UNITED STATES OF AMERICA
FOOD AND DRUG ADMINISTRATION
+ + + + +
CENTER FOR DRUG EVALUATION AND RESEARCH
+ + + + +
ADVISORY COMMITTEE FOR PHARMACEUTICAL SCIENCE
+ + + + +
MANUFACTURING SUBCOMMITTEE
+ + + + +
OPEN PUBLIC HEARING
+ + + + +
TUESDAY,
JULY 20, 2004
+ + + + +
The above entitled Meeting was conducted at 8:30 a.m., in the CDER Advisory Committee Conference Room, 5630 Fishers Lane, Rockville, Maryland, Dr. Judy P. Boehlert, Subcommittee Chair, presiding.
PANEL MEMBERS PRESENT:
JUDY P. BOEHLERT, Ph.D., Chair, Manufacturing
Subcommittee
HILDA F. SCHAREN, M.S., Executive Secretary,
Advisors and Consultants Staff, CDER, FDA
PATRICK P. DeLUCA, Ph.D., Professor, Faculty of
Pharmaceutical Science, University of Kentucky
DANIEL GOLD, Ph.D., D.H. Gold Associates
DAVID HOROWITZ, Esq., Director, Office of
Compliance, CDER, FDA
AJAZ HUSSAIN, Ph.D., Deputy Director, Office of
Pharmaceutical Science, CDER, FDA
PANEL MEMBERS PRESENT:
KENNETH M. MORRIS, Ph.D., Department of Industrial
and Physical Pharmacy, School of Pharmacy,
Purdue University
GARNET PECK, Ph.D., Industrial and Physical
Pharmacy, Purdue University
JOSEPH PHILLIPS, Regulatory Affairs Advisor,
International Society of Pharmaceutical
Engineers
G.K. RAJU, Ph.D., Executive Director, MIT/PHARMI,
MIT Program on the Pharmaceutical Industry,
Massachusetts Institute of Technology
NOZER SINGPURWALLA, Ph.D., Director, Institute for
Reliability and Risk Analysis, Professor of
Statistics, George Washington University
HELEN WINKLE, Director, Office of Pharmaceutical
Science, CDER, FDA
ALSO PRESENT:
JOHN BERRIDGE, Ph.D., Vice President, Pharmaceutical
Sciences, Pfizer, Ltd.
GARY BUEHLER, R.Ph., Director, Office of Generic
Drugs, OPS, CDER, FDA
PAUL FACKLER, Ph.D., Senior Director, Product and
Biopharmaceutics Strategy Development, Global
Generic Research and Development, Teva
Pharmaceuticals
DONALD MARLOWE, FDA Standards Coordinator, Office of
Science and Health Coordination, Office of the
Commissioner, FDA
TOBIAS MASSA, Ph.D., Executive Director, Global
Regulatory Affairs, Operations/Chemistry,
Manufacturing and Controls, Eli Lilly & Co.
MOHEB NASR, Ph.D., Director, Office of New Drug
Chemistry, OPS, CDER, FDA
FREDERICK RAZZAGHI, Director of Technical Affairs,
Consumer Healthcare Products Association
C-O-N-T-E-N-T-S
AGENDA ITEM PAGE
Call to Order...................................... 3
Conflict of Interest Statement..................... 4
Introduction to Meeting............................ 7
Topic Updates:
ICH Q8...................................... 26
ICH Q9...................................... 49
Life
Cycle Management for Process and
System
Control: An Industry Proposal........ 66
ASTM
E55 Committee: Pharmaceutical
Applications
of Process Analytical
Technology.................................. 83
Introduction to the Bayesian Approach............. 92
Research and Training Needs: The
Industrialization
Dimension
of the Critical
Path
Initiative............................ 130
P-R-O-C-E-E-D-I-N-G-S
(8:32
a.m.)
CHAIR
BOEHLERT: Good morning. It's 8:30.
I call this meeting to order and welcome members of the committee and
all other participants that are going to be presenting in this two-day
session. We have an interesting program,
updates on a lot of topics we've addressed in the past. One that I'm particularly interested in is
finally hearing, you know, some discussion on Bayesian statistics. We've touched on it many times in our
discussions, so Nozer, I'm looking forward to that. You finally get your chance.
DR.
SINGPURWALLA: You'll be tested.
CHAIR
BOEHLERT: That's what I was afraid
of. It's not a topic that's in my area
of expertise but I expect to learn a lot today.
With that, I'd like to turn the meeting over to Hilda for the conflict
of interest statement.
MS.
SCHAREN: The following announcement
addresses the issue of conflict of interest with respect to this meeting and is
made a part of the record to preclude even the appearance of such at this
meeting. Based on the agenda, it has
been determined that the topics of today's meetings are issues of broad
applicability and there are no products being approved at this meeting. Unlike issues before a committee in which a
particular product it discussed, issues of broader applicability involve many
industrial sponsors and academic institutions.
All special government employees have been screened for their financial
interest as they may apply to the general topics at hand. To determine if any conflict of interest
existed, the agency has reviewed the agenda and all relevant financial
interests have been reported by the meeting participants.
The
Food and Drug Administration has granted general matters waivers to the special
government employees participating in this meeting who meet prior waiver under
Title 18 United States Code, Section 208.
A copy of the waiver statements may be obtained by submitting a written
request to the Agency's Freedom of Information Office, Room 12-A-30 of the
Parklawn Building. Because general
topics impact so many entities, it is not prudent to recite all potential
conflicts of interest as they apply to each member and consultant and guest
speaker.
FDA
acknowledges that there may be potential conflicts of interest but because of
the general nature of the discussion before the meeting, these potential
conflicts are mitigated. With respect to
FDA's invited industry representatives, we would like to disclose that Gerald
Migliaccio is participating in this meeting as an industry representative
acting on behalf of regulated industry.
Mr. Migliaccio is employed by Pfizer.
Dr.
Paul Fackler is participating in this meeting as an acting industry
representative. Dr. Fackler is employed
by Teva Pharmaceuticals. In the event
that the discussions involve any other products or firms not already on the
agenda for which FDA participants have a financial interest, the participant's
involvement and their exclusion will be noted for the record. With respect to all other participants, we
ask 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.
CHAIR
BOEHLERT: Thank you, Hilda. To get the meeting started, Ajaz -- if I turn
on the mike and you can actually hear me.
To get the meeting started, Ajaz will provide an introduction.
DR.
HUSSAIN: Good morning, and welcome to
Rockville. The Manufacturing
Subcommittee for the Advisory Committee for Pharmaceutical Science, I think
this is the third meeting after the key subcommittee ended and we have
discussed many of the developments with this committee and we'd like to sort of
use this meeting to bring forward the concepts that have been developed and the
challenges that we are overcoming in trying to implement some of the concepts
and seek your input in a number of questions that have been posed to you.
Just
to recapitulate, at the Advisory Committee of Pharmaceutical Science in 2001
July, we had used the CGNP initiative and that was the starting point for
discussion on manufacturing in a very focused manner that led to the CGNP
initiative for the 21st Century and later on we have two other initiatives
defined, one on molecular innovation and one on critical path. In some ways I look at all these initiatives
as a desire to define a desired state which more efficient, which is more
effective in meeting the needs of the customer, that's a patient and so forth. So the desired state that FDA is trying to
articulate in a shared manner for the US patient is in many ways very forward
and I'll focus many on manufacturing with regards to manufacturing and utilize
the six dimensions of our pharmaceutical quality for the 21st Century
Initiative as a means to share with you how this meeting agenda was organized.
Although
we called our initiative CGNP for the 21st Century, we realized that is was
probably a mistake to just call it the CGNP initiative, because it is an
initiative which is dealing with all aspects of pharmaceutical quality. It applies to CMC Review Process as well as
the CGNP inspection. So often we refer
to that as the Pharmaceutical Quality for the 21st Century initiative instead
of just CGNP Initiative.
The
six dimensions for this initiative are foremost, strong public health
protection. We want to maintain that and
strengthen that function of FDA. We want
to bring an integrated quality systems orientation to our activities and our
programs that could simply mean better communication between different
organizations within the agencies, the industry and so forth but also a more
systematic approach to pharmaceutical quality and more integration and
collaboration between different parts of the organizations that deal in
pharmaceutical quality.
Science
based policies and standards, risk based orientation and international
cooperation. Those are the five pillars
of this initiative. The sixth dimension
is time and the time we decided was for two years. We will work on this initiative trying to
define the desired state, trying to define the issues to be addressed in the
two years. The two years time comes to
an end next month, but that doesn't mean the initiative ends. It means that you would now move into a
regular routine of trying to implement all these activities.
And
in September we hope to announce how this process will become a more permanent
model within the agency. So the
initiative was for two years to define the issues to be addressed and identify
issues to be addressed and come up with a way to address those but that doesn't
mean that we will have completed all the objectives.
If
you look at what we have been engaged in, I call those directional vectors, we
would like to insure regular review and inspection policies based on state of
the art pharmaceutical science and create new technological advances and create
risk based approaches that focus both industry and agency attention on critical
areas, facilitate modern quality management techniques, including
implementation of quality systems from within the agency as well as outside the
agency and industry, and have the consistency and coordination of FDA's quality
review programs, in part by integrating enhanced systems approaches into the
agency's business processes and regulatory policies concerning review and
inspection activities.
If
you look at how we are covering these topics, we can visualize this as a
three-dimensional aspect, science, risk and system integration concepts, we
started with the PAT initiative. We have
a draft guidance. That guidance will be
finalized in the next month or so. We
took some of these concepts to ICH and now we have a number of topics in ICH
and that will be a subject for discussion this morning. We wanted to move to a more flexible approach
to post-approval changes and move it from change being bad to change being
viewed as an improvement, and we struggled with delegating a compatibility
protocol that would be user friendly, useful in many ways.
And
we're still struggling with that and I think that tomorrow you'll hear some
aspects of the struggle with that protocol.
Aseptic processing I think is an important guidance that will be
finalized soon. Guidance on CFR Part 11,
probably this is one of the major accomplishments of this initiative is to
address some of the challenges of Part 11, better integration to collaboration
and cooperation between inspection and review staff, products specialists and
inspection, the PAT model is evolving and this is working nicely and we're
trying to expand that beyond the PAT model.
Pharmaceutical
inspectorate is another major accomplishment.
Over the next several years we will have a core group of pharmaceutical
inspectorate staff, in ORA who will spend most of their time or 80 percent of
their time inspecting pharmaceutical plants and they will have a high level of
training and certification to accomplish that.
Dispute
resolution process is also a major aspect of this because in a large system
such as our regulatory system, when you start moving towards a different
approach for dealing with regulatory aspects, you have to have an efficient
dispute resolution process. And clearly
pre-approval inspection compliance program was one of those but you will hear
tomorrow from David Horowitz and Larry and others a risk based approach to
inspection, site selection, where do we inspect, where do we put our resources
where the risks are and so forth. So
these are some of the activities that sort of cover risk science and system
integration approaches that we outlined for us.
But
quickly, I'd like to summarize why we felt the time was right to move forward
here. There was the scientific
opportunity. And this was a sensitive
document to sort of bring up and simply stated that pharmaceutical development
and manufacturing is evolving from an art form that is now based on science and
engineering based. Effectively using
this model in regulatory decisions when we establish specifications and we
evaluate manufacturing processes can substantially improve that efficiency of
manufacturing regular processes. That was the initial hypothesis that we
started as a basis in 2001 and hopefully you'll see that some of the activities
that will be discussed at this meeting we can move forward and put a conceptual
framework around them.
The
other dimension was the risk and the risk mitigation and communication
opportunity was clearly an opportunity because there are many risk approaches,
risk mitigation approaches which have matured, have been utilized within the
agency and outside the agency. For
example, within the agency, on the food side there is more effective analysis
on the devices side have been utilized for a number of years and other
industries have utilized some of these.
And we sort of brought the concept up and done some designs by quality,
by design, again a phrase which is a very old phrase but brought a dimension to
this to focus on reliability and risk mitigation and hopefully we can
communicate this better, we can find leverages for reducing regulatory but the
third dimension of opportunity was the quality systems opportunity. Again, if you look at the evolution of
quality, you start with sampling plans, and so forth and GNPs came in there and
many of the quality systems are based on other GNPs and what we are hoping to
do is to sort of in a jargon free way, adopt the practices in all of these
quality systems into our system and we are moving towards a general quality
system framework for the agency and hopefully support that for external
industry also.
So
for the two-year journey, which is coming to an end next month, from the
perspective of defining the issues and defining the training and conceptual
framework, to what is the destination. I
often use this slide, the book by John Guaspari, "I know it when I see
it", is to me an excellent reflection of the current state. I often say the person in that picture is our
CMC reviewer because they often do not have information that they need to make
the decisions with respect to risk and so forth. So often the answer is, if you want to change
the site of manufacture, I need three batches of separate data. The only decision they can make is when they
see the three batches of separate data.
So we can move away from that to a vision 20/20, I can see clearly now,
which is part of the desired state. And
we define the desired state as follows.
The
part quality and performance is achieved by design of effective and efficient
manufacturing processes. Correct
specifications based on mechanistic understanding of how formulation and
process practicing factor on performance, again, that's missing from the
current state. We don't have this
information in the submissions. And move
towards a continuous assurance of quality.
The primary motivation for the third bullet was you achieve that only if
you gain a high level of process understanding.
You cannot achieve that without that and when you achieve that, that
brings a more efficiency continuous manufacturing and so forth.
But
to facilitate that, our policies, that is regulatory policies need to be tailored
to recognize the level of scientific knowledge supporting applications, process
qualification and process capability, and we started emphasizing the process
capability because product are validated but many are not capable, so there is
a missing element. Validation does not
insure capability but shows a missing link.
So risk base review relates to the level of scientific understanding of
how formulation and manufacturing process effect product quality and
performance and the capability of process strategies to prevent unmitigated
risk of producing a poor quality product.
So that was our way of saying, we can facilitate moving toward a desired
state by providing regulatory incentives.
So
this meeting -- the primary objective of this meeting is to seek input and
advice from you and from the public on charting the most efficient part of the
desired state and the discussion focuses on review assessment of chemistry,
manufacturing and control sections of submissions and I deliberately I sort of
wrote the CMC as its written in our regulation, chemistry, manufacturing, and
controls. The reason for stating it that
way, that's as it's written in our CFR, is we often focus only on the
chemistry, the manufacturing controls part is -- doesn't get the attention it
deserves. And that is an opportunity, I
think, that Q8 and Q9 sort of bring forward.
Risk-based
procedure inspections, you will hear a pilot program on selection of
manufacturing site inspections. There
are elements of risk which says if a process is well, well controlled, there is
a way to reduce the risk for those sites and so forth. So you'll hear the discussion tomorrow. You will also hear updates on a number of
topics but I just wanted sort of put up the Q8.
What do we wish to accomplish with Q8?
As
an example, we hope that Q8 will facilitate movement towards the desired state
that we have articulated. We believe
this is important because this will help us better understand the proposed
product and process design and its relation to independent review. Improved process of establishing regulatory
specifications, this is the heart of the key here. This is the voice of the customer. FDA is the customer defining the voice,
making sure the quality is there because the GNPs then have posted so if you
don't get the specifications right, the problems linger on.
And
four, we could identify and understand critical product and process practice,
again, this is not well understood today in the part of the type of information
we've seen in the submissions. Allow us
to do a risk-based approaches and recognize good science and facilitate
improvement, improve communication and system thinking and be a advocate for
public health, regular and industry.
I'll
skip this. John Berridge graciously
agreed to come and talk to you about how we are approaching Q8, but there is a
question that we have posed to you and this is the reason I'm showing this
slide. One of the concept that has
evolved in a harmonized way to move forward is the concept of continuous improvement
and the concept for design space. And
this is a part of the question that I think, we have asked you to address. The key factor here or key concept here is
that if you have understood the critical formulation basis, the critical
process basis and you have charted your design space, within that design space
movement is not a change any more and I think that's an important point.
So
how do we define this design space is a key element. It is a multi-dimensional space that will be
defined by critical vector of product in performance. One of the examples of
such critical vectors, vectors that define robust manufacturing processes,
consistent ability of meeting its
specifications, different manufacturing options. Here is a graphical presentation of what this
design space might be. Currently, much
of this is a black box, especially with respect to raw material properties,
processing conditions, and so forth. So
we have very limited information about what are the critical factors and so
forth.
We
hope the future will be sharing of pharmaceutical knowledge and that shows us
where things are critical and not critical and therefore, be more rational in
moving through different things and improving model. We are very confident that many companies
already have this information. We have
met with several companies. They've come
and met with us, shared with us this information and we believe it's already
there. So for many companies this is not
any additional work. It is simply
sharing this information at the right time, in the right way, and for us to
move forward.
So
in many ways you'll see the discussion.
We're hoping Q8 brings a level of understanding which is not there today
and the key aspect here is the company has its own quality system. We have post-approval changes and the current
system says changes is bad because it's uncertain or risky. We don't know what that change might be, that
if fact, if that is true, additional testing.
Yet you have the CMC regulatory oversight, you have the CGNP regulatory
oversight, you have a perceived or real risk out there and all of our
activities are focused addressing all of this and aligning this in such a way
that we move forward to serve the patients in a more efficient manner.
The
process of understanding, you align all of this together, you have an
opportunity that would say post-approval changes is not bad, it's actually
good, it's a continuous improvement and that leads to significant risk
reduction on a continuous basis. So
that's what you're going to accomplish and I see it Q8, Q9 and the proposed Q10
are graphically in my mind, going together this way and you will hear
presentations on this. So the meeting
today in many ways, I look at that as moving towards a desired state. We seek your input on how best to do
this. Day one, you will hear updates on
our current efforts on ICH Q8, Q9 and the proposed Q10. We are moving forward with ASTM in a
significant way, especially for the PAT standards and in aspects, which I think
ASTM is a paradigm shift because we believe that to do it right, you have to
bring this standards and the umbrella of understanding.
For
example, if you have a chapter in the USP that lacks the process understanding
dimension, so that's a live base mentality to those standards. Those standards are not yet useful as when
they are within the framework of understanding.
So the ASTM model is a basis for moving forward in that direction. I think the
awareness topic that we are introducing today is to fill some of the gaps
that we think exist even in spite of all the work that we have done and some
research planning. Bayesian approaches
in chemistry manufacturing control, I think is a significant topic. It's a
topic that has been discussed recently at an FDA John Hopkins University
co-sponsored workshop but more on the clinical side. We would like to start the discussion on how
we can bring some of these concepts to bear on CMC decisions. Professor Nozer is an expert on reliability,
Bayesian approaches and so forth, so we requested him to give us a talk. This is simply an awareness topic at this
point but I think you want to build on this.
There a critical path initiative that we talked to you about but then we
move on to some more significant discussion.
It's how do we start moving toward implementing the concepts that we
have developed in the Office of New Drug Chemistry and Office of Generic Drugs.
Moheb
Nasr and Gary Buehler can share some of the parts. Our focus mainly has been on Office of New
Drug Chemistry right now to bring all of these concepts to bear over time of
the Office of Generic Drugs and all of these offices we have will come together
in this. But it's going to take some
time. But to set the stage for this
discussion, what we have is two introductory lectures or presentations. One on the manufacturing science and
knowledge. G.K. Raju will do that, and
to share some thoughts on quality of design and setting specifications. Then we'll have Moheb Nasr and Gary Buehler
share some of their thoughts and then we have invited Ken Morris. Ken Morris has been working with our CMC
leadership at the agency in moving towards the question of the CMC review
process and we asked him to share some of his thoughts with you after Moheb and
Gary have shared their thoughts.
Day
two will focus on risk based CGNP inspections.
You will hear about the study being conducted on industrial
practices. Then we'll have a significant
discussion on pilot model or selection of manufacturing sites for inspection,
how do we identify the risk factors but also I think an important topic which
is -- will be a substantial topic, Joe and Moheb will talk about this, CGNPs
production IMBs. I think this is going
to be a significant topic but mostly a wellness topic. Then I think we'll sort of wrap up this
discussion is trying to sort of identify some of the challenges that remain and
some of the things that are working well as well as some fascinating continuous
improvements and reduction in the need for product food supplements. They use the PAT as an example of how we are
bringing the review and inspection people together, the staff together, to make
decisions without having to have supplements, compare the quality topic that we
discussed and John will also discuss some of his parts on this but they're
challenges because some of the concepts are within the old system and some of
the concepts are happening with the new system.
So
with that, sort of that's a broad discussion on the meeting. What we have tried to do is to share with you
or ask you some questions. For example,
you agree that current activities within ICH and the ASTM has been to move
toward the desired state. We also seek your recommendation on how to insure
these activities are synergistic and simply on risk basis recommendation in the
new paradigm. We have a number of
questions. The flexibility for you to
address your discussion around these question will help. I'm not sure whether the committee would like
to come together to sort of address this in brief summary if they could but the
topics that we have for day one, the questions apply to all the topics, so
towards the end if you could summarize some of the parts, that would be very
useful for us. And you already have this,
I won't spend more time on that, with that I'll give it back to you.
CHAIR
BOEHLERT: Thank you, Ajaz. Is there comments or questions from committee
members?
Okay, before we go on with the next speaker,
there's one thing I neglected this morning and that's to have our committee
members introduce themselves. I think
this is important for the benefit of committee members who may be new. So we'll start with Dr. Fackler, introduce
yourself and your affiliation, please.
DR.
FACKLER: Paul Fackler with Teva
Pharmaceuticals, representing the generic drug industry.
MR.
MIGLIACCIO: Gerry Migliaccio, with
Pfizer representing PhRMA.
DR.
SINGPURWALLA: Nozer Singpurwalla, George
Washington University.
MR.
PHILLIPS: Joe Phillips, Regulatory
Affairs Advisor, International Society of Pharmaceutical Engineers.
DR.
RAJU: G.K. Raju, MIT Pharmaceutical
Manufacturing Industry.
DR.
DeLUCA: Pat DeLuca, University of
Kentucky.
DR.
MORRIS: Ken Morris, Purdue University.
CHAIR
BOEHLERT: Judy Boehlert, Consultant to
the Pharmaceutical Industry.
MS.
SCHAREN: Hilda Scharen, FDA.
DR.
PECK: Garnet Peck, Purdue University.
DR.
GOLD: I'm Dan Gold, D.H. Gold
Associates.
DR.
HUSSAIN: Ajaz Hussain, Deputy Director
Office of Pharmaceutical Science, CDER.
MS.
WINKLE: Helen Winkle, Director, Office
Pharmaceutical Science, CDER.
CHAIR
BOEHLERT: Okay, thank you,
everyone. Our next speaker is going to
discuss ICH Q8. Ajax has introduced us
to these topics, and John Berridge, Dr. John Berridge, will make the
presentation.
DR.
BERRIDGE: Thank you, Judy. Good morning, ladies and gentlemen and thank
you for the opportunity to present to you today on the topic of Q8,
Pharmaceutical Development on behalf of the expert working group and some
additional thoughts, of course. What I
would like to do today is to give you a little bit of background to the topic
and address the opportunity for change.
Look too, at the progress that we've made so far and then to round off
by considering some of the implications, implications for the future as a consequence of this guideline.
So
at the very highest level, the purpose of the ICH Q8 topic is simply to provide
guidance, that is harmonized guidance, on the section P.2 which is entitled
"Policy for Development" of the comment technical document format. And its scope is very clearly outlined in the
concept paper and it is all the products that are pertinent to the CTD. Of course the CTD is not mandated in the US
but I think it's true to say that the majority of applications for new entities
are actually using the CTD format today.
So
that's the very highest purpose but I think it's pertinent to actually look
deeper, look underneath and to say what are actually the drivers, why would we
really want to do this? So if we go back
and think about life as it is now, life before we actually get to the Q8
state. In the United States the amount
of information that industry submits in its NDAs is variable partly because
some of that information may have been submitted through the IND process. Some companies go to a different extreme and
actually submit the report that they would present in Europe.
Others,
the information is distributed around the new drug application in various
places, but even so, there is variable information that is presented and part
of is driven by industry concerns. If we
provide a lot of information, we get a lot of questions. So there is sometimes reluctance to provide
information that would give a full understanding. It's slightly different in Europe where there
is the -- there has been traditionally and still is the development of pharmaceutics concept which describes how formulations
are designed and the manufacturing process is put together all in one sanction
and the home for that in the CTD is P.2.
Japan,
there are very limited expectations. So
we can see that there is a varying degree of expectation and a varying degree
of implementation around the world. Is
there anything wrong with that? Well, I
think there is because right now there's a lot of focus in an NDA on the future
regulatory commitments, a reluctance to describe how the product was truly
designed. When you put those things
together with the worry about the future and the regulatory commitments, it
creates what are called a "check-list" mentality. E go around providing and reviewing
submissions in a ticking the box process.
Where a development report is written it tends to focus on successful
preapproval inspection.
And
the other major driver, I think, is we've heard in the earlier presentation
from Ajaz a desire for international cooperation. So we have disharmony. There is a P.2 section in the CTD but we
don't have any guidance on exactly what we would put there. When we look at the regional implications
where development of pharmaceuticals is the cornerstone of the European
submission, I think there is some missed opportunities. And this all tends to result in the very
limited regulator incentive to truly understand how products and processes to
describe that understanding and then to move into a process of their
optimization.
Q8
brings with it an opportunity for a significant change, a change that moves us
from simply providing huge amounts of data and what happens when you get huge
amounts of data? It tends to get checked
and boxes get ticked or they don't get ticked because there's a mistake. So let's move from that, move from these huge
boxes of data to a situation of information and knowledge. And we can express that in a different way
which is basically a manufacturing sciences based approach to submission and
approval.
And
if we agree to that, then we see the creation of a significant new
paradigm. It's a new paradigm for both
parties. It's a new paradigm for
industry and a new paradigm for the regulators and a significant set of
positive opportunities. You've seen
this slide almost. The first two points
are as age asset. Some people discuss
the word "mechanistic". We
could substitute the word "scientific", but we're certainly talking
about a true understanding of our products and processes. And we're trying, through Q8, with the full
support of the expert working group, to get to that state which allows us to
effect continuous improvement and opens the door to continuous real time
quality assurance.
So
if we look at the guideline itself and the mechanics, the processes underlying
the development of the guideline, the topic was actually adopted back in
October 2003 and the expert working group has met three times since then, have
produced at the meeting in Washington just a few weeks ago, a third version of
the guideline. This is under
consideration by the experts themselves with input from their various
associations, but we're aiming to get the document out for public consultation,
this is ICH Spec 2, after our November 2004 expert working group meeting in
Yokohama, in Japan.
I
think we're cautiously optimistic that that timeline will be met. Q8 itself is a guideline that's being
conceived in two parts. Part 1, the core
document, describes baseline expectations and optional information. I'll come onto this a little bit more in a
moment but describes a concept of regulatory flexibility. Again, I will discuss that a little more in
just a moment. And as I've indicated, we
hope to get to Step 2 later this year.
The
second part, which has not been started yet and which is still subject to
discussion, relates to annexes of
specific dosage forms and the possibility to include in it appropriate examples
of risk management. And in that sense,
the Q9 guideline that provides a toolbox of risk management examples, provides
useful input into the QA guideline. I
think we can stick to our intended time line.
Then we should be able to start back in November of this year.
So
I've talked about baseline, other expectations, and it is clear in the
guideline that not all the information is mandatory. But the guideline is carefully constructed to
insure that this doesn't create any misunderstanding. What it does is describe one system with
different levels of focus. And there's a
complex phrase here "process understanding and predictive ability"
that actually is intended to describe this continuum, not a two state system
but a continuum. What we mean is that
the more that the process is described and understood, the more one provides
for the future regulatory flexibility.
The
less you give, the more rigid the subsequent approach is. And so it doesn't actually describe a
mandated content, it describes an opportunity.
So if we look at that in the context of quality by design, which has
also been mentioned as a concept today, we're looking at on the left-hand side,
understanding that we have a well-characterized product. We understand the process. We've looked at the risk, and taken
appropriate mitigating actions and we understand how we're going to monitor our
process in the future.
If
we put those four components together, it drives the framework for continuous
improvement. In fact, you can put
together the sum, if you like, the product and process knowledge together with
appropriate risk management and that can comprise the manufacturing
sciences. Well, if we drive towards that
framework of continuous improvement with the knowledge as indicated, on top of
that, so I should say that the first three are part of the Q8 topic, so the
first three are critical elements of the ICH Q8 topic. If we put those together, then we can build
on top of that this concept of regulatory flexibility. So Q8 is really a major engine driving
towards the opportunity for regulatory flexibility. If you look at this in the context of the
variable space, you can take a couple of hypothetical vectors and Ajaz earlier
talked about what some of those might be.
Traditionally, industry has focused on a very narrow understanding or at
least described a narrow understanding, even if it knew more, intended to do
that three batch validation and any move away from this situation created a
post-approval change.
What
we're saying now is if we consider the overall boundary and we have a good
understanding of the impact of these variables on product and process quality,
and we can look at elements of risk, that we should be able to move within the
space that's described by this rectangle and optimize our processes and this is
not a change because it's within a pre-agreed and described variable
space. We understand the
implications. So we can now move to this
new paradigm of continuous improvement.
We don't need to keep submitting post-approval supplements.
So
it creates a kind of if and then process for the future. If industry can provide and regulators agree
that there is a appropriate relevant scientific understanding and earlier a
couple of concepts were put forward such as stability and availability, if we
can show that is understood, if we can show the ability to predict the impact
of movement within our defined vector space to predict the impact on quality
and performance, if we're confident that we understand the control of product
and process critical variables with an ability to be able to assess the impact
of change, if we can show a degree of high competence in the value of our
specifications and the validity and reliability and reproducibility of our
processes, then we get to a new state where first cycle CNC approval is much
more likely.
We
can continue to optimize our processes without seeking prior approval and we
can work to improve the dialogue and assist the risk based inspection process
because we understand what the critical quality parameters are. Of course, this carries some implications for
the future. Both industry and the agency
will need to think differently. Industry
submissions will need to change and the agency reactions and behaviors for both
submissions will also need to change.
There are some issues that we need to resolve as we move the guideline
forward, of course. Industry, what do we
put exactly in P.2. What is the depth of
the discussion that we would put? Well,
we said it could be a continuum. Looking
at it in terms of the agency, how do we construct a consistent review of the
section. Because the amount of
information is going to vary, it's not in Section P.2 going to be a compliance
document. It's an information and
understanding document. We want the
reaction that gives flexibility and an incentive, not a reaction that is
ticking the box.
We've
said that this document can have utility for both review and inspections, so we
need to define exactly what the separation overlap of roles and
responsibilities is likely to be. And we
need to think about how we might update this document. What would trigger an update to this
particular section? Why would we do it,
how would it be submitted. Now, if we
can get these resolved and I think we can, we get to a future state vision
which demands change on both parts.
Hopefully with an agency perspective, we get the more open communication
about our understanding. We're able to
work with the reviewers in an engaged way looking at the science and the agency
accepts a change of content of
applications which encourages this knowledge sharing and encourages elimination
of simply providing data. We would
encourage that agency to move to science and risk based evaluations and that
will, of course, reduce post-approval change in regulatory matters. The quid pro quo of course is that industry
needs to be transparent. It needs to
share the information. Sometimes we have
the information, sometimes it needs to be generated.
We
need to understand that our regulatory agencies have needs and we need to
provide them with those needs. If the
agency is willing to accept a different content, we have to provide a different
content, a content which shares the knowledge, a content which focuses on the
science and our understanding of products and processes and a content which actually
talks about assessment of risk and its mitigation.
Putting
that all together means that we need to provide an insight into our
manufacturing processes if we want to achieve that regulatory flexibility. But I think if we drive Q8 to a successful
conclusion, it does, indeed, open that door to the new state and it compliments
the other initiatives that have been talked about here today. Thank you for your attention.
(Applause)
CHAIR
BOEHLERT: Thank you, John. Are there any questions or comments from
committee members?
DR.
GOLD: Judy, may I?
CHAIR
BOEHLERT: Yes, Dan, please.
DR.
GOLD: First, I'm very much in support of
inter-group knowledge in the development of processes. I've long felt that we too often rush our
processes because of commercial considerations and do not explore members' base
sufficiently, so I'm very much in favor of this, but I am confused about a few
issues as explained here. I will get
your slide 12, which is parameter space, variable X, variable Y and you show a
small explore space in the upper right-hand -- left-hand quadrant showing a
rather narrow evaluation of the parameters and then you show a rather large
space to the right, parameter space to the right. Is it your thinking that this second
parameter space would be explored and defined in the initial filing? And if that were the case, why would we not
have enlarged the total allowed parameter space in the initial filing?
DR.
BERRIDGE: Well, I think each one builds
-- it depends on the scale of your understanding because I could have drawn
this with a little rectangle around what's in the right-hand area.
DR.
GOLD: Of course, of course. What I'm really asking is, if you -- in this
development, in this enlarged -- am I getting feedback?
DR.
BERRIDGE: No, it's okay.
DR.
GOLD: If in this enlarged development of
parameter space, you already know the efficiency of the variables and the
variables are acceptable to produce a product that will be fit for use, why
would you not include it in the original definition of the allowed parameters?
DR.
BERRIDGE: Well, I think that you would
include in your original submission a description of the impact of let's say
the extremes of this parameter space.
You might not have explored every increment within this parameter space
but you will know that moving around the extremes does not have an adverse
impact on product quality attributes.
You might then move instead of let's say the upper left-hand quadrant,
your consent is one where let's take a blending operation as an example. In the upper left-hand quadrant of this
picture it really represents a process that says, "Blend for 10
minutes". Now, as you move to the
future state, you change that time based concept to an -- actually, to a
material attribute concept and you talk
about blend to uniformity.
And
you then move within this parameter space, to a blend to uniformity
criteria. Now the exact space -- the
exact point you're going to be on here is one that you can -- that you monitor
and control in real time. And for
example, you may include process analysis tools to actually monitor that
attribute and you could be moving around in this space on a batch by batch
basis, depending upon your material inputs for example.
But
you can't define exactly where you're going to be at any particular point
because you've moved now to a different paradigm, not one that is rigorously
controlled, but one which moves within a bounded space that you is not a
problem provided you are within it.
DR.
GOLD: I understand, but then why would
you not include that in the additional filing?
DR.
BERRIDGE: You could include the boundary
in the --
DR.
GOLD: In the initial filing.
DR.
BERRIDGE: -- initial filing but not
necessarily the exact point that you're going to be on a batch by batch basis.
DR.
GOLD: As a manufacturer won't you be --
won't you have an advantage if you included this larger parameter space in the
initial filing --
DR.
BERRIDGE: Well, as I say --
DR.
GOLD: Excuse me, and obtain approval for
this larger space and use POT to define when an acceptable end point would be
reached?
