1

DEPARTMENT OF HEALTH AND HUMAN SERVICES

FOOD AND DRUG ADMINISTRATION

CENTER FOR DRUG EVALUATION AND RESEARCH

ADVISORY COMMITTEE FOR PHARMACEUTICAL SCIENCE

Tuesday, October 19, 2004

8:30 a.m.

CDER Advisory Committee Conference Room

5630 Fishers Lane

Rockville, Maryland

PARTICIPANTS

Arthur H. Kibbe, Ph.D., Chair

Hilda F. Scharen, M.S., Executive Secretary

MEMBERS

Patrick P. DeLuca, Ph.D.

Paul H. Fackler, Ph.D.

Meryl H. Karol, Ph.D.

Melvin V. Koch, Ph.D.

Michael S. Korczynski, Ph.D.

Marvin C. Meyer, Ph.D.

Gerald P. Migliaccio, Ph.D. (Industry

Representative)

Kenneth R. Morris, Ph.D.

Cynthia R.D. Selassie, Ph.D.

Nozer Singpurwalla, Ph.D.

Marc Swadener, Ed.D. (Consumer Representative)

Jurgen Venitz, M.D., Ph.D.

SPECIAL GOVERNMENT EMPLOYEES SPEAKERS

Judy Boehlert, Ph.D.

Gordon Amidon, Ph.D., M.A.

FDA Staff

Gary Buehler, R.Ph.

Lucinda Buhse, Ph.D.

Jon Clark, M.S.

Jerry Collins, Ph.D.

Joseph Contrera, Ph.D.

Ajaz Hussain, Ph.D.

Monsoor Khan, R.Ph., Ph.D.

Steven Kozlowski, M.D.

Vincent Lee, Ph.D.

Qian Li, Ph.D.

Robert Lionberger, Ph.D.

Robert O'Neill, Ph.D.

Amy Rosenberg, M.D.

John Simmons, Ph.D.

Keith Webber, Ph.D.

Helen Winkle

Lawrence Yu, Ph.D.

C O N T E N T S

PAGE

Call to Order

Arthur Kibbe, Ph.D. 5

Conflict of Interest Statement

Hilda Scharen 5

Introduction to Meeting

Helen Winkle 8

Subcommittee Reports - Manufacturing Subcommittee

Judy Boehlert, Ph.D. 26

Parametric Tolerance Interval Test for Dose

Content Uniformity 53

Critical Path Initiative

Topic Introduction and OPS Perspective

Ajaz Hussain, Ph.D., 64

Research Opportunities and Strategic Direction

Keith Webber, Ph.D. 105

Informatics and Computational Safety

Analysis Staff

Joseph Contrera, Ph.D. 117

Office of New Drug Chemistry

John Simmons, Ph.D. 165

Open Public Hearing

Saul Shiffman, Ph.D. 192

Critical Path Initiative--Continued

Office of Generic Drugs

Lawrence Yu, Ph.D. 204

Office of Biotechnology Products--Current

Research and Future Plans

Amy Rosenberg, M.D. 248

Steven Kozlowski, M.D. 282

C O N T E N T S (Continued)

PAGE

Office of Testing and Research--Current

Research and Future Plans

Jerry Collins, Ph.D. 316

Lucinda Buhse, Ph.D. 338

Mansoor Khan, R.Ph., Ph.D. 362

Wrap-up and Integration

Jerry Collins, Ph.D. 410

Challenges and Implications

Vincent Lee, Ph.D. 419

Committee Discussion and Recommendations 428

P R O C E E D I N G S

Call to Order

CHAIRMAN KIBBE: Ladies and

gentlemen--welcome. I want to take a little page

from the coach at the New York Times, who says that

a meeting that starts that eight o'clock actually

starts at five minutes before. And to get us

rolling in about 30 seconds, ahead of time.

Do we know--

[Comment off mike.]

--he'll be here tomorrow. All right.

So--Dr. Amidon, my co-pilot here, will be here

tomorrow.

I'd like to call you all to order for my

last go-round as Chairman of this August body. And

the first order of business, of course, is to read

about all of our conflicts.

Conflict of Interest Statement

MS. SCHAREN: Good morning.

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

committee 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 interests as they may apply to

the general topics at hand.

To determine if any conflict of interest

existed, the Agency has reviewed the agenda and all

relevant financial interests reported by the

meeting participants. The Food and Drug

Administration has granted general matters waivers

to the special government employees participating

in the meeting who require 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 may apply

to each member, consultant and guest speaker. FDA

acknowledges that there may be potential conflicts

of interest, but because of the general nature of

the discussions before the committee, these

potential conflicts are mitigated.

With respect to FDA's invited industry

representative, we would like to disclosed that

Paul Fackler and Mr. Gerald Migliaccio are

participating in this meeting as a non-voting

industry representative, acting on behalf of

regulated industry.

Dr. Fackler's and Mr. Migliaccio's role on

this committee is to represent industry interest in

general, and not any one particular company. Dr.

Fackler is employed by Teva Pharmaceuticals,

U.S.A., and Mr. Migliaccio is employed by Pfizer,

Incorporated.

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 participants' involvement and their

exclusion will be noted for the record.

With respect to all other participants we

ask, in the interest of fairness, that they address

any current or previous financial involvement with

any firm whose products they may wish to comment

upon.

Thank you.

CHAIRMAN KIBBE: Thank you.

And now we'll hear from the Director of

the Office of Pharmaceutical Sciences, Ms. Helen

Winkler.

Introduction to Meeting

MS. WINKLE: Good morning, everyone.

All right, I want to welcome everybody

this morning to the Advisory Committee for

Pharmaceutical Science. This is, I think, a very

important meeting, and I"m really looking forward

to the discussion. But before we get there, I want

to welcome all of the members. We have one new

prospective member, Carol Gloff--Dr. Gloff--has

joined us. And we have two other prospective

members who we're having a little complication with

in getting on board. So we're working on that.

We also will have a number of SGE's here

today; Dr. Boehlert, Dr. Amidon and several others

who are going to participate with us in a number of

things. So I want to welcome everybody.

I also want to thank Dr. Kibbe. This is

his last time as Chair. It will break all of our

hearts to see Dr. Kibbe go out of this position.

He has been very, very enthusiastic as the Chair of

this committee, and I think all of us have enjoyed

working with him. But he's not to go very far.

We've already told him that we anticipate him

coming back to a number of meetings and helping us

with some of the discussion in the future. So we

really want to, again, thank him for all he's done.

Dr. Cooney--Charles Cooney--has agreed to

be the chair of the committee for the next two

years. Unfortunately, Dr. Cooney couldn't be

here--after he accepted, he couldn't be here today.

But he will be here at the next meeting. So--he's

been very gracious to accept this position. He and

I have talked at length about some of the issues we

want to cover on the Advisory Committee, and he's

very enthusiastic about moving ahead for the future

of the committee.

The agenda for the meeting today: there's

a number of things we want to take up. I'm going

to talk a little bit about next year--2005 being, I

guess, this fiscal year--and some of the things

that we plan to take up with the Advisory

Committee, where we're going in OPS, just to give

the committee a little feel about some of the

things that we're looking at.

I also want to give a quick--and I mean a

quick--update of the cGMP Initiative for the 21

Century. We're also going to have an update on a

number of the subcommittee and working groups. Dr.

Boehlert is going to talk about the Manufacturing

Subcommittee meeting that we had several months

back. It was a very, very--we accomplished a lot,

I think. It was a very good meeting. And Judy can

fill us in on some of the highlights of that

meeting. Also Bob O'Neill is going to talk about

the Working Group with IPAC RS, and some of the

accomplishments--or the focus that we've had in

that Working Group.

We're also going to talk about the

Critical Path Initiative. And I think this is a

really important discussion that we can have with

the committee today. Critical Path is, of course,

one of the main initiatives in the agency now, and

what we would like to talk about with the committee

is give you some idea of our thoughts, as far as

Critical Path; some of the things that we're doing

in the Critical Path Initiative, in the office of

Pharmaceutical Science in the various product

areas, and get some input from you as to what

direction we need to go; if there's other things we

need to be thinking about; and if there's other

types of topics that we need to be taking up, we'd

like to do that.

Dr. Woodcock talked about the Critical

Path Initiative when she introduced it, saying that

FDA was really in the best position to identify

those areas, or those gaps, in drug development,

and to work with others--collaborate--on how we

could get the data necessary to fill those gaps.

So this is really what we're looking for

doing under the Critical Path Initiative. And we

need to be certain that we are identifying the gaps

correctly, and that we are able to do the types of

research that needs to be done to fill those gaps.

Of course we can't do everything, so I think some

of what we want to talk about and think about, too,

is how we can prioritize some of that research.

Tomorrow, we're going to talk about

manufacturing, and moving toward the desired state.

As I said, we had a very productive meeting of the

Manufacturing Subcommittee. A number of things

were identified at that meeting that we need to

discuss further; that we needed to look at and

determine how we're going to do it. A number of

questions that we need to answer--and we're looking

at possibly having a subgroup to do some of that--a

fact-finding group. So Judy will talk to that.

But there are a number of things, too,

that we want to talk about with the committee

today; a number of--the gaps that we recognize that

we have in OPS and the agency, in moving toward

that desired state.

So several of us are going to talk about

those gaps. We're going to talk about the

organizational gaps, the science gaps, and the

policy gaps--all of which are important if we in

the agency are going to be prepared as the

manufacturers and others move toward that desired

state.

So I think that will be a really

interesting issue, and I think there are a number

of things that the committee can help us with in

identifying how best to address these answers and

to address the gaps.

We also have a number of bio-equivalence

issues that we want to discuss. We want to

continue the conversation from the last Advisory

Committee we had on bio-equivalence. And Dr. Yu

and some of his staff are going to talk about some

recommendations from that. And we're also going to

bring up a new topic on gastroenterology drugs.

So--moving on to OPS in 2005. I think

2004, we had an extremely busy year, mainly focused

on the GMP Initiative, and all of the aspects of

that initiative--especially the areas concerning

manufacturing science and how wee were going to

really address those issues and concerns, and how

we were going to incorporate those into the

regulatory framework.

As we move into 2005, I think we still

have a lot of issues that we have to handle under

Pharmaceutical Quality Initiative. We've already

said that that's going to be some of what we take

up with the Advisory Committee today. But we

really need to pursue those next steps. In doing

that, though, we also need to be looking at

continuing to streamline the review processes. We

continue to get more and more products in for

review, and there's got to be some way to offset

that increasing workload. And streamlining the

review processes seems to be--we're moving in that

direction, and it seems to be the answer to

handling some of the enormous workloads that we

have.

Also, we need to incorporate best

practices. We've added the Office of Biotech

Products in the last year. They joined us in

October of 2003, and they have a lot of practices

in their review that I think can be very helpful as

we move forward in looking at ways to improve--both

in out office of New Drug Chemistry, and our Office

of Generic Drugs.

So we're going to be looking at

incorporating best practices across the entire

organization.

Supporting the Critical Path

Initiative--I've already brought this up. It's a

very important part of where we're going. I think

much of our research is going to be done there, and

I think we're talking about much more than

laboratory research. I think there's a number of

activities that we hope to take on in 2005 where

we're looking at improving on how we do the

regulation, and in actually working through the

Critical Path Initiative to get some of this done.

So we'll talk more about that as we get into

Critical Path and some of those projects that we're

looking at doing.

We're looking at further integrating the

whole Office of Biotech products. There are still

some things that need to be accomplished there. I

think there are still a number of questions that

the Advisory Committee can be very helpful to in

answering. So you will hear more about this in the

next fiscal year.

And, last of all, I think there still

continues to be a number of regulatory on follow-on

proteins, as well as a number of general scientific

issues that we'll want to discuss with the

committee.

So I think we have a lot on our plate

during the year, and I look forward to working

closely with the Advisory Committee in the next

fiscal year to help us identify some of the--other

things that we need to look at, as well as help us

with the issues that we already have identified

ourselves.

Okay. As I said, I'm going to talk real

quickly about the CGMP Initiative for the 21st

Century. I think most of you all have probably

read the background material, which included the

report. We've actually come to the end of the

first two years of the initiative. And I"d like to

emphasize: I don't think that's the end of the

initiative. I think it's just the beginning. I

think that the initiative helped us identify a

number of things that we need to be looking at in

review, that we need to be looking at in

inspection. We still have a lot of changes to

make. I think we've made a lot of progress--and

I'll talk a little bit about some of that progress.

But I think we've got a lot more that we have to

focus on.

So that was only, in my mind, the first

step.

But I thought it would be helpful just to

step back real quickly and look at what the goals

of the initiative were. Because I think you can't

really appreciate the accomplishments without

really understanding what the goals were.

So there were basically six major goals.

The first one was to incorporate the most

up-to-date concepts of risk management and quality

systems approaches; secondly, was to encourage the

latest scientific advances in pharmaceutical

manufacturing and technology, ensure submission

review program and the inspection program operating

in a coordinated in synergistic manner; apply

regulation and manufacturing standards

consistently; encourage innovation in the

pharmaceutical manufacturing sector; and use FDA

resources most effectively and efficiently to

address the most significant health risks.

And you can see, when you look back at

these initiatives, the role OPS has had to play in

all of these goals. I think they're very

important, not only to the agency, but important to

us at OPS, and important to the industry and others

involved in the manufacturing of pharmaceutical

goods.

So, quickly, through the

accomplishments--again, you can read the report.

You'll get a lot more out of the report. But I

just want to emphasize that there was an awful lot

done in the last two years; a lot that will affect

how we move forward in the future, in the 21st

century. So I wanted to highlight those.

The first thing was Part 11. We did a

last in the last two years to clarify the scope and

application of Part 11. There were quite a few

questions; quite a bit of complication in

implementing Part 11. And I think we've moved

forward in trying to eliminate some of that

complexity and complication. We issued two

guidances during the two-year period that have

helped in that clarification.

Technical Dispute Resolution Process--this

was also a very important part of the initiative.

And it really has had a very positive effect, I

think, on the industry, and a positive effect on

how the field has dealt with inspections and has

increased the time and effort that the inspectors

are putting into the inspections, and the time and

effort that they're spending with industry when

they go in and do these inspections. And it has

really been the basis of much discussion in the

inspection process. And the outcome--we have not

had any technical disputes. We have a very good

process--as I said, the process has sort of set the

framework for opening up the discussion. And so I

think that it has had a really positive effect.

I'm actually a co-chair of that group. I kept

waiting for disputes. I thought we were just going

to have tons of them. We have a pilot program, and

I thought in the 12 months of the pilot we'd be

able to figure out how best to run the program.

But not having any disputes, we haven't learned a

whole lot of lessons.

But, again, it's had its very positive

effects. So I think that it has really been useful

under the initiative.

The GMP warning letters--this was an issue

that was handled very early on. And we

accomplished the goals that we wanted under this

particular working group of the initiative; and

that's that warning letters now are reviewed by the

Center to ensure--in the Center before they go out

to the companies--to ensure that they have adequate

scientific input. Many of the warning letters that

went out in the past were not reviewed to make sure

that the issues were scientifically sound. So that

has changed now. And I think that's had a very

positive effect.

International collaboration--I won't go

into that, but we have spent a lot of effort in

ICH, and Q8, Q9, and hope to do a lot in Q10. And

also one of the things we are planning on doing is

getting more involved with PICS, which is looking

at inspections on a worldwide basis.

Facilitating innovation--including doing

standards and policies--we were very fortunate to

put out a number of different guidances under this

part of the initiative; the aseptic processing

guidance--which industry is very familiar with.

They've been waiting for this guidance for a long

time. And I think it addresses many of the

questions that have been out there in industry's

mind. So I think it's a very, very positive part

of the initiative that we were able to accomplish.

The next guidance that was put out--I

think many of the people--in fact, everyone on the

Advisory Committee is very familiar with this

guidance, because we did have a subcommittee on the

PAT--the Process Analytical Technologies--under the

subcommittee, and we were able to put, under Dr.

Hussain and others in the group, we were able to

put out a guidance to industry which has had an

extreme effect, I think, on how industry and others

are looking at manufacturing in the future. I

think it's been probably one of the best parts of

the whole initiative. It really has promoted the

two--the team approach to doing work; working on

standards. We've worked with ASTM under E55. And

I think, all in all, this has been an extremely

successful initiative under the GMP initiative.

The last guidance that we've had, that was

comparability protocol. That guidance is still in

limbo. We're trying to make sure that before we

issue the guidance that we're not increasing the

regulatory burden--which I think many of us felt

when we read the original draft guidance. So we're

busily working on that to make sure that what we

come out of is very beneficial to industry and to

FDA, and that we don't put any additional resource

requirements on either part of the regulatory

system.

Manufacturing science--the desired state

under !8 of ICH has become a very important aspect

of where we're driving to. And, of course, we're

going to talk to that tomorrow morning; continuous

improvement and reduction of variability have been

an important part of manufacturing science, and

areas that we need to explore more in the future,

and assure that we can accomplish that, especially

being able to open up in the agency and allow more

continuous improvement for manufacturers.

Product specialists--this includes

enhancing the interactions between the field and

the review. We're looking at a team approach, in

having our reviewers all out on inspections. And

we're looking at best practices from both the PAT

team and Team Biologics. I think there's a lot of

best practices there that we can incorporate in out

thinking in the future on how we handle review and

inspection.

Integration of approval and

inspection--this is more of that. We have

developed the pharmaceutical inspectorate, and

we're looking also at changes in pre-market

approval program.

Quality management systems--there's a

number of things that we've worked on here. They

take a number of directions. We've developed a

standard quality systems framework; a quality

systems guidance. We've worked on GMP

harmonization, analysis process validation, and

good guidance practices--none of which are going to

go into in detail, but I think all very beneficial

to helping us in the future in the 21

st century.

Risk management--risk management, I had

thought--we did introduce a site-selection model

for inspection under this part of the initiative.

I believe there's a number of other things that we,

especially in Review, need to focus on as far as

risk management, and have a much better idea of

what the risk of products are, and how we're going

to mitigate those risks. And I think this is

something that we will bring up in the future at

the committee.

Team Biologics was to look at a number of

initiatives that were already underway, and adopt a

quality systems approach.

And last of all was the evaluation of the

initiative, which hasn't been completed yet, but

it's a very important part of what we've done.

So that, in a nutshell--I mean, that's a

lot of effort, obviously, that we've done. And if

you, again, will read the report I think you'll get

a much better feel. But I felt like, since we've

talked about it so much during the last few years,

that it was very important to sort of wrap up what

has happened in the last two years with this

committee.

So that's all I have to talk today. I'm

going to give it back to Art, and I look forward to

very lively discussion on a number of these issues,

and look forward to working with you for the next

two days.

Thank you.

CHAIRMAN KIBBE: Thank you, Helen.

We now have a report from the chair of one

of the subcommittees--the Manufacturing

Subcommittee.

Judy?

Subcommittee Reports

Manufacturing Subcommittee

DR. BOEHLERT: Good morning, ladies and

gentlemen. Before I just get started here--I tried

pressing down, and--aha. I need an SOP for how to

operate the slides.

[Slide.]

It's a pleasure for me to be here this

morning to update you on the Manufacturing

Subcommittee. We met in July. And I think you'll

find that a lot of the topics we discussed tie in

very well with what Helen was talking about this

morning, and also with some of the topics that are

going to be on your agenda.

[Slide.]

We met for two days in July. Just a brief

overview of the topics that we discussed: quality

by design--we've heard that this morning;

introduction to Bayesian approaches--and we'll talk

just a little bit about that; research and training

needs--the industrialization dimension of the

Critical Path Initiative--another topic we heard

about this morning; manufacturing science and

quality by design as a basis of risk-based CMC

review; and risk-based CMC review paradigm.

[Slide.]

On the 21

st: introduction to

pharmaceutical industry practices research study; a

pilot model for prioritizing selection of

manufacturing sites for GMP inspection; cGMPs for

the production of Phase I INDs; and applying

manufacturing science and knowledge, regulatory

horizons.

What I'm going to do is just go over,

briefly, some of the topics that were discussed,

and also the comments that were made by committee

members.

[Slide.]

Quality by design: topic updates. This

addressed three guidances that should be coming out

of ICH. The first of ICH Q8, which is a guidance

on pharmaceutical development section of the Common

Technical Document. It's going to describe

baseline expectations and optional information;

requires FDA and industry to think differently.

Industry needs to be more forthcoming with

information in their submissions, and FDA needs to

look at the review process; focuses on process

understanding and predictive ability. And if you

really understand your process, you'll gain

regulatory flexibility. It's a framework for

continuous improvement. And Step 2 is expected in

November this year. That means it will be out for

public review and comment.

[Slide.]

ICH Q9 is quality risk management. It

looks at risk identification--should link back to

the potential risk to the patient, because, after

all, that's what's important; risk assessment--what

can go wrong? What is the likelihood? What are

the consequences?

Risk control--options for mitigating,

reducing and controlling risks; risk

communication--between decision makers and other

shareholders. And this may also reach step two in

November of this year, although that was a bit

questionable.

[Slide.]

And then we're going to talk about quality

systems needed to recognize the potential of !8 and

Q9. And this is ICH Q10: monitor and evaluate

processes with feedback groups in a manner to

identify trends and demonstrate control or the need

for action; manage and rectify undesirable

occurrences; handle improvements; management,

implement and monitor change.

This is currently on hold, not because

it's not a good topic, but primarily because all

the resources that would address Q10 are tied up

with Q8 and Q9.

[Slide.]

We also talked about the ASTM E55

Committee. And Helen mentioned that this morning.

Their involved in the development of standards for

PAT. And the important things here are consensus

standards, with input from industry, academia and

regulators. There's an established process, with

an umbrella set of rules. And ASTM is recognized

worldwide.

They have three functional subcommittees

on management, implementation and practices and

terminology. But one of the concerns expressed by

the committees is are they going to duplicate other

initiatives. There area lot of people right now

working on PAT initiatives, and are they going to

duplicate some of that. So we need to make sure

that everybody gets on the same page.

[Slide.]

All right. Now, this topic I'm going to

be reluctant to say a whole lot about, but we had

an introduction to Bayesian approaches. Dr. Nozer

Singpurwalla was kind enough to give us an

introduction to the topic. So, Nozer, I apologize

if I mis-speak when I summarize--[laughs].

You know--so it's with fear and

trepidation--he's threatened us a quiz--

DR. SINGPURWALLA: You've already done it.

DR. BOEHLERT: Yes, I know. [Laughs.]

That's what I was afraid of. But I didn't think I

could leave it out, or you'd get after me then,

too.

Okay--Reliability for the Analysis of

Risk." Reliability--the quantification of

uncertainty. And I'm just going to say a few words

here: utility--costs and rewards that occur as a

consequence of any chosen decision. These are the

things that Nozer talked to us about--risk

analysis--process assessing reliabilities and

utilities, including an identification of

consequences. We talked about scales for measuring

uncertainty--for example, probability.

[Slide.]

Now this is a quote, so I have to be

careful here. "When the quantification of

uncertainty is solely based on probability and its

calculous, the inference is said to be Bayesian."

I am not a statistician, so I'm certainly not a

Bayesian statistician. And then there is

discussion of use of Bayesian approaches for ICH

Q8, Q9, Q10 and the use of prior information.

[Slide.]

Industrialization--dimension, the Critical

Path Initiative. We heard about that this morning.

We'll hear about it in the next two days:

examining innovational stagnation. Everybody needs

to take a look at what we've been doing in the past

and get things moving forward in a new environment,

with new technologies.

Critical path--has been inadequate

attention in areas of new or more efficient

methodologies and development research.

Industrialization--goes from the physical

design of prototype up to commercial mass

production. And Education and research

infrastructure needs improvement. And this

education and research applies to industry; the

education also applies to the agency. We all need

to learn how to go forward in the new environment.

[Slide.]

FDA has a strong interest in computational

methodologies to support chemistry and

manufacturing control submissions. They're putting

together a chemometrics group. There's a new FDA

research program focusing on industrialization

dimension. And there's training needs. AS I

mentioned before, particularly with the

pharmaceutical inspectorate. That's started.

There is an inspectorate now of trained

investigators. There need to be more.

[Slide.]

Manufacturing science and quality by

design--it's a basis for risk-based CMC review.

Companies share product-process understanding with

regulators. And this is a new paradigm, if you

will, that companies will share more of the

information that they have available than they have

in the past.

Specifications should be based on a

mechanistic understanding of the process; there

should be continuous improvement; and real time

quality assurance. You shouldn't have to wait

until the end of the process to know that your

product is okay.

[Slide.]

Science perspective on

manufacturing--define current and the desired state

and the steps to go from here to there; define

terms--and this is going to be important going

forward--things like "manufacturing science,"

"manufacturing system," "manufacturing

capability"--what do they really mean?

Real case studies will help. This came up

time and again in the committee discussions. It's

nice to talk about all these theoretical concepts,

but give me a real case study that I can look at

and see what it really means.

Testing is mostly non-value added.

Quality by design is the desired state.

[Slide.]

Risk-based CMC review--from the Office of

New Drugs--should provide regulatory relief by

incorporating science-based risk assessment; more

product or process knowledge shared by the

industry--and I've said this several times; more

efficient science-based inspections; focus

resources on critical issues; and specifications

are based on a risk-based assessment.

[Slide.]

Quality assessment rather than a chemistry

review--in the past it's been a strict chemistry

review: go down the list and check off the boxes;

conducted by inter--and I see some smiles on the

parts of agency folks--conducted by

interdisciplinary scientists--so it could be a team

approach. It should be a risk-based assessment;

focus on critical quality attributes and their

relevance to safety and efficacy. They have to

rely on the knowledge provided by applicants. If

industry doesn't submit the information, the agency

has nothing to make their decisions on. And the

comparability protocols are an important part of

this review.

[Slide.]

Role of process capability in setting

specifications will need to be addressed. Very

often, those kinds of process controls that you

have may have no clinical relevance. The knowledge

base at the time of submission can be an issue,

because very often you don't have that much

information at the time you submit. It's a

learning process as you go through early

marketability and commercial production.

Specifications should not be used as a

tool to control the manufacturing process. And we

might need to expand the Quality Overall Summary

going forward.

[Slide.]

AS I said before, the extent of product

knowledge is key. Risk-based decisions should be

based on supportive data. Voluntary--all of these

new initiatives are voluntary. And that needs to

be made very clear to the industry. These are not

requirements that everybody drop what they've been

doing in the past and start over with new

approaches--strictly voluntary.

Supplement need is based on the knowledge

of the risk of the change. And there should be a

clear rationale for the selection of

specifications.

[Slide.]

Identify critical parameters for product

manufacturing and stability; train FDA staff and

regulated industry--this came up a number of times.

We all need to learn what the other is doing;

should give us--industry--greater flexibility in

optimizing the process; should lessen the

supplement burden, which is good for industry and

good for the agency. And, once again, real

examples would be an asset.

[Slide.]

In the Office of Generic Drugs--generic

industry's focus is on producing a bioequivalent

product. Often patent issues--to design around.

They may not have the flexibility as the new drug

folks. Workload in OGD is a significant issue, and

committee members made a number of comments on this

when they heard how many submissions there are, and

how far behind they are. We were impressed by the

workload.

Provide advice to industry on improving

quality of DMFs--those are "drug master

files"--very important to the generic

industry--also to the new drugs, but to a lesser

extent.

[Slide.]

Desired state--include needed data in a

filing; process and product design; identify

critical attributes; identify process critical

control points. And this is the difference from

the past. Analyze data to produce meaningful

summaries and scientific rationales; and reviewers

assess the adequacy of the submission by asking the

right questions.

[Slide.]

Okay--some additional committee comments

that came out of the Day One discussion: ICH and

ASTM appear to be synergistic, but ICH needs to be

very aware of the ASTM focus. There was some

concern they might not be tied into what's going on

there; some concern that FDA, internally,

themselves, may be getting ahead of what's

happening on an international basis. So they may be

a little ahead of ICH Q8, Q9 and Q10. That's not

necessarily a bad thing, by the way.

Need concrete examples--that came up time

and time again; need to clearly demarcate "minimum"

and optional information--you know, just what do

you mean by "this is the minimum you need," and

just what is "optional" information? And

"optional" information comes in degrees. The more

you make the more you know. So you may not have as

much information at submission as you will down the

road after you've been in commercial product for a

number of months or years.

[Slide.]

Need to avoid implying there are two

different quality concepts. We don't want to say

that products made in the conventional way---the

way we've always done it--are different than

products that may be made according to some new

paradigms. Bring in new training programs--and

Helen mentioned we're talking about forming a

working group under the Manufacturing Subcommittee

to address some of the issues, particularly case

studies.

We need to find better terms than

"minimal" and "optional;" and focus on process

first, and then the tools that we're going to need.

[Slide.]

We had some reports on an FDA research

project that's being done by Georgetown University

and Washington University, and their goal is to

identify attributes that impact inspection

outcomes. They're compiling and linking FDA

databases. They're looking at variables for

product-process, facility, firm and FDA. Right now

they're collecting data. CDER is just about

completed, and CBER is ongoing--although by now it

may be even further down the road. This was July.

[Slide.]

Focus--are cGMP violations related to

managerial, organizational and technical practice?

And then interviewing manufacturers. They have an

internet-based questionnaire that went out in the

fall of 2003. They're looking at U.S. and European

manufacturers. And their data collection is near

completion.

[Slide.]

There's concern with just looking at

numbers of deviations or field alerts, particularly

when investigation may have shown little cause for

concern. You can put in a field alert and then

find out later on that--oh--you know, we figured it

out. It really wasn't a problem. So if you just

look at numbers, you get those as well as the ones

that are true issues.

Also it was pointed out that if you're a

company with a very detailed SOP you have a much

bigger chance for deviating from it than your

company with a really poor SOP that sort of allows

you to do anything, where you're hardly ever going

to deviate. But who's to say which one is better?

India and China are not include in the API

manufacturers. And we saw this as a downside to

that survey, because they are major manufacturers

of APIs.

[Slide.]

We talked then about risk ranking and

filtering, where risk ranking is a series of

decisions to start to rank within a class or across

classes. Tools may be customized for each

application. And filters may be used to reflect

resource limitations and/or program goals.

[Slide.]

There's a pilot risk-ranking model to

prioritize sites for GMP inspections, using ICH Q9

concepts to define risk; Site Risk Potential--a new

term for us--SRP--includes product, process and

facility components.

Look at probability and severity

components that make up harm; and look at other

risk-ranking models, for example those used by EPA

and USDA; and then using the CDER Recall database.

[Slide.]

Comments--from the committee--focusing on

volume at a site may be misleading because, in

fact, when you have a high volume your process may

be better controlled than if you have small volume.

We need to also consider the risk of the

loss of availability. If you're a single-source

drug for a life-threatening condition perhaps that

needs to come into the equation.

Look at "hard to fabricate" products, or

products with difficulty controlling uniformity.

Investigator consistency will be--and has been--an

issue, but with the pharmaceutical inspectorate

that should be better. And it was suggested by at

least one member that maybe they should look at

high personnel turnover in a plant, because that

might be indicative of problems--although it was

recognized that that might be hard information to

come by.

[Slide.]

Committee members wanted to know if the

sites are going to know how they are ranked. That

would be very useful information for management to

know about. Right now self-inspections are a

critical part of the quality system but the value

of these would be diminished if that information

were to become available to FDA. This has been a

longstanding concern of industry. You know, you

don't want to share your self-inspections because

then they lose their value to you.

[Slide.]

Next talked about GMP guidance that's

proposed for the production of Phase I drugs. CMC

review to ensure the identify, strength, quality

and purity of the investigational drugs as they

relate to safety. This draft guidance is in

process. It's a risk-based approach. No regular

inspection program, but these Phase I drugs are

looked at on a "for cause" basis.

I want to point out that it was noted

during that discussion that for Phase 2 and Phase

3, those drugs still fall under the GMP

regulations--21 C.F.R. 210 and 211.

[Slide.]

Also had an update on the PAT initiative.

As Helen indicated, that guidance was recently

finalized, in September. It should be expanded to

cover biotech products. And, of course, it

requires continued training of FDA staff.

[Slide.]

We also talked about--we had a full

agenda--comparability protocol. We had an update

on guidances, The goal is to provide regulatory

relief for post approval changes. It requires a

detailed plan describing a proposed change with

tests and studies to be performed, analytical

procedures to be used, and acceptance criteria to

demonstrate the lack of adverse effect on product.

Many comments have been received from the public.

That was FDA's comment on this. We did not see

those.

But the committee had comments, as well.

[Slide.]

Single use protocol has limited utility.

It's more utility if you're going to have

repetitive changes--if you're only going to do it

once it may not help. Specificity of the protocol

may limit repetitive use. Just how much

specificity is needed? And for a well-defined

protocol, an annual report should be sufficient.

That really will lessen the regulatory burden.

[Slide.]

Some general conclusions from our two

days--and we've heard the first one several

times--general principles are good, but case

studies are needed to facilitate understanding.

That came up time and time again. Case studies

should cover all industries; for example, dosage

form, API, pioneer and generic.

The committee expressed concern on what

appears to be understaffing in OGD.

[Slide.]

Failure Mode &Effect Analysis can be

linked with risk-based decision-making wherein the

results feed into decision trees; training and

education of both regulators and the industry in

the new approaches is going to be key; historical

inconsistency in regulator findings may limit the

utility of surveys. In the past, you know, not all

investigators have investigated in the same manner,

so it's difficult to compare results.

And that's the end of my presentation. I

thank you for your attention, and would be happy to

address any comments, now or later.

CHAIRMAN KIBBE: Are there any questions

for Judy?

DR. SINGPURWALLA: I have some comments,

but I probably would wait until all the

presentations are over, and then make comments.

Would that be acceptable?

CHAIRMAN KIBBE: Whichever way you want to

do it, as long as it's within one of the two tails

of the Bayesian distribution we're all right.

[Laughter.]

DR. SINGPURWALLA: You are confused, Mr.

Chairman. [Laughs.]

CHAIRMAN KIBBE: On a regular basis.

[Laughter.]

You had a question?

DR. MORRIS: Actually, just one comment to

add to what you'd said, Judy, about the Georgetown

study.

I think they had made sort of a plea that

the reason that they hadn't been able to go to the

Indian and Chinese manufacturers was strictly a

resource issue. It wasn't that they had ignored

that as an area of concern.

DR. BOEHLERT: Ken, thank you for that

clarification.

CHAIRMAN KIBBE: Go ahead.

DR. KOCH: I guess, looking around on the

schedule, I'm not sure if we're going to talk any

about training. You mentioned it in several

different ways: the continuation, the inclusion of

industry, etcetera. But will that come up as a

discussion topic at some point?

DR. HUSSAIN: Not in this meeting. I

think we will eventually bring that back at some

other meetings, though.

MS. WINKLE: Actually, when I talk about

some of the organizational gaps I'm going to bring

up training as part of that gap. So if you want to

comment then, it would be fine.

CHAIRMAN KIBBE: Anybody else?

DR. SINGPURWALLA: Well, maybe I'll speak

now. I just--we--this is a question more to

Ajaz--about case studies and specifics.

We've been through many sessions of the

Manufacturing Subcommittee meetings. Has there

been any concrete plan made to start seriously

undertaking some case studies? And, if so, would

you be kind enough to let me know?

DR. HUSSAIN: Yes. Dr. Boehlert's

presentation to this committee--she's the chair of

the subcommittee--and the decision was made to form

a working group under that. And after this meeting

we'll start populating that working group and

create a working group under that committee to

start addressing that.

In addition to that, I think we're also

looking at other parallel tracks to create case

studies. One such case study has just started to

take shape, with Ken Morris, and then Purdue is

working with our reviewers to actually develop a

case study also.

So we hope in the next several months we

will have examples and case studies to outline the

framework.

CHAIRMAN KIBBE: Anything else?

DR. SINGPURWALLA: Yeah. One other

matter. After the subcommittee meeting, some

minutes were released, and I had made some comments

about the minutes. I did not receive an update of

the minutes--update of the revision.

Has--is there any reason for that?

Because the normal protocol--the normal protocol is

you put out the minutes, people give comments on

the minutes. You either incorporate those

comments--and if you don't, you let us know why.

And then you issue a final document of the minutes.

And then the entire committee, or whoever it is,

says "Yes, we go along with these minutes." And

they should become a part of the record.

I was wondering if this was done, because

I did not have access to that.

CHAIRMAN KIBBE: I think the final draft,

or the final copy of the minutes is posted on the

web page--FDA website--so that after the draft goes

out to the members of the committee and the

corrections come back in, they update to reflect

the suggestions from each of the members, and then

they post it.

So if you wanted to check the website you

could see whether--you know, how well your

suggestions were incorporated in the final minutes.

DR. BOEHLERT: I would just add, also,

that I reviewed comments that were made to the

minutes before I made this presentation, and I

tried to make sure that they were all incorporated

in what I said today.

DR. SINGPURWALLA: I thought so.

DR. BOEHLERT: If they were not well

reflected in the minutes, they should have been

reflected in my comments today. So--

DR. SINGPURWALLA: I thought so, but I

wanted to see what the protocol was.

DR. BOEHLERT: Okay. Thank you. That's

fair.

CHAIRMAN KIBBE: Okay?

DR. WEBBER: One quick question.

CHAIRMAN KIBBE: Go ahead.

DR. WEBBER: That will be okay?

You mentioned the pharmaceutical

manufacture and research study, and I'm looking at

the dates there. It seemed like it was fall of

2003. And I just wanted to confirm whether or

not--that was during the period of transition of

products from CBER to CDER. Were our products in

OBP--the biotech products that transitioned

over--were they--are they completed now within

CDER? Or are they considered part of the CBER.

DR. BOEHLERT: Yes, I think Ajaz

DR. HUSSAIN: No, Keith, that's

not--that's an external study that's focusing on

all of manufacturing. So all products--CDER and

CBER--products are under. It doesn't matter

where--

DR. WEBBER: Where they were--just all

products--okay. Thank you.

CHAIRMAN KIBBE: Anybody else? Good, that

will keep us pretty well on schedule.

I have now a "Parametric Tolerance

Interval Test for Dose-content Uniformity"--Robert

O'Neill.

Parametric Tolerance Interval Test for

Dose-content Uniformity

DR. O'NEILL: Magic button. There we go.

Good morning. I'm Bob O'Neill. I came

before at the last meeting--I was asked to be the

chair of a working group that you all blessed, and

I'm here to give you an update on where we are on

this issue of addressing the specifications for the

delivered dose--uniformity of inhaled nasal drug

products.

[Slide.]

Just to refresh your memory, the folks on

the left-hand side are the FDA folks who are part

of this working group, and some are more active

than others--some of them, in blue, are part of a

sub-group that has been put together that is

working on more specific issues that I'll address

in a moment; and the folks on the right--Michael

Golden, in particular, who is a colleague on the

industry side, who is coordinating our efforts in

that area.

[Slide.]

The objective of this working group--as

you probably know--is to develop a mutually

acceptable standard delivered dose uniformity

specification--that's both the test and the

acceptance criteria--for the orally inhaled nasal

drug products, with a proposal to come back to you

all. And that's the time frame that I'm talking

about right now.

So there's been a lot of work going on in

the past few months, and that's what I just wanted

to bring you up on.

[Slide.]

There have been three full working group

meetings, where the folks on that previous

slide--and some others--have come together at FDA

for two, three hour sessions, and to go through

information that has been presented to--primarily

by the industry--to us to chew on. And we have

spent a lot of time internally talking to

ourselves, and coming up with some additional

issues and proposals, and we met the last time with

the working group, and FDA had a proposal that we

felt was moving in the direction of what everybody

wanted.

Subsequently, there's been a working group

that will now be chewing on what was presented to

the last joint meeting, and they're meeting

November 4 th. And

there's a lot of statistical

issues; there's data analysis issues. But I think

what we're all on the same page with regard to is

that the need to reassess the FDA--the past FDA

recommendations, and I think there's--as we

indicated the last time we briefed you--that the

parametric tolerance interval approach is an

improvement in a value-added type of testing

strategy, over and above the zero tolerance

interval strategy that's been used for awhile.

So the next steps are the following.

[Slide.]

This working group is meeting--the

sub-group is meeting in November, and we hope that

they will then come back to the full working group

by the end of the year, and we will evaluated the

iteration between the FDA modification to the

proposals that have been made by IPAC--and this has

a lot to do with the placement of the operating

characteristic curve for the acceptance criteria.

Essentially, there have been many operating

characteristic curves that have been shown to you,

some of which are more steep, some of which are

more shallow. But where the proposal is being

evaluated right now is: how good is it at getting

from an acceptance or rejection perspective, those

assays that essentially are off target mean. You

can look at the performance characteristic, or an

operating characteristic curve of a testing

strategy if you assume that it's 100 percent on

target. But the more you move away from 100 percent

on target, the more you look at how well does it

grab that, and how robust is it to allowing you to

be a little off 100 percent?

[Slide.]

And so we're in the stages of looking at

the statistical performance characteristics of

that, and we hope that the working group will

evaluate this proposal in more detail, and come

back to you in the spring of 2005, with a final

recommendation to discuss with you. So that's sort

of the game plan.

And Michael Golden is here. He's my

colleague on the working group from the industry

side, and we'd both be willing to take any

questions if you have them.

CHAIRMAN KIBBE: Questions?

Nozer?

DR. SINGPURWALLA: Well, I guess Jurgen's

hand went up before mine. So--

DR. VENITZ: Okay, let me go first.

DR. SINGPURWALLA: He may ask the same

question.

DR. VENITZ: Maybe.

In your draft proposal--or what you're

considering so far to be a draft proposal--

DR. O'NEILL: Yes.

DR. VENITZ: --are you considering the

intended use when you look at statistical

characteristics of your operating curve, for

example?

DR. O'NEILL: Well, certainly that has

been discussed, both from an emergency--a

one-time-only, a chronic use, a medical risk

involved--

DR. VENITZ: Right.

DR. O'NEILL: --so, certainly, Dr.

Chowdhury is involved, and others are involved, in

considering this issue. So--

DR. VENITZ: And I would encourage you to

do that because, obviously, in my mind, it is

different whether you're looking at inhaled

insulin--

DR. O'NEILL: Right.

DR. VENITZ: --and you're looking at the

performance of a drug product, versus a beta

agonist, for example.

DR. O'NEILL: Yes.

CHAIRMAN KIBBE: Go ahead.

