1

DEPARTMENT OF HEALTH AND HUMAN SERVICES

FOOD AND DRUG ADMINISTRATION

CENTER FOR DRUG EVALUATION AND RESEARCH

ADVISORY COMMITTEE FOR PHARMACEUTICAL SCIENCE

Volume I

Tuesday, May 3, 2005

8:30 a.m.

CDER Advisory Committee Conference Room

5630 Fishers Lane

Rockville, Maryland

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

Charles Cooney, Ph.D., Chair

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

Committee Members:

Patrick P. DeLuca, Ph.D.

Paul H. Fackler, Ph.D., Industry Representative

Michael S. Korczynski, Ph.D.

Gerald P. Migliaccio, Industry Representative

Kenneth R. Morris, Ph.D.

Marc Swadener, Ed.D., Consumer Representative

Cynthia R.D. Selassie, Ph.D.

Nozer Singpurwalla, Ph.D.

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

Special Government Employees:

Carol Gloff, Ph.D.

Arthur H. Kibbe, Ph.D.

Thomas P. Layloff, Jr., Ph.D.

Marvin C. Meyer, Ph.D.

FDA Participants:

Gary Buchler, R.Ph.

Lucinda Buhse, Ph.D.

Ajaz Hussain, Ph.D.

Mehul Mehta, Ph.D.

Vibhakar Shah, Ph.D.

Helen Winkle

Lawrence Yu, Ph.D.

C O N T E N T S

PAGE

Call to Order

Charles Cooney, Ph.D., Chair 5

Conflict of Interest Statement

Hilda Scharen, M.S., Executive Secretary 5

Introduction to Meeting--OPS Update

Helen Winkle 7

Opening Remarks

Charles Cooney, Ph.D. 16

Establishing Drug Release or Dissolution

Specifications:

Topic Introduction

Ajaz Hussain, Ph.D. 18

Dissolution Measurement System: Current State

and Opportunities for Improvement

Lucinda Buhse, Ph.D. 45

Questions by Committee Members 76

Overview of Guidance Documents

and Decision Process:

Biopharmaceutics Section

Mehul Mehta, Ph.D. 95

Questions by Committee Members 128

Establishing Dissolution Specifications:

Current Practice

Vibhakar Shah, Ph.D. 138

Questions by Committee Members 156

Open Public Hearing:

Will Brown, USP 162

Questions by Committee Members 171

C O N T E N T S (Continued)

PAGE

Establishing Drug Release or Dissolution

Specifications: (Continued)

Factors Impacting Drug Dissolution and

Absorption: Current State of Science

Lawrence Yu, Ph.D. 179

Questions by Committee Members 198

Summary of Tactical Plan

Ajaz Hussain, Ph.D. 208

Committee Discussion and Recommendations 229

Clinical Pharmacology Subcommittee Report

(via teleconference)

Jurgen Venitz, M.D., Ph.D. 284

Questions by Committee Members 301

P R O C E E D I N G S

Call to Order

DR. COONEY: I would like to welcome

everyone to this morning's meeting. We have an

opportunity for an on-time start. I am Charles

Cooney, the new chair of this committee. I am

delighted to welcome everyone here, both the

committee members as well as the guests. We have,

not surprisingly, a full agenda this morning and we

will begin with addressing the conflict of

interest.

Conflict of Interest Statement

MS. SCHAREN: Good morning. The Food and

Drug Administration has prepared general matters

waivers for the following special government

employees, Charles Cooney, Patrick DeLuca, Carol

Gloff, Arthur Kibbe, Michael Korczynski, Thomas

Layloff, Marvin Meyer, Kenneth Morris, Nozer

Singpurwalla and Jurgen Venitz who are

participating in today's meeting of the

Pharmaceutical Science Advisory Committee to, one,

receive an update from the Clinical Pharmacology

Subcommittee and, two, discuss and provide comments

on the general topic of establishing drug release

or dissolution specifications.

This meeting is being held by the Center

for Drug Evaluation and Research. Unlike issues

before a committee in which a particular product is

discussed, issues of broad applicability, such as

the topic of today's meeting, involve many

industrial sponsors and academic institutions. The

committee members have been screened for their

financial interests as they may apply to the

general topic at hand. Because general topics

impact so many institutions, it is not practical to

recite all potential conflicts of interest as they

apply to each member. 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

representatives, we would like to disclose that Dr.

Paul Fackler and Dr. Gerald Migliaccio are

participating in this meeting as non-voting

industry representatives, acting on behalf of

regulated industry. Dr. Fackler's and Dr.

Migliaccio's role on this committee is to represent

industry interests in general and not any one

particular company. Dr. Fackler is employed by

Teva Pharmaceuticals and Dr. Migliaccio is employed

by Pfizer.

In the event that the discussions involve

any other products or firms, not already on the

agenda, for which FDA participants have a financial

interest, the participant's involvement and

exclusion will be noted for the record. With

respect to all other participants, we ask in in the

interest of fairness that they address any current

or previous financial involvement with any firm

whose product they may wish to comment upon. Thank

you.

DR. COONEY: Thank you. Now Helen Winkle

will provide an update.

Introduction to Meeting--OPS Update

MS. WINKLE: Good morning, everyone. I

would like to welcome all the members of the

advisory committee and to especially welcome Dr.

Charles Cooney as our new chair of the advisory

committee. We, at FDA, have worked with Dr. Cooney

as a member of the committee and have really felt

that he has provided a lot of input into the

committee's activities, and feel that working with

him in the next two years as chair is going to be a

very important step for all of us.

Before I talk about the agenda for this

session of the advisory committee, I would like to

talk a little bit about our current focus at the

agency or what we are calling a paradigm shift. I

think it is important for all of us to understand

clearly the changes that we are making in the

agency and the role of the advisory committee in

assisting us in making these changes. Based on

recent initiatives in FDA, including the

Pharmaceutical cGMP Initiative for the 21st

Century, the PAT Initiative and the Critical Path

Initiative, you can see the shift in FDA's thinking

about regulating product quality.

Specifically, there is a focus in these

initiatives to place more responsibility on

industry to ensure the quality of their

pharmaceutical products rather than rely solely on

regulatory scrutiny to maintain that quality. This

is really the paradigm shift, a sharing of

responsibility for drug quality with emphasis

placed on industry to understand their processes

and the underlying science of those processes.

Why would we want to make that change?

There is no evidence that the products out there on

the market are bad products. There is no evidence

that the agency has done a bad job in serving as a

surrogate for ensuring good quality products for

the consumer. And, there is no evidence that

industry is not focused on quality as an important

attribute to manufacturing products. However,

times are changing. As we enter the 21st century

we have an excellent opportunity to begin to

prepare for how we will handle pharmaceutical

regulation in the future. The time is ripe for us

at FDA to invest in that future and to ensure that

the direction we are going in is adequate to handle

the changing world of pharmaceutical development

and manufacturing while we continue to be able to

serve the consumer. It is the right time too to

ensure that our regulatory involvement does not

hinder the innovation and continuous improvement in

manufacturing and ensuring the quality of

pharmaceutical products.

So, FDA has begun a journey towards this

paradigm shift. I want to say it is a long

journey. It started several years ago but we have

a long way to go, and we have numerous challenges

along that way. However, with these challenges

come opportunities and I think this is the

important thing for us and the advisory committee

to remember, that we need to take advantage of

these opportunities. It is important not only to

take advantage of the opportunities to help us

improve on how we regulate product quality, but

also to ensure that we provide for modernization

both at FDA and within industry for the 21st

century.

The guiding principles of the

Pharmaceutical cGMP for the 21st Century, which

include risk-based orientation, science-based

policies and standards, integrated quality systems

orientation, international cooperation and strong

public health protection, serve to help us in

developing the pathway to restructure the oversight

of the pharmaceutical quality. As each of you

knows, there are a number of forks in that path and

you, as members of this advisory committee, are

really here to help us determine the right path in

the road to go from a scientific perspective, and

to help and advise us on how to fill the gaps which

exist in the FDA. These include gaps in

organization, gaps in science and gaps in policy.

The committee has already participated in

discussions on a number of scientific issues which

have helped in formulating a strategy for

addressing many of the questions that have emerged

as a result of this paradigm shift. We have

already discussed a number of issues which have

significance as we develop our future regulatory

paradigm, including such issues as polymorphism,

bio-inequivalence of generic products, and we have

worked together to support such initiatives as the

Process Analytical Technology Initiative. The

committee has also been extremely helpful in moving

toward this new paradigm with other discussions

that we have had on various topics. However,

again, the journey has only just begun.

The agenda for the next two days was

developed to provide an opportunity to discuss two

other scientific topics which are important to us

to better understand and manage in order to move us

steadily along the path of change. The first topic

is on establishing drug release or dissolution

specifications. Obviously, how we set

specifications is important to the future. As we

move to the desired state of pharmaceutical quality

we want to ensure that specifications are based on

mechanistic understanding of how product and

process factors impact product performance. We are

currently in the process of developing a tactical

plan for setting dissolution specifications. As

you will hear from the presentations today, we have

developed the fundamentals for this plan which

include a systems view of setting specifications,

ensuring that all factors which affect dissolution

are considered; basically ensuring that we connect

all the dots in CMC to ensure a more comprehensive

and systematic way of setting specifications.

FDA recently held a specifications

workshop in co-sponsorship with the Product Quality

Research Institute. The workshop indicated a need

for additional efforts to move toward better

setting of specifications in general. Some of the

specific points that were brought out at the

workshop included a lack of globalization on how

specifications are set; a need to better define

what we should do versus what we can do; a need to

better define the role of the compendia in the new

paradigm; and a need to revisit the decision trees

in ICH Q6A.

Our discussion at the advisory committee

today is not designed to address all the workshop

issues and concerns on setting specifications. The

discussion today will, however, help us finalize

the tactical plan for setting dissolution

specification and will lay the foundation for our

thinking in setting specifications for CMC and

addressing the specific issues that were identified

at the workshop.

We would appreciate the committee's

comments and suggestions as to what data is needed

to support our plans. This would include looking

at statistical methodology, etc., and how we might

improve on our thinking in our tactical plan and

specifications setting in general.

At this meeting we will also discuss, as

our second big topic, quality by design and

pharmaceutical equivalence. As you will remember,

at the last meeting we set the stage in our

discussions on bioinequivalence and bioequivalence

testing of locally acting GI drugs. At this

meeting our goal is to modernize our general

thinking about pharmaceutical equivalence and to

explore how quality by design can be leveraged to

ensure more rational approaches to decision-making

so that we can move from a reactive environment to

a proactive regulatory scheme of assessing

equivalence.

We will discuss a number of relevant

topics, including biopharm. classification system;

using product development information to address

highly variable drugs; and we will revisit the

concept of decision trees for ensuring a rational

approach to determining bioequivalence for topical

drug products. We look forward to the committee's

feedback on these extremely important topics, and

that discussion is for tomorrow.

There are a number of other topics we will

cover at this meeting, including an update from the

working group on parametric tolerance interval test

for dose content uniformity. Bob O'Neill will give

that update. And an update from the Clinical

Pharmacology Subcommittee. We will also discuss

with the committee our perceived need to establish

a working group for the review and assessment of

OPS' research programs. Our goal in looking at our

research programs is to ensure a common approach to

all laboratory work and to ensure that our research

aligns with our overall mission.

Now that we have two laboratories in OPS,

a biotech. laboratory and a lab focused on small

molecules, it is extremely important that this

alignment takes place and we really look forward to

your input on how we can better align these two

laboratories.

As you can see, we have a full agenda but

I think the topics to be discussed are really of

great interest to us as we move down the path to

the desired state for pharmaceutical quality, and I

look forward to a very interesting discussion on

each of these topics. Thank you.

Opening Remarks

DR. COONEY: Thank you, Helen. I would

like to just add a couple of comments, if I may, to

get us started. I am delighted to have the

opportunity to work with the FDA and to work with

this committee during the coming two years. It is

a particularly exciting time because as we look

forward, as Helen has indicated, there are very

important new initiatives on the table with the

cGMP Initiative and the Critical Path Initiative

and these, indeed, lay a foundation that we all

need to work within. In fact, it is an exciting

opportunity to work within those initiatives to

look at how we can better address some of the

challenges going forward.

Certainly, as we look forward there are

more challenges than there have been in the past.

We are facing a world of increasing molecular

complexity; a world of increasing demands by

consumers; a world in which we have increasing

complexity not just in the molecules but in the

delivery formats of these products and the role of

this committee is very important in helping to

provide advice to the FDA and to the division to

deal with these problems. I must say, I applaud

the forward-looking and the proactive stance that

is being taken on these issues.

We have some challenging goals today and

tomorrow, not the least of which, of course, is to

stay on time. But the reason that the challenge of

staying on time is a challenge is because of the

very high content of material that we need to deal

with. So, I will do my best to try and keep us

within the proper boundaries.

Again, I look forward to working with

everyone. This will be a very interesting two days

and I see it as an important step in what will be a

continuing series of discussions and activities and

recommendations that we will need to work on with

the FDA. With that, the first presentation and

discussion this morning will be by Ajaz Hussain,

and we will begin by digging in to establishing

drug release and dissolution specifications. Ajaz,

please?

Establishing Drug Release or Dissolution

Specifications Topic Introduction

DR. HUSSAIN: Thank you, Dr. Cooney. I

think topic one is entitled quality by design

approach for quality control and assurance of

dissolution rate. In the background packet, as

well as in the presentations, I have tried to keep

the terminology dissolution rate all along to

illustrate the one challenge which we will not be

discussing today, and that is the metrics for

dissolution rate itself. We express dissolution

rate as a Q factor which tends to be confounded

with the variability of the assay itself. So, that

is not the topic for discussion today but I just

wanted to alert you on why the word "rate" keeps

coming back and back again. So.

Topic one: Our goal is to seek your

recommendations and endorsement of the proposed

regulatory tactical plan. With this tactical plan

we hope to start moving towards putting together a

set of regulatory tools and policies that will help

us define elements and details of the elements

necessary to realize the goals of quality by

design.

The question that we are posing to you is

are the tactical steps outlined consistent with the

goals that we have shared with you? What initial

steps and/or changes would you recommend to improve

this plan? What additional scientific evidence do

you feel would be necessary to support the

development and implementation of this plan?

General considerations for identifying and

developing statistical procedures and, in

particular, I want to emphasize for this discussion

today that we have left out a discussion on

statistical procedures because our experience has

been that if we start with that as a topic it leads

to protracted debate, and you will hear one report

on that debate from Bob O'Neill tomorrow, the

debate on parametric tolerance interval that has

been going on for years and hasn't come to any

resolution. We feel that if you approach it from

scientific foundations first, statistics is simply

a tool to implement the scientific decision

framework. So, that is the reason we have kept

that out of discussion for today. Clearly, we are

seeking your specific recommendations and other

recommendations that you may have including how we

should prioritize our work to develop this tactical

plan to a full proposal, which we hope to bring to

you at a subsequent meeting.

What are the proposed steps? The proposed

steps are to develop an alternate regulatory

approach to demonstrate the suitability of

dissolution measurements system; introduce and

utilize the concept of reproducibility and

repeatability study using the actual pharmaceutical

product for which we set specification. Here, the

proposal is to consider using the pivotal clinical

lot or the bio. lot as a basis for identifying how

sensitive, or lack of it, it is to a dissolution

test method and estimate the variability in the

method and, therefore, of the product.

So, the first two steps are sort of

together. But the next three steps are also sort

of in one clump. We want to move towards a

system-based decision tree for establishing

dissolution rate specification. Within that

framework I think we would like to utilize

opportunities to utilize the PAT approach for

controlling dissolution rate and development of

real-time quality assurance strategies. Also, a

decision tree for design-based concepts articulated

in the draft ICH Q8 guideline, which is in your

background packet.

So, those are the decision trees which we

would like to develop. At the same time, when we

come back with the full proposal to you, we would

like to bring to you a side-by-side comparison of

new and generic drugs because we think this is an

opportunity for both sides, and some of the

frustrations the generic industry feels can be

addressed with this proposal, and I will explain

that towards the end of the day, and explain why

the level of quality assurance or quality control

confidence in the proposed approach will be higher

than what is achieved in the current system. There

is no doubt in my mind but, clearly, you have to

agree with that.

We also seek today your recommendations on

how we should approach the statistical aspect of

this and then what will help you to discuss this

proposal when it comes to you. So please give us

your recommendations on how we should prepare a

detailed proposal for a subsequent ACPS meeting.

The other step that I think is important

and is very timely, because this Friday or this

Saturday I leave for Brussels for our ICH meeting,

is that we do intend to seek the discussion here,

and utilize the discussion here, to seek

harmonization of the approach we are proposing

under the ICH, especially the ICH Q8 Part 2, and we

will start developing that guideline in Brussels

next week.

ICH Q8 draft guideline essentially brought

a basis for getting and considering pharmaceutical

development information in a structured way for

pharmaceutical decision-making in the CMC arena at

FDA. The guideline was constructed with this

figure, on your right, in mind. You have to focus

your design efforts on the intended use of the

product, the patient population and so forth, that

leads to a product design and that product design

dictates the design specification, which are

customer requirements, and these requirements, some

of them if they are critical, become regulatory

specifications. Then the product design and design

specification dictates or leads to design of a

manufacturing process to reliably and predictably

deliver those specifications back to deliver the

intended use.

In a systematic way, if you approach

pharmaceutical development in a structured way, you

get some benefit, we believe. You achieve a higher

degree of process understanding and give regulators

high confidence of low risk of releasing a poor

quality product; high efficiency through continuous

learning and improvement. And, I think it helps us

to address some of the gaps we have in the current

system. I have tried to illustrate the current

process within FDA and the manufacturing and R&D

process within industry. Research and development

will develop the products, transfer them to

manufacturing and then we have, by law, a separate

quality unit to maintain quality assurance. You

have all the specification results and you have

products which don't have all the specification

results.

In approaching and assessing the quality

we bring a team approach, a multi-disciplinary team

approach which includes pharmacology, toxicology,

CMC review, clin. pharm., bio. pharm. review and

the clinical assessment. And the decision

collectively is a risk-benefit decision that leads

to an approval of a product. The approved product

is then transferred--it is called technology

transfer--and the process is validated.

The validation process includes

qualification criteria and so forth, but there is

an element of that which is process qualification.

Process qualification is essentially, in my

opinion, the interaction between materials and

equipment and environment that you really have to

study. In the current state that essentially is

judged on your ability to repeat it three times.

Since the pharmaceutical development information is

not available for CMC reviewers, the quality of

that and the understanding containing that is not

well understood in the regulatory sense. So, we

are losing an opportunity to make more rational

decisions.

Now, we have a divide, an organizational

divide within the agency between, say, review and

GMP inspection. The cGMP process is helping

bridge. The PAT is an example of how we have

bridged it. Our experience or learning from the

initiative clearly identified a need for a quality

system orientation. I would be wrong if I said

that I really did not understand what this really

meant or really didn't care about what quality

system issues were because I was looking from

several years ago. But I think I have gained a

much deeper understanding of the importance of the

quality system orientation.

So, here is a representation of that from

a paper that we published on innovation and

continuous improvement in pharmaceutical

manufacturing. Say what you do, do what you say,

prove it and improve it are the elements that make

up a quality system. Consider the way what you do

is your application to FDA. So, that is a CMC

review assessment process. Now, do what you say

can be considered as are you able to manufacture to

the commitments that you have placed in your

application? Now, there is a gap since our

reviewers don't have an idea that they really do

what they say. That is a GMP function so there is

uncertainty there. And prove it. How do you prove

it? I think metrics for proving it could be

process capability and the recalls, and this and

that. If you are unable to prove it you need to

have a collective action and a preventive action.

Our experience has suggested that in most

cases root cause is unknown or a poor analyst is

blamed. So, we actually don't get to a root cause

generally. Does the current system support or

facilitate getting to the root cause? I think that

is the question. In many ways I think say what you

do and do what you say, if you take that ratio is

process understanding and your ability to prove.

So, in many ways I think you have to think about

that.

Now, a modern quality system has a

dimension of improvement, continuous improvement

and innovation. The dotted line simply says that

is an option that should be available for industry

to do. It is dotted because that is not a

requirement per se, but the rest are all

requirements. So, I think we are trying to address

some of these gaps along the way.

Now, the definition of continuous

improvement is interesting and it really sets the

stage for this discussion. I have taken the

definition from QS-9000 to illustrate the challenge

we face for continuous improvement. For those

product characteristics and process parameters that

can be validated using variable data, that is

continuous data, continuous improvement means

optimizing the characteristics and parameters at a

target value and reducing variation around the

target value. So, in a sense, you need a target

value and you need to have an estimate of variation

to start thinking of continuous improvement. In

our specification setting often we don't even have

a target value. So. And forget variation. So.

But the second bullet is more important.

For those product characteristics and process

parameters that can be only evaluated using

attribute data, pass/fail, continuous improvement

is not possible until characteristics are

conforming. If attribute data results do not equal

zero defect it is by definition a non-conforming

product. Improvements made in these situations are

by definition corrective actions, not continuous

improvement. And, we have clearly distinguished

between corrective action, which is a risky

scenario, and continuous improvement, which can be

managed differently.

Continuous improvement in processes that

have demonstrated stability, acceptability,

capability and performance--continuous improvement

really is only possible for those products that

have demonstrated stability. Process validation

today does not give us the assurance that the

process is stable. So, that is another element.

Acceptable capability, we don't have an estimate of

the capability value.

Now, the reason for finding this out is

that I think we don't use compendial methods as

release specifications. Actually, the compendium

approach to specifications is right. That is the

way they should be. There is nothing wrong with

this specification criteria for the market

standard. It is perfectly all right. But it is,

as Janet Woodcock says in her paper, different from

release specification and that is the

distinguishing feature that I think is the problem

here. If you use market standard as release

specification, then you have all the elements that

hold back continuous improvement. So, you really

need to distinguish between standards and

specifications. Unfortunately, in the current

paradigm specifications equals standard. So, what

we are moving towards is a control strategy that

will allow you to have your market standard but

then have a control philosophy that allows you a

risk-based decision process.

A recent proposal from USP I think is a

step in the right direction. It is essentially a

similar proposal to the parametric tolerance

interval test to take dissolution specification

criteria towards more of a tolerance interval

approach. But as you will hear tomorrow from the

parametric tolerance interval discussion, you

cannot approach it as hypothesis testing for every

product batch, and that is one of the discussions

that we will have. And, there are many challenges

before we even can get to that, and that is a part

of this discussion. We believe one has to start

with a pharmaceutical science discussion before

developing appropriate statistical tools.

One other challenge for continuous

improvement is the mind set--and this is a major

challenge not only within the U.S. but

globally--that corrective actions is the only way

to force improvement of quality on industry. This

is direct current paste from the paper that we

issued. Some would argue that corrective actions

provide the necessary constancy of purpose for

improvement, necessary since manufacturing is a

stepchild of industry because the difference

between cost of manufacturing and price of drugs is

large. Keeping the system of corrective action

provides the leverage for ensuring improvement, to

ensure the cGMP.

That is a fundamental challenge. How do

we achieve that? If you improve your manufacturing

process by reducing variability your regulatory

acceptance criteria will be narrow. So, that takes

things into a way for continuous improvement. So,

that is another challenge that we will start

addressing.

The argument has some validity but it is

based on an assumption that current practices,

including measurement systems and product

specification, provide efficient means for

identifying, understanding and then reducing

variability. For quality assurance in the 21st

century we need a sound basis to verify such

assumptions in the current system.

To emphasize this point further, we

discussed the case of dissolution and that is what

we present to you today. Let me illustrate an

example, a real case example. This is an example

of an approved and validated manufacturing process

at a major pharmaceutical company. I will read the

middle portion of this. This is the warning

letter: There is no assurance that the production

in process control procedures established--this is

controlled-release product--to produce a product

that has the quality it is purported to have or

represented to possess. How did we approve it?

How was it validated? So, this is after the fact.

The duration of each coating cycle is determined by

the pan operator but is based on visual

determination that the coating solutions are evenly

distributed before proceeding to the next step. It

is noted that literally 50 percent of the batches

are thrown out every year because of dissolution

failures, and then you have partial release

occurring too. Doesn't this undermine the entire

credibility of our system? And, this was

catastrophic for the company.

Now, inability to resolve our

specification observations I think undermines the

credibility of our decision system. It raises

questions of adequacy of the current decision

system. It increases the risk of releasing an

unacceptable quality product to the consumer, and

contributes to low efficiency.

Now, corrective action, preventive

action--there are some challenges. There are

difficult questions faced by manufacturing groups

and regulators since we have a calibrated system

that we use for dissolution and a calibrated system

is a tablet similar to any other tablet that we

use, and the quality is an issue there. If you

choose to use a calibrated tablet for gauge R&R

study, reproducibility and repeatability study,

what you see there is that the calibrator

variability and its manufacturing process is

confounded within that system. I am not going to

go through the equations but it is simple algebra.

In addition, we have another challenge.

The challenge is that the assumption of independent

variable cannot be really verified because the

hydrodynamics of the vessels are such--I see our

colleagues from Health Canada here who have been

criticizing this for a long time. Thank you for

coming, sir. So, how representative is the

suitability for that product is an issue.

But the need for improvement is not

limited. We need to be confident of our analysis,

of surveillance samples, consumer complaints, other

investigations. One of the frustrating jobs that I

have is where we get consumer complaints; we do

investigations; we do dissolution--no answers. I

mean, you really don't get to the root cause.

I think the basic philosophy that Walter

Shewhart sort of proclaimed years ago is very

important. Pure and applied science have gradually

pushed further and further the requirements of

accuracy and precision. However, applied science,

particularly in the mass production of

interchangeable parts, is even more exacting than

pure science in certain matters of accuracy and

precision. That is the basis of this discussion.

Is the current approach to calibration

adequate? Dr. Cindy Buhse will share with you her

challenges--as one of the premier labs, probably

the world standard for dissolution at FDA and

elsewhere, and Tom Layloff had started some of

these processes and he is here too--dissolution

testing of the USP wants to require diligent

attention to details, mechanical and chemical.

Dosage forms can respond definitely to small

variations; large differences in dissolution

results are possible unless all parameters are

carefully controlled. Differences in

reproducibility can often be traced to improper

mechanical calibration or degassing. Much of that

is mechanical. When you only have suitability

criteria just based on a tablet, it hides some of

this variability.

We had a rude awakening to this ourselves.

This is really when I started realizing the

confounding nature of the problem that we have.

Just to illustrate how frustrating this experience

was, our marines were contracting malaria when they

were in Liberia and we were asked to see whether

this was a quality problem. We faced significant

challenges in analysis because I had insisted that

two labs would do this because this was a grave

situation. Unexpected inter-laboratory differences

highlighted limitations of current calibration.

Here is just a quote from our DPA lab: We are at a

loss to explain the difference between DPA and the

Philadelphia district office initial results. Then

we started tracing it back. It had to be

mechanical differences and degassing.

Well, I think that is not the only issue.

I think the bigger issue that we are confronted

with is that we need to better understand the

sources of variability in product performance and

quality so as to establish the most appropriate

design specifications for the product that support

continuous improvement and address the increasing

complexity of product designs.

This is another concern. We are moving

towards drug eluting, towards nano materials,

towards other complex devices and, yet, we don't

have good measurement systems for these products.

We want measurement systems for products intended

for non-oral administration and non-oral drug

delivery systems; develop and implement globally

harmonized proactive regulatory decision system,

including Q6A and Q8.

I just want to sort of lay the foundation

for other aspects that Mehul and Vibhakar will

share with you. Pharmaceutical development and

dissolution specification without pharmaceutical

development information creates more challenges.

Decisions focus only on dissolution test data.

Tests are often used for both in-process control

and final product testing. Decision

characteristics focused only on the mean value will

deal with variability indirectly. Variability

managed indirectly using "disconnecting test

conditions" and acceptance criteria leads to

deterministic interpretation of specifications and

ignores background variability and, as Dr. Woodcock

has said, we need to move towards a probabilistic

decision system. Specifications are standards and

standards don't give any room for uncertainty or

risk-based decisions. If you don't meet the

standards, you are off the market. It is as simple

as that. So. And you have event trees as opposed

to decision trees. It is difficult to resolve

specification observations which could be related

to how we set specifications, and post-approval

changes and optimization in continuous improvement

is difficult.

This is simply an illustration of the gap

that we base all of our decisions on test-to-test

comparison, in vivo to in vitro, and there is an

opportunity to use the design information to make

rational decisions. Just to illustrate this, again

this is from Health Canada which has been very

proactive and pushing this agenda and I am sorry we

just didn't react more quickly--here is an

illustration of the false-positive and

false-negatives that you get. The reference

product dissolves 95 percent in 15 minutes, and the

reference AUC, Cmax. But if you look at product F,

it dissolves very slowly in vitro but, yet, in vivo

it meets the criteria--it is almost identical to

that.

So, there is a formulation attribute that

does this. For example, if you have a large amount

of organic or insoluble excipient it is a

hydrodynamic effect. That doesn't happen in vivo.

The in vivo media, the surface tension, the

hydrodynamics are completely different. So, you

tend to see this but you also get false-positives

and false-negatives. If I look at product C, it

has only 62 percent dissolution compared to product

F and has half the Cmax.

