. And

I don't have it in front of me, but I think elderly

was one of those. And there's some measure of dose

reduction in the elderly. And I'd wondered if

there's any information on either--well, the

information on what was used to lower that dose,

and is there any exposure data in elderly that

would be related to the exposure you would see in

genotype?

In other words, I'm trying to draw an

analogous situation between lowering the dose in

terms of elderly, because they have a certain

exposure of SN38 area under curve and then

comparing that to the area under curve that we see

in the genotypes and see if there's any logic to

using that as a guide to what dosing reductions

would be done.

DR. PARODI: If I can answer the first

question, and maybe Dr. Morrison can answer the

122

second, with regard to the correlation between

baseline bilirubin levels and the neutropenia,

actually, the Innocenti paper models this

correlation and actually, in the electronic version

of the paper, they offer to deposit the data at the

publicly available genomics database.

I have checked the database, and it was

not publicly available yet. So it might become

available. And then, once that data is available,

we could attempt the modeling. They report a

multivariate analysis adding a genotype with

baseline bilirubin, and in their modeling, both

genotype, baseline bilirubin and sex were

determinants of the correlation.

MR. MORRISON: Regarding the second

question, I can't comment on exposure to Irinotecan

or SN38; however, the decision to use clinical

judgment to perhaps decrease the dose level in the

elderly was based on an increase in late diarrhea

that was seen in that population.

DR. PAZDUR: But didn't Upjohn do a study?

Perhaps Dr. McGovren could comment on this, on the

123

elderly, and it had a PK component?

DR. MCGOVREN: Yes, yes.

DR. PAZDUR: And I believe I was one of

the authors on that.

[Laughter.]

DR. VENITZ: Would you introduce yourself,

please?

DR. MCGOVREN: For various reasons, that

data has been a long time being put into a report

form, and--

DR. PAZDUR: Yes, I know that.

DR. MCGOVREN: And in fact, that data will

be filed with the agency very soon. Age is

probably not the best example to go into here,

because in fact, in that study that you

participated in, there really was no association

between age and diarrhea or age and PK.

DR. VENITZ: Can you introduce yourself

for the record?

DR. MCGOVREN: I'm sorry; it's Pat

McGovren from clin pharm group at Pfizer.

DR. PAZDUR: And for the record, that long

124

time was how many years?

DR. MCGOVREN: It was probably about five

or six years, yes.

DR. VENITZ: Okay; thank you.

DR. MCGOVREN: For the record, do you want

to indicate what that gesture means?

[Laughter.]

DR. VENITZ: Speaking about the record, I

think it's time for a break. We'll take a break

until 10:45 and reconvene for the Committee

discussion. So at 10:45, we'll reconvene.

[Recess.]

DR. VENITZ: Okay; welcome back, everyone.

Our next and last speaker for our first topic is

Dr. Mark Ratain. He is one of the authors of the

paper that was discussed in detail earlier on, and

he's going to give us his perception and

perspective as to how to use the test for

UGT1A1*28.

DR. RATAIN: Good morning. Thank you. I

very much appreciate the opportunity to speak here.

I want to thank Dr. Lesko and Rahman from FDA for

125

inviting me. I really want to thank Pfizer and

their predecessor companies for providing drug for

my trials. Actually, it goes back into the early

nineties with our initial trial. And I thank the

indulgence of the Advisory Committee.

Some of you are totally overexposed to

Irinotecan already.

[Laughter.]

DR. RATAIN: And that goes long before

this particular meeting.

Now, I've been working on this drug since

the early 1990s, and I was specifically asked today

to speak to you as a clinician, as a medical

oncologist, and how I would see from with obviously

a biased view how this test could be used to

enhance the treatment of patients.

So, many of you have seen a poster child

before. The poster child is the one on the right

here, 6-mercaptopurine. And you're very familiar

with this, as was alluded to in the important work

that's been done by many, particularly the group at

St. Jude, and this is a figure from the Nature

126

Reviews cancer paper by Mary Relling and her

colleague showing the relationship of TMPT

polymorphisms to the therapeutic index of

6-mercaptopurine in children with acute leukemia.

Now, here's another poster child. This is

from pharm GKB. This is the Website of the NIH

pharmacogenetics research network, and this was our

poster child. This was the first pathway that went

up. This is a pathway that some members of this

advisory committee have agonized over, and I would

urge you to take a look at it. It is a clickable

interface. And in theory, it should get you the

data that you're looking for; at least that's what

we're told. Some of us have actually been able to

retrieve the data out of this database.

But I think it's pretty clear that this is

a very complex drug. You see the parent drug here.

The parent drug is inactive. The only way the

parent drug becomes active is when it's hydrolyzed

by carboxylesterases, and there are

carboxylesterases within cells. So you can, if you

expose cells to Irinotecan and the cells contain

127

carboxylesterases, the drugs will get hydrolyzed to

SN38 and become activated. But by itself, it is

inactive. And the primary enzyme responsible for

the hydrolysis is CES-2.

Irinotecan is also a substrate, as shown

here, for CYP-3A. Both CYP-3A4 and CYP-3A5,

although CYP-3A5 has a relatively minor role; the

major oxidative metabolite is APC. That is formed

only by CYP-3A4, and I will add we have looked at

CYP-3A polymorphisms in our studies. We have not

been able to correlate it with anything.

SN38 is a substrate for glucuronasil

transferases. As far as the glucuronasil

transferases that are expressed in the liver,

UGT1A1 far and away is the most important enzyme.

There is probably a minor contribution of UGT1A9.

We have not been able to demonstrate any role for

UGT1A6 in the metabolism of SN38.

And there are other UGT1s that are

expressed in the gut, particularly UGT1A7 and

UGT1A10 that do have the capability of

glucuronidating SN38.

128

And finally, SN38 as well as the parent

drug and the glucuronide are all substrates for a

variety of transporters, and we are actively

looking at the relationship of polymorphisms in

these transporters to the pharmacokinetics and

toxicity of the drug. We have not been able to

find any relationship between polymorphisms and

MDR1 or ABCB1 and clinical outcomes, but we have

some preliminary data that was presented at the

American Society for Clinical Oncology meeting this

year on a polymorphism in ABCC2, also known as

MRP2, and the pharmacokinetics plasma

concentrations of parent drug as well as APC and

SN38 glucuronide, and we currently have some work

in progress looking at possible relationships of

subtleties such as haplotypes and other clinical

outcomes, but that is all work in progress that has

not even undergone any internal statistical review.

So I really want to focus you on the

subject of today, UGT1A1. Because as you've heard,

this is a polymorphism that, one, is common, and

two, for which there have been multiple studies.

129

Now, this is a study, and I can't remember whether

this was a pharmacy or an Upjohn study, the mass

balance study of Irinotecan. And I think that this

paper, published in Drug Metabolism and Disposition

in 2000, gives you some idea of what happens to the

drug.

And in a mass balance study, 55 percent of

the drug is found, is excreted as parent drug.

Nine percent is SN38; 3 percent is SN38

glucuronide. Only 11 percent is this oxidative

metabolite APC. Only 1.5 percent as NPC, another

oxidative metabolite. So as far as metabolites,

you can see that this pathway, down SN38 and SN38

glucuronide is pretty important, but also, there's

a lot of parent drug that comes out unchanged,

which does make you wonder about the importance of

polymorphisms and transporters.

Now, as you've heard, Irinotecan is a

cytotoxic agent approved in the United States for

metastatic colorectal cancer. It is usually

administered these days in combination with 5-FU,

and I will add is also active in many other

130

malignant diseases, and it's commonly used

off-label for other solid tumors.

Its usage is definitely limited by

toxicity, both actual toxicity and perceived

toxicity. The toxicities include life-threatening

neutropenia and associated infection. This appears

to be most common on the every three week schedule.

And the other major toxicity, clearly the one that

is more problematic when it occurs, is severe or

life-threatening diarrhea, requiring parenteral

fluids and/or hospitalization, and this occurs

primarily on the schedule, the weekly schedule,

which consists of four weeks on, two weeks off.

And in our hands, at least, this diarrhea

really is not very common on the every three week

schedule. So clearly, we have different

pharmacodynamics going on on these two different

schedules. And so, it is very important not to

lump studies together, particularly when looking at

the diarrhea, because of the schedule-dependent

effects, as well as the confounding issue of

concomitant drugs such as 5-FU, which commonly

131

causes diarrhea.

Now, I think from a clinician's

perspective, Irinotecan is one of many FDA-approved

choices for metastatic colorectal cancer. And the

discussion to date has focused on if one chooses to

give Irinotecan, what does one do? And I think

from a clinician's perspective, one has to also ask

the question: might genotyping help the clinician

decide among the various choices?

So for first-line therapy, you have 5-FU,

which nowadays is always given with leucovorin,

folinic acid, a modulator of 5-FU, Irinotecan,

oxaliplatin, which is a platinum analogue that has

a totally different mechanism of action from any of

the other approved drugs; capecitabine, which is an

oral fluoroprimadine and is very similar to 5-FU,

and bevacizumab, Genentech's monoclonal antibody

against veg-F. And these are all approved for

first-line therapy in various combinations.

For second-line therapy, one has a choice

of Irinotecan, oxaliplatin, 5-FU with leucovorin,

or cetuximab, the monoclonal antibody against EGFR

132

marketed by Inclone and Bristol-Myers-Squibb. So

again, one has many choices.

So how might clinicians choose among

various choices? Well, one is clearly personal

experience. Two is interpretation of phase three

data; three, marketing influences; four,

reimbursement; five is a very controversial piece,

chemosensitivity testing. There was a recent story

in the Wall Street Journal suggesting that this

should be done. The American Society for Clinical

Oncology has reviewed this and really, there are no

good data as to how one might use chemosensitivity

testing in an infectious disease kind of model to

decide among treatments. And then, the one we're

talking about today, genotyping, whether one can

predict toxicity or one can predict activity or

efficacy. These might influence how a clinician

would choose among the various options.

So I want to review with you some of the

clinical data, so you can see the dilemma. So this

is a study from the North Central Cancer Treatment

Group, and this was a prospective randomized study,

133

three different regimens: N9741. And you see that

two of the regimens, IFL and IROX, included

Irinotecan, and one regimen, FOLFOX4, did not

include Irinotecan.

And this study demonstrated that the two

Irinotecan regimens were both--that FOLFOX4 was

superior to IFL, p value 0.0001, and that IROX was

superior to IFL, so that the conclusion of many

clinicians from this study was that FOLFOX4, a

regimen that does not contain Irinotecan, was the

preferred first-line therapy.

Here is a more recent study from the New

England Journal of Medicine. This regimen used

IFL. This is a regimen that was shown to be

inferior in the previous study, and combined it

with Genentech's monoclonal antibody, bevacizumab,

and this study showed that IFL plus bevacizumab is

superior to IFL, and this study led to the approval

of bevacizumab for the first-line treatment of

metastatic colorectal cancer. And the label does

not say in combination with this IFL regimen; the

label says in combination with any 5-FU

134

leucovorin-containing regimen.

So again, the clinician is still

struggling with what to do. The only published

data is in combination with IFL here.

And then, there is this trial by

Tournigand, a European trial, published in the

Journal of Clinical Oncology this year. This was a

randomized trial that compared FOLFIRI, an

Irinotecan-containing regimen, to FOLFOX, a regimen

that does not contain Irinotecan. Prospective

randomized trial; 113 patients per arm, and then,

second line patients crossed over to the

alternative therapy.

And what this study showed was that

basically, for first-line therapy, the two

regimens, the Irinotecan and the non-Irinotecan

regimen, were comparable from the standpoint of

response rate. Again, bringing the clinician back

to wonder what's appropriate first-line therapy?

And when one looks at survival, again, you get the

same survival no matter what you start with.

So the clinicians treating colorectal

135

cancer need all the help they can get.

Now, Oncoscreen, a German company, has

taken advantage of this dilemma and is marketing a

commercial test for UGT1A1 genotyping, and you can

go to www.oncoscreen.com, I think, and you can

read--part of it's in German, and part of it's in

English, and part of it's in misspelled

English--about the side effects of Irinotecan, also

known as CPT11, and polymorphisms in the promoter

region of UGT1A1. And it gives you the address,

and you can send blood here. I've never tried, and

I have no idea how well they're doing, but they've

taken advantage of this opportunity to actually

market the test.

And this is the data from our study, the

Innocenti study that you've heard about, shown in

greater granularity. And this was 66 patients

enrolled prospectively as you've heard. And the

study was powered around trying to show a trend, a

significant trend, 6/6, 6/7, 7/7, although the

original study design was powered to look for

diarrhea, which at the time we started the study,

136

we did not understand the schedule dependent

differences in the diarrhea, and so, we ended up

looking at neutropenia as the endpoint.

As you've also heard, there are other

polymorphisms: allele 5 and allele 8. Allele 5

has been suggested to have higher glucuronidating

activity than a 6, and allele 8 has been suggested

to have lower glucuronidating activity than the 7.

And in response to the question previously asked

about ethnicity, the study was primarily caucasian,

not exclusively caucasian. I believe there were

one or two Asian patients, and there were certainly

some African-American patients in the study, but

there were certainly not enough within any

population subgroup to stratify for that.

And you see that there was a significant

trend with the 7/7s having a lower absolute

neutrophil count nadir than the other two groups,

with the 6/7 being intermediate, but clearly, the

difference between 7/7 and 6/7 is greater than the

difference between 6/7 and 6/6.

I will also add that if you want to

137

translate absolute neutrophil count nadir to grades

of neutropenia, grade three neutropenia is less

than 1,000. That's probably not clinically

significant, although it can affect subsequent

dosing. It might result in delays of treatment if

you develop grade three neutropenia.

Grade four neutropenia is an absolute

neutrophil count nadir of less than 500, and a

patient who has grade four neutropenia, who

develops a fever, is essentially automatically

admitted to a hospital and treated with parenteral

antibiotics. And so, it is very common to get

fevers when you're neutropenic, and so, that that

is a real morbidity and a real cost issue.

So others have addressed the issue of

sensitivity and specificity, and I'm just going to

again go through our data, and again, this is a

single study: 350 milligrams per meter squared

every three weeks and looking at grade four

neutropenia, the clinically significant

neutropenia, this is the extreme, less than 500.

And again, we agree with the Pfizer analysis. The

138

sensitivity is 50 percent of patients who have

grade four neutropenia who are 7/7. Specificity:

95 percent of patients who do not have grade four

neutropenia are not 7/7.

And the positive predictive value: 50

percent of patients who are 7/7 have grade four

neutropenia, and the negative predictive value, 95

percent of patients who are not 7/7 do not have

grade four neutropenia.

Now, let's put this into the context of

without testing and with testing. Without testing,

100 percent of patients are treated, and 10 percent

have grade four neutropenia. If you chose not to

treat the 7/7 patients, with testing, 90 percent of

patients are treated, and approximately 5 percent

would have grade four neutropenia. You would have

a 5 percent absolute reduction. You would test 20

to protect one.

So I put out what is my bias but I believe

is still hypothesis that pharmacogenetic testing

will improve outcomes in metastatic colorectal

cancer. That's really what we're here to discuss.

139

And I believe it will allow the clinician to select

a drug regimen based on patients' genetic, and now,

we're talking germ-line polymorphisms, genetic

characteristics, that this will lead to reduced

toxicity and potentially will lead to increased

efficacy, something that we've not previously

talked about.

My opinion is that sufficient data exist

to recommend that patients who are homozygous for

the star-28, the 7/7, should not receive Irinotecan

at standard doses. Some might say that you could

treat at standard dose and accept greater toxicity.

Some might say you could reduce the dose. I

believe that these patients would be most

appropriately treated with an alternative regimen

such as an oxaliplatin-based regimen that has the

same survival outcome as an Irinotecan-based

regimen. I cannot sit here and recommend reduced

dose, because we have no clinical data to show that

patients treated with 7/7 at a reduced dose have

comparable activity and comparable survival

outcomes to patients treated with alternative

140

regimens.

On the other hand, the optimal treatment

of patients who are at reduced risk of Irinotecan

toxicity star-one, star-one, or 6/6 is unclear.

Should they be treated with standard

Irinotecan-based regimens? Should they be treated

with high-dose Irinotecan-based regimens? There's

one European study that took patients treated with

standard dose, escalated patients who did not have

significant toxicity; they escalated them up from

350 per meter square to 500 per meter square. It's

a single-arm study, but it's got the highest single

agent response rate of any study in the literature,

and so that this may be an opportunity to reexplore

dose in a low-risk group of patients representing

50 percent of patients that are candidates for this

drug.

Or is oxaliplatin the best regimen for

these patients? We have no data to support that,

particularly for the low-risk patients.

I want to contrast this with other drugs

and other polymorphisms, because I think this is a

141

great opportunity to use pharmacogenetics to

individualize treatment of colorectal cancer. And

this is Bob Diazio's Website, www.dpdenzyme.com,

where you can learn about screening patients for

DPD enzyme deficiency. Oncoscreen also offers this

test.

What do we know about this test? And

here, you see the Oncoscreen Website, and it says

this test is supported by the German health

insurance companies. Actually, the German health

insurance companies initially--the German oncology

group initially recommended this test and then

retracted the recommendation, which is kind of

interesting. There's a history there.

And the most common mutation in DPD is an

exon 14-skipping mutation. This has an allelic

frequency of approximately 1 percent. The star-7

polymorphism has an allelic frequency of

approximately 35 percent. So there's a big

difference in allelic frequency here. DPD testing,

if you test this exon 14-skipping mutation, and

your endpoint is grade four, life threatening 5-FU

142

toxicity, without the test, all patients would get

treated, and approximately 3 percent of patients

have this toxicity. So only 3 percent of patients

have grade four toxicity from 5-FU as a single

agent.