DR. SINGPURWALLA: Dr. O'Neill, we had

this discussion when you made the first

presentation, so I'm going to back--

DR. O'NEILL: Right.

DR. SINGPURWALLA: --to the same point

again.

I agree with you that tolerance interval

approach is to be preferred to the zero tolerance,

or something to that effect.

DR. O'NEILL: Right.

DR. SINGPURWALLA: But in your description

of the next steps, you have talked about operating

characteristic curves, and performance

characteristic curves. Of course those are not

indicative of any Bayesian thinking towards this

particular area. And while you're in the process

of formulating your plans, I strongly encourage you

to incorporate that into your thinking. You may

not want to adopt towards the end, but at least it

should be evaluated.

And the second comment I'd like to make is

that--and I'm certainly not volunteering and, if

asked, I would refuse--the working group members

consists of individuals from the FDA and from the

pharmaceutical industry. It would be good to have

some neutral people on the working group--people

from industry or people from government agencies

that are not connected with the FDA, so that you

get some sense of balance. Otherwise, it seems to

be--you know, it seems to be a self-serving group.

So I would like to encourage you to expand

your membership.

DR. O'NEILL: Yeah.

DR. SINGPURWALLA: And I want to

emphasize: I'm not available.

DR. O'NEILL: Well--no, the last point--I

mean, this is hard work. The people who are doing

this work are spending a lot of time, and there's a

lot of evaluation--a lot of data evaluation going

on. We were presented with information from the

IPAC group that consisted of a huge database.

And one could look at, well, how much time

do you want to spend on evaluating a huge database?

I mean, it's an electronic database, and lots of

different--and where I'm going to on this is the

Bayesian argument. The Bayesian argument is very

much a sensible argument--or a sensible framework

when you can look at empirical data that allows you

to feel pretty comfortable about what your priors

are, and what the distribution of information is.

That is not always accessible to the agency. It

may be accessible to a sponsor.

So the strategy of being in-process and

out-of-process, and being in control, and what's

acceptable variability is very much--very much--a

Bayesian framework, and very much within the

context of how you may want to be looking at this,

in terms of looking at in-process validation, as

well as acceptance criteria.

The extent to which that carries over into

the type of testing we have to be very clear about.

And it's--at the point we're at right now, we're

essentially most interesting, or most concerned

about how far out can you push the acceptance curve

so that it has a proper balance between accepting

and rejecting--particularly when we don't have, or

no one can show us empirically, what the

distribution of off-target means are, for example.

How far away from 100 percent does the mean have to

be before you want to maybe ratchet in this

operating characteristic curve?

So, I certainly could see the value to

external folks' helping us out. The more the

better. And I believe that this is a

time-intensive effort. And just as, you know, you

would not like to volunteer, we would have to go

and find folks who could invest the amount of time

that is necessary, in the time frame that we're

talking about, so we can get where we want to be.

That's not to say that more brains are

not--and independent brains--are--but this is--I

would say we're pretty much trying to meet in the

middle of this whole thing with resources that

we've thrown out it that we feel are fair and

objective.

DR. SINGPURWALLA: Let me clarify.

I'm not volunteering because I'm making

the suggestion.

DR. O'NEILL: Yes. Yes.

DR. SINGPURWALLA: And that's the proper

thing to do.

What I would like to encourage you is to

involve at least two Bayesian's on your group--two,

because they need support--

[Laughter.]

--from the point of view of simply guiding

a framework, or guiding the concept, and things

like that, rather than get involved with the

nitty-gritty.

And the two individuals--or perhaps

more--need not come from two stratified groups.

They should come from somewhere else.

So I'm making two suggestions: one is to

have people with expertise in Bayesian statistics

involved, and to have people from outside these two

communities also involved--perhaps in a limited

way. This will give you a broader perspective and

will not subject you to criticism two years down

the line.

And that's the suggestion.

DR. O'NEILL: Okay.

CHAIRMAN KIBBE: Anybody else?

Ajaz, do you have something to say?

Reaching for your mike?

DR. HUSSAIN: I think the point I was

going to make was, I think, at this point in time

it's going to be difficult to add more people to

the working group. But the point is well taken

that I think you do need to bring that perspective.

And I'm hoping this Advisory Committee, and some

other format, could be sufficient to sort of bring

that framework for that--that perspective to bear

on the progress of this working group.

CHAIRMAN KIBBE: No one else?

Thank you Dr. O'Neill. Appreciated your

presentation.

Dr. Ajaz, perhaps you could begin our next

topic, and then we can take a break, because we're

running slightly ahead, and it will give us a

little flexibility as we move on.

And so we're going to talk about Critical

Path Initiative.

The Critical Path Initiative--Challenges

and Opportunities

Topic Introduction and OPS Perspective

DR. HUSSAIN: Yes, I think I'm pleased

that we have more time, because many of the

presentations here are very lengthy

presentations--[laughs]--including mine.

I'd like to sort of introduce the topic of

Critical Path Initiative--the challenges and

opportunities.

[Slide.]

The goals that we have for the fiscal year

2005--and the initiatives, and the strategic goals

at FDA level and the Department level are shown on

this slide. And the slide is from the "State of

CDER" address by Steve Galson and Doug

Throckmorton.

Today, our discussions will primarily

focus on the Critical Path, the cGMP initiative,

focused on risk management and innovation. And the

goal at the Department level is to increase science

enterprise research. But also, I think the follow

on biologics, follow-on proteins, I think is

interconnected to all of these discussions.

[Slide.]

My focus today is to introduce you to the

topic of Critical Path, and also outline a proposal

that we are contemplating at the OPS immediate

office level as an umbrella proposal for all the

discussions you'll hear today by scientists from

different parts of the Office of Pharmaceutical

Science.

But at the same time, some of the

discussions in here also impact, say,

counter-terrorism effort and other efforts that are

ongoing. And not all projects that we'll discuss

are Critical Path projects today.

[Slide.]

What is Critical Path? It's a serious

attempt to examine and improve the techniques and

methods used to evaluate the safety, efficacy and

quality of medical products as they move from

product selection and design to mass manufacture.

[Slide.]

In the continuum of drug discovery and

development, you really go from basic research to

prototype design or discovery, to preclinical

development, clinical development, to an FDA filing

and approval. You have a focused attempt, say, for

example, at the National Institutes of Health on

translational research. The Critical Path research

does overlap with some of the aspects of the NIH

translational research, but it covers predominantly

the drug development aspects of the entire

sequence.

In our White Paper, we identified some of

the challenges for Critical Path. The drug

development process--the "Critical Path" is

becoming a serious bottleneck to delivery of new

medical products.

[Slide.]

Our research and development spending has

been exponentially increasing. And as an index of

1993, you can see the exponential increase from

1993 to the current 10 years--increase in both

private and public spending on research.

[Slide.]

However, new product submissions have

remained flat--or, some would argue, are on the

decline.

[Slide.]

Why is FDA concerned? FDA's mission is

not only to protect but also to advance public

health by improving availability of safe and

effective new medical products.

[Slide.]

FDA has a unique role in addressing the

problem. FDA scientists are involved in reviewing

during product development--they see the successes,

failures and missed opportunities. FDA is not a

competitor, and can serve as a crucial convening

and coordinating role for consensus development

between industry, academia and government. FDA

sets standards that innovators must meet. New

knowledge and applied science tools needed not only

by the innovators must also be incorporated into

the agency's review process and policy.

[Slide.]

The challenge is how do we proceed? It

should be a science-driven and shared effort,

drawing on available data, need to target specific,

deliverable projects that will improve drug

development efficiency. It cannot just be an FDA

effort. We can identify problems and propose

solutions. Solutions themselves require efforts of

all stakeholders. We have issued a Federal

stakeholders, and we have received a number of

suggestions, and we are working through those

suggestions as we formulate our strategy for a

Critical Path research program.

[Slide.]

This is a significant initiative, and the

Department of Health and Human Services' Medical

Technologies Innovation Taskforce is providing

broad leadership. Dr. Lester Crawford is chair of

this Medical Technologies Innovation Taskforce, and

it includes CDC, CMS, NIH and FDA.

This taskforce is working on finding

additional funding to meet the needs of the

Critical Path program. It is meeting with external

stakeholders to identify opportunities, enlist

allies, and so forth.

[Slide.]

In summary, I think from a Critical Path

perspective, the present state of drug development

is not sustainable. We believe FDA must lead

efforts to question any assumptions that limit or

slow new product development: are these

assumptions justified? Are there more efficient

alternatives? If so, why are the alternatives not

being utilized?

[Slide.]

As we sort of focus on the discussions

today, I'll remind you that the Office of

Pharmaceutical Science is predominantly focused on

one aspect: Chemistry Manufacturing Control--or

the initialization dimension. But the Office of

Pharmaceutical Science also supports many other

aspects, from pharmacology, toxicology to clinical

pharmacology research and so forth. So, although

our review responsibilities predominantly are on

the quality side, our research programs are

interconnected to every aspect of the drug

development process.

So you will hear presentations coming from

all aspects--all three dimensions of the Critical

Path.

[Slide.]

The three dimensions are: assessment of

safety; how to predict if a potential product will

be harmful; assessing efficacy; how to determine if

a potential product will have medical benefit; and,

finally, industrialization--how to manufacture a

product at commercial scale with consistently high

quality.

[Slide.]

Our discussions, to a large degree, have

focused on the third dimension. And I think you

will see, today, many of the projects within OPS

that also impact the other two dimensions.

[Slide.]

In our White Paper, we defined the three

dimensions and the connections to the Critical Path

as follows: safety, medical utility, and

industrialization. An every aspect--every box that

is there has a need for improvement and research to

support that improvement.

Applied science is needed to better

evaluate and predict the three key dimensions on

the Critical Path development.

I just returned from Europe--spending a

week there last week--and with respect to the

industrialization dimension, I came back somewhat

depressed. The amazing work I saw coming out of

the University of Cambridge in the area of

industrialization of pharmaceuticals--the approach

to new technology, in terms of manufacturing, novel

drug delivery systems and manufacturing processes

itself, was astounding. I don't see any of that in

the U.S.

So my concern is, much of the R&D and

innovation is going to come from Europe and Japan,

probably. And unless we really improve our

infrastructure, we are going to be lagging behind

in a very significant way. And I think that

concern keeps growing on me, and I think I do want

to sort of emphasize that.

[Slide.]

Office of Pharmaceutical Science programs

and Critical Path Initiative--the discussion today

is to seek input from you and advice, on aligning

and prioritizing current OPS regulatory assessment

and research programs, with the goals and objects

of the Critical Path Initiative. Please note that

not all research programs and laboratory programs

are intended to focus on "Critical Path." There

are equally important other aspects--bio-terrorism

and so forth--which may not be considered as part

of the Critical Path Initiative, but they're

equally important. So all of our programs and

projects are not likely--or should not be part of

the Critical Path. There are aspects. So you have

to distinguish that.

We hope that you'll help us identify gaps

in our current program; identify opportunities for

addressing the needs identified by the Critical

Path Initiative.

[Slide.]

What I'd like to do today is--before I

introduce Keith Webber--he took the lead on putting

this program together--I'll share with you an OPS

immediate office project that Helen and I have been

developing. These are our initial thoughts of how

an umbrella project, within the OPS office, will

help to sort of bring all of this together.

So let me share some of our thoughts on a

Critical Path project that OPS--Helen and I are

sort of developing right now.

An immediate need in OPS is to ensure

appropriate support of general drugs--the growing

volume and complexity of applications. That's the

challenge. You saw the numbers increasing.

In the New Drug Chemistry, the new

paradigm for review assessment and efforts to

support innovation and continuous improvement goals

of the cGMP initiative--Office of New Drug

Chemistry has taken the lead to be the first office

to sort of implement all of this. So they have

significant need for support.

Biotechnology products--complete

integration into OPS, and the evolving concept of

"follow-on protein products"--although I have put

follow-on protein products under this, we don't

know exactly how the regulatory process will

evolve. It could be--let's say, a work in

progress.

And, clearly, alignment of research

programs in OPS to meet our goals and objectives.

[Slide.]

So what are our thought processes, from

our immediate office perspective? To develop a

common regulatory decision framework for addressing

scientific uncertainty in the context of complexity

of products and manufacturing processes in the

Offices of New Drug Chemistry, Biotechnology

Products, and General Drugs.

Regardless of the regulatory process,

regardless of regulatory submission strategies and

so forth, we believe we need a common regulatory

decision framework--a scientific framework--for

addressing the challenges.

[Slide.]

What are the motivations here?

Uncertainty--whether it's variability or knowledge

uncertainty--and complexity are two important

elements of risk-based regulatory decisions. A

common scientific framework, irrespective of the

regulatory path or process for these products, will

provide a basis for efficient and effective policy

development and regulatory assessment to ensure

timely availability of these products.

That's the overreaching OPS goal, is to

provide the common framework. Although the

submission strategies might be different, the

science should not be different.

[Slide.]

How are we trying to approach this

challenge? We know that there are no good methods

available for developing a standard approach for

addressing uncertainty. That means you need

different approaches for different assessment

situations. [Laughs.] All right, let me complete my

thoughts.

So what we are thinking about--a decision

framework for selecting an approach for addressing

uncertainty over the life cycle of products is what

is needed. So you may have different approaches

and so forth, but a common decision framework will

help us identify the right approach.

[Slide.]

Project 1 is to create an "As Is"

regulatory decision process map for the Office of

New Drug Chemistry, Office of Biotechnology

Products, and Office of Generic Drugs. Much of

this work will be done through a contract--we plan

to have a contractor come in and work with us on

some of these things.

We think a representative sample of

product applications could be selected for mapping

the scientific decision process in the three

offices.

[Slide.]

Determine regulatory processes efficiency

and effectiveness, using metrics similar to that

what we have learned from the manufacturing

initiative; and identify and compare critical

regulatory review decision points and criteria in

the three different offices; evaluate, correlate

and/or establish causal links between review

process efficiency metrics and critical decisions

criteria, and available information in the

submission--that's the mapping process; and, also,

evaluate the role of reviewer training and

experience, and how it bears on some of these

decisions.

[Slide.]

Summarize available information on selected

products; collect and describe product and

manufacturing process complexity, post-approval

change history, and compliance history--including,

when possible, adverse event reports that come

through MedWatch and other databases; describe

product and process complexity and uncertainty with

respect to current scientific knowledge;

information available in submissions; reviewer

expert opinions and perceptions; and, if feasible

or possible, seek similar information from the

sponsors or company scientists on these same

products that we might select.

[Slide.]

What we hope to do is aim for the

following deliverables: organize Science Rounds

within our office to discuss and debate the "As Is"

process map, and the knowledge gained from the

study; identify "best regulatory practices" and

opportunities for improvement--these may include

opportunities for improvement of filling the

knowledge gap, develop a research agenda for all

OPS laboratories based on what we learn.

What is, I think, missing today is a

common scientific vocabulary. There's a need to

develop a common scientific vocabulary to describe

uncertainty and complexity. There can be--each

come from a very different perspective right now.

Develop an ideal scientific process map

for addressing uncertainty and complexity; adapt an

ideal scientific process map to meet the different

regulatory processes.

In the following--I think the three

projects that we're thinking about are not actually

fully independent. They're all connected together.

[Slide.]

Project 2 is to sort of focus on a systems

approach. We believe that without a systems

approach to the entire regulatory process--that is

from IND to NDA--Phase IV commitments and cGMP

inspection, the broad FDA goals under the cGMP and

the Critical Path Initiatives will not really be

realized.

[Slide.]

So the team approach and the systems

perspective that evolved under the cGMP Initiative

only addressed a part of the pharmaceutical quality

system. Quality by design and process

understanding to a large extent is achieved in the

research and development organization.

Pharmaceutical product development is a complex and

a creative design process that involves many

factors, many unknowns, many disciplines, many

decision-makers, and has multiple iterations in the

long life-cycle time.

So we have to treat it as a complex system

optimization problem.

[Slide.]

Significant uncertainty is created when a

particular disciplinary design team must try to

connect their subsystem to another disciplinary

subsystem--for example, clinical versus chemistry,

or CMC to GMP. When you bring those connections,

there's significant uncertainty.

Each subsystem can have its own goals and

constraints that must be satisfied along with the

system-level goals and constraints. It is possible

that goals of one subsystem may not necessarily be

satisfactory from the view of other subsystem and

design variables in one subsystem may be controlled

by another disciplinary subsystem. Impurities is a

good example. Pharmtox, CMC, and how you bring

that together.

[Slide.]

So the Project 2 proposal that we're

developing is to use ICH Q8 as the bridge between

the cGMP Initiative and the rest of the regulatory

system, and to develop a knowledge management

system to ensure appropriate connectivity and

synergy between all regulatory disciplines. Can

that be done? I mean, that's the feasibility

project that we are trying to develop. So--connect

Pharm/Tox, Clinical, Clinical Pharmacology,

Biopharmaceutics, CMC, Compliance all together.

[Slide.]

The current thinking is to approach this

problem as connecting every section within the ICH

Q8 CTD-Q, within the same document, but to all

other sections in an NDA, in some way or form. For

example, each section within the P2 can have an

impact on the other P2 sections and, similarly,

other sections of a submission and to cGMP.

By recognizing this as a complex design

system that involves multiple attributes, goals,

constraints, multidisciplinary design teams,

different levels of uncertainty, risk tolerances,

etcetera, we wish to find opportunities to identify

robust designs and design space that provides a

sound basis for risk assessment and mitigation.

So this would be a scientific framework.

It was a regulatory tool that could come out of

this. And with the case studies and everything

coming together, this might be a way to bring and

connect all the dots.

[Slide.]

What we have been looking out is outside

pharmaceuticals. We believe that a significant

body of knowledge exists. Example, in mechanical

engineering, as it applies to the design of

aircrafts, that addresses some of these challenging

points that we have discussed. These are three

examples that I have selected as just illustrative

examples of how multidisciplinary optimization

methods and system-level problem solving tools can

be thought about in the drug context.

[Slide.]

Just to illustrate this point, let me

create an example here. The applicability of

multidisciplinary optimization methods for solving

system-level problems and decision trade-offs will

be explored in an NDA review process. That's what

we're proposing.

For example, in the Common Technical

Document for Quality--the P2 section, which is what

ICH Q8 will define--critical drug substance

variables that need to be considered in section

2.2.1, which is "Formulation Development" are

described in section 2.1.1. So there's a drug

substance, and there's a formulation. They're two

different sections.

Information for "Drug Substance," has a

bearing on that of the "Formulation Development."

So how do you connect the two together?

For example, the current language in ICH

Q8 for "Drug Substance," states: "Key

physicochemical and biological characteristics of

the drug substance that can influence the

performance of the drug product and its

manufacturability should be identified and

discussed."

So that's describing the information

content in section 2.1.1. that we will hopefully

receive whene ICH Q8 is done. So how does this

have a bearing on the "Formulation Development"

section?

[Slide.]

I'll skip this and just show you a figure.

[Slide.]

You have the API--or drug substance

manufacturing process. The X(1.1) is the design

variable; the f(1.1) is the objective function to

be addressed; and the g(1.1) is the constraint for

that manufacturing process that delivers the drug

substance. Okay?

Since this is not part of ICH Q8, what

will be part of ICH Q8 is section 2.1.1., which

will identify what are the critical variables for

the drug substance, as they relate to the

formulation aspect. But that becomes the input for

what--how it connects to the "Formulation

Development" aspect. And that link is through a

linking variable.

Since my means and standard deviations

have become finger-pointing and so

forth--[laughs]--so you know--you have a design

variable, you have a linking variable, you have an

objective function, you have constraints around

which you define your design space. You have mean

objective function--that's your target. You have a

standard deveiation that you sort of bring to bear

on that. And deviation range of the design

solution, or the design space.

So all of this sort of has to come

together for this to be meaningfully connected.

And, for example, if you start with a simple design

of experiment, you may have mathematical models,

which are empirical, but then they provide that

connectivity. So it's a start of a very formal,

rigorous approach to dealing with uncertainty,

knowledge gaps and complexity.

So this might be a useful concept. So

that's the process right now, to see whether this

could be a feasibility project that we could do.

[Slide.]

So the potential deliverables of using

this approach could be significant. Since we are

moving towards electronic submissions, in

conjunction with electronic submissions, this

project can potentially provide a means to link

multidisciplinary information to imporve regulatory

decision--that is, clinical relevance to CMC

specifications. We may not all have all that

information, but the links--the structure--will be

there as we grow, as we improve our knowledge base,

or will it be refined, the links could get

populated, and this might be an approach for

knowledge management within the agency.

Creating a means for electronic review

template and collaboration with many different

disciplines; provide a ocmmon vocabulary for

interdisciplinary collaboration; create an

objective institutional memory and knowledge base;

a tool for new reviewer training; a tool for FDA's

quality system--and, clearly, it can help us

connect cGMP Initiative to the Critical Path

Initiative.

So that's the project that we hope to

develop. We really want to get some feedback from

you, and develop this as a project under the

Critical Path Initiative.

[Slide.]

But the third aspect of this--it all could

happen in parallel--explore the feasibility of a

quantitative Bayesian approach for addressing

uncertainty over the life cycle of a product. The

most common tool for quantifying uncertainty is

probability. The frequentists--the classical

statisticians--define probability as "limiting

frequency, which applies only if one can identify a

sample of independent, identically distributed

observation of the phenomenon of interest."

The Bayesian approach looks upon the

concept of probability as a degree of belief, and

includes statistical data, physical models and

expert opinions, and it also provides a method for

updating probabilities when new data are

introduced.

The Bayesian approach may proivde a more

comprehensive approach for regulatory decision

process in dealing with CMC uncertainty over the

life cycle of a product. It may also provide a

means to accommodate expert opinions.

And I think there's a connection here.

The evolving CMC review process may be a means to

incorporate expert opinions. And I think that is a

significant opportunity.

Using the information collected in Project

1--that I described--you would seek to develop a

quantitative Bayesian approach for risk-based

regulatory CMC decision in OPS.

So that would be a project that will run

in parallel to the other two approaches that we are

moving forward.

[Slide.]

So, I'll stop my presentation here with

sort of summarizing, in the sense--I think OPS,

from its goals and objectives, has to have an

overreaching project that sort of connects all the

dots together. And the proposal--the first one

clearly is a process map--"As Is" and so forth.

But the two others are feasibility projects that we

want to look at the Bayesian approach and a complex

system optimization problem.

The knowledge exists outside. It's simply

adapting and adopting it in our context.

What you'll hear--after the break, I

think. Or--unless you want to start earlier--after

the break, is other immediate office projects;

Office of Biotechnology projects, Office of New

Drug Chemistry project, Office of Generic Drug

projects on Critical Path, and Office of Testing

and Research.

What we have done is Keith Webber will

introduce the reset of the talks. You will hear

each group's perspective. And we have requested

Jerry Collins to come back and sort of

summarize--after his talk on the Critical Path--the

entire Critical Path Initiative from an OPS

perspective and pose questions to you.

And we have also invited Professor Vince

Lee, who is now part of FDA--who used to be the

chair of this committee--who has been with agency

for almost a year now, to come with his perspective

on how--what are challenges he sees. So you will

hear sort of presentations and some opinions from

people who have been at the agency and been looking

at this challenge for some time.

So, again, the discussion today is to seek

input and advice on ACPS; on how to align, identify

gaps, and identify opportunities.

I'll stop here and entertain questions on

my part of the presentation.

CHAIRMAN KIBBE: Are there any questions?

DR. SINGPURWALLA: I have comments.

CHAIRMAN KIBBE: Okay. Thank you.

DR. SINGPURWALLA: I just--what you say is

music to my ears. You have good vision about some

of the things you want to do. But I think it's now

time that the dance should begin.

We should get back--take concrete problems

and address them. I've said this before.

But let me just make some specific

comments on some of the things you've said. And,

of course, I'm going to question some of the things

you said.

The first argumetn I want to make on your

slide on page 7, about efficacy and safety:

generally, those tend to be adversarial. Drugs

that give you benefit may have side effects. So

the important issue is to do a trade-off. For that

you need to talk about assessing utilities: what

is the utility of the benefit, and what is the

dis-utility of the harm? That's a part of the

whole package of thinking about these problems, and

I encourage you to look into it.

Now, I take strong objection to some of

the things you have said. You have distinguished

uncertainty into stochastic and epistemic. I have

seen that distinction before. I claim it's totally

unnecessary. Uncertainty is uncertainty, and one

doesn't--one should not pay much attention to the

source of the uncertainty--

DR. HUSSAIN: Right.

DR. SINGPURWALLA: --whether it is

regulated allatoire uncertainty, or epistemic, does

not matter.

CHAIRMAN KIBBE: Right.

DR. SINGPURWALLA: The Bayesian approach

does not distinguish between the two. And since

you've been talking about it, I think--

You also say that there are no good

methods for devleoping standard approach for

addresing uncertainty. I think that's the wrong

slide to put up. That's liable to do more harm

than good.

DR. HUSSAIN: Okay.

DR. SINGPURWALLA: There are methods

available. So I would not encourage you to put it.

And the other thing is: I don't like your

linking uncertainty and complexity. They're two

different issues.

And you also say that there is no common

scientific vocabulary. Well, I claim there is a

common scientific vocabulary, and that is

probability.

Now, as far as recommendations are

concerned: I'd like to suggest--and, again, I'm

not volunteering since I'm making the

suggestion--that you have your people exposed to a

tutorial on Bayesian methods and Bayesian ideas, so

that you get a better appreciation of what it's all

about. And the best way to do this is to take a

simple example and work through it; work through

your expert opinion notions that you're saying.

Go through an example, and you'll get a

better appreciation of what it's all about. And

once you get that appreciation, you'll be tempted

to remove some of the other things you've said.

Those are just comments. Thank you.

DR. HUSSAIN: No--the point's well taken.

And we actually have a project right now with the

University of Iowa, looking at our stability data

from a Bayesian perspective. So we're just

starting to put a real-life example on that. So

that's--

With regard to the utility, Jurgen and the

Clinical Pharmacology Subcommittee has been sort of

bringing that up. So we will connect to the

Clinical Pharmacology group.

Jurgen, do you want to say anything about

that?

DR. VENITZ: [Off mike.] Well, other than

the fact that--other than the fact that we're

discussing it. It is a controversial issue,

because you're really trying to map, then, a lot of

different things into a uniform scale. Personally,

I don't see an alternative, and I think it's

already done. We're just doing it intuitively, as

opposed to expressedly.

So it is being discussed. We have to see

where it goes.

DR. HUSSAIN: And, regarding, I think, the

common vocabulary, I think it's a common vocabulary

in the context of when we speak from a pharmacist

to a chemist to an engineer--we have very different

interpretation--that's what was referred to.

DR. SINGPURWALLA: That's why you need a

tutorial.

DR. HUSSAIN: That's exactly--

DR. SINGPURWALLA: Put people together.

Because about 15, 20 years ago, the Nuclear

Regulatory Commission was facing similar problems.

And one of the things they did is they had lots of

tutorials to get everyone on board, talking the

same language. Otherwise, you'll have a doctor

talk to an engineer, and those two talking to a

lawyer--and you know what can happen.

[Laughter.]

VOICE: [Off mike.] Lawsuits.

CHAIRMAN KIBBE: Another question?

DR. KOCH: I guess, just to build on the

last comment--when you get into all those

multidisciplinary functions--and particular when

the ICH Q8 is going to serve as a group, together

with the implementing the cGMPs--there's a couple

of organizations out there I think could serve as

very valuable resources. One we've heard about a

couple times today in the ASTM 55, as a body to

help at least standardize the terminology. And the

other one is the ISPE, which could serve as a

multidisciplinary conduit that, working together

with ICH, could probably facilitate some of the

multidisciplinary issues.

DR. HUSSAIN: I think we do plan on

extensive training and team building and coming on

the same page. If you look at the PAT and the

manufacturing signs White Paper that we issued, we

actually laid out a lot of these things in there,

including the role of ISPE, ASTM, PQRI, and so

forth.

So we have been thinking about this in

that context, and at the ICS meeting in

Yokahama--on Wednesday, I think, the date is

set--we will be updating on that. So I'll get a

chance to talk about ASTM to ICH in Yokahama,

Japan, also.

So, we're aligning everything together.

So that's happening.

There was one point that I wanted to

respond to: the reason for keeping uncertainty, in

terms of variability in knowledge--keeping the

distinction, at least as we think about this,

was--and the link to complexity, also--clearly,

complexity and uncertainty are two independent

things. But, unfortunately--well, the challenge we

face is this--in the sense we have a very complex

product. We have simple products--within the same

office, in OPS and different regions. Yet today, I

think, from a variability perspective, we're not

very sophisticated in how do we deal with

variability.

And, for example, in our manufacturing

science White Paper, we don't even deal with

variability of our dissolution test method. We

don't even know how to handle it. So we have

challenges today where simple variability--we don't

have a good handle on.

So that was the reason for keeping

variability and knowledge-based uncertainty on the

table.

CHAIRMAN KIBBE: Ken?

DR. MORRIS: Just a quick question: on

your identification of the gaps in the current

programs, are you thinking more in terms of

technical gaps--as in science that needs to be

done? As opposed to logistical gaps within--

DR. HUSSAIN: Both. Both.

DR. MORRIS: So, with respect to the

scientific gaps, are thinking, then, to take it one

level more--basically, are you talking more about

new science that needs to be created? Or science

that needs to be communicated more

effectively--within the agency?

DR. HUSSAIN: Well, I think the immediate

need would be to communicate the existing science

and bring all the existing knowledge to bear on

that. And, clearly, in the long term there are

fundamental issues--and most of the new science

would be needed. So I think it's an issue of

timing.

DR. MORRIS: Thank you.

CHAIRMAN KIBBE: Anybody else? Nozer, you

wanted to--

DR. SINGPURWALLA: No, I just wanted to

say that this distinction between allatoire and

epistemic has been artifically created by

frequentist statisticians. And Bayesians don't buy

it.

DR. SELASSIE: I have a question.

CHAIRMAN KIBBE: Yes, please.

DR. SELASSIE: You know, in your graph on

R&D spending--has there ever been a breakdown in

100

how much of that spending can be attributed to the

"R" and how much to the "D?"

DR. HUSSAIN: I don't have that--I'm sure

that information's--I don't have it. So I'm not

aware of it.

DR. SELASSIE: Because would one parallel,

you know, the flatness?

DR. HUSSAIN: One was the public funding;

one was more private funding, so--

DR. SELASSIE: Yes, but they're both going

up.

DR. HUSSAIN: Yes.

DR. SELASSIE: But I'm wonder if, you

know--because you look at your product submissions

are flat. Now, is that because there's not been an

increase in development funding? Or--

DR. HUSSAIN: I don't think so. But I

don't have an answer.

DR. SELASSIE: Yes.

DR. HUSSAIN: So let me say that.

CHAIRMAN KIBBE: Marvin?

DR. MEYER: Ajaz, you don't seem like a

101

depressive kind of guy--

DR. HUSSAIN: [Laughs.]

DR. MEYER: --but you said you were

depressed last week.

DR. HUSSAIN: Yes.

DR. MEYER: Can you give us just a real

short synopsis of where you see Europe doing things

right, and us doing things wrong?

DR. HUSSAIN: [Sighs.] [Laughs.]

No, I mean, again, I'll focus on what I

see happening in Europe--especially in the

U.K.--and how they're translating academic

research--academic finding research--into

entrepreneurial business--in particular in

manufacturing, in particular in dosage form

design--the pharmacy-related ones.

Look at Bradford, particle engineering.

And the one I saw--I saw a beautiful manufacturing

system for coating. Forget coating pans. This is

electrostatic coating; precise, automated, complete

on line, and so forth.

Nothing of that sort is happening

102

here--within my domain.

CHAIRMAN KIBBE: We have a couple more

comments, and then we're going to have to take a

break.

Go ahead.

DR. MORRIS: Yes, just to follow up on

that. I think there's--I just came back from

Europe depressed, as well, but I was in

Scandinavia. So maybe that had something to do

with it.

[Laughter.]

Yeah, it's pretty dark up there.

But, in any case, I agree with Ajaz in

that there are a couple of caveats and, in fact, if

you look at our latest hires, they're one from--via

Bradford, another one via Bath. My post-doc is

from Nijmwegin, another post-doc from Roger Davies

Group in the U.K.

And we're not training people--number one.

So, aside from not transferring the technology

effectively we're not training people to do it very

much any more. There are few places--represented

103

at the table--that still do it to some degree.

But that stems back to one of your earlier

slides, which is trying to muster NIH and NSF to

fund this sort of research. Because some of you

have been a lot closer to deanships than I. If

there's no overhead money, it doesn't get a very

kind reception. And the fact of the matter is is

we haven't had it.

So, this is--I'll stop here, because this

is my old soapbox. But I lay this at the door, in

part, of NIH and NSF for not recognizing, in the

face of overwhelming data, that there is a crisis

that needs to be address.

On the upside, there are some people in

Europe doing some things--and Japan, as well.

CHAIRMAN KIBBE: Pat, go ahead.

DR. DeLUCA: Just a quick follow up on

that, too. I know from my trips to Europe, too, if

you just look at the colleges--the pharmacy schools

in Europe--I mean, they all have departments of

pharmaceutical technology. I mean you'll be

hard-pressed to find pharmaceutical technology as

104

an area of focus in an American college of pharmacy

now. Certainly you won't see any departments of

pharmaceutical technology.

So I think it's been--and it wasn't that

way 20 years ago. But, I mean, it certainly has

changed, though.

CHAIRMAN KIBBE: Anybody else?

Good--I think we're at a nice break point.

And if we could take perhaps a 10 minute

break--because Ajaz has managed to get us--use up

all of our lead time.

[Laughter.]

And we can get Keith to start his talk at

about 10:22, that would be great.

[Off the record.]

CHAIRMAN KIBBE: 22 minutes after 10 has

arrived, and one way or another we're going to get

back on process.

Dr. Webber, are you prepared to get on

process?

He's on the way to the podium.

Those of you walking around with cakes in

105

your hands, and sodas, you want to sit down.

Nozer. Here we go. Good luck. We gave you 10

minutes to do that.

[Pause.]

You snooze, you lose, as the old saying

goes.

So, Dr. Webber, shall we start our

Strategic Critical Path?

Research Opportunities and Strategic Direction

DR. WEBBER: Okay. I guess we're about

ready to get started on this session, regarding

research activities and our strategic goals for the

Office of Pharmaceutical Science.

I'm Keith Webber, with the Office of Biotechnology

Products. And let me--

[Slide.]

--there we go.

Ajaz went through a very good

presentation, I think, on the Critical Path. And

I'm not going to really address very much about the

Critical Path Initiative itself. But, in my view,

this--I've sort of summarized things into the Drug

106

Development Path, which begins with discovery of

potential targets--or potential new drugs; and then

you have to have a period where one evaluates the

candidates and makes a selection of what candidate

you should carry forward into the pre-clinical

study, where one looks for potential toxicities and

potential efficacies in an indication of interest.

If all goes wlel, one moves into clincal

studies, and if all goes even better, into

commercialization. And then, once you're on the

market, there's always the period of post-approval

manufacturing optimization--or we would like to see

that, from the FDA's perspective, anyway.

And then, often, we get new

indications--we see new indications being developed

for drugs that are on the market. And that

essentially starts the process back up again--often

at the clinical studies stage.

[Slide.]

The--I didn't bring a pointer. Is there a

pointer here?

VOICE: [Off mike.] Just use the mouse.

107

DR. WEBBER: Just use the mouse. Okay.

That will work. Right there is the mouse. Okay.

I guess, historically, FDA interactions;

have occurred primarily ion this area here, from

clinical studies on. Prior to that, we have had

very little influence, I think, until we receive a

submission which contains information regarding the

pre-clinical studies.

But I believe we have opportunities to

have an impact on this entire process in the

future.

Let's see.

Essentially, I guess, sort of the essence

of the Critical Path is the--in my mind--is the

view from empirical versus guided drug development.

And drug development has to be a learning process

in order to make intelligent decisions regarding

such issues such as your candidate selection; what

dosage form you're going to have and what the

formulation should be; in choosing clinical

indications, you need to know what patient

population is going to be the best selection for

108

your product. And then when you're evaluating

clinical endpoints, one needs to know which are the

most appropriate endpoints to evaluate in the

clinical studies, and are there surrogate endpoints

that are more appropriate than others, if you can't

look at an endpoint which is directly related to

survival or efficacy in the more normal manner.

And, of course, with adverse event

monitoring, any clinical trial is going to monitor

particular parameters, and you need to have a good

knowledge base in order to understand which adverse

events we should be looking for, and the best way

to evaluate those.

And then, finally, the manufacturing

method certainly is a major concern because that

has to do with the ability to improve the

manufacturing process post-approval and

pre-approval, as well as avoiding issues that can

come up with regard to safety and efficacy of your

product.

[Slide.]

The goal of industry, as well as the

109

agency, I believe, is to establish a knowlege base

and the tools that are necessary to predict the

probable success of any given product, and the

manufacturing methods that are appropriate to it,

and then to foster the development of products that

are going to have a high likelihood of success,

throughout clinical development and on the market.

[Slide.]

Now, for this late morning's presentations

and this afternoon's presentations, we'll be

hearing from a number of groups within OPS. One is

the Informatics and Computational Safety Analysis

staff, which is in--essentially in the immediate

office of the OPS; and then Office of New Drug

Chemistry, Office of Generic Drugs. And the first

three here are the groups that do a lot of

relational and database analyses as part of their

research activities. There are, in some cases,

collaborative research going on with laboratories,

per se. But it's the groups on this--the last two,

the Office of Testing and Research, and Office of

Biotechnology Products, that have actual

110

laboratories where research at the bench is going

on.

[Slide.]

Let's see--within OPS's Critical Path

Research, I think we can address--or can address

the issues regarding candidate selection, based

upon an understanding of the structure and activity

of the relationships that we see, and the products

that ocme down the line, as well as what's reported

in the literature.

Dosage form development and evlauation I

think is an important area that we're working in.

Toxicity predictions for products is--we're

amenable to that, so our research can address that

through, again, structure activity-type

relationships and structure-function issues, as

well as knowledge of the impacts that a particular

disease state might have on physiological function

that may lead to toxicities that wouldn't be

present in all populations.

Bioavailability and bioequivalence

predictions are certainly important for all of our

111

products, but particularly for the Office of

General Drugs, they're quite critical. And I think

with regard to the follow-on products as well, it's

a major area of concern.

Metabolism prediction is something that

is, I think, crucial because products, once they

enter the body, as you know, they don't remain in

their initial state. And the metabolism can impact

toxicity, it can impact efficacy, it can impact the

bioavailability and biofluence of the products

themselves.

Immunogenicity is another area that is of

large concern, particularly for protein products.

And there we need to evaluate and understand, not

only the caues of immunogenicity, or the impacts of

various structures in the proteins on

immunogenicity, but also the impact that the

patient population has on immunogenicity; what

impact the indication that's selected can have on

impacts of immunogenicity as a safety concern.

Often, as I mentioned earlier, you have

biomarkers that you're looking at for

112

pharmacodynamic parameters, or for surrogate

endpoints. And a good knowledge of the validity of

a particular biomarker, and our ability to evaluate

those, as well as industry's ability to select

those, is dependent upon the knowledge that they

have of the biology of the disease that they're

studying, or that they're trying to cure or that

they're trying to treat.

The mechanism of action of the drug is

certainly critical when you're looking at the

potential. One area is with regard to drug-drug

interactions. Oftentimes we've been looking

primarily at metabolism for drug reactions, but

certainly there's a concern that I think is

building for utilization of multiple drugs that

impact on the same metabolic--not metabolic

pathways, but the signaling pathways, let's say, at

the cell surface, which are getting the

treatments--you know, getting a treatment into the

cell, or that are resulting in the clinical

effect--is what I'm trying to say, in a very poor

way.

113

Let's see--the pharmacogenomics is a new

area that we're getting involved in, but it's very

important with regard to patient selection, as well

as the potential for certain populations to be

impacted by drugs in a unique way, that can impact

not just efficacy, but also the safety.

And manufacturing methodologies are an

area that we have research programs in within the

office, and those are important for developing and

understanding of the robustness of various

manufacturing processes, and the ability to

implement new paradigms, such as process

technologies in the manufacturing process of

pharmaceuticals

[Slide.]

Out strategy here is to coordinate

cooperative research activities. And, as I

mentioned, we have predictive modeling programs.

And these are generally based upon information from

regulatory submissions that we receive, as well as

from laboratory research that's going on within the

agency, as well as outside and in the published

114

literature.

One area which, I think, we need to build

is our abilities to get information from industry

that we don't get in a our regulatory submissions,

and that they don't publish, and finding a means to

have them help us to gain knowledge of that

information so that we can implement it into the

decisions we make and share that--basically the

conclusions that come out of that with industry as

a whole, to address the Critical Path.

[Slide.]

There's also laboratory research going

on--you'll hear from the Offices of Testing and

Research, Applied Pharmacology Researhc, and

Product Quality Research, and Pharmaceutical

Analysis--and also from my office, Biotech

Products, from our divisions of Monoclonal

Antibodies and Therapeutic Products--it should be

Therapeutic Proteins. Sorry. Typo there.

There's also research going on in other

FDA centers that we can collaborate with, and do

collaborate with, as well as outside, to gain

115

information from academia, industry and other

egoernment agencies, as well.

[Slide.]

Now, I think we can gather all this

information, but it's critical with regard to how

we're going to use it, and how we're going to

disseminate it, such that we can have an impact on

the Critical Path.

There are a number of avenues to get to

academia and manufacturers, and those include the

public forums, where we can present the conclusions

and recommendations. We certainly write guidance

documetns that can help in this manner, as well.

And then, when industry comes to meet with us at

the regulatory meetings, such as pre-IND, and

pre-NDA meetings--pre-BLA meetings--we can interact

with them at those points, as well.

But we also need to change, to some

extent, our review processes within the agency,

and--so the information has to go to the reviewers,

as well. And we can do that via training programs,

as well as the guidance documents that we do write.

116

They're used a great deal by the reviewers.

Then, again, mentoring programs, to bring

up the new reviewers in an understanding of the new

paradigms and new concerns, or lessen their

concerns for particular issues that relate to

pharmaceutical manufacturing, or clinical issues.

And then all of this together should help

to enhance the application of your process from the

reviewer's standpoint, and with regard to the

manufacturers should help to remove some of the

hurdles and obstacles we see in the Critical Path.