There are other differences in how we

approach specification setting. The difference

between the U.S. and Japan--we included a paper of

the Japanese perspective on this in your background

packet. Because of the new restrictions I took the

names off. I had to go back and erase those. This

is a published paper so I was surprised that I

needed to take the names off.

The point here is this, all are basic

drugs and this is a rule of thumb that has been

known for 30 years, if you have a drug with PK

between 4-6 the best media to illustrate in vivo

performance is that of the PK value. That is where

the dissolution is slower. So, the Japanese have

been in that direction. All our specifications use

0.1 normal, here. Is that important? Well, the

Japanese think so because they are very concerned

with hypoacidity in the subjects. If I really look

at it, with antacids and H2 blockers most of us are

hypoacidic too. So, is this a gap that we need to

fill is the question that I think we will address

as we go along. So, you can see the dramatic

difference in dissolution as pH 1.2 to pH 7.2 and

the resulting blood levels.

So, in a sense, the opportunity we are

trying to realize is ICH Q6A actually had it quite

nicely captured in this quote: The quality of a

drug substance and drug product is determined by

the design development, in-process controls, GMP

controls, process validation and by specifications

applied throughout development and manufacture.

So, you have the goal; you have the decision

characteristics; and you have the life cycle. The

design development was the missing element in our

decision characteristics. Now we have an

opportunity to use it more effectively.

This is how ICH Q8 captured that

opportunity, to bring the development and design

information not only to ask the right question but

also to realize the opportunities of flexibility

that might bring. So, design and development

should impact positively on how we set

specifications in process controls and have more

confidence in process validation and GMP controls.

With that as a background and the reason

for this topic for discussion, in many ways the

tactical plan is an attempt now to go back ten

years and to see how we can do better with our new

information that could come through the PAT process

and the ICH Q8 process. In many ways we are

reexamining the SUPAC guideline, the dissolution

guideline for '97, the biopharmaceutics

classification assumed in ICH Q6A. The vector for

the desired state is that we are adding another

layer of variability assessment, identification

assessment and utilization of variability in our

decision-making. So, the basic fundamental is that

the quality of decisions can only be better so the

current system is the minimum level of quality that

we achieve.

So, for the discussion today Cindy Buhse

will share with you her proposal on measurement

system, how mechanical calibration will be better

and that is what we want to use. Mehul Mehta will

share with you the general overview of our decision

process in our guidances. Lawrence Yu is one of

the leading experts I think in sort of modeling

dissolution and in vivo absorption. So, I have

asked him to share a perspective on the current

state of science. Then I will come back and

outline the steps of the proposal. I have a number

of slides in your packet but I will not be using

those slides. I will be using only the first 16

slides to give you ample opportunity for

discussion. Those are backup slides. If there are

questions I will come back to them.

In your background packet I specifically

identified one person by name for his

contributions, and that is Dr. Vinotcha [ph.]. I

think the work he has done in particular--the

reason I am pointing him out today is because he

has decided to retire and I want to recognize his

contribution to dissolution. He has brought it to

this level and I think taking it beyond that, and I

thank him for that and he is here today. A number

of people are there from DPA who are experts in

this and I will recognize them at some other point.

With that, I will stop and invite Cindy to

share her thoughts with you. Any questions before

I leave?

DR. COONEY: Thank you, Ajaz. We will

certainly have time for extensive discussion later

but I think, particularly since we are right on

schedule, if anyone has any questions for Ajaz

right now, particularly for clarification of any of

the points he has made, this would be a very

appropriate time to take a moment for this. Ken?

DR. MORRIS: Yes, just one quick point on

an early slide where you were talking about the

development process, it actually goes from the

intended use through to development. I would just

say for clarification, because this is something

that I get quite a lot, what we really want to get

across I think is the idea that when you have the

intended use and the characteristics you really

select your process first.

DR. HUSSAIN: Yes.

DR. MORRIS: And then come back to the

formulation. So, it doesn't necessarily change the

order but it adds a level because that is a

constant source of confusion, particularly when you

are talking about building in dissolution

characteristics.

Dissolution Measurement System:

Current State and Opportunities for Improvement

DR. BUHSE: Thank you, Ajaz. It is going

to be my job to tell everybody a little bit about

dissolution. Some of you, I know, are very

familiar with it but some of you may never have

experienced it or seen it done and it is kind of a

very different way of testing so I am going to show

you a little bit about the different apparatus you

can choose to do dissolution testing; talk a little

bit about how we currently determine instrument

suitability in terms of calibration, both

mechanical and chemical; and also validation of

dissolution of test methods and what we typically

see in our lab when we take a look at method

validation packages. Then I am going to show you

some sources of variability within dissolution,

show you examples of how some formulations are

sensitive to some parameters and some formulations

are sensitive to others and we really need to

understand for your particular formulation where

your sources of variability are coming from. Then

I will just briefly talk about some opportunities

for improvement, many of which Ajaz already alluded

to in his talk.

If you go to USP, there are seven

different dissolution apparatus listed. They are

all up here. You can see that the ones I am going

to talk about today mostly are apparatus 1 and 2

because those are the two that are used the most by

most pharmaceutical companies. We do see some of

the other apparatus occasionally. Apparatus 3,

reciprocating cylinder, can also be set up for

apparatus 7 so those are actually the same piece of

equipment. The flow-through cell is used more in

Europe than it is in the United States. We don't

see much with that here. Then, apparatus 5 and 6

are used for transdermal delivery systems and they

are actually a modification of apparatus 1 and 2.

What I am going to talk about most today

is apparatus 1 and 2, which is actually the same

piece of equipment and what you are doing is you

are changing the shaft on the different vessels to

change it from apparatus 1 to apparatus 2.

Actually shown in the picture there is apparatus 2

and you can see there are paddles above each one of

the about 900 ml vessels there. The way

dissolution works is that you are actually testing

6 tablets at once. I think Ajaz showed that in the

specifications there usually is a specification

which says 6 tablets have to have a certain

dissolution value and if one of those 6 fails you

go to 12 tablets and then you go to 24. So, you

start with just 6 and if everything goes right,

then you will be done after the 6 tablets.

So, you essentially have 6 different

pieces of apparatus here because each one of those

vessels acts independently. You would fill each

one with whatever media it is that you want to test

in, whether it is 0.1 normal HCL or water or

simulated intestinal fluid. There are all sorts of

ranges of media that people use. So, you put

500-900 ml in these vessels and then for apparatus

2 you just lower the paddle down and start it going

at whatever rpm you decide. Certainly, that is

another variable you can manipulate. Then you drop

your tablet or capsule in and then you take a

sample out of the media at whatever time point your

specification is. If your specification may be 80

percent dissolved after an hour, then after an hour

you would withdraw a small portion of the media and

then you would determine how much the drug has

dissolved. Usually the determinative step there is

HPLC. So, you do that for all 6 of these vessels

and then, hopefully, everything dissolves in the

right amount of time and you will be done.

the basket--similar. You just change the

shaft and you put a basket on and you actually put

the drug in the basket and then you lower the

basket and start it spinning and you go through the

same procedure.

Just so you can see what it looks like,

this shows you what apparatus 3 looks like, which

you can also turn into apparatus 7 by changing the

holders. You actually would put the tablet or

capsule inside each one of those up at the top.

What it does, it comes up and down inside each one

of these little vessels down at the bottom. What

you can do with this apparatus is you can change

the media so in every row you can put a different

medium if you want. So, if you want to start your

capsule dissolving at 0.1 normal HCL and move it to

simulated intestinal fluid, in the first row you

could put acid. In the next row you could put

intestinal fluid. In the next row you could put

whatever you want. Then you can move this

apparatus up, you know make it go up and down for

an hour in one and then move to the next and go up

and down. So, that is how you could do it with

apparatus 3.

This is apparatus 4, and I think I

mentioned we don't see a lot of this one. This is

a flow-through cell. You can see over there, on

the far side, that is what the actual cell looks

like. So, if you had a capsule or tablet that

didn't completely disintegrate you could put it in

this cell and actually flow through, somewhat like

actually happens in humans--flow through a media

and change it as you go. You can either recycle it

around or you can actually have a one pass through

media as well and then analyze the media as it is

coming out to see how much drug is dissolved. For

this one there is also a bunch of different cells,

different geometries that you could put in this. I

kind of show examples of that there.

Most of what I am going to talk about

today is apparatus 1 and 2, and that is because

that is the majority of what we see in methods that

are given to us for method validation. When they

use apparatus 1 or 2 they use the USP criteria for

setting up the equipment and for calibrating the

equipment, and I will go over what those parameters

are. Then, as I think Ajaz said, most tablets and

capsules have a one point acceptance criteria. For

immediate-release products we see anywhere from 2

to 4 time points, maybe 1 hour, 4 hours, 8 hours,

24 hours depending on the product.

The first thing you are going to do if you

have one of these apparatus, you are going to run a

test method. You need to ensure that you have

instrument suitability. The first point I have up

there is which one of these 7 instruments you are

going to use. What we find is that most people use

1 and 2. Most people believe that that is what the

FDA wants to see. I have been to many different

dissolution conferences and, you know, consultants

and companies will get up there and say if at all

possible use apparatus 1 or 2 because that is what

the FDA wants. I have heard many people say that

so a lot of people try to use 1 and 2.

Then, once you have chosen your

instrument, you need to make sure it is set up

properly for mechanical calibration. You can see

by the picture that if your shaft is not quite

centered, or if your vessel is not quite seated

right, your rpm aren't calibrated, etc., you can

imagine that you can get different hydrodynamics

from vessel to vessel or from time to time. You

need to really carefully make sure that everything

is set up properly. Then, once you have everything

set up properly, you can then run a calibrator

tablet provided by the USP to see if you get within

the range that the calibrator tablet says you

should get. Then that gives you some measure of

confidence that perhaps you have set this thing up

properly with mechanical calibration. I think Ajaz

has mentioned that the calibrator tablets actually

are U.S. phenomena and they are not used either in

the European or Japanese pharmacopeias.

Once you have instruments all set up, then

you can certainly do method development/method

validation, and I will talk a little bit about what

we see and what is actually given to us, as the

agency, when it comes to validating the dissolution

method.

Here is an example of some of the

mechanical calibration parameters out of the USP.

Some of them have specific values. For instance,

the shaft has to be 2 mm from the centerline, which

means you actually have a 4 mm spread because you

can have one direction and then it spins around to

the other. You can see there are other parameters

which don't really have any hard numbers associated

with them, such as the wobble--no significant

wobble and that is kind of nebulous there, or no

significant vibration. So, those are the some of

the USP criteria for setting up the basket and

paddle methods.

The actual calibrator tablets--actually,

our lab in St. Louis had a lot to do with

calibrator tablets coming into being. It is

certainly the current 10 mg one that is used today.

But they came around in the 1970s and there are two

different calibrator tablets. One is

disintegrating and one is non-disintegrating. So,

one pretty much falls apart when it goes into the

dissolution apparatus; the other stays together as

a tablet throughout the calibration procedure.

In 1997 a 50 mg prednisone tablet, the

disintegrating one, was replaced with a 10 mg

tablet which was manufactured at the University of

Maryland, here, and was based on the formulation of

a product that our lab had found was sensitive to a

lot of the parameters of calibration, including

degassing and mechanical calibration, so we thought

it would be a good calibrator tablet.

Actually, last year the working group at

the USP was actually looking for a replacement for

the 10 mg tablet. It does have quite a bit of

variability associated with it and some stability

issues so they would like to see if they can find

something else.

So, if you are actually calibrating your

apparatus what you would do, if you use your

equipment for both basket and paddle which is what

we do in our lab--a lot of pharmaceutical companies

will have one that will always stay paddle and

another will always stay basket but we go back and

forth. If you are using the same instrument for

both paddle and basket, what you would do is you

would do 4 different calibration runs. You would

do both calibrators with the paddle installed and

then you would turn around and do both calibrators

with the basket installed to make sure that your

instrument is set up properly.

How often do you do these? In our lab we

do it every 6 months. We do the calibration using

the prednisone 10 mg tablet. Here is the actual

data on the current lots of calibrator tablets.

The O lot, which has been in effect now for almost

two years I think--you can see there are different

dissolution criteria depending on whether you are

running it in the basket or the paddle method. You

see there is a fairly wide range. You can see that

for the basket as long as you are anywhere between

53-77 percent for each vessel you are going to pass

calibration. So, you have your 6 vessels and this

one, over here, can be 53 and this one, over here,

could be 77 but you are still going to pass

calibration. Actually, late last year they changed

the ranges of the prednisone tablet because there

were stability issues and a lot of failures in the

market, and you can see that the range is even

wider now, 51-81 percent.

I have also included up there the values

we get in our lab for at least the prednisone

tablet. For the basket method we get 72.6. You

can see we run very much on the high end of that

range. In fact, we do quite often fall out on the

high end. You can see we tend to run on the low

end of the range on the paddle method for these

calibrator tablets.

The salicylic acid tablet has a much

narrower range. It is also much less sensitive to

many of the parameters that you set for dissolution

testing so it is not sensitive to degassing; it is

not sensitive to mechanical calibration setup.

The problem often with running these

calibrator tablets is if you do get an out of

specification value, then what do you do? You

check your mechanical calibration. It can be

difficult to decide whether the issue is the actual

calibrator tablet itself or the issue is some way

that you set up the instrumentation.

The other problem with the calibrator

tablet is that you can see it has a fairly wide

range. It can often interfere with a continuous

improvement process. If your vessels can be

anywhere from 51-81 percent and you are still

passing, what does that say when you are running

your own product and you want to try to narrow down

the variability of your product? You don't have

much room here I guess to try to keep everything

consistent.

I am going to talk just a little bit about

development and validation. We don't see a lot of

development data but we do see the validation data

in our lab. Obviously, when you are developing a

dissolution method you have to decide about all

these different parameters, a lot of which I have

alluded to, and you want to develop a method that

is going to be discriminatory. You want to be able

to tell between good product and bad product. You

want the method to be repeatable. You would like

the method to give you the same results no matter

which lab you are running it in. I think Ajaz said

we had some trouble with the malaria drug in trying

to get two different labs get the same results.

You have to decide which instrument to use. Like I

said, most people try to pick 1 and 2 if at all

possible; then what media to run it in. A lot of

the test methods we get either are in 0.1 normal or

HCL; a lot of them are just plain water. Then you

have to decide whether degassing is going to be

important or not for your product; and decide

whether or not you need sinkers. Some products

don't automatically go to the bottom of the vessel

if you are using the paddle method. You can buy

commercial sinkers, which are these little devices

that you put the tablet in that will actually make

it fall to the bottom, or you can just wrap a wire

around, which is what is in the USP, to make it go

down to the bottom.

Once you have decided all these

parameters, you still need a determinative step,

and that is what the main focus of validation is

for most companies. So, when we get validation

packages in from companies on their dissolution

test methods, their validation really focuses on

the determinative step. They do a lot of work on

varying the parameters on the HPLC method but less

data do we see on varying the parameters on the

actual dissolution method. So, we see more on the

determinative step and less on the actual

parameters that are associated with the dissolution

apparatus.

You can see that there are a lot of places

here where variability can be introduced, and

certainly when developing a product if you want to

have a test method that is going to allow you to

continuously improve your product you really need

to understand what all the sources of variability

are going to be.

This is one of Ajaz's slides. I think he

showed a similar one earlier which is basically a

slide just to show you that the total variability

you are going to see in any test method is going to

be the variability that is inherent to your product

and your manufacturing process and the variability

that is inherent to your test method. For

dissolution the variability inherent to the test

method can be quite large, especially if you don't

understand how all the different parameters can

affect your product.

I am going to just show some examples of

some of the variability. You can see I have a lot

of information up on this slide, and every single

one of these bullets can be a source of variability

when running a dissolution test method. You have

to make sure your operators are well trained. You

have to make sure you have set things up properly.

You have to make sure that you understand how all

the different media and equipment parameters,

sinkers etc., can affect the variability of your

specific product. So, there are a lot of places in

here where, you know, if you add a tenth or so, or

a percent or two of variability by the end you have

quite a wide range of potential dissolution

parameters you could get even with the same lot of

material.

When it comes to mechanical calibration, I

think I showed some of the USP parameters earlier

and what I want to show you here is actually that

in our lab, DPA, we use more stringent mechanical

calibration than what is listed in the USP. A lot

of the criteria we use come directly out of the

PhARMA recommendation. I think that paper is in

your packet. It came out in the '90s, where they

did a collaborative study to take a look at

mechanical calibration a little more closely to see

if tighter mechanical calibration might reduce

variability when running the calibrator tablet.

Because we run so many products in our lab

and we don't necessarily have the time to stop and

see if this product is really sensitive to

centering or not, etc., we just try to be very

careful about how we set up our equipment. Some

tools are now available to very easily set these

parameters much tighter than what is currently in

the USP. So, you can see that for quite a few of

these we are tighter, and for others we have added

criteria that are not actually in the USP as

specific numbers. For instance for shaft wobble

and vibration, we actually measure those and set

criteria for those.

Degassing is one of the things I think

that really got us into trouble--I don't want to

use that word, but with the malaria drug the

different labs were degassing in different ways and

this drug happened to be very sensitive to

degassing. So, typically in the past the way you

decided whether your media was well degassed or not

is that you ran the calibrator tablet. The 10 mg

prednisone is very sensitive to dissolved gasses in

the media so if you weren't sure if you were

degassed or not you could just run that calibrator

tablet to see if you were in range and then decide

if you were degassed properly.

Well, it turns out that there is some

equipment on the market that you can use to

actually measure dissolved gasses so this is

something we have done recently in our lab. We

have taken this meter, which is actually used in

other industries and not in the pharmaceutical

industry, and used it to try to determine how much

dissolved gases are left after using different

degassing techniques.

There are many different ways in which

people degas their media. The reason you need to

degas your media is because there are some products

that if you take a vessel and you drop in a tablet

or capsule, what will happen is you have gases in

the media. The bubbles will form around this

tablet or capsule and oftentimes will prevent it

from dissoluting. So, you actually need to get the

gases out of there before you start.

Here is a little graph of the different

ways people degas and the results we got with the

total gas meter, measuring both total gas and

oxygen. You can see that for the first bar over

there that is obviously atmospheric pressure and

atmospheric oxygen in the media. These are all

done in just plain water. The next bar is the way

we degas at DPA, which is point of vacuum at less

than 150 ml of mercury with agitation, and you can

see we get rid of about a little more than half of

the total dissolved gases and quite a bit of the

oxygen.

The USP method is also very good. There

you are heating up to about 41 degrees and

aspirating to remove the dissolved gases. They

also get half the total gone and about half the

oxygen.

Some people actually helium sparge and you

can see helium sparging and although you do reduce

the oxygen significantly you do not reduce the

total dissolved gases.

So, does this matter or not matter? You

know, this all depends on the product you are

testing. So, I just want to show you some examples

here. These are 3 different products, called

product 1, 2 and 3 so I don't give any product

names. You can see that for product 1 and product

2 there is a huge difference between non-degassing

and degassing. For both of those graphs I have

shown two different ways of degassing. One is the

USP and DPA method, both of which give similar

results. The other is helium sparging. You can

see in both cases that the helium sparging does

give slightly higher results than either the USP or

the DPA method. Certainly, for product 2 helium

sparging gives much more variable results than the

DPA degassing. You can see that on the helium

sparging line which is kind of the green-yellow

one.

You can see that product 3 doesn't really

care whether you degas or not. One of those lines

is non-degassed and one is the DPA method which had

the lowest percent of dissolved gases. You can see

that you get essentially very similar dissolution

whether you degas or not.

Larger than just degassing is the actual

composition of the media. I think as Ajaz

mentioned, Japan is looking at what type of media

you actually want to be using. We see a lot of

acid here and some buffers. Here is a product and

the dissolution method is pH 7.2. So, 7.2, as you

can see on your left I guess, is the media that is

used in this product. It also turns out that with

these 6 different tablets there is some variability

between the 6 but they all passed the dissolution

specification for this particular product.

This is a product where we wanted to take

a look at some lower pHs just because there are

some patients who happen to use this drug who may

have lower intestinal pH than 7.2 and so we went

down to 6.8 and, lo and behold, every single tablet

looked different to us and no two tablets were the

same. We repeated this over and over again, trying

to figure out what is going on. You can actually

do a lot with dissolution by just watching your

product. There is nothing like the human eye

sometimes.

If you watch this product in the vessel

what you will see is that it sits there and does

not dissolve and you get no dissolution until you

see the coating split open. Once the coating

splits open, then it dissolves fairly quickly. So,

taking a look at that we were trying to figure out

what could be the sources of variability of this

product. Is it the way we are handling it when we

put it into the dissolution vessel? Are we

damaging the coating in some way? Are these tablet

differences real or is this the manufacturing

process itself? Do we have instrument variation?

These 6 tablets are in 6 different vessels so is

there some difference in these vessels where maybe

we have improper calibration or something?

Well, after much investigation, what we

found is that this is actually a product problem.

If you cut open these tablets and take a look at

the coating, not all of them have uniform coating.

You can see there, on the left, one of the tablets

that has a very uniform coating thickness. Then

every once in a while you ran across a tablet that

had a void between the drug and the coating. The

drug is actually on the left side here; it is kind

of the yellow sparkly stuff and the red is the

coating. So, some of the tablets had very uniform

coating; some of the tablets had defects. These

defects were dissolving much faster or were

breaking open, splitting open much faster than the

ones that didn't have defects. This is a situation

where perhaps dissolution could help this

manufacturer make a more consistent product if they

were doing their dissolution at a slightly

different pH or doing a dissolution test method at

several different pHs to try to make sure they were

making a consistent product.

I was just going to mention sinkers

because I talked about them and also because they

do make a big difference. The graph up there

actually has nothing to do with the sinkers but it

shows you what happens if you don't get your tablet

at the center of your vessel. The bottom blue line

is product 1, right down at the bottom of the

vessel, centered completely. The green-yellow line

is if it is off center by 1 cm. So, if it is just

off center by a centimeter you can see that it

dissolves much faster. There are different

hydrodynamics in that area than at the bottom of

the vessel. So, if you have a tablet that is

fairly light and is not going to stay put, then

often you will put it inside a sinker.

Traditionally, in our lab we have used the

sinker at the top to the right. That is the one

that we have used in our lab. It is very easy to

use. It has a spring load and you just pull back

the spring and drop the capsule or tablet in and it

is, you know, very convenient I guess. The USP

method is to use a wire and wrap the wire three

times around the tablet or capsule.

Well, we did run across a product--this is

what I talked about, that you have to understand

your product and how it reacts to different

variables--that was sensitive to this actual

commercial sinker. This is the product we tested

and with the commercial sinker that I just showed

you it failed dissolution. The specification here

was 80 percent at 30 minutes and you can see that

all 6 tablets failed. Of course, we thought the

product was perhaps a failure but it actually

turned out that if you visually looked at what was

going on, the product was being trapped. It was

swelling up and getting trapped inside that

commercial sinker and so it could not essentially

dissolve.

We went back to the USP method with three

wire turns around the tablet, and you can see that

the product passes wonderfully with no problems

whatsoever. So, we no longer use commercial

sinkers in our lab but a lot of people use them so

I just wanted to make you aware of the fact that

something as simple as a sinker can affect the

individual product that you are looking at.

So, what I have tried to show you is just

some data that illustrates the fact that different

products are sensitive to different parameters when

you are doing dissolution, and there are obviously

a lot of places where you can introduce variability

in your test method. What we would like to propose

is an alternate approach to calibration and

validation which includes complete understanding of

how dissolution and the measurement system in your

product specific variables affect variability, and

try and understand the relationship between your

product properties and your dissolution results.

This includes understanding the dissolution

apparatus that you are using, why you are choosing

it and why you are choosing the media you are

choosing, and determine, hopefully, the best method

to give you opportunities for improvement and to

ensure that the quality of your product is good.

You can see that because of the way

dissolution is currently set up there are a lot of

things you have to control, and perhaps there are

new approaches we can also use to get the same type

of information that might have inherently less

variability. Then, obviously, a part of this whole

process needs to be communication and training. If

people are out there saying that FDA wants us to

use apparatus 1 and 2, then that is what people are

going to do. So, the FDA is trained in a more

open-minded look at other things. If people feel

that way at least, then they might be willing to

look at other approaches.

When it comes to alternative approaches to

dissolution calibration validation, I think as I

told you in our lab we do more stringent mechanical

calibration because some products are very

sensitive to how the apparatus is set up and,

certainly, if you set it up properly your

variability will be less than the variability of

the calibrator tablet. Certainly, when you are

using your specific product itself, you need to ID

and control all the source of variability that you

are going to see. You need to determine how your

product is sensitive to things like the apparatus

type, the setup parameters and the media, both type

of media and whether it is degassed or not. There

is an interaction between the instrument you use

and your product, and understanding that is going

to also help you reduce the variability in the

dissolution test method. People like to use

calibrator tablets. I think it gives them a

measure of confidence that they set everything up

and their system is suitable.

So, what we are proposing is that

certainly the USP calibrator can be used if

somebody wants to take a look and see that they

have set up properly. Perhaps it also might be

useful to set up an internal calibrator maybe based

on a bio. batch or clinical batch to make sure of

system suitability. The calibrators dissolve in a

certain way or are sensitive to certain things and

not sensitive to certain things, the USP ones, and

those parameters may not be the parameters that

your particular product is or is not sensitive to.

So, creating your own internal calibrator and

understanding how your product is sensitive to all

the parameters is going to be perhaps better than

an outside product that may not have the same

sensitivities that yours does. Obviously, you need

to confirm the suitability of your internal

calibrator using some kind of a gauge R&R study so

you can really understand what the variability is

in your product.

Ajaz mentioned gauge R&R a little bit. If

you pick a lot of product or a piece of a lot to

maybe set up as an internal calibrator you need to

carefully characterize that and determine what its

variability is. You want to make sure it is

representative of your manufacturing process. You

want to make sure that it was manufactured while

your process was under control. Obviously, when

you are doing a gauge R&R you need to take a look

at what variability is introduced instrument to

instrument, vessel to vessel. As you can see, each

instrument is like 6 individual little instruments.

And variability from personnel to personnel and,

obviously, media and whether it is degassed or not.

We need to understand the benefits and

limitations of the different dissolution apparatus.

I showed you that there are 7 different ones in the

USP. We also sometimes get ones that are non-USP

apparatus when people submit test methods. So,

there are a lot of different things out there to

choose from and, better than just choosing one that

someone thinks maybe the FDA wants to see, maybe

try to understand how the hydrodynamics work; try

to model your system. Actually, I have been told

by people who do modeling that apparatus 1 and 2

are difficult to model so there may be some better

systems out there where we can do some better

predicting of what is going to happen as we change

physical parameters of our product, and take a look

at some other things we might be able to do.

Of course, what would even be better is

just quit doing dissolution as it is known today

and maybe find some other ways to assess product

quality. People have done some work in our sister

lab here, in White Oak, to try to correlate

dissolution with NIR. There is a lot of

spectroscopy out there that can be used online as

part of a PAT feedback loop, and perhaps good

correlations and good models could be developed

between those and quality and in vivo availability

and we can dispense perhaps with the current

dissolution test method, which has all of its

parameters--things that can go wrong and need to be

set very carefully. Obviously, key to this is

going back to the first principles and modeling and

understanding your formulation, and how each

component of your formulation contributes to the

quality of your product.

So, that is all I had to say and I just

wanted to acknowledge Terry Moore, who is actually

here today, who probably knows more about

dissolution than anybody in the world. He is

sitting over there, if you want to know more about

dissolution. Then, Zongming Gao is also in our lab

doing dissolution; and Lawrence who also knows a

lot about dissolution; and Ajaz all helped with

this. So, thank you.

DR. COONEY: Thank you very much, Cindy.

There certainly is time for questions. Gerry?

Questions by Committee Members

MR. MIGLIACCIO: Cindy, first I applaud

your last comment about using alternate methods. I

just want to point out that you made several

comments about the use of apparatus 1 and 2 and,

speaking I think for most companies, we don't use 1

and 2 because we think FDA wants us to use them.

You did a great job of pointing out the variability

of the different parameters that can impact

variability. It is very important when you are

testing thousands of batches a year that you have a

really well trained work force that knows how to

use this apparatus, and that you have consistency

in the way you test because if you are switching

from one apparatus to another it presents another

level of complexity. So, it is really the

consistency. Because of the variability that is

inherent here, it is the consistency that drives us

to apparatus 1 and 2 and not a lack of desire--

DR. BUHSE: To try something else?

MR. MIGLIACCIO: --but, you know, it is

complicated enough so it is really consistency that

drives us there.

DR. COONEY: Marvin?

DR. MEYER: The data you showed from your

lab versus the specs on prednisone, and you said in

one case you tend to be high and some cases fail,

when you do fail the calibration what do you do

about it? Is it the calibration that is no good?

Is it the USP specs that is no good? Is it the lab

that is no good? Or, do you just keep going until

you have 36 samples?

DR. BUHSE: Well, historically what we

have done is double check your mechanical

calibration and then you really run the calibrator

tablet. So, was the original failure the tablet?

Rarely do we find something to adjust when we check

the mechanical calibration. We do the mechanism

calibration much tighter than the USP anyway so

essentially you rerun. We actually don't run them

anymore in the lab, the USP calibrator tablets.

DR. MEYER: That solves that problem!

DR. BUHSE: That solves that problem! We

have an internal calibration tablet that we use now

that we have characterized ourselves in our lab

that has lower variability. We stopped using this

one probably at the end of last year. The data I

showed was the data from 2004, 2003.

