Wednesday, May 8, 2002

8:30 a.m.







5630 Fishers Lane

Rockville, Maryland



Vincent H.L. Lee, Ph.D., Acting Chair

Kathleen Reedy, R.D.H., M.S., Executive Secretary


Gloria L. Anderson, Ph.D., Consumer


Mary J. Berg, Pharm.D.

Joseph Bloom, Ph.D.

Judy P. Boehlert, Ph.D.

Patrick P. DeLuca, Ph.D.

John Doull, M.D., Ph.D.

Arthur H. Kibbe, Ph.D.

Marvin C. Meyer, Ph.D.

Nair Rodriguez-Hornedo, Ph.D.

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


Tom Layloff, Ph.D.


(Robert) Gary Hollenbeck, Ph.D.

Michael S. Korczynski, Ph.D.

Kenneth R. Morris, Ph.D.


Leon Shargel, Ph.D., R.Ph.

Efraim Shek, Ph.D.


Thomas Garcia, Ph.D.

Tobias Massa, Ph.D.


Ajaz Hussain, Ph.D.

Richard Adams, Ph.D.

Nicholas Buhay

Yuan-yuan Chiu, Ph.D.

Peter Cooley, Ph.D.

Steve Miller, Ph.D.

Bryan Riley, Ph.D.

Helen N. Winkel



Call to Order: Vincent H.L. Lee, Ph.D. 4

Conflict of Interest: Kathleen Reedy 4

Process Analytical Technology

Introduction and Overview:

Tom Layloff, Ph.D. 9

Subcommittee Reports and Next Steps:

Ajaz Hussain, Ph.D. 30

Rapid Microbial Testing - Update

Introduction and Overview:

David Hussong, Ph.D. 70

Bryan Riley, Ph.D. 71


Michael S. Korczynski, Ph.D. 76

Open Public Hearing:

Jeanne Moldenhauer 104

Scott Sutton, Ph.D. 108

Blend Uniformity

Introduction and Overview:

Ajaz Hussain 120


Tobias Massa, Ph.D. 131

Thomas Garcia, Ph.D 137

Regulatory Issues Related to Crystal Habits


Introduction and Overview:

Yuan-yuan Chiu, Ph.D. 176

ONDC Issues:

Steve Miller, Ph.D. 179

OGD Issues:

Richard Adams, Ph.D. 200

(Invited Guest, Kenneth R. Morris, Ph.D.)

OPS Updates and Topics for the Next Meeting 223


1 P R O C E E D I N G S

2 Call to Order

3 DR. LEE: Good morning. I am Vincent Lee,

4 Chair of the Advisory Committee for Pharmaceutical

5 Science. I am calling the meeting to order.

6 The first order of business is the

7 Conflict of Interest. I ask Kathleen Reedy to read

8 us the statement.

9 Conflict of Interest

10 MS. REEDY: Acknowledgement Related to

11 General Matters Waivers. Advisory Committee for

12 Pharmaceutical Science, May 8, 2002.

13 The Food and Drug Administration has

14 prepared general matters waivers for the following

15 special Government employees: Drs. Marvin Meyer,

16 Mary Berg, Judy Boehlert, Jurgen Venitz, Gordon

17 Amidon, Vincent Lee, and Patrick DeLuca, which

18 permit their participation in today's meeting of

19 the Advisory Committee for Pharmaceutical Science.

20 The committee will discuss: (1) receive

21 summary reports and provide direction for the

22 Process Analytical Technology Subcommittee; (2)

23 discuss and provide comments on regulatory issues

24 related to crystal habits - polymorphism; (3)

25 discuss problems and provide comments to form a


1 scientific basis for establishment of acceptance

2 limits for microbiological tests that use newly

3 developed technologies that do not rely on colony

4 counts, and their application as process controls

5 and product release criteria; and (4) discuss the

6 current status of, and future plans for, the draft

7 FDA guidance entitled "Guidance for Industry,

8 ANDAs: Blend Uniformity Analysis."

9 Unlike issues before a committee in which

10 a particular product is discussed, issues of

11 broader applicability, such as the topics of

12 today's meeting, involve many industrial sponsors

13 and academic institutions.

14 The committee members have been screened

15 for their financial interests as they apply to the

16 general topics at hand. Because general topics

17 impact on so many institutions, it is not prudent

18 to recite all potential conflicts of interest as

19 they apply to each member.

20 FDA acknowledges that there may be

21 potential conflicts of interest , but because of

22 the general nature of the discussions before the

23 committee, these potential conflicts are mitigated.

24 We would also like to note for the record

25 that Drs. Leon Shargel of Eon Labs, Inc., Efraim


1 Sheik of Abbott Laboratories, Thomas Garcia of

2 Pfizer, Inc., Tobias Massa of Eli Lilly & Company,

3 Aziz Karim of Takeda Pharmaceuticals North America,

4 and Jack Cook of Pfizer Global Research and

5 Development are participating in this meeting as

6 Industry Representatives, acting on behalf of

7 regulated industry.

8 As such, they have not screened for any

9 conflicts of interest. With respect to FDA's

10 invited guests, there are reported interests which

11 we believe should be made public to allow the

12 participants to objectively evaluate their


14 Dr. Kenneth Morris reports that he serves

15 as a consultant, speaker, researcher, and has

16 contracts and grants from multiple pharmaceutical

17 companies. Dr. Gary Hollenbeck reports that he

18 owns stock in the University Pharmaceuticals of

19 Maryland and Aerogen, Inc. Dr. Hollenbeck would

20 also like to disclose that he serves as a

21 consultant and scientific advisor to University

22 Pharmaceuticals, as well as other pharmaceutical

23 companies.

24 Finally, Dr. Hollenbeck reports that there

25 are numerous companies contracts within the


1 University Pharmaceuticals of Maryland. Dr.

2 Michael Korczynski reports that he serves as the

3 Senior Vice President of Mikkor, which has a

4 fiduciary relationship with Afton Scientific

5 Corporation.

6 Dr. Korczynski also reports that he owns

7 stock in Abbott, Johnson & Johnson, Pfizer, and

8 GlaxoSmithKline. In addition, he serves as a

9 consultant for Lighthouse Instruments, LLC and

10 Afton Scientific. Finally, Dr. Korczynski speaks

11 for AAI and is a scientific advisor for Afton

12 Scientific Corporation.

13 In the event that the discussions involved

14 any other products or firms not already on the

15 agenda for which FDA participants have a financial

16 interest, the participants are aware of the need to

17 exclude themselves from such involvement and their

18 exclusion will be noted for the record.

19 With respect to all other participants, we

20 ask in the interest of fairness that they address

21 any current or previous financial involvement with

22 any firm whose product they may wish to comment

23 upon.

24 DR. LEE: Thank you, Kathy.

25 I would like to go around the table and


1 have the members introduce herself or himself,

2 beginning with Mary Berg.

3 DR. BERG: Mary Berg, College of Pharmacy,

4 University of Iowa.

5 DR. DOULL: John Doull, University of

6 Kansas Medical Center.

7 DR. DeLUCA: Pat DeLuca, University of

8 Kentucky.

9 DR. MEYER: Marvin Meyer, Emeritus

10 Professor, University of Tennessee.

11 DR. KIBBE: Art Kibbe, Wilkes University.

12 MS. REEDY: Kathleen Reedy, Food and Drug

13 Administration.

14 DR. ANDERSON: Gloria Anderson, Callaway

15 Professor Chemistry, Morris Brown College, Atlanta.

16 DR. BLOOM: Joseph Bloom, University of

17 Puerto Rico.

18 DR. VENITZ: Jurgen Venitz, Virginia

19 Commonwealth University.

20 DR. BOEHLERT: Judy Boehlert. I have my

21 own consulting business to the pharmaceutical

22 industry.

23 DR. RODRIGUEZ-HORNEDO: Nair Rodriguez,

24 College of Pharmacy, University of Michigan.

25 DR. SHEK: Efraim Shek, Abbott


1 Laboratories.

2 DR. SHARGEL: Leon Shargel, Eon Labs.

3 DR. LAYLOFF: Tom Layloff, Management

4 Sciences for Health.

5 DR. KORCZYNSKI: Mike Korczynski, Mikkor

6 Enterprises.

7 DR. HUSSAIN: Ajaz Hussain, Office of

8 Pharmaceutical Science, FDA.

9 DR. LEE: Vincent Lee, University of

10 Southern California.

11 Before I talk about the agenda, I have one

12 clarification to make for the record. When I

13 summarized the meeting yesterday, I gave somebody a

14 heart attack by what I said, not intentionally.

15 Let me see if I can get it right this time.

16 Otherwise, a new Chair.

17 That is, for Class I BCS Type 1 products,

18 since in vivo B was waived for fasting conditions,

19 the committee feels they ought to be waived for the

20 fed conditions, as well. That is for the record.

21 Today, we are going to be talking about

22 four issues. The first one is on Process

23 Analytical Technology, and so on, and so forth, and

24 I think it would be appropriate for me to invite

25 Ajaz Hussain to come up to the podium and give us


1 the introduction.

2 DR. HUSSAIN: Vince, what I would ask is

3 that Tom start the report to you guys for the

4 Advisory Committee, and then I will follow Tom and

5 gave sort of a progress report on what we have done

6 at FDA and propose some next steps, and then we can

7 have a discussion.

8 Process Analytical Technology

9 Introduction and Overview

10 Tom Layloff, Ph.D.

11 DR. LAYLOFF: Good morning.

12 [Slide.]

13 I would like to talk to you today about

14 the Process Analytical Technology Initiative, which

15 is an FDA Initiative and for which I serve as the

16 chair of the committee.

17 [Slide.]

18 In pharmaceutical development, the first

19 thing that happens is an active pharmaceutical

20 ingredient is identified to be a therapeutic agent.

21 My symbols didn't work out. That is supposed to be

22 an alpha on the far left, which is the beginning of

23 a process, and the question mark is supposed to be

24 an omega, which is the end of the process.

25 So we have an active pharmaceutical


1 ingredient, which is a therapeutic agent, which is

2 processed through a series of steps to give us a

3 therapeutic endpoint in the body.

4 The technologies we normally use for the

5 assessment of the active pharmaceutical ingredient

6 typically involve determinations of impurities,

7 looking for the active ingredient, and those

8 technologies often move forward into development

9 and into control, and then again show up again in

10 the body fluid analysis.

11 The question is, is that appropriate.

12 [Slide.]

13 Is the API an appropriate process quality

14 surrogate marker for a process for manufacturing?

15 The focus has been on the API without regard for

16 excipients and processes, so that if you look at a

17 pharmaceutical process, you take the active

18 pharmaceutical ingredient, the technologies that

19 you use, and the identification of a moiety, and

20 you move it into development, and you keep tracking

21 the pharmaceutical ingredient using the same

22 assessment technologies.

23 Now, as you add excipients in blending,

24 you continue to watch the active pharmaceutical

25 ingredient, and the question is, is that a


1 reasonable thing to do. It is a univariate handle

2 on a polyvariate problem and the question is, is

3 the API a good surrogate marker for the process, in

4 many cases it is, and cases it is not.

5 [Slide.]

6 The PAT is to change the paradigm. Look

7 towards broader product quality dimensions. Use

8 new assessment technologies and new product

9 assessment targets. Shift from interrupted unit

10 operations to on-line/in-line assessment on

11 continuous process streams.

12 Now, the way it is done currently, you

13 blend, stop, sample, test, move to the next step,

14 so you have a series of steps, staircase steps as

15 you move to product. The intent of PAT is to move

16 technologies on-line/in-line, so that line streams

17 out to a smooth presentation.

18 [Slide.]

19 This is one of Ajaz's slides which I

20 borrowed, and it shows that if we go to the new

21 technology in-line, you actually have the device

22 sampling the process itself and monitor it to a

23 performance endpoint rather than sample, take it to

24 the laboratory, hold everything until the results

25 come back, and then proceed.


1 [Slide.]

2 The assessment tools and support systems

3 and technologies are available to improve the

4 consistency, reduce bad production and recalls.

5 [Slide.]

6 The consistency assessment tools that are

7 available now for it are spectrophotometric methods

8 like near infrared, laser-induced fluorescence,

9 Raman, various ATR, attenuated total reflectance

10 methods when you are using crystals or optics,

11 fiberoptics.

12 There are other technologies, such as

13 acoustic monitors, image field analysis,

14 thermometers, pH meters, oximeters, on-line chip

15 analyzers, many of which are already used in the

16 biotechnology industry. We will come back to some

17 of these.

18 Image field analysis is quite interesting

19 because it's like preparing a stew. You put

20 everything in the bottom and then you stir it up

21 and see if it's uniform. In image field, you look

22 at the image at intervals and see if it's

23 consistent, so it's basically a variance of image.

24 When the variance is reduced, then, it's blended to

25 its completion.


1 [Slide.]

2 These are the members of the PAT Committee

3 - Gloria Anderson, Joe Bloom, Judy Boehlert, Art

4 Kibbe, and then a bunch of individuals who had

5 applied through the Federal Register announcement

6 and are on the committee. The rest of them are

7 listed there.

8 They come from various organizations and

9 industries, and the PAT was developed into four

10 working groups. There is Applications and

11 Benefits, which was chaired by Art, Chemometrics by

12 Mel Koch, Process and Analytic Validation by Leon

13 Lachman, Product and Process by Judy Boehlert.

14 I have asked Judy to give a few remarks on

15 her committee, if she would. Did you want to say

16 anything, Judy?

17 DR. BOEHLERT: When you put it that way,

18 Tom, I would be happy to say a few words. What I

19 did this morning when Tom asked me to make some

20 comments was to go back over the conclusions that

21 came out of my committee.

22 We have a very productive discussion, and

23 this was once again the Process and Product

24 Development Working Group, and I am going to just

25 read some of the conclusions that we came to.


1 First, the group agreed that the benefits

2 are under-utilized, there needs to be some selling

3 done here. People don't realize what PAT can do for

4 them. It would apply to most areas of the

5 manufacturing process, but there are different

6 levels of maturities for some of these techniques

7 that Tom mentioned.

8 Some are ready to go now maybe, others are

9 a bit further away. It may not work in all cases.

10 There are instances where PAT is not going to help

11 you very much. The feedback controls that you have

12 on the process may mean that when something goes

13 wrong, you don't lose the entire batch. You have an

14 opportunity to make corrections while the batch is

15 processing, and that is a good thing.

16 Guidance that FDA comes out with shouldn't

17 be limited to when you think about the use of

18 alternative methodologies and technologies. Tom

19 had a list, but there may be other techniques we

20 haven't even thought of yet, that will be applied,

21 and there needs to be a mechanism for putting those

22 in place.

23 What we are doing is going to a

24 multivariate approach. The variables that may be

25 more pertinent to the process, they might not be


1 the ones we looked at in the past, and there needs

2 to be a regulatory way to be able to submit those

3 because, in fact, the parameters that you filed may

4 not be the ones you are measuring now, and, in

5 fact, you may not comply.

6 There will be engineering issues involved

7 with this approach because if you start trying to

8 apply new technologies to old systems, you may need

9 to look at those old systems. It may only apply to

10 some process or some operations in a process, and

11 that is not necessarily bad. There may be

12 incremental advantages to being a part of the

13 process.

14 Very often, whether you implement PAT or

15 not is going to be a business decision. People

16 didn't really see technical down sides, but it is

17 timelines, how does it impact on timelines, do you

18 have the resources available to implement these

19 technologies. It has potential for reducing OOS,

20 potential for perhaps predicting product

21 performance, things like dissolution.

22 It is not going to do away with stability

23 studies. It may predict that you have a more robust

24 product, but people don't see that that is going to

25 go away.


1 So the conclusion of the group in general

2 was that there is an advantage to using PAT because

3 it leads to a consistent, more high quality

4 product, however, there was a down side, and I left

5 that to last. There are a number of people in

6 industry say it is not broke now, don't fix it. It

7 works.

8 DR. LAYLOFF: Art, did you want to comment

9 on your committee?

10 DR. KIBBE: To respond to your question,

11 the committee was filled with wonderful people, but

12 we were looking at applications and benefits, and

13 for us, we started looking at the broadest

14 application. We felt there was application for the

15 technology of PAT to be applied to practically any

16 product in any environment.

17 We thought that there were going to be

18 great benefits and that there would be a learning

19 curve, and as companies began to use technology and

20 put it in place, and begin to see the benefits,

21 that the next step, and the next step would come a

22 little easier.

23 All four subcommittees responded to some

24 general questions, and I think therein lies the

25 direction that we want to give to the agency in


1 terms of its guidance. We started out as a

2 complete group, a definition of PAT, and our

3 subgroup added the word "critical."

4 We felt like just because you can measure

5 it, doesn't mean you should measure it. You should

6 be measuring what is important to the outcomes.

7 One of things that we were concerned about is that

8 the process, the user of PAT generally generates a

9 huge amount of data as opposed to the current

10 method of sampling and getting discrete answers

11 about the concentration of actives in the sample,

12 or so on.

13 So the question arose in terms of the

14 guidance what do you do with all this data, and is

15 that data going to be much tighter than our general

16 requirements for any individual product, what is

17 the agency going to do about that data, and so on.

18 So the issues surrounding the guidance

19 boil down to hoping that, first, the agency and the

20 industry could work cooperatively towards improving

21 the manufacturing process in every area, that it

22 would not be viewed by companies as a means for the

23 agency to become more restrictive on their ability

24 to manufacture or make things. It wouldn't be

25 viewed as something other than it is, which is a


1 way of encouraging or empowering the industry to

2 move forward in a very open and honest way to

3 improve its own bottom line and hopefully, as a

4 result, that benefit the patients and the general

5 cost of health care.

6 We have a set of guidance suggestions,

7 which I think -- I don't know whether we want to go

8 through them now or you want to hold them until --

9 DR. HUSSAIN: I think it will help if we

10 could walk through them. That was an excellent set

11 of points, recommendations that came out I think.

12 DR. KIBBE: First, the guidance must allow

13 the development of PAT whose endpoint is a

14 signature of the quality of the process, because

15 the data that we capture is going to be interesting

16 in terms of the way it looks especially to some of

17 us old hands who expect to see, you know, some

18 nice, discrete numbers, and we get this pattern.

19 If you look at near infrared or some

20 others, you get a pattern of what the process looks

21 like and you say, okay, what does that mean, I have

22 got all these wonderful curves. We are going to

23 have to start accepting a signature or an pattern

24 as an endpoint rather than a discrete number.

25 It implies that we use it in an


1 environment of continuous improvement without undue

2 regulatory burden. While we see this as a really

3 beneficial process going forward, as with any

4 technology, when you are early on, you are not

5 going to ever get the best result, and we think

6 especially when most of the data that we looked at

7 was near infrared, and there are lots of other

8 techniques available, we are going to see great

9 improvement as times goes on.

10 All products have critical attributes,

11 quality attributes that need to be assessed, but

12 not everything needs to be assessed. Process

13 variables exist that can be controlled and

14 maintained, and these critical quality attributes

15 within acceptable limits, PATs are applied to

16 achieve both understanding and control of the

17 process and are causally linked to the product's

18 critical quality attributes.

19 We think it's a great way of making sure

20 that you never have to eliminate a batch, and we

21 were talking about continuous manufacturing

22 processes, how do they batch them now. Well, if

23 they are in a continuous process, they just take an

24 arbitrary time and they say everything produced

25 today is one batch, everything produced tomorrow is


1 another batch.

2 I think we are going to have to change

3 some of our thinking about how we lot and batch and

4 examine things, but that is going to be something

5 to look on down the road.

6 There are new and developing measuring

7 tools and guidances should not be limited to a

8 selection of a tool, and that is clear if you look

9 at the technology.

10 We want to encourage companies to move

11 away from current univariate prescriptive testing

12 to multivariate process focus measurements. We

13 want to identify the essential or critical factors

14 that should be considered.

15 PAT can apply to all six of the

16 manufacturing subprocesses, which is inbound

17 logistics, active ingredient manufacture, bulk

18 formulations, fill and finish packaging and

19 outbound logistics, and the only thing -- and I

20 agree with Judy -- the only thing we left out was

21 short-term and long-term stability studies, because

22 they happen in parallel to the in-process.

23 Getting back to my example of a continuous

24 manufacturing process, you would then take samples

25 off the line and put them into stability testing on


1 a regular basis throughout the process, so you

2 would always have samples of your well-controlled

3 process at different stages of stability to look

4 at, to make adjustments, and so on.

5 It really I think will be a powerful tool

6 for that end of it, too. The guidance should

7 recognize that new insights is the process which

8 does not affect the quality of the product for its

9 intended use, should not require mandated changes

10 in the process.

11 Because we are going to use these tools,

12 we will learn a lot about blend mixing and what

13 happens, we will learn a lot about the process

14 itself and the individual steps, and if we are

15 still making a product well within the ramification

16 or the quality rubric that we have now for the

17 product, that shouldn't make the companies have to

18 do master reworks.

19 Now, I think that companies, when they

20 learn more and more about their product, are going

21 to make changes because they will see the benefit

22 of it, but if they are getting into this in fear of

23 the agency coming in on them and making them do all

24 sorts of, quote, unquote, "unnecessary" changes, I

25 think that is going to be detrimental.


1 We allow for the replacement of current or

2 classical methods with PAT for routine testing

3 methodologies. The guidance should recognize PATs

4 will, in large measure, replace current validation

5 measurements, and the guidance has to define what

6 records have to be kept and for how long.

7 Because we have the ability to take

8 real-time measurements on a continuous basis, how

9 many hard drives do we want to fill up and keep,

10 and how much of that is worth keeping, and is it

11 worthwhile, then, to establish a snapshot

12 recordkeeping system for an in-process.

13 Those kinds of questions we didn't have

14 real answers for, but we know that it is possible

15 for large manufacturers, who are making 30 or 40

16 different products, to fill up computers with data

17 and no one look at it, and no one need it, but

18 because of the system we have now, requiring all

19 the data to be kept, and how do we involve the FDA

20 in the PAT development. I think that is another

21 important aspect.

22 Classically, regulated industries don't

23 like the regulator in there helping them improve

24 their process, and I think this is an opportunity

25 for it to happen in a productive way.


1 Let me just say that the group of people

2 that we worked with on the subcommittee -- and I am

3 sorry I don't have all the names -- but it was

4 really a wonderful working experience, and I really

5 enjoyed it a lot. We had some good thinking and

6 some input from industry and from academia, and

7 some good statistical look to see what is going on.

8 I think that a lot of the results of that

9 are in the minutes of those meetings and really

10 worthwhile.

11 DR. LAYLOFF: Thank you, Art.

12 I have asked Ajaz to comment on the

13 Chemometrics.

14 DR. HUSSAIN: I would sort of summarize

15 the Chemometric discussion, not go through the

16 presentation that you already have in your handout.

17 The Chemometric Working Group was chaired by Dr.

18 Mel Koch from the University of Washington at

19 Seattle, and it was a mix of several individuals

20 from different backgrounds.

21 The classical chemometrics, I think what

22 chemometrics is was the sort of first business

23 point that the group focused on, and generally, we

24 tend to think of chemometrics are statistical

25 principles applied to chemistry and tools, such as


1 partial least squares, principal component

2 analysis, and artificial neurometrics are generally

3 considered to be part of chemometrics, but we will

4 have to look at chemometrics very broadly.

5 I think the key point that was raised

6 before was moving towards a multivariate approach

7 for assessing quality and performance of products

8 and move away from the current univariate system.

9 That itself is part of the chemometric paradigm

10 that we will have to develop.

11 Another most important part of the

12 chemometrics would be the design of experiments,

13 statistical design of experiments and how we use

14 that information to optimize formulation, and so

15 forth.

16 The group actually stressed quite heavily

17 on the need for design of experiments. I had

18 raised some concerns at that meeting, and I will

19 bring those concerns back to you also. The aspect

20 I think which would be important in the

21 chemometrics would be how do we validate some of

22 the software and statistical tools that would have

23 to be used in using multivariate approaches.

24 One of the concerns that was raised by one

25 of the speakers that we had invited from


1 GlaxoSmithKline was under the current system of

2 computer validation, for example, the perception

3 out there is we can't validate well-established

4 tools, such as MATLAB or SAS. Those commercial

5 software packages have some validation issues, and

6 so forth.

7 I think that would be a challenge, and I

8 think we will have to address software validation,

9 as well as validation of the statistical models

10 themselves.

11 The discussion tended to be more on a

12 concern, I think concerns were raised with respect

13 to these, but I looked at that from a very

14 different perspective. The reason is I think we

15 make very critical decisions in Office of

16 Pharmaceutical Science based on modeling, PK/PD

17 modeling, and all, I think we have had tremendous

18 experience with pharmacometrics.

19 So my concerns with chemometrics were not

20 truly reflective of the group's, and the reason for

21 that was I was coming from the pharmacometrics

22 background, and I can see many different ways of

23 validating and being very pragmatic approaches to

24 validation.

25 The other suggestion that I had at that


1 meeting was to look at how the Center for Devices

2 approaches software validation, off-the-shelf

3 software validation, and you have very good

4 pragmatic approaches for validation of software.

5 So the concerns were with respect to validation,

6 but I feel we have potential solutions to address

7 that, and I will probably bring that up for

8 discussion at the next subcommittee meeting.

9 The issue of experimental design. The

10 reason experimental designs were brought up and

11 discussed at length were because you really would

12 need to understand the causal links between

13 formulation of process variables and it is best to

14 do this in the development area where you can

15 actually design an experiment and sort of identify

16 the critical process variables and formulation

17 variables, and then that becomes a basis for

18 identifying which of those should be controlled,

19 and so forth.

20 The concern I raised was I think at the

21 University of Maryland, our research adopted a lot

22 of those principles, and I think Gary is here.

23 My concern was a survey that Professor

24 Shangra [ph] had done in '93, and it was published

25 in Pharmaceutical Technology, and one of the


1 questions there was how prevalent is the use of

2 statistical designs in R&D work in pharmaceutical

3 industry, and the number was very disappointing.

4 Less than 5 percent companies use design of

5 experiments in their development work.

6 That was '93. Has the situation changed?

7 I don't think so, and, in fact, my concern is more

8 and more the pressure on R&D has increased to a

9 degree that development itself is rate-limiting now

10 and more and more, people are going towards drug

11 bought in a bottle, and not even formulating until

12 you have some Phase I/Phase II data.

13 So more and more, I think attention to

14 process and product development has shift towards

15 end of the clinical or towards the end of the

16 development studies or, in some cases,

17 post-approval.

18 So we will have to work through it, and if

19 statistical designs are essentially necessary, I

20 think we have to look at it from a different

21 paradigm, so that we can provide a means of

22 understanding your processes better and

23 understanding how to model some of the systems, but

24 when will that occur is going to be a great

25 challenge.


1 My sense was, in the sense, the

2 Chemometrics Group was heavily dominated by people

3 who were thinking mainly in terms of the absolute

4 need for multivariate analysis, and one thing which

5 we did not pay much attention to in the

6 Chemometrics Group, but was done in Art's group,

7 and so forth, was a process signature, because you

8 could actually not have to do a lot of chemometrics

9 if we find a way to address process signatures and

10 control those process signatures, so that would the

11 first step that we could take before further

12 understanding could be developed.

13 A point that was made by Joe Famulare in

14 the sort of closing remarks, which was in the

15 Validation Group, is I think we would need to have

16 very flexible approaches for validation, not only

17 of the processes and analytical methodology, but

18 also of the computer software and models. So I

19 think we will have to think very differently and be

20 very pragmatic about how we bring this to bear

21 without adding undue burden.

22 So I think that is sort of a nutshell of

23 what the discussions were, and you have the

24 specific points in your handouts.

25 DR. LAYLOFF: Thank you, Ajaz.


1 I wanted to mention a few comment on the

2 validation issues, which is one of my favorite

3 topics.

4 [Slide.]

5 If you talk about validation, like to

6 think of buying a horse. When you go out to buy a

7 horse, you define what functions you want the horse

8 to perform. You know also that visual acuity is an

9 issue, but you frequently don't measure the acuity

10 of eyesight of the horse. You know that kidney

11 clearance is important, but we frequently don't

12 challenge the kidney and look at clearance. We

13 look at the overall performance of the creature to

14 see if it performs to meet our intent needs.

15 Signatures are going to bring a new

16 paradigm because people are comfortable looking at

17 the API, because they look at it at the beginning

18 and they look at it at the end, and they want to

19 keep looking at it, but that paradigm is not useful

20 for multivariate processes especially these

21 fingerprint technologies.

22 Validation is going to be a very key

23 feature because our concepts of validation have

24 been built around the API and those separation

25 technologies. Moving to signatures is going to be


1 disquieting, but to get people that buy horses to

2 come in and look at those computer systems because

3 when you start talking about 11 megalines of

4 coding, there is no way that you can plow through

5 that except as an animal.

6 Now, the assessment technologies, near

7 infrared, laser-induced fluorescence and things

8 like that are going to bring new problems for us in

9 validation because we have traditionally looked at

10 single variable processes.

11 Moving to multivariate detection is going

12 to be a big problem, and validation is going to be

13 a keystone in this whole business, and data stream

14 that Art mentioned, the data that you generate to

15 reach the endpoint is voluminous, huge volumes of

16 it, and unless you focus just on the endpoint,

17 validate the endpoint and then just look at the

18 endpoints, it is going to bury everybody.

19 I would say, all in all, it was a really

20 very exciting and interesting experience. I really

21 enjoyed it.

22 [Slide.]

23 The working groups addressed guidance

24 document issues and helped build consensus on

25 objectives.


1 [Slide.]

2 It is a world which is opened up to us by

3 the computers. It is useful to note that I think

4 cGMPs, as we have them now, can't accommodate the

5 changes. There is much to be done. The keystone

6 to the whole thing will be education and training

7 especially in the FDA. The barriers are going to

8 be conservatism in industry, as we saw with the

9 BCS, conservatism in FDA, which is always a hidden

10 stone, but we have a great beginning to move

11 forward from.

12 [Slide.]

13 The FDA initiative to come up with a

14 guidance document will help diffuse industry

15 conservatism, give a focus to the FDA reviewers and

16 investigators, which will help move the whole

17 thing. A well-trained cadre of FDA people to

18 address the new issues of validation and

19 performance-based process quality systems will be

20 critical.

21 [Slide.]

22 I like to think of it like making a good

23 cup of coffee. You buy beans, which are raw, and

24 you roast them, and you roast them to a certain

25 pyrolysis temperature, the color, you roast them to


1 light, dark. We can measure that by measuring

2 temperatures. We can also look at the color of the

3 beans and use that to monitor a process.

4 Grinding beans, if you have a little

5 grinder at home, you find out you can tell it

6 acoustically. You can listen to how the blade is

7 striking the particles and tell how fine the grind

8 is. So you can envision putting a microphone on

9 there and setting a specification that you are

10 going to grind it to a certainly sound level, at a

11 certain frequency, as an endpoint for the grinding.

