DEPARTMENT OF HEALTH AND HUMAN SERIVCES
NATIONAL INSTITUTE OF MENTAL HEALTH

NIMH-MATRICS: NEW APPROACHES CONFERENCE

Thursday, September 9, 2004
8:07 a.m.
Bolger Center
9600 Newbridge Drive
Potomac, Maryland 20854

PRESENTERS

WAYNE FENTON, M.D.
STEPHEN MARDER, M.D.
CAMERON CARTER, M.D.
ALAN BREIER, M.D.
JOHN JONIDES, Ph.D.
JAMES HAGAN, Ph.D.
TREVOR ROBBINS, Ph.D.

TABLE OF CONTENTS

.
Agenda Item Page
I. Welcome and Introduction 3
II. NIMH's Approaches to Treatment Development5
III. Overview of MATRICS and TURNS14
IV. Introduction: Thomas Insel 23
V. The Promises and Perils of Moving from Clinical Neuropsychological to Cognitive Neuroscientific Approaches to Measuring Treatment Effects on Cognition in SchizophreniA 25
VI. Needs and Approaches for Proof of Concept and Clinical Validation Studies of Putative Treatments for Cognitive Deficits in Schizophrenia 55
VII. Advancing Clinical Science Through the Translation of Cognitive Neuroscience Theory and Methodology 99
VIII. Needs and Prospects for Preclinical Predictors of Efficacious Treatments for Cognitive Deficits in Schizophrenia129
IX. Synthesizing Schizophrenia: A Bottom-up, Symptomatic Approach156
3

PROCEEDINGS

I. WELCOME AND INTRODUCTION

 
          3             DR. GEYER:  If people can take their

          4  seats, we can get this road show on the road.  We

          5  definitely need a collection of roadies up here

          6  with all the technical complications that we are

          7  having, but I think that will sort itself out.

          8             I want to welcome everybody to this,

          9  which is the sixth and last of the

         10  consensus-building conferences of the MATRICS

         11  program, with which I know this audience is quite

         12  familiar and will receive much more introduction

         13  this morning, but I do want to thank you for

         14  coming.  It's an opportunity, I think, for us

         15  to--and the goal is by the end of tomorrow--to have

         16  identified a series of specific recommendations for

         17  a research agenda for the future, both in academia

         18  and industry and within government efforts to

         19  further the research on new cognitive treatments

         20  for the--new treatments for the cognitive deficits

         21  with schizophrenia where we can identify compounds,

         22  validate them, and evaluate the patients.

 

                                                               4

          1             We do want to come out of this

          2  enterprise with some consensus product about the

          3  most important immediate priorities for research.

          4  So this morning, we will hear from a group of

          5  essentially representatives of the major consumer

          6  groups who will use this research agenda, and then

          7  this afternoon, each of you should sign up, if you

          8  have not already, for one of the four breakout

          9  groups of the early afternoon and another breakout

         10  group in the later afternoon.

         11             If you haven't gotten the printout of

         12  the breakout group detail information, Tanya has

         13  them outside at the desk.  They also indicate which

         14  room each of the breakout groups is in.  One is

         15  this room.  One is in Room 200 below us.  One is in

         16  Room 400 above us.  And another one is in Room 111,

         17  which there will be signage to direct you to.  And

         18  then at about 3:00 or 3:30, we will have a coffee

         19  break and people will shuffle and shift to the

         20  appropriate rooms for their second breakout group.

         21             And be sure to sign up for those if you

         22  have not already.  Some of them may fill sooner

 

                                                               5

          1  than others, and the sign-up is outside at the

          2  registration desk.

          3             We are hoping and there is a good

          4  possibility that Tom Insel will be joining us.

          5  Okay.  But being flexible, as we must be, and

          6  exercising executive functions, I will move ahead

          7  to Wayne Fenton who--no.  Who is next?  Yes.  It is

          8  Wayne, who will give you a little bit of background

          9  on where we've been in the past one and

         10  eight-tenths years.

         11             Wayne.

         12    II.  NIMH'S APPROACHES TO TREATMENT DEVELOPMENT

         13             DR. FENTON:  Thank you.  I suspect the

         14  reason that Tom Insel hasn't gotten here yet is I

         15  don't think he's ever been to the Bolger Center

         16  since moving to NIH, but I would bet he will

         17  probably be here sometime in the next 15 minutes.

         18             I guess I wanted to welcome you on

         19  behalf of NIMH to this sixth and final of the

         20  MATRICS meetings, and I really wanted to start off

         21  by thanking a number of people from the NIMH side.

         22  Bob Heinssen and Bruce Cuthbert have taken the lead

 

                                                               6

          1  in organizing this meeting and of course Mark Geyer

          2  from UCSD in putting this together.  I also wanted

          3  to thank Steve Marder and the UCLA team, many of

          4  whom are here and who have really overseen this

          5  process over the last almost two years.

          6             I guess in my opening remarks, I wanted

          7  to give a little bit of background and summarize, I

          8  think, where we are with respect to our thinking

          9  about treatment development, perhaps a little bit

         10  about what's been accomplished with the MATRICS

         11  process and where we go to from here.

         12             Just to start off, as you know, the

         13  mission of NIMH is indicated here, to reduce the

         14  burden of mental illness through research on mind,

         15  brain, and behavior.  Our priorities are set

         16  substantially based on scientific opportunity and

         17  public health importance, and the area of cognition

         18  schizophrenia is really an area that met both of

         19  these criteria.

         20             Increasingly, we are re-assessing all of

         21  our programs, in fact, reorganizing our programs,

         22  as maybe Dr. Insel will say when he gets here, to

 

                                                               7

          1  focus more tightly on the mission, and I think in

          2  particular to focus more tightly on trying to

          3  achieve results that will help patients with mental

          4  illness more rapidly than otherwise would be the

          5  case.  Obviously in this context, treatment

          6  development has emerged as really an imperative,

          7  and we will be looking for ways that we as a

          8  government agency can play a role in facilitating

          9  the development of new treatments.

         10             The overall problem I think you're all

         11  familiar with, this slide shows the trend in

         12  spending over the last decade.  The top line is

         13  spending in billions by U.S. pharmaceutical

         14  industry, R and D, in the bottom line, the total

         15  NIH budget.  So we really are coming off a period

         16  of doubling in the NIH budget and really very, very

         17  substantial investments in research and

         18  development.  At the same time, over this same

         19  decade, what we see is actually a decrease in NMEs

         20  received by the FDA.  So it appears as though we

         21  are in a situation of investing more and somehow

         22  getting less of a yield in terms of new and better

 

                                                               8

          1  medicines, medicines for patients.

          2             This just really looks at this in a

          3  little bit more detail and shows the number of NMEs

          4  received by the FDA in green, but if you look at

          5  novel molecular targets down here in purple, it's

          6  really even much fewer.  You see the cuts, who in

          7  terms of PD-5 in 1997, but I think everybody here

          8  is really, really quite familiar with the problem

          9  that many of the medicines that are being

         10  developed, particularly in psychiatry, are based on

         11  medicines or actions that were discovered

         12  serendipitously many years ago.

         13             This is a slide that actually Ed

         14  Scolneck brought in and really shows the impact of

         15  these issues for psychiatry, for depression and

         16  schizophrenia compared to heart disease.  We had in

         17  the 1950s two mechanistically distinct drugs for

         18  depression and for schizophrenia, three for heart

         19  disease.  I guess since the 1950s, beginning with

         20  the Farmingham study defining that elevated

         21  cholesterol is associated with a elevated risk of

         22  heart disease, the discovery of the enzyme and the

 

                                                               9

          1  synthesis of cholesterol and then the development

          2  of stantons and other advances in fundamental

          3  understanding of pathophysiology, we really now

          4  have 15 distinct mechanisms in our treatment of

          5  heart disease along the pathophysiologic

          6  architecture, whereas progress in depression

          7  schizophrenia has really, as you all are very

          8  aware, has really not kept pace.

          9             This is essentially where we are with

         10  respect to the antipsychotic medications and the

         11  second generation antipsychotic medications.  We

         12  have six over the past now almost 15 years, again,

         13  drugs which fundamentally have relatively minor

         14  differences in their molecular targets, and I think

         15  that, you know, we all have the sense, certainly

         16  those of us who treat patients as well as do

         17  research, that we're reaching a ceiling with

         18  respect to what we can expect in terms of patient

         19  improvement from drugs of this sort.

         20             The MATRICS initiative was started about

         21  two years ago.  I think our first meeting may have

         22  actually been here in the Bolger Center to address

 

                                                              10

          1  this issue, and it appears as though cognition and

          2  schizophrenia was really a very, very ripe target

          3  for translation.  We had a 25-year history of basic

          4  research focused on mechanisms.  We had probably at

          5  least half a dozen credible theories with respect

          6  to potentially new molecular targets.  There were a

          7  number of barriers, really, to development that the

          8  MATRICS program was designed to try to overcome.  I

          9  won't really reiterate these in detail.  I think

         10  that everybody here is quite familiar with it, but

         11  this has evolved really into three distinct

         12  programs in MATRICS, which the goal was to really

         13  try to clarify the methodology and actually working

         14  with the FDA to clarify the pathway to drug

         15  registration for medicines targeting cognition

         16  schizophrenia with the underlying belief that if we

         17  could achieve clarity, that would create an

         18  incentive for greater investment in this area.

         19             The TURNS program, which Steve Marder

         20  will talk a little bit more about shortly, has been

         21  awarded over the last couple months.  We now have a

         22  network of six clinical performance sites that are

 

                                                              11

          1  engaged currently in trying to identify nominations

          2  for compounds and will actually be using some of

          3  the methods developed in the earlier conferences to

          4  test compounds targeting cognition schizophrenia,

          5  both further refining the methods and also to try

          6  to get the field jump started.

          7             Finally, we anticipate publication of a

          8  standardized NIMH battery to assess neurocognition

          9  as an influence in schizophrenia, probably sometime

         10  in the next six months.

         11             These, then, were the major products of

         12  the MATRICS process to date.  We do have a

         13  neurocognitive assessment battery specifically

         14  designed for schizophrenia.  We have prioritized

         15  targets and compound lists.  FDA and NIMH have

         16  collaborated on guidance for a trial design, and we

         17  have actually a draft of that guidance posted on

         18  the MATRICS web site, and we anticipate publication

         19  of that in a journal sometime over the next six

         20  months.

         21             I believe that we have succeeded in jump

         22  starting government, academic, and regulatory and

 

                                                              12

          1  industry partnership to target this therapeutic

          2  area.  In addition, we have set up a clinical trial

          3  network.

          4             I just wanted to end by looking forward

          5  and to call your attention to anyone here who may

          6  not have seen this.  This was a report from the FDA

          7  that was actually released in March 2004, which

          8  represents their perspective on the pipeline

          9  problem with respect to medications.  Essentially,

         10  they, the FDA authors of this report, were in the

         11  position of having seen both the successes and the

         12  failures in terms of new medication development

         13  over the last 30 years and try to summarize their

         14  recommendations for what we can do to really move

         15  things forward in this report, and if you haven't

         16  read it, I would recommend it to you.

         17             But, in essence, in FDA's view, the

         18  applied sciences needed for medical product

         19  development hasn't kept pace with the advances in

         20  basic science.  The new science isn't being used to

         21  guide the technology development process in the

         22  same way that it's accelerating the technology

 

                                                              13

          1  discovery process.  In other words, in many cases,

          2  developers of new medicines have no choices but to

          3  use the tools and concepts of the last century to

          4  assess this century's candidates.

          5             I think one of the major criticisms of

          6  the MATRICS process has been that in a way, by

          7  definition, it has been backward looking.  In

          8  creating a neuropsychological battery, we've had to

          9  rely upon existing batteries where there's evidence

         10  of reliability and validity in putting together a

         11  battery.  I think one of the criticisms that has

         12  been valid is that we may have ignored important

         13  findings from cognitive neuroscience in this

         14  process.  It's really in that context that this

         15  final conference was designed, to try to look not

         16  at how we'll measure cognition next month or next

         17  year as we initiate clinical trials, but what sort

         18  of research we need to do over the next five years

         19  or the next ten years to get us to the next

         20  generation of more sensitive assays that will then

         21  facilitate, you know, things in a way that exceeds

         22  what we can do today with medication development.

 

                                                              14

          1             I think I will probably stop there and,

          2  once again, on behalf of NIMH thank you for your

          3  participation.  There is a great deal to needing a

          4  structure in terms of breakout groups, and we look

          5  forward to your recommendations and to review them

          6  carefully and take them seriously as we design

          7  initiatives going forward over the next several

          8  years to continue to facilitate the process of

          9  medicine development, both in this area as well as

         10  in other areas.

         11             So thank you and welcome.

         12             [Applause.]

         13             DR. GEYER:  Thank you, Wayne.

         14             Next up, I would like to introduce Steve

         15  Marder, who has been the principal investigator,

         16  together with Michael Green at UCLA, of the MATRICS

         17  project.

         18             So Steve.

         19          III.  OVERVIEW OF MATRICS AND TURNS

         20             DR. MARDER:  Thank you, Mark.  On behalf

         21  of the UCLA group, I'd like to welcome all of you

         22  to this sixth MATRICS conference and one that

 

                                                              15

          1  actually comes at just about the two-year

          2  anniversary of the initiation of the MATRICS

          3  program.

          4             Okay.  As Wayne mentioned, there are two

          5  main goals of the MATRICS initiative as I see it.

          6  One of them is the very practical issue of trying

          7  to facilitate the development of

          8  cognition-enhancing drugs for schizophrenia, and I

          9  think you'll see that we've made substantial

         10  progress in that, but I think the second is to

         11  develop models for drug development that assure

         12  that findings in basic neuroscience that come out

         13  of academia can be translated into improved patient

         14  care and improved patient outcomes; and in that

         15  sense, every MATRICS meeting, all six of them, have

         16  been experiments in how to develop that

         17  collaboration and improved communication.

         18             I believe that one of the by-products is

         19  that many neuroscientists and neuropsychologists

         20  and clinicians now understand much more about the

         21  process of drug development and the concerns of

         22  industry than ever before, and I think we've set

 

                                                              16

          1  the stage for that kind of collaboration.

          2  Similarly, I think the relations with industry have

          3  been enhancing, that they understand much better

          4  the concerns of academia.

          5             Now, this is the way I view the status

          6  of MATRICS as of today.  Under the leadership of

          7  Michael Green and Keith Nuechterlein and the

          8  neuropsychology committee, we have the beta version

          9  of the MATRICS battery.  It's on the MATRICS web

         10  site.  It's been evaluated in the past study, which

         11  has actually met its recruitment criteria.  It's

         12  been done, if you can believe it, more rapidly than

         13  thought with a lower dropout rate and under budget,

         14  and the results, the final MATRICS battery, should

         15  be announced sometime next month on the MATRICS web

         16  site.

         17             The process for the development of this

         18  battery will be described in upcoming issues in an

         19  article in Biological Psychiatry that's in press,

         20  and a broader discussion of those issue will be in

         21  a special theme issue of Schizophrenia Research,

         22  which I believe will be published in December, in

 

                                                              17

          1  December of this year.

          2             Our other goal was to prioritize

          3  molecular targets.  Many of the group here were at

          4  a meeting at the NIH clinical center last June.  At

          5  that meeting, we prioritized molecular targets.

          6  The results of that meeting, which was run by Mark

          7  Geyer and Carol Taniga, will be posted will be

          8  posted on the--well, it's already posted on the

          9  MATRICS web site, and it's described in a special

         10  issue of Psychopharmacology which has already been

         11  published.  If any of you don't already have it, we

         12  can get it to you very quickly.  Send me an E-mail,

         13  or Mark.  I think that most of the people here have

         14  it.

         15             Another priority was to address the

         16  issue of what types of trial designs will be most

         17  useful for studying neurocognition in schizophrenia

         18  and cognition-enhancing drugs.  We had a joint

         19  NIMH-FDA meeting.  The results of the--the

         20  transcript of that meeting are naturally on the

         21  MATRICS web site.  A lengthy article describing the

         22  decisions from that meeting has already been

 

                                                              18

          1  written and is on the verge of being submitted.

          2             Next.  But there's other things to come.

          3  First, the past study is still ongoing.  They're

          4  developing normative data, and I think that that

          5  should be completed within the next several months.

          6  Of course, the next thing is this meeting, which I

          7  think plays a vital role in MATRICS because it

          8  addresses the issue that Wayne described in that

          9  the MATRICS process, in order to form a consensus

         10  and to facilitate drug development, have needed to

         11  take the field and freeze it at a moment in time

         12  and develop a consensus about what's the optimal

         13  battery or optimal way of measuring cognition as an

         14  outcome and to prioritize molecular targets, that

         15  there is a difference between the world of science,

         16  which deplores a consensus and actually enjoys the

         17  competition of ideas, and the world of drug

         18  development and the kind of information that FDA

         19  needs which actually has to freeze the world at a

         20  moment in time.

         21             This meeting, I think is looking forward

         22  and looking at sort of the next steps in the NIMH

 

                                                              19

          1  research agenda which can facilitate treatment

          2  development in this area.

          3             I think that some of the lessons that

          4  we've learned in this process we want to apply to

          5  other domains, and Will Carpenter and Wayne Fenton

          6  and I are in the stages of organizing a meeting to

          7  see if some of these methods can be applied to the

          8  issue of negative symptoms in schizophrenia, which

          9  I believe is an area in which the conceptualization

         10  of the domain needs to be re-evaluated just as we

         11  re-evaluated cognition in schizophrenia.

         12             As Wayne mentioned, the TURNS contract

         13  was finally awarded over the summer.  It's going to

         14  be coordinated at UCLA with myself and Michael

         15  Green.  The NIMH project officer will be Wayne

         16  Fenton again.  There is a trial management unit

         17  which will be headquartered at the University of

         18  Maryland under Bob Buchanan and a scientific

         19  operations unit which resides in North Carolina

         20  under the leadership of Jeff Lieberman.  In fact,

         21  the main activities thus far have resided in the

         22  scientific operations unit.

 

                                                              20

          1             Next slide.  These are the TURNS

          2  performance sites.  As you can see, they're at

          3  Harvard General, Nathan Kline Institute, University

          4  of Maryland, University of Carolina and Duke,

          5  Washington University, and UCLA.

          6             Next slide.  And these are the TURNS

          7  activities that are currently ongoing.  As I

          8  mentioned, they reside mostly in Jeff Lieberman's

          9  unit, which is identifying the compounds that

         10  should be used within a clinical trials network.

         11  We submitted a solicitation for nominations, which

         12  I believe has been very successful.  We have had a

         13  number of compounds nominated using different

         14  molecular targets.

         15             Jeff, do you remember how many

         16  targets--we haven't closed yet, but I thought it

         17  was eight or ten.

         18             DR. LIEBERMAN:  [From audience,

         19  inaudible.]

         20             DR. GEYER:  Could we use the microphone

         21  so we can get it on tape, Jeff?

         22             DR. MARDER:  I wanted you to hear this

 

                                                              21

          1  because this is sort of breaking news.

          2             DR. LIEBERMAN:  Sorry about that.  But

          3  we had received, handed off by Mark and Carol, a

          4  list of vetted compounds that was paired down to 43

          5  as being worthy of consideration.  Subsequent to

          6  that, a solicitation was posted in a variety of

          7  biotech and pharma news services and an ad put in

          8  Nature, calling for nominations, and that was put

          9  in in early August, and since then, we've gotten 26

         10  nominations from a variety of big pharma, mid-size,

         11  and biotech companies.

         12             DR. MARDER:  Yes, and I think the

         13  important message from that is that big pharma,

         14  medium-sized, and small pharma are all interested

         15  in working with this particular network.  The

         16  compounds that have been nominated are being

         17  evaluated by a group of target experts, and many in

         18  this group are going to be called for their advice

         19  to us.  We've already developed the criteria for

         20  evaluating nominated compounds, and these compounds

         21  are going to be prioritized in November of this

         22  year.  Once we've decided what compounds we're

 

                                                              22

          1  going to study, we're going to initiate in November

          2  the actual design of these clinical trials with the

          3  idea that the first trials will be initiated in May

          4  or June of '05 with a period of subject entry and

          5  initiation of trial.

          6             The first TURNS trials may include large

          7  multi-site clinical efficacy trials, smaller

          8  pharmacodynamic or pharmacokinetic trials, proof of

          9  concept trials.  Those things remain to be seen.

         10             I think that's my final slide.  In other

         11  words, I think the process is moving forward, and I

         12  think we'll certainly be informed by the results of

         13  this meeting.

         14             Thank you very much.

         15             [Applause.]

         16             DR. GEYER:  Thank you, Steve.  We're

         17  not, obviously, taking questions at this point

         18  unless there are some very burning ones, of course,

         19  because two-thirds of this meeting is already

         20  extemporaneous.  So I don't think you'll end up

         21  saying there wasn't a chance to be heard.

         22             Now, on time for his 8:30 scheduled

 

                                                              23

          1  arrival, Tom Insel will introduce the conference

          2  for us, and then we'll return to the program as

          3  listed.  We've invited Tom here first as a

          4  scientist, by the way, and only secondarily as the

          5  director of NIMH.

          6            IV.  INTRODUCTION:  THOMAS INSEL

          7             DR. INSEL:  So in light of being a

          8  scientist, for that part of the invitation, I

          9  didn't wear a tie.  As director, I decided to

         10  arrive late.  I was thinking as I walked in, as my

         11  schedule actually said 8:30, was I had a similar

         12  thing happen in medical school with a professor of

         13  physiology who arrived late.  The way he started

         14  the lecture was to say that he had recently picked

         15  up a young woman in a local bar in Boston, and she

         16  said to him--he arranged to meet her in his room,

         17  and she said, That's great, let's meet at 8:30.

         18  But she said, If I'm not there, you start without

         19  me.  So I'm glad you had the wherewithal to do just

         20  that.