DR. MEYER: The other question I have or

comment is that on one of the slides you suggest

using perhaps an internal calibrator, a bio. batch

or some known that you have produced.

DR. BUHSE: Right.

DR. MEYER: How do you know that that

product, over the lifetime of the product being

manufactured, hasn't changed? Dissolution doesn't

change, you are satisfied your equipment is in good

order when, in fact, it isn't because you couldn't

pick up the change--

DR. BUHSE: Stability is a big issue.

Stability is an issue with the current USP

calibrator. It is known to drift down I believe

with the paddle method over time, or whatever. Do,

you want to talk about that, Ajaz?

DR. HUSSAIN: Yes. Marvin, I am going to

go over that in detail. The gauge R&R is actually

for three purposes. It is to establish and

benchmark the variability. I think the proposal

actually is that mechanism calibration actually is

sufficient. The gauge R&R is an opportunity to

establish your target. You benchmark your

variability and then use that variability for

setting specifications, and so forth. But then you

have that and then you can keep the system stable.

I think stability of the system has to be based on

mechanism calibration. That is what other

countries do anyway. So, I will go over that in a

bit more detail. So, the opportunity is more than

just the internal calibrator. So.

DR. MEYER: One follow-up, I kind of joked

that you made the problem go away because you are

not using it anymore. What if you are a company

and had in your NDA or ANDA that you would

calibrate your dissolution using the prednisone and

USP and you started to fail, your dissolution

couldn't meet the calibration? They don't have the

luxury of just saying, well, we are going to use

our own now because they are stuck with using what

they said in the NDA, right? What should a company

do about that?

DR. BUHSE: You want to talk about that,

Ajaz?

DR. HUSSAIN: Well, I think this meeting

is step one to start addressing that in a sense.

Here is an alternate procedure. So, I think if the

advisory committee will sort of endorse this and we

move that way, we will put that in policy and there

are many different ways to implement that. So.

But from the compendia perspective, I think you

have to comply with the compendia so that is a

different challenge that the industry and companies

have to deal with. So, all we are doing right now

is creating an alternate regulatory decision

pathway and our enforcement strategy based on that.

DR. COONEY: Nozer?

DR. SINGPURWALLA: Slide number 13, I

thought you said it was Ajaz's slide. Therefore,

it is wrong!

[Laughter]

DR. BUHSE: Yes, it was Ajaz's slide.

DR. SINGPURWALLA: Well, how do you

distinguish between repeatability and

reproducibility?

DR. BUHSE: Well, I was going to say with

a destructive test it is very difficult.

DR. HUSSAIN: See, this is gauge R&R for a

destructive test. You really have to have design

experiment and I was going to cover that in my

talk. What this does is, it actually ensures that

the lot you choose is stable and in a state of

control. That is the only way you can actually

move in this direction. So, that achieves that

target. The destructive gauge R&R is a very formal

experiment and it is a nested design which does get

an estimate of whether a practice or an operator

can repeat it. That is repeatability.

Reproducibility is the variability associated with

that.

DR. SINGPURWALLA: So, the repeatability

refers to a physical thing. The other thing is I

don't know how important it is for you to manage

variability but if it is important to you to manage

variability, then my sense is that as the product

variability increases the measurement variability

will also increase. Therefore, there will be

correlation and, therefore, the sigma squared total

that you have will be underestimated the way you

have put it down. If it is of any importance, you

may want--

DR. HUSSAIN: I think it is. That is the

reason the leverage--the quality by design having

the pharmaceutical development information starts

to allow us to dilute some of this. But the

variability that you are observing you are

observing to the eyes of the measurement system so

the measurement system and variability in the

product are together. I will try to come back and

sort of explain some of that.

DR. DELUCA: I apologize for my voice.

You very nicely pointed out the multitude of

variables that are involved. There is instrument

variability as well as product variability so you

have interaction. You mentioned degassing. But

you are using a set agitation in your system. When

you start degassing, are you not sparging? Now,

you can create agitation or sparging during the

test?

DR. BUHSE: No, it is done beforehand.

You do it before you start and you put the media in

the different vessels and there is no degassing

during dissolution itself. Questions come up,

especially for extended-release products, where

actually the dissolution test method lasts for 24

hours per product, and the question then becomes

what happens to the gas level over that time. We

hope to test that with this meter. The one I

showed you here is actually one that has a probe

that is, like, 3 inches around so you have to put

it in a giant vessel. They are making a new probe

that is small and will fit inside the dissolution

vessel so we can see what happens actually in the

dissolution vessel over time. Like you say, with

some of these test methods at high rpm, 100 rpm, we

are getting a lot of agitation. So, that is a good

question.

DR. DELUCA: And I was worried about the

product and how product variation can affect--so,

you have an interaction between the instrument and

the product where particle size might influence,

you might have a set agitation rate. If the

particle size changes then it is going to change

the result.

DR. BUHSE: Right, unless you have a

method that can discriminate that if it is

important to the acting of the drug.

DR. DELUCA: You have talked about

modeling, I mean you mentioned it. Maybe it is

going to be covered later on, but I wondered if you

include anything here to look at profiles, release

profiles.

DR. BUHSE: We haven't done a lot of

modeling yet in our lab. I don't know if we are

going to talk about that specifically later on or

not today.

DR. COONEY: Ken?

DR. MORRIS: Just a couple of things. One

is that given the sort of lag--I guess I just have

a philosophical problem with calibrator tablets in

that if you are looking at a process and want to

independently establish that it is in control or

that it is doing what you think it is doing--we are

producing these the same way we produce the tablets

for testing--

DR. BUHSE: That are no better.

DR. MORRIS: What is that?

DR. BUHSE: That are no better.

DR. MORRIS: In fact, there are some data

that I think we will see to day that there are some

liabilities. I think maybe this is something we

will talk a lot more about, I am sure, but I think

one of the things that may come out of this is that

calibrator tablets just don't have a prominent

role. What I would say is that if you look at an

immediate-release system--and we will also get into

BCS exemptions--then the issues become sort of

treatable in other ways. If you are looking at

sustained-release or modified-release, such as

enteric or extended, then my argument is that you

ought to be controlling the coating process and

that sort of activity is really much more advanced

than it was. I mean, you have your example of the

tablet that has the air pocket but probably what

was more important was the difference between the

80 and 50 micron coating thickness. This is

clearly a failure of reproducibility of coating and

the dissolution may catch it or may not. I mean,

the statisticians--I don't know, there is the

Bayesian argument but I have talked to Sandy

Bolton, for one so, you know, if you have high

variability dissolution maybe 6 tablets is enough

to pick it up but, depending on what constitutes

high variability, you know, it is in the laps of

the gods whether you get it or not. So, to the

extent that things are surface-based alternate

methods--I mean, in the first place, you want to be

controlling the coating processes and then, to the

extent they are surface-based, have you considered

things within the group like the combination of

that and, like, IGC to look at surface free

energies or something that is at least a little

less subjective? I don't know if you have talked

about it because everything else is a correlated

technique.

DR. BUHSE: Right.

DR. DELUCA: Whereas, something that

actually measures surface free energy, even though

there is no practical instrument right now, is a

direct measure.

DR. BUHSE: We haven't done that with that

particular product. We have tried to do some

spectroscopy correlations.

DR. HUSSAIN: If I may?

DR. COONEY: Yes.

DR. HUSSAIN: I think you make a good

point, and I think the goal that we have, number

two, desired state, specification based on

mechanistic understanding--so, if the mechanism is

controlling the dissolution based on a coating

thickness, if you are able to measure the coating

thickness reliably, and so forth, that should be

sufficient. So, I think that is the direction we

wish to move in, and some of the new technologies

and science sort of helps that.

There is another point I think which I do

want to make and this is my graduate school

training; this is biopharmaceutics 101. When we

approach trying to develop a product we first think

about the patient, and so forth. Prof. Richard

always insisted you don't even think about an in

vitro test. You first try to get initial

information in humans and then say, all right, what

sort of testing will we need. So, you establish

your formulation, human connection or patient

connection first before spending time in an

artificial way. In my consulting role before I

came to FDA, one company I worked for carried out

53 experiments, screening and so forth; they had no

idea whether dissolution was useful or not. They

spent all this development effort trying to

optimize a hypothetical, what they thought was the

dissolution rate and the first experiment they did

was completely off. So, all the experiments were

actually off target.

So, there is a tendency within industry to

assume that in vitro dissolution is going to guide

them to a formulation without even understanding

its relevance. I think Marvin knows that company

very well. We actually had a paper on that issue

together. So, there are challenges I think. So,

quality by design actually forces us to think what

is the patient and then think about the tests so

that is what we are trying to achieve here.

DR. COONEY: Ajaz, perhaps we can capture

that as a point, that the purpose of formulation

development is to optimize patient care, not

dissolution assay. We hear you.

DR. FACKLER: Could I just make a point?

Dissolution can function for a number of different

purposes and on one of your slides you suggested

that finding a discriminating method might be

useful, and I would agree under certain

circumstances.

On the other hand, if you look at that

enteric-coated product really the purpose of the

enteric coating is to protect the tablet for the

first hour, or whatever time it might exist at the

very acidic condition of the stomach. Whether or

not coating breaks open at one hour, two hours or

three hours might have no relevance to the in vivo

performance of the product.

So, I think it is important, as we talk

about the future of dissolution testing, to

recognize what it is intended for. If it is

intended to predict in vivo performance, that is

one thing and a predictive or correlative method

then I think would be the ideal. If it is to look

for product quality and to reduce the inherent

variability in products, well, then a more

discriminating method that might have no relevance

to in vivo performance might be our goal. I think

we just need to keep that in perspective as we

think about the future of dissolution testing.

DR. HUSSAIN: If I may since we have time,

I think this is one of the first steps in our

tactical plan. Since we have time, if we could

engage the advisory committee to make sure is this

an acceptable step further discussion is needed.

So.

DR. COONEY: Tom?

DR. LAYLOFF: Yes, I was going to say

because of my concern with the problem with

degassing--I never degas my stomach before I take

my medication--

[Laughter]

DR. FACKLER: You probably don't swallow

900 ml of water either.

[Laughter]

DR. MORRIS: Just a couple of comments.

First, when we validate equipment we have to

understand what tests we are doing and what we are

trying to validate, and the standard tablet just

doesn't--intuitively, it doesn't get there for me

because we are looking at validating a piece of

equipment and all of a sudden the variables that we

are throwing into the pot include what is the

dissolution medium and how we handle that; what is

the size of tablet and how we handle that when we

are trying to validate a piece of equipment. So,

probably the first step is saying what validates

the equipment, and anything else we do is a waste

of time.

Then, the next step, to get right to what

Paul said, is what is my dissolution test telling

me because I am a manufacturer and I want to keep

my process under control, or am I predicting what

is happening in people? We have seen for 35 years,

as far as I know, that dissolution doesn't predict

the human results in terms of bioavailability or

bioequivalency. You can't do it that way. You

have to get that data and then try to correlate.

So, if we are using dissolution for quality

control, for process, fine, then there is a set of

variables and we do it that way. But if we are

trying to say that I can do a dissolution study

and, therefore, I will know that my formulation is

going to work in a person I think we are really

biting way more off than we can chew.

DR. HUSSAIN: I think I agree with you,

but in may aspects you do establish correlation.

Actually, Lawrence, in his talk, will actually make

that same proposal as you did. So.

DR. COONEY: Tom?

DR. LAYLOFF: The early work done on

digoxin was designed to go for in vivo/in vitro

correlation for about 35 manufacturers, and that is

how that standard was set. Prednisone subsequently

was done the same way. In reviewing that, it would

determine that if the FDA continued down that path

it would eventually take all the resources of the

FDA to do it because of the cost of performing

those in vivo/in vitro correlations. Then the

dissolution standard was just arbitrarily applied

across the board.

DR. COONEY: Marvin?

DR. MEYER: Ajaz, I think you ask if you

are on the right track and I think you definitely

are. You know, when you first said we are going to

revisit dissolution I said, oh, my God--

[Laughter]

--so, I think you are on the right track.

I mean, for me, when I used to do some dissolution

just in a university laboratory, I loved the wide

range for the calibrators because then my equipment

always passed and I didn't have to worry about it.

But now, sitting around this table, I have a

different hat on and it is shocking, 51-81 percent.

How can you have a calibrator--if somebody comes in

with analytical data like that you would say go

away; this is a very poorly controlled analytical

procedure. So, I think that revisiting the issue

is very important.

DR. HUSSAIN: Marv, in may ways, you know,

I was blind to this. I actually was not fully

aware of the scenario, and Cindy will attest to

this. When I started writing this paper I put

Lawrence through hell. I said how could this

happen? Because our standard criteria for

specification is plus/minus 10 percent and the

instrument is this way so there was a disconnect

that I was not aware of and I have to apologize for

that.

DR. COONEY: Are there any other comments

or questions at this point?

[No response]

What I would like to suggest is that we

take a break for 15 minutes and reconvene at 10:25,

and we are in good shape for continued discussion

and I have no doubt there will continue to be more.

[Brief recess]

DR. COONEY: I would like to now welcome

Dr. Mehta to speak to us about an overview of the

current guidance on the documents and decision

process in biopharmaceutics.

Overview of Guidance Documents and Decision

Process: Biopharmaceutics Section

DR. MEHTA: Good morning. As you can see

on my slide here, I am asked to give an overview of

guidances documents and decision processes from a

biopharmaceutics perspective.

Before I start, I want to acknowledge a

couple of people in my division, Dr. Ramana Uppoor,

she is sitting in the back in the audience, and Dr.

Patrick Marroum, team leaders in neuro. and

cardiorenal in my division and some of the experts

in biopharmaceutics in my division.

This is the outline of my presentation. I

am going to give you an overview of

biopharmaceutical aspects of dissolution-related

guidances. That is a formidable task. My first

draft that I sent to Ajaz had 100 slides and Ajaz

replied by saying an excellent overview but cut it

down. So, I am now down to 60.

[Laughter]

But I still intend to finish in time.

Then with a quick overview I will take you through

some examples from our NDA reviews of

immediate-release and modified-release products,

and share with you my perspective on opportunities

for improvement.

These are the guidances I am going to

quickly take you through. Chronologically they are

different but in terms of science, the way the

ideas are represented I have shifted them around.

I am going to first start with the BCS guidance.

In parentheses are the references. I will follow

that by the immediate-release dissolution guidance

that came out in 1997. The BCS guidance was

finalized in 2000. The IR dissolution guidance

invokes BCS principles and that is why I have

arranged it that way. That will be followed by a

quick overview of the IVIVC guidance and that is

for modified-release products, in vitro/in vivo

correlation. Then a couple of slides on general

bioavailability and bioequivalence guidance, which

was finalized in 2003.

I will quickly switch to something known

as scale-up and post-approval changes for

immediate-release products and modified-release

products, and the topics covered there.

So, let me start with the BCS guidance

summary. Maybe it is known to everybody, but just

for the sake of completeness let me point out the

highlights of the BCS guidance. This guidance

takes into account three major factors that govern

the rate and extent of drug absorption from the

immediate-release solid oral dosage form.

These are the solubility and intestinal

permeability of the drug substance, and dissolution

of the drug product. So, based on the solubility

and permeability characteristics of the drug

substance the drugs are classified into four

categories: high solubility, high permeability; low

solubility, high permeability; high solubility, low

permeability; and then the fourth category, low

solubility, low permeability.

The third bullet is the central idea, the

central concept, a very sound scientific concept of

BCS which is, you know, if a drug product is BCS

class 1, and for different formulations of this

class 1 product if they are rapid and similarly

dissolving you can give a biowaiver for the test

formulation without requiring an in vivo

bioequivalency assessment, provided you show

similar dissolution profiles over the physiological

pH range.

The last important point about this

guidance is that in this guidance we have defined

what determines rapid dissolution, and we say if

your drug product dissolves 85 percent in 30

minutes over the pH range absorption should not be

dissolution limited. So, that is all for BCS.

Moving on quickly to the immediate-release

dissolution guidance summary, and again I will do

my little bit of acknowledgement here, Dr. Shah and

some members on the panel here have contributed to

this guidance and, from my personal perspective,

this is scientifically a very well written document

although it was almost ten years ago.

These are the topics covered in this

guidance. The guidance lays out approaches for

setting dissolution specifications for a new

chemical entity. As I said, it takes into

consideration BCS nature of the drug product and,

depending upon that, you can have minimal

dissolution requirements in setting specifications

or more stringent.

Another very important point from my

perspective is that this guidance has outlined

something known as mapping or response surface

methodology. Again, this is supposed to be for

immediate-release products. The guidance says that

undefined clinical manufacturing

variables--manufacture your products at the

extremes of CMVs and in vivo performance and, if

you have that information, you will have a very

100

sound rationale for coming in with appropriate

dissolution specifications.

Finally, in this guidance there is a

discussion of how do you compare dissolution

profiles of two products. One of the approaches

that I recommend is known as the f2 or the

similarity factor which essentially looks at the

differences in dissolution at each time point, with

a range of 0-100. An f2 of 50 or greater than 50

indicates similarity of the dissolution profiles.

As we have said in that guidance, dissolution

specifications are established in consultation with

Biopharmaceutics and the CMC review staff. The

general bioavailability/bioequivalence guidance

summary, again limited only to dissolution

considerations, we have a section in there that

talks about what should be submitted in an NDA or

an ANDA in terms of a dissolution method. There

should be a dissolution method development report

for an NDA, new drug application. It should

contain a pH solubility profile of the drug

substance; dissolution profiles generated at

101

different agitation speeds; and dissolution

profiles generated on all strength in at least

three dissolution media. Essentially you want to

see the in vitro performance of your product over a

variety of conditions, including different media

and different agitation; and select the agitation

speed and medium that provides adequate

discriminating ability, taking into account all the

available in vitro and in vivo data.

For ANDAs, abbreviated new drug

applications, the guidance states that one should

start with an appropriate USP method if it is

there, in the USP. For some reason, if it is not

there for this product, then if the FDA method is

publicly available, utilize that. If that is not

available, also publicly available, then submit the

dissolution method development report, as described

above for a new drug application.

Again, for modified-release products for

ANDAs the dissolution profiles use the appropriate

USP method, if available, otherwise use the FDA

method for the reference listed drug if available.

102

In addition, and I think this is probably because

you could have for a generic similar or different

release mechanisms, so additional dissolution data

in three different media.

Now switching to the IVIVC guidance which

is, you know, in vivo/in vitro correlation for

modified-release products, again from my

perspective, this is a very useful guidance also.

The main purpose of this guidance was to provide an

outline for waiver of bioequivalency studies for

modified-release products if one was able to

establish an in vivo/in vitro correlation, a

quantitative correlation.

The guidance defines correlation in

different categories, A, B, C and D. Level A

correlation is most quantitative, and I have listed

in my presentation just the level A discussion.

Level A correlation is supposed to be a

point-to-point relationship between the in vitro

dissolution and the in vivo input rate of the drug

from the dosage form. Usually this is a two-stage

process, meaning that you take your dissolution

103

data, convert that into dissolution rate, and you

take your in vivo data and convert that into

absorption rate and correlate the two. Generally,

this relationship is linear but non-linear

relationship is also acceptable provided it is

adequately characterized.

So, this is an example of how level A

IVIVC would look. On the Y axis you have percent

of drug absorbed and on the X axis is the percent

of drug dissolved; your linear relationship over

the range and this establishes your correlation.

For the purpose of obtaining biowaivers, you need

validation of this level A correlation. From the

point of view of setting dissolution

specifications, that level of validation is not

necessary, and I will get into that subsequently in

my examples.

In the IVIVC guidance for modified-release

products we have some general concepts laid out for

what the dissolution specification should mean.

Ideally, as we say in the guidance, all lots within

the lower and upper limit of the specifications

104

should be bioequivalent. At the minimum, those

lots should be bioequivalent to the clinical trials

lots or an appropriate reference standard chosen by

the agency. In other words, you have your

reference performance and the upper limit should be

similar to the reference and the lower limit should

be similar to the reference. Ideally, the extremes

should be bioequivalent.

Some further considerations are that

variability alone should no longer be a primary

consideration in setting specifications for

modified-release products. Specifications wider

than 20 percent are acceptable only when evidence

is submitted that lots with mean dissolution

profiles that are allowed by the upper and lower

limits are bioequivalent. In other words, you can

have specifications wider than 20 percent if you

have a correlation, a quantitative correlation.

If you don't have an IVIVC and you want to

set dissolution specifications for modified-release

products, these are some of the characteristics of

what the data should be. The profile should have

105

at least three time points. The last time point

should be the time where 80 percent of the claimed

labeled amount is dissolved. Specifications are

set to pass at stage 2, meaning that there are 12

dosage units.

As I mentioned a while ago, for setting

dissolution specifications with the IVIVC, external

validation is not required and, as I already

mentioned, wider specifications based on what the

correlation predicts can be done.

This is graphically presenting that. On

the left panel you see that in the middle is the

performance of your product, the variability around

the mean dissolution profiles. The blue line is

the upper limit of the specification. The red line

or orange line is the lower limit of the

specification. You take that data using your in

vitro/in vivo correlation model. You predict the

plasma concentration based on the two limits.

On the right panel, the diamonds are the

actual blood levels, the predicted blood levels at

upper and lower limit, and the predicted level for

106

Cmax and AUC should not be greater than 20 percent.

Back in '97, what we could come up with was setting

the consideration based on the mean difference.

So, the upper and lower limit would not differ on

the mean AUC and Cmax by 20 percent. We could not

build into this consideration the variability

aspects and, as we have already heard in an earlier

presentation today, that is an opportunity for

improvement for future consideration.

Switching gears, I am going to quickly

tell you about what the SUPAC guidances mean as far

as immediate-release and modified-release products.

There are also a few guidances that came out

subsequent to the issuance of the SUPAC in 1997,

which is called equipment addendum, FDAMA and the

changes approved to an NDA or ANDA guidance in

2000. Again, I am going to try to capture this

very quickly.

Conceptually speaking, these guidances

identify what are the changes or what are the

variables that are covered in terms of

manufacturing considerations. The level of changes

107

for these variables, what are they? They are

defined; and then how do you deal with that?

So, the variables covered in this

guidance, manufacturing related, are composition

and components. For excipients it is

non-release-controlling as well as

release-controlling. The non-release-controlling

aspect is what is the part of the SUPAC-IR

guidance. That is taken as it is into the SUPAC-MR

guidance and then what is added is the

considerations for release-controlling excipients.

Other variables covered are site, batch size,

meaning scale-up and scale-down, manufacturing

equipment and manufacturing process.

I am going to take you through only one

set of variables here and show you how the levels

are defined and what are the related tests

recommended and what are the related filing

requirements.

Essentially, the idea is this, the

guidance has defined the level of change into three

categories, level 1 is the minor change; level 2 is

108

the moderate change; and level 3 is the major

change. So, moderate could have an in vivo impact

on level 3 or major changes likely to have an in

vivo effect.

Related to those changes, the tests go

along with them in terms of document evidence. The

lowest level, level 1, would usually require only

application of compendia tests and stability data.

Level 2 change would require extensive in vitro

dissolution and release data. That typically means

that for immediate-release products you require

profile comparison in five different media. Then,

for modified-release you need profile comparison in

three different media. Level 3 is the most

significant change and that will be allowed only if

you have an in vivo bioequivalency study or you had

established in vitro/in vivo correlation.

The filing requirements, again going from

minimal to most which is annual report, changes

being effected supplement, or prior approval

supplement. In the subsequent discussion I will

just focus on the first two bullets, which is level

109

of change and the tests. I am not going to touch

filing documentation at all.

Here is an example of how the guidances

break down changes into different levels. For

SUPAC-IR excipient levels excipients are listed for

level 1 change, level 2 and level 3. If you look

at glidant, for example, for talc, plus/minus one

percent change is allowed. If you look at the top

of the right-hand column, it is percent change

weight of the change of the excipient over the

weight of the total unit. For talc it is

plus/minus one percent. Other glidants would be

plus/minus 0.1 percent. So, that is the lower

limit of change, plus/minus 0.1 for talc. If you

look at filler, for example, it is also plus/minus

five percent change. So, this defines level 1

change, minimal change.

If you go to level 2 the ranges double.

So, you go from plus/minus 0.2 to plus/minus 10

percent. Anything beyond 10 percent is considered

a level 3 change. Again, this is for

non-release-controlling excipients.

If you go down to release-controlling

excipients for modified-release products, the

criteria are more stringent. Now, the change is

110

measured as a percentage of the total

release-controlling excipients and not the total

dosage form unit so your denominator is a smaller

number. The percentage allowed is smaller for

release-controlling excipients.

For level 1 change, that means that the

total additive effect of all release-controlling

excipients should not be more than plus/minus 5

percent. Level 2 should not be plus/minus 10

percent. Changes beyond plus/minus 10 percent are

considered level 3.

So, this is a summary of what we have

recommended in the SUPAC-IR and MR guidances.

These guidances define the tests; filing document

recommendations; level of changes in composition

and components, release-controlling and non-release

controlling excipients; site changes; batch size

changes; equipment and process changes.

The following changes either need a bio.

111

study or an established IVIVC: Level 3

release-controlling and level 3 non-release

controlling change; level 2 release-controlling

change for NTR drugs; and level 3 site change and

level 3 process change. All of those changes,

meaning level 2 changes, would require comparable

dissolution documentation, meaning, as I said,

profile comparison in several media.

As I mentioned in the title slide for

these guidances, the equipment addendum came out a

little later and there we identified equipment by

class and subclass for all major unit operations,

and a change to a different class is generally

considered a change in design and principle. So,

if you have equipment changes within the same

design and operating principle it is considered a

minor change. If you go to a different design and

principle it is a major change. Finally, the

changes guidance allows for multiple different

level changes. As we all know, these changes do

not occur only one at a time; it is a composite of

changes for any change. So, if you have, say,

112

several level 1 changes and one level 2 change for

your new product you would be held to the most

restrictive individual change of level 2, and

whatever requirements go with that level of change.

So, that was a quick overview of the

guidances. These documents are available on the

web, and if you have any questions please look them

up. Let me switch gears here and take you through

some examples of the way the specifications are

set.

But before that, let me share with you

generally what we see in an application in terms of

information available for setting specifications.

The data that are available for a typical

immediate-release product in an NDA are as follows:

Dissolution results under a variety of agitation

and media conditions. Then typically what we see

are several methods. One method is selected by the

sponsor which generally provides you with a rapid

dissolution profile. Using that method, we have

data of 6-12 units and that is the limit of data we

have for any given lot. So, that is the range of

113

variability that you would typically see for a

particular lot. Using that method, you have

dissolution data from the bio. batch, the batch on

which bioavailability has been characterized, plus

few to several production lots under this

condition. Again, as I said, these batches are

usually in very large quantities, hundreds of

thousands to million units. We see the data on

6-12 units.

Then we do have a lot of bioavailability

data on this product. Actually, bioavailability,

relative bioavailability, bioequivalency trials and

dissolution data of lots used in efficacy trials

and stability data. So, we look at all this

information and try and come up with a meaningful

specification.

What do we do when we consider setting

specifications? These are the factors that are

taken into consideration when setting specs. for an

immediate-release product. The in vivo behavior of

a drug product, particularly how rapidly the drug

is absorbed and an indicator for that is Tlag time

114

or what is the Tmax of your product. Since the

issuance of the BCS guidance we look at the

permeability data very closely. In vivo

permeability would be based on mass balance studies

as well as absolute bioavailability studies and

that, in my mind, is the gold standard by which you

define whether a drug is highly permeable. If it

is quantitatively absorbed, then you say this high

permeability, along with your high solubility data,

puts the product into BCS class 1. Then that

carries its own benefits. I have an example of

that to show you a little later.

That is what one pays attention to, in

vivo behavior of the drug product from a

bioavailability point of view. We look at

dissolution behavior across all conditions in vitro

and then we try to come up with an adequately

discriminating method, taking all this data into

consideration based on any quantitative or

qualitative in vitro inference.

What is very helpful for evaluation of an

NDA is if you have data like this where a solid

115

dosage form in vivo is compared to something that

is even more rapidly dissolving, meaning your solid

dosage form's performance in vivo with respect to,

like, a solution. If we have this data, this tells

us a lot about what is the in vivo dissolution of

your solid dosage form and that can help us

evaluate the in vitro considerations for setting

specifications for that product. So, this can

guide how discriminating the in vitro method needs

to be.

As I said, we look at all the available

dissolution data and pay particular attention to

the lots that have in vivo data, and then discuss

with our chemist colleagues about what is available

in the stability domain, the data there and the

specifications we are considering. If we see a

significant change or time with stability

performance, that will have to be resolved by a

bioequivalency study.

Possible outcomes in terms of setting

specifications, one is everybody is happy.

Sufficient data are submitted and specs are

116

finalized. It is possible that insufficient data

are submitted. Based on the product's indication,

the product needs to be approved with reset interim

specs. We agree with the sponsor what additional

data needs to be generated. We agree upon a

time-line. We evaluate the specs and we finalize

the specs. In the rare instance where there is

insufficient data submitted--I have not seen this

happen in my lifetime where we have withheld

approval for a drug product because of insufficient

dissolution data. At the least, we will set specs

on the clinical trial product. So, if insufficient

data are submitted and specs can't be finalized

even including interim specs, then we have to

resolve that prior to approval.