With the test, you would treat 98 percent

of patients, and approximately 2 percent of

patients will still have toxicity, a 1 percent

absolute reduction. You would test 100 to protect

one, so much lower efficiency of this test.

And then, there's another important

polymorphism that may predict for toxicity and

efficacy of fluoroprimadines, and that's a

polymorphic repeat sequence in the thymidylate

synthase gene that has been suggested to affect

translational efficiency but not gene expression.

And this is quite polymorphic.

Here, you see the population distribution

of this 28-base pair repeat. This is data from

Howard McLeod's group. And you see that the three

repeat is more common than the two repeat, and that

there's also a four repeat present in African

143

populations.

So with parting words: oncology is widely

anticipated to be the best model for demonstrating

the clinical importance of pharmacogenetics as it

relates to germ line polymorphisms. Colorectal

cancer is an important model, because of the large

number of active agents. We have candidate genes,

candidate polymorphisms and abundant clinical data.

And I want to thank my colleagues in the

PAR group, pharmacogenetics of anticancer agents

research group. I want to thank my colleagues in

the PGRN, pharmacogenetics research network, those

sitting here today, those I've collaborated with,

and those who have had to sit through far too many

discussions of Irinotecan.

So, thank you.

DR. VENITZ: Thank you, Dr. Ratain.

Any questions or comments by the Committee

before we start our overall discussion?

Paul?

DR. WATKINS: Just a question about the

UGT1A7, which is in the gut, and we've heard that

144

diarrhea is probably a bigger issue than

neutropenia. What work has been done looking at

UGT1A1 polymorphisms and diarrhea?

DR. RATAIN: It's a very difficult

problem, because there are definitely polymorphism

1A7 that have shown to be functional that are

strongly linked to UGT1A1, because it's all one

gene. And, in fact, the linkage in UGT1A1 goes

five prime at least down to UGT1A9. So to actually

distinguish the independent effect from 1A7 from

1A1 requires a very large study.

One would not--since 1A7 is not expressed

in the liver, one would not expect it to have a

significant effect on the plasma pharmacokinetics

or on the neutropenia, but it certainly is a

candidate gene for gastrointestinal toxicity. But

we really need a lot more data, because this really

will require haplotype based analyses of the whole

UGT1 gene.

DR. SADEE: Mark, this comes back to my

earlier question about dosage escalation in

populations. You mentioned that here that in

145

patients who are apparently protected against the

cytotoxic effects, you can go to higher doses and

get higher efficiencies. So I think that really

sets an important example to pursue that.

Do you have any other examples where that

has been pursued, so rather than looking at the

negative side, one would look where you want to

avoid things. You exploit the patients that really

should get a different dose.

DR. RATAIN: Well, I mean, you know, there

have been some studies in oncology where patients

sometimes get intraindividual dose escalation, but

there's really not a large data set on that. I

mean, Dr. Pazdur may have some comments.

DR. VENITZ: Howard?

DR. MCLEOD: Mark, we heard from Dr.

Rahman's talk about how the current package insert

includes data on age and bilirubin and some other

factors that I'm forgetting, public radiation and

one other thing, as risk factors and with a need

for dose reduction.

I wonder if you could put the 7/7 genotype

146

into the context of those existing risk factors.

DR. RATAIN: Well, we've looked at age in

our data set, and we have not found at least in our

study of 66 patients a significant impact of age.

So I would say from the standpoint of neutropenia,

genotype is certainly more important than that.

Bilirubin, in our hands, is a pretty good poor

man's genotype, but this is a single institution

where the bilirubin is collected in a standard way.

Once you get into multiple laboratories, and

bilirubin is tested at various times of the day

with various degrees of fasting, you're going to

really obscure the relationship between genotype

and bilirubin.

And so, I think that yes, patients with

higher bilirubins, particularly if it's

unconjugated, are very likely to be 7/7, because

many patients within the normal range of bilirubin

are 7/7. So, but I think even there are some

subtleties. I think again, patients with very low

bilirubins probably are not 7/7, and I've used that

in my clinical practice to help determine dosing in

147

the absence of an approved test.

DR. GIACOMINI: Yes, Mark, in your study

in which you documented the neutropenia, did you

also measure pharmacokinetically the SN38, and was

it higher in those patients with the 7/7?

DR. RATAIN: Yes, we did measure SN38.

SN38 is higher in the 7/7. SN38 correlated with

neutropenia. As I said, we have some evidence that

polymorphisms in ABCC2 through our collaboration

with Deanna Krebs may relate to the

pharmacokinetics of SN38 glucuronide, which makes

it difficult to interpret SN38 to SN38 glucuronide

ratios, which we previously assumed to reflect

solely glucuronidation. It quite possibly is

determined by both glucuronidation as well as

excretion.

DR. GIACOMINI: Let me ask a followup on

the bilirubin thing. Does bilirubin actually

competitively inhibit the glucuronidation of the

SN38 to SN30 mechanistically? Is it a competitive

inhibition, so when the levels of bilirubin are

low, it's telling you two things, one, about the

148

genotype but also about just direct

competitive-competitive inhibition?

DR. RATAIN: You're asking me does

bilirubin inhibit--

DR. GIACOMINI: Yes.

DR. RATAIN: We've not looked at that. I

don't know of any data. I would not expect it to.

And there is certainly some evidence not for--it's

possible that SN38 could inhibit bilirubin

glucuronidation if the levels are pretty low, but

there are certainly examples of other drugs,

particularly the protease inhibitors, that inhibit

UGT1A1 and do competitively inhibit bilirubin

glucuronidation.

DR. GIACOMINI: Okay; but you wouldn't

expect the bilirubin and the high bilirubin levels

to be inhibiting the SN38.

DR. RATAIN: Not in--not in--I don't think

so; I mean, Dr. Watkins would have a better feel

for that.

DR. DERENDORF: I'd like to come back to

your mass balance slide. If I understood it right,

149

only about 9 percent of the parent drug gets

converted to the SN38. So what do we know about

the other metabolites?

DR. RATAIN: Would you like for me to put

that back up?

DR. DERENDORF: Yes, you can. It's the

number five.

DR. RATAIN: Messed it up; sorry.

So, this is the mass balance slide you

were referring to. And I'm sorry--

DR. DERENDORF: Only 9 percent gets

converted to the SN38, right?

DR. RATAIN: Well, 9 percent is found as

SN38, and 3 percent is found as SN38 glucuronide.

DR. DERENDORF: Oh, okay.

DR. RATAIN: And again, this is a limited

number of subjects. These subjects were not

genotyped, but approximately 12 percent, I think

it's fair to say, goes down that pathway. I think

that's a reasonable estimate. And you see 55

percent in this study was--the parent drug was

excreted unchanged. About 12 percent is oxidative

150

metabolites, metabolites known to be formed by

CYP3A, and then, we don't know the rest of this.

And again, I was not an author of this study; just

presenting it for perspective.

DR. WATKINS: Just to address that issue

of can bilirubin itself interfere with the

glucuronidation of SN38 or any other drug, in

theory, that's possible. It certainly works the

other way around. There are some drugs that will

inhibit glucuronidation in patients who have a

genetic predisposition of Jolbert's. But I'm

unaware of any studies that have looked the other

way, so I don't think I can address that.

But the question I wanted to ask myself,

one of the concerns with using genotypes of the

host as opposed to the tumor in cancer chemotherapy

is the fact that genotype and phenotype don't

always go together, particularly in an ill cancer

patient on multiple drugs with cytokines, and

certainly, if their liver is completely replaced by

tumor, genotype is irrelevant.

And one of the very unique things here is

151

this particular phase two enzyme has an endogeneous

substrate, so in effect, you have a phenotype

measurement. And my assumption up until what you

just said was that that endogeneous probe for

UGT1A1 was not very good. But what you're saying

at your institution, it's in fact very good.

DR. RATAIN: There are two studies that

address this. There's our study that within a

single institution, all patients were on a research

protocol; the bilirubins were collected at, you

know, in a fairly consistent way just by nature of

our research practice, and it looked pretty good.

There's also a study that I was a coauthor on that

relates to a Pfizer study in which a large data set

was analyzed, and bilirubin really wasn't a very

good predictor, and this was just published in the

Journal of Clinical Oncology this year by Meierhard

is the first author, and the company may want to

elaborate on that further.

DR. WATKINS: Because if I can just follow

up, I mean, the key question is what does

genotyping add to the existing tool kit of the

152

oncologist? And my assumption in all the

background reading was that it adds a significant

amount. If an alternative is just standardizing

indirect bilirubin measurements, that's another

option that could be considered, I think.

DR. RATAIN: Again, you know, from an

analytical perspective, there's a gray zone, and it

doesn't--bilirubin doesn't, you know, in a large

data set may not correlate as well as something

that is a discrete answer like a 7/7 genotype. I

think also, it might be tough to distinguish 6/6s

from the 6/7s, and I think that the 6/6s are

potentially appropriate candidates for phase four

clinical trials looking at higher doses of

Irinotecan which I think is another important

reason to find a way to get this test in the hands

of the clinical oncologists and the research

oncologists to help further explore the

dose-response of this drug.

DR. HALL: So in part to follow on from

that, then, in your hands, what would your

recommendation be for the heterozygotes? Are they

153

to be treated or given an alternative?

DR. RATAIN: I would see no reason not to

treat the heterozygotes, given the data today.

DR. MCLEOD: One of the things you

commented on was that neutropenia is an important

toxicity, and I certainly second that. And during

the discussion, hopefully, we will elaborate on

that more, because it's a common problem that is

less of a worry but probably more of a problem to

the patients.

The question I have for you is you made

the comment that you thought the 7/7 patients

maybe--might represent a data set that should get a

different drug, oxilaplatin or something like that.

But with the current state-of-the-art and the one

for probably the foreseeable future is that every

colorectal cancer patient at a decent center will

get Irinotecan. If they don't get it first line,

they will get it second line.

And so, we can't really avoid the issue

that Irinotecan is going to appear. This is a real

drug for colon cancer. And they're going to get it

154

at some point, first line, second line, third line.

So it would be worthwhile, either now get your

comments or in the discussion, trying to think

about that issue, because the drug is approved in

both these settings, and so, we do have the remit

to actually talk about it in first line, second

line, et cetera.

DR. RATAIN: Well not everybody gets it

second line. There's some patients--

DR. MCLEOD: Not everybody does, but they

should.

DR. RATAIN: What I'm saying is if--some

patients don't get to second line. If you were to

give Irinotecan first line and have a grade five, a

fatal event, they won't get to second line. So you

would say, ideally, the clinician would like to

reserve the more toxic drug for second line rather

than first line.

DR. MCLEOD: But I totally agree with your

thinking behind it. I'm just--we can't avoid it.

I mean, the patients that make it through first

line because they didn't get Irinotecan, and they

155

were 7/7, the selection now, you've just taken one

drug off the table. It's now either single agent

Irinotecan or Irinotecan and bevacizumab, whatever.

We may not be able to come up with those

things, because like you said, there is no

prospective studies. But certainly, it is an

important issue.

DR. RATAIN: You know, I think as you

think about it as a clinician, you know, if you

have a discussion with a patient, and you sit down

with them, and you talk about Irinotecan versus

oxaliplatin as first line therapy, when you talk

about Irinotecan, you have to talk about

neutropenia, diarrhea, which can be severe,

life-threatening or even fatal. And as you talk

about oxaliplatin, you have to talk about

neurotoxicity that can be persistent.

And, you know, patients have to make

choices, and I think being able to inform patients

about their relative risk, particularly of the

toxicity that scares a lot of both patients and

clinicians, which is the neutropenia/diarrhea

156

complex, I think, is useful, and I think if one

could reassure a clinician that their patient is at

relatively low risk of these toxicities of

Irinotecan, a clinician would be more interested in

prescribing Irinotecan first line.

And so, I think that this actually is

helpful, very helpful to the clinician and I think

potentially very helpful to the sponsor who is

marketing the drug.

DR. BARRETT: You mentioned the lack of

prior appreciation of the scheduled dependence on

diarrhea, and I wondered, back on your

recommendations as far as not reducing the dose,

would you feel the same with neutropenia and the

diarrhea? And I guess the followup question there

is most of this data has been summarized outside of

the time dependency, so do you feel if that kind of

information is brought to light through either dose

reduction in the context of managing toxicities

that you could perhaps devise a schedule for one

toxicity versus another?

DR. RATAIN: Well, the diarrhea is a lot

157

messier to model; no pun intended. And the

neutropenia is pretty well correlated with plasma

SN38 exposure, and we can understand that both in

the context of these studies as well as in the

context of other studies, anticancer agents and

neutropenia.

The diarrhea is not fully understood, and

we've tried to model it in the past to somehow try

to model biliary excretion of SN38, and we have one

study that actually came up with a surrogate

endpoint, or, actually, biomarker would probably be

a better term for it, which was the CPT11AUC times

the SNC38AUC over the SN38 glucuronide AUC.

But it's very complicated. I think that

the thing one is most concerned about is the

neutropenia/diarrhea complex and particularly in

the context of schedules that are more frequent

than the every three weeks, which is where you see

this problem.

DR. VENITZ: Mark, as you know, the

Committee is asked to review the evidence to see

whether we would concur with the recommendation to

158

include pharmacogenetic testing. I'd like to know

what the competition is. In other words, I'd like

to know right now what is being done to come up

with a starting dose. How do you choose a starting

dose for Irinotecan with the information right now

without genetic testing?

DR. RATAIN: Right now, people just rely

on clinical evidence, which is one size fits all

based on body surface area, even though body

surface area has been shown not to correlate with

the pharmacokinetics of this drug.

DR. VENITZ: Followup question: how do

they adjust the dose once the patient is being put

on Irinotecan?

DR. RATAIN: I think clinicians do it to

some extent by the package insert and some extent

by their personal experience.

DR. VENITZ: Okay; thanks.

Larry, I think you had the last question

and then maybe frame the questions for the

Committee so we can start the questions.

DR. LESKO: Okay; thanks.

159

Yes, Mark, and I think you may have

answered this in the last couple of minutes, but I

was looking at the relationships that Atik had

presented looking at the probability of neutropenia

and diarrhea respectively as a function of SN38

area under the curve.

They're remarkably similar, although they

were based upon mean data. So the question is is

there an indirect benefit in reducing the risk of

severe diarrhea from paying more attention to,

let's say, the neutropenic problem? In other

words, do they go together, and to what degree do

they go together?

DR. RATAIN: Yes, there's certainly an

association of the two, but they don't always go

together. But a patient with neutropenia is more

likely to have diarrhea and vice versa.

DR. VENITZ: Good. Thank you. We

appreciate your comments.

Larry, why don't you frame your questions

for us?

DR. LESKO: How about if we bring them up

160

on a slide?

DR. VENITZ: That's fine.

DR. LESKO: I'll just scroll through

these.

Okay; so, as the time proceeds, I'll

scroll through the individual questions, but the

first question that we have for the committee

discussion is regarding the scientific and clinical

evidence that we're all aware of at this point. So

the question is is the evidence presented

sufficient to demonstrate that the homozygous

star-28 genotypes or 7/7s, as we call them, are at

significantly greater risk for developing a,

neutropenia, and b, the acute and delayed diarrhea

that we've heard about as an adverse event?

DR. VENITZ: And you would like for us to

vote on this?

DR. LESKO: Yes.

DR. VENITZ: So as far as the Committee is

concerned, any comments, discussion items for FDA

before we vote? And by the way, the vote is going

to be by voice vote. I'm going to call your

161

individual names, and you're going to have to tell

me whether you're a yes, no or abstain for the

individual questions as we go along.

DR. SINGPURWALLA: Jurgen?

DR. VENITZ: Nozer.

DR. SINGPURWALLA: Would you consider

removing the word significantly?

DR. LESKO: What would you suggest as an

alternative? Just--are you thinking of it because

it's a statistical--how would you convey a small

risk versus a large risk?

DR. SINGPURWALLA: Well, I'm not sure if I

could subscribe to the view that the risk is

significantly larger.

DR. LESKO: Could we use markedly greater?

Clinically important?

DR. SINGPURWALLA: I made my point.

DR. LESKO: I think the question is

intended to convey a magnitude of risk. If we want

to say clinically important, markedly, I think it's

fine. I think it conveys the same thing.

DR. SINGPURWALLA: You mean you insist on

162

an adjective?

DR. LESKO: I think a qualifier would

help.

DR. VENITZ: Any further comments?

DR. SADEE: Are we to consider these two

together, A plus B, or A separate from B?

DR. LESKO: I think we need to, based on

the way the evidence was presented today, it's

probably better to consider them separately.

DR. VENITZ: So we'll have two votes.

We'll have one on neutropenia and one on diarrhea.

DR. LESKO: Yes.

DR. VENITZ: Any other comments before I

call for the vote?

[No response.]

DR. VENITZ: Okay; then, the first

question is is there sufficient evidence of a

greater risk of developing neutropenia. And as I

said, you have three choices: yes, no or abstain.

So let me go down my list.

Dr. Barrett?

DR. BARRETT: Yes.

163

DR. VENITZ: Dr. Capparelli?

DR. CAPPARELLI: Yes.

DR. VENITZ: Dr. D'Argenio?

DR. D'ARGENIO: Yes.

DR. VENITZ: Dr. Davidian?

DR. DAVIDIAN: Yes.

DR. VENITZ: Dr. Derendorf?

DR. DERENDORF: Yes.

DR. VENITZ: Dr. Giacomini?

DR. GIACOMINI: Yes.

DR. VENITZ: That doesn't count.

DR. VENITZ: Dr. Hall?

DR. HALL: Yes.

DR. VENITZ: Dr. McLeod?

DR. MCLEOD: Yes.

DR. VENITZ: Dr. Sadee?

DR. SADEE: Yes.