         21             I'm only going to take a minute, because

         22  we're already going into the material, to say that

 

                                                              24

          1  this is a very high priority for the institute.  As

          2  all of you know, the mission of NIMH is, as it's

          3  coded in our web site and now in our brains, is to

          4  reduce the burden of mental illness through

          5  research on mind and brain behavior.  The problem

          6  of cognitive deficits in schizophrenia, we're

          7  really at the very core of the burden of the

          8  disorder.  As we've said in previous meetings, and

          9  this is now the last of a series that began in

         10  mid-April of 2003, there's a clear recognition that

         11  even though we can have some impact on the positive

         12  symptoms, a little less on the negative symptoms,

         13  on the cognitive deficits in this disorder, we

         14  really have not made a very clear or consistent or

         15  reliable dent, and yet only about 30 percent of

         16  people with this disorder are able to work, at

         17  least with the current treatments that we have.

         18  Only a small fraction are able to really take on

         19  life in the community, which we would envision as

         20  being recovery.  And that's very much for us a

         21  challenge, I think over the next decade, is how we

         22  can do that.

 

                                                              25

          1             What's exciting for us at NIMH about

          2  this process is not only that this has become now a

          3  poster child for what we want to do, this is

          4  probably one of our absolute highest priorities,

          5  but also the way in which this problem is

          6  approached is a little different for us.  It's not

          7  only the partnership with industry, which is just

          8  obvious looking around this room, as it has been in

          9  all of the meetings that we've had, but it's also

         10  the partnership with the FDA and other parts of

         11  government, and actually for us what is kind of

         12  wonderful to see here is the partnership even

         13  across the institute.  So I see in the audience the

         14  leadership and program officers from across all of

         15  the different parts of the NIMH as well.

         16             So this is very much a huge

         17  collaborative effort for a very important target.

         18  It is a very, very high priority for us.  This is,

         19  of course, one way to approach the whole problem of

         20  coming up with new medications, that is to identify

         21  a target and look at what's out there and try to

         22  hit it in the way that Steve just described.

 

                                                              26

          1  There's also other things going on.  We have a very

          2  large small molecular repository that we're putting

          3  together under the road map initiative.  In that

          4  case, we're going to be looking for molecular

          5  targets, doing very high treatment screening, and

          6  at some point in the future, we hope these kinds of

          7  efforts come together so that five years, ten years

          8  from now, we can look back and say, my goodness, we

          9  really have been able to make a difference for

         10  people with this very serious disease.

         11             So I'm here really just to say in the

         12  last of this series of MATRICS meetings, as we

         13  begin to make the transition now to actually

         14  getting the work done, thank you for your working

         15  with us up to this point.  I especially want to

         16  thank Bob Heinssen and Bruce Cuthbert who on our

         17  side are working with Mark Geyer to put this

         18  particular meeting together and to recognize Wayne

         19  Fenton as well, who in many ways has been the brain

         20  child for this whole project from its inception.

         21             So good luck with the meeting.  I'm

         22  going to be here almost the entire day, which

 

                                                              27

          1  almost never happens for me, but this is something

          2  that we all are very invested in and I look forward

          3  to seeing the fruits of your labor.

          4             Thank you.

          5             [Applause.]

          6             DR. GEYER:  Thanks, Tom.  The real work

          7  begins this afternoon.

          8             Now we're going to get into some of the

          9  scientific background that address the content more

         10  than the process, and with that, I welcome Cam

         11  Carter, who has been a very active and contributing

         12  member of the New Approaches Committee, as we call

         13  ourselves.  Over the past six months or so, we've

         14  been having a whole series of conference calls from

         15  around the world, and it's a real pleasure to have

         16  been able to work with him and thereby get to know

         17  Cam.  And I'm not even going to try to introduce

         18  his title for you.

         19  V.  THE PROMISES AND PERILS OF MOVING FROM CLINICAL

         20    NEUROPSYCHOLOGICAL TO COGNITIVE NEUROSCIENTIFIC

         21     APPROACHES TO MEASURING TREATMENT EFFECTS ON

         22               COGNITION IN SCHIZOPHRENIA

 

                                                              28

          1             DR. CARTER:  So the title was a

          2  collaboration between Mark and I.  As you can see,

          3  it's already a very productive collaboration.  I'm

          4  also not used to working so closely with Bob

          5  Heinssen.

          6             So traditionally the development of

          7  pharmacological treatments has involved an

          8  interplay between work using animal models of human

          9  cognition, and eventually that's resulted in

         10  clinical trials targeting various disturbances in

         11  behavior, cognition, and emotion in patients.  Over

         12  the last ten years, I think particularly over the

         13  past several years, the field has seen some very

         14  dramatic developments.  First of all, we know a lot

         15  more about how the brain works from animal models,

         16  increasingly sophisticated animal models of human

         17  behavior, and we know also know a lot more about

         18  human behavior through the development of

         19  increasingly sophisticated models of human

         20  cognition and approaches to study it.

         21             In addition, we now have this amazing

         22  translational bridge that's afforded us by the

 

                                                              29

          1  availability of basic functional brain imaging.

          2  It's possible to take sophisticated animal models

          3  and actually have the resources.  You can do

          4  invasive studies at the same as you are do imaging

          5  studies of the kind that we were doing.  It's also

          6  possible to take patients, to image them, and to

          7  test the effects of drugs directly on brain

          8  activity associated with the emotional and

          9  cognitive processes in the diseases.  So it's a

         10  remarkable possibility, but as Wayne noted today,

         11  there's not necessarily been a rapid implementation

         12  of this approach from a discovery process,

         13  particularly as it might relate to the development

         14  of new treatments for diseases such as

         15  schizophrenia and targeting treatment refractory

         16  aspects of the disease such as cognitive

         17  disfunction.

         18             So today we begin to focus both on the

         19  promises and on perils of taking this approach.

         20  I'd suggest that today while we're going to address

         21  and we've got both of those avenues up on the

         22  screen, there are two general avenues that we can

 

                                                              30

          1  take for integrating cognitive neuroscience into

          2  the drug discovery process for impaired cognition

          3  in schizophrenia.  One, which I think is a much

          4  more longer term objective, would be to actually

          5  take this approach and apply it to clinical trials,

          6  outcome measurement, as per the initial part of

          7  MATRICS.  I think it's very unlikely that we will

          8  use neuroimaging as outcome measurements in

          9  clinical trials in the very near future, but it

         10  certainly is possible if we develop more specific

         11  and sensitive measures of a cognitive function that

         12  we could use experimental cognitive measures in

         13  clinical trials and outcome studies.

         14             I think the most immediate application

         15  of cognitive neuroscience through the drug

         16  discovery process for indicating cognition in

         17  schizophrenia will be through the development of

         18  methods for undertaking proof of concept studies in

         19  compounds that might have shown progress or promise

         20  in animal models in cognitive schizophrenia.  To

         21  make this transition, there really are two kinds of

         22  challenges.  The first is to make the transition

 

                                                              31

          1  from using methods that have been more traditional

          2  in the study of cognition in schizophrenia,

          3  measures from clinical neuropsychology, to using

          4  measures from experimental cognitive psychology

          5  which are the substrain which most neuroscientists

          6  and cognitive neuroscientists use in their studies,

          7  and then there are a second set of measurement

          8  initiatives that relate to the use of non-invasive

          9  imaging methods and a number of complexities and

         10  compounds that are related to using these tools.

         11  And the example that I think we'll talk about a lot

         12  over the next couple of days is pharmacofMRI.

         13             Can I have the next slide?

         14             So to make this transition from a

         15  cognitive--towards a cognitive psychology of

         16  schizophrenia involves a shift, I think, in

         17  emphasis.  As it's been noted, the work that was

         18  done in the first round of MATRICS has been

         19  appropriately conservative, and most of the

         20  measures that have developed the MATRICS have been

         21  measures that have been adapted from the various

         22  batteries of clinical neuropsychological tests.

 

                                                              32

          1  There are many advantages to this approach, and

          2  I'll skip through a few of those.  There are also

          3  some limitations.

          4             The challenge for us is what to study in

          5  cognitive neuroscience.  If we're going to use

          6  cognitive neuroscience, if we're going to bring its

          7  constructs and tools to bear on the problem of

          8  developing treatments for impaired cognition in

          9  schizophrenia, we're going to need to make this

         10  transition towards cognitive psychology, and I

         11  would note that there will be some challenges and

         12  perils, and we'll talk a lot of about those,

         13  particularly in the breakout groups this afternoon,

         14  but there also may be some advantages with regard

         15  to the measurement of cognition in schizophrenia.

         16             Clinical neuropsychology, as we know,

         17  uses standardized tests and administration methods

         18  to evaluate mental abilities to aid in the

         19  diagnosis and to provide a baseline for

         20  rehabilitation, and most of the measures in the

         21  current MATRICS initiative are subtests or tests

         22  from the various studies that have been used over

 

                                                              33

          1  many years in the clinical assessment of patients

          2  with schizophrenia.  These tests have many

          3  advantages.  The first is that they have a

          4  standardized form and administration.  In fact,

          5  many clinical psychologists do advanced fellowships

          6  to learn how to do this right, to get this right,

          7  to administer the tests and to interpret them in a

          8  rigorous way.

          9             There's extensive knowledge data

         10  associated with the clinical neuropsychological

         11  tests, and these tests have well-established and

         12  very robust psychometric properties.  So they're

         13  good for measurement.  They have high test

         14  reliability and they're not confounded by floor and

         15  ceiling effects.  They have a test that measures

         16  the constructs that will perform well both within a

         17  testing session and across testing sessions and in

         18  which subjects are not performing at ceiling, so

         19  that there is no room to show no improvement, or

         20  they're not performing at floor, that is to say

         21  they're not performing the test.  It's really a key

         22  prerequisite for any measure that's used to detect

 

                                                              34

          1  a change in a clinical trial that's focused on

          2  impaired cognition in schizophrenia.

          3             There are some limitations to this

          4  approach though, and perhaps this is where I will

          5  try to convince you to buy into the idea that there

          6  may be some advantages as we turn to the future in

          7  the measurement approach to cognition that

          8  cognitive psychology provides.  One of the key

          9  issues with regard to these measures is that they

         10  are very complex and it's not possible to relate

         11  performance on most standardized clinical and

         12  neuropsychological tests to specific cognitive

         13  mechanisms.

         14             So actually you'll probably do about six

         15  clicks here.

         16             So the task might be the Wisconsin Card

         17  Sorting Task.  I think John Jonides is probably

         18  going to have a similar slide to this in his talk.

         19  It involves having subjects sort cards based on one

         20  or another feature, and initially subjects have to

         21  learn a certain the rule, sorting rule, that they

         22  do that based on feedback, and then once they've

 

                                                              35

          1  learned the rule, it gets switched on them and they

          2  have to learn a new rule, not only the old rule,

          3  but then a new rule.  And this task is a very

          4  reliable task for distinguishing between patients

          5  with schizophrenia and healthy control subjects.

          6  It has good psychometric properties, and it engages

          7  a wide range of cognitive processes, even a couple

          8  more.

          9             So one has to process the feedback.  One

         10  has to learn from it.  One has to be able to shift,

         11  as Mark so nicely typified earlier.  One has to be

         12  able to select the relative dimensions to respond

         13  to, to both select the appropriate response and

         14  outcome for each of the rewarding responses.  One

         15  has to use working memory to keep all this

         16  information in mind in order to make the right

         17  decision, and one has to coordinate all these

         18  processes.  And on the one hand, having a global

         19  case like this will either provide a psychometric

         20  ability, but on the other hand, there is a danger,

         21  and the danger is if you are trying to remediate

         22  the specific cognitive deficit that's related to

 

                                                              36

          1  functional outcome in schizophrenia, and it's

          2  related to one of these processes--let's say,

          3  hypothetically, it's related to set shifting.  One

          4  could actually have a drug that could change

          5  performance on set shifting in a way that could be

          6  positively correlated to functional outcome, and

          7  yet because this process is embedded in a whole

          8  range of other processes, one might not detect that

          9  change with a measure like this.

         10             A second concern about clinical

         11  psychological tests is that for the most part,

         12  these initiatives are unable to distinguish between

         13  the generalized performance deficit and one that's

         14  associated with a specific cognitive deficit.  So

         15  this concept, which has been talked about a little

         16  bit in some of the MATRICS meetings, relies on the

         17  fact that different tasks have different properties

         18  with regard to their sensitivity to individual

         19  differences.  So this construct is often referred

         20  to as discriminating power, and one of the concerns

         21  related to this is that a task that has very good

         22  discriminating power is very sensitive to

 

                                                              37

          1  individual differences, and it will be very

          2  sensitive to a generalized deficit performance.

          3             Now, what do I mean by generalized

          4  deficit performance?  What I mean is poor

          5  interaction, poor test-taking skills, feeling bad,

          6  feeling dysphoric, not being engaged in the task.

          7  So let me give you an example.  This is a bad

          8  example from my own research.  A number of years

          9  ago, actually a really long time ago, inspired by

         10  work by the late Phil Hosman and the late Matt

         11  McKeesh in Stone Park, we did a study of spacial

         12  working memory in schizophrenia.  In this study, we

         13  showed subjects a central fixation cross, and in

         14  the next frame, a probe came on at a random

         15  location around a circumference, and following

         16  that, a random letter appeared and subjects had to

         17  respond by naming the letter which identified the

         18  location.  And there was either no delay between

         19  the response and the onset of the probe, so no

         20  spacial working memory demands, or there was a long

         21  eight-second delay.

         22             Show us the next slide.

 

                                                              38

          1             In this study, we were able to show that

          2  in addition to performing overall poorly,

          3  schizophrenia patients had a bigger deficit in

          4  performance with the delay than in the non-delay,

          5  and this was statistically significant, and we

          6  would have liked to have been able to claim this

          7  benevolence for spacial working memory deficit in

          8  schizophrenia, but here is the problem.  The long

          9  delay condition had better discriminating power

         10  than the short delay condition, and so if patients

         11  weren't motivated, if they just kind weren't into

         12  doing the task, they, in fact, will do worse in the

         13  long-delay tasks.

         14             So the notion here is in the face of a

         15  generalized deficit--I'm really starting to get

         16  into this coordinated thing now--in an easy working

         17  memory task that had low sensitivity compared to a

         18  hard-working memory task that has high sensitivity

         19  will result in a pattern of performance that will

         20  look like this.  So you get a relatively small

         21  effect for an easy task and a relatively larger

         22  effect for the hard task.  What you do to an

 

                                                              39

          1  alteration in a working memory project, what you

          2  will be detecting is a generalized deficit.

          3             And again, I want to make the point that

          4  in the same way that identifying specific deficits

          5  amongst an array of other cognitive mechanisms that

          6  might be engaged by a task, distinguishing between

          7  a generalized deficit and a specific deficit is not

          8  just a theoretical advantage.  It's not just sort

          9  of cognitive creativeness.  It actually speaks to

         10  the issue sensitivity and specificity of

         11  measurement.  If, indeed, there are specific

         12  deficits, and I think most of us believe there are,

         13  and they are very generalized deficits, your

         14  ability to detect change with the task that is

         15  compounded by both components is going to be

         16  reduced.  There is a big danger when you're testing

         17  the molecules in this way that  you'll have a

         18  Type-II here.

         19             There is another reason why I think it's

         20  particularly helpful to take this more

         21  reductionistic approach and to pass specific

         22  deficits from generalized deficits.  This comes,

 

                                                              40

          1  actually, from many conversations that I have had

          2  with my colleague and friend Jim Gold about the

          3  fact that it might be that a very important and a

          4  salient aspect of schizophrenia is not just the

          5  specific deficits that we might be able to measure,

          6  but also the generalized deficits.  We all know

          7  that patients have problems with motivation.  We

          8  know that the treatments sometime appear where

          9  there's no assistance.  Systems such as motivation

         10  have now become tractable and very popular, in

         11  fact, for investigation in cognitive effect in

         12  neuroscience.

         13             Next one.  There is an identifiable

         14  neurocircuitry which involves regions such as the

         15  orbitofrontal cortex that's involved in subjects

         16  with motivation and pharmacology, and this could

         17  just as well be a viable target for drug

         18  development that we'd target and something that we

         19  would consider to be a component of the generalized

         20  deficit, but, in fact, but being strongly related

         21  to functional illness.

         22             So I'm not a psychometrician.  I'm just

 

                                                              41

          1  a psychiatrist.  There are solutions to this issue

          2  of generalized deficit.  There are some experts

          3  developing these solutions that are here in the

          4  audience, people like Dalton Strauss and Steve

          5  Sorenstein.  I'll refer you to those, but the key

          6  thing with regard to this issue of standardized

          7  tests versus experimental cognitive tests is that

          8  most of the solutions, particularly psychometric

          9  matching, developing tasks, specific tasks and

         10  controlled tasks, for the equivalent discriminating

         11  power or process disassociation approaches where

         12  you have an indexed menial task that is associated

         13  with good performance and an index that might

         14  distinguished a generalized deficit from a specific

         15  deficit, these kinds of approaches require task

         16  modification.  You can't do a standardized task.

         17  You have to change the task.  You have to develop

         18  multiple tasks in order to be able to make these

         19  distinctions.

         20             Now, this idea that experimental

         21  cognitive psychology is a useful construct that

         22  provides us with a useful set of tools to

 

                                                              42

          1  psychopharmacology is really not a new one.  This

          2  is a paper that Notch Callaway published in 1988

          3  that made this case.  Notch is one of the pioneers

          4  of this field.  In fact, he was doing studies in

          5  cognition in schizophrenia and publishing them in

          6  the year that I was born, and he also was using the

          7  same tasks that we're currently using.  Notch makes

          8  the case in this article that experimental

          9  cognitive psychology can, in fact, provide us with

         10  the tools and constructs that we need for drug

         11  discovery as it might relate to such things as

         12  impaired cognition in schizophrenia.  I think he's

         13  very active in his retirement.  If you send him an

         14  E-mail, he'll send you a PDM of his paper, which I

         15  recommend.

         16             Cognitive psychology takes a very

         17  mechanistic approach to understanding human

         18  cognition.  It tests mechanistic models of the

         19  human cognitive architecture, and it does this

         20  through primarily experimental task manipulation.

         21  This is also not a new psychology.  It's been

         22  around since before World War II.  There's 60 years

 

                                                              43

          1  of tons, gigabits, of experimental behavioral data

          2  that have developed not just the constructs that we

          3  use to test hypotheses, but the alterations in the

          4  human cognitive architecture in schizophrenia, but

          5  also a very sophisticated technology for measuring

          6  cognition in a connotative way.

          7             What's particularly compelling, I think,

          8  about using cognitive psychology and constructs and

          9  tools from cognitive psychology is that over the

         10  last 10 to 20 years, with the wide spread

         11  availability of non-invasive imaging, pretty much

         12  all of the major theories and approaches in not

         13  just cognitive psychology, but decision science and

         14  social psychology, have been subject to the test of

         15  whether they can fit it into what we understand

         16  about how the brain works, and increasingly, the

         17  human cognitive architecture has also become a

         18  neuro architecture.  That's something that's very

         19  appealing to those of us who are interested in

         20  developing drugs that affect the brain in a

         21  positive way when it comes to treatment and

         22  refractory aspects of schizophrenia.

 

                                                              44

          1             Another thing that's worth recognizing

          2  about experimental cognitive psychology--maybe just

          3  click three more times--is that this discipline

          4  doesn't necessarily assume that there is

          5  separability of domains across such areas as

          6  attention and language, working memory.  It's a

          7  highly mechanistic approach that is associated with

          8  a particular general circuitry that involves

          9  lateral or medial prefrontal cortex in the parietal

         10  regions, each of which contributes specific

         11  computations to these overall overarching cognitive

         12  control processes.

         13             Two clicks.  Go back.

         14             And in turn contributes to the

         15  regulation of a range of cognitive systems.  So it

         16  starts delaying working memory and sight memory,

         17  language processing.  Various levels of demand

         18  across these systems can engage this general set of

         19  control structures in different ways, and the key

         20  thing here is that by focusing on specific sets of

         21  computations associated with the components of this

         22  network, say, for example, the cognitive control

 

                                                              45

          1  functions that might be contributed by the dorsal

          2  lateral prefrontal cortex, it may be possible to

          3  develop treatments that will have positive effects

          4  across a whole range of what might be considered

          5  two separable domains, and this is likely to have

          6  highly robust effects on functional outcomes.

          7             Okay.  So those are the promises.

          8  Cognitive psychology can bring us tasks that

          9  provide a delay that is designed to allow us to

         10  investigate specific cognitive mechanisms tied to a

         11  particular task.  We can design versions to

         12  differentiate specific cognitive deficits from

         13  generalized performance deficits, and both of these

         14  approaches can increase our sensitivity in order to

         15  detect the affects of drugs in cognition in

         16  schizophrenia, but they don't allow--they don't

         17  assume cognitive domains.  So we can isolate the

         18  elementary mechanisms that will support possibly a

         19  range of domains in which patients might be

         20  affected and drug treatments that might have

         21  cascading affects across those domains.

         22             And last, but by no means the least,

 

                                                              46

          1  this is the psychology that is used in cognitive

          2  neuroscience.  This is the psychology that's

          3  largely used in developing animal models.  So if

          4  we're going to use these new tools, we will need to

          5  understand and take command of this particular

          6  process study in cognition in schizophrenia.

          7             Okay.  So those are some promises.

          8  There are great perils, of course.  For one thing,

          9  there are no standard tasks practiced.  So if you

         10  we're trying to develop a standardize approach to

         11  do this or a set of recommendations, you'd need to

         12  get people together to agree on a general range of

         13  standard tasks.  I actually don't think this would

         14  be a very hard thing to do.  This wouldn't take a

         15  great deal of work.  In fact, there are

         16  people--Trevor is one, and there are others who

         17  have developed ways of trying to developed

         18  standardized like tests that tap specific cognitive

         19  mechanisms under neuropsychology.  So I don't think

         20  this is particularly challenging.