Now let me take you through some specific

examples, starting with simple to a little bit more

complex. This is an immediate-release drug product

A. The drug is highly soluble over the pH range of

1.2-6.8, or 6.9 in this case. Based on the

bioavailability and the in vitro permeability, we

established that the drug is highly permeable. So,

117

we have high solubility, high permeability criteria

met. The drug product is rapidly dissolving over

the pH range of 1.2-6.8. So, we have seen this.

We are sure of these characteristics and we say

okay, this is BCS class 1.

We have dissolution results of the

bioavailability lot and the clinical lot so all

that data is utilized in setting the

specifications. There was stability data also

available that was taken into consideration. It

turned out to be a straightforward case. The

sponsor's proposal was that they use a USP 1

apparatus at 100 rpm in 900 ml 0.1 normal

hydrochloric acid; specs of 80 percent in 30

minutes. We agreed with the sponsor.

Just as a note, Ajaz and I didn't exchange

notes beforehand but in this case the sponsor chose

apparatus 1 to avoid coning effect. Ajaz had an

example from the Canadian database where that was

the reason why you saw a big investigator

difference compared to the reference, but the in

vivo data turned out to be fine.

Another example for an immediate-release

drug product, product B, the drug is a free base

with 2 pKs of 5.4 and 7.2. It is highly soluble at

118

pH 1 but it is practically insoluble at pH 7, and

the solubility drops sharply between pH 4-5. I

have a graph that shows that clearly. The drug is

absorbed slowly, at Tmax ranging from 3-5 hours.

The half-life is long, 45 hours. It is not highly

permeable. The fraction absorbed is around 0.75.

So, what do we do with this? This is the

dissolution behavior across the pH ranges. As you

see, below pH 5, which is the third curve from the

top, dissolution starts dropping rapidly as the pH

increases. The sponsor chose the dissolution

method at pH 5, and showed that the clinical and

to-be-marketed formulations had similar profiles.

This is what that comparison is at pH 5.

We had bioequivalency data on these two

formulations and that turned out to be clearly

bioinequivalent in vivo for the test, meaning that

to-be-marketed product showed a clear difference in

Cmax. The Cmax was 17 percent lower. We

119

interacted with the sponsor and they optimized

their method to come up with an adequate

discrimination condition to evaluate this

formulation further.

This is what they came up with, 5 percent

volume Tween 80 and the same two formulations that

were clearly bioinequivalent in vivo, they were

able to identify their in vitro performance and

show that, indeed, they were different. This was

verified further by taking the two formulations

that were bioinequivalent in vivo and the method

showed that they were similar in vitro.

This was the availability of dissolution

data across several batches. All I want to point

out to you is that, as I said, dissolution data for

different batches, from 6 units, mean and range is

available and if we look at the right-hand column,

the lowest range is 86 percent.

Taking all that data into consideration,

the sponsor proposed the specification with

apparatus 2 at 50 rpm and 1000 ml to Tween 80 in

water; Q of 75 percent in 45 minutes. We

120

recommended no changes in the condition but a Q of

80 percent in 45 minutes. Here is an example of

availability of in vivo data optimizing the

specifications.

The final example I have is for a drug

product, a modified-release drug product with in

vivo/in vivo correlation. For this drug product a

level A correlation was established. Correlation

was obtained from in vivo data from 6 different

studies, and the media consisted of pH 1.5 for the

first 1.5 hours and then pH 6.8 for the remainder

of the 24 hours. This is a once a day product.

These are the results. I think this was

excellent work on the sponsor's part. We worked

with them and we were very happy to figure out the

specs with them. Look at the hatched region. That

is the observed range of dissolution data. That is

the extent of variability across the entire

manufacturing experience for this sponsor. So, the

hatched area is the dissolution variability,

dissolution range the product showed in vitro. The

specs we agreed upon are the two dotted lines above

121

that hatched region. So, those were the

specifications proposed and we agreed with them.

The best part is that if you look at the

third level of curves, which are the topmost dotted

lines, the topmost and the bottom, those are the

predicted in vitro dissolution behaviors of two

formulations that would be comparable in vivo. So,

the specifications were set within the limits of

what products would be bioequivalent, so a good

IVIVC that could lead to meaningful specifications.

Now let me conclude with some personal

comments on opportunities for improvement. Before

I get into my own suggestions, I want to cite this

article that Ajaz already mentioned from Dr. Janet

Woodcock, a clinician who has written beautifully

on pharmaceutical quality. I am just going to cite

two quotations out of this article. I mean, I can

stand here and tell you a great deal about all the

complexities involved in clinical trials but I

think Dr. Woodcock has summarized this very well in

this first bullet, which is, as she says, for the

purposes of clinical use, the established drug

122

quality attributes are generally adequate because

they achieve much tighter control of the level of

variability than could be detected in patients

without extensive study.

These are part of all the variabilities,

specially manufacturing variability. It can be

done but it is a difficult task and it would be

very extensive, and that is not the paradigm

currently used.

But maybe even more important, as she

points out here in the very second line of the

previous quotation, in contrast, for regulatory and

manufacturing processes, the lack of detailed

understanding of the real-world importance of

quality attributes is a serious problem, leading to

many disputes that might be resolved easily were

relevant information available on the relationships

between various quality parameters and clinical

performance. I personally couldn't agree more with

that concluding comment.

So, clinical performance, if I were to

dissect that further--everybody talks about

123

variability and this is my share of what are the

different types of variability in therapy. You

start with manufacturing variability, then you have

variability associated with the drug exposure and

then you have variability associated with the drug

response. You have compliance issues. You know, a

lot of people can actually add more bullets to this

and provide a complete picture of how complex the

system is when a patient is being treated in vivo.

But I have taken a shot at just making a

point on exposure-related variability and

manufacturing aspects associated with that. The

next table is a snapshot. We have an internal BCS

database of almost 200 NDAs. That is in the

process of being audited and we hope to publish

that soon. So, what I requested Dr. Uppoor to do

is to randomly select a few drugs and prepare a

table that would show variability in AUC and Cmax

and the exposure parameters of different BCS

products.

Again, this is tentative because this is

not fully audited so that is why I have starts in

124

this table. This is BCS class 1, 2, 3 and 4 across

the top horizontal line. You have the permeability

associated with the AUC parameter and the Cmax

parameter for these products. As you can see,

staring with class 1, we have variability in the

range of 17 to about 24 percent. Class 3 shows

maximum in vivo variability.

So, if I want to take this tentative class

information further, the point I want to make--the

numbers might be off when we have the actual

publication coming out, but this is the point I

want to make, that if I assume that the clinical

trial formulation for this product was

optimized--if it is not optimized, I think it is in

the interest of the sponsor to optimize that so

that even a little bit of manufacturing variability

does not reflect in the in vivo performance at

least from a drug exposure point of view. But

assuming that this formulation is optimized, even

for BCS class 1 products you do see a decent amount

of variability in vivo. Again, this is reflecting

how the drug is handled by an individual and the

125

variability of handling that across individuals.

This information can be utilized by a sponsor to

come up with rational specs.

These are some of my thoughts in terms of

opportunities for improvement. The first point is

nothing earth-shattering but I still think it is a

point that has to be made, to select an appropriate

dissolution method based on physicochemical in

vitro and in vivo characteristics of the drug and

the drug product.

It would be useful to have an estimate of

in vitro variability for low solubility and low

permeability. Estimate of variability of lots used

in pivotal efficacy trials would facilitate setting

of rational specifications. For modified-release

products estimate the in vitro release

variability--the example I showed where if you had

a handle on the variability across your entire

manufacturing process, then you can bring that into

setting a meaningful specification. As I already

mentioned, right now the IVIVC current guideline is

based on the limit mean estimates only and if you

126

can build in the variability aspect and in vivo

performance based on estimate of mean as well as

variability, I think that would lead to more

rational specs, maybe even wider specs compared to

what we are doing now.

The things that I see in the near future

are new technologies like PAT. Hopefully, it can

provide in vitro and in vivo relationships based on

the performance of an individual dosage form unit.

I mean, this would be a non-destructive method.

You would be able to assess the dissolution

performance of a unit without breaking it up and

then you would administer that to an individual and

you would get that individual's exposure parameters

so you would have correlation relationship on an

individual dosage unit form in an individual

patient taking it. I think that would be a very

powerful set of data to set meaningful

specifications.

We are getting more and more complex

products like drug eluting stents and liposomes.

For these complex dosage forms I think it would be

127

essential to study drug elution, drug release using

mechanistic models and new techniques in imaging

and fluid dynamics. Hopefully, future

specifications will be based on in vitro mean and

variability estimates.

Moving from a science point to a process

point--I didn't know our good friend Dr. Chuck

Hoiber [ph.] would be here but this is in those

days when Chuck and I were on the same floor and we

started implementing this which is that from the

process point of view there are also a lot of

opportunities to optimize setting of specifications

and that, from my perspective, is come and meet

with us early. A meeting would be useful if you

have good quantity and quality of data. As we have

done on several occasions, we have had separate end

of Phase II meetings with CMC Biopharmaceutics and

colleagues on our side and the industry, going over

the development plan and that has led to a quicker

review and arriving at meaningful specifications at

the time of NDA approval.

Finally, I do personally believe that good

128

homework will always bring dividends. If you have

good data, please share them with us and we will

work with you to come out with rational

specifications. Thank you.

Questions by Committee Members

DR. COONEY: Thank you very much. Some

questions from the committee? Ken?

DR. MORRIS: Two things. I was a little

surprised to see the high variability with BCS 3.

In principle, you would expect BCS 3 to be a good

candidate for waiver because, as long as your

driving force doesn't change, you would expect that

the absorption is rate limiting and falls into the

same basic concept as 1.

DR. MEHTA: That is a very good

observation. We are looking at the data carefully

ourselves, but I think it is maybe one product that

is--

DR. MORRIS: Driving the variability?

DR. MEHTA: Yes.

DR. MORRIS: Or is it that the absorption

itself is just variable?

DR. MEHTA: Again, we can think about it

but it is a question if you have a class 3 high

solubility, low permeability drug and if low

129

permeability is not leading to the same conditions

in vivo that is going to take away some of your

high solubility benefit.

DR. MORRIS: Not the same conditions on

which side? Are you talking about in the gut?

DR. MEHTA: Yes.

DR. HUSSAIN: Sorry, if I may, I think one

of the challenges is that this was always a

question when we were deliberating the BCS

guideline, high solubility. But the in vivo

dissolution actually is more sensitive for low

permeability drugs and we actually have published

on this with Lawrence--

DR. MORRIS: Right.

DR. HUSSAIN: So, people often say this is

high solubility so dissolution is not rate limiting

but in vivo dissolution behavior is quite complex.

Plus, you add site-specific absorption of these

compounds that adds to all the sources of

130

variability.

DR. MORRIS: Right. I guess that is my

point in a sense. Shouldn't the compounds be

segregated into site-specific and passive absorbed

compounds to really do a valid experiment?

DR. HUSSAIN: I fully agree with you. We

came up with the classification system and those

four classes are beautiful but there is nothing

that black and white. Greater than 90 percent

permeability, highly permeable, but there is a

gradation of that and, you know, we have to take

that into account. You know, there are, like,

windows of absorption. So, we need to subclassify

those four classes and then come up, you know, with

better--

DR. MORRIS: Yes, but it would be nice if

you could identify some more waiver-worthy classes.

DR. HUSSAIN: Yes.

DR. MORRIS: Just another quick comment is

that I am sure it won't surprise you but, you know,

with the general BA/BE guidance people, because of

what is in the guidance, are actually doing pH

131

solubility profiles of non-ionizable compounds.

[Laughter]

DR. MEHTA: That is taking us too

seriously!

DR. COONEY: Paul?

DR. FACKLER: I have a comment and a

question. The comment had to do with the slide

where you suggested there might be about 20 percent

variability for even BCS class 1 compounds. I

would suggest that that is vastly understated, that

the variability is much higher than that because I

am guessing that your data comes from

bioequivalence studies where all of the subjects

take exactly the same amount of water, the same

amount of food. None of them are BMI greater than

a particular number. If they are old studies they

were all men. I would just say that in the general

population with the way pharmaceuticals are really

taken--some people run three miles, come home and

then swallow their tablets; some people roll out of

bed and swallow them without water--the variability

even for class 1 is significantly higher than 20

132

percent. But it is just my opinion.

DR. MEHTA: That would just add to the

thought I had which is, you know, use that

information to evaluate your in vitro

specifications. That will help you.

DR. FACKLER: The question I had had to do

with that same chart where you looked at 17 drugs

that were randomly pulled out of the pool of 200.

It was interesting to see that the class 3 is N

equals 7. I am just wondering if the distribution

of these 17 in any way represents the distribution

of the 200 drugs.

DR. MEHTA: I don't think so. The whole

idea was to see if we can get a handle on what is

the exposure variability for these products. A few

years ago I presented this database at one of the

APS workshops what was surprising is that we saw a

lot of NDAs falling into class 4 category. If it

is a class 4, then you would see very few drugs,

low solubility, low permeability. You know, they

would fall out of drug development. But, as I

mentioned a little while ago, the way classifying

133

we have created these four classes, 90 percent of

data goes in class 1 over this 85 percent

absorbed--you know, it is still low permeability.

So, I don't think when we come out with this

information, all audited, that there is going to be

a majority of them falling in class 3. I don't

think so.

DR. COONEY: When you presented the table

of the 17 samples, your intent is to expand that?

This is just a piece of work in progress?

DR. MEHTA: Yes, very much so.

DR. COONEY: So, the idea is to really

address the question that was just asked, that is,

to have an analysis that is representative of that

whole set?

DR. MEHTA: Yes. I mean, right now we are

going through each drug and making sure, to our

level best effort, that the data available

classifies that drug product in the appropriate

class. We have the information put together and

now it is like careful auditing going on.

DR. COONEY: Good. Marvin?

DR. MEYER: I did come up with a couple of

questions. It always bothered me that the BCS

system had this quadrant drawn and then the lines

134

kind of floated depending on how you wanted to

define high and low--

DR. MEHTA: No, it is rigid right now.

DR. MEYER: I know it is rigid but the

rigidness was arbitrary.

[Laughter]

It is arbitrarily rigid.

DR. HUSSAIN: I will defend it tomorrow;

don't worry!

DR. MEYER: Okay.

DR. MEHTA: We started out with a

conservative position and now with the availability

of more data we want to expand that rationally with

proper evidence.

DR. MEYER: It also bothered me that this

permeability goes all the way from a very rigorous

intubation of humans to a K2 cell to looking at

Tmax. So, how it is defined or determined can be

another source of variability in where it falls in

135

this rigorously arbitrary quadrant. So, I think

that may be a reason in part why the class 3 seemed

to be more variable than 2. One drug in that would

have expanded the range.

DR. MEHTA: That is just the way those

drugs got pulled out. That is why I have that

range. That may not be reflective of what it is.

I don't want to take up too much time, but we look

at permeability assessment now very carefully and,

in my mind, hopefully, if we have data on the NDA

side, which is mass balance data and

bioavailability data, that is the maximum way in

terms of assessing, you know, whether the drug is

90 percent absorbed or not. Sometimes we have an

issue with that. Then we utilize the in vitro

methods for that decision.

DR. MEYER: One last question. Do you

feel that the f2 test has been rigorously

evaluated?

DR. MEHTA: A good question, Marv.

[Laughter]

There are people in the audience that--

DR. MEYER: Do you feel--do you feel it

has been rigorously evaluated so it will detect

differences when they should be detected and will

136

allow passage when it should be allowed?

DR. MEHTA: Well, I mean we do state in

our guidances under what conditions this approach

should be employed. You know, if your variability

is very high in dissolution on each formulation

this is not the right way of comparing those

profiles so then you need to get into more complex

assessment, and all that. If it is done properly,

yes, I do myself.

DR. COONEY: Pat?

DR. DELUCA: In the BA/BE guidance summary

for modified-release products you are saying that

they should profile using at least three other

dissolution media and water. Why do you need three

others if you have a correlation?

DR. MEHTA: No, it doesn't say that there

is a correlation. This is just a question--well,

usually correlation is release formulation

specific.

DR. HUSSAIN: It is just for that product.

So.

DR. MEHTA: It is right now.

DR. COONEY: Nozer?

DR. SINGPURWALLA: When you don't

understand something you start asking technical

137

questions.

[Laughter]

You showed a picture of linear correlation

long ago, one of your early slides--

DR. MEHTA: Yes, level A correlation.

DR. SINGPURWALLA: Level A correlation. I

have two comments. The first is that you are

looking for relationships between the percent of

drug dissolved and the percent of drug absorbed so

correlation only measures linear relationships.

You may have dependence which may be not linear but

still of value to you, but correlation does not

measure that. So, I just want to say that as a

comment.

The second more serious comment is that

that particular correlation misses the time index.

138

What you really need is a third axis also showing

the time at which all these happen. For you to do

that, you want to look at these two as what we

would call stochastic processes or time series, and

you want to cross-correlate the two time series.

So, if you want to improvise on that particular

theme, you may want to look not at correlation but

what I would consider cross-correlation where you

also introduce the time axis. That is the only

comment I want to make.

DR. MEHTA: Thank you. That is helpful.

DR. SINGPURWALLA: Do you want to

challenge me now?

DR. MEHTA: No, I didn't say that.

DR. COONEY: Are there any other questions

at this point?

[No response]

Thank you. The next presentation will be

by Dr. Shah establishing dissolution

specifications.

Establishing Dissolution Specifications:

Current Practice

DR. SHAH: Good morning. Mehul gave a

nice overview on the BCS guidance and other

guidances which are used in setting dissolution

139

specification from a biopharmaceutics perspective.

My job today is to cover the CMC aspects of setting

the dissolution specifications. In this

presentation I am going to start with an overview

of the current practice, and in that overview I am

going to cover the CMC assessment and bring in some

of the ICH Q6A principles, how we evaluate the ICH

Q6A principles in our CMC assessment, and then I

would like to talk about a case study example for

extended-release oral suspension and in that

example I am going to cover the drug development

strategy by the applicant, the dissolution results

obtained based on that development strategy, then

what we identified as critical issues, followed by

our recommendations and based on those

recommendations, what were the improvements

implemented by the applicant and what was the

outcome out of those implementations. I would like

to end my talk with some concluding remarks based

140

on this example as well as general remarks in that

aspect.

As Mehul suggested in his presentation, I

want to reemphasize that establishing dissolution

specification is a shared responsibility between

the Office of New Drug Chemistry and the Office of

Clinical Pharmacology and Biopharmaceutics.

In the next three slides I have presented

the considerations that should be given during that

development, as well as the focus of CMC assessment

during the NDA review, and what forms the basis of

setting the dissolution specifications from CMC

perspective.

As I have pointed out here, it is a known

fact that physicochemical properties of the

formulation components, such as drug substance and

other excipients, such as the solubility, pKa,

particle size distribution, polymorphic forms and

there may be some others, have a significant effect

on the dissolution. The physicochemical properties

impact the processibility of the formulation

components, as well they may affect also the

141

safety, efficacy and stability of the drug product.

In addition to that, the manufacturing processes,

especially those having the potential to influence

the release profile of the drug substance also

should be studied during the development. And, the

control strategy of the critical process parameters

and in-process testing also should be developed

during the development, and those are the focuses

of the CMC assessment.

During the drug development one should

expect that there should be a relationship of

in-process testing to the critical quality

attributes, such as dissolution of the drug

product. Some of the in-process testing that may

be carried out might be particle size distribution;

release rate; and the compression force, tablet

hardness and friability in the case of solid oral

dosage form.

In addition, during the CMC assessment we

focus also on the development and validation

aspects of the proposed in vitro dissolution

method. Cindy already covered some of these

142

aspects in terms of how the methodologies are being

developed and what are the validation criteria that

need to be covered, especially pertaining to

specificity, linearity, accuracy, precision,

ruggedness, etc. In addition, we also focus on the

release time point intervals and what should be the

adequate tim point intervals.

Once we have this information we need to

see or need to provide during development, as well

as the NDA submission data, what is the

relationship between the in vitro dissolution data

from development, clinical, bio. and primary

stability batches, and also identify a discerning

trend on storage. We also evaluate the proposed

shelf-life of the drug product on the basis of the

stability data analysis of dissolution, as well as

other drug product attributes.

In the end, it is in coordination with

Office of Clinical Pharmacology and Pharmaceutics

that appropriate dissolution specifications are

recommended and these specifications are reflective

of the dissolution data from various batches

143

including clinical, bio., stability and other

batches.

In terms of the ICH Q6A document, ICH Q6A

discusses the potential relevance of particle size,

polymorphic content and polymorphic changes, and

how it affects the dissolution.

Here I have these three decision trees

just for reference. I just wanted to point out

that CMC assessment very well integrates these

principles in our assessment for the quality

assessment of the drug product. This is about the

particle size distribution and the decision tree

guides you on how to set acceptance criteria.

This is in terms of polymorphic content.

That also guides you on how to set acceptance

criteria. The next one is how to set the

polymorphic change acceptance criteria in the drug

product.

Now I would like to focus on the case

study example for extended-release oral suspension

for the remainder of my talk.

Let me give you just some background.

144

This was submitted as a 505(b)(2) application. As

a result, there was no clinical trial required

because the safety and efficacy of the proposed

active ingredients for the proposed indication was

established through immediate-release products

available under the tentative OTC monograph for the

same indication. The proposed dose was a single

dose given every 12 hours to patients 6 years of

age or older. That was equivalent to the nominal

OTC monograph which was given every 6 hours twice.

In terms of the formula, the drug product

contained two different active ingredients, and I

will call them drug substance 1 and drug substance

2. For proprietary reasons, most of the data I am

going to discuss here are well concealed and they

are masked but the data are real. Drug substance 1

is anchored to a drug carrier support and coated

separately with semipermeable polymer to prevent

dose dumping and to impart the extended-release

profile. Drug substance 2 binds the drug carrier

support in situ during the manufacturing process,

but it is not coated. Both active ingredients,

145

along with other excipients, are suspended in

aqueous solution.

The concerns we had here arise from the

safety implications due to the potential dose

dumping, and efficacy implications due to

insufficient rate and the extent of release of the

actives. These concerns were brought to the

applicant's attention during the end of Phase II

meeting as well as pre-NDA meetings, and they were

very mindful of those two concerns.

This was the strategy adopted by the

applicant in the beginning. They wanted to

demonstrate bioavailability of the drug product

formulas, and that was coated with 6 percent

coating of drug substance 1, to a reference drug

which as an immediate-release solution, and it was

containing the same two active ingredients. They

had no other choice but to start with the

immediate-release solution because there was no

existing extended-release product containing these

two ingredients.

Their plan was to formulate three

146

experimental drug formulations, each differing only

by the coating level of semipermeable polymer on

drug substance 1. They were low coating, for

example, 2 percent; medium coating, example, 5.5

percent; and high with 9 percent coating on drug

substance 1. They labeled them as fast-release

solution, intermediate-release formulation and

slow-release formulation. The approach was to

establish IVIVC for each active among these three

experimental formulations, and establish

dissolution specifications for both actives based

on generated dissolution profiles from the slow-

and fast-release drug product formulations.

In the NDA the data submitted include five

formulations of the drug product containing drug

substance 1 coated with varying levels of

semipermeable polymer, 2 percent, 5.5 percent 9

percent, as well as 6 percent and 10 percent. They

performed the following PK studies, multi dose

bioavailability studies with immediate-release

solution and single dose food effect study

containing 6 percent polymer coating, and single

147

dose IVIVC study containing three formulations, 2

percent, 5.5 percent and 9 percent polymer coating.

In support, there were PK results from four batches

and stability results from four PK and five

stability batches.

Based on these PK studies, these were the

applicant's claims, that level A IVIVC was

established for both actives of the ER suspension.

The mean individual level A IVIVC models for drug

substance 2 met the FDA validation criteria and, in

their opinion, it can be used for setting

dissolution specifications and biowaivers.

The mean and individual level A IVIVC

models for drug substance 1, which is coated,

failed the FDA validation criteria in that the

predicted values had a larger error than

recommended. However, if the dissolution criteria

remain within dissolution profiles tested in IVIVC,

they proposed that the drug substance 1 results can

serve as a mapping study for the formulations.

Now let's see what was the agency's

finding in terms of the PK results. On the

148

bioavailability and food effect studies, which was

the 6 percent coating of drug substance 1, the

agency found that systemic exposures of both

actives were favorable between the extended-release

suspension and multi dose of reference

immediate-release solution, and there was no food

effect on both actives.

However, in terms of the IVIVC study,

where the drug substance was coated with the 2

percent formulation, 5.5 percent formulation and 9

percent formulation, with respect to drug substance

1, the agency found that it failed to establish the

in vivo/in vitro correlation, and observed more

than 20 percent of difference in Cmax for

formulation of fast and slow dissolution profiles.

With respect to drug substance 2 that was

not coated, level A IVIVC was established, however

it failed to validate the IVIVC. The formulations

used in the IVIVC study were found to be

bioinequivalent, that is to say the Cmax of the

formulations used in the IVIVC study were different

by more than 20 percent. The proposed dissolution

149

specification and the approach to set a dissolution

specification based on IVIVC by mapping was found

unacceptable.

Now let me share the stability results

analysis. This is what we review in our CMC

assessment. What we found was contradictory

release profiles observed between drug product

formulations containing 6 percent and 9 percent

coated drug substance. Drug substance 2 showed

more decrease in dissolution than drug substance 1,

and we observed substantial decrease in dissolution

at 1-hour, 3-hour and 6-hour time points for both

actives from the corresponding initial values among

all batches, including bio. and primary stability

batches, at all storage conditions. The decrease

in dissolution was most notable at 3-hour and

6-hour time points. The decrease in dissolution is

minimum at the 12-hour time point and the decrease

in dissolution for both actives levels off by 9

months on storage.

This is displayed on this slide. This is

the dissolution results of drug substance 1. For

150

clarity purpose, I have labeled the coating for the

dissolution curves. The yellow bar shows the 6

percent coating that was used in the

bioavailability study. The purple is the 2

percent. The middle one is blue, which is 5.5

percent coating of drug substance 1. The red one

is the 9 percent coating of drug substance 1.

Now, what I explained in the previous

slide is what you can see is a decrease in

dissolution profiled for all the solutions. You

would expect the 9 percent would be showing a slow

dissolution compared to the 6 percent but it is

quite the other way.

If you look at drug substance 2, the

decrease is more compared to drug substance 1,

which is shown basically from the least point and

at the 18 months time point. That is more than

about 20 percent decrease in dissolution over time.

So, based on this analysis these were the

critical issues discussed with the applicant, and

they concerned the raw material controls,

manufacturing processing and in-process controls

151

and controls related to particle size distribution

and dissolution method.

I just want to point out over here that

these discrepancies in the results showed that the

coating process was not in control and we discussed

that issue with the applicant. They decided to

reformulate the drug product and decided to abandon

the idea of the IVIVC approach to set dissolution

acceptance criteria; conduct PK studies on

commercial scale bio. batch containing drug

substance 1 at the specified target coating level,

rather than a range, and compare it to the

reference IR solution; manufacture additional 3

pilot scale primary stability batches of the drug

product containing drug substance 1 at the same

specified target coating level; and propose

dissolution acceptance criteria based on in vitro

dissolution profiles obtained for both actives from

the bio. batch.

These were the process improvements

implemented. They coated the drug substance with a

specified target coating level of semipermeable

152

polymer; revised the coating and subsequent

manufacturing processes; instituted appropriate

process controls to stabilize binding of both

actives to the drug carrier support in the

suspension; and manufactured one commercial scale

bio. batch and three pilot scale stability batches.

They instituted appropriate particle size

measurement method, for example laser diffraction,

for drug carrier support and coated drug-bound

carrier particles. They revised particle size

distribution acceptance criteria for the drug

carrier support, coated drug substance bound

carrier support particles and suspension

stabilizing excipients.

Based on these results, they conducted

three PK studies utilizing the drug product

formulation with coating of drug substance 1. They

conducted BA/BE assessment; PK at steady state; and

food effect studies. The results showed that the

PK profiles of drug substance 1 and drug substance

2 from test extended-release suspension were found

comparable to the reference IR solution following

153

single and multiple dose administration, and food

had no effect on bioavailability of both actives.

Now let me share with you the stability

results analysis. After the implementation of the

improvements in manufacturing process for coating

and instituting adequate process controls in terms

of particle size, we observed stable and consistent

release profiles at 1-hour, 3-hour, 6-hour and

12-hour time points for both drug substance 1 and

drug substance 2 on storage within each of the bio.

and three primary stability batches. There was no

discernible trend in release profiles of drug

substance 1 and drug substance 2 and on bio. and

primary stability batches at all storage

conditions. And, there were comparable release

profiles for both drug substance 1 and drug

substance 2 among bio. and three primary stability

batches.

This is displayed in this graph for drug

substance 1. You can see, as opposed to the

dissolution rates that we saw before and after

implementation of manufacturing processes. This is

154

with respect to drug substance 1, which was coated.

This is the bio. batch and these are the three

primary stability batches. Most of the

dissolution, as you can see, ranges between 5-7

percent.

This is with respect to drug substance 1

dissolution profile. This is the bio. batch and

you can see these are the three primary stability

batches and you do not see any discernible trend

and most of the dissolution ranges between 5

percent if you compare it to drug substance 2 prior

to the implement.