DR. VENITZ: Dr. Singpurwalla?

DR. SINGPURWALLA: Yes.

DR. VENITZ: And Dr. Watkins.

DR. WATKINS: Yes.

DR. VENITZ: Okay; then, the second part

164

of that question is is there sufficient evidence to

substantiate a significantly greater risk for the

delayed diarrhea and acute delayed diarrhea? Oh,

before--I'm a yes, too. So we have unanimous.

Okay; second question, then, what about

diarrhea? Does the Committee feel there is

evidence to support significantly the increased

risk?

Dr. Barrett?

DR. BARRETT: No.

DR. VENITZ: Dr. Capparelli?

DR. CAPPARELLI: No, not yet.

DR. VENITZ: Dr. D'Argenio?

DR. D'ARGENIO: No.

DR. VENITZ: Dr. Davidian?

DR. DAVIDIAN: No.

DR. VENITZ: Dr. Derendorf?

DR. DERENDORF: No.

DR. VENITZ: Dr. Giacomini?

DR. GIACOMINI: No.

DR. VENITZ: Dr. Hall?

DR. HALL: No.

165

DR. VENITZ: Dr. McLeod?

DR. MCLEOD: No at the moment, but the

data looks like there's something there.

DR. VENITZ: That counts as a no.

Dr. Sadee?

DR. SADEE: No.

DR. VENITZ: Dr. Singpurwalla?

DR. SINGPURWALLA: I abstain.

DR. VENITZ: Dr. Watkins?

DR. WATKINS: No.

DR. VENITZ: And I would add my no, but it

does appear not only that there might be something

but it may be limited to patients that have

colorectal cancer; in other words, diarrhea may not

be present in patient populations that don't have

it.

Okay; any other comments about question

number one? So we have a unanimous vote on the

first part, and we have an almost unanimous part on

the second part of that question.

Okay; Larry, you want to present us with

the second part of this question?

166

DR. LESKO: Yes, I think on this second

question, if I can propose that the way this is

worded, it's not in a sense a votable question,

because I think we're looking for discussion; for

example, what would be the risks and benefit, what

is an appropriate study design, and it strikes me

that it doesn't lend itself to a vote. So if I can

propose that we look at this question and address

the questions that are posed on the slide in a

discussion context as opposed to a voting context,

I think that would be useful to us.

DR. VENITZ: Okay; then, I open the

discussion.

DR. GIACOMINI: Yes, one of the things I

didn't see, and I don't even know who I'm

addressing this to, this question to, but did we

ever see any data and, like, Kaplan-Meyer curves?

I don't even know if--Kaplan-Meyer curves, where

they've factored, you know, where they've looked at

survival data over time and then genotypes, put

them in categories, like the people with the star-7

genotype or the star-28 genotype are they having a

167

better survival or worse survival?

Can I ask Mark, or is he not allowed to

talk? Can I ask this to Mark? He can't talk?

DR. VENITZ: Yes, you may.

DR. GIACOMINI: Mark, I mean, just to get

an idea of the benefit to--

DR. RATAIN: There are no published data.

Dr. McLeod has a data set that may provide some

insight into the answer to your question.

DR. GIACOMINI: Oh; Howard? I guess what

I'm trying to do is get a feel--

DR. MCLEOD: Yes, there are no published

data on colorectal. There are two studies included

in the papers from--provided by Pfizer that looked

at the UGT1A1*28 genotype and survival. They were

both in the context of non-small lung cancer, if I

recall correctly. And one of the studies found

that the 7/7 genotype group from the star-28

homozygotes had a poorer survival. The other study

didn't separate the groups quite the same way, but

the group that contained the 7/7 genotypes had an

improved survival.

168

Now, neither of them met statistical

significance. They were all 0.06, 0.07 type

things; small studies, no covariance; I mean, a lot

of different issues. So my interpretation is we

really don't know the effect of UGT1A1 on survival.

In the context of colorectal cancer, there

is sufficient data, in my mind, to show that any

one study is not really going to have the full

answer on patient survival. So if you look at

patient survival for the impact of first-line

therapy is confounded by the presence of good

second-line therapy, good third-line therapy, et

cetera. And the Chornagon study demonstrated that.

Didn't matter what you gave first from what Mark

showed. If you gave the other one second, then, it

was a wash in the end.

And so, in the context of response, there

is some data that UGT1A1 may have an influence on

response, although the numbers were small and not

definitive. But there was no impact on time to

progression or survival, and so, it's inferior

data. It's as good as we have at the moment.

169

There are large studies in the cooperative groups

that are going to be able to address this in a much

more aggressive fashion because of sample size.

So I don't really know the answer. It

appears there may be an influence on response, but

there certainly does not appear to be an influence

on time to disease progression, so time until the

tumor grows again, or survival influence.

DR. VENITZ: Jeff?

DR. BARRETT: In thinking about dosing

this agent, I'm struck with Dr. Pazdur's original

comments when he talked about the fragility of the

original dose selection of this compound and the

modest response rate. So while I think most of the

discussion is focused on managing toxicities, the

loss of efficacy looms very high with this

compound. But I guess the other curious thing I

had in my mind is do we have any of the historical

data in which dose reductions were, in fact,

monitored where you could look at the

responsiveness of these markers or responses as far

as diarrhea and neutropenia go relative to a dose

170

reduction, so you can get some sense of, you know,

how, in fact, responsive those toxicities are to

dose reduction?

DR. MCLEOD: So that would be dose

reduction regardless of the cause?

DR. VENITZ: Atik?

DR. RAHMAN: I'd just like to comment on

something that we have in the package insert

already. We have data on 100 milligram per meter

square weekly dose, 125 milligram per meter square

weekly dose, and 150 milligrams per meter square

weekly dose, and what we have seen is that there is

not a whole lot of differences in the response

rates, although numbers are very small, so you

cannot do a cross-study comparison here.

But the observation that we have from the

package, and also, it is in the package insert is

that the survival, median survival across those

dosage groups is not a whole lot different so is

not the response rates.

DR. PAZDUR: The additional point is that

we do not know the relationship between response

171

rate as a surrogate for survival in this situation.

I would like to point out that this drug had in the

5-FU refractory disease population a 15 percent

response rate, yet it was able to show an overall

survival advantage compared to best supportive care

in two trials, which indicates to me that perhaps

disease stabilization or some influence on time to

progression is far more important than simply tumor

reduction size.

DR. VENITZ: Dr. Williams?

DR. WILLIAMS: I think one of the most

important questions to answer is what you're going

to base your dose selection on for the 7/7s. Mark

suggested, you know, not treating them as one

option, but obviously, you are going to have to

treat them. You cannot base it on a survival

observation. You just don't have enough patients

to make that observation.

So you're going to have to decide what to

select a dose on, and your new study, perhaps,

that's going to be done to look at dosing in that

population. So what are you going to base it on?

172

I was sort of interested with Mark, would you base

it on a targeted dose of AUC of SN38? That's a

little unsettling, because the slide that Atik

showed suggested that even with the same AUC, these

patients had a higher degree of myelosuppression.

You know, they were having grade four neutropenia

all along the bottom of that graph up and down the

AUC spectrum, so that's a little mysterious, and

then, you ask, well how about from the

pharmacodynamic standpoint for the tumor? Is the

tumor equally sensitive to the same AUC of SN38?

So, I mean, I think you're going to have

to target something. You can't, you know, you just

can't look retrospectively at toxicity, and I'd be

interested in what the Committee thinks when you do

this new study to try to individualize dosing for

these patients, what are you going to target?

DR. RAHMAN: I'd like to make a comment

about the starting dose. What I have shown in my

presentation that there is already a nice algorithm

for a starting dose for standard therapy and

continuous therapy and dose modification based on

173

toxicities in the package insert. And as I have

mentioned that we have some predictive factors

already in the package insert which are bilirubin

levels, prior radiation therapy, performance

status. Those are already indicating, recommending

a dose level lower than the standard dose as a

starting dose.

And then, if the patients do not have any

complication with that, the package insert allows

to go up to the standard dose and then move on with

that. So here is the starting dose that we can be

thinking about that can we do anything different

for the UGT1A1 patient, 7/7 patient, I mean?

DR. MCLEOD: And to follow up on that, I

wonder if maybe Dr. Morrison or one of the Pfizer

team could comment on whether dose reduction is a

covariant in terms of outcome, time to progression,

whatever your favorite is, coming back to try to

get at Jeff's initial question.

MR. MORRISON: Maybe if I could defer to

Pat to comment on that, because this was actually

before my time.

174

DR. MCLEOD: Lucky Pat.

DR. MCGOVREN: Yes, I don't have an

answer. It has not been modeled. So I don't

know--I think that the various risk factors were

arrived at very empirically, and dose reduction was

not done in any systematic way.

DR. MCLEOD: Well, Atik clearly and

correctly mentions that there is a range of doses

that seem to be equal. Those patients may have

declared themselves as being different not only in

their sensitivity to the drug but also for other

factors, and so we can't say that just because we

can start low, that means that people who are

sensitive will still do well.

DR. PAZDUR: But there is an inherent bias

in looking at the data that patients that may get

the dose reduction are poor performance status or

other issues that lend themselves to poor either

responses or poor survival outcomes. I think it's

clear, though, you know, having worked with this

drug before I came to the agency and have had a

long history, I think it would be fair to say that

175

we do not have a really good handle on what is the

dose in its relationship to the eventual outcome.

You know, could we have achieved a similar

outcome with a reduced dose? Remember, this drug

was developed in a time when oncology had the

mantra more is better, more is better, more is

better, and we kind of were hitting toward what is

the absolute highest dose that we could deliver,

and this is common in many of the oncology drugs

that we have developed over the past decade, and

now, we're trying to step back and ask this

question, which is very difficult to do.

You know, should we look at, for example,

at these patients that have this genetic mutation

to do just simply a phase one study, as we

suggested, sometimes through the company to take a

look at what would be the appropriate dose,

starting out at an artificially dose reduction and

seeing actually what the dose, because we really

don't have a good handle, even in the general

population, of what is a dose response for this

drug. And we're basing it on toxicity, basically,

176

and that's--we have to be realistic on the

development of this drug. That's how it happened

over the past decade.

DR. VENITZ: Let me make a--

DR. MCGOVREN: Yes, go ahead.

DR. VENITZ: --followup comment that gets

to item number B. I think right now, the concern

is that if you reduce the dose, we might compromise

efficacy. Well, but can you not turn that argument

around? If you improve tolerability and compliance

on a long-term treatment, don't delay treatment as

a result of a lower dose, you might actually

improve efficacy, not just compromise it. So to

me, I don't know which way to go. As you pointed

out, this drug was developed under the paradigm of

an MTD.

So by actually backing off of the dose,

you might get improved efficacy just by keeping

more patients on drug.

DR. MCGOVREN: Yes, none of the trials, I

don't think, were large enough to actually dissect

out the efficacy in patients who started at the

177

standard dose and continued on the standard dose

until their tumor progressed versus patients who

started with the standard dose, were dose-reduced

because of toxicity and then continued on a reduced

dose versus those who started on a reduced dose

because they had a risk factor at the time they

went on treatment, performance status or whatever,

and then continued on that reduced dose or even had

that dose reduced because they couldn't take the -1

level dose.

So it's just very difficult to tease out

of the available trials all of these factors which,

of course, complicates how do you design to

determine the appropriate dose for the 7/7s?

DR. VENITZ: Then maybe let me focus the

Committee on the third part of this question: what

would be needed, what would need to be done in

order to figure out what to do with those patients

in terms of coming up with a starting dose for

patients that are 7/7 genotypes?

DR. SINGPURWALLA: Jurgen? As an outsider

looking at this, the question is what is an

178

appropriate study.

Now, I can't answer that question as to

what is an appropriate study, but one thought goes

through my mind: electrical engineers use control

theory to control the movement of something or to

control the behavior of something. Has any thought

been given to using a similar kind of a paradigm in

this particular business? You start with a certain

dose; you make a prediction as to what the effect

of the dose will be; then, you observe the actual

outcome and make a correction subsequent to that

and keep on doing it in some kind of a filtering

scheme.

That is a suggestion that I would like to

put forward.

DR. MCLEOD: There are study designs that

have used a variation on that theme, both in terms

of trying to reduce the number of patients required

to study in early evaluation and also try to make

them more rapid. They've had variable success, and

in the end, we've kind of fallen back to the status

quo. But people are certainly aware of sort of

179

iterative-type processes. We just haven't figured

out how to do them very well.

DR. SINGPURWALLA: Well, I'm surprised

that you've said you've fallen back, because

control theory is one of the most successful

applications of process control, which is really a

part of this, and I'm surprised why the study

failed or why they regressed.

DR. MCLEOD: Well, it's a very successful

theory in many industries. Biomedicine is not one

of the areas where it has been a blazing success.

And so, I think with the greater understanding of

systems biology, it will be successful.

Currently, the endpoints that we talk

about in phase one are incredibly crude, and crude

endpoints don't lend themselves to intricate

approaches such as what you described.

DR. SINGPURWALLA: Are these studies

published? Is there any way I can read up on them?

DR. MCLEOD: Certainly.

DR. WILLIAMS: Let me sort of restate my

question earlier. I think from a practical

180

standpoint, certainly, you could do a phase one

study in these patients, right? And I think the

question would be that's probably what you should

do, right? Do a phase one study in the 7/7s; you

find a reasonable toxicity.

But then, what are you going to use to

provide yourself with the assurance you're in the

right place? Will it be that you have the AUC that

everybody else had with SN38? Would that provide

you assurance, or would there be some other route?

So I guess I'm just throwing out the possibility

that you would do a normal type phase one study

looking at toxicity and pharmacokinetics.

But then, what would you do, you know, to

assure yourself that you're where you want to be?

DR. VENITZ: Marie?

DR. DAVIDIAN: I just wanted to bring up,

related to that, there's been some recent work in

the statistical literature by Peter Fall, who is at

M.D. Anderson, and I was just wondering if any of

his work would be relevant in looking at toxicity

and efficacy jointly?

181

DR. WILLIAMS: I mean, I wonder, you know,

in this setting, do you think that--I guess the

only thing you could look at would be response rate

in that kind of setting. I don't know--we have to

talk to the statisticians, whether you could really

study enough patients to do that.

DR. DAVIDIAN: I agree.

DR. VENITZ: Just a followup to your, I

guess, subquestion here. I think that's the only

way you can approach it with what we know right

now. Doing a phase one dose escalation study and

see what area do you accomplish? What's the

corresponding dose?

DR. PAZDUR: But we really don't have a

good pharmacodynamic relationship between any

parameter and efficacy, either if one tries to look

at response rate or any other clinical endpoint

with this drug, and we have to be realistic about

that. And I think also, it depends on what type of

dose reduction that we're talking about with these

7/7 patients. If we're talking about a 75 percent

dose reduction, that could be quite problematic.

182

If all we're talking about is a 25 percent

dose reduction, I would feel a little more

comfortable.

DR. WILLIAMS: Let me push back to my boss

here. But what if it was a 25 percent dose

reduction, and you had the same AUC? Would that

help you? That's what I'm sort of trying to get

at.

DR. PAZDUR: Yes, but we don't know as far

as that would involve some comparison here, but we

don't know, basically, that AUC correlates with

response rate or doesn't.

DR. MCLEOD: And Atik's data says that AUC

is not AUC in terms of risk of neutropenia.

DR. WILLIAMS: For these patients, anyway,

right? There somehow seems to be a little bit of a

difference in these particular patients'

sensitivity to AUC than the other patients. That's

what I took away from this graph.

DR. PAZDUR: But, Grant, I think that

would give you a degree of some comfort here to

have some parameter that you're achieving.

183

DR. VENITZ: Final words, Larry?

DR. LESKO: If we're thinking of dosing

adjustments, whether it's a drug in oncology or

not, we generally try to bring a quantitative

analysis to the probability of an adverse event

based on exposure, whether it's a renally-impaired

patient or whether it's a drug interaction, and

depending on the outcome of that analyses, we would

reduce the dose to achieve similar area under

curve. This is pretty much how labels are derived

in terms of dosing adjustments for specific

populations.

So that concept isn't all that foreign.

It's actually the first principles of the way that

drugs work. So I think in any study that would be

conducted, the area under curve of the species that

has been shown to correlate with neutropenia to

date, which is the SN38 area under curve, would be

extremely important to measure and then use as a

guide along with other measures to determine what

the appropriate dose would be.;

I also think you don't necessarily need a

184

prospective study. There are many studies

discussed today, and not knowing the details of all

of them, one could imagine that a study would be

conducted in which efficacy or safety would guide

the treatment and then having some genotype

information in a retrospective fashion to associate

the dose that was given and the appropriate outcome

that was previously decided upon would be an

extremely powerful correlation to have, coupled

with area under curve to figure out what the right

dose reduction ought to be.

MR. BELLO: Hello. My name is Akintunde

Bello. I work for Pfizer clinical sciences. We

just thought it's interesting and important to

point out that there is actually a study that's

going to be starting shortly that's actually

looking at different doses and will be looking at

genotyping as well as exposure, PK exposure for

various moieties related to CPT11. So this is work

that's ongoing. There's a study that's forthcoming

and may give us the answers that we're looking for.

DR. VENITZ: Howard?

185

DR. MCLEOD: In the context of the

cooperative groups, this issue has also come up,

not just so much from a regulatory standpoint but

from a clinical trial standpoint, and I'm wondering

whether Dr. Ratain would be able at this point to

comment on some of the discussion that's going on

in the context of these patient genotypes. It may

be too early for that, Mark, but if you want to

comment, here's an invitation.

DR. VENITZ: Are you willing, Mark?

DR. MCLEOD: And if I put you on the spot,

I'll buy you a beer.

DR. RATAIN: Thank you.