12 Of course, in preparing the drug extract

13 that we like to drink, the temperature, time,

14 extract volume, air exposure, and stability are all

15 issues because if you make a pot of coffee and let

16 it sit there for two days, stability is a big

17 issue.

18 So I think that we have in our world these

19 technologies already. If you grind coffee, you

20 could hear it. We roast beans to color. You buy

21 dark roast, light roast, and those can all be done

22 visually or they can be done by machines, it can be

23 done in batch or on line.

24 So we do interact with these things, but

25 you talk about now how do you validate a microphone


1 to show that it is meeting its performance

2 specification on a daily basis.

3 It is going to be very interesting.

4 Now, I will turn it over to FDA.

5 Next Steps

6 Ajaz Hussain, Ph.D.

7 DR. HUSSAIN: What I would like to do is

8 give sort of a progress report and sort of next

9 steps. Following that, I would like to have the

10 committee discussion and recommendation on two

11 major issues.

12 One is we are planning the next PAT

13 Subcommittee meeting on the 12th and 13th of June.

14 I would like your suggestions and help in framing

15 the agenda for that. In fact, we are working after

16 this meeting to put a packet together that needs to

17 go out by the end of this week, so the timing of

18 the discussion is very appropriate, so I seek your

19 input on what should the second meeting focus on.

20 I think after my presentation, you have a

21 better sense of the direction on that, so that

22 would be one point.

23 The second aspect would be discussion on

24 an outline for the guidance, the draft general

25 guidance that you are preparing, and your


1 recommendations on the next steps, are we on the

2 right track, what should we do more, and so forth.

3 In the discussion that will follow, I am

4 hoping to get input on those aspects from you.

5 [Slide.]

6 Let me give sort of a progress report and

7 next steps.

8 To summarize the motivation from an FDA

9 perspective why did we start this, and what is the

10 sense of urgency, let me just explain that.

11 When we started looking at the

12 manufacturing processes and the issues related to

13 manufacturing processes that we are facing today,

14 we felt there was a significant potential and also

15 a need for improving the efficiencies of

16 pharmaceutical manufacturing and associated

17 regulatory processes.

18 When I say "efficiency," I am talking

19 about FDA efficiency, as well as industry

20 efficiencies, not just industry.

21 We felt technological opportunities were

22 available to realize this potential, and PAT, or

23 process analytical technologies, are just one

24 example of the opportunity that we have.

25 We heard at our FDA Science Board,


1 industry is reluctant to take advantage of such

2 opportunities due to regulatory uncertainties or

3 risk of uncertainty when you come to FDA, and has

4 preferred to adopt a "Don't Use" or a "Don't Tell"

5 approach.

6 Under the "Don't Use" approach, they

7 actually have corporate policies not to do this, so

8 I know of several companies where there is a

9 corporate policy not to include PATs in the U.S.

10 The same companies have done it outside.

11 The "Don't Tell" approach essentially is

12 you do this, but not register it, and you would

13 actually use that data to better understand your

14 processes and control your processes, but for the

15 FDA, you will provide the routine testing that you

16 do.

17 Again, both situations are undesirable

18 from a public health perspective, as well as from

19 an industry economic perspective. I think if

20 regulatory uncertainty is the cause of the low tech

21 aspect of pharmaceutical manufacturing, and I use

22 the phrase very carefully, and the phrase was used

23 by Dr. Ray Scherer, the senior VP for

24 GlaxoSmithKline, at our previous Science Board

25 meeting.


1 Our industry is very high tech in the R&D,

2 but from one perspective, it is low tech in

3 manufacturing, and that perspective is from the

4 perspective of process understanding, and so forth.

5 I think we have much more than equipment, and so

6 forth, so I am not talking about low tech in terms

7 of the equipment, and so forth, but the thought

8 process, the understanding, and so forth.

9 Again, those are very difficult concepts

10 for FDA to deal with, and I think we are dealing

11 with those in a win-win situation, a win-win

12 approach, so that we all benefit, and it is not

13 criticizing one part or the other part. I think we

14 will have to look at this as a mirror for all of

15 us, not just industry.

16 [Slide.]

17 Why PAT? We think PAT provides an

18 opportunity to move forward and to move from the

19 current testing to document quality paradigm to a

20 continuous quality assurance paradigm that can

21 improve our ability to ensure quality was built-in

22 or was by design. We think this is the true spirit

23 of cGMP.

24 The "c" in cGMP, current Good

25 Manufacturing Practices, I think Dr. Woodcock has


1 said several times I think we cannot use

2 enforcement to help in a way of bringing innovation

3 to manufacturing, it has to be a different

4 approach.

5 So the true spirit of cGMP is what we want

6 to capture in this PAT initiative, and I think we

7 can do that.

8 Why PAT? I think At/On/In-line

9 measurements of performance attributes are

10 feasible. When we measure or test in-process

11 material right now, you will hear about plan

12 uniformity. We test for drug substance, as Tom

13 pointed out. It may or may not always give you a

14 performance measure, but I think in a collective

15 multivariate way, we need to look at in-process

16 information that you can start predicting

17 performance.

18 We have real-time or rapid feedback

19 controls, which is not the case now. We currently

20 test and if the test results are acceptable, we

21 proceed to the next step or we throw away that

22 material and start again.

23 Real-time feedback controls are not truly

24 ingrained into pharmaceutical manufacturing.

25 Real-time feedback control moves us to a prevention


1 mindset. I think that is an important aspect.

2 We will get greater insight and

3 understanding of process. When I talk about

4 win-win, I think improved process understanding is

5 I think a key to be one part of the win-win,

6 because that is how you would improve quality and

7 improve efficiency, and decrease regulatory

8 concern.

9 Why would PAT help in terms of improving

10 of process understanding? Current methods, as Tom

11 correctly pointed out, we focus on impurities, we

12 focus on chemistry, wet chemistry. Functionality

13 of excipients, the physical attributes are not well

14 understood, are not well managed, and not well

15 controlled, because we don't truly have the tools

16 that have focus in this area, so PATs allows you to

17 bring physics and chemistry together to address all

18 those issues.

19 Potential for significant reduction in

20 production and development cycle time. You have

21 seen Professor Rogers or Dr. Rogers' presentation

22 to you at least once where he showed you the cycle

23 times of current manufacturing of simple tablets,

24 and the numbers were quite disturbing.

25 Reduce regulatory concern and potential


1 for remote inspection strategies. I say that with

2 a -- I will just skip that.

3 [Slide.]

4 Goals and objectives of the PAT

5 initiative. We are using PAT as a model

6 technological opportunity, to develop a regulatory

7 framework to facilitate introduction of new

8 manufacturing technologies that enhance process

9 efficiencies and understanding. I underscore

10 "understanding," because that is how public win-win

11 comes from both industry and us.

12 To do this, we need to identify and

13 eliminate perceived or real regulatory hurdles.

14 The more I listen, there are more real hurdles in

15 this issue. We have to develop a dynamic,

16 team-based, scientific approach for regulatory

17 assessment, review and inspection of these new

18 technologies.

19 I think, as I mentioned to you yesterday,

20 the Manufacturing Subcommittee is trying to bring a

21 mechanism for providing technical dispute

22 resolution for GMP manufacturing issues. We don't

23 have any mechanism. PAT will help us move in that

24 direction very quickly.

25 Clearly, we have to go for international


1 harmonization.

2 [Slide.]

3 The strategy that we adopted was very

4 simple. It has to be a win-win approach, and this

5 was my first presentation to you on the 19th of

6 July, if you remember that presentation. I said

7 this has to be a win-win, otherwise, this will not

8 work.

9 We started with input from you, Advisory

10 Committee for Pharmaceutical Science, but didn't

11 stop there. We went to the FDA Science Board, and

12 the reason for that was this is not just a CDER

13 issue or an OPS issue. It is an issue that

14 addresses all manufacturing and associated

15 organizations within FDA - Office of Regulatory

16 Affairs, Office of Compliance, Office of

17 Pharmaceutical Science, so you have to have the

18 highest endorsement for this project.

19 For that reason, as well as the potential

20 paradigm shift that occurs if this project is

21 successful, because this changes the entire

22 manufacturing system potentially.

23 So that was the strategy, to seek approval

24 from the highest levels at FDA. We set up internal

25 collaborations between CDER and Office of


1 Regulatory Affairs, and that collaboration has now

2 been effective for some time, and it is in the form

3 of a PAT Steering Committee. I will show you the

4 membership of that in a minute.

5 We needed external collaboration. We

6 don't have the technical knowhow to do this alone,

7 and we need to work together, and we use

8 subcommittee model to do this under your direction,

9 the PAT Subcommittee which Tom chairs. This was

10 the first report to you.

11 Clearly, PQRI is another mechanism and

12 soon we would like to sort of develop a program

13 under PQRI for research and technical issues that

14 need to be addressed here, and also build other

15 relationships, Dr. Gordon at the Pharmaceutical

16 Engineering Program at Michigan. There are many

17 other opportunities also available to do this.

18 We started with two parallel tracks. One

19 track is a general guidance on PAT. This guidance

20 will only focus on general principles and the

21 regulatory process. It will not focus on any given

22 technology, because I don't think we want to

23 identify a preferred technology, and so forth,

24 because one technology may not work for everything.

25 So at this time, we only want to issue a


1 general guidance which will delineate the

2 regulatory process and remove the regulatory

3 uncertainties. One important aspect of that

4 uncertainty is a Safe Harbor concept that we

5 discussed at the last FDA Science Board meeting,

6 and I will come back to that in a minute.

7 The Safe Harbor concept simply means that

8 as Art was mentioning, you may find something which

9 would indicate that the process may not truly be

10 under control, it may be a completely validated

11 process right now, and its intended use, so we

12 don't want to sort of penalize and then create a

13 situation where doing the right thing can get a

14 company into trouble, so you want to create a safe

15 harbor and move forward on this.

16 We are encouraging submissions now and for

17 that we are planning a team approach for review and

18 inspection during development and implementation of

19 PAT. What that will do is essentially as a company

20 decides to implement PATs now, they would have a

21 reviewer and an inspector working with them, so

22 that concerns, regulatory concerns are identified

23 and addressed, and not wait until a submission, so

24 that the investment a company will do and have to

25 wait for an answer which might not be acceptable,


1 so we want to sort of partner and help companies do

2 this now.

3 [Slide.

4 So in terms of a progress report, I used a

5 timeline of our meetings. The first meeting was on

6 the 19th of July, and then we had two Science Board

7 meetings, the 16th of November and 9th of April, so

8 that is essentially what the progress there is, and

9 the next step follows this meeting and the next

10 subcommittee meeting on June 12th and 13th.

11 [Slide.

12 So the progress so far has been we have a

13 wonderful collaboration between CDER and Office of

14 Regulatory Affairs right now, and this is in the

15 form of PAT Steering Committee. The members are:

16 Doug Ellsworth, he is the District Director for New

17 Jersey District; Mike Olson and Diane O'Brien from

18 Division of Field Science. They represent the

19 field labs. Then, you have Joe Famulare from

20 Office of Compliance; Moheb Nasr from Office of

21 Testing and Research; Yuan-yuan Chiu from Office of

22 New Drug Chemistry; Frank Holcomb from Office of

23 Generic Drugs, and myself.

24 So that essentially is the Steering

25 Committee. The difference here is we don't have a


1 working group. The guidance is being developed

2 through the PAT Subcommittee and the Steering

3 Committee without a working group, so that is a

4 deviation from our norm, because I don't think we

5 want to put a working group of internal groups to

6 write this guidance.

7 We have had several consensus building and

8 awareness activities within the Center. We had a

9 Center for Drug Evaluation research scientific

10 rounds where we discussed, debated a lot of these

11 issues. We had several seminars, and we just

12 completed a Visiting Professor Lecture Series. We

13 had several invited guests who came and talked to

14 us about the PATS including folks from industry.

15 [Slide.

16 With respect to external collaboration,

17 the PAT Subcommittee, you already know about that,

18 and the PQRI is something that we will pursue the

19 next few months, but I want to focus on an academic

20 collaboration that we have put together that is

21 developing a curriculum right now.

22 We have selected three National Science

23 Foundation process centers. These are the major

24 centers on Process Analytical Technologies. One is

25 a Pharmacy School, University of Purdue. One is a


1 Chemical Engineering School, University of

2 Tennessee, Knoxville, and the other one is the

3 University of Washington, Seattle, is a Center for

4 Process Analytical Chemistry.

5 What we are doing with these three groups

6 is to develop a training and a certification

7 program. I think the certification program is

8 something that have not made the final decision on,

9 and there are several reasons for that. Also, we

10 would need a continuing education program for all

11 the reviewers and the inspectors. These training

12 programs are for internal reviewers and inspectors.

13 Just one more point. At the next

14 subcommittee meeting in June, we will propose and

15 discuss this curriculum. We won't finalize this.

16 We will have this discussion at the next

17 subcommittee meeting.

18 [Slide.

19 The general guidance that we are working

20 on, the goals are as follows. General principles

21 and terminology essentially bringing the community

22 on the same page. I think we start with the

23 definition of what is on-line, in-line, and going

24 to chemometrics and all other definitions and

25 terminology.


1 Address issues related to regulatory

2 uncertainties. That will include a Safe Harbor

3 concept whereby, for example, now if a company

4 wants to apply PATs on-line to an existing

5 manufacturing line, they fear that numbers or data

6 coming out of that could be misused by us, FDA.

7 You might see its strength, and so forth.

8 So you want to think about a Safe Harbor.

9 So a Safe Harbor concept, I think would be one of

10 the agenda items for a discussion at the next

11 subcommittee meeting, what is a Safe Harbor that

12 would allow a company to investigate, and not fear

13 a negative regulatory in back of that, so how

14 should we define the Safe Harbor.

15 In addition, other issues with regulatory

16 uncertainties would be validation, I think computer

17 validation, batch, recordkeeping, and so forth, are

18 all regulatory uncertainties, so we will have to

19 deal with those in a general sense. I am not sure

20 we will be able to provide detailed resolution of

21 all of those issues in the first guidance, but I

22 would like to hear from you how should we address

23 that.

24 We have to clarify regulatory process. My

25 way of thinking with PAT is you have the current


1 system, which is adequate for intended use. With

2 PAT, we potentially are creating a totally new

3 regulatory system for manufacturing CMC. So that

4 is the level of potential difference we are seeing.

5 So we may have a completely new review

6 inspection system for PAT, which is distinct from

7 the current system. That is one possibility. I

8 think we have to look at that. The other

9 possibility is I think technical dispute resolution

10 is how we would do that, and maybe create a

11 technical resolution team which would address

12 disputes between reviewers, inspectors, and

13 industry that may come about in this process.

14 So those are the type of thoughts that we

15 are thinking about, and we will bring some of these

16 to the second meeting of the subcommittee.

17 We also hope we have other tangible

18 benefits of this general guidance. We think, we

19 hope it will serve as a tool for building

20 within-company consensus. I do want to sort of

21 emphasize I did not appreciate the lack of

22 consensus within companies until I started visiting

23 many companies.

24 Manufacturing, R&D, regulatory affairs are

25 not on the same page with respect to PAT. I think


1 you will see champions of PAT in the manufacturing

2 area, and the R&D area do not want to touch us,

3 they don't want to be bothered with it. So it is a

4 huge challenge.

5 I think you also have to think about this

6 in a cultural setting. This, I hope will not lead

7 to a disciplinary fight. What I mean by that is a

8 traditional pharmacy for development was chemical

9 engineering. I think you are looking at different

10 cultural aspects because traditional pharmacy

11 schools, industry pharmacy programs, they will have

12 to learn how to use the near IR signatures and

13 other things, and use that to optimize their

14 formulations. They don't have the knowhow, and so

15 a multidisciplinary team concept comes about this,

16 so that is a major challenge.

17 We don't have pharmacy schools, in fact,

18 most pharmacy schools have cut down on their

19 industry pharmacy program. Where will these people

20 come from, who would do this? One interesting

21 aspect, as I have been talking to companies, many

22 companies would prefer to do this manufacturing

23 outside the U.S. This is one of the reasons, the

24 talented pool of qualified people, where will they

25 come from.


1 So other tangible benefits of building

2 within-company consensus, hopefully, when you

3 reduce regulatory uncertainty, that will help that

4 process.

5 We also need to promote research and

6 development activities in this area. I think this

7 is just starting from scratch. We plan to work

8 with NISD, National Science Foundation to make a

9 case for public funding in this area.

10 [Slide.

11 Options for introducing PAT. I think this

12 is an important aspect. I would prefer to see PATs

13 being developed during the R&D process, but that is

14 a dream I think. In the current situation and

15 timeline pressures, I think we have to look at all

16 different options of when a company can bring PATs

17 to apply.

18 There are several options that we plan to

19 discuss and introduce in the draft guidance.

20 Option 1. A company might decide to use a

21 currently marketed, quote, unquote, "robust"

22 product that helps the company to focus on just PAT

23 issues, and not process-related issues, and then

24 apply PATs for improving efficiency and for

25 probably learning at the same time.


1 Several companies have done this. One I

2 have visited in Plankstadt, Germany, AstraZeneca.

3 They have done it for the reason of very important

4 compounds, and I will just mention that as I wrap

5 up my presentations.

6 Here, you would essentially bring on line,

7 and at some point, then, routine end product

8 testing may not be necessary if you bring

9 everything on line. The broken green line

10 essentially indicates that you will do routine

11 testing for shelf life, as we discussed.

12 A company, I think this will be a

13 challenge, but a company experiencing significant

14 manufacturing problems now where the product is

15 highly variable, and that is my depiction of the

16 variable product, a company might choose to get a

17 handle on that product, a step-by-step fashion,

18 where they will focus on each unit operation and

19 eventually have that process under control and move

20 towards on-line analysis of that.

21 Ideally, this should occur in a new

22 product development situation, but being pragmatic,

23 I think this will take time. Delay in drug

24 approval, the fear is so great that I don't think

25 companies would be ready to do that. I wish they


1 would, but let's see how.

2 At the same time, I think Dr. Woodcock,

3 when we were preparing for the Science Board

4 presentation, felt that this was not included in my

5 presentation to the subcommittee, is this is going

6 to take time. I think emotional high and emotional

7 excitement is building with PAT, but that has to be

8 followed with intellectual high at the same time,

9 and if you don't keep the two together, you have a

10 potential problem.

11 So what can we do to keep the momentum

12 going? Her suggestion was to think about unit

13 operation by unit operation, and to bring PATs and

14 provide guidances where you can just do one unit

15 operation at a time. You won't go for the entire,

16 but one at a time. So that is a concept that I

17 will just mention in a brief.

18 [Slide.

19 Let me just tell you about Track 2, which

20 is not part of the PAT Subcommittee, but Track 2 is

21 to encourage companies to provide submissions now.

22 For that, they have to contact the Office of

23 Pharmaceutical Science, and we will work with the

24 companies to set up the meetings necessary with the

25 compliance office of the field, and so forth, to


1 create a team effort to address issues and concerns

2 for individual applications.

3 So you are looking at a concurrent

4 development review-inspection strategy that we are

5 trying to develop right now. To date, I am happy

6 to say that we have received two formal requests.

7 The first company submission meeting is next week,

8 so I am happy to say we are moving on this already.

9 So two major U.S. companies have sent us letters to

10 request the meeting, so we are moving on the Track

11 2 also.

12 [Slide.

13 Track 2a is Dr. Woodcock's suggestion to

14 encourage established PAT technologies now. To do

15 this, we will encourage application of selected

16 on/in/at line measurement tools for unit operations

17 and/or as alternate tests.

18 For example, unit operations, such as

19 blending, drying, I think a lot of literature,

20 publications are already existing, data exist, so

21 we can use that and move with this right away.

22 The technologies could include near IR,

23 Raman, chemical imaging, on-line HPLCs, and so

24 forth, so well-established technologies for each

25 unit operation, we will try to encourage that now.


1 But to do that, we have one option, I

2 think, or several options. One of the options is

3 to start including PATs in the existing guidances

4 that we are working on.

5 For example, the draft Blend Uniformity

6 Guidance document that we will talk about, we will

7 focus on revising the guidance to bring in the

8 stratified sampling, the PQRI proposal, but at the

9 same time, it could include a section which could

10 say on-line has this and this benefit, and how you

11 might do that. So that is something I would like

12 your thoughts on also.

13 [Slide.

14 So just to give you a sense of timeline,

15 what activities are going on and what is happening,

16 I don't have a laser pointer here with me, but the

17 Track 1, the two boxes you see are the ACPS-PAT

18 Subcommittee. We already had one meeting in

19 February.

20 I think we got valuable information

21 especially from the Benefits Group that laid out

22 the outline for that guidance. I already have a

23 draft available for internal use here. We are

24 actually meeting with the PAT Steering Committee

25 tomorrow, after this meeting, to go with the draft


1 and use that internal draft to put in an agenda for

2 the next subcommittee meeting.

3 We will not share that with the PAT

4 Subcommittee. I think that was reported in one of

5 the magazines that we will present that draft to

6 the PAT Subcommittee. We cannot do that. So the

7 draft is only for internal FDA use.

8 We will use that to set up our agenda, and

9 so forth. I had only planned for two PAT

10 Subcommittee meetings. We may need a third one, so

11 that third meeting is not shown on this, and I

12 think the members of the PAT Subcommittee, many of

13 you are here, and you may suggest that I am right,

14 that two may not be enough, we may need a third

15 meeting to draft all the issues here.

16 So using the PAT Subcommittee to get the

17 information on issues with regulatory

18 uncertainties, when the draft comes out is not on

19 this chart because it is a very difficult

20 predictor.

21 We are planning a training program for PAT

22 reviewers and inspectors this summer. In May, we

23 are starting with the Track 2 with the first

24 company meeting.

25 The other information there, we have made


1 several presentations to companies and other

2 institutions, and I am happy to say that we have

3 visited International Federation of Process

4 Analytical Chemistry we visited Aventis,

5 Bristol-Myers, a PDA in Basel, and then when we

6 went to Basel, we took advantage of that and

7 visited Pfizer in Friberg, Germany, and then

8 visited AstraZeneca plant in Plankstadt.

9 So what we are talking about, PAT on-line,

10 it is not a theory, it exists at AstraZeneca, so it

11 was quite impressive to walk through that

12 manufacturing facility. Joe Famulare from Office

13 of Compliance, I, and Helen walked through that

14 plant, and I think it was quite interesting and

15 good to see that this is not a theoretical thing

16 that we are talking about.

17 We had several such meetings, I will not

18 describe all those meetings, but the consensus is

19 building, and I think the highest level of support

20 from Dr. Woodcock, the FDA Science Board, and so

21 forth, I think has helped tremendously with this

22 effort.

23 [Slide.

24 So the next steps are, for internal, we

25 are establishing a CDER-ORA PAT team for joint


1 review/inspection. We have just started the

2 process of selecting our reviewers, and we will

3 hopefully find the inspectors to go along with

4 that.

5 The plan is to have four reviewers and

6 four inspectors to be part of the first team, so it

7 is not training the entire group, it just focusing

8 on four reviewers and four inspectors.

9 We plan to recruit expert consultants. We

10 already have started the interview process for a

11 process or chemical engineer, process analytical

12 chemist, a chemometrician, and we already have an

13 industrial pharmacist who is working this. That is

14 Roger Poole. I don't see him in the audience here.

15 So we are going to fill those positions,

16 and these would be part of the technical consulting

17 staff, as well as people who would help develop

18 technical guidances and may be part of the

19 technical dispute resolution team.

20 [Slide.

21 So developing a training (and

22 certification) program for PAT review and

23 inspection team. The reason "certification" is in

24 parentheses is there are a lot of drawbacks for

25 certification.


1 One drawback is those folks would be hired

2 by industry the first day we certify them. So that

3 is a challenge since we will lose them very

4 quickly. That is not such a bad thing, I think we

5 can work with that.

6 The proposed curriculum will be discussed

7 at the June meeting of the PAT Subcommittee, and

8 then we will put the training program together.

9 I just last week looked at several

10 abstracts that we submitted. We have a tremendous

11 number of abstracts on at least at-line use of near

12 infrared to predict dissolution and predict other

13 attributes, so we have had a quite a good success

14 in using near infrared, chemical imaging for a

15 number of applications including prediction of

16 product performance, that is, dissolution, so we

17 will present those papers at AAPS meeting.

18 We will expand that research. Moheb is

19 here, and we are trying to do some work on direct

20 work and get this program expanded.

21 We will publish the draft guidance as soon

22 as possible after that.

23 [Slide.

24 Wrapping up next steps, a public workshop.

25 We already have the program developed for the Arden


1 House Conference for this year or for the year 03.

2 The U.S. Arden House will be in January and will

3 focus mainly on technical details and technical

4 aspects of PAT.

5 The program for UK will be a more big

6 picture, economic benefits type, and that is a

7 collaboration between AAPS and Royal Pharmaceutical

8 Society, FDA and MCA. We have been working very

9 closely with MCA. We have had some contacts with

10 the German authorities, so informally, we have

11 started talking about issues with harmonization,

12 but I will come to that in a minute.

13 A FDA/AAPS PAT workshop is being planned.

14 We had one meeting, but have sort of held this back

15 because there are so many issues that are unclear,

16 so we are hoping to understand the issues better

17 before we finally put this program together. We

18 will aim for April 2003 for this. Hopefully, the

19 guidance might be available at that time.

20 We need to formalize efforts toward

21 international harmonization, as I said, currently,

22 informal communications with a few European

23 regulators.

24 For example, when we visited the Pfizer

25 plant in Friberg, we had invited the German


1 authorities to visit with us, so we had a lot of

2 discussion, and so forth, but it has been informal.

3 There is no mechanism right now under ICH to

4 discuss this, so we will have to start working on a

5 formal mechanism for harmonization.

6 One of the sense of urgencies we had was

7 when I presented this to you in July of last year,

8 we felt Europe was ahead of us in many ways. In

9 some ways they are, but our thought process, I

10 believe, has matured to such and such a degree

11 right now, I think we probably will regain the

12 leadership in this again.

13 I had better stop and look for your

14 discussion.

15 DR. LEE: Thank you very much, Ajaz.

16 How do you want to take the questions?

17 First of all, I think I would like to alert

18 everybody that quite a few committee members have

19 been schedule to leave before 5 o'clock, so I am

20 going to tighten the discussion, and I am planning

21 on a 4:00 p.m. adjournment at the latest.

22 I hope that some of the other presenters

23 later on in the program are here or they would be

24 contacted, so that there will be presentations.

25 Certainly, this is a very exciting


1 project. I think what I would like to see is to

2 keep the momentum going. I have identified a very

3 small subcommittee within this committee to help

4 digest and lead the discussion.

5 I have Art Kibbe, Judy Boehlert, and

6 Efraim Shek to ask the questions in case there are

7 no other questions from the groups.

8 Efraim, would you like to take the floor?

9 DR. SHEK: Maybe I will start and taking

10 the risk, maybe something melodramatic, but I

11 really believe personally that PAT will bring a

12 revolution, and I believe a revolution to the way

13 we are manufacturing and the way we are treating

14 it. If you look really at that aspect and how it

15 is going to influence us, in industry, in R&D, as

16 well as the manufacturing.

17 I think we have to look at it from this

18 point of view. It is a major revolution. Saying

19 that, okay, and talking about Tom, I like

20 analogies, too, but we are not buying horses, and

21 we are not grinding coffee, we are not making

22 cookies.

23 [Laughter.]

24 What we really develop and manufacture is

25 pharmaceutical products, which I believe, even


1 without the PAT, that, in general, they are in high

2 quality.

3 Also, I think I can say in public, because

4 PhRMA is basically the organization that basically

5 supports this approach. Saying that, I believe the

6 devil will be in the details, as you indicated, you

7 know, a few aspects.

8 Number one, which I think we have to look

9 at very carefully, is the training, both training

10 of FDA, as well as training in the industry. That

11 will be extremely important.

12 How we do the training, who does the

13 training, today, most of the experience is in the

14 industry, those people who have really used it, and

15 I think we have to learn from those companies who

16 already started it.

17 When we look at the curriculum, we have to

18 be very, very careful there. We can go on and

19 list, you know, of details, but if you have to look

20 at just one example, if there is, like any other

21 computer system, a sensor will malfunction, what

22 happened to the batch that you are manufacturing,

23 and those things I believe will happen.

24 The concept of validation, I think will

25 have to be reviewed, but I am encouraged. I


1 personally think, coming from an IND, I think it is

2 stuff that we should get involved with, because in

3 this case, I am using an analogy, the horse is out

4 of the barn, and we should look at it very, very

5 carefully. The opportunity is so great.

6 One that may be fruit for some thoughts is

7 how do you encourage, okay, the industry, their

8 conservatism both the FDA and the industry, and I

9 think we have to think about some ways that we can

10 encourage companies, both there is an economic

11 aspect there, to take this risk, okay, or to

12 invest, and if we can come up with some thoughts,

13 how can we encourage like any other unit forces to

14 make those things happen.

15 I believe that most of the facilities, the

16 issue will be, you know, are there any technical

17 issues, okay, implementing PAT into existing unit

18 facilities

19 DR. LEE: Let me maybe focus discussion a

20 little bit. First of all, I would like to see

21 whether or not these committees like what was

22 presented and are we on the right track, and also I

23 think that we need to, it seems to me from a

24 committee member, I would like to some kind of, not

25 open-ended process. Are we at that stage where we


1 can begin to estimate how long with that take.

2 Is it a process that we would like to see

3 uniformity, I mean can we tolerate two systems, at

4 what point we would like to see one system, and

5 more importantly, how can we put ownership in these

6 stakeholders, what hurdles. Is it realistic that

7 within 10 years, everyone will be doing this

8 on-line monitoring?

9 So those are the issues that I would be

10 interested to hear from the committee, and see, are

11 we on the right track. Are the horses guided?

12 Yes.

13 DR. KIBBE: Let me try to respond to some

14 of those questions, and I want to throw a couple of

15 things into the pot. One of the problems with this

16 is the name of it says process analytical

17 technology, and "analytical" generally conjures up

18 in most of our minds a specific kind of activity.

19 We toyed with, in our subcommittee or our

20 small working group, changing it to "assessment,"

21 because we are really assessing a process using

22 whatever technology is available and whatever tool

23 is available, so that we don't have to do extensive

24 end-stage testing, that we know the process did the

25 job, and if the process does the job, the result


1 must have done the job. I mean the result must be

2 there.

3 One thing that I am concerned about,

4 whenever we talk about harmonization, I think in

5 terms of making everything down to the least common

6 denominator. In this situation, what we want is to

7 bring everybody with us up to a different level of

8 expertise.