Then I would like to conclude my

presentation with the following remarks. We were

able to identify probable causes of discrepant and

inconsistent dissolution results for drug substance

1 and drug substance 2, and recommend corrective

measures to address the issues. The outcome was

consistent manufacturing process; acceptable BA/BE

results; stable and consistent release profiles

without any discernible trend on storage for both

drug substance and drug substance 2. Dissolution

155

criteria which were set were better reflective of

the data. There was a substantial improvement in

the quality of the drug product and there was a

significant improvement in assurance of the safety

and efficacy concerns.

However, the case study example

highlighted two significant points. There was a

lack of or poor understanding of the raw material

properties and manufacturing processes that were

critical to be controlled for consistent quality

and thereby desired performance, for example,

extended-release dissolution of the drug product.

It also identified inadequate efforts invested by

the applicant during the drug development to

understand the causal links of dissolution

failures.

The case study example stresses a dire

need for improvement to the existing drug

development efforts to understand the relationship

between the raw material properties of formulation

components and critical quality attributes of the

drug product; the effect of raw material properties

156

of formulation components on their processibility

for selected manufacturing processes, and the

effect of manufacturing processes and associated

critical process parameters on the critical quality

attributes of the drug product.

I would like to end my talk with the last

remark that there is no substitute to a systematic

and scientific approach to drug development for a

safe, efficacious and quality drug product. Thank

you.

Questions by Committee Members

DR. COONEY: Thank you. There is an

opportunity for questions. Nozer?

DR. SINGPURWALLA: Just a point of

information, you repeatedly used distribution,

particle size distribution. What particle size

distributions do you use in your activities?

DR. SHAH: I am not following the

question.

DR. SINGPURWALLA: Particle sizes are

random.

DR. SHAH: Correct.

DR. SINGPURWALLA: They are not the same.

So, they have a probability of distribution.

DR. SHAH: Yes.

157

DR. SINGPURWALLA: Now, there is a lot of

literature, perhaps not in your business, on what

should be the distribution of particle sizes. This

morning we heard the viable distribution attacked

by my colleague here, but the log normal

distribution is often used as a distribution of

particle sizes. My question is what distributions

are used in the pharmaceutical industry for

particle sizes, or is this a completely different

scenario?

DR. SHAH: I am not sure how to answer

that question, but I will tell you what we practice

in CMC review. We ask for the applicant to

identify the particle size range in D10, D15 and

D90. That means 90 percent of the particles--

DR. SINGPURWALLA: Right.

DR. SHAH: And we ask for the span,

basically the ratio of D10 to D90 divided by D15

and that gives you where the distribution lies.

158

Basically, that kind of gives control of

consistency of the particle size distribution.

DR. SINGPURWALLA: Actually, you answered

my question. What seems to be not there in your

industry is you are just looking at the percentiles

and if the distributions are skewed one way or the

other it makes a big difference what they are when

you simply work with the percentiles. So, I am

just encouraging you to look into that.

DR. SHAH: I agree. Thank you.

DR. COONEY: Ken?

DR. MORRIS: I think one of the things

that occurs is that people don't control the

distributions. They tend to be log normal sort of

in a general sense but people don't intentionally

control this. They usually control to a mean,

which is a real big problem--

DR. SINGPURWALLA: If you control the mean

you start to control the distribution.

DR. MORRIS: Yes, you try to control the

mean but there is no real--and I am not sure what

historically the reason is for that but that is

159

sort of the case.

DR. SINGPURWALLA: You need to know what

it is.

DR. MORRIS: But you need to know what it

is.

DR. SINGPURWALLA: You can't control it.

DR. MORRIS: That is right. My question

is do you think that there problem was control

simply of film thickness or was it perhaps

incorporation of one of the compounds into the film

unintentionally during the coating process?

DR. SHAH: No, that was definitely the

coating process, and this was like black art in

that they were mixing and matching and they never

had a handle on the coating process itself.

DR. HUSSAIN: One point that I think I

wanted to illustrate with this presentation was

that really to control, to achieve a state of

control, and so forth, you have to get down to

upstream activities, starting with raw materials,

and so forth.

The point I also wanted to sort of

160

emphasize was that just focusing on a test, even

when you have a correlation, which is just a

correlation and may not be causal, I think is that

gap that we are also trying to fill with focusing

on the CMC part of the manufacturing controls.

Without that the system really--the method is

weakened. So, the quality by design aspect is to

emphasize that part of it. So.

DR. COONEY: I think another dimension

with this particular case is that there is a

significant amount of complexity because you are

dealing with multiple products, complexity both in

the process as well as in the product itself. This

is I think a particularly good example where

quality by design can have a greater impact with

these more complex processes and products, and the

processes and the products need to be thought

through together, which is your point. It is very

clear.

I think we are actually going to begin

lunch ten minutes early. However, beginning lunch

ten minutes early does not mean that you get an

161

extra ten minutes for lunch. We will reconvene at

12:50--guess what, you can get an extra ten minutes

for lunch. We will reconvene at one o'clock.

[Whereupon, at 11:50 a.m., the proceedings

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

162

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

DR. COONEY: If I could have people's

attention, welcome back from lunch. I hope that

everyone appreciated the extra 9.5 minutes that you

had for lunch. It is one o'clock. It is the

opening period for open public hearing. We have

one presentation for this afternoon by Will Brown

from USP, and he will speak with us on USP and

dissolution testing. Thank you. Welcome

Open Public Hearing

DR. BROWN: Thank you so much, and I would

like to thank the various FDA staff members for

giving a staff member at USP the opportunity to

speak before this committee. I am a member of the

staff of the Department of Standards Development at

USP, and I serve as one of the liaisons to the

Biopharmaceutics Expert Committee.

This is breaking news. USP reorganizes

itself once every five years, and part of that

reorganization is the election of the chair of the

Council of Experts. We have a reelected chair,

Thomas Foster, for the Biopharmaceutics Expert

163

Committee. You can see on this slide the

membership, and you will see names you recognize

hopefully.

USP and dissolution--well, we are terming

dissolution one of the performance tests.

Performance tests currently mean dissolution or

disintegration test, and by test I mean part of the

specification. The ICH definition, and it is very

easy to use terms loosely, says that a

specification is a list of tests, associated

procedures and acceptance criteria. So, that is

kind of the idea of the USP dissolution. It is

part of the specification. You will find the

public specification in the USP monograph.

The general dissolution test is found in

the general chapter, 7-11 on dissolution, and that

gives a general description of the techniques that

are available, with the understanding that those

techniques can be modified. We saw this morning

what the modifications might represent. They might

represent the appropriate medium or agitation or

apparatus as determined by the applicant and the

164

FDA.

Now, the study design that is embedded in

the dissolution test and the analysis is in three

stages. We have a fixed number of samples tested

at each stage and there are acceptance criteria

again that are determined by the applicant and the

agency, and then communicated to USP by what I am

terming the sponsor, who is the same party as the

applicant.

The general approach is to test by

attribute. In other words, a product is either

good or bad. It either conforms or it doesn't and

that is a fairly decent approximation and

convenient for application by an independent

analyst but it doesn't necessarily address

underlying distributions of performance.

In the USP test by attributes there is a

control on the spread of the data. By example, at

the S3 level where you tested 24 units there is a

limit that says that no individual unit value can

be below Q-25 percent. So, there is an

acknowledgement that there may be an underlying

165

distribution at least on stability.

For the Biopharmaceutics Expert Committee,

in this cycle the expert committee is working on

revising general chapters to include performance

tests by dosage form, by route of administration.

The current approach to applied dissolution is

typically two oral products and some transdermals.

The routes of administration that USP has

identified were discussed in a stimuli article in

Pharmacopeia Forum, in September, 2003 and

basically identified five basic routes of

administration, topical dermal, gastrointestinal,

mucosal, by injection and by inhalation. It is

just a way to cut the universe.

The intention is to work with the FDA and

industry as appropriate but to facilitate this work

the Biopharmaceutics Expert Committee has asked for

the formation of advisory panels, which have been

formed. They were formed in the last cycle and

they are currently meeting.

My general feeling is that meetings may be

productive but oftentimes they are not. I have two

166

examples of meetings that are productive. In 1993,

I am told that the predecessor to this committee

met and out of that ultimately, in '97, came the

immediate-release and extended-release guidances

that were talked about this morning. Another set

of meetings that happened in that same time frame

are the meetings of the Pharmacopeia discussion

group. The Pharmacopeia discussion group includes

the Japanese pharmacopeia, the European

pharmacopeia, the USP and the World Health

Organization. In the process of harmonization,

there actually has been a common statement with

respect to system suitability. It doesn't talk

about calibrators, however there is a provision to

have national text and in the national text portion

of system suitability the USP continues to describe

calibrators as part of the system suitability

determination. The general chapters are currently

at stage six and that information can be found in

the current PF and the corresponding Japanese and

European documents.

I was told that I only had ten minutes so

167

this presentation is briefer than I usually intend,

but I would like to draw the committee's attention

to possibly a useful document. This document

article by Walter Hauck and a group at USP talks

about oral dosage form performance tests, new

dissolution approaches. It is in the recent

Pharmaceutical Research, I think February 22.2. It

talks about an approach that has explicit

hypothesis testing. Parametric tolerance interval

is involved. It gives an improved way, or at least

the authors believe that it is an improved way to

set dissolution acceptance criteria, and allows

more flexibility in the design of a protocol. So,

I will just point you at that resource. It may

have some value.

It allows the industry representatives

more control on study design; allows the

opportunity for tiered testing. It doesn't

specifically talk about tiered testing but allows

that there may be an opportunity for some kind of

successive testing on failing to meet the criteria

at the first level. It allows some flexibility in

168

the number of units that are tested within each

tier, and it allows the possibility that the test

protocol, the test design could be changed from

manufacturer to manufacturer.

The idea is to set a probability of

passing units from a batch where the clinical

properties are known. So, you characterize the

batch for in vitro dissolution; determine, in some

kind of a discussion with the agency--again, I am

speaking from industry perspective even though I

don't represent any industry perspective--sets the

fraction of the units in this idealized reference

population or this actual reference population that

must conform to the standard.

This approach, and I won't be able to

describe this more fully, the authors believe will

allow the consumer and producer risks to be clearly

assessed, managed and communicated. Ideally, if we

continue with the model of dissolution for

performance assessment, this could be communicated

publicly in the compendium. The basic underlying

approach conforms to the approach for uniformity of

169

metered dose inhalers that I believe this committee

will be talking about tomorrow.

Finally on to calibrators, the system

suitability determination is written into the

general chapter and, as I interpret it, is part of

the performance of any dissolution test. So, if a

dissolution test is performed for compendial

purposes, currently USP requires that the apparatus

is demonstrated to be suitable, and the

demonstration of suitability includes successful

performance of the calibrators.

In actual point of fact, the use of the

calibrators has a GMP function. Test apparatuses

need to be demonstrated to be suitable twice a

year. So, that is the actual application of what I

believe to be more comprehensive suitability

determination. I don't currently work in the lab

but when I was in the lab if there were critical

dissolution experiments to be performed, they were

performed on an apparatus that was calibrated

before and after so that the integrity of the data

was not suspect on the grounds of an unsuitable

170

apparatus. The idea of calibration is not to focus

on the performance of the apparatus but to rule out

unsuccessful or unacceptable apparatus, so rule out

apparatus on the extremes.

The extremes--there is a range of

acceptable results that is determine from a

collaborative study, and we try to cast the net as

widely as possible so that we can capture the

sources of variability in properly operating labs.

Inter-laboratory variability is a major component

of the ranges. I would submit that any one

dissolution apparatus or assembly, because the USP

looks at the apparatus as a single vessel, single

spindle combination but, in fact, we have

assemblies, groups of apparatus. So, that is part

of the wideness of the range. We can talk about

that if you wish.

Calibrators, what we do with calibrators,

USP is aware of problems. Salicylic acid has

elegance problems. And, we go into unit packaging

in the latest batch. Prednisone tablets, the

prednisone tablets that we distribute are a

171

scale-up from the University of Maryland batches

that were intended to reproduce the NCDA2 10 mg

prednisone tablet and we have a new batch in

production. Theophylline beads are calibrated for

apparatus 3, were deleted partially in response to

requests or concerns by users. And, thank you

very much.

Questions by Committee Members

DR. COONEY: Thank you. I would like to

now open up your presentation to questions from the

committee.

DR. MORRIS: Just one point. I guess if I

sort of put aside for the moment the specific

hesitation I have about calibrator tablets, if I

just look at the criteria by which you would

reproducibly generate a standard for calibration it

seems to me that the things that are missing are

not that unattainable. I mean, I think that you

have to have a particle size solid fraction as

opposed to weight and, you know, the normal

controls you have on tableting. I don't think

there is solid fraction control in the calibrator

172

tablets, if I am correct. Art, do you know?

DR. KIBBE: No, and I want to go back and

have a solid cylinder without any particle size

where you only have dissolution for the surface.

The point that I was going to make sure I was right

about is that it is my understanding that if you

claim a USP product, that product must meet USP

testing if USP testing is done by anybody. But it

doesn't require you to do USP testing on your

product if you have a better way of controlling the

quality of that product. Isn't that right?

DR. MORRIS: That is correct. That is

clearly stated I think in the Journal of Medicine.

DR. KIBBE: Right. So, if I was a company

who was really heavy into PAT and had a really good

control on my product, once I established that my

product met USP guidelines for that product, I

would never have to do that test again unless I am

challenged.

DR. BROWN: That is correct.

DR. KIBBE: Right.

DR. MORRIS: Can I just ask is solid

173

fraction controlled in the calibrator tablets?

DR. BROWN: I can't speak to the

manufacturing of calibrators but I know that I have

seen formulas where the solid fraction beat

particle size. My concern about a disintegrating

dosage form, my personal concern is the rate of

disintegration and how fast the active is exposed

to the medium. That is critical and I take your

advice on the solid cylinder. I think that we see

very nice dissolution from salicylic acid tablets

which, again, are being essentially remanufactured

and repackaged, and I think will have the potential

to show some value that we don't currently extract

from them.

DR. COONEY: Tom?

DR. LAYLOFF: I was going to comment on

Art's question. You know, if you demonstrate that

your product complies and you cross-validate

against that, then you can enter it in the columns

because you would say it meets the specifications

and it is up to someone else to demonstrate that it

doesn't.

DR. KIBBE: Right.

DR. LAYLOFF: I mean, if you are

challenged you don't have to retest it. Their

174

challenge has to include the test.

DR. KIBBE: Right, that is what I said. I

mean, as a manufacturer I don't have to continue to

do this test at all.

DR. LAYLOFF: And if challenged you don't

have to repeat it again. They have to show that it

doesn't comply. The FDA has to go out and seize

it.

DR. COONEY: Marv?

DR. MEYER: In one of your slides on page

three you said inter-laboratory variability is a

major contributor to the width of ranges. If so, I

wonder why that is the best way of doing it. Why

not just have a really good machine in a really

good laboratory and do it? FDA has a beautiful

guidance on analytical methods for bio. studies and

they specify variability, and they specify

precision, and they don't say, well, this only

applies to University of Tennessee but Pfizer has

175

to do this and Teva has to do that. They say this

is what works; this is what we will accept. Why

not have a methodology that doesn't include inter;

the inter is their problem it should be the USP

problem. You should just have a good set of data

and everyone has to match it.

DR. BROWN: One of the things that you see

in dissolution labs is that there are apparatuses

that have tendencies. I am not sure what

constitutes a really good apparatus. I am not sure

that we have nailed it down even with the arbitrary

limits that were given in Dr. Buhse's presentation

this morning. I am not sure what constitutes

active variables in the dissolution. I am not sure

that any of us are. So, I am not sure what that

really good apparatus would be and would it reflect

reasonable expectations from various

well-manufactured products, apparatus products.

DR. COONEY: Tom?

DR. LAYLOFF: The dissolution experiment

is not a mystical science enterprise. Many years

ago we did an 11-lab collaborative study of FDA

176

labs where we required them to meet the system

suitability requirement before they continued the

test. Now, what that paper said--and this was done

I guess about 1990 maybe, what it said was that the

between lab results variance was about 2.6 percent;

the within lab was about 1.6, which is absolutely

consistent with the retrospective study that was

done by Bill Horowitz over about 50 studies in

AOAC, collaborative studies among labs. The among

lab results for various analytical techniques for

pharmaceuticals was 1.6 and the among lab data was

about 2.6. The ratio is about 1.7. And the

collaborative study for dissolution testing with 11

FDA laboratories, all required to meet a

suitability requirement, was in the same range,

which meant that the sample preparation with the

dissolution medium had the same variance as

analysis with other techniques.

DR. COONEY: Ajaz?

DR. HUSSAIN: I think the challenge we

face is if inter-lab variability is a concern,

which I am not sure--and to some extent I am

177

following what Tom is saying--I don't see how the

calibrator really does anything. It adds probably

the majority of the variability to that study. If

we are going to make a calibrator using same type

of excipients, raw materials with control of

particle size based on D50 percent and so forth,

how can that really give us any information when

the quality of the product being tested--the real

product might be far superior? That is the

fundamental flaw here.

At the same time, what is the quality

control strategy for the dissolution calibrator

tablet? It is another dissolution test. So, that

goes in circles. So, that is the challenge we

face.

DR. COONEY: Tom?

DR. LAYLOFF: One other thing, as was

noted previously, the calibrator tablet isn't

calibrator because you can't use it to calibrate

anything. It is a system suitability test tablet.

As Ajaz mentioned, it in itself can be a moving

target and your product may be less of a moving

178

target. If you have good control of your product,

if you do have control over your particle size,

excipient, polymorph--you control all those

variables you are home free.

DR. COONEY: It seems to me we are coming

back to a point that we talked about earlier, and

that is to have clarity in what it is we want to do

with the test that we are using, and to make sure

that we are not just simply taking a methodology

developed at some point in history and applying it

to meet some regulatory or perceived regulatory

requirement, and we need to really understand what

we are testing for; what we want to measure; and

what is driving the variance in those measurements.

Are there any further comments from the

committee? Tom?

DR. LAYLOFF: One comment, it is a

regulatory requirement and people lose product on

the market because of it. I mean, it is a very

real thing that is out there.

DR. COONEY: Thank you very much. This is

the only presentation we have for the public

179

period. Since there are no others we can begin to

go back to the regular agenda. The next

presentation, by Lawrence Yu, is on establishing

drug release and dissolution specifications.

Factors Impacting Drug Dissolution and Absorption:

Current State of Science

DR. YU: Good afternoon, everyone. The

assignment today is to talk about the factors

impacting drug dissolution and absorption: current

state of the science. This certainly is a big

title. My talk will cover three aspects, basically

in vitro dissolution testing. I am going to share

with you some of the limits to oral drug absorption

which has relevance to in vitro dissolution, and

finally I want to share with you some thoughts on

challenges to regulatory evaluation of dissolution

which have been discussed this whole morning.

This morning we discussed the variability

of dissolution testing. We discussed how to set

specifications from a biopharmacist and

pharmacokinetics perspective. We discussed it from

the manufacturing and control perspective. The

180

question is why are we doing dissolution testing?

Why do we dissolution testing for almost all single

solid oral dose drug forms, as well as the majority

of dosage forms like parenteral dosage forms?

Let's review the basic process of oral

drug absorption when a patient takes a tablet or

capsule. This tablet or capsule will disintegrate

or dissolve in the stomach. Dissolved and

undissolved drug will be emptied from the stomach

into the small intestine where dissolution

continues to occur. The dissolved drug will cross

intestinal membrane, will pass through the liver

and reach the systematic circulation.

So, from the mass transport perspective,

look at the processes, the fundamental processes

going on here. We have gastric emptying, transit,

dissolution, permeation, and metabolism. These

processes determine the rate and extent of

absorption, all of which we call bioavailability.

Because of the significance of

bioavailability with respect to safety and efficacy

of product, as you can see, dissolution becomes an

181

essential or critical step, the first step for drug

absorption. That is part of the reason why.

Another reason why we normally conduct dissolution

test.

Now, the question is what are we

measuring? Usually we use the classical equations.

I understand those equations are not perfect.

Sometimes they do not fit the dissolution profile

well but they give you a flavor of what are

critical variables involved with respect to

dissolution.

We have particle size which we mentioned

this morning. Basically, particle size effects the

surface area, so surface area affects the

dissolution. The larger the surface area, the

smaller the particle size, the faster the

dissolution.

We have solubility. Now, a number of

factors impact solubility, for example pH for

ionized compounds. For example, polymorphism

impacts solubility. So, those factors affect

solubility which eventually affect dissolution.

182

That is why we say the particle size, polymorphism

sometimes have impact on dissolution.

Finally, there certainly is bulk media.

Bulk media will determine the pH and pH, in turn,

impacts the solubility and, again, impacts the

dissolution. Therefore, if we review the factors

involved, we have particle size; we have solubility

and certainly bulk solvent.

In terms of utility, normally we have two

utilities involved, two kinds. Certainly under

each kind there are different kinds of utilities.

The first is for quality control from chemistry

manufacturing control perspective. I will give you

an example here. There are two polymorphic forms.

The polymorphic II automatically translates into

polymorphic I, for whatever reason such as

manufacturing, storage, and so on and so forth. It

will impact dissolution. So, therefore,

dissolution is a tool for quality control.

Another utility which we mentioned this

morning is so-called in vivo performance

evaluation. Now, what I show you here is very

183

beautiful, perfect examples. I have to say we have

a every chance to be successful in showing level A

in vitro and in vivo correlation. Nevertheless,

this is a real example where one of the companies

was able to come out with a beautiful IVIVC which

certainly can be used for waiver in the future by

current studies, which certainly can be used for

setting specifications, and so on and so forth.

But it is not easy. It is a challenge to establish

in vitro and in vivo correlation.

Now, despite its utility for dissolution,

there are many limitations. When we talk about the

limitations in the CMC or bio. area dissolution is

always a hot topic because it is so easy to

criticize. For example, you will say the

dissolution is over-discriminating, which means

that in vitro dissolution you find a significant

difference and you find no difference at all in

vivo. But in some cases you find no difference in

in vitro dissolution, yet, you find a significant

difference in vivo. So, we call it

over-discriminating.

So, in order to understand those phenomena

from the perspective of oral drug absorption, we

have to understand the oral drug absorption first

184

and particularly the limits to oral drug

absorption. That is what I want to talk about,

limits to oral drug absorption.

First let's review drug substance factors

and then we will review the drug product factors,

which is the disintegration. When we review drug

substance we mentioned that particle size and

solubility are two major factors which impact the

in vitro absorption and certainly sometimes, again,

in vivo absorption. So, in order for us to

understand the limits we have to define what are

the limits here. Yes, we have dissolution but in

vivo absorption certainly involves more than just

dissolution, as we discussed. In terms of process,

we have transit, gastric emptying. We have

permeation and metabolism.

But from an in vitro perspective, from in

vivo dissolution perspective what we define as the

limits to oral drug absorption is dissolution

185

limited absorption and solubility limited

absorption and permeability limited absorption.

Now, it is very easy to understand what is called

permeability limited absorption. It is simply that

the drug across membranes is very, very slow

compared to other process. That is why it is

called limited.

Now, with dissolution limited absorption

we have to talk about dissolution so the

dissolution limited absorption seems very easy to

understand. People ask me very often what is

called solubility limited absorption. Why do you

distinguish solubility limited absorption from

dissolution limited absorption? This is because in

the reality setting, especially for drug discovery

and development, we have to understand that the

concentration in vivo or in the gastrointestinal

tract of a patient cannot exceed the solubility, as

is shown here in the middle. Concentration cannot

exceed solubility. Now, with in vitro dissolution

testing we always have simulated conditions so,

regardless of the amount of drug, it can always be

186

dissolved but not necessarily in vivo. Think about

it, if you dump tons of drug into the human body,

they are totally saturated and the concentration

cannot exceed solubility regardless of how much you

dump into the patient or subject. It is not quite

useful anymore. This is the distinguishing

difference between solubility limited absorption

and dissolution limited absorption.

Mathematically, with dissolution limited

absorption we generally refer to particle size as a

major factor. Solubility limited absorption refers

to solubility. Now, solubility, of course, in turn

impacts dissolution but from an in vivo

perspective, because the concentration cannot

exceed solubility, that is why we define solubility

limited absorption. It has clinical implications

with respect to permeability limited absorption,

solubility limited absorption and dissolution

limited absorption. Let's look at it.

Theoretically, in order for us to define

under what conditions is dissolution limited, under

what conditions is solubility limited we have

187

designed some numbers and we have dose volume; we

have dissolution time; we have absorbable dose.

The dose volume is the amount required to dissolve

the dose. The dissolution time is minimal time

required to dissolve a single particle under the

same conditions, while absorbable dose is the

maximum amount of drug that can be absorbed under

certain physiological conditions such as solubility

and transit time.

Based on those parameters, we can define

under what conditions is dissolution limited; under

what conditions is permeability limited; and under

what conditions is solubility limited.

Let's look at the comment here. For

dissolution limited absorption or for permeability

limited absorption the absolute amount of drug

increases with the increased dose. So, it is very

simple. When you require high exposure in vivo

simply give more drug. That is usually the

practice in drug discovery and probably

development, particularly during animal toxicity

studies.

However, for solubility limited absorption

the absolute amount of drug absorbed does not

increase with increase of dosing. That is part of

188

the reason that for solubility limited absorption

you have to rely on other approaches to improve

bioavailability, to improve the absolute amount of

drug absorbed. Very often industry and our fellow

friends in toxicology department they often give

more and more drug and when they cannot see high

exposures they ask us why. The reason is that in

many cases it is because the absorption becomes

solubility limited, therefore, if you give more and

more drug you cannot see absolute increase of the

amount absorbed.

In terms of prediction--and this is the

permeability limited absorption, there is certainly

fraction of dose absorbed, although percent of drug

absorbed can be reasonable predicted with

mathematical equations. The data in this slide is

from human permeability but sometimes rat

permeability can be used to give us very good

results. K2 cell permeability can be used but does

189

not necessarily give us good results, in other

words, whether in humans the compound has high

permeability, can always high permeability in other

systems, including the rat.

This is another beautiful example to

determine how particle size impacts absorption. As

I mentioned, in this theoretical framework we

define dissolution limited absorption is because of

particle size so, as you can see, for digoxin, yes,

it is very poorly soluble; yes, the solubility is

only 60 mcg/ml. You are actually able to get 100

percent bioavailability. In this case there is no

metabolism, therefore, 100 percent absorption

equivalent to the bioavailability.

This morning we discussed particle size

distribution, and I have to tell you that in

simulation in these slides I simply used mean

particle size. Yes, there is literature out there

using particle size distribution in order to

understand how particles impact absorption. In

fact, in my publication we attribute that at 100

mcg prediction is not that good because we used

190

mean particle size instead of particle size

distribution because even though mean particle is

100 mcg, a number of small particles can be full

absorbed. That is part of the reason that you see

higher percent absorption, much higher than what

has been predicted.

Griseofulvin is a very classical example.

If you look at 100 mg griseofulvin, the solubility

is 95 mcg/ml. As you can see, when you give a

higher dose the fraction of dose absorbed has not

been increased, and part of the reason is when you

reduce particle size. It explains the utility of

solubility limited absorption.

This is a graph never published but I

thought I would share it with you because it is the

interplay of bioavailability, solubility,

permeability and hepatic clearance, showing in the

drug discovery setting of the drug is above the

surface. This means the bioavailability most

likely is lower than 30 percent. Below this, the

bioavailability is most likely above 30 percent.

So, if you want to increase the bioavailability,

191

depending on where the drug molecule is located,

you can increase the solubility; you can increase

the permeability; you can increase hepatic

clearance and increase the bioavailability. So,

that is the utility in industry settings.

This is a very brief overview of how

factors impact the absorption from the drug

substance perspective. This will give a very brief

overview about the drug product factors which

impact the absorption. Certainly, the

manufacturing process could impact the particle

size but I will mainly focus on disintegration, in

other words, how a tablet disintegrates and becomes

smaller particles.

In order to illustrate how the

manufacturing process could impact dissolution, I

will just give you an example here. This is a drug

which is highly soluble and highly permeable. So,

if you give oral solutions you can see the Tmax at

about 1.3 hours and it very quickly reaches the

Cmax, about 13 hours. This is basically the

solution curve.

Now, when you give a different dosage

form, because of slowing of disintegration you

could have a different dissolution profile, slow,

192

medium and fast. When we translate this in vitro

dissolution into in vivo here is the plasma

concentration profile. Obviously, the slower the

solution translates into lower Cmax and lower AUC.

Another case is when you have a different

coating system, this may be dissolved at different

pHs and in in vitro dissolution we are able to see

a significant difference and you also see it in

vitro, as you can see in this slide.

So, those slides basically illustrate how

the dosage form impacts dissolution, impacts

absorption. Certainly, there are many factors in

in vitro dissolution that impact, for example,

manufacturing process, compression force, and so on

and so forth. Those will impact the dissolution of

a product.

I want to share with you some thoughts

about challenges or opportunities which we are

facing today. First of all, when dissolution is

193

very rapid, outcome rate determining steps, drug

levels in the blood and plasma may not reflect

dissolution differences at all, as you can easily

understand. Because dissolution is not limited in

vivo, therefore, in vitro how much different they

are doesn't matter because you cannot see it in

vivo from the in vivo perspective. Yet, for

dissolution limited absorption sometimes we can get

IVIVC and sometimes actually in vivo fraction

absorbed is not always possible because of lack of

IV.