DR.
BERRIDGE: Exactly. That opportunity is there to describe this
boundary absolutely. That's what we're
trying to encourage, a description of the boundary and an ability for you to
move within that space without having to go to the agency and say I want to
move three points to the right because it's actually not a change. It's within the agreed process and product
parameters that have been submitted in that original application.
DR.
GOLD: I'm fully in favor of this but I
believe that what you're describing may be a rather trivial example. A more pertinent example, perhaps, would be
where you have explored different particle sizes for excipients and have shown
that when you have a change in excipient particle size, and that occurs to many
of us at various times, you can still achieve a successful blend by modifying
the conditions appropriately and upon your knowledge of the particle size and
how it interacts with the blending circumstance. Perhaps that's a more significant approach to
exploring parameter space in a beneficial way.
DR.
BERRIDGE: I absolutely agree with
you. In the time I was here today, I
couldn't give you a set of illustrations of all the things but absolutely. So as I said, you could move within this
space and it may be that one of these axes is particle size and excipient dense
and another axis could be lubricity of magnesium stearate.
DR.
GOLD: Correct, correct.
DR.
BERRIDGE: And then based on the input
material attributes, you then as you're monitoring their impact on the process,
your actual process itself, the timing or whatever you do with the process, is
actually moderated by your assessment of the input attributes and I could have used
that as an alternative example.
DR.
GOLD: Yes. If I may have one more minute, Judy.
CHAIR
BOEHLERT: Okay, one minute, because we
have another question.
DR.
GOLD: Okay. And that is if we are going to allow
enlargement of Section 3 of the CTD, there's no mention in any of this yet of
enlargement of the expert report that accompanies the CTD. Is that visualized as part of the extension
of Section 3?
DR.
BERRIDGE: Well, I would have to somewhat
disagree with you. We're actually
thinking that the body of data of the CTD could change, not necessarily
enlarge, but it changes because its focus becomes different. It's information not simply huge amounts of
data. In terms of what you call the
expert report, there is no longer an expert report. What we do have is a quality overall
summary.
DR.
GOLD: I'm sorry, I'm misusing the term,
correct.
DR.
BERRIDGE: And I think there is an
opportunity and FDA itself has been describing the potential for an opportunity
to look at how that quality overall summary can act as a good distillation of
both manufacturing sciences so it's concisely embodied in that single
document. Now, what that looks like has
not been discussed within the framework of the CTD group but I think it
provides an opportunity that we're beholden to look at.
DR.
GOLD: And that is one of the objectives
that will be coming forth?
DR.
BERRIDGE: Certainly, it's one of the
topics that we should be considering.
DR.
GOLD: Thank you very much.
CHAIR
BOEHLERT: Ken, did you have a question
or a comment?
DR.
MORRIS: Yeah, a little of both,
actually. Following on Dan's point, I
think part of the issue with respect to margining space to use your example,
Dan, is the fact that when you're in development, you may not have the range of
raw material characteristics in order to define that fully. So you may not have the opportunity to file
against the whole range would be one comment.
Which
certainly leads into the question or to the thought is that one of the things
we are always struggling with in the new -- in your new paradigm is now the
three batches and out is the rule which we all agree has flaws. How do we define it so that there are
criteria that will let industry know when their product is ready to file, I
think is the question. I'm not
sure. Do we have the answer to that?
DR.
BERRIDGE: Sure, I want to delve into the
answer to that but yes, that's a pertinent question.
DR.
MORRIS: But I think that's something
that we have to discuss as we are discussing, of course, outside this meeting
as well, but it's something to be taking an issue, I'm assuming that Q8 will --
DR.
BERRIDGE: I'm not sure that Q8 will
actually attempt to define what product set validation should look like.
DR.
MORRIS: Yeah, I wasn't thinking so much
of validation in the strict sense as I was just the scientific basis for a
decision. Somebody else may have a
comment.
CHAIR
BOEHLERT: G.K., did you have a comment?
DR.
RAJU: Sure. I have a question, actually, John, reflecting
on Ajaz's comment earlier today on what you're going to put in this
section. To what extent is your thought
process and maybe all thought ICH about generating new science and data
knowledge as opposed to simply taking what you already have and putting it into
a submission? To what extent is the
about putting what you have in, in a different way or generating a new kind of
knowledge, a new kind of understanding?
DR.
BERRIDGE: Well, I think there will
always be elements of both, but I think a good start would be to provide in the
initial submission what is already there.
I think there could well be more that's available that's not necessarily
being encouraged to be shared. I think
we need to also, to come back to Dr. Morris' point, think too about the state
of knowledge at a particular phase. So I
think there will be an amount of knowledge that exists in the initial
submission, which is fit the purpose and then as the product moves into the
commercial manufacturing phase, a whole new set of information and
understanding can then be generated and I think there's an opportunity then to
build on the initial R&B knowledge with the knowledge that's acquired
through the manufacturing of scale to describe a still greater understanding of
the manufacturing sciences and it's probably -- could well be at that second
stage that we really get to a more stable situation where we described what we
call the band width within which we can truly effect that ongoing continuous
improvement.
CHAIR
BOEHLERT: Okay, Garnet, then Ajaz.
DR.
PECK: In reflecting through your slides,
there is the element of what is done in Europe and the complete understanding
of the formulation. What is the
objective of a particular product and going back to Slide 12 and flexibility, I
still see and I like this, is the material science of the material that we're
bringing together into a particular dosage form. That's highly significant and will aid us and
we're approaching a better field and you've already mentioned excipients and
particle size, that's one element of it.
The
second part of what's in the flexibility is the understanding of the processing
of what we're trying to do and I look at your diagram as an extreme vertices
type of thought and you have in the center of the extreme what you want but you
do have limits and that guides you and I think we can look towards that kind of
guiding rather than just the simple three-batch concept. It gives us space.
DR.
BERRIDGE: Yes.
DR.
PECK: And I think you've also emphasized
the space part. I think that's
important.
CHAIR
BOEHLERT: Ajaz?
DR.
HUSSAIN: I think this discussion is very
helpful but at the same time the comments consider different ways of defining
the space. For example, the aspect of
how much information we have on expedients and their functionality at the time
of new product development. Some might
be limited but you can bring that know how to bear on that because I think we
have established a way to say all right, the physics might not be different, so
the use of prior knowledge, better use of prior knowledge, I think, is a key
opportunity and I think -- so that the company has made 300 different
formulations of a drug. The chemistry of
the drug might be different but the physics of the powders are not that
different. So how can you bring that to
leverage an opportunity?
CHAIR
BOEHLERT: Okay, any last brief
comments? If not, John, thank you very
much.
Our
next speaker is Fred Razzaghi, and he's going to provide an update on ICH Q9.
MR.
RAZZAGHI: Good morning, Dr. Boehlert and
good morning, Committee. I'm here to
give you an update on the status of the Quality Risk Management Doctrine
developed at ICH called Q9. I've been
talking to you about what quality risk management is. I'll give you some background. Initial steps in guide development,
development of the guideline. The
guideline starts off as to the scope, the process and the tools and how it's
integrated into operations and what the next steps are.
This
team is defining progress management as a process in assisting of well defined
steps which when taken in sequence support their decision making by
contributing to a greater insight into risks and their impacts. And the steps in the process could include
identification of risks assessment, education, elimination and communication of
risks. There's some understanding in the
committee, in the group, that risk is a combination of property of occurrence
and severity of the harm that this caused.
Here's
some background for you. Last October
you were presented with three presentations.
One, use of management from simple manufacturing, then you provided with
a process risk assessment model and then the relationship between risk and
knowledge and how to apply them pre and post-approval, e.g. scrutiny and post-approval
changes and the variety of GMP areas. IN
April at the OPS meeting, one of the objectives that were stated was that OPS
will implement a review quality system and procedures that will recognize the
level of scientific knowledge, supporting private complications, plus process
capability, apply a risk base rate to scrutiny that will relate to level of
scientific understanding of how formulations from manufacturing processes
factors besides product performance and then the capability of process control
strategies to prevent or mitigate risk of poor product performance.
Some
background in the ICH process to date; there was a meeting in July in Brussels
where groups came together to discuss whether or not there were merits to
moving ahead. Following that, there was
a meeting in Osaka, Japan in November of last year where the concept was
developed and approved by the steering committee. In between November and June we snuck another
meeting in there in March in London, where we drafted an outline and had a
discussion for two or three days about what is the general approach to actually
making this happen.
And
then we had some significant progress made in Washington in June where a first
draft of a guideline was issued by the team and it's been distributed to the
constituents for review. A few words on
the approach here; in July of last year, this statement was agreed upon by all
parties, "To develop a harmonized pharmaceutical quality system applicable
across the life cycle of the product emphasizing an integrated approach to risk
management and science".
The
ICH process is unique in that it requires consensus by all the parties and it
has its own varying process because of
that. We also agreed in March that we
would keep a few things in mind. We want
to approach this with a process oriented thinking in mind. We want to be practical about it. We want to find where we can use available
risk tools and apply them appropriately.
We want the product that they exercise to give us some predictability. We want to approach it in a flexible manner
because we want it to apply to as many places as possible.
We
expect it to be consistent and integrated.
Initially the goal was to come up with a risk based approach here and we
sat down and went through a list of why are some of the reasons we need to have
a risk based approach here. The document
-- these are some of the reasons and I won't go through them. I will kind of run down the benefits for
you. We thought that enhanced patient
confidence in this to assure quality is a benefit. We expect to promote more effective use of
regulatory agency and industry resources.
Establish a systemic and well-informed thorough method of decision
making which leads to greater transparency and predictability. Increased knowledge of exposure to risk, and
as Ajaz mentioned, we expect this to foster quality by design, continuance
improvement in the technology embracement.
The
scope of this document is as follows; this provides the framework that may be
applied to all aspects of pharmaceutical quality, including GMP and submission
of new processes throughout the life cycle.
It applies to APIs, drugs, biologics, vaccines and excipients of
packaging material. It does not include
pharmacovigilance.
The
process is as follows. First, the
process will be initiated, assessed, risk has to be confirmed, communicated and
then follow-on review. Some guiding
principles here are the evaluation of the risks should ultimately impact on the
potential risk to the patient. The extent
of the risk management process should be commensurate with the level of risk
associated with a decision. The more
robust dissent would be to lower a certainty.
It is essential to have a clear delineation of the risk question. Risk management should be a iterative
process. People who apply risk
management should be trained and use it appropriately. A risk management process should be
appropriately evaluated and verifiable.
Now,
once we embark upon starting a process like this, this is some of the thoughts
to keep in mind. Define a specific risk
management problem or question including the assumptions leading to the
question. Assembling background
information and data under hazard where
human health impact relevant to the assessment.
Defining how the assessment information and conclusions will be used by
the decision makers. Identify the
necessary resources. Members of the team
will have the appropriate expertise with a leader clearly identified.
The
idea here to do a good job of risk assessment you need a team of experts that
can bring knowledge and information but there's a need for someone who can
exercise a tool, that's aside from the experts on the specific scientific
topics. Ask and direct life risk
assessment questions. State clearly the
assumptions in the risk assessment.
Assessing the quality and sufficiency of relevant data and specifically
a tie line of deliverables for the risk assessment.
Now,
I'm going to go through the process. The
first is risk assessment and three questions are posed. What can go wrong, what is the likelihood,
which links back to the original relationship and what are the
consequences? It breaks down into two
pieces. Risk analysis is a suspended use
of information to identify specific sources of harm and to estimate the
risk. Risk evaluation compares the
estimated risk against given risk criteria using a quantitative and qualitative
scale to determine the significance of the risk.
The
next step is risk control. It describes
the actions of the risk managements decisions.
The questions here might be what could be done to mitigate and reduce
risk? What options for controlling risks
are available? What are the impacts of
current risk management decisions on future options risk management? This too breaks down into three steps; risk
mitigation focusing on reduction of severity of harm, risk reduction focusing
on the reduction of probability and occurrence of harm and detection of harm
and risk acceptance is a decision to accept risk, i.e., no additional risk
control activities are necessary at the time the decision is made. In other words, once risk control is
completed the decision to make the move ahead but the next event you will see
allows the opportunity to come back.
The
next step in the process is to communicate the risk. Risk communication is the exchange or sharing
of information about risk and risk management between the decision maker and
other stakeholders. Information can
relate to the existence, nature, form, probability, severity, acceptability,
treatment, detectability and other aspects of risk to quality. The communication of one's stakeholders
concerning quality risk management decisions can be made through existing
channels. In other words, in each region
currently there are ways where industry and regulators communicate on a variety
of risk issues.
And
this is a piece about coming back to the decision. All risk management processes are dynamic or
iterative. Quality risk management would
apply to benefit from new knowledge with each decision cycle and used to
enhance future decisions allowing for continuous improvement. In other words, when the team exercises that
process of going through a risk decision, the outcome of that would be
something that would be useful next time a risk decision is required.
Here
is a proposed process flow. I just went
through it. There's an initiation step,
there's an assessment step, there's a risk control step and then a
communication step and then a look back or review. The we've listed some risk management tools
and in this section, what the team -- what we tried to do was not to go out and
re-invent the wheel, and was to look around for what are some of the best tools
out there that are available keeping in mind that a lot of these tools are used
in other industries and we need to apply the original criteria for retrofitting
it to the particular circumstance that we're dealing with in pharmaceuticals.
But
one thing that we thought we are going to put on that list is process mapping,
which is the orientation of thinking when it comes to risk. Most of the places we're thinking of applying
this, we're talking about a process where the knowledge of events prior and
following are important to realize. And
there's a list of them here and Ajaz mentioned HSSN (phonetic) and FMEA. All of these have a variety of attributes and
are used in different places.
A
complimentary list to that list is the use of statistical tools that give you
information that allow you to make a good discussion and there's a list of them
here. Design of experiments is something
that was mentioned already, so now this part talks about how we take these
concepts of risk management and use of the tools and where could they be used
and here's a list of them. Risk
management or risk assessment could be used in product development, e.g., a
discussion of the risks and the limits of knowledge or the specification being
set during development. Regulatory
authorities can use risk assessment and risk management when they do regulatory
pre and post-approval. It could be used
as a component of quality system. In
other words, in auditing complaints, recalls and changed management, there is
always a component that could benefit from the use of risk management.
And
there's a list of other applications. It
could be used in facility management, it could be used in supply chain
management, in other words, materials management, assessment of suppliers, that
sort of thing. It could be used in
production. It could be used in validation. It could be used in laboratory controls,
packaging and at the end we put Regulatory Authority Activities which applies
to some of the other regions. It is
quite active in this committee and they put forth some valuable information to
this product -- to this document. David
is going to talk about it tomorrow. The
risk granting and the process that's proposed comes from that.
Our
next step is apparently the draft document is out there to the parties that are
involved in ICH to review this thing and give their comments back. In September, we plan to get together and try
to consolidate those comments and take a Step 2 document to Yokohama, Japan for
the steering committee to approve. I've
listed here for you the organizations that are participating in this working
group. As you can see, it's quite
diverse and it's a challenge to work the consensus process and it has benefit
of leadership from PhRMA, the FDA and from the regulators. And we've gotten very good technical feedback
from the European industry. This is the
beginning of the list of definitions.
This list is expected to grow as we get a little more detailed into the
document.
And
then finally, I have some references for you.
Thank you.
CHAIR
BOEHLERT: Thank you, Fred.
(Applause)
CHAIR
BOEHLERT: Are there questions or
comments from the subcommittee members for Fred. Yes.
DR.
RAJU: Fred, as you look forward, how do
you see the Q9 and the Q8 processes in --
MR.
RAZZAGHI: That's a good point. What we've talked to Q8 about so far is Q9 is
basically a tool that needs substance in it.
In other words, the real value of risk management is what is the process
of working through a decision for example, to come to a decision. But that vehicle would be hollow if it's not
filled with information. So the best use
of this tool is involvement would be if the relationship between knowledge and
lack of knowledge and development can be
explored as a risk that using this process will allow us to move to the
next phase, continue to make progress and collect more information and --
DR.
RAJU: And that's something that will
take -- is the thinking about connecting the science of this with the science
of --
MR.
RAZZAGHI: Yes.
DR.
RAJU: -- approaches that would happen
after the --
MR.
RAZZAGHI: Yes, some of these things are
working in parallel. Q9 is working to
develop the document and we're kind of working closely with Q8 to find out what
the synergies are and we'd like to do that same thing. We have in our section about integrating. The real value of this tool is going to be
how to be used in a variety of places.
So the criteria that we have used for selecting and using would be for
it to be flexible and simple but maintain the poignant parts of it.
CHAIR
BOEHLERT: Other questions? Dan?
DR.
GOLD: Yes, Judy. Can you explain why you're not developing the
severity concepts that are related to all this?
MR.
RAZZAGHI: That's a good point. We have looked at some models and we haven't
quite gotten to the point where we're going to negotiate or discuss how that
ranking is going to be done but the preliminary thinking is it is fair to
stakeholders to discuss it and come to an agreement as to what the ranking --
what the appropriate ranking should be in the absence of one that's out there
that could apply to everybody. There
isn't one out there that applies to everybody.
In
other words, given a certain process, given a certain product, in the context
of the science of that product and process, you can discuss and come to an
agreement what the appropriate ranking could be or if generally speaking, the
risk is low, the person who's using the tool can do a risk ranking on their own
and then explain it, you know, in an appropriate setting.
DR.
GOLD: Have you seen any differences in
the three regions in evaluation of severity levels or concerns for severity
level differences and viewpoints?
MR.
RAZZAGHI: Yeah, I think John is a lot
more gracious about it than I. I put
that bullet in my slide. It's quite a
challenge. It's quite a challenge to
work with a topic like this in ICH. And
there are a variety of -- I mean, risk is understood in a variety of ways by
all participants. And work off of a
template that says let's look at these principles that we're trying to
implement every time you look at a topic, look at a specific issue, it's
helping us make progress. And as I said,
the regulators especially FDA has come forward with a lot of information and
they're really helping to move the process along.
PhRMA
has done a good job of providing leadership and kind of moving it along. So I would say that the chemistry within the
team is working pretty well but we have no illusions about the feedback we're
gong to get once the document is out for comment.
DR.
GOLD: Thank you.
CHAIR
BOEHLERT: Okay, Nozer?
DR.
SINGPURWALLA: Yeah, I have two
comments. On your slide entitled
"Risk Assessment", the first comment I'd like to make is that there
is a difference between what we mean by probability and what we mean by likelihood
and to articulate that difference is going to take me an hour but for the
record, I don't think you should use the two words interchangeably.
The
second comment is that your definition of risk analysis is circular and let me
tell you why. You define risk analysis
in terms of risk but you've not defined risk.
So --
MR.
RAZZAGHI: No, that's okay. I have the definitions in the back and I kind
of flew through it, but --
DR.
SINGPURWALLA: Okay, and the third
comment is your catalog of supporting statistical tools is very
incomplete. You can have --
MR.
RAZZAGHI: It is.
DR.
SINGPURWALLA: -- a long catalog, but the
more important elements that should go into that catalog should be a
elicitation of probabilities and elicidation of debilities. That seems to be the very important function
that one needs to do a risk analysis.
Design of experiments, I'm not going to argue with you but I don't think
that it should be an important tool.
Control charts, it's accumulated some charts -- it's cumulative some
charts, not accumulated some charts. So
these are just academic quibbles for the future. You may want to look at these slides more
carefully.
MR.
RAZZAGHI: I would be interested in those
two points that you raise because one of the challenges we've issued to the
team in I think it was in Osaka, was that in order for this thing to work, we
need to go back and do some homework.
It's -- you know, we really have to manage the dynamic --
DR.
SINGPURWALLA: I'm delighted to go to
Osaka and tell you what it's all about.
(Laughter)
MR.
RAZZAGHI: You're certainly welcome.
CHAIR
BOEHLERT: Other questions or comments?
If not, thank you, Fred.
MR.
RAZZAGHI: You're welcome.
CHAIR
BOEHLERT: Our next speaker is Dr. Tobias
Massa and he's going to be talking about an industry proposal for life cycle
management for processes and system control.
DR.
MASSA: Good morning. What I'm going to talk about now is not
formally an expert working group at ICH.
It's a proposal made by the three regional industry groups to look at
what quality systems need to be in place in order to realize the potential of
Q8 and Q9. We are looking at this is how
we can utilize science and risk based quality management systems to enable
post-approval change and improvement. So
we're trying to take what have we learned in Q8, what do we know about the
process, how do we apply risk management tools to it and be able to operate in
an environment that allows us to make continuous improvement, make
post-approval changes, but the important thing is trying to operate within that
box that Dr. Berridge described, define what the box is so that we don't have
to get into a loop of continually having to make supplements in order to
implement that change.
So
what we want to be able to do is define what are the quality systems that we
need to have in place that give both ourselves as well as industry the
confidence that we're looking at our manufacturing and control processes
appropriately and that based on the knowledge that we gain during development
as well as during commercialization, that we are appropriately using all of those tools, collecting all the
data appropriately, evaluating all that data appropriately and then
implementing change in a controlled manner.
What we want to do is put this into -- put this process into a guidance
because there are different expectations about how this should be done that
vary region by region and inspector by inspector.
One
of the things that both of the previous speakers talked about was that there is
disharmony in what some of these expectations are and the goal here is to try
and create a harmonized guidance of how do you apply this tool. What we would like to have in this document
is a description of how you monitor your process and controls to identify
trends. Now, those trends may tell us
you're in control and you don't have to do anything further. You just continue to monitor or they may tell
us that we need to do something to get things to an appropriate level of
control or improve the process.
We
also want to have a system that allows us to look at what we've called the
undesirable occurrences, the things that we need to react to, such as
deviations, product complaints, audit or inspection findings, or the results of
our root cause analysis and how do we incorporate those into a technical agenda
for the particular product we're talking about?
We also want to have a system that allows us to take our proactive
activities into account. We know that at
the time we go to commercialization, we may not be optimized. In most cases we are not optimized. So we go into commercialization with a
knowledgeable technical agenda. How do
you -- what quality systems are you going to use to make sure that those are
appropriately worked into the quality plan for that particular product?
What
we hope to do by having this guidance in place -- and it's important that these
things need to be linked. What we're
talking about needs to be linked to Q8 and Q9, is that we have a standard that
allows us to realize the full potential of Q8 and Q9. We have a standard that encourages industry
to make changes. I'll show you some
slides at the end of my presentation that explains why this industry is
discouraged from making changes.
We
also need to give the regulators confidence that we have the appropriate
quality management systems in place to handle this. And as Fred mentioned in his presentation, we
want to be able to provide product to the customer and insure that we have a
continued source of supply for these valuable products to the customer. One of the things that we've looked at over
time or what some of the concerns have been out there relative to our products,
and this is a slide that I think Ajaz may have actually presented here at the
beginning of our discussions about product quality and GMPs.
And one of the key concerns was that we had variability that creates an
increased risk. What we don't know about
the product and how variability impacts the product creates risk. As a result of that risk, we have more
compliance. You have to test more, we
get inspected more but that's absolutely the opposite of what we're trying to
accomplish.
What
we want to do is have quality by design; design these things into the process,
into the control of the manufacturing process rather than testing to assure
quality. And I think this slide kind of
gets to what Dr. Peck and Dr. Gold were talking about. In our typical GMP process, we have raw
materials coming into a process that's controlled by process variables that
lead to some product that meets some determined set of specifications. During development, what we currently do or
at least the perception of what we currently do is we concentrate on the process
variables and we don't look at the variability of the incoming raw
materials. So we concentrate on the
process variables and we try to optimize those during the development process
and during commercialization, we concentrate on controlling those process
variables. But when we get variability
in the raw material, to Dr. Peck's point, you know, some of the physical
attributes of these materials, we end up with an impact on product.
So
what we're trying to do is change this paradigm. Now, I'll leave this for you to read but
Deming, you know, 50 some odd years ago made a comment about variability in
inspections and testing quality in as opposed to designing quality in and the
key things to take away from his comments so that you have to understand the
process. We were talking about this 50
some odd years ago, we still talk about it today. And we want to be able to predict quality
from upstream activities and measurements, not on final product quality
attributes. And we want to do all of
this by working toward reducing variation.
Well, that's exactly what quality by design is.
Dr.
Hussain, today, presented information that was also on a slide that Dr. Nasr
gave at a presentation at DIA just last month, talked about FDA's desired state
is. Well, I think you can take the FDA
off the top of that and put industry's desired state up there as well, because
these are exactly the same things we want to achieve. We want to have quality by design. We want to be able to set specifications
using mechanistic understanding. We
want to be able to have continuous improvement and we want science and risk
based regulatory policy that allows us to undergo continuous improvement.
These
are the same things that we want. This
slide has been shown before. It's one
that started out in PhRMA and the quality by design paper. It was adopted by the ICH industry groups as
we started making our pitch to the regulators about what we were trying to
accomplish with Q8, Q9 and Q10 and I think it's been used by I don't know how
many people in various presentations.
The concept here, quite honestly, is that the more you know about your
product, the greater your level of manufacturing science knowledge is, the less
risk there is that variability is going to have an adverse impact on your
product. So as manufacturing science
knowledge increases, and that's not necessarily during development. That can be during commercialization.
To
the points that have been made before, we've probably learned just as much or
more about our process during commercialization than we do during
development. So we should take that
accumulated knowledge, the risk that associated with that product and processes
and controls should decrease over time as that manufacturing science knowledge
is obtained. The goal here is to have an
appropriate level of regulatory oversight that matches up with the level of
manufacturing science and the level of risk that you have. So the more manufacturing science you have,
the less regulatory oversight you should need particularly in the area of
post-approval changes. The key to that
it having the right quality management systems in place that control how you're
doing change within your company for that particular product because having
that flexibility doesn't decrease the oversight that you have to have as you
are implementing change.
It's
just, what we're trying to talk about here is what's the level of regulatory
oversight, what's the level of prior approval that you need in order to
implement changes. So how does this come
together? How does this work and I'll go
through this with words and then show you some diagrams of how we envision
this. It starts in development with
quality by design, using data rich experiments to identify the critical quality
attributes of a product and the process.
To the points that have been made before and what Q8 is all about is
taking this development data and getting it appropriately into an application
for review.
The
point has been made before that several of us in industry have met with FDA to
share what the data base is that we have going into an application and it's, I
think, true that we've been reluctant to submit all of the information or more
information than what we currently do in an application because all of that is
looked at as a regulatory commitment.
It's not looked at as here's the data that got us to what the actual
regulatory commitments are. How did we
identify what the critical process parameters are? How did we identify what the in-process
controls and specifications are? It all
gets looked at right now as a regulatory commitment. So there needs to be a give and take on both
industry's part and the regulator's part what information gets submitted and
how those data are reviewed and looked at.
All
of that data leads us to our validation protocol. What are the critical process parameters that
you're actually going to validate and monitor during commercialization? That will lead to your validation report and
both of those things, I think, are appropriate to be submitting as part of this
piece of data that you're going to be giving the agency. One of the things we haven't talked about,
we've talked about setting specifications based on a mechanistic understanding
of the process but what we haven't talked about is interim specifications. In other words, what are the specs based on
your development data and how might they change as a result of the accumulated
commercial data that you get over in your initial manufacturing process.
Q6
actually talks about setting interim specifications but it doesn't go into how
you go from -- or how you set interim specs and then how you convert those to
long term specifications. So that's
something that we should be able to think about and work out. Dr. Hussain talked about comparability
protocols and hopefully the final document that comes out on comparability
protocols will be broad enough to encompass the types of changes we're talking
about here. All of this leads to the
point of having continuous improvement in supplements without prior
approvals. Having the science and the
risk management piece of that is one part of doing that. Having quality management systems in place to
control that process of continuous improvement is what we're trying to
implement with Q10 or propose in Q10.
And this is kind of what we, at Lilly kind of call our radiator diagram
that depicts what we envision this process to be, starting in development and
driving towards a development history report based on the information that's in
there, you start to develop an integrated validation master plan.
We
also have what we call a process flow document which gets very specific about
how you make and control the product.
It's very specific to what equipment is used, what are the operating
parameters, raw material specifications, all of that. Ultimately that leads us from working in the
pilot plant to transferring this process to the ultimate site of commercial
manufacturing where we undergo qualification and validation using today's
parlance.
We
then get to commercialization, what we're calling execute and monitor, where
we're accumulating information and getting ready to make two types of changes;
one what we can technical evaluation changes.
That's the prospective part. How
do we want to optimize the product based on what we've learned during
development and what we've learned during commercialization. The reactive part is the GMP or quality
evaluation and that's the response to things like out of specs, deviations,
product complaints, adverse events. Both
of those would go through the same type of risk analysis that Dr. Razzaghi
referred to and we would develop a quality plan for that product at that site. And that would cycle back into the process,
maybe even going back into further development and then working its way down
through the chain again.
But
in order for that to happen, and what we're showing here are two parallel
processes. The top process is really
the scientific part. The bottom part of
this, the bottom three boxes or the bottom half of this diagram, really refers
to having the appropriate quality management systems that allow the science to
drive forward. So you can't have one
without the other and that's what we're trying to drive through with Q10. So again, coming back to Dr. Gold and Dr. Peck's
concerns about how the process should work, if we are concerned about physical
attributes, for example, of raw material coming into a process, if we're using
PAT to measure those raw material attributes, we can adjust the process
variables on a feed forward basis. By
the same token, we can look at the product and measure critical quality
attributes that are being accumulated for the product and feed back on those
process variables. And the combination
gives us better control of the product and this is exactly what, I think, we're
talking about when we talk about quality by design and operating in that box
that Dr. Berridge referred to because making these changes to these operating
parameters, these process variables, if you've defined them appropriately in
the box are not really manufacturing changes and they're not things that need
to -- certainly not things that would have to go through a regulatory approval
process.
What
you need to have are the appropriate quality systems that allow you to monitor
these things and keep track of how they're occurring and determine what changes
need to be made based on that monitoring.
Q10 is only part of the solution to post-approval changes. Now, part of the deal here is that if we're
identifying the box appropriately that we talked about, we don't have to get
into a lot of post-approval changes because they would be considered part of
the process. But I still think and I'll
make the pitch, that the regulatory process needs to be changed. And the reason we say that is that regulators
regulate regionally. Manufacturers
commercialize globally. There is
definitely -- we talked about disharmonization before. There is definitely a lack of harmonization
on the regulations that govern manufacturing changes.
Every
region has a different set of rules that we operate under and these differences
can include the regulatory mechanism for filing the same change, what the
review cycles are for the review of the dossier, data requirements and even
interpretation of the same data. Over
time, this has resulted in this reluctance on industry's part to make changes
because the regulatory hurdles are high.
And just as an example, if you consider an API in Product A, we start
off submitting one CMC dossier for that product. That gets submitted and we'll just talk about
four regions right now that result in differences in the specifications, in
process control, shelf life and in some cases can even impact how you're
actually making the product.
One
of my colleagues related to me that for the same product they actually have
three different manufacturing processes that came out as a result of the review
process. So now you've got, instead of
having one product, you've got four different bulbs that are regulated
differently because of the differences in the review process. If you now start to make process improvement
changes, you start getting differences in those products as well or that review
process results in different APIs there as well.
So
now you've got two different processes running three to five different
products. It creates an absolute
logistical nightmare to do this. Now, if
you take that and magnify it even more, saying that you're making a change in
an API that effects three different formulations of the same product, you can
see that this becomes a real logistical challenge to make change. And just by way of a simple example, we at
Lilly had a change which was an extension of an expiration date based on real
time data, based on an approved protocol.
We had to file over 100 supplements or variations and it took over two
years to get all of that approved. And
that's a simple change that certainly in the United States is an annual report
filing, but because of regional differences, we had to go through a rather
extensive regulatory process. So I think
the combination of what we're trying to do with Q8, Q9 and the proposed Q10 and
a change in some of these changed regulations will get us to a point where we
have much better use of our resources, much better use of the regulator's
resources, and a system that allows us, a quality management system, that
allows us to do continuous improvement.
Thank
you.
CHAIR
BOEHLERT: Thank you, Dr. Massa.
(Applause)
CHAIR
BOEHLERT: Any comments from the
committee? We're using up a lot of time
very rapidly. Ajaz?
DR.
HUSSAIN: Well, I think I wish to thank
Toby for coming, especially today's is his wedding anniversary and --
DR.
MASSA: Thank you. My wife will thank you if the plane gets home
on time.
CHAIR
BOEHLERT: Ken?
DR.
MORRIS: Just one question, maybe you
said this but what's the timeline of this?
DR.
MASSA: Well, that's an interesting
dilemma for us. One of the issues we're
running into, particularly with the EU and Japanese regulators, is they have
said they don't have the resources to devote to Q8, Q9 and Q10
simultaneously. So we're kind of on hold
at this point. The ICH steering
committee has given a tentative approval to the Q10 concept but we're not going
to be able to form an expert working group until we get to Step 2 for either Q8
or Q9.
Now,
give credit where credit is due, I think in FDA we're trying to drive this
forward independent of the ICH guideline.
So we may be able to lead the way here and try and push the EU and
Japanese regulators to see the benefit of this.
CHAIR
BOEHLERT: Other questions or
comments? If not, thank you, Dr.
Massa. Our last speaker before we take a
break this morning is Don Marlowe, who's going to be talking on the ASTM E55
committee.
MR.
MARLOWE: Good morning, Madam Chairman
and committee. I appreciate the
opportunity to speak to the committee about the development of standards for
PAT. What I hope to do this morning is
give you a very brief history of where we've been. It's been about a year since we've started
doing this and try to give you a feel for the framework that we're operating
within and please, as I go along, if there's any questions about where we are,
don't hesitate to jump on me here.
First
of all, I hope to leave you with these four points when we get done and to a
later or lesser degree I can do this and get you out in time for your scheduled
break. Why use consensus standards,
first of all, for PAT? Consensus
standards provides an opportunity for all interested parties to participate in
the discussion as equal playing partners, so that they members of the regulated
industry, academic experts and people from the agency can all come to a non-threatening
forum and sit down and talk about the issues and talk about what the important
topics are and agree on what approaches should be to accomplishing the
objectives that everybody wants to achieve and it's a balanced discussion. If you operate within the voluntary standards
community in the United States and particularly if you operate with an ANSI
accredited standards developer, you are guaranteed that the process discussion
will be a balanced discussion. That
means that no sector of the community will have a dominant voice in the
discussion and we'll talk about that as we go along this morning, but for
example, the FDA is just one partner at the table. The regulated industry and the academics are
partners at the table but nobody can dominate the discussion.
Due
process is an important consideration.
The ANSI, American National Standards Institute, basically has an
umbrella set of rules within which all standards are developed in the United
States and they follow closely to the WTO Code of Practice and the TBT Agreements on the handling of documents
within the standards process and one of the key attributes is due process. Everybody has an opportunity to be heard and
nobody can summarily dismiss a discussion.