9 I think one of the problems we see when we

10 start depending on in-process is the reliability of

11 the process monitors. I liken our situation to the

12 early days of NASA and the way that they made sure

13 that the process monitors worked is redundancy, and

14 I think we are going to see a good company will do

15 redundancy.

16 Instead of one microphone monitoring, they

17 will have two or three, and then they will say

18 these two said it was good, this one was off, we

19 are going to clear the process, keep it rolling,

20 and we will check out the monitor that was out of

21 sync. I think we can look at that.

22 Now, every process that we put in place,

23 every technology will have a different need for

24 redundancy and a different need for validation of

25 the different robustness, and the more we get used


1 to that particular monitor or method of analysis,

2 the better we like it.

3 I don't know how many people check their

4 balance, you know, between each weighing with a

5 standard weight to make sure it hasn't changed. We

6 all know it balances pretty well, and we will do

7 the same thing here.

8 One of the things that is driving this

9 whole process is the rate of expansion of

10 computational informational technology. We could

11 never have even envisioned doing this until a

12 computer can digest the types of information flow

13 that we are talking about, and it's humongous, but

14 at the rate the computational power goes up, which

15 is doubling every year, the chances of being able

16 to digest these things, if you remember -- maybe I

17 am the oldest one to remember this -- but we used

18 to test for tablet hardness by picking a tablet out

19 of the batch and snapping it and listening for the

20 sound.

21 If it had the right sound, it was probably

22 the right hardness. Well, when Tom said grinding

23 coffee, we tuned engines to sound, but now we have

24 gotten to the point where we have technology that

25 does it.


1 Well, that is the same thing we are

2 dealing with here. We are going to look for that

3 fingerprint, that general look of the product. I

4 know that makes people uncomfortable, but we are

5 going to get there where we can say that looks like

6 a finished product and the reason we are

7 comfortable with that is we have done this 20

8 times, and every time it looks like that, it is a

9 good product, and we will be able to move forward.

10 Incentivizing the industry. The industry

11 is incentivized by one thing, and that's money, and

12 I will predict for the industry, and they can check

13 it with their stockholders, but 10 years from now,

14 the pharmaceutical manufacturers who aren't using

15 this are going to be out of business, because the

16 potential for improvement and economic savings and

17 quality of the product is so great that over a

18 10-year development with the improvement in

19 computational technology going along with our

20 improvement in sensoring devices and our ability to

21 accept that we don't have to do end-stage testing

22 if we did everything right, you don't have to test

23 if you reach New York City, if you followed the

24 directions.

25 You don't make another quality control to


1 make sure you have made it, because you have driven

2 there. If we can accept that change in the way we

3 think about how we manufacture, so what I am really

4 getting down to is the technology is going to race

5 ahead of our human ability to be comfortable with

6 the technology, and the companies and the people

7 who run it, who get comfortable with it first, are

8 going to have an edge on everybody else.

9 DR. LEE: Well, Art, you touched upon a

10 very important issue. I think now manufacturing is

11 done in ways that are most economical. Obviously,

12 this process might be cost-saving, cost-effective,

13 and would there be forces opposing that for being

14 implemented?

15 DR. HUSSAIN: The key aspect is the first

16 thing that we established with the Science Board

17 was this is totally voluntary, nobody has to do

18 this as a requirement, and the second thing we

19 established with the Science Board was the Safe

20 Harbor concept, the ability to use a risk-based

21 approach to address problems, some that we may find

22 when they apply more close scrutiny of the process.

23 So the concerns, the two major concerns

24 that industry has, I think has been addressed with

25 the Science Board already, and I think how, the


1 details, I think Efraim mentioned the details. I

2 am hoping the PAT Subcommittee would help us

3 articulate some of those details that will be

4 necessary for the guidance.

5 DR. LEE: Tom.

6 DR. LAYLOFF: A couple of comments. I was

7 going to say in support of Art that it would be

8 redundancies and also orthogonal measurements, so

9 that you are actually looking at several

10 parameters, signature parameters at the same time

11 onstream.

12 I think the efficiency will come a lot

13 from a compression where you actually are using the

14 equipment closer to 100 percent of the time, so

15 that the cost, that will be a big driver for it in

16 addition to the reduced analytical load, because

17 these on-stream devices are going to be very

18 inexpensive.

19 So I think it is going to be a tremendous

20 advantage, and it is going to be a revolution

21 driven by the computational abilities and the

22 ability to handle redundancy in orthogonal system

23 signature systems.

24 DR. HUSSAIN: One aspect which I would

25 like to share with you, a question was raised by a


1 FDA Science Board member, what is the incentive

2 especially today when discovery is not

3 rate-limiting anymore, development is

4 rate-limiting, and that is the reason people are

5 shying away from doing extensive development

6 because of the high failure rates of compounds in

7 early clinical trials, it is what incentive and why

8 would a company invest this upfront in the R&D.

9 The answer I think, which I was impressed

10 by the answer of Ray Scherer from GlaxoSmithKline

11 said is most of these unit operations are not

12 specifically new drugs, so a lot of the information

13 would already be existing, and essentially, when

14 you have a new molecular entity, you can actually

15 compute and actually predict what the conditions

16 would be.

17 So, essentially, once you have understood

18 all your unit operations, you will be actually

19 doing very few experiments, but predicting and

20 confirming the experiments, so mathematical

21 modeling understanding brings a level of

22 understanding that will help actually development

23 itself.

24 So today, development is rate-limiting.

25 DR. LEE: Other opinions?


1 DR. LAYLOFF: One other comment. I hit a

2 paradigm wall when I was working on this. I kept

3 thinking of a single process stream in which you

4 had devices to process it through, and I was

5 visiting with one of the industrial presenters, and

6 he told me they were getting ready to move PAT into

7 a manufacturing site where they did 200 different

8 products.

9 So they were looking at basically the PAT

10 in the specific technology areas, which they then

11 would merge into the product lines. I thought it

12 was going to be more hardened, you know, like a

13 single plant, but they are actually moving to do

14 200 -- a plant where they manufacture 200 different

15 items, which I was really dumbfounded with.

16 DR. LEE: Yes, Pat DeLuca.

17 DR. DeLUCA: I guess that Art mentioned

18 about the title PAT, and certainly the success of

19 this is going to depend on -- this is sensor

20 technology which is developing right now -- and I

21 wondering why somehow maybe that couldn't be worked

22 into kind of the identity of this, that it is a

23 sensor technology.

24 I go back to the mid-eighties when we used

25 NIR in the early stages for looking at moisture in


1 an intact vial, freeze-dried product. It was an

2 end product. But one of the things we learned,

3 too, was presentation of the product to the sensor

4 is very important, and that I think is something, a

5 technological problem that needs to be overcome,

6 and I think engineering that can be voluntary.

7 I guess with regard to immediate release

8 products, I see the application here, you know, in

9 the manufacturing area, and I agree with Art, I

10 think 10 years from now, that this is something

11 that behooves all manufacturing companies to adopt

12 this for survival, but I think the development, it

13 seems to me to be done in the manufacturing area

14 for the immediate release products, where one can

15 build up a history and experience and know just

16 what it is that you need, what parameters you need

17 to be monitoring there and what is the robustness

18 of this and how does that play a role in that.

19 I guess, to answer Ajaz's statement with

20 regards to development, I can see this in the

21 development area for extended release products

22 where you are now dealing with maybe products that

23 are six-month or a year type of products, and you

24 really can't wait around six months or a year to

25 release the product. This would be very beneficial


1 to be able to have this type of technology that

2 would allow one to release a product, you know,

3 without having to go through some release test or

4 some test that is extended over real time in that

5 manner.

6 DR. LEE: Thank you, Pat.

7 Yes, John.

8 DR. DOULL: One of the problems we are

9 having in bringing our students up to speed in

10 genomics and proteomics, and so on, is we find it

11 isn't really enough to teach them molecular

12 biology. You also really have to teach them

13 informatics.

14 We can't teach that. We have to bring

15 people in to teach that because that's a

16 specialized field. When we talked about this the

17 last time, it seems to me we asked that question

18 about informatics and whether you have sufficient

19 expertise, Ajaz, on your groups to bring that

20 discipline clearly in. You know, rather than

21 bringing it from in-house, you may have to go

22 outside and get special kinds of techniques and

23 what have you to really use -- Art mentioned they

24 use the computers, and that is a special area, and

25 I am not sure that we have that in-house


1 information. You may need to go outside and get

2 more of it.

3 DR. HUSSAIN: I think informatics, I sort

4 of put that under chemometrics. I think

5 chemometrics is broadly defined as chemical

6 informatics, and so forth. Surprisingly, FDA

7 probably has a lot of expertise, probably cutting

8 edge expertise with pharmacometrics, I think the

9 tools like Jurgen and others would use in PK/PD

10 modeling, and so forth, are essentially similar

11 tools here.

12 Our toxicologists have done extensive work

13 on informatics. So bits and pockets of information

14 is there, but it has not been applied to chemical

15 problems, and that is the reason we felt that we

16 will hire a chemometrician, a chemical statistician

17 to handle this.

18 You are right, I think we need to hire,

19 and I think the four individuals that we are trying

20 to hire, one of them would be chemometrics and

21 informatics.

22 DR. LEE: What about Judy, do you have

23 something to add? As a member of the working

24 group, do you feel that we are moving in the right

25 direction?


1 DR. BOEHLERT: There is one other point

2 that Ajaz addressed in his presentation, and that

3 is the lack of consensus within companies among

4 different groups.

5 Clearly, right now it is seen as a

6 manufacturing initiative, and it should be a

7 multidisciplinary initiative, and I don't know to

8 what extent, when you are planning workshops, and

9 things of that sort in the future, you can bring

10 these different groups together, because clearly,

11 the roles and responsibilities of some of these

12 groups are going to change.

13 For example, the Quality Group, they are

14 going to evolve to a situation where you don't have

15 off-line testing. The functions that they perform

16 now are going to go more into the audit kind of

17 mode, and I think we need to begin to educate these

18 folks, not that they are going to lose their jobs,

19 but their jobs are going to change, and start

20 bringing them on board, and if we don't bring

21 everybody on board, regulatory affairs,

22 engineering, product/process development, and

23 quality, then, it is not going to have a high

24 chance of success within companies.

25 That is part of the selling job right now.


1 I would agree that in 10 years, the big companies

2 will be there. I think it is going to be more of a

3 challenge for the smaller companies, and it is a

4 resource issue.

5 You know, they don't have anybody on staff

6 that knows much about informatics very often. They

7 don't have the technical expertise.

8 DR. LEE: Maybe by that time industry will

9 be out-sourcing everything.

10 DR. BOEHLERT: Right, and it is quite

11 possible a whole new industry is going to develop

12 to support this kind of initiative.

13 DR. LEE: So many things are happening,

14 and I would just like to ask the committee, you

15 know, what is the advice to the subcommittee on

16 PAT, do we put a point to prioritize, do we put

17 some effort into certain areas more so than others?

18 DR. HUSSAIN: I think in terms of the

19 focus for the next meeting, we will focus more on

20 the regulatory uncertainty, defining the Safe

21 Harbor, defining all those processes, and sort of

22 wrap the second meeting with that.

23 We had planned for a two-day meeting, but

24 there are many issues with respect to computer

25 validation, with respect to validation itself.


1 From that perspective, I had come to you before, I

2 said we will probably use two meetings.

3 My thinking is we may need one additional

4 meeting about this, even for the general guidance,

5 and then sunset that subcommittee after the third

6 meeting, because their job to provide information

7 for the general guidance will be over.

8 We will have need for more technical

9 guidances, and so forth. What I was hoping is we

10 will sunset the subcommittee and bring on the

11 Manufacturing Committee as a committee under the

12 ACPS, and then address some of the technical

13 science issues either on the PQRI or other

14 mechanisms. That is the thought process for the

15 next few steps.

16 DR. LEE: Gloria, you are the

17 representative of the Consumers. Any comments

18 about this development?

19 DR. ANDERSON: No, I don't really have

20 anything to add other than to say that I think it's

21 a big step forward, and I am particularly pleased

22 with it. I would like to ask if you could take a

23 minute and tell us a little bit about what you

24 observed when you made your visit, what you

25 observed in terms of the technology that was being


1 used in-line, on-line.

2 DR. HUSSAIN: I think, broadly speaking,

3 and I don't want to focus on one company --

4 DR. LEE: She is validating that you were

5 actually there.

6 DR. HUSSAIN: Yes. What I was surprised

7 was, as I said, I was at Aventis, Bristol-Myers,

8 AstraZeneca, Pfizer, and last week Merck, and I was

9 amazed in terms of how much work is already

10 ongoing.

11 In general terms, use of Raman emerging as

12 inspector methods for controlling particle size and

13 polymorphism at the crystallization stage of the

14 drug substance. I think many companies are working

15 on that in the very mature area in terms of

16 controlling the polymorph that you produce, and so

17 forth.

18 In terms of blending, obviously,

19 laser-induced fluorescence, man companies have been

20 working with that, and near infrared is very, very

21 common. When I sort of talk, and so forth, I bring

22 my bias, the solid dosage form, but the thing I

23 have to guard against myself is there are many

24 technologies.

25 When we visited another company last


1 summer, with on-line HPLCs, on-line GECs,

2 everything has been used for several years, but

3 again have not had the regulatory applications. I

4 mean they are doing in addition to what they do for

5 regulatory purposes.

6 So my sense, and what I am very hopeful

7 is, there is a lot of activity that is already

8 ongoing within many companies. The concern is

9 sometimes when we went to these companies, the R&D

10 and the Regulatory Affairs folks for the first knew

11 what was happening, so that is the reason for, as I

12 said, is a disconnect. So my presence or our visit

13 helped them talk to each other, so I am hoping we

14 can do that more.

15 DR. ANDERSON: I would just like to say

16 that I think that anything that improves

17 efficiency, and hopefully cost effectiveness,

18 certainly should be good for the consumer.

19 DR. LAYLOFF: One other remark. I think

20 that the pharmaceutical industry, the traditional

21 drug industry is very conservative, it is very

22 staid, but if you look at the more dynamic

23 industries in the biotechnology areas where there

24 is a rapid flux, you find that there is a lot more

25 assimilation of trying to be more efficient, keep


1 everything moving, because they haven't really

2 stabilized.

3 I guess they will grow old and

4 conservative also eventually.

5 DR. LEE: Other comments?

6 DR. HUSSAIN: Final talk for the record, I

7 think there is an omission that Tom did and I did

8 in my previous presentation, and so forth, is for

9 some reason, Efraim and Leon, we missed their names

10 on the list, so we just want to acknowledge they

11 participated in part of that PAT Subcommittee.

12 DR. LEE: It seems to me that the

13 committee is gaining some momentum, and I think

14 it's on the right track, and we are going to hear

15 from you again at the next meeting.

16 Thank you very much.

17 Let's take a 10-minute break.

18 [Break.]

19 DR. LEE: We are going to start out with

20 two individuals from the agency, Dr. David Hussong

21 and Bryan Riley.

22 Rapid Microbial Testing - Update

23 Introduction and Overview

24 David Hussong, Ph.D.

25 DR. HUSSONG: I simply wanted to introduce


1 today's speakers. We are kind of lucky because of

2 the backgrounds of these people.

3 Dr. Bryan Riley is a review microbiologist

4 at CDER. He has as spectacular background in

5 clinical microbiology, and, of course, clinical

6 microbiology is where a lot of the microbiology

7 rapid methods came along.

8 We also have Dr. Mike Korczynski, who has

9 a very good background from Abbott Laboratories

10 where he headed up a major program, and he has

11 since gone on to be an independent consultant.

12 Dr Korczynski also had great input in a

13 technical document produced by the Parenteral Drug

14 Association on the introduction of new methods in

15 microbiology.

16 We are also blessed that later today, we

17 will be hearing from Jeanne Moldenhauer and Scott

18 Sutton, who were also on that committee.

19 I would like to introduce Bryan Riley.

20 Bryan Riley, Ph.D.

21 DR. RILEY: Thank you, David, and good

22 morning.

23 As an introduction to this session, what I

24 would like to do is give you a quick look at some

25 of the methods that are used for microbial limit


1 testing.

2 [Slide.

3 I would like to start out with the

4 compendial methods, which around here means USP,

5 Chapter 61, Microbial Limit Tests, and there are

6 essentially two methods that are used. They both

7 rely on the growth of the organisms.

8 The first one are called plate counts,

9 which give us colony-forming units. In that

10 method, you apply the sample either onto or into

11 the solid medium. You incubate the medium, allow

12 the colonies to grow, and then you count visible

13 colonies.

14 The second method, which is less accurate

15 than the plate count method, is called the MPN or

16 most probable number method. In this, a series of

17 multiple dilutions are made in a broth culture, a

18 liquid medium. These serial dilutions are

19 incubated. At the end of the incubation period,

20 you look at the different tubes in each dilution

21 that show growth.

22 You then refer to an MPN table, which will

23 tell you what the most probable number of organisms

24 was in the original sample.

25 [Slide.


1 The advantages of the compendial methods

2 are they are very simple, easy to do, are tried and

3 true methods. They can be done by any

4 microbiologist in any microbiology lab. They only

5 count the viable or living organisms because we are

6 not really interested in the dead organisms at this

7 point. We just want to know what's alive, what to

8 grow in the product cause product problems for you,

9 the product quality, or the patient who takes the

10 product.

11 The disadvantages are the incubation time,

12 which can be up to seven days for a yeast or mold

13 culture, as well as not all organisms will grow on

14 a single medium, and so you are really sort of

15 getting a subset of whatever organisms that might

16 be present in that sample. So that is a little bit

17 of a drawback.

18 [Slide.

19 I will say a little bit about microbial

20 viability again. We talked about the compendial

21 methods only will detect living organisms that can

22 grow, and that is all we are interested in.

23 Therefore, any new or rapid method will have to

24 have some way of differentiating between the live

25 and the dead organisms.


1 To do that, you need some sort of marker

2 for viability. As an example of this, for the

3 rapid test, I would like to give examples of a

4 couple of markers that are used for these tests.

5 [Slide.

6 The first one is esterase detection. The

7 esterases are enzymes that are ubiquitous in

8 microorganisms, and it works by the reagent that

9 you apply to the sample being cleaved by the

10 esterases and releasing a fluorescent compound

11 which can be detected in the sample

12 [Slide.

13 The method works as follows. The sample

14 is filtered, the filter membrane is exposed to the

15 reagent, and after an incubation period or short

16 incubation period, the membrane is analyzed by

17 laser scanning, and you get a count of the

18 organisms that are present on the filter.

19 [Slide.

20 The next method is the ATP

21 bioluminescence. ATP is a primary energy source

22 for all organisms, so it is going to be present in

23 any living organism that you have in there. The

24 reagent, which is a combination of luciferin, which

25 is a substrate and luciferase, which is an enzyme,


1 along with ATP will react to produce visible light,

2 which can be measured.

3 [Slide.

4 The procedure works as follows. The

5 sample again is filtered, much like the esterase

6 test. The membrane is placed on a solid medium for

7 a brief incubation period to amplify both the

8 numbers of organisms, as well as the ATP content,

9 to allow it to be detected.

10 The cells are disrupted to release the

11 ATP. The bioluminescent reagent is added to the

12 membrane, and then a coupled device is used to

13 detect the light, and the results are analyzed by

14 computer to give you again the number of organisms

15 in the sample.

16 [Slide.

17 The advantages of the rapid methods are

18 they are rapid, they are fast, less than 24 hours

19 in some cases, much less than 24 hours, 2 or 3

20 hours in some cases. They are very sensitive. As

21 I said, they don't necessarily rely on growth, and

22 so they can detect any organism that is present in

23 the sample, whether it can grow on a medium or not.

24 The disadvantages would be increased

25 complexity for these methods. They are much more


1 complex than just plating out a sample on solid

2 medium. They are expensive, both originally the

3 setup costs of the equipment, validation, et

4 cetera, so they are not cheap.

5 In some cases they can be too sensitive,

6 getting back to the increased sensitivity of the

7 test, in some cases you may have an

8 out-of-specification result, because you are

9 detecting more organisms than you would with the

10 compendial method, and that could be a problem.

11 My final slides are going to be a couple

12 of questions for the committee to ponder.

13 [Slide.

14 First of all, should or could the agency

15 do anything to encourage industry to use these new,

16 rapid micromethods?

17 [Slide.

18 Finally, getting to the sensitivity issue,

19 since to address the sensitivity issue or the

20 increased sensitivity issue, you may need to change

21 the specification for some of these drugs, and what

22 could be considered as a loosening of the

23 acceptance criteria by making the numbers higher,

24 how can we address this both from a scientific

25 standpoint, as well as a regulatory standpoint,


1 what can we do to make this work smoothly and still

2 provide adequate microbial quality for the

3 pharmaceutical products?

4 With that I would be happy to take any

5 questions that the committee might have.

6 DR. HUSSAIN: Just an additional

7 perspective on this, we had a discussion on rapid

8 micro, the same meeting we first had the PAT

9 discussion, and at that point, I think the

10 recommendation was to either form two separate

11 committees, one for rapid micro, and for PAT.

12 We didn't make progress on rapid micro, so

13 I asked them to come back to this committee to sort

14 of examine these questions, but in addition, to see

15 -- a lot of the issues that we deal with here are

16 the same issues with PAT, so from a general

17 perspective, the general guidance that we have on

18 PAT could incorporate a lot of these issues by

19 themselves.

20 So, the general guidance on PAT probably

21 could cover the process, regulatory uncertainty,

22 and risk-based approach that we would need to

23 address the sensitivity of some of these methods,

24 and then follow that up with more technical

25 guidances as you do for the PAT.


1 So that is sort of the perspective as you

2 ponder and discuss this.

3 DR. LEE: Maybe we can hold off questions

4 until the end. Thank you.

5 Dr. Korczynski.

6 Michael S. Korczynski, Ph.D.

7 DR. KORCZYNSKI: I would like to say I am

8 really pleased to have been invited to speak.

9 During my industrial career and thereafter, I saw a

10 number of occasions to use rapid methods. I think

11 they have a real place in industry, and I think we

12 should really all be part of moving those

13 activities forward in the industry.

14 [Slide.

15 For all of the reasons that we already

16 heard this morning, efficiency, improvement of

17 efficiency, improvement of productivity, shortening

18 corrective action time, so you can respond more

19 immediately rather than waiting a week basically,

20 better utilization of your personnel, more

21 efficient use of your personnel, and finally, we

22 all hope some sort of cost reduction for the

23 industry because that is basically what they are

24 going to be looking for in the implementation of

25 these methods.


1 Now, I heard the word "conservatism" this

2 morning. I think that is a good word. I would like

3 to read some thoughts that I had written down

4 regarding the impetus that must be provided. My

5 analogy, we need to start rolling the stone down

6 the hill basically. Someone has to start to be a

7 champion of these methodologies.

8 While many companies appreciate the

9 potential of rapid methods in microbiology, and are

10 willing to institute these methods, it isn't going

11 to happen unless companies feel that the FDA

12 supports these methodologies.

13 Some companies with the resources and

14 technical expertise are bold enough to be risk

15 takers and have the resources to present the data.

16 You know, they will call a meeting with the FDA,

17 they will eventually do their R&D homework, and

18 maybe eventually that will turn into a supplement

19 and an NDA, but that is mainly some of the major

20 firms.

21 I think what we find in many cases, the

22 smaller companies and smaller companies without the

23 resources are holding back, you know, where is it

24 going, is the FDA presenting this information

25 publicly, does it appear that they are supporting


1 this technology, where does the USP stand, is there

2 an in-process revision for this methodology.

3 So that is the conservative attitude that

4 someone else has to get it moving before others

5 will follow. Therefore, I think that once FDA

6 public endorsement appears to occur, many companies

7 will start employing the specific methodology, and

8 the FDA indeed is a significant factor in

9 introducing rapid methods to the industry.

10 [Slide.

11 Now, I would like just for a historical

12 basis, the FDA has -- and I think some of us have

13 forgotten about this -- the FDA has played a role

14 in the introduction of some microbiological methods

15 in the industry.

16 Now, one method, some of you may not even

17 recall this, it's about 1974 to 1977, in that era,

18 data was taken from academia and eventually moved

19 over to industry, and that was in the Limulus

20 amebocyte lysate test, which is a test for

21 bacterial endotoxin with a lipopolysaccharide

22 associated with gram-negative cell walls.

23 That is a method that tests the cellular

24 component. Prior to that, the industry had to have

25 huge animal colonies, that you test for LPS's,


1 fibro/fibril producing, or pyrogen, and so in all

2 your lots, you had to do pyrogen testing, mandated

3 that you had a huge animal colony, that that animal

4 colony was inspected. It took time to run

5 basically.

6 Well, the FDA, I think this is really a

7 hallmark, it was a hallmark activity, and that the

8 FDA worked with USP, the industry, academia, and

9 developed a protocol of how to move that technology

10 forward and for a while there was some finished lot

11 testing, I believe by the FDA, but I think it was

12 model system in terms of introducing a rapid method

13 to industry. That is why I spent a little time on

14 that particular topic.

15 The other one was membrane filtration.

16 Many people forgot that it was Dr. Francis Bowman

17 back in the 1960s that was a proponent of membrane

18 filtration for sterility testing. Prior to that,

19 you did a dilution type of test, it would take you

20 14 days. The membrane filtration now allowed you

21 to filter product and test for 7 days basically.

22 It became accepted, it became part of the

23 USP. Now, however, the pendulum is swinging the

24 other way, just as an anecdotal comment, and even

25 though you filter for aseptic fill products, you


1 are incubating for 14 days. EP wants 14 days

2 incubation even though it is filtered, fraternally

3 sterilized product.

4 So here is an issue now, and that issue is

5 Dr. Riley presented some rapid methods, sterility

6 testing to obviate that 14-day test, is an ideal

7 opportunity for one of these viable methods that

8 was just presented, so we need to think about that.

9 The other item that I thought was helpful

10 to the industry was the FDA, I think it was Bureau

11 of Drugs back somewhere in the seventies,

12 introduced a protocol that allowed you to shorten

13 your incubation time for biological indicators that

14 are used to monitor sterilization processes,

15 providing statistically your data fell into the

16 mode presented in that protocol, and that was very

17 helpful because in many cases, that shortened

18 process time testing from, in some cases, 7 days to

19 5 or 4.

20 I have gone through the role of the FDA as

21 an advocate of some of the methods.

22 [Slide.

23 On thing I might mention in terms of

24 microbiology, rapid methods in microbiology, one of

25 the rapid methods that does exist is the


1 identification of microorganisms. You used to have

2 to go through laborious test tube reaction types of

3 tests. There is now identification equipment where

4 you can inoculate cards or wells, and actually have

5 a much faster readout.

6 So that has been fairly widely accepted in

7 industry, and that does exist as a rapid method.

8 [Slide.

9 Some of the important aspects of rapid

10 methods you have heard, you have heard a lot of

11 this, but it is real-time analysis, process

12 real-time analysis. We are getting close to it in

13 microbiology because sometimes you still need a

14 dwell time, you need some type of incubation time,

15 but you could minimize corrective action time,

16 which is very important.

17 You might be processing something and we

18 find the water supply is over action level, you

19 didn't wait days to find that out, perhaps you

20 could find that out that very day, stop the batch,

21 make the improvements.

22 Again, you are going to increase assay

23 sensitivity in many cases. That was already

24 described. Therein lies a problem, but I think, as

25 scientists, you have to deal with the data.


1 So if technology moves you forward and you

2 now have a more sensitive test, you have to figure

3 out basically how to do that. You know, maybe in

4 some cases you are going to have to modify your

5 alert and action levels, but you can't hide from

6 the facts, and the new procedures will be more

7 sensitive in many cases.

8 Of course, it is going to remove some of

9 the operator performance, and we hope to see more

10 improved reproducibility. I think there will be a

11 more efficient utilization of personnel, and that

12 is why I say efficiency and productivity, because

13 now you can take those people who were spending too

14 much time on some of the longer assays and have

15 them doing other things basically.

16 I think there is an opportunity for cost

17 reduction after the initial capital investment.

18 Many of these rapid methods have equipment that you

19 have to purchase basically and then once you

20 establish and have that equipment, you have to buy

21 the commodities to keep it going.

22 So it is that initial capital investment

23 that is going to cost, and I think to have an

24 appeal to the industry, you probably need a payback

25 of, you know, maybe in a five-year period.


1 What does all this mean? It means

2 potentially, reduction of product release time.

3 Now, maybe not all companies have this, but, you

4 know, there is an element that is moving towards,

5 and it is already there, just-in-time

6 manufacturing, reduce your inventories, manufacture

7 to the orders, so therefore, you are trying to

8 remove all the lag times and delays out of your

9 processing, and shorten that product release time,

10 and these micromethods have that potential.

11 [Slide.

12 Dr. Riley already went over several

13 methods, but just to categorize these for you,

14 there are about four basic methods, growth based,

15 and I might say for every one of these methods,

16 there are about three to four rapid methods that

17 can support those four entities.

18 So if you have a growth-based assay, you

19 could perhaps utilize ATP bioluminescence, maybe

20 reduction of CO 2,

maybe head space pressure in a

21 container, and, of course, under growth-based

22 technologies, one would find the various

23 biochemical ID assays that I talked about.

24 Then, I am going to show a slide on

25 viability-based studies, which I think have


1 probably a fairly wide application and utilize some

2 very sophisticated instrumentation that is

3 appearing out there, and also has some costs

4 associated with it.

5 The next technology would be cellular

6 component. This would trying to look for entities

7 associated with a cell, so you could have fatty

8 acid detection, you could have mass spectroscopy

9 that kind of gives a fingerprint. You talked about

10 signature. Well, it sort of gives a signature of

11 your different genus, maybe species, of

12 microorganism.

13 You could have enzyme-linked immunosorbent

14 assay that looks for antigens or antibodies

15 associated with a cell, and also you could have,

16 under cellular component, I just talked about LAL,

17 that is a rapid method in a sense because it takes

18 just a couple of hours and you can determine

19 whether you have an endotoxin of LPS concentration

20 present.

21 The next item would be the nucleic acid

22 technology, and there we are talking about DNA

23 probes, ribosomal tying, and PCR, polymerase chain

24 reaction. So you can see there are technologies

25 out there, there are methodologies out there that


1 could begin to fit these required technologies.

2 [Slide.

3 I am not going to talk about this to any

4 great length, but example of rapid viability

5 methods. It is using again perhaps you filter the

6 solution, you bring the cells in contact with a

7 dye. Internal esterase enzyme within the cell

8 cleaves that, it becomes fluorescent. You either

9 look at it via a microscope or it goes through a

10 photomultiplier tube, put that information in the

11 computer. Basically, you have counted cells.