Yes, the CLGB has a study in

development--CLGB is Cancer and Leukemia Group

B--in which patients will be genotyped; patients

with 6/6 genotype will be enrolled in a trial to

establish the safety or potential safety of higher

doses, as high as 500 milligrams per meter squared,

based on the evidence from this European trial that

some patients can tolerate 500 milligrams per meter

squared, and the hypothesis that these are 6/6

186

patients.

And if, indeed, we can establish that, the

next step would be a prospective randomized trial

in the 6/6 genotype patients of 500 versus the

standard 350.

I also, since I'm up here, I thought I

would follow up on some of the previous comments

about the pharmacodynamics of the activity. I

don't think we know what correlates with activity.

I think there is a fair amount of evidence that

it's not the SN38 AUC, and in fact, in a study that

we've completed and is in press in Clinical

Pharmacology and Therapeutics, we have modulated

the pharmacokinetics of Irinotecan with cyclosporin

A as an inhibitor of transport and also inhibits

CYP3A as well as phenobarbital is an inducer of

glucuronidation, and activity is preserved in some

patients despite very, very low SN38 AUCs.

So I would not recommend that one titrates

dosing to--from an efficacy perspective to

particular SN38 AUC. I think that is useful to

guide neutropenia considerations but not from the

187

standpoint of efficacy.

DR. VENITZ: Okay; Larry, you want to move

along to the next--

DR. LESKO: Yes, I'll just say the next

two questions are obviously up there, and the first

is not, I don't believe, a voting question, but

nevertheless, it would be useful, again, to have

some discussion of a context. Some of this has

been covered already, I think, in the prior

discussions, but if there's any other remaining

comments on the question number three in terms of

how a genotype could be complementary to

preexisting information on risk and how it might be

integrated into predelivery of the drug or

simultaneous delivery of the drug would be

beneficial.

Question four refers to some of the terms

we had about performance and probability content

information of a test, and one of the things that

hasn't been discussed is the relative value of the

expressions of a performance of a genomic test to

clinicians in terms of understanding. We've heard

188

sensitivity, specificity, predictive value, odds

ratio. There is one other, and that is the

likelihood ratio, all of which are used in the

literature to different degrees for these tests as

screening tests, basically, and any discussion or

comment people have on the relative value of these

different tests in conveying the probabilistic

nature of these genomic tests would be useful.

DR. VENITZ: Okay; then, I start with any

comments to question number three. How would you

incorporate PG information?

DR. WATKINS: Well, the point I made

before is that this is, to my knowledge, a unique

situation where you've got a

xenobiotic--polymorphic, xenobiotic metabolizing

enzyme that has an endogeneous substrate. So one

of the biggest concerns in using host genotypic

information to predict dose, particularly to

escalate dose in the 6/6 individuals is that in

fact, there might be a nongenetic factor or

additional polymorphisms that would make that

person susceptible.

189

But in this case, you have an endogeneous

marker. You've got indirect bilirubin, which is a

safety valve. So if you're missing environmental

reasons or other snips, the bilirubin should go up,

with the caveat that in total liver failure, serum

bilirubin only rises about a milligram and a half

per deciliter per day, so this would not be a

sensitive measure of acute changes.

So, apropos question number three, I think

the main unique situation here is there is an

endogenous, built-in marker for the rare

individual, or it doesn't matter how rare, the

individual that would be 6/6 genotype but in fact

would have low activity.

DR. VENITZ: Any other comments to

question number three?

DR. MCLEOD: Just to follow up on that,

Paul, I mean, it should be a surrogate marker,

biomarker, bilirubin, but, I mean, from some of the

data that was presented and some of the discussion,

it doesn't seem to be a good biomarker. I mean,

genotypes seem to offer something beyond the

190

current approaches. And I know there are

approaches out there where you give a single dose

of rifampin and then six hours later take a

bilirubin level, look at induction or induction but

the increase in glucuronidation. And there's other

tests like that.

But in terms of something that could be

used in clinical practice, baseline bilirubin in

the context of multiple centers, from what Mark

described in his JCO paper, wasn't a good marker.

But yet, genotype wouldn't be influenced by those

things. So genotype wouldn't be the answer, but it

seems like an achievable answer.

DR. WATKINS: No, and I think the overall

data is that bilirubin is not as good a marker as

genotyping in this case. That wasn't the point I

was making this time, which was the comforting

thing is there's a built-in marker for someone

who's very deficient in UGT1A1 but genotypes as

having normal activity. So there's a built-in

safety valve, which is really unique, to my

knowledge, to this situation, which is very

191

reassuring and, I think, makes it easier to go

ahead and push genotyping, knowing that there's a

safety factor involved.

That wasn't implying bilirubin is better;

it's just a safety valve here.

DR. VENITZ: Steve?

DR. HALL: I think one of the features of

the UGTs is they don't have a high degree of

specificity, you know. So maybe the 1A1 is a major

determinant of bilirubin conjugation, but many

others contribute a small part, and in the absence

of one, they kind of all contribute something to

the remaining activity.

So I don't think it would be surprising

that the bilirubin wouldn't work as a good index of

the enzyme, and I think the 1A family of the UGTs

is complex. They're all this single locus. They

have highly related polymorphisms that probably all

contribute in some part to the overall bilirubin

thing. So I don't think it's likely to be the

surrogate for that single enzyme defect.

DR. BARRETT: I think if the question is

192

how to use this information relative to the other

factors, I mean, you have a clear idea with the

bilirubin and these other factors in conjunction

with genotype as far as the directionality goes

with Irinotecan, so as far as using it, I mean, I

think there is a practical guidance that could come

out of this, independent of the fact that it's not

a perfect correlate.

So, you know, where you are today in terms

of your understanding of this polymorphism, there

is a directionality there. Whether or not people

use it is another thing. I mean, I think the

comment from Dr. Raitan was very interesting. You

know, for the most part, there is a default to

what's in the label as far as dosing guidance, but

there's still a lot of empiricism. So the extent

to which you can provide educated information to

that empiricism, you should do it.

DR. VENITZ: Okay; then, let's move on

to--Wolfgang?

DR. SADEE: Since you bring that topic up,

I have to agree. Clearly, we have, for the star-28

193

allele, we have good information that it does make

a difference. But what is missing is the

information on what is the variability within these

genotypes? And I don't think it's all that

difficult to get. In fact, while we're listening

to it, motivated to maybe look into this and maybe

provide definitive numbers as to in a population of

500 people, when you look at 1,000 alleles, how

often do you see that one is less than the other?

And what's the variability within this one

genotype?

I see this again and again with

pharmacogenetics that translate into clinical

trials, where there's a single genotype that's been

isolated; for instance, the LPR for the serotonin

transporter. And every single clinical study is

using this, and there isn't even evidence that it

does make a difference in where the gene is really

expressed.

So I think in this case, there's very good

evidence that we have a clear difference. We still

haven't defined here how much does this difference

194

really cover of what's actually happening in the

body? And so, I would like to really see that we

take the first step, and we have all agreed that

there is a correlation already with neutropenia,

but the second step must be--we must have

quantitative information: how often is this

predictive? How soft is this information? And I

think this needs to be clarified.

DR. VENITZ: Okay; then, last question,

Larry, do you want us to vote on this, question

number four?

DR. LESKO: Yes.

DR. VENITZ: Okay; so, first, let's have a

discussion on do we believe as a Committee that

current test has sufficient sensitivity and

specificity to be used. And I'm assuming response

predictor means toxic response predictor.

DR. LESKO: Yes, it does, and if there's

any discussion of the question or other measures we

haven't brought up to the Committee, that would be

appreciated as well.

DR. VENITZ: Let me start making my

195

comment first. Obviously, the numbers that we've

seen, that both Dr. Raitan and the Pfizer group has

presented, the positive predictive value looks

pretty low: 50 percent. The negative predictive

value, 83 to 90 percent, very high. So now, we

have to use my favorite concept of the utility

concept. In other words, we have to use judgment,

not just statistics.

So is a 50 percent positive predictive

value, is that something that we deem clinically

important? In other words, is that enough

assurance for a patient and/or a health care

provider to start treatment or vice versa? Is a

negative predictive value of saying I'm 90 percent

certain that with a 6/6, you're not going to

develop neutropenia, is that comforting enough?

And in my assessment, it is, based on clinical

judgment, not based on the statistics empirically

per se.

DR. PAZDUR: Could I just ask a question

before we go on? And maybe this is to Larry or

Atik: the meaning of this question, are you trying

196

to implicate in this question that all patients,

before they go on Irinotecan, should have their

status known? Is that what we're after here?

DR. LESKO: That's a different question.

I don't think that's the question we were intending

to ask. It was more directed towards if, as a

physician, I'm going to treat a patient, and I'm

going to use this test, I have to have some

information about what the test conveys in terms of

probabilities.

I think we heard about the sensitivity and

specificity, and then, we moved to predictive

values and odds ratios, so it's, for example, the

question on the likelihood ratio would be if I

tested positive, I would have a ninefold greater

chance of becoming neutropenic. That's what the

likelihood ratio would say for this. Now, what

would that mean to the clinician in terms of, a,

using the drug; monitoring the patient; using a

lower dose; making other decisions, coupled with

the knowledge of the bilirubin or other preexisting

risk factors?

197

And I think it's important not to take the

test in isolation in making these decisions but

coupled with and complementary to the other

information that would normally be at the disposal

of the patient and the physician to make a

decision. So I don't know if that answered the

question. I don't think it's asking is there a

need to prerequisite do the test before deciding to

give the drug, but it certainly would seem to be

useful, very useful a priori information.

DR. WILLIAMS: One of the points we had

discussed internally was to look at the current

label and some of the information that suggested

you might want to dose-reduce based on these

things, such as age, et cetera. And realizing

that--thinking about, I wonder what the basis of

that was?

So some things, we may have put in the

label. You might wonder about how strong the

evidence was there. I don't know if that relates

to this. It seems to a little bit.

DR. VENITZ: Howard?

198

DR. MCLEOD: My followup question, the

reason I asked Mark that question is is there

performance data for these other factors, for

bilirubin, prior pelvic radiation, performance--

DR. PAZDUR: It's very poor.

DR. MCLEOD: Okay.

DR. PAZDUR: And a lot of this has to do

with how the clinical trials were done that led to

the registration of the trials, because what we put

in the labeling usually reflects the patient

eligibility of the clinical trial that was done.

For example, age was put in the label because the

European trial restricted entry based on age.

Whether or not that would occur now, I don't know,

and we've heard from Pat that that probably doesn't

make a lot of sense, and we need to revisit this.

So the data on this are probably not as

robust as what we're seeing here, to be honest with

you.

DR. BARRETT: You're going to appreciate

you framing the question, because if I had to

answer number four the way it's written, I would

199

say no. However, if you said to me would I vote

for a test in which the negative predictive value

was greater than 90 percent as far as an aid to

dosing, I would say yes.

DR. VENITZ: Larry?

DR. LESKO: Yes, another way we tried to

think about this question is really the question

that we're trying to ask: are we trying to rule in

a risk or rule out a risk? And I think that really

reflects on the usefulness of the predictive

values. If we're trying to rule out a potential

risk with a high specificity, that would seem

valuable to know that in terms of making judgments

about the therapy with the drug as opposed to

trying to rule in someone with toxicity. It gives

an indication, but it's a little bit softer because

of those predicted values.

So I think there's a context for these

tests that have to be what is the question we're

asking?

DR. VENITZ: Well the positive predictive

value in my mind is so low because you have low

200

prevalence. It is only an average 10 to 20 percent

of neutropenia. So you have to have a very

specific, I mean, very, very sensitive test to have

a high positive predictive value.

DR. LESKO: Well, the other question is

you're exactly right: the predictive value is a

function of the prevalence, and we know that's

relatively low. Another way to think about the

question is how does it perform in the context of

other tests that are used routinely in

therapeutics, in particular in oncology, where some

of the predictive values are down around 10

percent?

Another way to ask the question is what is

the predictive value in sensitivity and specificity

if I want to detect a variant allele, namely, a 6/7

or a 7/7 patient? Now, you have a prevalence of

about 50 percent, and then, you begin to look at

predictive values; they're probably moving up on

the positive side to 85 or 90 percent at that

point, with a 50 percent prevalence.

DR. VENITZ: When you've seen Mark present

201

that it takes 20 patients tested in order to avoid

one bout of toxicity relative to TMPT, where it

takes 100 patients that need to be screened.

DR. LESKO: It was 300 to find one in the

TMPT, so this is fairly efficient.

DR. VENITZ: Any additional comments in

terms of question four before we vote?

Nozer?

DR. SINGPURWALLA: I was not sure whether

you wanted some kind of a reaction to odds ratio

versus likelihood ratio. Is that correct?

DR. LESKO: Yes, I think that would be

useful, because both are used in the field of

testing, of screening tests, and we'd like to hear

what the Committee thinks or what you think about

that.

DR. SINGPURWALLA: Well, perhaps I'm

wrong, and maybe Marie can correct me, but I

thought that the likelihood ratio is, in fact, the

odds ratio when the model is a binomial model. So

I think they are the same thing. And I was

wondering why you wanted a comment on the

202

distinction between the two.

DR. LESKO: Because I'm not a

statistician.

[Laughter.]

DR. SINGPURWALLA: Okay; so, here is my

contribution to this meeting: they are the same

thing.

[Laughter.]

DR. VENITZ: Any other comments,

statistical or not?

DR. WILLIAMS: Another comparison that

might be of interest is the design of a phase one

oncology study. We usually consider, you know, you

have one toxicity in three patients, and then,

maybe you should look at a little more; or you have

less than that, it's okay; you have more than that,

it's not.

Well, here, I think 50 percent is--of

grade four toxicity, it is kind of interesting.

That really is above what we would say is the MTD,

and so, to that extent, you might consider it's

relevant. You're getting a patient population here

203

saying is above the MTD. So from that extent, you

might consider it relevant.

DR. VENITZ: Okay; any other comments

before we call for the vote?

[No response.]

DR. VENITZ: Okay; so, we are voting on

question number four. The only friendly amendment,

predictor means toxic response predictor, right?

Because we're not talking about efficacy.

Okay; so, you have three choices: yes,

no, or abstain, and I'm going to call your name.

Dr. Barrett.

DR. BARRETT: Yes.

DR. VENITZ: That's a yes but, right?

DR. BARRETT: Yes but.

DR. VENITZ: Okay; Dr. Capparelli?

DR. CAPPARELLI: Yes.

DR. VENITZ: Dr. D'Argenio?

DR. D'ARGENIO: Yes.

DR. VENITZ: Dr. Davidian?

DR. DAVIDIAN: Abstain.

DR. VENITZ: Abstain?

204

Dr. Derendorf?

DR. DERENDORF: Yes.

DR. VENITZ: Dr. Giacomini?

DR. GIACOMINI: Yes.

DR. VENITZ: Dr. Hall?

DR. HALL: Yes.

DR. VENITZ: Dr. McLeod?

DR. MCLEOD: Yes.

DR. VENITZ: Dr. Sadee?

DR. SADEE: Yes.

DR. VENITZ: Dr. Singpurwalla?

DR. SINGPURWALLA: I'm afraid I have to

abstain.

DR. VENITZ: The statisticians abstain.

DR. SINGPURWALLA: Well, I'll make a

comment that if somebody starts with a yes, there

is a high probability that the yeses will--

[Laughter.]

DR. VENITZ: Okay; Dr. Watkins?

DR. WATKINS: I'm going to abstain. I

mean, we've all agreed that the test predicts

neutropenia, but this is somewhere between that

205

answer and do we think all oncologists should be

doing it, and I'm just not sure where the question

really is in that spectrum, so I'm abstaining.

DR. VENITZ: Okay; so I'm going to vote

yes, so we have three abstentions and nine yes, for

a total of 12.

And I think that does conclude our morning

session, so I appreciate you all's contribution.

We'll take a break until 1:00 for the open

public hearing, and the Committee members have a

room for lunch reserved in the restaurant right

here in the hotel, Martindale's.

[Whereupon, at 12:12 p.m., the meeting

recessed for lunch, to reconvene at 1:00 p.m.]

- - -

206

A F T E R N O O N S E S S I O N

[1:08 p.m.]

DR. VENITZ: Our next agenda item is the

open hearing, and we do have one letter submitted

by Dr. Rowling, a member of the Committee who was,

unfortunately, not able to attend. She submitted a

letter for your information that is attached to

your packages and will be posted on the Website.

Other than that, we have nobody here for

public hearing.

Then, our next order of business is

conflict of interest statement.

Hilda?

MS. SCHAREN: Hello.

The following announcement addresses the

issue of conflict of interest with respect to this

meeting and is made a 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

207

subcommittee in which a particular product is

discussed, issues of broader applicability involve

many industrial sponsors and academic institutions.

All special Government employees have been screened

for their financial interest 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 general matter 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 12A30 of the Parklawn

Building.

Because general topics impact so many

entities, it is not practical to recite all

potential conflicts of interest as they apply to

each member, consultant and guest speaker. FDA

208

acknowledges that there may be potential conflicts

of interest, but because of the general nature of

the discussions before the subcommittee, these

potential conflicts are mitigated.

With respect to FDA's invited industry

representative, we would like to disclose that Dr.

Paul Fachler and Dr. Gerald Migliaccio are

participating in this meeting as nonvoting industry

representatives acting on behalf of regulated

industry. Dr. Fachler's and Mr. Migliaccio's role

at this meeting is to represent industry interests

in general and not any one particular company.

Dr. Fachler is employed by Teva

Pharmaceuticals USA, and Mr. Migliaccio is employed

by Pfizer. 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's involvement

and their exclusion will be noted for the record.

With respect to all of the participants,

we ask in the interest of fairness that they

address any current or previous financial

209

involvement with any firm whose product they may

wish to comment upon.