It has to be considered and evaluated by all the partners.
And
finally the NTTAA, the NTTAA is the National Technical -- Technology Transfer
and Advancement Act. It was passed about
1995 and has been implemented by the Office of Management and Budget in a
guidance document, A-119 which basically tells the federal agencies to use the
standards developed through this process, through a voluntary consensus
process, wherever possible. So in order
to comply with our responsibilities under NTTAA, we are using this standards
developed in ASTM as an engine for accomplishing this activity. And ASTM is, as I said before, an ANSI
accredited standard developer with all of the baggage and attributes of an
accredited developer. They have more
than 100 years of experience. They were
formulated. They were developed in 1989,
specifically at that time to improve fatigue, what we now believe to be the
fatigue resistance of steel rails for the railroad industry.
But
they have many years of experience in all kinds of committees. There's more than 130 committees operating
within ASTM. The agency works with more
than two dozen different ASTM committees and E55 is just the most recent of the
committees that FDA has worked with in ASTM to accomplish our standardization
objectives.
ASTM
is a recognized developer of international standards. If you look around the world, more than 44
countries have written ASTM standards into the national codes, so we believe
that ASTM is an engine for accomplishing an awful lot of the objectives the
previous speaker mentioned. The
difficulty with the resources problem in many parts of the world is that the
resources are scarce to accomplish the changes that everybody wants to achieve
here and with ASTM being a globally recognized developer, we hope that some of
these will be eased.
Finally,
their offices are close. They're up in
-- just outside Philadelphia, up between Philadelphia and Valley Forge and it's
a speed run up the road. We can be there
in about three and a half hours, so it enables us to go up there, consult with
the staff managers up there on activities that we need for standards
development and also we've held several meetings in their facilities up there
and it's a speed run up and down the road for staff.
The
history of our working with these folks is very brief. It's almost a year that we've been working
with ASTM to develop standards for PAT.
You can see the calendar here, it really got organized in February of
this year and it took about four months to have the first standard published
through this consensus process. There's
a terminology standard, E2363 and there's more than 70 terms already agreed to
by the consensus process in this standard and the standard is being revised as
we speak. More terms are being added,
terms that their needed discussion in the first cycle of approval for that
standard are being revised and added as we speak.
The
next meeting well be in November here in Washington and it will be part of the
standard ASTM committee week. It will be
over at the Omni Shorham Hotel over on Calvert Street. And I encourage anybody and everybody who is
in the room and wishes to participate in the process to get engaged and I'll
have a slide at the end that tells how.
This is how the committee is organized.
There's really three functional committees; Sub 1 on management, Sub 2
on implementation and practices and Sub 91 on terminology. The Sub 90 committee is just a kind of
organizational thing that you need to do to have a committee to keep the train
running on time. But the activities of
these committees, when we sat down to talk about this, we realized there were a
few activities that were easy to talk about, to break out as separate entities,
materials, operating equipment, control the environment, people training,
analytical equipment and control systems, transport and storage of packaging
and package parts and packaged pharmaceuticals as well as the management of the
processing and packaging and obviously the systems infra-structure at the
bottom, what the plant needs to make it work.
I
didn't see the second -- the operations and maintenance systems, there's an
awful lot of things of that type that can be standardized. The initial work items, you see there on the
left-hand margin there is a work effort ongoing. WI is work item. There's a work item to develop some standards
for raw materials, another one for manufacturing equipment and finally, a work
item on instrumentation. These are the
three active work areas. We anticipate
that as some of these things are accomplished, that the committee will move
onto some of the other activities. So
the items that you see there in detail are all managed in E55-02. This was the implementation and practices
subcommittee and overall, there's an over-reaching management system being
discussed, a management systems standard being discussed in E55-01, the first
of the subcommittees that I mentioned and the objective here, obviously, is to
have a unified system and E55-01, an umbrella document and a bunch -- several,
many E55-02 implementation documents to reflect the best practices.
Some
of the -- the overall effort will be to describe and accomplish a work plan and
describe and accomplish and enable the outcomes. As I mentioned before, we would like people
to participate. The agency has a pretty
heavy commitment to making this work. I
am the chairman of E55 and some of my colleagues here in the Center for Drugs
are active on the three committees that I mentioned previously, the three
subcommittees. Interestingly the senior
management of all the other subcommittees are industry people. They are members of the regulated industry
and actually have taken their responsibility every seriously about their roles
in managing the activities of development of standards for PAT. And contacting Pat Picariello at ASTM is a
clean easy way to get into the system.
They have a website also, astm.org and E55 has a link in the website, so
that if you want to go see what's being done and what the status of things is,
it's an easy access to the information.
And I'll answer any questions.
And I missed by a minute.
CHAIR
BOEHLERT: Very good, nevertheless. Any questions in E55? G.K.?
DR.
RAJU: Don, I really value the due
process in which you operate and we certainly hope to live up to the
expectations in all these meetings. As
you look at the rest of our discussion around ICH and you look at the
compliment in terms of resources internationally, on one side that's a positive
thing. Do you see any duplication
possibly in the future in terms of ASTM doing things and ICH doing things given
how long it takes for the government.
It's tough to look at duplication after the product is over and if there
is some, does it get managed with the people more common or is it done
structurally with people like you?
MR.
MARLOWE: It is actually -- I think it is
actually done best with the people that are involved. I think that it's unlikely that there would
be some kind of a super management system of the whole thing, but I do think
that the exercise within ASTM will be a detail exercise, not an overall quality
system management discussion. So there
will be a discussion on best practices for management of PAT within a
manufacturing facility where the regulated industry, the firms, get to share
their best practices but the overall impact of that and the overall role that
that would play in a quality system management plan for a firm will not be on
the table in ASTM. That will be an ICH
discussion.
CHAIR
BOEHLERT: Any other questions or
comments?
MR.
MARLOWE: Thank you, ma'am.
CHAIR
BOEHLERT: Thank you, Don.
MR.
MARLOWE: Appreciate it.
CHAIR
BOEHLERT: And we are right on time so I
thank you for your effort on our behalf.
We're scheduled for a 15-minute break and we'll reconvene promptly at
10:45.
(A
brief recess was taken.)
CHAIR
BOEHLERT: Okay, it looks like we have
all members present and accounted for.
Now we're going to change directions a little bit and we're going to be
educated hopefully, on Bayesian statistics.
Nozer, ti's all yours.
DR.
SINGPURWALLA: Well, thank you for the
opportunity or the imposition to give this talk. Let's see, okay, thank you. Well, the good news is that I've given this
talk about -- this is the third time I'm giving this talk in the last two weeks
which is fortunate because I was asked a few days ago by Sandia Labs to give a
talk on Bayesian statistics and things like that. Then I was in Iran giving the same talk and
Los Alamos Labs wants me to give this talk again, so I've got a package that I
can keep talking and talking and talking about.
Now,
the motivation why I was invited at Sandia to give this talk is that there is a
large group of individuals who are thinking in terms of alternatives to
probability. And so they wanted me to
talk about this topic and particularly as applied to reliability and
fortunately, Ajaz mentioned the word "reliability" so I feel slightly
comfortable talking about it but basically the title of this talk is
"Reliability for the Analysis of Risk" and it is a Bayesian
perspective.
So
these are my coordinates and this is mostly based on a book that I'm working on
for a long time. So first let me start
with proper definitions. Everything has
to be defined so that there is no confusion of vocabulary. So the first question is, what is reliability
and why is reliability relevant to this particular community? It's -- this is some spy is ringing the
phone.
Okay,
so it's the quantification of a certain type of uncertainty associated with the
efficacy and safety of a large complex system to include biological systems
where it goes under the name of "Survival Analysis". So your drug -- sorry? The drug manufacturing is also a large
complex system and it doesn't matter what the complex system is, but basically,
reliability is the quantification of uncertainty.
The
next question is, why do we need reliability, why reliability? Well, it is one of the two necessary
ingredients for making logical decisions in the face of uncertainty connected
with the efficacy and safety of large systems.
So reliability is one of the two ingredients that we really need to make
logical decisions no matter what the decision is, whether to administer a
certain drug or whether to manufacture a certain drug or how to manufacture the
drug, it doesn't matter.
What
is the other ingredient? The other
ingredient is utility and utility is a very difficult concept to essentially
make precise but most of the time when we talk about utility, we talked about
costs and our rewards that occur as a consequence of any chosen decision. So every time you make a decision, there are
going to be consequences and associated with the consequence there is either
going to be a risk or -- I'm sorry, I shouldn't use the word
"risk". There is either going
to be a cost, a penalty, or there's going to be a reward.
So
the next question comes up, is what do we mean by risk analysis? We've heard this term used repeatedly in
this particular audience and I think I would like to see risk analysis as the
process assessing reliabilities and utilities and it should include the
identification of the consequences. So
risk analysis is the process of assessing the reliability and the utility and
think of reliability as a probability.
Think of reliability as a probability but let's keep it specific.
And it should include the identification of all
the consequences.
The
next question comes up why must we quantify uncertainty? Why this business of quantification? Why not just do things? Managers essentially make decisions without
quantifying, you know. Generals make big
wartime decisions without quantifying.
Why not just go ahead and do it based on whim. Well, I'm not saying that by doing things on
whim you won't do the right things but essentially formally, by quantification
we mean the measurement of uncertainty and by measurement we mean a comparison
against a scale. For example, we use
feet for distance and pounds for weight, so what we need to do is really we
need to come with a scale to measure uncertainty. We are uncertain, we have to have a scale to
measure uncertainty because we want to quantify uncertainty.
And
measurement is a necessary ingredient for invoking the logical method and
mathematics is a logical method and I'm sure there may be others but I only
know of one. Because without measuring,
we cannot talk about it as said very nicely by Lord Kelvin several years
ago. So we need to quantify so that we
can invoke the logical method and without quantification, we really can't talk
about anything systematically. Thus, to
quantify uncertainty we need a scale of measurement.
So
the basis problem is we are uncertain about certain things. We need to quantify it and to quantify we
need a scale and so the question comes up what is the scale. What is the scale for measuring, what is the
fortrula (phonetic), what is the weight for measuring uncertainty? So what are the scales for measuring
uncertainty? Well, probability is the
oldest and perhaps the most commonly used case.
There are alternatives to probability that are popping up on the horizon
with a lot of passion and with a lot of debate but sometimes without much
content. And these alternatives are
possibilities and as this community gets more and more into this game, I won't
be surprised if 10 years down the line, there will be an Ajaz Hussain standing
up and saying, "We should use possibility", so I want to caution you
that there is a scale that's lurking on the horizon.
There
is also another scale, it's called belief.
There is another scale called plausibility. There is another scale called fuzzy
measures. Confidence limit and point
estimate is also a scale, but probability is the oldest and perhaps the most
commonly used scale. Well, the
questions comes up is if you are advocating probability as a scale, why it
should be the scale, what about these other possibilities and beliefs and so on
and so forth? Confidence limits, the FDA
uses them. Point estimates, the FDA uses
them. We think these are alternatives to
probability. So what are the strengths
of probability?
Well,
the first strength is it has a foundation that is firmly grounded in coherent
behavior -- coherent betting and the axioms of coherent behavior. So there is a foundation behind probability
that is firmly grounded in coherent betting.
Coherent betting simply means you don't go to Las Vegas purely with the intention
of losing money. You're hoping to come
out ahead. So any time you gamble, there
should be a fair chance of also winning.
And axioms of coherent behavior it's a long story but human beings
behave in certain ways and the calculus of probability is grounded.
But
the more important reason, particularly germane to this particular activity, is
that it's calculus leads us to a prescription for decision making under
uncertainty. Most of you in business and
industry are decision makers. So you
need to make decisions and you need to make logical good decisions, how you're
going to do it. Well, it says that the
calculus of probability and I'll tell you what the calculus means, leads you to
a prescription for decision making under uncertainty. The others to the best of my knowledge, do
not have a similar prescription.
So
the next question comes up, if that be the case, why are there alternatives to
probability? Well, this is a technical
issue and I won't go through the details of this but the axiomitization of
probability, the legitimization of probability from a mathematical point of
view is based on a certain structure which some people find is very rigid and
therefore, they propose alternatives to probability, but we won't go into the
details but to the best of my knowledge, the alternatives do not have a
behavioristic foundation and do not lead to a prescription for making
decisions.
Also
some alternatives lead to answers that are inadmissible. That simply means you get silly answers,
answers that fall flat in terms of common sense. But I'd like to make some qualifying comments
and slowly we should get to that. As a
word of caution, the axioms of coherent behavior upon which probability and its
calculus are based are set to be normative.
That means they tell us how to behave.
In actual individuals may not behave according to the dictates of
normative behavior. We have plenty of
examples. People do silly things. I like to drink alcohol every day in the
evening. I know it's bad for me but I do
it. So that's not normative
behavior. I'm told not to do it, but I
do it and there are other examples. I've
done some recent work with my colleague, Jane Booker, who is at Los Alamos
Labs, and we have been able to overcome some of these objections. Again, I won't go through the details.
All
right, so that much for some background.
And now the main question, what is Bayesian inference which is what you
all want to learn or those of you who know about it simply find all of this
very trivial. Those of you who don't
know, wonder why all this is happening.
So what is Bayesian inference?
Well, the answer is very simple, Judy, extremely simple. And the answer is this; when the
quantification of uncertainty is solely based on probability and its calculus,
inference is said to be Bayesian. So to
be a Bayesian simply means following probability and the rules of
probability. And of course, it's not
easy to understand the rules and it's not easy to work with it, but as a
general statement, if you are purely going to describe uncertainty, and measure
uncertainty using the calculus of probability, you are a Bayesian. Any time you violate from that, you're not a
Bayesian.
In
other words, a Bayesian is strict in his or her adherence to the rules of
probability. That's it. It's not very hard to be a Bayesian. Well, of course, within the class of
Bayesians there are categories and I'm just putting this down. One are called Objectivists and the other are
called Subject Matter Specialists. The
Objectivists, the spokespersons for that particular school were Jeffreys, a
British astronomer, mathematician, philosopher.
Jaynes was a an American physicist, passed away recently and LaPlace,
you all know who he was.
So
they wanted everything to be objective and they simply said, "We should
quantify uncertainty using probability but we should not have any personal
opinions coming into the picture and what we need are standards by which we can
work". Of course, this particular
school was criticized. In fact, La Place
was severely criticized for doing this and essentially La Place suffered a
tarnishing of his reputation. Then there
are the subject matter specialists and the biggest proponents of that school
are De Finetti, Savage and Lindley, who happens to be my co-author and
friend. They basically are of the
opinion that to quantify uncertainty you really have to understand the subject;
drug manufacturing, engineering, economics, physics, whatever have you. You have to really get into the guts of the
subject in order to be a good Bayesian.
That was basically the idea.
There
is a long debate about it and a long -- so, what is non-Bayesian
inference? Well, it's the opposite of
Bayesian inference; any process of uncertainty quantification that does not
fully subscribe to the calculus of probabilities so labeled. Well, of these, Frequentist Inference is the
most prevalent. In the FDA and in NIH
and in government, Frequentist Inference is the most prevalent. All your military standards; 404, 105D, a lot
of your control charts, quality control procedures, the old ones, the Shoehart
(phonetic) chart, Quinsome (phonetic) charts, they all Frequentist and a
Bayesian would reject them, including Deming, who at some times rejected them
not because he was a Bayesian but he was using his common sense.
Now,
why is there Frequentist Inference?
Frequentists, while subscribing to the notion of probability as a metric
for quantifying uncertainty, interpret probability in such a way that sometimes
they have to forsake probability as the sole basis for quantifying
uncertainty. Well, I just mentioned
probability but there are many ways to interpret probability and if I had the
whole day, I would go into all those particular issues but I've been given only
45 minutes. Fortunately, they are during
the morning. I was scheduled to speak in
the evening when all of you would be either gone or asleep or if you were
awake, you would fall asleep. But we've
been moved up and there is a long reason why all this happens.
So
I'm just going to put up one little picture as a schemata of what's going
on. So here we have the quantification
of uncertainty and we basically have two groups. One group says probability is the
metric. Then there is another group that
has possibility, belief, confidence intervals, and all as metrics for confine
uncertainty. Within this particular group,
you have the objective Bayesians, we have the subject matter Bayesians and then
we have the Frequentists or Sample Theoretic people and it's a kind of a
strange box here because this box has an arrow here and also an arrow
here. This particular proponents of
this, most statisticians that I know and I was trained as a Frequentist,
essentially use probability as a metric but the interpretation of probability
at some point in time drives us into this particular box. So that's the schemata.
Well,
that's an overview and that's a general statement. Well, the best way I can illustrate all this
is by a very simple example and in the course of the example I will define what
I mean by risk and I will also define what I mean by utility and this is what
you would call risk based decision making or whatever verbiage you use. The simple example that I will use is the
simple example that I've always been successful using for the general audience
because everybody flies, takes an airplane, including myself and you're faced
with a decision. And what brought this
to my attention is I was on a committee of the National Academy of Engineering
or Science or whatever on certification of aircraft and I was dealing with a
lot of people who manufactured huge, big, powerful engines which take this
plane up and the particular individual who was on this committee was a very
fine gentleman from Boeing who was responsible for putting two engines on the
Boeing 777. So that was a big decision
why they built this plane with only two engines when the classical jumbo jets
had four engines, so how did they make this decision to use two engines?
Well,
they didn't use decision theory to be quite honest with you. They didn't use what I'm prescribing but I
had to talk to him and tell him this is how I would go about doing it. So I'm going to give you that example. So the example here is should we outfit a
newly designed airplane with one engine or with two engines? Now, when you're manufacturing drugs at
Pfizer or wherever have you, I'm sure you have a lot of decisions to make. You can translate this into your own
particular problem. So how should we go
about looking at this particular problem?
Well,
I'm not going to put numbers because I'm very uncomfortable with numbers, so
let's C1 be the cost of acquiring and installing an engine. Risk analysis is the most important thing are
two ingredients, probabilities and utilities.
Utilities are costs.
Probabilities are probabilities no matter how you interpret them, those two things are the
most important elements of making risk informed decision making or whatever
verbiage you use. So C1 is the cost of
acquiring and installing an engine. This
is slightly loose. C2 is the loss
incurred due to an aircraft failure. So
if the airplane fails because you don't have enough engines, you're going to
suffer a big loss, I just called it C2 and I call this C1.
And
I'm assuming that C2 is much bigger than C1 because a loss, if an airplane goes
down, is going to be much more than cost of putting an engine. You know, it keeps running and running and
running. Let C1 be the reward received
upon successful flight. So every time
you carry passengers from Tehran to London, which is what I flew and then back
from London here, they collect effort from.
So this is -- just measures the air flow. All right, now comes the next component and
again, it's all laid out in notation, P1 is the probability of failure of an
engine during its mission. There's a
probability that the engine will fail and I do faultry analysis, failure modes
on effects analysis. I do all kinds of
things, collect data, collect expert judgment, talk to the fellows who design
these engines, blah, blah, blah, and come up with a number P1 as the
probability that a single engine will fail.
Well, I have two engines so P2 is the probability of failure of both
engines. So you know, one engine can
fail and there is a certain probability, and P2 is the probability that both
engines fail and when both engines fail, we have a bit of a problem. How do we calculate P2? It's a big complicated question. I have simply multiplied P1 by P2 which is
what old-fashioned individuals in the industry were doing.
They
were assuming that the chances of failure -- that the failure of one engine
doesn't increase the chance of failure of the second engine. So they were just multiplying it and they got
into ridiculous problems doing this. But
I've just put P2. Now, the next question
is, so this is a part of risk analysis, getting this P1 and P2 and C1, C2 and C
is all a part of risk analysis. But also
a part of risk analysis is the consequences to each decision. What is the consequence in this simple
example? Either we succeed, which is S,
or we fail which is F. So there are two
consequences. It again, illustrative. There are other ways to look at this in much
more detail but I'm just giving you a general sense of what needs to be
done. If you want to move forward in
this business, these are the kind of thinking that should come into play.
So
we start by constructing a decision tree.
Again, there is fancy vocabulary here used by different people. The last
meeting we had they used some other term which was more acceptable to others
but basically you had a constructive decision tree. So let's look at the decision tree. And this is the guts of everything. We have to make a decision and that's called
a decision node, D. The decision maker,
the engine designer, the airplane designer has to make a decision. So she has two choices. She uses a single engine, which is decision
D1 or she uses two engines which is decision D2. So those are the only choices she's allowed.
Now,
as soon as she makes her choice, nature comes into play, that's called R1 to
denote the random node. What is nature
going to do? Either it's going to result
in a success or it's going to result in a failure. This is not a game because you're making a
decision against a benevolent nature. In
a game, when you want to use this in the context of strategic issues, you have
an opponent who is kind of active, but this is a passive opponent. So you either result in a success S, or an a
failure, F. Then you have to outline your
utilities. The utility of a success when
you make decision D1 is USD1. The
utility of a failure when you make Decision D1 is UFD1. P1 is the probability of failure, remember I
did that before and one minus P1 is the probability of success.
Again
the rules of probability say that if this is P1, this should be one minus P1,
so you have the utilities here.
Similarly, you do this at this node, the second node, that is you have
chosen two engines and then at least one engine survives. We assume that with one engine the airplane
can fly and we assume that with both engines failing the airplane comes
down. In actuality, it doesn't, it
glides down, but we just assume that it's a failure and then there is a utility
associated with those two. So any risk
informed decision making you want to make, if you're not going to come up with
a good decision tree, then you're just doing it in a haphazard way. This is the important step that you have to
go through and the important step involves a lot of important sub-steps.
You
have to calculate your probabilities.
You have to calculate your utilities and you have to outline what the
consequences are. Here I have only two
consequences, S and P1. Well, the rest
of it is all mechanical calculations but I'll illustrate what the calculations
are. At this random note, R1 we compute
the expected utility of decision D1.
This expected utility is called, by definition, the risk. This is the risk of decision D1. It's the
expected utility. What do we mean by
that? It's the -- the following
calculation is the utility of success
when you choose decision D1 multiplied by the probability of the success plus
the utility of the failure when you use Decision D1 multiplied by the probability
of failure.
So
this expected utility is calculated by R1 and I, of course, should also write
here saying that this is what is called the risk of decision D1. This is the definition of risk, expected
utility. And that's why I made a comment
to the previous -- one of the previous speakers what it means. Well, these are just numbers. You don't have to worry about it but these
utilities I have set down in terms of costs.
Of course, discomfort to a passenger also is a form of utility and we
need to quantify that and that's going to be very important especially in the
drug business where you can take some kind of a medicine and have side effects. It cures your disease, but you feel
lousy. How do you put a value to
it?
Well,
that's the more difficult part but somehow you had to come up with a value and
I have simply used dollars and cents to encapsulate this. Similarly, you do this at R2. You do exactly the same at R2 and you compute
the expected utility at the second node.
And again, I have these numbers.
You don't need to go through the details but you have to compute the
expected utility at this node and at this node.
Then the beauty of all this is this principle of maximization of
expected utility, MEU. It says, use
decision D2, namely two engines if the expected utility, that is the cost
multiplied by the probability, added over the two consequences, exceeds the one
for decision D1. Otherwise choose
decision D1 which is a single engine. So
all you have to do is construct the tree, not easy but this is where you have
to work with the correct scientists and the correct people, elicit
probabilities, which is where most statisticians would play a role, elicit
utilities which is where economists, managers, marketers and others would play
a role and compute expected utilities and choose that decision for which the
expected utility is the highest.
Well,
these are some notes because the audience I was talking to was engineers who
always like to dabble with numbers and always like to pull out their
calculators and punch a few digits, seven digits off to the decimal point and
brag about it. We don't want to do this
here. So here's a commentary. The role of probabilities and utilities in
making decisions is clear. The more
important point is this; that it is the calculus of probability that leads us
to the maximization of expected utility as a prescription for taking
action. So there is probability and
there are rules of probability which would be the next topic if I were
continuing this talk but just so that you may know that it's the rules of
probability that lead you to the maximization of expected utility.
The
alternatives to probability need to provide a similar prescription. I don't think they have one and they need to
come up with one before the alternatives could be. Okay, the above plus the fact that the
calculus of probability has an axiomatic foundation that is grounded in
coherent betting and behavioristic rules is the strongest argument in favor of
the Bayesian paradigm so why should we be a Bayesian? Because it's the calculus of probability that
leads you to a prescription for making decisions and that the calculus of
probability and probability as a metric for measuring uncertainty had a
foundation that is grounded in so many other things.
Now,
a lot of people like to be Bayesians. The fact that the Bayesian recipe can
address problems like one of a kind system.
Suppose you've designed a new airplane where there has been not trial
runs, do you make a decision to fly it or if you want to send a spaceship to
the moon, you have not sent spaceships before, should you decide it or not,
that's a one of a kind system.
Information
fusion, the Bayesian -- the calculus of probability allows you to fuse
information systematically, rather than doing it in an ad hoc way, the ability
to make predictions. Savings on sample
size are simply desirable by-products.
There are a lot of by-products that are very desirable but the key
argument for advocating a Bayesian point of view is philosophical and
mathematical. The key reason is to have
a sound philosophy and sound mathematics.
The fact that there are some nice by-products should not be the driving
argument. That's just something which is
desirable.
But
there are some issues and what are those?
But this philosophic disposition also entails a price to be paid. It takes the form of two issues. So you want to be philosophically clean and
clear but you have to pay a price. And
what is the price? The actual behavior
of humans is not always normative. We
don't do things which we are supposed to do always. We get more pleasure doing things that we are
not supposed to do and maybe pleasure is
a part of our utility but we -- there is a specification of the prior that
comes into this business, posses difficulties and my previous meetings here at
the FDA and other places the big flag raised against the Bayesians is the
prior. You'll hear the word, "But
the prior, where do we get the prior for".
Because to get the prior you need to understand the underlying science
and engineering or the economic theory so statisticians don't like to get
involved, at least some of them, don't like to get involved in physics or
chemistry or pharmacy or economics. They
just want to do what they're trained to do.
But
this particular paradigm requires that you start talking to your scientific
colleagues in other disciplines and that becomes an issue. The other more important issue is that the
priors may not be unique. My prior and
your prior may not agree and, again, it's a big topic of discussion. Why it is so, we won't go into it but the
Bayesians have an answer to this and what is the answer of the Bayesians? Well, the answers are the following. The first answer is that the behavioristic
axioms dictate how one ought to behavior.
They're a prescription for rational behavior. The Bayesian says, "This is how you
should behave. The fact that you don't
shouldn't be a criticism of the paradigm".
The
second more harsh reaction is that one has no business working on a problem
that one does not understand, thus studying the underlying science and
engineering is a desirable thing. And
the basic argument is whether you're testing an engineering unit for success or
failure should not be viewed in the same vein as studying the sex of newborn babies,
whether they are male or female. There
has been studies, you know, what proportion of newborns are males or females
and there is also this same similar issue of testing for success or failure. You shouldn't look at those, both those
problems in the same vein. One has
genetics and biology; the other one has physics and other things going into it.
Non-Bayesian
methods lead to inferences that are inadmissible and this is a heavy price to
pay. And here are some examples of non
-- of inadmissible answers. You can get
estimates of failure rates and densities that are negative. You're estimating something which by
definition is a positive quantity and you can produce estimates that are
negative and engineers will simply reject that answer. Well, you get confidence limits that are
silly. I won't go through the details,
and you can also get into this trap.
Perhaps more important for the FDA, you are testing some kind of a drug
for acceptance or rejection, approval or non-approval. A capricious individual, a capricious
organization, can manipulate the process in such a way that you will accept bad
things.
So
in the context of military standard 781(c), sequential live testing, there is a
nice example where a manufacturer of bad products can sell the government the
bad product by completely following the rules but by behaving in a certain
capricious manner. I won't go through
the details but just as an example.
Thus,
as a matter of principle, some do not use procedures that could lead to a trap,
even if the alternative procedures demand more of the user such as specifying a
prior. And the other point I want to
make is there's no known situation wherein the use of a Bayesian approach has
resulted in an inadmissible solution or an inconsistent estimate. In other words, the Bayesian solution is a
safe bet once a prior has been agreed upon.
And the main important problem is eliciting priors which seems to be the
main job of a statistician; namely, you elicit priors to estimate probabilities
and of course, it's a big enterprise which is what we need to work on and I
have -- this talk goes on for the whole day but I'm not going to punish you,
nor am I going to give you a test which is what I promised Helen, so I'm going
to spare you in the hopes that -- okay.
I
have all this on a disk which the people at Sandia were kind enough to
transcribe to a disk, so I'm not going to give you all the 80 slides. I can provide 19 slides but this is just a
casual conversational overview. There is
a lot behind this and there is a lot that needs to be done but my only advice
to you, Ajaz, is unless you take a specific problem, simple as it is, and work
it through, you cannot lead the way. We
are simply otherwise talking about what needs to be done. Sit down, take a problem and work it step by
step perhaps in collaboration with industry, get the whole group together, just
to see how this needs to be done, thank you.
Bye.
(Applause)
CHAIR
BOEHLERT: Thank you very much,
Nozer. Are there any questions or
comments? Yes, G.K.
DR.
SINGPURWALLA: G.K., yes.
DR.
RAJU: Nozer, have you seen people at
Boeing or -- ever use this successfully?
DR.
SINGPURWALLA: Oh, yeah.
DR.
RAJU: And what is the benefits, what has
been their experience?
DR.
SINGPURWALLA: Well, since you asked the
question, the subject of reliability was invented at Boeing. They invented the idea of fault trees. Well, it's part of the game.
DR.
RAJU: Is it strictly Bayesian influence,
Bayesian decision?
DR.
SINGPURWALLA: Well, I'll tell you what,
Boeing Laboratories closed about 20 years ago.
So I can't talk about Boeing any more.
I can only say one thing, that they invented the Bayesian -- I'm sorry,
they invented the fault trees, failure modes and reliability. They contributed fantastically to it. Where I see this happening mostly is right
now at the labs, at the national labs, there is a lot of passion one way or the
other, for this and there is a lot of activity going on in this.
Of
course, people in business have used it quite a bit. People in oil exploration have used it, you
know. There are pockets of resistance
but I think the pockets of resistance are losing the battle.
DR.
RAJU: The arguments from the purists is
the traditional --
DR.
SINGPURWALLA: No, no, we are the
purists. The argument from the impurist,
okay. As long as we get it right.
DR.
RAJU: The FMEA that the aerospace
industry started are not truly Bayesian and you can't really multiply them
because they haven't really been formulated as probabilities. I mean, this is --
DR.
SINGPURWALLA: FEMA?
DR.
RAJU: FMEA.
DR.
SINGPURWALLA: Failure modes and effects
analysis is a strictly engineering function.
What they do is they say that the airplane has failed, why did it fail? Was it the engine or was it the pilot? If it was the pilot, why did the pilot
fail? Did he have an alcoholic drink or
was he upset and if it's the engine, was it the wings? You know, they go through and trace the whole
process. So that's the failure modes and
effects analysis.
Now,
when you design a new airplane, you want to calculate the probability that it
will be successful, so you have to first lay out the whole scenario, that's the
failure mode and effects analysis, then work your way up calculating the
probabilities. Now, people make
mistakes. What's the biggest mistake
they make? They multiply probabilities
when, in fact, they shouldn't. So here's
a classic example. Take the Boeing 777,
the Boeing 747. It's got zillions of
parts. Each part had a probability of
failure. If you multiply all those
probabilities, then the probability of success of the airplane goes down to
zero, yet the airplane flies. So
immediately the reaction was something is wrong with our calculations. So the big criticism is not Bayesian methods
and don't confuse Bayesian methods with calculating probabilities, you
know.
If
you don't calculate your probabilities correctly, you are going to get silly
answers. So I think the big question is,
how to do it correctly. It's very difficult,
time consuming and demanding, but there are certain rules which have been --
obvious rules which have been violated and that is the biggest problem.
Any
other comments? Ajaz?
DR.
HUSSAIN: No, I think if I recall the
discussions we had at one of the previous meetings, probably the main advisory
committee meeting where we discussed the zero tolerance and we discussed the
traditional confidence and --
DR.
SINGPURWALLA: That's right, that's
right.
DR.
HUSSAIN: -- confidence and criteria for bio-equivalence
and so forth.
DR.
SINGPURWALLA: And we changed?
DR.
HUSSAIN: Right, and I think could you
put this in that framework? What are the
advantages of moving away from that type of approach to something that uses a
Bayesian type of approach?
DR.
SINGPURWALLA: Well, let me give you an
example of why you shouldn't use confidence limits, okay? And this is going to be a quiz, Ajaz because
that's how you're going to learn. I have
-- X is the height of all men in this room.
And suppose X is distributed normally with some mean -- don't even worry
about normal, there is some mean MU1. Y
is the height of all women in this room and they're also normally distributed
like us, thank God, otherwise we'd be accused of sexism, and their mean is
MU2. Okay, the height of men is MU1, the
height of women is MU2.
And
for some reason, some crazy statistician wants to estimate the ratio of MU1
over MU2, the height of men over the average height. And he calls that RO and he computes
confidence limits on RO. Does all his
calculations nicely, computes confidence limits and he comes and says the
following. "I've computed the
confidence limits. They are minus
infinity to plus infinity and the probability of coverage is 99
percent". Will you buy that? No, it has to be one. That's the kind of answer you'll
produce. So there are certain traps that
you can fall into. Now, it doesn't mean
that you'll always fall into a trap.
Sometimes you'll fall into a trap but once you fall into a trap, you
have to be careful because you don't know where the next trap is.
There's
another reason. The meaning of
confidence limits is itself a convoluted idea.
A confidence limit when you calculate, doesn't tell you anything about
the particular scenario. It says, if you
repeated this process over and over and over again, 95 percent of the time
you'll get what you want, whereas the Bayesian response is, "Well, I'm not
interested in the, you know, 99 other scenarios that I have not seen. I'm more interested in this particular
scenario". That's why you should
get away from them. But they are very
strongly ingrained into our culture and that's the reason why we have this. So the particular meeting that we had, we
advocated a decision -- they used another language, but basically this is what
they were doing. Thank you. Any other comments?
DR.
HUSSAIN: Just one more, one of the other
aspects, we had a two-day workshop on this, at FDA at Johns Hopkins University.
DR.
SINGPURWALLA: I'm familiar with
that. I mean, I'm familiar with the
characters that go to the workshop.
DR.
HUSSAIN: Right, and our sister center
CDRH has been using Bayesian approaches.
DR.
SINGPURWALLA: For equipment only.
DR.
HUSSAIN: Right. In the context of what we are talking about
ICH Q8, Q9 and so forth, I think one of the attraction that leads me to seek
more information and probably more research in this area for myself and for FDA
is use of priors because prior knowledge and use of prior information to make
better decisions is the opportunity, I think, I see of --
DR.