For solubility limited absorption, as well

as permeability limited absorption in vitro

dissolution will not always reflect into in vivo

because they are limited by solubility, because

they are limited by permeability, therefore, any

difference in vitro may not be translated into the

in vivo at all. So, we normally call that

over-discriminating.

Now, here we have f2, very useful,

similarity factors. Usually we use 50 or above

because 50 or above reflects average 10 percent

194

difference or less. I have to say this f2 value

certainly works fine or works well in most cases

but we can always see a criticism here because in

some cases the f2 value is not reflected in vivo

because, as I said, the f2 value is basically

average 10 percent difference. When you average 10

percent under different limits to absorption, those

may be reflecting in vivo or may not--may not see

in vivo situation.

So, as I said earlier in the discussion of

the presentation, there is a role of dissolution

for quality control and for in vivo performance

evaluation. So, the question that comes to my mind

is are these two goals always consistent? They may

not be. That presents an opportunity for us. For

example, under the hydrodynamic conditions--and we

have a paper from Rutgers University to show that

Reynolds number for chemical engineering for 50 rpm

as well as for 100 rpm--under those conditions we

may see some in vivo/in vitro correlations. In

fact, the paper suggests we may lower the rate so

that under laminar flow we could see significant or

195

more difference in the amount of dissolution. Yet,

when we go to the lower levels number and in

laminar flow, even though we may see in vitro, but

those may not translate in vivo because in vivo the

hydrodynamics is much more higher, the Reynolds

numbers--could be a lot higher. In fact, in the

stomach the motility is highest--educated guess, it

is difficult to see in vitro laminar flow in the

same situation.

And then there is the media. The media is

always an issue because in the real setting we will

evaluate a number of the media and we are trying to

define the dissolution a significant difference, or

most of the difference when we select in vitro

dissolution testing. If you make a survey of some

of poorly soluble drugs, for example, sometimes we

saw hydrochloride. Sometimes we see pH 7.4

phosphate buffer--these are all dissolution tests.

The question that comes to us is are those

dissolution tests or dissolution media reflected in

vivo? I have serious doubt that our human

beings--sometimes the pH could be 7.8 in the

196

stomach or sometimes could be 7.4, or sometimes the

concentration could be 2 percent or sometimes could

be zero percent. I think there is a wide spectrum

maybe in vivo. This is true but it may not change

as much as we see in in vitro dissolution

evaluation.

So the question that comes back is should

these two objectives for dissolution testing be

separated? I think this is the challenge in front

of us. We do not have a solution yet but I think

we ought to seriously look at the possibility and

the value and drawbacks as well as benefits.

For example, for dissolution quality

control, hopefully, hydrodynamics and media are

chosen for reproducibility and detection of product

changes, for example, particle size changes; for

example, polymorphic changes. Certainly, design of

in vitro dissolution test and for quality control

are not constrained by a desire to mimic in vivo

conditions.

But for in vivo biorelevant dissolution we

may choose a battery, a number of tests which

197

pretty much covers what is going on in vivo in the

human body. So, those tests will, hopefully, get

reasonably good correlations to in vivo. Yes, FDA

has lots and lots of dissolution data. Yes, we

have a lot of products approved that have required

dissolution data. Yet, when we look at dissolution

data and try to transfer those dissolution data

into knowledge, unfortunately, we almost get

nothing because every single drug, every single

product has used similar or even different

dissolution media. It has been difficult for us to

get to some kind of in vivo/in vitro correlation

even though we have lots, and lots, and lots of

data because the difference among in vitro

dissolution tests almost cannot be translated in

vivo. That is the difficulty.

So, in summary, believe me, I have

discussed in vitro dissolution testing and

discussed the limits to oral drug absorption.

Again, the limits to absorption could be

solubility; could be dissolution; and could be

permeation. We have briefly overviewed the

198

dissolution profile comparison and discussed the

future role of dissolution with respect to quality

control and in vivo performance evaluation. Thank

you very much.

Questions by Committee Members

DR. COONEY: Thank you. I would now like

to invite questions from the committee. Ken?

DR. MORRIS: I guess, Lawrence, one thing

that occurs to me is that really the distinction

between your dissolution and solubility limitation

seems to me--or, I guess this is a question, are

you really differentiating the source of the

solubility attenuation? In other words, aren't you

really looking at whether or not it is the activity

coefficient or the lattice energy that is

controlling the solubility?

DR. YU: I guess when we talk about

solubility, certainly there are two parameters. In

this difference, when we distinguish the solubility

limited absorption and dissolution limited

absorption, we infer that solubility is the

solubility final number, whether it is caused by

199

lattice energy or not.

DR. MORRIS: But that is not really the

case, is it? Right? I mean, there is a

distinction. I think it only changes the

terminology; I don't think it changes your

conclusions but maybe we can talk about it

off-line.

DR. KIBBE: I follow him a lot better than

where you are going.

[Laughter]

DR. YU: Thank you, Art.

DR. KIBBE: What he is talking about is

absorption and what drives absorption, and not

dissolution. What drives dissolution is a

different set of parameters.

DR. MORRIS: I don't think so. If you are

talking about the absolute value--I mean, you have

the same solubility but different causes for--

DR. KIBBE: No, he is talking about

different absorption with different solubility.

DR. MORRIS: But that gives you the

differences for the driving force. The driving

200

force is determined by the concentration or the

activity--

DR. KIBBE: Right.

DR. MORRIS: --no matter what.

DR. KIBBE: What he is saying is that you

have a series of events and if the rate of

dissolution gets you to a relatively low solubility

maximum quickly, then the whole thing is driven by

the limited solubility in terms of absorption. We

can do the numbers off-line.

DR. MORRIS: Yes. Yes, I see what you are

saying. I still think that your dissolution is

solubility limited. It is still just a different

solubility limitation, but we can talk off-line.

DR. COONEY: Ken, I think the point you

are making relates to one segment of the possible

space where you are solubility limited.

DR. MORRIS: Yes, for low solubility.

DR. COONEY: Yes.

DR. YU: I guess the solubility limited

absorption as well as dissolution limited

absorption is with respect to in vivo absorption.

201

You have to look at in vivo absorption when we talk

about those limits because with in vitro

dissolution evaluation you always same conditions,

at least based on FDA guidance. Therefore, you

always can see dissolution. But in vivo there is

dissolution going on but think about when the whole

small intestine is saturated by the drug, under

this condition we call it solubility limited

absorption. I don't know if I can explain it to

you better.

DR. MORRIS: I think I understand what you

are saying.

DR. YU: It is just terminology I guess.

DR. COONEY: Marvin?

DR. MEYER: Lawrence, the dissolution for

quality control is distinct from dissolution for in

vivo, that comparison. Why have two? If your

quality control is irrelevant to in vivo, I don't

see any particular relevance to it. Let's say you

want to have something you can detect changes in

the product, but what if you needed 25 percent

methanol in water to detect a change in coating

202

thickness, and you needed 10 percent methanol in

water to detect a change in particle size, and you

needed some other medium to detect a change in the

combination of the two and then, all of a sudden,

you would have three or four dissolution tests,

none of which may make any difference when it comes

down to the various product differences in the in

vivo setting? Why hot just focus in on the

biorelevant dissolution?

DR. YU: Well, I think the question that

comes back is that, first of all, in the 40 years

of dissolution history it is quite unusual. As I

mentioned, you really need a lot of laughing,

sunshine and good luck to get in vivo/in vitro

correlation, and even if you get it today it may

not exist tomorrow if you change the formulation a

little bit. That is why the famous words from Ajaz

are that IVIVC is formulation specific.

The question is why do we need to

separate? What is the value? I guess this was

presented to you today. What I would argue is the

following, when we talk about in vivo dissolution

203

testing it is quite complex. It is not so

difficult to do. If you think about it, you have

100 products, or one company has, you use 900

vessels and each and every day you consume a lot of

acid. So, if you think about it, those dissolution

tests can be replaced by simple quality control

dissolution tests of, say, water it is a worthwhile

effort simply from an economic perspective and from

a convenience perspective. However, you have to

say whether those tests are sensitive enough to

detect any significant changes in vitro, for

example particle size changes, polymorphic changes,

and so on and so forth. But in vivo dissolution

tests, hopefully, is a fixed battery of dissolution

tests which pretty much capture what is going on in

the physiological conditions. Now, whether we

actually can develop those or not we are waiting to

see. This is simply a proposal presented to you to

see what you think about it.

DR. HUSSAIN: May I? I think just to add

to what Lawrence just said, Marvin, the current

tendency is to use--the phrase "performance test"

204

was used to some extent--is to capture all the

other controls that are missing. Let me explain

that. We currently have a univariate focus on

quality. What it means is we do all the testing on

the content, rate and everything of the drug

substance only. But the excipients, distribution,

and so forth, is also relevant. So, dissolution

tests for quality purposes capture all those

aspects. Actually, I have an example in my slides

that I could show you later on.

So, that is the current philosophy of

needing a dissolution test even if it is not

relevant from an in vivo perspective to do that. I

think quality by design would say that if you are

controlling all aspects that are relevant, then you

may not even need one. I mean, we opened that

possibility to you.

DR. COONEY: Art?

DR. KIBBE: Which is kind of where I

wanted to go. The dissolution testing and a lot of

our terminal testing was developed when we couldn't

characterize our product effectively. The easiest

205

way of determining whether your tablet was hard

enough was to snap it and listen to the sound of it

cracking it. Friability was you took a handful and

dropped them on the floor, and if none of them

broke then your tablet was okay. That is where I

started in this business so, hopefully, we have

come a long way from that.

If we have a process under control that we

can characterize and we know the factors that are

affecting the way it performs in vivo because we

have looked at them with data in people, then you

say to yourself do I still have to have a test

which was originally thought of as a surrogate for

the in vivo performance at all? If so, what is it

getting me in terms of information to help control

my process?

The question I asked USP was, well, can we

still be USP without doing this test? Of course,

you can. You don't need to do this test if you

will eventually pass it whenever you felt like

doing it. You could do tests that are much more

useful and instructive in terms of whether your

206

process is under control. It reminds me of the law

that is still on the books in New York City that

when the car was invented the requirement was that

a man walk 20 paces in front of the car with a red

flag to warn everybody that the car was coming.

Now if they tried to enforce that law in New York

he would get run down by the 40 cars that were

coming. Perhaps we ought to look at dissolution

testing as the guy with the red flag, warning us

that we might not be in control anymore when we

really have much better controls on the system.

DR. COONEY: Are you suggesting we change

the red to a blue flag?

[Laughter]

DR. KIBBE: I was going to go for Claude

Raines and make the flag invisible.

DR. MORRIS: Actually, it is Kevin Bacon

these days.

DR. KIBBE: That shows you my age, right!

DR. MORRIS: I tried a Claude Raines in

class and nobody knew what I was talking about. I

guess to that point, if you can characterize it, I

207

mean sort of looking at statistical significance

without getting into discussion of statistics, you

can have a much larger fraction of the population

sampled during development than you are going to

ever have when you are doing manufacturing. So,

once you establish that design space, then by the

time you get to manufacturing, in principle you

should never have to do dissolution, which I am

sure would make everybody a lot happier if not a

little nervous.

DR. COONEY: Tom?

DR. LAYLOFF: I can see dissolution

testing as a regulatory tool to assure that the

manufacturer is putting out a product which

conforms to the specifications, but there is no

recovery from a failure in dissolution tests, short

of destroying the batch. I think that there have

been in place for a very long period of time the

tools to control the processes so that you can

deliver a product which would meet the test if

tested. So, I don't see the dissolution test as a

useful release test because if it fails that

208

release test you have failed the lot. It means the

process has failed, everything has failed. So, I

don't see it as a utility for a release test. I

see it as a regulatory tool, enforcement tool

because you can see using that to hammer somebody

but I can't see it as a release tool. If you have

to use dissolution testing as a release tool it

means you don't have everything in control.

DR. COONEY: I think what we will do is to

move forward to the next presentation. There is

clearly going to be more conversation and

discussion and that will be most appropriate after

Ajaz has taken us through his first 16 slides.

[Laughter]

Summary of Tactical Plan

DR. HUSSAIN: Yes, I will be brief and

just lay out the steps of the tactical plan and

stop at that point and open this for discussion. I

do have a number of slides in your handout which go

on to sort of illustrate some more deeper part what

we are thinking about and some illustrative

examples.

But just to summarize, I think what we are

trying to do is seek your recommendation on are our

tactical steps outline consistent with the goals we

209

are trying to achieve? What additional steps, if

any, of changes would you recommend to improve our

plan? What additional scientific evidence is

necessary to support development of the plan when

we come back to you to make the proposal next time,

hopefully? General considerations for identifying

and developing statistical procedures; any other

specific recommendations you may have.

Now, the proposed steps are in one clump

focusing on the measurement system, an alternate

regulatory approach, suitability of dissolution

measurement system which will rely on a rigorous

mechanical calibration and, when necessary,

measurement of degassing concepts that Cindy talked

about. This will be coupled with a

characterization of your clinical pivotal lot or

the bio. lot in terms of a gauge

reproducibility/repeatability study where you could

look at how sensitive this formulation is to

210

conditions of measurement systems such as

degassing, such as operator, such as the apparatus,

and so forth, in a structured design in an

experimental way since this is a destructive

sample, and there are means to do that.

Then we will focus on developing decision

trees, and I have put them in one clump. The focus

of the decision tree would be to establish

dissolution rate specification. I have requested

Lawrence to give you a flavor, just a flavor of how

we can start thinking from a mechanistic

perspective of setting mechanism-based

specification. He just gave you a snapshot of the

mechanisms that affect absorption. There is a

whole other set of mechanisms that define the

release of drug from the dosage forms, and so

forth, and depending on the types of release

mechanisms so that combination will allow us to

start thinking about how you approach a

mechanistic-based specification because that will

drive us to what are the critical factors that

affect release, and then work around that.

This is also an opportunity because often

people are concerned that when you talk about

mechanism this will restrict--that only few

211

mechanisms are acceptable. No, I think it actually

says you can have multiple different mechanisms but

then you will modify your specifications to that

specific mechanism and not force companies to one

set of dissolution specifications.

Clearly, I think the opportunities for

utilizing a control philosophy, quality by design

and our technologies under the umbrella of PAT will

replace these methodologies. As I mentioned, the

methodology would be part of the decision tree. As

part of the decision tree also would be the level

of process understanding and control to achieve--

to essentially create a concept of design space and

how this might be used for post-approval changes,

the type of changes that Mehul sort of illustrated

to you, and I will actually pick that up in my talk

tomorrow and explain that further.

Also, I think it is important for us to

make sure our decision trees are compatible and

212

equally open and transparent to all. Therefore, we

would like to develop a side-by-side comparison of

our decision tree for new and generic drugs. I

think one of the key aspects would be to come to

the committee to get an endorsement for the level

of quality assurance and quality control confidence

that we will have with our decision trees and our

control strategy, quality by design, would be

higher than what the current system is. To me, it

is a given but I think all of us need to be

convinced of that.

Today we will specifically seek

recommendations from you on the statistical

procedures, how do we want to proceed, and I have a

couple of slides on that. After this meeting we

will get busy and develop a detailed proposal for a

subsequent meeting for discussion. The timing of

this meeting was very important. We rushed and

tried to get this meeting because, as I said, this

weekend I go to Brussels and we are starting a

discussion on decision trees for dosage forms in

ICH and I needed this discussion behind me, and the

213

comfort of knowing, arguing and making the point

because we want to achieve harmonization in Europe,

Japan, and so forth. So, this meeting and your

recommendations I think will play a part in our

discussions.

Just to summarize what we intend to

accomplish with all this--improve our ability to

identify sources and type of variability, and to

ensure quality by design. Vibhakar illustrated one

simple example to you. If you only rely on

dissolution, even though you have IVIVC, a

correlation may not be causal. I think you have to

bring a control system perspective or control

strategy to assure quality, not just a test.

Obtain global estimates of variability to

use in regulatory decisions. Our current approach

can be improved. You saw our approach and

opportunities for improvement there. And, we would

like to use this information on variability,

sources of variability in how we set regulatory

specifications and process controls so that we

focus on controlling the real source of variability

214

and, for example, if the concern is that this

excipient is not uniformly distributed and that is

important, today the only test we have is

dissolution. If you have other means you can move

in that direction. We also want to use this

information for assessment of adequacy of proposed

material and manufacturing process control

strategies.

Facility, assessment and communication of

technology knowledge transfer and assurance of

state of control and production operation--this is

the current big gap between CMC and GMP, and I

think this will help us to bridge that gap.

Clearly, the basic philosophy is if you can

demonstrate a state of control that opens the door

for continuous improvement flexibility.

The inspiration for the proposal--there is

not a single thing which is new or unique in this

proposal. It is well established and we have

simply borrowed it from other sectors. The

inspiration was the DMAIC concept of defined

measure analyzing improvement control of the Six

215

Sigma concept. So, what we are proposing is in

practice in every other sector, literally.

But there are many challenges, and this is

an important challenge that I think we have

discussed several times, the pharmaceutical

quality, because of the challenges--the consumer,

the patients or the physicians cannot judge

quality--creates some challenges. Now, the key

concepts in Six Sigma are that you need to know

what is critical to quality attributes, attributes

most important to the customer. This is the

quality to clinical challenge. It affects failing

to deliver what the customer wants. In this case,

failing to meet your specifications or deviations

in GMP practices unless these really are focused on

critical to quality attributes, and so forth. This

is an opportunity. It is in green. We can get to

more critical variables and focus on signs rather

than following what our practices have been.

Process capability, what your process can

deliver, this is again an opportunity because the

concern from a CMC review perspective is that

216

process validation may not be adequate, and with

the sample size that we have from a USP market

standard perspective as a release test is a

significant concern, whether it is right or not. I

think that is debatable. But the market standards

are perfectly find. In my opinion, I think the USP

market standard is fine. There is nothing wrong

with that. I think the practice of using them as

release tests and as in-process tests, that is

where the challenge is.

But the challenge I think is variation,

what the customer sees and feels. We don't know

this so we have to go with what our signs say and

what the best practices are. Dr. Woodcock, in a

paper, called it market failure because you cannot

get the feedback from the customer really unless

there is a dramatic failure.

Stable operation is an opportunity, and

this is a significant opportunity for continuous

improvement because the regulatory agencies around

the world and industries also--corrective actions,

the only leverage for continuous improvement--we

217

have to get away from that mentality. And,

demonstration of stable operations can provide a

logical and scientifically rigorous way to

alleviate this concern.

Design for Six Sigma I think is designing

to meet customer needs and process capability, and

here the fundamental premise is that you design

your product for your patient, not for your

dissolution test. I mean, that distinction really

has to come through. And, many times I think if

FDA is asking the right question we can make sure

that happens. If FDA doesn't ask the right

question they will design to what FDA wants, and I

think that is a fundamental challenge. In some

ways that is a specifications capability gap that

can exist. So, here are the opportunities and

challenges and I think we have to address those as

we develop our decision trees.

Step number one is measurement system

suitability, and this is clearly honing down on the

target value, the mean value for your measurement.

It focuses on mechanical and media factors, but I

218

am not sure it is actually an independent step so I

have also put step number two, gauge R&R which is

essentially a qualifying one that you do once as

part of your approval, using a clinical pivotal lot

or a bio. lot. Here you have to think about

analysis of variance and the factors that might

contribute, like apparatus, dissolution media,

operator, clinical pivotal lot, and this a

structure design of experiment.

Now, clinical pivotal lot or gauge R&R has

some considerations. It is not just automatic. It

has to be supported by pharmaceutical development,

stability and sampling. By stability, here I mean

the process was stable from start to finish, not

from the conventional stability perspective also.

It is a statistical control.

Can we also do this for currently marketed

products is a question mark. Information that

would come from these studies would help us

facilitate a shift from a deterministic design

culture because we want to move towards assessing

variability and using variability. So, this is

219

step two. This could be an independent step but I

just wanted to sort of repeat that here.

The decision trees--I am not going to walk

you through in detail of the decision trees. I do

have some examples later on. How should we

consider moving towards decision trees? Lawrence

and the Office of Generic Drugs has started working

on a question-based review process, and the whole

thing is asking the right questions in a sequence

that drives you towards quality by design. So,

clearly, as part of the decision trees is what are

the key questions we should be asking, and so

forth. So, asking the right questions would be a

consideration. Clearly, beginning with the end in

mind, which is the intended use, keep the focus on

intended use as we develop these questions.

Systems base, connecting the key

disciplines and regulatory submission section--at

the previous advisory committee, as part of the OPS

Critical Path Initiative, I presented a proposal on

how to connect different parts of the section that

will be part of this consideration. Vince Lee

220

actually has started working on that and I think we

will see how that connects.

We also wish to facilitated structured

product development process. The traditional trial

and error of one experiment at a time will not

really cut it. You really need to have a

structured product development. Yet, we do not

want wish to dictate the specific process. That is

the challenge.

Pre-approval changes and bridging

studies--on average in a new drug application there

are three to six bioequivalence studies done on new

drugs--three to six. Some we don't even review.

If you really look at it, if you leverage that

information to bring in considerations that every

experiment you do is a hypothesis you have so many

opportunities to evaluate that hypothesis. So,

bridging studies would be a leverage.

The decision trees would be cumulative in

terms of leading to a decision but also support use

of prior knowledge. For example, we have wonderful

approaches to predict the impact of particle size,

221

and so forth. So, as you are in your early part of

development studies you can predict the impact of

particle size on manufacturability as well as

dissolution and bioavailability. So, if you have

such a system in place for the next product that

comes along you will make the prediction and

evaluate that so you postulate your next experiment

that you go into in vivo that this is what I

expect. Once you do that, you start setting up a

learning system that at some point becomes very

useful.

That means that we will move towards a

scientific hypothesis format. I have removed those

slides. Those slides come later on if you have any

questions.

How do se address the challenge, the two

challenges of market failure and quality to

clinical gap? In a sense, one of the challenges I

think we have is we think about risk and

uncertainty and we confuse that, and this is

again--I learned from this committee and I will

share with you a table of how we can separate

222

uncertainty, variability and risk tomorrow. I

shared it at the training session yesterday with

some of you.

I think our current assumption that we

work under is that our methods that we use are most

discriminating, therefore, risk is mitigated. So,

if that is the case we are not dealing with risk;

we are dealing with uncertainty and how you

approach that opens up a whole new set of

approaches. So, if you are dealing with

uncertainty lack of knowledge is the challenge.

So, improve the knowledge that leads you to that.

So, from that basis, product specification based on

mechanistic understanding provides a means to

address uncertainty and that will be a progression

that we will use.

So, if you really look at it, the way we

set specifications is because of uncertainty.

Uncertainty management without pharmaceutical

development information is a challenge. We focus

on a discriminating test concerned with in vivo but

without the pharmaceutical development information.

223

The discriminating part comes from the method

measurement system. Is it really discriminating

the formulation variables that really impact?

Sometimes you have a disconnect there.

So, often we have a shotgun approach. Dr.

DeLuca was sort of raising that question, why do we

ask for three or four dissolution media sort of

blindly? If we had an understanding of what the

physicochemical aspects were, if we knew what

dissolution media or conditions would be most

discriminating why would we ask for more? We

wouldn't. We shared that. And, the rule of thumb

that I shared with you, which is a 30 year old rule

which we still do not practice in industry is when

you have certain acids or bases, however sensitive

PK is, 4-6 or 3-6, go with the PK8 value. That is

the most discriminating. That is a well

established rule of thumb but we don't utilize

that.

So, often it is the shotgun approach, 3-5

different media. We just focus on pH. Now, the

surface tension of the dissolution media is about

224

70/cm. The surface tension of the GI fluid is

about 30-50. It makes a big difference. So, why

aren't we doing it at that surface tension? So, it

opens up all the questions that Lawrence raised.

In practice, I think the frequent tendency

is to utilize 0.1 normal HCL and I illustrated the

concern from the perspective of the Japanese. If

that is a real concern, then we have to think about

how to address that.

Quality assurance versus in vivo relevance

debate, I think that will be part of the decision

tree process. But one aspect that we have to think

about is mechanistic understanding. We haven't

defined it. And, I am proposing that we will use

the ICH Q6A concept that we have already accepted,

and here is that concept. For example, this is

section 3.3.2.3 of ICH Q6A, particle size

distribution testing may also be proposed in place

of dissolution testing when development studies

demonstrate that particle size is the primary

factor influencing dissolution. So, if you are

able to make that decision for a parenteral

225

preparation why don't we make that decision for an

oral preparation? the reason here is because the

parenteral suspension is a suspension. In a solid

dosage form you have compaction, and so forth, but

we have technology that can even address that so

this becomes a meaningful way to move forward.

But, again, as was pointed out, the particle size

distribution should be well characterized and well

represented.

So, mechanistic understanding as a

proposal is that identification and scientific

justification of causal, physical or chemical

relationship between pharmaceutical materials

and/or process factors that impact quality. Here I

want to draw a distinction. Establishment of

correlation may not be causal. So, we want to be

careful and that is the reason why in the PAT

guidance we said correlation may not be sufficient.

We need causality and that is process

understanding.

I think this is a significant debate and I

think we may be able to achieve it in the U.S. but

226

it has to occur in this. Specifications equals

standard. That is the philosophy now. That is the

reason why release tests are USP market standard.

Now, this is well ingrained. I would propose we

need to start thinking about changing that. Market

standards have a value, as Tom pointed out. Market

standards are fine. You need them from the

perspective that Tom laid out. But then the

release specifications and control studies should

be different and be risk-based. And this is the

challenge.

So, specifications are standard.

Non-conformance means you have to reject or recall

a lot. Now think about this, most companies have

this but we haven't utilized it. You have to start

thinking about a control limit with a target value

and an acceptance range around a common cause of

variability, with special causes of investigation,

special causes to investigate. So, if you think

about this, one of the reasons why the concern is

if you start reducing your variability the

regulators around the world will say you have

227

narrowed your specification is this phenomenon

because, if you don't, they may not investigate and

if there was a special cause that might be a signal

for something else. So, if you move towards a

control philosophy and if you bridge this gap

between CMC and GMP this is a way forward. But,

again, there is no consensus around the world on

that.

Step six, general consideration for

identifying and developing statistical procedure.

In my presentation I also identified the recent

contribution from the USP. I think it is a step in

the right direction. But it is, again, very

similar to the parametric tolerance interval test

concept and we have been debating that for at least

the three years that I have been involved; I think

it is more than that. And, you will hear a

progress report on that tomorrow from Bob O'Neill.

This is where I really need the

committee's help. We have to start thinking

differently. Testing a hypothesis on every

production batch is not the right way of thinking

228

about this when you are in a state of control. You

are not testing every production batch as a

hypothesis. The gap between CMC and GMP I think is

validation. Hypothesis testing--my proposal would

be to be limited to the validation part of it, and

you really test the hypothesis that you have

transferred the technology appropriately, and so

forth. So, specification setting and standards and

hypothesis testing, such as parametric or

non-parametric tolerance interval--and we have been

working on that. This is a recent hypothesis

structure. There was a recent proposal from our

Office of Biostatistics for dissolution. It should

be limited as the time of approval and validation

that our CMC reviewers make that call, and so

forth.

But then in production you have to move

towards control. So, you have to start thinking

about control charts of variable, not attributes,

where you really have a focus and target value and

risk-based upper and lower limits; process

capability analysis, not hypothesis testing on

229

every lot. This is important. This is a very

important principle on which I need your help.

So, I will stop here. The questions are:

Are the steps outlined consistent? Any additional

steps you might recommend? How should we prepare

to come back to you? What other scientific

evidence is necessary to support this? General

consideration for identifying and developing

statistical procedures and any other thoughts or

comments you have on improvements.

Committee Discussion and Recommendations

DR. COONEY: Ajaz, thank you. I would

like to take some time now for the committee to

raise questions and discussion. We will probably

break at some point, have a chance to think things

over and then come back and have more time for

discussion, especially given the importance of the

recommendations that you seek us to make. Nozer,

you had your hand on that button quickly.

DR. SINGPURWALLA: I have had it ever

since he stood up.

[Laughter]

DR. COONEY: The floor is yours.

DR. SINGPURWALLA: Well, I think you have

come a long ways from the last meeting when I

230

believe there was a discussion on mechanistic

considerations. What I would like to add is,

Lawrence, you put up a differential equation.

Remember that? You put up a slide; you put up a

differential equation.

DR. YU: Differential equation, right.

DR. SINGPURWALLA: Now, what struck me

when you put it up is certain things. The first

thing is you didn't define everything, which is a

bad way to put up an equation but I won't punish

you for that.

[Laughter]

But what Ajaz said has some relevance to

what you said. That differential equation you put

up is a mechanistic equation.

DR. HUSSAIN: Yes.

DR. SINGPURWALLA: It is a mechanistic

phenomenon.

DR. HUSSAIN: Yes.

DR. SINGPURWALLA: And there is

variability.

DR. HUSSAIN: Yes.

DR. SINGPURWALLA: And one of the elements

of your equation was the particle size. We have

had some discussions here about the particle size

231

not being a variable entity. Therefore, one of the

things I would recommend doing is--you asked what

additional scientific evidence you need--I would

encourage you to look at that particular equation

and make it stochastic so that it becomes a

stochastic differential equation, and I would use

whatever knowledge you have about particle size

distributions, and my colleague from Pfizer has

given me some clues about what could be particle

size distribution, and somebody else also. I would

like to suggest that you merge the two, the

deterministic, mechanistic equation and the

particle size equation.