         21             I think the biggest challenge is that we

         22  have almost no data on the psychometric properties

 

                                                              47

          1  of these tests.  People who do experimental

          2  cognitive studies do their studies.  They have

          3  relatively small numbers of subjects.  Unless the

          4  reviewers demand it, they tend to not do analyses

          5  of test reliability and stability and what have

          6  you.  So this will need to be taken on, and this

          7  probably wouldn't be the most exciting work for a

          8  cognitive psychologist, doing what would be

          9  essentially kind of normative and stability testing

         10  in large populations of patients; however,

         11  fortunately from MATRICS, there are precedents for

         12  doing this.  Even the standardized tests that were

         13  adopted by MATRICS have been subject to additional

         14  rigorous evaluation.  So it's doable and it has to

         15  be done before we can, I think, seriously consider

         16  that this would be a mainstream approach.

         17             So there is a set of challenges for

         18  cognitive psychopharmacology.  We need an

         19  agreed-upon set of domains, and I'm sure we'll do

         20  this in a few hours in the Jonathan, Deanna, and

         21  John Jonides's breakout group this afternoon,

         22  together with developing a set of more or less

 

                                                              48

          1  standard range of parameters.  We need for each of

          2  these domains to address this issue of specific

          3  generalized cognitive deficit.  I sort of my framed

          4  my opening of cognitive deficit in terms of the

          5  limitations of experimental cognitives, but

          6  actually--in terms of limitation of standardized

          7  tests, but, in fact, most of experimental cognitive

          8  studies don't address this issue either.  It is

          9  addressable within the framework of experimental

         10  cognitive approaches, but it's largely ignored by

         11  the field.

         12             Test, re-test reliability, and practice

         13  effects for administration need to be examined with

         14  measures that are developed from within this

         15  approach, and we clearly need psychometric studies

         16  in clinical populations.  I'm not sure how

         17  important it is--I put norming in here in healthy

         18  subjects, but actually we're really interested in

         19  cognition in schizophrenia.  I don't know that we

         20  need to care that much about norming these tasks

         21  across the general population.  If you're

         22  interested in aging and development and other

 

                                                              49

          1  things, then you certainly do.  I think that task

          2  was just to develop tools to be used in measuring

          3  change with treatment in cognition in

          4  schizophrenia.  I think the key task at hand is to

          5  do the psychometric evaluation of these measures.

          6             So moving on to the second set of

          7  measurement challenges, the idea that we can

          8  non-invasively measure brain activity, that we can

          9  actually directly target circuitry interest, so the

         10  notion would be what's the best model to detect

         11  cognition in schizophrenia, it would be cognition

         12  in schizophrenia.  We can study that directly with

         13  drugs that could change the circuitry.

         14             Go ahead.  However, there actually are

         15  lots and lots and lots of conceptual as well as

         16  methodological problems that I think have impeded

         17  progress so far, and again, in the breakout groups

         18  this afternoon, we're going to talk about these and

         19  hopefully come up with an agenda for some of these

         20  problems.  In addition to the issue of domain

         21  selection and task selection and psychometrics,

         22  which have to come first, there are inherent

 

                                                              50

          1  problems with BOLD imaging in particular.  BOLD

          2  imagining is not quite detective.  We can't measure

          3  the baseline brain activity using fMRI.  We can do

          4  it with other MRI-based methods that are being

          5  developed, and again, I think you get slightly

          6  different views on whether these approaches, such

          7  as arterial stimulation and have you are really

          8  prime time, but I think we can partially get

          9  quantitative signals that will give us baseline

         10  information, but we can't get, you know, fast whole

         11  brain signals at this point.

         12             Drugs may affect the coupling blood flow

         13  to the neuroactivity, so it enhances the basic

         14  premise here that we're imaging something that's

         15  related to blood flow and increases related to

         16  neuro metabolism, and drugs can change that

         17  coupling.  If they change that coupling, it's going

         18  to be very difficult to interpret the results of

         19  the pharmaco MRI studies.

         20             There are many issues within test

         21  reliability.  In fact, when I talk to people from

         22  industry, this is perhaps the biggest worry, and we

 

                                                              51

          1  have very little data about test reliability in

          2  schizophrenia.  There are some published studies.

          3  Those are small studies that are assuring, but

          4  there really, I think, no definitive data that

          5  establishes the current state of test reliability

          6  either at the box level or at the region level.

          7  There are some groups who are working on this, and

          8  it's very clear this is a very high priority.

          9             Finally, in the drug development

         10  process, time is of the essence.  So most studies

         11  that involve patients require a multi-center

         12  approach, and there are still some major issues

         13  with doing multi-center studies.  There are a

         14  couple of initiatives around the world.  The BYRD

         15  Institute, the BYRD initiative here in the U.S., is

         16  a multi-center fMRI study of first-episode

         17  schizophrenia that's underway in Germany.  People

         18  are developing solutions to this, but there is

         19  still actually quite a lot of work to do be done

         20  before we'll able to confidently acquire data at

         21  different sites on different schemes and different

         22  data formats and pool across sites, and this is

 

                                                              52

          1  clearly going to be something that needs to be

          2  addressed if we're going to use this tool in the

          3  process for these other tests.

          4             So we're going to spend the next day and

          5  a half talking about this and many other related

          6  issues.  We have breakout sessions this afternoon

          7  in fMRI, on ERPs, on cutting cognition at its

          8  joints, what are the key constructs, how do we go

          9  about pursuing it.  What I would like to throw up

         10  here is a very preliminary agenda.  Hopefully by

         11  the end of the next couple of days, we'll have a

         12  more definitive agenda.

         13             The first batch is to refine our

         14  understanding of cognitive deficits in

         15  schizophrenia in a mechanistic manner.  I think

         16  there's still quite a lot of work to be done here

         17  taking this cognitive psychological approach, as I

         18  talked about today.  The second is we need to take

         19  on the challenge of performance psychometric

         20  analysis, and this will require the testing of

         21  large numbers of subjects on multiple occasions

         22  with batteries of selected tests.  The challenge

 

                                                              53

          1  will be narrowing the batteries and agreeing on the

          2  premise for the tests and then actually doing very,

          3  very hard work that involves doing this kind of

          4  psychometric study in large populations of all

          5  characterized studies--patients.

          6             We need a systematic analysis of the

          7  sensitivity and reliability of methods such as

          8  fMRI, and this will probably need to be done on a

          9  relatively large scale, and in addition, something

         10  that I didn't talk about in the background, but

         11  something that I think we will need to continue to

         12  do, is to optimize our ability to combine data from

         13  non-invasive imaging studies.  We often hear about

         14  the potential that, you know, multi-host studies

         15  where co-registration is used to maximize the

         16  resolution of tests, such as EEG and MEG and

         17  spacial resolution of fMRI.  There are methods for

         18  doing this, but this is also an extremely complex

         19  problem, and there are many caveats to it, and I'm

         20  sure this is something that we'll also talk about

         21  in the breakout groups.

         22             So I'm going to finish now, and before I

 

                                                              54

          1  do, though, I think it's very important to

          2  acknowledge the work that has been accomplished so

          3  far in MATRICS.  I don't think anyone imagined that

          4  in two years we would come as far as we have and

          5  that so much would get done, and I think that the

          6  people in the program, such as Wayne and Bob and

          7  Bruce, also the group at UCLA in San Diego, really

          8  deserve commendation, because we have gone from a

          9  situation a couple of years ago where if you had

         10  talked to people in the industry, the likelihood of

         11  ever having a serious effort focused on improving

         12  cognition in schizophrenia seemed exceedingly dim.

         13  I think at this point, it's kind of to the realm of

         14  possibility, thanks to the work of these people.

         15             Thank you.

         16             [Applause.]

         17             DR. GEYER:  Thank you, Cam.  That was

         18  great.

         19             Our next and last presentation before

         20  the coffee break is from Alan Breier, and it's

         21  entitled "Needs and Approaches for Proof of Concept

         22  and Clinical Validation Studies for the Putative

 

                                                              55

          1  Treatments for Cognitive Deficits in

          2  Schizophrenia".

          3             Alan, it's a pleasure.

          4  VI.  NEEDS AND APPROACHES FOR PROOF OF CONCEPT AND

          5  CLINICAL VALIDATION STUDIES OF PUTATIVE TREATMENTS

          6        FOR COGNITIVE DEFICITS IN SCHIZOPHRENIA

          7             DR. BREIER:  Thank you, Mark.  It's a

          8  pleasure to be here.  I understand that this was

          9  one time a functioning chapel, and I'm not sure if

         10  it's fitting that we're having the last MATRICS

         11  meeting in such a religious setting or not, but I

         12  do understand that there are confessionals along

         13  the side that are still operative.  So if the

         14  breakout groups go over the line a little bit, you

         15  may want to cue up over there and get some things

         16  off your chest.

         17             It is really a great pleasure to be

         18  here.  I've got this long list.  This is like, you

         19  know, a retirement party, but I need to credit

         20  Wayne Fenton's vision and work on bringing this

         21  consortium together.  This is really, truly an

         22  extraordinary effort.  I have never seen anything

 

                                                              56

          1  like this before that delivered so much, who

          2  brought so many important groups that may have not

          3  necessarily worked together on a regular basis in

          4  this structure called MATRICS and then to deliver

          5  what's been delivered, I think is extraordinary,

          6  and Steve Marder and UCLA group and the others that

          7  have made this happen.

          8             So from a model perspective, I hope

          9  there is a real deep dive into what worked best

         10  about this process, what didn't work as well, and

         11  then continue this process on with other

         12  significant problems that affect our field, because

         13  I think it's really a great one.

         14             We'll go to the first one.  I think this

         15  just is sort of centering, at least for me, the

         16  next day and a half, this slide, in that there has

         17  been a pretty tremendous amount of progress in our

         18  field, but I think it would be an understatement to

         19  say that we still have a long ways to go

         20  particularly in terms of the impact illnesses like

         21  schizophrenia have on individuals and their ability

         22  to progress.

 

                                                              57

          1             The next slide, I guess to me, would

          2  sort capsulize what the problem is.  The problem is

          3  that after five decades of modern pharmacotherapy,

          4  schizophrenia functional outcomes remain poor.  So

          5  the work has been good.  The progress has been

          6  important.  The atypicals have contributed I think

          7  importantly, but the overall state of the state is

          8  not good, that the outcomes are still poor.  Only

          9  about 20 percent of patients work independently.

         10  All schizophrenic patients have cognitive deficits.

         11  It is, indeed, the core of the illness.  Maybe it

         12  defines the illness.  Maybe it defines the other

         13  symptom parameters of the illness, including

         14  hallucinations and delusions as being the absolute

         15  core of the illness and other things being

         16  secondary.

         17             Cognitive impairment appears to be a

         18  major determinant of the functional deficits.

         19  Current treatments have modest effects on

         20  cognition.  So that is the problem that we're

         21  really trying to solve.  At the end is improving

         22  the functional outcomes of patients and do that

 

                                                              58

          1  through this pathway or this road called cognitive

          2  impairment.

          3             Next.  Again, that the contributions of

          4  MATRICS is really extraordinary.  It brought key

          5  groups together, FDA, academia, NIH, industry and

          6  structural workers, the clear deliverables.  The

          7  Rand model was very, very important in terms of how

          8  you do this process of getting groups together with

          9  the discipline of coming to clear-cut decisions on

         10  tight time lines, etc., I think was quite good.

         11  This was really important, establishing cognitive

         12  impairment schizophrenia as a legitimate target to

         13  drug development, having Tom Logren and FDA in the

         14  room to hear about cognitive impairment

         15  schizophrenia to understand the science behind that

         16  and to appreciate that, indeed, it is a very

         17  important issue and is, indeed, a legitimate target

         18  for drug development with very, very important

         19  contributions and then creating this so-called

         20  clear path for registration.  From a drug

         21  development perspective, without that clarity of

         22  path, one doesn't go forward.  The expense and time

 

                                                              59

          1  is too great, frankly, and without the assurances

          2  that if one stays on the path and meets the

          3  milestones, that there will be a successful

          4  registration, one goes a different direction.  So

          5  getting agreement on what that pathway looks like,

          6  where the guardrails are, what are the key

          7  milestones was, again, very, very important.

          8             Having determined that co-primaries will

          9  be important, cognitive battery, we at Eli Lilly

         10  and Company will use that cognitive battery in our

         11  cognition trials, and then the functional outcome

         12  piece.  I'm going to actually come back a bit to

         13  this relationship in a moment, spend a few minutes

         14  talking about the cognitive functional

         15  relationship, understanding what the trial

         16  durations will look like, that there will be

         17  clinically stable patients in the residual phase,

         18  and then clarity around some key baseline

         19  parameters and inclusion criteria is also very,

         20  very important.

         21             This has really been a sticking point on

         22  so many areas, particularly in mental health

 

                                                              60

          1  research, the pseudospecificity, in getting good

          2  discussions and a really kind of critical look at

          3  what does this mean for cognition in schizophrenia.

          4  I think coming to the agreement that this issue can

          5  be adequately addressed in add-on treatments,

          6  adding an agent to an existing antipsychotic drug

          7  to improve cognition, will address

          8  pseudospecificity.  How we would develop

          9  monotherapies, though, I think is still unclear,

         10  and that still remains very, very important.  The

         11  path of choice in drug development is monotherapy.

         12  So I think that's going to require more thinking,

         13  more thought about how one might go forward with

         14  monotherapy in this area.

         15             I think one of the outshoots, then, of

         16  MATRICS is TURNS.  We've heard about that.

         17  Critical from an industry perspective is addressing

         18  the IP issues, and it looks like that is addressed

         19  or is being addressed in TURNS.  TURNS is a dream

         20  come true for small biotech companies who have

         21  smart people and are developing interesting

         22  molecules that don't have a clinical apparatus to

 

                                                              61

          1  test them.  With IP being addressed, TURNS will be

          2  a very desirable option to companies besides Eli

          3  Lilly and Company.  We have more molecules that we

          4  can develop with our own resources.  Forming

          5  partnerships and alliances in the development

          6  process will become very important.  A, from the

          7  practical aspects of getting the work done, but, B,

          8  particularly in this area in terms of the

          9  intellectual collaboration is going to be very,

         10  important.

         11             Most of the molecules, as I'll talk

         12  about in a little bit, that will go into TURNS will

         13  fail, and I think going into that, that sort of

         14  sobering reality is very important, and I'll show

         15  you the numbers, but the majority of these

         16  molecules will not end up becoming successful

         17  drugs.  So then it becomes very important to design

         18  these experiments, learning tools, so that a failed

         19  study on the primary is still vital information

         20  accrued for the next experiment so that it's a

         21  building process and those clinical trials that are

         22  built with adequate measures and tools that inform,

 

                                                              62

          1  and if it's a one-shot, maybe we'll get lucky with

          2  this one and we'll improve cognition and we'll go

          3  to Phase III, that's just not realistic.  It's

          4  likely for the vast majority involved just not to

          5  happen.

          6             So I think the accruing of information

          7  and knowledge, deepening the knowledge base as we

          8  go along with each experiment becomes critical.  We

          9  have had unfortunate experiences in my tenure at

         10  Lilly where we've gone to proof of concept studies

         11  with cognitive enhancers in substantial studies, a

         12  couple hundred schizophrenic patients, double

         13  blind, well controlled, had failed studies, but

         14  could not answer still fundamental questions, did

         15  the drug get to its target in brain, were the doses

         16  appropriate, did we underdose, overdose,

         17  fundamental questions like that so you complete a

         18  proof of concept study of a couple hundred

         19  patients, you don't separate on your cognitive

         20  measure, but if you don't have the information to

         21  design the next experiment, then you've really

         22  wasted a time lot of time and resources.  And

 

                                                              63

          1  again, unfortunately, most of the studies will not

          2  likely be positive on the primary end point if

          3  that's cognition.

          4             And this area, I'm going to really build

          5  on, and maybe it's one of the central messages of

          6  at least my presentation, at least from a drug

          7  development perspective, is the importance of

          8  biomarkers and improved animal models.  We use

          9  radial maze in our labs at Lilly.  That's pass go.

         10  So a drug needs to look good in an animal model of

         11  cognition.  We, frankly, have been disappointed in

         12  the predictive power of at least that test in

         13  humans.  So we feel we need to get to humans with

         14  the best data we can, but the test of the question

         15  happens at the clinic.  If we had more predictive

         16  animal models, animal models that would predict the

         17  likely occurrence in humans, it would revolutionize

         18  drug development in this area.

         19             So I can't underscore the importance of

         20  that.  We've spent a lot of time talking about

         21  biomarkers and what we mean by biomarkers and why

         22  are biomarkers so important, and I know we just had

 

                                                              64

          1  a brief conversation, Wayne, about potential

          2  initiatives on biomarkers.  I would strongly

          3  encourage that.  I would love to see a MATRICS II.

          4  I don't know what the acronym would be for

          5  biomarker development.  We have very substantial

          6  work going on in biomarker development, both in

          7  basic science and clinical, and need more partners,

          8  more smart people, more people working together on

          9  this problem, because as I'm going to show you,

         10  changing the equation of drug development where it

         11  really counts in this area, I think it's going to

         12  happen with biomarkers.

         13             Now I'll just spend a minute on this

         14  functional cognition relationship on the next slide

         15  and just tee-up a few questions that I think are

         16  still important, and this is not going to be the

         17  focus of the next day and a half, but I think

         18  they're still important questions.  One is does

         19  cognitive impairment warrant remediation

         20  irrespective of functional impact.  I understand

         21  the FDA's perspective in needing something more,

         22  wanting co-primaries, wanting to see that the

 

                                                              65

          1  change in a cognitive test has some real median

          2  impact on an individual, and so I understand that

          3  and I think that's important.  What that other

          4  measure will be, what that functional measure, that

          5  impact measure will be, I think requires a lot more

          6  thought and discussion.

          7             I will say from my clinical experience,

          8  having treated schizophrenia patients my entire

          9  career and continue to make rounds and see

         10  patients, I do see folks whose cognitive impairment

         11  alone leads to an enormous amount of subjective

         12  distress, and relieving that subjective distress of

         13  a muddled mind, irrespective of that ultimate

         14  impact on that person's ability to leave the unit

         15  or take on a job or something, I think is still

         16  very important and laudable.  So I think that there

         17  is a number of different facets to this.  This was

         18  discussed very thoroughly in MATRICS.  It's a key

         19  issue, and I think we're clear going forward about

         20  how we need to design our studies and the

         21  importance of the FDA's position.

         22             If cognitive improvement creates the

 

                                                              66

          1  readiness for functional gains, will psychosocial

          2  therapies require to show a functional effect?  If

          3  you have your neurons firing better, but you still

          4  stay at home and watch TV, how do you demonstrate

          5  that that correction in cognition translates into a

          6  functional gain?  So do we have to think about

          7  trials that have a very kind of robust component, a

          8  work rehab or some other form of psychosocial

          9  treatment so that we could really appreciate those

         10  benefits.

         11             This one maybe is just a bit provocative

         12  and outdated.  It isn't to kind of derail the

         13  focus, but if functional improvement is the golden

         14  ring, should drug development focus there?  Should

         15  we bypass cognition and get right to the heart of

         16  the matter, functioning?  When I think about drug

         17  development and the impact of a drug, showing a

         18  cognitive improvement in a schizophrenia patient

         19  will be very important, but show me a drug that can

         20  take that man on the first slide from a subway to a

         21  job, and you can grab the golden ring.  So while

         22  cognition looks like it's an important pathway,

 

                                                              67

          1  it's a bit of a surrogate.  Maybe we should be

          2  thinking about cognition as the surrogate end

          3  point, something we would use in early development,

          4  but the real test in Phase III maybe is cognition.

          5  So we go back to a single primary, and that's

          6  functional outcome measures, because that's really

          7  what counts and what's going to be most important,

          8  so again, just something more for thinking about.

          9             Given the early onset of schizophrenia,

         10  elements of adult functioning may not have been

         11  acquired, so how is its effect demonstrable on

         12  these functions?  And I don't think that's

         13  insurmountable, but for the very early onset, the

         14  very chronic patient, the patients who spend a good

         15  part of their lives in an institution, having a

         16  functional effect or functional impairment when

         17  you've never had it the first time is very

         18  different than in Alzheimer's disease where you

         19  were fully functioning, you've lost something, and

         20  then you've regained it.  And so it's a very

         21  different dynamic, and we've made a lot of

         22  parallels to the Alzheimer's clinical trials, and

 

                                                              68

          1  I'm going to show you some, because most of our

          2  cognitive agents start there with maybe Alzheimer's

          3  in mind and make their way down circuitous paths

          4  from there, but I think that's an important

          5  consideration.

          6             I'm going to come to this point on my

          7  last slide, but what also is going to be

          8  fundamental is a pretty robust educational effort

          9  with practicing clinicians.  You have a drug for

         10  cognitive effects.  How does the practicing

         11  clinician know that their patient has a cognitive

         12  impairment if they can't recognize it and measure

         13  it?  How do they know that the drug was successful

         14  if they can't map change?  So where cognitive

         15  assessment has been the domain of specialists, in

         16  order to bring drugs to market, it's going to need

         17  to be part and parcel of the way we train

         18  psychiatrists in residency programs to use these

         19  drugs; otherwise, you're not quite sure what

         20  patient to choose to select, and then after a

         21  six-month trial, how do you know to keep your

         22  patient on that drug if you can't really mark that

 

                                                              69

          1  their cognitive improvement has occurred.

          2             So I think that whole concept of getting

          3  that to the public domain as we have treatments

          4  available is something we can't get around.  I

          5  don't know how you can introduce a drug if the

          6  practicing clinicians writing the prescription

          7  can't identify it, measure it, know it responded,

          8  it didn't respond, etc., dose, to follow cognitive

          9  signal if they can't measure it and recognize it.