Thank you.

DR. VENITZ: Thank you, Hilda.

The second topic of today's meeting is in

regards of drug-drug interaction and will be

introduced by Dr. Shiew-Mei Huang, who is the

deputy director for sciences of the Office of

Clinical Pharmacology and Biopharmaceutics.

DR. HUANG: Thank you, Jurgen.

Good afternoon. Before I talk about

relevant principles of drug interaction concept

paper that is published as part of the background

information for this Committee's discussion, I'd

like to briefly summarize some of the publication

and discussion that happened to lead to a revision

of this guidance.

Back in 1997 and 1999, we in CDER, with

CBER, published two guidance documents for

industry: the 1997 on in vitro drug interactions

and 1999 on in vivo drug interactions, focusing on

study design, data analysis, and recommendations

210

for labeling. Subsequent to the publication of

these two guidance documents, we had various public

workshops discussing different topics related to

drug interactions.

We also had a lot of internal discussions,

including CDER-wide scientific round discussions.

There is one example of publication on one of the

public workshop, and you have heard from Dr. Lesko.

We have various internal documents. Some of them

are published, such as the good review practices,

where we have included important drug interaction

questions to ask during the review of the

applications. And we also have drafted a MAP,

which is Manual for Policy Procedures about

cross-labeling and also about in vitro evaluation

of drug interactions.

PhRMA has published a white paper last

year on general drug interaction issues, and as Dr.

Lesko summarized earlier this morning that this

advisory group, the Advisory Committee for

Pharmaceutical Sciences and the Clinical

Pharmacology Subcommittee, at a meeting last year

211

in April, we discussed the proposal of classifying

CYP3A inhibitors, and we also touched upon PGP

inhibition-based interactions, and in November, we

talked about some of the emerging important

enzymes, such as CYP2B6 and 2CA and their role in

the evaluation of drug interactions.

So based on these discussions, the CDER

working group with the contribution from CBER, we

have drafted an interaction guidance, which is in

internal review right now. And this will be

published soon as a draft for comments, and when

the guidance is finalized, this will replace the

two in vitro and in vivo guidance documents

currently posted on the Internet, where we have

updated information and recommendations on drug

interaction evaluation.

We also have this guidance to address some

of the recent labeling rule change. In 2000, we

had published a proposed rule about professional

labeling of prescription drugs. The final rule

will be published soon, with accompanying various

guidance documents to talk about various segments

212

of the labeling.

So I'd like to talk about some of the

principles that we discussed and the drug

interaction concept paper which was released for

discussion purposes only.

In this concept paper, we stress the

importance that metabolism and drug interaction

information to benefit-risk assessment for new

molecular entities prior to market approval. We

have learned our lessons from recent U.S. market

withdrawal from 1998 to 2003. Note that this table

was constructed prior to the withdrawal of Vioxx,

so we did not include Vioxx in the table.

However, if you look at these 10 drugs

that were withdrawn between 1998 and 2001, where

they had been approved between 1985 and 1999, these

10 drugs with different characteristics and use;

there are some antihistamine or cholesterol

lowering. But if you look at these, the risks,

five of the 10 drugs, the risk of drug interaction

has contributed to withdrawal. And out of these

five drugs, if we look at Terfenadine, Astemizole,

213

Cisapride, Cerivastatin, these are substrates of

cytokine p450 enzymes or other enzymes or

transporters, while Mibefradil is an inhibitor of

CYP enzymes, PGP and possibly other transporters.

So these examples demonstrate that it is

important to evaluate other drugs' effect on the

new molecular entity and the new molecular entity's

effect on other drugs. We have a recent example

where a new molecular entity is a CYP3A inducer,

and the risk of drug interaction has contributed to

that drug's nonapproval. So again, we want to

stress it's important to evaluate inhibition-based

interaction as well as induction-based interaction.

Second principle I'd like to talk about in

the concept paper is to talk about an integrated

approach to evaluate drug interaction in vitro, in

vivo, specific and population pharmacokinetic

studies where when you look at the totality of data

to estimate the potential for drug interaction, and

this, hopefully, will reduce the number of

unnecessary studies and to optimize our knowledge.

In the concept paper, we discuss that for

214

the evaluation of metabolic interactions, as far as

evaluating the new molecular entity as an

inhibitor, we said it's important to study the five

major CYP enzymes: 1A2, 2C9, 2C19, 3A and 2D6. As

far as evaluating as an inducer, since the 2D6 has

not been shown to be inducible, here, we're

stressing the importance to study the other four

major CYP enzymes.

We know it's important to study other

drugs' effect on the new molecular entity, so it's

important to evaluate the metabolic profile of the

new molecular entity. We think it's important to

evaluate those five CYP enzymes, but when none of

these enzymes are found to be responsible for the

metabolism, it may be important to evaluate other

CYP enzymes such as CYP2B6, 2C8, rarely 2E1 or

other phase two metabolizing enzymes.

This morning, we have discussed how the

genetic variation would affect a drug with a

substrate for UGT1A1. Unless we know this drug is

metabolized by UGT, we probably won't know how the

genetic component would affect its metabolism and

215

its clinical response.

As far as inhibition, we have included an

appendix to talk about how to evaluate in vitro,

and we have indicated one parameter to look at

possibility of in vivo inhibition based on in vitro

data is to look at the I over KI, I as the

concentration of an inhibitor, which we like to use

a CMAX at a steady state at a highest dose and

compare to a KI of get a five major CYP enzymes.

The PhRMA paper indicated I over KI of 1

or 0.1. More than 1 is likely to be an inhibitor.

We did not specifically indicate what ratios,

although we did mention when the I and the KI were

separated by a very large--such as 50, then, it's

not likely to have an interaction. However, we

also indicated that we could rank order the in

vitro data to determine and prioritize the in vivo

studies.

For example, this is one new molecular

entity. And here, the five major CYPs are

evaluated. We like to look at the KI value.

Sometimes, we don't have the KI values, because

216

when you have very high concentration, you still

don't see inhibition. Sometimes, IC50 will be

expressed as higher than the concentration being

evaluated. So in this case, if you look at I and

KI, you would say, well, this is a very likely

event, and this one falls into probable, and this

may not be likely.

In order for us to--we don't have a

definite number to work with, I over KI ratios, so

the suggestion would be to look at the KI in rank

order, and you probably want to evaluate the CYP

that's most potently inhibited first in vivo. If

the results are negative, then, you wouldn't have

to evaluate the other that's less potently

inhibited, but if the results are positive, we

couldn't extrapolate, and we need to evaluate the

other CYP enzymes.

As far as induction, we have a new message

in the concept paper. We say that induction can be

addressed with in vitro methodology. In our

previous guidance documents, we mention that

induction can only be evaluated in vivo as a

217

technology for evaluation has not--there's

insufficient data to support a use of in vitro. So

we said we would look at the induction data based

on in vivo at this point.

I have mentioned earlier, it is important

to evaluate CYP1A2, 2C9, 2C19 and 3A. However, we

are suggesting that the initial in vitro evaluation

can be done with 1A2 and 3A. Part of the reason we

thought the 3A could be coinduced with 2C9 and

2C19, so if the results from CYP3A is negative,

then, you don't have to evaluate 2C9 and 2C19. Dr.

LeCluyse is going to show us some data to support

that argument later.

Again, we say negative results may

preclude in vivo evaluation of the other important

CYPs that we have mentioned that are important to

evaluate in a submission. Unlike the inhibition

study where we only say a positive control is

optional, for induction, we say a positive control

is recommended. For example, if you're evaluating

CYP3A, we think it's important--we could use

revamping as a positive control.

218

We think it's very important, since our

recommendation is if the data is negative, then,

you don't have to do in vivo. If it's positive,

then, you need to do an in vivo study. So it's

important how we define when it's positive. The

original concept paper, we said we can either use a

40 percent of positive control as a cutoff or

twofold of the negative control.

With subsequent discussion that when we

look at both 3A and 1A2, there may be too much

false positive if we use the twofold negative, so

we have dropped it right now, and we are

discussing, we are asking the Committee to comment

on the appropriateness of using a 40 percent of

positive control to suggest a possible induction,

and this 40 percent number was based on the PhRMA

white paper.

Ever since we have started to discuss the

appropriateness of using in vitro induction

methodology to evaluate induction, we have received

quite a few comments. Well, then, it's now a need

to conduct in vitro inductions for all new

219

molecular entities. And our answer is no.

However, it's important to address induction. You

can either use in vitro or in vivo. It's important

to address, but you don't necessarily have to use

in vitro, but it may be a good approach to start

with in vitro. And then, if the results are

negative, then, you're done, but if it's positive,

then, you continue. This is sort of what I just

said. Positive in vitro needs to be followed with

in vivo.

And I want to mention that induction can

be part of evaluation of in vivo inhibition

studies. Oftentimes, we have seen inhibition

studies carried out with Midazolam when we're

evaluating the possibility of inhibition of CYP3A

with Midazolam. And when the sponsor conducted a

study with multiple dose, multiday evaluation, when

the results are negative, you could claim that this

is not an inhibitor. At the same time, you could

also say it's not an inducer.

Study design data analysis is key and

should be well thought out so that we can provide

220

important information for proper labeling. In our

concept paper, we said we need to design a study to

maximize seeing an effect. And we said that when

you are starting with an inhibitor, we'd like to

use the highest dose, shortest-dosing interval of

an inhibitor.

A common question is always, well, if we

are evaluating inhibitor effect using ketaconazole

to evaluate a CYP3A inhibition, should we use 400

or 200 milligrams? Many of our submissions use

multiple doses. And so, the question is really

whether what is the dose level that should be

employed.

The literature data has many studies using

400 or 200 milligrams. However, they have varied

study design. The difference in study length,

timing of coadministration or different

populations, so it's difficult to compare

intrastudy. And that's why later on, I will show a

study where we compare within study, where the

subject was given both 200 and 400 and make a

direct comparison.

221

However, in one literature data, one

publication has shown that ketoconazole CMAX

concentration appeared to show a correlation with

the inhibition effect on Midazolam. If you look at

the AUC ratio where Midazolam was given with

ketoconazole versus when it's given alone, you can

see the ratio increase as ketoconazole levels

increase.

This study was conducted only with one

dose of ketoconazole, but this is the initial base

of our recommendation to sponsors that we should

use a higher dose of ketoconazole when conducting

interaction studies, however, we did include a

study to evaluate 200 versus 400 milligrams of

ketoconazole as part of a collaborative research

and development collaboration with Indiana

University, and Dr. Steve Hall is the principal

investigator for the collaboration.

And this is a preliminary result that was

shown from that study, where Midazolam, after IV

and oral were compared when it's given together

with 200 milligram dose of ketoconazole or 400

222

milligrams of ketoconazole given for six or seven

days. You can see that after IV administration,

the extent of interaction is smaller as compared to

oral. It's about fourfold after the 400 milligram

dose, and it's about threefold after the 200

milligram dose.

After oral administration, the extent of

interaction is much higher. The AUC ratio is about

15 after the 400 milligram dose; it's about tenfold

after the 200 milligram ketoconazole dose. So

based on the literature data and the study

comparison that I just showed you demonstrate that

CYP3A inhibition after ketoconazole is dose

dependent with 400 milligram dose having a higher

effect than a 200 milligram dose. And we believe

that inhibition studies with ketoconazole should be

conducted at a 400 milligram dose.

However, we have seen in many applications

that a study is already being done with 200

milligram doses. So questions always come up:

well, if you're already studying at 200 milligrams,

do you need to conduct another study with 400? And

223

there are several cases where the sponsor went back

and conducted a 400 milligram dose, and it showed a

difference. The 400 milligram produced a higher

extent of interaction.

There is also a case where a 200 milligram

dose was already demonstrated to have a very large

extent of interaction and is likely to result in a

contraindication. In that need, may not need to

have an additional study, because if you already

know what 400 milligram results will--data will

result in what kind of labeling; it's probably very

similar. It's a contraindication. So in that

case, you don't need to conduct another study. We

need to look at the results and other information

such as exposure response before we automatically

request an additional study.

What about other study design issues?

This one was not directly addressed in the concept

paper, but it was frequently asked: can we use the

cocktail approach where, in vivo, a mixture of pro

substrates for three to five of the major CYP

enzymes were given together with the new molecular

224

entity to evaluate the new molecular entity as an

inhibitor or inducer?

We say yes, they can be used if they are

properly designed; probes are specific; they do not

interact with each other, and there are a

sufficient number of subjects that are used in the

evaluation and if the results are negative, then,

we could preclude further evaluation. However,

many of these cocktail studies used a ratio such as

metabolic ratio in the urine or plasma level, and

it's difficult to extrapolate to assess what would

be the extent of interaction, unlike the studies

that we used where you look at AUC ratios, where

you know it's a fivefold increase or a tenfold

increase. In that case, then, we may need

additional evaluation to provide some quantitative

information.

And again, we have seen cases where some

of the older cocktails were used, and one of the

probes may not be specific, and it may interact

with one of the--it also affect the other CYP. The

other--the data from the other CYPs can still be

225

used, and it could be used in combination of other

in vitro-in vivo data. It could still provide

useful information. Again, we don't automatically

throw away data from a study just because it's not

well designed and certain parts of a design.

Again, the design issue, we were often

asked what kind of substrates or inhibitors or

inducers that should be used both in vivo and in

vitro? What concentrations of substrates should be

used in vitro? We've been asked so many questions,

and this happened always in a sponsor meeting. So

the working group thought it would be good if we

can provide tables in the concept paper on some of

the proven or good in vivo and in vitro probe

substrates, inhibitor inducers.

Earlier, we thought this may be too

proscriptive, and the tables may be outdated

frequently. And we thought we could address it by

using a Web link so we can provide more frequent

updates of the tables than the guidance itself.

And this is just one example of in vivo

probes that we have included in the concept paper.

226

You can see, in addition to the five major CYPs, we

also included information on 2B6, 2C8, since these

are emerging, and 2E1, and you can see that in some

of the well-defined polymorphic enzymes such as

2C9, 2C19 and 2D6, we also think that the

evaluation of pharmacokinetics in poor metabolites

of those enzymes and compare that to the extensive

metabolizer, and this could be done in lieu of a

drug interaction study.

We also indicated that for 1A2, since we

couldn't find a good inducers, since omeprazole has

not been consistently providing induction effect

based on some of the criteria that we mentioned in

our table that made these drugs onto the list, so

we provided that perhaps the pharmacokinetic

evaluation of smokers versus nonsmokers could be

conducted in lieu of an induction study.

And this, I already mentioned, that the

kinetic evaluation in poor metabolizer or smokers

can be used, and we also mentioned, we put a

statement that it may be important to evaluate

interaction based on a pathway in poor metabolizers

227

of enzymes, of the other pathway, which is

considered to be major and the extensive

metabolizer. For example, if the drug is a

substrate for both CYP2D6 and 3A, then, in poor

metabolizer, the CYP3A may be an important pathway,

and you may want to consider the evaluation of that

pathway.

In addition, based on information that we

know about herb, dietary supplement interactions,

juice, food interactions, we thought it's important

to also start to look at the protocols, and we

provided some sample language that should be

included in a clinical protocol when we evaluate

drug interaction, so that when we look at the

interaction results, they're not compromised by the

unknown factors that are contributed by these other

factors.

The concept paper not only discussed

metabolism-based drug interactions, but it also

included transporter-based drug interactions,

although right now, we focus only on PGP-based

interactions. In our concept paper, we mentioned

228

that if a new molecular entity is an inhibitor of

PGP in vitro, then, we think a clinical study using

digoxin may be appropriate. And we have discussed

this in the April meeting last year, and this was

just a summary of some of the data that are

presented at that time.

This is the digoxin plasma AUC or

steady-state concentration that's the ratio when

it's given with these drugs or without. And you

can look at some of the known inhibitors of PGP:

quinidine, retonavir, verapamil, has increased the

ratio to 1.5 to 2.5-fold. Here, grapefruit juice,

aprepitant did not show an interaction.

The known inducers of PGP, St. John's

wort, rifampin, has shown to reduce the plasma

concentration by 20 to 30 percent. And we'd like

to ask the Committee to comment on this point

again.

So we talk about the new molecular entity

as an inhibitor. What about it as a substrate?

And we thought it's important to discuss it with

the status of its CYP3A, whether it's a CYP3A

229

substrate or not. So we said in a concept paper if

a new molecular entity is a substrate for PGP and

CYP3A, and we have a lot of cases like this, then,

the clinical study with a dual inhibitor or a

multi-inhibitor may be appropriate. We just put in

ritonavir as an example, because ritonavir affects

multiple pathways, and here, we're just using

example data from vardenafil labeling, where you

see the AUC ratio of vardenafil when it's given

with these compounds as compared to when it's given

by itself.

And you can look at ritonavir, indinavir,

ketoconazole. Vardenafil is a CYP3A substrate, and

you can look at the strong CYP3A substrates have

shown a large degree of interaction. It's more

than tenfold, and here, ketoconazole is only given

as 200.

The moderate inhibitor, I will explain

about in the classification on CYP3A inhibitors

later, but erythromycin has shown a little bit

lower than fivefold increase in vardenafil.

This should show even these three

230

compounds are classified as strong 3A inhibitors,

but they did show some differential effect. And

so, there's a possibility that ritonavir, because

of its effect on other pathways, in addition to PGP

and other transporters that contributed to a much

larger effect on the substrate.

So we say if a new molecular entity is a

substrate of PGP but not a substrate of 3A, then, a

clinical study with regular known PGP inhibitor may

be appropriate. Again, it's hard to differentiate,

because some of the compounds that are listed here

are also 3A inhibitors, but they're not as strong

an inhibitor.