SINGPURWALLA: That's right.
DR.
HUSSAIN: How can -- can you share some
more thoughts on that?
DR.
SINGPURWALLA: Yeah. First is, I want to criticize you with no
prejudice, of course. You should not be
a Bayesian because you could use prior information, no, no. You should be a Bayesian because it's
logically closed and coherent. Now, the
fact that it allows you to use prior information is certainly a big advantage
because you're going to save on the amount of testing and so on and so
forth. The danger is bad prior
information could also lead you astray.
So getting an honest and honorable period is going to be an activity and
there are methods by which you elicit prior information from people who are
subject matter specialists and experts and codify it very carefully.
There
are methods and there is a large body of literature to do it. There is also a philosophic position and that
is the following. That any Bayesian
analysis is the analysis done by either an individual or as a group of
individuals -- as a group acting as a whole and it's their best judgment. So it's completely possible that given the
same data, and given the same information, one group can come up with a certain
conclusion and another group can come up with a different conclusion because
they have different prior knowledge and different priors.
That
people find objectionable. They want one
answer to run across the board. So that
is a criticism. So I don't know if I've
answered your question but getting prior information is an essential step and
there have been efforts to get away from this ever since the days of La Place
and right now there is a large body of Bayesians in this country actively
growing, who are trying to get away from the prior information and come up with
canned priors.
Most
of the pure Bayesians reject them as, you know, not being to the spirit of what
is intended here. So there is a big
activity but there are methods by which you can elicit and quote prior information
and that's where the research effort should be going. The prior information plus the data gives you
the probabilities. Utilities is another
big very important subject, particularly in the drug context because there are
side effects which are uncomfortable.
The drug industry has a very serious problem in terms of utilities. It's not just dollars and cents. It's more than that, so I think those two are
the key important steps. Another
question?
DR.
MORRIS: Yeah, so if I understand
correctly then, so if your risk is the weighted average of the utilities
weighted by the probabilities --
DR.
SINGPURWALLA: That's right.
DR.
MORRIS: -- and if we don't really have
priors, as you say, if there's an absence of priors or in some cases maybe the
data that have been collected aren't really critical attributes and don't
really reflect the utility or -- so you can't really calculate a probability I
guess, then are you basically saying that that's -- you can't really apply the
Bayesian methods until that's the case?
DR.
SINGPURWALLA: No, thank you.
DR.
MORRIS: So --
DR.
SINGPURWALLA: Okay, I got the gist of
your question.
DR.
MORRIS: Okay.
DR.
SINGPURWALLA: And it's because I just
didn't make one thing clear. It is true
that you have to calculate the weighted average, utility multiplied by
probability. How do you get the
probability? There are two schools of
thought, the Bayesian and the non-Bayesian.
Okay? The Bayesian says you must
have a prior to calculate the probability.
The non-Bayesian says well, the priors could be subject, non-unique and
therefore, we should only have data to calculate it. The Bayesian said, you need the prior and the
data to calculate the probability. The
non-Bayesian simply says, you only need the data and no prior, okay?
But
once you've calculated the probability, both the Bayesian and the non-Bayesian
will use the same prescription. The only
flaw here is that the non-Bayesian, in using the prescription, essentially uses
the calculus of probability and the calculus of probability demands that you
have a prior. It's slightly, you know,
elaborate to explain but both will do the same thing.
Decision
theory has been practiced even by non-Bayesians, okay, but the foundation for
it comes from the Bayesian thought process.
Simply being a Bayesian means following the rules of probability.
CHAIR
BOEHLERT: Okay, any other questions or
comments? Nozer, thanks very much.
DR.
SINGPURWALLA: Sure.
CHAIR
BOEHLERT: Our last speaker before lunch
is Dr. Ajaz Hussain.
DR.
HUSSAIN: I wanted this to be sort of
filling the gap to some degree but I think it's an important topic. It's again, an awareness topic that we wanted
to sort of put on your radar screen. We
probably will discuss this in detail at a subsequent meeting but I do want to
sort of bring an awareness of this initiative to you as a critical path
initiative and I'll focus on the industrialization dimension.
The
key aspects here I think, I hope you had an opportunity to at least look at the
executive summary of this initiative document or the White Paper that we issued
recently. The key focus area is
innovational stagnation. I think we are
trying to examine this and challenges and opportunities on the critical path to
new medical products. The finding that I
think as a nation both private and public funding for research and biomedical
research has been growing quite significantly over the years but the
translation of all that basic research to products for the patients seems to be
not in sync and that's what we were trying to examine and at the same time the
cost of new drug development seems to keep skyrocketing.
And
there are different figures out there, 800 to $1.7 billion and so forth. So from a regulatory perspective, I think
what we feel is the critical path which is from the prototype design to the
approval of that, is not receiving adequate attention from the research
community and even from the academic community and this could become or is
becoming a bottleneck to new drug development.
So the critical path that we have identified and defined is an area
which has not been receiving the attention with respect to new methodologies,
more efficient methodologies, and research in development of drug products and
medical products.
So
if you really look at it from a new drug development improving efficiency of
drug development and review, new development is a high risk and highly costly
enterprise and it's often due to the high failure rate that we see. And can we do better? And I think we must do better is the theme
that we are trying to move forward. There is a plan to issue a list of projects
that we thing are the high priority projects in both safety efficacy and
industrialization. And the feeling is
strong that the current process is not sustainable if you want to maintain a
robust pharmaceutical industry to meet the public health needs of the U.S.
With
that in mind, I want to focus on the three dimensions of the critical path
initiative, the one on industrialization which goes from the physical design of
the prototype to characterization small scale production, manufacturing, scale
up and mass production. If you really
look at the challenges we face today in conventional materials and dosage
forms, tablets, capsules and so forth, the functionality of exigence, the
availability of exigence, the characterization is still a big gap, we don't
understand all of those things, but as we move forward to its nanotechnology,
nanomaterials, the physics becomes more and more important and we are not able
to address physics adequately for our conventional materials, so a challenge in
the complexity is going to be much greater.
So
how does -- this is simply to sort of remind us what the current state is. Research and training needs from both the
national perspective as well as the perspective of FDA, I think, is clearly a
topic for discussion that we want to sort of bring forward and have it in a
public forum. The question I have is our
nation's education and research infrastructure, is it adequate to meet the
critical path challenges? To me that
answer is a clearly sounding no. And I
say that from two perspectives.
One
is before coming to the agency, I came -- spent nine years in teaching so very
familiar with the academic situation in the U.S. The society essentially has decided that the
role of pharmacy from pharmacists in the U.S. is going to be of a drug
information, patient care. So the
schools of pharmacy which used to have a program and the rigors of physical and
analytical sciences in those programs has completely be gone. So schools of pharmacy, the pharmacy
graduates coming out of schools of pharmacy in the U.S. actually often do not
qualify to fund the PhD program. In
fact, I would prefer not to use them because they don't have the physical
grounding necessary.
So
schools of pharmacy, the industrial pharmacy programs in the U.S. are really
incapable of meeting the needs of this nation.
And I've said that. Some people
have disagreed with that but I think that's -- I strongly feel about that. And I think there is a need to focus or take
our focus on more of a pharmaceutical engineering type of curriculum and there
is a need for center for excellence in pharmaceutical engineering, education
and research.
Now,
how do we sort of promote that? Several
schools have contacted us, schools of pharmacy and in collaboration with
schools of engineering have contacted FDA that they would like FDA to work with
them in developing such a center. And I
think we have a strong interest in that and we will meet and we are meeting
with these schools to see how we can support this. But clearly, the critical path initiative
document was intended to bring this issue at a level for public discussion, debate,
so the society can decide how well to fund this area because a lot of this
information, a lot of the science and a lot of the knowledge that needs to be
created has to be a public data base. It
cannot be a private enterprise.
So
I think I would like you to sort of think about and if you have towards the end
or right after my talk how should FDA support the case for a focused effort on
pharmaceutical engineering? We have met
with ISPE, International Society for Pharmaceutical Engineers, and politely I
said, there's not much pharmaceutical engineering there. So we need to bring more pharmaceutical
engineering in that and actually have a workshop on the topic of pharmaceutical
engineering and the national need for this focus. So please think about this and please share
your thoughts on how we should proceed.
We
will be meeting with schools of pharmacy and engineering who have interest in
this and try to explore this possibility.
I think clearly from an internal FDA perspective, I think next several
months we will have to put a research agenda together. We are right now focused on the Office of
Pharmaceutical Science on the industrialization dimension, so what are the
research and training needs of FDA?
I
think from a research perspective, we have been sort of collecting a set of
topics, projects, or topic areas for research and realignment of our research
programs and clearly the PAT research program that we have initiated, some
internally, some on collaboration, for example, the collaboration with Pfizer,
we are exploring other collaborations with other companies, too, that will be
part of this critical path initiative but we are, as I speak, have a group of
people meeting with NCI, National Cancer Institute, on looking at collaboration
on physical characterization of nanomaterials and physical and biological
characterization, so we're moving in that area of physical characterization of
nanomaterials.
Clearly
we have an interest in computational
methodologies. Office of
Pharmaceutical Science has a wonderful group of bioinformatics with respect to
toxicology. We have done some work with
respect to use of prior information and use of export systems in terms of
formulation but that has been limited.
There's an opportunity for that.
There's an opportunity -- actually, we are putting together a very
strong chemometrics group. We already
have a few people. We're hiring a few
more to include computational fluid dynomix (phonetic) and include all elements
that I think would be needed to bring a sound computational basis for CMC
aspects. I think our other aspect is
support for generic drugs, efficient methods for bioequivalence (phonetic) is
clearly one of the aspect, but I think as we move forward in the critical path,
I see blurring or actually increasing the challenge of what is pharmaceutical
equivalence and how do you define bioequivalence, so our focus will be on that
and in fact, we will have to probably take up the topic of what is
pharmaceutical equivalence soon because I think there is an opportunity to
align that and to streamline that and to actually make it more simpler because
today a tablet is not pharmaceutically equal to a capsule but if you put a
tablet inside a capsule, it's pharmaceutically equal. So we have logical ways of defining
this. I think we need to sort of pick
that up.
So
all of this sort of comes together as a research program that we have Mon
Surhan (phonetic) in the room. We just
hired him from the University of Texas and I think he and Cindy Busey
(phonetic) are focusing on the industrialization dimension. So this year's program planning for research,
I think, we will really focus on this.
Jerry Collins and others are clearly focused on the clinical side of
it. So here's an opportunity but you
also have a chance to sort of give us your thoughts, what are the project
topics that we really should consider, these are the broad areas that we are
working on and developing a research program to meet these needs.
Clearly,
the training needs are equally important, the pharmaceutical inspector training
program, the critical elements development that we start next month, the
training program, but also training of the CMC review staff that Moheb will
talk to you about. And I think we will
have to have a systematic way of doing that, especially if you have to
alleviate some of the concerns John Berridge raised and how do you address
these things.
So
just I'll stop here and put these two questions on your radar screen. Anytime you have suggestions and so forth,
please send these to us. Thank you.
(Applause)
CHAIR
BOEHLERT: Thank you, Ajaz. Any questions or comments for Ajaz? Ken?
DR.
MORRIS: Yeah, just a comment and this is
not news to Ajaz. I'll apologize in
advance for repeating it but to get it as part of the record, I think one of
the historical issues has been separate from FDA or industry and that is that
NIH and NSF just don't view the kind of research that we're talking about as
fundamental enough to be treated by them and they expect the pharmaceutical
industry to shoulder the burden of that and that's historically, I think, why
the departments, particularly at the graduate level, have had to abandon the
sort of research so that they could maintain funding in other areas.
So
I think that's -- not to just express regrets but to say that in the future if
we can bring pressure to bear as I know you guys have already talked to --
Helen and Ajaz both have already talked to folks in the other agencies, but if
we can bring pressure to bear so that they understand that significance of this
both financially and in terms of public health, can only help.
CHAIR
BOEHLERT: Any other questions or
comments?
DR.
SINGPURWALLA: Yeah, I was just going to
pursue the point that was raised by my colleague here. I'm just curious. Work that needs to be done which is of
interest to the FDA, why should the NSF put money into it? Am I correct in articulating that?
DR.
MORRIS: Well, I guess what I would say
is that the disconnect hasn't been that it's work that's needed by the
FDA. The disconnect has been their
recognition of this as a relatively fundamental set of research topics that
need to be addressed in general. I
think, just as we draw largely on material science and biology and the other
disciplines to bring into pharmaceutics, there are specific aspects of -- in my
particular case, of course, I'm narrowed by the scope of my research. For instance, if you look at material science
literature, very little of it deals with small molecular organic molecules.
So
it's not like you can go to the book and grab the fundamental theories to be
able to be used on these sorts of compounds necessarily. And they've just not historically recognized
the value of this and the broad significance of scientific endeavor.
CHAIR
BOEHLERT: Any other questions or
comments?
DR.
PECK: Yes.
CHAIR
BOEHLERT: Yes, Garnet?
DR.
PECK: Well, that's not what I was going
to talk about but I'll say it anyway.
Several years ago we applied to NSF a rather, what we thought a rather
good grant proposal to study the fundamentals of corn starch. And the only way that that grant was
eliminated was the fact that corn starch is not a uniform material. We were trying to find out why it wasn't
uniform and what kind of physical properties we could measure and we had a methodology
that was proposed but they couldn't fathom why we would look at this very
variable material, how important was it to our particular endeavors and at the
time that was the major disintegrating agent in most of our pharmaceutical
tablets.
We
simply wanted to understand it more. So
NSF turned us down and we did something else.
Concerning what Ajaz said, I have to be very careful, Ajaz. You may know of my feelings and some of them
are historical. I'm not convinced that
our solution is in pharmaceutical engineering.
If we consider basic engineering programs, at least the ones I'm aware
of, the amount of biological education that is provided those individuals is
very limited. You hit on something that
has to do with pharmaceutics and the fundamentals of pharmaceutics which gave
us those tools to bring along new drug delivery systems for the patient.
But
it was aided by this sensitivity to where the products were going. I'm having trouble right now coping with
pharmaceutical engineering programs.
There are so many excuses why we cannot open up the programs and that's
going to be a major, major hurdle with doing what is needed. As you have noted, we have to change in our
fundamental programs, in our graduate programs.
So you have identified the needs and that's great, but some of those
that have control over what we can do have to loosen up. That is a concern that I have.
DR.
HUSSAIN: I think your point is well-made
and well-taken that just an engineering approach is inadequate and not sufficient. I totally agree with that. And therefore, I think the pharmaceutical
engineering curriculum itself will have to sort of bring together the key
elements and not looking at that as a purely engineering discipline. It has to bring the fundamentals of
chemistry, biology, and engineering all together and that's something which is
not present in our curriculums in the U.S.
But
if I start looking outside the U.S., you see a very strong push for these
comprehensive programs, especially in China, and more so in Japan coming
through quite vigorously in a sense. So
I think the challenge here is this; the community, the pharmaceutical community
is a very small community. If you look
at the American Institute of Chemical Engineers, it's a huge community. If you look at American Association of
Chemists, it's very huge, but the subset that is interested in the
pharmaceutical industry is often small.
So you need to maintain that identity.
The industrial pharmacy programs and the pharmacy school programs were
successful in sort of meeting those needs, but now the societal needs and the
societal demands, supply and demand is such that look at BS degrees that you
have either in chemistry or even pharmacy BS degrees that Purdue has, you
create a scenario where the professional pharmacist and their salary structure
is so dramatically different so it's not sustainable from attracting the
strongest candidates to your program.
The
pharmaceutical engineering as a team provides a means to create that identity,
provides a means to create that resource structure and attraction for students
then focus on that. So we will have to
develop the curriculum that is needed to meet the needs. So your point is well-taken, Garnet.
CHAIR
BOEHLERT: Any other questions or
comments? If not, I'd like to thank all
of this morning's speakers. We are right
on time. We will break for lunch and
reconvene again at 1:00 p.m. I'd just
mention, members of the committee, we've made arrangements for lunch and Bob
King will be escorting us, right, to the place, our destination.
(Whereupon
at 12:01 p.m. a luncheon recess was taken.)
A-F-T-E-R-N-O-O-N S-E-S-S-I-O-N
1:03
p.m.
CHAIR
BOEHLERT: Good afternoon everybody and
welcome back. We're going to start this
afternoon with some introductory comments by Ajaz Hussain followed immediately
by Ajaz and his presentation. G.K. will
come after Ajaz.
MEMBER
HUSSAIN: I think the afternoon session
will hopefully provide the Committee with more information and more substantial
information to help answer some of the questions we have posed.
The
thought process of putting this afternoon session was to still take a look at
some of the opportunities and what we have been able to accomplish with respect
to the concepts that we have developed.
And
then have share some thoughts from Moheb Nasr and Gary Buehler because these
individuals are responsible for managing the day-to-day activities and some of
the challenges they face. And what are
they planning to do in helping us moved towards the "desired state".
So
you'll see different levels of activities while they're trying to manage the
day-to-day activities, how do we move towards the "desired state".
And
following that discussion, I've invited Ken Morris to come and speak to you
about his experience in helping us think about this and helping us move towards
what I would like to call a question-based review process.
And
I like that term because it helps us to hopefully focus on asking the right
question. Question-based review process
is actually in place in our Office of Clinical Pharmacology and
Biopharmaceutics. And I actually like it
quite a bit where you simply, clearly identify what are the questions to be
addressed; and then focus your review around those questions.
And
I think we have an opportunity in the CMC world to do the same. And so Ken has been working at it with the
CMC leadership within the Office of Pharmaceutical Science, debating,
discussing. So he, I think, is the right
person to share some of his thoughts with you before you get into your
deliberations and discussions.
So
that's the agenda for this afternoon.
After
listening to the discussions this morning, especially with respect to some of
the discussions with the design space and some of the opportunities, for
example, Dan, you raised the issue if you have understood this range of
conditions work fine, why don't you sort of take advantage of that?
And
I think that's what we are trying to do in Q8.
I see as Q8, from my perspective, is trying to harmonize different
regulations in Europe, U.S., and Japan with respect to changes or variations by
changing what is -- how we define change.
For
example, if you have a range of studies done, and you understand the range is
not critical, so why not redefine that as not being a change? So that's, I think, what we're trying to
achieve.
So
keeping that in mind, I'll share with you some of my thoughts on
specifications. Some challenges and
opportunities in the enhancement of CMC sections of NDA's quality by design is
how to set specifications.
And
I was planning to speak after G.K. Raju but I think listening to this might
help you because I think G.K. is going to talk to you about the wonderful
opportunity we have from the knowledge-sharing perspective.
So
with that in mind, I'd like to sort of again repeat that I think the
opportunity is therefore companies that acquire extensive understanding about
the product and manufacturing processes and share this with the regulators,
that helps us to be -- enhance our science and risk-based regulatory quality
assessment in setting specifications, reduction in volume of data to be
submitted replaced by more knowledge-based submissions, and flexible plus
continuous improvement.
In
fact, I think our goal is to move towards the state where you have one cycle
review, CMC review, and essentially everything is banished in the GMP
inspection site in the continuous improvement.
And
I think this is the desired state that is possible. And I think Dr. Woodcock's presentation,
which is in your handout, she has continued to think about this. And her latest presentation on quality by
design I think is quite telling in the sense this is a proactive approach on
how you approach the development and how you approach specification settings.
Quality
by design stipulates or postulates key performance parameters early in the
development process. Now this is based
on what we know at that point plus your prior information. And then you design product and processes to
be robust around for these parameters.
But
the challenge today, as John Berridge
discussed some of this in his presentation, without adequate product and
process development and/or knowledge sharing, you have high levels of
uncertainty with respect to critical attributes, what is critical and what is
not critical.
And
when you have that high level of uncertainty, we often have to make decisions
conservatively. If you don't know,
everything is critical then.
And
also I think the questions that we struggle with is is the sample size
representative, in representative disc samples and adequacy of risk
coverage? Example, compendial discs to
assure batch quality.
So
those are the regulatory risks or concerns that our reviewers are trying to
minimize through their approach to specification setting. In absence of extensive understanding of
product or process factors, you have to make a conservative decision.
I
think reduced concerned risk by covering all apparent attributes with
acceptance criteria based on capability of test methods and/or manufacturing
process plus very inflexible SOP that sort of follow from that. So that's the current way of sort of doing
business of current regulatory risk mitigation strategy.
But
I think I wanted to illustrate some of this using a current situation on
distribution attribute. Now there are
many, many guidances that sort of you have to look at to glean this information
but I won't be able to do all of that for you today.
But
I just wanted to sort of share with the ICH Q6A decision tree and how it sort
of addresses the resolution and why I think the current state tends to be
testing to document quality. And in ICH
Q8, we're moving toward the desired state where we are trying to get to a quality
by design.
The
biopharmaceutics classification system, the BA/BE guidance, the SUPAC guidance,
and the dissolution guidance itself are sort of interconnected. Unfortunately we don't have the time to share
with you all of those connections that I have sort of worked out.
But
let me start with the Q6A decision tree.
The first question that we asked in this decision tree does the
dissolution significantly effect bioviability?
If the answer is yes, we develop test conditions and acceptance criteria
to distinguish batches with unacceptable bioviability.
If
the answer is no, we go down this decision tree to say do changes in
formulation or manufacturing variables effect dissolution? If the answer is no, we go down to adopt
appropriate test conditions and acceptance criteria without regard to
discriminating power to pass all clinically acceptable batches.
But
if the answer was yes, are these changes controlled by another procedure and
acceptance criteria? If the answer is
yes, we come back to the previous result.
If the answer was no, adopt test conditions and acceptance criteria
which can distinguish these changes.
Generally, single point acceptance criteria acceptable.
Now,
I have inserted some questions.
How? How do we know dissolution
significantly effects bioviability?
Okay? There are wonderful studies
that are done in Phase I, Phase II which actually show you so much information.
For
example, one typical that is carried out is a related bioviability study
solution was established. We actually do
not use that effectively in our decision-making. Often you will see a solution perfectly
superimposable to a solid dilution essentially saying dissolution is not great
limiting. Okay? So we know that that happens in many cases
but not in all cases.
But
then the solubility, the particle size, dissolution rate, all can give you the
signal. We don't utilize that
information today.
So
often our answer is yes, dissolution is an important attribute. It is an important attribute and we have to
control it using a dissolution test.
And
the many questions that how, what, why, and so forth that you see on this chart
are not fully addressed but not only the information is scattered throughout
the NDA submission but also I think we often don't have time to pull all this
together in a concise way to answer these questions.
And,
therefore, we often set specifications because that is the tradition.
Suppose
I go down this route, dissolution does not significantly effect
bioviability. Should we be asking the
question do changes in formulation or manufacturing variables effect
dissolution?
Why
would we ask that question?
The
answer is for over shelf life, over the period of shelf life, there might be
change which might not be apparent in the release.
All
right. But then we establish a
dissolution criteria using a dissolution test.
So that's the current situation.
And
here are three examples, more recent examples of how we set
specifications. Now these are three very
recent examples. And here are the
reviewer comments, three different ideas.
Without
adequate product development and/or knowledge sharing, we debate
frequently. So one of the last decisions
we might do is this is your specification to the end of the NDA review cycle,
this is what our decision might be. And
you often have no choice but to accept it.
So
here is the first comment. The reviewer
recommends tighter dissolution specification.
Q of 80 percent in 30 minutes.
And in this case, it was based on, you know, what the clinical batches
were.
And
if you go down the list, they say the sponsor-recommended dissolution
specification method was unacceptable.
We simply say the sponsor's Q of 70 percent is too low. The direct product that releases only 70% is
likely to be bioequaled -- is less likely to be bioequaled than a product that
releases 100 percent.
Therefore,
we recommend a Q of 80 percent.
Sometimes that Q of 80 percent may not be actually a profile point in
this, the total example. Therefore, we
propose the sponsor's specification of Q of 80 percent at 60 minutes should be
changed to specification of Q of 80 percent in 30 minutes.
Much
of this discussion is based on three, four, five batches that we see in the new
drug development. We do not bring the
systematic thinking with respect to the physical, chemical properties of the
drug, the formulation, the disintegration mechanism of any of them. That's not really fully utilized today.
And
then what I say here is we have cGMP problems.
Here is a warning letter. An
inspection of your drug facility, blah, blah, blah, there is no assurance that
written production and process control procedure established for coating are
sufficient to produce a product that has the quality it purports or represents
to possess.
The
duration of coating cycle as determined by the pan operators is based on a
visual determination that coating solutions are even distributed before
proceeding to the next step.
It
should be noted that it was hundreds of batches. So the numbers are not small here. A number of batches made in `97 or `98 were
rejected due to in-process distribution failures.
And
then you go on to the partial release of various products even though there was
not data to invalidate all the specification results. And so forth, and so forth. This is catastrophic. And this is not a small company. This is one of the major companies.
So
what happens is, I think, every aspect of our regulatory is
interconnected. And G.K. Raju, his data
has always shown us in the sense that, you know, all the specification results
are a significant -- they contribute a significant increase in cycle times and
so forth.
And
here is a couple of examples that I took from his slides. But here also many of these are physical
attributes, dissolution. And what I
would argue is many of these are physical attributes that -- where we have
struggled with.
Dissolution
is not the worse case scenario. I would
say when it comes to particle size, gasket compacture, and others, you have
significant measurement variability that you have to deal with.
But
let's look at this. Testing to document
quality clearly requires a less variable test method. Here is the data from our lab in St.
Louis. The current USP 10 milligram
Prednisone caliber tablets exhibit slow dissolution over time. It's not a stable caliber. It keeps changing.
So
if the acceptable test equipment calibration limit is 28 to 54, and if you
often live with our F2 criteria, which is an average of six, and then you
compare the two average profiles, and that average profile should not be more
than ten percent different between the pre-change and post-change, what do we
see?
The
calibration limit far exceeds that but that's what we have been practicing for
years. So what can we say about the use
of F2 criteria where the mean profile difference that we accept is ten percent
or less as a way to document and change quality?
And
if you look at the table there, the table from two different data sets, the
shift in the stability of this calibrate, so if I look at calibration as a
means to say this is my target, that's giving me a target value, even the mean
estimate, the point estimate is questionable with this method.
And
just to summarize the dissolution experience at the FDA's Division of
Pharmaceutical Analysis, dissolution testing with USB Apparatus 1 and 2
requires diligent attention to details, both mechanical and chemical, dual
response can respond differently to small variation in apparatus setup or
degassing, large difference in dissolution results are possible unless all
parameters are carefully controlled.
The
experience at Division of Pharmaceutical Analysis, FDS St. Louis indicates that
differences in reproducibility can often be traced to improper mechanical
calibration and/or degassing.
And
we have a situation where we often have to reject, recall batches, because of
minor dissolution failures. And we have
no good means of getting out of that trap that I think we are in.
And
this is not new. Our Canadian
colleagues, Health Canada, has been talking about this for years. And this is from 1992. We often see false positives and false
negatives in some of our measurement systems.
And
here is just one example. I'm not going
to explain that but Ian Miggelri, and he used to be at Health Canada, published
this some time ago.
Now,
just to continue the thought processes that are so entrenched in testing to
document quality, and we often ignore all the prior information and we focus on
the test results, is a reason for thinking -- of major thinking.
Here
is another example from the Q6A decision tree.
I just want to illustrate two points from this. Now the question I want to illustrate from
this is do we always need a dissolution test for every solid dosage form? The answer is yes currently.
But
I think Q6A opened the door to say not necessarily. Although I'm not too pleased what Q6A
recommends, a disintegration test instead of dissolution, which is probably far
worse than that. I think there is a
better way to deal with this.
In
Decision 3, No. 71, it says the product is not modified release, the drug has
high solubility, the product has rapid dissolution, then you ask the question
has a relationship been established between disintegration and dissolution?
If
so, then you might want to go to a disintegration test instead of a dissolution
test. Now in Europe, this is
acceptable. I don't think we have
approved a single one in the U.S.
But
the point I want to make here is just pay attention to that. We're focused not on understanding the
product, not on understanding the process.
We are trying to create comparison between two different tests, a
disintegration test and a dissolution test.
And
the reality is this. If you are familiar
with the disintegration test, you have a cube with a 10 mesh screen which goes
up and down. So you put a tablet and
that goes up and down. And you just look
at the time when all the tablet fragments have passed through the sieve.
So
in this case, the table is disintegrating into larger chunks to small chunks at
a point where you stop and say the tablet has disintegrated. The total dissolution of that is throughout
the surface, larger particles, smallest particles and so forth. So dissolution can continue after
disintegration is over.
Now
the point to illustrate here is this, they're twofold. One, we are comparing apparatus --
dissolution apparatus to that of a disintegration apparatus. The hydrodynamics are different and the
medium might be different. That's not a
true comparison. That's not a quality
comparison per se. But that's fine.
The
other aspect here is, I think, the hesitation that we often have is now, yes,
there is a risk associated here by moving to a disintegration test because
dissolution continues even after the disintegration time is over. The reason a risk could be polymorphic
transitions. You may see polymorphic
transitions and a disintegration test might not ever pick it up, correct?
So
these are sort of the questions that I think with good science, what we have
talked about in Q8, we can address some
of these questions in a submission. But
not today because we don't have all this information to really make a rational
decision.
Testing
to document quality, the face has many dimensions. It is applied as in process and end product
release and stability testing. So the
reliability of specification is a key question because, I think, we look at
that in absence of process understanding.
Managing
post-approval change and continuous improvement is a challenge. And I showed you just one aspect of the F2
metric and what challenges it poses.
Product and process knowledge acquisition and generalization is also
challenged because now you are relying on a traditional wet test to -- and if
you're trying to do a design of experiment, that's a humongous resource
commitment in the time it takes to do these tests.
So
how can pharmaceutical development knowledge help? How can we demonstrate quality was designed
in, specifications based on mechanistic understanding, continuous "real
time" assurance of quality, and flexible continuous improvement.
I
think the Q8, the Q9, and the overall Q10 are all trying to move in this
direction to answer these questions.
What I would hope to see, and this is for debate, discussion, and so
forth.
This
is the same decision tree that I showed you earlier from Q6A but now, from a
design -- quality by design perspective, dissolution significantly effect
bioviability, that's a design question.
You
postulate that based on the characterization of your API or drug
substance. You know the solubility and
you know the pKa, you know so you have a knowledge based on how this molecule
might behave. And so you postulate.
And
then throughout your development program, you confirm based on mechanisms
and/or empirically. So that product
design applies to those two decision trees that you have.
At
the same time, from a risk perspective, if we understand the PKPD of this, we
will have a better focus maybe towards the end of the drug development process,
not at Phase I, Phase II, but towards the end of the NDA submission process,
what is acceptable? What is not
acceptable bioviability?
Today,
the answer is anything outside of 80 to 125 is, by virtue, an acceptable
bioviability. And that's a wonderful
clinical pharmacology question of what that question is.
So
once you have that, you start answering this question, design for manufacturing
and controls or design of manufacturing and controls and how reliable are these
because the second decision diamond that you have, do changes in formulation or
manufacturing variables effect dissolution?
Right?
If
the answer is yes currently, are these changes controlled by another procedure
or acceptance criteria? If the answer is
yes, you still go back to the dissolution test step. My answer to that question would be is that
really necessary? With this scientific
knowledge base and so forth, can we do better?
So
those questions can be brought to bear on this.
An overall, risk-based CMC, why can be asked. I think a reviewer should ask why do need
this? Why do you need redundant
system? What is the value of this? And so forth.
But
also I think we need to find ways to answer the question so what. Now if the virtue of, I think, what I have
learned from a quality system is you have to focus on the voice of the
customer. Now if dissolution does not
significantly effect bioviability, if a drug is highly soluble, and this is a
rapidly disintegrating drug, is that a critical variable?
I
think today we'll answer always yes.
Dissolution is an important attribute, no doubt about that. But a test?
Is that important? I think we
have to start thinking about so what?
And the so what has many, many connections. What is acceptable? And so forth.
I
think overall CMC systems approach that Moheb will talk to you about, I think
he's starting to think about this as a quality systems assessment program. And it's to sort of bring the connections and
the Q8 offers that opportunity to link the morphic form particle size stability
failure mechanisms to ask those questions why and then how.
So
based on quality of pharmaceutical development knowledge, can we not evaluate
overall CMC systems approach, that is link to morphic form particle size
stability failure mechanisms and address the concerns and risks? Is dissolution specification needed? Instead of wet dissolution test can we use
disintegration test?
I
don't like that personally but that's a valid question. Real time release and stability based on
process controls and say NIR tests, capsules and so forth.
The
key is, I think, we all understand that not all information is mandatory. We are okay with this. And we are work in the ICH to avoid a two
different system model. Instead we are
moving towards one system with different levels of quality by design.
And
you'll see that, I think, in different offices you'll have different levels of
process understanding. And so forth.
The
challenges we face. Common approach to a
more clear articulation of not all information is mandatory. We seek your help on that, I think, in the
questions we have posed.
Improved
process understanding and control technologies may afford reduction in
regulatory requirements. That's the
design space concept that that's coming about.
And
I think the key is and in most relationships it is expected between
effectiveness of the quality by design and risk to patient being exposed to
product that is not fit for use. That's
something that will need to evolve.
And
I think what we are moving forward is hopefully ensuring continuous improvement
and a process for continuous learning and updating of this knowledge base.
So
with that, I'll stop. And I have --
invite G.K. to share his thoughts on it.
CHAIR
BOEHLERT: Are there any questions or
comments for Ajaz?
Yes,
Nozer?
MEMBER
SINGPURWALLA: Ajaz, I'd like to make a
comment.
MEMBER
HUSSAIN: Yes?
MEMBER
SINGPURWALLA: Just to keep the notation
and the language clearer and clean. What
you have is not a decision tree. What
you have is an event tree. A decision
tree is one where you make a decision.
What you have is a flow of event as they occur.
So
just so that we don't, in the future, confuse, you should really call it an
event tree.
MEMBER
HUSSAIN: Unfortunately, I can't change
the ICH.
(Laughter.)
MEMBER
SINGPURWALLA: Change it.
CHAIR
BOEHLERT: Any other questions or
comments?
(No
response.)
CHAIR
BOEHLERT: Okay, before we begin with
G.K., I would just like to note for the record that there is no open hearing
this afternoon because, indeed, there were no people that requested time.
So
having said that, G.K., it's all yours.
MEMBER
RAJU: Thanks, Judy. And thanks, Ajaz, for the opportunity to
present today.
I'm
going to try to talk about manufacturing science and knowledge and in some ways
build on what I presented before in this general audience. And I think in many ways compliment the
presentations of here today.
The
outline for my talk, I'm going to frame manufacturing and science within a
broader social context. Once I have a
frame, I'll define some vocabulary.