12 These methods are turning out to be quite

13 sensitive. I believe the fluorescence flow

14 cytometry, we are measuring, looking at

15 microorganisms in flowing solution. I think that

16 requires somewhat of a higher count, but at least

17 on the filter scanning types, you can detect down

18 to 1 to 10 cells.

19 The literature is in your booklets, right?

20 So it is there. I think these perhaps have lots of

21 promise, could be quite sensitive.

22 [Slide.

23 I put together some thoughts on what might

24 be ideal attributes of a rapid viable microbial

25 quantitative method, and could process variable


1 sample sizes. You want it to detect more microbes

2 than general plate counting. You want to detect

3 low levels of viable cells, and there is an issue

4 going on, on unculturable cells. We never heard

5 this, you know, 10 years ago, that word wasn't

6 used.

7 We are finding in certain systems where

8 the microbes are stressed and just about able to

9 survive. They are there, and when you use

10 conventional culture media, you can't recover them,

11 because the media is too enriched for the

12 environment that they were used to.

13 This method should probably have that

14 capability, and it has to be able to differentiate

15 between artifact and actual cells. Sometimes you

16 may stain, in your staining procedures, some type

17 of debris, may stain and it is not a cell. You

18 have to be able to differentiate that in these

19 methods.

20 In industry, it is very important for us

21 to have portable systems, your environmental

22 monitoring, you are going from aseptic suite to

23 aseptic suite, you need something portable. So

24 this system should be portable.

25 Very important, you should be able to


1 corroborate the data by some other method, and that

2 is going to lead us into validation. Then, you

3 want to see a return on the capital equipment.

4 [Slide.

5 Now, pharmaceutical acceptance, you know,

6 where is the USP on these issues? Maybe my

7 colleague, Dr. Scott Sutton, may have something to

8 say about that, but I am mainly using an example a

9 little bit of sterility testing here in that there

10 is some polymerization trying to occur between USP

11 and EP.

12 Relative to sterility testing, it is still

13 the accepted, you know, it's a 14-day test even if

14 you are filtering the product. As I said, it may

15 get to get 14 days for products that are

16 aseptically filled, and yet there is no -- the USP

17 will talk about the possibility of available

18 methods, but there is no rapid method mentioned in

19 USP or EPA that basically addresses, say, sterility

20 testing.

21 So my opinion is -- just anecdotally, I do

22 sit on a USP committee -- but my interpretation is

23 it is rather slow or nonexistent, and what I mean

24 by that, is the encourage of new rapid microbial

25 methods, and kind of why.


1 I think the reason for it is there is

2 really no provision for the validation of these

3 methods, and there is an uncertainty of how do you

4 validate some of these new rapid methods and by

5 whom, you know, in order to get a reasonable cogent

6 interpretation of that data. You can't have the

7 supplier just presenting their validation data. It

8 has to be done by some type of outside peer group.

9 [Slide.

10 I might add, though, the USP, when I said

11 slow to move, does recognize, in some of the

12 chapters they recognize that alternatives can

13 exist, and such methods should be validated if they

14 are used, and, of course, you should have

15 equivalent reliability and when dispute arises, the

16 compendial method is conclusive. So while it is

17 recognized, there are no specific examples of

18 alternative rapid methods.

19 [Slide.

20 I might add that, if you are interested, a

21 very good resource document is the PDA Technical

22 Report 33 that addresses evaluation of validation

23 of rapid microbial methods. It basically lists

24 these items, it defines them, gives them

25 definitions, but it is a fairly demanding and


1 arduous task to validate a new microbial method, it

2 will be.

3 [Slide.

4 Now, if we look for guidance, you know,

5 how can we validate these methods. Some of you are

6 probably aware of more of these, but there are two

7 guidelines for validation of chemical methods, and

8 one is USP Chapter 1225, Validation of Compendial

9 Methods; ICH, Validation of Analytical Methods.

10 However, some of us just off-line had a

11 conversation. I am not so sure how often we can

12 take chemical validation procedures and apply them

13 to the microbial method scenario. I don't know how

14 much we can draw upon that.

15 In terms of the microbial methods

16 validation, I think it is fairly weak. We have an

17 ASM verification and validation of procedures in

18 the clinical micro lab. That is good, but in

19 industry, we are not in a clinical micro lab, not

20 in the manufacturing environment.

21 The other item is USP Chapter 1227,

22 Validation of Microbial Recovery for Pharmaceutical

23 Articles. Good, but it doesn't directly apply to

24 some of the rapid methods that we are considering.

25 [Slide.


1 Now, I have a proposal for validation and

2 review, because I think that is one of the keys in

3 the system of these rapid methods.

4 [Slide.

5 This is my own viewpoint, so no one in any

6 group or agency advocated this, but it just seems

7 to me if we can only go back and look at what it

8 took to implement the LAL method in industry, that

9 could serve as an ideal model, and I think we can

10 establish a protocol for evaluation of microbial

11 methods.

12 We should involve industry in generating

13 the data and then we can covalidate perhaps --

14 perhaps at an FDA laboratory and/or USP labs, and

15 as I said, use the LAL test method as acceptance,

16 and have a joint peer review by the FDA, USP, and

17 industry scientists.

18 Now, I thought about this. This sounds

19 nice, but there is going to be a challenge in there

20 in that every company that supplies a rapid method

21 or makes rapid method equipment are going to line

22 up and say, you know, are you putting my method in

23 queue for validation testing.

24 So you are going to have to deal with

25 those issues, and if you selected one viable


1 method, why didn't you select the viable method

2 manufactured by another company basically. So

3 those are the challenge would lie ahead.

4 [Slide.

5 So, in total, if we look at an initiative,

6 and I think this initiative is very compatible with

7 what I heard this morning relative to the chemical

8 side, and that is to facilitate the technical

9 transfer of valid rapid microbial methods to the

10 pharmaceutical industry, resulting in the use of

11 consistent and accurate assay methods that will

12 expedite corrective action, reduce manufacturing

13 time, increase productivity, and, of course, reduce

14 expenses, and hopefully, those will be passed along

15 to the consumer.

16 Thanks for your time.

17 DR. LEE: Thank you very much.

18 I will now open the floor for questions.


20 DR. HUSSONG: I think one omission on my

21 part, I forgot to introduce Peter Cooney. Peter

22 Cooney heads the Microbiology Group in the Office

23 of Pharmaceutical Science, and this group has just

24 moved to direct under the Office of Pharmaceutical

25 Science, as I mentioned yesterday. He will


1 participate in this discussion.

2 DR. LEE: Peter, would you have any

3 comments to make?

4 DR. COONEY: I just want to say that the

5 Microbiology Group in OPS is on board with Process

6 Analytical Technologies. We look forward to

7 cooperating with the industry, and we encourage you

8 to make submissions because we are not opposed to

9 approval of these methods.

10 DR. LEE: Great. I think that is one

11 question posed to the committee - is the PAT's

12 program sufficiently broad to address the general

13 issues related to the introduction of rapid

14 microbial testing. So that is the question. We do

15 have representation by Dr. Riley, and do you have

16 any specific questions for him or for Dr.

17 Korczynski. Yes.

18 DR. SHARGEL: I would like to bring up

19 another issue that occurred to me, and it probably

20 impacts on PAT, as well. This is a case where, in

21 many cases, there is an older product with low

22 sales volume, say, $10 to $20 million, that a large

23 company may feel for business purposes that we

24 drop, and this is happening right now.

25 A smaller company may feel it wants to


1 pick up the product, but accordingly, and let's say

2 it's an antibiotic product, that the NDA is pretty

3 old, it's safe, it's efficacious, it's a low cost

4 product, in order for a small company to pick it

5 up, they would have to do, say, a fair amount of

6 analytical testing and validation, whereas, the

7 original antibiotic NDA may not even have had a

8 chemical assay. It could have had an antibiotic

9 assay, which was total actives by some approach.

10 How will this impact, this new technology,

11 on old products and particularly the fact that

12 right now we are losing some very good products

13 just currently because of business decisions, and

14 smaller companies cannot sell these products

15 because of the economic costs of validation and

16 development are too high?

17 DR. HUSSAIN: I was hoping, in a sense,

18 that what we are talking about is not a

19 requirement, and so forth, so I think it gives a

20 company a choice to use whatever approach is

21 appropriate, so that is a reason I felt that the

22 issue was sort of addressing that.

23 DR. SHARGEL: I would just like to follow

24 up. Currently, using the state of art now, there is

25 a requirement to have stability indicating assays,


1 chemical methods, and all of that, so we are losing

2 products right now currently as the regulations

3 occur. So we are adding now more methodology.

4 DR. HUSSAIN: Leon, I think -- I haven't

5 given thought to exactly the issue you just raised

6 -- but I didn't see that as a sort of PAT issue per

7 se, but let me go back and think about that. I am

8 not sure I have the answer for you today.

9 DR. LEE: Tom.

10 DR. LAYLOFF: I think the one he just

11 mentioned is the one where the technology changes

12 and the bar raised, so when the chemical assays

13 came in, they were more sensitive to some of the

14 product quality dimensions, and the bar actually

15 raised on the products.

16 Going back to one of Mike's comments, I

17 don't think there is going to be a change in the

18 base legal status, you know, legal definition of

19 what is a sterile product. These will probably be

20 validated replacements which are open in USP, that

21 you can cross-validate that there will still be a

22 legal bar, and that is probably not going to change

23 unless there is something really extreme happening,

24 but I think setting up validation criteria to bring

25 in new test methods is a very reasonable thing to


1 do.

2 DR. KORCZYNSKI: Regardless of whether you

3 use the rapid method or conventional method, the

4 conventional method for a small company may be less

5 expensive at the beginning basically.

6 The final result is the final product.

7 The final product is sterile whether they use the

8 conventional method or someone used a rapid method,

9 but the person that used the rapid method enjoyed

10 some efficiencies and productivity perhaps that the

11 smaller company didn't use. I think that's the way

12 maybe the outset, the way it is going to be.

13 In other words, you could take an analogy

14 to that. Aseptically filled product is sterile,

15 and some companies will fill in an isolator, some

16 companies will fill in a sterile suite, and you

17 still have some hand-filling going on, but that end

18 product is still sterile.

19 So I think, at least maybe in this next

20 decade, there is some larger core for variability

21 in how we do these tests. Does that make sense?

22 DR. COONEY: Let me just make a statement

23 about sensitivity levels, because that is what

24 everybody is concerned about. In other words, you

25 know, in chemistry, sometimes when the ASA


1 sensitivity got better and better, the limits got

2 tighter and tighter.

3 I personally do not intend to do that in

4 microbiology, and the reason for that is we will

5 only do it if there is an associated increased

6 risk, and it has to be risk based, so that if you

7 have one of these methods that have been talked

8 about, and let's say you detect in terms of

9 microbial limits, not sterile products, but

10 microbial limits, let's say you detect 10 times the

11 number of organisms in the sample that you would

12 have detected using compendial methods, using

13 trypticase-soy agar.

14 Well, that doesn't mean the product is 10

15 times worse than it ever used to be. An example of

16 that is, for example, if you use trypticase-soy

17 agar and do bioburden assessments or assay to

18 water, you know, people accept a limit of 10

19 colony-forming units per 100 ml, and I think that

20 is even mentioned in USP, if I am not mistaken, but

21 if you use a different agar and different kind of

22 culture conditions, you can get 10 to 100 times

23 more colony-forming units in the same sample. That

24 is in the literature, and it is seen a lot of

25 times.


1 We expect that to happen with these new

2 methods, as well, and this is what Bryan and Mike

3 were alluding to, what that means is you have a

4 different measure. You have X number of ATPase

5 units, or you have X number of esterase units, and

6 how does that correlate to what you see in old

7 methods.

8 It doesn't mean that now the limit of

9 acceptability will be 10 or 100 times less just

10 because you can detect 10 or 100 times more,

11 because like I said, it's a risk-based assessment.

12 DR. LEE: Peter, let me ask you a

13 question. It might be a silly question. Would you

14 envision that we would expect to identify the

15 bacterium? I think soon the genome would be

16 sequenced, but that's overkill.

17 DR. COONEY: Well, in the manufacturing

18 environment, when they establish microbiological

19 control of a manufacturing process for a certain

20 product, identification of the microorganisms is

21 part and parcel of that. Some people do more, some

22 people do less, but it is important to know what

23 kinds of organisms are associated with the

24 environment, with the product, so that if there is

25 a change, you will know that something is


1 different.

2 There are a lot of rapid microbial ID

3 methods, as well, that Mike spoke to. One thing to

4 think about for the Advisory Committee and the

5 subcommittees, or any future subcommittees, is

6 there is a difference between microbial limits and

7 sensitivity in microbial limits where something is

8 supposed to be there, and sterility testing where

9 something is not supposed to be there.

10 The question one might ask from a

11 technical standpoint is what happens if 50 percent

12 of samples tested using the technologies turn out

13 to have something in them that you never saw before

14 using USP sterility tests, what do we do then? So

15 that would be an interesting question for the

16 future.

17 DR. LEE: Yes, Tom.

18 DR. LAYLOFF: That is what I was

19 interested in because USP, in the General Notice,

20 it says that you can do all type methods, but there

21 is one definitive legal standard. Now, if you

22 stick with that, then, you don't raise the bar.

23 You can change the technology, but the

24 legal standard platform is still there, and I

25 thought that is what I heard you say you wanted to


1 do with microbiological analyses, and if you went

2 with modern technology, you were going to keep the

3 legal platform where it is, which is an alternate

4 method approach.

5 DR. COONEY: Well, it is an interesting

6 statement, and maybe we could talk later, but the

7 USP sterility test doesn't say that a product is

8 sterile. All it says is it meets the requirements

9 of the test.

10 DR. LAYLOFF: That's correct, but in the

11 General Notice section, there is a statement on

12 alternate methods, that you can use alternate

13 methods for sensitivity, speed, whatever, but that

14 the legal platform is that one that is defined.

15 DR. LEE: Let's hear from the other

16 members of the committee. Any comments, Judy?

17 Efraim?

18 DR. BOEHLERT: I definitely think this is

19 a topic that falls under the umbrella of PAT,

20 however, there are specific differences, and those

21 have been mentioned by Tom and Mike and Peter.

22 In this case, we are talking about a new

23 rapid technique where a referee method already

24 exists, and it is not brand-new. That referee

25 method has legal consequences. You are talking


1 about a number of different techniques which would

2 need validation perhaps to show that they are

3 equivalent or better than that referee technique.

4 You might want to give consideration to

5 working with USP to see whether that is the

6 appropriate referee technique going forward because

7 if indeed a manufacturer uses one of these rapid

8 techniques, their product gets out in the field and

9 it is challenged for whatever reason, and they

10 haven't used the referee method, you know, then,

11 they are in trouble, and there is some risk

12 involved there.

13 So that area needs to be addressed

14 upfront.

15 DR. COONEY: You know, I mean these are

16 really interesting issues and we have thought about

17 them for a long time. One example of a process

18 analytical technology and microbiology, maybe it

19 was stated before by somebody else, maybe it should

20 be called process assessment technology, is

21 parametric release, that is releasing products

22 without a sterility test.

23 Now, we first approved that in 1985, and

24 since that time, I made the statement once before,

25 just a gross estimate is over 5 billion individual


1 product units have been released without a

2 sterility test with no problems in 17 years.

3 Yes, the USP sterility test is the referee

4 test, and I suppose if you took one of those

5 products and you tested it and it failed, you would

6 be in trouble, but as far as in terms of product

7 release, I mean there isn't any microbiologist in

8 the world who now believes that test is worth

9 anything.

10 So working with the USP might be a good

11 idea to change the referee test. In fact, the

12 question is, is there, in fact, an end product test

13 that can detect nonsterility and at what level.

14 In sterility, we know you insist the

15 probability that any one unit is nonsterile, is

16 about 1 in a million, and the sensitivity of a USP

17 sterility test is about 1 in 14, 1 in 10, so you

18 are talking about a 100,000-fold difference in

19 sensitivity.

20 There, you run into what Mike was

21 referring to, is how do you validate the methods,

22 and do you validate the new method against the old

23 method if the old method is 100,000-fold less

24 sensitive than the new method.

25 So all those are interesting things that


1 need to be discussed in the future.

2 DR. LEE: John, you are shaking your head.

3 DR. DOULL: Well, I am a little concerned.

4 The focus that you have given us this morning is

5 really on diagnosis, and you have showed that you

6 can improve diagnostic methods a lot and do it a

7 lot faster, and so on, but the real issue, and I am

8 glad you brought up risk, because that is the real

9 issue, what you want to do is predict risk, and we

10 are focusing on developing methods and we are not

11 asking at the same time the question we should be

12 asking, is how do we improve our ability to predict

13 risk from these microbiological exposures.

14 It seems to me that somehow we need to

15 figure out how we can pay a little more attention

16 to predicting risk than simply developing new

17 methodology. You talked about chemical comparisons

18 and the fact they don't really go, and I agree, but

19 a lot of microbiologists are looking at that issue

20 of how really do you predict risk from

21 microbiological exposures.

22 I think there are some developments, and

23 there are particular molecular biological

24 developments that would give you a way to begin to

25 get ahold of that issue of predicting risk. I


1 guess what I would say is hopefully, that would --

2 I don't know whether that fits into PAT or not, but

3 somehow it ought to at least fit into what we are

4 doing in microbiology.

5 DR. HUSSAIN: In terms of PAT, I think the

6 risk can be looked at from two different

7 perspectives. One is risk assessment or risk

8 management for existing products that have proven

9 and have a track record in terms of when you find

10 something new, which had already been there, I

11 mean, but since you are using new tools, how do you

12 manage. That's one perspective on risk.

13 But on the other hand, I think I would

14 like to look at risk management, the PAT is a tool

15 for minimizing risk, because essentially you are

16 controlling every process, every step much, much

17 more carefully, and because of the availability of

18 new technology, you are preventing the bad product

19 to be manufactured in the first place, so it is a

20 prevention mentality. So it is a risk minimization

21 approach that I think PAT brings.

22 DR. DOULL: Ajaz, that is only half of it.

23 The risk assessment is the other half of it, and

24 you could develop some new procedures in risk

25 assessment which would help you before you get to


1 the management stage.

2 DR. HUSSAIN: I totally agree with that.

3 I think the aspect which I think I sort of

4 underscored in one of my slides was understanding,

5 process understanding, and prevention and

6 identifying risk factors, I think if you want to

7 look at it from that point.

8 If we do a good job in identifying all the

9 critical variables that affect quality and

10 performance, then, that is one assessment of risk

11 factors and how we control that, so that will go in

12 that direction.

13 DR. LAYLOFF: I have one question for

14 Peter again. I think the surrogate testing for

15 sterility, the target is 1 in a million, and I was

16 wondering what failure rate might be required to

17 find it out in a population where it is being used.

18 In other words, is the epidemiology sufficient that

19 you could pick up 1 nonsterility in 100,000?

20 DR. COONEY: Not on a bet.

21 DR. LAYLOFF: You said not on a bet?

22 DR. COONEY: Not on a bet. The false

23 positive rate, even in a well-constructed sterility

24 testing facility, and Mike can chime in here,

25 ranges in the area of about 0.1 percent, 1 in


1 1,000, and you are looking for an event that is as

2 infrequent as 1 in a million, so it kind of doesn't

3 work.

4 We actually in the beginning, you know,

5 you talk about conservatism in the agency and in

6 the industry, in the beginning, when we first

7 approved parametric release for terminally

8 sterilized products, we asked the company to do

9 sterility testing along with parametric release for

10 two years, and they accumulated all this data, and

11 we didn't learn anything.

12 So that kind of a validation isn't really

13 right. We have to think about a better scientific

14 way to do that, and there will be cases. I am sure

15 there will even be cases in chemistry -- and Ajaz

16 can attest to that -- I mean you are talking about

17 content uniformity, and stuff, and you do 10

18 tablets versus checking every one, so it is kind of

19 like apples and oranges.

20 DR. LAYLOFF: But you do a surrogate model

21 to validate the system, the sterility process, but

22 there is no way to tell if it meets 1 in a million

23 or not, even in population use.

24 DR. COONEY: That is part of the

25 validation process.


1 DR. KORCZYNSKI: I just want to make a

2 comment in terms of the current compendial USP

3 sterility test, if I recall, my probability tables,

4 you could have 5 percent contamination, and you

5 would only recover that about 64 percent of the

6 time. There is room for improvement.

7 DR. LEE: I think we are getting into very

8 technical issues. Let's hear from Pat.

9 DR. DeLUCA: I just wanted to comment on

10 the sterility test. I know Peter implied it is not

11 an effective test. It is a very good test, it is

12 effective. The problem is the sampling plan and

13 the fact, I think that Mike just brought out,

14 actually, to be assured of 95 percent sterility,

15 you have to have about 15 percent contamination in

16 the batch.

17 I think you were saying 5 percent, 64

18 percent of the time. If you want to really be 95

19 percent sure, you have got to have 15 percent

20 contamination in order to get that.

21 I guess the on-line, I think it would be

22 good to have some rapid tests certainly to make

23 decisions with regards to release, I think that

24 would be very, very beneficial, and there certainly

25 are DNA probes that would allow rapid testing, as


1 Mike I think pointed out, within a 24-hour period.

2 I think the other area, too, is with

3 regards to assessing bioburdens. I think this is

4 very important in an operation, if these rapid

5 methods, and I don't know how the PAT could be

6 adapted to this, to be able to monitor bioburdens

7 in a manufacturing operation. I think this would

8 be very valuable.

9 DR. LEE: Is Pat speaking the sentiment of

10 the committee? Should we put this on the plate of

11 the PAT group, to identify what are some of the

12 hurdles, what is to be done?

13 Well, hearing no objections, Pat, you have

14 another assignment.

15 DR. LAYLOFF: Great.

16 DR. LEE: One more horse.

17 So that concludes the agenda item on rapid

18 microbial testing. Thank you very much for all the

19 presentations and discussion.

20 We are now into the open public hearing.

21 We have two volunteers. Their names were mentioned

22 already. That is Jeanne Moldenhauer from Vectech

23 Pharmaceutical Consultants, and Dr. Scott Sutton.

24 They each have been told that they have 10 minutes

25 each.


1 Open Public Hearing

2 MS. MOLDENHAUER: Currently, if you look

3 at the microbiology testing lab, this is the

4 typical testing performed for each product.

5 [Slide.

6 Depending on your product, some tests may

7 not apply but this is the typical kind of

8 microbiology testing. The majority of it is

9 retroactive testing where it is performed after the

10 fact, and so improving all these things to the PAT

11 is great and wonderful, but you won't be releasing

12 product any quicker unless we also apply rapid

13 methods to microbiology.

14 [Slide.

15 Within these methods, there is three basic

16 types of tests that are performed.

17 Presence/Absence tests, is there positive sterility

18 tests, is there not? Is there a specific organism

19 present, is there not?

20 Enumeration tests where you are actually

21 looking for counts of organisms, and

22 identification/characterization tests, what

23 organism is it that is present?

24 [Slide.

25 When we talk about rapid microbiology


1 technologies, Dr. Korczynski gave a general

2 overview of the technologies that are available,

3 but within the growth-based and viability-based

4 technologies, you would find methods to do -

5 presence/absence tests, enumeration tests.

6 If we go on the additional methods that

7 are available, there is methods to do presence of

8 organisms. There are additional items to identify

9 and characterize organisms. These methods, in

10 general, are superior to the methods that exist

11 today, and as a industry person, the difficulty

12 arises, then, the USP method allows me to provide

13 equivalence, but these methods are superior.

14 The technologies may not even be remotely

15 similar, and the ability to appropriately show

16 equivalence is extremely difficult. In addition,

17 the problem that one faces is training your local

18 investigators to understand that these new

19 technologies are different and how you are going to

20 address the fact that we are not raising the bar,

21 but this sterility test is positive, and this one

22 is negative, and the risk to the product, and

23 that's not an easy thing to address or resolve as

24 an industry person.

25 [Slide.


1 So we run into some difficulties. I would

2 like to say publicly that CDER's microbiologists

3 are probably the best in the agency as far as

4 resolving or providing guidance on rapid micro

5 methods. I work with several different divisions

6 of the agency, and they are the most responsive and

7 knowledgeable in helping us in trying to get these

8 things approved through industry.

9 [Slide.

10 But the technology currently is available.

11 There is over 300 companies that are in some degree

12 of commercialization of rapid microbiology

13 technologies of as last degree.

14 It is available for all the types of tests

15 that we need to do. They are generally superior,

16 but the other thing is no one system is going to

17 handle all our problems. You may find that in the

18 typical pharmaceutical company, when these methods

19 are employed, that they may end up with four, six,

20 or eight different technologies to accomplish all

21 the tests that they need to do, and that is one of

22 the problems that people face in looking at it, is

23 looking for one system that is going to solve

24 everything or the fact that one system is validated

25 or approved will solve all their needs. It just


1 will not do that.

2 They allow decisions to be made in a more

3 timely fashion. It is a very difficult, as a user,

4 to say that I am going to find out three days from

5 now that I probably should have shut down

6 production three days ago because I have a problem.

7 It doesn't help, it is very difficult in resolving

8 the cost of the batch, how much additional cost you

9 placed into it which results in end product that is

10 more expensive for the end user.

11 In addition, it is harder to convince

12 people why you have to throw away product on the

13 possibility that it might have maybe been

14 contaminated. Those things are difficult in

15 resolving and addressing.

16 In addition, most of these things will

17 reduce batch release time. Dr. Cooney gave an

18 example of parametric release. I can tell you that

19 eliminating the sterility testing, going to

20 parametric release, realized a more than $3 million

21 annualized savings.

22 That same savings is available, if not

23 more, to go to a rapid sterility test because again

24 you reduce the inventory hold time by that same

25 amount. In the case of aseptically filled drugs,


1 there is no parametric release allowed, so

2 eliminating a 14-day inventory hold time realizes a

3 very significant financial savings to the user and

4 the end consumer and makes it worthwhile.

5 I do have one caution that wasn't on my

6 conclusions before, but it is my concern as of this

7 morning. I am very concerned as an industry person

8 about issuing a guidance document by chemists, for

9 chemists, that is going to be applied to

10 microbiologists without sufficient microbiology

11 review and input because there are many of these

12 documents that get out there, and it makes it

13 extremely difficult for microbiologists to get

14 resolved both in their company, how to get these

15 methods approved and provide appropriate guidance,

16 and there are significant differences between rapid

17 chemical systems and microbiology systems

18 specifically dealing with differences in

19 sensitivity and counts, and how you are going to do

20 these things.

21 Thank you.

22 DR. LEE: Thank you very much. Are there

23 questions?

24 If not, Dr. Sutton.

25 DR. SUTTON: I appreciate the opportunity


1 to address the Advisory Panel.

2 [Slide.

3 The issue that I would like to talk about

4 is the existence currently of documents that bear

5 on this process, rapid microbiology methods. There

6 are four of them I am going to talk about. There

7 are others I could talk about. Dr. Korczynski

8 talked about the ASM document. I am not going to

9 mention that.

10 There are also a couple of ICH documents

11 dealing with validation of new chemical methods. I

12 am not going to talk about that. I am going to

13 talk about the domestic documents primarily in this

14 setting.

15 The first one, of course, is Chapter 1225

16 in the USP, which everyone is fairly familiar with,

17 and Chapter 1223, which is a new chapter,

18 Pharmacopeial Preview Stage, that just appeared in

19 the January-February Pharmacopeial Forum.

20 I also want to talk a little bit in depth

21 about PDA Technical Report 33 and USP Chapter 1227.

22 [Slide.

23 I am going to go over this quickly. I am

24 sure most people in the room are far more familiar

25 with this chapter than I am. It was designed by


1 USP to help the USP staff in analyzing incoming

2 methods from industry and from other sources on

3 chemical assays.

4 After they developed this analysis scheme,

5 they realized it would make a very good general

6 information chapter, and so published it as such.

7 It is from here we get the different terms,

8 accuracy, precision, specificity, detection limit,

9 and so on.

10 [Slide.

11 There are also in this chemical analysis

12 of a new method or an analysis of a new chemical

13 method, certain data elements that are required.

14 These are broken down into different categories

15 depending on the use that this chemical assay is to

16 be set to.

17 [Slide.

18 These data elements then are further

19 broken down into performance characteristics, and

20 depending on the category, certain characteristics

21 are needed or were not needed.

22 [Slide.

23 That is basically where we stand as far as

24 the chemical assays, but what became clearer to

25 many of us who were looking at how to handle


1 microbiological assays, it's that that system

2 doesn't work for microbiology. It just flat-out

3 does not work.

4 There are three kinds of micro assays that

5 we are talking about, and these have been alluded

6 to by other speakers. First, there is qualitative,

7 is there something there, are there viable

8 organisms present, and a quantitative, and the

9 topic of this morning's discussion, by my

10 understanding was plate counts, so I have kind of

11 highlighted that one, how many microorganisms are

12 present, and then there is identification, if there

13 is something there, what is it.

14 Now, one thing on the quantitative that I

15 did want to mention. Dr. Riley mentioned that

16 there is going to be changes in the counts that you

17 get from some of these methods. He used the

18 esterase method as an example. That is a very good

19 example.

20 If you grow cells up on TSA,

21 trypticase-soy agar, or on R2A agar, if you are

22 doing water testing, you are going to get a certain

23 number of organisms that grow. If, however, you

24 change the definition of viability from "ability to

25 grow and form a visible colony" to "does it have


1 esterase activity and an integral cell membrane,"

2 that is a completely different definition of

3 viability, you are going to get a completely

4 different response from that test.

5 In fact, I will share with you that I have

6 been playing with some of the city water down in

7 Fort Worth, Texas in R2A agar. It's good stuff,

8 less than 10 colony-forming units per ml. However,

9 if you move over to this esterase activity,

10 question, you get several thousand viable cells per

11 milliliter.

12 Now, this is kind of, of academic interest

13 at this point, and it is really not that important

14 as far as the pharmaceutical process, because as it

15 passes through the water treatment, these counts go

16 way down inside the plant. However, it does

17 underscore the problem of how do you set specs when

18 your method changes. The other main question is

19 how to demonstrate equivalence to a compendial

20 method.

21 [Slide.

22 There are, as we have mentioned, several

23 different types of alternate methods out there. I

24 have broken them down by slightly different scheme

25 as I want to approach this topic from the proposed


1 USP Chapter on Validation of Alternative Methods.

2 I have broken them down on the basis of qualitative

3 tests, are there something there, quantitative

4 tests, and identification.

5 I have heard a lot of talk this morning

6 about an alternate sterility test. None of these

7 methods is going to provide an alternate sterility

8 test. That is not the correct way to think. With

9 apologies to everyone present, that is not the

10 correct way to think of this problem.

11 What we are doing is we are taking a

12 certain subset of the sterility test, that portion

13 of the sterility test where you take a membrane

14 filter and ask are there viable cells on that

15 membrane filter, and we are changing that part. We

16 are asking the viability question differently.