And here, this is the same table I have

listed earlier with digoxin, so you can see one of

these PGP inhibitors could be used when we have a

new molecular entity which is a substrate of PGP

but not 3A. So we're asking the Committee to

consider whether CYP3A status should be a key

factor when we decide what kind of inhibition study

to conduct, when the drug is a PGP substrate and

also whether we have sufficient data to recommend

231

routine evaluation of PGP interaction if a

substrate, if a drug is a substrate of PGP.

Finally, the last issues regarding study

design: we put in some statement in the concept

paper about the use of multiple inhibitors or

multiple impaired system. When we evaluate the

possibility of a serious adverse events such as we

use the QT prolongation to assess the probability

of trassar DuPont's, we have recommended in the QT

concept paper, actually, it's an ICH document right

now, to use perhaps a strong inhibitor of the major

pathway.

In addition, we have seen examples where

either the reviewer has recommended or the sponsor

has conducted that multiple inhibitors--this is

different than multi-inhibitor. It's a multiple

inhibitors to attack different pathways or, using

one inhibitor for one pathway in poor metabolizers

of the other pathway in the evaluation.

And we have examples such as

telithromycine. An inhibitor such as ketoconazole

was used in the evaluation of a QT prolongation to

232

obtain maximum exposure. We also have cases where

a strong inhibitor was used, for example, when we

evaluated vardenafil. A separate study prior to

the QT evaluation was conducted to estimate what

the maximum exposure that's attainable with a

strong inhibitor; then, use that information to

design a high dose study to evaluate QT

prolongation.

Finally, though not directly related to

this issue, we think the use of multiple inhibitors

of one pathway is also important. Particularly,

right now, we're talking about possibility of

classifying CYP3A inhibitors to moderate inhibitors

and possibly monoinhibitors. That was suggested in

the PhRMA position paper, and we have research

ongoing again with Indiana University, looking at

multiple moderate inhibitors' effect, whether they

would be additive or synergistic or producing an

effect like you're giving a strong inhibitor.

Next point I'd like to stress is this is

the same point that we have stressed in the

previous guidance in 1999, that it's important to

233

establish a therapeutic equivalency boundary for

the new molecular entity, so we will be able to

interpret the extent of interaction based on

interaction studies and what to put in the

labeling.

And here, I am going to present a

hypothetical case where we use combined data from

different applications. This new molecular entity

was given with ketoconazole, a strong 3A inhibitor.

This new molecular entity is a 3A substrate. And

you look at the CMAX increased by fourfold.

The moderate inhibitors: erythromycin,

verapamil, increased by threefold. The AUC showed

similar effect. I put CMAX here because one of the

adverse events was believed to be related to a

maximum concentration.

And we look at exposure response data,

where from the safety and efficacy database, we try

to relate the exposure to one of the endpoints for

efficacy, and one of the endpoints was adverse

events. Here, I simplified the outcome. Actually,

we have several endpoints for both efficacy and

234

safety. And based on the data, between 15 and 60,

the exposure, consider that the drug will be

efficacious and safe. However, because of

ketoconazole's effect, it's varied. It's very

large. We think it's important to advise against

abusing strong inhibitors with this drug.

For moderate inhibitors, if you approve

the dose of 15 and 30, since if you give 30

milligrams, and the moderate inhibitors will

increase the exposure to outside the safe and

effective exposure range. So we would recommend to

use a lower dose.

My final point is that labeling language

needs to be useful and needs to be consistent. In

our concept paper, where we said that if a drug has

been determined to be a sensitive substrate or a

CYP3A substrate with a narrow therapeutic range,

and I'll explain a definition later, and it does

not need to be tested with all strong or moderate

inhibitors of 3A in order to warn about it in the

labeling.

And in the concept paper, we gave

235

examples. We have many tables. And one table is,

well, strong, examples of strong 3A inhibitors or

moderate CYP3A inhibitors. Here, the strong 3A

inhibitors, we have included. The definition is

any substrate, any--if an inhibitor, if that caused

more than fivefold increase in the area under the

curve of a CYP3A substrate. And that's not limited

to Midazolam, then, we listed it as a strong

inhibitor.

The PhRMA paper specifically talks about

Midazolam. But since there are many strong CYP3A

inhibitors, we do not have Midazolam data, and we

think it's important to include these strong

inhibitors in the table, since we do have

information from the other.

The moderate inhibitors, we have similar

definition with a PhRMA white paper, except we

added some specifics. We said that a moderate

inhibitor is one that caused a more than two but

less than fivefold increase in area under the curve

of a sensitive substrate. It has to be a sensitive

substrate, and the inhibitor needs to be given at

236

the highest dose and lowest, shortest dosing

interval, so that we won't misclassify a strong

inhibitor because a study was conducted with a low

dose, a long dosing interval, or it's not--it was

conducted not with a sensitive substrate, so you

may underestimate the extent of interaction and

therefore misclassify.

And one example I've already shown that

even the study was only conducted with ketoconazole

for a strong inhibitor but it does not prevent us

from labeling it with other strong inhibitors. And

for moderate inhibitors, even only done with

erythromycin and verapamil, we will be able to

label with the other additional moderate

inhibitors.

In the concept paper, we also mentioned

that if a drug has been determined to be a strong

inhibitor of 3A, it does not need to be tested with

all sensitive substrates or substrates here

specific about CYP3A with a narrow therapeutic

range. And in the concept paper, we included

examples of sensitive substrates or substrates with

237

a narrow therapeutic range.

This is a new definition. The PhRMA white

paper did not discuss a sensitive substrate in the

definition. And here, we defined that sensitive

substrates are drugs that AUC will increase

fivefold or more with an inhibitor. It doesn't

have to be a strong inhibitor; any inhibitor.

A CYP3A substrate with a narrow

therapeutic range: this would be applicable to

drugs that are not a sensitive substrate. However,

the increase in exposure because of

coadministration with a CYP3A inhibitor may result

in serious safety concerns, such as trussar DuPont,

so you can see there are quite a few drugs:

cisapride, astemizole, terfenadine; these were

removed from the market but are included in the

table just for illustration purposes.

An example of a labeling based on this

table would be--I'm using telithromycin as a case.

This drug, when, it's given with Midazolam,

increased the area under the curve by sixfold, so

in definition, it's a strong inhibitor. So in the

238

labeling, we said telithromycin is a strong

inhibitor of the cytochrome p4503A, and we also

said the use of simvastatin and other statins here

concomitantly with telithromycin should be avoided.

We also said that the use of telithromycin

is contraindicated with cisapride and pimozide.

And you will notice, based in the information in

the summary of our study, we did not evaluate all

of these drugs that are listed here. For sensitive

substrates, we only evaluated with simvastatin, but

it does not prevent us from listing other sensitive

substrates.

For substrates, CYP3A substrates with

narrow therapeutic range, the pimozide was not

evaluated. But again, because of what we classify

it as a substrate with narrow therapeutic range, we

put it in our labeling. Right now, we have various

discussions on how to label strong inhibitors; what

sensitive substrates to put in the labeling when we

are evaluating one, and we may come up with a

different list. Therefore, we think it's important

that we publish the labels and constantly update it

239

so that we have consistency among the labeling of

different drugs.

So in summary, we think metabolism drug

interaction is key to benefit-risk assessment, and

I think based on today's discussion, we probably

will add some transporter information as well. An

integrated approach may reduce the number of

necessary studies and optimize our knowledge.

Study design data analysis is important and

information for proper labeling, and we have

devoted many pages of our concept paper on study

design, and we've also added an appendix on the

conditions of in vitro evaluation: what are the

study design and data analysis issues?

The thing we need to establish,

therapeutic equivalency boundaries, so we can have

proper interpretation of the clinical outcome and

put it into a useful information in the labeling,

and we have added tables of classification of CYP3A

inhibitors, substrates, to hopefully that we have

consistent and useful labeling.

And I'd like to acknowledge the drug

240

interaction working group. It consisted of many

members from our office, the Office of Clinical

Pharmacology and Biopharmaceutics; from individuals

outside our office, the Office of Pharmaceutical

Science; individuals from CBER; some of them have

joined us after the reorganization and also from

the Office of Medical Policy.

I think my time is up. Do you want to

take any clarification questions?

DR. VENITZ: Thank you.

Any comments or clarification questions

for Shiew-Mei by the Committee?

DR. JUSKO: I have one.

DR. VENITZ: Go ahead.

DR. JUSKO: Shiew-Mei, that was very clear

and impressive. With the study of the ketoconazole

interaction that you showed, I didn't see that

using 400 milligram ketoconazole was that much

better than 200, and I would have come to the same

conclusions with either dose. Why are you so firm

on 400, where there may be some additional negative

aspects as opposed to 200?

241

DR. HUANG: The example, since we use a

sensitive substrate with Midazolam, you probably

can see, well, 200 milligrams already gives you a

tenfold increase, and we probably will classify to

say not to use it together with some more sensitive

substrate already.

If the exposure response data are such

that, then, you don't need to do another study.

However, we have a lot of compounds where CYP3A is

part of the pathway. So in that case, the results

are not clear cut. We did have one example where a

threefold and fivefold difference, from these two

different doses, and it would make a difference.

For example, one of the examples that I showed you,

the moderate inhibitor and strong inhibitors, one

shows fourfold; one shows threefold, and we do have

a different proposal for labeling, because

depending on the compound's exposure response,

fourfold increase is going to take you outside that

safe and effective exposure range; then, you would

contraindicate. But if it's threefold, it may

still be within the range, and you can either using

242

a dose reduction in the labeling to address that

issue.

So for less sensitive substrates, the

difference, three to fourfold or versus one to

twofold, it will make a difference in the proposal

and the labeling.

DR. GIACOMINI: Do I have that--yes,

Shiew-Mei, that was an excellent and very clear

presentation. I just have a couple of comments

related, of course, to transporters and how we have

to begin thinking of not really metabolism but more

metabolic pathways, which would include maybe an

influx transporter, the enzyme and then an eflux

transporter, which may be all part of a pathway.

So when you've indicated here, and you've

got particularly sensitive substrates, and you have

examples of inhibitors, and many of these are

dirty; they'll inhibit different things along the

pathway, and I think it would be helpful in this

paper at some point to at least indicate what may

be a dual substrate and a dual inhibitor, and are

you planning to do that dual, triple, whatever?

243

DR. HUANG: There are a lot of

publications that did suggest this, and what we

want to put in the guidance is where it's going to

be useful in the study design or in the labeling,

what's going to be translated to a clinical

setting. So any information that may not be

helpful; for example, if we say we evaluate this

drug as an MRP substrate, and we know it's both

CYP3A and MRP.

However, we really couldn't recommend to a

sponsor a certain type of study to conduct besides

a PGP. We do have some proposals; or we don't know

what to do with the data, and how would that help

prescribing a physician or health care provider's?

Then, we don't think that that will belong to the

guidance. It will belong to the literature, and we

have enough information to make a recommendation

under this case, what should you do in your study

design? Then, we will include that in the

guidance.

DR. GIACOMINI: I mean, I agree with you.

I hear what you're saying, but it seems to me like

244

if somebody has done some kind of an inhibition

study, they may make an interpretation; they're

going to use that to make some kind of an

interpretation, and you're focused mostly on the

interpretation as it relates to the dose of the

drug. But I'm just wondering about even a

mechanistic interpretation by at least indicating

that certain substances may, certain compounds may

be inhibitors of both a transporter and an enzyme,

that some caution in the--especially, you're going

to extrapolate, right? Because you're going to say

well, now, that we showed this, you better be

careful for all of these, all of these compounds,

which may also be substrates of CYP3A4 when, in

fact, the transporter was the bigger part of the

interaction, and that wasn't--

DR. HUANG: Yes, well, I would welcome the

Committee's discussion, because I did have some

question to see what other transporter that should

be evaluated. But you will notice, even we put all

the tables, when we want to translate one study to

the others and put in the labeling, we only provide

245

some very specific information. For example, we

say ketoconazole, an inhibitor, and whatever

happens with that result, you can translate to the

others.

When a study was conducted such as a

cyclosporin study, and with rosovostatin, when we

couldn't translate that in the labeling, we only

said when cyclosporin increased rosovostatin by

sevenfold, and therefore, the initial dose should

be this, and we do not translate that to others.

So until we know more, then, I think we

will be able to put in the table and put in the

guidance you're suggesting.

DR. DERENDORF: I'm very happy to see that

induction is addressed, and that was overdue in

both in vitro and in vivo. Now, in the in vitro

part, I have a question: it says if 40 percent of

positive controls suggest possible induction

potential, does that mean, first of all, 40 percent

of what? What will be the measure? And the

positive control will be defined, because

otherwise, you can change, you know, the percentage

246

based on your control.

DR. HUANG: Yes, you will hear more from

the subsequent speakers, but I can say in our

guidance, we--sorry, concept paper, we did

recommend that, for example, with CYP3A, you look

at revamping induction, and we use the enzyme

activity, the fold increase in enzyme activity. So

if it's increased tenfold, then, 40 percent would

be fourfold.

DR. DERENDORF: With respect to the in

vivo, you have the classification of strong and

moderate with two and fivefold increases in

exposure. If we apply that to induction, would

that mean that a fivefold decrease in exposure

would also be the border between moderate and

strong?

DR. HUANG: Well, we did propose that at

the April Committee meeting last time, and the

comments from all of you was that we don't have

sufficient data to indicate which one is a strong

inducer, and we just don't have the information.

But I'll be happy to revisit that if the Committee

247

thinks it's proper that we do that.

DR. HALL: Yes, could you comment

on--you've done a really good job of talking about

when to conclude something is an inhibitor, but

when it came to concluding that it was not an

inhibitor, you somewhat skirted around that. And I

think, you know, there are clearly labeling

advantages to being able to conclude it's not. You

mentioned you have no recommendation, but the

former working group did come up with a

recommendation.

And it seems like that would be an

important thing to address.

DR. HUANG: Well, I'd love to hear the

Committee's recommendation. The PhRMA white paper,

as far as inhibition, it says if I over KI ratio is

more than 1, it's likely; if it's between 0.1 and

1, it's probable, and I believe it's--when it's

less than 0.1, it's at least--well, I don't

remember the exact words, but it's not likely. And

we do have, we have cases where the ratio of 0.1,

you still see some interactions.

248

So it is difficult to say--I mean, rare

occasions to say that it directly translates,

especially for 3A, since the inhibition, I mean,

the induction could happen--I'm sorry, fall back;

inhibition could also happen in the GI tract. So

when you use the equation, you might be able to

come up, to derive an equation to say, well, 0.1,

it's going to result in very small extent of

interaction.

But if you consider the other components

in the GI and also the uncertainty of the

concentration in the hepatocyte as compared to what

we are using right now, plasma concentration, and

that's why we are using a more conservative

approach. We did not use exactly 0.1, although we

did mention when you have a large gap between I and

KI, and we put in the numbers and say if it's

fiftyfold, then, it's not likely there's

interaction. I know it's a very conservative

number.

And that's why we're proposing perhaps we

could use a rank order evaluation. Any time in

249

doubt, you probably want to study with the one with

the strongest inhibition, that is, the smallest KI.

And if the in vitro data, in vivo also show no

interaction, then, you do not have to do the other

studies. If there are other alternative

approaches, the working group will be very willing

to listen to the suggestion.

DR. SADEE: Shiew-Mei, I have a comment

and maybe a question about HIV therapy, which takes

advantage of many of the agents that you have shown

to be interacting, so we expect a lot of

interactions. Now, in that case, physicians use

retonavir to, in fact, as a dosage sparing agent;

in other words, you block probably PGP; you block

3A4 and a number of other cytochromes by adding

retonavir; then, you systematically adduce the

other agents because of that knowledge.

Unfortunately, these patients are also

given statins. They're given antineoplastic

agents, antidepressants, you name it. So this is

the inverse usage of the information of inhibitors,

and it appears to me that it has tremendous effect

250

on the adverse effects that are pretty prevalent in

HIV patients. So is this something you also want

to look at, or I was personally very surprised to

hear these relatively nonchalant views of the

inhibitors to spare other agents, whereas, to me,

it would induce a lot more problems.

DR. HUANG: Yes, you mentioned kalitra,

which is a combination of retonavir and lopinavir,

and a low dose of retonavir was used to increase

the exposure of lopinavir to its HIV therapeutic

effect. In that case, there's no difference in how

we treat the evaluation of kalitra as an inhibitor

or inducer if it's submitted today. So we have the

package included many interaction studies based on

that we already know retonavir is a CYP3A

inhibition, and there are many studies that were

conducted because of the nature of the HIV therapy,

and many of these studies, the results were

summarized in a table format, and there's also

certain for kalitra, I think most of the study

results were summarized in tables. I don't think

there's extrapolation of the conducted study.

251

And this is true for most of the HIV

therapy. When you submit a new molecular entity,

this drug's effect on others will be evaluated as a

standard procedure almost the same as what we have

described in the concept paper. If you're going to

evaluate multiple inhibitors, how that would

effect, and I think there's some ongoing research

project that we hope to conduct using modeling and

simulation and just see how different the various

inhibitors or various pathways will result in an

extent of drug interaction.

And the study that we're conducting with

Indiana, and it's only a very first step to looking

at multiple inhibitors of one substrate, how would

that conduct? How would that compare to a

simulation outcome? And what you envision is much

more complex.

DR. WATKINS: You know, that was a great

presentation, and I think it's a great idea to try

to merge the two old documents and come up with new

guidelines, but I suspect to industry, it's not

going to be reassuring that the reason to do this

252

is that we can now Web link the different

substrates that could change week to week is sort

of the implication.

But I think the document then needs to

stress the fact that unfortunately, this is still a

work in progress. We really haven't refined the

probes, for instance, for PGP and the issues Cathy

was bringing up of separating out transport from

metabolism and the interactions of transport and

metabolism. And the last thing you would want is

with the publication of this guidance for upper

management in a major pharmaceutical company to

feel that this had been solved and that the

scientists could be put onto other projects,

because there's so much work left to do in the

area.