Hopefully, Nozer will approve of the vocabulary. And use that vocabulary to then describe the
desired --
CHAIR
BOEHLERT: G.K., G.K., you may need to
get closer to the mike.
MEMBER
RAJU: Sure, okay.
Once
I've defined the current and the desired state, I then use that vocabulary to
define leverages to go from here to there.
Implications of those leverages, possible next steps given those
implications for the leverages and, of course, acknowledgments because we stand
on broad shoulders.
The
frame that I'm going to use for the rest of my talk is to say that
pharmaceutical manufacturing is not really something you do inside a plant in a
company. It really is a social
capability that has resulted from a set of choices that we have made, all of us
as patients.
So
all of us here are patients. But all of
us as patients have made decisions about risk, what is a good release? How does it work? And how much are we going to pay for it?
The
government, who has decided to fund certain kind of research, and if Ken wasn't
happy that they didn't fund other kinds of research, the pharmaceutical
industry that has decided to focus on product innovation and in doing so has
made a tradeoff about process innovation.
And academia, who has decided to have all their tenured professors to
focus on everything except pharmaceutical engineering.
(Laughter.)
MEMBER
RAJU: And so all of us are stakeholders
in this broader society as if we could go with what Ken said. And in the end, inside that plant, in the
broader social structure, somebody is making these drugs that we consume.
So
I'm going to try to frame it in that sense and now let's look deeper with that
frame. Given that frame, let's define a
vocabulary. The first set of vocabulary
is around science. It was the first
thing that Ajaz wanted to include in my talk.
And
interestingly science is both a noun and a process, in some ways something
active and it's doing. And there is the
process of scientific inquiry. And there
is an extent of science which is what you know at any point in time.
Given
that, and we've defined manufacturing science in the past, you can then go to
the next word in your definition and say once you have science down, how about
the word system. A system is a set of
processes and broader systems, including people, with a common material and
information flow.
The
way I define system, the manufacturing system is very much connected with the
quality system. They're not two
different things. They're almost the
same thing although there are reasons to be different in that particular
industry. So the second piece of
vocabulary is now in place.
Here
is a set of manufacturing systems that you could have. I'm going to call them A, B, C, D, and E. That's pretty obvious. That's how I learned the alphabet. And given these classes of manufacturing
systems, let's look at what our manufacturing system looks like as we go to the
rest of the talk and move forward.
The
third piece of vocabulary is the word capability. I'm going to define given the frame that
pharmaceuticals is a social capability to have manufacturing capability to be
defined consistent with that frame.
Manufacturing capability is the ratio of the voice of the customer to
the voice of the process.
And
we, as a society, have focused on how much, what is important to us in terms of
one, the patient who says what is important to him is safety, efficacy, and
availability, the regulator who we as a society decided that their role is to
assure that safety, efficacy, and availability, the head of that operation who
only wants to do better because that's how his job is really about, the CEO who
focuses on not only the effectiveness, that is he wants all of these customers
to get what they want but he also wants to do that with an efficient allocation
of resources, and the scientist in all of us, not just the academic who simply wants to understand
because that's just the reason why he exists.
And
so we have a hierarchy of customers, each of which has a voice. And we as a society decides which of these
voices will be heard and we invest. And
we make the investment.
What
shows up after many, many years, is the voice of the process. That simply said, this is what you've
invested. This is what the society is
giving you back in terms of its inherent variability of its process.
That
then is the manufacturing capability in the world, in the United States, in our
group of industries, in our segments of industries.
With
those three pieces of vocabulary, which is manufacturing science, manufacturing
system, and manufacturing capability, let's now define where you want to be in
the context of this desired state that we heard five or six times on the
previous slides earlier today.
What
did we say the desired state was? We saw
the FDA desired state. And we had the
industry come up and say you can put industry here. We want the same thing. What is that same thing? That same thing is that we give the customer
what he wants with a deep amount of understanding in our designs to make sure
he always gets it. So we're not even
worried about him any more at the bottom of the pyramid.
It
says we understand the mechanistic basis why something happens and we try to
understand the first principles of that knowledge. You can argue that this is an unreachable
state. We're still trying to find out
first principles. We believed in Newton,
in Isaac Newton. Here comes all of these
new things with nanotechnology that says maybe Newton misled us. But at least he took us so far.
So
this is an evolving thing. It's about a
domain. It's about a set of
questions. This is the first principles
for pharmaceutical manufacturing as we know it.
The
desired state is dynamic. That is we
want to be at that level in society but we get to that level one product at a
time based on the product we're making now.
And that's the development process and that's the continuous improvement
process.
Strategically,
you'd like to have society have laid the foundation of that knowledge so that
you already start with the generic mechanistic understanding, understand the
basic causal variables. You adapt it to
your own new drug.
You're
already starting so high and then you do a little bit of development here and
then you're at that level. You should
have no supplements to file. That's the
design space that you saw in the earlier presentation from Ajaz.
The
other alternative is to say society has laid all of that foundation great but
I'm not going to invest in going too high too far ahead of time because this is
enough to ensure safety and efficacy.
I
am going to work with my commercial plant and I'm going to continue to
improve. Because the basic foundations
are in place, I may or may not have to make any submissions even in this case
because the foundation of mechanistic knowledge is available in the greater
social structure.
That's
where we'd like to be. If you translate
that's where we'd like to be from a knowledge point of view into what do we
want our manufacturing system to be, I'd like to argue that we'd like to have
much simpler processes.
Today,
much of our processes look like System B.
What we'd like is processes that have few steps. They have a lot of automated control. And maybe we won't even have to do the final
product release testing if we've laid the foundation of knowledge that has been
institutionalized into our system and shows up in our capability.
The
current state, however, seems to reflect -- at some point this is personal
opinion, of course, that the level of our knowledge in pharmaceutical
engineering is at a basically correlative and descriptive level.
It's
a consequence of the broader social investment in it that shows up in academia
and, therefore, in research, and a greater industrial investment and a customer
prioritization about what he wants in a pharmaceutical and its regulation and
what he thinks the FDA should do if it needs to exist in the first place.
Given
this is where we are from a knowledge point of view, what is the dynamics of
that knowledge? The dynamics of that
knowledge is we stay at that level of knowledge and we stay at that level of
knowledge from the beginning to the end.
So this is what I call a social structure that has a learning
disability.
And
we need to overcome this learning disability by saying from a system point of
view, this is what our system looks like.
We have a system where the causes are far away from the effects. And we can't correlate them. And so we can't get to causality and so we
can't climb this family of manufacturing systems.
We
spend 25 days testing here. And we have
a cause organization that is separate from it.
We need to transform this system which is the result of social decisions
made in the past.
What
is that transformation about? It's about
two choices which really are about when you do that transformation. You could do that transformation in
development, which is the strategic leverage, which is learning before
doing. You do all this improvement,
change your manufacturing system to be E or D during development.
Or
simply the other alternative is to do that during commercial manufacturing if
we've laid a body of knowledge already in place, you might still be able to do
that.
But
what shall we do today when we haven't laid that body of knowledge in
place? And our processes look like
this. And we all agree on the desired
state that we want to look like this.
What are the leverages that make us go from this unsatisfactory position
to here? And you saw Ajaz present the
benefits of getting to this higher state.
That
is how are we going to all work together during manufacturing or during
development given this body of knowledge to climb up this portfolio of
manufacturing systems? Why is it
important to do? One of the leverages --
and I'll take one leverage.
In
this case, I'm going to choose the tactical leverage instead of the strategic
leverage, which, I think, you've heard a lot about in the morning. Let me talk about the tactical leverage.
That
is let's climb this set of manufacturing systems during manufacturing if we
can. Why is this important? A number of pharmaceutical companies have
warning letters. And what is the most
cited component of these warning letters over the last few years? It's about the quality. It's about investigations of the broader
quality system.
Let's
think about an investigation around some real data. Here is the solution. And as a broader social structure, you have
to first ask the question is this a critical quality variable? If we had asked this question and socially
invested in the answering of this question, I would have either had yes here
for this graph or I wouldn't have a graph because we wouldn't have this
variable.
But
because we didn't answer this question over the last 25 years, I have this
graph. And I have this question on the
graph. First question.
Second,
because I'm not even sure about this, the next question that remains, what is
its specification? If this and this had
been laid in place ahead of time, I wouldn't even have to show them on this
graph. So let's put them out because
they are strategic leverage questions.
Let's
ask the tactical leverage question. The
tactical leverage to climb up the pyramid, not the question that's about
releasing a batch. I'm not asking the
question should you release a batch. The
answer to that in today's vocabulary is easy.
I'm
asking the question if we are going to use knowledge as a basis for changing
the way regulation is done in our social structure because we can't pay the
price for it, we've got to climb up the knowledge pyramid and here are the
questions we have to answer for ourselves to be able to climb that knowledge
pyramid.
First,
are these data representative of the underlying reality? Is this the solution really the dissolution
of the one million capsules it's meant to represent?
As
part of that, there's a sampling question but it's also a measurement question
that Ajaz talked about. Is that
measurement an appropriate measurement of dissolution and the way it was done? So this is a sampling and testing
measurement.
Two,
have I seen this before? Learning
disability is about seeing the same thing and giving the same reaction and not
able to separate that you haven't understood and prevented it. That's a knowledge management question. That's about have I seen it before? Can I go back to a past answer?
What
is this variation? Is this somehow
inherently different from all of this variation? Or is this simply a little bit of variation
put together showing up in a general pattern that regresses that? Is this a special cause? Or is this just natural or common cause or
normal cause?
This
is the whole basis of SPC and Shewhart's theory where he spent many, many
decades of his life teaching us about how to answer these questions and how to
ask these questions.
Is
this process capable? Capable of meeting
which customer's needs? The patient's
needs? The regulator's needs? The head of manufacturing who simply want to
do better? The scientist's needs who
want to understand why?
And
then have we put in a place an effective, corrective, and preventative action
here so that this doesn't happen here?
If
we're going to use a knowledge-based and science-based approach to
manufacturing in the future, then answering each of these questions should be a
piece of science just list each of the clinical trials and their publications
are pieces of science. That is if we are
to climb that pyramid, it should be based on building blocks that have
significant scientific quality.
Small
scientific studies about sampling and testing not for release but for process
understanding. How much you sample and
what should be your measurement technology to climb that pyramid? And you are going to come up with very
different questions when you're asking the climb the pyramid question versus
release question.
How
do I know? What is the body of
knowledge? What is the scientific study
that I have to put in place to say this is special cause variability versus
common cause variability?
What
is the basic building block of science in this overall pyramid that allows me
to put in place effective, corrective, and preventative action that makes me
climb up to System E so I don't see that again.
And
the bigger questions that I put in a different color are how do we answer and
put pieces of science ahead of time in the development context? Investigations, small leverage in
manufacturing. And that's 90 percent of
our products today. We must focus on the
strategic leverage one part at a time.
But
the opportunity in manufacturing is to build these blocks of science around
investigations, around technology transfer, around process
characterization. And this is the basis
on which we get regulatory relief. But
beyond that, satisfy the higher customers in our overall social capability
structure.
This
is one way of climbing up this family of manufacturing systems and reaching one
that is much more independent of the broader social structure, much more
independent of the operators as this is highly automated.
And
that is the basis for completely eliminating any of those warning letters or
even having to see the investigator because no one want to really see him.
Implications
of the vocabulary and the leverages are first the vocabulary provides a
positive position. It doesn't matter
what word you use but if you use the word science, the customer likes it, the
regulator likes it, the patient loves it, the government likes it, maybe NSF
doesn't like it in some cases. They all
like it. It's a positive word. And so is capability.
It's
an enabling vocabulary because it's something that's so general. And we all like good science. And it's all about a broader community of
understanding that I think it is the basis of collaboration among all these
four stakeholders. To work together for
this broader social structure of understanding.
Three,
it's a basis for a very different relationship with the regulators. If you think about academia as saying let me
start with some general glass beads instead of reality and try to understand if
I can explain reality that is starting with first principles and trying to see
if they explain any data and really industry that starts with today's data and
try to understand it better.
Causal
knowledge in the middle is the middle of the top down and the bottom up
strategy that says let's look at using some of these research exemptions and
these safe harbors that are put in place in the PAT guidance to really work
together between the regulator and the regulated to truly understand the root
causes in these investigations including bringing in new measurements to do
that.
In
doing so, that would lay the foundation to climbing up the pyramid and making
of the regulator quite irrelevant. But
while doing so, this is the opportunity and the guidance is an opportunity to
start going deeper than today's root causes.
And
guess what? That fits perfectly, that
vocabulary fits perfectly with the current momentum around the FDA, cGMP in the
21st century. The critical path takes it
further as well.
Not
only is this one of the components of their four-pronged components for the
21st century, but it is the fundamental basis for risk. Risk analysis is a scientific process. It is a fundamental process through
manufacturing system for modern quality management techniques and science.
And
you heard the Q8 and the Q9 discussion.
What did they say? They said we
can get this a lot more harmonized. This
is a lot more difficult to harmonize.
Remember
what Fred said? Science is the
underlying theme that is also going to be the more powerful framework in which
to harmonize because of the very reason that everybody has a positive, enabling
view about it. And this is a very
powerful foundation that the FDA has laid.
Five,
science is a basis for the collaboration among competitors. It's very difficult to climb that pyramid in
development when you are in a hurry to push out a product. You are always going to hear every company
say that.
What
is missing in that conclusion is the presumption that you can't learn from all
your past products and you can't learn from all the other companies that do the
same set of things again and again. That
is can you learn through science and publications about excipients that are
more than 50 percent of your products that you all share?
Could
you learn from the fact that you've been doing this for 12 years in a row? And can you capture that knowledge which is
your priors?
Science
to collect to your past and to collect with your competitors to get out of that
dysfunction that says I only have a year so I can't move up the pyramid. You only have a year in the boundary that
you've drawn for yourself.
And
finally, science is about going into the very process that gives us all the
rewards that we want as regulators. It
is the benefit. It is the fastest way to
generate the products that we need. It
is the basis for true process understanding, for the academics, for the
regulators, and the broader CEO to ultimately get back his economic rewards as
well.
Those
economic rewards lay the foundation for enhanced manufacturing capability that
allows all of the different stakeholders to achieve all their needs, that is
the voices of the different customers, and lays the foundation as a social
structure for a complete reversal of where we spend resources.
If
you go back to the last 25 years and you look at where we, as a society, are
spending resources in terms of QC and QA and regulatory people, and the FDA,
and the investigators, you could say maybe the qualitative direction is clear. Maybe the units are tough to figure out. This is clearly on the wrong track.
And
when we design it then, which is quality by design, let's spend the next 25
years reversing back, go back to the same basic level so that all these
resources, including the industry, can focus on bringing in new products.
The
next steps for the next 25 years, given that the cGMP initiative is coming to
its two-year cycle and an end in a month that is based on many years of history
before that, first is to broaden the shared vision. We saw the FDA put us a vision. We had the industry come back and say I agree
with that vision.
We
can now connect this vision to the CEOs.
If this is a social capability, how are we going to bring them into
this? With the government, which might
impact decisions about funding, for example, a long-term social map.
At
this time, we have good intentions.
We're beginning to have a common vocabulary. The PAT guidance is a guidance but we need
much more of a map into the future. A
lot more of science and knowledge has to be characterized. And the implications are there in terms of
benefits, rewards.
And
what do I do next has to be clarified over the next few years in the real economic
case. And I believe that could be the
basis to broaden the shared vision and maybe get funding at a social level for
some of this research that is badly needed and has been for a while.
Something
that came up earlier today, we need some real case studies. In terms of the PAT submissions that are
coming, they're still let me just test the waters, in my opinion, however
little I know about it. Let's do
something real now that we've trusted each other and we've learned to trust so
that we can really turn things around in the next 25 years.
Besides
case studies of real data and the fact that I presented those slides to you
shows that I'm willing to go as far as I can but I'm not somebody who generates
these data and they can go further than me.
Pilot
the future. Just like you have a new
Medicare, a Medicaid program that's piloted in a state before you push it out
to a broader society, pilot something about this science-based manufacturing,
knowledge-based manufacturing into the future where nobody loses. It's a fish bowl for the broader society.
And
I know a number of academics who would probably play a lead role in that. And I've thought about it as well.
Acknowledgments,
of course, I must start by acknowledging the Consortium for the Advancement of
Manufacturing that has funded a lot of my research. MIT and Purdue, Ken Morris is here. I stand on big shoulders which Charlie and
Steve as well. And Janet, Helen, and
Ajaz, who have been an unbelievable help for society. And I've really benefitted from all.
And
if you just look at this list, you can see that it has got industry, academia,
and regulatory. You can't do it without
all four of those -- did I count -- I missed one. I didn't work enough with the customers, I
think, because I am one.
Bottom
line, to end, I introduced a frame that said it's a social capability. And what we see today is the result of the
social choices, of all of us together equally responsible for the good and bad.
I
said science, system, and vocabulary are three words that we can all share to
describe the desired state and the current state. Given that we seem to agree on the desired
state and we seem to agree that the current state is not satisfactory, we had
to then talk about leverages to go from here to there.
I
took one case, a very tactical case, and a strategic case would be actually a
much more powerful story, and let's say investigations is one of them. And you could take technology, transport, you
could take characterization. Let's build
a body of science around it, science of processes to climb up the pyramid.
What
are the implications? And what are the
next steps? And, of course, thank you to
all those who have helped me along the way.
That's
my talk.
(Applause.)
CHAIR
BOEHLERT: Thank you, G.K.
Are
there any questions or comments from members of the Committee? Yes, Kenneth?
MEMBER
MORRIS: G.K., as the sort of keeper of
the statistics in general, are there any estimates of the number of
non-value-added tests, real or perceived, that we do in the course of releasing
material?
MEMBER
RAJU: First, tests are non-value added.
MEMBER
MORRIS: Right.
MEMBER
RAJU: If they're designed in, you don't
have to do the tests. So that's the
amazing part. Even if you count the
tests as value added, by most computations in the literature, about five
percent on a time basis is value added in our industry. Ninety-five percent is non-value added in all
the paperwork and all the waiting time because we haven't designed in the
quality. And that's because of our
social investment or the lack of it.
There
would be a time when the number would grow if you include testing but let's not
even go there. Let's go over the body of
knowledge that we have to put in place.
And we deal with the consequences but maybe we said I'd rather fund
genomics than this. And we deal with the
consequences of making that choice.
CHAIR
BOEHLERT: Any other comments? Questions?
(No
response.)
CHAIR
BOEHLERT: If not, thanks, G.K. for a job
well done.
MEMBER
RAJU: Sure.
(Applause.)
CHAIR
BOEHLERT: We have a speaker with two
ovations so I don't know what that means.
(Laughter.)
CHAIR
BOEHLERT: You know the next topic is
risk-based CMC review and we're going to look at it from two perspectives, the
Office of new Drug Chemistry and the Office of Generic Drugs. And first Moheb Nasr will be speaking on the
ONDC perspective.
DR.
NASR: Good afternoon. Can you hear me okay? Can you hear me now?
(Laughter.)
DR.
NASR: I don't know why I'm hear.
(Laughter.)
DR.
NASR: I think we'll find out
collectively. I think many presentations
were made this morning that very much convey why we are here. I think we talked about the principles behind
Q8 and Q9. Ajaz articulated his vision
of the desired state.
And
G.K. did his always wonderful job even though he did something I asked him not
to do and that is his insistence in using pyramids. I think being an Egyptian,
I'm entitled to use of pyramids but G.K. always uses pyramids.
What
I would like to do today is to share with you where we are and where we are
heading. What I'm sharing with you is a
roadmap into the future. Without any
exaggeration, I think we are changing the paradigm of how to assist quality of
pharmaceuticals in the U.S. and in the world.
I'm
going to share with you where we are, why we are changing, some of the
high-level thoughts, and by the end of my presentation and Gary Buehler's
presentation, our combined effort, hopefully we'll illustrate to you where the
Agency is heading. And then we can open
the floor for discussion and seek your input.
I
will appreciate hearing from you all after my presentation because we are
working at a very fast pace in order to make this change happen. And we would like to make this happen in a
matter of weeks and months, not years and so forth.
These
are the topics that I will try to cover within 25 minutes but Gary and I have
an hour so I may use a little more time, Gary.
I
would like to share with you where we are.
I would like to update you on what we had before, which we called the
CMC risk-based approach or initiative. I
want to tell you that we are changing from chemistry review into a new quality
assessment paradigm and describe to you what I mean by that.
I
would like to summarize in a few slides the difference that I see between
chemistry review and the quality assessment.
And I would like to share with you some of our pilot programs and
supplement review and so forth.
CMC
review, as we all know, is intended to assure the identity, purity, quality,
and strength, an potency as related to safety and efficacy for drugs throughout
their life cycle from IND to NDA, most of all through the ANDA process.
This
is an organization chart of ONDC. You
see how simple it is. We have about 130,
135 review chemists and scientists spread out through 19 chemistry teams co-located in 15 clinical
divisions. It's very difficult to manage
such an organization. We are not
managing well.
I
hope in the future when I come next time, if Ajaz invites me, to share with you
our new organization and how it will not only compliment the future product
assessment but manage the losses within the agencies much better than it's
being managed today.
This
illustrates how much work we do in the office.
The in the last fiscal year, we reviewed 159 NDAs. We had close to 1,000
INDs. We had about 2,000 supplements. That's a lot of work. And if continuing in that direction, we are
going through a viscous cycle for when every time we approve a drug, the number
of the supplements increase, our workload increases, and we create a problem
not only for ourselves but for efficacy in the public as well. And there is a crying need for a change.
To
summarize our current CMC review practices, when it comes to the application
that we receive, the quality of this application varies considerably. Some are much better than others.
The
applicants don't always seek consultation and meetings through the review
process or follow some of the recommendations that we make and agreements we
make during the review process and during the submission.
And
sometimes they have, sometimes they don't have, but in many cases they do not
provide enough pharmaceutical development information that I consider to be
essential in order for us to do what we call risk-based CMC review.
What
about our review? We evaluate all CMC
information and data that comes in the application without doing too much as
far as differentiating between what is critical and what is less critical.
We
evaluate all the information that comes to us.
And that evaluation does not necessarily utilize the vested training and
background of our reviewers. Basically
we have one CMC reviewer, for most part a chemist, who conduct the entire
evaluation.
And
if you don't have enough knowledge, they try to do the best they can. They are trained while they are doing the
review. And there is good mentorship
throughout the process. It's a value
list-based review. I think someone today
called it a check-list review. It's not
really a check-list but it's a value list-based review.
We
don't do enough in-depth review of process information and that's in part not
totally because of the center field agreement.
We have tight specification, I have to admit to that. But the specifications are set based on the
limited data we receive.
This
is the information we get, and based on that information, we set the
specification with our goal is to assure that consistency of manufacturing
process. So basically the specification
is a way to control the manufacturing process.
Often
we have late and voluminous CMC amendments that lead to delay in review. And as you all know, we have problems with
the cycle of review and approval.
The
decisions are made based on submitted data and the individual experience. There is a lack of critical information
pharmaceutical development. Guidances,
for the most part, are established to provide regulatory relief but at times
create an increased number of supplements and that creates problems for us at
the agency and for industry as well.
What
are the problems with the current system?
For us at the agency, it is very resource intensive. You have seen our organization chart and you
see the workload. We have to deal with
recalls and drug shortages at times.
For
you all in the industry, there's a perception that because of the existing
regulatory system, it discourages continuous improvement. Regulatory burden, what's the value of all
the supplements and all the review we do?
And what is the consequences of being out of specification that require
investigation, recalls, 483s, warning letters, and so forth.
What
about the public? High cost drugs maybe
and delay in drug approval at times.
In
the middle of this, with all what we are doing, with all the problems, we are
facing some major challenges. In trying
to outline these challenges in this slide here, we have the GMP initiative
which, I think, many of us agree is really a product quality initiative for the
21st century.
How
can we fit the existing regulatory system into the new way? How can we do that? There is a conflict. How to deal with first cycle approval? The heavy workload. How can we address the consistency issues and
problems and difficulties that exist among the 19 chemistry teams in 15
clinical divisions?
We
are attempting to do that through the guidance process. It helped some but created different kind of
problems.
We
have problems with the guidance and policy development. There is a lack of expertise in many critical
CMC areas, many sites of pharmaceutical development. We are dealing with novel, new delivery
systems, combination drug products, new technologies.
Because
of all these, what we have done before and attempted to do it with some success
is react rather than have a proactive proposal of how to deal with issues in
the future.
I
want to spend a couple of minutes talking to you about the standards of the
risk-based CMC initiative that started in the year 2000 and went on until last
year when I came here to this shop. That
initiative was evolved over many years.
It's
multi-tiered. If you look at the
initiative, it was outlined as a three-tiered process. When everything was said and done, it was a
five- or six-tiered because every tier split into two sub-tiers. We would start with Tier 1A and talk about
three years. So if you go through the
five-tier process, it would have taken us many, many years. That's okay.
The
whole initiative was product specific.
It addresses and deals only with what we are very comfortable with and
that's mainly synthetic drug substances.
Characterization must be done using traditional analytical techniques
that you can clearly see. It applies only
to very specific products such as immediate release or dosage and so forth.
That
initiative was intended to provide regulatory relief by incorporating
science-based and risk-based assessment in CMC review. But one thing that became obvious with the
GMP initiative is the relevance of that initiative with our new product.
This
is something that we have to deal with only for a small class of drugs and in
very special cases or if there is some merits for better utilization of
science- and risk-based to apply that for everything we do, from that
pre-marketing into the post-marketing.
So
now we are dealing with more progressive and expanded initiative that was focus
on the totality of quality assessment.
The risk-based quality assessment has a variety of advantages. And what I have done in these two slides is
summarize some of the excellent findings that were obtained after the PQRI
Conference about a year ago. The PQRI
Conference that Toby Massa co-chaired.
The
benefits of the policy assessment risk is the quality assessment for the
patient for the increased availability, faster approval, and the patient will
continue to receive our quality products.
So we are not going to sacrifice the product by -- that may result from
a reduction of regulatory oversight.
It's basically more focused on our regulatory process rather than reducing
regulatory focus.
For
us at the Agency, there will be more product and process knowledge that is
shared by industry, more efficient resource allocation, increased trust and
better communication. And for industry,
there will be more efficient science-based inspection, faster -- and you will
hear more about that. I think David
Horowitz will talk to you all tomorrow about the new paradigm in GMP
inspection.
There
will be faster, more consistent review, a potential for reduced regulatory
burden, ability for you to manage the changes without very strict regulatory
oversight from the Agency, focus our resources on critical issues, flexibility
to focus on what should be done not what can be done, improved communication
with the Agency.
And
I think that the striking element of what we are trying to do today is if you
look in the past, the Agency changes regulation. The industry we had. The industry raises the bar because of new
delivery system and newer technology.
The Agency react. But in this new
paradigm, we are working together in order to head in the right direction.
When
we talk about the new quality assessment paradigm, I would like to make clear
to everyone here today that this is not a single initiative to address one
dimension of a multi-dimensional, often complex quality assessment process. This is not a streamlining effort.
It's
a new paradigm of quality assessment for new drug applications. And Gary will share with you his thoughts
about generic drug applications as well.
But that covers for the new drugs the entire or the totality of quality
assessment from pre- to post-marketing activities.
With
that we have to change our vision and our mission. And that is part of where we are heading with
our new organization. I'm going to focus
here on a couple of things because I think -- I do believe that the vision and
the mission should clearly indicate to us, to our staff and to the public,
where we are heading.
Our
new vision indicates very clearly that this is a scientific organization that
services the center, the Agency, and the public through leadership and innovation and international
collaboration. I do believe in
international collaboration. I do
realize that we are dealing with global industry. And our efforts here have to be done under
the umbrella of harmonization with other international agencies.
As
far as our mission, we no longer continue to do chemistry. What we will be doing is for our office to
assist the critical quality attributes of manufacturing processes for new
drugs, establish what is the standards to assure safety and efficacy and -- and
that's very critical here and that's why we need to work together to be a
partner to facilitate drug development.
Some
of the future elements that we need to work on and we started working on our
assessment will start with a comprehensive quality overall summary. And I think you had some questions and some
comments about that this morning. And
that is something that we need to work on.
Review
practices should be based on good scientific principles. There will be considerable increase in
emphasis on manufacturing science. The
CMC review and the quality assessment functions we do will be critically
reviewed by our colleagues and staff and scientists at the Agency. And we must integrate our review functions
with the inspection. And that goes under
the umbrella of Q8, !9, and potentially Q10.
When
it comes to CMC's specification and there will be another time for a larger
group for another discussion about how we set the specification and why we set
it and how it should be set but the main principles are specification has to be
risk-based -- based on risk-based
assessment, clinical relevance, safety considerations, process capability,
knowledge gained from pharmaceutical development reports, and better
utilization of modern statistical methodologies.
There
is such a thing as regulatory relief.
Such relief will be provided based on the following three criteria.
One
is process understanding and control.
And that what you can share with us through the pharmaceutical
development reports, assessment throughout the manufacturing process, and your
ability, because of your understanding of your process, and your plans to
continue to improve the process. So
these are three criteria that has to be there in combination in order to
provide an assurance of your ability to continue to improve the process. One of these elements by itself is
insufficient.
Pharmaceutical
development reports may facilitate meeting for a cycle approval, science-based
specifications, risk-based GMP inspection and regulatory relief from
post-approval activities.
What
we do at the Agency is done by people, not by machines and computers only. And that's why it's very important that we
invest in our staff and provide the correct work environment and resources to
support our staff. So it's very
important for us to provide better work environment to our staff to facilitate
superior performance and job satisfaction.
During
the CMC restructure, we are in the process of reorganizing the office. The reorganization is intended to facilitate
the implementation of the new quality assessment paradigm. What I'm saying is we are not moving 15 or 19
offices from one place and put them in another place. The organization will be there for one
purpose and that is to facilitate the new paradigm and to facilitate the
implementation of the new quality assessment.
I
may come back to you later on on this one but I just want to give you heads
up. We are considering establishing a
CMC Scientific Advisory Board and some of the functions of this Board would be
to provide scientific consultation when needed.
There
is no way we will have enough expertise in house to address every regulatory or
scientific issue we deal with. The Board
will oversee the ONDC regulatory research program, restructure and modernize
the ONDC training program, and also develop regulatory science seminars.
We
are in the process of recruiting and hiring and training pharmaceutical quality
assessors with expertise in drug discovery, analytical chemistry,
pharmaceutical development formulation, and pharmaceutical engineering. I think there are so many people here in this
room, if you know of anyone whose is looking for a challenging opportunity, I'm
all ears.
(Laughter.)
DR.
NASR: We have several vacancies both in
the review side, on the technical side, and in management as well. And I'm serious of inviting you to help us
help yourself by sharing some of the talent that is out there that we need in
the Agency.
ONDC
is building a strong and independent scientific organization to better serve
the public and our internal stakeholders.
And if you see where we are today, we are co-located with the 15
clinical divisions.
Linkage
with clinical division is very important but it is one of many linkages that
must be there in order to assure appropriate quality assessment. So we will maintain the linkage with our
clinical colleagues but we will have to work closely with our colleagues in the
Office of Compliance and the Office of Generic Drug as well. And with industry and other scientific
organizations.
Our
re-engineering effort is intended to work on problems that have been identified
in order to meet expectations and to establish a modern equality with appropriate metrics to measure the
quality of CMC review and performance.
This
is very important here and we are working very hard to do that. It's very easy to have metrics to count
beans, how many reviews, how many supplements, how long it takes you to do
that. But we need to identify the
appropriate metrics to measure the quality of the work we do and that input of
our review into drug development. This
is something we need to work on.
Before
I go to these two slides, I'd like to remind you all that we have a very large
quota of competent, dedicated, hard-working scientists. But what I'm sharing with you today does not
necessarily indicate in a negative way that our organization is not functioning
well. But we are shifting our paradigm.
So
I want to describe to you where we are today and where we are heading. And I think I can best describe that in these
two slides.
Here
is what we do today. What we do is
chemistry review. This is not something
-- I've used a term that I intended that everyone is using that term around the
agency. The review is conducted by
chemists. There is extensive data
analysis in order to generate the necessary knowledge and summary reports of
CMC issues. That's what we do.
We
get a lot of raw data, stability data, validation data. We use -- we review everything that is
submitted to us. And generate summaries
in order to be able to have a story to tell about the product itself.
One
would question is it us who should be developing this story or is it the
industry or the sponsor who developed the product that they can come and tell
us their story?
It's
a guidance-based review. There is more
focus on chemistry and specification issues and there is less focus on process
and manufacturing. There is no clear
emphasis on what we consider to be critical CMC issues. We do not have a peer review process to
evaluate the quality of the work we do at the center or in the office.
Quality
assessment is a very different thing, assessments conducted by
interdisciplinary scientists, chemists, pharmacists, engineers, and others as
needed. There is more reliance on
knowledge provided by advocates and that includes pharmaceutical development
report and comprehensive quality overall summary.
It's
a risk-based assessment. It's not
everything. Focus on critical quality
attributes and developments to safety and efficacy and these are some of the
critical attributes that we must focus on.
It's a question-based review and there is a greater utilization of peer
review process.
I
want to spend the next two slides to briefly summarize where we are with some
of these changes we are making. You will
hear tomorrow from Steve Moore, a team leader in our office, talking about
comparability protocol.
I
think comparability protocol can serve as a bridge or linkage between the
existing system and the new quality assessment paradigm. And that's why it's taken us more time in
reviewing the comparability protocol guidance before we put it out because when
we put it out, we want to make it more useful and more meaningful and to
facilitate the changes that we are all trying to achieve.
Comparability
protocol utilizes and applies quality by design principles. It should facilitate continuous improvement
with risk regulatory oversight from the Agency.
It provides scientific basis for expecting, understanding, managing, and
addressing changes.
It
brings more focus of what is critical and what is less critical. It has a great potential for down-regulating
CMC supplements. The bottom line is with
the workload that I described to you earlier in the first few slides, we can no
longer continue to have a quality review of the large volume and that
application information we get within the existing system we have.
We
are exploring ways not only to down-regulate but potentially eliminate certain
types of CMC supplements that have many potential to adversary effect on
identity, quality, purity, safety, strength, and potency as they relate to
safety and efficacy. So we are looking
why do we have supplement? What role
they serve?
ONDC
is developing in our new organization ways to manage the supplement review more
efficiently to facilitate continuous post-marketing product improvement and to
provide more resources for new NDA review.
I think if we understand what you are doing and you share with us your
understanding, and we'll do that at the pre-marketing stage, we have great
confidence in your ability to manage your own change.
You
can go ahead and manage that. That will
provide more resources for us to be more of a partner during drug development.