[Slide.]

You'll hear the coming presentations. So

there are some questions we'd like you to keep in

mind, that we'll be bringing up later for

discussion.

And first is: are we focusing, within the

office, on the appropriate Critical Path topics?

And are there other topics that we should be

addressing through our research programs? And it's

both the database relational type information or

research programs as well as the laboratory

117

programs.

And then, in the future, Critical Path

issues may change. So how should we identify

Critical Path issues in the future. And we'd like

recommendations on how we should prioritize those.

Because we're really--at this point, we can't do

everything that needs to be done with the current

resources, and so we're going to have to prioritize

now, and in the future we'll need to prioritize, as

well, and we'll need some guidance on that.

That ends my presentation. We'll move

into the first talk--to stay on time--which is

going to be--let's see, I'll bring it up here--Joe

Contrera.

Informatics and Computational Safety Analysis

Staff (ICSAS)

DR. CONTRERA: Okay. I'm the director of

the Informatics and Computational Safety Analysis

group. Our main mission, really, is to make better

use of what we already know; material or safety

information, toxicology information that's buried

in our archives; and also in the scientific

118

literature and in industry files.

Our group develops databases and also

predictive models. You can't develop models

without the databases. So they go together.

We have develop our own paradigms for

transforming data, because traditional toxicology

data is textual, and converting into a weighted

numerical kind of a scale that is amenable to be

processed by computers, and also to be modeled.

And we encourage, promote and also work

with outside entities to develop QSAR--qualitative

structure activity relationship software--and data

mining software, for use in safety analysis.

We don't work alone. And you'll hear more

about this in my talk. We leverage, very much, and

cooperate, and collaborate very much with

outside--with academia, with software companies and

with other agencies. And we do this through

mechanisms such as the CRADA--the Cooperative

Research and Development Agrement--which is really

a buisness agreement--and also we do it with

Material Transfer Agreements, for an exchange, quid

119

pro quo exchange, with software and other

scientific entities outside the center.

[Slide.]

The Critical Path Initiative--you've all

been, and you're going to be hearing more about it,

and you've heard a lot about it. I'm focusing on

what is relevant to my group, and that is: the

problem is that we have not created sufficient

tools to better assess safety and efficacy. We're

still relying on toxicology study designs that were

designed 50 or sometimes 100 years ago. And it

doesn't mean that they're inferior, but maybe there

are better ways of doing this now.

So we need a process to develop better

regulatory tools. And it was really a controversy,

to some extent: whose misison is this? And in the

past, the agency didn't consider it as the agency

mission to develop these tools--necesarily. It was

academia. And academia said, "No, it's the

industry." It wasn't--it was vague as to who was

actually responsible for developing new analytic

tools that can be used for regulatory

120

enpoints--especially in safety endpoints.

[Slide.]

So now how d we connect with the citical

path? I think we were doing Critical Path research

well before there was a Critical Path Initiative.

I mean, we've been in operation, in one form

another, for over a decade in the Center, at a time

when people were questioning whether this was the

mission of the agency in the beginning.

We developed databases and then predictive

tools that are used by the industry--by the

pharmaceutical industry--more and more to improve

the lead candidate selection. And the question

was: why should the agency supply industry with

better tools to select lead candidates? Well, it's

in our interest that they develop lead candidates

that have fewer toxicology or safety problems.

Because when they come to us, in the review process

and submissions, they can said right through with

very few issues. Otherwise, they'd bog down the

system. And we have multiple review cycles, and

there are issues to be addressed. And it would be

121

wonderful if they could just slide through.

And so also to facilitate the reiew

process internally, by having reviewers having a

rapid access to information that is usable for

"decision support," we call information; that they

can use to make judgments on a day-to-day basis.

And we hope that also this could reduce testing;

reduce the use of animals. And also encourage

industry--software companies--to get into the

business of developing predictive modeling tools.

[Slide.]

And we see this three-dimensional diagram

for the Critical Path. Well, the computational

predictive approaches are identified in two of the

three pathways. And so we feel we're right in step

with what the future goals of the agency are.

[Slide.]

What have we accomplished already? Well,

again, we do two things: databases and predictive

modeling. And this sort of summarizes some of the

accomplishments; the first being we've developed

predictive software for predicting rodent

122

carcinogenicity, for example, based on the compound

structure. It's being used by the pharmaceutical

companies. It's distributed by small software

vendors.

We are also--obviously, we cannot screen

industry's compounds in the agency. That would be

a conflict of interest. But our software is being

used. We have an Interagency Agreement with

NIH--NIH has a drug development program--we have a

contract with NIH. NIH sends us compounds that

they're screening in their drug development program

for treating addiction. And so we are, in our own

way, practicing what we preach, in terms of using

our software in lead selection in drug development.

We also--software is being used--and we

lay a consulting role, within the Center, for

evaluating contaminants and degradants in new drug

products and general drugs, to determine--to

qualify them, and determine limits. So we feel

that our software could have much more application

in that realm.

And decision support for review divisions.

123

We collaborate very closely with the Center for

Food Safety. And, in fact, we're training their

scientists, and have shared our software with them,

and they're using our carcinogenicity predictive

software to screen food contact substances. Because

they're working under the new FDAMA rules that

place the burden on the agency; in other words, the

agency has to, within 120 days, decide whether

there is a risk. The agency has to give cause why

a substance is a risk. It's a reverse of sort of

what drugs are.

So in order to meet those kinds of

deadlines, they had to go to predictive modeling to

ascertain whether there's a potential risk of a

food contact substance--within 120 days.

EPA is looking at our--and we work with

them. And the software also can be used in

deciding whether we have a data set that is

adequate; whether there are research gaps that need

to be filled.

[Slide.]

So we talk about the FDA information. We

124

get submissions, we review them. There's an

approval process, and then the post-approval

process. We extract information from this process.

We extract proprietary toxicology data,

non-proprietary toxicology and clinical data. And

we build proprietary and non-proprietary databases,

so we can keep information that can be shared with

the public through Freedom of Information and

information that will not be shared--or cannot be

shared legally--into two different databases.

And we use these databases for a variety

of functions: for guidance development, for

modeling. And also for decision support fo the

review; and also it feeds back on industry, because

much of this information can be shared with the

public, because it's under the Freedom of

Information Act.

[Slide.]

We have leveraging initiatives in both

realms. We leverage to get support from outside to

help us develop databases, so that we don't rely

entirely just on FDA funding.

125

And the objectives are to creat specific

databases--endpoint specific. They could be mouse

studies, three month, 90-day studies, one year

studies; the toxicology databases that people are

interested in.

These database initiatives are funded and

supported through CRADAs and other mechanisms. We

have a CRADA with MDL Information Systems, which is

a part of Reed Elsevier publishing company. They

are interesting in building a large information

system, and so they're helping, supporting, our

effort. We have CRADAs in the works with Leadscope

that has a wonderful platform for searching

toxicology data. And also we have a CRADA in

process with LHASA Limited, in England--University

of Leeds in England--that has a system also--an

interest in these kinds of databases.

What we--our databases are

constructed--the center of our database is the

chemical structure. It is a chemical-structure

based database. And the structure is in digital

form so that it can be teased--it's a

126

chemoinformatic database. And the digital form is

called the .mol-file structure, and it's a common

structure used in industry for over a decade. So

the chemical structure, as well as the name is the

center search point.

And then once you have a structure that's

in digital form, you can not only ask a simple

question about, "Can I find substance x," but you

can also query and ask whether--"I'd like to know

everything--all the compounds that are like it."

And that's such a powerful tool--regulatory

tool--that I think is another--puts us in another

dimension.

It's not that I want--"Tell me about

acetaminophen," but I want to know compounds that

are 90 percent like acetaminophen in a data set.

And we're able to do that now--really easily--with

the system.

So once we have this system, then we tie

in--the databases are linked to this search engine.

We have our clinical databases that we

model--post-marketing adverse event reporting

127

system, and also the tox databases. And we use all

this--what we're really interested in is modeling;

computational predictive toxicology.

And the sources of that data on these

databases come from reviews. We extract

information from the regulatory reviews and from

other databases.

[Slide.]

So, now, getting into our modeling

operation, we transform the data. We supply the

chemical structure data, and our collaborators and

software companies supply the software. And we

work with them on an iterative basis to improve and

make these things work, and develop software for

these endpoints.

We've also, I think, are probably the

first group that have developed a way of using

chemical structure to predict dose. And so we have

a paradigm for predicting what the maximum daily

dose of a compound might be in humans, within a

statistical, obviously, error bar, in humans.

So, currently, in our prediction

128

department, you might say, we have access to five

or six different platforms. And they represent

very different algorithms. And this is the

point we want to have interactions with software

companies that have approaches that are different

from one another. And then we evaluate and work

with them to try to develop models, using our data

sets.

So we have two CRADAs on board right now,

with multi-case and MD/QSAR, and we have others in

the works. And we also have interactions with

other prediction approaches from the statistical.

[Slide.]

In terms of the models that we're working

on now, the objective is to model every single test

that's required for drug approval. And so we

started with carcinogenicity, because that was the

most--the highest profile, in terms of preclinical

requirement; and teratology would be next. These

are endpoints that cannot be simulated in clinical

trials; mutagenicity, gene tox--all these are

models, either have been created or are in the

129

process of being created and being worked on.

We're also attempting to model human

data--the adverse event reporting system;

post-marketing human data. This is an enormously

difficult data set; very dirty data set, but it's

enormous, in terms of its size.

[Slide.]

And we have had some success, preliminary

modeling, of hepatic effects, cardiac effects,

renal and bladder, and immunological effects in

humans. These are still works in progress, but we

have made progress.

And in terms of the dose related

endpoints, we have made really good progress. We

were surprised, ourselves, because we didn't really

think this would work. We've been able to

successfully model the human Maximum Recommended

Daily Dose--you know, that's the dose on the bottle

when you get your drug. It says "Don't take more

than 10 milligrams a day for an adult. Well, we

modeled that, because that comes from clinical

trial data. That is really human data. And it

130

represents an enormous scale--I don't want to get

into it--but it's like an eight-block scale of

doses, and we have 1,300 pharmaceuticals that are

either--that we've modeled, in our database. And

we were able to successfully model this--and I'll

get back to that in a moment.

[Slide.]

The other question that came up was

proprietary data and sharing industry data. It

would be nice to get their data, especially in

areas that we know the industry has a great deal of

experience in, like gene tox data. Right now we

can't have access to data that was not in

submissions. And so we need a way of doing this.

Chemoinformatics gives you a way of at least

getting there partially. We're able to share the

results by not disclosing the structure and name of

a compound. You can disclose the results, but you

say "What good is disclosing results, or using the

results, without knowing where they came from?"

Well, you can use descriptors--chemical

descriptors--that can be used in modeling, but

131

cannot be used to unambiguously reconstruct the

molecular structure. But they contain enough

information to model.

And so you're sort of at least halfway

there. You can share some information that can be

used in modeling. And so this is a feasible

approach and, in fact, it's already being

accomplished--legally. It's gone through our

legal--our staff at the agency and it's

incorporated in some of these softwares.

[Slide.]

And this is an example. This is 74 MDL

QSAR descriptors for the compound methylthiouracil.

Now, these descriptors are used in modeling, and

ocntain a great deal of scientific information, in

terms of modeling. But all of these descriptors

will not unambiguously recreate the structure of

methylthiouracil, because there's a lot missing.

It's like a pixel pictures. You know, you have a

photograph--a digital photograph--if you've only

got 70 pixels, you'll get a rough picture of what

it is, but you won't know it's your uncle. It's

132

just a person--you know. But if you had 10,000

pixels, you'd know exactly who it is. It's the

same idea. So you can share this crude image.

[Slide.]

Getting back to modeling the human maximum

daily dose--at present, we have to go through many

steps to arrive at a starting, Phase I clinical

starting dose, in a drug that's never been into man

for the first time. We start with animal

studies--multiple dose studies in multiple species.

So already that's a lot of cost. Then you estimate

the no-effect level--has to be estimated from this.

Then you have to decide which species is closed to

man by looking at the ADME and, you know,

metabolism and everything. And then you have to

convert that to a human equivalent dose using

allometric scaling. And then, on top of that, you

use a little--the uncertainty factors, dealing for

inter-species extrapolations--finally come up with

a dose that you might try for your first dose in

human--in clinical trials.

Well, if you could model, on the basis of

133

structure, the maximum recommended daily dose, you

get a predicted dose in humans--because that's

human data. You take one-tenth, or one-hundredth

of that, just to be on the safe side, and you have

a dose.

And what's the benefit? There's no

testing in animals. There's no lab studies.

There's no inter-species extrapolation, because

you're using human data. And we think it's more

accurate, because animal studies don't predict

whether a drug is going to cause nausea, dizziness,

cognitive dysfunction. Animals can't tell you

that. But yet that appears in labeling for old

drugs all the time.

So we feel that this is a good approach.

Everyone acknowledges that the estimation of the

first dose in clinical trials is a bad--but it's

the only thing we know how to do. So this has got

to be better, because it's better than nothing.

You know, because right now what we're doing is a

very crude approximation.

[Slide.]

134

What's another application? And--in

conclusion--the two-year rodent carcinogenicity

study--in mouse and rat. It costs $2 million. It

takes at least three years to do. And there's

always controversy about the outcomes of these.

Yet it has an enormous effect on the drug's

marketability.

Is it necessary to do these studies for

all drugs now? Can computational methods replace

some of them? I'm not saying we're getting rid of

all testing. But if we know a lot about a

particular compound, based on the experience of the

past, perhaps with predictive modeling there may be

a subset of compounds in which we don't have to

test as vigorously. And those which we know very

little about--and the computer can tell you that;

that the compound is not covered in the learning

set, and therefore you better do all the studies.

But if a compound is another--you know,

antihistamine, maybe there's a lesser path because

a structure that's so well represented in the data

set, that it's sort of silly to keep testing it

135

over again, just to meet a regulatory requirement.

So we're hoping that this would reduce

unnecessary testing and put the resources where

they're needed; testing things that we really don't

know anything about, and that are new--that are

really new compounds.

[Slide.]

So the challenges for accepting predictive

modeling: we need accurate, validated--and that's

always--you know, what we mean by "validation" is

always arguable. But we need to develop that.

That's part of our mission.

Standardization of software; experience

and training--it's not something that's going to go

on a reviewer's desktop ever, because it requires

interpretation. It's a really special skill.

We need more databases; adequate sharing

of proprietary information; the bigger the

database, the better. But we need, also,

regulatory mangers and scientists that are willing

to consider new ideas--consider; don't have to

adopt--consider. That makes a big--you know, opens

136

the door for innovation.

And then the ned for an objective

appraisal of current methods. It's the emperor's

clothes. How good, really, is what we're doing

now? And that is something that's painful, but

it's something that needs to be done. Compared to

what? Is it better, worse--compared to what?

[Slide.]

In PhRMA 2005 meeting that occurred

several years ago--and I think it was very

farsighted--Price Waterhouse Coopers had a

paradigm. And they said, "Right now you have

primary sciences: the lab-based, patients--you

know, clinical trials; and the secondary is the

computational--what the call "e-R&D"--that there

will be a transition where they'll reverse from

primary to secondary. And the primary science

maybe in the next generation, will be the modeling

and predictive science, and the lab and clinical

will be the confirmatory science.

So, with that, I'll end my talk. We've

published much of what we've done. A lot of it is

137

in press right now. We have a web site: our

maximum recommended daily dose database is on our

website, and a lot of people are working with it,

and we're happy to say that they're getting the

same results--which was nice.

And I'll end my talk here.

CHAIRMAN KIBBE: i'll take the prerogative

of the Chair and ask the first question. And then

we'll get rolling.

Your database looks wonderful when you're

dealing with toxicity. Have you also done a

similar thing with clinical effectiveness, or

utility, of compounds? Some way of looking at the

structure, and then looking at the effect, and

being able to predict how effective one structure

is relative to another?

And then follow up with that--if that's

true, can we plug into the opposite end of your

program and go back the other way, and just bypass

drug discovery?

[Laughter.]

DR. CONTRERA: [Laughs.] Well--no fair.

138

I'll start with the last one--but you'll be only

discovering what we already know. There may be--

CHAIRMAN KIBBE: But I was thinking of

plugging in different parameters--

DR. CONTRERA: Yeah.

CHAIRMAN KIBBE: --in the toxicity and

outcome: lower toxicity, higher efficacy--

DR. CONTRERA: Oh, yes. Yes.

CHAIRMAN KIBBE: --and then go backwards.

DR. CONTRERA: Yes, that's possible.

CHAIRMAN KIBBE: Thank you.

DR. CONTRERA: But getting back to

efficacy--yes. In fact--I mean, industry is using

it as an efficacy tool all the time. That wasn't

our mission. But potentially--certainly

applicable. And sometimes we stumble on those

things. But that isn't our mission.

And you know where research--we've got

four people in this unit. And then we have

contractors. And then we get students. So we're a

small, tight unit. And you have to be very

focused, in terms of your priorities, and doing

139

what is feasible first, and less--and so we didn't

get into efficacy. No.

CHAIRMAN KIBBE: Who have I got down here?

I've got everybody on the right side.

So we'll start it at the end, and work our

way down.

Go ahead.

DR. SELASSIE: Okay. I have a couple of

questions for you.

First of all, with your database, you have

in-house data that you're generating for your

toxicology?

DR. CONTRERA: Yes.

DR. SELASSIE: Do you ever go to the

literature and get information from it?

DR. CONTRERA: Yes. Actually, that could

be a much more complicated slide. But we mine

everything. We mine other databases; the NIH

databases; literature. And, in fact, we're

using--we're using our CRADA with MDL--because MDL

owns almost every journal in the world

now--practically. Elsevier owns almost everything.

140

And so--and they have access to data that's

enormous.

So, using the leverage with a publishing

company, we have a pipeline now to the literature.

Yes.

DR. SELASSIE: Okay. I have another

question.

DR. CONTRERA: Yes.

DR. SELASSIE: When you're inputting the

structures, do you all ever use the SMILES

notation?

DR. CONTRERA: Yes, we use SMILES. There

is some ambiguity. In fact, the software will use

either one.

But, the .mol file--you know, you could

add a lot more: three-dimensional components and

other--you know, .mol file has the capability of

doing a lot more than SMILE. But the software will

run on both--both systems.

DR. SELASSIE: Okay. And noticed, in

using your descriptors, or using the e-state

discriptors--

141

DR. CONTRERA: Yes, e-state.

DR. SELASSIE: Do you ever use log P in

there? For partition coefficient?

DR. CONTRERA: Oh, yes--log P is part of

the MBL QSAR package. It's also part of the MCASE

package.

For carcinogenicity--I will be frank--for

carcinogenicity predictions, log P doesn't have any

role at all. We took it out because it didn't do

anything. It didn't help.

CHAIRMAN KIBBE: Jurgen?

DR. VENITZ: Yes, I wanted to commend you

for your efforts. Obviously, this is exactly where

the FDA can something contribute that nobody else

can--

DR. CONTRERA: Yes.

DR. VENITZ: --because you're in the

possession of all this proprietary piece of

information, you can perform meta analysis using

qualitative methods.

A couple of comments: the first

one--right now toxicity is your main endpoint.

142

DR. CONTRERA: Right.

DR. VENITZ: You're looking for predicting

toxicity--

DR. CONTRERA: Right.

DR. VENITZ: --or doses. You might also

want to use similar methods to predict

biopharmaceutical characteristics, such as

bioavailability, metabolic stability, permeability.

DR. CONTRERA: Yes.

DR. VENITZ: Because, I mean, in the sense

of the Critical Path method, where you're trying to

screen out, in silico, potentially bad candidates--

DR. CONTRERA: Right.

DR. VENITZ: --that's, I think, number

one or number two on the list why drugs fail. They

don't get absorbed, or they get metabolized.

DR. CONTRERA: Yes, right.

DR. VENITZ: So that if you wanted to use

your resources, other than toxicology, that would

be one thing to do.

DR. CONTRERA: Right.

DR. VENITZ: The second comment is maybe a

143

little less--or more farfetched, I guess: and that

is to look at things like biosimulations, that

don't use empiric models but, rather, mechanistic

models to predict what might happen with new

chemicals. In other words, you're trying to mimic

physiology--and, again, I think is think this is

still in the infancy, in terms of predicting

certain kinds of--

DR. CONTRERA: Right.

DR. VENITZ: --toxicity. But it may come

in handy, in addition to those more statistical

empiric predictive models.

DR. CONTRERA: Well, in terms of your last

point, with the mechanistic data, that's why we

have a collaboration with University of Leeds in

England. Because they have an enormous amount of

experience with human expert rule-building, and

LHASA Ltd. And they have a--their Derek program is

used all over Europe for predictive modeling, and

that's based on getting data and trying to--and a

human committee coming up with mechanistisc rules,

based on--and so--but we felt that was out of our

144

expertise, but it was way--it was exactly what

they're doing. And that's why we're developing a

CRADA with that group. Because they are probably

one of the best, in terms of taking statistical

modeling--Bayesian modeling--and teasing out

rules--mechanistic rules.

And in terms of the ADME--of

bioavailability--you know, Dr. Hussain has already

brought that up as a wave of the future, and we

actually had discussions with Simulations Plus, and

Ray Bolger, to get into that.

But we're going to do that with those

people--within our group--that have expertise in

that area.

My group is really, mostly toxicologists

and chemists. So now we've got--and we don't just

leap into a new area until we develop alliances

with people that are experts in another field.

DR. VENITZ: One--can I make one last

comment?

CHAIRMAN KIBBE: Go ahead.

DR. VENITZ: It's not related to chemistry

145

as much as looking at biomarkers; and that is

relating biomarkers to outcomes--either

pre-clinical or clinical outcomes, where you could

use similar methods to--

DR. CONTRERA: Yes, I think it can be.

This is--you know, this is in its infancy, but I

think it's an emerging science. It's great. It's

really exploding.

CHAIRMAN KIBBE: Dr. Koch?

DR. KOCH: I just wanted to comment that I

think it's a very impressive approach. Will there

be a follow-up, in terms of using this type of data

as a way to enhance new drug discovery, or some

examples when something some together? Or is there

a possibility that it actually raises the bar on

new drug discovery, because of predictions?

Maybe a suggestion--unless you've already

done it--maybe tie in with what Art has

suggested--but if you put into that model some

already-approved past generation

pharmaceuticals--maybe some simple things--

DR. CONTRERA: Yes.

146

DR. KOCH: --like acetaminophen or

aspirin or some steroids--and see how you would

have predicted their--

DR. CONTRERA: Sure.

DR. KOCH: --present day efficacy.

DR. CONTRERA: Sure. Sure. To some

extent that's part of what we do--what we call our

internal validation, where you take compounds out

of the system, then you have the system predict

them--and not only predict them, but then show you

what clusters of compounds that were in the

database it used to make the judgment of whether it

was going to be carcinogenic or not.

And, actually, that's the most, I think,

enlightening tool, in terms of the scientists.

Really, it's an interface. What we're tryign to do

is develop an automated expert. You know, when you

go to an expert, what does an expert do? He says

he thinks--he has a good deal of experience, and he

says, "You know, I've seen that before in my years

of experience." And also, he goes to the

literature, and he--and so all we're trying to do

147

is, to some extent, automate that, speed up that

process.

We're still going to have the human

interface, but people are so--you know, get a

little bit suspicious of the machine, but we're

asking the machine to do what we ask our human

experts to do. But maybe it can do it a little bit

more thoroughly, you know. But you still have to

evaluate the output of the machine.

So one thing is good about many of the

softwares is that you get the basis for the

conclusion. And then you can judge and say, you

know, "This doesn't make sense. It says it's

carcinogenic, but the top 10--the compounds that it

modeled in the cluster of compounds that it used to

make the model, none of them are--"--you know. So

you say, "This is junk. There's something wrong."

So you still--so you need good trained

operators to be able to interpret.

CHAIRMAN KIBBE: Ken?

DR. MORRIS: Yes, thanks. This is really

a nice presentation. I think it's pretty exciting.

148

The first thing an expert tells you, of

course, is their rate--by the way.

[Laughter.]

My question actually deals more with

mayabe what will be in the future, I guess, because

at least as I understand from the presentation,

that your descriptors are all based on the

molecular structure.

DR. CONTRERA: Yes.

DR. MORRIS: And then responses--

DR. CONTRERA: Right.

DR. MORRIS: --which is the typical QSAR

approach.

I guess--and we were talking about this at

breakfast this morning--the thing that sort of

comes to mind is the opportunity--or is there an

opportunity, I guess is the question--to use the

targets--that is the receptors or whatever it is

that stimulates it, and do a more--what would be a

more traditional, I guess, molecular simulations

approach to actually backing into--the reason the

rational drug design in many senses didn't meet its

149

promise was because of the statistics, as well as

the lack of knowledge of efficacy; whereas here,

your same database should give you significantly

more data--if you can identify the targets, and if

there's--

DR. CONTRERA: The targets aren't

necessarily well-defined. And there are better

laboratories than us out there that are doing

target, you know--modeling targets. And in the

pharmaceutical industry, that is their domain.

And what we wanted to do is what no one

else was doing.

DR. MORRIS: There are people modeling

targets?

DR. CONTRERA: Oh, yeah. Yeah. They have

three-dimensional modeling of receptor targets in

order to develop drug molecules--

DR. MORRIS: Oh, no, no, no--I don't mean

to develop drug molecules.

DR. CONTRERA: Oh, okay.

DR. MORRIS: I mean, to use the database--

DR. CONTRERA: Yes?

150

DR. MORRIS: --with targets, particularly

if you have structures for the targets--

DR. CONTRERA: yes.

DR. MORRIS: --to be able to go back and

calibrate this. Because the problem with the

people that you're talking about, and the problem

they face every day is the vagaries in their force

fields, as well as some of the other tools they

use.

So, with this as an anchor, so that you

actually have the data with which you could

calibrate those in a sort of semi-empirical

fashion--

DR. CONTRERA: Yes, that may--

DR. MORRIS: --it seems like you'd have a

big leg up.

DR. CONTRERA: Yes, maybe there would be a

complementary--you know--

DR. MORRIS: Yes--no, I don't think--I'm

not saying you should--

DR. CONTRERA: --yes, we stayed away from

that type of--but you're right. Yes.

151

CHAIRMAN KIBBE: Pat, do you have

anything?

DR. DeLUCA: Just--certainly impressive,

what you're doing. And I guess I'm wondering about

applying it to the product development part of drug

development, in that once something is, you know,

discovered--knowing it's a weak base, or a weak

acide, knowing the PKA, solubility--some of those

parameters that can be plugged into the database

that would then a lot right in the formulation

aspects--is there a salt form, if you're looking

for a higher concentration that you may not--is not

soluble in the form, the weak base; what salt form

might be performed, a drug made?

So if the database can help in that

product development scheme, to look at formulation

aspects, I think that would be very helpful.

DR. CONTRERA: Right. I think,

again--that's something we got involved in--I think

we get involved with, because I know it's a big

problem for industry. It's one of the reasons why

drugs fail, in terms of bioavailability and

152

solubility.

CHAIRMAN KIBBE: Najer--we're working our

way around the table. So I don't want to--

DR. SINGPURWALLA: Well, this is not a

criticism of you--[laughs]--but it's a criticism of

the Price Waterhouse Coopers slide that you put up.

DR. CONTRERA: Yeah?

DR. SINGPURWALLA: I think that slide is

very misleading. And I'd be very reluctant to put

it up. And it's because of a slide like that that

our Chairman raised the question that he raised.

The slide seems to give the impression

that computers are going to address these issues,

and it's going to make the primary science

secondary. Now, the reason why I take objection to

this is because of the following: that any

model-building endeavor involves three elements.

Element number one is the basic science--that's the

physics, the chemistry, the pharmacy--whatever have

you. The second thing it involves is data, if

available. And the third thing it involves is the

judgment of the scientist--even in pure theoretical

153

physics, the judgment of the scientist plays a very

important role.

So, what the computer--and then, there is

a theory, which helps you put all these together.

So there are two theories: there is the theory of

the science, and the theory of the fusion--how to

put all these things together. And the computer's

role is simply to facilitate the putting these

three all together.

So I think one should be very careful in

trying to highlight the role of the computer here.

There is a parallel in what you're doing, and what

is done elsewhere, in the context of nuclear

weapons. Similar problems are faced.

DR. CONTRERA: Sure.

DR. SINGPURWALLA: We can't talk much

about them, but I think you may want to look at

what else is going on in that area, and downplay

the role of computers, and not use this Price

Waterhouse Coopers slide, because obviously they

are a consulting firm, and they're going to push

computers.

154

DR. CONTRERA: Well, I don't know--they

also are--I imagine, are involved in all kind of

research beside computer research. They do

everything. They just look at markets in general.

But--and maybe there's--calling it

"primary" and "secondary" science, people that are

lab-based would say, "Oh, you've made me a

secondary citizen" kind of thing. And you can

change the term.

All we're saying, that the emphasis is

goign to change. There's going to be more emphasis

on trying to model and predict; before you spend a

lot of money on an experiment you better make sure

the experiment's worth doing--or it hasn't been

done before. And that's what we've been wasting

money for a generation.

CHAIRMAN KIBBE: Marvin Meyer?

DR. MEYER: Have you had any successes

yet, where the computer and the software predicted

no toxicity, and the agency therefore did not

require certain toxicological testing? And I

assume the answer is "No, we haven't."

155

DR. CONTRERA: We--

DR. MEYER: How close are you to that?

DR. CONTRERA: No, we haven't applied it

that we. We're very careful about saying

that--we're not using this to make a regulatory

decision. This is a decision support.

DR. MEYER: But you could.

DR. CONTRERA: But down the road maybe it

will be. But right now we're not there yet--by any

means there yet.

But right now, it's being used more and

more heavily by the pharmaceutical industry, in

terms of their screening process. That's where the

big role is.

And, you know, it's just like--I don't

know if you're familiar with--but when Bruce Ames

came out with the Ames test--you know--for

mutagenicity, all of a sudden everyone started

using it. It was an easy test. It was relatively

inexpensive. The drug companies started mass

screening of all the compounds. And before you

know it--you know, we don't get Ames-positives

156

anymore in the agency. Whereas we used to get

Ames-positive tests that were compounds. They're

gone. So that tells you that a testing paradigm

could have a big effect.

And so these programs that predict

carcinogenicity will filter out those rodent

carcinogens that are really--major rodent

carcinogens will disappear. And eventually people

are going to say, "You know, we've been doing this

test. We never get much positive anymore. You

think we should--"--that's where I want it to go.

It won't happen by fiat, it's going to happen

by--but it's going to happen, you know.

CHAIRMAN KIBBE: Judy has a quick one.

DR. BOEHLERT: Yes, Judy has a quick

one--going out of order.

When adverse drug experience reports come

into the agency, is anybody going back to your

database and saying, "Could this have been

predicted? Does it look like this is real? Or

could this be a fluke?"--you know. "I wouldn't

expect it for this molecule."

157

DR. CONTRERA: They do. Actually, they do

come to us. They come to us a great deal when

there's ambiguity--in test data, and they can go

either way; you know, there's some slight positives

in one test, it's like negatives on the other. And

they'll use it sometimes, again, to try to come in

and weigh on one side or the other. And that's

what we call "decision support."

It happens a great deal in the

contaminants and degradants area. Now, a compound

comes up really late in development--all of a

sudden they scale up, and there it's over x-percent

that the ICH level, and a company said, "Oh, it's

harmless--"--you know. And we say, "I don't know.

You've got to lower it."

And then what usually happens--because I

was a reviewer for 10 years, and I was a team

leader during that period of time. So I sort of

came up from the review ranks. And many times a

chemist would come running to me and say, "Oh,

we've got to do something about--tell me everything

you can do, as a pharm tox. What is it? And is it

158

bad?" And I said, "How do I know?"--you know.

And, you know, you look at it and you say,

"Well, is it like something that's real bad?" And

then you'll tell the company that they have to do a

tet, because you've got to close the regulatory

loop. I'd say, "Oh, do a two-week rodent study,

and if it's clean you can go on." "And do an Ames

test." If it doesn't show a positive, then they

could probably go with over 2 percent.

Now, that's an answer, but the chances of

getting any positive toxicity in a two-week study

is zero to none. And they've already done an Ames

test probably, so you do it again.

So what I'm trying to do is have a

rational basis for regulation, where you go to the

computer, where you do a predictive model; the

model gives you 20 compounds that are 90 percent

sinilar, and what their regulatory or testing

history is. Now, you bring that to a reviewer and

you say, "You know, I think there may be a problem

because this compound is like a teratogen. It's 90

percent similar to a known teratogen." So now you

159

can go to the company and say, "Look, either you

can reduce it, because we have reason to believe,

based on the literature, that it's close to

teratogen. But if you don't think it is, do

a--"--now I can tell you exactly which test to do.

"Do a segment 2 teratogenic study. And if it's

negative, you're clear." Or reduce the level.

But I think that's a rational basis of

regulation.

CHAIRMAN KIBBE: We need to start to close

this up. So--because we've been having lots of fun

with this talk.

[Laughter.]

Go ahead.

DR. KARO: Okay. I havea comment, and

then two questions.

First, I would take exception to something

that you said early on, that we're still using

toxicity tests from 50 years ago. You know, as a

toxicologist, we've made a lot of progress.

DR. CONTRERA: Sure.

DR. KARO: And there are some new

160

tests--especially in sensitization; that we're not

using the old tests.

The other is that with QSAR, the quality

of the database is absolutely essential to know.

How do you evaluate the quality of the various

databases that you're using?

And, secondly, you mentioned validation.

And that is, you know, critical. If you have a

human database, how do you validate the predictions

from the human database?

DR. CONTRERA: Well, human database

validation is probably the--that's the most

difficult. And we're not sure yet how to best

validate that. We're right now trying to devleop

models that are stable, and we validate those by

looking at the cluster of compounds on which the

decision was based to see if a human expert would

agree that they did represent aspects of the test

compound that made sense--you know?

In terms of data quality, that's always a

problem. And that's why we try to rely on data

that's already been screened by committee. In the

161

case of--that's why--and one of the good things

about carcinogenicity data is that we have a

carcinogenicity assessment committee within the

agency. And the committee meets and decides on

what the study said. Because there's a lot of

ambiguity within the studies. And so we base it on

the calls of the CAC committee--calls in our files,

going back many years.

And in terms of other databases, we try to

base it on committee-based data sets--you know.

Teratology--the tera agonist--there's a lot of

organizations that have already, you know, reviewed

a lot of this data and have published it.

But often, you know--that is always a

problem with data mining. And my bottom line is to

predict a performance. Because if there's really a

lot of junk in the database, predictive of

performance will go down. But if the data set has

good predictive performance, then you have

somewhat--

DR. KARO: It's primarily prediction?

DR. CONTRERA: Yes, the predictive--and

162

how we validate, we do it two ways. We keep

compounds out. They're never in the learning

set--to use later, to see how well it predicts.

And also we take compounds out of the data set a

little out of time--

DR. KARO: Right.

DR. CONTRERA: --model and then, you

know--which is the traditional way QSAR people do

it.

DR. KARO: Let me share and experience.

DR. CONTRERA: Yes.

DR. KARO: I developed a model for skin

irritation, using a human database--

DR. CONTRERA: Yeah.

DR. KARO: --that, using this internal

validation, was at 90 percent predictive.

DR. CONTRERA: Yeah.

DR. KARO: We then went and tested it on

humans, and it was like 30 percent predictive.

DR. CONTRERA: Right. And that's what

we've always been afraid of. And that's why we use

external validations a lot. And that involves--the

163

best external validations come from--in areas where

there's a lot of data--you know? But most of the

time people try to put all the data they can find

into the model, and then you have nothing to test

it with--you know?

But because we're in the agency, compounds

keep coming in. So we stopped collecting at a

certain point for the database, so we have 1,200

compounds. We wait two weeks--or a year--we'll

have 24 new carcinogenicity studies. So we'll test

it against those, you know. And they represent new

drugs.

And so that's the best sort of real-world

kind of testing that we try to do.

DR. KARO: And then you readjust the

model--

DR. CONTRERA: Yeah. Yeah. And then we

go to the model. And so with our collaborators, we

tell them on a yearly basis, we have to give them

an updated, you know, software.

CHAIRMAN KIBBE: Nozer is going to get the

last word in--I cant see it. And then we're going

164

to have to move on, or else we'll be here 'til

midnight.

DR. CONTRERA: Okay.

CHAIRMAN KIBBE: You're doing a great job.

We're really enjoying it.

DR. SINGPURWALLA: Well, the comment is:

the new paradigm, you said, is modeling and

prediction. I would like to suggest that the new

paradigm be fusing of information from dierse

sources, so that you get good predictions.

DR. CONTRERA: Yes. Yes.

DR. SINGPURWALLA: I think the focus

should be changed.

DR. CONTRERA: Using it from everywhere

that you could possible find. And that's where

leveraging and collaborations are essential. You

cannot do this alone. No one can.

CHAIRMAN KIBBE: Thank you. Okay, thank

you very much.

Keith?

DR. WEBBER: The next speaker is Dr. John

Simmons, who is the Director of the Division of New

165

Drug Chemistry I, in ONDC. And because we have to

start the open public hearing at 1:00, we may want

to consider saving the last speaker--Lawrence

Yu--until after lunch, perhaps.

CHAIRMAN KIBBE: Okay, thank you. John?

DR. SIMMONS: Yes, how much time do I

have?

CHAIRMAN KIBBE: John's slides are being

handed out as we speak. Don't go looking for them.

You have one-and-a-half milliseconds. But just go

ahead.

[Laughter.]

DR. SIMMONS: I'll try to keep it as

focused as possible.

Office of New Drug Chemistry

DR. SIMMONS: I guess, just a little

background. You know, the Office of New Drug

Chemistry is really where--is the incubator for

this journey of change. And we'd like your

constructive comments and your input, because we

are trying to change some paradigms, and that's not

always a clear path.

166

[Slide.]

I just wanted to highlight four things

that I'm going to talk about before I leave. One

is the Critical Path Initiative, and where we're

at--what our role is going to be; what our current

regulatory research is--and I'll explain that a

little bit more as we get to it; then, as we look

to the risk-based initiatives, as a paradigm for

review; and, lastly, what some of our future goals

are going to be.

[Slide.]

Ajaz did a very good job of outlining the

basic Critical Path components. And, obviously,

where our biggest impact is is on that lower arrow.

We can certainly step in and help folks that are

developing beyond discovery, but all the way up

through large-scare manufacturing, and that's going

to be our focus, I think.

Likewise, if you look at

industrialization, down at the bottom, that's our

home; that's where we feel most comfortable. The

Office of New Drug Chemistry looks at small-scale

167

production, manufacturing scale-up, refinement and

selection of specifications; and then, finally,

large scale. And after that, post-approval changes

and refinement, once a product is up and running.

[Slide.]

now, as regulators, and as a regulatory

body, and as a person that's been involved in both

the research and review and approval of drugs,

along this Critical Path, if you looke at some of

the areas where we can have a large impact, I'd

like to draw your attention to the pre-IND phases.

More and more, successful companies are companies

that shorten their Critical Path by coming in and

talking with us, and meeting with us.

There are invariably questions that can be

raised, discussed--scientific issues--that will

shorten their journey. And we certainly encourage

people to do that.

As you move fruther down the clinical

development, once the IND is submitted and the

phases start, certainly the end of Phase 2 meeting

is probably one of the more Critical Paths along

168

that Critical Path. And a firm that is wise, a

firm that would like to minimize the amount of work

that's done over and above what's necessary, would

come in to an end-of-phase meeting and meet with

all the disciplines--but certainly with CMC.

Oftentimes I see, on a day-to-day basis,

oftentimes products that are exciting, that

companies are trying to develop in a rapid fashion.

Oftentimes their development gets ahead of the

manufacturing. And I think this is an area where

firms can come in and meet with us, pose questions;

we can give some guidance. And I think it helps

them.

Another area would be prior to submitting

an appliation. There is no way that we can review

and approve a new drug application in a short

amount of time, unless we have interacted very

thoroughly and very intimately with firms along

that path. And I think that's something that I

always enocourage people to do when I speak at

scientific meetings, and gatherings of the

regulated industry.

169

Now, the Office of New Drug Chemistry also

gets involved in research--usually initiating

research. And I have to be honest with you,

oftentimes it's very reactive; oftentimes it's very

inefficient; and oftentimes it's very focused.

The Office of Pharmaceutical Sciences has

had the foresight to ut in place a rapid-response

team, which helps us in that venue. When you're

reviewing an application, or you've just reviewed

an application, or a problem has arisen

post-approval, oftentimes we need to look at

scientific issues that the firms simply no longer

are interested in--or simply aren't equipped to do,

or simply refuse to do. And our rapid response

team has done a very nice job of being able to take

very focused regulatory projects, put them into

place as a research project, report back the

findings, and help us make a decision. And that's

something that we want to continue to do, but I

think we want to do it in a more proactive way; in

a way that helps us anticipate, rather than be

reactive.

170

And that's one of the reasons we're here. If you

drop down to that last point, I think-- we're

seeking your input, we're seeking your guidance.

This is a journey that we are embarking on, and I

think that's one of the strengths of a committee

like this, is to validate and direct.

[Slide.]

Just as an aside, you know we're currently

developing new paradigms. The office is

reorganizing. We've started a journey where, if

you look at chemistry, manufacturing and controls,

we are trying to balance CM and C. We've spent an

awful lot of time looking at the chemistry of

things, and now we're looking more closely and the

manufacturing and the control of that

manufacturing--as an integral part of this process.

So that is a journey that we're not afraid

to take, but it will take some guidance.

We're also looking at a review focus:

what should our review focus be? And we're also

looking at the research focus: how can the

research be focused to help us make regulatory

171

decisions in a timely way?

[Slide.]

Just to illustrate some of what I've been

giving you a prelude to: here are four topics that

have involved either regulatory or regulatory

research activities. And I'll give you some

illustrations after I walk through some of the

examples.

Conjugated estrogens--difficult problem

for us; complex drug, mixture of actives, not

always consistent. We need to look at ways to

fully characterize and establish criteria for

pharmaceutical equivalence. And we've gone to our

laboratory research groups--we've got one in St.

Louis and we've got one here in the metropolitan

D.C. area--that have been very helpful in that

area. And I"ll illustrate that shortly.