DR. HUANG: Right, we--the idea of the Web

link tables is, I think, because the last guidance

was published in 1999, and this is not likely to be

finalized until 2005. So it's a six-year gap. And

with the Web link, I think we can do maybe more

frequent than every six years.

253

DR. REYNOLDS: I just wanted to address

the ritonavir issue. When a company is evaluating

an HIV droug, and ritonavir will be part of the

regimen, we really consider the drug plus ritonavir

the drug. So if the drug will be given alone, or

if it will be given with ritonavir, we expect them

to look at it both ways, and we consider the

interactions very seriously.

DR. SADEE: On that also, interactions

with statins, for instance, which are very often

given or metabolized and transported by very

similar gene products.

DR. REYNOLDS: Right, so we would expect

to understand the interaction of the protease

inhibitor plus ritonavir on the statin.

DR. BLASCHKE: I think coming back to

Wolfgang's point, as you're saying, I think,

virtually all patients who are getting a protease

inhibitor for HIV are also getting ritonavir, and

what I think it speaks to is probably the

importance--and they're also getting these multiple

other drugs, whether they're statins or CNS active

254

drugs and so forth, and it probably really speaks

to the importance of most drugs in which we think

there's any possibility of using that in

HIV-positive patients, that the ritonavir should be

one of the drugs that is studied rather than some

alternative.

DR. VENITZ: One comment: as you know,

I'm very much in favor of using this approach to

minimize the amount of studies that need to be

done. One concern that I have, and I mentioned

that before, is whenever you talk about dose

adjusting based on either inhibition or induction

data, you're basically trying to match areas under

the curve or something like that for the parent

drug.

What you don't necessarily consider, and I

suggest you incorporate that in your paper, in your

guidance, the change in the metabolite profile.

You're reducing the dose. It's not the same as

inhibiting a particular pathway. You all of a

sudden have a metabolite in higher concentrations

than it would be, okay? So I'm not sure whether

255

that's relevant for specific drugs, but it may well

be, depending on whether the metabolite contributes

to activity, meaning safety or efficacy.

But it's something that I haven't seen in

any of the documents that you've provided us.

DR. REYNOLDS: Correct, yes.

DR. VENITZ: Okay; any other comments or

questions?

[No response.]

DR. VENITZ: Then, thank you, Shiew-Mei,

and our next speaker is Dr. Keith Gottesdiener from

Merck, who's going to give us the scientific

perspective.

DR. GOTTESDIENER: Thanks very much for

inviting me here today. Before I start, I'd just

like to also let you know that many Merck

colleagues helped me to put this talk together, and

I just wanted to acknowledge some of the people who

had actually worked on this talk as well.

It's a real pleasure to be here today.

I'm in charge of early development and clinical

pharmacology at Merck, and to a great extent, what

256

I do or a major part of what I do every day, every

month, every year is really looking at this

question from inside the industry as opposed as to

from outside. Of course, the FDA is very

interested that the packages we put together and we

submit for registration of a drug be complete.

In a sense, I get to do that sometimes

months and years ahead of the FDA, and it's really

my job to really make sure that package is robust

and to try to put it together. And so, in a sense,

I think that both myself and the people in the

industry who do these kinds of things have a very

unique perspective. We get to see a lot of

compounds that never really make it past this

evaluation stage as well as those that actually go

forward to filing, and I hope I'll share some of my

thoughts with you today.

I can assure you that senior management

does not think that this problem is solved yet

today, and I'll point out some of the issues. I

wasn't able to really participate in the last

meeting where you talked about induction, but some

257

of the questions that came up to Shiew-Mei are

exactly very similar to the kinds of questions that

I would pose as well: how difficult it is

sometimes to do this in a real life situation.

So what I'm going to do is talk just a

minute about the approach to assessing drug

interactions. I'll talk about the many areas of

agreement with the concept paper that exist, which

I really have to applaud. It's a real step

forward. I'll mention a couple of areas where I

think there's really some further discussion, and I

just pick three today: induction, transporters,

and this issue of multiple inhibitors, multiple

impaired.

I'm not going to cover specific comments

on the concept paper. I do have many. I've shared

some of them with Shiew-Mei as well. The study

designs, the tables, et cetera; when that comes out

as a draft guidance, I'm sure I'll have plenty of

opportunity to comment, and nor was I going to

spend much time talking about specific comments on

the questions to the Committee.

258

What I wanted to do was focus on the

approach to some of these issues. When I think

about approaching, assessing drug interactions, I'm

really probably saying this slide or two, I'm

probably talking to the wrong audience. I often

have to explain to people what the approach is.

But clearly, the issue is how should we adjust the

dose of a substrate drug in the presence of an

interacting drug? And which DDIs and which drug

interactions to study, how to answer that question?

And clearly, we're moving from the past,

when this choice was largely empirical by the

likelihood of coadministration, clinical

consequences of the interaction towards a

science-driven approach, particularly where

feasible. You know, we're using preclinical in

vitro studies to determine in vivo studies, in vivo

studies using probe substrates and really robust

study designs. But clearly, I think there's ideas

where the science is evolving and the necessary

tools and the probes are still lacking.

We also think it's important, again, like

259

the FDA, that there be prespecified criteria to

compare the PK or PD measures to the drug in the

presence and the absence of the interacting drug,

and clearly, this is based on the safety and the

efficacy profile of the substrate drug, the

therapeutic index, the clinical context of the use

of the drug, which I think actually is quite

important and very hard to capture in the guidance

and the concentration and response data for the

substrate, which is obviously something this

Committee is very interested in, and so are we.

One thing, though, I do want to emphasize

is often, this is not clearly positive or negative.

It's very difficult if something is or is not an

inhibitor; it does or does not have a clinically

relevant effect on one drug or another, and

actually, probably, the one comment I'd make about

the questions today, it will probably be the only

one, is the questions are really framed as

either-or. If it is an inhibitor, this is what you

should do.

And in many cases, I think the guidance

260

the Committee is going to give, and it's going to

be quite interesting; my problem is trying to

decide is it an inhibitor or not in many ways, and

I know that the FDA struggles with that very

question. For the NCE or the NME, the data is

often quite limited; concentration response info is

always better for efficacy than it is for safety,

and I still think there's many areas for probe

substrates, where there really isn't much

consensus, even though I think we've come quite

far. And as I'll point out today, induction in

particular is problematic.

But let me talk first about all the good

points. I think that the integrated and scientific

approach is clearly the right step forward. I

think we've made a lot of progress and clarity on

CYP interactions, especially the in vitro-in vivo

correlations and the clarity on the substrates,

inhibitors or inducers, though I do have to

comment: somehow, simvostatin is on the list of

sensitive inhibitors twice, and I didn't know if

that was a hint from the agency to Merck or not or

261

whether it was just a typo.

I'd certainly agree with the use of PK in

poor metabolizers where appropriate. I think the

robust study designs is really important, and in

many ways, I applaud the efforts and the question,

the slides that Shiew-Mei showed, for example, on

doing a ketoconazole interaction study. Having

read that literature for many, many years and

struggling with that issue day by day, the issue of

whether you're going to do 200 and 400, whether

you're going to dose one day, three days, five days

or a week, those are real issues that have real

impact on how the results come, and it's all too

easy to pick a study design that really will, in a

sense, manipulate the result so that it comes out

the way you'd like it to rather than the way it's

most scientifically correct.

So I think really, we're going to be in

much better shape as we start to look at robust

designs and receive, you know, gain a little bit of

clarity on which ones really give us the best

information.

262

I also applaud useful and consistent

labeling language. Part of my job is to read to

every new label that comes out and see what the FDA

says about every new drug, and I understand the

desire to be consistent, really quite hard in the

field of drug interactions, and I'm always so

struck about how difficult that is. And of course,

then, I also get to compare it to what happens in

the EU and the rest of the world, where a whole new

variety of approaches come forward as well.

So I think that this is a real step

forward, but I do think there still needs some

discussion on how to label moderate inhibitors, how

to define sensitive CYP substrates, and I must

admit: all the devil is in the details. So while

I agree with the principles the FDA has said, it's

really going to be what's in the tables and how

it's translated into labeling language, I think,

that I'm going to be very curious to see.

So let me talk now about some of the

issues that I think are worth discussing, where I'm

not sure I fully agree with the concept paper or

263

some of the issues that you may grapple with today.

And I think about things very, very practically.

You know, where are we today with in vitro

predictions of in vivo drug-drug interactions? And

I think ready for prime time. You know, we really

understand where we are; things related to CYP

inhibition, particularly for the five major CYPs.

I think almost ready for prime time are

some of the PGP interactions, UGT and some of the

other CYPs, I won't talk about them today, and CYP

induction. And there, actually, I think there are

some issues with the tools we can use in vitro, but

I also think there's many more issues actually in

the in vivo studies that follow, because in the

end, as I look at this, it isn't only an interest;

it isn't only of interest to me to try to predict

from in vitro what is going to happen in vivo, but

there's also the issue of how do I interpret what

happened in vivo, obviously, into something that's

useful, so that we can actually use the drug

properly?

And then, I really think that many things

264

are not ready for prime time, though I applaud the

science moving forward. Most transporters, to me,

still are in this gray, murky area where I have a

very difficult time understanding how to use them.

And I'll also point out some of my difficulties

with multiple paths of inhibition as well.

So induction, I don't have to talk to you

about. There's a lot of concerns for induction.

Mostly, it's related to the reduction in

therapeutic efficacy. Auto-induction is also a big

concern; rarely the imbalance between toxification

and detoxification. It's dose and time dependent.

The study designs become really quite important

here. It's dependent on clearance and route of

administration. Again, a study design issue, and I

should also point out it's a concern with both

initiation and discontinuation of an interacting

drug.

We have many models or many tools now to

talk about CYP induction, and animal models were

previously used. You know, it wasn't that long ago

that that happened. When I arrived at Merck nine

265

years ago, the only way we really assessed

induction was by doing high dose, short-term

studies in rodents instead of looking at liver

weights and science of induction.

You know, nine years is a very short time.

And today, we're talking about, really, a whole new

category of tools. Dr. LeCluyse is going to talk

about that. But clearly, those were poor

predicters. We looked at those, and we shrugged,

and we went ahead into the clinic, and we had no

idea what to do with the information.

Nowadays, the in vitro models are much

better: the assays, the primary culture of human

hepatocytes, and they're very, very--clearly, very,

very helpful in the selection of drug candidates.

And in fact, in many ways, that's where their most

helpful nature is. We rule out enormous numbers of

candidates, because they're really positive in

these assays overall.

But I will also tell you it isn't always

that easy. It sounds like it's great: you set a

criteria; you cross off a drug candidate, and you

266

move on. As targets become more complex, the

chemistry becomes more complex; the size of

molecules increases, more and more whole areas of

structures actually carry some risk of induction.

So in many cases, we're not able to cross off those

candidates, and we have to bring them forward into

the clinic.

But the problem I have is still is it

quantitative in vitro-in vivo prediction possible

for induction? And I think there's many factors

that complicate that in vitro-in vivo

extrapolation, particularly inter-individual

variability; plasma protein binding; multiple

mechanisms as well.

Now, I wanted to share with the Committee

some idea of what I actually see as a

vice-president of clinical, you know, pharmacology,

drug development. This is the kid of data I see.

These are hypothetical drugs. None of them is

real. But they're all based on drugs that actually

have made it into the clinic. And as you start to

look at this, you can see some of the things that

267

come forward and some of the complexities.

I picked five drugs where the mouse

five-day study was actually negative, okay, just to

sort of get that off the table. You can see the

human PXR data, the mRNA data. Those are percent

activity of a rifampin control at 10 micromolar;

the enzyme activity in human hepatocytes, and I put

over on the side something that I also think is

important is really what the CMAX concentration was

in the clinic. Sometimes, that's a predicted

value; sometimes, it's an actual value when we get

into the clinic, and we understand efficacy.

And you can see, if you looked at any one

of these drugs here, the question about should we

or shouldn't we do an in vivo induction study is

really quite difficult. Now, I wish I could tell

you what these five drugs were. More importantly,

I wish I could tell you what the results of, for

example, Midazolam studies were for those five

drugs. Many of them have not progressed far enough

in the clinic to have that evaluation, but this is

the kind of data we grapple with every single day.

268

And the question I ask myself is which

ones really need an in vivo study? And I don't

really know. So at present, I could probably only

say that we can predict a likelihood of CYP

induction; highly possible on one hand; less likely

on others. And I kind of gave some examples of

things where at least they fell into the possible

range going forward, and most likely, we would have

done an in vivo study to follow up what's going on.

I should also point that clinical data

does sometimes help. For example, we see--often,

we see evidence of autoinduction, which helps to

clarify the issue in a particular clinical dose.

But I think once again, it really depends to some

degree on what kind of exposures one has in the

clinic, and that helps in some ways to really

interpret the data.

Now, of course, I think that's the easy

part. I think the hard part is actually

interpreting an in vivo study. I think there's

less consensus on probe substrates, their clinical

interpretation. I wasn't here for the Committee

269

deliberations last time about the issue of

induction, but I find these questions quite

difficult.

And this is just a slide showing the

percent of baseline exposure for a variety of known

inducers. I don't think you have to pay attention

to the specific data. Many of these were studied

on numerous occasions. But this is the effect on

oral Midazolam. And I look at this data, and I'm

asking myself really, is this where we think the

bar should be? This is just about a fivefold

decrease, or should the bar be here? We know the

glucocorticoids and St. John's worts do have

clinical effects on certain drugs, or where should

it be?

And in the end, I still struggle with

really the interpretation of induction, whether

it's the in vitro or in vivo going forward. Now,

more recently, the role of transporters has been

recognized, and I think there's clear examples of

transporter-mediated drug interactions. A couple

of years ago, I don't think I would have actually

270

said that there were clear examples. I would have

said hypothetical or potential examples.

And I certainly think the understanding of

peak lack of protein is advanced greatly. But the

in vitro methods are not really standardized, and

they're not really quite as available as we'd like.

And I think a quantitative in vitro prediction of

in vivo relevance is still quite difficult. And of

course, it's complicated by the fact that the

transporters really are just not an issue of

metabolism but also absorption, tissue

distribution, excretion; as someone said earlier,

the whole pathways can be involved as one is

looking at this influx, eflux, et cetera.

Now, probably peak lack of protein is the

best study, and I don't have to say much to this

particular Committee about that. But even there,

the in vitro methodologies are not quite what I

would like. The transgenic MDR naga mice are a

very powerful tool, but we have numerous examples

where the human and rodent differences occur.

The in vitro tools are clearly becoming

271

more sophisticated, but some of the PGP substrates

don't follow simple kinetics. There's a lot of

overlapping substrates between PGP and CYP3A4, many

inhibitors affect both, and of course, there's the

issue of other transporters as well. The tools are

most useful to identify PGP substrates. We can

certainly identify PGP inhibitors, but it's still

comparatively laborious and time consuming to do

so.

And what might I, as a clinical

pharmacologist, actually get as an evaluation of a

PGP substrate? This is a paragraph, actually, from

a real drug. I changed the numbers once again to

make it a little hypothetical, but you can read

this. What we see is what happens in the MDR mice.

We can look at transport ratios going back and

forth. I have a B to A ratio from the two sides of

1.7. I'm not really sure if that's a substrate or

not; what should I do with that particular data?

Now, if an in vivo study is indicated, and

I've told you I struggle with what that actually,

you know, how do I actually decide that, I think

272

for assessing a potential PGP inhibitor, digoxin

clearly is a suitable probe. I do think that most

other probe PGP substrates are less than ideal.

But I've borrowed a slide from Shiew-Mei

just to talk about some of the difficulties I have

about thinking about assessing PGP substrates.

This is the same slide that Shiew-Mei showed

before. These are all PGP inhibitors, and you can

see in vardenafil what the tremendous difference

there might be between all those, and of course,

part of the difference is that ritonavir, as we

pointed out, oops, doesn't work, ritonavir, as we

pointed out, clearly is an inhibitor not only of

PGP but of CYP3A4, but in this particular case,

it's also an inhibitor of 2C9, which is probably

the other pathway by which vardenafil is actually

metabolized.

And you can see there's a widely divergent

variation in terms of the results one would see.

I'm not sure here that I think that that's really

the kind of data I'd like to be generating to help

understand how to extrapolate datas to new

273

situations.

I think the situation in terms of

inhibitors and doing a study on a substrate with an

inhibitor is even worse. If you look at the

compounds that we have that are inhibitors, there's

quinidine, ritonavir, verapamil, cyclosporin, okay?

These are all very difficult compounds to work

with. We don't use quinidine in volunteers. We

really think it carries too much risk. Ritonavir,

as I've already mentioned, is a 3A4 PGP substrate,

inhibits 2C9.

We've had studies where we've seen

tremendous induction of UGTs, a whole variety.

Some have gone up, some have gone down. In the

end, when we do studies with ritonavir, what we do

is we conclude this is what ritonavir does and

nothing else, because we really just don't know how

to interpret the individual data.

Verapamil, also very complicated; similar

on PGP and CYP3A4, but we stopped using it in our

clinical trials. When we gave it to volunteers,

Dr. Vago in the back there who used to be at Merck

274

did studies for us where we showed that we clearly

saw, you know, PR lengthening in all of our healthy

volunteers, and we just really thought that the

risk-benefit really wouldn't allow us to do that.

And of course, cyclosporin has myriad

effects. The interpretation is difficult, but it

carries a significant risk to volunteers, and we've

been unwilling to do more than single dose studies

in volunteers because of the effects it has on the

kidney and on immunosuppression. As a matter of

fact, this is very real to me. Last Friday, Merck

spent a couple of hours, a whole crowd trying to

design a study requested by the agency to really

try to understand the effect of a model PGP

inhibitor, cyclosporin, on one of our drugs, and we

just found it almost impossible to design a study

that we thought would really be able to answer the

question without significant patient risk.