17 That is all we are doing, we are still

18 sampling 20 units, we are not improving the

19 sensitivity of the sterility test, we are not

20 changing anything about that test except the

21 recognition of viable microorganisms, are they

22 there or not, and by changing that, we can recover

23 14 days in process time and perhaps increase the

24 sensitivity of the test, but not changed the test

25 in any fundamental way.


1 [Slide.

2 Okay. We have talked about 1225, the

3 chemical assays. A companion to 1225 in USP is a

4 brand-new chapter I mentioned, 1223, Validation of

5 Alternative Micro Methods, which is a close

6 companion to the PDA Technical Report 33, which has

7 been discussed at some length already.

8 [Slide.

9 Now, similar to the Chapter 1225 in

10 Chemistry, there are different components depending

11 upon the type of test you are looking at, whether

12 it be an identification test for microbiology or a

13 qualitative type of test for microbiology or, in

14 this case, a quantitative test that would apply.

15 [Slide.

16 However, when we talk about accuracy in a

17 micro test, perhaps we are talking about something

18 slightly different than we are talking about for a

19 chemistry assay. These changes in definitions -- I

20 apologize, I am going to breeze right through

21 these, but they are available in the press -- these

22 changes in definitions are out there both in the

23 PDA Technical Report and in this proposed USP

24 Chapter.

25 [Slide.


1 They need to have some attention paid to

2 them, are we looking at the correct way of

3 analyzing a micro assay. The changes that we have

4 made from Chapter 1225, are these changes

5 appropriate, and we need input, and I am hoping

6 that that is one thing that the Advisory Panel can

7 help out with here.

8 [Slide.

9 They range in ruggedness, and so on.

10 [Slide.

11 Robustness.

12 [Slide.

13 The final one has been mentioned briefly,

14 1227, Validation of Microbial Recovery from

15 Pharmaceutical Articles, and this one is very

16 different from the chemical questions.

17 [Slide.

18 On this one, the question really is can

19 you recover viable microorganisms in the presence

20 of compounds that are inhospitable to microbial

21 growth, can you neutralize preservatives, can you

22 neutralize the antibiotics.

23 Secondly, micro assays are notoriously

24 variable. The accuracy of the plate count is a

25 serious question. One of the problem with the


1 out-of-spec results when you are looking at

2 something like the esterase activity model is that

3 when you ask for a colony-forming unit, everyone

4 assumes that a colony-forming unit is a cell.

5 Well, it is not a cell. In diplococci, it

6 is two cells. In most of the strep, it could be a

7 string of anywhere from 5 to 50 individual cells.

8 For many of the staph, those clusters can get into

9 the hundreds of cells, all of which yield 1

10 colony-forming unit, but on the esterase activity,

11 they would lead to 2, or to 5, or to 10, or to 50,

12 or to hundreds of counts, a very, very different

13 way of looking at the same question.

14 [Slide.

15 So plate counts are going to be a very

16 real problem here.

17 [Slide.

18 I am going to move through here. I am

19 running out of time, I apologize.

20 One thing I do want to spend time on,

21 though, is validation of the recovery, as described

22 in 1227. There are three different types of

23 recovery strategies that are used in traditional

24 methods, recovery in agar, recovery in liquid, and

25 recovery in membrane filtration.


1 Sterility tests, which you talked about a

2 lot this morning, doesn't use recovery in agar. It

3 does for a direct transfer use recovery in liquid.

4 You directly inoculate your sample into a liquid

5 and ask does the liquid turn turbid after 14 days.

6 Cells in liquid grow very, very

7 differently than cells on agar. Membrane

8 filtration is yet a third way, and this is the

9 other main way that the sterility test is used.

10 These three main types all have different concerns

11 that need to be addressed.

12 The final issue in this validation is the

13 recovery of injured organisms, which Dr. Korczynski

14 talked about earlier.

15 [Slide.

16 The other question on plate counts is that

17 you have a very narrow range. If you have much

18 more than 300 colonies, these are very small

19 colonies on a plate, you start seeing crowding

20 effects, you start seeing competition for

21 resources, you start seeing a depression in the

22 number of cells that you should see.

23 So you really can't count more than 300

24 colonies of very small, well-defined colonies on a

25 plate, and you can't really count much fewer than


1 30 because you start getting errors on the other

2 side, so you have to do dilutions.

3 [Slide.

4 In summary, technologies for these

5 alternative methods are not new. Everything we

6 have talked about this morning, esterase activity,

7 bioluminescence, these have all been in the

8 research labs and in academia for decades. This is

9 not new stuff. This is stuff that we are trying to

10 get into the pharmaceutical industry.

11 Guidance documents exist or are under

12 development to help us do that. We need some help

13 in developing these, and we need some guidance.

14 Hopefully, this Advisory Panel will help us out

15 with this.

16 Finally, microbiology validation studies

17 are very different than chemistry validation

18 studies, and I would have to echo Ms. Moldenhauer's

19 concern that what would be really bad at this point

20 -- I would have to phrase it that way -- would be

21 to have yet another chemistry assay that is forced

22 to apply into a microbiology laboratory, because

23 they just don't fit, they are different beasts.

24 Thank you very much.

25 DR. LEE: Thank you. Any questions?


1 DR. HUSSAIN: Just one sort of observation

2 for the committee. I think that the proposal that

3 we presented to you was to address the general

4 issues, not the technical issues, so I think the

5 concerns expressed here are not really the issue

6 from my perspective, because the regulatory process

7 and the Safe Harbor concept, and so forth, I think

8 are the general philosophy that the guidance would

9 go, and I think will be followed by technical

10 guidance.

11 I just wanted to clarify that for the

12 committee.

13 DR. LEE: Thank you.

14 Leon?

15 DR. SHARGEL: Nothing to say.

16 DR. LEE: John?

17 DR. DOULL: I do have one comment. I

18 appreciate what you are saying about chemical and

19 microbiological. You ask three questions - is it

20 there, how many, what is it. I would add a fourth

21 question, will it hurt me, and I would like you to

22 answer that question with science, and not with

23 policy.

24 In order to answer that question with

25 science, we have got to explore somewhat how we are


1 going to answer that question with science. All I

2 am saying is we are grateful because on detection,

3 we don't have much focus on prediction. Somehow we

4 need to get that in there, at least that concept.

5 DR. LEE: Okay. We have a volunteer.

6 DR. HUSSAIN: I think that is an excellent

7 suggestion and I think in the framework of risk

8 management and how we link quality issues to

9 safety, efficacy, and risk, I think we probably

10 should consider bringing that as a topic for

11 discussion later on.

12 I don't have an idea exactly how we will

13 do that, but I think it's an excellent suggestion.

14 DR. LEE: There are several observations

15 that I made along those lines. Number one is that

16 science moving forward, are the tests moving along

17 with it. We have an opportunity of expanding to

18 implement the database informatics. Are we in a

19 better position to assess risk?

20 Obviously, there are technical issues,

21 there are philosophical issues, and I think what

22 this subcommittee is going to be asked to do is to

23 think about the philosophical issues, and if the

24 philosophical issue is given the green light, then,

25 the technical issues will be addressed in due


1 course, I would hope.

2 Tom, you are chair of the PAT

3 Subcommittee. Do you want to say anything?

4 DR. LAYLOFF: I was just going to comment

5 on John's comment or Ajaz. The question is how

6 many organisms, which ones, who is exposed, and

7 when.

8 DR. LEE: On that note, I am going to

9 conclude this morning's session. We are doing very

10 well on time. Lunch is here. We will continue at

11 12:30. Thank you.

12 [Whereupon, at 11:30 a.m., the proceedings

13 were recessed, to be resumed at 12:30 p.m.]



2 [12:30 p.m.]

3 DR. LEE: The afternoon is going to be on

4 two issues - blend uniformity, this is an issue

5 that was discussed last time, and with an update.

6 Ajaz, would you like to introduce the

7 topic?

8 Blend Uniformity

9 Introduction and Overview

10 DR. HUSSAIN: In some ways, we probably

11 will look back at this meeting as sort of a

12 historic meeting, and this is the first

13 recommendation of PQRI, and I really thank Tom,

14 Toby, Sid, and others of PQRI, who have really

15 worked hard over the last two years to build

16 consensus in this proposal.

17 As Helen mentioned in her opening remarks,

18 we have already started the process of withdrawing

19 the drug guidance in anticipation of incorporating

20 the recommendations of this PQRI recommendations

21 into our next draft guidance that will be issued

22 again as draft for public comment.

23 Today is the opportunity to sort of have a

24 public discussion on the recommendations itself.

25 Within the agency, we have formed a separate group


1 to look at this and adopt this as it goes along.

2 This group is different from the group which worked

3 on the blend uniformity proposal that is coming

4 from PQRI. So this is sort of a separation as a

5 peer review process.

6 My own role sort of changed when PQRI was

7 started. I served as the technical director, and I

8 think Tom's working group was essentially getting

9 started, make sure true scientific dialog starts

10 between FDA folks on the committee, as well as the

11 industry members.

12 From that point, I sort of moved away from

13 PQRI process in terms of creating a distance for a

14 rational evaluation process.

15 [Slide.

16 Let me give you a background information

17 on blend uniformity. Many of the slides you have

18 already seen in the previous meeting, but I think

19 it helps to bring that into focus. The slides you

20 see, again, you have seen before.

21 Blend uniformity analysis is an in-process

22 test we do today. It has been the subject of

23 intense debate for about 10 years, and I have sort

24 of criticized the whole processing that we have

25 talked about, but haven't found a solution.


1 I think we have a solution at hand now to

2 deal with it. The controversies and the debate

3 have focused on sample size, sampling errors,

4 segregation following blending operations, lack of

5 correlation with content uniformity, and so forth.

6 Also, I think the issue has been whether

7 it is a cGMP issue or a review issue. Clearly, we

8 had at the Blend Uniformity Workshop, complaints,

9 and so forth, that there has been inconsistent

10 enforcement, some do, some don't, and so forth.

11 The draft ANDA guidance was issued in

12 1999. I will share with you the motivation behind

13 that, and then the PQRI came about to solve some of

14 these issues.

15 [Slide.

16 If I look at the draft Blend Uniformity

17 Guidance, the motivations for these were to sort of

18 improve the consistency in the review process

19 itself. Clearly, when we issue a guidance, we

20 highlight some concerns reviewers have, and the

21 reviewers, from my way of looking at that guidance,

22 raised the issue of some concern that the content

23 uniformity needs to be emphasized.

24 Why would that be, I think in the generic

25 applications one could look at and say there is


1 insufficient information to assure the quality was

2 by design, but that is not a generic issue. I

3 think that issue is a general issue for new drugs,

4 as well as ANDAs from my perspective.

5 Why would one say that? If you look at

6 the submissions, how do we define blending, define

7 the blender type, capacity, operating speed, and so

8 forth. Again, a lot of these issues I think are

9 issues of disconnect between how we operate between

10 the field and review. I think a lot of this

11 information that is already generated are with

12 companies, so in many ways, some of these issues

13 are communication issues rather than science issues

14 per se.

15 The draft guidance recommended for which

16 products blend uniformity testing is needed and

17 how, to some degree, you have to do this, some

18 indication of sampling size and criteria was

19 incorporated into this draft guidance.

20 [Slide.

21 From a science perspective, I think these

22 are sort of my thoughts, which again I have shared

23 with you before, is from a pharmaceutical science

24 perspective, what is the science issue. I think

25 blending is a process which is quite complex when


1 it comes to physical process itself, and

2 performance of a physical unit, performance of a

3 unit operation depends on material characteristics,

4 particle attributes, equipment design, operating

5 conditions.

6 I think we address all of these through

7 our validation in different ways, but I think we

8 lack clear way of making a case that this was by

9 design.

10 This is how an engineer might look at that

11 to optimize the performance of a blender or any

12 other physical unit operation. That is from AICHE

13 Journal in 2001.

14 [Slide.

15 If you continue to argue that today, trial

16 and error is the norm, and have done that on many

17 occasions, the question from a reviewer perspective

18 is do SOPs reflect established heuristics at least.

19 Here are some of the heuristics that we have

20 learned over the years. Again, this is from the

21 same article that the previous slide was from.

22 Here, you look at attributes of material

23 that result in segregation problem either during

24 blending, after blending, and so forth, and many of

25 these are at least not apparent in the SOPs that


1 are practiced today, and the general assumption is

2 the validation process would have addressed this.

3 [Slide.

4 Limitations of current approach, I think

5 unit operations are intended to produce in-process

6 materials that possess optimal attributes for

7 subsequent manufacturing steps.

8 In general, I think the testing mentality

9 instead of control mentality, I would say that the

10 current controls always ensure consistent quality

11 of in-process material. I think one way we do that

12 is to reject through testing.

13 But I think the point again, I keep

14 harping on this, but it's an important point, the

15 physical attributes of pharmaceutical raw materials

16 can be highly variable, and that is one

17 contributing factor to the concern that we tend to

18 express with some of our guidances.

19 [Slide.

20 Again, from a review perspective,

21 limitations of process validation leads to such

22 concerns. Again, a quote from a published paper,

23 "In the spirit of cGMP and how we practice, there

24 may be a disconnect." In some cases, I think,

25 again, this is not a general observation, but


1 illustrates what can happen when quality was not

2 built in.

3 So validation in many ways, some people at

4 least would view that has become a well-rehearsed

5 demonstration that manufacturing formula can work

6 three successive times.

7 That is not the true intent. A lot of the

8 development efforts go towards the whole validation

9 process, but again, it is a perception leading to a

10 concern.

11 In their experience, Harwood and Molnar,

12 "validation exercise precedes a trouble-free time

13 period in the manufacturing area only to be

14 followed by many hours, possibly days or weeks, of

15 troubleshooting and experimental work after a batch

16 or two of product fails to meet specification.

17 This becomes a never-ending task."

18 There is, if you recall, data, that G.K.

19 Raj presented points to some of that thought

20 process.

21 [Slide.

22 With respect to blend uniformity, reliance

23 on end product testing, I think the whole issue

24 here is a question of representative sample. Here

25 is an example where I think the stratified sampling


1 will really help.

2 This is a case study that was sent to me

3 and Helen by a company, an individual in a company

4 who had done this work for PQRI, but I looked

5 through the database, I couldn't find this in the

6 database example, so this probably never got

7 submitted.

8 Here is an example of a validated product

9 which has been on the market, blend sample

10 analysis, I think wonderful results, percent RSD of

11 less than 1, it passed. USP content uniformity

12 Stage I passes. But when you do stratified

13 sampling, you see a trend, in this case, towards

14 the end of the production run, where you have a

15 deviation.

16 So the stratified sampling in this case

17 picks it up, whereas, the blend sample, how the USP

18 sample did not.

19 [Slide.

20 So the draft BUA Guidance that we have,

21 and I think this is what Tom pointed out, changing

22 the focus a bit right now, I think the questions

23 that we asked in the draft guidance and our

24 regulations, and what we have practiced for years

25 and years and years have been focused on drug


1 substance alone, and we assume that demonstration

2 of adequacy of mix with respect to drug alone is

3 sufficient.

4 Is this generally a reasonable assumption

5 with a few exceptions, I think that question comes

6 up, but I think it is, because the manufacturing

7 history, the recall, and so forth, the small

8 numbers would bear that. In a sense, I think it's

9 a reasonable assumption.

10 [Slide.

11 But with respect to PAT, we can ask that

12 question and actually build quality in upfront, and

13 here is an example from Steve Hammond at Pfizer,

14 how he can even look at magnesium stearate in his

15 distribution whether it's homogeneous or not.

16 Currently, we don't. Currently, we look at

17 dissolution as a surrogate for some of these

18 attributes.

19 [Slide.

20 But I just want to sort of put this on the

21 radar screen, is homogeneity with respect to drug

22 substance alone sufficient? In some cases, if the

23 development efforts are not there, it may not be

24 the case.

25 Here is an example of a drug product which


1 does not need blend uniformity testing, does not

2 need content uniformity testing because the drug is

3 95 percent of the product, but homogeneity of

4 magnesium stearate became an issue, and you can see

5 stratified sampling in this case picks up even

6 dissolution failures at the beginning and the end

7 of the run.

8 So the point I am making is I think the

9 stratified sampling brings more rationality to the

10 issue of representative sample and can help improve

11 the whole process of quality assessment, and so

12 forth.

13 [Slide.

14 So, in summary, the PQRI recommendations

15 on blend uniformity analysis, I think analysis of

16 in-process dosage units , collected using the

17 proposed stratified sampling plan, may be used as

18 an alternate to routine blend sample analysis to

19 satisfy the requirements of 21 CFR 211.110(a)(3).

20 That is essentially the letter Toby Massa

21 sent to Janet Woodcock. That is essentially the

22 summary of the recommendation.

23 [Slide.

24 In your handout, I will not read through

25 all of this here for 211.110, but I just provided


1 you the entire section for which this guidance

2 applies. Instead of reading it, I am just going to

3 skip and, for the sake of time, have Tom and others

4 to come and speak to the proposal itself.

5 [Slide.

6 The other issue simply here is it goes to

7 (3), adequacy of mixing to assure uniformity and

8 homogeneity, but the whole section applies to

9 dissolution, disintegration, weight variation, and

10 so forth, also.

11 [Slide.

12 I am going to skip that. This was just

13 for you to have it available to the discussion. I

14 hope you can read it. I think the copies were made

15 on a smaller print.

16 [Slide.

17 The questions that we pose to you, and I

18 hoping that as Toby Massa and Tom Garcia walk

19 through that, they will address that and the

20 committee could discuss and give their

21 recommendations on this.

22 The same questions that we posed before, I

23 am just refining, adding a few clarification points

24 here.

25 PQRI proposes that blend uniformity


1 analysis may not be necessary following development

2 and validation studies. During routine production,

3 dosage unit testing, going back to 10 and 20, are

4 sufficient to document "adequacy of mix," based on

5 a study about sampling plan now.

6 The proposal utilizes a criteria or a

7 classification system which says some products

8 readily comply when the RSD is less than or equal

9 to 4 percent. Some "marginally comply" when the

10 RSD is still less than 6, but greater than 4, but

11 for those, you do additional five consecutive lots

12 before you can sunset the blend uniformity testing.

13 For "readily comply," you sunset routine blend

14 uniformity testing almost immediately.

15 [Slide.

16 So the questions to you are: Do you

17 consider the PQRI proposal appropriate for

18 inclusion in a planned revised FDA guidance? You

19 may consider the following point.

20 Supporting simulation studies assume a

21 normal distribution, is this a reasonable

22 assumption?

23 Was the retrospective data mining

24 sufficient to conclude that "blend uniformity

25 testing in routine in routine manufacture is not


1 predictive of the uniformity of dosage units"?

2 Is this conclusion a necessary condition

3 for regulatory application of the PQRI proposal?

4 [Slide.

5 If the proposed stratified sampling and

6 analysis plan is limited only to bioequivalence and

7 validation batches, how should adequacy of mix be

8 ensured for routine production batches?

9 In a sense, this question goes to the

10 point is the classification, that is,

11 "readily"/"marginally" comply, and proposed

12 additional assessment to justify deleting routine

13 blend uniformity analysis justified?

14 In the absence of blend uniformity

15 analysis, is stratified sampling plus limited, that

16 is, going back to 10 or 20 depending on how many

17 products you test, product testing sufficient to

18 assure content uniformity of the entire batch?

19 [Slide.

20 In some ways, I think the letter Dr. Massa

21 wrote, and I think the recommendation applying to

22 CFR 211, probably has already addressed this, but I

23 just want to be clear that our thinking here is:

24 Should the planned revised FDA guidance only focus

25 on generic drugs or should it be a general


1 guidance, that is, for both new and generic drugs?

2 Our thinking is it is a science issue, and

3 it is not an ANDA versus new drug issue, and it

4 should apply equally on both sides. When we do

5 that, I think we will have to rethink, since the

6 recommendation is only focused on ANDA

7 bioequivalence and validation batches, we will have

8 to go back and see what would be an appropriate

9 counterpart of that in the drug development.

10 With that, I will stop and invite Toby to

11 give the opening remarks followed by Tom Garcia.

12 Tobias Massa, Ph.D.

13 DR. MASSA: Thanks, Ajaz. It is a

14 pleasure to be here today. As Ajaz said, this is a

15 special day for us because this is the first of

16 what we hope will be many recommendations from PQRI

17 for the agency to consider.

18 Although I work at Eli Lilly, I am here

19 today really as the chair of the PQRI Scientific

20 Steering Committee, and as a member of the board.

21 [Slide.

22 If you are not familiar with PQRI, we were

23 formed three years ago, and our charter is to serve

24 as a neutral forum for academia, industry, and the

25 agency to conduct pharmaceutical product quality


1 research and to develop recommendations that lead

2 to public standards.

3 [Slide.

4 I think that the real benefit that PQRI

5 brings is that we are fostering good science to

6 good regulation. I think we have heard variations

7 on that theme this morning, and certainly we have

8 heard people like Jane Haney, the former

9 Commissioner, as well as Janet Woodcock, the

10 current Center Director, talk about the need to

11 emphasize that regulation has to come from good

12 science.

13 We hope that our research will serve as

14 the basis for reducing burden, not only for the

15 industry, but also for the agency, so that our

16 resources can be used more efficiently and used in

17 places where they really do need to be used, and at

18 the same time, we want to either maintain or

19 increase product quality. Reduction in burden

20 doesn't mean reduction in product quality.

21 [Slide.

22 The advantage we have is that we have

23 industry, FDA, and academia cooperating on

24 identifying what the specific product quality issue

25 is, helping design a protocol that addresses that


1 problem, analyzing the data, writing the

2 recommendation, all together and all agreeing that

3 yes, this is indeed the way we need to be going.

4 Additionally, although we don't write the

5 final guidance, guidance is openly discussed in the

6 PQRI forum, and unlike many other instances where

7 there is FDA-industry cooperation, this is one of

8 the few, if not the only, place where guidance will

9 be discussed, and that was by design.

10 I think you have heard earlier, in some of

11 our discussions this morning, that when we are

12 talking about PAT, that the subcommittee is not

13 going to be the place where guidance gets written.

14 It is going to be written within FDA without the

15 industry present.

16 This is a case where we are openly

17 cooperating on guidance. Ultimately, the agency

18 independently has to assess that recommendation.

19 We don't obviously write the final guidance. We

20 might like to write the final guidance, but it

21 doesn't work that way.

22 [Slide.

23 I put this chart of the organization up,

24 not to discuss the organization, but to demonstrate

25 for you the rigor of the PQRI process. If you look


1 at the part here, the working groups, that is where

2 all of the hard work occurs, and all of the PQRI

3 organizations, as well as academics who are

4 appointed from AAPS, work to develop the issue,

5 develop the protocols, review data, and actually

6 write the recommendation.

7 Those recommendations are then reviewed by

8 the Technical Committees. There are three

9 Technical Committees within PQRI, and Blend

10 Uniformity was part of the Drug Product Technical

11 Committee.


13 Each of the member organizations has a

14 representative on this Technical Committee. They

15 review the recommendation, as well as disseminate

16 that document within their own member organization,

17 so all the member companies at PhRMA, for example,

18 were asked to comment on this.

19 Any comments that came back that warranted

20 being addressed had to be addressed by that working

21 group, so it is a pretty rigorous process.

22 Ultimately, the recommendation will go to the

23 Steering Committee, and the Steering Committee role

24 is really to manage the research portfolio and deal

25 with the policy issues of the institute, not really


1 the scientific aspects of it, but we ultimately

2 will send the recommendation to FDA for their

3 consideration.

4 [Slide.

5 These are the members of the group and you

6 can see that it reflects the diversity of PQRI.

7 There are generic company members on there,

8 innovators, as well as a number of people from FDA

9 representing new drugs, generics, as well as

10 compliance.

11 [Slide.

12 Now, the charter, as Ajaz has mentioned,

13 of the Blend Uniformity Group was to address the

14 specific part of the guidance that said blend

15 sampling is required to demonstrate adequacy of

16 mixing, and they specifically referenced Part 211.

17 They undertook a rather long process, and

18 this was our first project within PQRI. We learned

19 a lot about how to do things and quite a few things

20 about how not to do things, as Tom can attest, in

21 the PQRI process, but they basically looked at the

22 gaps between where the regulation was and where we

23 thought current science was, and to come up with

24 alternative approaches.

25 [Slide.


1 Tom will go into the nitty-gritty of the

2 recommendation, but basically, we came away saying

3 that blend uniformity is not always predictive of

4 mixing and that stratified sampling and testing of

5 in-process dosage units is a viable alternative to

6 the use of blend uniformity analysis.

7 [Slide.

8 Now, these study results, as I had

9 indicated before, go through a review process

10 within the institute, and ultimately get submitted

11 to FDA. As per the bylaws of PQRI, there are two

12 options that the agency has. They can either

13 accept our recommendation and change the guidance

14 or regulation as appropriate, or they will reject

15 our conclusion, and if they do, they will respond

16 to us in writing indicating where the faults were

17 in the data and the recommendation that we made, so

18 that we can address this and hopefully come back

19 with a recommendation that is acceptable.

20 Again, that is one of the beauties of

21 PQRI, because that doesn't happen anywhere else.

22 [Slide.

23 Now for us, obviously, success is that the

24 agency will accept the recommendation. We think

25 that we are well on the way for that to happen. To


1 address one of Ajaz's questions, although this

2 guidance was specifically dealing with ANDAs or

3 generic products, we feel that this applies to new

4 drug chemistry situations, as well as the Office of

5 Generic Drugs, as well as the GMP part that would

6 be dealt with, with the Office of Compliance.

7 We think this recommendation ought to be

8 applied wherever blend uniformity is an issue.

9 With that, I will turn it over to Tom

10 Garcia from Pfizer, who chaired the Blend

11 Uniformity Group, and he is the guy who will answer

12 all your questions about the science stuff.

13 DR. HUSSAIN: Vince, may I make a comment?

14 Something Toby said, I just want to make sure I

15 clarify that for the record.

16 Good guidance practices would not allow us

17 to even share a draft guidance that we are working

18 with PQRI, so that doesn't apply. I think we

19 discussed a published draft guidance, so that was a

20 different scenario. We have to follow the good

21 guidance practices, and that would also apply.

22 DR. LEE: Thank you.

23 Thomas P. Garcia, Ph.D.

24 DR. GARCIA: Today, I want to address two

25 things, just to briefly go over the recommendation


1 that PQRI came up with. It is very similar to what

2 I presented back in November of last year, so my

3 apologies for those that are going to hear it

4 again, but for those of you that weren't here in

5 November, you get a shot to see it.

6 The second thing will be the results of

7 the data mining effort that PQRI conducted to

8 support the recommendation.

9 [Slide.

10 The recommendation that we are proposing

11 does a number of means. First of all, it

12 collectively considers the uniformity of the powder

13 blends and the dosage units that are manufactured

14 from them.

15 We really believe that you couldn't look

16 at the two kinds of uniformities in a silo, they

17 are interrelated.

18 The second thing, it acknowledges the best

19 way to assess blend uniformity may be indirectly

20 through analysis of the subsequent dosage units

21 coming out of that batch. Mainly what we are

22 talking about here is a sample bias issues that

23 arise when we try to take blend samples.

24 The test is very simple to use. It

25 maximizes use of all the data and acceptance


1 criteria that we are presenting are very easy to

2 interpret. They don't involve detailed statistical

3 analysis or anything like that, very basic

4 calculations.

5 Finally, this recommendation is very, very

6 discriminating when you have a poor quality batch.

7 If you have some segregation that occurs during the

8 dosage unit manufacture, it will pick it up, and

9 there is a good chance that you will end up

10 rejecting that batch.

11 [Slide.

12 The first part of the recommendation

13 starts out with process validation for the blend.

14 Basically, what you do is you have to take at least

15 10 locations out of the blender, take triplicate

16 samples from each of those locations. You test one

17 sample from each location. Your acceptance

18 criteria is the RSD is less than or equal to 5

19 percent, and all individuals are within plus or

20 minus 10 percent of the mean absolute.

21 This is different than 90 to 110 percent

22 because there were a lot of instances that the

23 working group members cited where you would have a

24 very consistent bias in the sampling technique

25 where you either preferentially sampled the drug or


1 the excipients. In other words, you may have a

2 mean of 110 percent, but an RSD of 2 percent. That

3 is all due to sampling bias albeit it in a very

4 consistent manner.

5 So, hence, we incorporated it's all based

6 on the mean.

7 If you pass, congratulations, the blend is

8 done. If the 10 samples initially tested fail, you

9 assay the second and the third day sample from each

10 one of those locations. Basically, what we are

11 doing now is perform some sort of investigation

12 into the cause of the failure of the first 10

13 samples.

14 If that cause is related to mixing, then,

15 you failed that validation batch. You have got a

16 problem, you need to go back to development and

17 figure out how to address that.

18 If the problem is determined to be

19 something attributed to non-mixing, a problem, for

20 example, may be an analytical error or sampling

21 bias, then, you proceed to Stage 2 testing of the

22 dosage units.

23 [Slide.

24 This is the second half of the validation

25 approach, and this is addressing the dosage units.


1 What you do is during your compression or your

2 filling operation, you take at least 7 dosage units

3 from at least 20 locations during the filling or

4 compression operation.

5 Stage 1 testing is examining at least 3 of

6 those dosage units. Your acceptance criteria are

7 the RSD of all the individual tablets or capsules,

8 whatever, is less than 6 percent. Here is the

9 critical one. Each location mean of these 20

10 locations up here, the mean has to be between 90

11 and 110 percent of label claim. That is the most

12 discriminating part of this approach.

13 Finally, all individuals have to be within

14 75 to 125, and what we are looking for in this last

15 point is just if by some remote chance, happen to

16 get a tablet at 126 or 130 percent, but passed

17 everything else, that batch should be rejected. We

18 are looking for superpotent or subpotent tablets.

19 If you find those, you should stop.

20 If you pass this, the process is

21 validated. If you fail, you progress to Stage 2

22 testing where you analyze at least four more

23 tablets or capsules, so a total of 7 dosage units

24 are now tested in combination with Stage 1 and

25 Stage 2.


1 The acceptance criteria are the same.

2 Obviously, if you pass, you are done; if not, the

3 batch fails. Really, Stage 2, all it does is it

4 gives you a second chance, if you have got a mean

5 value of, say, at 89 1/2, to bump that up. It gives

6 you one more chance to pass it.

7 [Slide.

8 Now, for routine manufacture, once again

9 we are advocating the use of stratified sampling.

10 The first question you have to ask -- and Ajaz

11 alluded to this in his presentation -- is do you

12 readily comply or does your product not readily

13 comply?

14 "Readily comply" is described in the box

15 up on the top. During your exhibit or validation

16 batches, all of the dosage unit RSDs were less than

17 or equal to 4 percent, all your means results were

18 within 90 to 110, and the 75 to 125 percent for

19 individual dosage units that applies.