         10  So I think that's going to be important.

         11             Then, lastly, how is functional change

         12  assessed.  We looked into the literature as we're

         13  kind of gearing our studies with co-primaries with

         14  functional outcomes.  While there's very nice

         15  functional scales, there aren't too many scales

         16  that have been validated as change measures that

         17  hit all the important domains.  We, consequently,

         18  next one, have endeavored to form a collaboration,

         19  an academic collaboration to develop such a scale.

         20  So we're in the midst now of developing reliable

         21  and validated functional outcomes in schizophrenia,

         22  record actual performance over a relevant period of

 

                                                              70

          1  time, assess change over time, and focus on

          2  long-term non-acute change.

          3             Next, this is just a group of some of

          4  the academicians who have been helpful in terms of

          5  their consultation on what this scale should look

          6  like.  We had a vendor who is skilled at making and

          7  developing scales working with us, and we may be

          8  able to work with CATI for some validation work.

          9             Next, this is what it looks like.  We

         10  want it to be no longer than 30 minutes to

         11  administer, semi-structured with these four

         12  domains:  Living situations, whole functioning,

         13  basic and instrumental activity of daily living,

         14  and social and recreational.  And this is the

         15  development plan.  There will be a focus group for

         16  content validity this month, field tests to assess

         17  psychometric properties, 150 stable patients with

         18  informants.  In the winter, we hope to finalize the

         19  instrument, on the first quarter of '05, translate

         20  and validate.  So it's something to keep your eye

         21  out for.  What we will plan to do is for our

         22  cognitive studies with schizophrenia, we'll use the

 

                                                              71

          1  MATRICS battery for cognition and this scale for

          2  functioning, provided it meets all the requirements

          3  it needs to be a legitimate scale.

          4             Okay.  Let's move on.  Now, in terms of

          5  the development of functional cognitive enhancers,

          6  this can be very relevant to the group at TURNS and

          7  I think the groups working over the next day and a

          8  half, and I would say that the major challenge from

          9  the perspective that I sit in, in a drug

         10  development company, is improving the probability

         11  of technical success.  I would wrap that up in a

         12  net shell as saying that's going to be the big

         13  issue and the big challenge in order to bring these

         14  drugs to market.  Let me give you some background

         15  about this, next one.

         16             I think you've all seen slides similar

         17  to this, the growing cost it takes to develop a new

         18  drug.  This 800 to a million is actually probably

         19  low.  The more recent data suggests it's probably

         20  about 1.2 billion over about 15 years.  This one

         21  billion 15-year equation is not sustainable, at

         22  least not for my company.  We can't do that

 

                                                              72

          1  anymore.  We have to find ways to decrease this,

          2  change this equation, shorter times, less cost in

          3  order to develop drugs.

          4             Now, the next slide I think is really,

          5  really important for this group, because it talks

          6  about this attrition problem, and candidate

          7  selection is a very important point in the

          8  development of a drug.  An awful lot of work has

          9  been done up to that point.  I know the NIH is

         10  planning on taking on more of this early work in

         11  candidate development.  If I were advising--and no

         12  one has asked my advice, but if I were advising, I

         13  think that the NIH has been so critical in helping

         14  to define the right targets, understanding the

         15  biology, hopefully the development of clinical

         16  tools and biomarkers, that there is no drug on the

         17  market today that has not been an academic-industry

         18  collaboration, none.  The SSRIs would not be here

         19  today if it were not for Julius Axelrod and the

         20  work he did and people from his lab and going

         21  forward.  So that is always the case.

         22             What we do really well is the grind and

 

                                                              73

          1  the crank work once you've got the target and you

          2  need to find the molecule.  We're a medium-sized

          3  company.  We have over a thousand medicinal

          4  chemists just to do that work.  It's very

          5  laborious.  It's very expensive.  And it's

          6  something I think pharmaceutical companies do very

          7  well.  So in a division of labor and divide and

          8  conquer, I think working together on the targets,

          9  the biology, tools becomes very important, but this

         10  pre-candidate selection work has traditionally been

         11  the domain of pharmaceutical companies.

         12             But then look at the attrition rate.  So

         13  one out of ten molecules make it, nine don't,

         14  ninety percent of the molecules.  Once you already

         15  have a validated candidate, once you have a

         16  molecule that looks really strong and has passed

         17  most of the hurdles, only one out of ten will make

         18  it, and the attrition really occurs here.  It's

         19  this proof of concept, this early first human

         20  dosing exercise.  Our experience at Lilly in Phase

         21  III is actually very good.  Most of ours go to

         22  market.  Partly it's because we do so much work

 

                                                              74

          1  here.

          2             And why do drugs fail in this zone?

          3  What's the major reason for attrition?  Well, the

          4  big one is toxicology.  All drugs, new drugs, need

          5  extensive assessments from a toxicology

          6  perspective, including a two-year oncogenicity

          7  study.  We've had drugs die in month 20 when tumors

          8  emerge that are unpredictable.  So toxicology is

          9  one.  Another earlier on has to do with the issues

         10  that we've been talking about, and that is target

         11  validation.  The drug may look beautiful in the

         12  radial maze, it looks beautiful in the preclinical

         13  laboratory, but until you get it into the clinic,

         14  fundamental questions around did it get into brain,

         15  did it get to its target, what is the right dose

         16  become sounding like mundane issues, but are

         17  absolutely critical and fundamental.

         18             And then here, did they demonstrate

         19  efficacy, and those proof of concept studies,

         20  again, a hundred, two hundred patients, become

         21  very, very important, and that's where a lot of the

         22  attrition happens, and the more, let's say, less

 

                                                              75

          1  depth in understanding of the disease pathology,

          2  target validation, etc., will lead to more and more

          3  and more attrition.  I'll show you in a minute that

          4  the bad news is that CNS has the highest attrition

          5  problems of any therapeutic area, which is probably

          6  not a surprise, but it's a reality.

          7             Now, what do you do here?  What are we

          8  trying to do?  We accept the fact most of these

          9  candidates will fail, but what we want them to do

         10  is fail very early.  We call it fail-fast.  So we

         11  want this curve to be maybe sharp here, but flat

         12  out here.  As more and more resources are being put

         13  to bear, we want the answers and know which of the

         14  molecules of the many molecules that we can choose

         15  from that we want to take further.  So we want a

         16  full attrition as early into the process as

         17  possible.  If the only way you can determine if a

         18  candidate looks like it's a viable molecule is a

         19  large proof of concept study, you're failing late.

         20  So you want those biomarkers.  We're developing P

         21  parts for glucose modulation for diabetics,

         22  hemoglobin A1Cs, fasting glucose levels.  There's a

 

                                                              76

          1  myriad number of measures one can look at in the

          2  very first human studies and get extraordinary

          3  information so that you're able to fail fast, but

          4  when you don't have those biomarkers, when you

          5  don't have a glucose level hemoglobin A1C, simple

          6  blood pressures for other drugs, etc., and you're

          7  relying on clinical responsiveness to your main

          8  outcome measure, and it's a noisy outcome measure,

          9  you're failing late.  So you're putting an enormous

         10  amount of resources only to fail.

         11             So the more we have the--the better

         12  biomarkers we have, the better early indices we

         13  have that we can bring very early into the process

         14  and fail early, is very desirable and will begin to

         15  change this equation.  There will be more molecules

         16  than we have resources to clinically test, and when

         17  you don't have the assurances with the biomarkers

         18  and your real only test is a patient responding,

         19  you want to find creative ways, partnerships and

         20  ways because there's many of those in the clinic

         21  with the right experiments that will allow you to

         22  make a call one way or the other.

 

                                                              77

          1             Next, this just shows you that the CNS

          2  portfolio of drugs has the greatest amount of

          3  attrition of all.  Again, I don't think that's a

          4  surprise, but if the industry average is once a

          5  drug is ready for clinic is one in ten, CNS is

          6  probably one in twenty.  If you think about

          7  cognition in schizophrenia and the challenges

          8  there, it's upwards of one in twenty.  Maybe I

          9  don't know what it is.  Maybe it's somewhere

         10  between one in twenty and one in forty.  So that's

         11  what I meant, that returns to understand that

         12  reality that the likelihood of so many of these not

         13  making it is important going in so that the

         14  experiments are designed to glean information for

         15  the next experiment as opposed to a one-shot, let's

         16  hope we get lucky with this one, because the odds

         17  are stacked against us.

         18             Next.  Now, how are we going to improve

         19  the technical--probability of technical success?

         20  You know, the absolute fundamental key is advancing

         21  the understanding of the basic biology of

         22  cognition.  I think that goes without saying, that

 

                                                              78

          1  that's really what's going to be most important.

          2  Establishing better animal models, which we've

          3  talked about, and I think your animal models will

          4  be more perhaps elegant in measuring important

          5  cognitive programs.  What we in drug development

          6  need are, again, animal models that are going to be

          7  predictive of human experience.  So the more

          8  variance we can account for in the animal, the more

          9  we can change that attrition curve.  So if we have

         10  faith in our animal models, we're taking a lot less

         11  molecules to clinic and then we're quickly moving

         12  forward with the ones that look strong.

         13             But again, our experience with tests

         14  like radial maze are that we've not been impressed

         15  with their predictive abilities.  So better animal

         16  models will be very important.

         17             And then this area, develop improved

         18  biomarkers, the target validation, we are left now

         19  with a number of drugs, and I'll show you some

         20  examples, where we don't have good ways of

         21  validating that they're reaching their target in

         22  humans, and that just creates tremendous amounts of

 

                                                              79

          1  lack of precision in understanding the data, so the

          2  target validation, dose selection, surrogate end

          3  points.  Perhaps, again, there is a particularly

          4  good pre-pulse inhibition, some measure that could

          5  be brought very early that could be a surrogate

          6  that would give the one the confidence to design

          7  the Phase III trial and to move forward.

          8             And then population segmentation, I

          9  think is other one.  Population segmentation is the

         10  rule of the game in oncology.  You do not develop

         11  an oncology agent without a pharmacogenomic

         12  strategy to segment the population, and as we

         13  develop more and more understanding of our

         14  segments, having a homogeneous segment of groups

         15  that are more likely to respond is very useful and

         16  very important.  So thinking about how we would

         17  segment a population of schizophrenic patients for

         18  cognition trials that might be more homogenous,

         19  more likely to respond to Drug A versus Drug B

         20  could be very, very useful in the drug development

         21  process and increase the probability of technical

         22  success.  Perhaps it's a profile with cognitive

 

                                                              80

          1  measures or something of that nature.

          2             Okay.  This just speaks to the basic

          3  biology of cognition and how complicated the task

          4  is going to be, but how critical it is.  So

          5  understanding that basic biology and cognition

          6  becomes very, very important and very, very

          7  fundamental.

          8             Now let's go to the next one.  What I

          9  thought I would do now in the remaining part of my

         10  talk is just talk about some of our experience at

         11  Eli Lilly and Company.  I'll take two molecules on

         12  two different platforms, a glutomatergic molecule

         13  and a muscarinic, two that were targeted early as

         14  potential cognitive enhancers, and take you through

         15  some of the work to give you an idea of some of the

         16  mistakes we make, some of the challenges that come

         17  up so that hopefully it will better inform some of

         18  your thinking as you go forward.

         19             So first let's talk about the

         20  glutamatergic platform.  Glutamate is the major--in

         21  the nervous system, involves many physiological

         22  processes in cognition.  How can we selectively

 

                                                              81

          1  modulate glutamatergic synapses transmission in

          2  CNS?  Can we selectively target the pathological

          3  process involving the nervous system?  Are the

          4  glutamatergic modulates candidates for cognitive

          5  enhancement?

          6             I think Wayne didn't say it, but in his

          7  presentation, I'll just take it a little bit

          8  further.  We probably don't need any more me-too

          9  drugs.  What we probably need are very novel drugs.

         10  So we want to look at platforms that are more novel

         11  if we really want to take that man from the subway

         12  station to independent living.  We're probably not

         13  going to get there through tweaking the old

         14  molecule.  So that means looking at new targets.

         15  The sobering part, again when you're looking at the

         16  targets, is your attrition goes up, and that just

         17  goes with the category.  When you take on a brand

         18  new area where there is very little work done in

         19  glutamatergic drugs, it's just harder and more

         20  difficult work.  You need more people working on

         21  the problem.

         22             Next, these are the clone glutamate

 

                                                              82

          1  receptors and transporters.  All of these are

          2  potential drug targets.  Our verb is "drugable",

          3  can we drug these targets, and the answer, by and

          4  large, is yes.  So every one of these transporters

          5  and receptor targets are potential candidates for a

          6  drug, and with our relatively primitive

          7  understanding of how these systems work in

          8  cognition, it would be difficult to even rule any

          9  of them out right now as a potential.  So what do

         10  you do from there?  How do you take so many

         11  different molecules; which ones can we decide to

         12  take forward to the clinic?  When we decide on a

         13  target, we automatically start developing back-ups

         14  of those molecules, because we know that there will

         15  be bio availability issues, toxicology issues,

         16  efficacy problems.  So when we embark on, say, a

         17  receptor site, we'll begin with one molecule, but

         18  then we'll already begin development of the second,

         19  third, and fourth, and all of the drugs in CNF that

         20  I'm aware of have gone that route, so that you've

         21  got backups so that if one may look like a total

         22  failure in the clinic, it doesn't mean that that

 

                                                              83

          1  target necessarily is not valid.  You have another

          2  backup to consider depending on how you want to use

          3  the resources.

          4             Next, so this is just an example of

          5  drugs in our labs and in other labs for these

          6  multitude of different glutamatergic sites, all,

          7  again, potentials for--most of them, anyway, I

          8  think have reasonable backgrounds that one might at

          9  least want to consider them in a cognitive

         10  enhancing paradigm, and when you're this early on,

         11  you're not thinking about cognition in

         12  schizophrenia.  You're thinking about cognition

         13  broadly.  Alzheimer's may be at the top of your

         14  list, and I'll show you one of our development

         15  programs that went that route, but so early on,

         16  you're thinking much more broadly, and then as your

         17  experiments go forward, you're starting to narrow

         18  your focus.

         19             I thought it might be instructive to

         20  take one of these and look at some work that we've

         21  done, particularly from a proof of concept

         22  perspective, on AMPAs and drugs that develop as

 

                                                              84

          1  so-called AMPA kinds or drugs that potentiate

          2  effects at the AMPA receptor.

          3             Next, AMPA receptors in CMS are key

          4  players to plasticity, the ability to learn and

          5  remember, capacity to recognize with injury.  The

          6  basic mechanism underlying plasticity is

          7  neurogenesis, activity depending on synapses,

          8  sprouting of synapses, and pruning of synapses.

          9  There's pretty strong data that AMPA plays a role

         10  in three out of four.  So it's a target for

         11  plasticity, for learning and memory.  AMPA is not a

         12  bad choice from a basic science perspective.  Basic

         13  mechanisms support brain plasticity, and then it

         14  has applications to treat cognitive disorders and

         15  neurodegenerative disorders.

         16             Let me show you some of our work.  Here

         17  is one of our AMPAs.  These are very highly potent

         18  compound, much more potent than those in some of

         19  Gary Lynch's earlier compounds.  Here, you see AMPA

         20  and then the AMPA kind being administered in

         21  enhancing the signal as we go forward.

         22             Next, this is from a rodent model of

 

                                                              85

          1  Parkinson's disease showing neuroprotective effects

          2  of this particular AMPA.  Here is the in tact road,

          3  a wedgy one, and then we see ones pretreated with

          4  the AMPA kind in tact and then the lesion

          5  preservation of neurons, demonstrating nice effects

          6  there.

          7             So, you know, the data builds.  It's

          8  looking interesting.  I'll go to the next step.

          9  We've already talked about the radial maze, but

         10  here is one of the AMPAs demonstrating nice

         11  improvement on the radial maze, and you can look at

         12  it in comparison with other potential drugs as

         13  well.  So at this point, it's looking like we have

         14  a candidate to consider moving into the clinic.

         15  But what do we don't know?  What are the things

         16  that we'd really like to know about this or other

         17  AMPAs in order to do this experiment?

         18             Next, and that then really gets into

         19  this whole idea of developing the appropriate

         20  biomarker.  Now, there are biomarkers that are

         21  disease related, disease progression related.

         22  Those are very exciting and very desirable.  We

 

                                                              86

          1  probably have a paucity of those because we don't

          2  have a deep understanding of our disease pathology

          3  to have a biomarker of the progression of the

          4  disease, but I think in terms of drug development,

          5  biomarkers for pharmacological response are very,

          6  very important and very valid to the work that we

          7  would do.

          8             So what we would clearly like in this

          9  project would be to have the tracer, a PET tracer,

         10  for AMPA receptors.  If we have a PET tracer for

         11  AMPA receptors, we would know that the drug is

         12  getting into the brain.  It's getting to its

         13  target, and then depending on how good the tracer

         14  is, we could probably ascertain dosing information,

         15  and then with that information design the right

         16  proof of concept study.  This has worked very

         17  nicely with the NK1 antagonist, first looked at in

         18  depression, some exciting information from Merck

         19  and then failed studies and a decision not to

         20  pursue an NK1 antagonist for depression.

         21             Some of newer work in anxiety, however,

         22  looks very, very promising.  Now that the tools are

 

                                                              87

          1  there, the work that we've done with collaborative

          2  work that Merck has done, we have a reliable way to

          3  test drug occupancy of NK1 receptors.  So with a

          4  PET study now, we can come pretty darn close to

          5  answering those important questions, even picking

          6  the dose to go into the trial as opposed to not

          7  knowing and having three or four or five or six

          8  treatment arms, each with a different dose in order

          9  to try to ascertain the dose.  So that was sort of

         10  the old way, very expensive, very laborious, but

         11  using an effective productive biomarker.

         12             So we put a lot of energy with

         13  collaborators into developing a PET legate for AMPA

         14  receptors and have been successful.  In fact, most

         15  of work we've done in developing novel

         16  tracers--again, we're looking not for me-too drugs,

         17  but novel targets.  That means developing novel

         18  tracers--has been difficult.  NK1 is a success

         19  story, but we've not been successful at the M-1

         20  agonist and others.  We have not been successful at

         21  this point with AMPA.

         22             So what do you do now?  You don't have a

 

                                                              88

          1  target validator; what can you do?  Well, you go to

          2  another approach, and what we decided to do was to

          3  look at energy expenditure or metabolism.

          4             Next slide.  Now, it's less desirable.

          5  It may tell you that your drug is getting into

          6  brain, but to ascertain more information from, say,

          7  an FDG study around dosing, around target

          8  validation, it's just less direct and therefore

          9  less helpful, but not not helpful.  So if you don't

         10  have a tracer for your receptor or your target

         11  site, then fMRI is a good approach.  Now, some of

         12  the imagers in the crowd may not agree with this,

         13  but resting metabolism in a healthy individual is

         14  fairly high.  If you're looking at agonist, you're

         15  then asking to show elevation on a signal that's

         16  already helpful.  That's hard to do.  It's easier

         17  to show a detriment or a pathological state to show

         18  an increase, shown here with lung cancer and brain

         19  mets.

         20             So a strategy that we thought about with

         21  the AMPA drug would be to lower metabolism and then

         22  show an effect on top of that, and one fairly

 

                                                              89

          1  reliable way to lower metabolism is alcohol, and

          2  we've also been interested, just kind of

          3  coincidentally, that AMPA receptors seem to have an

          4  interesting effect on alcohol intoxication, not

          5  working through the AMPA receptor.

          6             So using that, we designed a strategy,

          7  next, to look at fMRI to detect the

          8  neuropsychological tests, and this was basically

          9  the design.  Again, when you're thinking about

         10  TURNS and thinking about studies you may want to

         11  do, these are just some design ideas that we've

         12  worked on and worked through:  Double-blind

         13  cross-over with subject with two periods, two

         14  conditions per period, one with washout between

         15  periods.  The two conditions would be saline and

         16  alcohol infusion, twelve volunteers.  Here is the

         17  study drug, and the dose that we would then try,

         18  and then your imaging parameters throughout, and

         19  we're doing this study at IU, in our backyard,

         20  Indiana University.

         21             Next, this just shows you some of the

         22  fMRI data from this experiment, and I think an

 

                                                              90

          1  experiment like this can, again, tell you that the

          2  drug is getting to the brain.  Before doing

          3  experiments like this, as I noted before, we would

          4  go forward with a drug like this, not even knowing

          5  the answer to that question.  Then, if you get your

          6  negative effect or no effect, you're wondering how

          7  much got into the brain, did any get into brain,

          8  etc., etc.  So this is important to at least answer

          9  that question.

         10             Next, let me take you through our

         11  experience with a muscarinic drug and muscarinic

         12  agonists, and again, just for more illustrative

         13  purposes, this is a long story at Eli Lilly and

         14  Company with the molecule, but not nearly atypical.

         15  Drugs look horrible and are put on the back shelf.

         16  They're resurrected and put on the back shelf.

         17  They're resurrected, and almost all drugs in the

         18  market today have gone that path, and this is one

         19  that went this path.  This is a drug called

         20  xanomeline.  It's a muscarinic agonist with

         21  relative selectivity to the M-1 and the M-4

         22  receptor, first evaluated in Alzheimer's disease

 

                                                              91

          1  and later in schizophrenia, and I think many drugs

          2  will probably go this path.

          3             Next, this just shows you that the

          4  preclinical work normally looks good.  Here is,

          5  again, radial maze data showing that there was some

          6  cognitive improvement on this test in the rodent.

          7  So that was encouraging.  The decision then was to

          8  go forward with the studies in Alzheimer's.

          9             Next, this is sort of a fairly

         10  traditional Phase III design for an Alzheimer's

         11  test, 343 out-patients, 17 sites, six months

         12  followed by a one-month washout period and

         13  co-primaries, which is the rule of thumb in

         14  Alzheimer's, ADAS-COG for cognition, civic plus and

         15  some other indicator of impact on the person's life

         16  with other measures as well.