Prussion Blue--very recent example of a

compound that was used as--is to be used as a

counter-terrorist measure; difficult problem to get

companies involved with. You know, these are

medications and countermeasures that may never be

172

used, or may only be used in a catastrophic

condition. Companies are loath to do all the basic

research that are involved in developing those

products.

During the review of this product, we

looked to shorten the crticial path, and we

involved our rapid-response team to look at

surrogates--in vitro surrogates--for binding of

this particular compound. It's a ferric cyanide

compound--a complex salt. It does a nice job of

binding some of the radioactive nuclides that are

around. And the company that was--the companies

that were involved in developing these products

certainly didn't havea lot of information, or

clinical human experience to go on.

There were issues about the binding

capacities, and what impacted those binding

capacities. There were also issues of the release

of free cyanide. What happens to these compounds

upon storage, or use; you know, do we generate

toxic--is the cure worse than the prevention.

Inhalation products--another area where

173

comparing products across products is not always

easy, and we invoked our research teams to look at:

how do we develop in vitro methods to establish

pharmaceutical equivalence? How can we look at

particle size, spray pattern and chemical imaging

as techniques to help us come up with standards by

which we can evaluate these products?

And lastly--and more of a guidance

venue--we're looking now at the marvelous

combination of drugs and devices. We're looking at

stents that are put in coronary arteries. We've

got a few on the market already. But in the

process of looking at athat it became painfully

obvious to us that the roles that the Center for

Drugs and Center for Devices played, and how we

could interact, needed refinement, needed focus,

and needed agreement. And we're working feverishly

on some joint guidances so that these products can

be approved in a more timely fashion.

[Slide.]

I said I wanted to illustrate a few

issues. Conjugated estrogens--when we asked our

174

research laboratories to get involved in these

products, we asked them to look at complex--look at

a complex mixture and tell us, in a systemic way,

how we can actually measure them.

And the laboratory out in St. Louis did

some marvelous work using LC mass spec combinations

to do just that. Here is a total ion chromatogram

of all the various components.

[Slide.]

And here are some of the individual

identities of those particular components. And

they can be identified and quantitated. And that

helped us in focusing some of the questions that we

would, in turn, ask our innovator companies

non-innovator companies.

[Slide.]

With respect to the Prussian Blue issue,

this was an area that was not too familiar to the

center. You know, Prussian Blue is an inorganic

therapeutic, and it's been a long time since we've

seen inorganic therapeutics in the agency.

We needed to have a better sense of what

175

to do with things that were largely insoluable; how

to look at those, how to evaluate those. So we

evoked the laboratory to take a look at them, and

they gave us a very nice idea of what to expect

when we look at APIs; what types of variations

could we see with time, as to binding; what are the

batch-to-batch variations--and, in fact, we saw

some. And it helped us focus some of the issues

that were involved in the approval.

[Slide.]

Likewise, this material can be dried.

And, as lots of inorganic salts, oftentimes water

is trapped in the matrix--in various matrix holes.

And the level of hydration can have a marked

difference on the ability to bind a nuclide.

[Slide.]

On to the issue of looking at inhalation

products. Our laboratory set up some very nice

work that helped us focus what plume dimensions

mean to a product; or what spray pattern--how could

spray patterns be chemically imaged so that we

could look, and compare products across product

176

lines to come up with some ocnsistent questions to

ask firms.

[Slide.]

Now, I'd like to move on to the risk-based

CMC review paradigm, and that's something that's a

little different than what we've been doing in the

past. In the past we've relied largely on the

science and the guidance--and by "guidance," I mean

guidances that we ourselves have writen, guidances

that have been written by international bodies,

such as International Harmonization--ICH. We're

moving away from that paradigm. We're tryign to

move from review by guidance, into review by

science and review by risk. And there are clearly

some benefits.

To patients, the obvious ones are faster

approval of products, increased availability,

continued supply. For the FDA, obviously, there's

more product and process knowledge; more efficient

allocation of resources. If we do risk-based

review versus guidance-based review, where does

that lead us? And obviously the one thing that

177

probably is the intangible that is hard to

evaluate, and that is the increase in trust and

understanding that occurs between companies that

are submitting new data to us, and the reviewers

and people that approve those products. I think

that's an invaluable aspect. If we keep things on

a risk and a science basis, I think it's much

easier to talk and come to conclusions.

[Slide.]

To industry, obviously it's more efficient

and science-based inspections. Now that's an

interesting paradigm, as well. Those of you who

are from the biologics venue have seen team

biologics, where reviewers and investigators go out

to sites. We've been exploring that in CDER for

small molecules, but not nearly to any organized

fashion. And I think you will see that in the

future. And I think there's value to that.

There are faster and more consistent

reviews. If the manufacturing and the science and

the chemistry are looked at in a more balanced

way--not only at headquarters, but also in the

178

field, there's potential for reduced regulatory

burden.

The issues of changes and nonconformance

requires less FDA oversight, if you draw it to its

extreme. We can focus resources on critical issues

that way. We can make judgments asto what's more

important.

And then there's flexibility on focus as

to what's to be done, rather than what can be done.

And I think Judy raised that issue. At some point

we have to tell people what we would like to see,

and that's not always an easy issue to come to an

agreement on.

And, obviously, it also improves

communication with the agency. You have to

communicate with the agency if you want to use a

risk-based approach.

[Slide.]

One of the paradigms that our Center

Director, at the time--Janet Woodcock, who is now

up at the Commissioner level--raised the issue to

us was: you know, how do we link quality

179

attirbutes to clinical performance? How do we link

values and specifications to safety and efficacy?

And how do we link our inspectional process to

those same issues. That's not always an easy line

in which to draw the dots.

[Slide.]

Under the new quality assessment paradigm

that we're currently lo9oking at, obviously

risk-based assessment is high on the list; clinical

relevance is high on the list; safety

considerations is high on the list.

The process capabilities are also high on

the list. At what point do process capabilities

become a limiting factor? At what point to process

capabilities give us a venue of guidance? One of

the problems that often happens in rapid

development of drugs is that firms don't have the

luxury of making large numbers of batches of

things. And I think process capabilities can be

used both as a sword and it can also be used as a

guide. And I think we're looking toward that

paradigm--that guidance paradigm.

180

The knowledge gained from pharmaceutical

development reports--you know, one of the wonderful

things about ICH is that we're into this paradigm

of sharing information and explaining how you came

to the conclusion that this was the optimum

formulation. And process development reports are a

window into that. And I think we would like to

utilize those better as companies move into that

paradigm.

And then, obviously, the better

utilization of statistical methodologies.

Statistical proces control, I think, is a way of

the future. I think companies are implementing it

in small ways now, but I don't think that firms

have had the luxury of developing it on a large

scale--at least not the drug industry in this

country.

We're looking at assessment, starting from

the comprehensive overall summary--something that

ICH has given us as a paradigm to look at. At what

point can we look to the firm to summarize some of

the issues that are involved, rather than us

181

looking at all the raw data and coming to our own

conclusions?

Good review practices, and good scientific

principles--current good scientific principles--I

think that's probably going to be something you'll

hear more and more about.

Increased emphasis on manufacturing

sciences--as we move into the new paradigm of the

Office of New Drug Chemistry, we are building a

manufacturing science team. We're currently

identifying and hiring people that have had

large-scale, hands-on manufacturing experience. It

will be very interesting to see how we incorporate

that into the review process. I'm looking forward

to it.

The use of critical and peer review of our

evaluations--you know, the paradigm up to now has

been one reviewer, on review, one product. I think

we're going to be working more on a team basis in

the future, and I think we're going to be looking

at critically evaluating ourselves as to what

questions were asked and what decisions were made.

182

And then, lastly, this integration of

review and inspection--I, for one, have always

encouraged people in my unit to accompany

investigators whenever possible. But there's a

different between accompanying an investigator and

being an integral part of making the scientific

evaluations on that site. And I think that's the

paradigm we're moving towards.

[Slide.]

If--my arrows disappeared. What happened?

These are all connected by arrows, but I

want to draw your attention to the lower boxes.

VOICE: [Off mike.] [Inaudible.]

DR. SIMMONS: One more click, you think?

By George, you're right. Let's see how many clicks

it takes.

[Pause.]

Great. Thank you.

Draw your attention to the lower boxes:

quality by design, product development report, and

comprehensive overall summary--quality summary.

We're looking at those to feed into

183

risk-based quality assessment, and reduce time

review. And, ultimately, if we want to reduce that

Critical Path we want to move towards first-cycle

approvals--especially when it comes to the

manufacturing venue.

We have little control over the toxicity,

little control over the efficacy, but we can

control some of the manufacturing issues--early on.

[Slide.]

What's in the regulatory future? I think

we see increased CMC-only meetings; by that, I mean

all disciplines certainly meet as a team with

manufacturers, but there are issues tha may involve

only the manufacturing, chemistry or controls, in

which we can meet with industry and discuss

specific issues, to shorten that Critical Path.

Quality by design initiatives; IND

Guidances--how can we better help firms formulate

what quality we'd like to see, at what levels as

you move through the graded phases of development.

Obviously, we have to be flexible on things like

this. And I think the more information that we

184

look at earlier on, the better off we'll be. But

it puts an awful lot of pressure on industry to

develop those data.

Process Analytical Technologies has abeen

a driver in the Center. We're looking more and

more at looking at in-line, on-line--or

at-line--analyses that have feedback loops to

manufacturing. We're seeing it more and more.

The integration of review and

inspection--I've already talked about that.

Strategies to facilitat first cycle

approvals--we'd like your input on that.

Combination products--we're now entering a

wonderful world in which devices and drugs are

being approved together; where the device is either

delivering the drug, or the device is carrying the

drug to prevent some secondary impact, as in

drug-eluting stents.

Also, with biological-type products--so

not that the proteinaceous drugs are within CDER,

we can look more closely at biological

small-molecule combinations. That's the way

185

they're used in real life, and I think now we can

start looking at them in a more coherent fashion.

Nono-particle technology--where will that

take us? How will we evaluate the size and shape

and impact of that type of technology on drugs--not

only how they're manufactured, but what the

toxicity and efficacy of those drugs are. We now

have in the pipeline nano-technology products, and

they present some very, very interesting questions.

And I don't claim to have all the answers, and I'm

looking to--I think we're looking to the committee

to give us some guidance on things like this.

[Slide.]

Some of our immediate next steps are

obviously implementation of the PAT

Guidance--Process Analytical Technology. I've had

the wonderful opportunity to work with teams of

people that we're training to send out to look at

these products. You know, we've just come off of a

very long journey where we had investigators and

compliance officers and reviewers exposed to the

same type of information, and trained as to what to

186

look for when you're looking at process analytical

technologies. And I think we're ready to start

seeing the fruits of that labor.

Revision of CMC guidances--can we make the

guidances more science based? Can we make them

more commonsense? Can we make them far less

checklist in nature?

Combination prodcut guidacnes--obviouskly

that's an area that we have to look at very

closely, And this integration of review and

inspection--what questions can be asked here? What

questions have to be asked and answered on a plant

floor?

[Slide.]

I think the two major future goals are:

to establish a meaningful regulatory program that's

science-based, that supports drug deevelopment and

review. I think we're partners in this process.

We're not simply a hurdle.

And I think the other one is: to explore

regulatory mechanisms to speed that process, or

shorten that Critical Path.

187

So I think I'd like to bring this to a

close, and open it up for questions, and ask you to

think broadly about some of those issues.

CHAIRMAN KIBBE: Are there any questions

for our speaker?

Good. Go ahead.

DR. MORRIS: this is a relatively short

question.

I think, when you're talking about the

integration of review and inspection, which is a

question I get a lot as I visit the companies--

DR. SIMMONS: Yes.

DR. MORRIS: --but is the limitation

organizational? Or resources?

DR. SIMMONS: I think both. I think what

we're seeing is that in the current paradigm, where

there's one reviewer and one application, and one

product, scheduling can be a terrible problem. I

think, as we move to separating pre-approval from

post-approval, and allowing people to focus on

developmental and NDA issues, I think we will see

more and more structural inspections involving the

188

reviewer.

I think the other issue is the resources

of the field. Obviously, to put two people or

three people together at a site requires intense

scheduling, the availability of resources--and

pre-inspection conferences. You can't go into a

plant without a plan.

DR. MORRIS: Yeah.

DR. SIMMONS: And I think that's the type

of thing that we're up against. And I think we'll

be--I'm pretty confident we'll be able to--

DR. MORRIS: But there's no inhibition

to--

DR. SIMMONS: I don't think so. I don't

think so. I think it's only limited by our own

resources and biases. Yes.

CHAIRMAN KIBBE: Joe?

DR. MIGLIACCIO: Just following up on

Ken's question--you talk about what question's

asked here, what questions on the plant floor.

Remember the scientists who develop the formulation

and the process are not on the plant floor.

189

DR. SIMMONS: Good point.

DR. MIGLIACCIO: So we need--

DR. SIMMONS: [INAUDIBLE] made available.

DR. MIGLIACCIO: Yes. Yes, they are made

available. But we have to have a good discussion

between industry and FDA about where the division

is.

DR. SIMMONS: Yes.

DR. MIGLIACCIO: What questions--

DR. SIMMONS: I agree.

DR. MIGLIACCIO: --are appropriate for

the plant floor.

DR. SIMMONS: I agree.

DR. MIGLIACCIO: We don't want to be

having detailed formulation discussions--

DR. SIMMONS: No. No.

DR. MIGLIACCIO: --with pharmaceutical

engineers on the shop floor.

DR. SIMMONS: No. I agree with that. But

on the other hand, I think it--a picture is worth a

thousand words. If you're looking at process

analytical technology development, you're looking

190

at the placement of sensors.

DR. MIGLIACCIO: Sure.

DR. SIMMONS: I think there's no

substitute for looking and touching those pieces of

equipemtn.

DR. MIGLIACCIO: And if I could just make

one more comment--you talked about statistical

process control--not heavily used. Actually,

statistical proces control is somewhat pervasive in

the industry. The problem is, the statistics are

being applied to data that is being gathered for

compliance purposes.

DR. SIMMONS: Yeah. Yeah.

DR. MIGLIACCIO: And I think we're

shifting away from that now; that we're now willing

to gather data for scientific purposes--

DR. SIMMONS: Right.

DR. MIGLIACCIO: --not compliance

purposes.

DR. SIMMONS: Well, thank you--good

clarification.

CHAIRMAN KIBBE: Anyone else?

191

DR. KOCH: John--I know you participated

in the training with the combination reviewers and

inpsectors. And that continues to come up. And I

know it's difficult for the scheduling, but

anything that can be done to encourage increased

involvement in the training, so that you have more

of a base to draw from for setting up the--

DR. SIMMONS: I couldn't agree more. I

think there's no substitute for that hands-on

experience. I think it's valuable.

CHAIRMAN KIBBE: Anybody else?

If there are no further questions--

thank you.

I have logistics question. We have one

speaker for the open hearing, and we are at noon.

And we have one more speaker that fits with this

set. So the question really is: shall we go ahead

and run long, and get Dr. Yu done before we break,

and come back late? Or do we want to fit him in

after the open hearing, before we start the next

set?

And what would make more sense?

192

DR. HUSSAIN: I think the open hearing

time cannot change. I mean, that's the

restriction.

CHAIRMAN KIBBE: Well, if we have only one

person on our list--so.

I mean, if we had an open hearing and the

time is used in 15 minutes and we're done, and

there's no one else, then we can put him in there.

DR. HUSSAIN: Yes, definitely.

Definitely.

CHAIRMAN KIBBE: All right. Okay.

So we will then apologize to our next

speaker, and have him have to give his presentation

on a full stomach--

[Laughter.]

--which, hopefully, will make him more

comfortable.

We will now be at recess until one

o'clock. And if the members of the committee will

hang around, we'll discuss with you lunch plans.

[Off the record.]

CHAIRMAN KIBBE: I see by the clock on the

193

wall that we have rapidly approached the one

o'clock hour, which means that we will entertain an

open-hearing presentation.

Open Public Hearing

CHAIRMAN KIBBE: Dr. Saul Shiffman?

Please identify yourself.

DR. SHIFFMAN: I will do.

CHAIRMAN KIBBE: And then you can go ahead

and do your presentation--appreciate it.

DR. SHIFFMAN: Well, thank you for your

time. I'm just going to take you on a brief

excursion to some fairly different territory than

what you've covered this morning.

[Music.]

My name is Saul Shiffman. In my day job,

I'm a research professor of pharmaceutical

sciences, psychiatry and psychology at the

University of Pittsburgh.

Ooop--but today I'm here as Chief Science

Officer of invivodata, inc., which provides

clinical diaries for--electronic diaries for

clinical trials.

194

And, in a sense, I want to shift the focus

for a moment from the focus on drug discovery,

screening and manufacturing, to the testing of drug

products and devices in human clinical trial; and

also, in a sense, to shift from the sort of

ambitious initiatives considered under the Critical

Path Initiative that require new science, new

technology, new regulation, toward an example of

some of the kinds of things that can be done with

current science, current technology, current

regulation.

So--briefly, I'm going to talk about the

use of diaries in human clinical trials, and the

different methodologies that are in place,

basically talking about the fact that paper

diaries, which are in wide use, have serious both

scientific and regulatory, as well as operational

problems, whereas newer technologies fall within

the regulations and solve these operational and

scientific issues; and that the FDA can facilitate

the development of those newer methodologies.

So, briefly, stepping back--while

195

obviously many clinical trials are run with hard,

biological endpoints, it's not uncommon that key

endpoints are what are call "patient reported

outcomes," either because they're subjective

states--such as pain, which can't be gathered any

other way--or because the patient is often, if you

will, the most privileged observer to report on

certain events which are objective, but which the

patient is in the best position to observe.

[Slide.]

And, in fact, patient report outcomes are

collected in nearly three-quarters of all trials,

across all four phases of drug development. An FDA

audit showed that they were present in about a

third of NDAs. And diaries, in particular, are

used in about a quarter of trials. And, of course,

the function of diaries is to get the data in real

time in order to avoid the pitfalls of recall.

The traditional method has been a paper

diary. And if you've ever done a diary study, this

may bring back some memories. Operationally, there

are a lot of issues. Diaries often contain errors.

196

They're often illegible and therefore, on both

accounts, fall under the regulatory standard as a

problem; but also operationally, in trials

containing diaries, the diary is usually the last

source of data that's processed. And so it becomes

literally the item on the Critical Path that slows

completion of the diary.

A number of academic groups, as well as

industry providers are providing electronic

diaries, and audits show that they reduce errors

and the need for data cleaning very

dramatically--by 98 percent--because of the ability

to filter the data at its source, and therefore

provide operational efficiencies.

But what's important is the potential for

the diaries also to provide enhanced validity.

And, really, the biggest concern about paper

diaries has always been that they're not completed

in a contemporaneous way. Anyone who's ever done a

diary study has probably seen patients filling them

out in the parking lot, or in the waiting room.

And, in fact, the field has coined a phrase of

197

"parking lot compliance."

That's been anecdotal. Let me show you

some more formal data.

[Slide.]

We did a study with pain patients. This

shows you the data that's usually available from a

paper diary. And it shows that the patients

returned the diary cards reflecting that 90 percent

of the diary cards had been completed in

inappropriately timely way. And the problem is

that all we have is--in other words, this is what

was noted on the card.

The innovation in this study is that we

had developed an electronically instrumented paper

diary that, with photosensors, made a record of

when the record was actually filled out, so that we

could try and verify the patients' report of timely

compliance. And the data were rather

dramatic--which is that if you look at the actual

records, only 11 percent could conceivably have

been filled out at the appropriate time; in other

words, 79 percent of the returned records were

198

either inaccurate or falsified.

Importantly, we observed hoarding, which

is to say on one-third of all days, the diary

wasn't opened the entire day, and yet 96 percent of

the diary cards were returned for those days.

What we never expected to observe, but did

observe, was forward filling; that is, that

patients would--

[Laughter.]

--today, on Tuesday, fill out their

reports for Wednesday, Thursday and Friday. It

made me think that I wanted to stock advice from

these folks--

[Laughter.]

--since they could tell the future.

So, clearly, there are very serious

problems that go both to meeting the regulatory

standard--accuracy and contemporaneous

completion--but also, as you'll see, go to the

issue of scientific validity.

And, in contrast, we had a group that had

been assigned to use an electronic diary. And, in

199

fact, they completed 94 percent of the entries in a

verifiably timely way. So there is a solution to

this problem of diary completion.

So what is the benefit, then, for clinical

trials of improving the methodology?

[Slide.]

And, if you will, the hypothesis--the

compelling hypothesis--is that by getting data in

real time you reduce error, which makes trials

statistically more efficient, with greater power,

and therefore you have both more efficient--that is

smaller--trials, and essentially more reliable

trials whose answers can be relied upon better.

And, in fact, to try and validate this, a

couple of groups have done analyses comparing paper

and electronic diaries--of the same phenomenon;

essentially parallel studies.

[Slide.]

And what you see is, in fact, a one-third

reduction in error variance; essentially a damping

out of the noise, which translates into roughly a

50 percent decrease in the sample size required for

200

those trials.

So this improvement in measurement can

produce smaller trials, more reliable trials, and

possibly fewer trials, in the sense that trials are

often re-done because the first one failed.

[Slide.]

So, in essence what we have here is a

situation where the science, the technology and the

regulations are already in place. You may not be

familiar with ALCOA--it stands for

"attributability, legibility,

contemporaneousness--"--I forget what the "O"

is--and accuracy. So, essentially, there are the

existing standards, but they haven't been applied

very systematically to diaries.

[Slide.]

So, what is needed? Really, what's needed

is not new regulation, but for the FDA to apply its

existing regulations in a consistent way. At the

moment, some of the older technologies are getting

a pass on the regulations, in terms of accuracy,

originality, all of those criteria that the FDA has

201

set. And essentially, it's not so much that FDA

has in any way ruled out electronic diaries, as it

has left room for FUD--is "fear, uncertainty and

doubt." Industry regulatory folks are not known

for being adventurous. And so without clear

statements from the FDA of its own policies, this

has hampered the methodological development of the

field.

[Slide.]

So, essentially, as I've said, there's now

not just anecdotal but quantitative and formal

evidence that paper diaries fail both to meet

regulatory standards and scientific and statistical

standards; that methods are available, and what is

needed, as a small step available today, is for FDA

to speak clearly about its interest in newer

methodologies.

[Slide.]

The issue of innovation has been with us

for a long time. This is a statement from a

scholarly journal you'll be familiar with: "That

it will ever come into gneral use, notwithstanding

202

its value, is extremely doubtful because its

beneficial application requires much time and gives

a good bit of trouble, both to the patient and

practitioner, and its foreign to our hats and

associations." This statement was made in the

London Times, in 1834, and it referred to the

stethoscope.

So, initially, most innovations are

resisted, simply out of inertia. And I think part

of the Critical Path Initiative has to be for the

FDA to facilitate the adoption of improved

methodologies.

Thank you very much for your time and

attention.

CHAIRMAN KIBBE: Thank you.

Anybody have any quick questions--clarify

the information?

Marv?

DR. MEYER: Two questions: one, do most

of the electronic diaries have a provision for an

open-ended response, or an adverse event that isn't

in the database?

203

And then, secondly, coming from the great

state of Florida--

[Laughter.]

--where I see a great hesitancy to launch

into this modern electronic voting--they much

prefer having paper--

[Laughter.]

--do some of the recipients of this device

that are participating in a study have resistence?

DR. SHIFFMAN: Let me take the questions

in turn. The diaries can have provisions for

open-ended text. And, literally, you can use

handwriting and record the visual image; or, more

commonly, you can provide a little keyboard, and

people can type small comments. It varies with the

protocol whether that provision is made available

or not.

And to, in essence, amplify what's behind

your question, sometimes, indeed, one of the

reasons paper diaries are so messy is that people

write marginal notes, and a few of those have some

clinical relevance. You'd like to be able to

204

capture those, as well.

In terms of patient resistence, that's

really been very little of an issue. I showed you

the data from this pain study. We replicated those

data in a COPD study, where the average age of the

patients was in the 60s, and we've done a study of

medications for prostate cancer, with average age

in the 70s. And, in general, we get not only good

acceptance, but, if anything, we've done analyses

to show that the performance of older patients is

actually better.

So I think we have a bit of ageist bias,

thinking that this is only going to be for teenage

computer nerds. But there's just a lot of evidence

that this is well accepted and well used.

CHAIRMAN KIBBE: Okay. Well, thank you

very much.

MS. SHAFFER: Thank you.

CHAIRMAN KIBBE: We now will finish up our

morning's activities.

Lawrence is ready to give us his 25-minute

presentation in 12-1/2 minutes--to show you the

205

level of efficiency, when we apply PAT to

presentations.

Critical Path Initiative--Challenges

and Opportunities - Continued

Office of Generic Drugs (OGD)

DR. YU: I think I have 45 minutes,

right? Until two o'clock. [Laughs.]

CHAIRMAN KIBBE: I do have a priority

button.

DR. YU: Okay. I've got it.

After 15 years' graduating from Ajaz, I

guess I still look at his students.

Good afternoon, everyone. Chair and

members of FDA Advisory Committee for

Pharmaceutical Science, and my FDA colleagues and

distinguished guests, it give me great pleasure and

privilege this afternoon to discuss with you FDA's

Critical Path to medical product development

opportunities to generic drugs.

[Slide.]

As discussed this morning, the FDA's

Critical Path encompasses three aspects, namely:

206

safety, efficacy and quality.

I want to emphasize that the path to new

drug development does not end with the approval of

the NDAs, but it continues with monitoring of

post-approval changes, post-approval manufacturing

optimization, and eventually the development of the

generic drugs. In fact, the generic drugs is an

integral part of the USA health care system, as

pointed out by our President Bush, on his October

8 th second Presidential debate: "Tahere

are other

ways to make sure drugs are cheaper. One is to

speed up generic drugs to the markeplace, quicker."

So U.S. government looking for generic drugs to

limit increase in drug price, while our fellow

friends--American consumers--looking for access to

low cost, high quality, efficient, same efficacy,

and same safety, generic drugs.

[Slide.]

So let's back to the Critical Path

Initiative, as Janet Woodcock pointed out--which

you saw this slide in the morning--the FDA's

Critical Path Initiative is "A serious attempt to

207

bring attention and focus to the need for targeted

scientific efforts to modernize the techniques and

methods used to evaluate the safety, efficacy and

quality of medical products as they move from

product selection and design to mass manufacture."

So, when we apply this to generic

drugs--let's define what is a generic drug.

[Slide.]

The generic drug is basically a

therapeutic equivalent to a brand-name product. So

it would equivalent is defined as a pharmaceutical

equivalent and bio-equivalent.

So in more term, is a generic drug is a

comparable to a brand-name drug products in dosage

form, strength, route of administration, quality

and performance characteristics and, finally,

intended use.

[Slide.]

When the Critical Path Initiative defined

the safety, efficacy and quality as applied to

generic drugs, we define as bioavailability,

bioequivalence and quality. As you know, that

208

generic drugs not only should high quality but,

more importantly--equal importantly, you know, make

sure they're equivalent in terms of pharmaceutical

equivalent and bioequivalent and eventually

therapeutic equivalent to brand-name products.

So, therefore, my talk covers the

following three aspects:

[Slide.]

Bioavailability and bioequivalence

modeling and prediction; bioequivalence of locally

acting drugs; product design, characterization and

in vitro performance testing.

Now let me talk on the first topic:

bioavailability and bioequivalence modeling and

prediction.

[Slide.]

Now, this is the sketch which I made a

couple years away for my talk with Gordon Research

conference. At this time I swear I think I

invented new term: e-ADME. One time actually I

asked my son to register e-ADME as a website, end

up like the web site was registered 24 hours ago.

209

So I lost that opportunity to register web site for

e-ADME.

The basic fundamental is connect with your

control this morning's talk is the e-R and

D--e-research and development. Here, ADME means

"absorption, distribution, metabolism and

elimination." So basically e-ADME is electronic

ADME.

In terms of predicting bioavailability and

bioequivalence, or bioavailability--if you look at

the approaches of predicating forecast the

bioavailability, bioequivalence, there's two

approaches to get there. One is experimental

approach. You measure solubility, you measure

permeability, you measure metabolism, you measure

protein binding, and you measure many, many others

as development scientists did in their discovery

stage.

From those pharmaceutical measurements,

you select the so-called pharmaceutical leads. The

leads will be--a number of select leads will go to

animals, hope from animal models to predict

210

bioavailability information for humans.

Now, another approach--which I will

highlight here--is computer modeling approach.

I use red here--biopharmaceutics

classification system; compartment absorption

transit model--or CAT model--and quantitative

structure bioavailability relationships. Now

this--I put this slide basically as those research

is going on in FDA, by no means incompatible,

because we know, for example, in this slide we did

not include one of the very well known approaches

from Pfizer, and in this case Rule 5.

So let me go through each one of them very

briefly--with I think Dr. Jugen Venitz discussed

this mornign.

[Slide.]

First, look at he biopharmaceutics

classification system. The biopharmaceutics

classification is a scientific framework to

classify drugs based on solubility and

permeability. These two parameters--solubility and

permeability--each parameter has two levels, you

211

end up with four classes, namely: class BCS Clsss

I, Class II, Class III and Class IV. Class I is

highest solubility, high permeability; Class II is

low solubility, high permeability; Class III is

high solubility, low permeability; and, finally,

Class IV is low solubility, low permeability.

Four years ago, in 2000, the FDA issued a

guidance to waiver of bioavailability,

bioequivalence studies for highly soluble, highly

permeable drugs--those rapidly dissolving,

immediate release dosage forms. With issuing the

guidance, does not necessary mean investigation

research within FDA stopped. In fact, we are

continually exploring possible bi-waiver extensions

for BCS Class III drugs, namely high solubility,

low permeability drugs; we're investigating the

effect of sepins on absorption. We're

investigating transporters--for example,

p-glycoprotein transporter absorption. We're

investigating refinement of the BCS classification

system.

So research is very active within FDA, as

212

is shown here. We have three publications so far

for this year alone.

We blieve the biopharmaceutics

classification system not only its utility in

regulations, but also has its utility in drug

discovery and development. This is because the BCS

system can help you to select a proper dose form;

can help you design a formulation; can help you to

see what could be issue down the road in the

development process.

[Slide.]

So, let's move on to next topic,

which--next, the model, is what we call the

"compartmental absorption and transit model." Now

this model has become a software which was

mentioned this morning, called "Assimilation Plus."

I have a disclaimer: I have no financial tie

whatsoever with Assimilation Plus."

This is a basic software based on this CAT

model, which originally developed by myself long,

long time ago at the University of Michigan, under

professor Kodio Miro.

213

This basically, basically as a mechanistic

model, describes how a drug gets into the blood;

how much it gets into the blood; and how fast it

gets into the blood. So it's considering the

impact of gastric emptying--for example, after

lunch, gastric emptying time's probably four hours.

Before the lunch, only 20s and half hours. We look

at--we incorporate the effect of the small

intestine transit time, blood flow, volume,

dissolution, permeability, metabolism, distribution

and conventional pharmacokinetics.

The research going on is continue to

identify critical bioavailability or bioequivalence

factors. For example, if you look at this

beautiful suface here, on left side--or right

side--this is what we call the "Surface of

preferable properties as a function of solubility,

permeability, hepatic clearance and potency." Now

this is surface of purely calculated, based on

computer model, basically give you some idea what

potentially bioavailability will be for a new

molecule which just even have not been synthesized,

214

based on the solubility and permeability and

hepatic clearance you get some idea what to the

degree of bioavailability of the drug itself, of a

compound above this surface--above this surface.

This means that bioavailability will likely below

30 percent; below the surface bioavailability will

likely higher than 30 percent.

Now this is the calculate of the

theoretical model has not been validated. We are

planning to use FDA data to validate this surface

for the benefit of the public health.

[Slide.]

The next--the slides basically show you

the quantitative structure bioavailability

relationship model. Now this model, if you look at

the top left, that's basically is the structure and

bioavailability relationship. It's based on 691

drugs whose human bioavailability actually is

available within the--in the public domain. If you

look at structure at the activity relationships or

bioavailability versus structure, you've got a

correlation coefficient .71. Now, if you look at

215

it statistically, that's .71 very low.

Now, we look at these 691 compounds--this

model--to predict the drugs which were approved

around 2002, which we have 18 drugs. These 18

drugs never been utilized to QSBR models. The

correlation coefficient is 0.62.

Now if you look at the bottom--look at the

rat and dog, how animal predicts human? The

correlation coefficient for rat is .41, while the

correlation coefficient for dog is .43. So this I

can--for this system, for this drug--for those

drugs which were evaluated, the computer model at

least will not be worth at all than the animal

model.

Now, if you look at the bottom two

figures, you will say, "Lawrence, you ought to have

a five or four points. Why was that?" You say,

"N=18." Very simple: because we use 18 data from

NDA jacket internal FDA database to verify this

model, but those data were not available in the

public domain, in the public literature. That's

why we say FDA's in unique place to do modeling

216

work, which we have the data that we believe

probably no one else has so complete database as we

do.

Well, we're unique place to develop models

for the benefit of the public

[Slide.]

So, to summarize, the bioavailability and

bioequivalence prediction--we discussed the

biopharmaceutics classification system. We're

continue investigating the bi-waiver extensions;

we're exploring classification refinement. We are

continue investigating the impact for transporters,

such as the p-glycoprotein impact and absorption,

using compartmental absorption and transit model.

We use the QSBR model is a quantitative structure

bioavailability model should be developed.

Unfortunately, at this point, has not been widely

used. We believe FDA is in unique position to do

this work for the benefit of the public.

[Slide.]

So now let me move on to next topic, it's

the bioequivalence method for locally acting drugs.

217

We all know the bioequivalence method for systemic

drugs is well understood, well developed, well

utilized. In fact, luckily, we have used them for

generic drugs over 7,000, the drug products.

However, well understood, well established, well

used for systematic drugs does not necessary mean

is well understood, well established, well applied

for locally acting drugs. That's key scientific

challenges, we believe, for those--can be best used

off of FDA's Critical Path Initiative for the

benefit of the public.

The key scientific challenges include the

following: topical dermatological products; nasal

spray and inhalation; gastrointestinal, vaginal and

ophthalmic products. Now, those products, because

a lack of the bioequivalence method--the

bioequivalence method often requires the clinical

testing, the clinical evaluation. The target of

research is to provide a scientific basis for in

vitro and in vivo bioequivalence method.

[Slide.]

Let's look at--give you example why is

218

clinical studies sometimes an issue. Now this is

for topical products--I'm sorry, what I want to say

is for locally acting drugs, why this issue here?

This is because for systematic drugs, the plasma

concentration usually relates to the safety and

efficacy of drugs, while for locally acting drugs,

the plasma concentration is not usually relevant to

local delivery of bioequivalence. Because of that,

we have to rely on other alternative methods; for

example, pharmacodynamics method; for example, in

vivo clinical comparisons--for example, in vitro

comparison and certainly any other scientifically

sound, well established method, which we think is

appropriate.

[Slide.]

So, as we discuss here, the clinical

method--clinical evaluation is always available for

establishing bioequivalence. The question comes

back why this is an issue here. Why? What's going

on?

Let's look at give example here. This is

a topic product. If you look at the cure rate,

219

different, if you look at the test, in the figure

you have n=number of subjects--in fact, the number

of patients. So 90 percent confidence interval

between test, and reference and cure rate have to

be plus and minus 20s. Now, clinical evaluation

usually has large variation. In this case

estimated variability is around 100 percent.

Look at the table, in the center. Utilize

463 subject; even with 463 subjects used, the

confidence interval is minus 8 and plus 20. It

barely pass; barely pass. Now if this is 400

subject, this study will fail. In fact, we were

told the many clinical trial studies fail because

improper power; inadequacy of the human subjects.

So that, in sumamry, for clinical trial

studies to document bioequivalence present

tremendous challenge for us; tremendous challenge

to the industry; tremendous challenge--certainly

difficult for consumers because the availability or

lack of availability of appropriate scientific,

reasonable bioequivalence becomes a barrier to

generic competition; become a barrier, in fact, for

220

process improvement, for product improvement, for

products optimization because many cases those

changes require documentation of bioequivalence

method--of reasonable, simple, scientific front,

bioequivalence method is not available and it will

be difficult to make any improvement or significant

changes.

[Slide.]

As we see here, clinical endpoints have

high variabilities, and we hope--we hope,

here--develop scientifically sound, reasonable,

simple bioequivalence method to reduce unnecessary

human evaluation, or human testing.

So this is the developed for the

discussion of bioequivalence of locally acting

drugs. Let me move on to the topics which are also

dear to our heart in the Office of Generic Drugs:

product design and characterization.

[Slide.]

I said it before. The generic drugs not

only show high quality, but also equally important

to show equivalent- to the brand-name products or

221

we could pharmaceutical equivalence--pharmaceutical

equivalence, this means the same drug substance,

same dosage form, same route of administration.

So, with respect to to "same drug substance," we

need to document that exactly same; for example, we

have lots, lots issues before with pharmaceutical

solid polymorphism. This issue is resolved. But

issues still can exist for complex drug substance.

For topical dosage forms, sometimes it's

difficult to define whether it's ointment versus

cream. So this also presents challenges. So it's

exceeding--in factors of the classification dosage

form, if those exceed being inside the

classification dosage form, how do we see they're

the same?

So, therefore, when you define, you give a

very clear definition what is called the dosage

forms.

And product quality--when your product

quality standards; for example, adhesion tests for

transdermal products--of course, appropriate

scientific, predictive, in vitro adhesion test not

222

only can be applied for generic drugs, but also can

be applied for innovator brand-name products.

Equally important, we need standards for

nasal and inhalation products and a novel drug

delivery system, such as liposomes, which was

mentioned by Dr. John Simmons this morning.

[Slide.]

Another typic that research--the topic I

wanted to mention is product performance

evlauation. Now, in vitro, dissolution testing has

been around for decades; has been very successful;

has been utilized for ensure the product

quality--give example, left figure, this in vitro,

dissolution testing has been around for decades;

has been very successful; has been utilized for

ensure the product quality--give example, left

figure, this in vitro dissolution method can

usually predict, for example, polymorphic change;

the top one polymorphic 1, the bottom is

polymorphic 2. So proper dissolution testing

ensures the product quality, able to detect the

inadvertent changes of pharmaceutical solid

223

polymorphism.

Nevertheless, it's a very simple

system--just compare to human gastrointestinal

tract. You have stomach, you have duodenum, you

have jejunum, you have ileum. The volume changes

back and forth, in and out. There's 14 leaders in

and out. There's different pHs, from 1.4 to 2.1.

Before the lunch, average pH is 1.4, 2.1; now after

lunch average pH is 6, or 4.5.

Look at the duodenum or jejunum--also more

complex is the transit time is changed. Sometimes

the gastric emptying time is only two or five

minutes, under fasting conditions; sometimes hours.

The fundamental message here is:

dissolution is very simplification of a human

gastrointestinal tract. That's part of the reason

why the very easy, we see the criticism say that

dissolution is underestimating, overestimating, and

in vitro, in in vivo dissolution methods is

formulation-specific. So on and so forth.

So how do we get from here?

[Slide.]

224

The dissoluation method, beginning was

used for quality control, lately has been for in

vivo evaluation, basically the dissolution test as

a product quality-control tool to monitor

batch-to-batch consistency of drug release form of

product.

It also has been used in vivo performance

testing as in vitro surrogate for product

performance that it can guide formulation

development and ascertain the need for

bioequivalence tests.

[Slide.]

When we look at complexity, for quality

control tool, you want to have a simple dissolution

test you can use every day for every batch.

However, those simple tests for quality control may

not be appropriate for in vivo systems. That's

part of reason why, where, at the beginning, we're

asking to ourselves if these two objectives are

consistent? If it's not, we need

investigator--when you develop a bio-relevant

dissolution method it's predict in vivo--I want to

225

say it again, dissolution method has been here, has

been very successful ensure the high quality for

consumers, but those dissolution methods may be

over simplification of in vivo system. That's part

of the reason why we believe in make an effort to

develop bio-relevant in vitro dissolution method to

be predictive of in vivo dissolution, to be

predictive in vivo phenomena going on in

complicated system.

[Slide.]

Before concluding my talk, I want to say a

few words on process identification, simulation and

optimization tools. You have heard enough--that

hisotrically, pharmaceutical products involves the

manufacture of the finished products using batch

processes, followed by excessive laboratory testing

and analysis to verify its quality.

However, the process identification,

simulation, and optimization tools need to be

developed for pharmaceutical batch processes so

that any manufacturing process failure can be

readily identified and corrected. When this

226

process means that a formulation has been

defined--has been selected. The product quality

ought to be assured by high quality of starting

materials, robust manufacturing processes, and

limited--not excessive--laboratory confirmation and

test or analysis.

[Slide.]

So when we're look in future, the Office

of Generic Drugs wants to continue--all go to

continue building world class scientific expertise

in predicting bioavailability, bioequivalence and

process optimization. We face many, many

challenges. We prioritize scientific efforts. We

will pursue collaborations. We cannot do it by

ourselves. Within FDA, we have Office of Testing

and Research. I think this afternoon it's the

Division of Pharmaceutical Analysis, Cindy is goin

to give a talk. She is providing a lot, lot of

support to Generics, and office of OTR--also, rapid

response teams.

We had a collaboration already in place

with academia--for example, University of Michigan,

227

University of Kentucky, Ohio State University,

University of Maryland, and Colorado School of

Mining.

We also have a collaboration in place with

National Institute of Standard Technology, while

pursue collaboration with other government

agencies. Finally--not least--with industry.

With that, I conclude my talk. Any

comments are welcome. Thank you.

CHAIRMAN KIBBE: Marvin?

DR. MEYER: Lawrence, two questions on

that slide on page--I guess it was slide 10, the

QSBR model. One--simply, you said you illustrated

the one down on the right-hand corner, I guess, as

illustrative of the FDA's problem in presenting

data publicly. And you had four data points shown

from an n of 18.

I wonder why--how revealing would be the

other 14 data points, if you're just plotting

percent f, human percent f dog? I mean, I have no

idea whether you're talking about aspirin or you're

talking about vitamin B-12.

228

DR. YU: Well, I guess, first of all,

Marvin, you have to believe me what I said, here.

[Laughs.]