20 If you meet this criteria, then, you

21 readily comply. What you do is you go during your

22 filling operations or compression operation, you

23 take at least 3 tables or capsules, dosage units

24 from at least 10 locations. What we are doing here

25 is we are trying to get 30 dosage unit per USP


1 testing, so 10 locations, 3 dosage units.

2 The acceptance criteria here is the mean

3 is between 90 to 110 percent, and that is weight

4 corrected, and the RSD is less than or equal to 5

5 percent.

6 If you meet that criteria for the first of

7 the dosage units from each of the locations, then,

8 you have passed, you have demonstrated adequacy of

9 mix.

10 If you don't, progress to Stage 2 and test

11 the remaining 2 dosage units from each of the

12 sample locations, and in this particular instance,

13 once again you rate and corrected all your results,

14 mean between 90 and 110 for all 30 now, and your

15 RSD less than 6 percent, so you get another 1

16 percent for that.

17 If you do not readily comply, you are

18 going to come right down into Stage 2, so you have

19 got to test 30 dosage units rather than being given

20 the opportunity to go to 10 plus 20.

21 If you come down this path, you don't

22 readily comply, and you meet the acceptance

23 criteria, up here, 5 batches in a row, then, in the

24 future, you can come down this level, the go to

25 Stage 1 test. So, just because you don't readily


1 comply at the first few batches after validation

2 doesn't mean you are locked in there forever. You

3 do have to prove, though, that you have made some

4 changes to your process and it's under control.

5 [Slide.

6 That is a very brief tour of the

7 recommendation. By the way, the recommendation and

8 the datamining report are located on the PQRI web

9 site, PQRI.org, going to Blend Uniformity Working

10 Group, I think, or something like that, and there

11 is a nice icon. You can click on it, and you can

12 get both of these documents there.

13 So the second part I want to talk about is

14 the results of our datamining effort.

15 [Slide.

16 These tables on the slide summarize the

17 data that was submitted. We got 149 batches

18 submitted from 8 companies, and the distribution of

19 how many batches each company submitted is on the

20 bottom plot here.

21 Of the batches, about half or over half of

22 them were less than 5 percent active ingredient,

23 which is what we wanted, because typically, your

24 more dilute blends are going to give you more

25 problems than when you have higher concentrations


1 of active, so that was good.

2 We did get a nice distribution between 5

3 and 15, 15 to 25, and greater than 25 for the

4 remainder of the batches, but really here is where

5 we were interested in, but we also wanted to make

6 sure that it served the purpose up there.

7 We had 12 direct compression products or

8 batches, 67 wet granulation, and 70 dry

9 granulation, so we had all three of those

10 manufacturing processes covered. We only had

11 tablet data submitted. We didn't get any capsules,

12 sachet powder packets, and I will address that in

13 the next slide.

14 Our batch sizes had quite a range. I

15 believe the less than 100 kilos, the smallest batch

16 was something like 25 or 30 kilos, I can't remember

17 exactly, but it was in that range. Then, the

18 greater than 400, I think this was up around 1,200,

19 1,300 kilos, so we had a very wide range when it

20 came to batch sizes.

21 [Slide.

22 Now, the big question there we wrestled

23 with -- is capsule data essential to apply this

24 recommendation?

25 There was a lot of discussion with the


1 working group about this, and what we decided, you

2 know, the scientific explanation we came up with is

3 general processes that you use to produce a blend,

4 whether it be for tableting or encapsulation or

5 powder filling, is the same.

6 The processes to make that blend do not

7 differ significantly. Thief sampling error is just

8 as likely if you have a capsule or a tablet blend,

9 and that is another thing we were addressing with

10 this recommendation.

11 The key thing is the capsule weights,

12 although they may be skewed, all the data that we

13 analyze in our recommendation is weight corrected,

14 so that takes care of that problem.

15 So our conclusion was, yeah, we would have

16 loved to have had some capsule data to test it, but

17 we don't feel that is necessary just because the

18 processes up to encapsulation or compression are

19 basically the same.

20 [Slide.

21 The acceptance criteria that I presented

22 in the previous slides were all generated using

23 computer simulation, Monte Carlo simulations. The

24 way we did this is we had a number of operation

25 characteristic curves that we generated, and each


1 one of those curves had a number of points on it,

2 probably about 25 or 30 points. Each point had

3 5,000 simulations, so we used extensive use of

4 Monte Carlo simulation to come up with these

5 criteria.

6 The one thing about the computer

7 simulation is it assumes that the data was normally

8 distributed, so one of the things we wanted to look

9 at is all the data we had submitted, how was that

10 distributed.

11 Eighty-five and 89 percent of the batches

12 were normally distributed, and what these two

13 numbers refer to is within a location and between

14 locations, we look at the distributions of both of

15 those. So you are looking at micromixing and

16 macromixing.

17 So, for the most part, there was normally

18 distributed. The instances where it wasn't

19 normally distributed was due to outliers.

20 Remember, we asked industry to submit us data to

21 challenge our proposal.

22 They send us some good data, but as

23 expected, they sent us some data they had problems

24 with, and that was exactly what we wanted to do,

25 how would this proposal perform.


1 If you sent us your best data, well, then,

2 anything will pass the good criteria, so we wanted

3 a challenge, and that is probably the cause of

4 seeing up to 15 percent non-normality, those

5 particular batches.

6 The other thing is even if you are in a

7 non-normal situation, the acceptance criteria in

8 our recommendation becomes much more discriminating

9 and restricted. You are going to fail more

10 batches, batches are not going to slide through.

11 So if it is not normal, it is harder to pass.

12 So the conclusion that we came up with,

13 our assumption is justified.

14 [Slide.

15 The second part of the datamining effort

16 was to test the hypothesis that blend uniformity

17 testing during routine manufacture is not always

18 predictive of the uniformity of dosage units.

19 [Slide.

20 This slide right here I put up back in

21 November, and it pretty much sums it up. What it

22 is, is the dosage unit uniformity is a function of

23 the blend RSD. If you had a 1 to 1 correlation,

24 the blend and the dosage unit RSDs will perfectly

25 correlate. You have this 45-degree line going up


1 here sort of in red.

2 From the points of the data, you can see

3 that correlation is lost. What we did is we

4 divided it up into three sections. That particular

5 part, RSD of the blend is between zero and 3

6 percent. We then looked at between 3 and 5

7 percent, and then greater than 5 percent.

8 [Slide.

9 What we found, and you actually cheat

10 ahead here and go to the next conclusion slide, if

11 the blend RSD is less than 3 percent, and we had

12 pretty good correlation between blend and dosage

13 unit uniformity, and, in fact, in some instance,

14 you could see that the dosage unit uniformity is

15 higher than the blends.

16 Potential causes of this, it could be

17 weight variability, the dosage unit, another

18 component are there. Also, what is just as likely

19 is you may have a uniform blend or a blend in the

20 mixer, as soon as you start transferring it to the

21 tablet press or capsule-filling machine, you are

22 getting segregation. As a result, your RSD goes

23 up.

24 So, in this region, it is of use. As you

25 start going up to the 3 to 5 percent range, though,


1 you start losing that correlation, you see a couple

2 points on the line, but down here, you are seeing

3 higher blend RSDs and low dosage unit RSDs. So

4 here is where you are losing your correlation.

5 By the time you get to the situation where

6 your blend RSD is greater than 5, there is no

7 correlation between blend and dosage unit

8 uniformity.

9 So that is the first conclusion that we

10 came out with, you know, whether or not the

11 hypothesis held.

12 [Slide.

13 We do believe, though, that there are many

14 instances where blend uniformity is value added.

15 For example, during process development, one of the

16 things we wanted to stress is our recommendation is

17 not a substitute to go out and do poor process

18 development and blend mixing development. You

19 still have to do that, and as you saw, the number

20 of tablets that you have to test in Stage 1 and

21 Stage 2 testing is significantly different, so it

22 is in your best interests to make sure you get a

23 uniform blend and one that you can sample.

24 Other instances where it is warranted is

25 in validation. The one exception that we


1 highlighted is if you have toxic products where you

2 have contained processes, it may not be in your

3 operator's or safety's best interests to break

4 those containers. We feel that stratified sampling

5 approach is still very discriminating to indirectly

6 measure the uniformity of the mix.

7 Finally, during investigations, blend

8 uniformity data is always valuable information to

9 have.

10 We also came to the conclusion that blend

11 uniformity testing is not necessarily the best

12 choice during routine manufacturing to demonstrate

13 adequacy in mix. The stratified sampling approach

14 is superior, we feel, to it for a number of

15 reasons. You eliminate all the sampling bias that

16 you potentially have.

17 The other thing is it picks up if you have

18 that segregation between the blending operation and

19 filling operation, it will detect that. The blend

20 can be perfect in the mixer, but what happens to it

21 when you transfer it to the press or the

22 capsule-filling machine. If you get segregation,

23 it's all for naught.

24 Another thing is regardless of how the

25 blend data looks, you are always going to go and


1 test the dosage units. If the blend data is good,

2 you are still going to have to test the dosage

3 units because you have got to make sure you don't

4 have segregation during transfer. If the blend

5 data is bad, you are going to test the dosage units

6 to see whether or not it is sampling error, so it

7 all comes down to the dosage units.

8 [Slide.

9 The final conclusions that we have or one

10 thing we want to emphasize is this approach is only

11 one of many ways that you could assess blend and

12 dosage unit uniformity. There is other ones out

13 there.

14 GMP should be flexible and allow a number

15 of different techniques to be used. For example,

16 if you want to go with some sort of process

17 analytical technology, such as NIR, that is

18 perfectly acceptable, as well. If you want go

19 continue going with traditional blend sampling and

20 dosage unit testing, that is fine, as well.

21 This is one way, but one technique that we

22 feel is very, very discriminating.

23 That concludes my presentation.

24 DR. LEE: Thank you very much.

25 Committee Discussion


1 DR. HUSSAIN: Just to sort of summarize

2 the process that we plan to use, as I said, we have

3 a separate group which is going through these

4 recommendations carefully and get back to PQRI if

5 we have any disagreements, and so forth.

6 We have also requested a statistical

7 consult on the statistical analysis. The

8 preliminary findings I think from the statisticians

9 have been that I think in terms of the

10 retrospective data analysis really can't conclude

11 the hypothesis holds. I think that is the initial

12 part that has come back to us from the statistical

13 group, so that is one of the issues, the questions

14 I pose to you.

15 In addition, so what we would like to do

16 is have you discuss the issues, and we will take

17 back this discussion and recommendations, and

18 re-discuss that and get back to PQRI and provide a

19 process for revising the draft guidance.

20 DR. LEE: Art.

21 DR. KIBBE: First, I would like to say

22 that I strongly support eliminating of a test which

23 not only doesn't predict, but might mislead.

24 I have some questions about some of the

25 recommendations and a couple of other little


1 things, so, first, when you decided on the number

2 of tablets to take at each stage during your

3 testing, did you take into account the batch size

4 when you decided that you were going to use those

5 numbers?

6 DR. GARCIA: No. We debated this issue.

7 We had Jerry Planchard, who was on the committee,

8 who is PhRMA statistician, and we said, you know,

9 we are saying you take 20 locations, 140 tablet, 7

10 per location, whether it's a 20-kilo batch or a

11 2,000-kilo batch, and do we want to look at that.

12 We decided not to because the number of

13 locations and the number of replicates in the

14 locations were all determined through Monte Carlo

15 simulations, and the statistical power behind those

16 numbers was established and one that the group felt

17 comfortable with.

18 DR. KIBBE: I understand you gave the

19 value of the batch based on the weight, but for me,

20 I sit here thinking is this a 2 million tablet

21 batch or is it a 25 million tablet batch, and now I

22 am thinking of are we taking 20 times 3, 60 tablets

23 to characterize a 2 million tablet batch and to

24 characterize a 25 million tablet batch, and that is

25 where I am coming from.


1 I mean I am sure your statistician is

2 doing something, but if you make 150-mg tablets,

3 and you make a 4-kilo batch, and you make 500-mg

4 tablets and you make a 4-kilo batch, you have got

5 completely things.

6 I just wonder if it's worth looking at it

7 in terms of number of tablets in the process. I

8 don't have any real basis for arguing four more,

9 but I wonder if it's overkill on a small batch and

10 underkill on the big batch.

11 DR. GARCIA: Right now we are looking at

12 30 per USP. So we are way beyond that. This is a

13 question, though, that we could get back and get a

14 formal response.

15 DR. KIBBE: When you didn't have any

16 capsules at all, and you said, well, okay, we

17 blend, but I don't know how many capsules you used,

18 what granulation before you blend, and what

19 granulation changes the ability of actives to

20 stratify the changes the way those blends are

21 re-mixed when you actually go to tablet. It

22 changes segregation patterns. Even dry granulation

23 can't, because when you start to bind up the active

24 with some of the inactives, then, they move in the

25 blender differently. You don't do that with


1 capsules.

2 I am not as comfortable as you seem to be

3 that only tablet data lets you make the connection

4 to capsule data. I don't know whether anybody else

5 feels that way.

6 DR. LEE: Art, are you proposing that this

7 proposal is not complete?

8 DR. KIBBE: I am asking questions. I hope

9 somebody has better answers than I have.

10 DR. GARCIA: Back to your previous

11 question, direct compression is covered in the

12 tablets, though, the data that we had. So direct

13 compression is probably your more common capsule

14 doing mix and process.

15 DR. KIBBE: Well, yes and no. I wouldn't

16 use freeze-dried lactose in a capsule necessarily,

17 and that has different stratification

18 characteristics, and also some of your directly

19 compressible excipients have an ability to absorb

20 active ingredient. I am not saying you are wrong,

21 but I am not yet convinced that using only tablets

22 is a guaranteed extrapolation to capsules.

23 I think those are the two things that I

24 was most interested in. I have got a couple other

25 little notes, but I can come back to it.


1 DR. LEE: Let's hold it for now. Let me

2 read the two questions.

3 DR. DeLUCA: Can I comment on what you

4 just said?

5 DR. LEE: I would like to read the

6 questions first.

7 The committee is asked to address two

8 questions.

9 1. Do you consider this proposal

10 appropriate for inclusion in a planned revised FDA

11 guidance? If no, there are a couple of things that

12 we need to consider. That is Question No. 1.

13 2. Should this planned revised FDA

14 guidance only focus on generic drugs or should it

15 be a general guidance?

16 Those are the two questions.

17 DR. HUSSAIN: The questions I posed, if I

18 could put them back on.

19 DR. LEE: Am I reading the wrong one?

20 DR. HUSSAIN: If you look at my

21 presentation, there are three set of slides for the

22 questions, the last three. They are essentially

23 the same, I have sort of refined that a bit more.

24 DR. LEE: Would you please read it.

25 DR. HUSSAIN: Question 1. Do you consider


1 the PQRI proposal appropriate for inclusion in a

2 planned revised FDA guidance?

3 Within that, sort of a question I am

4 asking: Supporting simulation studies assume a

5 normal distribution, is this a reasonable

6 assumption?

7 Was the retrospective data mining

8 sufficient to conclude that blend uniformity

9 testing in routine manufacture is not predictive of

10 the uniformity of dosage units?

11 Related to that, is the above conclusion a

12 necessary condition for the regulatory application

13 of the PQRI proposal?

14 Question 2. If the proposed stratified

15 sampling and analysis plan is limited only to

16 bioequivalence and validation batches, how should

17 adequacy of mix be ensured for routine production

18 batches?

19 In relation to that question is: Is the

20 classification "Readily" and "Marginally" comply

21 and proposed additional assessment to justify

22 deleting routine BUA justified?

23 In absence of BUA, is stratified sampling

24 plus going back to the 10/20 limited product

25 testing sufficient to assure content uniformity of


1 the entire batch?

2 Those are the two questions related to

3 that.

4 The third question was generic ANDA for

5 all. That has already been addressed, so we don't

6 have to go to the third question.

7 DR. LEE: Let's focus on the first two

8 questions. I have asked Pat to give lots of

9 thought to those questions and apparently he is

10 ready to address those questions.

11 DR. DeLUCA: I just want to compliment the

12 committee. It certainly was a thorough study

13 albeit just focused on tablets.

14 I guess in answer to Art's question with

15 regards to capsules, I think if the recommendation,

16 in my mind, if the recommendation was that blend

17 uniformity was predictive, and based on just the

18 tablets, I agree with you, I would say that I think

19 that you needed to include capsules in here also to

20 make that recommendation.

21 I think because the recommendation is that

22 it is focusing in on stratified sampling analysis,

23 and of the dosage units, then, I think that

24 precludes that possibility, that there is a

25 problem, and capsules need to be included in here.


1 I guess in going directly, and we haven't

2 had really a chance to I think dialogue with the

3 committee members, I just very briefly talked with

4 them, I guess to answer the questions that are

5 here, do you consider the proposal appropriate for

6 inclusion in a planned revised FDA guidance, I

7 would have to say yes, and I think Ajaz has

8 indicated that that is something you are going to

9 be dealing with.

10 So it looks like that question there is

11 answered. I had a couple of questions with regards

12 to the -- when you say, in the dosage units,

13 locations, what do you mean by a location? I

14 understand it from taking a blend sample with the

15 thief, where the location is in the blender, but

16 when you are on a compression, when you are

17 compressing, what do you mean by a location?

18 DR. GARCIA: It's a sampling time during

19 that run, anywhere from T equals zero, start out at

20 the end.

21 DR. DeLUCA: I thought it was that, but it

22 wasn't clear and I wanted to make sure that you

23 weren't taking some blend locations and then

24 compressing those separately.

25 DR. GARCIA: It's a function of the


1 percentage of the batch that a dosage unit is made.

2 DR. DeLUCA: Would that answer Art's

3 question with regards to size of the batch, the

4 locations?

5 DR. GARCIA: No, he is questioning the

6 number of locations.

7 DR. KIBBE: No, I am questioning as far as

8 number of dosage units.

9 If you have a 25 million batch, and you

10 take it during 20 times, then, you are taking 10

11 or 7 tablets out of a million.

12 DR. GARCIA: No. Oh, wait, a 25 million

13 batch, okay.

14 DR. KIBBE: Right? If you have got a 2

15 million batch, you are taking 7 sample out of

16 100,000, and the question is, in my mind,

17 especially after this morning's microbiology thing

18 about how well we can predict anything from that,

19 you know, 10 out of 100,000 is a much bigger

20 percent sample out of the population than 10 out of

21 a million, and why don't we adjust for that.

22 I am not a statistician, so I don't know

23 how much power we are gaining or losing in the

24 process. That was my question.

25 DR. GARCIA: I am going to defer that


1 question. We will give you answer from the PhRMA

2 statistician on the group, but my guess, and this

3 is just an opinion, is when you get to a certain

4 point, your statistical power, the gains you

5 achieve after you get to a certain level diminish,

6 and I believe that is where he was coming from on

7 this. But let's let the statistician answer that.

8 I am confident that they will be able to provide

9 you a suitable answer.

10 DR. LEE: I wish that our statistician was

11 here, but he is in Houston today.

12 Let me also at this time introduce two

13 guests, Ken Morris from Purdue and Gary Hollenbeck

14 from Maryland. Please feel free to contribute your

15 thoughts.

16 DR. HOLLENBECK: Would now be a good time?

17 DR. LEE: Please.

18 DR. HOLLENBECK: I would just like to ask

19 a question. Is there any reason why you could

20 speculate there were no capsules submitted as part

21 of the dataset?

22 DR. GARCIA: I don't have any idea.

23 DR. MASSA: Even with the data we had, we

24 had to work very hard to get data from our industry

25 colleagues. This is not an easy process at all.


1 DR. HOLLENBECK: No, and you should be

2 commended for that. For those of us who were

3 around when PQRI started, this is a special moment

4 I think when we look at what you have accomplished.

5 Tom, in your slide on routine manufacture,

6 I am not sure I understood what happens after five

7 consecutive batches have been successful. What

8 happens at that point?

9 DR. GARCIA: This is for those products

10 that did not readily comply. You have to go down

11 to testing 30 dosage units immediately, right off.

12 If after five consecutive batches you meet the

13 "readily comply" criteria, then, for the sixth

14 batch, you do down testing 10 plus 20.

15 DR. MORRIS: Tom, I had one question, and

16 we sort of touched on this at the meeting, whenever

17 it was, a year ago, but I agree that the normal

18 distribution is probably not a valid assumption,

19 it's very often, but as you say, it is the most

20 conservative stance, how concerned is the committee

21 based on the data you have seen that it is going to

22 be too harsh, that is, that you are going to be

23 failing things more often than you expect based on

24 its not being normal?

25 DR. GARCIA: Part of the datamining report


1 that is available on the PQRI web site shows the

2 percentages of batches that passed various

3 criteria, USP, PQRI, PDA criteria. There was, for

4 this subset of data, the PQRI was a little bit more

5 selective than the USP test.

6 The one that was really rejecting the

7 batches was the PDA approach. For the details of

8 that, to answer your question, I refer you to that

9 report on the web site.

10 DR. HUSSAIN: The question, I think Gary

11 raised that in terms of the five. Again, in my way

12 of thinking, in a sense, what is different with the

13 five consecutive batches, and so forth, what

14 happens from a mechanistic perspective, because I

15 think somehow the thought process, the numbers are

16 there, but how do we justify those numbers? I

17 think some discussion on that would be helpful.

18 DR. GARCIA: Ajaz, if I could address

19 where the 4 percent came from, on the operation

20 characteristics curves, for those of you that have

21 the document, I think it is Attachment 2 or 3, if

22 you look at when the dosage unit RSD approaches

23 about 3.8 percent, somewheres in there, you start

24 sliding down, and the number of batches that passed

25 the criteria significantly decreases. This is


1 where you are getting on the steep part of the

2 curve. That is where the 4 percent came from.

3 The five consecutive batches gives the

4 formulators of the process scientists a chance to

5 go and improve the process. We felt if you are

6 passing with an RSD of 5 or 6 percent routinely for

7 dosage units, your product really ought to be

8 reexamined.

9 So we said if you have made some process

10 improvements post-approval, and you demonstrate

11 that you get that RSD below 4 percent consistently,

12 then, you shouldn't be punished beyond that side of

13 the ladder, and that is where that came from.

14 DR. LEE: Pat has a recommendation for

15 this committee and I am not sure whether the

16 committee feels the same way.

17 DR. RODRIGUEZ-HORNEDO: I just have a

18 comment and a question regarding the first

19 question, and this is to Tom.

20 Regarding the datamining, I find it

21 interesting, and I was present at your seminar here

22 in November where you presented these, probably

23 with more time to present, is there any correlation

24 between the method of whether it's granulation or

25 dry granulation or direct compression, or anything


1 that correlates with a high RSD for the blend?

2 I am certain you thought about that, so

3 that it doesn't correlate with the dose RSD, was

4 there anything?

5 DR. GARCIA: We didn't break it down to

6 that degree. A lot of that would probably be in

7 your formulation and process development.

8 Different companies are going to do different

9 degrees of the science that goes into that.

10 If you do a lousy job of selecting your

11 formulation excipients, particle sizes, grades,

12 things like that, and don't control your drug

13 substance, you are going to have some problems with

14 RSD, but we did not break that out, take the

15 analysis to that degree.

16 DR. LEE: Ken?

17 DR. MORRIS: Just following up on that

18 point actually, in the dataset you have, where you

19 have already analyzed the data in terms of

20 reconciling the unit dose versus the blend

21 uniformity, do you know, can you break those data

22 out by DC versus wet granulation?

23 DR. GARCIA: Not now. I believe it's in

24 an Excel table, if you wanted to go back and look

25 at it.


1 DR. MORRIS: I think that would really be

2 worthwhile doing, I think.

3 DR. LEE: It seems to me, just posing, I

4 am stir the pot or something, it seems to me that

5 maybe this should be put to peer review like

6 journal, conflict of interest.

7 DR. DeLUCA: Let me comment on that now

8 that you have brought that up. That is in the

9 plan. These have been put in the form of

10 publications, both studies, and the recommendation

11 and the datamining.

12 We are going to put it out with a review

13 on this, and I am not finalized on this. I will

14 after hearing some dialogue, I did talk with Tom

15 and with Ajaz on how to proceed with this, but the

16 thought was that we put it out into the journal,

17 and invite comments from the readers on this, that

18 could be fed back to the committee and FDA.

19 DR. HUSSAIN: The thought process, I

20 discussed this, I think it would be a good idea,

21 but I think the timing probably won't work for this

22 case. I think the peer review in a sense would

23 need to occur within the FDA and with this case

24 from that perspective to maintain the timeline that

25 we have.


1 I think for the future projects, I think

2 we could time it in such a way that both could

3 occur. So I had some concern with the timing of

4 that process and how does that link to FDA process.

5 In my mind, I think the FDA process would

6 sort of move on with an internal review,

7 statistical evaluation, and so forth.

8 DR. LEE: I can turn it around in two

9 weeks.

10 Yes, Art.

11 DR. KIBBE: Just a couple of more things.

12 First, this report is basically an indictment --

13 that is a bad word -- of the traditional blend

14 uniformity methodologies, so that your conclusions

15 that blend uniformity should be waived in place of

16 stratified sampling might not necessarily be true

17 if we were using an in-process uniformity system

18 that could do a good job of making sure the blends

19 stay uniform, and if that process then was

20 reapplied to, or the same sensing system was

21 applied to the fill line, so that we knew that the

22 fill process was going in, and that I think is the

23 direction that PAT would like to see go.

24 Blend uniformity, the way it is being done

25 is what is not predictive, not necessarily that we


1 couldn't come up with a new methodology that would

2 be predictive.

3 DR. GARCIA: Go back to the last slide I

4 had. We said this is just one of many ways to do

5 it. GMP should be flexible.

6 DR. KIBBE: And I agree with that.

7 DR. GARCIA: We had to come up with a

8 recommendation that could be applied across

9 industry. A lot of companies probably don't have

10 that technology in-house, but, yes, I agree with

11 you, that PAT is a separate issue that could be

12 used in lieu of this approach, which could be used

13 in lieu of current approaches.

14 DR. MASSA: There is one other thing to

15 add, is that even if you had an in-process

16 analytical method that would measure the blend,

17 that wouldn't account for post-blend segregation at

18 the time of compression.

19 DR. KIBBE: No, but you can imagine the

20 same sensing system on a fill line to monitor

21 post-blend segregation as the tablets were being

22 filled and how close to the final compression would

23 you have to be.

24 I mean I can imagine the system doing away

25 with a whole bunch of unit dose assays, but a


1 second small thing is because of the way you

2 sample in your stratified sampling system, it is

3 possible to determine that in a given batch with a

4 given product, that 80 or 90 percent of the

5 products that is manufactured are within compliance

6 because they are right up there, and the last 10

7 percent of the batch falls out, and would that be

8 grounds for a company just separating as the

9 material came off, and sub-lotting, and then

10 getting rid of the last?

11 DR. MASSA: Unfortunately, that is not

12 consistent with cGMPs. It might be if we got to

13 PAT later on and had a way to do that, but right

14 now you can't do that.

15 DR. GARCIA: If that's the case, I think

16 you need to take a look at that formulation and

17 process.

18 MR. BUHAY: But if there is data which

19 supports that approach, that shows the processes so

20 consistent, and that variation begins at a

21 measurable point, it is consistent with GMP.

22 DR. HUSSAIN: One point I wanted to make

23 for the record in the sense I think the survey that

24 was done by PQRI, and the discussion at the

25 workshop, and even the data, blend sampling is in


1 thieves, I think can pose a problem, does pose a

2 problem in about 10 percent of the cases. In a

3 large percent of the cases, it works very well. I

4 just wanted to make that perspective is also.

5 DR. MORRIS: Can I just add to that in a

6 sense, and somebody had said it earlier, and I

7 apologize for not remembering who, but the idea

8 that you use, whether it is PAT, but particularly

9 for PAT-type approaches, during development is when

10 hopefully you find the sort of systems that you are

11 talking about with PAT, so that when you get it to

12 the floor, you are in the situation where it is not

13 one of the 10 percent. Otherwise, you just are

14 tracking, so you say, yeah, here is where it goes

15 back, but if you can do that every time, it may be

16 accurate, but it doesn't -- what is that?

17 DR. KIBBE: If you can do that every time,

18 you ought to fix it.

19 DR. MORRIS: Exactly, exactly.

20 DR. KIBBE: That's true. I just was being

21 a troublemaker.

22 DR. LEE: So the answer to the first

23 question is okay, looks okay, and the agency is

24 going to do these statistical validation scrutiny,

25 evaluation.


1 DR. HUSSAIN: The question is I think the

2 preliminary discussion we have had with the

3 Statistical Group essentially, I think I have to go

4 back and carefully read some of this again, but the

5 concern that was raised was the hypothesis at

6 least, the claim of the hypothesis testing that

7 blend uniformity is not a value-added test or

8 predictive, I don't think the initial analysis that

9 this approach can prove or disprove that

10 hypothesis.

11 So that is not the question, and that is

12 reason I raised the question, does that really have

13 to be part of the recommendation at all, because is

14 that necessary for moving forward. That was not

15 the purpose of the recommendation.

16 The point I am making here is essentially

17 in the sense, when you go stratified sampling, you

18 are focusing more on the end product which is given

19 to the patients although I don't like end product

20 testing for the sake of testing, and that goes

21 totally opposite to the PAT, but again we have two

22 systems to consider.

23 So regardless of the old system, testing

24 to document quality, let's keep that system, is

25 when you do blend uniformity, clearly, you have


1 other processes that occur post-blending -

2 discharge, flow, and so forth, that can induce

3 segregation.

4 From that perspective, focusing on

5 uniformity does provide relevance from a clinical

6 perspective. It is more relevant from that

7 perspective.

8 DR. LEE: Let's move on to the second

9 question.

10 DR. DeLUCA: This was should the proposed

11 stratified sampling analysis be applicable only for

12 the bioequivalence batch and validation batches, is

13 that right?

14 DR. LEE: Yes.

15 DR. DeLUCA: Okay, because there were a

16 couple of versions here of questions.

17 The recommendation, as I saw it, is that

18 certainly blend uniformity testing is very valuable

19 during the development stage during validation,

20 troubleshooting, and certainly should be continued

21 in those, and the ANDA exhibit batches, but it

22 should be applicable then to not just generic

23 batches, as well. It should be across the board, I

24 think, for ANDA products or not. Wasn't that the

25 intent of the recommendation?


1 DR. MASSA: I think you are addressing the

2 last question, and clearly we said that it ought to

3 apply across the board, it shouldn't just be

4 related to ANDA products.