         17             Next, this just showed that on ADAS-COG,

         18  it normally would look really good.  It would be a

         19  good cognitive enhancer.  The data was very, very

         20  encouraging from a cognitive perspective.  Next, on

         21  civic plus, same thing.  Here is the washout period

         22  here, six months.  So from an efficacy perspective

 

                                                              92

          1  for Alzheimer's disease, this drug looked

          2  promising.

          3             Next, and interestingly, and this is why

          4  it got resurrected later, it was quite good on a

          5  subset of symptomatology that included delusions

          6  and hallucinations.  So not only was it looking

          7  good on the cognition, but it also looked good on

          8  delusions and hallucinations.  So that caught the

          9  eye of the investigative team, but what happened

         10  were side effects that really prohibited its

         11  marketability, its commercialization in Alzheimer's

         12  disease, particularly GI side effects, vomiting.

         13  So that was a no-go.  That was not going to be a

         14  viable candidate for Alzheimer's disease.  It went

         15  to the back shelf.  As we started thinking more

         16  about cognition and the field evolved, we did more

         17  work with muscarinic sites.  We started saying

         18  maybe we ought to dust off xanomeline and take

         19  another look.

         20             So we did some more work, next, and this

         21  is pre-pulse inhibition work, thinking that this

         22  could be maybe a better predictive model for human

 

                                                              93

          1  experiments in schizophrenia, and it shows here

          2  with apomorphine, the effects on pre-pulse and then

          3  the xanomeline counteracting the apomorphine

          4  effects shown here.  There are some nice effects in

          5  that model, and we thought that it would be

          6  reasonable or interesting to look at in

          7  schizophrenia.

          8             Next, so we designed a very small proof

          9  of concept study.  Because it had effects on

         10  delusions and hallucinations, and we were kind of

         11  interested in testing the hypothesis, could a drug

         12  with no D-2 effects whatsoever have a dual effect

         13  or perhaps monotherapy for the core symptoms of

         14  delusion and hallucination and also cognition.  So

         15  that was the first study, double blind, placebo

         16  controlled, very small sample, also done at Indiana

         17  University.

         18             Next, and we saw some promising effects

         19  on cognition.  It was a comprehensive battery.  The

         20  sample sizes were very small, ten and ten or

         21  something like that, but at least there was

         22  something there that looked encouraging.

 

                                                              94

          1             Next, we also some saw some very early

          2  signs that maybe there could be some symptom

          3  response.  Now, I think this underscores the

          4  pseudospecificity problem that a mono therapy is

          5  going to have, particularly in a design where you

          6  have washouts and placebo controls.  That is going

          7  to be very hard to untangle the delusions and

          8  hallucinations effect from the cognitive effect.

          9  So I think that kind of underlines this particular

         10  problem.  It seemed to be very better tolerated in

         11  schizophrenic patients.  They're younger.  I don't

         12  know if that has anything to do with it, but next

         13  where we would go if we were going to continue this

         14  or other M-1 agonists would be as a primary

         15  consideration, an add-on to a typical cognitive

         16  impairment using a MATRICS guidelines, use stable

         17  residual-based patients, use the MATRICS cognitive

         18  battery as a primary end point and perhaps a

         19  functional scale, and one of the nice things about

         20  some of the atypicals, like olanzapine, is they're

         21  excellent anti-amidics.  So the concern around the

         22  vomiting that one saw in the elderly Alzheimer's

 

                                                              95

          1  patients could potentially be mitigated with an

          2  atypical.  So this could be a nice match where some

          3  side effects are mitigated.  One could have stable

          4  patients, residual patients, and then look more

          5  squarely at cognitive improvement.

          6             But we were also encouraged, at least

          7  the data was very preliminary, but perhaps this

          8  could be a target for a new kind of monotherapy for

          9  an anti-psychotic drug, and we're pursuing that

         10  concept as well.

         11             So let me close with this slide, areas

         12  for future academic industry, NIH collaborations,

         13  and animal models for cognition.  I can't

         14  understate the importance there.  I'm very

         15  intrigued by the possibility of NIH activities in

         16  the biomarker marker area, again that pulling that

         17  attrition forward, having experiments that will

         18  tell us what we need to know before designing a

         19  large Phase III trial, very, very important,

         20  clinical trial networks for sure, but also later

         21  trials.  There are good examples in cardiology and

         22  oncology of large academic networks that work very

 

                                                              96

          1  closely with pharmaceutical companies.  ME is an

          2  organization out of Harvard that we work

          3  extensively with in cardiovascular drugs in Phase

          4  III, and so I think there's lots of other

          5  opportunities.  And then, lastly, the point I

          6  raised earlier is that we're going to need

          7  education and training for practicing clinicians on

          8  how to detect and assess change in cognitive

          9  deficits.

         10             Thanks very much for your attention.

         11             DR. GEYER:  Thank you, Alan.

         12             We have time for a coffee break, and if

         13  we can be sure to be starting promptly at 10:30.

         14  That clock is actually slow, so 10:25 on that clock

         15  would be good.  It will save time for lunch in

         16  order to get to the breakout sessions.

         17             If you have not signed up for the

         18  breakout sessions, please do so at this time.

         19             [Recess.]

         20             DR. GEYER:  I have three administrative

         21  details to apprise you of.  One is that when

         22  you--if you didn't print it out from your E-mail,

 

                                                              97

          1  when you arrived, you may have picked up a document

          2  of about 11 pages that describes in more detail the

          3  content and the bullet points about the various

          4  breakout groups.  There was one batch of maybe 20

          5  or 30 of those that had only every other page

          6  copied.  So they have now been thrown away and

          7  replaced with complete copies, but the remaining 15

          8  or 20 of them are still floating around.  If you

          9  have one that skips every other breakout group or

         10  has every other page, you might want to seek a

         11  replacement for that.

         12             On that document is also, together with

         13  the bullet points developed by the participants of

         14  each breakout group, the actual location, room

         15  location, of each of the breakout groups.  So when

         16  this session ends, we will then all go to lunch in

         17  the Bolger Center, and if you're not on the Bolger

         18  Center meal plan associated with your room, then it

         19  would be advisable to purchase a lunch ticket from

         20  Tanya's desk out there at the reception.  She has

         21  lunch tickets for sale with receipts, and so it

         22  would be advisable to get that rather than do that

 

                                                              98

          1  at the luncheon center at the Bolger cafeteria.

          2             And also, I would like to request that

          3  all of the leaders and co-leaders and reporters

          4  from the eight breakout groups spend three to five

          5  minutes with Bob Heinssen and me up here at the

          6  front of the room right at the end of this session,

          7  prior to lunch, just so we're all on the same page

          8  about how to run your breakout group and the fact

          9  that you would will need to condense from your

         10  breakout group at least the three highest key

         11  priority areas.  The major product of your breakout

         12  group effort is to identify the three highest

         13  priority research areas from your domain of

         14  interest and orchestrate how you're going to

         15  present those to the group in tomorrow morning's

         16  session.

         17             I think that's all of my administrative

         18  duties for now, and I'll try to save enough time

         19  for Dr. John Jonides.  It's a pleasure to have you.

         20  I'm not going to read this title, because I think

         21  you have a better title of your own.

         22     VII.  ADVANCING CLINICAL SCIENCE THROUGH THE

 

                                                              99

          1   TRANSLATION OF COGNITIVE NEUROSCIENCE THEORY AND

          2                      METHODOLOGY

          3             DR. JONIDES:  I have actually changed

          4  the title from the one that you have in your

          5  program, but it connotes the same material.  I

          6  consider myself an amendment, an addendum to Cam

          7  Carter.  So I'm going to present myself that way

          8  for the next half hour or so, because Cam has done

          9  a wonderful job of setting up the single issue that

         10  I'd like to address in this half hour's session,

         11  and that is a relative comparison between the kinds

         12  of complex instruments that are used in

         13  neuropsychological batteries traditionally and what

         14  we might call the more refined cognitive

         15  instruments that are available today and that might

         16  serve some purpose if they were further developed

         17  for use in batteries that can be used to evaluate

         18  compounds in the treatment of various diseases,

         19  schizophrenia of course among them.

         20             So let me start with dessert first.  I

         21  want to give you the final message of my talk and

         22  then kind of trace you through what is going to be

 

                                                             100

          1  the substance, which is three examples that I'm

          2  going to march through in some detail.  What I am

          3  essentially doing here, by being the addendum to

          4  Cam Carter, is to take this message and to move it

          5  down one level of specificity so that you can see

          6  some of the innards of the tasks that might be more

          7  refined examples of cognitive tasks, not the only

          8  ones by any means.  I had an opportunity here just

          9  to bring forth three examples that might be of some

         10  interest.

         11             The general point, which is the one that

         12  Cam made quite nicely, is that many psychological

         13  functions are complex mixes of component processes.

         14  Long-term memory is certainly one example of that.

         15  Cam had talked about some others as well, in

         16  particular in some detail the Wisconsin Card

         17  Sorting task, and I want to remind you that that

         18  task is one of my examples.

         19             Standard neuropsychological instruments

         20  are often too coarse to penetrate what some of the

         21  component processes are of these complex

         22  psychological tasks like long-term memory.  For

 

                                                             101

          1  example, you can even at a coarse level of

          2  description consider separate processes of

          3  encoding, storage, and retrieval of information

          4  from long-term memory already recognizing that

          5  long-term memory is a single entity measured by one

          6  number, say a percent, correct.  It's just not

          7  going to be captured very well because there are

          8  subprocesses of some importance.

          9             What we need, of course, is theories

         10  that parse what these component processes are,

         11  theories, of course, including computational

         12  theories that are precise in their specification of

         13  what those processes are, and then following on

         14  that, we need assessment instruments that isolate

         15  these processes for further measurement, using

         16  refined behavior tasks and now using the new

         17  technology that Cam introduced us to before, using

         18  neuroimaging and also lesion evidence, including

         19  temporary lesions performed by techniques like

         20  transcranial magnetic stimulation in order to

         21  penetrate what some of these isolated processes

         22  are.

 

                                                             102

          1             So what I would like to devote the

          2  entire rest of this talk to is going through three

          3  examples, two of which I culled from the

          4  literature, one of which I culled from my lab, to

          5  show you examples of complex psychological

          6  processes and how we're making some progress in

          7  understanding them in some greater detail by

          8  looking at some of the individual processes that

          9  make them up.

         10             Let me just start by motivating the

         11  search for isolated processes as opposed to looking

         12  at more complex measures.  So if you find a

         13  deficit--I think Cam, again, had made this point.

         14  If you find a deficit on the gross test of

         15  performance, the Wisconsin Card Sorting Task being

         16  the one that Cam described, what this doesn't

         17  specify is the individual processes that are

         18  down-regulated as a function of this deficit.  So

         19  what you need is more analytic instruments that

         20  will then target the specific processes that might

         21  be affected by whatever the change agent is.

         22             So testing drug interventions, for

 

                                                             103

          1  example, is best done with knowledge of these

          2  specific processes in order to find out what the

          3  drug agent is mitigating specifically and without

          4  effects being swamped by other processes because

          5  any task like the Wisconsin Card Sort in the hands

          6  of Cam as composed of some seven or eight processes

          7  and in the hands of others by even more, if you're

          8  isolating any one of two or those, the effects of

          9  those could be swamped by the effects of other

         10  processes in the task of interest.

         11             Another reason to motivate the isolation

         12  processes is to make better contact with tasks that

         13  are used for animal models, because the very best

         14  animal model, the very best tasks used with animal

         15  models, are ones that come from your more isolated

         16  processes that we tend to study in the cognitive

         17  laboratory, and vise versa, that is we have made a

         18  great deal of mileage in cognitive neuroscience by

         19  taking exquisite cognitive research.  In my own lab

         20  from the--for example and picking out the essential

         21  aspects of those tasks in order to study processes

         22  involved in things like working memory.

 

                                                             104

          1             So let me start marching through some of

          2  the examples, a little bit of introduction about

          3  executive functioning, a little bit of a reminder

          4  about executive functioning.  We know, of course,

          5  that there is no conceptual definition of what

          6  executive functions are.  Instead, there are

          7  various lists of them.  Here is someone's list--I

          8  called it mine because it's my slide--including a

          9  number of processing like focusing attention,

         10  shifting attention between stimuli, managing

         11  multiple tasks in which you're engaged, monitoring

         12  ongoing performance to be sure that performance is

         13  meeting the goal that you had set up, detecting

         14  conflict between internal processes that are

         15  ongoing or between responses that you might be

         16  seduced into making that may be correct in the

         17  present context, and, finally, inhibiting

         18  irrelevant information internally or irrelevant

         19  responses and irrelevant stimuli in the outside

         20  world.

         21             So if executive functions are that

         22  complex, that is if they're composed of a number of

 

                                                             105

          1  processes of that sort, that helps explain why some

          2  of the tests that had very good sensitivity for

          3  executive dysfunction, things like the Wisconsin

          4  Card Sorting Task and trail-making A and B and

          5  fluency tasks and others, those tasks have very

          6  good diagnostic specificity--diagnostic

          7  sensitivity.  They had diagnostic sensitivity

          8  because what they're measuring is a collection of

          9  processes, some or all of which might be impeded by

         10  whatever the deficit is that's under consideration.

         11             The down side of that sensitivity,

         12  however, is that the tests each have poor

         13  specificity, that is there are component processes

         14  that make up each of these tests, and the analysis

         15  of these component processes is thin with these

         16  complex instruments because the instruments are

         17  complex, as I've asserted.

         18             So the example that Cam introduced you

         19  to was the Wisconsin Card Sorting Task in which

         20  patients or subjects are given a deck of cards and

         21  then have to sort them into a number of piles

         22  according to some criterion, and unbeknownst to

 

                                                             106

          1  them, the criterion changes at some point and they

          2  have to change their sorting.  As Cam described on

          3  the slide that he showed you, the Wisconsin Card

          4  Sorting Task is composed of a number of individual

          5  processes having to do with shifting attention,

          6  receiving and analyzing negative feedback or

          7  positive feedback, switching processes, working

          8  memory processes that hold one set of rules in mind

          9  and then reinitialize those with another set of

         10  rules in mind and so forth.

         11             Any analysis of some of--or reasons that

         12  you might have executive dysfunction that relies on

         13  the Wisconsin Card Sorting Task is, of course, not

         14  going to be particularly analytic about which one

         15  or ones of those processes is imperiled by some

         16  manipulation, and so various investigators have

         17  tried penetrating the Wisconsin Card Sorting Task

         18  to see whether we can analyze in some more refined

         19  detail what some of these processes are and,

         20  importantly, what the brain mechanisms are that are

         21  affected by those processes.

         22             So I brought along just one example from

 

                                                             107

          1  the literature of a computerized version of the

          2  Wisconsin Card Sorting Task that separates out

          3  various phases of the task when subjects are doing

          4  it and then compares them to a control task,

          5  various phases in a control task, that the subjects

          6  in this experiment are also doing.  The task is

          7  really modelled on the Wisconsin in a fairly

          8  straightforward way.  The subjects receive four

          9  alternative piles, as it were, electronically on a

         10  computer screen and they're given one card to sort

         11  into one of those piles.  In the Wisconsin version

         12  of this task, the sorting criteria are as they are

         13  in the standard WCST.

         14             In the controlled version of the task,

         15  the subjects are simply matching a card against the

         16  one item in the pile that matches it exactly, and

         17  what you can do now is isolate in this rarified

         18  version of the WCST individual processes that might

         19  be important to the overall performance of the

         20  task.  For example, one aspect of the Wisconsin

         21  Card Sorting Task is that occasionally subjects

         22  receive negative feedback.  You're told that that

 

                                                             108

          1  sorting pile is wrong and now you have to do

          2  something about it.  So there could be brain

          3  activations and individual psychological processes

          4  involved in receiving and evaluating that negative

          5  feedback.

          6             There is, of course, performance that

          7  has to follow that negative feedback on the next

          8  pile.  So now that I've been told that I was wrong,

          9  now what do I do?  I need to switch to some other

         10  rule or try another hypothesis, do a sorting based

         11  on the different criteria and see how it goes from

         12  now on.  And here again, you can isolate out

         13  performance on the next trial compared to the epic

         14  where you're actually receiving the negative

         15  feedback.  Also, you can examine what happens when

         16  subjects receive positive feedback and what happens

         17  when they're engaged in matching following the

         18  positive feedback.

         19             So here is a case, here is a model task

         20  that I think nicely picks apart the more complex

         21  Wisconsin Card Sorting Task and asks the question

         22  can we identify separable psychological processes

 

                                                             109

          1  and also separate neuro networks that might be

          2  implicated in various phases of the task.  In this

          3  experiment, they did very nicely, I think, isolate

          4  out various components of the task, and I brought

          5  only a small piece of the data along.  There are

          6  activations in dorsal lateral and ventral lateral

          7  prefrontal cortex and other structures as well,

          8  basal ganglia and thalamus, that are involved in

          9  responses to the negative feedback that subjects

         10  receive in the task, suggesting sites that are

         11  sensitive to the presence of conflict, that is you

         12  made a response that you thought was correct;

         13  you've now been told that it's incorrect and you

         14  have to do something about, or shifting attention

         15  involving some cortical, slash, subcortical.

         16             Positive feedback, by contract, when you

         17  receive positive feedback in the rarified task, it

         18  activates a very different sort of cortical network

         19  that's not shown on this slide.  It's a network

         20  involving largely the structures that are engaged

         21  when people are involved in verbal and working

         22  memory tasks, and you can imagine how that would

 

                                                             110

          1  play out.  Your subject can think the appropriate

          2  sorting criteria is color.  The experimenter tells

          3  you, yeah, that response was right, color, and so

          4  now you try to keep in mind color is the

          5  appropriate sorting criteria, and you're probably

          6  using verbal strategies to do that.  And there are

          7  imaging data that I haven't shown you on this slide

          8  that isolate out, I think quite nicely, the effects

          9  of positive feedback from the effects of negative

         10  feedback in a way that you never see if you just

         11  looked coarsely at the overall performance in the

         12  Wisconsin Card Sorting Task, whether behaviorally

         13  or in the context of a scanning environment as

         14  well.

         15             Some of my colleagues have gone on to

         16  even pick apart some of these processes even

         17  further.  I have done a metananalysis with a former

         18  student of mine, Tor Wagers now at Columbia, where

         19  we have isolated out processes that are involved in

         20  the switching operation that's involved in the

         21  Wisconsin Card Sorting Task where you have to go

         22  from one sorting criteria to another sorting

 

                                                             111

          1  criteria, asking the question do various kinds of

          2  switching operations hang together as a piece, that

          3  is if you're switching among attributes of the

          4  stimulus or switching from one stimulus to another

          5  or switching from one task to another, is there any

          6  coherence in the patterns of activation that you

          7  get in the brain among various switching

          8  operations, and now there have amassed in the

          9  literature some 40 to 60 imaging studies that allow

         10  us to take a look to see whether there is any

         11  concordance among those, and, happily, there is

         12  concordance, and I brought along just one panel to

         13  show you this:  There are activations in dorsal

         14  lateral and also ventricle lateral prefrontal

         15  cortex that appear as a function of switching

         16  operations regardless of what it is that is the

         17  domain of the switching, suggesting that the

         18  Wisconsin task which has a certain kind of

         19  switching in it might generalize to other sorts of

         20  situations in which you're shifting your attention

         21  from one thing to another.

         22             So this could, once again, be a model of

 

                                                             112

          1  trying to pick apart processes and isolate out

          2  attention-shifting components that might then be

          3  the subject for some kind of intervention.  In

          4  fact, the story gets deeper that this even.  In

          5  collaboration with a--I'm sorry.  This animation

          6  may take a bit to unfold.  In collaboration with a

          7  current graduate student of mine, Steve Lacy, we've

          8  examined whether switching processes themselves are

          9  immutable or whether the brain activations as a

         10  function of switching are due to the particular

         11  context in which the switching occurs.

         12             So what we did was we provided subjects

         13  with a switching task in which they had to make a

         14  decision either about a letter that appeared in a

         15  display or a digit that appeared in a display, and

         16  they had to alternate back and forth whether they

         17  were making a decision about the letter or the

         18  digit depending upon the color code of the objects

         19  they were looking at.  So here, they'd be

         20  responding to the letter.  Then they would have to

         21  switch to respond to the digit, stay on the digit

         22  because they're color-coded the same, switch back

 

                                                             113

          1  to the letter, and so forth.

          2             And we provided two experimental

          3  contexts in which this occurs.  One, the only

          4  thing--the only relevant stimulus in the display is

          5  the one about which they have to make a decision, a

          6  letter here, a digit there, a digit there or a

          7  letter there, and so forth.  So even though they're

          8  switching back and forth, there is no competing

          9  information to seduce them into working on the

         10  wrong thing, whereas in what we call a cross-talk

         11  condition here, both sorts of information are

         12  present.  Each of these stimuli is what you might

         13  call bivalent.  They could make the decision about

         14  a letter or a digit for each stimulus, but we have

         15  cued them by color which one to focus on, and the

         16  issue here is whether switching processes which are

         17  present in both of these situations yield identical

         18  patterns, identical signatures of brain

         19  activations, because the underlying switching

         20  operations are fundamentally the same, or are the

         21  switching operations modulated by the presence of

         22  interfering information.

 

                                                             114

          1             And as it turns out, they are modulated

          2  substantially by the presence of interfering

          3  information as you can see in this panel.  Here are

          4  activations in the switching condition compared

          5  to--on switch trials compared to non-switch trials

          6  in the presence of neutral competing stimuli that

          7  don't themselves have any probative value in the

          8  task itself.  Here is what I call the cross-talk

          9  case where you're switching and information is

         10  present on the other valence of the stimulus as

         11  well.  And you can see that the pattern of

         12  activation, especially in dorsal lateral aspects of

         13  prefrontal cortex, is dramatically different in

         14  these two cases, suggesting, once again, that even

         15  switching operations themselves need to be further

         16  subanalyzed into processes that can be understood

         17  in even simpler contexts.

