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
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