5 I think what Question 2 is trying to

6 assess is where specifically are we using this

7 stratified randomization. I think what we are

8 trying to say is that it is used in concert with

9 blend uniformity as you are doing development and

10 validation, but once you have established that you

11 are validated, that you do away with blend

12 uniformity, and are only using the stratified

13 sampling on your routine batch release.

14 DR. HOLLENBECK: Can we just clarify that

15 for sure, that you are proposing that you will use

16 the stratified testing on a regular basis for

17 routine release of product?

18 DR. MASSA: Yes. What we are saying is

19 the stratified sampling is to be used for routine

20 release and that the only time you would use blend

21 uniformity analysis is during development and

22 validation, and for troubleshooting should you run

23 into an issue later on.

24 DR. HOLLENBECK: Well, then, to me, that

25 answers your question, Ajaz. That is how adequacy


1 of mix will be ensured doing routine production.

2 DR. VENITZ: It answers the subquestion to

3 your previous one in terms of that the dosage form

4 is more important than the blend uniformity

5 regardless of the datamining.

6 MR. BUHAY: I would just like to comment

7 that I would not want to over-endorse that concept,

8 that if you go to that concept, in-process testing

9 becomes everything, and there is no role at all for

10 in-process testing.

11 DR. MORRIS: I think it makes Art's

12 argument a lot stronger in terms of validating or

13 verifying that the Monte Carlo simulations are good

14 at all scales. If you are really going to do that,

15 I don't even know the algorithms that we use, but

16 it just puts a lot more pressure on that being

17 true.

18 DR. KIBBE: I am comfortable with the

19 two-step process that I think I hear happening,

20 which is once we have got a process that we have

21 run a few times and we know that we have got

22 consistency, and we have a known problem with the

23 way we sample blend now in some cases, but if they

24 have got a product that has been run a few times,

25 and they have done blend analysis, and they have


1 done tablet analysis, and they seem to be

2 consistent, and the process is holding, producing

3 good batches on a regular basis, then, why do a

4 test that is, at best, redundant?

5 It is a diagnostic test if there is a

6 problem with a tablet, because the tablet is what

7 you have to give to the patient. Now, if the test

8 is truly predictive in that you would do the test

9 and then kill the process completely because that

10 is where your problem is, then, you have to keep it

11 in there. In my estimation, you have got a bad

12 manufacturing process then.

13 DR. HUSSAIN: Art, I think related to your

14 comment right now, I think the question then

15 becomes using stratified sampling, doing the USP 10

16 tablets --

17 DR. KIBBE: That's my real problem.

18 DR. HUSSAIN: That's the question.

19 DR. MORRIS: But there is another exposure

20 issue in the sense that if you have to find an

21 assignable cause, if you have a failure and you are

22 only testing tablets, I mean that is a risk that

23 the companies have to accept then because you are

24 not going to be able track back and find an

25 assignable cause if it happened in the blender, for


1 example. That is a justifiable risk.

2 DR. HUSSAIN: One of the reasons for

3 raising the subquestion, the issue of "readily" and

4 "marginally" comply, the reason I raised that

5 question is I think when we had some internal

6 discussion, two issues had come up. I just want to

7 point out those issues to you.

8 One is many of these decisions are being

9 made on a sample RSD or standard deviation, and its

10 ability to reflect with confidence the population

11 RSD depends on the sample size, and so forth. So,

12 there is that uncertainty. So the population,

13 standard deviation could be smaller or larger, so

14 there is uncertainty there. So that was sort of

15 one concern that was raised internally.

16 The second question was I think in terms

17 of you are classifying "readily" and "marginally"

18 on the basis of those numbers with certain

19 uncertainty, and when you get into routine

20 production, the materials that you have, you have

21 different lots of excipients, and so forth, and

22 there is no sort of linkage between that

23 variability to what the future variability might

24 be.

25 So those are the sort of underlying


1 concerns that we had discussed internally.

2 DR. DeLUCA: I would like to also share

3 what Ajaz was saying about the way the

4 recommendation was worded. Certainly, I don't feel

5 that the blend uniformity is not a good predictor,

6 I mean it certainly is not a good predictor of the

7 end product, the quality, as doing the dosage

8 units.

9 But it seems in the validation procedure,

10 you are going to be correlating the blend

11 uniformity with the end product, the dosage units,

12 so as long as that is done and followed, it seems

13 then one could rely then on the dosage units as the

14 end test rather than doing blend uniformity.

15 It seems to me that, one, we shouldn't

16 overlook the importance of putting the blend

17 uniformity testing into the PAT, but that would not

18 be a sampling of the blend, but it would be an

19 in-process type of test with some sensor.

20 I think the problem with the blend

21 uniformity is the sampling itself and any

22 segregation that occurs after the sample is taken

23 and analyzed that's the problem here.

24 DR. LEE: Thank you.

25 Judy.


1 DR. BOEHLERT: The thought just occurs to

2 me in looking at this RSD of 4 percent, to what

3 extent did you consider the precision of the

4 analytical method when you are looking at an RSD of

5 4 percent? Some analytical methods are better than

6 others. That might not give you much room for

7 error in your product if the method is variable.

8 DR. GARCIA: In one of the OC curves in

9 the recommendation, I believe we included a 1.5

10 percent analytical error, so we did play around

11 with that a little bit, but that is something else

12 that the company has to address. Your analytical

13 method has a wide variability, that's the price you

14 pay.

15 DR. BOEHLERT: And that might be an area

16 where the FDA will get comments when this gets into

17 revised guidance when people begin to take a look

18 at this and say, well, wait a minute, my precision

19 is 2 percent or 2.5 or 3, and what does that mean

20 therefore, you know, I am never going to pass that.

21 DR. LEE: Other members of the committee

22 wish to express an opinion?

23 DR. HUSSAIN: Everything has been said,

24 Gloria said.

25 DR. LEE: Okay. Pat, will you please for


1 the record the answer to the three questions?

2 DR. DeLUCA: Well, do you consider the

3 PQRI proposal appropriate for inclusion in a

4 planned revised FDA guidance? Yes.

5 If yes, should the proposed stratified

6 sampling analysis be applicable only for

7 bioequivalence and validation, the answer is no on

8 that.

9 What else do we have? Then, should it be

10 applied to -- I think there was a third question

11 with regards to go beyond the generic products, and

12 the answer to that was yes.

13 DR. LEE: Okay.

14 DR. HUSSAIN: Just to clarify, the answer

15 to the second question you said was no. In

16 essence, the proposal is that for a low variable

17 product, which readily complies, we will sunset

18 that, and for one which has a bit higher

19 variability, you do additional testing for at least

20 five, and then based on that, decide. So, could he

21 just clarify that?

22 DR. DeLUCA: I guess maybe this is worded

23 in a way that -- I think what we are saying here

24 that we are going to use the blend uniformity

25 testing for validation ANDA exhibit batches to


1 bioequivalence batches and troubleshooting. I

2 guess that is what we talked about here, that blend

3 uniformity testing would apply to that.

4 DR. HUSSAIN: In essence, the committee

5 feels that the PQRI recommendation to sunset for

6 routine production, blend sample collection and

7 analysis, and rely on stratified sampling is

8 acceptable.

9 DR. LEE: John?

10 DR. DOULL: Sounds good to me.

11 DR. LEE: Let me say that as a scientist,

12 I would like to see this work be put through peer

13 review, and I think this is a very feasibility

14 study. I don't think that we tried to address all

15 these situations that might encounter, but I think

16 we have a good starting point and let's see how

17 things evolve.

18 DR. HUSSAIN: To sort of summarize the

19 next steps from FDA perspective, I think we will

20 take your discussion and recommendations, and as I

21 said, we have a separate group, we will work on

22 that, and I think, in my mind, I think we have to

23 have peer review, and peer review that will occur

24 through the internal process and the statistical

25 evaluation.


1 Then, I think as we go through that

2 process, we will provide information to the PQRI

3 with respect to all concerns if there are, and then

4 we will start working on the draft guidance, so

5 that is essentially the steps we will have.

6 DR. LEE: For the record, we do have an

7 expert in statistics on the committee, but he

8 happened not to be able to be here, so there is no

9 redundancy in this committee.

10 DR. KIBBE: We are going to get some

11 statistical understanding of whether we need to

12 have a different number of tablets for different

13 batch sizes, right?

14 DR. HUSSAIN: We will have our

15 statisticians, we will work with our statistician

16 to focus on that question as we go through our

17 review.

18 Also, as I had mentioned this morning, I

19 think the draft guidance that will come out, we

20 probably will include a section on on-line

21 planning. Again, those are two different issues,

22 the current system and the future systems, because

23 I did not overlap the PAT with this discussion.

24 Those are two separate issues.

25 So we will go back and proceed in that


1 fashion.

2 DR. LEE: Okay. Leon.

3 DR. SHARGEL: I just wanted to comment on

4 the last question. Just as a general concept, my

5 own feeling is that for the NDA and the ANDA-CMC

6 Section, they both should have the same quality

7 standards, so as we move forward on this issue, as

8 well as other areas, I was quite concerned when the

9 guidance came out and said blend uniformity for

10 ANDAs.

11 I think in general terms, we should look

12 at quality standards for both general, for both

13 ANDAs and NDA.

14 DR. LEE: We are kind of on schedule, so

15 let's take a 10-minute break.

16 [Break.]

17 DR. LEE: First of all, I have to watch my

18 time, as well, because I was told that my cab would

19 be here at 4:15. I thought I had about 5:15 to

20 enjoy the facilities. Therefore, I am going to use

21 my electronic gavel for the first time.

22 The last session is on polymorphism/

23 We have Dr. Chiu to introduce the topic,

24 and we have two presentations. This is an

25 awareness session.


1 Regulatory Issues Related to Crystal Habits -

2 Polymorphism

3 Introduction and Overview

4 Yuan-yuan Chiu, Ph.D.

5 DR. CHIU: Good afternoon. We are going

6 to introduce a new topic - polymorphism to the

7 committee. Today, we would like the committee to

8 make a general discussion. We are not going to

9 have, you know, issues on specific products. We

10 are not having any questions for the committee to

11 address. However, I would want to bring this topic

12 to the committee in preparation of future

13 questions.

14 We do plan to come back to the committee

15 on this topic with specific questions after the

16 agency does a little more work, because we are in

17 the process to draft a guidance document to address

18 a lot of challenges, regulatory and also legal

19 challenges for generic drugs.

20 [Slide.

21 As you can see, polymorphism has a

22 scientific definition. The scientific definition

23 is the ability of a substance to exist in two or

24 more crystalline forms that differ in the

25 arrangement of the molecules and/or confirmation of


1 the molecules.

2 So this is what we are talking about is

3 the crystallines as habits or crystalline lattice

4 of a substance. However, there is a broad

5 definition under ICH Q6A, which is the certain

6 specification for chemical substances.

7 This definition then actually encompass

8 solid state forms in addition to crystalline forms.

9 So it is stated the occurrence of different

10 crystalline forms of the same drug substance, and

11 this may include solvation or hydration products,

12 also known as pseudopolymorphs, and amorphous

13 forms, because amorphous forms are not really

14 crystalline lattice.

15 [Slide.

16 The polymorphism is very important to

17 final product quality. I am going to give you a

18 few examples to illustrate the different physical

19 properties exhibited by different polymorphs.

20 The packaging properties of the substance

21 can be different, which includes molecular volume,

22 density, refractive index, hygroscopicity, and the

23 conductivity.

24 The thermodynamic properties can also be

25 different. They can include melting point,


1 solubility, heat capacity, potential energy.

2 The third one would be the kinetic

3 properties of the crystalline forms. Different

4 crystalline form may have different dissolution

5 rate, and the rates are solid-state reactions, as

6 well as the stability.

7 [Slide.

8 In addition, the surface properties of

9 polymorphs can be different, the shape of the

10 crystals, and the surface free energy, and

11 interfacial tensions.

12 The last, and not the least, is the

13 mechanical properties, which includes hardness,

14 tensile strength, compactibility, which is the

15 tableting property, as well as handling of the

16 substance including flow and blending.

17 [Slide.

18 Because of the differences of those

19 properties will affect the product qualities,

20 therefore, for the patients we usually, you know,

21 have information on the polymorphism of the

22 substances.

23 So today we are going to have Steve Miller

24 to present to you how we handle the information for

25 new drugs. Then, Richard Adams will talk about the


1 regulatory and legal challenges of generic drugs.

2 Originally, we planned to invite Dr.

3 Brittain to present experts' views, but he could

4 not attend, so therefore, we will have two talks.

5 Without further ado, I ask Steve to come

6 to give his presentation.

7 ONDC Issues

8 Steve Miller, Ph.D.

9 DR. MILLER: Thank you. Good afternoon.

10 [Slide.

11 I am the team leader in the Antiviral Drug

12 Division of CDER. I will be presenting an overview

13 of polymorphism, the regulatory aspects of

14 polymorphism from the perspective of the new drugs.

15 Most of what I talk about today will be an

16 overview of a workshop that was organized by the

17 American Association of Pharmaceutical Sciences two

18 months ago. I will talk about that workshop a

19 little specifically in a moment.

20 [Slide.

21 When the 1987 Drug Substance Guideline was

22 written, it was recognized that for some drug

23 substances and for some drugs, the solid-state form

24 of the drug substance or active ingredient can have

25 an important effect on drug release and on


1 bioavailability.

2 As a consequence, during the development

3 of most solid oral dosage forms, suspensions, many

4 other drugs, there will be a dialogue during the

5 IND phase between the review chemist and the

6 pharmaceutical sponsor to discuss what is known

7 about the solid-state form of this particular drug

8 and what impact it may have on dosage form

9 development.

10 [Slide.

11 A second guidance that deals with

12 solid-state form is the BACPAC I Guidance, which

13 was issued more recently. This is the Bulk Actives

14 Postapproval Changes Guidance, which is used when a

15 change is made to an approved product in the

16 manufacturing steps of a synthetic compound, a

17 synthetic drug substance.

18 Essentially, the BACPAC I Guidance is a

19 change control protocol that outlines for many

20 different changes, for example, change to a new

21 solvent in some portion of the synthesis, or change

22 in the synthesis itself with new intermediates, for

23 those changes, what data would be expected to

24 demonstrate equivalence post-change to pre-change.

25 It is recognized that when it is the drug


1 substance that is being measured to look at

2 equivalence, the two important solid-state

3 parameters are polymorphism in its broadest sense

4 and particle size.

5 Here, you will note in this guidance the

6 definition of morphic form is the broad definition

7 that Yuan-yuan mentioned, which is also in Q6A. It

8 encompasses hydrates, solvates, and well as

9 non-crystalline amorphous forms.

10 [Slide.

11 Now, the guidance that covers polymorphism

12 in greatest detail is the ICH Q6A Guidance

13 developed under the International Conference of

14 Harmonization, and it provides guidance to U.S.,

15 European, and Japanese regulators and

16 pharmaceutical sponsors regarding setting

17 specifications for the active ingredients and for

18 the dosage forms.

19 By setting specifications, we are

20 referring to the tests that will be performed on

21 each batch. For example, a drug substance, each

22 batch before the drug substance is used to make

23 drug product, must be tested according to the

24 specification, and the Q6A guidance outlines how we

25 will go through and determine which tests are


1 appropriate, and when a test is determined to be

2 appropriate for the specification, what the

3 acceptance criteria should be, acceptance criteria

4 formerly widely called the limits, the numerical

5 limits, frequently numerical.

6 The same set of questions occur for the

7 dosage form, for the drug product, and Q6A

8 encompasses both forms.

9 [Slide.

10 I wanted to show some typical

11 specifications or some of the attributes that might

12 be part of typical specifications. You will see

13 that in some cases, there are what are essentially

14 universal tests that would be present in all cases,

15 identity and assay would be typical.

16 Then, there are other tests that may be

17 appropriate for a particular drug. Morphic form

18 tested in the drug substance, and more rarely in

19 the drug product, is one of the optional attributes

20 that could be appropriate for some drugs, may not

21 be necessary for other drugs.

22 The Q6A guidance goes into quite a bit of

23 detail about morphic form and polymorphism in the

24 form of some decision trees.

25 [Slide.


1 The decision trees themselves are broken

2 up into three parts for polymorphism. The first

3 part deals with whether multiple polymorphic forms

4 exist. The second part deals with whether a

5 regulatory specification in the drug substance is

6 appropriate. The third part of this decision tree

7 relates to whether a regulatory specification in

8 the drug product is appropriate.

9 Now, even with this amount of detail,

10 there is still quite a bit of gray areas, and the

11 recent AAPS Workshop was an attempt to formulate

12 unified understandings of the Q6A document as a

13 whole.

14 I was one of the four breakout moderators

15 for a session that dealt with solid-state forms.

16 [Slide.

17 I will now go into the three decision

18 trees that are part of Q6A. I will try to focus on

19 the questions that we brought forward at the AAPS

20 Workshop and the conclusions that came back from

21 the discussion groups that formed to discuss these

22 topics.

23 The first part of the decision tree

24 relates to whether there are multiple polymorphic

25 forms possible. Essentially, it instructs that


1 screening should be done in all cases to look for

2 the possibility of multiple polymorphic forms.

3 If there is an absence of any evidence of

4 multiple solid-state forms, then, essentially, you

5 move to the first NO, and that's the end of the

6 process.

7 In many cases, there will be several

8 morphic forms available, and in that case, the

9 latter part of this flow chart shows that those

10 forms should be characterized. It outlines some

11 representative tests that could be performed, some

12 X-ray techniques very commonly applied.

13 The second set relates to melting point or

14 DSC, differential scanning calorimetry, some

15 thermal methods that can be used. There are also

16 spectral methods that frequently can assess whether

17 a particular morphic form is present or measure its

18 levels in many cases.

19 Infrared spectroscopy in many cases may be

20 appropriate to differentiate between known

21 polymorphs, solid-state and MARS, another emerging

22 technology.

23 When we brought this forward to the AAPS

24 Workshop, one main question was what is a

25 reasonable polymorph screen for different


1 situations, different drug substances, different

2 dosage forms.

3 On the next slide, I will show some of the

4 results that came out of our discussions.

5 [Slide.

6 I would have to say that there was a

7 diversity of opinion among the participants. Some

8 people felt that really the screen should be

9 limited, should focus on the solvents that are used

10 in the drug substance manufacturing process.

11 Other participants felt that there was

12 value to be gained in exploring more broadly many

13 different solvents of different polarities,

14 hydrogen bonding propensities.

15 I would say there was quite a universal

16 agreement that when a particular condition could be

17 relevant to the drug product manufacturing

18 situation, for example, if the drug product is

19 formulated through a wet granulation process, then

20 aqueous solvents should be explored in the

21 screening for polymorphic forms because clearly,

22 formation of a hydrate would be something that

23 could occur during the drug product formulation.

24 One general concept was that reviewers at

25 FDA generally want to see some evidence of what was


1 carried out during the polymorphic screen in the

2 application.

3 After the second bullet, you will note

4 that there is a note GRP document. I have tried to

5 mark up on a number of these slides where we, in

6 CDER, hope to record some of these recommendations

7 and hopefully, some future recommendations, as

8 well, about polymorphism.

9 We are in the process of formulating a

10 good review practices document for the chemistry

11 manufacturing and controls aspect of a drug

12 application. We feel that some of this could be

13 incorporated into that document.

14 The third bullet relates to a separate

15 topic, which deals with solution drug products,

16 where in many situations, there is going to be very

17 few issues about polymorphism related to a true

18 solution, however, it is noted that some

19 information about the concentration of the drug in

20 the drug product solution relative to the limiting

21 solubility, the solubility of the least soluble

22 form.

23 That can be very valuable. It would

24 essentially tell you what the likelihood that the

25 drug substance could crystallize out of the dosage


1 form. If the drug load is very low relative to

2 saturation, there is essentially no concern there.

3 If it is very close to saturation, there could be a

4 concern.

5 A subpoint under that was the idea that in

6 the situation where you are very close to

7 saturation, it may be relevant to include the

8 dosage form vehicle in the screen for polymorphs to

9 ensure that the form that would crystallize out of

10 the dosage form is a form you have already

11 discovered through other processes, and is not a

12 new form. That is a point that we will try to

13 incorporate into the next drug product guidance,

14 which is currently in internal draft stage.

15 [Slide.

16 Now, I am going to go on to the second

17 part of the Q6A decision tree, which dealt with

18 whether an acceptance criteria is appropriate in

19 the drug substance.

20 The first decision diamond relates to

21 whether there are different properties of the

22 polymorphs that are known. The second diamonds

23 says if there are different properties, are those

24 properties likely to affect performance of the drug

25 product.


1 If the answer to either of those is no,

2 the properties are not different, it is unlikely to

3 affect the dosage form, then, essentially, it is

4 the end of story, and the Q6A recommends that no

5 acceptance criteria would be established in the

6 drug substance.

7 If however, yes is the answer to both of

8 those, an acceptance criteria is thought to be

9 appropriate for the drug substance.

10 Now, in terms of the discussion at the

11 AAPS Workshop, there were several points -- I hope

12 this will show up on the slide -- in the first

13 decision tree it says, "Are there different

14 properties," and a main question was, "What is

15 meant there, how different is different?"

16 It was felt that some case studies could

17 be very valuable to try to say how is this being

18 interpreted both in the industrial side and in the

19 review side.

20 [Slide.

21 A second question related to the second

22 decision point, and it related to the question of

23 when would you make the determination that even

24 though there are different properties, the dosage

25 form performance would not be affected.


1 It was thought that in many cases,

2 solution drug products would fall into this

3 category, but there was some discussion as to

4 whether there would be other categories, as well,

5 and I think that is another area where clearly,

6 some additional discussion could be very useful.

7 [Slide.

8 Finally, a third question regarding this

9 particular part related to the final diamond. When

10 you have decided that an acceptance criteria is

11 appropriate for measuring polymorphic form in the

12 drug substance, are you thinking in terms of

13 quantitative control, quantitating the amount of

14 the two polymorphic forms, or would there be cases

15 where a qualitative control, verification of

16 identity would be what was needed.

17 [Slide.

18 With regard to that point, we have some

19 clear decisions brought back from the workshop, a

20 qualitative test was thought to be appropriate when

21 the drug substance manufacturing process controls

22 which polymorphic form is produced, and you have

23 demonstration that morphic form is controlled

24 adequately by the drug substance process.

25 In that case, a qualitative, for example,


1 an identity test, which could be a melting point or

2 an IR test if those have been validated to be able

3 to show the morphic form could be all that is

4 needed in the drug substance specification.

5 There was also discussion about sunsetting

6 these tests when enough manufacturing experience

7 was gained through a certain number of batches or a

8 certain number of years, eliminating this

9 qualitative control.

10 There was also discussion of whether skip

11 testing, testing only one in a certain number of

12 drug substance batches would be appropriate.

13 It was also widely recognized that there

14 are many, much more complicated situations that

15 would need much more control and usually much more

16 information to tell you what type of control is

17 necessary. In many cases, you may need to know how

18 the dosage form will perform using different

19 morphic forms. That is elaborated a little further

20 on the next slide.

21 [Slide.

22 These complicated situations could involve

23 situations where the drug substance polymorph is

24 changed during the manufacturing process or on

25 stability, where the drug product contains multiple


1 polymorphs and it is known that they do affect

2 bioavailability, or where there is a significant

3 amount of amorphous drug substance intentionally

4 introduced in the drug substance.

5 It was widely recognized that it is

6 important to discuss these issues at the end of

7 Phase II meeting for new drugs, which would

8 typically occur a year or more before the NDA is

9 filed in order to have time to plan what studies

10 are needed, what type of documentation will be

11 available to make these decisions at the time of

12 the review.

13 [Slide.

14 I am now going to go on to the last part

15 of the flow chart, which unfortunately is the least

16 readable of the sections. This is the section that

17 deals with whether a control in the drug product

18 specification is appropriate.

19 I want to first show a couple of caveats

20 that are brought forth in the text of the Q6A

21 document. The first says that you would really

22 only be dealing with this part of the flow chart

23 when you have already established that there are

24 multiple polymorphic forms of the drug substance

25 and that they are likely to affect the performance


1 of the drug product.

2 Another caveat states that it is

3 frequently very difficult to measure a polymorphic

4 form in the dosage form, and if it is possible to

5 obtain equivalent quality control through other

6 testing, for example, dissolution testing, that

7 that is the preferred approach, and testing of

8 morphic form in the dosage form is more or less a

9 method of last resort if that does not apply.

10 The first decision diamond really relates

11 to this latter point. It asks whether other

12 performance tests, for example, dissolution

13 testing, provide adequate control when the morphic

14 form ratio changes, and if dissolution testing or

15 another performance criteria provide adequate

16 control, then, there would generally not be a need

17 to look at the polymorphic form in the drug

18 product.

19 However, if it is determined that no

20 performance testing of the drug product will serve

21 that role, then, the decision tree recommends

22 developing a method to look at morphic form in the

23 drug product and measuring it on stability.

24 The second decision diamond, halfway down,

25 says, "Has there been a change in the morphic form,


1 and if there has been a change, it instructs then

2 that in that situation, a control in the drug

3 product specification is appropriate.

4 [Slide.

5 Now, there is one specific question that

6 we brought forward in the AAPS Workshop, and that

7 related to the word "change" in this second

8 decision diamond, because it is written in the

9 context of stability testing, and it implies that

10 the change would be on stability, but the question

11 that was discussed was does this change also

12 encompass changes during manufacture of the drug

13 product.

14 I felt that we had a clear decision from

15 the AAPS discussions that yes, this change would

16 encompass change during manufacture of the drug

17 product or change on stability.

18 It is kind of a complicated situation, but

19 an example might be if you had a wet granulation

20 solid oral dosage form, you knew that there was a

21 change to the hydrate during the wet granulation,

22 and you also knew that the ratio of the hydrate to

23 the anhydrous form was critical to performance, was

24 critical, say, to bioavailability, but

25 unfortunately, it was not possible to develop a


1 dissolution test that would be sensitive to that

2 effect.

3 In that case, the decision tree would say

4 there has been a change during drug product

5 manufacture, and you would measure the amount of

6 the hydrate and the anhydrous form in the drug

7 product even if there was no change on stability.

8 That ends the overview of the decision

9 trees. I have one final slide in which I have

10 tried to sum up how the workshop ended.

11 [Slide.

12 I think there was a general feeling that

13 we could benefit from some additional discussion,

14 possibly an industry-FDA discussion group to put

15 together some concrete examples, some case studies,

16 and to try to bring forward some of the common

17 understandings then of the polymorphism approach in

18 Q6A.

19 I think that in particular, we would

20 benefit from some additional work related to when

21 dosage form should intentionally be made with

22 different polymorphic forms of the drug substance.

23 I think in both cases, those could perhaps be

24 incorporated into our ongoing good review practices

25 document.


1 I would like to summarize by saying I

2 think we made some significant progress in terms of

3 establishing some common understandings about

4 polymorphism in specification setting at the AAPS

5 Workshop. I think there is an opportunity to make

6 even more progress if we continue the dialogue, get

7 some additional feedback.

8 I would like to thank the three other

9 workshop organizers with me - Ivan Santos from

10 Merck, Tim Wozniak from Eli Lilly, and John Clark

11 from CDER, and I would like to thank you all for

12 allowing me to make this presentation today.

13 DR. LEE: Thank you very much.

14 Are there any burning questions for Dr.

15 Miller?

16 Yes, Ken.

17 DR. MORRIS: Yes. I have more of a

18 comment. In the screening for polymorphs, there is

19 a big gap there that deals with purity, because in

20 my experience, the biggest variable as you go from

21 the bench to the kilo lab to full scale is the

22 purity of the material you are getting. That will

23 bite you in the rear end faster than most of the

24 other things that are on that list.

25 DR. MILLER: That is a point actually


1 which was not discussed during the workshop.

2 DR. MORRIS: That may be the biggest point

3 really. I mean it is not the biggest point in terms

4 of finding them necessarily when you are screening,

5 although it may be there, as well, but in terms of

6 what changes as the chemists get more and more

7 refined processes and they hopefully increase their

8 yield in purity, they change that a lot, and

9 impurities, and for years and years used to control

10 all sorts of things indirectly by just people

11 skilled in the art at large scale. So I would say

12 that that is something that really has to be

13 included.

14 DR. MILLER: Very good, and I think that

15 was one point that was brought up during the

16 discussions was that the relevance of the data to

17 the whole life of the product is very important,

18 and I think that aspect would play into it very

19 much.

20 DR. MORRIS: Absolutely, and it is easy

21 enough to do screens, more focused screens. I mean

22 you can't put full resources against it every week,

23 but, you know, to do a more focused screen as you

24 go up, particularly if you know you have forms of

25 significantly different free energy because that is


1 what you really care about.

2 DR. CHIU: I think we recognize that.

3 That is why in the BACPAC, you make manufacturing

4 changes, it is important that you reestablish

5 whether you have the same polymorphic form, and

6 especially when the impurities change, you have

7 the different new impurities, then, you may

8 influence that because everything is based on the

9 crystalline seeds. You have a different seed, you

10 get different forms.

11 Thanks for pointing it out.

12 DR. RODRIGUEZ-HORNEDO: As I studied the

13 guidance and listened to your very good

14 presentation, it strikes me that we are focusing on

15 detection of polymorphs and that there is very

16 little attention, at least it seems to me, to the

17 kinetic events during screening and

18 transformations, and only time-dependent phenomena

19 I see is too late from my perspective to take

20 action.

21 It is either during dissolution, which is

22 very relevant, or it's also during stability

23 testing. So I think the solvents are very

24 important, and it is one of the variables that

25 determines the appearance of a polymorph, but doing


1 kinetic studies, they don't need to be very

2 rigorous, but they would be indicating what are the

3 levels of risk of a new form appearing during the

4 parameters of relevance of a given process.

5 I see that lacking. If you identify three

6 polymorphs or two polymorphs to me, and I have to

7 work with them, I would like to know what are the

8 rates, what are the time frames for transformations

9 all the way from the process, you know, all the way

10 from screens, and it is not only solvent.

11 I think this may be one of the unit

12 operations that is very applicable for PAT to

13 consider. I would like your comments on the time

14 events.

15 DR. MILLER: I would be interested in

16 hearing your thoughts. Are you thinking more

17 towards solid-state kinetics, transformation in the

18 pure drug substance or the dosage form, or are you

19 thinking of as kinetic experiments performed, sort

20 of ripening experiments in solvents?

21 DR. RODRIGUEZ-HORNEDO: I was thinking

22 more of experiments in solvents, so they could be

23 in suspensions, I think the induction times, the

24 presence of one of the polymorphs, how quickly does

25 it transform to the other polymorphs during the


1 conditions that are relevant to the system.

2 I also think that the dissolution is very

3 important. We think of water, for instance, when

4 we are designing on what granulation process in

5 which the solvent is water, but ultimately, all

6 these dosage forms are going to heat water, they

7 are going to be in the water environment, so how

8 quickly does the transformation occur in water, and

9 are our dissolution methods sensitive enough to

10 capture that.