         18             So to finish up this first example,

         19  complex tasks like the Wisconsin as measures of

         20  attention shifting depend on several processes that

         21  we reviewed, that Cam and I have both reviewed now.

         22  It's possible to parse these component processes

 

                                                             115

          1  with more refined tasks, and neuroimaging measures

          2  I think give traction to the measure of some of

          3  these component processes.  And it's possible to

          4  localize specific--localizing specific deficits in

          5  this way by having more refined tasks may provide,

          6  again, some traction again in understanding

          7  specific drug interventions for, say, switching

          8  deficit or processing negative feedback or

          9  processing positive feedback or whatever.

         10             Let me turn to a second example,

         11  assessing long-term memory.  We know from 150, at

         12  least, years of research that will there are a

         13  number of instruments that can be used to assess

         14  the fidelity of long-term memory.  One, for

         15  example, is the Hopkins Verbal Learning Test that I

         16  think was considered at one point for the battery.

         17  It might actually be present in the battery

         18  currently, even in its final form, in which you're

         19  presented a number of words and there is a recall

         20  test after a delay and there's a recognition test

         21  after a delay.

         22             Well, a test like this will yield a

 

                                                             116

          1  final set of performance measures.  It will yield

          2  recall accuracy, the kinds of errors that we make,

          3  intrusion errors, for example, in recall.  It will

          4  yield recognition accuracy, say a measure like

          5  de-prime that tells you how sensitive recognition

          6  performance is.  These measures, being unitary in

          7  form, make it seem as if long-term memory is a kind

          8  of unitary construct, but of course we know that

          9  that's not the case.  We know, as I said before,

         10  that it can be decomposed at least into coding

         11  storage and processes, and, furthermore, we know

         12  from another line of literature that I'm not going

         13  to review that there are explicit and implicit

         14  forms of the influence of long-term memory as well.

         15             So, for example, here is a study,

         16  probably the first one or at least one of the

         17  earlier ones in the literature, from John

         18  Gabrieli's work whose goal was to try to

         19  disassociate components of long-term memory

         20  processing that might be involved in encoding or

         21  retrieving, and what he did was to scan through a

         22  small portion of the brain, looking to see whether

 

                                                             117

          1  the hippocampal and parahippocampal regions that

          2  are activated at the time of encoding and at the

          3  time of retrieval might differ from one another.

          4  The task was to encode or to retrieve complex

          5  visual scenes and to scan subjects at the time of

          6  encoding, scan subjects at the time of retrieval,

          7  and what you find is different patterns of

          8  activation.  This is still a matter of some

          9  controversy, but there are, nonetheless, some

         10  confirmatory studies in the literature for the

         11  general conclusion that encoding operations seem to

         12  engage, seem to recruit more posterior mechanisms

         13  in the medial temporal lobe complex than retrieval

         14  operations do.  In fact, I think I inserted another

         15  slide here that demonstrates in this bottom panel

         16  in more protal slices.  In more protal slices,

         17  you'll see activations here as a function of

         18  encoding operations, and in these more anterior

         19  slices, you see activations as a function of

         20  retrieval operations.

         21             So encoding seems to yield activation in

         22  parahippocampal regions and also in the right

 

                                                             118

          1  frontal cortex compared to retrieval, which seems

          2  to yield activation in more anterior areas,

          3  suggesting that long-term memory as a kind of

          4  unitary concept itself can be broken down into

          5  subprocesses, and, in fact, there are tasks that

          6  can tap those subprocesses that can themselves be

          7  validated and coordinated by collecting both

          8  behavioral evidence and also imaging evidence and

          9  using those two in concert.

         10             Here are data from a person present in

         11  the audience, Anthony Wagner, on this very same

         12  issue.  This is a functional MRI experiment, also

         13  looking at encoding and retrieval operations for

         14  word stimuli and both pictorial stimuli, and you'll

         15  see here activation for word stimuli in more

         16  posterior areas of the medial temporal lobe complex

         17  in addition to activations in frontal cortex, again

         18  implicated in the Gabrieli study as well.  So once

         19  again, imaging evidence I think taken in context

         20  with refined behavioral evidence, I think allows us

         21  to go a step beyond analyzing working memory as a

         22  kind of unitary construct.

 

                                                             119

          1             So our conclusions here about long-term

          2  memory are that it requires more than a gross

          3  measure of accuracy of retrieval or the particular

          4  errors that you make or de-prime some sensitivity

          5  measure or your bias to making certain kinds of

          6  errors.  Rather, we can separate out the encoding

          7  and retrieval processes and even other processes of

          8  memory using neuroimaging measures which implicate

          9  different circuitries, and to the extent that there

         10  are different circuitries implicated, of course,

         11  that might also implicate different

         12  neurotransmitter systems that might be involved.

         13  For example, the dopamine system may be more

         14  effective in ameliorating strategic encoding and

         15  elaboration deficits based at least on the simple

         16  observation that there's frontal cortex that's

         17  involved in encoding, whereas focusing on the NMDA

         18  receptor function may yield a very different kind

         19  of intervention that might be presumed if you were

         20  interested in retrieval operations.

         21             Having not yet outstayed my welcome, let

         22  me move on to the third and final example.  This

 

                                                             120

          1  one has to do with processes involved in resolving

          2  interference.  There are many clinical symptoms of

          3  the dysexecutive syndrome and frontal lobe insults

          4  generally that involve failures to resolve

          5  interference, for example, perseverative behavior

          6  of the sort that you find in schizophrenic and

          7  other patients as well, and these kinds of

          8  perserverative behaviors and failures to resolve

          9  interference appear in a number of different

         10  pathologies, both focal pathologies and also more

         11  in defuse ones.

         12             There are reasons to be concerned about

         13  these in the normal world because they affect

         14  normal social processing.  They affect normal

         15  aging, but also, of course, in the world of

         16  pathology as well because they're important markers

         17  of brain dysfunction as well, and again, there are

         18  a number of neuropsychological instruments that

         19  appear to be reasonable assays, difficulty in

         20  interference resolution, and here's some of them,

         21  but once again, these assays seem not to be very

         22  specific about the processes that are involved.

 

                                                             121

          1             For example, you can ask is there such a

          2  thing as interference resolution as a unitary

          3  construct.  Well, in fact, if you look through the

          4  literature, as I and my colleagues have done, you

          5  find a number of areas of cortex that are activated

          6  by different kinds of allegedly simpler

          7  interference resolution tasks.  In fact, if you do,

          8  as we have done, a metananalysis of the imaging

          9  literature on all the tasks that engage

         10  interference resolution, what you find is a pattern

         11  of brain activations that looks like this,

         12  including activations outside of the head.

         13             [Laughter.]

         14             DR. JONIDES:  I guess I wouldn't take

         15  that one seriously, but I would take the red dots

         16  within the head seriously, suggesting that a task

         17  like the Stroop task, go-no S-R compatibility

         18  flanker tasks, signing tasks, and some others as

         19  well.  It looks as if there is no coherence to

         20  these data, as if they're just isn't a single

         21  construct.  That on some days might induce a kind

         22  of depression in the experimenter who is concerned

 

                                                             122

          1  with these things, but that depression is relieved

          2  by doing more careful analysis on those patterns of

          3  brain activations, and realizing that there is

          4  coherence to them, you do find structures like

          5  dorsal lateral prefrontal cortex in the right

          6  hemisphere, pre-motor cortex, anterior singular

          7  cortex, aspects of parietal cortex, aspects of

          8  anterior frontal cortex, hanging together in

          9  clusters of activations that seem repeatedly to be

         10  activated by a number of tasks.  In fact, Cam

         11  Carter and John Conconan and colleagues have, in

         12  fact, taken some of these data and amassed together

         13  a theory that brings together aspects of conflict

         14  monitoring and execution of functions to solve that

         15  conflict, engaging aspects of frontal cortex.

         16             Well, our attack on this problem is

         17  to--is what I characterize as little steps for

         18  little feet, that is we were looking for a task

         19  where we might be able to isolate an inhibitory

         20  process in a refined kind of way and then see

         21  whether we can find the brain mechanisms that are

         22  engaged by that inhibitory process, and the way we

 

                                                             123

          1  did that was to take a task that essentially

          2  involves no inhibitory processes and adding one, as

          3  it were, to it.  So let me just kind of march you

          4  through the experiment.  It's actually quite

          5  simple.  A standard task in the cognitive

          6  literature, as most of you know, is one invented by

          7  Saul Sternberg in the early 1960s in which a

          8  subject memorizes the number of letters, although

          9  it could be spatial positions or visual objects or

         10  faces or a number of things.  There follows a

         11  retention interval of several seconds, and then the

         12  subject is given a probe, in this case a letter,

         13  and has to answer yes or no whether that probe

         14  matches one of the letters that's stored in memory.

         15  The task is quite simply, obviously involves

         16  encoding, retention, and retrieval, but very little

         17  by way of the kinds of things you would typically

         18  identify as executive processes.

         19             So we take a task like this, and we

         20  induce in it an executive process of a particular

         21  form.  So here is an example of a task--of a trial

         22  that does that.  Here is the set,, the subject

 

                                                             124

          1  memorizes a retention interval and a probe.  Here

          2  is a second trial of the experiment, another set,

          3  retention interval and a probe.  Notice that this

          4  probe demands a no response because it's not part

          5  of that memory set.  Can you see that even in the

          6  back of the room?  Whereas, notice also that this

          7  probe, although demanding a no response, was a

          8  member of the previous trial's memory set, and so

          9  the subject might be seduced into answering yes to

         10  this probe because he has seen this item recently

         11  even though the proper answer is no on this trial.

         12             So we've taken a task that's light on

         13  executive processes and introduced into it a

         14  conflict component that needs to be resolved, and

         15  the obvious control condition here is by having a

         16  task in which there is no conflict between the

         17  present response and the previous trial compared to

         18  one in which there is conflict.  And what we find

         19  behaviorally and by way of neurosignature is that

         20  there is an interference affect caused by the

         21  high-conflict trials compares to the low-conflict

         22  trials, and that interference effect seems to

 

                                                             125

          1  produce a neurosignature in inferior and middle

          2  aspects of prefrontal cortex on the lateral side,

          3  restricted to the left hemisphere, that is an

          4  effect that's quite robust and has been replicated

          5  not only by our laboratory, but by others as well.

          6             Now, all I've shown you so far is a

          7  correlation between a brain activation and pattern

          8  of behavioral activity.  We would like to go beyond

          9  that to show that that there's cause and effect,

         10  and we have gone beyond that in a couple of ways,

         11  and I'll finish up with this.  We've identified a

         12  patient who has an injury due to AVM some 20 years

         13  previous to our testing of this patient that

         14  affects this region of prefrontal cortex, including

         15  inferior and middle frontal gyrus, and we compared

         16  that patient's performance on this very task to a

         17  number of control subjects to find out how well

         18  this patient does.  Presumably, having a lesion in

         19  this region, if this region is critical to

         20  performance, ought to cause at least a noticeable

         21  drop in performance.

         22             And here are the data.  Here is RC's

 

                                                             126

          1  performance:  Response time elevated substantially,

          2  error rate between 15 and 20 percent compared to a

          3  number of control groups, controls who are age

          4  matched who are not patients, frontal controls,

          5  elderly adults, and some young adults, and you can

          6  see that RC, in fact by several orders of

          7  magnitude--this is where statistics are no longer

          8  even necessary.  I'm sorry.  By several standard

          9  deviations his performance is worse than any of the

         10  control groups.

         11             So this begins to establish a link

         12  between a particular brain region and a particular

         13  pattern of behavioral activity and you can go even

         14  beyond this.  The normal elderly, as you might have

         15  seen on the previous slide, do worse on this task

         16  than the young people do.  They show a greater

         17  interference, in fact, and, in fact, so you would

         18  expect that if looked at their brain signatures in

         19  this critical region of left prefrontal cortex,

         20  they ought to show a difference compared to the

         21  young, and indeed they do.  I've highlighted here

         22  for you just to call your attention.  These are the

 

                                                             127

          1  data from the young adults that I showed you

          2  previously where you find activation here as a

          3  function of these interference resolution processes

          4  tracked down now to the data from the older adults,

          5  and see you see that they're not showing activation

          6  in a comparable region of the brain, suggesting

          7  that they're not activating this region, in turn

          8  suggesting that they're not engaging in the

          9  psychological processes within.

         10             So to conclude about this third example

         11  about interference resolution, there are many tasks

         12  can be used to assess skill in resolving

         13  interference.  There are different stages and

         14  processing in these tasks at which interference can

         15  be resolved, and, in fact, we think that

         16  neuroimaging evidence provides leverage in

         17  understanding what the stages are in which

         18  interference gets resolved, and the task now is

         19  picking nature apart at its joints, trying to find

         20  out what the stages of processing are and how

         21  interference can get resolved at these tasks, and

         22  there are still heavy lifting to be done on that.

 

                                                             128

          1             So I'll finish up by giving you dessert

          2  again.  The final message, as the initial was, is

          3  there are many psychological functions that are

          4  complex mixes of component processes.  I've talked

          5  about just three of them.  I could have marched

          6  forth several other examples.  There are standard

          7  neuropsychological instruments, but unfortunately

          8  they're often too coarse to penetrate the component

          9  processes involved in any complex task.  We need

         10  theories to parse these component processes, and

         11  then we need assessment tools that can isolate a

         12  process for further measurement, and I think these

         13  assessment tools are best conceptualized as some

         14  combination of behavioral measurements and imaging

         15  measurements.

         16             Thank you.

         17             DR. GEYER:  Thank you, John.  That

         18  indicates a domain of work that we will need to

         19  pursue in the future.

         20             It is a pleasure to introduce Jim Hagan,

         21  who has come all the way from the United Kingdom to

         22  join and us and present something about the needs

 

                                                             129

          1  and prospects for preclinical predictors of

          2  efficacious treatments for cognitive deficits in

          3  schizophrenia.

          4             Jim, thanks for coming.

          5      VIII.  NEEDS AND PROSPECTS FOR PRECLINICAL

          6  PREDICTORS OF EFFICACIOUS TREATMENTS FOR COGNITIVE

          7               DEFICITS IN SCHIZOPHRENIA

          8             DR. HAGAN:  While we're waiting for the

          9  slides, let me just say good morning to everybody

         10  and thank you on behalf of myself and my

         11  organization to the organizers, particularly to

         12  Mark for giving us the opportunity, giving me the

         13  opportunity to address what is I think is a very

         14  important meeting.

         15             What I had hoped to talk through is

         16  really somewhat of a personal view.  I've had an

         17  opportunity to discuss this with colleagues, and

         18  but I can't pretend and shouldn't pretend it

         19  represents anything like a consensus view from the

         20  industry.

         21             Okay.  The areas I want to talk to are

         22  outlined on the agenda slide.  I want to briefly

 

                                                             130

          1  discuss the drug discovery cycle and summarize the

          2  cognitive domains that have been agreed by the

          3  consensus committees and reintroduce you to the

          4  idea of the translation gap and then really get to

          5  the bulk of the talk, which is going to be to pull

          6  out some of the issues which I think face the field

          7  in terms of integrating preclinical and clinical

          8  research.  Some of these themes have clearly been

          9  touched upon by earlier speakers.  The title also

         10  asked me to consider some future prospects, and so

         11  I picked out a couple of examples where I

         12  think--which I think give us real optimism and some

         13  prospects of hope here, and finally to draw

         14  together some summaries and conclusions.

         15             I should have said at the beginning that

         16  my focus will be on almost entirely on behavioral

         17  influence.  This isn't to assure the value and the

         18  insights that will come from considerations of

         19  electrophysiology and biomarkers and so on.  I

         20  think these will be very valuable contributors to

         21  the field as we move forward, but just in the

         22  interest of time, I thought it was better to focus

 

                                                             131

          1  on some of the behavioral work just to draw out

          2  some of what I think are the practical issues which

          3  face us.

          4             On the discovery cycle, this is a highly

          5  stylized version, but we're a mixed audience, so I

          6  thought it would be good to go through at least the

          7  rudiments and talk you through what happens from

          8  the proof of--sorry--from the beginning of the

          9  concept where we start to become interested in a

         10  protein as a pharmacological target, the work which

         11  goes on through to demonstrating the fact that that

         12  approach lives up to its proof of mechanism both in

         13  vitro and then hopefully in vivo.  In this

         14  particular area, moving forward to demonstrate

         15  precognitive effects in rodents and non-human

         16  primates, then moving to translation models in

         17  human volunteers, translation models in patients,

         18  and then into the Phase II proof of concept

         19  followed by a Phase III.

         20             Now, for any particular compound in the

         21  discovery cycle, this is essentially largely a

         22  linear process as it moves through the various

 

                                                             132

          1  steps, but when we're constructing the cascade,

          2  this is a highly intrative and highly circular

          3  process, for obvious reasons, and I want to focus

          4  here on the links between rodent models and primate

          5  models and building the bridges across to the

          6  models which we can apply in schizophrenics and in

          7  human volunteers.  Now, to my mind, what the

          8  MATRICS committees have achieved so far is to put

          9  real meat behind the concept of developing new

         10  tools and concepts for looking at cognitive effects

         11  both in human volunteers and in patients and really

         12  is giving us a frame work of moving forward into

         13  Phase III and registration.  What we're essentially

         14  doing here today is--why I'm trying to do here

         15  today is take a step back and consider what the

         16  implications are for the preclinical area, and I

         17  hope to convince you about why I think that is a

         18  very important step in this.

         19             The seven cognitive domains, which we

         20  agree, haven't yet been formally mentioned by

         21  previous speakers, so they're listed on the slide

         22  as working memory, attention vigilance, verbal

 

                                                             133

          1  learning and memory, visual learning and memory,

          2  reasoning and problem solving and speed of

          3  processing, and certainly cognition is then added

          4  following the April 2000 meeting.  I'm grateful to

          5  Dr. Green for sending me a pre-print of the summary

          6  paper which will be coming out later on in the

          7  year.  I'm not going to discuss the requirement for

          8  a co-primary outcome measure.

          9             Next slide, the assumptions which are

         10  driving this or the premises have been alluded to

         11  by other speakers, but I think it's worth

         12  emphasizing them, and that is that the cognitive

         13  deficits are core features of the disease and

         14  they're relatively common in schizophrenic patients

         15  and, mostly importantly, that they're amenable to

         16  pharmacological improvement.  This is a point which

         17  we'll discuss in more detail as the presentation

         18  proceeds.  And finally, that cognitive deficits

         19  relate or correlate heavily with daily function.

         20             And this is a slide of uncertain

         21  heritage.  I got it from a colleague, Herb Harris,

         22  who attributes it to Wayne Fenton who may attribute

 

                                                             134

          1  it to Dr. Green.  But it makes an interesting

          2  point.  It emphasizes the negative correlation

          3  between cognitive impairment and functional

          4  outcome.  I want to use it to make a slightly

          5  different point, and  I think this is now widely

          6  recognized, but from the point of view of drug

          7  development, we really have to consider how these

          8  domains with hang together.  Alan alluded earlier

          9  to the fact that the preferred course for drug

         10  development is to have single target molecules, and

         11  I think by and large that is true; however, if we

         12  are in a position where we have to design

         13  pharmacological treatments which need to address

         14  delusions, hallucinations, and thought disorders in

         15  addition to the cognitive impairment, and that's a

         16  likely outcome, then as an added layer of

         17  complexity as we develop those molecules, we really

         18  need to try and understand how that particular

         19  mechanism of action will interact with other

         20  compounds, other drug therapies, which are designed

         21  to address these other deficits.  So I think that's

         22  an important discussion which we need to hold.

 

                                                             135

          1             This slide lists the tests which are

          2  undergoing validation in the beta version, and

          3  we're expecting some preliminary conclusions by the

          4  end of October.  So I think it's timely now that

          5  the preclinical community starts to consider what

          6  the implications of this consolidated view of

          7  mission is going to be for our preclinical work.

          8             I just want to remind you of the concept

          9  which was highlighted by Hyman and Fenton in the

         10  Science edition in 2003, what they call the

         11  translation gap, and this has been discussed in the

         12  context of the MATRICS initiative, a simple plot to

         13  the number of publications over the years on

         14  mechanisms related to cognition in schizophrenia

         15  and preclinical literature and the number of human

         16  clinical trials.  Now, clearly, when we consider

         17  the cycle times required for drug development,

         18  there's going to be a hysteresis between these two

         19  curves, that is it's going to take some times,

         20  years in cases, to translate novel mechanisms into

         21  clinical trials, but it's not just a matter of

         22  time.  I think we have to make a real effort to

 

                                                             136

          1  make sure that we put the intervening stages, the

          2  intervening processes into place to ensure that

          3  these potential new mechanisms really do get

          4  translated into therapeutic benefit.

          5             I've listed here the outcome of a very

          6  superficial and rapid review of the literature, the

          7  number of potential mechanisms which have been

          8  considered, and these appear in publications,

          9  journal publications, and in the patent literature,

         10  and it's very incomplete.  But you can see from the

         11  list that we're facing a potential deluge of novel

         12  approaches as pharmacology develops and as the

         13  fields of genetics and transcriptional analysis

         14  develops.  We can expect that the list of potential

         15  targets will increase very rapidly, and this will

         16  be predictable on the basis of our understanding of

         17  the human genome project.

         18             So the issue for us now is how do we

         19  address this issue of which mechanisms to

         20  concentrate on and how do we identify those

         21  approaches which are going to be most promising in

         22  order to deliver novel treatments as rapidly as

 

                                                             137

          1  possible.