DR. MEYER: Okay. [Laughs.]

DR. YU: Secondly, in this indeed is very

simplification modeling, and I can show you slides

with actually 18 drugs--their specific name--

DR. MEYER: Okay.

DR. YU: Those 18 drugs were approved in

2001 and 2002. The human bioavailability data for

all those 18 drugs were available, actually in

public domain--the majority either from the

Physician Desk Reference. However, for animal

data--for example, if you look at rat, we only

have--I only was able to find five drugs whose

animal data--rat bioavailability--that were

available in the public literature. The

rest--basically, that's 13 drugs--were not

available in the public domain.

DR. MEYER: My statement really deals with

agency paranoia, is: why can't you show us the

data points without saying, "This is a Pfizer

229

product, this is a Lily product, this is a Teva

product." Just say, "These are products that are

marketed." Or "These are analgesics." Or "These

are antihistimines," or--

DR. HUSSAIN: I think the key is this:

the animal data may not be in the public domain.

The human data would be on the label and so forth.

So if you are able--if you can trace back

what the drug was. That was the reason.

DR. YU: If I showed all 18 drugs here--

DR. MEYER: Mm-hmm.

DR. YU: --basically, I disclose all the

animal data, because you're able to see it. And

then--

DR. MEYER: But if you don't tell me what

the drug is--

DR. YU: Yes--

CHAIRMAN KIBBE: You're obviously not a

lawyer, Marv.

DR. MEYER: Oh, okay.

[Laughter.]

DR. MEYER: I'll pass on that.

230

The second question--

DR. YU: I guess I don't want to get

myself in trouble.

DR. MEYER: Yeah, I know--well, that's

paranoia, isn't it.

[Laughter.]

CHAIRMAN KIBBE: It's only paranoia if

it's unreasonable fear.

DR. MEYER: Yeah.

DR. YU: Marvin, you're SG, you can see

all this data.

DR. MEYER: Well, then I'll have to be

quiet about it. So I don't want to do that.

[Laughter.]

CHAIRMAN KIBBE: And that's really hard to

do, too, eh?

DR. MEYER: Maybe a less philosophical

question: if I look at the upper left and the

upper right, and I draw a line at, let's say, 70

percent f--on the y axis--

DR. YU: Mm-hmm.

DR. MEYER: --I have a range that goes

231

anywhere from 30 to 100 percent, as experimental or

observed--in both cases.

DR. YU: Mm-hmm. Mm-hmm.

DR. MEYER: So even though the r-squared

may be acceptable, I say you don't have very good

predictability--at least at that level of percent

f, which would be one of interest I would think--70

percent.

DR. YU: Marvin, you have--indeed, you

have an excellent question.

DR. MEYER: [INAUDIBLE]

[Laughter.]

DR. YU: I guess I can answer it two ways;

twofold.

First of all, that's part of the reason

that the quantitative structural relationship, as I

stated, that the FDA follow in Biologics meeting,

follow-on protein biologic product meeting that one

professor expert state, it's unrealistic at this

point--maybe in the future--as you also point out

this morning, the QSBR alone--alone--can be provide

for regulatory decision-making. In other words,

232

quantitative structure activity relationship will

be used for supportive information, but however

cannot provide a conclusive data for regulatory

decision-making--at least today.

DR. MEYER: There's kind of a line

between--I tend to agree, it's maybe better than

nothing--maybe. But if I were in a company, and I

went to management and I said, "Well, I can predict

the experimental bioavailability," and my vice

president says, "Well, what will it be?" "Well,

somewhere between 30 and 100 percent."

[Laughter.]

I better start looking for another job, I

would think.

DR. YU: Actually, if you look at it, when

you place 100 drugs--supposedly, at this point, you

have 100 compounds. You have $1 million. The job

is: give me maximum information you can with this

$1 million. No, 100 drugs you're available, you

can blindly pick up 100 compounds, you pick let's

say 10, for example, for human evaluation--okay?

And then probably a couple of them--for example,

233

the bioavailability is 0 or 5 percent, so you

failed. So at least failure rate, instead of

you--your test, you got a 7. However if you use

computer model, you pick the 10 with $1 million,

likelihood you got nine. You're getting a lot with

this simple computer model, you're only cost

$10,000 versus $1 million, you benefit

tremendously.

CHAIRMAN KIBBE: I think we have some

comments on that.

Ken? And then Nozer.

DR. SINGPURWALLA: I would like to pursue

this slide, and the previous slide. So why don't

you put up number nine first, please?

DR. YU: Okay. Please.

DR. SINGPURWALLA: I'm a little

intrigued with it. You have four variables:

surface permeable properties as a function of

solubility, permeability, intrinsic hepatic

clearance, and potency. You have, actually, five

variables, and you're portraying them in two

dimensions.

234

So I don't know what's the purpose of that

particular illustration. I don't get a sense of

what it is supposed to convey.

And the second point is: irrespective of

my first point, what was the basis of your computer

models? A computer model is based on some theory,

or previous data, or a combination of it. So it's

not clear to me what is the basis of that model?

DR. YU: Well, I'll try and answer the

question.

This bsis of the computer model is a

mechanistic model--okay? If you look at

absorption, you basically have four fundamental

processes going on. One is gastric emptying and

the intestinal transit; second is the dissolution;

third is permeation across membrane; fourth is

metabolism. So this model consists of about 100

differential equations encompasses all these

processes going on. Is basically what we call the

physiology model.

And this physiologic model--if you look at

the key parameters impact those mathematical

235

equations--you have solubility, you have

permeability, you have clearance, and you have

dose. So the reason important your dose is here,

because how much input into the body will impact

the dissolution.

Now, another I think important terminology

is bioavailability. So, basocially,

bioavailability is a function of solubility,

permeability, hepatic clearance, and effective

dose. Of course many, many other factors, but

here, simplification is basically theses four,

five--four basically are fundamental parameters

which ipact the bioavailability.

So, therefore, when you look ata those

four parameters, if you know effective dose, the

potency, your educated guess, if you look at this

surface, you get some idea what likely

bioavailability will be in humans before you even

actually doing it.

So the advantage is the same for the early

stage that leads to selection. If you have a huge

number of subjects--which when I gave my--I say

236

100--in fact, we have 1,000, for example--the

candidates for human evaluation. You need to--for

human evaluation which one you select? So this

surface will help you, which one has a likelihood

to be successful--likelihood to be success.

DR. SINGPURWALLA: But you have three

variables labeled--

DR. YU: Mm-hmm.

DR. SINGPURWALLA: --so this illustration

only pertains to three variables. And you said you

had five variables, and a hundred differential

equations.

DR. YU: It was--yes, we have a

hundred--the way--do have a hundred differential

equaltions. But a differential equation is a key

parameter here is solubility, permeability and

hepatic clearance--and dose. That's why I say dose

is 1.0. In fact we have a series plot--for

example, dose 0.1, 0.5, 1.0, 5 and 10--a--plot. So

when you select a specific dose, and then you look

at this plot, and this plot--you have three

parameters, basically--solubility, permeability and

237

hepatic clearance. And then from there you see

which is more appropriate candidate for human

evaluation.

DR. SINGPURWALLA: I think I made my

point. You see three variables here. There are

two others--I'm sorry, three parameters here. You

have two other parameters. You need another

picture to connect these with those. And I won't

pursue the matter.

Let's go to number 10--

DR. YU: I think this talk about hours,

all the mathematics from one stepwise.

DR. SINGPURWALLA: No, there are certain

principles.

DR. YU: Yes.

DR. SINGPURWALLA: You can't show, in two

dimensions, more than three dimensions.

DR. YU: Okay. Thank you.

DR. SINGPURWALLA: All right.

Number 10--picture number 10.

DR. YU: Okay.

DR. SINGPURWALLA: Now, you know the

238

correlation coefficient, r-squares--

DR. YU: Mm-hmm.

DR. SINGPURWALLA: --only measures a

linear relationship.

DR. YU: Yes.

DR. SINGPURWALLA: You could have two

dependent variables that are non-linear--

DR. YU: Mm-hmm, mm-hmm.

DR. SINGPURWALLA: --and completely

dependent on each other, which r-square doesn't

capture.

So, I go back to the point raised by

Marvin, here--and previous people. There are only

four or five points. They don't look linear to me

at all. And you can't claim a correlation--you

can't claim any meaningful correlation of point

.43. It doesn't have any meaning.

DR. YU: Actually, you made excellent

point. I guess I did not make it clear in my

presentation: the point I want to make here is

animal model are not predictive of all human being.

DR. SINGPURWALLA: Okay. So--

239

DR. YU: that's the key.

DR. SINGPURWALLA: Okay. So don't put

r-square. Okay? Just put it that way.

And the top one doesn't make sense--the

r-square of .71.

DR. YU: Uh-huh.

DR. SINGPURWALLA: It seems approximately

linear to me--notwithstanding Marvin's comment.

[Laughs.]

So the first one does make sense. The

second one--I don't know how many--you show a lot

of observations--

DR. YU: Yes, there's 18 points.

DR. SINGPURWALLA: No, the second one--the

QSBR model.

DR. YU: Okay--yes, this is 18 points.

Yes.

DR. SINGPURWALLA: I think you have more

than 18.

DR. YU: 20.

DR. SINGPURWALLA: Okay--whatever it is.

Again, r-square doesn't make sense there--does it?

240

DR. YU: Well, I guess--you know, I said,

you know, when you look at r-square, .6 or .7,

statistically probably is not meaningful at all.

But, I guess, from physiological, pharmaceutical

perspectives, that at least gives us some

indications what could be potentially correlation

coefficient; that whether it's good or bad.

I hope I answered your questions.

[Laughs.]

DR. SINGPURWALLA: Yes. Fine.

CHAIRMAN KIBBE: Ken, you want to wrap

this up?

DR. MORRIS: A general question, I guess,

Lawrence--you know, the charge of looking at how

we're adjusting the Critical Path or, how, you

know, that the Critical Path Initiative is being

addressed--given that a lot of what you're talking

about isn't really generic drug-directed--sort of

taking that as a given for the moment, if it's

adding to the overall Critical Path, it's probably

still valuable.

But if you look at the larger picture, and

241

you look at, like, your CAT slide, which turns out

to be a popular slide--you don't have to put it

up--but I guess the thing that jumps out--and maybe

this is jumping forward to tomorrow a little bit,

is that this all presupposes that the dosage form

consistency is there to begin with when we're

talking about the bioequivalence.

VOICE: [Off mike.]

CHAIRMAN KIBBE: Oh, you're mike's off.

DR. YU: You're absolutely correct. And

this scenario, where I'm not looking--it's useful,

these slides, we have not looked at how formulation

impact. Impact, if you look at formulation impact

for immediate-release dosage form, you have a

suspension, different particle size. I can talk

hours.

In terms of your first question, is this

absolutely generic? Probably not. It's actually

apply equall for drug discovery and development

innovators. I guess my Director at the bureau is

so nice he did not criticize, allow me to

[INAUDIBLE] here. So that's--I have to say it

242

comes out sometimes in my research--not my mission

to talk about some of the prediction

bioavailability, bioequivalence.

DR. MORRIS: Yeah, I didn't really--

DR. YU: Well, same mission, which is to

protect and advance public health. I'm sorry--go

ahead.

DR. MORRIS: No, I didn't mean it as a

criticism. I was just saying that--I'm just not

sure that the immediate applicability of this is

with the generics. But--

DR. YU: Yes, this is equally applied to

innovators. I guess, no matter where I am, whether

it's in the Office of Generic Drugs, or my previous

position, Office of Testing and Research, our

mission is to protect and advance public health.

That's why--is part of the reason, I guess, why my

director, so he's so nice, did not correct it.

CHAIRMAN KIBBE: Okay. I think we need

to--

DR. HUSSAIN: Clarify one point, which I

didn't--

243

CHAIRMAN KIBBE: I guess we don't need to

go on.

[Laughter.]

DR. HUSSAIN: No, I think, in listening to

the talk, the message that Lawrence was delivering

with respect to dissolution for quality, and

dissolution for predicting performance, just to

further clarify what I think I thought process is,

I think--for the last 20 years we have sort of

merged the two together. And essentially what

we're looking at is separating those out. There's

a quality-control function, and there's a function

for performance prediction. And those have to be

addressed differently. That's the message that

Lawrence was giving.

DR. YU: Thank you, yes. That made it

very clear.

Thanks.

CHAIRMAN KIBBE: Thank you. Thank you,

Lawrence.

Jurgen, you're not going to let us end

244

here, huh? All right.

DR. VENITZ: Is it on? Okay.

DR. YU: You have two minutes.

DR. VENITZ: I do? Okay.

DR. YU: [Laughs.]

DR. VENITZ: Okay, you have to count.

First, again, the same comment ethat I

made earlier today--I obviously commend you for

using quantitative methods to predict, as opposed

to always requiring measure, measure.

DR. YU: Thank you.

DR. VENITZ: I do concur with the

previous--with Ken's basically, statement that here

you're talking about drug substances when you do

your quantitative structure activity.

DR. YU: Yes.

DR. VENITZ: Given the fact that you are

at OGD, I think you should also focus on

excipients, and products; in other wordsthe , what

is formulation effect? And I'm not sure whether

you can have those nice models that you showed us,

that are very meaningful to come up with NMEs and

245

figuring out what the chemical structure may be,

related to bioavailability.

The second--so, excipient effect and food

effect, to me, is something in terms of Critical

Path that's important--not just predicting drug

substances.

I do urge you to continue to work on the

BCS, because I'm pretty sure in a couple of years

you're going to come to this committee, or the next

generation of committee members, for Class III, and

you might make the same recommendation for Class

III that you just, four years ago, made for Class I

drugs.

DR. YU: Ajaz made, yeah.

DR. VENITZ: Or Ajaz made.

Two more comments: clinical

bioequivalence--that's obviously something that

this committee pointed around for quite some time.

And you made the observation--which is a true

observation--that clinical bioequivalence means you

need a large number of patients, because you have

lots of variability.

246

But I would take that argument around, and

I'd say two things: number one, you're now testing

the product in the intended population. So you

have the benefit of getting away from healthy,

usually male, volunteers, where you assess

bioequivalence.

DR. YU: That's correct.

DR. VENITZ: Number two, what is the magic

rule that requires you to have confidence in the

value of 80 to 120, or 125--as we do for areas

under the curve? Why can't you/clinicians define a

minimum difference that is perfectly acceptable?

We do that all the time for non-inferiority

trials--in the clinical area. So why can't we use

that to assess this concept of clinical or

therapeutic bioavailability to get around this

sample size that is going to go up exponentially?

The last comment--the question that you

had on the dissolution testing, where you asked

what is--is this just monitoring product

performance, or is this something that is more

meaningful? Well, the short answer is: it

247

depends. If you have an in vitro-in vivo

correlation, it is not only something that you can

monitor, but it's something that actually can be

translated in in vivo performance.

So part of what you--maybe as part of your

research--want to look at, under what circumstances

do you have IV, IVC for simple dissolution test, at

a single pH? And the complex GI tract--actually

we'd use this to a beaker with solution in it?

Anyway--thank you.

DR. YU: Thank you. Do I have time for

comment?

CHAIRMAN KIBBE: Thank you, Jurgen.

Yes, you have time for comment. We are

planning, now, to extend the meeting this afternoon

to 7:30 p.m., so--

[Laughter.]

DR. YU: [Laughs.] I guess the excipient

effect I will show in my BCS slides, not show in

bioavailability prediction slides.

In fact, the first publication, Molecular

Pharmaceutics, 2004, is deal with food effect. So

248

where I just want to say that we're investigating

effect of excipients on absorption, on

bioavailability and bioequivalence.

And your--I guess I forgot your other

comment, so I wouldn't have to comment on that.

[Laughs.]

CHAIRMAN KIBBE: That's nice.

Let me just throw out that I agree with

Jurgen's penultimate comment.

VOICE: [Off mike.] Figure out what that

means.

[Laughter.]

CHAIRMAN KIBBE: We have an opportunity

here to show a real sense of cooperation between

academia, industry and the FDA.

We have a series of speakers, all of which

are claiming they're going to use 30 minutes.

Lawrence said he was going to take 20. It was 47.

[Laughter.]

If the other speakers are on the same

track--mostly because we ask lots of questions--all

really good ones--we will, indeed, be here 'til

249

7:30.

So, let's try to focus ourselves on the

talks at hand, move through them quickly. And

anybody who has more than 25 slides should be

embarrassed.

[Laughter.]

All right?

We're going to start out with Dr.

Rosenberg, on the Critical Path Initiatives--the

Division of Therapeutic Proteins' perspective.

Office of Biotechnology Products--Current

Research and Future Plans

DR. ROSENBERG: It's a pleasure to talk to

you about our perspective on Critical Path.

So--I think it's important to start with

why--how the Critical Path Initiative evolved. And

it evolved, of course, because of the dramatic

decrease in novel drug and biological product

license applications.

[Slide.]

so what you can see here is that, from the

mid-'90s there's been a steady overall decrease.

250

And more than, I think, just the decrease

in numbers, we've really had a failure to develop

therapeutics and vaccines to address difficult

diseases. There's some diseases for which there

hasn't been an improvement in therapy for over 30

years.

So, coupled with this general decrease in

novel product development, there's really been, of

course, a high candidate drug failure rate.

[Slide.]

And it's pretty dismal to look at these

statistics. So, I mean, the last two--so a drug

entering Phase 1 in the year 2000 is less likely to

reach market than one entering Phase I in 1985.

And more sobering I think, in fact, is the fact

that about 50 percent of Phase III studies fail due

to lack of efficacy.

So there's really a lot of uncertainty by

the time many compounds are entering Phase III

trials. And Bob Temple will go on about why that

is, and how to improve that. But that's not the

subject topic here.

251

[Slide.]

So this isn't--this sort of dismal picture

isn't for lack of trying. What you can see in this

slide is that, in fact, there's been an enormous

amount of money and effort dumped into research and

development, starting in the early '90s, and that

it certainly outstrips, dramatically, the number of

new chemical entity approvals.

[Slide.]

So there are many factors that contribute

to this decline in new product applications. And

certainly one that has been cited is the failure of

novel methodologies and treatments to achieve

practical application. So, you know, all of the

wonderful technologies that have come up in the

past 10 or 15 years--many of them have really not

seen very much in the way of a practical

application.

[Slide.]

And I think getting industry's sort of

post mortem analysis on this is very important and

interesting. So, a comment from Roche was that "I

252

think we got too enamored of technology and lost

focus of what to do. The 1990s were really a boon

for in terms of science, but we forgot that we

needed to link all of that to disease."

And the second comment--from Adventis--"We

though we would very quickly validate targets that

were critical to disease and agonize or inhibit

them as a way to start to find a drug...and what we

found, in fact, is that validating targets takes a

lot of time. And this is one of the big

disappointments of this era."

So, I think, nowhere is this--I mean, it's

key that we have a sort of naivete about product

development. And I think this is what the Critical

Path is trying to address--to take this naivete, to

do some good science, and to perhaps shorten the

length of time it takes from a great idea to

commercialization.

[Slide.]

And I think nowhere is this better

illustrated than in the development of a product

that we regulate in the Office of Biotechnology

253

Products, and that is monoclonal antibody

development.

And this timeline is a little bit warped,

in the sense that it doesn't start at the

beginning; because the beginning of this timeline

is 1975, when Kohler and Millstein developed the

hypodermic technology that would make it possible

to produce monoclonals.

And so what you can see is there's about a

20-year lag period before you have a real flowering

of products. And so I think that Critical Path

asks a question, and that question is: can we

shorten this time?

And it's--I don't think it's an assured

thing, but I think it is certainly worth a valient

effort.

[Slide.]

So let's focus a little bit more now on

biotechnology products, and biological

therapeutics, which is the group of products that

our office regulates.

So, one of the reasons that there has been

254

a decrease in numbers of these products is that

there has been a dramatic increase in the length of

clinical development time. And you can see here,

from the 1980s, through 2002, you know just this

linear increase in development time. And that's

coupled with, pretty much, preservation of the

approval--the length of time it takes to approve

these products.

[Slide.]

And that differs from the case of small

molecular drugs, where in both the clinical phase

and review times have diminished or pretty much

leveled off since the early 1990s.

[Slide.]

So what is it about biological

therapeutics that has caused such a length in

development time? Well, for one, there's a

really--a big shift in disease indications since

the mid-1980s, late 1980s. More and more, chronic

diseases are being assessed. And, of course,

longer trials are necessary in the case of chronic

diseases, to both the assess the efficacy of the

255

product, but as well as the durability of responses

is key.

And even more important, I think, there's

been a shift to therapeutic products whose

mechanism of action and toxicities were less well

understood. So, what was encountered in these

clinical trials were unexpected and difficult

toxicities, as well as a difficult in developing

appropriate surrogate endpoints that would allow

for shortening and greater efficiency of clinical

trials.

[Slide.]

So how can FDA help? As I said, I think

Critical Path is a great tool to try and address

the enhancement in product development efficiency.

But I think it's important to realize that FDA and

industry still will have different roles.

According to this review, the ultimate

goal, of course, of FDA and industry is the same:

to provide patients with access to new, safe and

effective treatments. And what's really at stress

here is that coordination and cooperation are

256

required.

And the comment here is that FDA can only

assist in the process. And I think Critical Path

is trying to take this "only assist" into "assist

greatly."

[Slide.]

In addition, we're not the only partners

here--and this has been mentioned before. There

are other players: disease-specific advocacy

groups, NIH, CDC, etcetera. And the NIH recently

has launched their "Road Map," which is very much

targeted for drug development. And they have, you

know, three basic initiatives within this Road Map.

And so FDA is going to have to work, not only with

industry, but also with NIH, as well as other

advocacy groups in moving this--in enhancing the

efficiency of product development.

[Slide.]

Now, this has also been mentioned--but FDA

is uniquely positioned to identify and overcome

challenges to product development. Reviewers can

identify common themes and systematic weaknesses

257

across similar products, and that based on such

knowledge, reviewers can formulate guidance

documents and clearly offer industry sage advice

about pitfalls.

Now, I think it's worth it to mention that

guidance documetns have actually be shown to foster

product development; that they improve the changes

of an initial success of a marketing application,

and they shorten time to approval. So there is

research that verifies that, and so I think it's

very critical to have scientific personnel that can

promulgate very helpful guidances.

[Slide.]

So what are FDA strategies for speeding

innovate therapies to market? The first one was

actually in 2002, and it was called "Improving

Innovation in Medical Technology: Beyond 2002."

And this one particularly highlighted the

importance of guidance documents in avoiding

multi-cycle reviews.

And, of course now we have Critical Path.

[Slide.]

258

So the Critical Path, as we all have

heard, it's a method to develop new tools, to

imiprove predictions regarding safety and efficacy

of new products in a faster time at lower cost.

And it essentially supports

research--clinical and otherwise--for applied

sciences needed for medical product development.

[Slide.]

You've all seen this. Critical path goes

to some translational research, through to product

launch. But actually, in our view, knowing the

trouble that biological therapeutics can get into

following marketing, and following licensure, we

think it goes well beyond that, into post-licensure

phases.

[Slide.]

Again, Critical Path involves issues of

safety, efficacy and industrialization. And our

scientists, in the Office of Biotechnology

Products, are very expert in all of these

aspects--or certainly in targted areas of all of

these aspects of product development. And, as I

259

say, underestimated here is post-licensure issues.

[Slide.]

So, what sort of personnel does one need

to negotiate this Critical Path? Well, for

biological therapeutics we think that the

researcher/reviewer is ideally positioned to

advance the Critical Path. So a

researcher/reviewer is a sort of hybrid species;

this is a person who does a lot of regulation.

This person is a producdt expert. They're

absolutely integral to the regulatory process at

all stages of product development, and they provide

scientific expertise on multiple levels: product

manufacture, including inspections--all of our

reviewers go on inspections; product--this is an

expert in product characterization, including

mechanisms of action, in vivo bioactivity and

toxicities. The researcher reviewer is also an

expert in some analytical methods, and in some

animal modeling. But the researcher/reviewer also

has a key role in policy formulation and

promulgating guidances.

260

[Slide.]

So the basis for the regulatory expertise

of the researcher/reviewer is engagement in a high

quality research program. So the

researcher/reviewer is required to maintain an

active laboratory reseaerch program in the field

relevant to the review area. This person must

publish findings in peer reviewed, high quality

journals, and they must undergo site visit

evaluations of their program every four years, and

yearly internal evaluations. And, in fact, our

promotions are promulgated more on our research

expertise, and our research accomplishments almost

than our regulatory accomplishments.

[Slide.]

So, interestingly, this requirement for a

regulator who is intimately familiar with

cutting-edge technology is very much in sync with

findings that a subcommittee of the FDA Science

Board made back in 1998, when they said, 'It is the

consensus of the Committee that FDA requires a

strong laboratory research focus and not a virtual

261

science review process; otherwise we risk the

potential to damage not only the health of the

population of the U.S., but also the health of our

economy."

And I think the health of both are clearly

in danger when we can't get new products out.

[Slide.]

So this group also went on to say that

regulators and policy makers require expert

knowledge and first-hand experience with the latest

technology being applied to biological products;

and that an intramural research porgram is required

to assess risks of new therapies, to develop assays

and new approaches to increase efficacy and safety,

and reduce risks. It sounds a lot like Critical

Path to me.

Moreover, I think a very strong point they

made was that a strong, well maintained intramural

research program provides the basis for a climate

of science and scientific communication with FDA.

They emphasized retaining high-quality scientific

staff, but I think the permeation of science into

262

the review process is absolutely paramount.

[Slide.]

Okay, let's go on--just skip this.

[Slide.]

So let's go to my division--the Division

of Therapeutic Proteins. This may be too small to

read, but the only point I wanted to make is that

all of our reviwers--and we do have some full-time

reviewers--are spread among three laboratories:

the Laboratory of Immunology, the Laboratory of

Biochemistry, and the Laboratory of Chemistry. And

we think that this is in keeping with keeping the

culture of science permeated into the review

process.

[Slide.]

Our division regulates an enormous

diversity of products. We have 37 total licensed

products; we have 30 novel molecular entities. We

have many naturally-derived products--mostly

recombinants, however; and really very minimal "me

too" products. We have several interferons, for

example.

263

We regulate many engineered versions of

prototype products that are designed to enhance PK

or other product characteristics; pegylated

products. Many of our products have site-directed

mutagenesis for hyperglycosylation, as well as

other enhancements.

Our products are produced in very diverse

cell substrates; from bacteria, yeast, insect

cells, rodent cells, human, as well as transgenic

animals and, soon to be, plants. And the

manufacturing process is unique for each of our

products.

[Slide.]

So the products that we regulate--I think

you're familiar with: interferons, interleukins,

thrombolytics, anti-thrombotics, therapeutic

enzymes; all the ematolic growth factors,

neurotrophic growth factors; chemokines--which are

a novel area for us; wound healing products;

toxin-fusion molecules; angiogenesis and

anti-angiogenesis agents; immunomodulators,

receptor antagonists, lectins; and, most

264

importantly, I left off cosmetics. We also have

botox. We're very proud of that product.

[Slide.]

So what are the principal scientific

issues--and regulatory challenges--for us?

We've got a lot of them in our division.

Comparability is always a paramount issue, because

there are no analytical techniques that will

precisely define the 3-D structure of our complex

proteins, we have to use a variety of techniques to

establish comparabilitiy. And sometimes that

actually requires animal studies and sometimes

clinical trials. And we're engaged in a great

exercise of this right now, in our follow-on

biologicals initiative.

All proteins are potentially immunogenic,

and so we have problems with immunogenicity. We

have hypersensitivity responses, we have

neutralizing antibody responses. And these can

really blow up in product development.

Potency assessments--as I said, because no

analytical technique--and one--is good at really

265

defining the 3-D structure, we use a potency assay,

which is an activity assay which gives you a clue

about product protein conformation. And that

differs quite a bit, in some respects, from small

molecule regulation.

Our products have been the subject of

product counterfeit--on both Neupogen and Epogen.

And so we're working th the Office of the

Commissioner in formulating responses to that.

We've also faced novel transgenically

produced products. We're going to get products

produced in chicken eggs, as well as plants. And

those raise very novel safety issues--and efficacy

issues, as well.

And we're always faced with infectious

disease transmission because of the way our

products are produced, and the materials that are

used to produce them.

[Slide.]

So, as product experts, we have a very

keen knowledge of pitfalls in product development,

from pre-clinical studies to Phase I and II

266

studies; immunogenicity, unexpected adverse events,

lack of appropriate animal models. Certainly,

mechanism of action, when it's not fully evaluated,

can be very problematic.

[Slide.]

in Phase III, the development of validated

potency assays are a real pitfuall in product

development, as well as changing manufacturing in

the middle of Phase III studies, which really

wreaks havoc.

And so we really--you know, we spend a lot

of time with sponsors trying to stear them away

from these pitfalls. And I think you'll see that

our style of communication is highly valued by

industry, who feels that it's, in fact, vital for

more efficient product development.

I'm just going to skip over some of the

clinical ones.

[Slide.]

So, our Critical Path focus for our

division is basically to support ongoing Critical

Path projects. And we think of those as pertaining

267

to entry of products with novel mechanisms of

action--and that would encompass research that

investigates mechanisms of action of new products;

research that establishes new animal models for

assessment of safety and efficacy; and research

that provides new or improved products to the

piplines.

[Slide.]

Moreover, we recognize very well the

barriers and hurdles to product development,

including immunogenicity and potency assessment.

And so we value research that overcomes these

barriers to product development; moreover,

activities to standardize assays--this is very

important when you're trying to compare across

different products.

Moreover, the last type of research we

think is highly critical-path appropriate is

identification of surrogate endpoints and

biomarkers for safety and efficacy. And so we

really value research that identifies novel

biomarkers, as well as activities to gain consensus

268

on appropriate surrogate markers.

[Slide.]

So, some of the programs that we have

really very much addressed directly with Critical

Path issues: one of them is the development of CpG

oligonucleotides as immunomodulators for infectious

diseases. Daniela Verthelyi is the principal

investigator, and so she investigates CpG

oligonucleotides as they interact with toll like

receptor, as well as other potential toll like

receptor ligands. And she studies primates; she's

interested in identification of surrogate markers

of immune protection, and development of novel TLR

agonists. This project also has high relevance to

bioterrorist situations; can we enhance the immune

response by fiddling with these toll like receptors

to bioterrorist agents?

[Slide.]

The second project that directly addresses

Critical Path issues is a research project that's

focused on chemokines, which are chemo-attractant

cytokines. And we're ver increasingly coming to

269

appreciate the fact that these products are

absolutely critical for cell migration in the

seetings of inflammation, metastasis, angiogenesis,

and atherosclerosis. Mike Norcross is the

principal investigator. And, within his research,

he is developing methods to assess the potency of

these products. Potency, as you can imagine, is

very difficult to assess for a product that's a

chemo-attractant product. Those are very squishy

assays; very variable. So, this has been a real

problem in product development.

He is, as well, trying to evaluate and

develop methods for non-clinical screening of

anti-viral biological products, as well as the

development and validation of biomarkesr and

surrogate endpoints for immune-based therapies for

HIV infection.

[Slide.]

And just to show you a little bit of a

schematic here--so you have bacterial products,

such as LPS, or CpG oligos that tickle toll like

receptors that are present on macrophages and

270

dendritic cells--antigen-presenting cells--that

cause them to emit chemokines, such as IL8,

MIP1-alpha, IP10, and these cause chemoattraction

of various immune mediators, as well as cause

trafficking of tumor cells to distant cites.

So it's a very exciting area, and I think

having such expertise is critical to the product

development.

[Slide.]

Dr. Donnelly also has a program which we

think fits directly into Critical Path. He is

focusing on signaling pathways of novel

interleukins and inferons; specifically, he's

defining signal transduction pathways for new

cytokines, new interleukins, ILs 19, 20 and 22, as

well as defining biological properties of a new

interferon, which may be significantly less toxic

than interferon-alpha. It's called

interferon-lambda.

[Slide.]

Dr. Beaucage--who many of you may

know--world-class chemist--basically has a program

271

to enhance the specificity and sensitivity of

oligonucleotide microarrays which, of course, are

used for myriad purposes. And so he has focused on

detection and quantification of bacterial and viral

nucleic acid contaminants in biologicals, including

blood products. This methodology would be helpful

for high-throughput screening of point mutations,

or single-nucleotide polymorphisms that might

dispose to human disease. And, of course, these

are used widely as gene expression assays to

evaluate potentially the safety and efficacy of

drugs.

[Slide.]

So, those are the projects we think are

directly relevant to Critical Path.

Others, I think, we conceive of as being

supportive of Critical Path; perhaps not as highly

targeted, but nevertheless, absolutely vital to

product development.

So, Dr. Shacter's program, and Dr.

Johnson's program are focused on novel anti-cancer

treatments. With Dr. Shacter, modulation of signal

272

transduction pathways to enhance tumor cell dealth

in response to chemotherapeutic agency, and the

investigation of antioxidants as potential

chemoprotective agents to limit side effects from

cehmotherapy. And Dr. Johnson is focused on

enzymology of epidermal growth factor receptor

signaling, as well as identification of novel

signaling molecules.

[Slide.]

Many of our programs are immunologically

oriented. And as I said, immunogenicity is a

critical issue along the Critical Path.

So, all of our proteins are potentially

immogenic. As I said, we can get hypersensitivity,

anaphylactic-type responses, or IgG antibodies that

will neutralize a therapeutic protein, or block the

action of an endogenous homolog of that

therapeutic. And immunogenicity has killed products

in development; certlain from epoeitin, CNTF,

GM-CSF-IL-3 fusion molecules, as well, it limits

the efficacy for many giological therapeutics, such

as therapeutic enzymes, interferons alpha and beta,

273

and asparaginase.

And it poes an ongoing concern for

licensed products followoing changes in

manufacture, packaging and clinical indication.

And I think most of you are aware of the situation

with Epo and the induction of pre-red cell eplasia,

due to changes in the packaging of Epresx.

AS well, there's a lack of standardized

assays for comparison across products in the same

class. And this is a problem.

[Slide.]

So I think, you know, for immunogenicity

most of us conceive of it as being capable of doing

the following, which is to block the development.

Actually, interesting--it was supposed to blow up.

So my Papa Haydn slide didn't work very well.

[Slide.]

So, the immunogenicity concerns and the

projects that address this have to do with

understanding the mechanism by which antibody

responses to proteins are switched to cause

anaphylaxis. And this also will have, I think,

274

some meaning for small molecular drug development,

because they are not without their hypersensitivity

response; research to develop better animal models

to assess immune tolerance and autoimmunity;

research to dissect immune responses to embryonic

stem cells; and we are also participating in

international efforts to standardize antibody

assays for erythropoietin products.

[Slide.]

Some new Critical Path projects that we

foresee, looking into the future: nanotechnology

is being highly toutedfor potential abilities to

deliver productsi n novel ways. This may also

actually present big problems immunogenicity for

vaccines that many of these approaches might be

terrific in enhancing immunogenicity, but they

could be devastating for therapeutic protein

products. And we think this is worth investigating

so that this technology--at least for biological

therapeutics is not stopped prematurely.

For therapeutic enzymes, the immune

response does limit efficacy, particularly of

275

life-saving products for patients who lack some

endogenous enzymes which are critical for life.

And so we think that tolerance induction should be

explored in that setting.

Protein aggregates are a perpetual problem

that induce immunogenicity. However, the

specifications for aggregates are essentially set

on manufacturing experience, not on risk. And so

we think it would be critical to evaluate the risk

of protein aggregates. What level of aggregates?

What kinds of aggregates? And how are they

delivered? What is responsible and what is

important in incurring risk?

And also the development of buidance

documents we think would be a very valid Critical

Path project.

[Slide.]

As well, some of our research--out of some

of our research has come an idea for a novel

product which would promote treatment of sepsis,

which is a disease that is notoriously refractory

to treatment. And Dr. Shacter's lab has identified

276

protein S as being critical for many functions,

among which are clearance of apoptotic cells. But

since activated protein C works in conjunction with

activated protein--with proten S, we think that our

research suggests that addition of protein S to the

treatment protocol that uses activated protein C

will improve efficacy. And so we would like to

develop that as a therapeutic protein. Of course

we would like to get that to a commercial entity

that would develop it.

[Slide.]

I'm coming to a close now. But we also

think that communication is a critical component of

Critical Path. And an industry survey done last

year that looked at good review management

practices, found that the kidns of communications

we had--and that were alluded to, I believe, in an

earlier talk--that is open, honest communication;

informal communiations; regular status updates;

timely communication of issues as they arise; and

clear and concise FDA responses with explanation of

positions--these were all review practices while we

277

were in CBER, and we have carried over to CDER, and

we certainly hope that, given that communication is

vital, that these will be carried on.

[Slide.]

And so I will skip through this. You can

read through it yourselves.

[Slide.]

Other DTP Critical Path activities involve

participation in ICH proceedings; and particularly

with regard to to comparability guidance. So Dr.

Cherney, who is the Deputy Division Director is the

lead on the ICH !5e, and so, again, the importance

of guidance documents can't be overemphasized, in

terms of enhancing product development efficiency.

Another one of our personnel, Dr.

Kirschner, is involved in standardization of

antibody assays for erythropoietin products, which

is an international effort. And, moreover, the

suport of risk-based approaches to GMP and

inspectional issues is something that we also think

is a vital Critical Path activity. We need to

switch from checklist approaches to GMP, to

278

risk-based approaches. And we're strongly

participating in that.

[Slide.]

So, in summary, DTP strongly supports

Critical Path efforts to facilitate development of

new products. We think that we have some projects

that are doing that now, and should be better

supported. We have identified new projects that we

think should be funded to enhance this process.

Other activities, including the

development of guidance, adoption of a risk-based

approach to GMPs, and maintenance of communication

form at with industry we also think are vital.

So--I'll end with that. And I hope I

didn't go too much over time.

CHAIRMAN KIBBE: Thank you very much.

Outstanding! All right.

You actually have allowed us five minutes

worth of question time. And we'll let Meryl have

it all.

DR. KAROL: Okay, thank you. That's a

very impressive summary of what you're doing.

279

I wondered if you're placing emphasis on

development eof biomarkers for not only

immunogenicity, but hypersensitivity, tolerance as

well? You know, could you tell us about those

efforts, to develop biomarkers to predict these

effects?

DR. ROSENBERG: Yes, I think--you know, we

do a lot of animal modeling. And so most of our

programs have to do with rodent models, and looking

at tolerance, and looking at immunogenicity,

particularly--Dr. Verthelyi's program.

Now, she has taken this one step higher,

to primate models, in trying to come up with

surrogate markers. And, you know, I think this is

something that we're putting an emphasis on. I

don't know that we have a real formal look at that

at this point, or we can really report on that.

But that is something that we would like to

emphasize better.

CHAIRMAN KIBBE: Ken? You really--

DR. MORRIS: I really have a question. I

really have one.

280

CHAIRMAN KIBBE: well, go ahead.

DR. MORRIS: But it's not as technical, I

don't think.

Given the Tufts projections, as well as

the statistics you showed on success, have you

attempted to factor the contribution of the various

thrusts that you're pursuing to determine--in terms

of prioritization?

DR. ROSENBERG: So--in terms of--yes. I

mean, I think that what we're trying to emphasize

are aspects that have been proven to do something.

So, certainly, communication is a critical aspect,

and development of guidance is--has been shown--

DR. MORRIS: Yes, actually, I guess I was

thinking more about your research thrust, but--

DR. ROSENBERG: Yes--so those are

emphasized.

The research thrusts--yes, I think that

what--you know, what we're looking at here is the

research projects that we have that absolutely

address Critical Path issues we would like to

expand. Of course, resources are limited, and

281

there's just a certain amount we can do. But the

one's we've identified I think are absolutely

critical for these novel emerging technologies.

And, as we've seen, you know, people can be very

naive about what one can expect from those.

So looking--you know, having looked at

that, and looking at what's coming ahead, we would

like to investigate, you know, the immunogenicity

concerns for nanotechnology. We would like to

look--you know--we would like to be able to look at

that. I can't do that with the personnel

limitations I have now. We would need funding and

personnel to do that.

So--as well as, you know, bioterrorism is

a very important factor now, and we think we have

the ability to address a treatment for that, which

would be the CpG oligonucleotides, or some similar

pathogen-associated molecular pattern ligand.

So those, I think--we have good models,

and we would like to push forward on those in

particular.

DR. MORRIS: Thank you.

282

CHAIRMAN KIBBE: Melvin, you want to--

DR. KOCH: Yes, just--excellent

presentation. I was just wondering, in many of the

needs you expressed--and they sound like ideal

candidates for CRADAs. And has that been explored

at all?

DR. ROSENBERG: Yes. We certainly try to

develop those where we can. It's a little tough,

given some constraints. Because, of course, as

soon as you develop a CRADA, you know, you're

limited in what you can participate with, in terms

of regulatory action. So, you know, you're always

sort of caught between a rock and a hard place.

But--yes, we are trying to develop CRADAs--for some

of those projects--and have been successful. Dr.

Beaucage has been successful, particularly, with

the micro-arrays over the years, and getting

CRADAs.

CHAIRMAN KIBBE: Thank you very much.

DR. KAROL: One more question? How are we

from developing SAR models for protein

allogenicity? Is that at all on the horizon?

283

DR. ROSENBERG: Whoa! That's a very good

question. And I don't know the answer to that. I

really can't tell you. I think I would have to

talk to somebody who's more of--more focused on

allergy.

CHAIRMAN KIBBE: Thank you very much.

DR. ROSENBERG: Thank you.

CHAIRMAN KIBBE: Next, we have Steve

Koslowski. Sever has 64 slides--

[Laughter.]

DR. KOZLOWSKI: Oh, I'm already in

trouble.

CHAIRMAN KIBBE: You're under the gun,

Steve.

DR. KOZLOWSKI: Well, thank you for having

me speak. And I will try and move quickly.

[Slide.]

So I'm going to talk about the Division of

Monoclonal Antibodies, which is the other

biotechnology product division. I'm going to talk

a little bit about quality, and I'm going to kind

of take the lead from one of Ajaz's slides, and

284

talk about connecting the dots; then about a

concept called biological characterization; the

reserch reviewer model, which we can go through

quickly, because Amy already covered that; the

organization of our division--our products; ongoing

research' Critical Path; and then sort of

summarizing Critical Pathways and directions.