We
have a pilot program for resubmitting the NDAs because we have to find ways to
reduce the resources and put the resources where they are the most needed where
a single CMC reviewer perform initial assessment. Initial assessment is being done in two
weeks. And relevant material are
requested.
An
assessment protocol is developed and then assigned to a primary reviewer. A primary reviewer will perform an in-depth
assessment as always done.
Streamlining
of resubmission will provide more resources for our original NDA review. Where I'm coming from is this, if from direct
resources and have enough and correct and enhance the level of communication
with the sponsors, that may lead to first cycle approval and potentially a
decrease of the number of resubmissions.
And
this slide here, this is my summary slide, this is my last slide, what I have
here on the left are some truths. These
are truths. We are working on
re-engineering supplement review, streamlining our review of resubmissions,
talking about quality by design for pharmaceutical development reports,
comprehensive quality overall summary.
The
re-engineering of the supplement will provide less regulatory oversight for post-marketing
approval changes and that may lead to more incentives for continuous
improvement. The same thing with the
other tools. They will provide more
resources. They will enable us to do
risk-based assessment. And there will be
less review time.
And
all this will lead or may lead to first cycle approval of new drugs. And putting all these things together, what
we will end up having is at the end better product available at maybe less
cost.
I
think I missed one slide. My last slide
that you didn't see, I would like to acknowledge Dr. Janet Woodcock and the
Steering Committee for providing a lot of insight, Helen, Ajaz, Chi-Wan, and
Guirag Poochikian for providing considerable input in this presentation.
Thank
you.
(Applause.)
CHAIR
BOEHLERT: Thank you, Moheb. You have some very ambitious endeavors.
Are
there any questions or comments? Gerry?
MR.
MIGLIACCIO: Well, I want to go back to
your CMC specifications to be based on, Slide 18. You say clinical relevance and safety
considerations, which obviously we all agree on. Then you follow that with process
capabilities. Can you elaborate? Those could be mutually exclusive.
DR.
NASR: I can elaborate but I think there
is time that will have to happen very soon, Gerry, where we will need to get
together. By we, I mean the Agency, the
sponsors, and others as well, to look at the ways we are setting specification.
The
way that specification are being set now is at times because of process
capability, that means if you can produce a product with a certain level of
impurity, that would be in the spec --
MR.
MIGLIACCIO: Right.
DR.
NASR: -- whether this is justified or
not. And even if that's not the spec,
what is the detection ability of a particular analytical instrument? We set specification at times because of
safety concerns for certain kinds of impurities because of some compendium
requirements.
What
I'm saying or suggesting in this slide that we have to exam all of these things
together in order to see how can we set specifications.
And
what we will end up having at the end of the day in my mind, and this is just
me and not the Agency speaking now, so I'm going to take off my FDA hat, is a
combination of all this. And it would be
more on a product by product basis rather than the more generic level of
setting a specification for all products, one size fits all.
So,
again, I did not answer your question.
But I think yes, many of these things are conflicting. And I think that's what you are saying. But we will have to look at all this -- two
weeks together and all these issues together to see how we can set
specifications in the future.
MR.
MIGLIACCIO: Well, just a follow on, I
mean conflicting yes but a highly capable process has generally very little
clinical relevance to slight changes in that process. And that's what the concern is is setting
specifications based on process capability.
There is no clinical relevance to that.
Secondly,
at the time that we're setting specifications, you have preliminary process capability. The knowledge base will increase
significantly in the first three to six months after commercialization. And so to base anything on preliminary
process capability is a concern.
DR.
NASR: I agree with you. And I'm not talking about specification the
way we do it now after the initial review of the NDAs. I think Toby talked this morning about
interim specification which, by the way, is something that we do now. It's not that novel of a concept.
But
what I'm trying to say in this slide that there is a crying need for us to have
a handle on setting specification. And
to have a specification that are most relevant for that particular product and
not use a specification as a tool to control the manufacturing process.
I
think what we have done before because we didn't know -- we don't know in many
cases how you are developing your manufacturing process and you know that,
Gerry, you know, the level of information vary from sponsor to sponsor.
We
try to have an assurance because we have our responsibility to the public that
the product that you will produce in the future have the same critical
attributes to the product that was used in the clinical trial. And that is by making sure that the level of
impurities, for example, are the same.
And even if they can be tighter, we tighten that so to make sure that
you continue to -- you have better control over your process.
Is
this the best way to do it? I don't
think so. But we will have to put our
thoughts together to see how can we set that in the future because what is
happening now in some cases is the specifications are too tight and they may
not be that relevant to clinical issues to start with.
And
that may result in disruption of the manufacturing, recalls, need for
investigations, and so forth.
MR.
MIGLIACCIO: Thanks.
CHAIR
BOEHLERT: It sounds to me like this is a
subject we might need to have some continuing discussions on because this whole
issue of manufacturing capability versus safety and efficacy is one I think
that drives industry a little nuts from time to time.
And
if you want to reduce the number of supplements, this may be an area that we
can take a look at because -- and you mentioned impurities. And it happens to be a subject that is near
and dear to my heart.
And
very often safety has been demonstrated at very much higher levels than are
approved as specifications. And if
something changes down the road, you shouldn't have to file a supplement if
it's well within those limits that have been established as safe as effective.
And
so I think it's a topic for a continuing discussion and an area we may be able
to relieve the regulatory burden.
DR.
NASR: That's a very good point,
Judy. Without stealing the thunder from
future events that will be taking place, we are currently working on having a
public workshop between the industry, the Agency, academia, and so forth, to
focus only on setting specifications.
And
all the issues I outlined on this slide what comes from analytical methodology,
from safety and efficacy, from clinical relevance, from manufacturing, all
these things will be raised because I think we need -- if we are talking about
the future paradigm and specifications that are more relevant and not one size
fits all, there is a need to do that.
And
we started the elementary discussions to get there.
CHAIR
BOEHLERT: Okay.
Ken?
MEMBER
MORRIS: Thanks. You know, Moheb, it hadn't occurred to me
until I saw it on your slide even though we've talked in general terms about
this, but in terms of metrics for determining the quality of the review process
in the future, do you have any ideas of what that is going to look like?
I
hadn't thought of it before you mentioned it but I can see whereas now you can
sort of count submissions or something like that, it's going to change in the
new system.
DR.
NASR: I think we started already,
Ken. Question-based review, the peer
review process that we instituted already.
And also we are looking in instituting a quality management system
throughout our new organization. Quality
assurance program and I also, as I indicated in one of my slides, am
considering the establishment of a Scientific Advisory Board.
So
I think we have several elements but what really needs to be done is to see are
these sufficient metrics? Are they
quantitative enough? Do we have a map
here where we can connect all these dots to have an overall system?
Once
concept that I've seen that's been used by other regulatory agencies, if you
wish, is sharing the review with the sponsor.
I mean if we are talking about scientific organization and dialogue
between industry and the Agency, how about if we share our assessment, if you
wish, and see how we can learn rather than judging the in-depth of the quality,
how can we learn from this to do a better job in the future?
CHAIR
BOEHLERT: Okay.
Dan?
MEMBER
GOLD: Thank you for a very interesting
talk. I think you're making a lot of
progress.
I
have a question related to an issue that came up during the last meeting of
this Committee where a representative pointed out that in Europe the quality
summary is -- it's a top-down approach to the review of the application. And they were pointing out that they thought
that in the U.S. it's a bottom-up review.
And that your reviewers are really not looking at the quality overall
summary.
Can
you comment on that please?
DR.
NASR: Yes, I can.
I
think, as you can see, that's one of the major elements in our future review
practices. Because of that, I spent
about two and a half weeks in Europe in April because what I've decided to do
is to expand my area of knowledge about other regulatory processes that proved
to be successful. And I went to visit
several national authorities and I participated in advisory Committee
discussions and so forth.
If
you are talking about the expert report which was used in the old system versus
quality overall summary which is currently part of the common technical
document, I can share with you the following.
What I'm talking about goes beyond the existing quality overall summary,
which has a very narrow scope.
I
think we are talking about more expanded
quality overall summary that has more pharmaceutical development
component into it. That's number one.
Such
a summary can serve as a summary because part of what we do now in our review
is creating the summary. So why don't we
have you, as a sponsor, as the one who developed the drug, provide us with such
summary?
And
then the focus of what we do is to be -- is to assist the critical areas that
in the application itself.
Number
three, such a summary will not be the only thing we review but it can be a
starting point to highlight what could be critical CMC issues that we expect to
see in that particular application.
And
then we will focus our efforts on critical issues but also since we have the
entire submission, we will go and be as detailed as we need to in order to have
complete understanding of some of these issues.
MEMBER
GOLD: So do I understand --
DR.
NASR: I forgot to add one thing if you
allow me. That also may require us
revisiting under ICH or under another way of how the submission is put
together.
MEMBER
GOLD: Do I understand you then to say
that if we put into -- if we submit a very good quality summary, this is going
to accelerate the review of the application and the more rapid approval of the
application?
DR.
NASR: Yes.
MEMBER
GOLD: All right. One second question if I could? I realize that the initiatives that we're
talking about are very new for the Agency.
Do you have any metrics that indicate the improvement using these
techniques that you have seen so far in terms of reducing application review
time?
DR.
NASR: I have some metrics and I'm doing
some experiments. As a scientist we have
to continue to do experiments. Some of
the knowledge I have is based on my experience talking to our European
colleagues. And when I talked to them
about utilization of quality overall summary and expert report, it does reduce
the review time. That's number one.
Number
two, we are currently experimenting with resubmission of NDAs in some of the
critical CMC review teams within some clinical divisions. And what we are trying to do is to start the
assessment process, as I indicated on one of my slides, by a high-level
evaluation of the application itself, and development of an assessment protocol
in order to -- before the assignment is made in order to facilitate the review.
That's
much better than the current practice where you have the many folders, as you
know, Dan, and you go through the entire review before you develop the entire
story.
I
think having a quality overall summary will facilitate the development of the
initial assessment protocol, if you wish.
MEMBER
GOLD: Thank you.
CHAIR
BOEHLERT: Any other questions or
comments?
(No
response.)
CHAIR
BOEHLERT: If not, Moheb, thank you.
DR.
NASR: Thank you.
CHAIR
BOEHLERT: From the Office of Generic
Drugs perspective, we have Gary Buehler.
DR.
BUEHLER: Thank you, Judy.
First
I'd like to thank Ken. Usually I'm last
to speak at just about everything I go to and somehow I don't know what you did
to someone, Ken, but thank you very much.
(Laughter.)
DR.
BUEHLER: It's really nice to not be
last. I was last at the GPHA meeting in
the wintertime. And I was right before
the golf tournament.
And
I started to speak and I heard all these cleats outside and everything. People were banging their bags around and
everything. So it's very nice to have a
nice quiet group here.
I'd
like to acknowledge Dr. Berridge's presentation. I have to say, Dr. Berridge, that was the
clearest explanation of this paradigm I've ever seen. I mean it was -- your slides were great.
And
actually I may be calling you for some of them.
After you see my slides, you'll understand but it was really a very
clear explanation of what we're trying to tell people today.
And
I have to admit there is a fair amount of repetition here. And I'm not going to be an exception.
Also
I have to say your English accent is great.
You know I am from Philadelphia.
I'm a Colonist. I haven't lived
there for 30 years but people still say I talk like a Philadelphian. And it's just so authoritarian. I'm hoping to be able to do this in that way.
Acknowledgments,
I have to say that a lot of my talk was furnished by Frank Holcombe and Vilayt
Sayeed. They're in the audience today so
if I say anything wrong, there they are.
Our
mission is really very simple. It is to
provide quality, safe, effective generic drug products to the American
public. I'm a nuts and bolts guy. This is what I have to do. And it basically is we have to review and
approve applications.
We
almost approved 400 applications last year.
That's what I do. And, you know,
this is a vision. This is a vision for
the future. And believe me we are fully
supportive of this vision in trying to make the quality of all drug products,
generic and innovator, better and the process much easier and much better for
both the industry and FDA.
But,
again, as you can see, my workload is increasing. And it has increased dramatically over the
past two years. In 2003, we received 449
applications. In 2004, we expect to
receive 566 full, original ANDAs.
You
don't -- I have about -- it's somewhere over 50 review chemists. It maybe 52 or 53. You don't need Bayesian statistics to figure
out that that is about 11 original applications per reviewer per year.
MEMBER
SINGPURWALLA: You'll get a better
estimate if you use that.
DR.
BUEHLER: Okay, thank you, thank you.
(Laughter.)
DR.
BUEHLER: That's a lot of work. We have a tremendous amount of work. It's increasing. It's increasing much faster than I can hire
people to review these applications.
So
we are looking for better ways to review these applications. We recently had an office-wide retreat for
the entire office to look at ways that we can cut down on our workload, become
more efficient. If we're looking at
something we don't have to look at, we don't want to look at it anymore. We're trying to identify anything we can to have
a more efficient operation.
Along
with our originals, and Moheb brought out the point that every time we approve
an original, we're looking at more supplements.
And if you approve 300 or 400 a year, you're looking at a lot more
supplements. So anything we can do to
reduce the supplement load, we're also very interested in.
Quality
-- and this -- I mean these posters you may see on buses. If you go to Los Angeles or Chicago, we've
actually had our posters on buses. The
waiting rooms in Eckerd's and I believe Giant had then in waiting rooms. So we are very proud of the quality of the
generic products that are on the market today.
We
believe your generic drug is safe, effective, and bioequivalent. We believe people should be able to take them
with full confidence.
So
the products out there today are not bad.
I mean they're good, safe, effective products. We're just looking at better ways to make
them, more efficient ways to make them so that the industry and the FDA will
have a less burden in reviewing the applications.
And
it gets to the definition of quality.
And Helen asked me, she said your quality slide is blank. And there are a lot of definitions of
quality. I know David will probably give
you one tomorrow. I think Janet Woodcock
has one.
And
to me quality is pretty much, you know, in the eyes of the beholder. You know when something is inferior in
quality. I had a 1976 Dasher a few years
ago. And it was the worst car I ever
owned. It wouldn't start. The air conditioner wouldn't work. And clearly my decision, based on the quality
of that car, was I never bought another Volkswagen.
And
all of you out there have stories about appliances, or electronics that you've
had, that really did not perform the way you thought they would. And your judgment on those were that they
were poor quality. And you probably
never bought that particular brand again.
That's your right to not do that.
Quality
with drug products is a different thing, though. Sometimes we can tell. If you have a patch that doesn't stick right,
that falls off when you take a bath, or if you have a bottle of pills that are
broken when you open then, you can make a sort of a consumer-based assessment
of quality there.
But
for the most part, you don't know if they're within specification. You open that pill bottle every day and you
take a pill with full confidence that it is going to make your cholesterol go
down or your blood pressure go down.
It's going to relieve your pain because you trust the FDA, you trust the
drug industry that what they say is in that pill is in that pill. And what they say that pill will do, they'll
do it.
So
that's where we come to play in. The FDA
has to be the person that helps to assure this quality. That's what we've been doing in reviewing the
applications to date and that's what we want to continue to do.
Now
the slide I showed previously, this one, our challenge with generic drugs is
that many people relate quality to cost.
And that's not a far stretch. A
Lexus costs ten times more than a Daewoo.
And they don't sell Daewoos anymore, yes. I mean but that's an extreme example.
I
actually rented a Daewoo once. It was a
horrible a car really. I see why they're
not around any more. But really that's
-- I mean that's a clear judgment people have.
You'd always rather drive a Lexus than a Daewoo.
But
with respect to generic drugs what we tell people is it doesn't matter that
they cost half as much. You should take
them with confidence, that they're made under the same quality conditions that
the innovator drug products are and you should be able to take them with
confidence.
That's
our challenge. And that's why Congress
actually asked us to start this campaign to make consumers aware of the quality
of generic drugs. And believe me with
the number of applications escalating that I'm getting, that's of primary
importance to me is to continue the quality of generic drugs.
Now
our current paradigm, and this is what we do today when we get an application,
we look at the quality standards. We
make sure that the standards are comparable to the reference-listed drug. We do look at the specifications of the
reference-listed drug established by Moheb's people in the CMC review in the
innovator products.
We
make sure the product is manufactured in compliance with good GMPs. And the process and specifications are
conditions of approval that require approval for any subsequent changes. Basically we lock in the specifications. If you want to change it, you've got to
submit a supplement to us.
That's
what we do now and we will probably continue to do that for a little while
longer.
Now
in original ANDAs, there's extensive negotiations over specifications. And we did an internal study in our office
recently where 40 percent of the original applications, the comments on the
first review cycle were all related to tightening specifications.
And
basically I don't blame the generic companies.
They come in, they base their specifications on the batch that they
made, that they submitted to us, and they don't know what the specifications of
the RLD are. It's a mystery. It's kind of a guessing game for them.
And
so they submit specifications based on their biobatch. And they try to, you know, make them as wide
as they think we'll accept because these are the specifications that is going
to lock them into their manufacturing processes for the next who knows how many
years.
And
we try to crunch them down a little bit according to, again, the references to
drug and what we think they can do.
And,
unfortunately, this takes time. And our
average review time for an ANDA right now is about 18 months. And we would like that to get down. Congress would like that to get down. And we're doing all we can to try to reduce
that number.
It
also necessitates a high number of supplements because once we lock in these
specifications, any time that the company wants to change one of these
specifications, they have to submit a supplement.
Now
in the new approach and, again, I harken to Dr. Berridge's presentation. You know, I feel like I should be like Mickey
Mantle and Casey Stengel when they went down to Congress and they were
testifying on the reserve clause in Congress and they asked Casey to give an
explanation of the reserve clause.
And
he went into this long explanation that, you know, went all around and around
and whatever. And actually the Congress
was kind of laughing at the end of it.
And then they went and asked Mickey Mantle if he could give his
comments.
And
he said I agree with Casey.
(Laughter.)
DR.
BUEHLER: So basically in this, I agree
with Dr. Berridge. The extent of product
knowledge is key. It drives the range of
risk-based decisions based on supportive data to assure a quality product. And that is a product with established
quality attributes, purity, potency and strength, identity, bioviability and
delivery, labeling, packaging, and physical performance.
So,
again, very general terms. You know
where is the specifics? And I said to
myself if I had to make a talk on the quality initiative, I want to be able to
provide good examples to the industry because the industry asks me what do you
want us to do?
And
I was hoping to be able to kind of have a slide where one side is this is what
you do now and the next slide is this is what we want you to do. And then the next slide will be like this is
what you'll get out of it, you know.
This is what you won't have to do because you've done the second part.
I'm
still not able to do that. We're still
working on that. And I will throw some
challenges out to you at the end of this presentation to try to help us to get
to that point because I believe we have to get to that point.
You
out there have to know what's in it for you.
You're a business. You're a
business to make money and the generic drug industry especially is a very
competitive business. And they want to
know, you know, how it can effect the way they manufacture drugs. And we have to be able to tell them that.
This
is voluntary. And I want to emphasize
that. I know that there are some
companies that are not ready for this.
And these companies are the companies that are submitting my 500-plus
applications to the Office of Generic Drugs.
We
will work with you. We will be glad to
work with you. We want to work with you
through your trade organization, the GPHA.
We will try to organize webcast presentations so that you can begin to
understand what we want from you.
It
will be a phase-in process probably. We
hope that certain parts of your application can use this paradigm if not the
entire application. And, hopefully, you
can do that through comparability protocols.
We
want to be able to move the generic industry into this paradigm but we know it
won't happen overnight.
We
don't want to unnecessarily impede optimization of manufacturing processes and
that's what people are accusing us of right now. They're saying that FDA is in the way of the,
you know, movement forward of the generic industry. And we realize many firms won't be able to do
this.
Gerry,
I'm going to pin you down. Do you make
Viagra 24 hours a day?
MR.
MIGLIACCIO: No.
DR.
BUEHLER: No? Do you have a dedicated facility for -- is
that because of the competition? Did you
make Viagra 24 hours a day?
MR.
MIGLIACCIO: No.
DR.
BUEHLER: No? Okay.
I thought those bathtub guys were giving you some competition.
MR.
MIGLIACCIO: They are.
DR.
BUEHLER: They are? Okay.
How do they get those bathtubs on the side of the mountain? Have you ever seen that commercial for the
bathtubs sitting on the side of the mountain?
(Laughter.)
DR.
BUEHLER: How do they put the water in?
But,
I mean obviously for a product like Viagra or Norvasc or some of your big guys,
I mean you are making -- you don't ever shut those lines down, correct?
MR.
MIGLIACCIO: Sure we do.
DR.
BUEHLER: I mean -- but I mean to just do
some maintenance on them but not to make another product.
MR.
MIGLIACCIO: Sure we do.
DR.
BUEHLER: Yes? Okay.
Really?
(Laughter.)
DR.
BUEHLER: I'm amazed. Okay.
I thought you just -- 24 hours a day.
No? Okay. All right.
MR.
MIGLIACCIO: Let's stop with this.
DR.
BUEHLER: Okay, I should. I should.
Well, all right. The innovators
are always beating me up. So I thought I
would pick on Gerry a little bit but he's got the answers so I can say for sure
generic companies don't make products 24 hours a day. And they don't even make products probably
week after week after week.
Some
very isolated products perhaps but most of your generic companies make numerous
products and they are breaking down their equipment and starting to make new
products, you know, weekly or monthly.
So it becomes more of a challenge for the generic company to implement
these process initiatives.
And
that's why I'm committed to work with the generic industry to try to phase
these processes in to how they make their products.
We
want to get a review completed in one cycle within the statutory time
frame. We'd like to get an approval out
within one cycle. That's pretty rare
right now but we are working to that.
We'd
like regulations based on knowledge and science that provide flexibility in
approval conditions. And we'd like the
need for supplements based on knowledge in the risk of changes effecting the
quality of the product, again, Dr. Berridge.
Now
we have made internal changes to enhance approvals. We're changing work assignments to optimize
review resources.
Right
now we have a system where we are assigning teams of reviewers to batched
applications or actually applications of -- we often get applications from
different sponsors for the same drug product.
And
so we are assigning actual review teams to those applications because we found
that many times the reviews kind of run along the same line. They use the same DMFs. And so it's much more efficient to actually
review the applications that way.
We
want to improve communications with the DMF holders. We actually want to work with GPHA to try to
do that. Many times the DMFs that we
have are deficient when we first review them.
We would like to remedy that because the DMF review is very critical to
our review process.
We
are incorporating the aspects of the CMC risk-based initiative. We want to identify CB supplements suitable
for expedited approvals. And what we
want to do here is when CB supplements come in to our office, we want to triage
them through the team leader. And we
want to issue an immediate approval if we can, if the team leader can make the
assessment on the spot that the supplement can be approved.
We
expect to deal with comparability protocols.
We expect that the generic industry will phase into this paradigm and
that we hope that they will do this through the comparability protocol pathway.
And
we want to utilize in-house knowledge for specific drug products to identify
new elements critical to product quality and to provide prior approval
supplement relief.
Now,
for the industry, formulation and process design based on inherent mechanistic
understanding of drug and its impact on product quality and performance. We need to have this information from
you. Sometimes we get some. Sometimes we don't get much at all. But that's what we're going to be looking
for.
And
I know, again, you're out there asking what are you going to do with it when
you get it? Well, I guess you're going
to have to trust us. We want to see it. We want to try to work within this paradigm
but we can't do it unless we have the information.
We
want specifications determined by the knowledge of the process or the
product. We want a clear rationale for
selection. And we have to confess that
we don't have that clear rationale right now.
Our rationale right now is based on the data we receive.
Process
understanding to mitigate risk associated with drug substance properties, we
want continuous process improvement. We
want to identify the parameters critical for product manufacture and product
shelf life for stability.
And,
again, we have to get together to do this because I know that you're not going
to send us a submission where you are going to try to guess at what we want
because that's too much of a risk for you.
So
we have to get together. And you have to
know what we want. And we have to
realize what we've asked for so that when we get these applications, we will be
able to review them efficiently.
Our
staff will follow guidances in current scientific literature. And the staff in OGD is very dependent upon
guidances. We don't have the one on one
interaction with the drug industry, with the generic drug industry, that they
have in new drugs.
We
don't have the end of Phase II meeting, the pre-NDA meeting, the little
fireside chats every once in a while when they have an issue. We just don't do that. With 550 applications we can't do that. And so we have to work within guidances and
formal guidances to the industry.
We
have to train our staff and we have to train regulated industry in what this
process is and what we expect. And we
have to get to the specifics.
This
represents a fundamental change in our thinking, in our culture of accepting
applications and reviewing applications.
And we have to be able to get away from this culture and into this new
paradigm.
We
need a review based on knowledge of the product and what manufacturing changes
will make a difference.
Why
should you do this? And this is the big
question that many of you have. Greater
flexibility in optimizing your manufacturing process. This is a good thing. This should be able to help you. And this should be able to help the industry
as a whole.
Lessened
post-marketing supplement burden. You
saw my slide where, you know, we're getting, you know, almost 3,000 supplements
this year. We have to be able to find
some way to lessen this burden for my office and for your industry.
And
reducing no assignable cause, results, and investigations. These are when you get your 483s and they
don't know why but something failed in your process. And there is no cause assigned.
Now
my ICH slides, I think I'm just going to blast through because actually Mr.
Razzaghi and Dr. Berridge have done a very good job in explaining how ICH fits
together with this particular paradigm and mine are just little summary slides.
Dr.
McClellan, our former Commissioner, stated that other high-tech industries have
achieved enormous productivity gains and we should expect nothing less from the
pharmaceutical industry. Yet the Wall
Street Journal said FDA regulations leave drug manufacturing processes
virtually frozen in time.
It's
true that regulations designed to protect the public's health make this a very
special industry. And they promote a
conservative risk-adverse mentality. And
FDA counters that the drug companies resistence to change is also partly to
blame.
You
don't want to risk changing. And we have
to admit that we're a pretty conservative bunch, too. And we sort of, you know, go with the flow
and we don't like to rock the boat too much.
But
here we've made the first step. We want
to encourage the use of equipment and protocols for continuous monitoring of
manufacturing processes, PAT. We want to
encourage moving to risk-based cGMPs to free the industry from rules that do
little or nothing to ensure quality. And
we're willing to facilitate initiatives as long as they improve the quality and
reduce the risk.
We
acknowledge the generic industry as experts in manufacturing. You manufacture hundreds of drug
products. And we know that you know how
to do this. And we know that you are
aware of the many processes, the many new processes that are available.
You
can identify and articulate the financial impact both for changing and for the
losses with current technology. And I
said before, I am sympathetic. I realize
you are a business. You do make
money. And the economic aspects of this
are important.
We
have to avoid the perception of a two-tiered quality product system once we get
into this. We don't want to have, you
know, the sort of, you know, the Level A quality people and the Level B quality
people. And I don't believe we're going
to get that. But we have to make sure
that that isn't a perception.
And
the partnership assumes product quality is about providing flexible regulatory
impact based on product understanding.
Because
this system includes a continuing of information, how this flexibility is
applied needs to be well understood to ensure even treatment and outcomes. That's what I'm saying. We have to be able to provide details to you
about how this will work.
FDA
is not in the business of manufacturing.
We don't manufacture. And your
question to us is what do we need to do?
And our question to you industry is what do you think needs to be done?
We
invite you to come to us either individually -- we know that sometimes you will
have issues where you want the entire industry to be present when you are
presenting your issues to us. You can
ask for a meeting on this and we will grant the meeting to discuss how you can
move forward.
We
also want to work with GPHA and my friend Gordon is in the back. We hope to be able to set up something with
GPHA so that we can talk about general principles and, hopefully, again,
because we have talked about general principles an awful lot. Hopefully we can get into the specifics of
how to do this problem, what we want you to do, and what we expect to see, and
what the effect will be upon you long term.
I
just have to finish with another slide, another bus slide. But I am very proud of the generic
industry. I'm proud of what we've been
able to do to alleviate the high drug costs in America today.
I
am a bit overwhelmed by the number of applications that we have in our office
right now but I'm also very pleased that the generic industry is sending them
to us. And we will happily review and
approve them hopefully.
Thank
you.
(Applause.)
CHAIR
BOEHLERT: Gary, thank you. Also some very ambitious initiatives.
Nozer?
MEMBER
SINGPURWALLA: Yes. I have a lot of questions and comments.
First
is I'm not sure whether you were addressing your talk to the Committee or to
the generic drug industry. I got the
impression that you were talking to the generic drug industry.
DR.
BUEHLER: There's a few of them here.
MEMBER
SINGPURWALLA: There's a few. Okay.
Well,
I would like to ask you a few questions and then I'd like to make some
comments.
First
is do you have any example wherein a generic drug is of better quality than its
non-generic counterpart?
DR.
BUEHLER: Better quality?
MEMBER
SINGPURWALLA: Yes.
DR.
BUEHLER: No.
MEMBER
SINGPURWALLA: So all --
DR.
BUEHLER: We say they're equivalent
quality.
MEMBER
SINGPURWALLA: Oh, equivalent. But there is never a counter example where a
generic drug is of better and more effective quality then a non-generic?
DR.
BUEHLER: Well, you know, it depends on
how you define quality.
MEMBER
SINGPURWALLA: Whatever way you want to
define it.
DR.
BUEHLER: Okay. Well, I mean --
MEMBER
SINGPURWALLA: Just yes or no.
DR.
BUEHLER: Yes.
MEMBER
SINGPURWALLA: There is?
DR.
BUEHLER: Yes.
MEMBER
SINGPURWALLA: Okay. Second, the approval time for a generic drug
you said is about 18 months?
DR.
BUEHLER: Yes.
MEMBER
SINGPURWALLA: How much is it for a
non-generic counterpart?
DR.
BUEHLER: Probably I think it's 12 to 14,
something like that.
MEMBER
SINGPURWALLA: So a generic drug takes a
longer time to be approved than a non-generic drug?
DR.
BUEHLER: Yes.
MEMBER
SINGPURWALLA: Well, I propose that if
you use Bayesian methods --
(Laughter.)
MEMBER
SINGPURWALLA: -- you will cut down on
both the generic and the non-generic approval time because if a generic drug --
if a non-generic drug has been approved, there is prior knowledge there --
DR.
BUEHLER: That's absolutely correct.
MEMBER
SINGPURWALLA: -- and that should be
translated to the non-generic -- to the generic counterpart and you should save
on --
DR.
BUEHLER: Well, Congress has made this
similar argument that you're making.
MEMBER
SINGPURWALLA: Well, Congress sometimes
is wise.
DR.
BUEHLER: Yes, sometimes.
(Laughter.)
MEMBER
SINGPURWALLA: Now, on your Slide 23, you cited two examples. One is by Dr. McClellan and the other one is
the Wall Street Journal, and you said that that was kind of a
contradiction but I don't see it as a contradiction.
One
was talking about productivity. That is
manufacturing. The other was talking
about the process of approval. They're
two different things. You know to
approve a drug, you have to look at its chemistry and all kinds of, you know,
biological features.
To
manufacture, it's a different process.
So I can see the two -- I don't see the two as being in conflict. I can see the two as being true because
productivity gain means how quickly you can manufacture, how efficiently you
can manufacture. Approval is a different
process.
DR.
BUEHLER: Well, I think the point being
Dr. McClellan said that the drug industry should do better but at the same time
the Wall Street Journal said that
we, the FDA, were holding back the drug industry.
MEMBER
SINGPURWALLA: Possibly true but on a
different matter.
DR.
BUEHLER: Okay.
MEMBER
SINGPURWALLA: And you say FDA is not in
the business of manufacturing. I
agree. But there are two comments. You monitor the manufacturing and secondly
this is the Subcommittee of the manufacturing.
So you do monitor the manufacturing process.
DR.
BUEHLER: But we don't manufacture.
MEMBER
SINGPURWALLA: Of course not. But you don't design the drug either.
DR.
BUEHLER: We monitor manufacturing.
MEMBER
SINGPURWALLA: You're just
monitoring. And anyway, my comment to
you is I think if you were to use Bayesian methods, you would save on time --
(Laughter.)
MEMBER
SINGPURWALLA: -- and you'd probably have
more time on your hands so that you can give more talks.
(Laughter.)
DR.
BUEHLER: Are the copies of your slides
available. I should be able to get
those.
MEMBER
SINGPURWALLA: Yes, but my slides are not
going to help you.
DR.
BUEHLER: I see.
MEMBER
SINGPURWALLA: They are just -- my slides
are not going to help anyone. They're
just going to tell you what it is all about.
To
really -- to be effective, you really have to go, take a specific example, work
it through very carefully, and make the case that this is what can be done.
DR.
BUEHLER: I agree. I absolutely agree. We need some examples to get through our
system and to be able to illustrate to everyone the economics of this and the
efficiency of this. And the fact that
there is benefits for the drug industry in doing this. I absolutely agree.
MEMBER
SINGPURWALLA: I'm done.
CHAIR
BOEHLERT: Ken, then G.K.
MEMBER
MORRIS: So, Gary, after you've
instituted the Bayesian analysis --
DR.
BUEHLER: Yes.
MEMBER
MORRIS: -- when you're talking about not
being able to have the same sort of end of Phase II meetings but in the face of
the extended, relatively extended review time, is there a possibility, because
it does actually in many cases, I know the direct contact really does speed up
the process by resolving issues that are quickly resolved when talking
scientist to regulator, et cetera, is there any chance at least for like
teleconference --
DR.
BUEHLER: Yes.
MEMBER
MORRIS: -- meetings and --
DR.
BUEHLER: We've actually instituted -- we
had -- believe it or not, you know, in past years, we had a system where we
didn't talk to anyone during the first cycle on the telephone.
MEMBER
MORRIS: Yes.
DR.
BUEHLER: And we are revising that
policy. And that was a policy that
instituted as the result of the generic drug scandal and trying to sort of
mandate this level of consistency across the entire office with respect to
review.
And
we have sort of broken away from those shackles and we are encouraging our
reviewers to talk, especially at the end of the first cycle. And to be able to discuss the deficiencies of
the first cycle.
One
thing that I did mention that we are trying to address are the DMF
deficiencies. We're highly dependent
upon, obviously, the DMF for the active pharmaceutical ingredient. And we are trying to do something where we
can either get those reviews done in an earlier time frame so that the
deficiencies can be set ahead of time and that they can be back in time for
when the application is reviewed.
Because
clearly we get many applications that could go out on the first cycle except
for the DMF deficiencies.
Yes,
G.K.?
MEMBER
RAJU: Coming back to -- you said you
like John Berridge's presentation from the morning. In his presentation he talked about the ICH
Q8 and Module 3 about pharmaceutical development.
DR.
BUEHLER: Yes.
MEMBER
RAJU: To what extent does that directly
translate? Is it different for the
generic industry, the importance of pharmaceutical development and what you
want submitted in terms of the whole ICH process and Q8 and what they're
putting into that section? Do you want
something from the generics? The
same? More or less?