          2             So I want to focus our attention for the

          3  next few moments on what we need to do to build

          4  these links between our understanding of the

          5  cognitive impairments in patients and the models

          6  which we have available to us to try to predict

          7  therapeutic benefit in animals.  On this slide,

          8  there's listed a number of experimental procedures

          9  which are available.  One of the characteristics of

         10  this field is that there is a huge variety of

         11  potential procedures and tasks that we can call

         12  upon.  I've listed them in terms of--set them up in

         13  terms of how they map onto the various cognitive

         14  domains and how they map onto the various clinical

         15  tests, some of which are being included in the beta

         16  testing battery.

         17             The list is incomplete.  One of the

         18  features of this list is that it contains

         19  experimental procedures which are derived from a

         20  number of different sources, largely from

         21  experimental psychology studies which have been

         22  focused on giving us a better understanding of

 

                                                             138

          1  cognitive processing in model animals and in

          2  volunteers and partly on the efforts of other

          3  therapeutic areas to develop models which have

          4  predictive therapeutic benefit in their cases, for

          5  example, in the case of Alzheimer's disease.

          6             We've put up a straw man proposal as to

          7  how these various models might map onto the

          8  cognitive domains which have been identified by the

          9  MATRICS organization so far, and again, this is a

         10  highly incomplete list, but I think what this list

         11  serves as is an agenda for the field for the next

         12  few years.  I think one of the important tasks we

         13  have to tackle is to really consider how these

         14  various tests map onto the cognitive domains which

         15  have now been agreed by consensus to be important

         16  in schizophrenia and to understand how those animal

         17  tests then map onto the battery of tests which are

         18  finally included in the agreed version.

         19             So this I think outlines what I would

         20  consider to be a potential agenda for the field

         21  over for the next period.  I think we're faced with

         22  a plethora of animal models of a lot of procedural

 

                                                             139

          1  variations.  We really need to consider a

          2  re-evaluation of those animal models in the light

          3  of the seven cognitive domains that have been

          4  agreed.  We need to get a discussion going.  We

          5  need to get clarity and at least understanding, if

          6  not agreement, on the sorts of criteria we should

          7  be using to select our models.  Face validity and

          8  construct validity are important.  Predictability

          9  is obviously very important for the drug

         10  development industry and pharmaceutical industry.

         11             Reliability of robustness is critical.

         12  One of the examples I'll talk about in a minute,

         13  which is the pre-pulse inhibition example, I think

         14  has made a big contribution to the field, and one

         15  of the reasons for that is because the procedures

         16  used in animals to test pre-pulse inhibition

         17  deficits are reliable and robust.

         18             For drug discovery, good throughput is

         19  an advantage, for academic science and for studies

         20  designed to understand and to explore underlying

         21  processes, that's probably less of a requirement,

         22  but certainly from our prospect, from our point of

 

                                                             140

          1  view, good throughput is certainly an advantage and

          2  it certainly accelerates the rates at which we can

          3  make our critical choices between alternative

          4  approaches and alternative molecules.

          5             And just visible at the end of the slide

          6  here is a proposal that I think we can facilitate

          7  progress in this area by following the example of

          8  the work which has been done so far in MATRICS and

          9  undertaking, first of all, the definition of the

         10  seven models and identification of the seven models

         11  which map onto the cognitive domains, and then

         12  considering what we do from a similar approach.  So

         13  we do beta testing to establish the reliability and

         14  reproducibility of those procedures.

         15             A very important aspect for us is

         16  pharmacological validation.  Within the drug

         17  discovery cycle, I guess what we're doing here is

         18  taking compounds which have given proof of concept

         19  in the clinic and bringing them back into our

         20  preclinical models to see if those models pick up

         21  those pharmacological activities.  I can't pretend

         22  to have done a comprehensive review of the

 

                                                             141

          1  literature, because it's absolutely enormous, but

          2  what I have done--I'm sorry.  Before I get to that,

          3  just to remind ourselves of the sort of consensus

          4  position we're at with respect to cognitive

          5  improvement with our currently available atypicals

          6  and typical antipsychotics, and these are data

          7  taken from the metananalysis by Harvey and Keith

          8  published in 2001 in which they looked at the

          9  outcome of 24 refereed reports, refereed studies in

         10  which by and large patients have been maintained on

         11  typicals and switched onto atypicals and then the

         12  change in cognitive performance has been measured.

         13             Now, a couple of things which are quite

         14  striking:  One is that the metananalysis shows that

         15  there is some degree of cognitive benefit across

         16  the piece, but also that the magnitude of that

         17  increase is relatively small.  What we don't know,

         18  of course, is how much we need to improve in order

         19  to get not just statistically significant changes,

         20  but also clinically significant changes.  And also,

         21  this slide, I haven't got time to go into some of

         22  the caveats and the problems associated with this

 

                                                             142

          1  kind of analysis, and that's done very elegantly by

          2  Harvey and Keith in their paper.

          3             And it may actually be that the switch

          4  from typicals to atypicals may be giving us

          5  actually a slightly misleading picture.  If the

          6  typicals are dosed in relatively high dose range,

          7  it may be that they're causing holes in the pattern

          8  through secondary mechanisms.

          9             But if we look at the some of the

         10  individual studies, and these are taken from a

         11  paper by Meltzer and McGurkin in 1999--they're

         12  studying the effects of olanzapine.  They're able

         13  to see statistically significant effects on

         14  executive function measured by a color-word

         15  interferences tasks, but are also able to see

         16  statistically significant effects on verbal

         17  learning and on verbal fluency, but not on visual

         18  learning and visual memory and on executive

         19  function as measured by the procedures.  Continuing

         20  the analysis of that study, they found the effect

         21  of olanzapine on working memory, although they did

         22  report effects on attention and reaction time.

 

                                                             143

          1             So looking at the level of individual

          2  drug response, it's possible to see evidence of

          3  cognitive enhancement as well as seeing the general

          4  picture in the metananalysis, and just to continue

          5  that theme, looking at the same paper,

          6  metananalysis studies on risperidone, again

          7  evidence of working memory effects, improvements in

          8  four out of four studies, and evidence for effects

          9  of verbal learning and memory with improvements in

         10  two out of two studies, with some evidence of

         11  effects on attention.

         12             So if we return to the animal studies

         13  now and look at a couple of paradigms, what I want

         14  to do is to look at a model, which is pre-pulse

         15  inhibition, which is relevant for pre-attendant

         16  processing and attention and vigilance functions

         17  and look at the preclinical data which have been

         18  generated there.  Now, pre-pulse inhibition deficit

         19  is very well described in schizophrenia, but also

         20  in Huntington's and OCD, and are thought to reflect

         21  disease-related deficiencies in central motor

         22  processing, but the clinical literature shows a

 

                                                             144

          1  mixed picture of the effects of antipsychotics.

          2  There are some reports that antipsychotics will

          3  improve the PPI of schizophrenics, but there are

          4  also no distinction between the typicals and

          5  atypicals in a couple of studies, and there are

          6  also some reports of greater efficacy for the

          7  typical, a paper by Kumari in 1999 and a follow-up

          8  in 2002.  But there are also reports that PPI is

          9  not affected by medication status and a lack of

         10  affect of risperidone.

         11             If we're looking at the preclinical

         12  literature, and there is some very, very extensive

         13  literature now and it's been reviewed by Mark and

         14  colleagues a couple of years ago, this has been one

         15  of the main focus of attention for the academic

         16  pre-clinical community, but also for people within

         17  the industry, and the broad picture which is

         18  emerging is that in model of PPI disruption in

         19  rodents, it's possible to get reliable deficits

         20  with a variety of dopamine agonists, direct and

         21  indirect, a variety of 5-HT2 agonists and NMDA

         22  antagonists and also by environmental innovations

 

                                                             145

          1  such as isolation rearing.

          2             And in the case of--when you look at how

          3  you can reverse those deficits, there is a reliable

          4  and robust reversal with typical and atypical

          5  antipsychotics, particularly with high D2 activity.

          6  There is a mixed picture with 5-HT2 agonists, a

          7  mixed picture with NMDA antagonists, and broad

          8  support for the idea that these compounds can

          9  reverse isolation-induced deficits.  Our experience

         10  with this model--I'm sorry.  On top of that--so in

         11  addition to pharmacological intervention, the

         12  isolation rearing procedure has been extensively

         13  used, and our experience with PPI-induced deficits

         14  to isolation rearing is that you can get reliable

         15  reversal with haloperidol, clozapine, and other

         16  atypicals, and you can also see reversals with

         17  2.701 molecules, selected D-3 antagonists, and

         18  100907, which is selective 5-HT2 antagonists, and

         19  so we're using this procedure as a way of trying to

         20  predict what the clinical response will be.

         21             The model also has huge potential in

         22  terms of--and has been used widely in terms of

 

                                                             146

          1  characterizing genetic models, and that's done for

          2  a number of different reasons.  Mice wich show a

          3  PPI deficit, can you give you clues as to the best

          4  pathways which are involved in controlling

          5  behavior, and it gives a robust and reliable and

          6  good throughput.

          7             The question is--I'm sorry.  Can you go

          8  back?

          9             The question is which is the most

         10  appropriate disease model that we should be using,

         11  and the second question which I think we need to

         12  resolve is how does the PPI model relate to the

         13  seven domains, can we reach agreement on how this

         14  rich pharmacology, this rich data set can be used

         15  in stepping back to try to predict clinical

         16  effects, and I think this is an example where we've

         17  got a very, very well characterized preclinical

         18  model and we really need to try now to

         19  understand--in the light of the agreed deficits in

         20  schizophrenia, we need to understand how this model

         21  will relate back to those deficits and how we can

         22  best use this data set.

 

                                                             147

          1             In the next slide, now I want to turn

          2  attention briefly to working memory deficits, and

          3  focus on some work using delayed decision tasks,

          4  and these are data taken from our own work, and you

          5  will recall from the metananalysis and also from

          6  some of the individual compound analysis that

          7  you're able to see improvements in working memory

          8  with some atypical compounds.  So we went

          9  looking--using this delayed match position model,

         10  we went looking in rodents to see if we can see

         11  evidence of improvement.

         12             Now, the graph doesn't summarize the

         13  whole experiment, but these animals are

         14  being--we're measuring choice accuracy in a

         15  discrimination task and introducing delays between

         16  the presentation of the sample and the response, so

         17  generating a delay function.  With none of these

         18  compounds were we able to see any specific effects

         19  on delayed function.  What we see instead is that

         20  as you go up in the dose, you start to see overall

         21  detriments in performance deficit and then choice

         22  accuracy starts to go down, and we see that to one

 

                                                             148

          1  degree or another with the four compounds, and when

          2  you look at the way in which the animals are

          3  behaving in this procedure, you see that what's

          4  happening is that the compounds are actually

          5  causing the animals to stop pressing.  They're

          6  having a substantially increased number of

          7  mistrials.

          8             So we're not able to see any evidence

          9  for improved cognitive performance in a delayed

         10  impairment procedure.  What we do see is an

         11  impairment.

         12             Next slide, so if we look across the

         13  literature of this and some other tasks, what we

         14  see in one study by Gemperle in 2003 where they're

         15  able to produce evidence for improvement by

         16  iloperidole, but clozapine and haloperidol

         17  impaired--I'm sorry.  Iloperidole improves a

         18  delayed match, but clozapine and haloperidol

         19  impair.  Again, with our work, we see an increased

         20  number of mistrials.  We're seeing impairments, and

         21  this is a picture which repeats itself in other

         22  publications.

 

                                                             149

          1             Looking at radial arm maze, we see a

          2  similar picture.  Looking with chronic drug

          3  treatments, we see a similar picture with water

          4  maze, and one of the most extensive analysis in

          5  this area is the Skarsfeldt publication in which

          6  all of these compounds have been shown to impair

          7  spacial learning.

          8             So the issue for these models is that by

          9  and large what we're seeing with compounds which

         10  give some degree of cognitive improvement in the

         11  clinic, what we're seeing is impairment rather than

         12  improvement.  This obviously raises a lot of issues

         13  about the format of the model.  It raises a lot of

         14  issues about what the most appropriate dosing

         15  regimes are, and it also raising an issue about

         16  pharmacological validation and questions about

         17  whether this is a problem with the model, that is

         18  are we looking at the wrong processes here, or is

         19  it a problem with the molecule, that is the

         20  cognitive benefit we see in the clinic just so

         21  small that we're unable to pick it up with these

         22  procedures, and I think that is an issue which we

 

                                                             150

          1  clearly need to come to terms with.

          2             Considering now some future prospects,

          3  there's been a little discussion this morning about

          4  the executive function deficits which you see

          5  reflected in the Wisconsin Card Sorting Task, and

          6  I'm pleased to pick up this example from our recent

          7  literature from our own work.  The effect of

          8  antipsychotics on this task is mixed.  There's good

          9  agreement that this is deficient in schizophrenics,

         10  but the response literature is mixed.  There is

         11  some evidence of performance improvement with

         12  risperidone and olanzapine and also with clozapine

         13  and olanzapine, although some authors find no

         14  significant effects.

         15             In a recent study published by Turner,

         16  et al., and Trevor is going to talk to this in a

         17  lot more detail, next slide, Trevor and his

         18  colleges, Trevor Robbins and his colleagues, have

         19  looked at the effects of modafinil in an adult

         20  study using 20 stabilized schizophrenics, using a

         21  version of the attentional set shifting procedure

         22  which is embedded in the Cantab cognitive battery

 

                                                             151

          1  and a series of working from a simple

          2  discrimination to reversal, compound

          3  discrimination, reversal, compound discrimination,

          4  and on to intradimensional shifts, reversal,

          5  intradimensional shifts and extradimensional

          6  shifts.

          7             And what you see is as this procedure

          8  progresses, schizophrenic patients stabilized on

          9  other drugs are starting to drop out.  A portion of

         10  patients which reach each stage is starting to

         11  reduce.  With the addition of modafinil, that's

         12  partially reversed, particularly on the

         13  extradimensional shift, and I think this is very

         14  encouraging and it finds a potential way forward.

         15             On the next slide, some work which has

         16  been done primarily in the University of St.

         17  Andrews by Brown's group has designed a task which

         18  is analogous using a rodent discrimination

         19  procedure.  Basically, the animals are being

         20  trained to discriminate on different dimensions,

         21  being trained to discriminate between two odors,

         22  trained to discriminate between different sets of

 

                                                             152

          1  data; and using this procedure--and again, Trevor I

          2  think is going to speak in more detail about the

          3  procedure.  Using this approach, it's possible to

          4  build an animal analog of the extradimensional

          5  shift procedure which has been used in the animal

          6  experiment and which mimics the deficits seen in

          7  schizophrenia.

          8             This is an example of a study by

          9  Tunbridge published very recently using tolcapone

         10  in which they're able to demonstrate highly

         11  significant and very specific beneficial effect on

         12  the extradimensional shift using this model, and

         13  tolcapone is of interest because of the biomet 158

         14  polymorphozone literature you'll see in the

         15  teaching.  And on the next slide, I just want to

         16  highlight some of the work that we've done in the

         17  5-HT6 receptor antagonist, 399885, in which we were

         18  able to see some improvements on the initial

         19  compound reversal, but also some improvements in

         20  the extradimensional shift, and we've been able to

         21  demonstrate this with this compound, 399885, but

         22  also at 271046, a molecule of similar pharmacology.

 

                                                             153

          1             So I think just to close, these are some

          2  of the issues which face the field and which we

          3  should be discussing later on this morning.  We

          4  need to identify those tests, those animal

          5  procedures which best predict for improved

          6  functional outcome in schizophrenia.  At the

          7  moment, we have a relatively poor understanding of

          8  the predictive validity of these tests, and Alan

          9  alluded to this earlier in the early experience.

         10  We have a relatively poor understanding of that

         11  area.

         12             We have a relatively poor data base of

         13  preclinical studies of antipsychotics.  There's a

         14  huge amount of literature published, but it's

         15  published in different model formats using

         16  different dosing regimes, and it's quite

         17  inconsistent.  We need to think about which of the

         18  various impairment models--and I haven't had time

         19  to go through those, but which of the various

         20  impairment models we should focus on in order to

         21  reproduce some of the cognitive deficits that we

         22  see or mirror some of the cognitive deficits that

 

                                                             154

          1  we see.  And we need to define what the most

          2  appropriate models are in schizophrenia and

          3  cognitive impairment.

          4             I think there are potentially a huge

          5  number of pharmacological opportunities.  I've

          6  alluded to some of this them.  Progress in

          7  pharmacological genetics and genomics is likely ti

          8  increase that list rather than diminish the list,

          9  and we have a plethora of animal models available,

         10  and so in order to close the gap and make sure that

         11  we're able to deliver on some of this potential, I

         12  think these are the things we need to address.  We

         13  need to focus debate on and achieve some sort of

         14  consensus on what the most appropriate models are,

         15  continuous cross-validation between animal and

         16  human test procedures, and at least from the point

         17  of view of drug discovery, we need quite robust and

         18  reliable animal test procedure.

         19             And in the longer term--this is

         20  something which we haven't gotten to--in the long

         21  term, we need to focus on models which more closely

         22  recapitulate the pathophysiology of schizophrenia.

 

                                                             155

          1             And this slide, I just want to

          2  reiterate, underline some of the points which Alan

          3  made earlier, draw to your attention why this is

          4  important.  These are the data taken from a recent

          5  review by Koler and Landis in Nature Revies.  They

          6  look at the percentage of success of compounds

          7  which enter man and finally end up getting

          8  registration.  Overall, the success rate is 11

          9  percent.  For CNS, that's actually lower at about

         10  eight percent.

         11             So greater than 90 percent of compounds

         12  which actually go into man in CNS indications fail.

         13  About 30 percent of those efficacies are due to

         14  failure--due to failures of efficacy.  Alan alluded

         15  to the problems posed by unpredictable toxicology,

         16  and that accounts for a large proportion of these

         17  failures, but a large proportion are due to

         18  failures of efficacy, and I think if we're able to

         19  get ourselves into a position where we can get a

         20  better understanding of these animal models,

         21  understand their elevation better, use them and

         22  identify more predictive models, we should be able

 

                                                             156

          1  to cut down that failure rate at least in the point

          2  of view of efficacy.

          3             And, finally, I'd just like to

          4  acknowledge the contributions of various colleagues

          5  and associates:  Declan Jones, Darrell, Jack Celia,

          6  Chris Kalinchev, Dr. Harris from GSK.

          7             Thank you for your attention.

          8             [Applause.]

          9             DR. GEYER:  Thanks, Jim.

         10             It's a pleasure to introduce our next

         11  visitor from the United Kingdom and a dear friend

         12  of mine, Dr. Trevor Robbins, who will help us

         13  discuss synthesizing schizophrenia.

         14     IX.  SYNTHESIZING SCHIZOPHRENIA: A BOTTOM-UP,

         15                  SYMPTOMATIC APPROACH

         16             Dr. Robbins:  Thanks very much, Mark.

         17  It's a great pleasure be here.  Thank you very much

         18  for inviting me.

         19             So synthesizing schizophrenia, putting

         20  it back together again, as it were--next slide,

         21  please--the way I see it, there are three domains

         22  of interest in this project.  I suppose we should

 

                                                             157

          1  focus on this one, the therapeutic drug problem.

          2  We have at a minimum, obviously, interest in

          3  antipsychotic drugs without disruptive cognitive

          4  side effects.  That's a challenge in itself.  It

          5  would be nice if antipsychotics had some additional

          6  cognitive enhancing effect, and maybe that's true,

          7  but of course we're also interested in

          8  non-antipsychotic cognitive enhancers potentially

          9  for different domains of cognition.

         10             Now, somehow this has got to make

         11  contact with a better understanding of what an

         12  animal model means in schizophrenia, contrasted

         13  aetiological models, and, of course, we don't

         14  really understand the molecular pathogenesis of

         15  schizophrenia that well.  So this is particularly

         16  difficult, and symptomatic models, models which

         17  will reproduce some of the cognitive deficits we

         18  see, but are not actually probably providing

         19  mechanisms that occur in schizophrenia.

         20             And the third element is having

         21  appropriate neurobehavioral phenotypes, and I think

         22  a very important point stressed by Cam Carter and

 

                                                             158

          1  John Jonides is the really importance of

          2  understanding of construct validity, the

          3  endophenotypes, the deep structure for cognition in

          4  the brain rather than simply operational measures

          5  which may or may not be responsive to drug effects.

          6             So the first point to make, perhaps,

          7  probably for the pathologists here is that they are

          8  diverse neuropathology over time, and of course

          9  they are not and they probably have different ways

         10  to produce the heterogeneity you see in this

         11  disorder, including at the cognitive level, and

         12  that's something else we also have to take into

         13  account.

         14             Next slide, in addition to the

         15  aetiological heterogeneity, there are probably a

         16  whole range of other factors which modulate the

         17  expression in schizophrenia, and some of these are

         18  captured in various models which range from

         19  genetics to lesions in the prefrontal cortex, and

         20  stress models.  Clearly, as to Jim Hagan has

         21  pointed out, we need to evaluate how some of these

         22  deficit models, as has been shown with PPI, respond

 

                                                             159

          1  to cognitive enhancers.

          2             Now I'm going to begin just by making a

          3  point which I thought--perhaps I'm missing

          4  something, and clearly the psychotic symptoms are

          5  distinct from some of the cognitive deficits in

          6  schizophrenia.  This is an extremely important

          7  discovery; however, I think we shouldn't lose sight

          8  of the fact that psychosis is actually a cognitive

          9  phenomenon or a cognitive behavioral phenomenon,

         10  and I think that's interesting because it gives us

         11  a clue about the nature of the cognitive

         12  impairments themselves.

         13             I certainly think it's possible to model

         14  some of these psychotic dysfunctions even in

         15  animals.  I'm prepared to defend this extraordinary

         16  claim that one can measure hallucinations in

         17  animals just as one can in people behaviorally and

         18  also model by looking at the building blocks of a

         19  assisted learning theory, such complex human

         20  elaborations as delusions, for example, as Kapur

         21  has recently written about.  I think a lot is

         22  understood about how the way the brain actually

 

                                                             160

          1  controls these basic blocks of learning and how

          2  they're manipulated by, for example, dopametric

          3  drugs.