So I want to put up a slide that Ajaz gave

me, about a way of looking at integrated quality.

[Slide.]

The fact that different disciplines in

review from clinical to manufacturing TO CGMPs to

PAT all need to be interconnected in a useful way.

And I'd like to take a little bit of a slice of

that figure--

[Slide.]

--and actually take a way a lot of points,

and basically leave the CMC relationship to

clinical attributes, and talk aout connecting one

dot to begin with: the chemistry of a product--or,

basically, its complete structure, to those things

that we control in evaluating it.

285

So, clearly, the characterization of a

product leads to what we eventually use as

classical specifications-0-or at least how we've

talked about productsi n the past.

[Slide.]

And there are ICH guidelines on this. You

need to characterize a biotech product in order to

pick the relevant specifications that you use for

quality control. You choose these specifications

to confirm quality--and obvious, you don't

recharacterize the product each time. But what's

critical is those molecular and biological

characteristics that are necessary for connecting

to safety and efficacy.

[Slide.]

And I think those are really the weakest

links, because the connections between what

structure really matters for clinical outcome--what

attributes are important--and what controls in

manufacturing, or what controls in regular testing,

confirming these things is a very hard link to

make.

286

And clearly you need to know these

relevant structural attributes to take advantage of

this CGMP, or more global way of looking at things.

So, again, what processes you need to control; what

structural attributes are important.

[Slide.]

That leads me to what I'll call biological

characterization.

So I'll start with talking about our

molecules. Amy certainly referred to the fact that

the biotech proteins--or products--tend to be very

large. And this is an example of a third of a

monoclonal antibody--an Fab section compared to a

statin.

[Slide.]

And so, clearly, the large molecule has

issues, not only of primary sequence, but higher

order structure, post-translational modifications,

and it is a very heterogeneous protein. In fact,

the variability in proteins--in fact in the desired

product--are greater in size than the size of a

statin.

287

And, again, comparing molecular weight:

150,000 to 400.

[Slide.]

So this leads to a problem--as Amy pointed

to--for complex molecules--as again, in the ICH

guidance--physiochemical information at least

presently insufficient to define higher order

structure.

[Slide.]

And so what we use--and it's an imperfect

thing--but we use biological activity as the

surrogate, sosrt of, for full biochemical

characterization.

And so biological actiivty is specific

capacity of a product to achieve a particular

effect. And potency is the way we measure that.

We use a variety of bioassays: animal based, cell

culture based, biochemical--sometimes receptor

ligand binding.

[Slide.]

There's a whole continuum of these assays,

which go from very simple assays to ones that are

288

very complex. The complex assays--like, ideally, a

clinical study--is true potency, but its

reproducibility and its utility as an assay is very

poor. On the other hand, simple assays are very

useful from a validation perspective, but may not

really reflect what you want to look for.

[Slide.]

So how do we choose the relevant biologic

activity as a surrogate for structure? So

assessment of bioological properties is an

essential step in the characterization. And so, by

characterizing the biological responses that are

generated by the product, one should be able to

pick a good assay.

[Slide.]

So just like you characterize the

structure to pick the physiochemical attributes,

you need to characterize the biological effects to

pick a potency assay, and to define those

characters that ensure safety and efficacy. And,

again, defning those characteristics--what

attributes really matter--are crucial to the ideas

289

of risk management, CGMPs for the 21

st century, and

PAT. Because if you don't know what to control,

you can't control it.

And that makes it also relevant to small

molecules and achieving this ideal state where

everything is connected, and you can avoid a lot of

testing at the end, and you can truly know your

process.

So I want to touch on two quick things:

molecular mechanism of action, and biological

plausibility.

[Slide.]

Molecular mechanism of action is--again,

you need it for potency assays for therapeutic

proteins. But for all CDER products, it will help

you pick relevant physiochemical properties;

sometimes predict toxicity, drug interactions and

efficacy; and can be useful in choosing animal

models and clinical monitoring early on, when you

don't have enough data to really know what a

protein or product is going to do.

[Slide.]

290

biological plausibility--we talked about

biomarker development and validation. You need to

be able to interpret early pharmacogenomic and

proteomic data. When you have a large enough

study, statistical data may be good enough. But

when early on you need to make a decision that

involves product development, biological

plausibility is a critical part of assessing a

biomarker. And one of the only ways to do that is

to really understand the mechanistic issues, and to

say that this marker or this gene makes sense.

[Slide.]

So if biological characterization is so

critical to these issues, why is there such little

guidance on how to do it?

If you look at the guidance of

end-of-Phase 2 meetings, it talks about having

"adequacy in physiochemical and biological

characterization." The term is used. However, if

you look in the parentheses: "peptide map,"

"structure," glycosylation"--no mention of what

biological characterization is.

291

Later on, it talks about "bioassays," and

metnions using a variety of materials in the

bioactivity assay, not just the product

itself--which is the beginning of biological

characterization.

[Slide.]

What you would ideally want--and, again,

this is difficult to do, and we're not saying that

this can be done or should be done--but binding of

the product; singal transduction pathways; cell

culture effects; tissue studies; and in vivo

studies--and sometimes multiple studies, because

the same protein or same product can have multiple

active sites.

To do this, you need relevant models.

That means you need the right receptor, the right

pathway, the right cells, tissues and species. To

pick those, you need to know the molecular

mechanism of action. However, if that's how you're

defining it, you have a circular problem. It

really is difficult to do this. There's no linear

algorithm to really biologically characterizing

292

something. And so, again--I'll use another term

from Ajaz--you really needs a systems approach.

You need a way of dealing with this information to

allow you to get the attributes that can allow you

regulatory relief from controlling them.

And there's also product specificity.

There's a lot more variability in a lot of these

biological assays. And it's very expensive.

[Slide.]

So one approach--and we have companies

who've actually done this--is to sort of have a

matrix. So for--and I'll talk about using one or

so lots, and using many lots for some of these

things.

So in initial development--the lots very

early on--you might look at multiple in vitro

assays, and really get a good feel for what your

developmental lots do; move on to testing some of

those in more complex animal assays--transgenic

models, sophisticated models that really try and

target the relevant attributes. And then, again,

in the end, when you have a validated bioassay, it

293

would be good to go back and look at all the lots.

Stressed lots--similar testing

plan--because this is likely to start giving you

variants that you can define as important or

unimportant.

And then for some of those variants you

might purify them, and then repeat some of othis

testing.

In clinical lot manufacture, you're always

going to have some lots that are at the extremes of

the ranges. And use of those lots in some of these

assays can also help you define this.

Ad, finally, the clinical lots, you know,

should be looked at in the validated bioassay and

sosmetimes in some of these other assays.

So having a sort of matrix approach to

what you're looking at may help define the

information that you need to help you avoid

retesting and looking at all these attributes.

[Slide.]

To do this, there needs to be biology

expertise. A biological characterization is only

294

as good as the data that supports it. Regulatory

decisions are impacted by this sort of

characterization. There needs to be a framework

for interpreting this data, interpreting the

assays, and defining what's needed.

And this expertise is going to become more

important over time; in fact, it may be useful to

actually have a guidance for how to approach

biological characterization for some of these

materials, and a mechanism for consulting people

with the right expertise in order to do this.

And with the recent consolidation, CDER

has now got some additional expertise in cell and

molecular biology, which could play a role in somse

of this.

[Slide.]

And, now, we talked about the research

reviewer model--basically, research reviewers do

both jobs. It's challenging. We're judged on

productivity. We have to go through site visits

and tenure committees, and we have the difficulty

of multiple workloads.

295

On the other hand, the research reviewer

model can serve as a form of catalysis and synergy,

because basically we know not all reviewers can

have active research program. It's economically

unfeasible, and it doesn't make sense.

[Slide.]

But if you have a small nucleus of

research reviewers, they can help encourage some

issues in biochemical and biological

characterization, process understanding, and

mechanism. And they can consult on key decisions,

and they can also network to NIH and other acadmic

groups, OTR staff, and the full time review staff.

[Slide.]

Research is organized in funny ways. So,

if you take disciplines like immunology, tumor

biology, neuroscience and developmental biology,

there may be people who have expertise in

cytokines, or cell hormones related to this;

adhesion related to these cells; and

differentiation or in signal transduction. And

there's a kind of a matrix. And you really can't

296

cover everything. But if you have a number of

researchers--as Amy talked about in her division,

and we have in our division--who cover a lot of

these areas and things, you can often find points

of intersection. And those points may involve

research related to your question; NIH journal

clubs that people participate in; academic

conferences; and, finally, the literature

itself--but this variety of networking that gives

you access to information.

[Slide.]

So, briefly, about our organization. So

we have three divisions: Molecular Development and

Immunology--Margie Shapiro's the lab chief; the

laboratory of Cell Biology--Kathleen Clouse is the

lab chief; and the laboratory of Immunobiology--and

I'm the lab chief. And each of these have three

principal investigators. They look at lymphocyte

and monocyte biology; tumor suppressors and

oncogenes; cell-cell and cytokine-receptor

interactions; signal transduction; and antibody

interactions--which are very relevant to our

297

products; and manufacturing process validation.

[Slide.]

And our products--if you look at them in

terms of indication--they tend to be either

immunology or inflammatory-related or

oncology-related. And therefore having expertise

in immunology and tumor biollogy is very relvant to

our products.

We have a number of approved products that

relate to immunology or inflammation; some of them

that share targets. We have to CD25s; we have a

variety of isotypes--different species of

antibodies, and anti-infective antibody products

against cancer--again, some of them share targets

like CD20.

[Slide.]

Our reviewers participate in inspections.

We have imaging agents that are radio-labeled, and

we're involved in developing guidance

documents--points to consider for monoclonal

antibody; plant transgenic products; orphan drug

status-monoclonal antibodies. And we're also

298

involved in Q5e, although Barry's the lead on that.

And we're involved in follow-on proteins.

[Slide.]

So, we have a research program. And this

I'm going to have to go through extremely quickly.

So, we have groups that have studied

particular chemistry. What I'd like to focus on is

antibody structure.

[Slide.]

This is a schematic, based on crystal

structure diagrams of an IgG molecure. The V

region on top, with the DRs are the variable

region, with a binding site which is a part of the

antibody that leads to finding its target.

There are variety of other regions in the

molecule which bind effector molecules like Fc

receptors, complemin, and a variety of other

receptors that mediate effector function. So the

antibodies have lots of different active sites.

They may be relevant for some things and not.

They're also glycosylated, some versions a lot.

IgG1 tends not to be as much, but this is also

299

relevant, in some cases, to PK and to effector

functions.

[Slide.]

So we have Margie Shapiro's lab that looks

at some of the things that generate antibody

diversity. There are new technologies in making

antibodies, like phase display, transgenic animals

that express human antibody genes. They may lead

to different binding sites--different diversity.

They're not.

[Slide.]

And if you look at immunogenicity of

antibodies--murine bio-similar, of which this looks

at 8--more than half of the patients who get them

develop antibodies against them no matter what.

They're highly immunogenic.

If you take away the Fc region, and just

have the top half binding site of the antibody,

that immunogenicity goes down. If you make the Fc

region human, and you leave the variable regions

mouse, you find that the immunogenicity also is

between 1 and 13 percent. Again, as Amy said, it's

300

almost impossible to judge these comparatively. So

you have to take this with a grain of salt, because

the assays vary.

But if you humanize the antibody--you make

all of it human except for the binding site area,

and some other amino acids, the immunogenicity also

is low--maybe a bit lower. If you take a fully

human antibody, which is one example of bi-phage

display, it actually doesn't have a lower

immunogenicity.

So the question really is: is the

technology of making these antibodies relevant to

how immunogenic they are. And she--and her lab is

studying this.

[Slide.]

Antibodies have effoctor interactions.

Complement recptors play a role. There's a new

family of Fc receptors--it was found through the

genome--which isn't known how it functions. And we

have Dr. Mate Tolnay, a new investigator, who's

going to look at whether or not those Fc receptors

play a role in antibody function.

301

And Dr. Gerry Feldman has looked at immune

complexes and how they signal responsiveness to

cytokines. Again, that may play a role in how

antibodies work.

[Slide.]

Now, in terms of the biological

characterization--I think I'm going to try and just

skip this. We have a lot of different projects

related to lymphocyte signaling, on HIV, sustaining

in reservoir; the EGF receptor--and all these

projects relate to products that we have.

So if you look at adhesion costimulation

molecules, we have a licensed antibody AGAINST

LFA-1. And that information is useful on how its

potency assay was looked at.

[Slide.]

We have antibodies herceptin against

tumors which signal through a molecule called Cbl,

and we have someone who works on that.

[Slide.]

And, again, I want to talk briefly about

Wendy Weinberg's project. She looks as skin as a

302

model of differentiation, and is interested in

p53--but not classic p52, but new members of this.

So here's an example of cells growing under low

calcium, and these cells have not differentiated.

If you increase the calcium concentration, they

differentiate; you see there's no more contact;

they're much less likely to grow; and there's a

decrease in the amount of proliferation. The S

phase is down by 43 percent.

But if you add a variant of a p63 gene,

which is a p53 family member, that no longer

happens. They continue to grow, despite the fact

that you're induced differentiation.

And the question about what these family

of genes do in cancers is relevant. And I'm going

to talk about that in a moment.

[Slide.]

We also have research regarding controls

and manufacturing process--and relating to

contaminants and process understanding.

[Slide.]

So the slide here is--you know, "This is a

303

brain; this is a brain on prions." You can see the

spongioform degradation. And this is an image of

how the prion protein is changed in conformation in

order to cause disease.

But it turns out peptide's a prion signal,

and they signal through the NF-kB pathway, and they

signal through inducing cytokines. And they have

different effects on different cell types.

And so information on this is useful in

designing, potentially in the future, cell-based

assays for prions--although they're not nearly

sensitive enough to do that now--and potentially

looking at the mechanism of the disease of this

common contaminant.

[Slide.]

Kurt Brorson, who works with Kathleen

Clouse's group, has made studies on retroviral

testing, using Q-PCR-based assays, which are much

easier and faster turnaround town; process

understanding, in terms of the unit operations, the

purification, chromatography.

[Slide.]

304

And here's an example of a bioreactor used

to produce many of our products. The question is:

what things can impact retrovirus expression? And

his work has shown scale and nutrients don't seem

to matter, but inducing agents and temperature do.

And, again, butyrate increases the expression of

retrovirus and increasing temperature does.

[Slide.]

Critical path--so, again, three

dimensions--we've all seen that.

[Slide.]

In terms of Critical Path projects that

should be defined as Critical Path--so you really

need to define a problem, state the dimensions, and

point out why the FDA--if the FDA's going to do

this research--is in a unique position to do so;

and what benefits go to what industry segments, and

the role we can play, and the impact of the

solution.

[Slide.]

So I'm going to go through this really

quickly, because there are a number of

305

investigators who I think have very clear Critical

Pathways. And one of them is the fact that anthrax

toxin is potentially a target for treating and

prophylaxis of anthrax, which is clearly a problem

we're all familiar with. But the bioassays for

anthrax toxin tend to be murine cell lines. And

they die. And, in fact, human cell lines do not

die from anthrax toxin. So is this r eally the

right model to be looking at the efficacy of things

that block anthrax toxin?

[Slide.]

And obviously this has medical utility and

industrialization issues, and also

counterbioterrorism. And the unique position of

the FDA is, is we're the only group that sees all

the INDs for anthrax therapeutics. And there's

another unique aspect here is the FDA also plays a

role in talking to the groups--like the CDC--that

are involved in BioShield. So the FDA's in a sort

of a funny role, because it's not only a regulator,

it's in some ways a stakeholder, since the

government is, you know, buying these things to

306

stockpile at some point in the future.

And, again, so David Fruct, who's doing

this, has shown anthrax lethal toxin activates a

particular pro-inflammatory cascade involving

cytokines. There's been a huge debate in the

literature of the role of cytokines. And I think

he's provided strong evidence that they do matter.

And he's also been able to show some

effects on human cells using this and, in fact, an

enzyme that drives this.

[Slide.]

And I have a quick schematic. So anthrax

toxin is composed of three components: the first

one, PA needs to bind a receptor. It forms a

hexomer. It then translocates the other toxin

units into the cell--this one, lethal factor,

effects MAP kinases, which are important to signal

transduction. And these lead to cell death.

But David Fruct's lab has also shown they

lead to IL-1b and IL-18 release. How this relates

to pathogenesis is unclear. But it already

represents a potential marker you could have for an

307

assay, for blocking the effect.

And he's also done some studies showing

some effects in human cells--which, again, might

make a more relevant bioassay.

[Slide.]

Wendy Weinberg--I showed you briefly her

slide, about looking at p53-like products. And I

think there is a huge lacking, in terms of goo

preclinical models to predict treatments for

cancer. have at least a number of products in

which Phase III studies were done for the wrong

indication, and later worked when the indication

was shifted.

So, clearly, the preclinical models used

to choose that first Phase iII study were in error.

And making models where you have mice, where you

have p53 knockouts, p63 knockouts--a variety of

mice in which you can mimic how human cancers

develop based on what genes are knocked out in them

might make a much more powerful way of picking that

first indication.

[Slide.]

308

And, again, Kurt Brorson, who looks at

process-related things--and I'm just going to skip

to a picture--

[Slide.]

--so a lot of our process uses a viral

removal steps, including nanofiltration. So we

filter away viruses. So how you test, and how you

validate these filters is tricky. You certainly

don't want to test the filter with a virus, if you

can avoid it. It's more difficult to do, and you'd

need containment procedures. And it's cumbersome.

But you wouldn't like to depend entirely on things

like gold particles, or a very poor surrogate for,

really, the ability of the filter to remove

viruses.

So Dr. Brorson's involved in using a

phase--a bacteria phase--which is easy to grow, in

testing these filters--and a good mimic for large

viruses. And so the phase he uses is PR772, has

been purified here by cesium chloride gradient, so

you don't get clumps. Clumps are very misleading,

because they look like they're cleared when, in

309

fact, your filter really can't filter out viruses

of the right size. And, again, this is showing the

purity by cesium chloride preps.

So, again, this has the potential to make

a better way of testing these filters and showing

that they really do the job they do.

[Slide.]

And there are a number of other

industrialization-related projects, in terms of

using gene arrays to look at cell culture changes.

We're also interested in databases of some of our

manufacturing experience. We sort of wanted a

comparability database for a long time, but it's

been a little slow to go.

[Slide.]

So to sort of summarize, in terms of

Critical Pathways, historically, cell and molecular

biology have always sat with basic research. But I

think now they have evolved so they are involved in

looking at clinical outcomes, in terms of

pharmacogenomics, and proteomics, and they're

involved in industrialization, because they offer

310

more sophisticated ways of measuring industrial

processes. And they're clearly important in

choosing the right pre-clinical development, the

right potency assay, and quality issues.

[Slide.]

And, again, by defining the biology of a

system better, you can pick the relevant

physiochemical properties--and that's critical for

cGMP and PAT--certainly, for our complex proteins.

There's a potential for this to affect toxicity,

drug interactions and efficacy, and even pick

early-in-development models. We've had cases

where, based on a biological effect that we would

predict from basic science about a protein, we've

talked to the clinical reviewers and we've said,

"Well, maybe this should be an exclusion criteria."

Or, "Maybe you should think about looking at this."

So this information really is relevant at

all stages of development--and, again, plays a

critical role in process validation and regulatory,

fewer failed studies.

[Slide.]

311

So, I think a critical direction is really

to better define "biological characterization."

Clearly, we're not asking industry to test every

lot of everything they've ever made, in every

animal model you can think of. But the point is,

by having good characterization of the mechanism of

action, using a variety of models, sthat really

leads into the fact that you can say "this

parameter," "this glycoform" doesn't matter, and

therefore avoid problems when you have

comparability issues--you know, in terms of a

difference there, and reduce, potentially, in the

future, the actual specifications you have.

Again, ideally, we'd want a guidacne on

thsi. It's--again, because it's non-linear it's

ckind of ocmplicated to think about how to do this.

And this plays a critical role for follow-on

proteins. The better you can characterize the

mechanism of action, the more confidence you are

that a follow-on protein is going to do what you

think it's going to do.

Again, we'd like to maintain this

312

biological expertise. We'd like to have research,

you know, across the relevant areas for out

products--which I'm calling "Critical Pathways,"

because it doesn't necessary fit the A, B, C, D, E,

F of Critical Path.

We'd like to facilitate access to OBP

biologists; interactions with other offices, with

pharmacology and clinical review groups. We

actually briefly had a conversation yesterday, in

terms of the pharmacogenomic review process, could

we play a role, in terms of helping define

mechanistic questions about correlating markers?

Again, Biotech Rounds with OBP and other clinical

groups; and mechanism of action journal clubs,

potentially, to talk about this; bioprocessing

journal clubs--and, again, eventually mechanisms

for consults on these issues.

[Slide.]

So--we also want to extend OBP into

Critical Path projects, like some of the ones I've

mentioned.

We talked about computers and

313

e-regulation. And I think relational databases are

a verty useful thing. We have, in our division, a

database Kurt Brorson is setting up for viral

clearance; a database that Patrick Swann, our

acting Deputy Director, has for review management,

USAN names and targets; one for specifications; one

for the risk of TSE. We have databases now we're

trying to capture internal meeting summaries;

workload databases; and, ideally, for monoclonal

antibodies, wehre there's tremendous similarity

between them, structural sequence information that

we could compare between our products, and link to

adverse events, would be very useful.

So, it seems all these databases--some of

the Excel spreadsheets, and some of them more

sophisticated--if we got somebody to make them a

relational database, where you could work all the

way through, that would be a very powerful tool,

and potentially aid in the Critical Path.

And, again, our ultimate goal would be to

use biological information, our research and our

regulatory review, to enhance safety and facilitate

314

regulatory relief.

Thank you.

CHAIRMAN KIBBE: Impressive.

Questions?

[Pause.]

Either we were all so impressed, or we all

need a break.

DR. KOCH: Well, I have a quick question.

If I understood correctly, where you have

the industrialization intersecting with the

Critical Pathway, you're inferring that's something

like surface plasma and resonance, or something

else that will actually be an interrogation of the

process?

DR. KOZLOWSKI: Well, again, I think--for

instance, surface plasma and resonance, for

instance, you can sort of do--I guess not in a

line, but you could look at binding off a process--

DR. KOCH: Right.

DR. KOZLOWSKI: --by filtering your two

to a bio-core chip.

Yes--I think that would certainly--that

315

would be very PAT-like, to actually look at--

DR. KOCH: Right--that's what I was--

DR. KOZLOWSKI: --the binding of

something on a biacore--say, straight out of a

fermentor--

DR. KOCH: Right.

DR. KOZLOWSKI: --and look at what your

conditions do.

DR. KOCH: Right. It becomes an

analytical or monitoring tool.

DR. KOZLOWSKI: Right.

CHAIRMAN KIBBE: Anybody else?

[No response.]

Okay. Thank you very much.

I would liek to propose a short break--10

minutes. And then we'll get started with Jerry

Collins at 3:13.

[Off the record.]

CHAIRMAN KIBBE: We need to get started.

And I see by our colleague at the podium, who is

now changing the entire proces, that we are almost

ready.

316

What's wrong?

DR. COLLINS: The cursor was stuck, but

it's back on now.

CHAIRMAN KIBBE: The cursor was stuck.

There's nothing like having a stuck cursor to kind

of ruin your afternoon.

[Pause.]

All right--Jerry Collins is going to talk

to us about Critical Path initiatives in lab-based

bioresearch of small molecules -my favorite kind of

molecules, because I can draw the structures.

DR. COLLINS: There will be a structure

quiz at the end, then.

Office of Testing and Research--Current

Research and Future Plans

[Slide.]

DR. COLLINS: If you haven't gotten tired

of trying to find something different in this

diagram each time, my only point here is to

emphasize that there are some areas of overlap

between what NIH does in translational research,

and what we consider Critical Path Initiative at

317

FDA; areas of overlap and areas of difference.

[Slide.]

We think about research--at least within

the Office of Testing and Research--along the same

three cornerstones that exist in drug development:

that's safety, efficacy and quality. And

throughout this talk and my final one, I'll try to

align our programs to those cornerstones.

[Slide.]

The divisions--in green, on your

left--represent our quality side in OTR, and on

your right, in blue, represent the biology side--or

mostly safety, a little tiny bit of efficacy.

About 75 percent of our staff is in the

left side, under "quality," about 25 percent is on

biology. I'm the director of the Laboratory of

Clinical Pharmacology, and also Acting Director of

Applied Pharmacology.

[Slide.]

There are three research programs in the

Laboratory of Clinical Pharmacology. For those of

you who've served on this committee in past terms,

318

we started a metabolism and drug-drug interactions

program in the mid-'90s. Our goal was to interpret

what was then a barrage--a virtual avalanche--of

data from in vitro systems, trying to predict

interactions between drugs, between drugs and food,

on the basis of metabolic pathways. This has been

a program in concert with the review staff that's

resulted in the production of several

guidances--I'll mention that in a minute.

A more recent project is hepatotoxicity.

Hepatoxicity, or idiosyncratic hepatotoxicity in

particular, has been a major cause of drug

withdrawals--we're actually trying to use our

expertise in metabolic processes and apply it to an

extension into liver toxicity. We'll come back to

that.

And, finally, the only project, really, in

the office that's related to efficacy, we're

looking at PET imaging for early therapeutic

assessment. It generates a number of interesting

consultation reviews, and it's really and extension

PK-PD issues that clinical pharmacology

319

subcommittee--this group--deals with regularly.

[Slide.]

I don't need to tell this audience that

adverse drug-drug interactions are a major

headache, and have been a major problem. I think

we have a very, very simple goal in approaching

this problem, and that's to improve the efficiency

and design of human clinical trials to

eliminate--or at least minimize to the smallest

possible degree--the potential for drug-drug

interactions.

They're relatively easy to find. They're

relatively easy to predict in advance. If you can

triage the worst ones--the potentially worst ones

in vitro, and study them in vivo, then you can gain

an incredible amount of confidence, rather than

looking under every stone for another drug-drug

interaction.

[Slide.]

In addition to our work in dealing

applications that come from drug sponsors, we also

work with the National Cancer Institute, which has

320

its own drug development pipeline, and we have a

memorandum of understanding to help them learn the

technology of drug metabolism so that they can

apply it in their pipeline. One of their employees

actually works in our laboratory on theses

techniques, and we've also participated in some of

their Phase I trials by analyzing the drug and

metabolism in vivo in their first in-human studies.

[Slide.]

This is the only flow diagram I'll show.

This essentially describes a decade-long process in

which we do metabolism studies using human liver,

in vitro in our lab. That led, initially, to a

guidance on how to do relevant in vitro drug

metabolism experiments. That guidance was enhanced

by the review experience, particularly from the

Office of Clinical Pharmacology and

Biopharmaceutics. We then extended that to the in

vivo situation, giving a guidance for industry on

in vivo metabolism, drug interactions designs, and

that also built upon collaborate clinical

studies--when we were able to do them--as well as

321

the metabolism-based drug-drug interactions that we

were able to study in the laboratory. And these

guidances are currently being updated in Clin Pharm

subcommittee of this parent committee, has been

active in reviewing it, sort of stage by stage in

some of the new areas.

[Slide.]

Hepatotoxicity, as I mentioned, is really

and extension of our expertise in drug metabolism,

because it's been known for decades, now, that some

of the most troublesome liver toxicities arise from

reactive metabolites--not from the chemical that

was swallowed or injected into the patient, but

from a metabolite that was formed right in the

liver, and the liver being the place where the

metabolite is first form, is also the first site of

potential injury.

So we're using our expertise in

understanding metabolism, to look specifically at

reactive metabolites. We aren't doing this in

isolation in the laboratory. John Strong, the PI

on this project is on FDA's steering committee.

322

There's a joint program with PhRMA to meet

regularly and discuss hepatotoxicity issues

together.

[Slide.]

Here's an example of the analytical

procedure that John and his group use. Glutathione

is the universal sponge for sweeping up reactive

metabolites as they form. So, by radio-labeling

the intracellular pools of glutathione, we can look

at what grabs onto it after the end of an

incubation with an unlabeled drug, and we've

labeled the reactive metabolite by its linkage to

glutathione--almost an operational definition of

what is a reactive metabolite: it's something that

wants to get together with glutathione.

Using the prototypical liver toxin,

acetaminophen, we were a bit surprised when we did

an inter-species comparison: it's a known

hepatotoxin in homo sapiens and in the rat. And

what we found is that the rat and the human have

one very large peak, eluting at about 13 minutes on

the HPLC tracing. The rat also has a second peak.

323

I think the important thing is not to get

distracted by inter-species differences in this

case; just a take-home message that, while you can

see it in rodents, since we have the ability to do

these experiments in human liver in vitro, that's

where the focus of our experimental work ought to

be.

[Slide.]

Just a minute t talk about efficacy. PET

imaging is intended primarily to look at an early

evaluation of drug action. And our involvement has

been to try to encourage innovation into this

process.

Those of you who've looked through your

background materials in the launch document from

March of 2004, "Noninvasive Functional Imaging" was

highlighted in the roll-out as one of the areas in

which FDA and industry have agreed to work together

to try to maximize its potential for finding the

winners and losers relatively early, streamlining

drug development.

We will be announcing a joint meeting with

324

BIO and with PhRMA, and with the Drug Information

Association early in 2005 to convene the community

of imagers and therapeutic developers to see how

the two can bring their tools to the table.

[Slide.]

In terms of why we're doing it, FDA, CDER,

has a very long history--as many of you have heard

in your service on this committee--in using

pharmacokinetic and pharmacodynamic principles and

their application to regulatory decision-making.

The disappointing thing, scientifically,

is we're always looking at extra-cellular fluid.

We do the absolute best we can with what we've got

to measure, but when it's just the circulating

plasma, we're only seeing part of the problem.

And the mechanism-based activity of the

drug is inside the cell. And the ability to see

distribution of a drug inside the cell, its

interaction with receptors, enzymes and

transporters, is more like what's done in drug

discovery, in terms of figuring out why this drug

was picked. So if we can find out in vivo [sic]

325

whether we have the right concept being applied in

vivo and select the dose, it could make our

downstream work a lot easier.

[Slide.]

A good example of this is a drug that was

reviewed by FDA's GI drug Advisory Committee last

year, and recommended for approval. This is a drug

called Emend, or aprepitant, from Merck. It's

intended for the reduction of chemotherapy-induced

nausea and vomiting.

And this is the classic curve that this

committee and other Advisory Committees--and our

review staff--are usually faced with in drug

development. The y-axis is a measure of

activity--Phase II data, not Phase III data--and

the x-axis is some plasma concentration of that

drug. So there's an attempt made to link

pharmacokinetics with pharmacodynamics. And what

was found was that at 40 milligrams there was some

activity, but it was sub-optimal. At 125

milligrams, we exceeded 90 percent of the activity;

at 375, of course, we were up on the shoulder, or

326

the plateau, of the curve.

So it was a molecule that certainly showed

dose-response, or dose concentration response. The

question is: did any of that activity relate to

why this molecule was chosen for development? Or is

just a me-too that acts by the same mechanism that

other drugs do?

Well, Merck Pharmaceuticals is one of the

leaders in applying PET imaging to the study of

drugs in their pipeline. And this is a tracer map,

on your left, of substance P-receptors in the

living human brain. And the color scale is that

red is the hottest concentration of receptors,

followed by yellow, followed by dark and then

lighter blue and then darker blue.

So that's the phenotypic map that can be

measured prior to treatment, or in a placebo arm.

And it's consistent with what's seen in the human

brain at autopsy--except that this subject is

living.

Subsequently, as you move across to the

right, the next image is what happens at 40

327

milligrams. 40 milligrams--we would interpret this

image as--is very effective at blocking the

receptor, so that when we give a probe--a

radio-labeled positron-emitting probe for substance

P, it no longer can stick to the receptor, and

therefore we don't detect it--non-invasively.

As we go up to 125 milligrams--the middle

image--there's a little bit better blockade. If

you do quantitative analysis, you can see

additional blockade. But clearly we're reaching

the plateau. And 375--and one higher dose, shown

on the far right--don't get you any extra benefit.

These information are supportive to

approval. The drug was recommended for approval by

the Advisory Committee, and approved by FDA, on the

basis of its activity in randomized Phase III

controlled trials. But the reason these data were

supportive, and presented to the Advisory Committee

were twofold. First of all, they relate the

activity of the drug, at a particular dose, to the

presumptive mechanism of action. And, number two,

they permit the lowest possible dose to be used.

328

Well, we all are familiar with that

concept: to maximize the therapeutic index you

want to minimize the penalty, in terms of adverse

effects. It turns out that the higher you go with

this drug--just like others--the more baggage you

bring in terms of adverse reactions.

In this case, there's a serious increase

in drug-drug interactions because aprepitant

induces and inhibits many metabolic systems. Since

these patients, by definition, are going to be

taking a bunch of other drugs at the same time,

minimizing drug-drug interactions by using the

lowest possible dose, consistent with preserving as

much anti-emetic potential as possible, helped in

choosing.

So, on the basis of this linkage of

imaging studies with Phase II data, the sponsor

chose 100 milligrams as their dose for the

randomized Phase III trial, and an add-on trial,

and it showed superiority in a placebo-controlled

test--an example of what we think is generalizable

in many therapeutic areas. So much for Clinical

329

Pharmacology.

[Slide.]

In our Applied clinical Pharmacology

Lab--which also might be called "Pre-clinical

Safety"--one of our elements is molecular

toxicology. And, like other labs at FDA, and in

university labs, we're very interested in

microarrays for their potential to show us a broad

range of signals, good and bad--in the case of

pre-clinical safety, early indications of possible

toxicity.

However, from a regulatory standpoint

we're very concerned--just like most of the

community is--in the chip-to-chip,

platform-to-platform, reliability and consistency

of microarrays. So microarrays are very impressive

as an 11,000 gene, one-page readout of most of the

relevant genome, but it's not the quality of the

image--the "awe factor" that we're interested in.

We want to know that if we take that same sample

and do the next 10 chips, with the same platform,

will we get the same picture? If we go from an

330

Affymetrix platform to an Agilent platform, will we

get the same kind of readout?

Those are the questions, if we're going to

make regulatory decisions on the basis of these

kinds of data, those kinds of cross-platform and

chip-to-chip reproducibility are what's important.

We can't do this by our own. We have

three people who are involved in this project. So

we partnered with the platform makers, the users of

these, and we're doing multi-laboratory ,

inter-laboratory comparisons of standards. And it

seems to be proceeding at the right pace.

[Slide.]

The second important aspect is not just

does the picture look the same, but how do you

analyze the picture? What kind of statistical

tests for robustness can you apply? And there,

we've enlisted a very good partnership with our

internal CDER statisticians--Bob O'Neill, who

joined us earlier in this meeting--and Bob's been a

very effective advocate, among the statisticians

across all centers at FDA, to form a partnership

331

between statisticians and biologists. So we'll

generate all the data that will help them develop

tests for figuring out multiple comparison,

correction factors, and all the things that they do

behind the scenes. And they'll help us develop

metrics for figuring out how reproducible the

quality is.

[Slide.]

In terms of gene markers of toxicity, not

surprisingly, we're interested in cardiotoxicity,

renal toxicity and, more recently, differences in

pediatric toxicity versus adults.

Again, this group is not acting in

isolation in our ivory tower in the White Oak

laboratory; closely connected to the Senior Science

Council--Associate Commissioner Alderson's group;

and several of us are on the Inter-Center Working

Group on Pharmacogenomics, chaired by Larry Lesko.

We've had two joint workshops, between FDA

and PhRMA, that we've participated in, and the

third one is in planning for 2005.

[Slide.]

332

In preclinical biomarkers--"biomarkers"

appears throughout the Critical Path document--what

we're interested in is trying to zero in on those

clinical toxicities that are particularly hard to

monitor or only develop late in the course.

And, traditionally, this has been

done--this is hardly a new field; we call it

different things--but biomarkers, in the past,

because of the technology, have been one at a time

events. Well, now that we have, you know,

multi-channel arrays of various sorts, coming from

olmics, genomics, proteomics, how do we bridge the

way we did these things in the past to the way we

do them in accelerating in the present?

[Slide.]

Well, anthrocyclines, such as doxorubicin

are known to cause cardiotoxicity. The slide at

the left, which is doxorubicin by itself, compared

to the slide at the right--doxorubicin plus

dexrazoxane--which is a cardio-protectant, we don't

need to be a histopathologist to see that there's a

difference, but we do have to have a piece of the

333

heart. And although you can get a piece of the

heart for valid therapeutic reasons, it's clearly a

difficult way to search through biomarkers. We'd

much rather have some kind of serum test.

[Slide.]

And, sure enough--for those of you who've

been following the New England Journal of Medicine

and other clinical papers--the troponin series has

been recognized--in fact, has been declared in

several recent articles--to be one of the major

breakthroughs in monitoring cardiotoxicity in human

beings.

Now, this particular study that's in front

of you this afternoon is looking at troponin T

levels in rats. So we took the signal from humans,

went backwards, in this case, to see whether we

would have picked it up a priori, or in advance,

rather than after the fact. And what we find is a

relationship between the cumulative dose of

doxorubicin on the x-axis, and the serum troponin T

circulating in the body.

[Slide.]

334

Well, that level of cardiac troponin T in

the serum does correlate very well with the

cardiomyopathy score, scored by a histopathologist.

So it looks certainly like it has the

characteristics of a good biomarker. But it's one

thing.

Is there some way to generalize this and

look more broadly?

Well, using expression arrays, we've

looked at a variety of different pieces of the

heart--pathways in the heart--that are known to be

affected by anthrocyclines, or have unknown effects

of anthrocyclines.

So, certainly, cardiac muscle function and

structure, you can imagine, is adversely impacted

by doxorubicin itself, and yet if you look at the

far right column, dexrazoxane has a protective

effect there.

We were unsure about fatty acid metabolism

and glucose metabolism, some aspects of immune

response. We get some mixed signals there--all of

which show changes in the treated animals with

335

doxorubicin, but in animals who get the same dose

of doxorubicin--in the middle--and get the

dexrazoxane as well, most of those changes are

modulated. The control arm is the right arm for

dexrazoxane by itself. And finally, it's not

surprising that something that's done this much

structural and functional damage also has

stress-induced genes that are highly overexpressed.

[Slide.]

One last example, from the safety

domain--recently, across many different therapeutic

areas, the phosphodiesterase inhibitors, among

sub-families 3, 4 and 5, as well as other

vasoactive drugs, have been shown to have bleeding

problems: vasculitis, vascular injury problems.

And although there are species differences across

the mammalian empire, rats, dogs, primates and

sometimes mice, have shown this phenomenon.

However, the only way you can see it is with

invasive testing. So, in keeping with our mission,

we were looking for biomarkers that might be

associated with it.

336

And a number of them have been studied.

Again, we're more into one at a time, or a few at a

time. We had a half dozen; here's four that fit in

a slide and might be still readable--

[Slide.]

--in which we can see a progressive

increase in circulating markers when the vascular

histopathology score is going up as well.

So, treating rodents with a variety of

phosphodiesterase inhibitors causes circulating

biomarkers to go up, and that increase in marker is

associated with the invasive test, which is looking

at histopathology.

[Slide.]

I guess the bottom line is that these

biomarkers represent a potential new tool for

evaluating preclinical safety, and as important an

endpoint as that is, I have to ask whether it could

be extended into humans, as well. And I'll talk

about that later in the day.

[Slide.]

In summary, on the biology side of the

337

Office of Testing and Research, our programs in

biomarkers, pharmacogenomics, noninvasive imaging

and drug interactions are certainly the template,

or the scaffolding that you could develop a

Critical Path Initiative around. We feel very well

aligned and prepared to charge into the Critical

Path Initiative projects that we think are quite

harmonious with its goals.

CHAIRMAN KIBBE: Questions.

[Pause.]

I don't see anybody jumping to the

microphone--go ahead.

DR. KOCH: I guess it's always in the

definition of "noninvasive," but with the PET, you

still need to inject the radioactive, short-lived

isotope. But that's noninvasive?

DR. COLLINS: Well, I guess my FDA

training would have me modify it to "relatively

noninvasive."

DR. KOCH: Oh, okay.

[Laughter.]

DR. COLLINS: We also included MRI

338

techniques in that regard. And if you don't have

to use a contrast agent--if you're doing the

standard T1 and T2 kind of paramaterization--you

actually do that--the only invasiveness is a

magnetic field. And it's--you know, particularly

where we come from, we're certainly not going to

blow off the risk of these kinds of things. But we

can quantify those risks in terms of other everyday

life activities. The radiation in a PET image is

less than that of a conventional chest x-ray, and

it can be made lower with more specific detectors

that are being detected now.

I forgot how many airplane trips back and

forth to Denver it would be equivalent to; the

radiation that you get at 35,000 feet.

So there are additional, incremental risks

that are undertaken, but in the context of everyday

risk, the local IRBs, human subject committees, and

the FDA have said, well, the benefit to society

versus the minor risk is okay.

But there are very strict dosimetry limits

on the amount that we can give as a radio tracer.

339

CHAIRMAN KIBBE: Anybody else?

[No response.]

You seem to have done a successful job of

presenting information.

And now we have Dr. Buhse.

DR. BUHSE: Okay. I'm Cindy Buhse,

Director of Division of Pharmaceutical Analysis.

And as Jerry mentioned we are on the quality side

of OTR labs.

My lab is mostly responsible for looking

at analytical methods that are used to test drugs.

And so my labs mostly made up of analytical and

physical chemists. And I'm going to go through

some of the programs we have.

[Slide.]

Let's see. Programs we have to support

the Critical Path Initiative--some of these you've

heard of this morning from John Simmons and

Lawrence Yu, because a lot of what we do supports

ONDC and OGD, in terms of trying to help them

determine how we can characterize novel dosage

forms and complex drug substances, not only to help

340

ensure we have the correct testing to approve a

generic drug, but also to ensure that we have the

right testing to approve changes in manufacturing

in innovator drugs.

We also have programs to measure and

identify micro and nanoparticles in drugs,

especially--it's often easy to measure the size of

a particle before you mix all your excipients and

drug together, and once you have a drug all mixed

together, what does that do to the particle size?

And we need ways to take a look at what's going on

in actually, final drug formulations.

We also establish--help establish

appropriate surrogate measurement techniques.