DR.
BUEHLER: Well, it sort of probably will
have a different focus. I mean and --
and Paul can maybe address this better than I but to me a generic firm in their
development report, the big part of their development is they want to develop a
bioequivalent formulation to the RLD.
That's sort of the big target.
And
how they do that with respect to, you know, if there is a patent that is in
their way and how they design around the patent, how they choose the inactives
for the particular formulation. And
then, you know, the development aspects of all of the formulating of that
product, we would be very interested in seeing.
And
so I think to us that would be our, you know, the development information that
we would want to see and all that was attendant to that.
MEMBER
RAJU: But the paradigm in which you
evaluate quality is bioequivalance. Then
your desired state in terms of mechanistic understanding is based on the
innovator's understanding? Or is it
based on getting a special -- a mechanistic understanding for the generic all
over again?
DR.
BUEHLER: Well, many times the
manufacturing processes are vastly different from the generic and the
innovator. So if we want to understand
the mechanistic, you know, the manufacturing process from, you know, A to Z or
whatever, it could be totally different than the innovator's.
We
certainly refer to the innovator applications for, you know, referencing and
actually looking at what they do and what problems they had. But with respect to the generic, we have to
look at their process and, you know, they would have to define the critical
parameters in their process.
MEMBER
RAJU: Okay. So as far as the product is concerned, it's
pharmacokinetics and dynamics. You take
that from the innovator because it's already out there. But in terms of the generic, not only
bioequivalance but you'd also look for some mechanistic understanding of their
formulation --
DR.
BUEHLER: Yes.
MEMBER
RAJU: -- to give them a specification
release.
DR.
BUEHLER: Yes. I mean some products are -- I mean like
extended release products have vastly different ways of manufacturing and
mechanisms. So --
CHAIR
BOEHLERT: Okay. Nozer, did you have another comment?
MEMBER
SINGPURWALLA: Yes, I'm sorry to come
back. I'm curious. Why does a generic drug take 18 months for
approval whereas a non-generic one takes 12?
Why less? Why more in the other
way?
DR.
BUEHLER: There we have almost 600
pending applications in our office right now.
MEMBER
SINGPURWALLA: Oh, so the cause of it is
you are overloaded?
DR.
BUEHLER: Yes, it's a queue system.
MEMBER
SINGPURWALLA: But it's kind of
unfortunate and unfair to the generic manufacturers that since the FDA is
overloaded, they have to wait, right?
DR.
BUEHLER: Well, yes and --
MEMBER
SINGPURWALLA: I don't own shares in a
generic drug.
DR.
BUEHLER: Well, well, no.
MEMBER
SINGPURWALLA: I just want you to
clarify.
DR.
BUEHLER: And that's an average,
too. And we do approve many applications
in eight months, nine months, ten months.
MEMBER
SINGPURWALLA: Oh.
DR.
BUEHLER: And they depend upon the
quality of the submission, whether it is a controversial drug or not, whether
we have patents to deal with, whether we've been sued on the particular
product.
Sometimes
when we're sued, well, Gerry, sorry, but Gabapentin, I mean there's still no
Gabapentin on the market. The patent
went out four or five years ago. We have
products in our office that have been pending for seven or eight years. Now what do you think they do to a mean?
MEMBER
SINGPURWALLA: Okay. So the bottom line is that it's not for
scientific reasons that you are taking a longer time --
DR.
BUEHLER: No.
MEMBER
SINGPURWALLA: -- to approve.
DR.
BUEHLER: I mean it's a -- we had a
generic drug scandal in 1990. So part of
that scandal was taking products out of order, taking preferential treatment to
certain companies. And so we have a
rigorous queue system in our office where we take things, you know, first in,
first reviewed. Not necessarily first
approved because it depends upon the quality of the submission.
And
they are stacked up in line. Each
chemist has a queue that goes down. Our
bioequivalents division has a queue of applications like, you know, 30 pages
long.
MEMBER
SINGPURWALLA: I got the message. I thought it was for scientific reasons. And if that was the case, then I'd be a bit
surprised because you already have knowledge from the poor non-generic drug
manufacturer who has done all the investing, you know, and done all the
work. You should be able to exploit
that.
DR.
BUEHLER: No, we acknowledge that. No, they do a good job.
MEMBER
DeLUCA: Along those lines, Gary, do you
want to comment on the future? Because
this is going to get worse as far as workload with the drugs, the biotech drugs
that are going to be coming off patents in 2005. You're going to have a very increased
workload in the generic area.
DR.
BUEHLER: Well, I probably won't comment
on the biotech drugs because that's a bit up in the air as to just who will be
doing those. But, no, from this slide,
obviously the trend is more work.
Moheb,
actually his slide, I think he said he had about 100 and some new NDAs, 115 new
NDAs. We got 102 last December, 102
ANDAs in December, in one month. So the
trend clearly is going up.
Like
I said, we did have an office-wide retreat about a month ago where we looked at
just about every one of our processes to try to determine where we could do a
better job in looking at fewer aspects of the application. And trying to identify really the critical
parts of the application that have to be reviewed.
And
at the same time, we're hiring people. I
mean every, you know, every couple weeks a new person comes on board. And we are trying to get to the point where
we have 60 review chemists, where we have three divisions of four teams each,
five chemists in each team. And we
believe that that will give us a good base to be able to address this workload.
CHAIR
BOEHLERT: Paul first, then we'll go
Garnet, and then Dan.
DR.
FACKLER: I just want to make a couple of
quick comments. One about the
pharmaceutical development reports.
They're admittedly different for generic drug development than they
would be for the innovator's product. We
have only a couple of targets that we need to hit.
We're
looking to have pharmaceutical equivalents.
And then we're looking to have dissolution comparability and
bioequivalents. So the development
reports for a generic product are focused, you know, certainly more tightly
focused than you'd have for the comparable brand product.
The
question about quality. Are there ever
generic products with better quality than innovator products? It depends on how you assess quality. We sometimes have a problem reducing
bioavailability on oral products to match an innovator. And you could argue that a better quality product
would have better bioavailability.
But
then we'd be coming out with a 15 milligram tablet to go against a 25 milligram
table innovator. It's not an
equivalent. We have to back off that
kind of a formulation.
And
the other kind of quality comparison is the variability that you see in the
bioequivalents study. And there is an
inherent variability in a drug substance but there's also a variability
associated with a drug product.
And
it's sometimes difficult to engineer -- for us using different release
mechanisms, it's sometimes difficult to engineer the same variability see in an
innovator product.
And
the last point I wanted to make was really a question about the review
time. We understand that reviews should
be -- or the first review should be completed in, I think, 180 days. And recognizing with the large number of
applications and the limited resources, we sometimes don't receive those within
180 days.
My
guess is that the review time is very short compared to new drug applications
if you discount the time that an application sits in the queue, if you will.
DR.
BUEHLER: Yes. And that time also reflects the time with the
firm. So if we send deficiencies to the
firm and the firm decides that this isn't a high priority application to
respond to and they have three others on their table that, you know, the patent
is going to go out in a month, they want to respond to, they will let the
application sit. And so that time counts
against us, too.
DR.
FACKLER: The other time that counts
against it is the 30-month stay.
DR.
BUEHLER: Yes.
DR.
FACKLER: So that if we've made an
application, we can't legally market a product for 30 months whether or not FDA
has approved our application.
MEMBER
SINGPURWALLA: An unfortunate system of
rules I should say.
(Laughter.)
DR.
BUEHLER: Well, it's a heavily legal influenced system.
CHAIR
BOEHLERT: Garnet?
MEMBER
PECK: I do believe that you mentioned
something to this effect that you will have an API that has an ANDA submitted
by multiple companies.
DR.
BUEHLER: Yes.
MEMBER
PECK: Yes. Just --
DR.
BUEHLER: Many times.
MEMBER
PECK: -- recently there were seven
companies got approval about the same day so are you trying to work those as a
unit?
DR.
BUEHLER: Yes, now we are. We didn't previously.
MEMBER
PECK: Through the Agency?
DR.
BUEHLER: Yes but now we assign them to
the same team if we can if it's a small enough number because many of the times
they utilize common DMFs so the DMF review, you know, can be utilized for a
couple of different applications.
And
also the issues related to the review of the application are many times common,
too. And so it helps to have a group of
chemists being able to discuss the issues with themselves and the team leader
in reviewing that.
And
we found that the review is much more efficient and actually done much faster
that way.
MEMBER
PECK: Yes.
CHAIR
BOEHLERT: Dan?
MEMBER
GOLD: Gary, you mentioned, I thought,
that some of the delays are caused by inadequacies in the drug substance DMF.
DR.
BUEHLER: Yes.
MEMBER
GOLD: I have not seen any publication by
the Agency or by the Generic Division as to what deficiencies they are finding
and what advice they might offer the industry in order to improve the quality
of the DMFs so that you can, thereby, take advantage and review, you know, and
reduce the review cycle time.
Why
not do that?
DR.
BUEHLER: Well, that's a good
suggestion. You are right. There aren't any that I know of. Frank?
No. DMF guidance? We don't have --
PARTICIPANT: Well, historically we've done this
periodically.
MEMBER
GOLD: I'm sorry. I cannot hear you.
DR.
BUEHLER: Frank said historically we've
done it with the industry.
PARTICIPANT: And probably 10, 12 years ago, there was a
series of DMF conferences within the Agency where there were a number of
instances discussed. Part of that long
series was here are the most likely things that you will find wrong, frequently
with DMFs.
And
it's not something that we repeat. It's
usually a special project when we go in and we look at them.
MEMBER
GOLD: May I suggest that you consider
putting out a type of document that other sections have put out such as
Q&As on --
DR.
BUEHLER: Sure.
MEMBER
GOLD: -- and this one directed to DMFs
to guide --
DR.
BUEHLER: That's a good suggestion.
MEMBER
GOLD: -- to guide applicants in that
area?
DR.
BUEHLER: Sure. That's a very good suggestion. And as I stated, we hope to have a meeting
through GPHA with some of the DMF holders also, a webcast where we can connect
people through telephone if they can't attend a meeting personally, and talk
about these deficiencies, too. We've had
these meetings on other issues within the office.
MEMBER
GOLD: And there's another issue here,
too, that since so many of the DMFs now are coming from overseas, I think the
estimate is of the order of 80 percent of the drug substances are coming from
overseas, I think we really have to broaden the approach we're taking in order
to reach all the applicants.
DR.
BUEHLER: Yes, we have --
MEMBER
GOLD: All the DMF applicants.
DR.
BUEHLER: -- we have to very often deal
with their agents in this country with our deficiencies and our communications.
MEMBER
GOLD: No, but I'm thinking in terms of
international meetings in order to expedite this because it is important to get
generic drugs on the market faster.
DR.
BUEHLER: Yes. Okay.
CHAIR
BOEHLERT: I have one last comment before
the break. Ken?
MEMBER
MORRIS: Yes, just to follow up on your
point. I think one of the problems that
gets lost in the shuffle with DMFs is that the companies, the drug companies
themselves often don't have access to much of the DMF so the audience for that
sort of a meeting is, of course, the DMF holders.
But
depending on their stake in the particular active that you're talking about for
the particular generic company, that may not be a compelling enough reason for
them to make a lot of changes or to be a very forthcoming.
So
I don't know the solution to that but I've run up against that before.
DR.
BUEHLER: Well, the drug industry is
clearly the customer -- or the DMF holder is the customer of the drug
industry. So, I mean, we sort of do look
to the drug industry, the generic drug industry, to actually pressure the DMF
industry to submit better applications.
That way their applications won't be held up.
MEMBER
MORRIS: No, I understand the point. My point is in terms of delays that are a
result or a manifestation of that, may not be something that lies within the
control of the generic company itself.
DR.
BUEHLER: No, you are -- that's
absolutely correct. They don't even know
what the deficiencies are.
MEMBER
MORRIS: Right.
CHAIR
BOEHLERT: Okay. Thank you all for very, very good discussions
this afternoon. We're going to take a
break now and reconvene at 3:45.
(Whereupon, the foregoing
matter went off the record at 3:30 p.m. and went back on the record at 3:47
p.m.)
CHAIR
BOEHLERT: Okay. Our last speaker of the day is Ken
Morris. Certainly last but not
least. He's already at the podium and
ready to go.
MEMBER
MORRIS: Well, that's because unlike
Gary, who was only facing people who were trying to go golfing, I'm facing
people who I'm the only thing between them and the bar. So -- what's that? Yes, when I hear the clinking of ice, I'll
know I've overstayed my welcome.
Well,
first of all, thanks for inviting me Judy, and Helen, and Ajaz.
The
purpose of this is to largely report to the Committee on some of the activities
that are going on with the senior CDER and DVM, and ORA folks to discuss and to
flesh out the ideas of question-based CMC review.
And
in the course of doing this, I'll try to differentiate my opinion from what
we've actually done. But in the first
half of that talk at least, what you'll largely see are the fruits of the work
that we've all done as a group to explore this and brainstorm.
These
are by no means final. And this is, as I
should point out, a work in progress. We
intend to continue this.
Lest
you choke at another current versus desired state, let me say that this is a
little bit different in that this is the assessment not only of ourselves and
the upper management but Directors, Deputy Directors, Team Leaders, and
Reviewers as well as the odd academic.
Right
now if you -- and you have these slides so the fact that they're animated isn't
going to mean much. I'll go through them
pretty quickly.
The
companies, as we've heard, may or may not have information. But it's not always in the filing. And there's not a lot of incentive for it to
be.
The
reviewers have to go through cycles of information requests and questions and
then wait for the responses. So the
companies may or may not have the clear scientific rationales for the choices
but, again, they're not always sharing it.
And
what this really results in is that the reviewers have to piece together data
and observations to discover, if you will, the rationale for a specification, a
method, a formula, or a process, et cetera.
And
really we're saying that the reviewers are in a large sense of the word, and
I'm not laying any blame here nor was the group, serving the function that
should actually be done in the company and may well be being done in the
company but just not shared.
In
a desired state, what we'd like to see, of course, is that companies would
include needed data with the filings and could share it prior to the filings,
the end of Phase II meetings being the sort
of the poster child for that concept.
They
would include the data analysis to produce meaningful summaries and scientific
rationales. So as opposed to the current
state where if there are data missing in the reviewer's opinion and you ask for
a data summary, in essence, and you get three boxes of chromatograms, that
doesn't really serve anybody's purpose.
The
idea would be to have meaningful summaries of the data, that is data that have
been analyzed and interpreted in the light of what the company believes is the
proper interpretation and shared with the reviewers and the Agency.
This
should lead to the specific or the scientific rationales, the product
development history sort of rationale we're talking about.
The
reviewers then would assess the rationales and the summarized data
presentations as satisfactory or not.
And in that scenario, what you see is the potential to gain all of the
things that we've been talking about all day and will continue to talk about
tomorrow.
We
had talked -- at Purdue, we had talked about sort of folding this into a
risk-based development concept. And now
I'll have to couch all this in terms of the Bayesian defensible risk and not. But I'll try to do that as I go along, Nozer.
First
of all, the idea, as I said, is a simple concept. And if you use sound scientific principles in
the design of the dosage form in the process, you've essentially met Phase
I. Not Phase I in the clinical sense.
You
have to identify the critical attributes for the raw materials, and we'll talk
a good bit more about this as we go, identify the process critical control
points for the processes, employ the proper analyses and process analytical
technology concepts for process understanding and control.
And
tie it all together with the appropriate informatics to feed the information
forward and backwards for quality by design and in continuous improvement,
which is the daughter of that. And that
all leads to innovation, which is supposed to and should reduce risk.
Now
we haven't talked very much about informatics today but clearly this is
something that is an inescapable and inexorably linked to all of these
initiatives. That it doesn't do you any
good to collect data if it's not used much less shared between the
organizations within the company, within the FDA, or between the FDA and the
companies.
So
what we'll do as we go along is expand on the righthand side of this list to
talk about the associated regulatory question rationale or rationales.
The
concept of risk-based development is really all about feeding forward, and I
would add backwards, but feeding forward at the outset. This is after a set of quotes that Ali Afian
had spewed at Arden House very passionately.
So
if you look at it with a little more detail, what we're saying is you can
explore the characteristics of the raw materials and possible variability in
the raw materials and processing that are expected, that is expected based on
some either previous knowledge or model, to impact on required dosage form
performance. And we'll come back to what
required means but, of course, that's another whole discussion.
Deciding
on a dosage form based on the first step and the business case and selection of
possible processes would be the logical next step. And what you'll see as a theme as we go
through this is pretty much what you would expect if you are in companies you
are doing now, and for the Committee, I would say that this is one of the focal
points of what we're going to talk about.
And I'll tender a hypothesis in a moment that's -- well, maybe I won't.
Then
deciding what data are necessary to assess the probable success of No. 2, that
is the dosage form, this can be from first principles, literature, design of
experiments, et cetera.
Collect
and analyze the data in the fourth step and you can see where PAT would play a
role here.
Then
Gap analysis and refining models as the development proceeds and finally the continuous
improvement, which starts the cycle over again.
I
wanted to use as an example here, and we sort of used this as an example in the
team as we met, Solid Oral Dosage Forms.
But, of course, we're not limiting any of the arguments or the
hypothesis to this dosage form.
But
there are really only two issues with Solid Oral Dosage Forms. One is does it work? That is the performance. And the other is can you make it? And that's the manufacturability.
If
you look at the subsets of each of these, for performance right now we have --
and when I say dissolution, I'm not talking about the dissolution testing. I'm speaking of it more as Ajaz was this
morning in that dissolution may be important whether or not dissolution testing
is measuring it is a different question, dissolution in vivo, absorption, and
stability.
And
then each of those have subsets which are logically defined by the physical,
chemical models that are around or that need to be developed. Where I have flags are places where we actually
have models in place. And if you look at
this really, the big unknown and the analogy here is on the old maps, when
you'd get to the end of the continents, they'd say and here there be dragons,
is the absorption, the clinical aspects.
But
really what we're talking about is the manufacturability for most intents and
purposes since we can't really fill that gap at this stage. And for manufacturability what I have here is
physical properties and processes and then those are broken down into their component
parts.
So
this is an overall example of what the requirements are for the dosage
form. They're really -- what we talked
about the required part in the first step here, the required dosage form
performance, that's really what we're talking about ultimately. But, of course, we aren't there yet.
Well,
how realistic is risk-based design, if you will? Or the whole concept we're talking about
really. And I start this by stating this
premise that as all good pharmaceutical scientists and engineers know, a
formula without a process is really a pile of powder if it's a solid oral
dosage form.
So
even during API characterization, developing a formula implies an expected
dosage form and a process or range of choices.
And the example is here you don't care about the compressibility of a
lyophile, for instance.
But
I'd submit that even at the very early stages, if you're sitting at a
pre-formulation desk and somebody throws ten milligrams of material on your
desk, you know exactly at that point what the dosage form is going to be.
Now
you may not know exactly what option within the dosage forms you are going to
have, but you're going to know if it's a tablet. If it's an analgesic, you're not going to
have an ocular injection is my standard example.
So
API characteristics are among the first information you need to feed
forward. So if we look at that though
for the people that are attending in the gallery as well as the Committee, you
have to be saying well what's different about this than what we do right
now. We do all of this now.
A
good formulator, a good scientist, a good engineer will just tell you right
away that this is the thought process they go through. But the difference is that we're not doing it
model based. We're not sharing and
feeding the data forward and backwards.
And there's no informatics to capture this in a meaningful way. In other words, it's the process.
The
process itself is what is new. And the
process itself is what's necessary to bring all of these ideas to fruition.
Just
as an example, just as dipping the toe into the pond of biology here for the
moment, even at very early stages when you receive just a molecule, you can --
even a molecular structure and a small amount of material, you can assess
solubility impact on pre-formulation on absorption using relationships such as
the modified absorption parameter, which does a fairly good job just based on
molecular structure and some estimates that you can make either computationally
or with simple experiments on whether or not even a low soluble drug will be
absorbed.
Well,
you've already seen this slide and I'm not -- this is actually from Rick Cooley
from -- that Toby showed this morning so he didn't have to cite it because it
was his company but Rick Cooley from Lilly actually presented this.
If
we look at this in terms of the overall variability of the process, a variable
input will lead to an invariable product if you hold everything in the middle
constant. This is just common sense.
So
the idea is is to be able to adjust it.
The catch here is what are you going to adjust, that is what are the
critical attributes as well as what are the critical process points, the
critical control points in the process?
And that's what we're going to talk about.
So
the example that we started with in the team was actually API selection. And the idea here was to explore the question
of how do you know what questions to ask?
So now you have people who are going to be looking at your filings as
they come in. And presumably now data
would have been shared early on.
And
the first question that we all agreed on, we did this as a team, was what's the
-- what the first question you'd want to ask if you had your choice is what
dosage form are you going to be using?
So the first thing I want to know is what's my dosage form?
Then
the questions went on. What's the second
thing, et cetera? And the hypothesis
that we're proposing here, and that the Committee can assess during our
discussions is that the development scientist and the regulator are or should
be asking many or all of the same questions.
So the same process that the scientist is going through in designing the
dosage form and designing the process should be the questions that the
regulators are responding to because that's ultimately what will determine
whether or not the dosage form has been designed by quality.
Well,
if you -- I got permission from Moheb to use the pyramid at Arden House so I
extended by non-exclusive license to it -- if you go through this pyramid of
questions, what you start with is what is intended dosage form, which is what
we just said. What's the intended
process? And then stepping up through
the various tiers of the pyramid to the point where you've actually identified
the critical attributes.
And
the other dimension here, much like Ajaz's sixth dimension, is time, of course,
because that will change. And this will,
in fact, be a cyclic process.
So
this is the hierarchy of questions that you might expect to see if you were to
make a filing and certainly if you were designing your dosage form. And it's really a fairly logical progression
of consideration of the physical chemical properties of the API.
If
you have an API, it will either be a solid, liquid, semi-solid biological. And what the question is is what are the
critical attributes of each of these?
Now
if you select one, we'll select solids here because that's what I know best, of
course, not that they have to limit it, then if your API is a solid, you go
down a logical process of deciding whether or not it's crystalline or
amorphous, whether or not it's a polymorph, a hydrate, or something else, and
when you've selected the one that it is or identified the one it is, there will
be certain criteria which will tell you what the characteristics ought to be
and then this might take you to not a decision tree but an event tree as Nozer
said, an event tree in the Q6A.
Then
this would at least give you the range of possible critical attributes, which
puts you back on the path of this thinking.
Now
I'm not saying that you have to follow necessarily this sort of a chart. I'm saying this is what most formulation
people will follow -- or pre-formulation people will follow intrinsically.
Then
you say well, if the dosage form is a solid, it's going to be a capsule, a
tablet, or other. There are no pills, by
the way, Gary, left any more.
So
you go from tablets to selecting which particular choice you have for
manufacturing the tablets, for the various critical attributes taken into
account. There's wet granulation, dry
granulation, dry compression.
Once
you've selected that then the attributes that are potentially critical should
be fairly well known. And this is a case
where maybe modeling gives you the prior knowledge in some cases.
This
is then cycled on data to determine what the risk really is. And in this case you might think of it in
light of what we heard this morning as generating prior knowledge. And then hopefully you identify the critical
attributes.
If
you move this on logically to the process design, you start from where we just
ended with the raw material critical attribute selection, take this up the
ladder so that what is the model for the process. And the what processes are viable, you've
already answered that in the raw materials section based on the mechanical and
chemical properties of your material.
What's
the model for the process critical control points -- say that three times fast
-- and then the basis, the possible PCCPs based on the raw materials and the
choice of response factors. And there
you go continually until you do your design of experiments and ID preliminarily
what the PCCPs would be, cycle back until you again optimize it.
So
what I would say is that all of these are logical top level questions. And the more detailed questions are the ones
that we were just going through in the raw material or the API selection.
Let's
use an example. I actually picked on Q6A
quite independently of Ajaz. He knew
that I was going to do this because I sent him the slides. But he didn't tell me he was going to be
doing it.
I'm
not picking on Q6A particularly but it was just a good example to use because
there are some good things and some not so obvious things in the event tree
that Q6A represents. And I think what
we're really talking about doing is changing it into a decision tree based on
what we're talking about here.
So
Q6A in the first table, I can't remember if this is 6 -- I think it's table --
I can't remember what it is but at any rate, the first question is can
different polymorphs be formed? Okay,
this is fine. If you understand the
solid state and know polymorphs are formed, you're done. So there you are at no and no further action.
If
there are forms, they must be understood.
So it's not enough to just say yes, let's characterize them. What you really have to ask is what are the
relative stabilities of the low energy forms.
And if you don't know the relative stabilities, at least explore what it
is that's -- what information you have that's possible to help you explain that
or at least elucidate it.
And
those are the right questions for the scientist and regulators. And as we go through the next few tables,
we'll try to carry out the same analysis.
The
second part of the table is do the forms have different properties, solubility,
stability, melting point, et cetera? If
it's no, then no further testing or acceptance criteria for drug substance
required.
So
that's okay. But when we're considering
the product, the logical first question should be quite different because the
answer here is if they do have different properties, the question is is drug
product safety performance or efficacy effected? Well, before you get to that question, you
really want to say based on what is known about the material and the process,
what, if any, change in form would be expected?
So
if I have something that is particularly soluble and I'm wet granulating it and
I know that there are form possible, then I might expect that I could either
change a less stable form to a more stable form during granulation or I might
trap a metastable form on drying. Those
are the sorts of questions you would ask long before you got to the point of
whether or not it actually occurred.
So
if the answer is none based on the scientific understanding, then a
confirmatory test during development should suffice because it is possible but
you're saying that there's no logic to say that it should happen.
Otherwise,
if there is a potential, the next question should be is the observed change the
one that you expected? Now we've just
gone through this. You should know what
change you expect to see based on your process and the properties of the
API. And the question is does the change
that occurs match what you expect?
And
then finally, this is the question that will give hiccups to a few folks I
suspect, is what was the rationale for selecting the processing step
responsible for the change?
Then
we're back to the tree again. So on the
third section, it says does the drug product performance testing provide
adequate control if polymorph ratio changes during the formation of the
product?
And
here it might be reasonable to ask instead does the performance testing relate
to the performance of interest? And this
is what we were talking about before.
Now
you may not have an answer for this but that's clearly the question that you
would want to know. If the change in
ratio makes no difference, then it may not be an issue. If it does, obviously you have to establish
acceptance criteria. And if the answer
is based on scientific understanding, we're back to here.
A
next question would logically be based on the understanding of the form's
behavior, what would the expected trend -- that should be expected trend in
transformation be? So if I have a ratio
of polymorphs and obviously one of them is more stable than the other, you
would expect against any other information that the metastable form would
transform to the stable form.
If
it doesn't then you've -- well, number one, you have the paper. But number two, it brings into question
whether or not you understand what's going on.
And
now these questions are pretty specific but these are the kinds of questions,
these are the level of specificity that you would really like to know in
advance of seeing the questions I'm assuming.
Going
to the second part of the third table is does a change occur which could effect
safety or efficacy? And here I would say
does the observed change correspond to an understood and expected
transformation? If not, the system is
not well understood -- is not as well understood as you thought it was.
And
if that's the question, then presumably you would have addressed these sorts of
issues early on but the value of this is that if you've addressed each of these
during development, then by the time it gets to the regulator and they're
essentially echoing these questions, you'll have answers for them and that
would expedite the process.
Virtually
all companies on the innovator side are doing polymorph screens. We've recommended focused polymorph screen
for generics. Because of the number of
companies, I don't know the relative ratios but as an example, this would be
the case.
Well,
let me use the last few minutes here to talk about a couple -- a specific
example. I may skip the last
section. Judy, just give me a high sign
if my time starts to run out. Six? Okay.
Yes.
Okay,
this is an example that is actually from -- largely from Greg Amadon at Pfizer
in Kalamazoo from talks that he's given over the years. But it illustrates one of the things that Dan
had raised and Garnet had raised with respect to excipients. And that's the mechanical properties.
We
treat table formulation more or less as a black box. Not so much from the chemical sense because
the chemistry is often well known by the time you get it. Certainly if it's generic you know it pretty
well but in terms of what you would use and how and what the ratios you would
use to give a tablet that had acceptable strength characteristics, counting
uniformity as well as performance characteristics.
And
if we look just at the mechanical properties elements or aspects of the raw
materials, there are several tools and I'm just going to introduce one here
which are the Hiestand Indices. And
Everett Hiestand, when he was at Upjohn years ago, developed indices for
bonding, brittle fracture, and strain measurements as a function of relatively
easy to get data from relatively small amounts of material.
And
these data are tensile strength, hardness, and things that you can get to
fairly easily. And I won't go into the
details but let me show you some of the results.
And
if you look at the overall range of materials that we are involved with in
normal manufacturing, I would say this extends even more so to biologicals, is
-- I should say even to biologicals, not more so -- everything we deal with if
you look at in terms of the mechanical properties and just focus on this column
for a moment where we're talking about the description, falls into the category
of moderately hard to soft. There's
nothing that's really hard. There's
nothing that's really soft.
So
everything falls into this category. And
here you see APAP at the top and starch at the bottom. This is from a great chapter by Rowe and
Roberts in mechanical properties.
So
we're really dealing with a fairly limited range. And we're dealing with a fairly limited
number of excipients. But the APIs, of
course, can change.
Well,
if you look at the importance of evaluating these indices up front, this is an
example of Phenacetin. And here we have
a case where in the compaction -- in this compaction, in the tri-axle
compactor, we have a compound in Phenacetin with a very low bonding index.
And
the result of that is that even though the Brittle Fracture Index is not too
bad, that in the dye it comes apart. Now
this is -- I'll show you a couple other summary slides but the point is that
there are threshold values, if you will, that I would imagine would lend themselves
fairly well to statistical analysis apriori that have to do -- that are shown
here actually that have to do with their relative properties that should
dictate this apriori.
So
if you look at that bonding index of excipients versus drugs, you see as no
surprise that microcrystalline cellulose has very high bonding index, right,
which is also why it's a compaction aid.
If
you look at drugs, they vary but drugs tend to be, on average, lower. And if you look at what it takes to make a
good tablet, you'll have to have some combination of those.
Brittle
fracture index is the -- is, I guess, in a sense one of the most dramatic of
the indices because when it fails, it fails spectacularly. Here is an example with a very high brittle
fracture indices.
High
brittle fracture is bad because it means that on expansion, the compact can't
maintain itself. And what you see here
is with a high brittle fracture, as soon as the compact is ejected, it
laminates. It just comes apart. And if you were to put an acoustic sensor on
it, you could hear it. I mean it's very
noticeable.
And
similarly, if you look at the brittle fracture index now across a series of
excipients, you can quite easily determine which ones have the brittle fracture
indices that are less desirable or more desirable.
And
as Garnet talked about earlier, corn starch being one of our formerly primary
diluents, had its own issues with respect to brittle fracture index, which is
why a lot of it was granulated, wet granulated.
So
put this together and what Greg had done here was to plot the brittle fracture
index versus the percent of drug mixed with an excipient for several compounds
listed here, Drug X, which is Pfizer, I'm assuming that's not one of the
bathtub drugs.
And
what it shows is that adding 30 percent of a non-brittle excipient makes a
mixture much less brittle and, in fact, quite compactible, which could be
predicted with grams of material. So
we're talking a long time before you get to a kilo lab and certainly in the
generic industry something you could do Day 1 with the proper equipment.
And
Greg went on to develop a semi-empirical model that shows how H here is any of
the indices or properties so here we have hardness, tensile strength, brittle
fracture, and bonding index, are all related via a logarithmic relationship so
that there at least is within products and within excipients a predictability.
So
if you think about this in terms of the scope of excipients that are available
to us, it's already been -- data has already been collected on most of these
excipients so these are available in literature.
So
there is the possibility of using these data as is to do prediction up front
with either very little measurement or at least feeding backwards.
As
Ajaz had said, if you have in a big company I don't know how many products a
big pharma makes. Over a hundred I
suppose, right? And generics can make up
to 500 at a plant, the amount of data you have is staggering. To be able to take these data and bring them
back, it's impossible to imagine that you couldn't perform some data analyses
that would give you your prior information for your Bayesian treatments, for
instance.
No?
MEMBER
SINGPURWALLA: There's a confusion of
concepts.
MEMBER
MORRIS: Except for the confusion of
concepts --
(Laughter.)
MEMBER
MORRIS: -- that's absolutely true. Right.
Yes, we'll get back to that.
Okay. So if we look at our beginning slide again at
this stage, then the questions that you would expect to be associated with
these steps so far would be what were the principles applied and were they
appropriately applied?
So
if you're using the bonding indices and the brittle fracture indices, were they
appropriately applied? And I would say
the answer is going to be yes in most cases for the folks who have been using
them.
How
are the critical attributes identified in the formula design? I mean this is Product Development History
101. It exists in many companies. Whether or not it's shared is a different
question.
The
next level, as we talked about on the second pyramid, is the identification of
process critical control points. And how
am I doing time-wise here? I'm over?
CHAIR
BOEHLERT: Not so hot.
MEMBER
MORRIS: Okay. I'll skip through this section.
CHAIR
BOEHLERT: Well, you know, we do have
some questions to address for Ajaz this afternoon.
MEMBER
MORRIS: I understand.
CHAIR
BOEHLERT: I don't mind keeping you late
but I think, you know, the rest of your Committee members might mind.
MEMBER
MORRIS: Yes, I've got two slides left.
CHAIR
BOEHLERT: Okay, good.
MEMBER
MORRIS: Because I'll skip the example
but I want to get in this point. And
that is if you look at the relationship between PCCPs and scale up with
monitoring, the basic approach is captured as two simple process understanding
premises.
First
is that PCCPs are preserved throughout the scale up process. That doesn't mean that the magnitude doesn't
change. It may. But the variables being monitored reflect the
state of the process.
And
second, as was alluded to this morning, and I can't remember who, I apologize,
is that monitoring material properties makes scaling less equipment dependent
so that even as you change equipment, if you're monitoring the same PCCP, the
value may change but the absolute -- or I shouldn't say the absolute but the
PCCP being monitored is the accurate one.
And
I'll just skip to the last slide which says that based on that example that you
just saw, that in addition the next questions are how did you identify the critical
attributes?
The
next question is how did you identify the PCCPs? What were the basis for the analyses
selection? What are the supporting data
for all of the above? And finally, the
product development history should reflect everything that you've said. And if it doesn't, it's a different issue.
And
asking the right questions at the right time, feeding forward and back between
disciplines, designing the product and process against meaningful metrics must
start in R&D. Development of
meaningful specs, of course, results only from the identification of the
scientific basis. Real-time monitoring
is a big advantage but not absolutely necessary.