          4             I certainly think it's also true that

          5  the negative symptoms, which of course are part and

          6  parcel of some of the cognitive deficits, can also

          7  be potentially modelled in animals, but one

          8  shouldn't perhaps expect drugs to effect this lot

          9  across the board in a common way.  This clearly can

         10  be specificity in a way which particular

         11  interventions will interact with particular neuro

         12  systems and will derail as a consequence of

         13  schizophrenia.

         14             I suppose as a passing comment, we

         15  really need a new neuropsychology, I think, to deal

         16  with schizophrenia just as much as recruiting the

         17  conventional neuropsychology and harvesting it to

         18  the best purpose.

         19             Next slide.  So we have these

         20  interesting domains from the MATRICS, and Jim Hagan

         21  has already done my job and pointed out that many

         22  of these, again, can be simulated properly in

 

                                                             161

          1  experiments with animals.  I think verbal learning

          2  is going to be a tall order, but most of the rest

          3  of them seem feasible.

          4             Next slide, we tried doing something

          5  like this between about 10 and 15 years ago with

          6  the Cantab project, and these are three of the

          7  tests in the Cantab battery, for those of you who

          8  haven't seen it, and they represent tests which are

          9  more sensitive to certain frontal executive

         10  dysfunction in humans and have different

         11  rationales.  So this testing, the test of planning

         12  at the top edge of the screen here is really

         13  inspired by the work of Gene Shalitz in cognitive

         14  science, cognitive neuroscience.  Interesting

         15  there, I think there are attempts now to show that

         16  monkeys can do these types of planning tasks, which

         17  is quite encouraging.

         18             This test of self-ordered spatial

         19  working memory, you'll see derives animal work from

         20  rats as well as monkeys.  It has been shown to be

         21  extremely useful in schizophrenia.  This test is

         22  inspired by the traditional approach of clinical

 

                                                             162

          1  neuropsychology, the Wisconsin Card Sorting task,

          2  but this is a Wisconsin Card Sort Task of the

          3  twenty-third century decomposed in a way that can

          4  be presented to a monkey.  So I think this is a

          5  very important element.

          6             When you give the Cantab battery to

          7  schizophrenics, cognitive deficit schizophrenics,

          8  you get the typical kind of profiles one sees for

          9  neuro batteries in a study in collaboration with

         10  Hutton and Joyce in London.  I just want to draw

         11  your attention to two or three interests points.

         12  You'll see that, as usual, the profile portrays

         13  some evidence of specificity, which we think is

         14  there because the tests are controlled in terms of

         15  difficulty level.  Interesting in these first

         16  episode of patients, the Wisconsin is not that

         17  disturbing.  The tests on learning, visual learning

         18  and spatial working memory are among the largest

         19  defects.

         20             I think this is very interesting because

         21  of something which hasn't been mentioned before--if

         22  it has, I'll stress it again, but there is

 

                                                             163

          1  particular heterogeneity in the schizophrenia

          2  population.  Just because a patient is bad in the

          3  impaired associates learning test, the visual

          4  learning test, doesn't necessarily mean to say that

          5  he or she will be bad in the spatial working memory

          6  test and vice versa, as we've recently shown in our

          7  submitted paper.  This suggests that patients

          8  actually have quite selective deficits in tests

          9  which probably are sensitive to damage to rather

         10  different parts of the brain.

         11             Well, the spacial working memory task is

         12  partly inspired from the classic pathological work

         13  that I won't re-describe for you, but this work

         14  which I think is one of the first papers really in

         15  cognitive neuroscience shows dopamine in the dorsal

         16  lateral prefrontal cortex produces a massive

         17  response.  Delayed response in spatial working

         18  memory is an abilation.  This effect can be

         19  partially ameliorated by dopamine agonists.  We set

         20  the agenda 25 years ago.

         21             Next slide, and the agenda, of course,

         22  continued to go there by showing that the effects

 

                                                             164

          1  of chronic neuro active drug treatment, chronic

          2  haloperidol, also produced cognitive impairment

          3  which were remediated by one dose of agonists with

          4  the hypothesis that this effect is a restoration of

          5  normalized prefrontal cortex function.

          6             How can we match this in human terms?

          7  Next slide, please.  Going back to our own

          8  self-ordered spacial working memory test, I don't

          9  have time to describe this in detail, but basically

         10  a subject has to search himself or herself for

         11  these wonderful rewards, these blue tokens which

         12  they collect and fill up this gap here, and they

         13  must learn not to return to a box which they were

         14  previously handed a token.  This test can be

         15  decomposed psychometrically in terms of spacial

         16  working memory capacity and strategic search, which

         17  appear to depend on probably different aspects of

         18  prefrontal function.

         19             Next slide, please.  So this early study

         20  PET study showed that the strategy evidence

         21  certainly activates the right dorsal lateral

         22  prefrontal area and spacial span capacity activates

 

                                                             165

          1  this right interior of the intralateral region.

          2             Next slide, please.  So that leads quite

          3  well with the animal agonists, but, of course,

          4  imaging evidence isn't precise because it doesn't

          5  demonstrate totality.  So lesions are still used in

          6  the study, and in these lesion patients, patients

          7  with frontal lobe damage to different areas,

          8  prefrontal, medial, dorsal lateral, large regions,

          9  we see that it's the dorsal lateral region which

         10  selectively produces the effect and performance on

         11  this task, matching the effects of very gross large

         12  regions.

         13             Next slide, the test is also sensitive

         14  to dopametric manipulation, and this is a study of

         15  dopamine in Parkinson's disease, which had a quite

         16  selective effective, which we now know with

         17  function imaging probably to be correlated with the

         18  restoration of the prefrontal cortex.  Just as an

         19  aside, and I think is an important initiative also

         20  to mention, it would be very nice if we had

         21  available in the drugs that industry could give us

         22  to explore the functions of the D-1 and D-2

 

                                                             166

          1  receptors in human tests in spatial working memory,

          2  but to my knowledge, there have been very few

          3  studies, for example, of the effects of D-1

          4  receptor antagonists which clearly will be

          5  predicated by the pattern under which each work,

          6  although we ourselves have evidence of the effects

          7  of sulphoride, which are detrimental.

          8             Next slide, the spatial working memory

          9  test, as you see, has some utility, largely in the

         10  writings by Chris Pantelis and others and Connie

         11  Joyce, around the deficits in cognitive

         12  schizophrenia, but perhaps more interestingly, the

         13  deficits showed earlier seemed to be present at the

         14  premorbid stage, before the first psychotic

         15  episode.  So this is another potential target for

         16  treatment, of course, in the same way that

         17  premorbid Alzheimer's disease might be.  There are

         18  various other interesting correlations to clinical

         19  variables which showed in some sense a validation

         20  of the test at that level.

         21             Next slide, now by contrast, the Cantab

         22  visual impaired associates learning test which

 

                                                             167

          1  involves subjects learning where objects are in

          2  boxes on the screen under a number of trials, and

          3  then they have--they're shown the objects and then

          4  after they've been shown the object in each box,

          5  they have to show the boxes and just by touching

          6  it, this test is sensitive in some groups of

          7  schizophrenics who are only impaired in spacial

          8  working necessarily over the IDD section.

          9             Next slide, please.  As a result of this

         10  test, the gain derives from the interplay of the

         11  classical neuropsychological work by Smith and

         12  Milner, patients' memory of the location of toys on

         13  trays and, of course, the monkey work of Bushke's

         14  showing the role of the medial temporal lobe system

         15  in controlling this type of object location and

         16  memory learning.

         17             Very interestingly, and this is a point

         18  I'd like to make, we've recently been able to show

         19  that this test actually predicts the diagnosis of

         20  Alzheimer's disease in patients with MCI, minimal

         21  cognitive impairment, to a very high degree of

         22  accuracy, much higher, for example, than this.  So

 

                                                             168

          1  clearly this is of interest.  This parallel between

          2  Alzheimer's and schizophrenia will be of interest,

          3  I think, in the development of drug targets.

          4             Next slide.  As you see, there is a

          5  version of the test that Mike Taffe developed, and

          6  this shows the very nice effects with schiponamine

          7  and ketamine.

          8             Next slide, I think the future will show

          9  that we're able to develop similar tests in rats.

         10  We've already developed tests of visual learning in

         11  rats using a similar touch-screen idea.  The rats

         12  simply go to the stimulus which it thinks is

         13  associated with the reward.  Perhaps again by using

         14  building blocks and simple vision discrimination

         15  data, we're able to come up with something as

         16  complex, possibly, as the PET associated task.

         17             Next slide, well, you've heard a lot

         18  about this one today, the Wisconsin Card Sort Task.

         19  A monkey is given a pack of cards and asked to sort

         20  into categories.  It's very hard to do.  They'd

         21  probably throw them up in the air and chew them or

         22  something like that.  So our task was to really

 

                                                             169

          1  decompose this very complex task, as has been

          2  mentioned by previously speakers, into a form which

          3  monkeys can do, next slide, elucidate.

          4             So here are two diametric shapes and

          5  lines which curve with one another.  The animals

          6  and the humans are trained to attend to one of the

          7  dimensions, that shape, and at some stage, they're

          8  shifted to lines.  That's the basis of the test.

          9  It's called an extradimensional shift, and there

         10  are other controls like an intradimensional shift

         11  where you change the set.  You have to shift to

         12  change the shapes.

         13             Next slide.  So there are a number of

         14  necessary building block stages which the animal

         15  and human have to learn according to some really

         16  generous criteria before they proceed to the

         17  critical stages.

         18             It's also worth pointing out that you

         19  can do some very interesting comparisons between

         20  this shifting here between shapes and lines and

         21  simply shifting between shapes and objects, which

         22  is the lower representation, and a reversal

 

                                                             170

          1  learning task, which is really redundant in mice

          2  and rats and monkeys.

          3             Next slide, now using exactly the same

          4  test stimuli as we use in monkeys, we begin to show

          5  the effects in schizophrenia, very large effects in

          6  chronic patients.  Now, remember back when I showed

          7  you the test scores early on, we certainly think

          8  the performance gets much worse as the disease

          9  progresses, to such an extent that they can't even

         10  achieve, they can't even get to the stage that

         11  qualifies as an intradimensional shift.  They're

         12  impaired in these basic building blocks and even

         13  quite apart from the extracognitive developmental

         14  shifting.

         15             So, in other words, it's fine for you to

         16  have some measure of these building blocks because

         17  if the building blocks are weak, the cognitive

         18  architecture will collapse, and that's probably

         19  what happens quite a lot with chronic

         20  schizophrenia.  The high-functioning subjects, they

         21  don't tend to show selective deficits in

         22  extradimensional shifting.

 

                                                             171

          1             Here are some clinical day from a

          2  collaboration with Chris Pantelis, the various

          3  stages of the task, percentage, control subjects

          4  there.  These are frontal patients who have a big

          5  problem with an ED shift.  Temporal lobe patients

          6  have their problems too, by the way, but this is a

          7  very large group of chronic schizophrenics who are

          8  impaired at the ED shift, but also very impaired

          9  even at the stage where they have generalize it all

         10  to do exempt tasks, the ID shift, which we think is

         11  a very basic measurement of the ability to extract

         12  and to produce working utilization.

         13             Next slide, please.  Well, it has some

         14  utility, we think, this test, because of the data

         15  which Jim kindly mentioned for us, the

         16  collaboration of Turner and others in Cambridge,

         17  showing that a single dose of modafinil does

         18  produce some improvement in this task.  It

         19  basically improves the subjects close to the end,

         20  which is encouraging and interesting and also

         21  challenging, because clearly we need to find other

         22  agents whose mode of action is much better

 

                                                             172

          1  understood to properly appreciate these data.

          2             By the way, if we go back one, the

          3  modafinil effect wasn't simply on this test, but it

          4  also--so you can generalize across two areas of

          5  executive dysfunction.

          6             Next slide, so here are the basic data

          7  on which these clinical data derive.  This is the

          8  study by Rebecca Dias, Andrew Roberts, and myself,

          9  showing the dysassociation that we see in the

         10  monkey on same test.  I've shown that lesions to

         11  the lateral prefrontal cortex, dorsolateral,

         12  selectively impair the extradimensional shift and

         13  then impair reversal, whereas the lesions to the

         14  orbitofrontal cortex have exactly the opposite

         15  effect, the reversal of shifting.  So this, I

         16  think, does show effectively the hierarchy of

         17  organization of responding to tasks and abstract

         18  dimensions even in the monkey brain.  It further

         19  shows the intradimensional and the basic building

         20  block is not effected by this prefrontal lesion,

         21  although it may be moderately affected in the

         22  orbitofrontal phase.

 

                                                             173

          1             Amazingly, a similar disassociation in

          2  rats, by Ferret and Brown, using this test not

          3  based on visual dimensions, but shifting between

          4  modalities, the same result essentially, reversal

          5  learning impaired by lesions and not shifting, as

          6  you see here, extradimensional sections impaired by

          7  medial lesions which may be homologous to the

          8  dorsolateral, maybe, but no effect on reversal.  So

          9  this kind of comparative psychology gives me a

         10  considerable faith in the fact that in these

         11  examples, these basic building block mechanisms are

         12  there and they have relevance in schizophrenia as

         13  we showed in the previous presentations.

         14             We've also been able to study in the

         15  monkey the effects of manipulation of one of the

         16  pathways into the frontal cortex.

         17             Next slide, and one thing you have to

         18  remember, which is an interesting complication, is

         19  when you deplete the dopamine in a monkey or rat,

         20  this is very interesting.  You have the possibility

         21  of a model of a cognitive deficit because of the

         22  dysregulation of the frontal cortex, but also this

 

                                                             174

          1  unregulated subcortical activity which may drive

          2  the perceptive salients and phenomena of delusions

          3  and hallucinations.  This is a very basic fact, I

          4  think, which extends further preparation, for

          5  example in the isolation of rats in that

          6  unregulated subcortical dopamine and not regulated

          7  frontal dopamine.

          8             Anyway, just to summarize what is a very

          9  complicated set of data, the study by Eric Crofts,

         10  Sandra Roberts, and myself on the section of the

         11  tasks and its very components, we found that

         12  dopamine depletion, interestingly, doesn't limit

         13  the effects of the lesions themselves in the

         14  frontal cortex in the animal, but they do effect

         15  the basic building block processes.  For example,

         16  they affect the intradimensional section itself.

         17  The basic learning is impaired by dopamine, and

         18  another thing that is impaired is the

         19  distractibility of the animal.  Now, this is a test

         20  we haven't yet used in humans, but it might

         21  illustrate the possible interplay that we can

         22  achieve between animal and human work.  If we

 

                                                             175

          1  change the background, irrelevance to this task so

          2  that they're distracted, this disrupts performance

          3  for a while.

          4             So the effects of dopamine loss in

          5  theses two cases, as I show here, is real data.  So

          6  this is a set of intradimensional sections--there

          7  are five actually here--where performance virtually

          8  improves because the monkeys selected the right

          9  direction.  Dopamine loss is very apparent on the

         10  first shift, but the fifth shift showed no effect

         11  whatsoever.  So they've been unable to form a set.

         12  So if they're unable to form a set, they're clearly

         13  going to have problems in shifting sets.  In fact,

         14  their sets may already be overloaded by the

         15  hypothesis based on the fact that in the

         16  distraction process, these monkeys are very

         17  distractible.  They take more errors--this is a

         18  square root scale to bring it all into the same

         19  range--many more errors to get back to normality

         20  when you introduce these background distractors in

         21  all animals or animals with dopamine depreciation

         22  in the prefrontal cortex.  It's actually

 

                                                             176

          1  interesting.  It shows exactly the opposite effect

          2  consistent with the dopamine change in the frontal

          3  cortex.

          4             So I think this is a really important

          5  point to make, that we need construct validation of

          6  this effect of dopamine loss if it's conceivably

          7  relevant to schizophrenia, and we've been inspired

          8  by the physiology studies of Sejnowksi and Seamons,

          9  which we interpret, in fact, we can argue about

         10  that this demonstrates the dopamine in the

         11  stabilization representations of the frontal

         12  cortex, that you get this distractibility at

         13  various set shifting.

         14             Moving to the rats to try and test this

         15  notion really of different species, we devised

         16  about 20 years ago this paradigm which was inspired

         17  again by human work, inspired, in fact, by

         18  continued performance tests used at the applied

         19  cycle homogenetics at Cambridge where the task was

         20  really to look at the effects of stressors and the

         21  biometric changes, temperature and so forth, on

         22  constant operation in human volunteers.  In this

 

                                                             177

          1  version of the test, the rats are confronted with

          2  these array of five apertures and have to detect.

          3  This rat test, we rigged up with a probe so we can

          4  measure neurochemical process while the animal does

          5  this.  We can talk about that if you like.

          6             The rats worked with selections and all

          7  were responding prematurely, impulsively or

          8  responding in a separative way after some deficit

          9  monitoring in response inhibition.  Interestingly,

         10  next slide please, the rat uses virtually the whole

         11  of its frontal cortex to do this task.  We see

         12  disassociations with this measures.  This is a

         13  summary, but basically lesions of the dorsal area

         14  impair selection accuracy and without much effect,

         15  for example, on premature sedentary responses,

         16  whereas some of these behaviors can be localized by

         17  region.  Premature responses are virtually

         18  associated with this area of the cortex, and the

         19  sedentary response, which one might predict,

         20  perhaps are associated with brain damage to the

         21  orbital frontal cortex.

         22             So here we have with the rat

 

                                                             178

          1  disassociations of executive processes which are

          2  according to prefrontal regions which have

          3  differential capability modulation as well and

          4  therefore subject to possible optimization and

          5  remediation by drug treatment.

          6             Next slide, this is one example of that,

          7  an intriguing result.  We studied rats with large

          8  modulations and had gross deficits in accuracy on

          9  this task, as you see here, and amounted to about

         10  65.  So the chance performance is 20 percent, a

         11  large range of opportunity to see changes.

         12  Intriguingly here, sulpiride, a D-2 receptor

         13  antagonist impairs performance, as Jim Hagan might

         14  have predicted in terms of some of his working

         15  memory data, virtually improved performance in the

         16  rats with frontal lesions.  The D-1 antagonist had

         17  a similar effect, although this is not significant.

         18             Next slide, please.  This D-2

         19  remediation is probably--as a concept of regulation

         20  of dopamine in this area, we get similar

         21  remediation.  So these are fairly small effects in

         22  the very context of the huge cognitive deficits in

 

                                                             179

          1  schizophrenia, but I think they're useful in

          2  limiting what the effects of the antipsychotic

          3  drugs might be.

          4             Next side, using the same task and

          5  injection into the medial prefrontal cortex and D-1

          6  areas, we began to show improvements in

          7  low-performing rats at least, specifically in the

          8  area of accuracy.

          9             Next slide, please.  Increasingly, 5-HT

         10  agents, for example the 5-HT antagonists, can

         11  improve performance even in normal rates as you see

         12  here.  We get quite a significant improvement,

         13  which might be relevant to interpreting the data

         14  that we heard earlier about olanzapinee affecting

         15  reaction times.

         16             So in this talk, I basically focused on

         17  the number of tests also that are relevant to

         18  attention dysfunction, which is something which is

         19  a very complex construct, widely acknowledged by

         20  MATRICS, but remembering that detection itself is

         21  very--there a very different types of attention.

         22  We see aspects of attention in the extradimensional

 

                                                             180

          1  set shifts tasks and the performance tests, but

          2  there are other tests of attention which affect

          3  very different neurosystems, a paper inspired by

          4  the work by Mike Posner in spatial attention which

          5  are pretty much dependant on--and of course there

          6  is the pre-pulse inhibition issue.  It is a very

          7  interesting basic building block of implicit

          8  inhibition which can mediated by circuits,

          9  including the subcortical dopamine sensitive areas.

         10             Next slide, please.  So in conclusion,

         11  and not to keep you from your lunch further, I've

         12  tried to argue that points of symptomatic models

         13  and aspects of schizophrenic function are feasible,

         14  but I do think the notion of a model needs further

         15  develop, which I hope we are going to do today and

         16  tomorrow in close association with neuropsychotics

         17  and neurogenetics and neuropathological studies

         18  around the disease itself.  The same even with the

         19  tests we have available, there is some promising

         20  evidence of changes to D-1 agents, 5-HT agents,

         21  which I haven't had a chance to show you, and also

         22  setting the stage for systematic analysis of the

 

                                                             181

          1  effect of the glutamatergic agents.

          2             Thank you very much.

          3             DR. GEYER:  Thank you, Trevor.

          4             So if I can have the breakout group

          5  leaders and co-leaders and reporters come just for

          6  a couple minutes--I promise not to keep you very

          7  long--we will then, after lunch, which is in the

          8  conference center cafeteria, not come back here

          9  unless you're coming to the breakout groups that's

         10  here, which you're welcome to do, but look in your

         11  program for the room for the breakout group to you

         12  which you signed up.  And then there will be a

         13  coffee break between breakout groups, and then

         14  after that, without any further instructions, you

         15  are to go to the second breakout group that you

         16  signed up for.

         17             We will then all reconvene as one group

         18  here tomorrow morning at eight in this room at

         19  which time we will then hear and discuss the

         20  results of the eight breakout groups today.

         21             Any particular questions about that?

         22             If you need a lunch ticket, you can buy

 

                                                             182

          1  a lunch ticket, if you don't already have the meal

          2  plan in the hotel, out at the registration desk.

          3             [Whereupon, at 12:16 lunch recess was

          4  taken, to convene breakout groups at 1:00 p.m. this

          5  same day.]

          6

          7

          8

          9

         10

         11

         12

         13

         14

         15

         16

         17

         18

         19

         20

         21

         22