Other transporters are far less

standardized and available. Many cell-based

systems contain multiple transporters, making it

hard to interpret in vitro. There's few

275

well-defined substrates and inhibitors. The

correlations are difficult. Many of the

interactions can't even be linked to a single

transporter system. And we just don't know how to

generalize these.

So I think that in general, my feeling

about other transporters are that the science

doesn't support an in vitro-in vivo correlation.

Clearly, we're moving that way. I have high hopes

that five years from now, if I stood in front of

the Committee, I'd have a different answer, but

this is what I think today.

Now, I wanted to just close and talk a

little bit about the multiple inhibitors, multiple

impaired. Just like with induction, Dr. LeCluyse

is going to be talking about that; someone is going

to be giving a presentation about this as well, and

I think there's some really elegant work there.

But I have to admit, I'm not a big fan of this

particular approach. I understand the agency's

desire for higher exposures when evaluating QTC

issues, which I think is probably the primary

276

driver for many of these studies.

But the new hurdle for QTC is very, very

stringent. And we certainly agree, I certainly

agree that the margins are critical. But the real

question I ask myself is for how many of the drugs

that are coming forward are extraordinary efforts

justified? And of course, not everybody would

agree with me that some of the things that we're

doing with multiple inhibitors are truly

extraordinary efforts.

But I want to just lead you through a

little bit about how this actually works in

practice you can understand why I take that

approach. First of all, the QTC effects of many

inhibitors are not well-characterized. That's a

solvable problem. And again, I hope that in a

couple of years, we'll know that ketoconazole and

itriconazole and, you know cyclosporin or whatever

we're going to use in these studies, ritonavir

really has no effect on QTC that could mess up

these definition QTC studies.

But the important thing to realize is

277

while the industry agrees with all the agencies

around the world that these studies are important,

these studies are extremely costly and difficult to

do, and the carrot to the industry, okay, is the

fact that if we do these right, and we get an

answer that really satisfies the agency that this

drug does not have a QTC effect, we really need

that information to prevent us from doing

extraordinary efforts in terms of monitoring in

phase three. So it kind of puts a limit on when we

need to really have this data to be most useful in

terms of designing a QTC study.

If it isn't available for us at the end of

a phase 2B study, honestly, it's much less valuable

to us overall. And so, we have to work hard to get

that in. Now, what do we have to do to do a

multiple pathway study? In most cases, we have to

do a clinical study first to really define the in

vivo metabolic pathways. It takes nine months to

set up, six months to analyze. We need clinical

data on each inhibitor separately, really, to

understand the usefulness to increase PK exposures.

278

We have to get that data, we have to model it to

make a prediction what is going to happen when we

look at multiple pathways.

And in most cases, I would argue, you

actually need to test the concomitant

administration of the inhibitors before the QTC

study because of the issue of safety and

tolerability. These studies are not really done at

sites that are really set up to carefully evaluate

sort of phase one type issues, and in many cases,

you'll be giving a new exposure to drug that no one

has ever seen before.

And of course, these QTC studies get quite

complex if people feel dizzy or nauseous or vomit,

have diarrhea, okay? I have to tell you: I am

senior management, but if I went back and said we

couldn't complete our QTC study and that a million

or two dollars are really down the drain because

people were unable to tolerate the drug, I would

not be well received. And so, we have to do that

stuff as well.

Special populations also are needed in

279

some cases, and I think there are some elegant

studies, like the telithromycin study, because

recruiting those particular people is really quite

difficult and time consuming as well.

Now, to make it even worse, those are

logistics issues. This is what I might see,

actually, from a particular drug where we're

considering a multiple pathway. And this is,

again, patterned after a real drug. In vitro data

incident that 3A4 plays the major role, but there's

10 percent from four other CYPs. What inhibitors

should we use? How should we actually design such

a study?

And lastly, I also question are we really

as smart as we think? Despite all our knowledge,

can we really predict the effects of inhibition of

multiple pathways? And I just wanted to give one

example of some data that will be presented at

ASCPT the next year about a study we were asked to

do, which I think was actually a very good study to

request, so I certainly support it, but one where

it showed me that I was a lot less smart than I

280

thought I might have been.

Aprepitant is a moderate CYP inhibitor.

It's used in combination with 5HT3 antagonists, and

dolasetron is a 5HT3 antagonist we had not studied

in our clinical program. It's metabolized by 2D6,

with 3A4 being an important pathway. And because

of the concern of QTC prolongation with the

dolasetron, we did a study at the agency's request

to conduct an aprepitant interaction study in 2D6

extensive in poor metabolizers.

All of the data that we had would have

suggested we should have had a remarkable effect.

And in that study, as you can see from that data,

if you take a look at what happens with dolasetron,

with aprepitant, APR, and dolasetron together, we

were able to show, in fact, yes, poor metabolizers

do have higher levels. But closing off the CYP3A4

route with a moderate inhibitor really had no

important effect at all on the levels. Very

reassuring in this particular case, but clearly not

what we would have predicted overall.

So overall, I think that we really are

281

making a lot of progress. I very much applaud the

efforts of this Committee and the FDA, and I think,

though, that we're really not there yet in all of

the things going forward, and this is sort of my

summary of what's ready today and what I hope will

be ready in the future.

Thank you very much for your patience.

DR. VENITZ: Thank you.

Any comments or questions?

[No response.]

DR. VENITZ: Thank you again.

Our next speaker is Dr. LeCluyse. He is

the chief scientific officer of CellzDirect, Inc.,

and he's going to talk about induction.

DR. LECLUYSE: Excuse me while we do a

technology switch here.

[Pause.]

DR. LECLUYSE: It worked.

Okay; I also would like to thank the

Committee for this opportunity to speak to you.

The way I interpret my role in all of this is that

I think I'm supposed to condense this labyrinth of

282

information that's out there on nuclear receptor

biology and what it all means in terms of human

gene regulation, P450 induction and how to do in

vitro screening for that.

So with that task in mind, let me start

out by just first putting up the questions that

it's my understanding that we are asked to address,

and this is very limited as compared to a number of

issues that Keith brought up and addressed in terms

of the in vitro-in vivo correlations, et cetera,

and some of the complications associated with that.

So I am specifically going to focus on

these questions that were placed in the paper or at

least suggested in the papers, questions that need

to be addressed, such as if a drug's induction

effect on 3A4 in vitro is negative, then, it is

acceptable to not recommend any in vivo studies

with substrates of 3A, 2C9, 2B6 and 2C19, yes or

no?

Also, the other question that was meant to

be addressed today is if the in vitro induction or

increase in enzyme activity is more than 40 percent

283

of the positive control, then, there is a need to

recommend an in vivo induction study, yes or no?

I'm going to focus predominantly on this first

question, because I think that's the one that's

most complicated and involves a little bit more of

a mechanistic understanding of what our current

understanding of regulatory of the human liver

genes, and this, you could argue, is as much of a

philosophical one.

So before we can address that specific

question, especially the first one, let's start off

by first reviewing the enzyme induction in humans

as we currently understand it or as observed in the

clinic.

So, for example, if you take compounds,

and certainly, this is not a complete or

comprehensive list, but it serves to represent the

point that for most drugs that are known to cause

clinically significant drug interactions, and

that's the point, our current understanding of

which CYPs are involved in their interaction is

pretty evident these days, especially by the number

284

of drugs that we use as probes as well as the in

vitro data to support that.

We also know the relevant plasma

concentrations at which we see a clinically

significant interaction event. And then, also,

now, we're very much aware of the particular

pathways that mediate these events. And notice

that for the most part, these center around three

receptors, namely, CAR, PXR and the AHR. And I'll

go into much greater detail on those in a second.

Another way to look at this is if you

actually look at the inducible P450 enzymes in

human liver, with the exception of CYP1A, which is

predominantly induced by aromatic hydrocarbons,

some dietary components and cigarette smoking and

with the exception of 2E1, which is basically

induced by solvents and drugs like isoniazid but

mostly involves a mechanism of stabilizing protein

and RNA, for the most part, the rest of these often

are induced by compounds represented by the

anticonvulsants, antibiotic rifampin, et cetera,

suggesting that there is some overlap or

285

commonality in their mechanism of regulation.

It's now fairly apparent that especially

for the regulation of the human hepatic enzymes,

that there's three major receptors that are

involved: predominantly the aral hydrocarbon

receptor, the AH receptor; constitutive androstane

receptor or CAR; and the pregnane X receptor, PXR.

And there's three main points that I want

to basically draw from this particular slide:

number one, each of these receptors contains a

ligand or drug binding domain which determines,

basically, which drugs are going to activate it,

and also, they contain a DNA-binding domain, which

determines which DNA sequences or response elements

that they're going to bind to upon activation by

drugs.

Now, the other point I want to bring out

is that these all form heterodymers with other

proteins, and for the most part, the AH receptor is

distinct, in the sense that it partners with a

protein called the aral hydrocarbon receptor

nuclear translocase protein. I didn't name it. It

286

was given that name a long time ago. The acronym

AHRNTP is given to that.

On the other hand, CAR and PXR both

heterodimarize with another receptor called RXR,

but basically, it's gratuitous in its function

here. It's predominantly driven by--it's the

partner CAR and PXR. Now, the other point I want

to make is that upon activation, each of these

nuclear receptors induces a number of genes, not

just a single subfamily or, you know, a limited

class of genes, but, for example, upon activation

of PXR, you're upregulating a number of phase one

enzymes; also, transporters as well as phase two

enzymes as well as others, including the

carboxylesterases, by the way.

So bear in mind that also, CAR and PXR

share a number of these target genes in common.

So, for example, 2B is upregulated by both CAR and

PXR; 3A4 and the 2Cs, beginning to suggest that

there are some common regulatory mechanisms of

these genes by these nuclear receptors. And we

actually understand now enough about the particular

287

promoter region sequences and the response elements

that are in the promoters of these genes to explain

mechanistically now that they can be and ought to

be coregulated by activators of these receptors.

So our first evidence for coregulation

that we observed in my lab, and we've been looking

at this for over a decade, I would hate to admit

now; basically, our first evidence was a study that

we set out to do to explore the effect of 14

different compounds that were known to induce 3A4

to various extents, either in vivo or in vitro, and

our intention was to relate that to their PXR

activation profiles.

Now, interestingly, when we extended those

studies to include 2B6 activity, we basically found

something very interesting, which is summarized in

these tables over here. So if you basically look

at the most potent or the strongest inducers of 3A,

you'll notice that clotrimazole, rifampin and

ritonavir are also very potent inducers of 2B6 in

this particular case.

Notice also in the 2B6 column that there

288

is a couple of others, including phenotone and

phenobarbital that are strong 2B6 inducers, but

they're either moderate or weak inducers of 3A;

however, upon more extensive evaluation, even these

compounds are known to induce 3A, here again

showing some common regulatory mechanisms.

Now, if we extend these studies to include

CYP2C9, we also find very similar profiles; for

example, potent inducers of 3A and 2B6 also induce

2C9. Now, this is represented nicely in this

particular slide, where we looked at the

coregulation of CYP2C9 and 3A4 by avasomid, which

we discovered to be a very potent PXR activator.

It's also been shown to interact clinically with

warfarin and midazolam and digoxin.

You can see here in two separate donors,

if you look at 2C9 versus 3A4 in hepatocytes from

one particular donor how the response concentration

curves basically are almost superimposeable. Also,

in a second donor, the same situation, suggesting

here common regulatory mechanisms via PXR in this

particular case.

289

Now, if we extend these studies to include

additional inducers of 2B6 and 3A, but then, look

at the induction now of multiple 2Cs, including

2C8, 2C9 as well as 2C19, in this particular case,

we're looking at RNA, not activity, but it still

exhibits the point that I want to make that

basically, all these compounds that are inducers of

2B6 and 3A via activation of CAR and PXR also

upregulate the three 2C enzymes.

The other point I want to make is that the

most efficacious inducers are actually

transactivators of these 2C9 genes have a tendency

to be rifampin and/or phenobarbital in all three

cases.

And finally, the other point that I want

to make, because it's going to play a role in terms

of why we're proposing looking at a limited number

of endpoints is if you actually look at the

induction of 2C9, it's basically between two and

threefold, with even the most potent inducers,

positive controls, if you will, suggesting that

it's actually not a very sensitive target gene if

290

you're trying to actually elucidate the induction

potential of a particular drug, and since we've now

discovered that there's a lot more coregulation

between these genes, we propose a more

mechanism-driven type screening strategy than what

has typically been proposed in the past.

So in essence, what this boils down to is

we think we're at a point now where we understand

the regulatory mechanisms of the relevant human

P450 genes to where we can now do a more

mechanism-driven screening strategy with a goal to

screen efficacious activators of these particular

dominant nuclear receptors and these clinically

relevant induction events where we propose

screening protocol using a sensitive endpoint for

each nuclear receptor being the goal with the

premise that potent activators of each of these

individual nuclear receptors will induce a number

of target genes but differentially.

So for example, potent PXR activators will

induce 3A, 2B, the 2Cs, even some of the phase 2

enzymes like 1A1, UGT1A1, transporters like MDR1,

291

but 3A4 is the most sensitive. Likewise, potent

CAR activators will induce a number of these same

genes that overlap with PXR, but 2B6 is the most

sensitive. And then, finally, potent AH receptor

agonists will induce 1A2, phase two enzymes such as

UGT1A1, GSTs. But 1A2 is the most sensitive in

terms of screening for that.

So, finally, an example protocol that we

would advocate, and we currently use, is to treat

human hepatocytes. That's a given with our

protocol. Treat with a new drug at three to four

relevant concentrations, especially where

clinically relevant concentrations are known; treat

for one to three days; include positive controls,

which is very important in terms of making

appropriate comparisons, so, for example, the most

robust 1A, 2B and 3A inducers ought to be used for

positive controls, in our opinion, where some sort

of maximum is obtained that's possible with the

particular preparations of cells.

Now, one has the ability to measure RNA,

certainly, protein as well as enzyme activity. We

292

would advocate that the enzyme activity is probably

the best representation of the induction response.

Protein content is semiquantitative at best, and

the relationship between RNA content and enzyme

activity is still yet to be completely

characterized, although I think we're nearly there.

And finally, the last point here is that a

major CYP target gene for each nuclear receptor

ought to be the focus of these initial screens; so,

for example, looking at CYP1A2 as an endpoint for

AH receptor agonists, 3A4 for PXR, and possibly 2B6

for CAR, although all 2B6 inducers and CAR

activators that we've come across thusfar also turn

out to be inducers of 3A4.

So finally, one other point that I want to

make in terms of looking beyond enzyme activities,

in this particular case, where we're looking at a

mechanism-based inhibitor such as ritonavir, if you

limit yourself to looking at enzyme activities,

which is the case in this particular study, where

we evaluated seven different inducers of 3A, you

can see that if you only evaluate things on the

293

enzyme activity, which is normalized enzyme

activity to the negative control, here, you can see

that ritonavir actually knocked out the activity

significantly in these microsomal assays that we

did. And as we all know, ritonavir is one of the

most potent mechanism-based inhibitors that we've

come across.

However, if you actually were to look at

its effects at the ability to upregulate 3A4 gene

expression at the RNA level, you'd find that

ritonavir is every bit as efficacious as positive

control rifampin at that level, suggesting that

it's actually a very potent PXR activator and

inducer.

So finally, some of the other important

factors to consider in terms of study design: the

interdonor differences in the control and basal

activity between preparations of hepatocytes can

often be a caveat. We suggest that that's why it's

important to compare it to a positive control

rather than fold over a negative control. We also

believe that it's possible that depending on how

294

high the basal activity is, it may exclude some

preparations of hepatocytes from maybe being

appropriate for induction studies.

Also, the relevant concentration range of

your drug is important, focused on plasma and

tissue concentrations; appropriate choice and

concentration of a positive control is an important

consideration; certainly, the major species

differences have to be acknowledged in terms of

nuclear receptor activation as well as induction of

specific P450s, so for example, it still surprises

me that some of the studies that I come across

where dexamethasone is still used as a positive

control in human hepatocytes, it's a very potent

inducer, as is PCN in for rodents, but there's

about an order of magnitude difference between

dexamethasone's ability to induce the 3A enzymes in

human hepatocytes compared to, let's say, a

positive control like rifampin.

Also, the expression of the data in the

relevant endpoints is very critical, and that's

been also an issue that's been addressed and

295

relating that to a positive control. Exposure time

is important, especially for the particular

subforms that you might be evaluating, shorter for

CYP1A, for example; longer for 3A; and then,

finally, one must bear in mind that solvent effects

on P450 expression and activity are observed.

DMSO, for example, is an activator of PXR itself at

sufficiently high concentrations, and also some of

the alcohols are known to even inhibit some of the

P450s.

And finally, just in summary of the key

points: our mechanistic understanding of enzyme

induction in human liver has increased markedly in

the past decade. Most inducible human P450s, UGTs

and transporters involved in DDIs are regulated by

a few receptors, namely PXR, AH receptor and CAR.

Screening for potential inducers during drug

development, in my opinion, can be achieved using a

single, selective and sensitive target gene for

each of these nuclear receptors through following a

3A4, 1A2 and/or 2B6, and activity data from in

vitro induction studies for a new drug should be

296

normalized to a negative control, compared to an

appropriate positive control at appropriate

concentration, considered significant when they are

greater than or equal to 40 percent of the positive

control, and that's actually a question that I'd be

interested to hear others' opinion on that and also

complemented with protein or RNA data if time

dependent inhibition is involved.

So, with that, I'll be happy to answer any