Lawrence talked quite a bit about this, and

dissolution is a big thing that goes on in our lab

in this area.

We also work a lot with the Office of

Compliance on drug authenticity and

anti-counterfeiting techniques. It's an issue--I

think if you watch the news at all--not only, like

Amy mentioned, in biologics, but its also an issue

341

with regular oral dosage form drugs.

And then we also--the last two, I'll

briefly go over--process analytical technology,

research we're doing, and some chemometrics, as

well, that ties into that. We're working with

DPQR--Mansoor Kahn's group--on those programs.

[Slide.]

To start off with the characterization of

novel dosage forms--some of the work that's

currently in our lab are things I think you've

already heard about from ONDC and OGD. We have a

program on liposomes, trying to characterize them

after chemical and physical change; trying to

determine how to--what analytical techniques work

best to detect changes in liposomes. And we have a

program with DPQR, as well, to take that a step

further and see if we can use cell-based assays to

see how these changes in the liposomes can be

detected in the cell-based assay.

Looking at transdermals--people call them

"patches" as well, patch products--and their

adhesive strength. How can we characterize the

342

adhesive strength and assure we have an analytical

method that can be used to no only compare a

generic to an innovator, but also can assure the

quality of a patch before it's released for sale.

John Simmons mentioned conjugated

estrogens. We have some LCMS techniques we've been

running. He showed some of that data. And we're

trying to improve those methods to make sure

they're very reproducible and can be used to

compare innovator to generic, or to compare an

innovator product after a change.

We also do some work with protein

products, trying to look at different analytical

methods to detect aggregation and degradation, and

assure we know the exact molecular weight and

distribution of protein products and can

characterize those.

Some of the regulatory accomplishments

we've had in the past in this area include input

into conjugated estrogen guidance, which is

currently out.

[Slide.]

343

This is just to give you an idea of the

kind of work we're doing on the liposome project.

We're looking at two different types of liposomes:

Pegylated and the convention--in fact doxo--the

very drug Jerry was just talking about, up there in

a liposome form.

We're looking at different stress

conditions, and then looking at different

analytical methods to determine how the liposome

was affected. Was the actual drug substance itself

affected? Or was the liposome affected both in the

lipid composition and in the amount of drug that's

encapsulated in the liposome?

And we determine what stress conditions

will give us a small amount of degradation, and

then Mansoor Khan's group will take those degraded

liposomes and see how they react in a cell-based

assay, see if we can see differences in their

uptake.

[Slide.]

Patches--there are different types of

patches out there on the market, and we're taking a

344

look at both kinds when we look at adhesive

properties.

One has actual drug in the adhesive--the

adhesive and the drug are mixed together and you

actually get your dosage by the size of the patch.

And then there's also reservoir-type patches. And

if you start looking into adhesive properties--and

we actually are jointly working with CDRH on this,

as you can well imagine, with things like band-aids

and medical tapes. There's a lot of variables to

look at when you're doing test method development

on adhesives. And I've listed some of the

variables down below that we're looking at to try

to come up with a method that could be reproducible

for patches.

[Slide.]

In terms of measurement and ID of micro

and nanoparticles, some of the projects in our lab

include looking an some of the sunscreens that are

currently being marketed as having nanoparticles.

We're trying to look at seeing what techniques can

be used to evaluate the size of these particles

345

once the sunscreen has been formulated.

And likewise, in nasal sprays, we want to

know what is the particle size of the active

ingredient once it's been mixed together,

especially nasal spray suspensions. And I'll show

an example of that in a second.

We've also done some evaluation of

Andersen Cascade Impaction, which is used to

determine fines--trying to determine how to improve

that test method. It's very variable, and are

there other options to using Andersen Cascade

Impaction to get a handle on fines in nasal sprays.

Some of the regulatory accomplishments

that have come out of our lab included input into

the nasal spray BA/BE guidance; and also we've done

some measurement work with cyclosporine particles

and helped ONDC and OGD with that.

[Slide.]

This is just an example of some of the

work we've done on nasal sprays. Here's some raman

chemical imaging. And you'll see--I guess it's all

the way on your left, up at the top, you have a

346

Brightfield, just microscopic image, of the nasal

suspension. You can see a lot of different

particles there. It's hard to tell which particles

actually are active. So if you're trying to

determine a particle size of your active within

this formulation, it's tough to tell just from that

picture.

You can kind of tell--you look at MCC,

kind of is that rod shape there. However, if you

actually can take the Rama spectra of each one of

those particles--which is what's shown just below

that, you can see that the spectra's very different

at each one of those particles, and you can look

for the Raman spectra of your actual active drug to

determine which one of those particles is your

active drug. And that's what's shown down at the

right--at the bottom. We're determining from the

Raman which one of those particles in that image is

actually the active drug. And you can see that

there's two of those particles that are active

drug, and there's a little bit of an active drug,

maybe, attached to some of that excipient.

347

And from that we can then get the particle

size of the active drug within the formulation, and

we can also get a feel for maybe if the active is

maybe sticking to some of the excipients, which may

actually change its actually size from what you

think you might have put it, from the formulation.

[Slide.]

Establishment of surrogate measurement

techniques--we've done quite a lot in the last year

on dissolution, trying to do quality of drugs.

We've worked with Office of Compliance on the

malaria drug mefloquine to try to figure out why it

was or wasn't working in the field, for the

military. And we've also done some work with

megestrol acetate suspensions, trying to compare

generic to innovator drugs, and figuring out the

best dissolution test method to use for that.

In general, we're taking a look at

dissolution testing because it is heavily used--as

Lawrence said--not only for quality control, but

also to try to--for bioequivalence, as well. And

so we're trying to make sure that the actual

348

dissolution test methodology can be as consistent

as possible.

For those of you who do dissolution, you

know it can be a very variable method.

[Slide.]

This is just some information on the

calibrator tablets that are issued by USP to check

set-up of your apparatus. And you can see that the

limits on the calibrator tablets are very high--28

to 42 percent is the range that you can get for the

Lot M. And lot N which was after that, was 28 to

54. And Lot O is currently out, and is proposed to

be just as wide, if not wider, than Lot N.

So if you use a calibrator like this to

test your apparatus set-up, you can see that any

variability that you're seeing in your test method

potentially could be due to apparatus set-up,

because you're not going to be determining it from

this calibrator tablet, because it's just too

variable. And so our lab is looking at alternative

ways to ensure set-up and reliability of

dissolution apparatus, other than using calibrator

349

tables.

[Slide.]

One of the areas that I think has become

very important lately is just anti-counterfeiting

techniques, and ways to ensure that the drug you're

taking is the actual drug you thought you bought.

And so our lab takes--keeps a close watch on

technologies that are out there for counterfeit,

and even to see how they can apply. We've been

involved in several projects with the Office of

Compliance to ensure the quality of not only the

active pharmaceutical ingredients, but also foreign

Internet samples.

So we've tested both of those in our lab.

And we used conventional techniques--like HPLC and

GC--looking for impurities, etcetera, to see

whether the drugs are the same as the U.S.

equivalent. But we've also taken a look at new

technologies, because some of these can be very

powerful, much faster ways to detect counterfeit,

or can actually show us new--maybe give us clues as

to where drugs may have come from if they are

350

counterfeit.

[Slide.]

As an example, I was just going to show a

little bit ratio mass spectrometry. This is a

technique which uses stable isotopes to try to

detect where chemicals may have come from, and also

to determine if things were made in the same plant

or not.

This is a plot of the stable isotope of

carbon--which C13, versus C12, and oxygen, which is

O18 versus O16. And you can see that Naproxen,

manufactured at different places in the world, and

different plants in the world, cluster together, in

terms of their stable isotopes, and that's because

stable isotopes aren't the same around the world.

And when you manufacture a product, your stable

isotope composition within that product is

dependent on the raw materials you use, where those

raw materials came from in the world, and also on

your manufacturing pathway. And so it can be a

powerful technique. You can see, if you have a

drug, and you can test it by IRMS, and then

351

determine potentially which plant it came from.

[Slide.]

In terms of PAT--as in

anti-counterfeiting, we try to take a look at the

technologies that are out there, either new

technologies or maybe new to the pharmaceutical

industry, and try to determine how they might be

used for PAT; what some of their limitations or

benefits might be so we can be in a position to

advise ONDC or OGD as needed.

We have a couple projects in our lab

taking a look at coating composition, how that

affects the ability to see what's going on within a

tablet, and also taking a look at excipients and

excipient-drug interactions within spectroscopy,

and how that affects the ability to use

spectroscopy for PAT.

[Slide.]

As an example I wanted to show you

Terahertz spectrometry. Terahertz--this is between

infrared and kind of your microwave. You can see

up there on the spectra on the right, there. And

352

one of the benefits of terahertz, it's like NIR;

it's non-destructive. But it also is a lot more

penetrating. The NIR can go deeper into a tablet

or into tissues.

So it's being looked at, not only for

quality control of drugs, but also as imaging of

biological tissue, especially skin cancers.

And I just want to show you a little bit

of the spectra we've gotten. These are

acetaminophen tablets. They're from 65 to 135 mgs,

and you can see that the terahertz spectra, which

is the one on the left--it's not much features

there. I mean, you would probably look at all of

those and say that they looked pretty similar. But

if you take that data and you run it through some

parametric programming, and compare it to content

by near-IR, you can see you get a very good fit

between the terahertz and the near-IR. But the

good thing about terahertz, it would have the

potential to go--to look past a coating, or to look

deeper into a tablet than near-IR.

The terahertz here was actually done with

353

transmission. So by detecting the radiation

through the entire tablet, and near-IR often you do

reflectance.

[Slide.]

Chemometrics is another project that we're

doing with DPQR, trying to understand the

chemometric software packages that are out there.

If we're requesting people to use PAT, and to use

more multivariate techniques, we want to understand

what their limitations and benefits are, especially

for model building--pre-treatment of data, things

like that. We want to be able to provide expertise

in that area.

Just as an example, the kind of things

that we've been doing. Here's some near-infrared

of those--actually the same--acetaminophen tables

that I just showed you with terahertz. But this is

all with near-IR. On the left is near-IR

reflectance, which is the full range of the

spectrum, from 4,000 to 10,000 forcipical

centimeters in the near-infrared. On the right

side is transmittance--okay? So this is where

354

you're trying to go actually through the tablet.

You'll see it's a little noisier in transmittance,

and you can actually only use about 8,600 to 10,000

because of the noise.

However, depending on how you treat the

data--you can see underneath the reflectance we

have--we've taken a second derivative, and we get a

good correlation between the near-IR and the

content measured by HPLC. However in the

transmittance data, we don't need to do the second

derivative. We can take the direct spectra and get

the same time of correlation with the content

measured by HPLC.

[Slide.]

I just wanted to put this up because

people talk about the St. Louis lab, sometimes.

That's us, I guess--Division of Pharmaceutical

Analysis. We are the only CDER lab located outside

of Maryland, so we have a small group of people at

White Oak, with Jerry Collins and Mansoor Khan.

But we also have our larger laboratory in St.

Louis. And so a lot of our interactions occur by

355

video-conference and telephone. But we still

manage to get quite a bit done out there.

So--hopefully the Cardinals will come back

in the next two games because, of course,

everyone's very depressed about that out in St.

Louis. So I'm not sure there's much work getting

done in the lab right now, after last night's

defeat.

[Laughter.]

So--I'm happy to answer any questions

about the Critical Path Initiative.

CHAIRMAN KIBBE: Questions? Michael?

DR. KORCZYNSKI: This is more or less a

comment. And I don't know whether you could

directly answer this--but, pharmaceutical

analysis--as you were speaking I was wondering:

most of the products that we're discussing are,

indeed, sterile products.

So is there a counterpart to your

activities in the microbiological areas, such as a

laboratory investing microbiological analytical

methods for even investigation of counterfeit

356

drugs, or bioterrorist activities? Is there some

type of microbiological analytical counterpart to

pharmaceutical analysis of products?

[Pause.]

Or maybe it's resourced out. I don't

know.

DR. HUSSAIN: Well, I think much of that

is done in our field labs. And Amy--and I don't

know whether we have a focused effort on

microbiological methods, but counterfeit efforts on

many of the injectable protects and OBP are being

carried out, too.

But, we actually do not have a very

focused broad quality microbiology lab within OPS.

CHAIRMAN KIBBE: Go ahead--Mike?

DR. KOCH: Yes--question, Cindy--on the

surrogate dissolution--

DR. BUHSE: Mm-hmm.

DR. KOCH: --you know, we heard this

morning of the different pHs, and time and

different things that go on there.

Over the years has there been, in addition

357

to the USP standard dilute hydrochloric acid

approach, has there been a way to simulate the

process, to try to come up with a dissolution test

that goes through a low pH, followed by neutral pH,

etcetera--to actually try to simulate.

DR. BUHSE: There's been a lot of research

done on dissolution, and there's a lot of research

in the literature and in academics. They have--one

of the dissolution apparatus is like a flow-through

apparatus, rather than the vessel, and some of the

studies done on those have been the type that

you've talked about. There, you don't recirculate

the dissolution media, you just continue--and you

can continue flowing it through the tube, and

you've got the actual pharmaceutical suspended in

the middle of the tube, and you can change the

media as it goes through--things like that.

So there are research programs out there

like that, and we're reviewing those and seeing how

they might be applicable--or maybe more applicable

than the vessel method.

CHAIRMAN KIBBE: Go ahead.

358

DR. MORRIS: Yes, just a question on the

chemometrics--looking at the well-executed, but

relatively traditional chemometric approaches in

evaluating the packages that are out there.

Looking at cross-process chemometrics in

sort of process-vector type work, or multi-block

systems to try to take into account more than a

single assessment of a product, as opposed to

looking at the product train?

DR. BUHSE: I guess--maybe Ajaz, who knows

a little bit more--

DR. HUSSAIN: Right--no, I think much of

the internal work has been focused on what we can

have.

DR. MORRIS: Sure.

DR. HUSSAIN: Because we're hoping the

CRADA with Pfizer, I think we're just starting to

get in the process and so forth--I think our

interest would be to get at process signatures and

so forth. But I think for that we need to have a

collaboration where we have that.

Mansoor is actually setting up the

359

manufacturing lab. And so once that is set, we

will have access to that. But most of the work

we're doing right now with in-house data is based

on chemometrics for products that we have our hands

on.

DR. BUHSE: Yes, and we've done a little

bit of that. Some of the data I showed you was for

one--like, for instance, for one compression rate.

We have similar tablets we've made--exact same

formulation, at different compressions, different

excipients.

So we, you know, try to throw more

variables into it. But I think some of the CRADA

and manufacturing efforts--make it more--give us

more the ability to do further work in that area.

I think Judy had a question.

CHAIRMAN KIBBE: Yes, Judy had a question.

DR. BOEHLERT: Well, I'm down here in the

corner.

This is sort of a general

question-comment. It applies to you and to several

of the more recent presentations this afternoon.

360

You have a number of research

projects--liposomes, characterization, adhesive

nature transdermal. To what extent do you interact

with industry? Because industry is also working on

these same factors, and looking at adhesive

strength, looking at the stability and

characterization of liposomes.

And, you know, I don't want to see people

going in two different directions to come up with

two different ways to do the same thing. So is

there synergy between what you're doing and what

the industry groups--or maybe even the academics

are doing?

DR. BUHSE: Yes, some of the projects we

do work extensively with industry; with the patches

project we've been working with--I think I

mentioned CDRH, our other center, but we've also

been working with 3M extensively, because they have

such a knowledge of adhesives, and they also

actually manufacture quite a few of the adhesives

for patches--as it turns out.

So, in some cases, we do work with

361

industry. A lot of cases we're not really able to

because what we're doing is trying to compare,

perhaps, two different products, or a generic and

an innovator, and there starts to become, you know,

some issues there where collaborating may be more

of a problem.

CHAIRMAN KIBBE: Follow-up, Ajaz? Go

ahead.

DR. HUSSAIN: no, not follow-up. I think

I just wanted to sort of emphasize--John Simmons

had mentioned the rapid response.

A lot of the activities in the St. Louis

lab are getting to solving problems that we face.

For example, the adhesive issue came up through

dramatic failures in adhesive performance on--we

manage, in the Office of Pharmaceutical Science, a

Therapeutic Inequivalence Action Coordinating

Committee. And then from the MedWatch, from the

consumer complaints--we were receiving a lot of

failures of transdermal systems falling off.

And then we looked at that and said we

actually do not have a good method, which is also

362

part of the stabalating program for many other

products. So that was an outgrowth of that.

And liposomes, for example--one of the

challenges was we were setting dissolution

specifications on liposomes--I'm not kidding. So

we said, "Let's understand some of that," and so

forth.

So, a number of projects that Cindy

does--immediate answers that are needed, and that

is a very critical element. So you have to keep

that in mind. So that's a very important lab, from

our perspective, in a sense, because immediate

answers are needed for John Simmons' Prussian

Blue-type work, and so forth, and so forth. So

that's--I just wanted to clarify that.

CHAIRMAN KIBBE: Anybody else?

DR. BUHSE: Quick questions?

[No response.]

CHAIRMAN KIBBE: I guess you're off the

hook.

DR. BUHSE: I guess it's on Mansoor.

CHAIRMAN KIBBE: Dr. Khan.

363

DR. HUSSAIN: Just as he comes on

board--he is new to FDA. So he came from academia.

So he's--

CHAIRMAN KIBBE: You're asking us to be

nice to him? Is that what you're doing?

DR. HUSSAIN: Yes, that's it.

[Laughter.]

DR. KHAN: Good afternoon. It's quite a

challenge to stay motivated and speak in the

afternoon, but I'll try to do my best here.

I'd like to thank Dr. Webber and his team

for giving me this opportunity. I would like to

thank the Advisory committee for your leadership

and the important role you play in this process.

I'd also like to thank the audience, who have been

extremely patient since morning--I've been

noticing. So--audience.

Most importantly, I would also like to

thank my colleagues from the Division of Product

Quality Research. Some of them are here, and some

of them that are not here, but they have given me

some of the slides to share with you, just to show

364

what goes on in the Product Quality Research.

[Slide.]

I will just briefly go over the outline.

People do ask me--I'm also new here, as Ajaz just

mentioned--that, you know, they asked me, "Okay,

what's the mission? What do you do?" So I would

briefly at least outline the mission and the reason

that we have here, then present to you the team, so

you'd get an idea of what our division is about,

and the current needs related to Critical Path and

the cGMP initiatives; some of the future

directions; and examples of "design space." It

comes about a lot, and I thinks morning, also, a

question was asked about the case study. I may not

be able to provide the case study, but at least I

can provide some examples of that one. And then

some questions about that. Okay?

[Slide.]

The teams--sorry, the mission first.

Advance the scientific basis of regulatory

policy with comprehensive research and

collaboration; focus/identify low and high-risk

365

product development and manufacturing practices;

share scientific knowledge with CDER review staff

and management through laboratory support, training

programs, seminars, and consultations; and foster

the utilization of innovative technology in the

development, manufacture and regulatory assessment

of product development. Basically, we would like

to stay aligned with OPS and the CDER missions.

The vision--we want to be recognized

leaders in providing support for guidance based on

science and peer-reviewed data; well trained staff

and state-of-the-art product quality laboratories

that is capable of providing any information sought

by reviewers, industry and the FDA leadership.

Culture--the way we live and act--one of

cooperation, mutual respect, synergy, professional

development with life-long learning opportunities.

Basically, this slide I derived from some of the

internal presentations. I just wanted to go over

it so that we are all on the same page.

[Slide.]

The division, we have about 19 scientists

366

currently working on this. We have three teams.

The fourth one is in the making: the

pharmaceutical/analytical chemistry team; we have a

physical pharmacy team; a biopharmaceutics team;

and a novel drug delivery systems team.

So I'll briefly go over what they do, and

share some of their slides with you.

[Slide.]

Pharmaceutical/Analytical Chemistry

projects--we have team leader Dr. Patrick Faustino.

He has done some work on this Prussian

Blue--basically, safety, efficacy and product

quality studies. John has presented to you this

morning some of those studies. And basically that

laboratory work was done in a DPQR.

Then we have the shelf-life extension

program, where we have the stockpile of drug with

the U.S. Army, so we look at some of the stability

issues of those drugs. And, then, very recently

they've also worked on isotretinoin--some of the

bioanalytical and kinetic studies they have done.

Basically I'm just going some of the

367

current work that is being done, and what we want

to do to make changes, with your recommendations on

this.

[Slide.]

Just one slide--he has already shared

these things with you, but I think this is just the

effect of pH we have seen, because this is a

compound of high interest--radioactive

decontaminant. It was releasing some cyanides we

have seen that the release of cyanide is much less

at a certain pH--I was just focusing here--that the

release of cyanide is much less at certain pH, so

safety at a certain pH--you know, it's much safer.

And then the efficacy--we have done some binding

studies of radioactive cesium. Also we are working

on thallium on this one. So the binding studies we

have seen as the pH goes up, the binding is more

here on there. So this gave us some idea about his

compound.

[Slide.]

The next team, the biopharmaceutics team,

headed by Dr. Donna Volpe. And it's a small team.

368

If we want to go in the area of bioavailability and

other issues, then I think this team needs to be

expanded a little bit.

They have broad activity on the BCS

guidance. A question came up this morning, also,

about the extension of this BC guidance. Donna

Volpe is working actively on these things.

They have also worked on--there was some

bioequivalence issues of levothyroxine sodium

products. So we have looked at the stability of

this. It was a huge project. A lot of people were

involved with this. We have just completed that

project and the final report is about to come out

on that one.

We are looking at the effect of

cyclodextrin, as well as some other excipients, on

the permeability of certain drugs. Dr. Volpe has

created a huge database where the permeability of

certain commonly studied drugs--like atenolol, some

metopralol--and, you know, we also looked at the

permeability of mannitol and the various factors

affecting the permeability of that drug. So we

369

have created a database of that.

And some uptake studies--I think Cindy

mentioned just some time ago about some of those

uptake studies of liposomes. So this is the work

which Donn's work group is doing.

[Slide.]

Just to give you an idea--I think that

this question came up about, you know, moving this

Phase II--BCS Class III drugs in the direction of

getting bio-waiver. This will give you an

illustration that if you have a high permeability

drug--you have a metapralol drug, a high

permeability drug, you get these two different

excipients there. There was no difference.

But if you change the excipient, where we

have this osmotic agent here--the sorbitol--then

you see there's a tremendous difference in the

availability. So this needs to be sorted out

more--you know, to seek the bio-waiver. So this

group has helped us study some of those things.

[Slide.]

The next group--the physical pharmacy team

370

that we have: Dr. Lyon, Robbe Lyon has done a work

in the PAT-related issues--the Process Analytical

Chemistry, I might say, because Chris Watts keeps

correcting us on this "PAT"--the terminology of

PAT. But what I'm talking about is mostly the

chemistry aspects of this PAT. And then Everett

Jefferson is the team leader of that.

And so I will highlight some of the

slides. They have given it to me--just to show you

what they are doing with these analytical sciences

that we have.

[Slide.]

You can see here--this is just with

near-IR profile of--

[Moves off mike.][Inaudible.]

This is a lot easier. You are right.

Trying to get where the mouse is. It will come

sometime? Okay. Okay. All right. Got it here.

So we have these acetaminophen powder

here, the avicel powder, and then you have a tablet

here. So you see--so we look at the contents of

this--the HPLC. We saw the content, we saw the

371

near-IR, we saw a correlation. I think Ken asked

some question as to what you do--what validation,

basically. You eliminate one thing at a time so

that you get a correlation where you can rely a bit

more on that. That's what they've done on this

one.

[Slide.]

Same thing we have done with Raman Spectra

here. We have it here in the laboratory. We have

this Raman spectra. I will tell you how we can use

it, later on, in some of the optimization studies.

We want to employ this. But at least now we have

the procedures in place to do some of those

studies.

Raman spectra--similarly, you look at some

of these peaks, and then see the correlation. You

see the HPLC content we have done with these

tablets, and we have correlated with Raman, partial

e-squares.

So you can see the correlation here. It's

fairly good here in this one, too.

[Slide.]

372

Likewise, we have also looked at the blend

uniformity. You can see, this is a formulation

that it clearly shows that this is well blended, as

opposed to this formulation which is not well

blended. You see the API, you can see that it is

not as well blended here.

So if you look at the near-IR spectra, the

spectra is very close to each other. You know,

it's likely that it's a good blend; you know, they

have mixed well. This is separating out. That

means, you know, they have not really mixed very

well. So it gives us some idea of this mixing.

[Slide.]

Now, hydration--this hydration--Robbe Lyon

has given me--hydration is not the process of

hydration, it's basically just an identification of

a product which either anhydrous, or a monohydrate,

or a some hydrate we can detect--whether the

product is hydrous, or anhydrous--anhydrous product

or a hydrated product.

[Slide.]

The next slide will show we have basically

373

two brands of product. We have brand 1, a capsule,

and then a Brand 2. You can see, this capsule

here, it has Core A and Core B. Basically, it has

two different cores--okay? And no some Brand 2 has

three cores, instead of two cores. Basically they

have two of this B core, and one of them is core A.

So if you want to detect how much of it is

anhydrous form, or how much of it is in the hydrate

form. So they look at some imaging here, and that

imaging, basically they are showing that in core A,

there's--these are nitrofurantoin capsules, by the

way--in core A you have more of the anhydrous

concentration. And basically they have estimated

it to be 8 percent. And actually, when they have

seen it, it was 9 percent there here in this one.

And, similarly, when they did it on this

brand, in core B you have seen this is a

monohydrate concentration, this has more of the

monohydrate concentration. The estimate was 50

percent, but the actual was 40 percent.

So, you know, it just gives us some idea

of see the current, and it's not just the drug.

374

But we can look at the different polymorphic forms

of the drugs themselves.

[Slide.]

This is the near-IR dissolution

correlation. We have looked at some of the

tablets. The tablets were prepared--I think Cindy

mentioned some time ago, you know, we had these

acetaminophen tablets. I think the next one will

show.

These are some tablets. Basically we

looked at the tablets. We predicted. We trained

the data. This is the training set.

[Slide.]

The one in the blue, and we looked at the

correlation. The correlation was .984. And then

we have this test data. We looked at almost 72

different tablet formulations, and we saw that it

was fairly--especially at a higher dissolution

profiles, and a lower level IC, that the curve is

off. Actually, if you take some of these data

points off, if you take--go for a higher

dissolution, when the amount is higher the

375

correlation might be much higher, if you take these

data points off.

[Slide.]

So if I have to summarize as to what's

going on currently in the DPQR, this is what it is.

We have the Drug Substance--we are

characterizing--trying to get the--

[Pause.]

Okay. Trying to get the mouse here--the

curser here. Okay.

So, basically, if you have a look--we have

drug substance. We are--currently, in the DPQR we

have this drug substance. We are characterizing,

and the process analytical tools that we have. We

have the analytical method. And the

biopharmaceutical groups brings some cell culture

work. And the drug product--we can characterize

the drug product that we are doing here. I have

shown you some of the work here that's being done.

And as well as the stability--as I

mentioned to you, the shelf-life extension program

is going on for some of the drugs of national

376

importance.

But, as a new kid on the block here in

FDA--I know, having worked in academia for a long

period of time--for about 12 years I worked in

academia--and then after coming here, I wanted to

see if what we are doing is enough for us. So what

I did, I started listening to the leaders here. I

started attending the meetings--some forums like

this--you know, the Advisory Committee--your July

Advisory Committee, I was here. And then I

listened to a lot of presentations of the leaders

of the FDA. I got to see what Dr. Woodcock has to

say. I got to see what Ms. Helen has to say in HR.

I've gone to a lot of their presentations, and I've

read a lot of internal reports. I attend a lot of

internal meetings, just to see some of the

directions.

To give you an idea, you have already seen

a lot of Critical Path slides, so I didn't want to

duplicate some of those slides. Initially, I had

that in the presentation, but I took some of them

out, seeing some of the speakers had those things.

377

But here--you know--because this is a

developmental type of research. I think Ken also

alluded to this in the morning, as to who will fund

this--the level of funding--and who will do this

research? If academia doesn't do it, if the NIH

doesn't do it, then who else will do it? And the

industry doesn't want to share the information.

So at least we will have some work--at

least we'll have some data in place so that the

reviewers don't have to operate in a total dark

box.

Since you have already heard Ms. Winkle's

presentation, the support for understanding of--the

process understanding and the Critical Path roles

is highlighted here in this slide.

[Slide.]

And the internal efforts have culminated,

really, in the articulation of this thing in the

desired state about ICH, as you can see. I don't

know if any other speaker had this, but previously,

in the manufacturing subcommittee Advisory

Committee that you had here, you had a lot of

378

people presenting this. Basically, product quality

and performance achieved and assured by the design

of effective and efficient manufacturing processes;

product specifications based on mechanistic

understanding; and ability to effect continuous

improvement continuous real-time assurance of

quality--that's exactly what we want to do in the

Drug Product Quality Research. So it becomes

easier to expand, it's easier to re-orient some of

the programs and expand some of the current

programs in BPQR.

[Slide.]

And this is what we intend to do--as I

have shown you before. This is the current work

that we were doing. Although we had this Chemical

Stability here, but we want to look at some of the

physical changes there, in that one, too. We want

to do that.

And what else we want to do is the

manufacturing aspects of this--you know, the role

of the excipient, the role of the formulation

variables, the process variables, the mechanistic

379

evaluations, optimization procedure--lot of it may

already have been done. We can gather it through

the literature, we can gather it through the

collaborator, we can do some of the in-house work.

But we want to provide this information. We want

to provide this training to our reviewers, and we

want to update ourselves on these things.

And just to give you an idea: we are

talking about--you know, if we want to talk about

the process variables--okay?

Now, just taking tablets in the picture,

there are so many process variables. You have this

mixing, they have milling, then you have your

granulation, the drying, compression, coating,

packing. Just by mixing you've--you know, the

blend, homogenating problems. And just by

granulation you might see a lot of problems there.

A lot of prime test data is not provided to the

reviewer. So to understand that, we need to have

some internal programs going so we will have this

understanding.

And this we are talking about just a

380

tablet dosage problem; a dosage form that is so

well know--or a capsule dosage form. They are so

well know. And when we talk these novel systems

that we are getting--

[Slide.]

--by the way, the bioavailability, also,

we wanted to either collaborate and do it in house.

The Novel drug delivery systems--we want

to have some of this program going in the

novel--the nanoparticles--there's a huge, huge

area; the liposomes; the sustained release, the

modified release; the transdermal systems; the

nasal pulmonary path; disintegration; the solid

dispersions--basically, we want to have information

of going in that direction; we want to have

information readily available, and the training

that is needed to evaluate those applications.

[Slide.]

Some of the newer projects: the novel

drug delivery systems, including nanoparticulates,

preparation, characterization, development of in

vitro procedures in DPQR laboratories--I will share

381

some of the data with you. We have already done

some work prior to me joining here. I'll share

that with you--some science-based projects, with

mechanistic understanding.

Process engineering with real time

monitoring and modeling. We have this particular

equipment--fluid bed--with near-IR probes attached

to it so we can monitor a lot of process here in

this one.

The SLEP-stability--and there are some

repackaging issues. We are working on some of the

stability of those repackagings. We want to work

on those issues. Basically they are

stability-related projects.

Generic drugs--I think Lawrence

highlighted some time ago. Tomorrow there's going

to be a presentation also. If you have some

locally-acting drug, what do you do with that?

Stents--again, combination drugs. You

have a device and you have a drug, and you have

some issues related to that. We can help in some

of thee issues related to that.

382

We already have CRADAs with companies, and

we are going to have some more CRADAs. Somebody

asked a question about collaborating with industry.

So we are collaborating with industry--and more

coming up. And we are very hopeful that we will

have some more CRADAs coming up pretty soon, so we

will turn in that direction.

Some permeability of these drugs.

[Slide.]

Now let me spend some time here on the

design space. I will give you a couple of

examples, and then I will also highlight the

importance of it. You have already seen that it is

important, but I will just share some of the

examples with you.

I will share one or two classic examples.

This is out of the text--you know, you have some

good statisticians, you have some good experts here

in this area. All I'm doing is I've just borrowed

something from the book, that people have been

discussing, and people have been having in the text

for a long period of time. So I think the time for

383

us is just to be able to adapt some of those

things, and show their relevance to the

pharmaceutical product.

So I will present one or two examples from

the literature, and then I will present some

examples of a design space in the laboratory

generated data that we have here in this one.

Now, here is a scientist--okay? A

scientist is trying to work. He has to run a

reaction, at a laboratory scale--he has to run a

reaction. There are two variables in this one--the

time and the temperature. So this scientist is

trying to--first of all, it does. So if you have

two variables there, first of all it does this, he

fixes the temperature here at 225 degrees; he fixes

the temperature at 225 degrees. He runs the

reaction for a certain length of time. He is

basically trying to get the yield of this

particular compound.

So, he fixed it at 225 degrees, and

then--and he ran it at different times, that

particular reaction. He got a yield, something

384

like 70 or 71 percent, and then he got that yield.

And then now that he got the time, then what he

did, he fixed the time here. The lower one will

show that 1--30 minutes--I can't see very well from

here, from this angle--but somewhere here it shows

this one, 30 minutes.

So he fixed the time here. Now he ran a

different temperatures--okay? So he came up with

different temperatures, and he saw that at 225

degrees, basically, he has this yield. So

basically he changed one variable at a time, and he

got the yield at 71 percent.

But if you have--if you listen to what the

statisticians tell us, what they show, if you

follow some of the examples that are already out

there, we can really perform the very design sort

of experiment, the same scientist, when he performs

the design sort of experiments--also I might argue

that a lot of times you will have less experiments

than you will have with so many, you know,

duplicates, and triplicates and quadruplicates.

So the same experiment, if we do with a

385

design set of study--

[Slide.]

--look at what he got. He basically

changed the temperature and the time

simultaneously. He was basically here in this

one--in this design--no matter how many experiments

you perform, no matter how many times you do it,

you're yield is likely to be around 70, 71 percent,

or somewhere in that neighborhood no matter how

much time you do. Basically, you are totally out.

And somewhere here you would not have gotten it.

Somewhere here, you see that he got this 90, 91

percent of the yield for this compound.

[Slide.]

Now I'll stop here for a moment, and I'll

change the gear a little bit. Let's assume that we

have an identical situation where instead of the

yield of this particular compound, we are looking

at some other response--this response could be a

dissolution response; some percent dissolve in

certain amount of time. It could be a

bioavailability area. It could be a hardness of a

386

tablet. You know, it could be any other response

that we are looking at.

But if he develops this kind of a

strategy--develop some experiments in the

laboratory, come up with something like this--but

if I have this particular product, if I have this

particular response--in the laboratory--I would

hesitate to go to the scaling up and to the actual

manufacturing, if I have this much a narrow window,

then talk about these problems here in scaling up

and product manufacturing. Then you say, well, you

know, the lab-based data is very different from the

manufacturing data. We don't want to do that

because it's variable. Yeah, if you are in such a

narrow window, any slight change you make, then

it's likely to have variability. Then we might

fall into a lot of difficulties. We might fall

into difficulties of--suppose you have some

out-of-spec situation. Then what do you do in a

case like this?

Okay? So how do you scale up? Huge

problem.

387

But if you take something around this

region--but if our product, if our optimized

product is somewhere here in this neighborhood, I

would feel more comfortable taking it for scaling

up, taking it for the manufacturing, because later

on you have--you don't really have a lot of

problems of scaling up. You don't really have a

lot of problems of out-of-specs. And even if you

have some out-of-spec situation, you can really

play around and improve that situation, because you

have something to go by.

And once we do this--I think if you really

look at this cGMP--the White Paper of cGMP--a lot

of these things are already described there in this

one. But if you have this formula, you can take it

for the manufacturing, and then what you can do, if

you are in manufacturing, then you can take some of

it and do the evolutionary--EVOP--basically the

next slide will show you that one. So you can play

around. You can fine tune and improve your

manufacturing process. That basically provides

some opportunities for continuous improvement and

388

innovation.

But if you have a product here, you took

it for the manufacturing, then really, you cannot

change the variables there, you know; anything, any

slight change in any variable might change the

product, and you don't know where to start.

[Slide.]

So, as I mentioned to you--this is a

different example--again from this book--this Box,

Hunter and Hunter, 1978, book--basically once you

have this optimized formulation, and once you take

this formulation, then here, in this case, you have

the stirring rate, you have this addition pan--you

have the solution pan, you can play around and

gradually you can play around. Because you know if

you are in manufacturing, you cannot afford to fall

outside the specification range. So your window is

very, very limited. So if you have a design space,

your window--you are well within your window to

play around a little bit. So you can gradually

work on this and improve the yield. And, finally,

you see in the last one, it doesn't improve any

389

more, you stop.

So this kind of data should be extremely,

extremely valuable, extremely useful. And I will

provide one or two examples of the laboratory data

that we have. I think one of the graduate students

had worked on it.

[Slide.]

And I have selected this for two reasons.

First of all, it's an extremely complicated

preparation; very complex preparation. Here you

have a protein, and you are trying to develop a

formulation of a protein; a lot of variables in the

protein, just to decide on this formulation study

itself, we had to do a lot of precharacterization

and characterization work. And you will see some

back-to-back--two back-to-back papers in J. Pharms

this year--February and March, there are two

publications--just to decide on the formulation

issue that we have to do.

After doing that, then we have decided

that, all right, we will try a dosage form--see

this salmon calcitonin is a peptide that we have

390

taken--polypeptide--salmon calcitonin. It was

degrading with enzymes. So what we did, we have

seen some turkey ovomucoids--a lot of work was

already done on turkey ovomucoid--basically it was

inhibiting the degradation of salmon calcitonin,

the different enzymes--you know; trypsin, the

chymo-trypsin, the elastase. It was inhibiting

their degradation. So we wanted to use this turkey

ovomucoid as one of the excipients to prevent the

degradation.

We also wanted to use this glycerotinic

acid, because it's protein, big molecule, doesn't

go through biological membranes. We have see that

glycerotinic acid--we evaluated--we screened almost

a hundred compounds. But finally we settled with

glycerotinic acid. We have seen that glycerotinic

acid enhances the permeation of this protein.

So we wanted to make the dosage form.

This is a bi-layered preparation, by the way, and

the top layer is very similar to your

procardia--you see this dosage form--this

bi-layered preparation; procardia, vomax and, you

391

know, these are osmotically-controlled bi-layer

tablets.

So here you have a protein, and then we

have this osmotic agent here. If you look at

it--so we make this--we compressed this tablet, we

made these bi-layer tables. We drill some opening

here. We provided some coating to it, so that it

releases drug in a particular fashion. It's a

dual-controlled release. You have a drug

protein--the polypeptide that's releasing, as well

as the ovomucoids that's releasing. Extremely

complex preparation.

The idea here is: you can see there are

so many variables here right now. What should be

the coating thickness of this one? What should be

the opening of this one? What should be the level

of the excipients that you use?

So you can see there's a lot of

variability here.

Now, a company that is manufacturing,

that's making dosage form, a lot of that

information they might have in-house as to, you

392

know, the coating thickness that is needed; some of

the process variables they might already have. And

if you don't have it, what you can do, you can

actually screen--we have just selected some of

them. We have screened some of those variables.

We could not do an extensive study--very expensive

proteins. We cannot do a lot of experiments. But

at least we screened those variables here, at two

levels each.

[Slide.]

And then the dependent--the response--the

previous one, the example that I gave you--the

yield of that compound was the response. But in

this particular case, we have the amount

released--salmon calcitonin release--in three hours

was our response. And then we can also place

constraints.

[Slide.]

Now, here, in this case we have placed

constraints at different dissolution time points,

so you can tailor a release. You can do that. Or

you can place constraints on tables. So you're not

393

interested in a tablet where the hardness is less

than 4 KP or more than 8KP. So you can place

constraints on hardness, constraints on some of the

parameters that you're looking for. So, here, we

placed constraints so that we can get the entire

release profile on this one.

So by placing constraints, we evaluated

that, and we looked at this--the development

equation here.

Now, again, as I said, this is just a

screening design. You cannot see the interaction

effect. The interaction effects are compounded;

the quadratic effects are compounded. So a lot of

information we are losing, we are missing. But we

gathered from here is: of those seven variables

that we looked, what are more important, what are

less important? Basically we screened those

variables.

So if we have to have a few experiments

you want to run--so what we did. So we selected

out of these three variables, and that we studied

at a slightly more detail--I will show you in the

394

next one.

[Slide.]

We have selected another response design

in this time. So basically we have seen the amount

of sodium chloride, the osmotic agent that is

needed in that particular tablet dosage form, and

the amount of coating, and the amount of

Polyox--it's the polymer that is required.

Basically, these are three variables, and we found

that these three variables are more important--at

least they're likely to have more effect on the

release of salmon calcitonin than other variables.

So we selected these variables. And this

was the dependent variable: salmon calcitonin

release in three hours--okay? And now we developed

this model.

[Slide.]

Now, believe it or not, this one equation

can talk more than probably 20 pages of slides, 20

pages of information. Really, it does say a lot.

It says how those variables affect the

response. It just shows how X1 changes the

395

response here; how X2--the coating level--if you

increase the coating level, dissolution decreases.

I know that. And if you increase or decrease the

coating level a little bit, immediately I can

calculate the response, without even doing an

experiment I can calculate the response. Same

thing, I can see the interaction effects of all of

them; the quadratic effect.

Basically, by this design sort of

experiment, finally we have used a process where we

have actually predicted the levels. We predicted

that. If you have this much of sodium chloride,

this much of coating thickness, and this much of

the Polyox levels, then we will get--this is the

kind of tailored dissolution profile. We predicted

that.

[Slide.]

And what we did, we performed an

experiment in triplicate--three, the proof, and

then with our product that we obtained was

identical to the product that was predicted.

So this is the case study that was done in

396

our laboratory by one of the graduate students.

I will not go into the details--oh, by the

way, this is the response-surface. You have

already seen the response-surface for the yield.

So here you know at what level you can get the

dissolution that you want.

[Slide.]

I will not go into the detail, but, you

know, we have also prepared some nanoparticles. We

have characterized by a lot of different methods.

You can see this publication--International Journal

of Pharmaceutics--highlighted all those

characterizations. But this one also--these

nanoparticles, also--we used a design set of

experiments where we have seen, basically, just a

formulation variable. We took it at three

different variables. After having gone through the

screening and all that, we have optimized it.