The comments you made about correlation

and causation are germane and correct. But the

reason is this, that correlation only measures

232

linear relationships and is not indexed by time.

Causation is a time index phenomenon. If I smoke I

will get lung cancer, assuming that that is the

causal. So, there is a time phenomenon.

One of the slides that you put up and you

quickly slid by is wrong. That hypothesis.

DR. HUSSAIN: As an example--

DR. SINGPURWALLA: That is wrong.

Somebody has to fix it. I wouldn't put it up

again.

DR. HUSSAIN: You have to tell that to Bob

O'Neill tomorrow.

DR. SINGPURWALLA: No way!

[Laughter]

DR. HUSSAIN: Thank God, it was not mine.

DR. COONEY: But, Ajaz, you have been

adequately warned!

DR. HUSSAIN: Yes.

DR. SINGPURWALLA: I will tell you what

the problem with that equation is. You have two

probabilities there. You have the null hypothesis.

You are making hypotheses on probabilities of

233

certain events.

DR. HUSSAIN: But that is the basis of the

parametric tolerance interval test that you will

hear tomorrow.

DR. SINGPURWALLA: Maybe I should stay at

home!

DR. COONEY: Ken?

DR. MORRIS: I mean, I think this is

clearly the right track and perhaps overdue,

hypothesis testing aside for the moment, but the

question I have is, is it within the scope of what

we will all jointly do to say that the dissolution

test, for example, is development activity as

opposed to a manufacturing control for either bio.

or control of process development?

The other thing is that if the decision

tree--I can't remember which one it was now--where

you had mentioned the design space section for

development, that is really the whole enchilada

here. Right? I mean, in a sense, if we do that we

have done everything.

DR. HUSSAIN: Right.

DR. MORRIS: With respect to laying out

quality by design development.

DR. HUSSAIN: Yes. No, I think, Ken, in

234

many ways the decision trees should sort of focus

on not directing what to do but more in terms of

what are the key questions that need to be asked

and direct what the decision should be. So, that

is the only way we can avoid interfering with the

development program. We don't want to do that.

So.

DR. COONEY: The use of the decision trees

acknowledges the fact that not all drug products

behave in exactly the same way--

DR. HUSSAIN: Yes.

DR. COONEY: --in a standard system, and

not all drug products act physiologically by the

same mechanism.

DR. HUSSAIN: Right.

DR. COONEY: As I understand what you put

forward, these decision trees would allow you to

have a standard process to develop methodologies,

but not constrained to always having a product work

235

the same way.

DR. HUSSAIN: Right.

DR. DELUCA: Your number three there--I

think you developed the first two very nicely.

Number three, what additional scientific

evidence--you know, once the plan is outlined, I

think that is the decision tree. Once the plan is

outlined and the steps are rationalized then, you

know, the additional evidence that is needed will

surface.

DR. HUSSAIN: Yes.

DR. DELUCA: And I think that will surface

from the decision tree.

DR. HUSSAIN: Yes.

DR. DELUCA: One thing I want to say about

knocking out the dissolution test--

DR. HUSSAIN: We are not; we are not.

DR. DELUCA: No, I know but I am saying,

you know, when you talk about particle size

distribution in parenterals for a suspension that

is very straightforward.

DR. HUSSAIN: Yes.

DR. DELUCA: I mean dissolution is going

to be related to particle size.

DR. HUSSAIN: Yes.

236

DR. DELUCA: But you are talking about an

order of magnitude of particle size distribution

here a lot different.

DR. HUSSAIN: Sure.

DR. DELUCA: I mean, you are in an oral

form and, in fact, I am not even sure that in the

parenteral size distribution is as important as

having a minimum particle size. You have to be

below a certain particle size; you have to get it

through the needle. So, the thing is it has to be

very small and so that makes a difference and it is

related then, so you don't need to do a dissolution

test if you have a low enough particle size. But

with the oral products you have so many other

factors involved--

DR. HUSSAIN: No, that is the reason I

sort of clarified that I used that as an example to

construct my statement of what mechanistic would

be. But the key aspect was, in a sense, if you are

237

able to achieve a state of control with some of the

new technologies--in fact, Jerry, at a science

forum a couple of days back, presented an imaging

approach to looking at dissolution prediction. I

didn't mention that to him but it completely

coincides with the percolation theory that Hans has

been progressing and just by looking at that you

could have said that. The purpose of that was to

simply construct a statement of mechanistic basis

for a decision to move forward. So.

DR. COONEY: Tom?

DR. LAYLOFF: I think it is quite striking

that we are having this discussion now. You know,

it has been about 30 years since we started looking

closely at content uniformity and putting market

standards on that, and looking at dissolution as a

means to assure bioavailability. It has been 30

years. And, I think that one of the things that we

have seen is what Cindy was saying, FDA likes

method 1 and 2; FDA likes content uniformity

testing, likes dissolution testing. So, the

industry, instead of focusing on quality systems,

238

directed research and formulas and production,

actually kept that focus on a 30 year-old concept

that dealt with the market issues 30 years ago

instead of building the quality systems that deal

with the present. I am happy to see those other

ones go away and let's go on with quality systems.

DR. COONEY: Art?

DR. KIBBE: Yes, since we are really deep

into discussing particle size, let me throw out

that really the key ingredient in understanding

dissolution is surface area measurements and not

particle size, and that particle size distribution

means different things to different kinds of

formulators at different times and we have a really

good understanding of the kinds of particles we

want when we compress a tablet and what percent of

them are fine, and so on. So, it is a different

game.

One of the issues that comes into play is

not only does the particle size matter but if you

have too small a particle size you start to get

thermodynamic forces acting to cause aggregates,

239

which give you a larger particle than you started

out when you measured it the first time. So, what

I am getting to is that the issue is complex and

the controls have to be put on for each system in

each situation, taking into account what we know

theoretically, and not impose as a general rule

that just sits on top of everything.

DR. COONEY: Ajaz, one of the

implications, as I understand it, of the proposal

is that one would develop methods based around

understanding of the science, the mechanisms that

are controlling the important phenomena. This

suggests that the responsibility goes right back to

the developer--the developer, CMC and the

manufacturer, cGMP, and you are going to bring them

together. But this responsibility goes back to

them to identify what those mechanisms are as

opposed to simply adhering to a standardized assay

of some type. This is the implication as I see it,

which will change the way the development work is

done in the first place.

DR. HUSSAIN: It may change. I think

240

maybe the industry colleagues--

DR. COONEY: That was an invitation, Gerry

and Paul, to respond.

MR. MIGLIACCIO: I mean, it is consistent

with the ONDC restructuring. It is consistent with

getting away from the check-box mentality of the

CMC section to put the appropriate science into the

CMC section. So, yes, it is going to change the

way you do things and, hopefully, it will change

the overall process of reviewing, approving and

then continuous improvement around our products.

So, yes, there is no doubt about it.

I guess, since I have the microphone, why

tactical step two? What is behind your question

around currently marketed products? Because while

quality by design for new products is starting to

be a reasonably well understood concept, you cannot

go back redesign a marketed product by those

principles. There is a lot we can do with

currently marketed products so why did you say

currently marketed products, question mark, when it

comes to tactical step two?

DR. HUSSAIN: So, you picked it up! Well,

I think clearly there is a hesitation there, and

the hesitation comes from the fact that what we

241

have here--the challenges we have are destructive

test and the hesitancy is variability of units in a

pivotal clinical batch to be used is a key concern.

So this is on the development side right now. And,

in a sense, what you do is you declare this as

acceptable. So, in the current scenario our

specifications might be tighter than the capability

of that and that is the means for imposing

continuous improvement.

So, the development information and the

stability of the batch really overcomes this

hesitation. That essentially is the key here. For

step number two, one of the conditions that becomes

is for this approach the clinical pivotal lot or

the bio. lot must be stable--stable, I am talking

about consistent--and its variation understood to

the extent that the unit may be sampled for a

destructive gauge R&R. That is the fundamental

principle of a destructive gauge R&R.

Now, we don't have this information for

currently marketed products so that was the reason

for the question mark. Can you overcome the

hesitancy without that? But I did put that, and so

forth, to illustrate that I think we need to move

in that and do it for those also.

242

Here is an example of what I mean by

stable process. Is this a stable process? This is

actually real data from a company. A friend of

mine sneaked it to me. So. The percent

dissolution as a function of manufacturing time,

and you see that in this case this was

non-homogeneous distribution of an excipient, which

we never check, was the cause of this. But most of

these are never caught. We approve those; they are

part of the system. I can show you many examples.

Validation may not always catch this because the

variation, as Gerry you pointed out, you learn more

in manufacturing than anything else. So, for the

gauge R&R step two to occur, you will need to

demonstrate complete stability of this lot.

DR. COONEY: Paul?

DR. FACKLER: To think about the

implications of what is being discussed here for

generic products in particular, it is gratifying to

hear that FDA recognizes that dissolution

parameters really are based on a formulation and

very often, in fact most of the time, generic

products have a different set of excipients than

the brand product that they are bioequivalent to.

It makes one wonder then why the generic company

243

would be held to the same dissolution

specifications that the brand product, with a

different set of excipients and different

formulation, might be held to.

So, I support a revisit of the whole

dissolution concept and, frankly, it is something

we have struggled with. We have a bioequivalent

product, for instance, yet can't meet dissolution

specs based on a different set of excipients.

DR. HUSSAIN: No, I did allude to that

fact in the background paper that you might have

seen. Dissolution specification is such a complex

performance test that depends on many factors, such

244

as hydrodynamics. If it appropriate for one

formulation design, it may not be appropriate for

the second. That creates that. In absence of

additional information such as further development

to understand how the formulation might behave, we

have very limited choice--I mean, that is a concern

that you see from the regulatory side. The way we

approaching it I think we can alleviate that

concern and actually address that scenario, and

this is one of the reasons to address that scenario

too. So, you are right.

DR. COONEY: Marvin?

DR. MEYER: Ajaz, not being familiar with

this area at all, let me ask a question anyway.

How easy is it going to be in all cases or all

complex formulations to develop a total mechanistic

understanding of the behavior of a given product,

or will we sometimes have to say, well, we made a

decent effort; let's go back to the old dissolution

and whatever?

DR. HUSSAIN: No, I think you will not

have a full level of mechanistic understanding, and

245

so forth, but if you approach it right from the

beginning and you start building it so you

development information really will add to that.

So, from a control strategy, in most cases you will

not have complete mechanistic understanding but

then at least you have enough understanding to put

in the right controls, and so forth, and not rely

on one particular test at the end. So, I think it

becomes a combination of the level of complexity

you have and the degree of uncertainty you have

that provides a means to say this is the control

strategy. So, you have to marry or arrange your

control strategy in light of the complexity and in

light of the uncertainty that is present with

respect to the intended use of the product. So.

DR. MEYER: I can certainly see from a

manufacturing point of view that a total

understanding of the product and the process would

be great, but it is also great to have a product on

the market that works.

DR. HUSSAIN: Sure, exactly.

DR. MEYER: So, the agency isn't proposing

246

that ultimately down the line I am sorry, you

didn't identify--

DR. HUSSAIN: No, no at all.

DR. COONEY: What I would like to suggest

at this point is that we take a break for 20

minutes--this is by design, an extra five minutes

but this comes with a price. As we think about the

questions that have been put before us, the six

questions, I would like for the committee to think

about, as they are sipping their coffee,

particularly questions two and three, what

additional steps or changes would you recommend to

improve what has been suggested? And, three, what

additional scientific evidence is necessary to

support the development and implementation of the

plan?

The reason for the extra five minutes is

because I think it is important that we think about

the impact of the recommendations that we are going

to address later today, beyond the discussion, are

there unintended consequences? What are the

impacts that we need to think about, not that we

247

need to identify them all, but what are the things

we need to think about as we go forward? So, for

that an extra five minutes of coffee break. Let's

reconvene at about 3:12.

[Brief recess]

DR. COONEY: I would like to welcome

everyone back. We have a period now where we can

have a discussion around the questions that are on

the table. The questions that we have been asked

to consider are summarized on the screen. What I

would like to do in the next hour approximately is

to have an open discussion amongst the committee on

these questions. I would like to get to a point

where we can take a vote as a committee in terms of

voting either--I will ask the voting members of the

committee to support the recommendations of the

committee going forward. I will do that at the

end. It will be yes, no or abstain. You will be

given options. The voting members of the committee

will be the committee plus the consultants that are

here.

But prior to that, I think the request

248

that we have on the table is a plan that is laid

out on the slide that is before us that has a

series of steps. The first, and I think the

central question, is are the tactical steps

outlined--and these are outlined in slides 8

through 15 of Ajaz's presentation--are they

consistent with the goal of quality by design? I

thought we might begin by talking about these

tactical steps first, have some discussion around

that, and then to work through the other questions

that we have been given. It is particularly

appropriate and important for the committee members

to take this opportunity and share with Ajaz and

the rest of the team their ideas and thoughts, and

it is my understanding that, if these

recommendations are approved, what it represents is

a step forward; the questions that have been

outlined will be examined and possibly as early as

our next meeting specific recommendations for our

discussion will be brought forward.

With that, let's take a look at question

one which really revolves around the tactical

249

steps, slides 8-15, that we have in front of us.

Then perhaps we can go through these steps with an

open discussion around them. Ajaz, could you go to

your slide 8, which is tactical step one? This is

really the first two steps in this plan. Given the

limitations of the dissolution assay as currently

practiced, and its relationship or lack thereof to

therapeutic efficacy and safety, to look at

alternative suitable methods and strategies in

order to evaluate the quality of drug products.

DR. SINGPURWALLA: Let me just get the

ball rolling. Steps one and two essentially

encapsulate what you have in mind in the boxed

items: Information collected should facilitate a

shift from the deterministic to a probabilistic

design culture. That is true for step one and step

two. And, I don't see any reason why we should not

endorse it. It is the natural thing to do and my

particular position on this is to go ahead and

endorse it, at least step one and step two, as I

see it.

DR. COONEY: Let me also put on the table

250

the question are there suggestions, modifications

that one might make to these steps? We can talk

about that as we go through this as well. Gerry?

MR. MIGLIACCIO: Yes, I mean, the whole

discussion today and steps one and two really focus

on dissolution. I guess I thought what we were

trying to get to is a more scientific measure of

performance, process performance being the quality

assurance measurement and product performance being

the in vivo. So, do we need to start with that

decision criteria first, Ajaz? I mean, the

tactical steps are all focused on dissolution

without saying is it really dissolution that we

should be talking about.

DR. HUSSAIN: Clearly, I think the

thoughts we had--why is step one the first step,

and I think that is the key. The way we see it is

the current dissolution test system, as we use it,

is an essential decision tool during product

development and for regulatory decisions so we have

it right now. So, step one and two are overcoming

some of the challenges that we have.

Gerry, what you are suggesting--we are not

eliminating that target at all and what we are

doing here is improving upon one tool that we are

251

using currently, and we anticipate using currently

and in the future also, improve it and then also

work on other ones, and other decision trees that

we are planning will build in aspects that I think

you are alluding to. So, our thought process of

why is step one the first step is that we

anticipate dissolution testing--that the

methodology will be with us for a long period of

time, and here the uncertainty with respect to the

suitability criteria and how we set acceptance

criteria--there seems to be a disconnect here and

there is a lot of frustration around these issues.

So, step one and two combined address the

immediate need, and also set up a system which is

more rugged, we can be more confident to start

building alternate methodologies.

MR. MIGLIACCIO: So, we are really talking

about more scientific approaches to dissolution.

DR. HUSSAIN: Step one and two, yes.

MR. MIGLIACCIO: I understand gauge R&R

well but I am not an expert. Can you get enough

information out of a gauge R&R on a single bio.

batch?

DR. HUSSAIN: One of the key aspects that

we are trying to get from this is that in most

252

simple cases I think we think the study will not be

a full-blown gauge R&R, and so forth. We think

that product development information will guide you

through that. Now, there are elements and there

are arguments out there that you need to know how

sensitive your particular formulation is to the

conditions of the dissolution test. So, from that

aspect, I think this will allow us to gauge the

sensitivity of your formulation, your particular

formulation to the chosen dissolution method.

Hopefully, if it is done early you would remove

that, and so forth. This actually then becomes

simply a study to benchmark variability. So, that

is the reason I anticipate that for most simple

cases this may not be an extraordinary effort

necessary to really do a full-blown gauge R&R.

MR. MIGLIACCIO: But you don't get the

batch-to-batch variability.

DR. HUSSAIN: Correct, but at least you

benchmark the product and then you could get

additional benchmarks from your validation batches,

and so on and so forth.

DR. COONEY: Art?

DR. KIBBE: I agree with my statistician

friend that we probably ought to move forward, but

253

I think Gerry raised a really important point, and

I was going to raise it too, that is, this whole

concept really started several years ago when we

started really talking about what is the essential

information we need to know to make sure that we

have good quality products for the citizens. That

started with PAT and it kept going. And, this

really is just a natural evolution of the

regulation of the quality of the process that gives

us pharmaceuticals for sale in the United States or

anywhere.

So, as long as we, as a committee, and the

FDA, with you as its spokesperson, understand that

254

this is just one more small step moving forward and

not a whole process--I mean, we are not about

dissolution testing; we are about eliminating

unnecessary testing and doing the correct testing

to make sure we have quality products. That is

one.

Second, the tremendous variability in

human response to a given drug product is just like

a thunder cloud on top of a small camp fire of

dissolution testing. And, if we get better and

better at controlling the process and if then the

batch is slightly out of that control, that doesn't

justify not selling that batch because it is well

within the goal posts that we have been working

with from the beginning. So, I think we need to

also fold into here the understanding that as the

process gets under tighter and tighter control

deviations from that control must be investigated

and must be understood but aren't necessarily a

justification for not releasing the batch. I don't

know how you fold that in. When we start talking

about dissolution as a terminal test, whether we

255

have surrogate markers that track the process and

tightly control it, and if we are a little bit out

we can allow the company to do a quick extra test

and say, yes, it is still good enough but we are

going to find out why it is out. If we don't, we

are going to scare off half the companies from

following us down this path to really tight control

systems.

DR. HUSSAIN: Your first point is well

made and in the break I was asked to sort of

summarize that, and you just did that for me. But

the second point is more challenging. If your

release test right now is a compendial test, that

is a market standard. You have no room for

anything there. That is the law. So, that is the

law. So, how you sort of address that I think is a

much larger issue. In the PAT guidance, if you

recall, we actually suggested in a sense that there

has to be a way for moving forward and we created a

system which we call research data so if it is an

alternate procedure that you are using that is not

your decision based on your compendial, you have a

256

way forward under the PAT guidance for that right

now.

But the point is, in a sense, for

standards you have to draw the line somewhere and

compendial standards draw the line. You have to

conform to that. That is the law. So, the whole

strategy then is that I think you have to move

towards what I would like to sort of share with

you, a concern that you expressed, and it is an

important concern--if I can find my slide--that I

think goes to another dimension.

I think we are moving towards a state of

control and demonstrating a state of control is

important but, at the same time, for some processes

we will never reach the state of assured

statistical control. So, we have to bring a

risk-based decision to that in the sense of how do

you sort of manage that because every deviation is

investigated. So, there is an element that we do

intend to discuss, and discuss internally as well

as I think in workshops, and so forth, the need to

debate engineering control versus process control.

257

There is an aspect to that.

DR. COONEY: Art, let me see if I have

heard your point clearly, that is, you are arguing

or you are suggesting that it is very important to

think through how this additional information, how

this different information will, in fact, be used

from a regulatory perspective.

DR. KIBBE: Yes, I think we have to be

careful, because we can narrow the goal posts, that

we don't necessarily do it if it doesn't gain us

anything clinically. We have talked about this

over and over again. The first point I made is

that this is just one more step in a process that

started several years ago and I think it is long

overdue, but we shouldn't focus only on dissolution

as the only thing we are doing because we are doing

lots of other things too.

The second is that in order to get the

companies to come along with us, they can't view it

as an opportunity for the regulatory agency to

nit-pick them. We talked about that before and I

just wanted to make sure it was kind of restated

258

for the record. That is all.

DR. COONEY: That is an important point

and that is why I wanted to get clarity, and it

should be part of the recommendation going forward.

Ken?

DR. MORRIS: Yes, to your scient point,

Art, and I think this may be a little bit out of

sequence, in my sort of image the way things would

ultimately be dissolution in some form--although I

would argue that there may be better forms and that

is a different discussion for the next question, I

guess--that is a develop tool. These are part of

the critical quality attributes that you want to be

able to build in with enough significance to have

faith in it. But then you are really in the

position of formulating the process variables that

are dictated by, hopefully, the first principles to

determine your design space. Once you have the

design space, then you are taking advantage of the

PAT guidance and others to do real-time release,

hopefully, but at least you will be releasing

within a parameter space that says if I control to

259

my endpoint, which should be a manifestation of the

proper attributes, then anywhere in that design

space that it takes me to get to my endpoint is

fine, and that is release. You release the product

based on the attainment of the endpoint as opposed

to a release spec after the fact. The only

criteria are that you are, first of all, within

your design space and that the design space has

been developed so that it does accurately represent

the process.

DR. COONEY: Tom and then Paul and Gerry?

DR. LAYLOFF: Yes, I don't think that we

will see a change in the market standards, the

limits that are there, because they are very

appropriate for among lab assessments within the

statistical bounds that you want to have for a

release. But the company release specifications

would be something more like a control chart I

think where you have the desired in the center and

then the bracketing on the outside. If you start

drifting away from that you start looking at kappa,

corrective actions to move it back down. But if it

260

goes outside those bounds, then you are probably

going to miss the market standards and if you miss

the market standards you have marketed an illegal

product. But I don't think the market standards

are going to change. If you look the among

laboratory test criteria and the expected

statistics, it is a very rational thing to do, to

have limits like that.

DR. COONEY: Paul?

DR. FACKLER: I agree. Of course, I

endorse pursuing examining whether there is a

better way to deal with dissolution. But I still

think fundamentally we need to ask why

dissolution--is it to hold variability on the

manufacturing process or is it to assure the

patient that when he takes the product it is going

to perform the way it is labeled to? We do content

uniformity for exactly that. If the tablet is 10

mg and you find that there are 7-8 mg you wouldn't

want to release that lot.

On the other hand, you do a dissolution

test and you are left scratching your head, saying,

261

well, I know that this dissolution test has nothing

to do with the way it is going to dissolve and be

absorbed by the particular patient, yet you are

forced to scrap the lot. So, really I would

encourage you to keep in mind what the goal of the

release test is above everything else.

DR. HUSSAIN: Point well taken, and please

keep in mind that the decision trees that we are

developing will actually address that aspect. The

key is this, we don't want to go with

one-size-fits-all, like you need a dissolution test

for everything. That is not the intention. The

decision trees will sort of guide us through when

it may be needed; when it may be not needed; and if

it is needed for quality assurance how do you

approach that; and if it is needed for, say,

characterization of a product, say, post-change

like we do in SUPAC, how would we do that? If you

are using it for biowaiver, as Lawrence suggested,

can we think about more biorelevant conditions that

might expand biowaiver decisions for BCS? So,

those would be the considerations.

So, step one and two are not intended for

that particular purpose but it sets up instrument

suitability criteria that makes the system more

262

stable for us to use, and it also provides an

experimental approach to start think of

characterizing the variability and doing more than

just six tablets but characterizing the variability

to start using that in our decision-making. So,

step one and two are just for that and all the

other aspects that you sort of mentioned are for

decision trees where we want to capture that. So,

that is how we want to approach it.

DR. COONEY: Gerry?

MR. MIGLIACCIO: Yes, I want to go back to

Ken's comments. Ken is always a visionary.

Unfortunately, the vision is a ways out.

DR. MORRIS: Yes, six weeks.

[Laughter]

MR. MIGLIACCIO: We might all be in

rocking chairs by that time. But, you know, he has

raised a good point which I think an industry

perspective needs to put out there, and that is,

263

you know, coming in with greater understanding of

product to process to come up with better

performance measurements will work if, when we are

sitting down and discussing these with the agency,

it is not just that plus the layer of the

traditional dissolution test and the traditional

this and the traditional that. If this is just

layering on more and more, then no company will

invest the resources to do it. We will do it

internally. We will have our internal controls

because that process understanding makes us more

efficient, more effective, but this process will

stop--Art has hit it right on the head, there has

to be an understanding that by coming in with this

we have to take a new path and not be mired in

history.

DR. MORRIS: If I can comment just real

quickly, I think the other thing is that, from the

agency side, adding another layer isn't going to

win any friends on the agency side either. Right?

You guys can speak for yourself but I think they

are already stressed to the limit.

DR. HUSSAIN: No, I think we clearly

understand that, and the key is, in a sense, if we

keep holding on to things which may not be adding

264

value and keep adding more the system doesn't work.

So, the decision tree will have to capture a

process which says these are the key questions that

we need to address and this is how we will control

this. If your control strategy meets that, for a

new product you might have a completely different

set of specifications which are non-traditional.

But keep in mind that if you already have approved

products you have those locked in and if you have a

compendial you have those locked in. So, the

future of specifications and controls could be very

different. So, we have to balance the two as we go

along.

DR. COONEY: Nozer?

DR. SINGPURWALLA: Ajaz, I am going to

comment on this slide.

DR. HUSSAIN: Please.

DR. SINGPURWALLA: There is no need to

debate. While this is a fictitious issue raised by

265

some of my statistician colleagues about 10, 15

years ago, what they did is if you have engineering

control then there is no need to do statistical

process control. Engineering control essentially

achieves control for you.

DR. HUSSAIN: Yes.

DR. SINGPURWALLA: This was some kind of

an article written by some of my colleagues, not

necessarily friends, in journals, trying to make a

distinction, and the distinction is completely

fictitious. So, I don't think you should spend too

much time debating this. I think engineering

control has elements of stochasticity in it and

that is about it. So, I would just not bother with

this question any more.

DR. HUSSAIN: I appreciate that because I

expect this to be a debate in the committee, and

the way I think we have constructed the QA

guidelines, especially how to define that, we

actually have approached it from an engineering

perspective. But I do have to put it on the table.

DR. COONEY: Mike?

DR. KORCZYNSKI: As most of you know,

pharmaceutical companies are very conservative in

nature and as long as the dissolution testing

266

appears in the USP, even though the USP may say one

can use alternative methods, I think there is going

to be a major reluctance to abandon dissolution

testing. It was already mentioned that a lot of

the emergent technologies, in fact, are sort of

relegated to segments of the process, not

necessarily the total process. So, somewhere I

would think some type of guidance document could be

available that cites opportunities relative to

dissolution testing, and maybe begin to provide a

bridge and have companies begin to think that, yes,

they can abandon the method if they use these

potential alternatives that are cited.

DR. COONEY: Gerry?

MR. MIGLIACCIO: Yes, speaking for the

innovators, you know, where there is no USP

monograph that we are dealing with, we obviously

can come up with any performance method that is

appropriate that I think many companies and FDA are

267

evaluating. I mean, the PAT guidance gives us the

open door to evaluate these, and there are a number

of them under evaluation now. None of them have

proven out but I don't think it is going to be that

much longer before they are. And then, I don't

know that we need guidance. I think the instrument

manufacturers will sell it for us. Once they

understand that the methodology and the equipment

that they are providing is a surrogate for

dissolution, I think they will be out there

marketing it very aggressively to the entire

industry.

DR. COONEY: Mike?

DR. KORCZYNSKI: Well, I think large

innovative companies all basically have the

wherewithal to develop their own emerging

methodologies and PAT systems, but there are many

smaller companies that would look for guidance

prior to, say, abandoning a USP methodology. You

know, we heard the words that, oh well, one can do

an alternative test but there is the reality of the

field inspector who may interpret that if you are

268

USP labeled you should be following the USP test.

That is why I think some companies we be real

reluctant to abandon that testing.

DR. HUSSAIN: One aspect is, I mean, we

are not asking for abandoning, and so forth. We

want to sort of move towards an appropriate control

or measurement that is relevant for that. I mean,

that is the whole intention. So, I think I have

some hesitation to use the word abandon. That is

not the discussion. I completely understand I

think the concern of the field and I think this is

one of the gaps which we are addressing as we are

moving along.

One of the bullets in Cindy's was training

and I think one of her jobs after this is to put

together a training program for our field

inspectors, and so forth, and Mehul and others I

think are really--I think the CMC and GMP aspect,

especially in how the quality assessment system is

involved in ONDC decisions. ONDC decisions have to

be shared with their colleagues and inspectors.

So, I think we are putting a systems perspective to

269

address all of these concerns so I think we do

understand the concerns there.

DR. COONEY: I think the specific point

being made here is one of being sure that there is

adequate communication of the work product of this

group to the broader community. It doesn't

necessarily need to be a guidance but certainly

adequate communication so that it can be used.

DR. HUSSAIN: One of the reasons for

putting a tactical plan--I mean, we could have

taken our time and brought a proposal to the

advisory committee but we felt that we wanted to

bring the tactical plan to engage and actually have

the committee debate and discuss this. When we do

bring a proposal, I think most of the committee

will at least have been engaged in discussing this

among themselves.