11 We may be considering developing

12 dissolution methods to identify or classify drugs

13 that may be high risk for what we may actually

14 initially perceive as erratic dissolution behavior,

15 but the transformation is occurring so quickly,

16 that miss it. So we may need to be a little bit

17 more creative in designing some dissolution methods

18 that are sensitive to these for screening purposes.

19 In the solid-state, I think it is very

20 important. I think we are already doing some of

21 that. I think that industry is doing some of that,

22 and the agency has focused a lot of solid-state

23 transformations, but I think the screening in

24 solvents needs to consider the kinetic events, as

25 well as the dissolution, because I think that is


1 where with a product or the actual substance, may

2 be very revealing.

3 DR. MILLER: I think the challenge will be

4 in creativity to determine the right type of

5 experiments for each particular case.

6 DR. LEE: Thank you.

7 DR. CHIU: I think that is a very

8 important, you know, good point, you know, the

9 kinetics. We are working, drafting a drug product

10 guidance in line up with the CCDQ. We could

11 consider to put that concept in the Pharmaceutical

12 Development section.

13 DR. LEE: I hope that you include dosage

14 forms beyond tablets and capsules.

15 DR. CHIU: Yes, we will do that.

16 DR. MILLER: I would have one more point

17 to add regarding kinetics. At the workshop, one

18 point was made that one reason for possibly doing a

19 broadly-based screen initially, looking at solvents

20 that might not really be used in the current

21 process, might be that kinetically, a particular

22 polymorph might readily form in one of those

23 solvents.

24 It might be the thermodynamically, more

25 stable polymorph in general, but you might have


1 more difficulty getting to it from the smaller

2 screen.

3 DR. LEE: Gloria.

4 DR. ANDERSON: My question is much

5 simpler, but before I ask it, I will say that in

6 terms of the crystal forms that you get, they are

7 certainly solvent-dependent. But my question goes

8 back to the beginning.

9 Why do you include amorphous substances in

10 your definition or polymorphs, when, by definition,

11 your definition, they are crystalline forms?

12 DR. MILLER: I think the next speaker, Dr.

13 Adams, will talk a little bit more on that point,

14 but I would say that because of the important

15 relevance in drug product manufacture, where

16 amorphous forms may offer very different properties

17 to the dosage form.

18 DR. ANDERSON: Absolutely.

19 DR. MILLER: Our intent is to bring it in

20 under that terminology. I think if we wanted to be

21 as correct as possible with current terminology, we

22 would probably just talk about solid-state forms,

23 which would then include hydrates, solvates, as

24 well as non-crystalline amorphous forms.

25 DR. ANDERSON: You can have crystalline


1 hydrates, but you can't have amorphous crystalline.

2 DR. MILLER: Many of you would not include

3 hydrates under the term "polymorph," because the

4 unit cell has different contents.

5 DR. MORRIS: If I can just comment,

6 Gloria, I think part of the reason is there is just

7 no other place to put it. I mean it's a condensed

8 phase, it is not a liquid, I mean it's a solid. I

9 just don't know where else you would put it.

10 DR. ANDERSON: In a separate category.

11 DR. MORRIS: Well, in a separate category,

12 but I mean in terms of this workshop.

13 DR. RODRIGUEZ-HORNEDO: One last comment.

14 You may be aware that there are some companies here

15 in this country, more of them in England, but there

16 are some in this country that have very closed

17 monitoring of their crystallization process by

18 Raman spectroscopy and other spectroscopic

19 techniques, so like what we have been discussing

20 this morning, it is very applicable to this.

21 DR. LEE: I can see that it is a very

22 interesting topic. I hate to cut it off, but in

23 order to give justice, in fairness, not justice, to

24 Dr. Adams.

25 OGD Issues


1 Richard Adams, Ph.D.

2 DR. ADAMS: Good afternoon. I would like

3 to talk about the issue of morphic form or I guess

4 we can call it solid-state or morphic form and its

5 issues that are particularly important in generic

6 drugs.

7 [Slide.

8 We have encountered active pharmaceutical

9 ingredients of different morphic form in the past,

10 but recently, the interest in it is intensified

11 because we have gotten a number of applications

12 which have been contested by the innovator

13 companies as inappropriately 505(j) because they

14 are different morphic form, different polymorphs,

15 and the innovator companies have objected.

16 We have gotten numerous Citizen's

17 Petitions and it has been extremely

18 resource-intensive in the Office of Generic Drugs.

19 Yuan-yuan correctly pointed out, when we were

20 speaking about this meeting, that amorphous forms

21 shouldn't be included, and that is certainly

22 correct.

23 The fact of the matter is that there are

24 many amorphous drug substances for reasons related

25 to the patent litigations and also because, in some


1 cases, crystalline drug substances are deliberately

2 converted to amorphous as in the case of

3 lyophilizations to increase the dissolution rates

4 and for other reasons.

5 [Slide.

6 In any event, we have gotten very familiar

7 with the issues. I would say that some of the

8 issues in generic drugs that are peculiar to the

9 generic review process have to do with the lack of

10 transparency often between the Drug Master File,

11 the supplier or the vendor of the active

12 pharmaceutical ingredient and the actual drug

13 product manufacturer, so that usually, I think the

14 relationship is different for the innovator

15 companies.

16 On the plus side, usually, there is a fair

17 amount of information known in the literature by

18 the time we get these applications, so in terms of

19 the screening for polymorphs, we certainly require

20 it, but usually, you can simply do a quick search

21 and you can find out a fair amount of information

22 about the morphic forms of the active

23 pharmaceutical ingredient.

24 [Slide.

25 In terms of our approach to it, the


1 regulations and the codifications of them are

2 fairly straightforward. The drug substance, the

3 active ingredient must be identical to the

4 innovator active ingredient, and in addition to

5 that, it must be the same dosage form, strength,

6 route of administration, and the conditions of use

7 must be identical.

8 If those conditions are met, that

9 satisfies the definition of pharmaceutically

10 equivalent, and then it just depends upon the

11 successful completion of a bioequivalence study to

12 achieve therapeutic equivalence.

13 In addition to that, there are some other

14 constraints. Labeling must be the same although

15 that has been interpreted rather broadly by generic

16 drugs, and depending on who is doing the

17 interpreting, the regulations are many different

18 things.

19 [Slide.

20 Review issues. The review of the

21 performance attributes, there are really only two

22 performance attributes that are critically reviewed

23 and felt to be critical to the performance of the

24 drug products and simply bioequivalence within

25 certain specified limits, subject to the ANDA, the


1 drug product must be equivalent to the innovator.

2 Of course, it must possess the stability

3 characteristics to remain bioequivalent over its

4 shelf life.

5 Other issues include the USP monographs.

6 They are relevant to our consideration of ANDA drug

7 products. Certainly, the monographs are the source

8 of specs and the drug product and drug substance to

9 the extent there are monographs, the dissolution,

10 identification, and assay specifications are

11 usually listed.

12 By regulation, USP is recognized as the

13 official compendium. To the extent that the

14 identity of an active pharmaceutical ingredient in

15 an ANDA is different from the monograph

16 description, then, it may not be labeled as such

17 unless it is explicitly disclaimed on the label.

18 Patents are prohibitive, but only in the

19 sense of the legal. They have nothing to do with

20 scientific conclusions.

21 [Slide.

22 The basis of industry arguments have

23 fundamentally claimed that the active

24 pharmaceutical ingredient that is of a different

25 polymorphics form or a different morphic form than


1 that in the innovator drug product, it doesn't

2 satisfy the legal requirement for identicality or

3 sameness, and furthermore, the point has been made

4 that existing guidances imply at least the lack of

5 equivalence when morphic form is different.

6 ICH Q6A has been cited although that

7 technically doesn't apply. It only applies to new

8 drug products.

9 There was also a final rule which is

10 relevant to this discussion in 1992, published in

11 the Federal Register, and that has been cited as a

12 reason for lack of identicality given the

13 definition of identicality given in that document

14 however, we use the same final rule as

15 justification, so it depends on what section you

16 read.

17 There was a court decision, Serono

18 Laboratories versus Shalala, which has been cited

19 by petitioners, in which the phrase or the ruling

20 that they cite is the variation of active

21 ingredient in a generic product should be permitted

22 unless, in addition to exhibiting clinical

23 equivalence to the pioneer, the generic must show

24 chemical identity to the extent possible, the last

25 phrase being the key.


1 So again that was a lengthy decision and

2 it depends on what portion you choose to excerpt.

3 [Slide.

4 An additional part of the argument,

5 although differences in formulation are permitted

6 to be different in the labeling, industry argues

7 that a different physical state or form is a

8 difference in active ingredient rather than just a

9 difference in formulation or any of the other cited

10 allowed differences.

11 Finally, because the generic is of

12 different physical form, it therefore doesn't meet

13 the standard of identity in the USP if it is

14 specified, and therefore, it would have to have a

15 different name, and a different name would

16 constitute different labeling and misbranding.

17 [Slide.

18 Industry has questioned the OGD process

19 with regard to I guess rigor and demonstrating the

20 lack of presence of all, or ruling out the presence

21 of all polymorphs in the active pharmaceutical

22 ingredient that is the subject of the application,

23 that is in the drug product that is the subject.

24 The acceptance criteria that we use has

25 been criticized for doing essentially only


1 evaluation of the bioequivalence after manufacture

2 of drug product without additional data points of

3 that performance attribute on stability or for

4 future batches.

5 Basically, the benchmarks that we use for

6 ensuring the sameness of the batch going forward,

7 they have been questioned, the process.

8 [Slide.

9 As I said at the outset, our arguments

10 have been rather simple, simply based upon the

11 belief that the identicality requirements are

12 satisfied by virtue of the same dosage form, the

13 same route of administration, conditions of use, et

14 cetera, and the physical form or the morphic form

15 is considered to be not relevant to the

16 pharmaceutical equivalence.

17 As stated in the Orange Book, the agency

18 considers drug products containing different

19 polymorphs of the same drug substance to be

20 pharmaceutically equivalent. So it only remains to

21 be demonstrated that it is bioequivalent to achieve

22 therapeutic equivalence.

23 [Slide.

24 Again, as Steve noted, in the 1987 Drug

25 Substance Guidance, although this doesn't speak to


1 solvates, it notes, essentially concludes that

2 although different polymorphs have different

3 thermodynamic energy content, they don't differ in

4 composition.

5 As far as the draft Drug Substance

6 Guidance, which has not been published, but as far

7 as I know, that doesn't specifically address the

8 issue of morphic form. As I said earlier, ICH Q6A,

9 we are not within the scope of that, but eventually

10 will be, so I think that that is a short-lived

11 point.

12 [Slide.

13 As far as labeling differences go, the

14 differences in labeling cited in 505 and the

15 subsequent codification of that, may include a

16 number of things. Those are noted on the slide.

17 The FDA feels that those are examples, and other

18 differences are not precluded, those are merely

19 examples of differences that are allowable.

20 Additionally, industry has cited, in Title

21 21, 229.5, wherein it is stated that to the extent

22 there are differences in the drug substance, the

23 pharmaceutical ingredient, it doesn't satisfy the

24 identity requirement, and the labeling must be

25 different.


1 But that same section goes on in Part C to

2 note that to the extent there are differences, as

3 long as those differences are specifically

4 disclaimed on the label, and the extent of

5 differences, it is allowable, and we have approved

6 applications with active ingredients that are not

7 the same, with disclaimers on the labels.

8 Furthermore, it is clear from the House

9 report on the original Hatch-Waxman Act, that

10 Congress did not intend for generic drugs to be

11 exactly identical in every respect to the innovator

12 drugs, and they included a number of other

13 qualifying differences aside from those noted in

14 the codification of that law including color and a

15 number of other things. Clearly, these were just a

16 non-complete list of things that could be

17 different.

18 [Slide.

19 As far as our process or assessment of

20 performance, the bioequivalence determination is

21 straightforward, and we have spent a lot of time

22 trying to ensure that the benchmarks that we have

23 with the pivotal batch or batches, that they are

24 adequate to ensure that the drug product will

25 remain the same from a bioequivalence standpoint


1 going forward.

2 That is not an easy task. Depending on

3 the drug product, there are issues, it is difficult

4 to determine whether or not it is going to be the

5 same with any surrogate test. In particular, often

6 the dissolution methodology or the specifications

7 are tailored to the application. We may have

8 several different sets of specs for dissolution

9 because the requirement is that it be

10 bioequivalent, so if the dissolution

11 characteristics are different, then, the

12 dissolution specs are changed.

13 That is the most critical issue clearly.

14 [Slide.

15 As far as our view of the USP monographs,

16 we have to certainly be aware of the standards of

17 identity as explicated in the monographs, when

18 there are monographs.

19 I just did a quick search, I guess the

20 results are on the next slide, but there are 97

21 monographs in the USP which have the identity term

22 crystallinity in them. There are 8 monographs with

23 amorphous present.

24 Most of those or 95 of those monographs

25 that have crystallinity in the description are


1 antibiotics curiously, but we have had applications

2 with different morphic form across the board,

3 though, they are not limited to antibiotics.

4 We do require unless in rare cases, we

5 require the applications to contain active

6 pharmaceutical ingredients that conform to the

7 identity description in the USP. What has been

8 done, we require the revision, the monographs to be

9 revised before approval.

10 The applicant contacts the USP and makes

11 the case for revising it, and it may be revised to

12 include inclusive of a different polymorph or

13 inclusive of amorphous, and, of course, since that

14 is an open process, the innovator companies

15 participate in the commenting during that revision

16 process, and it is lengthy usually.

17 [Slide.

18 So, in summary, 21 CFR 314.92 provides the

19 regulatory basis for determining the suitability of

20 drug products for filing in an ANDA. It reads in

21 part, Drug products that are the same as a listed

22 drug, and for determining the suitability of an

23 ANDA, the term "same as" means identical in active

24 ingredient, dosage form, strength, route of

25 administration, and conditions of use.


1 The agency considers drug products

2 containing different polymorphs of the same drug

3 substance to be pharmaceutically equivalent.

4 Additionally, in order to be therapeutically

5 equivalent, the drug product must be shown to be

6 bioequivalent to the listed drug upon which it is

7 based, and it must possess requisite stability

8 characteristics to retain that bioequivalence

9 within a reasonable shelf life.

10 It also should contain the same labeling

11 and it should conform to the identity and specs and

12 tests listed in the compendium if it is concluded.

13 [Slide.

14 The achievement of those requirements, the

15 critical questions are has the applicant

16 demonstrated diligence in screening for possible

17 polymorphs or hydrates or solvates. There have

18 been some rather practical solutions to some of

19 these problems.

20 In one case, the drug product was known to

21 be plagued with I believe seven different hydrates

22 and of differing solubilities and differing

23 bioavailabilities, so the solution was simply to

24 use a wet granulation, literally wet, and

25 overhydrate it, maximally hydrate it, and then dry


1 it down to a speck of less than a half a percent

2 water, and there were no further questions about

3 the hydrates.

4 As it happened, that anhydrous form was

5 acceptably bioequivalent and acceptably stable.

6 That begs the question of the potential for

7 interconversion on shelf life.

8 We have seen two applications that very

9 successfully used Raman spectroscopy for monitoring

10 of the drug product on shelf life, calibrated with

11 X-ray powder diffraction studies, which was very

12 good work

13 So, bioequivalence having been established

14 and shelf life, adequate shelf life satisfies our

15 constraints

16 [Slide.

17 There are complicated situations, such as

18 ANDAs, as I mentioned, with mixture of polymorphs,

19 and the formulation process may result in

20 interconversion during formulation requires an

21 assay which will work to discriminate among the

22 various morphic forms.

23 Generic Drugs gets a lot of change of

24 source of drug substance, so this is an ongoing

25 issue. In the previous slide, the first question,


1 this diligence, the screening for possible

2 polymorphs, this is something that we have been

3 much more attentive to recently. I think it is

4 going to become a routine process.

5 DR. LEE: Thank you very much.

6 Are there questions for Dr. Adams? Leon.

7 DR. SHARGEL: I want to thank you for a

8 difficult area, in the generic industry, the

9 concept of sameness is a major issue. In fact, I

10 think in terms of the generic manufacturer, if he

11 is not being sued by an innovator, he is not

12 successful.

13 It stands to reason that an API

14 manufacturer for a generic firm is going to use a

15 different synthetic pathway or approach that

16 doesn't infringe on patents of the innovator, so

17 therefore, the API manufacturer is more than likely

18 going to be somewhat different in terms of

19 polymorph or anhydrous or what is the hydration,

20 and things of that sort.

21 I think the main thing to consider is the

22 dissolution in vitro and bioequivalence in vivo,

23 that we are talking about the same molecular

24 entity, that once it is in solution or in the body,

25 that we really have a medical product. There is


1 much to-do about whether anhydrous and crystalline

2 form are different. I realizes in Chemistry 101 in

3 high school that they are different and have

4 different properties.

5 I think in my mind as we do stability, we

6 do dissolution, we do in vivo bioequivalence, how

7 much are these differences in crystalline

8 properties going to be different in terms of what

9 we expect in a therapeutically equivalent generic

10 drug product.

11 DR. LEE: Thank you.

12 DR. ADAMS: One more point I forgot to

13 make. I guess that one of the reasons why Raman

14 spectroscopy is particularly easy to implement in

15 the cases we have seen is because they have been

16 antibiotics, and they are largely drug substance,

17 which makes it a lot easier.

18 DR. LEE: Joe.

19 DR. BLOOM: If you have a product that has

20 many polymorphs, do you consider it to have just

21 one active ingredient?

22 DR. ADAMS: Yes.

23 DR. BLOOM: If the product becomes

24 bioequivalent, so that if an ANDA has polymorphs in

25 it, becomes bioequivalent, it should pass your


1 specs?

2 DR. ADAMS: Yes, if it's bioequivalent and

3 it's physically acceptable from an elegant

4 standpoint and stability, that is all that is

5 required. In other words, polymorphism or morphic

6 form differences do not preclude the conclusion of

7 sameness as far as composition goes.

8 That satisfies our requirement for

9 pharmaceutical equivalence.

10 DR. BLOOM: Then, a question is hydrates

11 or polymorphs are considered the same as the active

12 ingredient or is it in another category?

13 DR. ADAMS: No, they are considered the

14 same as.

15 DR. BLOOM: The hydrates or polymorphs are

16 considered the same?

17 DR. ADAMS: Yes.

18 DR. MORRIS: And amorphous.

19 DR. ADAMS: And amorphous, yes. OGD

20 regards morphic form, maybe the performance

21 characteristics may be different, and that, of

22 course, is another question, an independent

23 question, but as far as satisfying the regulation

24 and the law under 505(j), that the active

25 ingredient be identical, we have stated in the


1 Orange Book, in the preface, it states that

2 polymorphs and hydrates and amorphous forms, the

3 drug substance are considered to be

4 pharmaceutically equivalent.

5 DR. HUSSAIN: Let me sort of address that.

6 I think with respect to different polymorphs,

7 amorphous, and so forth, I think the key is

8 designing your products to meet the performance

9 attributes. I think you could start with different

10 polymorphic form, chemically identical one, in

11 solution, and essentially how you design your

12 process and how you design your product in terms of

13 bioequivalence and other attributes which link to

14 clinical safety and efficacy essentially is the way

15 we move forward on that.

16 That is how we handle that.

17 DR. MORRIS: A couple of points. I don't

18 think there is any real question that if you have

19 the drug molecule and you change its forms, that

20 it's the same drug molecule, the integrity of the

21 molecule.

22 Strictly from the technical and regulatory

23 and leaving the legal out of it because I don't

24 pretend to know anything about that, but there are

25 a couple of issues that are different when you go


1 to a generic, and I think you have hit on a couple

2 of them.

3 The first is the idea of what constitutes

4 a screen, an appropriate or a sufficient screen,

5 and it is different because of all the reasons that

6 we had talked about earlier with respect to

7 screens, and that is, that you have different

8 impurities, you have different solvents, you have

9 changes in vendors, and I can tell you that that is

10 one of the biggest problems that I have run into,

11 so that you really do have to do as good or more

12 broad a screen as the innovator does. Every time

13 you change a vendor, you are really safest to redo

14 a screen and considering the resources that it

15 takes, it is a no-starter to not do it, I mean you

16 just expect to do it.

17 The other thing is, is that I disagree a

18 little bit with what you are saying, Ajaz, and that

19 is because if I have a given form or let me use the

20 example of amorphous form since it is in this

21 category, if you look a work by Zagraffi [ph] and

22 others, you don't always get the same polymorph

23 depending on the conditions under which they

24 recrystallize, so if I start with a different form,

25 you may think, well, everything is fine and I will


1 monitor it through development, and then, you know,

2 sunset it, but if when it converts it converts to

3 an unknown but more stable form -- and this has

4 happened, not necessarily in solid dosage form --

5 then, you are not within the purview of the

6 stability and safety histories that the innovator

7 generated, so the generic is now putting a new set

8 of conditions on.

9 So, I agree that the molecule is the

10 molecule, but I disagree --

11 DR. HUSSAIN: You raised a good issue, but

12 I think in absence of any of the tests you have,

13 that becomes correct, and what you are saying is

14 right, but now I think the question I would sort of

15 rephrase that and say isn't the dissolution test

16 that we have set up - those are the questions.

17 DR. MORRIS: But the question really is --

18 when I pass my bioequivalence, am I passing it at

19 the margin or am I passing it well in the zone? If

20 I am passing it at the margin, then, I think you go

21 into your other point, which I was going to raise,

22 was with time, I think that is where this becomes

23 an issue, Ajaz, is with time to be able to make

24 sure --

25 DR. HUSSAIN: With time, I think the


1 monitoring of the dissolution, the stability

2 program essentially, then, the question becomes is

3 the stability program that we have protecting

4 against that.

5 DR. MORRIS: I fully agree that is the

6 issue. I am not sure that there is a good answer

7 to that question, though, is the problem, because

8 depending on where you live in that zone, there are

9 various differences.

10 To be frank, I think that formulation

11 differences make more of a difference typically

12 than crystal form if it's a polymorph you are

13 talking about, but if you are talking about going

14 from amorphous to crystalline, or anhydrous to

15 hydrated, the free energy differences are not

16 negligible, in fact, I would argue that the hydrate

17 formation and amorphous crystalline formation far

18 outweighs the typical differences you see in

19 polymorphic interconversion and are much more

20 common.

21 It is much more common in water,

22 everything gets exposed to some level, moisture,

23 you know, you put the cans in it, but it is in a

24 bathroom, so, you know, this is the old story.

25 DR. LEE: I want to save some time for


1 some philosophical discussions. I will now take

2 three more questions - Judy, Leon, and Mary.

3 Judy?

4 DR. BOEHLERT: Mine is a fast question.

5 Is OGD now using the ICH Q6A decision trees for

6 deciding when tests are done for polymorphs on

7 active ingredients?

8 DR. ADAMS: No.

9 DR. BOEHLERT: Has that been considered

10 because the issues may be the same, particularly

11 when you are talking about those critical

12 substances that are mixtures, and things of this

13 sort. While I grant that the solution will very

14 often pick up differences, you don't want to find

15 out that you have got an incoming material that is

16 different after you have made the dosage form.

17 DR. ADAMS: Well, again, I would point

18 out, first of all, much is known about the drug

19 substances, which are the subject of ANDAs, either

20 in the literature or -- so the problems are

21 somewhat different, and I would say one thing, that

22 as a result of that difference, I think that the

23 ICH Q6A decision tree No. 4 would be somewhat

24 different if it were inclusive of ANDAs, but that

25 is not an issue that I can really speak to


1 authoritatively.

2 DR. CHIU: However, we do have our

3 internal drug substance, drug product guidances

4 right now under development. Those guidances will

5 follow ICH CDDQ format, so therefore it will

6 include a Pharmaceutical Device section. These

7 guidances apply to generic drugs as well as new

8 drugs, so therefore, the same standard will be used

9 with regard to polymorph.

10 DR. LEE: Leon.

11 DR. SHARGEL: I think the innovator also

12 has similar issues as a generic and API. Very

13 often the API is changed, and that is why the

14 BACPAC guidance go back to the postapproval

15 changes. Both the innovator and the generic has to

16 requalify the API or the API manufacture.

17 The second issue is this idea of

18 marginally bioequivalent just hit me that whether

19 it's an innovator changing formulation or a

20 generic, I think we have a yes/no answer on

21 bioequivalence. We are either there or we are not

22 there. I don't know where we talk about marginally

23 bioequivalence.

24 DR. ADAMS: That is a difficult concept

25 actually to accept the statement or the truth that


1 anywhere within the window of acceptable

2 bioequivalence that the clinical endpoints are the

3 same, so would say "marginal" is a term that it is

4 not used.

5 DR. SHARGEL: It applies to both sides of

6 the industry.

7 DR. ADAMS: Yes.

8 DR. LEE: Mary.

9 DR. BERG: Just one further point for

10 clarification. In other words, you are saying if

11 your product conforms to the regulation by being

12 pharmaceutically equivalent, in other words, it

13 meets the standard as such, but then literally,

14 though, the product can be bio-inequivalent as

15 such.

16 That is what you are saying. So I think

17 that second question becomes important in different

18 ways that people are saying here, that one must

19 really look at that question of bioequivalence

20 because obviously, you are then affecting further

21 down the chain as such what is happening

22 pharmacodynamically as such.

23 That obviously gets into the safety

24 toxicity issue or efficacy issue, so that second

25 question really becomes very important even though


1 you have stated upfront, well, it's met the

2 regulation as such, but that second question is a

3 very good question that you have formulated to look

4 further.

5 DR. ADAMS: Of course those are the two

6 independent legs of the requirements, and it can't

7 get in the door unless it satisfies the number one,

8 and then once it's in, it must be bioequivalent to

9 be therapeutically equivalent, and therefore

10 approvable as a 505(j).

11 DR. LEE: Are there any other comments? I

12 think certainly you have heightened awareness of

13 this topic on both sides. Now, maybe I should get

14 some guidance or education about where the agency

15 is going from here.

16 DR. HUSSAIN: I think we just wanted to

17 bring this as an awareness topic from a very

18 different perspective. If you look at the previous

19 discussion we had with Christopher Rhodes, it

20 focused on physical steps, and so forth. I think

21 we want to come back. I just want to mention we

22 have a one-day symposium, internal symposium on

23 this topic in June, I believe June 6th.

24 We will go through an extensive look at

25 polymorphism from a science perspective, and so


1 forth, and eventually bring back a topic for

2 discussion, which will focus on physical stability.

3 I think that has not received the attention.

4 Again, when we do that and how we do that,

5 I think we will flesh it out as things go by. In a

6 sense, I think the draft guidances that we are

7 working on, and we will go through the process of

8 finalizing the drug substance, drug product, and

9 even BACPAC II as that comes along. Other

10 guidances that we have under work that will be the

11 focus of the discussion in terms of the physical

12 stability, and so forth.

13 DR. LEE: Are we all invited to that

14 symposium?

15 DR. HUSSAIN: Some of you are, and I think

16 if you are in town, you are more than welcome.

17 DR. LEE: This is a half-serious question.

18 Perhaps you know if the committee members who might

19 be providing some advice on this issue, then

20 perhaps I think ought to be there.

21 Nair will be. Are you speaking?


23 DR. MEYER: Let me just ask one question

24 in the context of waiving in vivo bioequivalence as

25 we are going to do for Class I drugs or are doing.


1 Is there anything peculiar about

2 polymorphism that would cause a difference in the

3 polymorphic composition of two products to show up

4 in vivo, but not show up in vitro? Can you

5 hypothesize a situation where that might occur?

6 DR. MORRIS: Are you asking me, Marv?

7 DR. MEYER: Anyone.

8 DR. MORRIS: I think Nair actually, you

9 spoke to that earlier, it's the kinetic question, I

10 think.

11 DR. RODRIGUEZ-HORNEDO: I understand that

12 but I think your question is more regarding could

13 we see polymorphs in vivo that we don't see in

14 vitro.

15 DR. MEYER: Yes.

16 DR. RODRIGUEZ-HORNEDO: I think we can see

17 crystallizations in vivo particularly with drugs

18 that are weekly basic, and I am not prepared to

19 talk about that, but we have seen in my lab some

20 investigations we have done, precipitation of drugs

21 by changing pH in aqueous systems that actually

22 create new solid faces that we have not seen

23 otherwise.

24 This is an in vitro method generally, but

25 it is not a method that would be used in chemical


1 development, so to address your question, yes, I

2 think there is a possibility for precipitation or

3 crystallization phenomena in vivo, at least in the

4 GI tract with weekly basic drugs.

5 DR. MEYER: But would you also see that in

6 vitro, and if you would, then we are safe.


8 DR. MEYER: If you would not, then, we are

9 not safe.


11 DR. LEE: To summarize, there are certain

12 words I learned in the last couple days, before I

13 turn it over to Helen or Ajaz on the next topic.

14 Our awareness about risk management resonates

15 throughout the two days, PAT, PCS, I think these

16 are all interconnected, and I am very pleased to

17 see the committee deliberate that we tried to see

18 how they might fit into those frameworks.

19 On that note, I would turn the podium over

20 to the OPS Updates. I am not sure who is going to

21 be on the floor.

22 OPS Updates

23 DR. HUSSAIN: I thank the committee for an

24 excellent deliberation and discussion. I think

25 this has been very useful. Reflecting back to


1 lessons learned from my perspective, I think we

2 could have done better with the agenda and

3 organizing some of that, so we learned a few things

4 we hope to use to improve the next meeting agenda

5 and then the background packet itself.

6 I think we will constantly work on

7 improving our meetings, and so forth, and always,

8 any feedback that you can provide will be very,

9 very helpful. I think the first day discussion, if

10 we had the BCS first and food effects second it

11 would really help, but I think it came out fine.

12 In terms of OPS update, I think Helen had

13 mentioned, and I will just reemphasize, that I

14 think we have moved the Office of Pharmaceutical

15 Science, and there have been some changes in terms

16 of new additions. One addition is an associate

17 director which would focus on rapid response

18 project, and I think at some point we will bring

19 that to the discussion.

20 In addition, we have several additions.

21 Dr. Sobel has joined us as Associate Director for

22 Medical Affairs, and I mentioned the Microbiology

23 staff moved to OPS, and we hope to have four more

24 individuals for PAT -- three more, we already have

25 one on board.


1 The next meeting would be in October. We

2 will work more diligently to make sure the agenda

3 and the background packet will be much better

4 designed to help you address the questions. Any

5 feedback more than welcome.

6 Thank you and have a safe trip back.

7 DR. LEE: Thank you very much. The

8 meeting is adjourned.

9 [Whereupon, at 3:20 p.m., the meeting was

10 adjourned.]

11 - - -