[Slide.]

And we have seen the dependent variables

here. And, again, the observed and the predicted

levels were identical in this particular one. It

397

just shows the levels, as I mentioned to you, about

the yield.

And here, if I--after developing this in

the laboratory--now, certainly, one has to feel

more comfortable taking it to the manufacturing,

because they know where they can play around. If

you select this particular product here for the

manufacturing, you manufacture it, you know you

have some room to play around. So you can do this

evolutionary operation and play around and improve

the product.

[Slide.]

So that is--with this, certain questions

that I had for the Advisory Committee.

As I said here, that I'm also learning.

I'm also just so new. I just want to orient our

programs, or orient our lab in such a way that it

reflects some of the agency's thinking, some of the

OPS thinking. We want to go in that direction. So

you are the experts in this. You have been

associated with this for quite some time, and if

there's anything that we are not doing you want us

398

to do, just let us know.

Does a systematic study with a designed

set of experiments provide opportunities for

reduction of--you know the post-approval, I did not

mention it at this time. But, you know--scale-up

changes--the post-approval changes--you want to

make some tiny change, you keep on getting these

post-approval submission documents. If you have

some window to play around, certainly, you know, it

can reduce. But if you don't agree, just let us

know.

Do you agree that the information on

design space, with a designed set of experiments

will reduce the out-of-spec situations a whole lot

more? You know, if you have a very tiny window,

any slight change--the speed of the machine, the

machine going on and off--just an operator just

coughed--you know, or you just change the operator

there, or anything might change that situation.

Do you agree that the research with

sell-designed set of experiments on lab scale with

create opportunities for continuous improvements

399

and innovations in manufacturing? So industry has

got to apply that and provide the data to the

reviewer, so that they are not operating under a

black box.

So, with this, I think you very much.

I'll be happy to take questions. Thank you very

much.

CHAIRMAN KIBBE: Any questions for--

DR. SINGPURWALLA: Mansoor, I was told to

go easy because you are new. [Laughs.]

[Laughter.]

So I will try and go easy.

DR. KHAN: I can only get something I

know.

DR. SINGPURWALLA: The design of

experiments--the questions you asked--my answers to

all of them is: yes, yes, yes, yes. Because

design of experiments is, you know, well recognized

and well accepted--particularly by the chemical

industry.

The question I have for you is: how do

you intend to use design-of-experiments in the

400

regulatory process? What you have described is the

use of design-of-experiments in manufacturing,

which is what the industry should be doing. I

suspect they are doing it. If they're not doing

it--shame on them.

[Laughter.]

But I'm sure they're doing it.

So how do you intend to use this in your

particular role as a regulator is what I'm eager to

see--or hear?

DR. KHAN: The regulatory questions, I

think--you know, some others will answer. You

know, it's beyond my understanding at this time.

But my idea here is to provide this

understanding to our reviewers; to provide this

understanding to our own scientists so they utilize

it. And also if we publish more papers--if we just

provide this information to others, a lot of others

might be more willing to use it.

And as far as the people in the industry

using it--you know, some of them are using, some of

them are not using. And people might be using it,

401

but at least they don't provide the information to

us at all in any significant way at this time.

CHAIRMAN KIBBE: Okay--

DR. SINGPURWALLA: I see big daddy is

coming to defense.

[Laughter.]

CHAIRMAN KIBBE: Go, Ajaz, go.

DR. HUSSAIN: Well, I think

design-of-experiments is--what?--a 60-year-old

technology that we're introducing. So it's not new

at all and so forth.

But at FDA, we don't have the ability to

say that somebody has done the work or not done the

work and so forth. So we have to assume that what

we see is the limited data that companies--many

companies do this and they don't share that.

But at the same time, I think surveys done

by Professor Shangraw, before he passed--at the

University of Maryland and so forth--and more

recent surveys, suggested the use of

design-of-experiments in pharmaceutical industry is

very low. About 7 percent of the companies we

402

surveyed through the University of Maryland said

they actually used design-of-experiment.

So that leads to the concern that we have:

if you haven't understood even the critical factors

and so forth, how can we allow them to change? So

we cannot allow them to change and so forth.

As a result, we have a static

manufacturing process.

So, for those companies that do this

routinely, that have this sort of information, if

this can be summarized as a means to demonstrate

what are the critical variables, to what extent the

validation ranges can be justified as wide as

possible, and so forth--so that provides a means

for regulatory flexibility--for those companies

that have this type of information and so forth.

For other who do not--not get the benefit

of regulatory relief at all. So--

So how would we use this in the regulatory

setting? That has been a continued discussion

internally. My thinking right now is this is not

an FDA policy and so forth. It's--what we would

403

simply need is to focus on the predictability and

reliability of the predictive power that you have

developed and so forth. And that should be enough.

We don't have to get into deep--there's volumes and

volumes and volumes of pages of how was this done

and so forth, because our job is to understand what

is critical; what ranges are acceptable; and then

what is the design space. And how well you know

that is through your predictability.

So it's more of a summary type of

information I'm looking for.

CHAIRMAN KIBBE: Go ahead, Ken.

DR. MORRIS: Yes, just to follow up--I

think part o this falls into the category of having

the reviewers understanding the process well enough

so that if they do get a good rationale of the

formulation and process design, and

design-of-experiments that they've really outlined

a real variable space, as Mansoor was talking

about, that they'll be able to appreciate it.

So part of that is, I think, ensuring, or

reassuring the companies that, you know, generating

404

these sorts of data, they'll receive the proper

reception when they get here.

CHAIRMAN KIBBE: Joe? And then I have

Melvin.

Go ahead.

DR. MIGLIACCIO: Well, to dispel any

myths--yes, we do use design-of-experiments.

Aggressively. Aggressively.

I think the issue is is that what we then

present is a proven acceptable range; univariant

proven acceptable range. That's been the

tradition. That's what has been expected.

As we move forward, using

design-of-experiments, coupled with the technology

we have now to, during those experiments, to

monitor the critical variables real-time--we'll

move from submitting a static process--a process

that is based on a range of time or temperature or

any other condition--to a dynamic process that

says: "If A, then B." And "if A then B" will be

based on rigorous design-of-experiments, with the

right multivariate analysis.

405

So I think that's--you want to respond to

that Ajaz? That's--

DR. HUSSAIN: No, I think--we have one

similar thinking on that. I mean, ICH Q8, I mean

that's the direction I see we're going.

DR. MIGLIACCIO: So it's not going to be a

fixed process.

One more--your third question, I have a

bit of, I guess--it implies something that I don't

think we want to imply: "Do you agree that the

information on design space, with a designated set

of experiments will reduce the OOS situations?"

You're implying there that you're going to

use the design space to set specifications. And

that--you know, specifications have to be based on

a mechanistic understanding of the formulation and

the process, and its impact on product

performance--not on the capability of the process.

And the design-of-experiments is helping

us to understand what's critical, and what the

process capability is. It should not be used to

establish finished-product specifications.

406

And your question there implies--you know,

if we set the specifications correctly, and we

understand the variability and the measurement

system, then yes, good design-of-experiments should

reduce--and establishing the design space--should

reduce OOS.

But on its own, it won't.

DR. KHAN: I agree.

CHAIRMAN KIBBE: Great.

DR. KOCH: I guess I'll just make a

comment that even though the field is 60, 70 years

old, in terms of Plackett-Burman and a lot of those

studies, it is surprising how little it's used.

And you can go into chemical, petrochemical and

other industries, and they have not used it very

well.

The reason behind it is often the cost of

analysis. To do a good study, where you're running

a number of variables, you've got a huge amount of

samples. And I know, just historically--I got

involved in several what they were called "big

projects"--that would be eight to 10 variables--and

407

it was always neck-back, based on perceived cost.

I think, in the future there's going to be

a lot more opportunity--addressing your last

question--with the development of better lab-based

equipment--microreactors, a number of improvements

in high throughput designs for other reasons. But

I think the equipment's going to become available,

and PAT is going to be a vehicle to be able to

monitor these things.

And, eventually, I think you'll get down

to where you can very effectively use these

techniques, often even on continuous processes,

where you can invoke feedback and feedforward so

you don't have to run a whole number of

experiments, but you can be analyzing in real time

and adjusting your parameters and filling out your

space much more adequately.

But I don't think it's been used very much

in industry.

CHAIRMAN KIBBE: Nozer? Ah, we get--you

had something else, there, didn't you?

DR. SINGPURWALLA: Yes, I just wanted to

408

react to Jerry's comment.

I was personally--my prior probability

that industry uses design-of-experiments was very

high. So I'm not surprised. And, basically, if I

was running an industry, I would use

design-of-experiments to maximize my own profits

and do my business more efficiently.

Industry A and Industry B can produce

exactly the same product, but one can do it very

efficiently by using design-of-experiments. And

the other can do it completely randomly and still

come up with the same answer, but you're spending

money.

So that was the only comment: that it's

more on the manufacturer who has to take advantage

of it. And I'm really surprised that they are not

using it--based on what I hear from you.

CHAIRMAN KIBBE: Judy?

DR. BOEHLERT: Yes. I mean, I would agree

with Jerry: they are using it. There are many

companies that are not. And another area where

it's used a great deal--particularly the

409

Plackett-Burman design--is in the optimization of

analytical procedures. And I see that in big

companies and small companies. They know how to do

it. They save their resources and they come up

with much better methods in the end.

It doesn't mean that everybody's doing it.

So I think to the extent that, you know, folks like

you can publish what you're doing, it helps those

that don't understand to get on the bandwagon.

But it is used, you know, in industry. It

hasn't been overlooked. But not everybody.

CHAIRMAN KIBBE: Anybody else? Comments?

DR. SINGPURWALLA: Well, the only comment

I want to make is I studied design-of-experiments

as a student. And perhaps it was the most boring

subject that I had to go through.

[Laughter.]

It is boring.

[Laughter.]

CHAIRMAN KIBBE: It's always good to have

Nozer's opinion on things.

[Laughter.]

410

CHAIRMAN KIBBE: I think we should more on

to Jerry. And thank you very much.

Jerry, your colleagues have managed to

leave you three-and-a-half minutes for your

20-minute presentation. And that will allow Vince

another 15 for his.

Wrap-up and Integration

DR. COLLINS: This is one person's

perspective on the day's events. And for those of

you who give talks a lot, it is very difficult to

stand up here without any of my props. I have no

slides.

I've been scribbling notes all day, since

9:30 this morning, when Ajaz was talking.

One of the most important thins on his

third slide was describing the Critical Path

essentially as not just another fad at FDA. Some

of us are a little shell-shocked by this management

agenda, or that initiative and so forth. We have a

commitment from the Commissioner--the Acting

Commissioner--the Deputy Commissioner for

Operations, and our Center Director, that this is

411

not something that's going away in six months. And

to turn the ship around and align it properly, we

need that kind of commitment from our leadership,

that we won't be thrown into the gulch to do

something else later. So, from the perspective of

the worker bees, that's very important.

Secondly, several speakers across the

board talked about relationships with NIH, going

back to the in silico talk from Joe Contrera; both

Steve and Amy talked about their relationships with

various parts of NIH; and my lab also has

cooperation with NIH. I've been at FDA for 17

years, and I spent 11 years at NIH before that.

I've never seen a better time for FDA and NIH to

collaborate and work together.

There's always been a little bit of "let's

make sure we know what our territory is." There's

overlap in our interest. There's also things that

are uniquely theirs, and things that are uniquely

ours. If we just focus on the overlap, I think we

really ought to take advantage of, again, what I

would call the golden opportunity here for

412

collaboration.

The other thing that's sort of been

missed--I'm surprised there hasn't--maybe I missed

it because I didn't get here 'til 9:30--but this

week, this month--is really the golden age of

quality. I mean, my computer screen didn't have

any disk space left a couple weeks ago, after

announcements on CMC, GMP, BAC, PAT--I mean, it was

just--there have been so many announcements about

the importance of manufacturing as an initiative

for FDA; about the success of the two-year

initiative; the roll-out of the implementation

phase. This really is a strong part--a strong era

of quality.

I hope it doesn't get lost in the Critical

Path. The Critical Path mentions quality issues,

but there are so many efficacy and safety issues

that we need to be vigilant, and not just rest on

our laurels.

In addition to getting your input, we've

asked the public for their input. The docket has

over a hundred responses. It's all in the public.

413

You don't--there's nothing secret. If you submit

something as a comment--we asked in April--and

there's over a hundred--on the website. And if you

have a really lot of time--because it's

clunky--over the weekend I looked at them all--and

it's very interesting. Almost all the comments

actually relate to efficacy. There's a few

comments that relate to safety, and a very small

number that relate to quality. And most of those

are actually for biological products of one sort or

another; either vaccines, blood-derived proteins,

or complex molecules from the OBP domain.

So we need to keep challenging the public

so that they recognize the importance of quality.

And we also need to look internally, that we're

responsive to--you know, our job is to either

convince them of the importance of quality, or to

re-align our resources.

As I mentioned, in OTR we're about 75-25

chemistry to biology. One of the excuses for

having this meeting is so the OTR folks can listen

to the OBP folks, and vice versa. And I'm still

414

learning about OBP. And I get more of a biological

flavor each time that I hear your presentations. I

don't know that I can fit your round peg into my

triangle of safety, efficacy and quality--but

that's part of the reason why we're here, so we can

learn each other's language, each other's culture,

and how it fits.

But I think, certainly, OBP is--actually,

the "B" is for "biology"--right? So, you know,

you're definitely more aligned with the safety and

efficacy side.

What about gaps in our program--various

places? Well, first of all, I mean the OTR-OBP gap

is really just about finding out about each other.

And one of the things that we probably discovered

today that would bridge the gap is the Critical

Path Initiative--is that all of OPS, and all of

CDER, and all of FDA is committed to going down

this route. So we all now automatically have

something in common, in that our programs must be

aligned to the Critical Path.

Now, Steve, I can't do that polygon stuff

415

that you borrowed from Ajaz, but in terms of a

bridge, I can think of the Critical Path Initiative

as something that connects two pieces.

The other thing is that product quality is

important--as everybody in this room thinks it is;

needs bridges to the clinical side, to the pharm

tox side, and to the clin pharm side. And so when

Ajaz talks about the ICH Q8 principles as one of

the ways that we can actually bridge these things,

this is really important. We can't do product

quality in isolation. And a hand-off from one to

the other has been covered in several of the talks.

But that's an area where we need to focus: on

making sure there isn't a gap there.

The other thing, in terms of keeping

reviewers and researchers together--we have two

distinct models that have been discussed here this

morning. John Simmons made a number of comments

bout the way Office of New Drug Chemistry interacts

with Division of Pharmaceutical Analysis, and

Division of Product Quality Research. Lawrence Yu

mentioned several projects that they've been

416

working on there. And then--and the OBP side, we

have the reviewer-researcher model that both Amy

and Steve articulated in their talks.

Those are somewhat different approaches.

In CDER, we have tried the reviewer-researcher

model, with very minor success. We found that

geography is a terrible burden and barrier--not to

mention use-fee deadlines and growing workloads on

the review side. So people who initially could do

both research and review eventually had their desks

swallowed up with all kinds of electronic copies of

documents, and found it hard to continue.

For the last 10 months OTR--a large part

of us--have been out at White Oak. And starting in

April, the immediate office of OPS--including the

in silico group--the Office of New Drug Chemistry

will all be there in the adjacent building. And

there is a physical bridge. It's just not a

conceptual bridge. The second floor of our

laboratory building is connected to the second

floor of their building. I think that will

facilitate reviewer-researcher models, because it's

417

location, location and location.

Now, it's not the whole thing. I mean,

the Office of Biotech Products is still on the NIH

campus for the foreseeable future. And our

laboratory in St. Louis is there for the

foreseeable future. So we don't have a

fully-integrated geographical solution to our gap

analysis, but it will be an interesting experiment

to see, particularly, how ONDC and the first floor

of the lab building interact, and whether that

improves the situation.

Last comment is that we're supposed to be

"science-oriented" here. And although the Critical

Path in drug development is a fact, it's

well-documented, it's only a hypothesis that we can

do anything about improving it.

We've laid out today--throughout the

day--a number of approaches that we've been

thinking about implementing, and have started

implementing, but it's only a hypothesis that

they'll work. The chances that they will work are

enhanced greatly by getting feedback from talented

418

people--from the public, from the industry, from

the Advisory Committee--taking that advice to

heart, and really giving it its best shot. Any

initiative fails if it's only a half-hearted

initiative, or if it's not well designed, or if we

don't have the right equations, or if we're 60

years behind in the technology. So--we appreciate

any forward-thinking ideas you may have in that

regard.

CHAIRMAN KIBBE: Okay.

Ajaz, help me here a little bit. Would it

be best for us to go ahead and let Vince do his

presentation and then take the three questions you

have sitting around here?

DR. HUSSAIN: Right--I mean, Helen and

Keith and I were just discussing that, in a sense,

because we have received constant feedback from you

throughout.

CHAIRMAN KIBBE: Right.

DR. HUSSAIN: Maybe after Vince's talk you

could just summarize, instead of getting into

answering all the questions in detail. But I

419

think, since we have received so much feedback, if

you could just summarize the Committee's thoughts,

it will be fine.

CHAIRMAN KIBBE: That means you're up,

Vince.

Challenges and Implications

DR. LEE: Okay, great. Maybe I can start

with the questions.

[Laughter.]

How am I going to work this thing?

[Laughter.]

Thank you.

Okay--thank you, Ajaz, and also thank you

Helen and Ajaz for giving this opportunity to work

at the FDA. It's an eye-opening experience, and I

recommend it to everybody. Because you get a

different perspective.

I was changing my talk as I was going

along, and that's why I was away for the first hour

of this afternoon; I didn't not that I would have

to make another copy that corresponds to my slides.

So that's something that also I learned.

420

Let me be more precise about what do I see

as the implications and challenges.

[Slide.]

The one thing is always to increase the

return on investment by fostering an innovation.

"Innovation" is the key word. And also, along the

way, we hope to improve the quality of life for the

patients, and lower the costs--the health care

costs--for society. In fact, as I was sitting

around the room, I wish that maybe sometime down

the road that we should include economists in the

committee to give us some assessments.

I wanted to look forward and see if we

were to follow this Critical Path Initiative, what

is the benchmark. What do we expect to see?

[Slide.]

And I'm trying to be cooperative, because

I have no clue about what should we expect. And I

don't know whether we can assume one number,

because each drug is different. But let's say that

maybe in five years' time--by 2010--then let's

commit to lower the development costs by 30

421

percent, shorten the development time by 50

percent, and increase success rates by a factor of

three. I have no idea if this is realistic or not,

but maybe we should start thinking about that.

And what else might happen? I would

expect that more drugs will be launched in a

controlled delivery platform when our

sustained-release system is used as a line

extension. So I'm proposing that his model will be

different.

Here comes the next point, is that the

sponsors of compounds might be forming a consortium

to share information and knowledge. This is

something that's not being done today. Obviously

it's because the conditions don't encourage that.

But we're in different times. And so maybe perhaps

we should think about different models.

And, moreover, maybe the sponsors will

subject their science to peer review for open

access in the global community. I would home that

maybe sometime down the road that equivalence, or

the genome project, would be reproduced in the drug

422

development arena. Now, this is something which is

quite naive, you might say. But I just want to put

it out there and see who would challenge that. And

I would be a bit worry that one reviewer, to make

judgment on one product--part time editor at the

same time.

[Slide.]

What else might be happening? Well, the

era of blockbuster might be over. I don't think at

this point in time few executives would believe in

it. And, frankly, I do not know how the agency can

confront this avalanche of applications if

everybody's looking at just specialized

populations. But I do think that a new era would

arrive where we'll be more realistic to look at

narrower indications, and then use the

patients--the users--to expand the knowledge base.

And I'm proposing that perhaps all of us would be

enticed to participate in a Phase IV study by using

the chips which are recently approved by the FDA.

This is subject of another big talk.

And then there will be a growing of

423

nano-sized assemblies with specialized

functionalities. Now this is something--I'm not

that fascinated by nanosystems. What intrigues me

about nanosystems is the capability, for the first

time, for the device circulating in our body,

collecting information, providing feedback to the

scientists. So I envision that maybe we can look

at nanosystems as satellites. This is something

that the body has never been exposed to. I have no

idea how the body would respond to it. But

intriguing to find out. Again, that would the

subject of another long presentation, talking about

diseases for which we need a way to assess the

early change. Cancer is very dreadful because by

the time we see the symptoms it's already too late.

Would it be possible to have a micro-chip

circulating in the blood stream, collecting

information that would report the

scenario--fingerprints characteristic of

disease--and that information would be fed into a

computer, and a database on that basis, a diagnosis

would be made.

424

So what I'm proposing is that maybe we're

approaching an era of preventive medicine, where

the patient would be at the center of the whole

process.

I'm going to just give a few slides in the

interest of time.

[Slide.]

This is a very intriguing slide to me,

because the reach limiting step--we talked about

changes, depending on the time. And depending on

the thinking of science at that time. 10 years

ago, in 1991, PK was a major problem. Now

everybody was focusing on PK, and now something

else popped up, a formulation, which was not a

major problem in 1991, becomes a major problem.

Who knows what it's going to be?

So what's the message? The message is

that we have to be always in touch with the leading

edge of science, and where the leading edge resides

is in the sponsors.

So what are the implications? The

implications are in four areas, as I see it.

425

[Slide.]

In terms of individuals, I think as

scientists that we can no longer focus on just one

thing that we're looking at. We have to have a 360

degree vision. And this is along the lines of what

Ajaz talked about--having a common vocabulary. I

don't know his name--but he's gone.

So the next point is the infrastructure.

How can we organize the scientists in such a way

they can respond to new opportunities on short

notice? I understand there's a SWAT team already

in place, but we need to have more of these in the

agency.

There have to be incentives, in terms of

incentives to reward innovation and teamwork.

Again, it's different times.

And finally, I see there should be some

kind of interrelationships--with the NIH--I agree

with Jerry, I think this is a golden opportunity

for NIH and FDA both being part of the HHS to

collaborate, to reinforce one another. I think

this is--and also, I think that the move to White

426

Oaks is very symbolic in the sense that for the

first time the agency's under one roof.

So I think that, whereas in the past

nobody talked to anybody, it's time for us to work

together, to exchange information. And certainly I

think the agency might consider sponsoring

projects.

[Slide.]

So what's next? I think that we need case

studies. This is easier said than done, but I

think there's a lot of information--data--in the

FDA archives. I don't where it is. I don't want

to volunteer to go look for it. [Laughs.] But I

think somehow we need the information, and

demonstrate that--under what conditions we can

categorize drugs in the same way as we do at the

BCS. And I think we need some kind of organization

to organize our thoughts.

We need some benchmarks, what should we be

looking for, if the Critical Path Initiative were

to succeed. I think it has succeeded.

Now, which sectors would apply this road

427

map to? Well, it was designed for big PhRMA. But

what about generics, biotechs and start-ups? And

who else? So we need to think about that.

And, finally, which drug class should we

begin with? And here we have no definitive answer.

But this is again a very interesting summary in the

nature of drug discovery, where it says that the

success--the percent of success--depends on drug

class--for obvious reasons. And I think that we

need to look at information such as this and do a

quick demonstration project to convince the

skeptics that it is the Critical Path concept is

viable.

[Slide.]

So what are the challenges to all this? I

think this is a recapitulation of what was said

throughout the day--communication. I think

everybody should understand what is meant by

Critical Path. And we should all follow the same

Critical Path. You go different Critical Path, I

think that we go nowhere. So broad understanding

and shared goal community-wide is important.

428

I think we should have a mechanism to

inspire the leaders among the scientists to create

new paradigms; and also to motivate the scientists

to adopt a new approach to decision

making--willingness to learn, and to unlearn, to

relearn--and learn. This is something that I'm

trying to do myself.

[Slide.]

This is Ajaz's favorite: the knowledge

management. When he first talked to me--not 11

months ago, but six months ago--I had no clue what

he was talking about. But finally I saw the light.

And we're definitely living in the

knowledge era--and there's no question about that.

And there was 200 years difference--200 years' span

between the industrial era and the knowledge era.

And the characteristic of the knowledge era is very

different from the industrial era. Where, in the

past we focused on single entities, now we're going

to have the ensembles. And why is that? That's

because in the past, we had no access to organizing

information; that we tend to think--we reduced

429

everything to a single entity. This may be the

physical chemistry influence on the formulation.

But when you find in the real world that

usually--not only single entity, but the things

work together as a team, an ensemble.

In terms of scientists, they no longer can

function as an individual; I mean, accomplish

everything individually, but has to have a network.

The success of science depends on the network of

all our colleagues.

Things are moving very fast in the

knowledge area. And things are dynamic. And I

think that we always are in view of sharing

information--sharing knowledge--whereas, in the

past, rewarded by being proprietary. Now this is

something which is very challenging, in my opinion,

to convince. Everybody think differently--because

we never know what the outcome will be. But at the

present time by protecting information, then we

move forward. But as a scientist, myself, I'm

always troubled by the duplication of efforts.

Oftentimes, you know, it's the failures that will

430

be useful, because at least I know that I will not

go down the same path.

And then it's very clear to me that--the

thing about my children, when they come along to

use this benefit of medicine in a major way, it's

definitely in a consumer-centered society, where

the consumers will know about health. And

hopefully, I think our government would promote

health education in the public.

[Slide.]

So what is the road map? It's very

simple--three things.

One is that we need to provide incentives

for industry and academia to formulate and test

alternative drug development schemes. And there's

no reason why drugs should fail in Phase III. If

they fail in Phase III, there must be a reason.

They must not be doing something right.

The second thing is that we need to think

about coordinating data mining worldwide for

forecasting hurdles to drug action, delivery,

formulation and manufacture. We can learn a great

431

deal from existing information.

And the third think was talked about,

again early this morning--is the computational

tools. I think that if we have access to

simulation models we can begin to test the weak

points--the critical parameters--and design

experiments properly, then we might be able to do

clinical studies more efficiently.

So this is the three things.

[Slide.]

So what are the--the three last points I

would like to leave with--the three areas--one is

outreach. I think that we definitely should

sponsor retrospective studies on the value of

sharing knowledge in accelerating drug development

and rendering it more precise. I think that we

can--although the past is no prediction of the

future, but at least we know what is the scientific

foundation.

The second proposal I have is to think

about convening a summit with industrial and

academic scientific leaders to identify the pros

432

and cons of what I proposed in the first, and to

understand the mechanisms to conduct data mining

without putting the innovator at a competitive

disadvantage. So I'm proposing we should think

about a strategic plan for drug development. This

is very far-fetched, but I think we should

contemplate this framework.

The second area that we should focus on is

the process. And, again, summarizing what was

talked about all day today--to examine. the current

review practices with respect to fostering

innovation and then propose necessary changes. And

the second point I would like to propose to be

looked at is to develop mechanisms for facilitating

continuous improvement in the quality of approved

products. I'm talking about the generics in this

particular point. There may be about eight years'

span between the launch of the innovator, and the

launch of the generic products. But science has

improved a great deal. Have we learned from it?

And how can we take advantage of these advances in

science.

433

And the third point is to be proactive in

identifying cutting-edge research of pharmaceutical

relevance that would fuel innovation. So, clearly,

the whole points of Critical Path Initiative is to

encourage innovation.

The last point is human resource. I think

it's something that, as a former academician,

education is of great value. And in fact, my

former university has a regulatory science program.

I don't think it's appropriate for the future. And

I dare to say that in front of my former dean. And

I think that we should do something differently,

because we should prepare the regulatory scientists

of the future.

In fact, I think it's very important for

us to think about the scientists on line five years

from now, and what do we need five years from now.

So I think the education of the regulatory science

programs--most of the programs in the

U.S.--perpetuates what we have today. So we need

to think differently.

And then the second point is a point

434

addressed to the agency, is the current practice of

recruiting scientists and retaining them, as far as

development of leaders from among the ranks. I

think this is central, and I do believe that

science has to drive the process, and research is

an essential component, and there's a lot to be

learned from the OBP part--the CBER--whether you

have research--where there's the opportunity for

research.

But the research that we do has to be

different--unique. And there's an unmet need. And

clearly it would be the bridge between academia and

industry research.

So--these are my thoughts. And certainly

if there's an interest, I will answer easy

questions.

[Laughter.]

Committee Discussion and Recommendations

CHAIRMAN KIBBE: I don't have easy

questions. I think you've said some very

interesting and thought-provoking things.

I've been thinking about everything that's

435

been going on today, and I know Ajaz suggested that

we might come up with some kind of a summary for

the questions today. And I really don't have a

good summary, but I have a lot more questions.

And what I think might be useful is those

of us who are staying for tomorrow to spend the

evening thinking about all of the things that we've

heard, and how it all comes together.

The human mind--as opposed to the

artificial intelligence that sits on our

desks--works in patterns and pattern recognition,

instead of sequences of computational paradigms.

But a couple of things come to mind that

I'd like to share with you, and then maybe we

can--I'll let you gentlemen ask questions if you

have any.

DR. LEE: Maybe I can add two points. One

point I should mention is, as a former chair of

this committee, that what--how could the committee

be more--let's see, I don't want to use the word

"useful," but since it's on the tip of my tongue,

I'll just say it--more of an asset to the office.

436

And this is something that I think I would be

interested to hear from this group, about how the

committee--how should the committee function

to--you know, in the Critical Path Initiative. So

that's one thing.

The second is that I think sharing

information is critical. And the way that things

work now is that information is passed from one

module to the next. I think that's in today's

world, the way that the human works is to

multitask. So any time information is available to

all the stakeholders in the enterprise.

CHAIRMAN KIBBE: Let me continue with some

thoughts. First, the question of the Critical Path

Initiative. Are we focusing on the appropriate

Critical Path?

The question I have is: is the output, in

terms of new and novel chemical drugs a result of

something that we need to work on in order to prove

the flow-through, or is it the result of a paradigm

that was begun early in the 20

th century and has run

its usefulness? Are we actually at that asymptotic

437

curve where we spend tremendous amounts of energy

to get a small breakthrough, but unless we have a

significant paradigm shift we're not going to get

there.

Are we asking ourselves that we new drug

entities? Wouldn't we be better off asking

ourselves that we need new and better therapeutic

ways of treating disease or preventing disease?

And maybe the shift that we went through,

away from surgeries and manipulations, to the use

of chemicals in the last century is over, and we

need to go into a different therapeutic thinking.

And if we can't make that paradigm shift, applying

tremendous amounts of energy to an old paradigm

that's running out of steam, isn't the way to get

there.

There's a lot of interesting new

technology on the horizon: computational power,

and what Vine talked about--which some people call

nanobots, are coming. And in 10 to 15 years we

will be at what some have characterized at a

singularity in our understanding of computational

438

power; a day when the ability of a desktop computer

to think in patterns and reason--as well in

patterns as it does in digital format--will allow

it to acquire data off the internet and come up

with answers we haven't even asked the questions

for.

And are we right now at that juxtaposition

where our traditional way of going about looking

for new therapeutic moieties is running into the

wall.

DR. LEE: Well, I think that we are,

because I think we leave the treatment with a

single compound may be on the way out. And more

likely that we are beginning to treat diseases with

combinations. Usually when disease, more than one

thing goes wrong.

Also, I too believe that with the day come

where you can hand in an application, then computer

will look at it and say, you know, yes or no. It

might--because, you know about is the pattern

recognition.

DR. MEYER: Yes, a couple of comments.

439

The agency loves acronym's, as we've seen

today. And I think it's interesting that "Critical

Path Initiative" is the same as "Consumer Price

Index."

[Laughter.]

They are related. We're trying to save

money, get things out sooner, make people weller.

So that's interesting.

Jerry used the term "hypothesis." And I

didn't hear what the hypothesis was necessarily for

all the things we've been talking about. And then

Vine, on page 5 said, "benchmarking." And I

think--well, he made up some--50 percent this, and

30 percent that in 2010--I think that would be

worthwhile, to show people where you intend to go.

Just a couple of comments that really deal

with the questions: prioritization in the era of

limited resources. Obviously, you have limited

resources. I think there's an impressive quantity

of work that was presented today. It was much like

going to an AAPS symposium. It was just

high-quality stuff.

440

And certainly I was brought up to learn

you'll be a more effective teacher if you're

involved in research--much like you'll be a more

effective reviewer if you're involved in research.

That was kind of my fair-and-balanced part. That

was the fair part. Now let's get to balanced part,

if I were a Senator on the Budget Committee.

We all know FDA has difficulty--a

difficult time with criticism about speeding up

approvals; difficult time with recalls of marketed

products; difficult time with a shortage of OGD

personnel--and a litany of other things.

So, given that era, I think it's going to

be critical to prioritize what you're doing in

terms of the Critical Path Initiative or any of the

other initiatives.

And let me just pose a couple of questions

that I would ask if you were telling me what your

priorities were: who else could do the work?

Could NIH? Could industry? Could academia? Could

CRADAs solve the problem? Who else could do the

research? Who else should do the research? Are

441

there really other groups that are better able to

do the work, rather than you re-inventing a

laboratory, and a process and equipment and

personnel etcetera? Are there other people that

should do it?

How can another resource outside the FDA

be encouraged--with a carrot--or forced--with a

stick--to undertake some of the things you're

already doing? I would use an example: you

publish a guidance, and before long there's all

kinds of people that are willing to train--for

money--industry; all kinds of people that are

welling to development instrumentation to help

industry. So you put an idea out there:

"Henceforth, in 2005, we will require that,"

somebody's going to figure out how to do it with

some piece of equipment, and market it, and that

will be good for the whole economy, and you won't

have to do it.

I would ask how does the research relate

to problems faced by FDA--not globally but, you

know, right now you have conjugated estrogens.

442

That's an issue. I don't know really who might do

that work. And then what is the importance of the

problem?

So I'd say: are there others capable of

doing the work? And what is the importance of the

problem? And how does the problem relate to

something closely involved with FDA.

CHAIRMAN KIBBE: Ajaz, what do you think?

Shall we farm it out? Outsource it?

DR. HUSSAIN: these are very, very

important questions. And I think the

benchmarking--the hypothesis--clearly, anything

that we do, unless we have a goal in mind, unless

we have a plan in mind, we're not going to get

there. And that's the reason the overreaching OPS

immediate office proposal we said was we will go

through some of the process, trying to map this

out, define the metrics and so forth. That would

be essential.

And clearly, I think, an initiative

umbrella creates expectations, creates a benchmark

that I think people will hold us to and so forth,

443

because nothing is free in life. So any

funding--anything that we get to support these

activities--will have an associated accountability

and efficiency in metrics.

So I think those are very important

questions that I think we will have to sort of

build into our thinking as we move forward.

CHAIRMAN KIBBE: Ken?

DR. MORRIS: Yes, just to follow up on a

couple points.

First, I think--to your point, Art--that

in the future I think therapies are going to be a

lot different; and, hopefully, significantly

different. But in the interim, between now and

then--given our 401(k)s and all--the thing that

strikes me most in your presentation Vince--other

than the eloquence, of course--is the Nature

Review's drug discovery article, and particularly

the attrition for each criterion, versus the

criteria.

And if you look at those from the '91 to

2000, what you see is that, in fact, tox has

444

certainly gone up significantly, but cost of goods

has gone from zero to 10 percent. Formulation has

gone from zero to 5 percent.

Commercial -"commercialization," I'm assuming--has

gone from 5 to 22 percent.

So I think those statistics really are

pretty much in line with a lot of what the 21

Century GMP initiatives, as well as the Critical

Path Initiatives were pointing out. I think,

overall, this is telling us that those are the

areas of opportunity.

The statistic you used about, you know,

decreasing the cost part by 30 percent is really

very consistent with what G.K. presented at the

manufacturing subcommittee last time where, if

you'd look at the current cost of goods sold as 25

or 26 percent of the current burden--if you can

reduce that by a third--say 30 percent--and apply

that to the discovery R&D--as long as we're still

in the paradigm of traditional chemical

discovery--that you can increase the discovery

budget by 50 percent.

445

DR. LEE: Oh, that's true. Yes.

DR. MORRIS: So that I think your

benchmarking is actually pretty--I mean, you know,

if not realistic--if it's not realistic we're in

trouble. I think it has to be realistic. I think

those are the goals we have to shoot for in the

short term, all the while keeping our eye on the

ball of the new therapies, I think.

DR. LEE: Your on the same lines that we

shift the responsibilities to--well, the

upkeep--the maintenance of the quality to the

manufacturers. So I would see that there might be

a reduction in the size of the regulatory

program--departments--and more resources that can

go into research.

CHAIRMAN KIBBE: I guess we're

getting--we're running out of time, and I think we

probably have--do you have something, Jurgen?

DR. VENITZ: [Off mike.]I always have

something.

[Laughter.]

CHAIRMAN KIBBE: I mean, do you want to

446

say something.

DR. VENITZ: [Off mike.] A question we

wanted to acknowledge or--

CHAIRMAN KIBBE: I don't know. Turn on

your mike.

DR. VENITZ: Two comments, then.

One has to do with the fact that I'm

concerned that we're trying to overreach. I mean,

FDA has only but so much impact on attrition rates,

on drug development. And I think the major part

of--not drug development, but the discovery part,

you have no control over. And you shouldn't have

any control over.

And my reading of those numbers that we've

seen, if I look at efficacy--30 percent fail

efficacy now, and they failed 10 years ago. Well,

maybe the wrong target was picked. Maybe we don't

know what the target does. Maybe we don't know how

the target is related to disease. That has nothing

to do with regulatory science. That has nothing to

do with product quality.

So I do think we have to kind of step back

447

a little bit and realize there's only so much of an

impact--no matter what your goals are, no matter

whether you reach them or not--that you can have an

impact.

The second one--and that's one that you've

heard me talk about for whatever--however many

years I've been on this committee--and that is to

really embrace this concept of risk; that risk is

something that is intrinsic to being alive. Being

alive is a risk because we're all going to die. So

the question then becomes: how can we quantify

risk? And how can we link that to--in your

case--product performance? And that, to me, is

really essential.

So all the rules that you come up with

cannot be driven by the ability to measure certain

things; certain what you consider to be critical

attributes. But they have to be really driven by

the fact that we think there is a reasonable link

between improving those attributes and some risk to

the patient; and that the stakes are high enough

for us to put all the resources in, in terms of

448

controlling that risk.

So--two comments; one, that there's going

to be a limited impact of whatever the Critical

Path Initiative that the FDA proposes will do;

secondly, that you really have to emplace this

concept of risk, and feed that back into your

critical attributes, and the whole cGMP change.

CHAIRMAN KIBBE: Ajaz has a comment.

DR. HUSSAIN: I think the discussion is

sort of coming together, in terms of giving us very

valuable insight in sort of the questions that we

need to pose.

If I may impose on the committee to--as

the Chair suggested--take the evening to think

about these.

But what I would sort of build on Marv's

and Jerry's presentation, and Vince's, is: I think

the key is the metrics, in the sense, I do believe

in this, since we don't want to overreach; we need

to understand where our impacts will be the most

positive, as Jurgen just sort of pointed out. And

we need to have some meaningful metrics to measure

449

whatever path we decide to walk on, and measure our

progress in that direction.

So if the committee members could think

about--from that--the discussion perspective now,

to sort of come back tomorrow to sort of summarize

some of their thoughts on some guidance on how we

should move forward here, it would be very useful.

For example, I think just building on

Jurgen's comments here, in the sense: where can

FDA have the maximum impact? And how can we

measure that? For example, I think--I look at this

slide here, and I say all right. Traditionally,

formulation was never an issue. Why is it showing

up as an issue now? Are the drugs more complex

that we're not able to--the product itself is so

complex? Or--so there are some indicators here

which were surprising, and so forth.

So if FDA has to have maximum impact, how

will we measure it? Multiple review cycles is one

measurement that we can look at.

For inhalation products, we have multiple

review cycles. If I look at our root-cause

450

analysis, the physical characteristics is a CMC

which leads to multiple review cycles as soon as

you have a drug and a device

combination--inhalation product. We cannot even

approve a generic product when it's inhalation

because of that level of complexity.

So--multiple review cycles, and reduction

of that could be a metric. I'm just asking you to

think about it.

Approval decisions--I think, with respect

to the example of PET imaging, how some of these

things impacted on approval decisions could be an

aspect that we could measure.

Clearly, I think, as we move towards

follow-on proteins, expand the generic programs and

so forth, within OPS we have a Congressional

mandated committee that we manage, which is a very

difficult task. It's the Therapeutic

Inequivalence. We don't have a good means to

manage that--reports that come in--because our

information is limited.

Keeping an eye on post approval reports

451

that come up--is that a means to measure that? I

don't know.

So I think if the committee members could

think about the discussion here, what metrics, how

can we measure this, and then come back with their

thoughts tomorrow, that will be wonderful.

DR. LEE: May I interject, also, I would

like to plead for the funding in the formulation

area. I think that was talked about this morning.

There's no department on pharmaceutical

technologies. And I think somebody should make the

case to support formulation in a big way.

CHAIRMAN KIBBE: Marv?

DR. MEYER: Just one suggestion--kind of

passing the buck, I guess--it seems to me it's a

little more efficient if some representatives of

FDA threw up a straw man tomorrow morning, because

they know what the problems are. They know what

potential solutions are. They've come up with the

Critical Path Initiative--throw up a straw man,

maybe with a couple examples, and let us hack at

that, rather than have us kind of out in a blind

452

somewhere try to come up with some harebrained

ideas that will be in the public record.

CHAIRMAN KIBBE: I plan on doing that

tomorrow, though. That was my whole thing with

tomorrow.

It would be nice to hear from our industry

reps, too. Because they have to live with the

challenge of finding better ways, and more

efficient ways of improving the quality of the

therapeutic moieties on the market, and doing it in

a constricted economic environment.

DR. MORRIS: Just to follow up--one think

I was thinking is that in some of the work we've

been doing with Ajaz's folks, and Helen's, we've

been looking at the CMC review process. And maybe

some of what we've been doing could be classified

as dividing it into opportunities for improving the

reviewing efficiency, versus real scientific

changes that have to be made to stimulate the

process--which I think sort of is reflected in this

slide here.

CHAIRMAN KIBBE: I think it's time to call

453

it a day.

You can turn off the tape, and then I can

say really weird things.

[Whereupon, the meeting was adjourned, to

reconvene on October 20, 2004.]

- - -