Process
understanding for quality control is known functionality; that is the models
against which data are used to control the mark. And I can't emphasize the model basis enough.
What
you get from this, I think we've heard quite a bit. I'll just -- this last point here is that in
tech transfer, you get a more realistic process to transfer, which is Gerry
Migliaccio's leg up statement from Arden House saying that we don't need a
final thing but we really could use a leg up so we're not starting from zero.
And
finally just to acknowledge Greg Amidon from Pfizer in Kalamazoo, CAMP, again,
with G.K. as our leader in CAMP, of course, Abhay Gupta is the graduate student
who did the example you didn't see. And
finally the team, which was headed up by John Clark but include Moheb and Rafad
and a lot of the people that are here as well so with that I'll end.
Thank you.
(Applause.)
CHAIR
BOEHLERT: Thank you, Ken. Any questions for Ken before he departs?
(No
response.)
CHAIR
BOEHLERT: Sorry we missed the
example. I was interested but, you know,
we are running out of time.
MEMBER
MORRIS: No problem.
CHAIR
BOEHLERT: Okay. Ajaz did you have a few comments?
MEMBER
HUSSAIN: No, I think what we have tried
to do is to give you a sense of what is happening outside FDA, especially in
ICH, ASTM, what is happening within FDA, especially from a more management
perspective but also from a science perspective.
And
we're hoping that I think if I could just put the slides on the questions --
this is a series of questions that we posed to make sure that we are on the
right track. And I'm hoping that your
discussion and general thoughts on some of these questions might be useful.
You
have a printed copy in your packet.
Usually we place this on the -- but maybe I can stand here and maybe
forward this for you. So it's up to you
how you wish to give us your feedback on these questions posed. So --
CHAIR
BOEHLERT: Well, I propose that we go
through these in order. First and third
are relatively short. The second one has
many subparts. So we'll start with the
first one.
Do
you agree that current activities within ICH and ASTM are helping us, FDA, move
toward the desired state? They seek our
recommendations on how to ensure these activities are synergistic. So I'm looking from comments from the
Committee.
Everybody
is saying yes. And particular comments? G.K.?
MEMBER
RAJU: I agree, very strongly agree. I'm not that familiar with the ICH process
but I did go to the ASTM process. And it
really is very synergistic as Don had said.
They're putting a lot more detail to it and bringing in a lot of outside
industry expertise.
So
I think they are synergistic. And
the synergies are happening with the
individual people. I don't know whether
there is a possibility for a more structural synergy among the people here but
in terms of what FDA does and what ASTM and ICH does, when I spoke to Don
earlier today and I asked him that question, he didn't think so.
But
there could be a time when it starts becoming really duplicative. I've seen people talk about pharmaceutical
development in at least five different organizations. Everybody's versions of what they want and
what is risk.
Everybody
has risk tool box, these five organizations.
So at some point, it's good to have a lot of people do it to get the
debate. But at some point it's probably
not.
And we're not there yet but we probably
will be in the future.
MEMBER
HUSSAIN: I think I'll just repeat what
Don had said in the sense, I think, the scope and the depth and the
details. These are two different
standards or guidances, whatever you want to call them.
If
you look at the E55 structure, what we're hoping to do there is to create a
framework. E55's focused primarily on
standards for PAT. And development is
clearly broader than that in a sense.
And
we're hoping the details that would come about through ASTM's standards would
be standards that can be cited. And we
really don't have to issue Agency guidelines on some of those things.
So
Q8, Q9, Q10 will evolve with a very different focus. And the ASTM would be more of a technical
standards rather than guidelines and so forth.
So I think there is a difference.
CHAIR
BOEHLERT: Ken?
MEMBER
MORRIS: That raises -- oh, I'm sorry,
were you not done, G.K.?
MEMBER
RAJU: No, I'm all set.
MEMBER
MORRIS: The one question and it actually
came up at the last ASTM meeting, will the ICH be able to -- or not be able to
but will they take advantage of the ASTM standards in citing them during their discussions?
MEMBER
HUSSAIN: Well, that's a very good
question. And I had brief discussion
with John Berridge about this a the sense.
The Yokohama meeting in November is probably when I would like to sort
of bring this topic up to ICH and keep them in the loop on this.
John
and I discussed this before the Washington meeting and felt that well, the ASTM
had not crystalized far enough to really share some of this. But I think starting in Yokohama in Japan in
November, we'll make sure that the ICH is fully aware of what's happening here
and seek that synergy.
Informally,
I have discussed this with all of our regular counterparts in Europe and
Japan. And there are a number of
European members on this ASTM and Japanese members. And I'll broach the subject of maybe the
regulators joining some of the ASTM groups also. That's a possibility.
MEMBER
GOLD: Ajaz, you're -- I just want to
ask, you're not going too fast in contrast to the regulators elsewhere, the
Japanese or the Europeans, are you?
MEMBER
HUSSAIN: I hope we are.
(Laughter.)
MEMBER
GOLD: Well, we have to move in
concert. And do you believe that you're
going to be able to move in concert?
MEMBER
HUSSAIN: That's the real --
MEMBER
GOLD: I asked the real question.
MEMBER
HUSSAIN: -- question. Well, move in concert in the sense we will
lay the foundation and hopefully they'll come and join us.
(Laughter.)
MEMBER
HUSSAIN: No, I think the ICH process,
the Q8, Q9, and so forth, clearly are toe to toe, we're moving together in a
completely harmonized fashion.
We
have plans I think with respect to the PAT process itself, we have an ongoing
dialogue with the European PAT Team. I
think we are fairly aligned in many ways the aspect which is of interest is our
European regulatory colleagues are hoping that a lot of the PAT concepts will
get incorporated in Q8.
And
the definition of PATs exactly we agreed in Washington will be the DA
definition. And Yokohama will get this
concept in Q8 in a very broad perspective.
So that's one approach we've got.
Plus,
I think, our PAT guidance is becoming final soon with announcements. And we are planning a series of workshops,
inviting our regulatory colleagues from Europe and Japan to participate in the
planning Committee. We're working with
ISPE in setting up some of these workshops in Europe and Japan. And the process has just started.
MEMBER
GOLD: Well, I want to say I'm very
impressed by what I've heard today and reading the black book that you sent
ahead of time. I just want to make sure
we're not so far ahead of the others that we are not going to have unanimity.
CHAIR
BOEHLERT: Okay. Joe?
MEMBER
PHILLIPS: I think definitely I support
everything that's been said by the previous commenters. But the activities in ICH and ASTM are
definitely moving us forward toward the desired state.
What
do we need? We need continued commitment
of the key players, many of whom are sitting in this room, from both sides of
the ocean. We have the regulators, we
have academia, we have industry on both sides.
And
from what I'm hearing, I have a lot of contact with industry and regulators in
Europe and Japan, there's a lot of interest in this activity. And they just want to be kept abreast of
what's happening. And I think the FDA is
to be commended for their efforts to keep everybody well informed. Office of Compliance has been working heavily
in some of these areas.
Any
time some of my colleagues in ISPE have had a question to raise, it's always
easy to get a direct answer from this team.
So I just hope that the same team stays committed and involved because
it takes prime movers and shakers, so to speak, to keep this thing going.
But
it's going very well at the moment. But
who would have thought two years ago we'd be at this point?
MEMBER
HUSSAIN: Okay. Should I move on?
CHAIR
BOEHLERT: Yes, let's move on to the next
one.
Nozer?
MEMBER
SINGPURWALLA: What is the desired state?
PARTICIPANT: California.
(Laughter.)
MEMBER
HUSSAIN: Well, I think I was getting
tired of showing my desired state slides, maybe I ought to keep these. It's prior knowledge.
(Laughter.)
MEMBER
HUSSAIN: So prior is all mixed up right
now so -- no, I think the desired state simply is to -- in a -- sort of a
conceptual way is to increase the level of scientific knowledge that is shared
between the agencies so that we can make more science- and risk-based decisions
which removes -- brings or removes the hurdles for continuous improvement and
reduces the burden on all of us. And
improves the efficiency of the whole system.
I
think clearly we believe that the quality of the products available to the U.S.
public is adequate for intended us. We
have an opportunity to improve the efficiency but at the same time a few years
from now, the complexity of our systems is increasing especially with
biophysics, nanotechnology, and others that are coming. And we are getting a better handle on
variability today through pharmacogenomics and so forth.
So
ten years from now, the current aspects of quality may or may not be
adequate. So I think it's preparation
for the future as well as improving the efficiency of today's systems.
MEMBER
SINGPURWALLA: Well, if that be the case,
then I'd like to comment on that particular issue.
Based
on what I've been hearing and what I've been seeing, I find the progress of
matters is rather academic and conceptual.
There are general principles, principles of quality control, principles
of management, principles of data analysis.
The focus has been a discussion of the principles.
Somehow
we have to get down to a demonstration of how these things work. And I believe I have said this before. What I think is really needed are some
concrete examples. And I'm proposing
that the FDA work in collaboration with industry, the drug manufacturers, both
the generic and what is it called, the creative, the original --
PARTICIPANT: Innovators.
MEMBER
SINGPURWALLA: -- the innovators --
actually I wouldn't like the word innovator if I was a generic drug
manufacturer but I think to work in collaboration with them and come up with
demonstrable examples of how these new ideas come to work.
Otherwise
it becomes like a lecture in a business school where they talk about everything
and need to follow up with case studies.
MEMBER
HUSSAIN: The point is well made and I
think well taken. That's a struggle
because, for example, with the PAT arena, we have about seven submissions at
different stages; one approved, one major complete PAT submission from start to
finish. We actually have a comparability
protocol in house right now. So -- but
it's proprietary. We cannot share it.
And
that's a struggle we often have is we are unable to share what we get because
we're not allowed to share it. We are
working with Pfizer, for example, through a collaborative discussion
development agreement so you will see some publications coming out on some
technologies through that collaboration.
We
are in discussions with two other companies on starting a collaborative
discussion development agreement so there will be publications but I think G.K.
Raju made that point also. I think we
have an acute need for a case study.
Otherwise this remains theoretical and I agree with Gary in the sense we
have been discussing concepts for the last two, three years.
But
at least we have agreed on the concepts.
It's time to move on to some tangible examples that are necessary. And we can do some through our research which
we are doing at Purdue and others. But I
think you really need a real life example.
Somebody has to step up and say we want to share this.
MEMBER
SINGPURWALLA: And there is a little bit
more to that. Not only should there be
an example but in the end, industry should come up to you, to the government,
and say thank you, government, because you made us do these things. We have benefitted and these are things we
would not have done on our own or it didn't occur to us. And you have paved the way and not only
improved our profitability but improved the general state of the art.
I
think you need something much more tangible so that industry can come back and
compliment you if that's possible.
MEMBER
HUSSAIN: I'll look at Helen. Let her answer that one.
(Laughter.)
CHAIR
BOEHLERT: Ken?
MEMBER
MORRIS: Yes, just -- we've talked about
this and in terms of the reduction to practice, if you will, not in the patent
sense, and we're doing things now not exactly in lock step with FDA but in
terms of developing processes by -- in the quality by design sense that
certainly will serve as a partial example, I think.
And
even though it's not being done under the -- it's not being funded by FDA but
they're participating in it so at least we'll get to the point of formulation
of process design, I think, which should be a concrete example that will be
publishable. And that's ongoing
now. So -- but I realize that's only one
and it's only partial but to the point.
MEMBER
HUSSAIN: Okay. I think going on to some other set of
questions, two has subparts. To
facilitate momentum with the desired state, FDA is providing incentives by
ensuring that use of new technologies and additional information about a
minimum acceptable submission standard will not be regulatory requirements.
Gary
raised that again. I think that's an
important point. But will be
opportunities for companies to demonstrate a higher level of process
understanding and risk mitigation. And,
therefore, a basis for regulatory flexibility.
That is example to reduce the need for prior approval of supplements and
so forth.
For
implementation of these concepts a clear demarcation of "minimum" and
optional information is necessary. And I
think this was a significant point of discussion at our ICH Q8. And as ICH Q8 goes to Step 2 in November, you
will see how we have tried to sort of address that.
But
I thought I'll pose this question to you in the sense this is a significant
challenge to sort of achieve this goal.
And especially because the European and the U.S. systems were quite
different. And the expectations in
Europe were different than what we have, the minimal expectation.
So
any thoughts that you can share or any insight that you can share on this would
be very helpful. But let me just
complete the question, Part B of that also.
Quality
by design and manufacturing science are considered foundation for rationale
risk-based decisions. Please recommend
how these principles should be linked to risk to suggest failure mode effect analysis. So we're looking for general principles that,
I think, you would wish us to keep in mind as we progress in this area.
MEMBER
SINGPURWALLA: I think I can respond to
Question B. Question B, of course, the
failure modes and effects analysis is basically a technology mostly based on
engineering or whatever subject matter discipline is at hand to essential work
your way up towards probabilities of certain undesirable events.
And
so those probabilities feed in to the decision, you know, to the decision
tree. So the failure modes and effects
analysis would be an event tree, which traces the course of events which lead
to failure.
And
superimposed on that would be the probabilities of the various sub-events which
lead to failure. And that probability
will be fed into the decision-making paradigm.
So those two are easily, you know, are easily put together as a
package. And that's the right way to go.
So
the question is a good one. And there is
an answer to it.
CHAIR
BOEHLERT: Gerry?
MR.
MIGLIACCIO: Let's talk about A,
Ajaz. I guess I have this -- it almost
implies and A or a B, one or the other.
And
when I think about the optional information, the optional information will come
in degrees, not you either have it or you don't, you know, we can't look at the
NDA as a line in the sand. So you may
get some of that optional information in the NDA and six months later, you may
get much more.
And
so the regulatory flexibility granted with the NDA is at a certain level. And the regulatory flexibility granted six
months down the road when we supplement with that greater process understanding
becomes greater.
So
I'm a little concerned about the clear demarcation statement that it's yes,
there is some information that will be optional. But the degree also has to be
understood. And I like Gary's if this,
then that, you know? If we could put
that map together.
If
you get this, then that's the regulatory flexibility that comes along with
it. And then if you get more of that,
that's what comes along with that.
MEMBER
GOLD: Gerry, I'm not clear -- I don't
see it your way. I interpret that
question as saying what more than we give presently would be advisable for
improving our knowledge or improving the knowledge of the process that we
provide to the FDA? And that I see this
as not asking for necessarily more than we're giving now in order to get
approval.
MR.
MIGLIACCIO: No, in fact, we're not
talking about, you stated more. We're
saying different. The knowledge that
we're providing is different. It's more
science based, more risk based.
MEMBER
GOLD: No, I understand. But I don't see that as asking for anything
more in terms of more science or more knowledge than we're currently supplying
in order to obtain an approval. There is
nothing in that that I see that requires us to elaborate beyond the information
we're providing currently.
However,
if we do provide more information, then this presumably allows us to make
changes with lower requirements, that is lower time limit requirements. So we may be able to go from a PAS to a CB30
or whatever. But I do not see that
statement as saying we must provide more.
MR.
MIGLIACCIO: No, and I didn't imply that
we must. What I'm saying is that what we
provide will be in degrees.
MEMBER
GOLD: Yes, I certainly think that's
possible.
MR.
MIGLIACCIO: There's an impression
sometimes in these discussions that it's all coming in the NDA. And it's not all coming in the NDA. It will be learned. It's a continuous learning process. It will be learned in the first six months of
commercial manufacturing.
And,
therefore, the flexibility has to be there to go back in with more process
understanding and, of course, get greater regulatory flexibility.
MEMBER
GOLD: But, Gerry, I've also seen
instances where companies have more information available to them at the time
of the filing that they don't believe they need to provide because the FDA has
not called for it. And so they just hold
it in their, you know, they hold in their own file.
MR.
MIGLIACCIO: Because the perception now
is if we supply it, it will extend the review period.
MEMBER
GOLD: Correct. Or may extend the review period.
MR.
MIGLIACCIO: That's correct.
MEMBER
MORRIS: Is this trying to get, though,
at the question we were talking about earlier which is, you know, instead now
we have, you know, three batches and then you file? Or is this saying that there's no set number?
MEMBER
HUSSAIN: No, I think -- well, let me
give you an example that I think might be relevant here. I think Gary had some of that information in
his slide in the sense, in particular on the generic side we have a tendency to
be quite conservative in terms of actually requesting an executive batch
record.
And
in some cases or sometimes, that executive batch record is your sort in process
control and so forth. So any change
requires a supplement.
But
that is because we often have limited information in how to establish
specifications, one biobatch and so forth.
So
that is the current way of thinking.
That's fine.
What
I might suggest is the optional type of information might be you have
pharmaceutical development information and other information that provides much
more flexibility that would not -- that would allow us to move away from that
executive batch record as the sort of a basis of sort of establishing something
to something more of process understanding basis.
So
that's how we're sort of approaching it.
MEMBER HUSSAIN: Gary and Moheb, any thoughts on this?
(No
response.)
MEMBER
RAJU: Judy?
CHAIR
BOEHLERT: Yes, G.K.?
MEMBER
RAJU: On the two questions, Ajaz, I go B
first and A second.
On
B, I believe that the priority should be since manufacturing science and
quality by design are both levels of performance and states of knowledge and
can be changed by processes, that on B the priority should be on defining those
levels and the processes that enhance it.
And
the tools -- so the tools only have context in that -- only have meaning in
that context. I do not want to say
please recommend how these principles should be linked to risk tools yet. We have to focus on the characterization and
the processes for it.
The
tools can be a tool set just like we have a lot of tool sets. The links shouldn't be made too early because
we haven't done the first step first. So
let's keep the tools in a portfolio of tools and understand them, bring them in
from outside the industry into ours.
Let's
focus on our industry and defining what we do transparently based on principles
of science. And then connect the
tools. So that would be my thought on
that.
And
there's a scientific process to the tools, too.
MEMBER
HUSSAIN: If I may --
MEMBER
RAJU: Sure.
MEMBER
HUSSAIN: -- suppose we remain with an
empirical approach to this so we don't have a mechanistic understanding and so
forth, so we are seeking causality or we're seeking correlation through an
empirical model approach, say design of experiment, okay?
Now
the number of potential factors that may be critical can be a large number
depending on the process. And an
approach could be is this is -- I'm basing this on the presentation by Amgen at
Arden House, is you start with a failure mode effect analysis based on all your
expert opinion information that's there based on historical know how to sort of
tease out what may be the critical variables.
And then design your experiments around that.
So
that is sort of another way of looking at it.
So that's -- there are many different options there because I think if
somebody wants to do a design of experiments, they really have to manage the
resources and their commitment very carefully.
Otherwise that can get out of hand.
So
that's one way of approaching that. But
the other way of approaching that is through screening experiments early on and
then sort of designing -- defining your design space and then doing a failure
mode effect analysis. So you need to
have flexibility of going either way.
MEMBER
RAJU: So this is Bob Sweeney's work at
Amgen?
MEMBER
HUSSAIN: Right.
MEMBER
RAJU: He did a nice job of saying this
is the process. Here are the variables.
And then he put fault modes into context.
MEMBER
HUSSAIN: Correct.
MEMBER
RAJU: Because he did that, it was a very
good story.
MEMBER
HUSSAIN: Yes.
MEMBER
RAJU: But it's not clear that that's
been done. And if it's not been done,
then it has to be done first before we bring the FME -- the tool only has
context within a goal and a process to get to that goal. So I think it works fine that way.
In
terms of A, I have a somewhat similar answer but at this point because the
demarcation, you said that what you get in a submission is variable. And you said you have some information, more
information, sometimes you have less information, and sometimes you have
different.
The
criterion of what is more and what is important has not been laid in place
yet. So it's somewhat dependent on the
company and their interpretation and their strategy.
It
seems like two things would help on A.
First, you said it was minimum and optional.
Probably
independent of the answer to A, to make sure that everybody believes -- that
everybody in the FDA believes that and will implement that is extremely
important because everybody -- I can hear a number of cases where people say I
know Helen and Ajaz, they would believe that.
But how do I know about the guy who is going to do my review? Or the person who is at the field, for
example, which may not be relevant in this case.
So
just making sure that what you believe in is somewhat uniform although we all,
as human beings, we'll never be uniform.
Second,
how about making it one of two possibilities?
Making it the company's choice because it's still somewhat not fully
characterized, to present to you here is minimum. And have here is the optional, what shall we
do with it? Either submit them both and
say you make a decision based on this and we can get a better deal based on
this?
Or
here is the optional. Can we discuss
with you whether we should submit it or not?
So
they make their first call on minimum versus optional. They decide to submit it. You start with the minimum and your
specifications get changed based on that.
But they don't pay the price for the optional because you say they
wouldn't.
You
get the minimum and the paying the price is more in the context of a
reward. And it could be done informally
first before it's formal. How about
that? It seems like -- just think aloud
now.
DR.
NASR: If you allow me to make a simple
comment here. I think this is very good
discussion. But in my mind the issue
before us is much simpler. And let me
elaborate a little bit.
I
think the existing system that we have is working. Why is it working? Because we have quality pharmaceuticals in
the market. So the existing system is
working.
So
in the future, I think companies, sponsors, will have to follow one of two
approaches. The existing regulatory
process and the regulatory framework with the guidances in ICH and the
submissions and the meeting or lack of or whatever.
And
we will continue on and when you make a change, you have to come to us, we'll
supplement. And we'll evaluate the
supplement and we'll make recommendation.
And you go ahead and you manufacture or not manufacture.
The
future paradigm we are describing and sharing with you today and Ajaz, I think
would agree with that over the years now is you share with us in advance, and
advance means either at the NDA stage or shortly after or long after, your
understanding of the manufacturing process, your ability to deal with the
change, and then back to such a change on the critical quality attributes.
And
based on that understanding, you're sharing in the form of pharmaceutical
development report or comparability protocol or whatever, we will give you the
freedom to manage your own change.
So
in my mind, it's very simple. You can
stay put and do what we are doing now and continue to have quality
pharmaceuticals in the market. Or if you
want to follow the quality by design and the new approach, which we believe is
beneficial to you, to us, and to the public, and that provide you with the
regulatory relief that you have been asking for for years and years to manage
your own manufacturing process.
So
in my mind, it's fairly simple.
MR.
MIGLIACCIO: Judy?
Moheb,
do you accept that we will have some hybrid situations?
DR.
NASR: We do. When I said comparability protocol, that's a
hybrid.
MR.
MIGLIACCIO: Yes.
DR.
NASR: When you talked about supplements
shortly after, that's a hybrid.
MR.
MIGLIACCIO: Right. So we'll have --
DR.
NASR: It's not a clear cut --
MR.
MIGLIACCIO: Right.
DR.
NASR: -- either or.
MR.
MIGLIACCIO: Okay.
DR.
NASR: And I think the third point that I
failed to make, Gerry, and I'm glad you made this comment, is I think our role
collectively is how to move from the existing system to the future paradigm.
So
we're going to have two different regulatory approaches. I hope we don't call this two different
quality system. One if more inferior
that the other. We will have two
different regulatory processes, the existing one and the one that fits better
with the future paradigm.
And
we should make products available to the public based on both processes. What we should work on collectively, because
I think from what I'm hearing today and I heard before, we are in agreement, is
how to move from the existing system to the future paradigm without penalizing
industry or the public.
CHAIR
BOEHLERT: Ken?
MEMBER
MORRIS: Something that's bothering me a
little is that, you know, what we've been talking about all along is that
industry should essentially be telling FDA what it thinks it needs to do in
order to justify its decisions in dosage form and process development
manufacturing. Not dictating but saying
here's what we think we should do -- which is sort of what we're saying in Part
A.
And
I don't have an answer to this. But what
bothers me a little bit is that if that's what we're really saying, then in
principle what you would expect is that the company would put together what it
considers necessary for itself in terms of a development report and share that.
Now
the question of what's minimum then really is almost a moot point because
minimum would have been passed long before you got to that point. Because if you're going to do minimum, then
you're using Moheb's other -- you're using your other eventuality where you're
just following the old system.
So
it seems like you've long passed -- that ship's sailed, I think.
MEMBER
HUSSAIN: No, Ken, I think the point
you're making is a good one. And I think
the only -- I think the primary reason for asking this question is because this
is the question that seems to come up again and again in our expert working
group discussions.
And
primarily I think I agree with Moheb in that at least in the U.S., with our
peer review process, with our quality system, it's not an issue within the U.S.
to manage this. I think we can easily
manage this with the new way.
It's
simply a question to sort of prepare ourselves for the future discussions in
say Japan in November. Is, I think,
Judy, if you would permit me, if John Berridge wants to come -- maybe I'll
invite him to comment also on that -- my thoughts were that in a sense the
uncertainty level seems to remain within the regulatory affairs, within the
industry itself. The hesitation to share
any information is there. So you still
have that.
And
what I'm hoping is we can find an opportunity to minimize that concern also at
the same time I think get to the right decisions, ask the right questions, and
get the right answers fairly quickly instead of going through an elaborate
process.
There
is a level of concern, hesitation out there, which is quite significant let me
tell you, I'll share this. It's trying
to minimize that.
DR.
NASR: Before John comes in, I want to
add one thing in response to G.K., who raised a very good point. Because what you heard from you that people
out there are saying Ajaz, Helen, David, Janet, and so forth believe in
this. How about the reviewers?
I
think, I hope I made it clear today that the Office of New Drug Chemistry has
made a commitment to change the way we do our work and reorganize in a way to
facilitate the implementation of the new paradigm.
And
this is not just me talking. I think we
have senior leadership here of the Office in attendance and the Office is
committed to do that. It is not just
Helen and Ajaz.
MEMBER
RAJU: And if you look at this
presentation you made and the one you made at Arden House, the amount of
changes you are making in the new drug chemistry seems to be really rapidly
different from a year ago. I've never
seen that kind of momentum in any place before.
It's clear.
CHAIR
BOEHLERT: John?
DR.
BERRIDGE: Yes. So I don't want -- I don't think there is any
point in my repeating the points that have been made. But I think there's one other thing to
consider about the communication and the way we get the new paradigm across.
One
of the things we discussed in the expert working group is to build on the model
that was designed, and Joe will probably be very familiar with this, that was
adopted by the Q7A Team, which was actually to construct an education process
that could be rolled out around the world, that would use a common set of
training materials that would be available to regulators and industry alike
that would clearly articulate exactly what it was we wanted to achieve and the
implications thereof.
I
think that would actually strengthen the understanding and remove the degree of
uncertainty and I'm almost bound to say fear that exists. And I think an element of the fear is driven
by the unknown.
So
the development of a training program, you mentioned particularly regulatory
affairs colleagues who haven't been quite as intimately involved in this
process as maybe their scientific counterparts, if we can get that adopted and
pushed out, I think that would be also a very valuable process for removing
some of the concerns that have been expressed this afternoon.
MEMBER
HUSSAIN: I think that's an important
point. And there are a number of
aspects, if I may see the Committee's thoughts and recommendations on this.
Helen
and I have sort of discussed this at length in the sense we have met with a
number of companies, one on one basis.
They have shared some of their ideas of how this report might be and how
the case studies might develop and what the criteria should be and so forth.
I
think meeting each company one at a time clearly is what we are going for, but
we're not getting something in the public domain which would be an example, the
case studies, and so forth.
The
proposal might be to the Committee just to consider maybe we form a working
group under this Committee to actually get to some of this tangible outcomes
quickly because I think we need a framework to work on this.
So
if the Committee would agree, I would propose that I think we might, following
this meeting, start the dialogue and put a working group under this Committee
to work on some of these aspects.
CHAIR
BOEHLERT: Any comments on that proposal?
MEMBER
SINGPURWALLA: It's a good idea as long
as I don't have to be on it.
(Laughter.)
CHAIR
BOEHLERT: Okay. Are we either ready --
MEMBER
HUSSAIN: Okay. I think there are a number of activities
going on in ONDC and OGD and we actually just talked about that. We have Office of Biotechnology Products also
gearing up for a number of things. And
if you saw the pharmaceutical technology report, you saw what Keith Beverly is
doing.
But
at this meeting, we didn't have time to bring him on board also. But what do you think, what advice, or what
recommendations do you have for Moheb and Gary that might help move them
further?
I
think they're doing a tremendous job already.
I think there is still aspects of communication, coordination, and so
forth that will occur. But anything you
can add would be a real help.
DR.
FACKLER: Judy? For a number of companies, somebody mentioned
just a minute ago the unknown. A delay
in an approval has a series economic impact on a company. And submitting more information than we have
been doing historically to an organization that confesses to being hopelessly
understaffed seems like a prescription for delaying one's approval.
MEMBER
HUSSAIN: Hopefully it is not more
information. It is less data, more
knowledge, and then more concise.
Hopefully we can transition to that.
DR.
FACKLER: Well, and that's what I think
needs to be clarified to companies in general is that not just a reassurance
that things will go smoother or faster but some -- certainly a concrete example
would be a good thing but it's too ill defined right now, I think, for
companies to risk changing something that they can measure right now.
You
know you make a submission and you have a fairly good understanding for when
you might get the first review back or the first approval. And it's the unknown that really is causing I
think a lot of hesitation in companies.
MEMBER
HUSSAIN: If I may, sort of building on
that, I think the whole thing begs for some concrete examples, criteria, and so
forth. That's what the next step logical
is. And I think to get there a working
group might be the best option to do that.
And
maybe I'll follow up with Judy and try to assemble a group under this Committee
that will report to this Committee.
CHAIR
BOEHLERT: And Pat?
MEMBER
DeLUCA: Yes, I got the impression that
in the submissions that there was information that was lacking. And the reviewers had to, at times, try to
tease out information or try to even decipher what the rationale was for doing
something.
And
I'm just wondering if that in moving from the existing system to the new
paradigm that the filings should include from the companies the rationale, the
summary, and then plans for improvement on the process that's going to take
place?
So
it's just not, you know, the process improvement should not be optional. It should be something that is expected even
after approval.
(Laughter.)
MR.
MIGLIACCIO: Judy?
CHAIR
BOEHLERT: Gerry?
MR.
MIGLIACCIO: In the ideal case that there
are no undesirable sources of variability in the process, why would you change
it in the ideal case?
MEMBER
DeLUCA: Well, you wouldn't change
it. I mean the only thing is is that you
should have some idea is there a way to improve the process. But saying that that's going to work but at
least you have some strategy that you can look into and investigate. And either prove or disprove it. If it is possible to improve the process,
then there is some effort to improve it.
MR.
MIGLIACCIO: An examination for example?
MEMBER
DeLUCA: That's right. I mean it's not compulsory that you improve
it. It's just that did you have a plan
or some strategy for improving it?
MEMBER
HUSSAIN: I think we probably will touch
upon this tomorrow also. I think the key
aspect is in terms of a decision to approve, I think in some ways you have to
look at that as a decision of an acceptable risk assessment that allows the
product to come out. Sometimes you have
to have special decision-making criteria for a very essential drug and so
forth.
But
generally a decision to approve means you have met the safety and efficacy
standard. And in many ways continuous
improvement the way I see it is is an improvement to improve efficiency,
improvement to bring new technologies, simply from a business case.
But
at the same time, there is a category of changes which are necessary. The process is not capable of meeting those
standards that we approve. Tremendous
failure and so forth.
So
when the process is not capable, there has to be a way to sort of improve
that. And we do it through enforcement
action today, concern degree and so forth.
So there is a category change which the FDA will come back to ask you
for the change.
So
-- but the other type of changes are, I think, are continuous improvement, to a
large extent efficiency improvements.
With that, I think -- oh sorry.
Go ahead.
MEMBER
RAJU: There are two presentations. The Office of New Drugs' presentation was
extremely impressive. One of the best
I've seen. The whole science into the
mission and the science principles were very powerful, knowledge gaining,
bringing in pharmaceutical development.
I
will echo, however, that probably process capability shouldn't be in
there. It should instead be process
stability because it's too early.
Process capability comes later.
It should be stable first.
But
if you include process stability, it would fit in beautifully.
In
terms of the overall piece, if you say special cause analysis before you go to
statistics, that's a beautiful place to bring in the FMEA actually. That's the right tool for that.
So
this is actually quite strong. I'd be
curious to hear your good scientific principles sometime in an offline.
In
terms of the Office of Generic Drugs, this is the first time that I've learned
about the size of the submissions and how long it takes. It doesn't seem acceptable from a social
point of view. I was really worried as a
citizen.
I
think there should be a synergy of leveraging the old innovator drug's
knowledge back here. But then there's a
whole other dimension of resources and prioritization that's beyond probably
the scope of this Committee or at least me that is extremely important that
there has to be something done about.
MEMBER
HUSSAIN: G.K., just a comment on, I
think it's a matter of semantics and vocabulary. I think process capability often we use it in
the new drug side from a slightly different perspective in a sense. How we often -- I'm very familiar with how we
set dissolution specifications. I use
that as an example.
If
you have say ten batches that you have used in the clinical setting, so you
have ten clinical batches over the clinical drug year. What we often will do is, I think, the
decision to set a specification and an acceptance criteria, mostly acceptance
criteria would be to maybe fail a couple of batches. That's what we often refer to. But it's not truly a calculated process
capability.
CHAIR
BOEHLERT: Garnet, did you have a
comment?
MEMBER
PECK: Just what I'm thinking about is an
overview without coming with specific recommendations or answers to Question 2
and 3.
I
feel that you have given in the beginning of Question 2 a great preamble. You have a number of suggestions here about
what might be done within a particular organization to demonstrate that they
understand the process, that they probably understand the product, the system
required to put together the product, which then would allow the Agency to have
this flexibility in terms of the regulatory affairs.
I
couldn't come up with something better than minimal or optimal. I think there's got to be another way of
expressing that. I don't think that's
the right way to do it. But there's got
to be some demarcation.
But
if we can have some feeling for PAT guidance, ICH newer thoughts, and we start
to apply these, it seems to me that we would have a total confidence in all
avenues that we were proceeding in be it new drug or be it generic.
And
I think the generic situation is a tough one because of the number of
filings. That is -- this number, I
hadn't seen this year's number and it's getting pretty large.
But
you are attempting to present, if you will, the possibilities of regulatory
flexibility with better understanding of the process and the product.
CHAIR
BOEHLERT: Anyone else?
(No
response.)
CHAIR
BOEHLERT: Ajaz, are you satisfied with
what you've heard?
MEMBER
HUSSAIN: No, I think this was a very
valuable discussion.
CHAIR
BOEHLERT: Okay.
MEMBER
HUSSAIN: And I think I was just kicking
myself for not bringing a piece of paper and pen to take some notes but the
transcript will have that.
But
again, thank you very much for the discussions.
CHAIR
BOEHLERT: Okay. Well, I'd like to thank everybody as
well. And if that's it, then we will
adjourn for this evening and reconvene tomorrow morning at 8:30.
(Whereupon,
the above-entitled meeting was concluded at 5:10 p.m.)