DR. COONEY: So far we have focused on the

tactical plan steps one and two, although we really

have moved into three through five, which are the

decision trees, as well. Without necessarily

curtailing discussion on any of those, I would like

270

to consider step six in your plan, which is general

considerations for identifying and developing

statistical procedures. I wonder if any one on the

committee might have any commentary on this point.

DR. SINGPURWALLA: Yes.

DR. MEYER: While he is formulating his

Bayesian response could I just say your proposed

steps, page two, and six bullets so I presume the

last one was number six.

DR. HUSSAIN: No, those are sort of

specific tactical, and the others are this meeting,

ICH and so forth.

DR. MEYER: Anyway, with respect to page

two, the sixth one, from a committee member

perspective I think that would be an excellent way

to present whatever you choose to change,

side-by-side comparison of new and generic, and why

the old wasn't as good as the new, and how you are

not really layering one thing on top of another but

you are replacing one thing with the other. It

would be extremely helpful I think to present the

information that way.

DR. HUSSAIN: Yes. Sorry, I think my

slide had a different number.

DR. SINGPURWALLA: Getting back to step

271

six--

DR. HUSSAIN: This one?

DR. SINGPURWALLA: Yes, on page eight. It

seems reasonable in the sense that in principle

what has been outlined is very reasonable. We may

have to disagree on details and detail steps. I

think you want to use control charts for variables

and not attributes. That makes sense. You lose

information when you use attribute data. Process

capability analysis, yes. You don't want to test

every lot. You don't want to focus attention on a

piece by piece; you want to look at the broader

picture. Yes, you need tolerance intervals. How

you are going to get them is a different matter.

And, maybe there are other things that you may want

to throw in which don't come to my mind

immediately, but I don't see why this is not a

reasonable step so, again, I would endorse it.

DR. HUSSAIN: Well, I think the challenge

272

is significant. This is the fundamental aspect

because tomorrow you have a proposal to actually do

hypothesis testing on every batch, and we have been

discussing that for the last three and a half

years.

DR. SINGPURWALLA: What is the alternative

to not testing every batch?

DR. HUSSAIN: No, I think as our CMC

reviewers are finalizing the specifications, and so

forth, and then if you need a hypothesis, that is

where it occurs. But then imposing a hypothesis

test on every production batch and then deciding

whether you met the hypothesis completely negates

the systems of GMP philosophy which says this is

validated and under control. So. But the

parametric tolerance interval test is exactly doing

that. So, that is the reason I think this sort of

comes back. Gerry, do you see that point?

DR. MORRIS: I don't know if it matters,

you were talking before, you know, that it is hard

to meet Short's criteria but there are more

advanced SPC or engineering control techniques than

273

those I guess older ones.

DR. HUSSAIN: See, the control chart and

the limits are connected to the hypothesis, and so

forth. They are not disconnected. But then you

are approaching it very definitely, the philosophy

and the system that you have to put in place is a

different system as opposed to hypothesis testing.

DR. MORRIS: That is what I am saying, if

you used a more sophisticated control system or

whatever you want to call it--a filter, doesn't

that suffice?

DR. HUSSAIN: It should. We can discuss

that as we debate that.

DR. SINGPURWALLA: I am presuming that

that is encompassed as general vocabulary.

DR. HUSSAIN: Yes.

DR. SINGPURWALLA: You know, even though

it is called engineering, there is a lot of

probabilistic thinking behind that. So, I endorse

those. If these techniques are becoming

old-fashioned, as my colleague says and he is

right--

DR. MORRIS: Long-standing.

DR. SINGPURWALLA: Yes, long-standing, I

think it is time to look at other things.

274

DR. COONEY: Are there any other comments

on this first question which is on slide two, that

is, the tactical steps which are outlined in slide

three of Ajaz's presentation? I think embodied in

part of that are the questions on additional steps

that you would recommend to improve the plan and

additional scientific evidence. I think we have

already made a number of suggestions to that end.

DR. HUSSAIN: In particular, I just want

to repeat what I think my understanding has been.

One aspect was that I think we have focused on

decision trees and really the science will evolve

from decision trees. In one aspect, one proposal

was from Cindy, the calibration, mechanical

calibration conditions that we are using in our lab

is what we are using and that will be a

recommendation to industry to use the same one. Is

that sufficient or does Cindy need additional

information to make her case for that? You could

275

think about it that way also.

DR. MORRIS: I mean, I think my opinion is

already pretty clear but I just don't see the

calibrator tablets as being value added. So, to

that end, I would say the alternative then is to

have a far more rigorous mechanical calibration, a

real gauge R&R in the sense of the word as you

proposed it.

DR. HUSSAIN: In addition, we are doing a

gauge R&R study for a couple of products that we

are evaluating. So, I think we will bring an

example of a gauge R&R, a real-life example.

DR. BUHSE: I think Ajaz was making the

point--some of you asked what would you do with a

gauge R&R? You know, it is a destructive test.

How would you carry it out? And, I think that is

what we are trying to figure out in our lab by

doing a few of them and kind of seeing what we get

and seeing how many tablets you actually have to

run, etc., etc., to try to get a feel for how much

work it would actually be to do something like

this.

DR. HUSSAIN: Right, and since we are

using a commercial product we hope to--because in a

destructive test, when the sample is destroyed the

276

selection of your reference material, which is your

clinical lot, really has to be very carefully

thought out and planned out and the conditions

under which you judge that they are acceptable.

That goes hand in hand with our approval decision

anyway. But we are using a commercial material to

do this so I think it will raise some aspects and I

think we will learn something about that.

DR. COONEY: Gerry?

MR. MIGLIACCIO: Well, since you ask the

question on calibrated tablets, after the last

industry study, which was in the background

material, the one remaining issue is vibration. Do

you feel in the FDA labs that that is now well

understood and controlled?

DR. BUHSE: We actually measure vibration

and we feel that we have a control over it

definitely, yes. We can measure it without having

to use a calibrator tablet.

MR. MIGLIACCIO: So, I think we all agree

that if we had a reasonable measure of vibration

and could control it that the calibrator tablets

aren't necessary.

DR. COONEY: Art?

DR. KIBBE: Just a quick point, even if

277

you didn't, the one that we are using as a tablet

calibrator we would have to redesign that, and I

think we are far better off with engineering

parameters in the long run.

One point, we never should feel married to

any test. I mean, many, many years ago the test

for quality of digitalis whole leaf was the pigeon

death test where we injected a pigeon with a

macerated solution of whole leaf digitalis until it

died and then we rated the number of units of

digitalis efficacy on that. We clearly don't do

that today. Perhaps the dissolution test as a

quality control test ought to go with the pigeon

death test.

DR. COONEY: Paul?

DR. KIBBE: See if you can follow that!

DR. COONEY: That is a tough one to

follow!

DR. FACKLER: I just want to remind

everyone of the difficult position the generic

industry is in with USP requirements imposed. You

know, part of what we do is try to obtain FDA

approval but we are also bound to USP

specifications. So, I would again encourage FDA to

work as closely with USP as possible to get both

278

organizations aligned such that we are not in a

position where we can satisfy one and not satisfy

the other. The whole industry is really at a loss.

DR. HUSSAIN: Gary, do you want to

respond?

DR. BUCHLER: Thanks, Ajaz. We certainly

can discuss with USP revising the requirements so

they are in line with our new paradigm that we want

to set up for quality. Clearly, we do have a

relationship with USP and I think we know a few

people over there so we should be able to discuss

things with them.

DR. LAYLOFF: A couple of things, first of

279

all, the industry can submit methods to the USP and

submit changes to the USP. That is one option. I

don't think it should be FDA. If FDA is very

concerned that a method is not suitable to protect

the consumer, the public health, then the FDA could

issue a standard and override the USP. They have

done that in the past. So, if the FDA has grave

concern about protecting public health they can go

over, but the industry can submit changes. You

said you have a product that fails the dissolution

but is bioequivalent to the innovator. It means

that the dissolution test is not a good surrogate

and you should submit an alternate one.

DR. FACKLER: And we do that. The problem

is the timing involved in getting a USP monograph

changed, and waiting for that often puts us at

tremendous disadvantage economically.

DR. LAYLOFF: You can also get from the

USP a letter waiving that requirement pending an

option.

DR. MORRIS: Can I ask what is the impact

if calibrator tablets were deemed to be unnecessary

280

by FDA? You would still have to use them? Is that

what you are saying?

DR. HUSSAIN: Well, I think that is a GMP

issue and I think we have other options of a

compliance policy guide. So, there are a number of

mechanisms to think about that. I think the words

we chose were very carefully chosen, an alternate

method. In my description of the vector for

desired state anything we do should conform to the

current standards in a sense. I think we are

trying to achieve a level of quality so that when

tested with minimal standards there should not be

an issue. So, that is the basic premise on which

we have developed the tactical plan. It says an

alternate procedure. So, that is an option too.

DR. COONEY: If there are no further

questions or comments from the committee, what I

would like to do is to consider moving forward with

a recommendation from the committee to the FDA that

the proposed regulatory tactical plan, as described

in slide two, be adopted and moved forward,

recognizing that our responsibility is to make a

281

recommendation to the FDA, it will then be

considered by the FDA as to how it actually is

implemented.

I would also suggest that, in addition to

the tactical plan, as described in slide two--

DR. HUSSAIN: This one? Right?

DR. COONEY: No, that plan and then, in

addition, the proposed steps in slide three are

incorporated in this recommendation.

I have heard from this discussion six

other additional points and I will just read these

quickly, acknowledging that this is one step in

quality by design, not just a focus on dissolution

testing. That one needs to think through how the

information will be used in both manufacturing and

regulation. That it is important to keep in mind

the ultimate goal of the release test, which is the

patient. The implementation plan needs to consider

the impact on the manufacturer and the regulator.

That may be redundant with the point I made above.

To develop a suitable communication strategy for

the work to be done, and to also work through the

282

implications to the generic products with respect

to USP requirements. These are additional factors

that I have heard out of the conversation this

afternoon.

DR. HUSSAIN: I think the next page, page

four, had one more step which Prof. Singpurwalla

really commented on. So, we will include that in

your recommendation. Developing detailed procedure

and harmonization--these are part of our tactical

plan but not necessary for discussion at this

stage. So, if you would include this as part of

your recommendation?

DR. COONEY: So, the request is to embrace

the content of slides two, three and four?

DR. HUSSAIN: Yes.

DR. COONEY: The procedure that we should

follow for a vote--we will go around the table and

we will begin with Art and I will fill in the last

step. You have three options, yes, no or abstain.

DR. KIBBE: Never give me three choices;

it is too hard on me. Yes.

DR. COONEY: Marv?

DR. MEYER: Yes.

DR. SINGPURWALLA: Yes.

DR. COONEY: Carol?

283

DR. GLOFF: I am supposed to vote?

DR. COONEY: Yes.

DR. GLOFF: Yes.

DR. DELUCA: Yes.

DR. MORRIS: Yes.

DR. COONEY: Mike?

DR. KORCZYNSKI: Yes.

DR. SWADENER: Yes.

DR. COONEY: Cynthia?

DR. SELASSIE: Yes.

DR. COONEY: Tom?

DR. LAYLOFF: Yes.

DR. COONEY: My vote is also yes.

Therefore, we have 11 yes, zero no and zero

abstentions for the vote. Wonderful! Thank you

all for your input and your discussion. The next

piece will be by video conference or

teleconference. It is scheduled for 4:30. We are

hopeful that we can get Jurgen on the line before

284

that. But would anybody object if I gave you a

ten-minute break? Ken has to stay here and work, a

10-minute break for everyone else while we set up

the teleconference.

[Brief recess]

Clinical Pharmacology Subcommittee Report

(via teleconference)

DR. COONEY: We are about to begin the

final part of today's program. Jurgen Venitz I

think is on line.

DR. VENITZ: Yes, I am here.

DR. COONEY: I believe that you can hear

everything that is said into a microphone.

DR. VENITZ: That is correct.

DR. COONEY: So, if anyone wishes to speak

please turn your microphone on. I also understand

that we have a set of slides with your name on

them. They are showing up on the screen, and I

would invite you to begin.

DR. VENITZ: Thank you for giving me the

opportunity to discuss on behalf of the Clinical

Pharmacology Subcommittee what transpired at our

285

last meeting.

As you can tell from the first slide, this

committee provides expertise to the parent Advisory

Committee for Pharmaceutical Science on

exposure-response modeling, pediatric clinical

pharmacology and pharmacogenetics.

The next slide shows you the topics which

were discussed at our most recent meeting in

November of last year. We had five topics being

discussed, one being an update on the progress the

subcommittee had made. Then we spent a

considerable amount of time on pharmacogenetic

testing of irinotecan, followed by a discussion on

drug-drug interaction potential assessment; a

tribute to Lewis Sheiner and, on the second day

discussed the role of biomarkers and surrogate

markers.

As you can tell, those slides are more

explicit than my usual summary, the reason being

that... [speaker phone problems;

inaudible]...exposure-response and simulation

guidances where the committee was having an impact

286

were used internally within the OCPB and continue

to have an impact on labeling recommendations. He

also told us that discussion

of...[inaudible]...risk assessment that had been

going on for about two years were on hold, the

reason being there was no consensus internally and

externally as to how to come up with an acceptable

way for setting up ...[inaudible].

On the next slide he also shared with us

that the [inaudible]...decision tree was continued

to be used within FDA and there was a research

project going on within OCPB that was trying to

assess different [inaudible]...in children and how

to assess and revise the pediatric decision tree.

The next slide... [inaudible]...in using

pharmacogenetic testing for TMP, an enzyme involved

in the metabolism of...[inaudible]...labeling

language to include that testing in 2003 and at the

time, meaning as of end of 2004, negotiations

between the FDA and the sponsors were ongoing to

analyze labeling language that would at least

recommend a test for TMPT. [Inaudible].

On the next slide, Dr. Lesko shared with

us that at the end-of-phase 2a meeting initiative

that he undertook late in 2003 was ongoing and that

287

a guidance was being developed and should be coming

out this year.

Lastly he told us that a QT-liability

discussion that has been going

on...[inaudible]...discussions were going on within

the agency to...[inaudible]...clinical study

designs and analyses.

The second part, as I mentioned before was

the role of pharmacogenetic testing for irinotecan.

[Inaudible]...Rahman, Dr. Parodi with respect to

Pfizer, and Dr. Ratan was invited as the consultant

to discuss...[inaudible].

Irinotecan is an oncological agent. It is

currently approved and used for first-and

second-line treatment of colorectal cancer. Like

most cancer drugs, it is limited in its use by

major clinical toxicities, primarily neutropenia

that cases infection and diarrhea. On the other

hand, CPT-11 has...[inaudible]...pharmacokinetics.

288

The drug is just a

prodrug...[Inaudible]...metabolized by SN-38. This

metabolite, SN-38 is further metabolized by an

enzyme called UGT1A1, which forms inactive

glucuronide. The peculiarity about this enzyme,

UGT1A1, is that it is subject to

pharmacokinetic...[inaudible]...has an allele

called 7/7 that is prevalent in...[inaudible]...to

form glucuronide. However, there are other enzymes,

such as CYP3A4 and other transporters such as P-gp

that are involved in irinotecan PK as well. Their

significance clinically is unknown as yet.

What was known at the time of the meeting

was that SN-38 as the active metabolite was

associated with the *28 genotype. So,

patients...[inaudible]...also have higher exposure

of active metabolite, which is what you would

expect. Furthermore, systemic experience to this

metabolite...[inaudible]...namely, neutropenia.

The risk of grade 4 neutropenia was felt to be 9.3

for patients that have 7/7 genotype. So,

patients...[inaudible]...to develop very severe

289

neutropenia. The second limiting toxicity,

diarrhea, was less clearly associated with the

genotype. [Inaudible]...label irinotecan

with certain known risk

factors...[inaudible]...responsible for

neutropenia, in particular, age, prior

abdominal/pelvic radiation, low performance status

and increased bilirubin. So, the discussion or the

vote the committee was asked to take was can we use

UGT1A1 genotype as a risk factor that would then

lead to a relabeling of...[inaudible]. So, we were

presented with a summary of clinical studies that

dealt with that issue. So, across four clinical

trials--the information on the next slide shows you

that the odds ratio for patients that have 7/7 gene

variant--the odds for neutropenia were raised from

2.5 to 16.7. So, obviously those patients were at

significantly increased risk of developing

neutropenia. On the other hand, diarrhea, as I

said, was much less clearly associated with this

genotype.

The limitations for all those studies were

290

the fact that they were not designed to assess the

strength of the genetic association. Nevertheless,

Pfizer presented the committee, on the next slide,

with a table that gives you the estimated

performance battery for this test. The sensitivity

of this test in those four clinical studies would

be 22 percent, however, the specificity would

expected to be 95 percent. This means that you

would expect negative fixed value of 83 percent.

This really means that if you take somebody and

somebody does have this allele there is an 83

percent chance that this individual would not

develop neutropenia. On the other

hand...[inaudible]...would be a patient that has

this neutropenia and does have the genotype. There

is a 50 percent chance of that. This is shown as

an overall incidence of neutropenia of

about...[inaudible].

This was further elaborated on

by...[inaudible] ...and he shared with the

committee that in his estimation without

pharmacogenetic testing 100 percent of the patients

291

were treated and 10 percent of those developed

severe neutropenia. On the other hand, if this

test is implemented, only 90 percent of patients

treated only 5 percent would develop severe

neutropenia. So, the test

actually...[inaudible]...many patients would be

tested in order to protect...[inaudible].

Now, there were some concerns raised

during the discussion about a role and implications

of...[inaudible]. The first was that current

studies were really limited in assessing the

strength of association between the negative

clinical outcome of neutropenia and the UGT1A1 *28

genotype. Pfizer also indicated to us that ongoing

clinical trials may help identify other variants,

other than UGT

...[inaudible]. Currently, there is no validated

algorithm for dosing that would allow us to adjust

the doses after PG testing has been performed.

And, there was some concern that if you reduced

irinotecan...[inaudible].

The committee was asked to vote.

292

[Inaudible]... that this 7/7 genotype is associated

with a higher risk of neutropenia and, as you can

tell, the committee unanimously agreed with that.

The second statement or the second question put to

us was that 7/7 genotype was associated with a

higher risk for acute or delayed diarrhea. The

committee didn't believe that there was sufficient

evidence to support that. The most important vote

was the last where we were asked does *28 PG

testing have adequate sensitivity and specificity,

and the committee voted 9 in favor and 3

abstentions.

During the discussion there were some

comments by the committee to include PG testing in

the label even though...[inaudible]. We noted that

there is lack of information in the current label

about the dosing regimen, which is not unusual. We

were...[inaudible]...that lower doses as a result

of the test may actually allow patients to stay on

drug...[inaudible]...as opposed to discontinuing

treatment because of neutropenia. We realized,

however, that additional clinical testing may be

293

needed to establish and validate a modified dosing

regimen. And there was some discussion about

bilirubin which, as I mentioned before is a UGT1A1

substrate, as a potential safety marker in addition

to the pharmacogenetic testing.

The second topic for discussion on the

first day dealt with the issue about drug

interactions. You can tell

...[inaudible]...starting with Dr. Huang, followed

by Gottesdiener, Lacluyse and Reynolds from OCPB as

well. The issues here were around the update of

the DDI guidance. In particular, we were

asked...[inaudible]...in vitro transfer studies;

how to integrate in vitro enzyme induction studies,

and whether there should be a requirement for

inhibitor/impairment in vivo studies.

[Inaudible]...and also eliminated by the kidney.

Then the question is should you study this drug in

patients who have...[inaudible]...as well as have a

genetic variant of 2D6 enzyme that makes them poor

metabolizers.

You can see in the next slide that we were

294

basically...[inaudible]...and the committee was

asked to vote in favor or against.

[Inaudible]...that has shown in vitro not to have

inhibition of 1A2, 2C9, 2C19 and 2D6 and 3A. Is

there a need for an in vitro study? And, the

committee pretty much unanimously agreed that there

wasn't any need to do an in vivo study. So, in

vitro would predict the in vivo.

The next question, if you have an NME that

is a P-gp inhibitor in vitro should there be a

requirement or should there be a follow-up in vivo

interaction study with a known P-gp substance?

Again by majority the committee was in favor of

doing an in vivo study as a result of in vitro

demonstrated inhibition.

The next question was a little more

complicated. So, here the NME, the new molecular

entity, is a P-gp substrate and a 3A4 substrate,

and the question was is there a need to do an in

vivo interaction study with a drug like ritonavir

which inhibits both P-gp as well as 3A4, and again,

by a smaller margin, the committee voted in favor

295

of that.

The next question related to an NME that

is a P-gp substrate and not a 3A4 substrate.

Should there be an in vivo study specifically to

look at P-gp inhibition and the committee, with a

very slim margin, voted in favor.

Then we moved on to new cytochrome P54

enzyme tests. The question was put to us would the

committee recommend in vivo drug interaction

studies for CYP2B6, 2B8 and for this UGT1A1 enzyme,

and the committee was virtually unanimous in favor.

then we moved along to look at transporter systems,

such as OATP and MRP, and we were asked would we

recommend in vivo drug interaction studies and, as

you can tell, the majority of the committee was

opposed to that, the main reason being that we

didn't think that science would really allow us to

draw any mechanistic conclusions from in vivo drug

interaction studies.

Then we moved to in vitro induction and

its predictive value. We were asked if an in vitro

induction study for a new molecular entity has more

296

than 40 percent ...[inaudible]...positive control

should an in vivo study be done. The committee, by

majority, voted against that.

The next question was if there was in

vitro demonstration of lack of that, there is no

need for any in vivo studies...[inaudible], and the

committee felt evidence exists currently

so...[inaudible]...by majority.

[Inaudible]...should inhibitor in vivo studies be

recommended to actually do that and the committee,

by virtually unanimous vote, voted against it.

Then we moved to the second day of our

meeting. That was started with a tribute to Dr.

Sheiner. That was given by Blaschka who is at

University of California at San Francisco. Lew

Sheiner, as some of you may

know...[inaudible]...worked as a consultant for a

long time. He was known as a seminal researcher

and teacher in the area of PK/PD exposure-response

and pharmacometrics, and a lot of terms and

approaches that he developed still are being used

quite extensively--learn and confirm cycles in drug

297

development; the role of empiricism versus

mechanistic approaches, and the issue of

frequentists versus Bayesian statisticians. He

also developed what is called "Sheiner's rules"

that deal with the certainty of knowledge that is

used or required.

[Inaudible]...the role of biomarkers,

surrogate markers and regulatory decision-making.

These presentations, starting with Dr. Woodcock of

CDER, Dr. Wagner, from Merck on behalf of PhRMA,

and Dr. Blaschke.

Dr. Woodcock led a discussion, and I guess

it was more a...[inaudible]...than it was a

discussion but she shared with us biomarkers

indicate biological processes and/or the

pharmacological responses to therapeutic

intervention. So, anything that changes as a

result of these or the drug is considered to be a

biomarker. On the other hand, clinical endpoints

measure how patients feel, function or survive and

are related to outcomes such as efficacy and/or

this of a drug. [Inaudible]...somewhat in between.

298

So, those can be measured and are intended to

replace clinical endpoints for efficacy and

toxicity.

Dr. Woodcock shared with us in her talk

that the rational use of those biomarkers can,

indeed, accelerate the drug development process and

the internal decision-making

within...[inaudible]...as well as regulatory

decision-making within the FDA. She also pointed

out that biomarkers can provide a mechanistic

bridge between preclinical studies that is

typically...[inaudible]...and the clinical test

which is typically quite empiric. However, in

order for this to really have an impact, a new

business model needs to be developed that allows

biomarkers to be developed in parallel with drug

development as part of a commercial enterprise.

[Inaudible]...and are rarely met in a

strict statistical sense. She pointed out future

clinical endpoints may not be a univariate as we

currently use them but...[inaudible]...and that

biomarkers may help get away from a mean analysis

299

to a responder analysis... [inaudible]...pave the

way for this individualization and personalization

of...[inaudible]...in clinical testing as well as

post-market.

Her presentation was followed by Dr.

Blaschke who did review use of HIV viral load

as...[inaudible]. He pointed out the validation of

those surrogate markers required sensitive assays

to detect...[inaudible]...as well as mechanistic

models about disease progression, both qualitative

as well as intellectual.

He pointed out that biomarkers need to be

causal path. That means they have to be

mechanistically... [inaudible] and proximal to the

disease endpoint in order to provide confirmatory

evidence to support the efficacy of new molecular

entities.

The last presentation was Dr. Wagner's

presentation which, as I mentioned before, spoke on

behalf of the work group that PhRMA put together do

deal with this issue of biomarkers and surrogate

markers. He reviewed what is called

300

"fit-for-purpose" qualification of biomarkers,

meaning the extent of qualification of a biomarkers

depends on its intended use. He reviewed four uses

for it. The first one would be the exploratory

use. So, here the biomarker is used as a research

tool and, obviously, the requirements in terms of

qualification are relatively minor.

The next level of qualification would be a

demonstration of purposes So, those would be

biomarkers that are likely or emerging as useful

biomarkers.

The next level would be characterization.

So, here you are using known or established

biomarkers to assess exposure-response

relationships and mechanism of action, for example.

The highest level of qualification would

obviously be surrogacy so here the biomarkers would

become a surrogate marker and would substitute for

clinical endpoints. That would obviously require

the highest level of qualification.

Dr. Wagner did lament the lack of

nomenclature in the biomarker area. He spoke of

301

the need for collaboration between PhARMA, FDA, NIH

and academia on two issues, number on, what are the

most useful biomarkers to pursue that might, in the

long run, provide a payoff and, more importantly,

how to decide what evidence can be used and how it

can be used to accept biomarkers and surrogate

markers as part of a regulatory...[inaudible], and

he also mentioned some of the hurdles associated

with extensive use of biomarkers such as

incentives, intellectual property rights and

funding.

[Inaudible]...that the discussion of

markers would continue in one of the next meetings.

And, that is all I have to report.

Questions by Committee members

DR. COONEY: Thank you very much. We have

an opportunity for questions from the committee.

Ajaz?

DR. HUSSAIN: Jurgen, thanks for reporting

the subcommittee report. I think Prof. Nozer left

before you mentioned the Bayesian aspect so you

missed that.

One question I had was I think in the work

we are doing also from a risk-based perspective

utility functions really have to be discussed and

302

we need to start thinking about that. What are the

challenges that you are facing that has started

that discussion?

DR. VENITZ: Well...[inaudible]...there

was no consensus on not only how to come up with a

uniform rating scale but also a process to follow

because you have to be involve all

stakeholders--patients, clinicians, sponsors,

regulatory individuals--in designing a process

before you can get utility...[inaudible]...results.

So, that is the reason why he told us that for the

time being that initiative has been put on hold. I

didn't get the sense that it was fatally wounded as

much as they were trying to reassess what to do

internally.

DR. HUSSAIN: One more comment that I

have, Jurgen, usually you are here but I think one

opportunity you have is if you really looked at the

PAT guidance and what you are doing in clinical

303

pharmacology has a lot of commonality, and the

biomarkers, surrogacy and so forth I think comes to

the same level of discussion in terms of

fundamentals to alternate testing and control

strategies that we are developing. So, keep that

in mind, and maybe have some discussion on the

commonality and how we might approach things that

might be useful at some future point.

DR. VENITZ: I would agree with that and I

would add, as was mentioned in the tribute that was

given to Dr. Sheiner, his approach, which I think

is very pertinent to the PAT approach as well, is

how much weight do you give empiric evidence by

testing and testing over again, and how much do you

give to mechanistic understanding whether it is a

manufacturing process or whether it is the

pathogenicity of disease. That is really what the

dilemma is. As I mentioned before, the

frequentists would just do things over and over

again if you are willing to make any mechanistic

assumptions. Of, if you are more on the

mechanistic side you might be willing to make

304

certain assumptions without having to empirically

repeat experiments and just base your confidence on

the reproducibility of the test.

DR. HUSSAIN: Thank you.

DR. COONEY: Thank you. Jurgen, the good

news and the bad news at the end of the day is that

it is the end of the day. Cynthia?

DR. SELASSIE: Jurgen, I have a question

for you. Do increased levels of P-gp result in

greater levels of neutropenia? Does it impact it

in any way?

DR VENITZ: As far as I know, there are no

clinical studies. All we know is that G-pg seems

to be involved in the kinetics of irinotecan.

Whether... [inaudible] is a very difficult question

at this stage to answer. The only thing we know is

that UGT1A1 has been shown to be associated with

neutropenia. None of those tests have proven to be

clinically relevant...[inaudible]... any

association between neutropenia levels of

irinotecan and those genetic...[inaudible]. So,

right now the answer is we do not know.

DR. COONEY: Thank you. I think we have

reached the end of our discussion, looking at the

way the people are sitting around the room. Thank

305

you very much for the summary and the update.

DR. VENITZ: Good luck. I am sorry that I

couldn't be there.

DR. COONEY: Thank you. We will look

forward to you joining us next time.

DR. VENITZ: Okay. Bye, bye.

DR. COONEY: I think we can adjourn the

meeting unless anyone else has anything pressing.

Everyone looks like they are ready to adjourn!

Thank you all very much. I look forward to seeing

you tomorrow morning.

[Whereupon, the proceedings were adjourned

at 4:42 p.m., the reconvene at 8:30 a.m.,

Wednesday, May 4, 2005.]

- - -